The Real Estate Market in Portugal: Prices, Rents, Tourism and Accessibility

estudos da fundação k 2022
Paulo M. M. Rodrigues, coord.
The real
The analysis of real estate market increasing in recent years, following
dynamics has gained interest in a relatively inexpressive evolution
recent years, following the US sub during the 1990s and until the
prime collapse in 2007, which quickly financial crisis of 2007. A detailed
estate market
spread worldwide and led to housing characterisation of the price
markets significantly impacting the dynamics and different dimensions
global economy. As such, of the housing market up until the
in Portugal
understanding price and rent beginning of 2020 is provided. The in
dynamics in real estate markets is of depth and comprehensive analyses in
foremost importance from a social, the different chapters of this study
economic and public policy represent an important contribution
perspective. The focus of this study to the understanding of potential
is the Portuguese real estate market. future paths for the housing market
House prices in Portugal have been in Portugal. Prices, rents, tourism and accessibility
estudos da fundação k 2022
The real
estate market
in Portugal
Prices, rents, tourism and accessibility
Paulo M. M. Rodrigues, coordination
Acesso rápido k Capa | Contents | Introduction | Chapter 1 | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | Chapter 6 | Conclusion | Notes /1

Largo Monterroio Mascarenhas, n.º 1, 7.º piso
1099‑081 Lisboa
Tel: 21 001 58 00
[email protected]
Director of publications: António Araújo

Director of the series Foundation Studies: Gonçalo Saraiva Matias
Title: The real estate market in Portugal: Prices, rents, tourism and accessibility
Authors: Paulo M. M. Rodrigues (coordenação), Luis Aguiar‑Conraria, Vera Gouveia Barros,
Paulo Batista, Pedro Brinca, Eduardo Anselmo Castro, João B. Duarte, Duarte
Gonçalves, René Huget, Rita Fradique Lourenço, João Lourenço Marques, Susana
Peralta, Victor Reis, João Pereira dos Santos e Maria Joana Soares
Proofreading and translation: TT Translations
Design: Inês Sena
Typesetting: Guidesign
© Francisco Manuel dos Santos Foundation

March 2022
ISBN: 978–989–9064–80–5
The opinions expressed in this publication are those of the authors and are
not attributable to the Francisco Manuel dos Santos Foundation or to the institutions
where they are affiliated. Permission to reproduce all or part of the contents must
be requested from the authors and the publisher.

Contents
Introduction 5 Chapter 2
References 8 Exuberance and contagion in the
Portuguese housing market: a perspective
Chapter 1
on disaggregate local residential prices 27
The importance of housing in Portugal:
2.1. Introduction 27
a macroeconomic perspective 9
2.2. Definition of housing bubbles and contagion 29
1.1. Introduction 9
2.3. The dynamics of the Portuguese housing market 30
1.2. Literature review on housing
2.4. Possible additional sources of
and the macroeconomy 9
exuberant behaviour 42
1.3. Relative importance of housing in GDP 11
2.5. Conclusions 44
1.4. The importance of housing
References 45
in Portugal's business cycles 13
Appendix A: the general supremum augmented
1.5. The importance of real estate in
Dickey Fuller test (GSADF) 47
household balance sheets 17
Appendix B: quantile cointegration test and results 48
1.6. Conclusion 24
References 25 Chapter 3
Synchronism in Portuguese housing 49
3.1. Introduction 49
3.2. Literature review 50
3.3. Data and methods 52
3.4. Results 54
3.5. Conclusions 61
References 63

Chapter 4 Chapter 6
Territorial patterns of housing Urban rents in Portugal: a decline
affordability in Portugal 73 with no return? 125
4.1. Introduction 73 6.1. Introduction 125
4.2. Social, economic and spatial elements 6.2. Census information 126
under housing affordability analysis 75 6.3. Rent control in housing markets 134
4.3. Measuring housing affordability 80 6.4. Social lease systems 144
4.4. Territorial patterns of housing 6.5. Is it possible to regulate housing supply
affordability in Portugal 89 through legislative action? 148
4.5. Concluding remarks 100 6.6. The segmentation of the lease market 151
References 102 6.7. Brief conclusion 152
6.8. Conclusion 153
Chapter 5
Abbreviations 153
Short‑term rentals and real estate
References 154
prices in Lisbon: boom and ban 107
5.1. Introduction 107 Conclusion 155
5.2. Portugal as an interesting case study 109
Notes 159
5.3. Data sources and empirical strategy 115
5.4. Results 118
5.5. Conclusion 121
References 123

Introduction
Residential property is the most important component of households’ as opposed to other forms of wealth accumulation, also matter, and
wealth, and is often used as collateral for loans (Englund et al., 2002). so does the average level of interest rates possibly related to the long
This said, changes in house prices may affect households’ consump‑ ‑term behaviour of inflation. The availability and cost of land, as well
tion and investment decisions, as well as their debt and ability to repay as the cost of construction and investments in the improvement of
loans (Case, Quigley and Shiller, 2005). Developments in real estate the quality of existing housing stock are also relevant (Poterba et
markets have far‑reaching implications on the economy as a whole and al., 1991, and Tsatsaronis and Zhu, 2004). For instance, the growth
on financial stability. of the housing stock can be constrained in the short run as a result of
a number of factors, including length of planning and construction.
It has been shown that changes in housing wealth can have a more There could also be shorter‑term drivers related to constraints in the
important effect on the economy than changes in wealth caused by growth of housing stock, prevailing conditions in the provision of
stock price movements (Helbling and Terrones, 2003 and Rapach and housing loans, or uncertainty about future prospects. Higher GDP and
Strauss, 2006). In fact, economic history suggests that some of the disposable income, more confidence in the economy, less unemploy‑
most severe systemic financial crises have been associated with boom ment, more labour and an increase in mortgage lending are expected
‑bust cycles in real estate markets (see e.g., Bordo and Jeanne, 2002, to have a positive impact on the housing market. In contrast, higher
Reinhart and Rogoff, 2013, and Crowe et al. 2013). interest rates are expected to drive borrowing costs up and demand
The analysis of real estate market dynamics has gained particular inte‑ down, leading to a subsequent fall in house prices and making alterna‑
rest in recent years, following the US sub‑prime collapse in 2007, tive applications of wealth more interesting. The same happens with
which quickly spread worldwide and led to significant impacts of residential investment, as if it increases, prices may go down.
housing markets on the global economy. Understanding price determi‑ This book focuses on the Portuguese real estate market. House prices in
nation processes in real estate markets and its dynamics is therefore of Portugal have been increasing in the recent past, following a relatively
paramount importance. inexpressive trend during the 1990s, until the financial crisis in 2007.
Traditional determinants of housing demand include growth in Over the last decade, the evolution of house prices has been highly
household disposable income and gradual shifts in demographics, differentiated, declining four per cent in the 2008–2013 period and
such as the relative size of older and younger generations. Permanent accelerating over six per cent up to the end of 2020. From the late 90s
features of the tax system that might encourage home ownership, and until 2007, residential investment recorded a two per cent contrac‑
tion while GDP only registered a slight acceleration. In the five years

following the financial crisis and until the recovery in 2013, the stron‑ This book comprises six chapters and provides a comprehensive
gest contraction was in terms of investment with over eleven per cent, analysis of several dimensions of the real estate market in Portugal.
compared to one per cent in GDP. The unemployment rate increased
sharply, and labour force declined, which may have been related to Chapter 1 — The importance of housing in Portugal: a macroeconomic
emigration flows. Between 2014 and 2019, amidst increasing confi‑ perspective, by Pedro Brinca and João B. Duarte, provides a description
dence, GDP accelerated two per cent in Portugal and residential gross of the housing market's importance for macroeconomic dynamics in
fixed capital formation (GFCF) increased four per cent. The COVID Portugal. The description is divided into two parts: the first looks at
‑19 pandemic caused an unprecedented GDP contraction of almost the importance of the real estate market in economic cycles; and the
eight per cent, although residential GFCF and house prices continued second analyses the relevance of real estate in the financial balance of
to increase. It is also important to analyse credit in detail, given its rele‑ families. Throughout the chapter, the Portuguese reality is compared
vance to the housing sector and the impact it may have on the cost of to that of the European Union, providing interesting different
financing. Data on bank lending indicates the existence of episodes of temporal and spatial perspectives regarding the real estate sector in
very high growth in mortgage loans between the mid‑1990s and 2007. Portugal, and how it affects the Portuguese macroeconomic context.
This annual growth averaged around fifteen per cent in Portugal, within Chapter 2 — Exuberance and contagion in the Portuguese housing
a context of declining bank loan costs and high and sustained growth market: a perspective from disaggregate local residential prices, by
in household disposable income, which resulted in an increase of family René Huget, Rita F. Lourenço and Paulo M. M. Rodrigues, specifi‑
indebtedness. The significant deceleration of housing credit from 2010 cally focuses on the significant issue of exuberant behaviour within
onward should be observed in the context of the international financial the Portuguese real estate market. This chapter seeks to answer two
crisis, which had a negative impact on supply, due to a significant tighte‑ important questions: the first is whether Portuguese house prices
ning in lending conditions, and on housing credit demand. have deviated from their fundamental determinants; and the second
Aiming to better understand the dynamics of the real‑estate market is whether local level prices (by district and civil parish) present
in Portugal, this book provides a detailed characterisation of its diffe‑ bubble‑like developments and if the exuberant behaviour is conta‑
rent dimensions, up until the beginning of 2020. Unfortunately, at the gious. To answer these questions, an aggregate and detailed local level
time some of these chapters were being written, not enough informa‑ analysis is performed, allowing us to understand the origins of the
tion was available regarding the impact of the COVID‑19 pandemic exuberance and whether this behaviour may have generated spillover
on the real estate market. However, the detailed analyses provided in effects across civil parishes.
this book will surely be useful to understand the potential paths the Chapter 3 — Synchronism in Portuguese housing, by Vera Gouveia
housing market may follow in a near future. Barros, Luís Aguiar‑Conraria and Maria Joana Soares, further

investigates the dynamics of housing prices in Portugal. Using wavelet interlinkages (spatial autocorrelation) and the expression of these
analysis, this chapter specifically focuses on housing price synchro‑ territorial features in housing affordability metrics.
nization between Portugal and other countries, detecting which
international markets are more deeply interconnected with the Chapter 5 — Short‑term rentals and real estate prices in Lisbon: boom
Portuguese market, while also investigating housing price synchro‑ and ban, by Duarte Gonçalves, Susana Peralta and João Pereira dos
nisation between Portuguese cities. Studying price synchronisation Santos, approaches another important dimension of the real estate
between different housing markets is of considerable relevance, as market which is of particular relevance to the Portuguese economy,
it may provide information on whether global investors and global namely short‑term rentals. This chapter specifically analyses the ban
financial conditions influence local housing price dynamics, which has implemented by the Lisbon City Council in some of its neighbou‑
implications on the effectiveness of a range of policy tools created rhoods in November 2018, which provides an ideal quasi‑experimental
to address imbalances in the housing market, such as affordability setup to estimate the causal impact of short‑term rental regulations on
problems (Hoesli, 2020; IMF, 2018). the real estate market. A quantification of this ban's impact on regis‑
tries, Airbnb prices and the real estate market in the city of Lisbon,
Affordability is precisely the topic of analysis in chapter 4 — namely the number of transactions and house prices, is also presented.
Territorial patterns of housing affordability in Portugal, by Paulo
Batista, João Lourenço Marques, Eduardo Anselmo Castro. This Finally, chapter 6 — Urban rents in Portugal: a decline with no return?,
chapter provides an integrated territorial analysis of housing affor‑ by Victor Reis, analyses the evolution of urban rents in Portugal based
dability. It not only reflects on the challenges posed to model market on existing traditional dwellings, as well as their forms and regimes
drivers by data collection and processing, but also (re)introduces the since the beginning of the 20th century. This important topic, which
concept of territory in housing affordability metrics. Here, a concept has received little attention to date, addresses the development of
of territory unbound from its common administrative expressions is the traditional housing stock, both nationally and within the cities
used (João Marques and Batista, 2021). In line with the debate on the of Lisbon and Porto, and their respective historic centres, since the
role of space developed by (Harvey, 2006; Lefebvre, 1991), the territo‑ 1960 census. The main legislative measures put in place over the
rial analysis of affordability considers a space where different forces last century, in terms of the liberalised rental markets, as well as the
interact and, thus, territorial patterns are not objectively fixed or various social housing rental systems related to public housing poli‑
bound a priori. To consider the notion of space‑territory, an approach cies, are also presented.
which integrates econometric tools was developed to observe the
expression of the spatial drivers of housing affordability, their spatial

References
Bordo, M.D. and O. Jeanne, (2002), Boom‑busts in asset prices, economic
instability, and monetary policy, (2013), The role of contagion in the last american
housing cycle. (8966).
Case, K.E., J.M. Quigley and R.J. Shiller, (2005), «Comparing Wealth
Effects: the Stock Market versus the Housing Market», Advances
in Macroeconomics, 5 (Article 1).
Crowe, C., G. Dell’Ariccia, D. Igan and P. Rabanal, (2013), «How to

deal with real estate booms: lessons from country experiences», Journal of
Financial Stability 9: 300–319.
Englund, P., M. Hwang and J.M. Quigley (2002), «Hedging Housing Risk»,
Journal of Real Estate Finance and Economics, 24, 167–200.
Helbling, T. and M. Terrones (2003), «Real and Financial Effects of

Bursting Asset Price Bubbles», IMF World Economic Outlook, Chapter, 2,
67–92.
Poterba, J. (1991), «House Price Dynamics: The Role of Tax Policy and
Demography», Brookings Papers on Economic Activity, 2(143).
Rapach, D. and J.K. Strauss, (2006), «The Long‑Run Relationship Between

Consumption and Housing Wealth in the Eight District States», Federal
Reserve Bank of St. Louis, Regional Economic Development, 2(2).
Reinhart, C.M. and K.S. Rogoff, (2013), «Banking crises: an equal opportu‑
nity menace», Journal of Banking and Finance 37(11), 4557–4573.
Tsatsaronis, K. and H. Zhu, (2004), «What drives Housing Price Dynamics:

Cross Country Evidence», Tech. rep., BIS Quarterly Review, BIS.

Chapter 1
The importance of housing
in Portugal: a macroeconomic
perspective
Pedro Brinca and João B. Duarte
1.1. Introduction
In this chapter, we describe the importance of the housing market description of the relative importance of housing in the Portuguese
for macroeconomic dynamics in Portugal. This description will be economy. In particular, this chapter focuses on providing stylized facts
divided into two blocks: 1) the importance of the real estate market from a selected number of dimensions previously identified by the lite‑
in economic cycles; and 2) the relevance of real estate in the finan‑ rature as sufficient statistics to provide a measure of the strength in
cial balance of families. First, we describe the real estate cycle in which macro policies interact with the housing market.
Portugal using data from the national accounts of Statistics Portugal
For the sake of comparability, it is important to note that during our
(INE), the Statistics Office of the European Union (Eurostat) and
analysis we will compare Portuguese characteristics with those of the
the Organisation for Economic Co‑operation and Development
European Union. This way, we bring not only temporal but also spatial
(OECD). Second, we will use microdata from the Household Finance and
variation into the description.
Consumption Survey (HFCS) of the European Central Bank to describe
the importance of real estate in the financial statements of families
in Portugal. Regarding national accounts data, we use data available 1.2. Literature review on housing
from the past three decades, starting our study from before the intro‑ and the macroeconomy
duction of the single currency (euro). For the second part, in which There is an extensive literature on macroeconomics developed over
we work with microdata, we use three surveys carried out by the the last two decades which demonstrates the relevance of the housing
European Central Bank in 2010, 2013, and 2017. market for business cycle dynamics (Piazzesi and Schneider, 2016).
To a large extent, these recent advances in the literature were motivated
Against the backdrop of the relevant nexus between housing and
by the event of the Great Recession, where the collapse of the American
macroeconomic dynamics, our contribution in this chapter is the

housing market gave rise to a devastating financial crisis, which in turn early 2000s, the works of Krusell and Smith (1998) and Castañeda et
caused the world economy to plunge into a deep recession. al. (2003) were the state of the art with regard to creating environ‑
ments that could replicate the joint distribution of income and wealth
On the one hand, this literature has revealed that economic cycles by resorting to incomplete markets, individual characteristics and
are strongly affected by real estate market cycles. On the other hand, permanent‑behaviour.
it has shown that macro policies, namely monetary policy (Iacoviello,
2005, Iacoviello and Neri, 2010, and Dias and Duarte, 2019) and credit However, despite the fact that these models had quantitative rele‑
access policy (Mian and Sufi, 2018) channel many of their effects to vance in the sense that they could replicate observed moments of joint
the rest of the economy through the housing market. For instance, distribution of income and wealth, aggregate variables showed little to
state‑of‑the‑art HANK models highlight the importance of the compo‑ no dependence from higher moments of such distributions. The main
sition of household balance sheets, where housing plays a key role result in Krusell and Smith (1998) is precisely that average wealth was
for monetary and fiscal policy (Auclert, 2019, and Kaplan, Moll and a sufficient statistic for agents to forecast aggregate dynamics.
Violante, 2018).
This led Lucas (2003) to claim that «for determining the behaviour of
This last role is particularly important for a number of reasons. First, it aggregates, [Krusell and Smith (1998)] discovered realistically modelled
plays into a broader picture of what has been the role of heterogeneity household heterogeneity just does not matter very much». A few years
in macroeconomics over the past 40 years. Following Brinca (2020), would pass until the Great Recession of 2008, where housing played
modern macroeconomics has expanded its focus solely from the dyna‑ a major role, and would contribute towards a greater sense that hete‑
mics of aggregate variables, such as income, consumption or savings, rogeneity mattered for aggregate dynamics. The issue with the second
to the dynamics of the distributions that aggregate up to those varia‑ generation of models that emerged in the 80s, was that while they
bles. The challenge was two‑fold: first to understand how heterogeneity were indeed able to replicate observed distributions of income and
could be relevant to aggregate dynamics; and second, to understand the wealth, they failed to replicate distributions of marginal propensities
distributional impacts of the dynamics of aggregate variables. to work and consume out of income. These models predict relatively
high marginal propensities at the bottom of the distribution, where
Early work exploring the role of heterogeneity can be traced back to credit constraints could prevent intertemporal optimization and lead
Fisher (1930) who focused on age, and Kaldor (1955) and Pasinetti to hand‑to‑mouth behaviour, i.e., marginal propensities to consume
(1962) on ownership of production factors. Later in the 1980s, incom‑ equal to one, for example. But rational dynastic agents would anti‑
plete market models in terms of uninsurable labour income risk would cipate the welfare losses of being close to the credit constraint and
emerge (see Bewley, 1980), and this second generation of models would thus accumulate precautionary savings to insure against such
became widespread and a vibrant research avenue. By the late 1990s, event, which leads to a very fast decay of marginal propensities.
This makes the average propensity to consume very low and close 1.3. Relative importance of housing in GDP
to 0.1, as households are still able to intertemporally smooth income
We begin our analysis with the following simple but essential ques‑
shocks and, consequently, very little change in aggregate behaviour
tion: what is the real estate market's contribution to the Portuguese
is observed. This is counterfactual, as studies have indeed shown
GDP? This contribution is fundamentally made through two channels:
that marginal propensities to consume are not just higher on average
but are also much less decaying with respect to income levels (see, • Consumption, through renting;
for example, Kaplan and Violante, 2014). This showed that the correla‑ • Investment, through the construction of new properties.
tion of income and consumption was much higher than the permanent
income hypothesis would predict, even for high levels of income, and In figure 1, we show the behaviour of these contributions
thus the term «wealthy‑hand‑to‑mouth» was popularized. to Portuguese GDP from 1995 to 2017.
This led to a third generation of models that explored ways in which

micro‑heterogeneity would allow for a stronger propagation of aggre‑ Figure 1Contribution of the real estate market to Portugal's GDP
gate shocks. Oh and Reis (2012), McKay and Reis (2016), and Kaplan, 25
Moll and Violante (2018) brought forth environments with nominal
rigidities, incomplete markets, illiquid assets and aggregate risk, which 20
improved upon the empirical plausibility of marginal propensities to
GDP Percentage
consume and work in those respective environments. In particular, 15
illiquid assets are a widely used feature to achieve such purpose and
housing represents the lion's share in that class of assets. 10
This is of special relevance for Portugal too as housing is typically 5
and by far the most important asset in Portuguese household port‑

folios, and therefore cannot be used for consumption smoothing. 0
1995 2000 2005 2010 2015
Consequently, and in line with the literature, high percentages of Year
wealth being held in illiquid assets lead, ceteris paribus, to stronger Imputed rents Rents Residential investment Utilities
responses of household consumption in face of aggregate shocks.
Source: Eurostat
This is consistent with the latest COVID‑19 crisis, for example, as in
2020 Portugal's Gross Domestic Product (GDP) contracted more than
the average in the Euro‑Area.
It is possible to observe that the real estate market accounted for Figure 2Contribution of the real estate market to GDP in the
around 20% of the Portuguese GDP in 2017. The housing market's European Union
contribution to GDP reached almost 25% in 2001 and then began 25
decreasing until 2017. Regarding the various parcels that make up this
part of the Portuguese GDP, we can see that from 1995 to 2002, the 20
major part of this contribution was made by residential investment,

which added up approximately 16% of GDP, followed by the value of
GDP Percentage
15
imputed rents, responsible for roughly 5%. However, by early 2000s,

when Europe experienced the introduction of the single currency, this 10
dynamic was altered. There was a sharp drop in the value of residential
5
investment, which reached its minimum level in 2014 — a date that
marks the end of the Troika's intervention— with 7.5% of GDP, stabi‑
0
lizing in the following years. This trend was inversely accompanied 1995 2000 2005 2010 2015
by an increase in the weight of imputed rents (transmitted through a Year
rise in rental prices, as we will see later). Therefore, while there was an Imputed rents Rents Residential investment Utilities
increase in the contribution of rents, there was also a decrease in the

Source: Eurostat
contribution from residential investment.
Although Portugal's value has been slightly below the European Union's
Comparing these dynamics with the European Union's average, it is
average, in terms of the total amount of the housing market's contri‑
possible to highlight significantly different behaviour. Figure 2 clearly
bution to GDP over the last few years, there are differences in each
shows that the contribution of the real estate market to the European
component's weight. There is an aspect that differs significantly from
Union's GDP has been relatively constant over time, amounting to
the Portuguese case: the proportions of each parcel remain practically
approximately 22.5%.
unchanged during the period in question, with slight changes during
the 2008 crisis. In this European scenario, the housing components with
higher weight are: residential investment with 11% and imputed rents
with 7%, followed by rents and utilities. Unlike Portugal, the average
value of European rents does not appear to have changed, nor has there
been a drop in investment in this market. It should be noted that this is
only an average and each country has different characteristics. In other
words, the European average may reveal different dynamics behind each Figure 3 displays the behaviour of three variables regarding trend devia‑
member country. Summing up, the property market in Portugal has a tions in Portugal: GDP, investment in real estate assets (residential), and
contribution to GDP in terms of rents similar to the European average. investment in non‑real estate assets. First, we highlight the fact that the
However, residential investment contributed around 7.5% to GDP in three variables behave similarly, or in other words, are strongly corre‑
2017, 2.5 percentage points below the European average. lated over time. Second, residential investment appears to have a slightly
later cycle than non‑residential investment, as the latter falls first and
1.4. The importance of housing also recovers first. However, it is worth noting that neither seem to
in Portugal's business cycles precede a fall or recovery in GDP. Third, with a standard deviation of
1.3%, GDP is clearly the least volatile variable of the three.
Given the high importance of residential investment in Portuguese
GDP, we will analyse the evolution of residential investment cycles in Another interesting point to be taken from figure 3 is the presence of
Portugal, contrasting them with economic and non‑residential invest‑ lead‑lag correlations between the three variables in question. Table
ment cycles. 1 displays a correlation matrix regarding the cycle components of
HP‑filtered data, with a focus on residential investment. As we can see,
residential investment is a coincident indicator of GDP, with a very
Figure 3GDP and investment cycles in Portugal
high correlation (0.81), even higher than non‑residential investment.
0.2 This highlights the importance of residential investment in business
cycle fluctuations and stabilization policy in Portugal.
Percentage deviations from trend
0.1
-0.1
-0.2
2000 2005 2010 2015
Year
GDP Residential investment Non−residential investment
Source: Eurostat
Table 1Matrix correlation between GDP, non‑residential and residential Figure 4Real house prices
investment 1.20
Yt Yt−1 It It−1 Ht Ht−1 Ht−2 Ht−3 1.15

Yt 1 0.87 0.71 0.55 0.81 0.74 0.68 0.57
Yt−1 1 0.73 0.71 0.73 0.81 0.74 0.69 1.10
It 1 0.71 0.57 0.62 0.66 0.67

1.05
It−1 1 0.44 0.55 0.62 0.66
Ht 1 0.81 0.67 0.56 1.00
Ht−1 1 0.82 0.69
0.95
Ht−2 1 0.82
Ht−3 1
0.90
2008 2010 2012 2014 2016 2018
Note: Yt — GDP at time t, It — non‑residential investment at time t, H — residential investment at time t. Year
PT EU
1.4.1. Price and renting dynamics Source: Eurostat
Regarding the assessment of price developments in the real estate

Note that, after the great recession of 2008, there was a clear decrease
sector, we need to consider two aspects: the price of real estate assets
in the value of real estate in both cases, followed by a recovery that
and the price of their lease. Starting with the first, the narrative in
started in 2013. Since then, Portugal has shown a sharp rise in prices,
Portugal since around 2014 has been that house prices in Portugal are
which shows no sign of deceleration. However, a price level similar
very high. In various media it is reported that «real estate is on the
to that of the time preceding the Great Recession of 2008 was only
rise» (Expresso, 2015), or that it has reached «maximum values since

reached in 2017. The «very high prices» narrative in Portugal is thus
the peak of the crisis in 2013» (Observador,
2017). Although this news
based on nominal rather than real prices — adjusted for inflation
is factual in nominal terms, it conveys a distorted view of reality, as
(a detailed analysis using real house prices is provided in chapter 2).
the important factor for economic agents’ decisions are relative prices
Price developments in the European Union are much smoother than
— housing prices adjusted for inflation. Since 2014, there has been
in Portugal. In comparison to the European Union, the decline shown
robust growth in housing prices, but if we take real housing prices into
in Portugal is considerably more accentuated, as is the recovery. Let us
account, the scenario is not so impressive. Figure 4 displays the evolu‑
note that, despite the years characterized by the sovereign debt crisis,
tion of real house prices, both in Portugal and the European Union.
when the difference between European and Portuguese prices was at
its smallest, this difference has increased considerably and shows no namely in 2014, which registered a sharp rise of the variable in question
signs of slowing down when we look at the last year considered. (for a detailed analysis of rents in Portugal, see chapter 6).
Has the rental market behaved similarly? Figure 5 shows the dynamics We were thus able to establish two distinct periods concerning the
of the renting market, once again in Portugal and the European Union. evolution of the rental price: before 2014, when this price was signi‑
ficantly lower than the European average, and after 2014, a period
during which Portugal demonstrated convergence with its European
Figure 5Real lease prices peers on this matter. Even so, it is relevant to once more point out
1.04 that, as of 2014, the real rental price was much more susceptible
to the effect of seasonality, which is the reason behind the high
1.02
‑intensity variance registered during this period. Combining this
1.00 graph with the previous one, we can see that the recovery in prices
coincides with the explosion of rents. The year of 2014 is marked by
0.98
a limited offer (since residential investment had been decreasing for
0.96 several years at the time) and high demand, due to the bullish beha‑
viour of the tourism sector.
0.94
As we explained in the beginning of this chapter, while comparing the

0.92
2008 2010 2012 2014 2016 2018
Portuguese reality with that of the European Union, one can observe
Year differences within the Union by looking at lease prices by country
PT EU in figure 6. While there is an obvious link between figures 5 and 6,
Source: Eurostat
the former hides several aspects that the latter displays. In fact, there
are asymmetries within each country. Nevertheless, the behaviour of
Here we observe a considerably different evolution. Firstly, rental values each market is not completely independent from the others. Notice
in the European Union prove to be less exposed to seasonal effects than that Portugal was not the only country with an increase in rent prices
Portuguese values. Second, after 2009, there was a slight drop in the between the period 2013–2015.
European Union until 2012, when the value of rents increased again,
also moderately, until returning to the value of 2008. On the contrary,
Portugal demonstrated much more accentuated falls and recoveries,
Figure 6Real lease prices by country government bonds (which have much lower returns). Table 2 presents
120 two variables which describe part of the rental scenario in Portugal.
100
Table 2Additional property and lease
80 Portugal Europe
2013
60
% of households with more than one real estate property 30,31% 24,25%
% of households that rent extra real estate 13,99% 30,66%
40
2017
% of households with more than one real estate 29,22% 22,16%
20
% of households that rent extra real estate 13,40% 41,14%
0 Source: HFCS and authors’ calculations

2000 2002 2004 2006 2008 2010 2012 2014 2016 2018
Belgium Austria Cyprus Estonia Greece From these numbers, we were able to make two major findings:
Spain Czechia Portugal Lithuania Malta
Netherlands Italy Germany Finland • On average, in Portugal, there is a higher percentage of families
Bulgaria Poland Latvia Ireland
owning two or more real estate properties than in Europe in general.
France Denmark Slovakia Luxembourg
Although this variable decreased for the two regions considered from
Source: Eurostat
2013 to 2017, the distance between them increased.
• On average, and despite the point mentioned above, in Europe, a
What does all this tell us about the profitability of real estate invest‑
considerably more significant portion of families who own real estate
ment? Being a medium/long‑term investment, is this type of asset
in addition to their primary residence, decide to rent it. Not only did
with little liquidity not attractive in comparison to different kinds of
this figure remain practically unchanged in Portugal for four years,
investment? No, on the contrary. Real estate investment continues to
but it also rose by more than ten percentage points in Europe. This
have high profitability levels in relation to the risk it pertains. Jordà
statistic shows that, in general, the Portuguese give their secondary
et al. (2019) show that residential investment generates an attrac‑
residence(s) another use (other types of business than short‑term,
tive return, taking into account its expenses when compared to other
long‑term or holiday rentals, etc.).
types of investment, such as equity (which has a higher level of risk) or
1.4.2. How the macro real estate cycle Figure 7Monthly instalment by NUTS III
determines regional cycles 1600
Looking at the regional segmentation of aggregate data, should we 1500
assume that local real estate markets are independent with regard 1400
to price developments? No. Despite their characteristics (supply and
Prestação média mensal (EUR)

1300
demand, strong presence/lack of tourism or foreign residents, etc.),
1200
the national and international macroeconomic scenario (namely deci‑
sions at the fiscal and monetary level) affects all regional markets, 1100
as we can see in figure 7. It is possible to draw a connection line 1000

between figures 7 and 5, where the former is a decomposition of the
900
latter by NUTS III.
800
700
2011 2012 2013 2014 2015 2016 2017 2018
Alentejo Beiras e Serra da Estrela Autonomous Region of Madeira

Alentejo Central Centre Autonomous Region of the Azores
Alentejo Litoral Mainland Portugal Aveiro Region
Algarve Cávado Coimbra Region
Alto Alentejo Douro Leiria Region
Alto Minho Lezíria do Tejo Terras de Trás-os-Montes
Alto Tâmega Middle Tagus Tâmega e Sousa
Ave North Viseu Dão Lafões
Baixo Alentejo West Lisbon Metropolitan Area
Beira Baixa Portugal Porto Metropolitan Area
1.5. The importance of real estate

in household balance sheets
We now move on to the description of the importance of real

estate in household balance sheets. Recent work by Mian, Sufi and
Verner (2017), Cloyne, Ferreira and Surico (2020), and Slacalek et
al. (2020) come to highlight how household balance sheets are not Portugal Europe
only decisive for economic cycles but also for stabilization policies. 2013
In this sense, this section contributes to describe household balance % relative to total wealth 79.62% 81.07%
sheets in Portugal. To this end, we used data from the European Monthly instalment 343 EUR 670 EUR
Central Bank, obtained through a set of questionnaires carried out Effort rate 18.49% 18.13%
in several European Union countries (Germany, Austria, Belgium, Amount outstanding EUR 72 723 91 466 EUR
Cyprus, Croatia, Slovakia, Slovenia, Spain, Estonia, Finland, France, % of ownership of the house 45.47% 49.64%
Greece, Netherlands, Hungary, Ireland, Italy, Latvia, Lithuania, 2017
Luxembourg, Malta, Poland and Portugal), the Household Finance % relative to total wealth 80.79% 81.73%
and Consumption Survey. (Note: Throughout the section, having this Monthly instalment 320 EUR 740 EUR
dataset as basis for the analysis, we will use European Union and Effort rate 16.16% 16.42%
Europe interchangeably). Amount outstanding 61 792 EUR 101 147 EUR

% of ownership of the house 44.02% 49.53%
1.5.1. Financing for the acquisition of primary residences Source: HFCS and authors’ calculations
We start with financing and indebtedness in the acquisition (and

It should be noted that, in Portugal, the weight of a household's
respective payment) of households’ main residences. Table 3 contains
primary residence in relation to the total value of its wealth is quite
information regarding Portuguese and European households’ main
similar to that of the European average. We also observe that, despite
residence for 2010, 2013 and 2017.
paying significantly lower instalments, the effort rate (percentage
of income spent on monthly payment in relation to total income) is
Table 3Financing of primary residence higher in Portugal. Regarding the amount outstanding on loans made
for the household's main residence, there is a significant difference
Portugal Europe
between Portugal and Europe, a situation that does not occur when
2010
we observe the percentage of ownership of the same real estate asset.
% relative to total wealth 80.21% 81.71%
In 2013, with respect to our first variable, the value of the main resi‑
Monthly instalment 391 EUR 661 EUR
dence in relation to the total value of assets, we concluded that it did
Effort rate 21.55% 18.39%
not show major changes, neither at a national nor European level. The
Amount outstanding 71 434 EUR 87 380 EUR
most significant difference in this seven‑year period was the decrease
% of ownership of the house 46.36% 47.66%
in the effort rate, both in Europe in general and in Portugal. However,
there are also signs of divergence, namely regarding outstanding loans period in Portugal following the sovereign debt crisis, with significant
acquired for main residences, from 2013 to 2017. While Portugal deleveraging and protracted national demand.
saw this value decrease, Europe saw it increase substantially by the
same magnitude. We can say that more houses for primary residence Instead of exclusively analysing the averages of the variables described
purposes were purchased in Europe (a fact that increases the amount above, we can consider the entire distribution regarding each one
due) than in Portugal. Another differentiating aspect is the behaviour of them (for example, if there is a strong concentration around the
of the main residence's monthly instalment, where Portugal once average, if the variables demonstrate high variance, etc.). The variable
again displays a downward trend, unlike the European Union. Finally, in question will be displayed on the horizontal axis and its density on
Portugal shows a slight decrease in the percentage of ownership of the vertical axis. Therefore, let us look at some of the distributions of
primary residences, while the European average remained virtually the variables presented above in the following figures.
unchanged from 2013 to 2017. This is consistent with the adjustment
Figure 8Monthly instalment

Monthly housing instalment (ENG UK) – 2010 Monthly housing instalment (ENG UK) – 2013 Monthly housing instalment (ENG UK) – 2017
.005 .005 .005
.004 .004 .004
.003 .003 .003

Density
Density
Density
.002 .002 .002
.001 .001 .001
0 0 0
0 1000 2000 3000 0 1000 2000 3000 0 1000 2000 3000
Portugal European Union Portugal European Union Portugal European Union
Source: HFCS and authors’ calculations
This first group of diagrams is an example where, by only taking the in Portugal, in nominal terms, we can now observe that the distribu‑
average value into account, we are left with an incomplete analysis. tion of this variable is much less dispersed (that is, very concentrated
Although we have already concluded that the monthly payment for around its average value) in Portugal, particularly in 2010.
the main residence loan was much higher in Europe in general than
Figure 9Effort rate

Effort rate – 2010 Effort rate – 2013 Effort rate – 2017
6 6 6
5 5 5
4 4 4
Density
Density
Density
3 3 3
2 2 2
1 1 1
0 0 0
0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1
With respect to effort rates, there are no significant differences have become more concentrated at lower levels of effort rates for both,
between Portugal and the European Union throughout the period which is most likely a consequence of macroprudential regulations
under analysis. However, we can see that the distribution seems to imposed on mortgage lending, such as higher loan‑to‑value ratios.
Figure 10Amount outstanding

Amount outstanding – 2010 Amount outstanding – 2013 Amount outstanding – 2017
1.2e-05 1.2e-05 1.2e-05
1.0e-05 1.0e-05 1.0e-05
8.0e-06 8.0e-06 8.0e-06

Density
Density
Density
6.0e-06 6.0e-06 6.0e-06
4.0e-06 4.0e-06 4.0e-06
2.0e-06 2.0e-06 2.0e-06
0 0 0
0 100000 200000 300000 400000 500000 0 100000 200000 300000 400000 500000 0 100000 200000 300000 400000 500000
In terms of amounts outstanding, as expected, there is a higher mass The difference between these two is that the former possess illiquid
of loans at lower values, which can come as a result of a number of assets (mainly made up of real estate) above a certain threshold. Table 4
different effects (lower prices and consequently loans, but also lower allows us to look at the proportion of each of these types of households
purchasing power). Nevertheless, it is interesting to see that since in the different European countries in 2010, 2013 and 2017.
2010 there is a trend for the difference in comparison to the EU to
grow, with Portugal's distribution of amounts outstanding being
increasingly concentrated at lower amounts in comparison. Table 4Percentage of ‘hand‑to‑mouth’ households
2010 2013 2017

1.5.2. Household characterisation according to liquid wealth Rich Poor Rich Poor Rich Poor
AT 4.77 8.8 3.66 8.65 4.34 7.94
The economic conditions of households are decisive for us to be
BE 11.38 6.86 10.24 8.84 10.6 9.55
able to anticipate the resilience of the real estate sector to aggregate
CY 31.13 7.36 28.27 10.85 34.57 13.83
shocks. In this sense, it is interesting to look at the financial structure
DE 11.75 9.98 12.33 12.54 12.39 11.99
of household assets according to their degree of liquidity. On the one
EE ‑ ‑ 26.35 8.93 28.43 3.62
hand, if households have little liquid wealth, significant income shocks ES 18.44 6.46 17.65 8.9 ‑ ‑
can lead to a need to liquidate some of the non‑liquid assets in order FI 13.16 9.95 13.16 11.31 13.49 11.9
to make ends meet. Since housing is a household's biggest illiquid FR 8.42 8.95 9.17 10.08 10.97 11.07
asset, this may lead to its forced sale. However, if done systematically GR 31.97 11.61 36.21 14.25 37 13.85
in reaction to an aggregate shock, this may lead to a substantial fall HR ‑ ‑ ‑ ‑ 56.11 3.13
in house prices and trigger a second aggregate shock in the economy. HU ‑ ‑ 30.29 6.23 42.04 8.99
On the other hand, the lack of liquid assets means that households IE ‑ ‑ 18.87 18.31 20.35 13.38
have to adjust their consumption when affected by income shortages, IT 14.57 8.72 13.5 10.05 13.2 9.52
thus generating a fall in aggregate demand and deepening an aggregate LT ‑ ‑ ‑ ‑ 37.94 2.76
negative shock (Corsetti, Duarte and Mann, 2018). LU 9.17 7.56 11.01 8.26 16.7 1.3
LV ‑ ‑ 50.15 15.86 59.82 5.74
Following the approach carried out by Slacalek et al. (2020), we deno‑ MT 6.95 3.57 7.44 2.61 6.48 4.36
minate these households as ‘hand‑to‑mouth’. According to the authors’ NL 11.75 7.81 10.92 8.63 5.06 9.54
definition, these households either have zero liquid assets or depend PL ‑ ‑ 34.75 7.44 29.43 0.8
on an unsecured credit limit. Just like the authors, we also distinguish PT 17.67 7.86 17.16 10.9 16.83 9.39
between ‘wealthy hand‑to‑mouth’ and ‘poor hand‑to‑mouth’ households.
2010 2013 2017 Figure 11Evolution of ‘hand‑to‑mouth’ households in Portugal and Europe
Rich Poor Rich Poor Rich Poor Percentage of ‘Hand to Mouth’ households
SI 37.09 6.96 37.26 11 34.56 8.43 25
SK 24.32 3.22 28.73 5.61 30.53 3.92

European 13.33 8.73 15.93 10.29 16.22 9.61 20
average
15
As of now, there is a marked heterogeneity within Europe itself. 10

Namely, most of the countries that were part of the enlargement of
the European Union in 2004 generally seem to have a higher propor‑ 5
tion of ‘hand‑to‑mouth’ households (that is, with a very limited net
worth and whose consumption patterns have a one‑to‑one rela‑ 0
2010 2013 2017 2010 2013 2017
tionship with net income) than the European average. Portugal
Portugal European Union
appears at the middle of the table, as it does not have as high a percen‑
Wealthy Poor
tage of this type of households as Greece or Hungary, but it is also far
from a lower level such as registered in Austria or the Netherlands. Source: HFCS and authors’ calculations
With around 25% of ‘hand‑to‑mouth’ households, Portugal has not

Before we finish this section, we consider it relevant to make a
shown any change in the total of ‘hand‑to‑mouth’ households over
brief analysis regarding the relationship between liquid assets and
these seven years (varying between 25 and 28%). Furthermore, from
the net wealth of households. The first concept is defined accor‑
2010 to 2013, there seems to have been a change in the proportion
ding to the previously used study (Slacalek et al., 2020): «Sight and
of poor and rich, where the first type has increased, and the second
saving accounts (deposits), directly mutual funds, bonds and stocks.»
has decreased. Then, in 2017 the number of poor ‘hand‑to‑mouth’
The second concept is the general definition of net wealth: assets
came back again. This tells us that there has been a decrease in illiquid
discounted from liabilities. The behaviour of the national averages of
wealth (in which real estate assets are often included) in some of the
these two variables for some of the countries previously referred to is
Portuguese households. Figure 11 shows us precisely that.
presented in figure 12.
Figure 12Net assets and net wealth of European households
Relation between household liquid assets Relation between household liquid assets Relation between household liquid assets
and liquid wealth − 2010 and liquid wealth − 2013 and liquid wealth − 2017
100000 100000 100000
LU LU
BE
80000 80000 80000
BE
LU BE
Liquid assets
Liquid assets
Liquid assets
60000 60000 60000
NL
MT
40000 AT 40000 MT 40000 FI
NL MT FI DE DE
CY AT AT
DE NL
FI FR FR CY FR
IT IT IT
20000 PT
20000 20000 PT CY
PT
GR GR SK
SI SI
SK SI GR
SK
0 0 0
0 300000 600000 900000 0 300000 600000 900000 0 300000 600000 900000
Liquid Wealth Liquid Wealth Liquid Wealth
One can verify that these variables have a positive relationship: coun‑ 1.6. Conclusion
tries where households have higher indebtedness (with lower levels of
In this chapter, we have described the importance of the housing
net wealth) display a lower amount of liquid assets. This trend appears
market within the Portuguese macroeconomic context. In general, we
to be unchanged for most of the countries considered. By combining
firstly concluded that the real estate market contributes to one‑fifth
these graphs with table 2, we see that the countries in the lower‑left
of the Portuguese GDP, and its cycles are strongly correlated with the
corner (Slovakia, Slovenia and Greece) are some of the countries with
business cycles experienced in Portugal over the last two decades.
the highest percentage of ‘hand‑to‑mouth’ households.
Secondly, we concluded that housing is a relevant asset — 80% relative
to total wealth — on the balance sheet of Portuguese households, and
that the composition of the latter is similar to that of the European
Union's average, especially in 2017. However, there is a great deal of
variation between countries within Europe. Portugal is in a group of
countries which has a high percentage of households without liquid
assets (25%), leaving the country more vulnerable to aggregate shocks.
References Iacoviello, M., (2005), «House prices, borrowing constraints, and monetary
policy in the business cycle», American Economic Review 95 (3): 739–764.
Auclert, A., (2019), «Monetary policy and the redistribution channel»,
American Economic Review 109 (6): 2333–67. Jordà, Ò., K. Knoll, D. Kuvshinov, M. Schularick and A.M. Taylor,
(2019), «The rate of return on everything, 1870–2015», The Quarterly Journal
Brinca, P., (2020), «Modern macroeconomics and heterogeneity», Notas of Economics, 134(3), 1225–1298.
Económicas (51), 7–20.
Kaldor, N., (1955), «Alternative theories of distribution», Review of Economic
Castaneda, A., J. Diaz‑Gimenez and J.V. Rios‑Rull, (2003), «Accounting Studies 23 (2), 83–100.
for the US earnings and wealth inequality», Journal of Political Economy
111(4), 818–857. Kaplan, G., B. Moll and G.L. Violante, (2018), «Monetary policy according
to HANK», American Economic Review 108(3), 697–743.
Cloyne, J., C. Ferreira and P. Surico, (2020), «Monetary policy when
households have debt: new evidence on the transmission mechanism», Kaplan, G. and G.L. Violante, (2014), «A model of the consumption res‑
The Review of Economic Studies 87 (1): 102–129. ponse to fiscal stimulus payments», Econometrica 82(4), 1199–1239.
Corsetti, G., J.B. Duarte and S. Mann, (2018), «One money, many marke‑ Krusell, P. and A.A. Smith Jr., (1998). «Income and wealth heterogeneity in
ts‑a factor model approach to monetary policy in the euro area with high the macroeconomy», Journal of Political Economy 106(5), 867–896.
‑frequency identification», Cambridge Working Papers in Economics 1816,
Faculty of Economics, University of Cambridge. Kay, A. and R. Reis, (2016), «The role of automatic stabilizers in the US busi‑
ness cycle», Econometrica 84(1), 141–194.
Dias, D.A. and J.B. Duarte, (2019), «Monetary policy, housing rents, and
inflation dynamics», Journal of Applied Econometrics 34(5), 673–687. Lucas, R.E. Jr., (2003), «Macroeconomic priorities», American Economic
Review, 93(1), 1–14.
Antunes, Marisa (2015), «Imobiliário em Alta. Veja as zonas de Portugal onde
as casas sao mais caras.», Expresso, April 2015, Accessed in July 2020, link Mian, A., A. Sufi and E. Verner, (2017), «Household debt and business
cycles worldwide», The Quarterly Journal of Economics 132 (4): 1755–1817.
Fisher R.A., (1930), The genetical theory of natural selection, Oxford University
Press, Oxford: 1–272. Mian, A., and A. Sufi, (2018), «Finance and business cycles: the credit‑driven
household demand channel», Journal of Economic Perspectives 32 (3): 31–58.
Iacoviello, M. and S. Neri., (2010), «Housing market spillovers: evi‑
dence from an estimated DSGE model», American Economic Journal: Observador, (2017), «Habitação. Preço das casas atinge valores máximos
Macroeconomics 2 (2): 125–64. desde mínimos de 2013», September 2017, accessed in July 2020, link.
Oh, H. and R. Reis., (2012), «Targeted transfers and the fiscal response to the
great recession», Journal of Monetary Economics 59, S50‑S64.
Pasinetti, L.L., (1962), «Rate of profit and income distribution in relation to

the rate of economic growth», The Review of Economic Studies 29(4), 267–279.
Piazzesi, M. and M. Schneider, (2016), «Housing and macroeconomics»,

Vol. 2, in Handbook of macroeconomics, 1547–1640.
Slacalek, J., O. Tristani and G.L. Violante, (2020), «Household balance

sheet channels of monetary policy: a back of the envelope calculation for the
euro area», Journal of Economic Dynamics and Control, 115, 103879.
Chapter 2
Exuberance and contagion in
the Portuguese housing market:
a perspective on disaggregate local
residential prices 1
René Huget, Rita F. Lourenço and Paulo M. M. Rodrigues
2.1. Introduction
According to Stiglitz (1990) «if the reason that the price is high today channel into the economy. These wealth effects are particularly impor‑
is only because investors believe that the selling price will be high tant because housing accounts for much of a household's investment
tomorrow — when “fundamental” factors do not seem to justify portfolio (Englund et al., 2002, Costa et al., 2020 and ECB, 2020). It has
such a price — then a bubble exists.» This definition comprises three also been observed that changes in housing wealth can translate into
dimensions often found in existing descriptions of asset price bubbles: higher consumption levels than with a similar increase in financial
1) assets trade above their fundamental value; 2) investors may act wealth (Carroll, Otsuka and Slacalek, 2011).
speculatively; and 3) a strong price increase is followed by a bust.
Second, house price bubbles may be associated with a substantial
Analysing house price exuberance is important for several reasons. increase in borrowing for capital accumulation due to the positive
First, house price bubbles generate wealth effects that may lead to expectations generated by high prices. Furthermore, as houses can
higher consumption during the boom and lower consumption during serve as collateral, rising prices during a bubble can lead to an increase
the bust phase. This behaviour stems from the permanent income of debt, creating a link to the banking sector. Banks’ balance sheets
hypothesis, relating stronger current consumption to the discounted look safe as long as asset prices are high and back up accumulated
value of expected income rather than to current income, because debt. However, when there is a bust, banks may react with tighter
households prefer a smooth consumption pattern (Friedmann, 1957). credit conditions, leading to a credit crunch which impacts economic
The discounted value of expected income does strongly depend on activity (Bean, 2004).
changes in wealth, and changes in house prices translate through this
Third, the fact that housing supply is not rigid can give rise to further growth and several press publications (as also indicated in chapter 1) call
welfare losses associated to bubbles in housing (Glaeser, Gyourko, for a Portuguese real estate or housing bubble.
Saiz, 2008). Strong positive house price changes can signal profitable
projects in construction, which may lead to oversupply. An example are The objective of this chapter is to answer two important questions:
the ghost cities in Spain which resulted from the housing bubble in the i) have Portuguese house prices deviated from their fundamental
years before 2007. During this period construction overshot housing determinants? and ii) do local level prices (by district and civil parish)
demand and lead to unoccupied and unused housing. A way to deter‑ present bubble‑like developments and are they contagious? These
mine whether an expansion of housing investment may have originated two questions are related in the sense that housing markets are driven
a «housing overhang», is to calculate the difference between current by local determinants. Therefore, a detailed local level analysis may
residential gross fixed capital formation (GFCF) and the average of provide insights on the origins of a nationwide bubble and whether
the previous 20 years (both as a percentage of GDP), and accumulate bubbles migrate.
that difference over a specific period. This indicator may be seen as This chapter contributes to existing literature in several aspects. Firstly,
the cumulative loss that would occur in domestic demand if residen‑ we use quantile regressions to assess whether a possible divergence
tial construction activity returned to its «long‑term» level (Gros, 2007). between house prices and their fundamentals has occurred in Portugal,
Hence, computing this measure for the period between 2000 and 2006, and if current exuberance is found as house price growth persists.
indicates that in 2006 there was an accumulated «excess» of about 30 Secondly, using local level price data for 18 Portuguese districts (and
per cent of houses in Spain, whereas for Portugal the housing stock was corresponding civil parishes), recursively estimated right tailed Dickey
negative (Lourenço and Rodrigues, 2014). ‑Fuller tests are used to detect exuberant behaviour and analyse
Fourth, besides the loss for investors who bought houses at a bubble whether the evidence reflects the aggregate picture of house price
price and sold them after the bust, the social impact of bubbles is also exuberance. Thirdly, possible contagion effects of local level bubbles
important, since a strong increase in the price for housing means that in Portugal are examined. Based on a time varying contagion coeffi‑
shelter (a basic necessity) becomes less affordable for households. cient, proposed by Greenaway‑Mcgrevy and Phillips (2016), we provide
evidence on whether the exuberant behaviour observed in Lisbon and
Studying the Portuguese housing market and examining possible exube‑ Porto has been contagious to their surrounding areas. Evidence suggests
rant behaviour is important as country specific empirical literature on that the contagion effect from the municipality of Lisbon was stronger
housing bubbles is scarce (except for the U.S.) and detailed information in its surrounding civil parishes comparatively to Porto.
for Portugal is even more limited. After the sovereign debt crisis, from
2014 to the present day, Portuguese house prices experienced a strong The remainder of this chapter is structured as follows: Section 2 briefly
reviews related literature on housing bubbles and bubble contagion;
Section 3 analyses the Portuguese housing market dynamics, and where δ = 1/(1 + r), Et is the expectations operator, Pt is the house price
provides results and discussion; Section 4 presents additional reasons in period t and Dt is the rent generated by the house in period t.
for bubbles; and Section 5 concludes.
Under the rational expectations’ assumption, (1) can be solved by
forward substitution,
2.2. Definition of housing bubbles and contagion
The literature on housing bubbles is extensive. In the following Pt = Ptf = ∑i=1

∞ i
δ EtDt+i (2)
section we focus mainly on rational bubbles in housing and on conta‑
Assuming that the transversality condition holds, i.e.,
gion. Works that incorporate plausible special features of the housing
lim δi EtDt+i = 0,
market will not be discussed in detail (see e.g., Head, Lloyd‑Ellis and i→∞
Sun, 2013 for a search model, Mian and Sufi, 2009, Stein, 1995, Glaeser a unique solution and therefore a unique fundamental house price, Ptf,
and Nathanson, 2014, for the role of easy credit, and Poterba, 1984, can be found.
Glaeser et al., 2008, for the role of housing supply and construction
costs). Neither of the models, which are presented in detail, depart In their essence, user cost models (such as Poterba, 1984, 1992 or
from the assumption of rationality (see e.g. Case and Shiller, 2003, Glaeser and Shapiro, 2002) are similar to (2) with Dt generally denoting
for evidence of optimistic expectations, and Piazzesi and Schneider, the benefits of owning a house (Glaeser and Nathanson, 2014). In other
2009, and Guren, 2014, for models with limited rationality). For studies words, Dt is usually associated with rental income (Himmelberg, Mayer
explicitly examining the Portuguese housing market and its drivers and Sinai, 2005) or with the benefits of living in a particular area, such as
after 2007, see, for instance, Lourenço and Rodrigues (2017) and Hill, income possibilities and amenities (Glaeser, Gyourko and Morales and
Lourenço and Rodrigues (2020). Nathanson, 2014, Head, Lloyd‑Ellis and Sun, 2014).
The mathematical foundation of an asset price bubble is that there is

2.2.1. Rational house price bubbles
another solution to (1) which, apart from the fundamental value, Ptf,
Following Cuthbertson (1996), we consider a model with risk neutral contains a bubble term, Bt , such that
and homogeneous agents, rational expectations, informational effi‑
Pt = ∑i=1
∞ i
δ EtDt+i + Bt = Ptf + Bt (3)
ciency (no information asymmetries) and a constant real return rate,
Et Rt+i = r. In this model, the price of a house is provided by the Euler The properties of Bt can be restricted to EtBt+m = Bt /δm (see Cuthbertson,
equation 1996, for an intuitive demonstration). This restricts the bubble beha‑
viour over time to a martingale, meaning that the existence of such
Pt = δ(Et Pt+1 + Et Dt+1) (1)
a term in the house price equation still satisfies the efficient market
hypothesis. For Bt ≠ 0 the Euler equation is satisfied but the transversa‑ channel through which shocks to financial markets are spilled over
lity condition is not because the bubble term is arbitrary. to the housing market corresponds to a time varying discount factor
in (3). Using logit and probit models provides evidence that financial
The rational bubble model just described can be extended to allow for market spill overs can cause explosive behaviour in house prices. This
strictly positive bubbles, which collapse almost surely in finite time contrasts with purely speculative rational bubbles and confirms the
(see, among others, Blanchard, 1979, Evans, 1991, Diba and Grossman, implications of their stylized model.
1988, Taylor and Peel, 1998, and Hall et al. 1999). Froot and Obstfeld
(1991) suggest a different formulation of bubble, one in which the Greenaway‑Mcgrevy and Phillips (2016) analyse contagious behaviour
bubble is tied to the level of dividends, known as intrinsic bubble. between national housing markets in New Zealand's metropolitan
Sola and Driffill (1994) extend Froot and Obstfeld's formulation by areas. To examine the timing of bubble behaviour in metropolitan
allowing for regime switching fundamentals. For further results on areas they use a time varying contagion coefficient. The authors find
regime dependent models see, inter alia, Hall et al. (1999), Funke et al. evidence for Auckland city centre leading the bubble phase and being
(1994), van Norden and Vigfusson (1998), Psaradakis et al. (2001) and contagious for its metropolitan area, as well as for further metropo‑
Nneji, Brooks and Ward, (2013).2 litan areas in the country. Hu and Oxley (2018) used the contagion
coefficient to provide evidence that in Japan, the 1980–1990 asset
2.2.2. House price bubble contagion bubble migrated to the housing market. Deng et al. (2017) conclude
that the house price bubble in China migrated to the housing market
Looking at contagion one may loosely formulate the phenomenon as
between 2005 and 2010. Gomez‑Gonzalez et al. (2017) used the
a bubble migrating from one market to another (Gomez‑Gonzalez et
Greenaway‑Mcgrevy and Phillips (2016) contagion coefficient consi‑
al. 2017, and Deng et al. 2017). Teng, Chang and Chen (2017) employ
dering 20 OECD countries and found evidence of contagion of the US
a state space model on data from the Taipei city centre and suburbs
bubble in a range of countries, although they did not find evidence of
finding a diffusion of bubble prices from the centre to suburbs. After
contagion from the UK bubble.
diffusion, bubbles in suburbs are found to be larger than in the centre.
DeFusco et al. (2012) find that contagion in the US market played
2.3. The dynamics of the Portuguese housing market
a role during the years before 2007, while not finding evidence for
contagion after 2007. 2.3.1. Dynamics of house price determinants
Martinez‑Garcia and Grossmann (2018) analyse how risk‑spread Before analysing the empirical results for Portugal, we first provide
shocks can cause exuberance in house prices even in the absence of a general characterization of the housing market in Portugal. In the
a bubble term in the house price equation. In their framework, the two decades preceding the year of 2007, house prices in Portugal grew
on average less than one per cent per year, in real terms, increasing Given its relevance for the housing sector and the impact it may have
one per cent since the crisis and up to 2019 (see figure 13). They did on the cost of financing it is also important to analyse credit in detail.
however mask a highly differentiated evolution over the last decade, Figure 19, on bank lending, indicates the existence of episodes of very
with house prices declining four per cent between 2008 and 2013, and high growth in loans for house purchases between the mid‑1990s
accelerating more than six per cent over the past five years. Looking and 2007. This annual growth was about 15 per cent on average in
at house prices measured in terms of bank appraisals rather than tran‑ Portugal, in the context of declining costs of bank loans and high and
sactions, and despite the cautiousness of banks in the period following sustained growth in household disposable income, which in which
the crisis, which probably moderated the upward price trend in recent triggered an increase in household indebtedness. The significant dece‑
years, the conclusions are similar. Figure 14 indicates that the number leration of housing credit from 2010 onwards should be seen in the
and value of residential dwelling transactions has increased since context of the international financial crisis, which had a negative
2014, particularly for existing dwellings, which represented 85 per impact on supply due to a significant tightening in lending conditions
cent of the total amount at the end of 2019. The average sale value and housing credit demand.
of a dwelling increased over the last decade and is higher for new
dwellings, especially in the metropolitan area of Lisbon (see figures 15
and 16). The metropolitan area of Porto currently represents 16 per Figure 13House prices and bank appraisals in Portugal
cent of total transactions, while the metropolitan area of Lisbon 120
corresponds to 35 per cent as illustrated in figure 17.

110
Analysing real residential GFCF, we see that the downward trend
initiated in the late 1990s proceeded between 2007 and 2013. 100
2017Q4=100
This class of investment contracted on average 12 per cent per year
compared to the one per cent average fall in GDP per year. Figure 18 90
shows that since 2014 there has been an improvement in GDP and
80
residential GFCF, the latter increasing almost five per cent yearly on
average. As to the conditions of the labour market, we observe in
70
figure 18 that after the crisis unemployment rate rose, reaching a peak 1988:2 1992:4 1997:2 2001:4 2006:2 2010:4 2015:2 2019:4
in 2013, and that there was a significant decline in the labour force, House prices (real) Bank appraisals (real)
probably as a consequence of increased emigration flows and ageing of
Note: Bank evaluations as average values in €/m². House prices based on transactions.
the population. Sources: Statistics Portugal, Banco de Portugal, Confidencial Imobiliário and authors’ calculations.
Figure 14Number of residential dwelling transactions in Portugal Figure 16Average value of a residential dwelling in Lisbon and Porto
50000 350
300
40000
Thousand euros
250
30000 200
150
20000
100
10000
50
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
0 Lisbon MA Lisbon MA – new dwellings
2009:1 2011:4 2014:3 2017:2 2020:1 Porto MA Porto MA – new dwellings
Existing dwellings New dwellings
Note: Average transaction value (in thousand euros). Yearly averages.
Source: Statistics Portugal and authors’ calculations.
Figure 15Average value of a residential dwelling in Portugal Figure 17Number of total residential dwellings transactions by region
50000
200
40000
150
Thousand euros
30000
20000
100
10000
50 0
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2009:1 2011:4 2014:3 2017:2 2020:1
Total dwellings Existing dwellings New dwellings Lisbon MA Porto MA Algarve Other
Note: Average transaction value (in thousand euros). Yearly averages. Note: Other regions include Madeira, Azores, Alentejo, Centre and North of
Source: Statistics Portugal and authors’ calculations. Portugal, excluding the Porto area. MA refers to Metropolitan Area.
Figure 18Economic activity Figure 20Credit market
110 260 9 160
100 220
6 120
90 180
2017Q4=100
2017Q4=100
2017Q4=100
80 140 3 80
%
70 100
0 40
60 60
1994:4 1999:4 2004:4 2009:4 2014:4 2019:4
Real GDP Real income per capita Real residential GFCF (right hand scale) -3 0
1994:4 1999:4 2004:4 2009:4 2014:4 2019:4
Sources: Statistics Portugal, Banco de Portugal and OECD.
Real money market rate Real loans to house purchase (right hand scale)
Figure 19Labour market Sources: ECB, Statistics Portugal and Banco de Portugal.
18 0,03
2.3.2. House price exuberance in Portugal
15 0,02
Ordinary least squares (OLS) regressions are typically used to esti‑
12 0,01 mate the conditional mean of a variable. Quantile regressions (QR) on
the other hand are used to estimate the conditional quantiles of the
yoy,Log
9 0
%
variable of interest. QR, first developed by Koenker and Bassett (1978),

6 -0,01 capture different impacts of predictors depending on the point in the
distribution of the dependent variable.
3 -0,02
The framework to describe asset price exuberance based on QR was

0 -0,03
1994:4 1999:4 2004:4 2009:4 2014:4 2019:4 suggested by Machado and Sousa (2006), who analysed stock price
Unemployment rate Labour force (right hand scale) developments conditional on a set of macroeconomic variables.
Gerdesmeier, Lenarčič and Roffia (2012) and Lourenço and Rodrigues
Sources: Statistics Portugal, Banco de Portugal and OECD.
(2015) apply the method to house prices in OECD countries. In the
QR framework, exuberant behaviour is evident when the observed housing becomes more affordable and thus, demand for housing
asset price exceeds the upper tail of the modelled distribution of the will rise, pushing prices upwards. The second variable included is the
predicted asset price, conditional on a set of regressors (Machado and three‑month money market rate, which captures the effects of credit
Sousa, 2006). The measure for extraordinary high prices should be availability to finance housing (by affecting the mortgage rate for
conditional on the macroeconomic environment, instead of the past dwellings and borrowing rates for institutional real estate investors).
behaviour of house prices. Gerdesmeier, Lenarčič and Roffia (2012) Higher interest rates are expected to drive borrowing costs up and
indicate that boom (exuberance) and bust phases in particular are demand down, leading to a subsequent fall in house prices, and make
known to be non‑linear, which can be captured by the QR framework. alternative applications of wealth more interesting. To capture demo‑
graphic effects that influence housing demand, the labour force is
In practice, one might estimate QR at the median, a high quantile and included as well.
a low quantile of house prices conditional on relevant covariates. The
high and low QR predictions can be interpreted as an upper and lower As we are looking at long run relationships, the supply of housing is
bound, determining the tails of the conditional distribution of house proxied through Residential GFCF. This variable captures investment
prices. When the observed house price moves into these tails, it either in housing, providing a good proxy for changes in housing supply.
signals exuberant behaviour or undervaluation, conditional on the Thus, when residential GFCF rises, the supply of housing rises, putting
selected macroeconomic drivers. downward pressure on house prices.
There is a vast number of studies that analyse the determinants of The house prices considered correspond to hedonic price data of
house prices. The findings in the literature indicate that models which newly built and existing dwellings purchased in the residential market
explain changes in house prices in the long run include a wide set of in Portugal (adjusted for quality measures of housing: square metre
fundamentals such as income, population, employment, borrowing price, size of dwelling and location of dwelling). The aggregate house
costs and returns on alternative assets on the demand side, and the price index is obtained from Statistics Portugal as from 2009 onwards,
availability and cost of land, as well as the cost of construction and and before that from Confidencial Imobiliário (a private and indepen‑
investments in quality improvement for existing housing on the supply dent databank that publishes statistical data based on real transaction
side (Poterba, 1991, Englund and Inoannides 1997, Tsatsaronis and prices). The empirical analysis spans from 1988:Q1–2019:Q3.
Zhu, 2004, Glaeser, Gyourko and Saiz, 2008).
The model we consider is
For Portuguese house prices at the aggregate level, and given the
length of the sample and available data, we consider disposable rHPt(τ│Ft−1) = α0 (τ) + α1(τ)rdict + α2 (τ)rmmit + α3(τ)rgfcft + α4(τ)labourt (4)
income per capita as a main driver of housing demand. If income rises,
where rHPt is the natural logarithm of the real house price index in Figure 21House prices and estimated fundamental quantiles
period t; rdict is the natural logarithm of real disposable income per 5.1
capita in period t; rmmit corresponds to the real 3 month money
5.0
market rate in period t; gfcft is the natural logarithm of residential
GFCF in period t; labourt is the natural logarithm of the labour force in 4.9
period t, and τ is the quantile at which the model is estimated.
4.8
As the variables used in (4) are nonstationary (confirmed by unit
4.7
root tests), quantile cointegration needs to be tested for. Following
Xiao (2009), quantile cointegration testing is performed using a 4.6
CUSUM test on the residuals of (4). Based on the results of this test
the null hypothesis of cointegration cannot be rejected (1.7470 for 4.5
1988:1 1992:3 1997:1 2001:3 2006:1 2010:3 2015:1 2019:3
τ = 0.2, 0.5950 for τ = 0.5, 1.6380 for τ = 0.8 the choice of these quan‑
Q 0.2 Q 0.5 Q 0.8 rHP
tiles is driven by the sample size as these CUSUM type tests in small
samples can become too sensitive when more extreme quantiles are Note: Y Axis denotes natural logarithm of real house prices: predicted quantiles and observed house prices are
plotted, yellow is the 0.2 quantile, green is the median, red is the 0.8 quantile, black is the observed series.
considered).3 Thus, (4) can be seen as a quantile specific long‑run Sources: OECD, ECB, Banco de Portugal and authors’ calculations.
equilibrium.
The persistence in faster growth leads the observed house prices
Figure 21 displays the model's estimation results, plotting the to move further into the upper tail of the estimated distribution.
predicted 0.2, 0.5 and 0.8 quantiles and the natural logarithm of the In 2017:Q4 house prices crossed the 0.8 quantile, marked by the
observed real house price index over time (the 0.2 and 0.8 quantiles are dashed orange line. Given this threshold, house prices are considered
chosen as limits due to the length of the series). One can see that the exuberant from 2017:Q4. Remember this result is conditional on the
estimated quantiles explain a lot of the variation in real house prices chosen macroeconomic fundamentals. Different model specifications
as these evolve between the estimated 0.2 and 0.8 quantiles for most were examined (including GDP per capita, unemployment, mortgage
of the series. However, towards the end of the series, from around rate, and real rents), and the result of house prices becoming exube‑
2016:Q2 onwards, house prices grew faster than the estimated condi‑ rant towards the end of the series is robust. Still, factors that might
tional quantiles. drive house prices above the fundamental benchmark determined
by macroeconomic variables remain (such as investment in tourism
dwellings, foreign direct investment, regulations and other observed or
unobserved economic variables; see discussion in Section 4).
2.3.3. Local level house price exuberance in Portugal More specifically, the aim of the GSADF is to test episodes of mildly
explosive behaviour based on recursively estimated right tailed
The analysis in the previous section has focused on the aggregate
Augmented Dickey Fuller (ADF) tests. The null hypothesis of a unit
real house price index for Portugal, in this section we perform a more
root versus the alternative hypothesis of a mildly explosive root is
detailed analysis based on data by district and civil parish. This is of
tested with the following recursive regression:
importance, since housing markets are known to be highly affected k
by local level variables. The price of housing in Lisbon is likely to Δyt = ar1 , r2 + βr1 , r2 yt−1 + ∑ψ
j=1
j
Δyt−j + ϵt , (5).
r1,r2
move differently from the price of housing in Leiria. Thus, detecting
exuberant behaviour when treating Portugal as one housing market
where yt is the time series of interest, Δ is the first difference operator,
may not be sufficiently informative, hence, local differences are inte‑
ar1 ,r2 is the intercept, βr1 , r2 is the autoregressive coefficient of interest
resting to explore.
(H0: βr1 , r2 = 0), ψ rj 1 , r2 , j = 1, …,k, are the coefficients of the lagged first
To capture these differences, we use local level data obtained from differences, Δyt−j , and ϵt ~i.i.d.(0,σ2r1 , r2 ). The subscripts r1 , r2 , such that
Confidencial Imobiliário, regarding the 18 Portuguese districts and 0 ≤ r1 < r2 ≤ 1, are the subsample start and end points, respectively, for
the 278 Portuguese civil parishes, from 1988:Q1 to 2020:Q1. Due to which the regression is estimated, i.e., r1T + 1, …, r2T .
the lack of time series for determinants at the local level, we apply
Using r1 = 0 and r2 = 1 provides the right tailed ADF test on the full
the univariate GSADF test to detect exuberant behaviour. This test
sample. However, since the power of the right tailed ADF to detect
is consistent with the rational bubble concept briefly described in
bubbles on the full sample is harmed by collapsing bubbles, it may
section 2.1, as the martingale characteristics of asset price bubbles
not detect the bubble even though bubble behaviour is present in
are captured. A drawback, when using the univariate approach is that
one or more subsamples of the time series. Addressing this issue,
explosive behaviour in the series can be caused by explosiveness in
Phillips, Wu and Yu (2011) use the recursive estimation of the ADF
the underlying determinants. When considering a model of rational
test, beginning with the smallest feasible window, starting from r1 = 0.
asset price bubbles with a constant discount factor, the determinant
Keeping the starting point fixed, the ADF is performed with r2, increa‑
is the expected rental income. Sustained explosive growth in expected
sing one observation at a time until r2 = 1. Phillips, Wu and Yu (2011)
rental income would mean periods of consecutive good news, provi‑
then compute the test statistic as the supremum of the set of ADF test
ding information which is better than that of the previous news,
statistics (SADF) obtained by recursive estimation. However, Phillips,
which might lack credibility (Greenaway‑Mcgrevy and Phillips, 2016).
Shi and Yu (2015a, 2015b) show that the SADF is less powerful in case
The test is conservative in the sense that a prolonged but not explo‑
of multiple periodically collapsing bubbles and introduce the General
sive growth of prices will not qualify as a bubble.
SADF (GSADF). The test uses an expanding window structure, which
is similar to that of the SADF, but allows r1 to be flexible. This leads to At first sight, bubble behaviour seems to be correlated across districts.
a set of ADF tests for all feasible subsamples.4 Before 2016, evidence for exuberant behaviour is rather sporadic in
the sample. Beginning in mid‑2016, first evidence for the subsequently
The GSADF test was performed with an initial window of 16 quarters prolonged and widespread exuberant behaviour is recorded for Lisbon
(4 years) and using four lags, supported by the Bayesian Information and Faro. By December 2017, exuberant behaviour is also detected in
Criterion. Critical values are derived by Monte Carlo simulation using four of the 18 districts considered (Braga, Bragança, Évora and Lisbon).
5000 replications. The test results for the districts are reported in
table 5.
Figure 22Evidence of exuberant behaviour in Portuguese districts over time
Aveiro
Table 5GSADF test results for Portugal's 18 districts Beja
Braga
Aveiro 3.4167*** Leiria 2.7588*** Bragança
Beja 3.5416*** Lisboa 2.6555*** Castelo Branco
Coimbra
Braga 2.2654*** Portalegre 2.7131*** Évora
Bragança 2.6060*** Porto 4.6269*** Faro
Guarda
Castelo Branco 2.7939*** Santarém 3.2570*** Leiria
Coimbra 4.5023*** Setúbal 6.2340*** Lisboa
Portalegre
Évora 2.2781*** Viana do Castelo 3.3685*** Port0
Faro 4.7140*** Vila Real 3.1258*** Santarém
Setúbal
Guarda 2.3573*** Viseu 2.7540*** Viana do Castelo
Vila Real
Note: *** denotes significance at the 1% level. Critical values are obtained by Monte Carlo Simulation. Viseu
Sources: Confidencial Imobiliário, Statistics Portugal and author's calculations.
2013 2014 2015 2016 2017 2018 2019
We have noticed that within the period considered, 1988:Q1–2020:Q1, Note: Areas represent the rejection of the null of the GSADF test in favour of mildly explosive
behaviour at the 10% (light grey), 5% (grey) and 1% (black) confidence levels.
evidence of exuberant behaviour is found for all 18 districts under Source: Confidencial Imobiliário and authors’ calculations.
analysis. To get a more systematic picture of the occurring episodes of
exuberant behaviour in recent years, figure 22 plots the timely course
of exuberant episodes in the 18 districts at the usual confidence
levels.5
Figure 23Evidence of exuberant behaviour in the Lisbon area over time Figure 24Evidence of exuberant behaviour in the Porto area over time
TOTAL TOTAL
Alenquer Amarante
Arruda dos Vinhos Baião
Azambuja Felgueiras
Cadaval Gondomar
Cascais Lousada
Lisboa Maia
Loures Marco de Canavezes
Lourihã Matosinhos
Mafra Paços de Ferreira
Oeiras Paredes
Sintra Penafiel
Sobral de Monte Agraço Porto
Torres Vedras Póvoa do Varzim
Vila Franca de Xira Santo Tirso
Amadora Valongo
Odivelas Vila do Conde
2013 2014 2015 2016 2017 2018 2019 Vila Nova de Gaia
Trofa
Note: Areas represent the rejection of the null of the GSADF test in favour of mildly explosive 2013 2014 2015 2016 2017 2018 2019
behaviour at the 10% (light grey), 5% (grey) and 1% (black) confidence level.
Source: Confidencial Imobiliário and authors’ calculations. Note: Areas represent the rejection of the null of the GSADF test in favour of mildly explosive
behaviour at the 10% (light grey), 5% (grey) and 1% (black) confidence level.
Source: Confidencial Imobiliário and authors’ calculations.
Figure 25Test results for exuberant behaviour in Lisbon These results agree with the aggregate analysis in section 3.2., in which
9,0 5,3 exuberant behaviour is detected starting in 2017:Q4, represented
by the dashed orange line in figure 21. Interestingly, despite Lisbon
6,0 5,0 leading the exuberance cycle, exuberance in Lisbon seems to have
ended in 2019:Q2. Also, the bubble in Porto is not evident towards the
Test results
3,0 4,7 end of the period (2018:Q4), while there is strong evidence for exube‑
log
rant behaviour in a wide range of districts.
0 4,4
2.3.4. Bubble contagion

-3,0 4,1
1991:4 1994:1 1996:2 1998:3 2000:4 2003:1 2005:2 2012:1 2014:2 2016:3 2018:4 The results of the previous section point towards the possibility of
Critical value 90% PSY Test – Lisbon municipality HP Lisbon municipality (rhs) contagion, as bubbles are correlated and some districts/civil parishes
Critical value 95% PSY Test – Lisbon district HP Lisbon district (rhs) present exuberant behaviour earlier than others. To analyse bubble
Critical value 99%
contagion, the contagion coefficient proposed by Greenaway‑Mcgrevy
and Phillips (2016) is used and the contagion response before, during
and after the exuberance periods is tracked.
Figure 26Test results for exuberant behaviour in Porto
9,0 5,3 As discussed in section 2.2., bubbles might be subject to contagion in
the sense that a bubble in one market migrates to another. To formally
6,0 5,0 describe this effect over time, Greenaway‑Mcgrevy and Phillips (2016)
develop a time varying contagion coefficient.
Test results
3,0 4,7
log
In a first step, the method estimates autoregressions similar to
0 4,4
equation (5) with a fixed window and no augmentation, leading to
sequences of autoregressive coefficients, {βi,ŝ }Ts=S , for each parish i,
-3,0 4,1
with S being the length of the fixed window, and s being the end of
1991:4 1994:1 1996:2 1998:3 2000:4 2003:1 2005:2 2012:1 2014:2 2016:3 2018:4 the subsample, s = S, …, T. The autoregressive coefficients are obtained
Critical value 90% PSY Test – Porto municipality HP Porto municipality (rhs) by ordinary least squares.
Critical value 95% PSY Test – Porto district HP Porto district (rhs)
Critical value 99%
With the estimated sequences the following regression is estimated, In our analysis, Lisbon and Porto are chosen as core markets because
s they represent the centre of the country's two largest metropolitan
̂ =δ +δ
βi j,s β̂ + ϵ , s = S, …,T (6)
1 j 2 j T − S + 1 core,s−d t areas. In Lisbon's case, the bubble also starts earlier than in any civil
parish. The lag order for civil parish j, possibly behaving contagiously
where j denotes a specific civil parish and core denotes the potential towards the core bubble, is determined by the quarters between first
contagion civil parish from which a bubble spreads, d is a lag para‑ evidence for bubble behaviour in the core and first bubble behaviour
meter for the lagged contagion effect from core to j, and δ2 j is the time in the tested civil parish. For Lisbon as a core market, all 15 surroun‑
varying contagion coefficient. The contagion coefficient is estimated ding civil parishes of the metropolitan area of Lisbon were tested for
from (6) using a local level kernel regression with bubble contagion. Regression (5) was estimated using local level kernel
regressions with local linear fitting. To estimate the autoregressive
∑Ts=SKhs(r)β̃j,sβ̃core,s−d 1 coefficients for the regression, a fixed window of 20 quarters (5 years)
δ̂2j (r ; h, d ) = and β̃j,s:= β̂j,s − ∑T β̂ .
∑T K (r)β̃
2 T − S + 1 s=S j,s was used. Results are presented for selected civil parishes in figure 27.
s=S hs core,s−d
1
Khs(r) = h K ( s/T−r
h
),
The contagion coefficients for Sintra, Mafra, Cascais and Oeiras are
2
where K(.) is the Gaussian Kernel, K(.) = (2π)−1/2e−1/2(.) and h is the positive and increase throughout the considered time period, while
bandwidth parameter; for further details, see Greenaway‑Mcgrevy and Sintra and Mafra present steeper slopes than Cascais and Oeiras. All
Phillips (2016). four civil parishes currently present the strongest contagion coef‑
ficients within the considered period. The contagion effects from
This framework lets the contagion parameter evolve smoothly over
Lisbon to Lourinhã, Odivelas, Amadora, Vila Franca and Loures
time and one can track the change in the contagion coefficient during
increase until reaching their maximum level around mid‑2018, before
bubble periods. The method helps visualize the shape of the conta‑
starting to decrease. Generally, the effects on Lourinhã, Odivelas,
gion impact over time in one civil parish and allows for a comparison
Amadora, Vila Franca and Loures are weaker compared to Sintra,
between them.
Mafra, Cascais and Oeiras.
Figure 27Time varying contagion coefficient from the Lisbon housing Figure 28Time varying contagion coefficient from the Porto housing
market to civil parishes in geographic proximity market to civil parishes in geographic proximity
1.75 0.75
1.40
0.60
1.05
0.45
0.70
0.30
0.35
0.15
0
2013:1 2014:1 2015:1 2016:1 2017:1 2018:1 2019:1 2020:1
Lisboa Bubble Amadora Lourinhã Oeiras Cascais Loures 0
Mafra Vila Franca Odivelas Sintra 2013:1 2014:1 2015:1 2016:1 2017:1 2018:1 2019:1 2020:1
Porto Bubble Gondomar Paços de Ferreira Valongo
Note: The shaded area indicates the Lisbon bubble period evident in section 3.2. Lines are Maia Paredes Vila do Conde
the plotted time varying contagion coefficients δ2j for the respective civil parishes.
Matosinhos Santo Tirso Vila Nova de Gaia
Source: Confidencial Imobiliário and authors’ calculations
Note: The shaded area indicates the Porto bubble period evident in 3.2. Lines are the
In a similar way, Porto is analysed as a core region having a contagious plotted time varying contagion coefficients δ2j for the respective civil parishes.
Source: Confidencial Imobiliário and authors’ calculations.
impact on the 17 civil parishes surrounding its metropolitan area. The
results are displayed for selected civil parishes in figure 28. All conta‑ In general, the results indicate bubble contagion because responses
gion coefficients of the considered civil parishes are positive and are positive during bubble periods. Nevertheless, the results are rather
present a positive slope in their contagion responses towards Porto. descriptive, using the kernel regression specification to track how the
The contagious effect from Porto towards its surroundings is strongest civil parishes’ autoregressive coefficients are related to the respective
in the most recent quarter. core market's autoregressive coefficients over time. To better explain
bubble contagion and commonalities, and differences in the response
Comparing Lisbon and Porto, one observes that contagion appears
functions above, further possible driver variables at the local level
stronger in Lisbon's case than in Porto's.
would need to be included.
The channels for contagion cannot be explored in detail in this work, Figure 29Real house prices, price‑to‑rent and price‑to‑income ratio
but a narrative for the findings is still interesting to provide. The for Portugal
reason for contagion might be that when speculation begins in one 5,2
place, say Lisbon, prices start to increase rapidly, leading first investors 5,1
to search for alternatives and invest in other property markets, where
5,0
they expect the next housing bubble to occur. Prices start to rise in
4,9
these regions, attracting more investors and may originate bubble
behaviour. Geographic proximity is an important channel, since bubble 4,8
behaviour in centres is likely to push people to the suburbs, pushing 4,7
prices up in those same regions. Even though bubble contagion can 4,6
be described and analysed after its occurrence, a general prediction
4,5
for bubble contagion patterns goes against the fundamental rule that
4,4
bubbles cannot be predicted (martingale behaviour). 1988:2 1992:4 1997:2 2001:4 2006:2 2010:4 2015:2 2019:4
rHP P/R P/I

2.4. Possible additional sources of exuberant behaviour
Note: Plotted here are the natural logarithms of real house price index (rHP) and natural
logarithms of Price‑to‑rent ratio (P/R) and Price‑to‑income ratio (P/I), 2015=100.
Exuberance behaviour may not just be a consequence of speculation, Sources: Banco de Portugal, Statistics Portugal and OECD
but it may also be caused by other variables. In this section, we discuss
different drivers for house prices, for which unfortunately we were
2.4.2. Tourism demand
not able to obtain sufficient data.
Apart from the demand variables already mentioned in section 3.2, a
2.4.1. Price‑to‑Rent and Price‑to‑Income ratios further important driver of housing can be tourism and investment
in tourism dwellings, especially in the Portuguese case. Due to online
Common variables considered in the literature are income and rent, platforms fostering access to clients and management for short
capturing affordability, the choice between renting and buying and the ‑term rental, housing investments to accommodate tourists became
return on housing investment. As can be seen in figure 29, the price more attractive for a wider range of investors. This not only increases
‑to‑rent and price‑to‑income ratios have been rising similarly to house demand in the housing market, but also the willingness to pay from an
prices from roughly 2014. Thus, at the aggregate level, it seems that a investment point of view, since short‑term letting to tourists (if the
rather small part of the increase in real house prices can be explained occupation rate is high enough) will lead to a substantially higher
by rising income or rising rents. return than permanent letting to the local population. The tie between
local income and return on housing is broken in this case. The argu‑ These factors present further demand for housing. Housing supply has
ment can be supported, to some extent, by the explosive growth in only grown slowly from 2014 onwards after a prolonged decline, resul‑
the number of firms associated to «furnished dwellings for tourists» ting in upward pressure on prices (see figure 30).
from 2013 and 2014 onwards (see table 6). The data ends before the
detected exuberance and bubble behaviour but it does give an indica‑
tion of the importance of investment in tourism dwellings. Figure 30Housing supply
4000 16000
Table 6Number of firms associated to furnished short‑stay dwellings 3000 12000

for tourists
2009 2000 8000
Short‑stay accommodation 2112

Furnished accomodation for tourists 340 1000 4000
Other short‑stay accommodation 1772
0 0
Source: IES — Banco de Portugal.
1994:4 1992:4 1997:2 2001:4 2006:2 2010:4 2015:2 2019:4
Real residential GFCF in million euro (left hand scale)

2.4.3. Foreign direct investment Building permits, new residential (right hand scale)
Further arguments on drivers for house prices include foreign direct Source: Banco de Portugal, Statistics Portugal.
investment in housing. Following the Golden Residence Permit

Programme, introduced in October 2012, foreign direct investment in 2.4.4. Bank appraisals vs market prices
housing likely accelerated. The regulation allows residence after inves‑
It may also be interesting to have a look at the differences between
ting more than €500.000 in property. The number of resident permits
bank appraisals and market prices. Appraisals provide a safety mecha‑
issued increased from 464 in 2013 to 1526 in 2014. In 2018, the number
nism for banks, as the minimum appraisal value and purchase price
was 1332, and in 2019 it reached 1160 (Serviço de Estrangeiros e
is pledged as collateral. The value pledged as collateral is also used
Fronteiras, 2020). According to Lourenço and Rodrigues (2017), the
for the calculation of the Loan‑to‑Value (LTV), as a key metric for
upward pressure on house prices due to foreign investment may have
risk measurement in lending. In the case of bubbles, LTVs are biased
contributed to contain the decline in house prices following the finan‑
downwards, skewing banks risk profiles. Nakamura (2010) argues that
cial crisis and until the end of 2011.
appraisals are systematically upward biased (above the market price), Together, the presented arguments can explain at least part of the
due to a conflict of interest, as banks hire the appraiser. detected exuberant and bubble behaviour. Unfortunately, due to
data restrictions it is not possible to quantify these impacts. Further
Figure 31 displays average bank evaluations against median sales values disaggregate data would be crucial to explain local differences in
for some important Portuguese housing markets. For Lisbon, average exuberance behaviour. However, it is apparent that the arguments
appraisals are substantially lower than median sales prices towards the leave space for rational speculation as a possible explanation.
end of the series. The picture for Cascais is similar but weaker. This
might be explained by banks’ precautionary behaviour towards the
2.5. Conclusions
strong increase in market prices. For Braga, appraisals are above market
prices throughout the series. Similarly, appraisals for Porto are higher This chapter has presented an analysis of Portuguese house price beha‑
than market prices but come closer together towards the end. viour. For aggregate data, we have conducted a quantile regression
analysis which, conditional on the chosen macroeconomic determi‑
nants, provides evidence for exuberant behaviour starting in 2017:Q4.
Figure 31Bank appraisals vs. market prices
3500
For an in‑depth analysis of the exuberant behaviour detected at the
aggregate level, we resorted to disaggregate data from 18 Portuguese
3000
districts and 278 civil parishes. The General Supremum Augmented
2500 Dickey Fuller test, of Phillips, Shi and Yu (2015a, 2015b) was used
2000
to detect local level exuberant behaviour. Broad based exuberant
behaviour is evident. Timely differences in exuberant behaviour can
1500
also be observed. Furthermore, to explore the question whether the
1000 exuberant behaviour in Lisbon and Porto was contagious for their
surrounding civil parishes, we use the contagion coefficient proposed
500
by Greenaway‑Mcgrevy and Phillips (2016). The results indicate that
0
Lisbon and Porto, although the latter to a lesser extent, seem to have
2011:1 2012:4 2014:3 2016:2 2018:1 2019:4
LisboaBE CascaisBE BragaBE PortoBE

had a contagious effect on surrounding housing markets.
LisboaP CascaisP BragaP PortoP
Clearly, exuberant behaviour in prices can be caused by a range of
Note: Y Axis is in €/m², dashed lines are market prices (P), Median value per m2 of reasons apart from speculative bubbles. Together, local and foreign
dwellings sales, solid lines are average values of Bank evaluations in €/m² (BE).
Source: Statistics Portugal. investment associated to tourism dwelling, foreign direct investment
and shifts in preferences, can explain some but supposedly not all of Clapp, J. and D. Tiritiroglu, (1994), «Positive feedback trading and diffu‑
the exuberant behaviour in the Portuguese house price market. sion of asset price changes: evidence from housing transactions», Journal of
Economic Behavior and Organization 24 (3): 337–355.
Future research on Portuguese local level housing markets is important
to better understand the country's housing market dynamics (as they Cuthbertson, K., (1996), Quantitative financial economics: stocks, bonds and
are known to be driven locally), even though data is less available than foreign exchange, New York: John Wiley and Sons.
in other countries. Also, a framework to analyse channels through
DeFusco, A., W. Ding, F.V. Ferreira and J. Gyourko, (2013), «The role
which bubbles can migrate is of interest to research. On a further level,
of contagion in the last american housing cycle», Working Paper University
it is important to research and improve techniques to detect bubbles
of Pennsylvania.
as well as to conduct research on feasible policy responses.
Dong, Y., E. Girardin, R. Joyeux and S. Shi, (2017), «Did bubbles migrate
References from the stock to the housing market in China between 2005 and 2010?»
Pacific Economic Review 22 (3): 276–292.
Banco de Portugal, (2018), Banco de Portugal: link.
Englund, P., M. Hwang and J.M. Quigley, (2002), «Hedging housing risk»,
Bean, C.R., (2004), «Asset prices, financial instability, and monetary policy», Journal of Real Estate Finance and Economics 24 (1): 167–200.
American Economic Review 94 (2): 14–18.
Evans, G.W., (1991), «Pitfalls in testing for explosive bubbles in asset pri‑
Brunnermeier, M.K., (2009), Bubbles, Edited by Steven N. Durlauf and ces», American Economic Review 81 (4): 922–931.
Laurance E. Blume, Vol. 2, London: Palgrave, Mcmillan.
Friedman, M., (1957), «The permanent income hypothesis», In A theory of the
Carroll, C.D., M. Otsuka, and J. Slacalek, (2011), «How large are housing consumption function, by Milton Friedman, 20–37. Princeton University Press.
and financial wealth effects? A new approach», Journal of Money, Credit and
Banking 43: 56–79. Froot, K.A. and O. Maurice, (1991), «Intrinsic bubbles: the case of stock
prices», American Economic Review 81 (5): 1189–1214.
Case, K. and R. Shiller, (2003), «Is there a bubble in the housing market?»
Brookings Papers on Economic Activity 34 (2): 299–362. Gerdesmeier, D., B. Roffia and A. LLenarčič, (2012), «An alternative
method for identifying booms and busts in the euro area housing market»,
Cho, M. and I.F. Megbolugbe, (1996), «An empirical analysis of property Working Paper 1493 European Central Bank.
appraisal», Journal of Real Estate Economics and Finance 13: 45–55.
Glaeser, E.L. and C.G. Nathanson, (2015), «Housing bubbles», Handbook of
Regional and Urban Economics 5: 701–751.
Glaeser, E.L., J. Gyourko and A. Saiz, (2008), «Housing supply and housing Hu, Y. and L. Oxley, (2018), «Bubble contagion: evidence from Japan's
bubbles», Journal of Urban Economics Volume 64 (2): 198–217. asset price bubble of the 1980–90s», Journal of the Japanese and International
Economies 50: 89–95.
Glaeser, E.L., and J.M. Shapiro, (2003), «The benefits of the home mort‑
gage interest deduction», «Tax policy and the economy» 17: 37–82. Koenker, R. and G. Bassett, (1978), «Regression quantiles», Econometrica
46 (1): 33–50.
Glaeser, E.L., J. Gyourko, E. Morales and C.G. Nathanson, (2014), «Housing
dynamics: an urban approach», Journal of Urban Economics 81: 45–56. Lourenço, R.F., and P. M. M. Rodrigues, (2015), «House prices: bubbles,
exuberance or something else? Evidence from euro area countries», Working
Gomez‑Gonzalez, J.E., J. Gamboa‑Arbeláez, J. Hirs‑Garzón and Papers Banco de Portugal.
A. Pinchao‑Rosero, (2017), «When Bubble meets bubble: contagion in
OECD countries», The Journal of Real Estate Finance and Economics 56: 546–566. Lourenço, R.F., and P.M.M. Rodrigues, (2017), «House prices in Portugal
— what happened since the crisis?», Working Paper Banco de Portugal.
Greenaway‑McGrevy, R. and P.C.B. Phillips, (2016), «Hot property in
New Zealand: empirical evidence of housing bubbles in the metropolitan Martínez‑García, E. and V. Grossman, (2020), «Explosive dynamics
centres», New Zealand Economic Papers 50: 88–113. in house prices? An exploration of financial market spillovers in housing
markets around the world», Journal of International Money and Finance 101.
Gürkaynak, R.S. (2008), «Econometric tests of asset price bubbles: taking
stock», Journal of Economic surveys 22 (1): 166–186. Mian, A. and A. Sufi, (2009), «The consequences of mortgage credit expan‑
sion: evidence from the U.S. mortgage default crisis», The Quarterly Journal of
Head, A., H. lloyd‑Ellis, and H. Sun, (2014), «Search, liquidity, and the Economics 124 (4): 1449–1496.
dynamics of house prices and construction», American Economics Review 104
(4): 1172–1210. Nakamura, L.I., (2010), «How much is that home really worth? Appraisal
bias and house‑price uncertainty», Business Review Federal Reserve Bank of
Himmelberg, C., C. Mayer and T. Sinai, (2005), «Assessing high house Philadelphia 11–22.
prices: bubbles, fundamentals and misperceptions», Journal of Economic
Perspectives 19 (4): 67–92. Nneji, O., C. Brooks and C.W.R. Ward, (2013), «Intrinsic and rational
speculative bubbles in the U.S. housing market: 1960–2011», Journal of Real
Homm, U. and J. Breitung, (2012), «Testing for speculative bubbles Estate Research (35): 121–151.
in stock markets: a comparison of alternative methods», Journal of Financial
Econometrics 10 (1): 198–231. Phillips, P.C.B., S.P. Shi and J. Yu, (2015a), «Testing for multiple bubbles:
historical episodes of exuberance and collapse in the S&P 500», International
Economic Review 1043–1078.
Phillips, P.C.B., S.P. Yu and J. Shi, (2015b), «Testing for multiple bubbles: limit Appendix A: the general supremum
theory of real‑time detectors», International Economic Review 56 (4): 1079–1134. augmented Dickey Fuller test (GSADF)
Phillips, P.C. B. and T. Magdalinos, (2007), «Limit theory for moderate The GSADF test statistic is given by
deviations from a unit root», Journal of Econometrics 136 (1): 115–130.
GSADF(r0) = sup sup ADFrr12 (B.1)
r1ϵ[0,r2−r0] r2ϵ[0,r0−r1]
Phillips, P.C. B., Y. Wu, and J. Yu, (2011), «Explosive behavior in the 1990s
NASDAQ: when did exuberance escalate asset values?», International
in which r0 is the smallest feasible window chosen. Under the null
Economic Review 52 (1): 201–226.
hypothesis, the limit distribution of GSADF(r0) is
Piazzesi, M. and M. Schneider, (2009), «Momentum traders in the housing 1 r [W(r )2 − W(r )2 − r ] − ∫ r2W(r)dr[W(r ) − W(r )]
market: survey evidence and a search model», American Economic Review sup 2 w 2 1 w r1 2 1
(B.2)
r1ϵ[0,r2−r0], r2ϵ[0,r0−r1] 2 1/2
99 (2): 406–411. rw rw ∫r1 W(r) dr − ∫r1 W(r)dr
1/2 r2 2 r2
Poterba, J.M., (1984), «Tax subsidies to owner‑occupied housing:

an asset‑market approach», The Quarterly Journal of Economics 99 (4): 729–752.
where W is a standard Wiener process and rw = r2 − r1 is the size of the
window.
Poterba, J.M., (1992), «Taxation and housing: old questions, new answers»,
American Economic Review 82 (2): 237–242. If GSADF(r0) is larger than the critical value from the limit distribution,
the null hypothesis of a unit root is rejected against the alternative
Serviço de Estrangeiros e Fronteiras, (2020), Serviço de Estrangeiros hypothesis of mildly explosive behaviour.
e Fronteiras: link.
Stein, J.C., (1995), «Prices and trading volume in the housing market: a model
with down‑payment effects», Quarterly Journal of Economics 110: 379–406.
Teng, H.J., C.O. Chang and M.C. Chen, (2017), «Housing bubble contagion
from city», Urban Studies 54 (6): 1463–1481.
World Travel Awards, (2019), World Travel Awards: link.
Xiao, Z., (2009), «Quantile cointegrating regression», Journal of Econometrics

150: 248–260.
Appendix B: quantile cointegration test and results Table B. 1Test results of ADFGLS and Phillips‑Perron
To interpret (6) as a long run relationship for each quantile, after confir‑ ADFGLS test Phillips Perron test
ming nonstationarity, (6) needs to be tested for quantile cointegration. Variable test statistic test statistic
Following Xiao (2009), quantile cointegration can be tested as follows. Z(t) Z(rho)
rHP ‑1.148 ‑1.301 ‑3.595
Looking at ψτ(u) = τ − I(u < 0) and the residual from the estimated quan‑
rgdpc 1.209 ‑2.942 ‑5.954
tile regression: εtτ = yt − Qyt(τ│Ft) = yt − Θ(τ)'Zt = εt − Fε−1(τ), one has Qεtτ(τ) = 0,
rdic ‑1.181 ‑3.087 ‑6.930
where Qεtτ(τ) is the τ‑th quantile of εtτ and Eψτ(εtτ ) = 0. Xiao (2009)
rmmi ‑2.540 ‑2.651 ‑13.122
proposes to test for cointegration by testing the stability of the process
rgfcf ‑1.505 ‑0.523 ‑0.777
of εtτ. Considering the partial sum process such as Yn(r) = ω*1ψ n ∑j=1 [nr]
ψt(εjτ),
labour ‑0.957 ‑0.862 ‑1.463
where ωψ denotes the long run variance of ψτ(εjτ), under appropriate
*2
rmrti ‑1.700 ‑2.327 ‑11.003
assumptions, that process follows an invariance principle and weakly
rrent 0.602 ‑3.147 ‑7.723
converges toward a standard Brownian motion W(r). Choosing a conti‑
unemp ‑2.008 ‑1.140 ‑2.564
nuous functional h(.) measuring the fluctuation of Yn(r), allows for a
Confidence level 10% 5% 1% 10% 5% 1%
robust cointegration test to be examined by use of h(Yn(r)). By the conti‑
Critical values ‑2.64 ‑2.93 ‑3.46 Z(t) ‑3.14 ‑3.44 ‑4.03
nuous mapping theorem and under regularity conditions and the null
Z(rho) ‑17.58 ‑20.80 ‑27.57
hypothesis of cointegration h(Yn (r)) h(W(r)). Under the alternative
hypothesis, the statistic diverges to ∞. Note: rdic, rmmi, rgfcf, labour are defined as in 5.1. rgdpc corresponds to the logarithm of real GDP per capita,
rmrti is the real mortgage rate, rrent is the logarithm of real rent index and unemp is the unemployment rate.
The ADFGLS test uses the Perron‑Qu method, and it is tested down from 12 lags. A trend is always included
If a quantile specific cointegrating relationship is present, the residuals except for rHP (the asymptotic p‑value of the ADFGLS without including a trend for rHP is 0.229 and the
critical values for the Phillips‑Perron test when not including a trend are –2.578, 2.888 and –3.501 for the 10%,
of the quantile regression only present fluctuations around a long 5% and 1% levels, respectively). As the null of a unit root is not rejected for all series (except for rrent at the
‑term equilibrium, reflected by a stable process of εtτ. 10% level using Phillips‑Perron Z(t)), the series are nonstationary and qualify for testing quantile cointegration.
Sources: OECD, ECB, Banco de Portugal and authors’ calculations.
Chapter 3
Synchronism in Portuguese
housing
Vera Gouveia Barros, Luís Aguiar‑Conraria, Maria Joana Soares
3.1. Introduction
In recent years, especially after the financial crisis, several advanced wealth accumulation, long‑term investment, and collateral for lending.
economies saw their house prices increasing (IMF, 2018). According to Correspondingly, there are two housing markets with blurred boun‑
OECD (2019), the gap between house prices and consumer price index daries: one for the investment good (housing stock) and the other
inflation widened in the early 2000s, lessened during the global finan‑ for the consumer good (housing services) (Arnott, 1987; Coakley,
cial crises, and then broadened after that. On average, house prices 1994; Fallis, 1985; Maclennan, 1979; Meen, 1996; Smith et al., 1988).
have grown twice as fast as inflation in the last two decades; and have Therefore, housing demand depends on permanent income, wealth,
also grown more quickly than median incomes. credit constraints, portfolio decisions, and financial‑market dynamics.
Thus, expectations about future price movements, risk diversification,
In this chapter, we are concerned with the dynamics of housing and return rates on alternative investments are important (Adair et al.,
prices in Portugal. To be more precise, we first study the housing 1999; Coakley, 1994; Fallis, 1985; Whitehead, 1999; Yang et al., 2018).
price synchronisation between Portugal and other countries, and
detect which international markets are more interconnected with the Between 1950 and 2015, housing assets in several advanced econo‑
Portuguese market. Second, we perform a similar analysis between mies displayed an average annual real return similar to that of equity
Portuguese cities. We will rely on wavelet analysis, which will allow investments, between 5 percent and 8 percent, but with a lower stan‑
us to examine the evolution and find relationships at different dard deviation (IMF, 2018; Jordà et al., 2017, 2019). Consequently, real
frequencies. estate has become more significant as store‑of‑value and housing
‑based wealth achieved records (Fernandez and Aalbers, 2016).
Housing has some peculiar characteristics, such as its durability.
A house satisfies the basic need for shelter, security, personal space, Another notable feature of housing is its spatial fixity. Gotham
and privacy, but being a durable good, it is also an instrument of (2009) defines spatial fixity as «a condition of non‑exchangeability,
non‑transferability, immobility, illiquidity and long turnover times started falling in 2006, causing the 2008 subprime mortgage crisis,
between buying and selling», which is why we traditionally look at which evolved into a global economic crisis and caused a worldwide
housing markets as local markets (Aalbers, 2016; Bardhan and Kroll, collapse of housing prices (Flor and Klarl, 2017; Hirata et al., 2012;
2007; Case et al., 2000; de Bandt et al., 2010; Ley, 2015; Pow, 2017; Kuang and Wang, 2018; Muellbauer and Murphy, 2008; Ryczkowski,
Vansteenkiste and Hiebert, 2011). However, real estate knows geogra‑ 2019). In recent years, housing has experienced a rise in its price throu‑
phic substitutability as an asset for investment and, actually, the housing ghout many countries. These developments have generated a research
market has increasingly become an international market (Büdenbender agenda on the co‑movement of housing prices across countries or
and Golubchikov, 2017; Clerc, 2019; Fields, 2015; IMF, 2018; Ley, 2015; cities (Beltratti and Morana, 2010; Gupta et al., 2015; Hoesli, 2020; IMF,
Pow, 2017; Rogers and Koh, 2017; Ronald and Dewilde, 2017). 2018, 2019; Miles, 2017).
The existence of global investors in housing markets is associated with Demand and supply determine housing prices. Therefore, in the
increased housing price synchronisation, especially at the city level. long run, housing prices should reflect market fundamentals. Several
Studying price synchronisation between different housing markets is, studies relate housing prices to household (permanent) income and
therefore, a very relevant matter. It suggests that global investors and wealth, mortgage interest rates, financial conditions, leverage, costs
global financial conditions may influence local housing price dyna‑ of land acquisition and construction, tax and other policies, and struc‑
mics, which has implications for the effectiveness of a range of policy tural factors as demographics or urbanization (Duca, 2020; Geng, 2018;
tools to address imbalances in the housing market, namely affordabi‑ Girouard et al., 2006; Gupta et al., 2015; Igan and Loungani, 2012; IMF,
lity problems (Alter et al., 2018; Duca, 2020; Hoesli, 2020; IMF, 2018; 2018, 2019; Kishor and Marfatia, 2017; Marfatia, 2018). Housing price
Katagiri, 2018). synchronisation may, therefore be the result of the co‑movement in
economic fundamentals, with either real macroeconomic or financial
3.2. Literature review variables, across countries and regions (de Bandt et al., 2010; Duca,
2020; IMF, 2018; Miles, 2017; Terrones and Otrok, 2004; Vansteenkiste
As an autonomous and separate area of study, housing economics
and Hiebert, 2011).
progressed in the 1970s and has grown significantly since then and
covered a wide range of topics (Arnott, 1987; Maclennan, 1979; In industrial countries, real housing prices are procyclical; reflecting
O'sullivan and Gibb, 2003; Smith et al., 1988). household's disposable income and employment prospects, they rise in
a boom and fall in a recession (Davis and Nieuwerburgh, 2015; Hwang
In the vast majority of OECD countries, house prices increased sharply
and Quigley, 2006; IMF, 2018; Leung, 2004; Terrones and Otrok, 2004).
between the mid‑1990s and the early 2000s (Girouard et al., 2006;
Between 1970 and 2000, for the OECD as a whole, business‑cycle and
Igan and Loungani, 2012; Miles, 2017). In the United States, they
house‑price turning points have roughly coincided (Girouard et al.,
2006). Some authors have documented an increase in business cycle transmit financial conditions across various countries, so that global
synchronisation (Aguiar‑Conraria et al., 2017; Aguiar‑Conraria and factors significantly explain them (Baskaya et al., 2017; Bruno and Shin,
Soares, 2011b; Belke et al., 2017; Bordo and Helbling, 2003; Fidrmuc et 2013, 2015a, 2015b; Caceres et al., 2016; Calvo et al., 1996; IMF, 2016;
al., 2012; Kose et al., 2008, 2012). Miranda‑Agrippino and Rey, 2020; Rey, 2015).
User cost of capital is another driver of demand and higher housing Furthermore, for some housing markets, cross‑border transactions
prices. We usually define the real user cost of capital as the after appear to explain housing price synchronisation (Alter et al., 2018;
‑tax mortgage nominal interest rate plus property taxes, insurance Badarinza and Ramadorai, 2018; Duca, 2020; Hekwolter of Hekhuis
costs, and physical depreciation minus expected house price apprecia‑ et al., 2017; IMF, 2018; Katagiri, 2018). Despite this conceptual
tion (Duca, 2020; Duca et al., 2011; Meen, 1990, 1996; Muellbauer and framework, housing price synchronisation, which refers to the more
Murphy, 2008; Poterba, 1984). Investors’ projections concerning house significant correlation or co‑movement of house prices across geogra‑
price appreciation in one country may be revised given expectations phic entities, is mainly an empirical question. It is the first step to
regarding other countries’ house prices, so synchronicity may be a assess the degree of synchronisation (Gupta et al., 2015; Hoesli, 2020).
consequence of harmonized prospects in the development of housing
markets across several countries (Beltratti and Morana, 2010; de Bandt The co‑movement of housing prices has been investigated at a regional
et al., 2010; IMF, 2018; Vansteenkiste and Hiebert, 2011). level, for which there is established empirical literature, concentrated
mainly in the United Kingdom. Within the framework of cointegra‑
On the other hand, co‑movement in housing prices may also be the tion, several authors have studied the structure of regional house
result of simultaneous changes in financial factors: global finan‑ prices within the United Kingdom, finding the existence of long run
cial conditions, portfolio channels, and expectations contribute to relationships between regional house prices and the ripple effect; that
housing price synchronisation (Alter et al., 2018; Claessens et al., 2011; is, the notion of a causal link between house prices in London and
Helbling and Terrones, 2003; IMF, 2018; Terrones and Otrok, 2004). the South East, and those in other regions (Alexander and Barrow,
Financial conditions refer to the costs, conditions, and availability of 1994; Cook, 2003, 2005; Giussani and Hadjimatheou, 1991; Holmes,
funds to the economy (IMF, 2017). They are an influential variable on 2007; Holmes and Grimes, 2008; MacDonald and Taylor, 1993; Meen,
housing prices (Agnello and Schuknecht, 2011; Baffoe‑Bonnie, 1998; 1999; D. Zhang, 2010); the same conclusion has been reached using
Cesa‑Bianchi et al., 2015; Englund and Ioannides, 1997; Favara and other methods and econometric approaches (Cameron et al., 2006;
Imbs, 2015). According to Del Negro et al. (2019), trends in real inte‑ Cascio, 2020; Cook and Thomas, 2003; Holly et al., 2011; Hudson et al.,
rest rates across advanced economies have converged over the past 2018; Kyriazakou and Panagiotidis, 2014; Morley and Thomas, 2011;
three decades, and country‑specific trends have all but vanished since Muellbauer and Murphy, 1994; Tsai, 2015). However, evidence on the
the 1970s. Other several studies have shown that different channels existence of regional house price convergence is mixed, and some
papers have failed to support the ripple effect (Abbott and de Vita, Terrones, 2003; Hirata et al., 2012; Hoesli, 2020; IMF, 2018; Katagiri,
2013; Ashworth and Parker, 1997; Chowdhury and Maclennan, 2014; 2018; Otrok and Terrones, 2005; Terrones and Otrok, 2004). For the euro
Drake, 1995; Hamnett, 1989; Rosenthal, 1986). area in particular, a different result has been attained by several authors
(Álvarez et al., 2010; Gupta et al., 2015; Henley and Morley, 2000; Miles,
From the late 1990s onwards, a few studies examining the interlinkages 2017, 2019; Vansteenkiste and Hiebert, 2011).
between regional and local housing markets in the United States, and
focusing on the general issue of price diffusion, have been undertaken,
3.3. Data and methods
also with mixed evidence (Apergis and Payne, 2012; Barros et al., 2012;
Canarella et al., 2012; Clapp and Tirtiroglu, 1994; Clark and Coggin, 2009; 3.3.1. Data
Flor and Klarl, 2017; Gupta and Miller, 2012a, 2012b; Holly et al., 2010;
To perform a cross‑country analysis, studying housing price synchro‑
Holmes et al., 2011; Kim and Rous, 2012; Payne, 2012; Pollakowski and
nisation between Portugal and other countries, and detecting which
Ray, 1997; Vansteenkiste, 2007; Zohrabyan et al., 2008).
international markets are more interconnected with the Portuguese
Housing price co‑movement has also been studied within several other market, we need a dataset on house prices that is as homogeneous
countries, including Australia (Bangura and Lee, 2020; Churchill et al., as possible. The Bank for International Settlements (BIS) publishes
2018; Liu et al., 2008; Luo et al., 2007; Tu, 2000), Canada (Grigoryeva statistics on residential property prices for several countries around
and Ley, 2019), China (Chiang, 2014; Huang, Li et al., 2010; Huang, the world. These derive their selected series from the detailed data
Zhou, et al., 2010; Weng and Gong, 2017; Zhang and Morley, 2014; set, which is harmonized according to the Handbook on Residential
Zhang et al., 2017), Finland (Oikarinen, 2005), Ireland (Stevenson, Property Prices Indices (Eurostat, 2013). They offer the closest indi‑
2004), Malaysia (Lean and Smyth, 2013), South Africa (Balcilar et cator to nationwide coverage for each jurisdiction and typically
al., 2013; Burger and Van Rensburg, 2008), Spain (Blanco et al., 2016; include all types of dwellings (new and existing). We opted for the
Larraz‑Iribas et al., 2008; Taltavull de La Paz et al., 2017), Taiwan Real Residential Property Price Index (2010=100), computed by defla‑
(Chen et al., 2011; Chien, 2010; Lee and Chien, 2011; Lee et al., 2014), ting the nominal residential property price series with the Consumer
and The Netherlands (Klarl, 2018; Teye et al., 2017). Price Index. The data set has a quarterly frequency.
Several other studies investigate housing market spillovers across coun‑ For Portugal, as the data from BIS only refers to the period from 2008
tries, namely industrial countries and advanced economies, provide onwards, we use the series obtained from Confidencial Imobiliário, starting
evidence that housing prices were often synchronized and the degree of in 1988, instead. This data set comprises the eighteen district capitals,
synchronisation has been rising (Alter et al., 2018; Beltratti and Morana, covering the period 1988:Q1–2020:Q1. Computing a simple average
2010; Case et al., 1999; de Bandt et al., 2010; Gros, 2006; Helbling and for those cities, we have determined a correlation coefficient of 0,94
with the BIS series. Therefore, in our cross‑country analysis, we have This chapter will rely on four main tools: the wavelet power spectrum,
employed this constructed Portuguese housing price index. Our sample the wavelet coherency, the wavelet phase‑difference, and the wavelet
includes eleven other European Union Member States. Data from 1988 de‑synchronisation matrix.
was available for Belgium, Denmark, Finland, France, Germany, Ireland,
Italy, Netherlands, Spain, Sweden, and the United Kingdom. A wavelet is a function with mean zero (implying that it has to wiggle
up and down) and well‑localized in time (e.g., shows fast decay), beha‑
We also used the data set from Confidencial Imobiliário to analyse ving like a small wave that loses its strength as it moves away from the
co‑movements between housing prices in Portuguese cities. centre, hence the choice of the term wavelet, meaning little wave. Here
we use the most popular wavelet, the Morlet wavelet. Aguiar‑Conraria
3.3.2. Empirical strategy: wavelet analysis and Soares (2014) show that this specific wavelet has optimal features
to study oscillations.
Fourier spectral analysis can be (and has been) used to determine
which frequencies play a predominant role in explaining the overall The continuous wavelet transform of a time series x(t) with respect to
time‑series variance. This analysis decomposes the observed pattern the wavelet is a function of two variables,
over time into a spectrum of cycles with different lengths. Similarly to
a prism, which decomposes light into a spectrum of colours of diffe‑ Wx(τ,s): Wx (τ,s) = ∫−∞
∞
x(t) 1s φ̅ t−τ
s dt.
rent frequencies. The term white noise, used to describe independent
The bar denotes complex conjugation, s is a scaling or dilation factor
and identically distributed variables, comes from the fact that such a
that controls the wavelet's width, and τ is a translation parameter
process has a flat power spectral density, which is a characteristic of
which controls the wavelet's location.
white light.
In analogy with the terminology used in the Fourier case, the (local)
The main advantage of wavelet analysis is that we estimate time
wavelet power spectrum is defined as WPSx(τ,s) = |Wx(τ,s)|2.
‑series spectral characteristics as a function of time, revealing how
its different periodic components change over time. To access the It gives us a measure of the time series variance distribution in the
synchronicity of housing price cycles, we will rely on the continuous time‑frequency plane. This tool, as with the Fourier power spectrum,
wavelet transform and several wavelet tools associated with it. We allows us to describe each frequency's contribution to the overall
refer the reader to Aguiar‑Conraria et al. (2012, 2013) for an intuitive variance. Additionally, we can identify when that contribution is at its
introduction to this technique and its applications to social science largest. In our pictures, the regions where the power spectrum, and
data. For detailed technical treatment, the reader may consult Aguiar hence the volatility, is larger, are depicted with lighter colours.
‑Conraria and Soares (2014).
The concepts of cross wavelet power, wavelet coherency and phase Finally, we will also measure the dissimilarities between the wavelet
‑difference enable us to deal with relations between two time series. transform of two time‑series proposed by Aguiar‑Conraria and Soares
The cross‑wavelet transform of two time‑series, x(t) and y(t), is defined (2011a). Several authors have successfully applied this tool to study
as Wxy(τ,s) = Wx(τ,s)Wy̅ (τ,s). The cross‑wavelet power of two time cycle synchronisation. The most relevant is Flor and Klarl (2017), who
‑series, |Wx(τ,s)|, depicts the local covariance. Compared with the also study the synchronisation of house prices (in metropolitan areas
cross‑wavelet power, the wavelet coherency, a concept akin to the within the United States).
correlation coefficient, has the advantage of being normalized by the
power spectrum of the two time‑series. In analogy with the concept of The closer to zero our measure of distance is, the more similar the
coherency used in the Fourier analysis, given two time‑series x(t) and wavelet transforms of x(t) and y(t). To be more precise, a value close to
y(t), one defines their wavelet coherency: zero means the two regions have a very similar wavelet transform. In
turn, this implies that the contribution of house price cycles at each
S Wxx(τ,s) frequency towards the total variance is similar in both regions; this
Rxy(τ,s) = , contribution happens simultaneously. Finally, the ups and downs of
S Wxx(τ,s) S (Wyy(τ,s)
each cycle coincide.
where S denotes a smoothing operator in both time and scale.

3.4. Results
Using a complex‑valued wavelet, we can compute the phase of each In this section, we present our results. We begin by estimating the
series’ wavelet transform, and thus obtain information about the Portuguese Housing Price Index's wavelet power spectrum, to get a
possible delays of the two series’ oscillations as a function of time sense of the dominant cycles we can identify in this data.
and frequency by computing the phases and the phase difference,
also known as the phase angle. A phase difference of zero indi‑ In subsection 4.2, we analyse how the Portuguese index is synchro‑
cates that the time series move together at the specified frequency. nized with a sample of other EU countries (including the UK). We first
A phase‑difference between 0 and −π/2 indicates that the two series estimate the «distance» between the wavelet transform of Portuguese
move in‑phase (positive correlation), with x leading y; while if the prices and other countries’ prices — one pair at a time. To assess if
phase‑difference is between 0 and −π/2, then it is y that is leading. synchronisation is statistically significant, we rely on Monte Carlo simu‑
A phase‑difference between π/2 and π indicates that the series move lations, considering two independent time series as the null hypothesis.
out‑of‑phase (negative correlation), with x lagging y; while if the
Then, we perform a more detailed analysis by estimating the wavelet
phase‑difference is between −π and −π/2, then it is y that is lagging.
coherency and the phase‑difference between Portuguese prices and
the prices of a sample of selected countries. For statistical significance,
we rely on Monte Carlo simulations. The interpretation of our econo‑ Figure 32Time series plot of House Price Index (on top). Wavelet power
metric results proceeds as follows: first, we identify the statistically spectrum (at the bottom).
significant time‑frequency regions, meaning that, in those episodes, House Price Index Portugal
we may confidently say that there has been a considerable co‑move‑ 20
ment of the variables for cycles within the indicated period; then, we
analyse the phase differences to detect whether the co‑movement 10
has been positive or negative and which countries were leading and
lagging for the statistically significant locations. 0
In section 4.3, the analysis is similar, but instead of analysing synchro‑ -10
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2016 2018 2020
nisation between countries, we do it between Portuguese cities.
Wavelet Power Sectrum
3.4.1. The wavelet power spectrum 1
In figure 32, on top we can see the Portuguese House Price Index (we
consider the year‑on‑year growth rate). At the bottom, we have the 2
wavelet power spectrum. Lighter regions correspond to areas of high
wavelet power (high volatility). We have the date on the horizontal
Period (years)
axis, and the cycle period on the vertical axis. They range from 1‑year 4
cycles (on top) to 16‑year cycles (at the bottom). The white stripes
correspond to the wavelet power spectrum's local maxima, providing
an estimation for the period of the dominant cycles. The thick black 8
contour designates the (5%) statistically significant regions.
16
1992 1996 2000 2004 2008 2012 2016 2020
NoteThe thick grey contour designates the 5% significance level. The shade code for
power ranges from black (low power) to white (high power). The cone of influence, which
indicates the area affected by edge effects, is the region outside the dashed line.
We can identify two primary cycles in the data which almost run the Portugal is almost synchronized with the housing market in the UK.
whole sample, and superimpose each other: one with a period of appro‑ Italy, Spain, and France come next. In figure 33, we have the estimated
ximately nine years and another of about 14 years. Note that these coherencies between Portugal and the first four countries. Wavelet
primary cycles are statistically significant. A shorter run cycle, an almost coherency is akin to the absolute value of the correlation coefficient
4‑year period, shows up around the 2008–2012 period, corresponding to but in the time‑frequency plane. Lighter regions correspond to areas
the international financial crisis and the subsequent sovereign debt crisis of high coherency. On the right, we have the phase difference for two
— although it is not significant at 5%, it is at 10% (not shown). frequency bands. The first frequency band, 1~8‑years, corresponds
to typical business cycle frequencies. The second band, 8~16‑years,
3.4.2. Synchronisation between Portugal and other countries corresponds to longer and more structural cycles.
In table 7, we show the dissimilarity between the Portuguese housing The country which has the largest regions of statistically signifi‑
cycles and several other countries. A value very close to zero means cant coherency with Portugal is Spain. Until the early 2000s, there
that the two countries have a very similar wavelet transform; this is a region of high coherency between the 4 and 8‑year frequencies.
implies that they share the same high‑power regions and their phases The corresponding phase difference is between −π and 0, meaning
are aligned. Intuitively, this means that the contribution of cycles at the correlation is positive, with Spanish prices leading the Portuguese
each frequency to the total variance is similar between both countries. prices. After that, we can see regions of high coherencies both at
This contribution happens simultaneously, and the ups and downs of higher and lower frequencies. In both, the phase difference indicates
each cycle co‑occur. It is in this sense that we say the two countries that Spain is the leader and Portugal the follower. The behaviour of
have synchronized cycles. house prices in Portugal and Spain is investigated by Lourenço and
Rodrigues (2014). Using three different approaches, these authors
draw some conclusions regarding the dynamics and contrast of house
Table 7De‑synchronisation between Portugal and other countries
prices in both countries.
Portugal Portugal
United Kingdom 0.17 Germany 0.28 1%
Legend
Italy 0.20 Denmark 0.35 5%
Spain 0.21 Ireland 0.35
France 0.23 Netherlands 0.39
Belgium 0.24 Finland 0.43
Sweden 0.28
Figure 33Cross‑wavelet coherency between Portugal and other countries
(on the left). Phase‑difference (on the right).
WCO(Portugal, United Kingdom) Phase-Differences WCO(Portugal, Spain) Phase-Differences
1 π 1 π
1~8 freq. band

1~8 freq. band
π/2 π/2
0 0
2 2 -π2
-π2
-π -π
Period (years)
Period (years)
1992
1996
2000
2004
2008
2012
2016
2020
1992
1996
2000
2004
2008
2012
2016
2020
4 4
π π
8~16 freq. band

8~16 freq. band
8 π/2 8 π/2
0 0
-π2 -π2
16 -π 16 -π
1992
1992
1996
2000
2004
2008
2012
2016
2020
1996
2000
2004
2008
2012
2016
2020
1992
1992
1996
2000
2004
2008
2012
2016
2020
1996
2000
2004
2008
2012
2016
2020
WCO(Portugal, Italy) Phase-Differences WCO(Portugal, France) Phase-Differences
1 π 1 π
1~8 freq. band

1~8 freq. band
π/2 π/2
0 0
2 2 -π2
-π2
-π -π
Period (years)
Period (years)
1992
1996
2000
2004
2008
2012
2016
2020
1992
1996
2000
2004
2008
2012
2016
2020
4 4
π π
8~16 freq. band

8 8~16 freq. band π/2 8 π/2
0 0
-π2 -π2
16 -π 16 -π
1992
1992
1996
2000
2004
2008
2012
2016
2020
1996
2000
2004
2008
2012
2016
2020
1992
1992
1996
2000
2004
2008
2012
2016
2020
1996
2000
2004
2008
2012
2016
2020
Note: Wavelet Coherency: the black contour designates the 5% significance level. The shade code for
coherency ranges from black (low coherency –‑ close to zero) to white (high coherency –‑ close to one).
The cone of influence, which indicates the area affected by edge effects, is the dashed line's outer region.
With other countries, the regions of high coherency are much scarcer. 3.4.3. Synchronisation between Portuguese cities
Between Portugal and the UK, and Portugal and France, the most
We will now focus on Portuguese cities and start by estimating the
impressive high coherency areas take place around 2008 at busi‑
de‑synchronisation index between them. Given that we are working
ness cycle frequencies. Interestingly, the phase difference for these
with 18 cities, we have 153 pairs. It would not be reasonable to display
frequencies tells us that Portugal is leading.
a table with all these values. Alternatively, we use the dissimilarity
Between Portugal and Italy, the only region of high coherency happens matrix as a distance matrix and map the cities in a two‑axis system.
at very low frequencies and lasts until 2004. During this time, the The idea is to reduce the dissimilarity matrix to a two‑column matrix,
Portuguese cycle led the Italian one. However, there are no more the configuration matrix, containing each city's position in two axes.6
regions of statistically significant coherencies. This algorithm results in the map we can see in figure 34.
We leave the task of trying to understand the determinants of such The first thing to note is that this map is not similar to Portugal's
synchronicity between countries for future research. Obvious candi‑ geographical map. We can confirm this impression by computing
dates are distance, business cycle synchronisation, general price the correlation between cycle distances and the physical distances
synchronisation, emigration destination, and tourism flows. between Portuguese cities. The Spearman's correlation coefficient is
–0.0266. Therefore, location does not seem to be a factor in explaining
Regarding the first three possible determinants, we did perform some synchronicity between Portuguese cities.
exploratory analysis. Given that we are working with 12 countries,
we have 66 pairs. The Spearman's correlation coefficient between The second result is that a few cities are quite de‑synchronized with
housing price cycle dissimilarities and physical distances between the rest of the country. Braga is the most obvious outlier, but the
countries (distances between capitals) is 0.15. We also estimated the same is true for Santarém, Portalegre, and Leiria and Setúbal — these
business cycle dissimilarities between countries (using data on real last two form a cluster of their own. Other cities are reasonably well
GDP) and found no correlation with housing price cycle dissimilarities. synchronized between themselves. It is also easy to identify some
Interestingly, the correlation between housing price and general price subclusters. Lisbon, Porto, and Faro, the three Portuguese district
cycle dissimilarities is 50%. This result comes as a surprise because, in capitals with international airports (we did not include the islands in
our data, house prices are presented in real terms. our sample) are synchronized. Like the subcluster formed by Aveiro,
Bragança, Coimbra, and Guarda, it is easy to identify with the hierar‑
chical tree clustering produced in figure 35.
Figure 34Multidimensional scaling map Figure 35Hierarchical Tree Clusters
0.3 0.65
0.60
Braga
0.55
0.2
Santarém 0.50
0.45
Portalegre
0.1 0.40
Guarda Aveiro 0.35

V. Castelo 0.30
Viseu Coimbra
0
Bragança Leiria 0.25
Évora Beja
0.20
V. Real Setúbal
-0.1 Lisbon 0.15
C. Branco Porto 0.10

Faro
Beja
V. Castelo
Castelo Branco
Évora
Vila Real
Viseu
Aveiro
Bragança
Coimbra
Guarda
Faro
Lisbon
Porto
Santarém
Portalegre
Braga
Leiria
Setúbal
-0.2
-0.3
-0.3 -0.2 -0.1 0 0.1 0.2 0.3
In figure 36, we have the wavelet coherency and phase difference for
a few pairs of cities. Lisbon‑Porto and Lisbon‑Faro (because these
cities form a subcluster with the Portuguese capital of Lisbon); Viana
do Castelo and Beja (because this is the most synchronized pair); and
Leiria and Viseu (the least synchronized pair).
Figure 36Cross‑Wavelet Coherency between Portuguese cities
(on the left); Phase‑Difference (on the right).
WCO(Lisbon, Porto) Phase-Differences WCO(Viana do Castelo, Beja) Phase-Differences
1 π 1 π
1~8 freq. band

1~8 freq. band
π/2 π/2
0 0
2 2 -π2
-π2
-π -π
Period (years)
Period (years)
1992
1996
2000
2004
2008
2012
2016
2020
1992
1996
2000
2004
2008
2012
2016
2020
4 4
π π
8~16 freq. band

8~16 freq. band
8 π/2 8 π/2
0 0
-π2 -π2
16 -π 16 -π
1992
1992
1996
2000
2004
2008
2012
2016
2020
1996
2000
2004
2008
2012
2016
2020
1992
1992
1996
2000
2004
2008
2012
2016
2020
1996
2000
2004
2008
2012
2016
2020
WCO(Lisbon, Faro) Phase-Differences WCO(Leiria, Viseu) Phase-Differences
1 π 1 π
1~8 freq. band

1~8 freq. band
π/2 π/2
0 0
2 2 -π2
-π2
-π -π
Period (years)
Period (years)
1992
1996
2000
2004
2008
2012
2016
2020
1992
1996
2000
2004
2008
2012
2016
2020
4 4
π π
8~16 freq. band

8 8~16 freq. band π/2 8 π/2
0 0
-π2 -π2
16 -π 16 -π
1992
1992
1996
2000
2004
2008
2012
2016
2020
1996
2000
2004
2008
2012
2016
2020
1992
1992
1996
2000
2004
2008
2012
2016
2020
1996
2000
2004
2008
2012
2016
2020
Note: Wavelet Coherency: the black contour designates the 5% significance level. The shade code for
coherency ranges from black (low coherency –‑ close to zero) to white (high coherency –‑ close to one).
The cone of influence, which indicates the area affected by edge effects, is the dashed line's outer region.
The coherency between Lisbon and Porto, the two most important quicker than median incomes. Simultaneously, the housing market
Portuguese cities, reveals that both district capitals are synchronized has become an increasingly international market, and there are claims
at lower frequencies (longer run cycles). The phase difference informs that foreign residential investments have been driving housing prices
us that the Lisbon cycle leads the one from Porto. However, at higher up. Consequently, there has been an intense public debate on housing
frequencies, the cycles may diverge from each other. We see a region affordability and global real estate investment.
of high coherency around the year 2008, in the –years frequency band.
What is particularly interesting is that the phase difference reveals a The existence of global investors in housing markets is also asso‑
negative correlation. Therefore, it is possible to have longer run cycles ciated with increased housing price synchronisation, especially at a
aligned while shorter run cycles are misaligned. The second picture city level. In this chapter, we relied on the continuous wavelet trans‑
shows that Lisbon and Faro became highly synchronized after 2005, form and several associated wavelet tools to study housing price
with Lisbon leading. synchronisation between Portugal and eleven other member states of
the European Union (Belgium, Denmark, Finland, France, Germany,
Comparing the third and fourth pairs, we can see the difference Ireland, Italy, Netherlands, Spain, Sweden, and the United Kingdom),
between the most and least synchronized pairs. Viana do Castelo and covering the 1988–2019 period.
Beja's wavelet coherency is consistently high at several frequencies.
The phase differences are very close to zero, suggesting the cycles are We detected two primary cycles in the Portuguese Housing Price
almost simultaneous in both towns. Index, one about nine years long and another covering a period of
approximately 14 years; these two cycles almost run the whole sample
In the case of Leiria and Viseu, the least synchronized pair, high and overlap. We also identified a shorter run cycle, almost four years
coherency regions are much scarcer. However, it is still remarkable long, corresponding to the international financial crisis and the subse‑
that coherency is very high and statistically significant at very low quent sovereign debt crisis (2008–2012).
frequencies, meaning that even these two cites display a common
long‑run cycle. We then estimated the «distance» between the wavelet transform
of Portuguese prices and other countries’ prices. The British housing
market is the one Portugal is most synchronized with. Italy, Spain, and
3.5. Conclusions
France come next. So, we have estimated the wavelet coherency and
In this chapter, we were concerned with the dynamics of Portugal's the phase‑difference between Portuguese prices and prices in those
housing prices, which have increased in recent years. This rise has been four countries. This more detailed analysis has revealed that Spain is
a feature of the economic recovery in several countries after the finan‑ the country with which Portugal has the largest regions of statisti‑
cial crisis. On average, in OECD countries, house prices have grown cally significant coherency. Until the early 2000s, there is a region of
high coherency between the 4 and 8‑years frequencies; after that, we subclusters, namely the one formed by Lisbon, Porto, and Faro (the
can see high coherencies at higher and lower frequencies. The phase three district capitals with international airports) or the subcluster
difference indicates that Spain is the leader and Portugal the follower. constituted by Aveiro, Bragança, Coimbra, and Guarda. Once again, we
In regards to the other three countries, regions of high coherency are have not tried to explain such synchronicity except in what concerns
fewer. They exist around 2008, at business cycle frequencies, with the physical distances, which are not correlated with cycle dissimilari‑
UK, and with France. ties, suggesting that location does not explain synchronicity between
Portuguese cities; hence, other determinants must be found.
Interestingly, the phase difference for these frequencies shows that
Portugal is leading. The Portuguese cycle also led the Italian one. Next, we have estimated wavelet coherency and phase difference for
However, the only region of high coherency between Portugal and a few pairs of cities. The two most important Portuguese cities, Lisbon
Italy occurs at very low frequencies and lasts until 2004, with no more and Porto, are synchronized at longer run cycles, with the Lisbon
regions revealing statistically significant coherencies. cycle leading over the Porto one. However, at higher frequencies, the
cycles may diverge from each other, and the correlation is negative.
We have not explored the determinants of such synchronicity. Therefore, it is possible to have longer run cycles aligned while shorter
According to the literature, distance, emigrant destination, and run cycles are misaligned. Lisbon and Faro became very synchronized
tourism flows (note that UK, Spain, and France were the Portuguese after 2005, with Lisbon leading. Viana do Castelo and Beja is the most
top 3 outbound tourism markets in 2019) are factors to be considered. synchronized pair, with consistently high wavelet coherency at several
However, we leave this examination for future research. How globali‑ frequencies. The phase differences are very close to zero, suggesting
sation and financialisation influence national housing price dynamics is that cycles are almost simultaneous in both towns. Regarding Leiria
also a very relevant matter, as it may have implications on the effecti‑ and Viseu, the least synchronized pair, high coherency regions are
veness of a range of policy tools aimed at addressing imbalances in the rarer. However, they still exist at very low frequencies, meaning that
housing market, namely affordability problems. even these two cities display a common long‑run cycle.
This analysis has also been conducted focusing on Portuguese cities. These results suggest that the Portuguese housing market is
To be more precise, we start by estimating the de‑synchronisation segmented and displays regional heterogeneities. Knowing that the
index between the 18 district capitals. Our results can be summa‑ housing cycles of Portuguese cities may be de‑synchronized, housing
rised as follows: Braga, Santarém, Portalegre, Leiria, and Setúbal are policies should be designed locally.
quite de‑synchronized with the rest of the country (the last two
form a cluster of their own); and despite other cities being relatively
well synchronized between themselves, one can distinguish some
References Aguiar‑Conraria, L. and M.J. Soares, (2011b), «The continuous wavelet
transform: a primer», NIPE Working Paper, 16, 1–43.
Aalbers, M.B., (2016), «The financialization of housing»,
In The Financialization of Housing. Aguiar‑Conraria, L. and M.J. Soares, (2014), «The continuous wavelet
transform: moving beyond uni‑and bivariate analysis», Journal of Economic
Abbott, A., and G. De Vita, (2013), «Testing for long‑run convergence across Surveys, 28(2), 344–375.
regional house prices in the UK: a pairwise approach», Applied Economics,
45(10), 1227–1238. Alexander, C. and M. Barrow, (1994), «Seasonality and cointegration of
regional house prices in the UK», Urban Studies, 31(10), 1667–1689.
Adair, A., J. Berry, S. McGreal, L. Sýkora, A. G. Parsa and B. Redding,
(1999), «Globalization of real estate markets in central Europe», European Alter, A., J. Dokko and D. Seneviratne, (2018), «House Price synchroni‑
Planning Studies, 7(3), 295–305. city, banking integration, and global financial conditions», In IMF Working
Papers (WP/18/250).
Agnello, L. and L. Schuknecht, (2011), «Booms and busts in housing
markets: determinants and implications», Journal of Housing Economics, 20(3), Álvarez, L. J., G. Bulligan, A. Cabrero, L. Ferrara and H. Stahl, (2010),
171–190. «Housing cycles in the major euro area countries» (Documentos Ocasionales
n.o 1001; The occasional paper series), link
Aguiar‑Conraria, L., P. Brinca, H.V. Guðjónsson and M. J. Soares,
(2017), «Business cycle synchronization across U. S. states», B. E. Journal of Apergis, N. and J.E. Payne, (2012), «Convergence in U.S. house prices by
Macroeconomics, 17(1), 1–15. state: evidence from the club convergence and clustering procedure», Letters
in Spatial and Resource Sciences, 5(2), 103–111.
Aguiar‑Conraria, L., P.C. Magalhães and M.J. Soares, (2012), «Cycles in
politics: wavelet analysis of political time series», American Journal of Political Arnott, R., (1987), «Economic theory and housing», In E. S. Mills (Ed.),
Science, 56(2), 500–518. Handbook of Regional and Urban Economics: Vol. II Col (pp. 959–988), Elsevier.
Aguiar‑Conraria, L., P.C. Magalhães and M.J. Soares, (2013), «The natio‑ Ashworth, J. and S.C. Parker, (1997), «Modelling regional house prices in
nalization of electoral cycles in the United States: a wavelet analysis», Public the UK», Scottish Journal of Political Economy, 44(3), 225–246.
Choice, 156(3–4), 387–408.
Badarinza, C. and T. Ramadorai, (2018), «Home away from home? foreign
Aguiar‑Conraria, L. and M.J. Soares, (2011a), «Business cycle synchroni‑ demand and London house prices», Journal of Financial Economics, 130(3),
zation and the Euro: a wavelet analysis», Journal of Macroeconomics, 33(3), 532–555.
477–489.
Baffoe‑Bonnie, J., (1998), «The dynamic impact of macroeconomic aggrega‑ Bordo, M. and T. Helbling, (2003), «Have national business cycles become
tes on housing prices and stock of houses: a national and regional analysis», more synchronized?», National Bureau of Economic Research, Working Paper
Journal of Real Estate Finance and Economics, 17(2), 179–197. 10130.
Balcilar, M., A. Beyene, R. Gupta and M. Seleteng, (2013), «“Ripple” Bruno, V. and H.S. Shin, (2013), «Global factors in capital flows and credit
effects in south african house prices», Urban Studies, 50(5), 876–894. growth», Working Paper 237, Griswold Center for Economic Policy Studies, link
Bangura, M. and C.L. Lee, (2020), «House price diffusion of housing Bruno, V. and H.S. Shin, (2015a), «Capital flows and the risk‑taking channel
submarkets in Greater Sydney», Housing Studies, 35(6), 1110–1141. of monetary policy», Journal of Monetary Economics, 71, 119–132.
Bardhan, A.D. and C.A. Kroll, (2007), Globalization and real estate: issues, Bruno, V. and H.S. Shin, (2015b), «Cross‑border banking and global liqui‑
implications, opportunities (Issue January 2007). dity», The Review of Economic Studies, 82(2), 535–564.
Barros, C.P., L.A. Gil‑Alana and J.E. Payne, (2012), «Comovements among Büdenbender, M. and O. Golubchikov, (2017), «The geopolitics of real
U.S. state housing prices: evidence from fractional cointegration», Economic estate: assembling soft power via property markets», International Journal of
Modelling, 29(3), 936–942. Housing Policy, 17(1), 75–96.
Baskaya, Y. S., J. di Giovanni, Ş. Kalemli‑Özcan, and M.F. Ulu, (2017), Burger, P. and L.J. Van Rensburg, (2008), «Metropolitan house prices in
«International spillovers and local credit cycles», No. DP11839; CEPR South Africa: do they converge?», South African Journal of Economics, 76(2),
Discussion Papers, cepr.org/active/publications/discussion_papers/ 291–297.
dp.php?dpno=11839
Caceres, C., Y. Carrière‑Swallow, I. Demir and B. Gruss, (2016), «U.S.
Belke, A., C. Domnick and D. Gros, (2017), «Business cycle synchronization monetary policy normalization and global interest rates», WO/16/195, IMF
in the EMU: core vs. periphery», Open Economies Review, 28(5), 863–892. Working Papers.
Beltratti, A. and C. Morana, (2010), «International house prices and Calvo, G. A., L. Leiderman, and C.M. Reinhart, (1996), «Inflows of capital
macroeconomic fluctuations», Journal of Banking and Finance, 34(3), 533–545. to developing countries in the 1990s», Journal of Economic Perspectives, 10(2),
123–139.
Blanco, F., V. Martín and G. Vazquez, (2016), «Regional house price con‑
vergence in Spain during the housing boom», Urban Studies, 53(4), 775–798. Cameron, G., J. Muellbauer and A. Murphy, (2006), «Was there a british
house price bubble? Evidence from a Regional Panel», In CEPR Discussion
Papers (No. 5619).
Canarella, G., S.M. Miller and S. Pollard. (2012). «Unit roots and struc‑ Churchill, S. A., J. Inekwe and K. Ivanovski, (2018), «House price conver‑
tural change: an application to US house price indices», Urban Studies, 49(4), gence: evidence from Australian cities», Economics Letters, 170, 88–90.
757–776.
Claessens, S., M.A. Kose and M.E. Terrones, (2011), «Financial cycles:
Cascio, I. Lo, (2020), «A wavelet analysis of the ripple effect in UK regional what? how? when?», WP/11/76, IMF Working Papers, link.
house prices», D/SEAS Working Papers, 3(1), 1–22.
Clapp, J. M. and D. Tirtiroglu, (1994), «Positive feedback trading and dif‑
Case, B., W.N. Goetzmann and K.G. Rouwenhorst, (1999), «Global real fusion of asset price changes: evidence from housing transactions», Journal of
estate markets: cycles and fundamentals», No. 99–03, Yale International Economic Behavior and Organization, 24(3), 337–355.
Center for Finance Working Paper.
Clark, S. P. and T.D. Coggin, (2009), «Trends, cycles and convergence in
Case, B., W.N. Goetzmann and K.G. Rouwenhorst, (2000), «Global real U.S. regional house prices», The Journal of Real Estate Finance and Economics,
estate markets — cycles and fundamentals» (Working Paper 7566; NBER 39(3), 264–283.
Working Paper Series).
Clerc, L., (2019), «Towards a global real estate market? Trends and evidence»,
Cesa‑Bianchi, A., L.F. Cespedes and A. Rebucci, (2015), «Global liquidity, In R. Nijskens, M. Lohuis, and W. Heeringa (Eds.), Hot Property (pp. 63–81),
house prices, and the macroeconomy: evidence from advanced and emerging Springer International Publishing.
economies», Journal of Money, Credit and Banking, 47(S1), 301–335.
Coakley, J., (1994), «The Integration of property and financial markets»,
Chen, P.‑F., Chien, M.‑S., and Lee, C.‑C., (2011), «Dynamic modeling of Environment and Planning A: Economy and Space, 26(5), 697–713.
regional house price diffusion in Taiwan», Journal of Housing Economics, 20(4),
315–332. Cook, S., (2003), «The convergence of regional house prices in the UK»,
Urban Studies, 40(11), 2285–2294.
Chiang, S. H., (2014), «Housing markets in China and policy implications:
comovement or ripple effect», China and World Economy, 22(6), 103–120. Cook, S., (2005), «Regional house price behaviour in the UK: application of a
joint testing procedure», Physica A: Statistical Mechanics and Its Applications,
Chien, M.‑S., (2010), «Structural breaks and the convergence of regional 345(3–4), 611–621.
house prices», Journal of Real Estate Finance and Economics, 40(1), 77–88.
Cook, S. and C. Thomas, (2003), «An alternative approach to examining the
Chowdhury, R. A. and D. Maclennan, (2014), «Regional house price cycles ripple effect in UK house prices», Applied Economics Letters, 10(13), 849–851.
in the UK, 1978–2012: a markov switching VAR». Journal of European Real
Estate Research, 7(3), 345–366.
Davis, M. A. and S. van Nieuwerburgh, (2015), «Housing, finance, and Favara, G. and J. Imbs, (2015), «Credit supply and the price of housing»,
the macroeconomy», In Handbook of Regional and Urban Economics (1st ed., American Economic Review, 105(3), 958–992.
Vol. 5B, pp. 753–811). Elsevier B.V.
Fernandez, R. and M.B. Aalbers, (2016), «Financialization and housing:
de Bandt, O., K. Barhoumi and C. Bruneau. (2010). «The international between globalization and varieties of capitalism», Competition and Change,
transmission of house price shocks». In O. De Bandt, T. Knetsch, J. Peñalosa, 20(2), 71–88.
and F. Zollino (Eds.), Housing Markets in Europe (pp. 129–158), Springer Berlin
Heidelberg. Fidrmuc, J., T. Ikeda and K. Iwatsubo, (2012), «International transmission
of business cycles: evidence from dynamic correlations», Economics Letters,
Del Negro, M., D. Giannone, M.P. Giannoni, and A. Tambalotti, (2019), 114(3), 252–255.
«Global trends in interest rates», Journal of International Economics, 118,
248–262. Fields, D., (2015), «Contesting the financialization of urban space: commu‑
nity organizations and the struggle to preserve affordable rental housing in
Drake, L., (1995), «Testing for convergence between UK regional house New York City», Journal of Urban Affairs, 37(2), 144–165.
prices». Regional Studies, 29(4), 357–366.
Flor, M. A. and T. Klarl, (2017), «On the cyclicity of regional house prices:
Duca, J.V. (2020), «Making sense of increased synchronization in glo‑ new evidence for U.S. metropolitan statistical areas», Journal of Economic
bal house prices», Journal of European Real Estate Research, Vol. 13 No. 1, Dynamics and Control, 77, 134–156.
pp. 5–16.
Geng, N. (2018), «Fundamental drivers of house prices in advanced econo‑
Duca, J. V., J. Muellbauer and A. Murphy, (2011), «House prices and credit mies», In IMF Working Papers (WP/18/164; IMF Working Paper).
constraints: making sense of the US experience», Economic Journal, 121(552),
533–551. Girouard, N., M. Kennedy, P. van den Noord and C. André, (2006),
«Recent house price developments: the role of fundamentals», In OECD
Englund, P. and Y.M. Ioannides, (1997), «House price dynamics: an interna‑ Economic Outlook, No. 475, Economics Department Working Papers.
tional empirical perspective», Journal of Housing Economics, 6(2), 119–136.
Giussani, B. and G. Hadjimatheou, (1991), «Modeling regional house prices
EUROSTAT, (2013), «Handbook on residential property prices» (RPPIs), in the United Kingdom», Papers in Regional Science, 70(2), 201–219.
In Handbook on Residential Property Prices (RPPIs), Publications Office of the
European Union. Gotham, K. F., (2009), «Creating liquidity out of spatial fixity: the secondary
circuit of capital and the subprime mortgage crisis», International Journal of
Fallis, G., (1985), «Introduction», In Housing Economics (pp. 1–26). Urban and Regional Research, 33(2), 355–371.
Grigoryeva, I. and D. Ley, (2019), «The price ripple effect in the Vancouver Hirata, H., M.A. Kose, C. Otrok and M.E. Terrones, (2012), «Global house
housing market», Urban Geography, 40(8), 1168–1190. price fluctuations: synchronization and determinants», NBER International
Seminar on Macroeconomics, 9(1), 119–166.
Gros, D., (2006), «Bubbles in real estate? A longer‑term comparative analysis
of housing prices in Europe and the US», No. 239/February 2006, CEPS Hoesli, M.,u (2020), «An investigation of the synchronization in global
Working Document, link house prices», Journal of European Real Estate Research, 13(1), 17–27.
Gupta, R., C. Andre and L. Gil‑Alana, (2015), «Comovement in Euro area Holly, S., M.H. Pesaran and T. Yamagata, (2010), «A spatio‑temporal
housing prices: a fractional cointegration approach», Urban Studies, 52(16), model of house prices in the USA», Journal of Econometrics, 158(1), 160–173.
3123–3143.
Holly, S., M.H. Pesaran and T. Yamagata, (2011), «The spatial and temporal
Gupta, R. and S.M. Miller, (2012a), «The time‑series properties of house diffusion of house prices in the UK», Journal of Urban Economics, 69(1), 2–23.
prices: a case study of the southern california market», Journal of Real Estate
Finance and Economics, 44(3), 339–361. Holmes, M.J., (2007), «How convergent are regional house prices in the
United Kingdom? Some new evidence from panel data unit root testing»,
Gupta, R. and S.M. Miller, (2012b), «“Ripple effects” and forecasting home Journal of Economic and Social Research, 9(1), 1–17.
prices in Los Angeles, Las Vegas, and Phoenix», The Annals of Regional Science,
48(3), 763–782. Holmes, M.J. and A. Grimes, (2008), «Is there long‑run convergence among
regional house prices in the UK?» Urban Studies, 45(8), 1531–1544.
Hamnett, C., (1989), «Regional variations in house prices and house price
inflation in Britain, 1969–1989», Journal of Valuation, 7(4), 342–360. Holmes, M. J., J. Otero and T. Panagiotidis, (2011), Investigating regional
house price convergence in the United States: evidence from a pair‑wise
Hekwolter of Hekhuis, M., R. Nijskens and W. Heeringa, (2017), «The hou‑ approach, Economic Modelling, 28(6), 2369–2376.
sing market in major dutch cities», No. 1, Occasional Studies, Vol. 15, link
Huang, F., C. Li and Y. Li., (2010), «Ripple effect of housing prices among
Helbling, T. and M.E. Terrones, (2003), «Real and financial effects of chinese deputy provincial cities based on an alternative approach»,
bursting asset price bubbles», In IMF World Economic Outlook, April 2003 International Journal of Business Administration, 1(1), 19–24.
(pp. 61–94), International Monetary Fund. link
Huang, F., Y. Zhou and C. Li, (2010), «Ripple effect of housing prices fluctua‑
Henley, A. and B. Morley, (2000), «European house price volatility and tions among nine cities of China», Management Science and Engineering, 4(3),
the macroeconomy: the implications for european monetary union», Royal 41–54. file:///Users/yangwen/Downloads/1341–1381–1‑PB.pdf
Economic Society Annual Conference, link
Hudson, C., J. Hudson and B. Morley, (2018), «Differing house price linka‑ Katagiri, M., (2018), «House price synchronization and financial openness:
ges across UK regions: a multi‑dimensional recursive ripple model», Urban a dynamic factor model approach», WP/18/208, IMF Working Paper.
Studies Journal Limited, 55(8), 1636–1654.
Kim, Y. S. and J.J. Rous, (2012), «House price convergence: evidence from
Hwang, M. and J.M. Quigley, (2006), «Economic fundamentals in local hou‑ US state and metropolitan area panels», Journal of Housing Economics, 21(2),
sing markets: evidence from U.S. metropolitan regions», Journal of Regional 169–186.
Science, 46(3), 425–453.
Kishor, N. K. and H.A. Marfatia, (2017), «The dynamic relationship
Igan, D. and P. Loungani, (2012), «Global housing cycles», WP/12/217, IMF between housing prices and the macroeconomy: evidence from OECD coun‑
Working Paper. tries», Journal of Real Estate Finance and Economics, 54(2), 237–268.
IMF, (2016), «The growing importance of financial spillovers from emerging Klarl, T., (2018), «Housing is local: applying a dynamic unobserved factor
market economies», Global Financial Stability Report, April (Chapter 2). model for the dutch housing market», Economics Letters, 170, 79–84.
IMF, (2017), «Are countries losing control of domestic financial conditions?» Kose, M. A., C. Otrok and E. Prasad, (2012), «Global business cycles: con‑
In Global Financial Stability Report (Issue April, pp. 83–108), link vergence or decoupling?», International Economic Review, 53(2), 511–538.
IMF, (2018), «House price synchronization: what role for financial factors?», Kose, M. A., C. Otrok and C.H. Whiteman, (2008), «Understanding the evolu‑
In Global Financial Stability Report April 2018: A Bumpy Road Ahead (Issue tion of world business cycles», Journal of International Economics, 75(1), 110–130.
April 2018), link
Kruskal, J.B., (1964a), «Multidimensional scaling by optimizing goodness of
IMF, (2019), «Downside risks to house prices», In Global Financial Stability fit to a nonmetric hypothesis», Psychometrika, 29(1), 1–27.
Report (Issue April), International Monetary Fund, link
Kruskal, J.B., (1964b), «Nonmetric multidimensional scaling: a numerical
Jordà, Ò., K. Knoll, D. Kuvshinov, M. Schularick and A.M. Taylor, method», Psychometrika, 29(2), 115–129.
(2019), «The rate of return on everything, 1870–2015», The Quarterly Journal
of Economics, 134(3), 1225–1298. Kuang, W. and Q. Wang, (2018), «Cultural similarities and housing market
linkage: evidence from OECD countries», Frontiers of Business Research in
Jordà, Ò., M. Schularick and A.M. Taylor, (2017), «Macrofinancial history China, 12(1), 1–25.
and the new business cycle facts», NBER Macroeconomics Annual, 31(1),
213–263. Kyriazakou, E. and T. Panagiotidis, (2014), «Linear and nonlinear causa‑
lity in the UK housing market: a regional approach», Economics and Business
Letters, 3(4), 288–297.
Larraz‑Iribas, B., J.‑L. Alfaro‑Navarro and F. Rodriguez‑Aragon, MacDonald, R. and M.P. Taylor, (1993), «Regional house prices in Britain:
(2008), «Asymmetric behaviour of spanish regional house prices: a multiva‑ long‑run relationships and short‑run dynamics», Scottish Journal of Political
riate approach», The Journal of Economic Asymmetries, 5(2), 81–104. Economy, 40(1), 43–55.
Lean, H.H. and R. Smyth, (2013), «Regional house prices and the ripple Maclennan, D., (1979), «Housing Economics — A Review and introduc‑
effect in Malaysia», Urban Studies, 50(5), 895–922. tion», Scottish Journal of Political Economy, 26(3), 325–329.
Lee, C.C. and M.S. Chien, (2011), «Empirical modelling of regional house Marfatia, H.A., (2018), «Wavelet linkages of global housing markets and
prices and the ripple effect», Urban Studies, 48(10), 2029–2047. macroeconomy», SSRN Electronic Journal.
Lee, M.‑T., M.‑L. Lee and S.H. Lin, (2014), «Trend properties, cointegration, Meen, G., (1990), «The removal of mortgage market constraints and the
and diffusion of presale house prices in Taiwan: can taipei's house prices implications for econometric modelling of UK house prices», Oxford Bulletin
ripple out?», Habitat International, 44, 432–441. of Economics and Statistics, 52(1), 1–24.
Leung, C., (2004), «Macroeconomics and housing: a review of the literature», Meen, G., (1996), «Ten propositions in UK housing macroeconomics:
Journal of Housing Economics, 13(4), 249–267. an overview of the 1980s and early 1990s», Urban Studies, 33(3), 425–444.
Ley, D., (2015), «Global China and the making of Vancouver's residential Meen, G., (1999), «Regional house prices and the ripple effect: a new inter‑
property market», International Journal of Housing Policy, 17(1), 15–34. pretation», Housing Studies, 14(6), 733–753.
Liu, C., Z.Q. Luo, L. Ma and D. Picken, (2008), «Identifying house price Miles, W., (2017), «Has there actually been a sustained increase in the syn‑
diffusion patterns among australian state capital cities», International Journal chronization of house price (and business) cycles across countries?» Journal
of Strategic Property Management, 12(4), 237–250. of Housing Economics, 36, 25–43.
Lourenço, R.F. and P.M.M. Rodrigues, (2014), «The dynamics and con‑ Miles, W., (2019), «Has the euro sustainably increased home price co‑move‑
trast of house prices in Portugal and Spain», Economic Bulletin and Financial ment?», Journal of Regional Science, 59(5), 931–961.
Stability Report Articles and Banco de Portugal Economic Studies.
Miranda‑Agrippino, S. and H. Rey, (2020), «U.S. monetary policy and the
Luo, Z., C. Liu and D. Picken, (2007), «Housing price diffusion pattern global financial cycle», The Review of Economic Studies, 87(6), 2754–2776.
of Australia's state capital cities», International Journal of Strategic Property
Management, 11(4), 227–242, link Morley, B. and D. Thomas, (2011), «Risk‑return relationships and asym‑
metric adjustment in the UK housing market», Applied Financial Economics,
21(9), 735–742.
Muellbauer, J. and A. Murphy, (1994), «Explaining regional house prices in Rey, H., (2015), «Dilemma not trilemma: the global financial cycle and mone‑
the UK», WP94/21, UCD Centre for Economic Research Working Paper Series. tary policy independence», Working Paper 21162, NBER Working Papers.
Muellbauer, J. and A. Murphy, (2008), «Housing markets and the economy: Rogers, D. and S.Y. Koh, (2017), «The globalisation of real estate: the poli‑
the assessment», Oxford Review of Economic Policy, 24(1), 1–33. tics and practice of foreign real estate investment», International Journal of
Housing Policy, 17(1), 1–14.
O'sullivan, T. and K. Gibb, (2003), Introduction, In T. O'sullivan and K. Gibb
(Eds.), Housing Economics and Public Policy (pp. 1–21), Blackwell Science. Ronald, R. and C. Dewilde, (2017), «Why housing wealth and welfare?»,
In C. Dewilde and R. Ronald (Eds.), Housing Wealth and Welfare (pp. 1–34).
OECD, (2019), «The rising cost of the middle‑class lifestyle», In Under Edward Elgar Publishing.
Pressure: The Squeezed Middle Class (pp. 102–133), OECD Publishing, link
Rosenthal, L., (1986), «Regional house price interactions in the UK,
Oikarinen, E., (2005), «The diffusion of housing price movements from 1975–81: a cross‑spectral analysis», Applied Economics, 18(9), 1011–1023.
centre to surrounding areas», No. 979, ETLA Discussion Papers.
Ryczkowski, M., (2019), «Money, credit, house prices and quantitative
Otrok, C. and M.E. Terrones, (2005), «House prices, interest rates and easing — the wavelet perspective from 1970 to 2016», Journal of Business
macroeconomic fluctuations: international evidence», In International Economics and Management, 20(3), 546–572.
Monetary Fund, mimeo (Issue May), link
Smith, L. B., K.T. Rosen and G. Fallis, (1988), «Recent economic develop‑
Payne, J.E., (2012), «The long‑run relationship among regional housing prices: ments in models of housing markets», Journal of Economic Literature, 26(1),
an empirical analysis of the U.S.», Journal of Regional Analysis and Policy, 42(1), 29–64.
28–35.
Stevenson, S., (2004), «House price diffusion and inter‑regional and
Pollakowski, H.O. and T.S. Ray, (1997), «Housing price diffusion patterns cross‑border house price dynamics», Journal of Property Research, 21(4),
at different aggregation levels: an examination of housing market effi‑ 301–320.
ciency», Journal of Housing Research, 8(1), 107–124.
Taltavull de La Paz, P., E. López and F. Juárez, (2017), «Ripple effect
Poterba, J.M., (1984), «Tax subsidies to owner‑occupied housing: an asset on housing prices, evidence from tourist markets in Alicante, Spain»,
‑market approach», The Quarterly Journal of Economics, 99(4), 729–752. International Journal of Strategic Property Management, 21(1), 1–14.
Pow, C.P., (2017), «Courting the ‘rich and restless’: globalisation of Terrones, M. E. and C. Otrok, (2004), «The global house price boom». In
real estate and the new spatial fixities of the super‑rich in Singapore», World Economic Outlook, September (pp. 71–89), link
International Journal of Housing Policy, 17(1), 56–74.
Teye, A. L., M. Knoppel, J. De Haan and M.G. Elsinga, (2017), «Amsterdam Zhang, D., (2010), «Testing convergence on UK regional housing prices:
house price ripple effects in the Netherlands», Journal of European Real Estate A Fractional Integration Approach», In L. Tsipouri, N. Tsounis, and A. Vlachvei
Research, 10(3), 331–345. (Eds.), International Conference On Applied Economics — ICOAE 2010
(pp. 845–852).
Tsai, I.C., (2015), «Spillover effect between the regional and the national
housing markets in the UK», Regional Studies, 49(12), 1957–1976. Zhang, F. and B. Morley, (2014), «The convergence of regional house prices
in China», Applied Economics Letters, 21(3), 205–208.
Tu, Y., (2000), «Segmentation of australian housing markets: 1989–98»,
Journal of Property Research, 17(4), 311–327. Zhang, L., E.C. Hui and H. Wen, (2017), «The regional house prices in
China: ripple effect or differentiation», Habitat International, 67, 118–128.
Vansteenkiste, I., (2007), «Regional housing market spillovers in the US:
lessons from regional divergences in a common monetary policy setting», Zohrabyan, T., D. Leatham and D. Bessler, (2008), «Cointegration analysis
In ECB Working Paper (No. 708; Issue January). of regional house prices in U.S.» In M. A. Gunderson (Ed.), Agricultural and
Rural Finance Markets in Transition, Proceedings of Regional Research Committee
Vansteenkiste, I. and P. Hiebert, (2011), «Do house price developments NC‑1014, October 4–5, 2007, St. Louis, Missouri (Issue July).
spillover across euro area countries? Evidence from a global VAR», Journal of
Housing Economics, 20(4), 299–314.
Weng, Y. and P. Gong, (2017), «On price co‑movement and volatility spi‑
llover effects in China's housing markets», International Journal of Strategic
Property Management, 21(3), 240–255.
Whitehead, C. M. E., (1999), «Urban housing markets: theory and policy»,

In P. Cheshire and E. S. Mills (Eds.), Handbook of Regional and Urban Economics:
Vol. III (pp. 1559–1594), Elsevier.
Yang, Z., Y. Fan and L. Zhao, (2018), «A reexamination of housing price

and household consumption in China: the dual role of housing consumption
and housing investment», Journal of Real Estate Finance and Economics, 56(3),
472–499.
Chapter 4
Territorial patterns of housing
affordability in Portugal
Paulo Batista, João Lourenço Marques and Eduardo Anselmo Castro
4.1. Introduction
Housing is a recurring topic when it comes to defining public policies. legal protection against forced evictions, harassment, and other
The right to housing is one of the fundamental human rights, and as threats is particularly relevant;
several authors argue (e.g. Allen, 2006), there has been a converging set • Availability of services, materials, facilities, and infrastructure:
of policies that focus on realising the legal duties arising from them. a house is appropriate if there is a set of elements and basic infras‑
Nevertheless, in the last few years, a growing debate has taken place tructures available within it, namely in what concerns access to
over how such policies translate into the practical realization of those drinking water, sanitation and collection/treatment of residues, and
rights, with a focus on the role of the housing market (see Marques energy (to cook, for warmth, lighting and food storage);
et al., 2020, for a brief overview on housing provision and housing • Availability of essential structural elements: housing is appropriate
consumption). This debate is supported by the claim that the right to if it ensures its sheltering capabilities — of physical safety and
housing should go beyond its usual interpretation, such as the focus protection from natural risks — and adequate space considering the
on structural and infrastructural housing conditions (Carmo et al., number and type of occupants;
2014; Ferrão, 2019; Wetzstein, 2017). • Adequacy to the physical needs of occupants: housing is appropriate
if it meets the specific needs of vulnerable, marginalized, or physi‑
The contemporary guidelines for the right to housing, developed as an
cally limited groups;
international consensus (see, for example, The Human Rights‑Based
• Adequate location: housing is appropriate if it provides residents
Approach to Housing and Slum Upgrading, 2017), establishes a multi‑
with essential services — education, healthcare, and access to other
dimensional approach which includes:
social protection services (for the elderly, for example,) — and equal
• Safety of usufruct: a house is not deemed appropriate if its occupants access to employment opportunities;
do not possess a certain degree safety in terms usufruct, for which • Cultural adequacy: housing needs to be considered part of the right
to individual expression and the cultural manifestation of identities.
However, in practical terms, efforts to identify and assess housing needs resources to buy a house than with assessing the satisfaction of
usually focus on a quantitative perspective, namely the number and type housing needs in their multiple dimensions. As a result, nowadays the
of dwellings compared to the population and its characteristics (such most relevant housing tenure status is that of owner (representing
as the number of households, size, or composition). A complementary around 70% of Portuguese households in 2011, a number in line with
quantitative perspective is concerned with an economic dimension, most of the OECD member countries), and a low potential for residen‑
from where the concept of housing affordability emerges. This pers‑ tial mobility has had important consequences on the satisfaction of
pective is commonly viewed as the outcome of a relationship between housing needs (Baker et al., 2016; Henley, 1998).
housing costs and household income. However, efforts to combine
affordability and larger dimension housing needs have resulted in a Moreover, the territorial features embedded in houses, and the
multitude of metrics that seek to measure the adequacy and satisfaction difficulty in fully identifying them, lead to relevant information asym‑
of households — see Mulliner and Maliene (2015) for a summary of a metries (Ambrose and Diop, 2018; Kurlat and Stroebel, 2015) and,
multidimensional vision of housing affordability. therefore, to markets which are usually far from efficient, while
also complicating the development of integrated policies. The terri‑
The spatial character of housing (Bogdon and Can, 1997; Butler and tory's spatial and socioeconomic characteristics offer an opportunity
Hamnet, 2013) has resulted in the decentralization of capabilities to pinpoint the diversity and multidimensionality of housing needs.
of local political administrative bodies (municipalities) to ensure However, this imposes the need to develop new metrics in order to
the effective monitoring and regulation of the housing market. intersect territorial diversity with housing market drivers.
The operational instruments have been pointed out as underdeve‑
loped and insufficient (Alves, 2020) but have seen a relevant boost in Seeking to contribute towards the spatial, social, and economic
recent years. Now, as part of a new generation of territorial planning features of the housing market, the work presented in this chapter
instruments, local housing strategies7 (Decree‑Law no. 37/2018) and makes an integrated and territorial analysis of housing affordability.
municipal housing charters (Act no. 83/2019) have defined a set of new It not only reflects on challenges posed to model market drivers by
housing intervention programmes8. These instruments are expected data collection and processing, but also (re)introduces the concept
to embody the principles of the equally recent housing act (Act no. of territory in housing affordability metrics. Here, a concept of terri‑
83/2019), allowing for the effective regulation of the housing market. tory unbound from its common administrative expressions is used
(João Marques and Batista, 2021). In line with the debate on the role
Public policy has been focused on improving households’ access to of space developed by (Harvey, 2006; Lefebvre, 1991), the territorial
housing through its participation in the open residential market. The analysis of affordability considers a space where different forces inte‑
state assumed a regulatory role in this context, with its interventions ract and, thus, territorial patterns are not objectively fixed or bounded
being more concerned with identifying households’ lack of financial a priori. To consider this notion of space‑territory, an approach was
developed integrating econometric tools to observe the expression of As an example, in territorial contexts where population is ageing, and
the spatial drivers of housing affordability, their spatial interlinkages usually associated with population decline, housing needs are linked to
(spatial autocorrelation) and the expression of these territorial the physical comorbidities of households. These conditions translate
features in housing affordability metrics. into very specific housing needs — e.g., on an architectural level —
which are not in line with available housing features, since they were
A synthetic and spatial measure of housing affordability was developed designed with another demographic context in mind. This example is
as an extension of the standard relationship between housing costs particularly relevant considering the demographic challenge Portugal
and household income ratios. This contributes to combining this clas‑ and the majority of European countries currently face.
sical approach with the endogenous and exogenous features of the
territory. To pursue this aim, the paper begins with a brief discussion of Several authors highlight life cycles as an important element in estab‑
the exogenous territorial patterns as candidates for explaining housing lishing housing needs (see e.g. Abramsson and Andersson, 2016, Bitter
affordability, in Section 2. In Section 3, we present the methodological and Plane, 2013, and Costa‑Font et al., 2009). Thus, while population
benchmarks used for proposing a territorialized housing affordability ageing and decline are generalized phenomena, they affect central and
index for mainland Portugal9, discussing the challenges related with data peripheral regions in different ways (Castro et al., 2020, and Marques
scarcity, the micro‑macro nexus, and small area estimations. Section 4 et al., 2021). Portugal, which fits the description of a peripheral terri‑
describes the results, namely the territorial patterns of housing affor‑ tory in the European context, faces a much fiercer decline and ageing
dability at a civil parish level, showcasing the explanatory factors of of its global demographic dynamics, also displaying highly contras‑
housing prices and household income. Lastly, Section 5 reports the main ting internal scenarios (Castro et al., 2015, and Marques et al., 2021).
findings and conclusions, focusing on the relevance of the identified Current demographic metrics show that only a few specific areas
geographical patterns for designing public housing market policies. within the cities of Lisbon and Porto, the Algarve region and some
other urban areas, are an exception regarding population ageing and
4.2. Social, economic and spatial elements decline (see figure 37 and figure 38). This internal contrasting demo‑
under housing affordability analysis graphic evolution is reflected in very differentiated patterns regarding
the housing market's demand and supply.
4.2.1. Demography and territorial transformation processes
If at first a decline in population (demand) can result in a surplus
Amongst the exogenous socioeconomic drivers of housing access
of available housing, according to these territories’ needs (in purely
conditions, demographic dynamics and territorial transformations
quantitative terms), the fact that the population's characteristics are
represent key elements. Characteristics arising from these pheno‑
generally different (due to population ageing) should also be consi‑
mena can be linked to the drivers of housing demand and supply.
dered, as well as the fact that the decline process usually originates
insufficient dynamics in the housing market. This also happens due
to low demand and expectations, and in turn will result in an even
further reduction in the future. In the Portuguese case, the metric of
the average age of residential housing units seems to, albeit partially,
prove this phenomenon (see figure 38, figure 39, and figure 40).
Figure 37Demographic evolution Figure 38Demographic evolution Figure 39Average age of Figure 40Elderly dependence
2001–2011 2011–2015 (municipalities) residential housing units (2011) figures
The overlap between urban transformations and different housing Thus, the urban transformations taking place within Portuguese
markets frequently uncovers inadequacies in the fulfilment of housing territory today seem to support a divide between coastline and coun‑
needs. These relations are well‑documented in scientific literature, tryside: for the countryside, the prevailing process is rural decline,
with the works of Bramley (2007), Fingleton (2008), Meen and Andrew which contrasts with the urbanization taking place in some civil
(2008) and Mulliner and Maliene (2013) making up some examples. parishes, particularly in the municipal seat (urban entrenchment); for
This highlights their robust territorial character, as well as, above all coastline regions, there is a declining pattern amongst more central
else, the spatial expression brought to housing affordability metrics. In civil parishes in large urban agglomerations (as happens in both metro‑
the Portuguese context, the localized dimension of both phenomena politan areas, and Braga or Setubal), contrasting with urban expansion
allows the creation of a common analysis strategy: the work of Wolf et processes in civil parishes neighbouring these older urban continuums,
al. (2020) approaches the geographic identification of different urban as well as with processes of suburbanization taking place at territory
transformation processes in an integrated manner, intersecting popu‑ level and already merging with more rural areas.
lation growth with artificial urbanity metrics (populational density,
proportion of artificial land use, etc.) and centrality (equipment, Geographical patterns identified by Wolf et al. (2020) will thus allow
services, economic activities, etc.). There is also the merit of using civil for the analysis and interpretation of housing affordability metrics
parish level analysis, as it allows for a more thorough territorial survey developed in this work. Processes of urban transformation promote
(see figure 41). the reading of affordability drivers in face of the defining characteris‑
tics of these phenomena, contributing towards the multidimensional
and integrated understanding of housing affordability, as has been
advocated in international policy forums.
Figure 41Processes of urban transformation in (mainland) Portugal's civil parishes 4.2.2. Housing and housing market
Urban expansion In Portugal, as in most European countries, the free market is the main
The growth of larger urban areas, leading
to a concentration of population, gatekeeping mechanism in housing access. The traditional assump‑
economic activities, or services. tion of free competing markets would want us to consider that
Urbanization housing prices are a mere result of the relationship between supply
The passage from low density land uses
to high density ones, mostly an outcome and demand. However, the housing market has its own characteristics,
of migration to existing urban areas.
which contribute to the emergence of much more complex outcomes.
Urban decline
Population decline in urban areas, leading
to the abandonment of infrastructures
The special characteristics of housing as an asset partially sepa‑
and housing, ageing or economic decline. rate it from the more standard economic theory (e.g. Bourne, 1981;
Rural decline Maclennan, 1977), mainly due to its unique mix of features: i) hetero‑
Population decline in rural areas, which
is generally characterized by agricultural geneity and singularity; ii) immobility; and iii) durability. These distinct
abandonment, the closure of services and elements of housing translate into two important market behaviours:
changes in the landscape.
a) as any decision over a durable asset, decisions on housing transac‑
Counter urbanization
Urbanization processes in rural areas tions are more easily postponed. Furthermore, economic decisions
beyond suburban or peri urban locations. are not only based on housing attributes themselves, but also ponder
Suburbanization investment or hoarding criteria; b) housing tends to incorporate the
The expansion of urban areas beyond
their former limits, normally marked by territorial features of its surroundings and the advantages (or disad‑
sprawl.
vantages) the location factor has within the urban context. Thus, the
housing market is vulnerable to influences of exogenous phenomena,
Note: Adapted from Wolf et al. (2020). whether these are produced through territorial planning or dynamics
arising from varied social and economic activities, which take (or do
not take) place in its surroundings.
Besides this uniqueness and complexity, housing is connected with

important macroeconomic multiplying effects (Meen, 2013). Thus,
global public policies seek to maximize the role of housing as an asset
in investment, or hoarding as a common strategy to foster global
economic growth (Aalbers and Christophers, 2019, and Santos et al.,
2014). This option tends to be implemented alongside unawareness Figure 42Trajectory of housing prices and income
regarding its potential effects on housing needs. This is a recurrent 150
strategy in Portugal's policy options, which can be emphasized by
a few examples from recent years: i) the promotion of the housing
125
market for tourism purposes — from the construction of touristic
Index (2015 = 100)

villages to the transformation of single housing units into touristic
housing, also called local accommodation (Cocola‑Gant and Gago, 100
2019); ii) the investment incentives targeting the housing market as

part of the foreign residence permit programme (Lopes, 2013; Rogers 75
and Koh, 2017); and iii) the implementation of favourable frameworks
for house purchases by large investment funds (Waldron, 2018, and
50
Wijburg et al., 2018), where one of the most notable examples is the
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
recent promotion of REITs (real estate investment trusts) as a special
Year
mechanism to attract global foreign investment to the real estate
Income Housing prices
sector in Portugal.
Although the distinction between endogenous elements in the

housing supply market and exogenous ones is difficult to measure, Table 8Residential housing and housing deprivation metrics in Portugal
it is important to bear in mind some of the consequences that have 2005 2015
been attributed to exogenous phenomena. It is possible to highlight: Residential dwellings by resident population: 0,53 0,57
i) the inflationary pressure on housing prices (Lourenço and Rodrigues, (Source: Statistics Portugal / own calculations)
2014, 2017), moving them away from the evolution path for household Overcrowding rate 16,5 10,3
(Source: Statistics Portugal — link)
income (OECD, 2020) (figure 42), thus contributing to worsening the
Severe housing deprivation rate 7,7 4,7
metrics of housing (un)affordability (Rodrigues et al., 2016; Xerez et
(Source: Statistics Portugal — link)
al., 2019), especially in the most vulnerable groups (table 8); and ii) the
Median of housing cost burden 8,4 13.4
prevalence of an apparent housing oversupply (a common pheno‑ (Source: Statistics Portugal — link)
menon in most developed countries, as highlighted by data compiled Housing overburden rate 4,3 9,1
by (OECD, 2020)) without translating into substantial improvement in (Source: Statistics Portugal — link)
housing deprivation metrics (table 8).
In fact, in Portugal, as shown in table 8, high housing availability (based 4.3. Measuring housing affordability
on the ratio of housing units per inhabitant) has not been reflected
4.3.1. Affordability — measurement challenges and
in the metrics of housing cost burden, which have been aggravating.
approaches
The metrics of housing deprivation and overcrowding, on the other
hand, have shown improvements, although they remain at high levels. Despite methodological challenges, the definition and measurement
In fact, the combined analysis of these more habitual dimensions of housing affordability has been a key challenge in supporting the
of housing shortages have suggested that the problem runs deeper, formulation and implementation of public policies: i) it is a key crite‑
leading to the National Survey of Rehousing Needs (Levantamento rion for the efficient allocation of public subsidies; ii) it allows the
Nacional Das Necessidades de Realojamento Habitacional, 2018), assessment of households’ abilities to bear the burden of mobilizing
conducted by the Institute of Housing and Urban Rehabilitation resources in order to gain access to housing; iii) it constitutes a metric
(IHRU), in collaboration with local authorities (municipalities). This that regulates real estate investment, especially the agents’ deci‑
initiative sought to provide a more accurate and effective portrait sions, be it public or private, to locate investments; and iv) it is a key
of the most serious housing shortages, having identified around dimension taken into account in the analysis of the population's living
26,000 households in unsatisfactory housing conditions, within the conditions and income, with special relevance in defining poverty
most disadvantaged territories and social groups. This survey was the thresholds.
basis for the design and implementation of the 1.º Direito (1st Right)
programme (Decree‑Law No. 37/2018, mentioned above), aimed at As mentioned by (Bramley, 2012), the scientific community has contri‑
conducting a more thorough10 survey and providing intervention buted with a fruitful discussion on the methodological strategies to
instruments (financial included). It should be noted that the 1.º Direito be adopted, and two main lines of methodological development can
programme (which is still under development) had, by June 2020, iden‑ be identified — the residual income perspective and the normative
tified situations of severe housing shortage that already exceeded the perspective.
values identified in the above‑mentioned survey by 77% (IHRU records
i) The residual income perspective
«an approximate 77%» increase in housing shortages, 2020)11.
In this approach, the definition of housing affordability is based on
The role played by the non‑housing aspect in setting housing prices
the construction of an «equivalence level», which allows the effects
naturally has a territorial dimension to it. Characteristics such as place
of household characteristics to be reflected (e.g., household size or
and location are relevant to these differentiated economic roles, thus
composition). In this approach, housing affordability is measured
contributing towards distinct territorial effects that should be taken
through a comparative analysis of the disposable income of several
into consideration when assessing housing affordability.
households, after deducting the necessary expenses considered
essential, excluding housing (Stone et al., 2011). The surplus value is Where the provision of housing is mostly carried out through the
then compared with housing access costs, taking into consideration market, the adoption of the (market) price becomes a trivial and
the most appropriate household characteristics. expeditious option to design this type of metric12. This perspective
also benefits from the consolidated development of techniques for
It is an exercise in which the multiple dimensions of housing rights measuring both house prices and household income: econometric
are more easily integrated, as the act of meeting basic housing needs methods provide an expeditious way of identifying market price (and
translates into allocating differentiated proportions of income by household income) factors in a scientific technical framework which
household income. Borrowman et al. (2017) refer to this approach as combines simplicity, (territorial) versatility, and (spatial) comparability.
a «housing stress» assessment, as it seeks to assess the budget balan‑
cing effort required from households to ensure minimum housing Given the degree of subjectivity at play here, it is important to
conditions, forcing them (or not) to forego the use of income for other realize some of the most common normative references, of which the
purposes. The residual income approach is, however, extremely diffi‑ following stand out:
cult to conduct. The specific details (consumption structures) and
those of the analysis units (the household, the housing unit) require 1. The proposal of consulting firm Wendell Cox Consultancy, in its
a detailed set of data (often inaccessible) and the implementation of renowned publication «Demographia International Housing
complex analytical methods. Affordability Survey» (Cox et al., 2020), adopts the ratio between
the median (total) price of transacted housing and the (median)
In addition to the reported difficulties, this methodological approach income of all households, in a set of different territorial units (309
proves very difficult to implement in a territorial analytical focus metropolitan areas, 7 countries). It associates a level (threshold) of
— which this work aims to achieve. The complexity and difficulty housing affordability with 5 classes: below 3 is affordable, between
in collecting data with the necessary (spatial) granularity make this 3 and 4 is moderately unaffordable, between 4 and 5 is severely
option impractical, not least because it makes the necessary regularity unaffordable, and finally, a ratio above 5 is considered greatly
with which this analytical exercise should be developed unfeasible, unaffordable.
given the very dynamics of the urban transformation processes already 2. The OECD yields a housing affordability metric with the clear
mentioned above. purpose of allowing comparisons between countries. In this case,
the ratio that confronts the average nominal value of housing
ii) The normative perspective prices (for sold housing units) with the metric of (average) nominal
The normative perspective refers to a more immediate and simple disposable income, per individual, is chosen.
method, where housing values are compared with household income. 3. Lastly, the European statistical system (Eurostat and Statistics
Portugal) yields a housing affordability metric that is used to design
European and national public policies: the housing cost overburden to extend the normative approach to the local level, where three
rate. In this case, the metric matches a proportion of households fundamental challenges will be answered: the decision regarding the
which are above a given normative threshold for a ratio that collates instrument of measurement to be adopted (models, assumptions and
housing costs with disposable income, calculated for each one of the functional specifications) and the data and variables (whether primary
households. This ratio is obtained through a representative survey on or secondary, objective or subjective, simple or composite, quantita‑
the population's living conditions, with the sample design providing tive or qualitative).
a maximum territorial detail corresponding to NUTS II. The norma‑
tive threshold for housing affordability is 40% (the maximum share 4.3.2. Methodology
that housing costs represent in a household's income).
The work in development aims to pinpoint a set of metrics that allow
The normative basis for housing affordability is the most common for housing affordability analysis regarding civil parishes in mainland
approach for country‑level (or region‑level) comparisons. Different Portugal. The normative approach is adopted due to its ease of use,
international and national, public and private institutions resort to simplicity and suitability as a support tool for designing public policies.
this type of approach as a way to provide expedient metrics which
The territorialized analysis of housing affordability requires esti‑
allow the comparison and assessment of public policy formula‑
mates to be obtained for the market price of housing units, as well
tions (Anacker, 2019; Hulchanski, 1995; and Li, 2015). These works
as for household income. In the first case, the role of the housing
allow us to realize that the spatial detail (disaggregation) of prices
market regarding housing supply in Portugal enables us to derive
and income metrics largely determine territorial granularity, through
estimates of the value of the entire housing stock, and consequently
which it is possible to analyse the housing affordability phenomenon.
of the (annualized) costs eventually associated with it, from market
For example, the housing cost overburden rate describes an accurate
prices observed in each territory. In the second case, we will make
metric, since it arises from detailed data for each housing unit and its
use of information on household income and its acknowledged
households. However, it has a very limited territorial scope (in Portugal
connection with factors such as qualifications, sectors of activity,
this reaches maximum spatial disaggregation at the level of the large
and occupations, as the most relevant explanatory element of terri‑
NUTS II regions), limiting its usability as an indicator to help local
torial variations in household income.
decision‑making regarding housing policy.
In greater detail, the analytical procedure can be broken down into 3
The complexity of the territorial patterns in Portugal briefly described
sequential phases.
in the previous section, suggests the need to adopt analytical instru‑
ments at the level where housing policies make more sense: at the Phase 1: identifying housing prices and household income drivers.
local/municipal scale. The methodological proposal shown here seeks
i) Housing price explanatory model market configurations on the determination of the market price. Thus,
we consider also including the following: a) a measure of the intensity
In the information provided by Statistics Portugal, the median value of tourist accommodations, b) a metric for the ratio of dwellings occu‑
of sales per square metre of dwellings (€)13 was selected as the market pied under rental contracts, and c) a metric referring to the number of
price metric. This has a territorial breakdown at the municipality level transactions for each territorial unit.
and a quarterly periodicity14.
The relation between the housing prices and housing attributes is
The model of housing price drivers (hedonic price model) can be rarely linear in nature — see, for example, the relationship between
expressed by the mathematical relation: housing prices and size. However, scientific evidence has allowed to
p (X) = b X + ε consider it reasonable to proceed, a priori, with transformations of the
initial variables, which easily establish the mathematical conditions
where: necessary to resort to estimates in a classical linear model. Among
the possible features that arise from transformations of the original
p (X) — housing market price;
variables, the log‑log15 model was selected. Here, the logarithmic func‑
X = (X1, X2, …, Xn) — housing attributes and explanation factors for tion is used for the transformation of the independent variables and
housing prices; the dependent variable. This transformation of variables also allows
for greater interpretative simplicity regarding the linear regression
b = (b1, b2, …, bn) — hedonic prices for each X attribute; model, since the coefficients (b) can thus be interpreted as percentage
changes in the dependent variable — p (X), arising from the 1% varia‑
ε — stochastic component.
tion of variable X — which, in economic terms, refers to the concept
Based on these considerations, the explanatory model of housing of price elasticity for the housing attribute represented by variable X.
prices that we propose to implement considers: i) intrinsic attributes
Finally, it should be noted that we intend to define a variable to explain
— dwelling's age and area; ii) neighbourhood characteristics — popu‑
annual housing costs, so that a direct comparison with income, which
lation density, typology of the surrounding urban area, rate of change
tends to be measured on an annual basis, can be drawn. It is assumed
in the number of housing units); and iii) other location and neighbou‑
that this variable derives directly from the observed market price metric
rhood attributes — defined by the region where the housing unit is
(in the year under analysis), assuming that in a housing market where
located. Considering that the housing market is not limited to the
there is a predominance of owner‑occupied housing acquired through
provision of housing and depends on different dynamics, it is also
credit instruments, it is admissible to consider the conversion of the
important to ponder a set of metrics that consider the impact of these
market value of housing into an annual cost corresponding to a 30‑year
period credit instrument, as an approximation for the annual cost of to adopt some assumptions, this metric allows the development of a
housing. The total amount to be amortized does not only correspond spatial econometric modelling process similar to that previously estab‑
to the house's market price but also to a portion reflecting the financial lished for house prices.
costs of the credit instrument/conversion of the house's market value
regarding the deferred amortization of that same value (interest rate, It should be noted that the approximation between income reported
risk premium, other administrative costs). in the IRS (income tax return) and the effective income of households
is an imperfect approximation, as it tends to underestimate the repre‑
It should also be noted that since the housing price metric shown is sentativity of households with income not always stated in the IRS
per square metre, the actual housing price from which the respec‑ form. On the other hand, the approximation between the concept
tive costs are derived requires the identification of an estimate of of household used in the statistical context, and that of the tax
the housing area. Within this specific work, assuming that the unit of household in the context of tax administration, may also create some
analysis corresponds not to one dwelling but to a set of reasonably biases. In this case, this is due to a selection of households, namely the
homogeneous dwellings located in a given spatial unit (from the muni‑ poorer ones, possibly not being represented as tax households in the
cipality to the statistical sub‑section), means working with the central context of income tax return, as they are exempt from handing in this
values of these variables — the median market price and the median statement due to low levels of income — a factor which will certainly
observable area in the spatial unit. influence the median declared income value.
ii) Explanatory model of household income Lastly, it should be noted that both models have a time gap between
the explanatory variables considered (mostly from census data from
To develop an explanatory model of household income, we identified 2011) and the metrics selected as variables to be explained — house
the census metrics referring to the qualifications, occupations, and prices and household income — which refer to data from 2015. It
respective sectors of economic activity of the employed population, should also be noted that although this differential may suggest a
as key factors. In fact, this set of aspects is consensually referred to in weakness, this is balanced with the fact that: i) the rates of residential
scientific literature as the main explanatory aspect of income distribu‑ mobility are generally low (even regarding the initially referred dura‑
tion, especially in studies that focus on territorial analysis. bility of good standards of housing, but also by other idiosyncratic
The choice of household income metrics benefits from Statistics factors of housing as an asset) and ii) the cycles of urban transfor‑
Portugal's recent divulgence of the metric median value of gross mation processes usually take place in time intervals longer than the
declared income per tax household16, available with spatial disaggrega‑ 5‑year intervals mentioned here (which translates into the reasona‑
tion at the municipality level. It should be noted that, despite the need bility of assuming a territory which generally retains its determining
characteristics, regarding both prices and income). In this context and
for the spatial partitions taken into account, it is considered that there Reconciling the availability of data at these two levels implies the adop‑
are no significant changes in the characteristics of households and tion of a challenging assumption: the explanatory models, developed for
dwellings with very significant influence on the identification of the the county level will have to be assumed to be reasonable approxima‑
main territorial patterns of housing affordability. tions of household income and housing costs drivers at the subsection
levels. That is, the weight (hedonic price) of the explanatory variables,
Phase 2: estimates of household housing costs and income for the estimated at the county level, is assumed to be similar (unchanged) for
statistical sub‑section analysis unit. the different measurement resolutions adopted (the county and the
The use of the above‑mentioned set of metrics, with the municipality statistical subsection units). This assumption ensures that the estimated
‑level resolution with which they are provided, is limiting. amount for housing costs and household income is assigned to any
As previously mentioned, a) housing affordability issues tend to arise dwelling and household located within its considered spatial unit.
most urgently at the household level and concerning the ability to The assumptions mentioned in the previous paragraph fall under
access a specific dwelling; and b) location and neighbourhood features well‑known challenges in spatial analysis. The issues of estimation
are drivers of housing prices (Can, 1992, and Marques, 2012) as well and inference at different geographic scales refer to the evidence,
as of income itself — as an example, Linneman (1980) identified that in many studies, that different scales capture different phenomena,
location and neighbourhood variables of housing surroundings repre‑ even though they may be related to each other — this is known as
sent a 15% to 50% variability in the prices of each dwelling. the Modifiable Areal Unit Problem, (see, for example, Amrhein, 1995,
To ensure a compromise between the resolution of the initial data Fotheringham and Wong, 1991, and Openshaw, 1981, for a detailed
and the desirable (territorial) resolution of the housing affordability discussion). This challenge can also be linked to the debate around the
metrics, we propose to infer housing costs and household income for Ecological Fallacy Problem, which refers to the impossibility of deri‑
spatial partitions determined by statistical sub‑sections. It is assumed ving global models by simple aggregation of individual models and vice
that, at the level of these spatial partitions, it is possible to make versa (see Openshaw, 1984).
assumptions about the homogeneity of the dwellings and households Aware of these challenges, it is considered that:
located therein. The availability of data on the features of the
dwellings and households, with a spatial breakdown at the statistical i) in the housing market, explanatory factors of housing prices are
sub‑section level — a spatial unit that is close to the urbanistic notion reasonably unchanged drivers for models at detailed geographic
of block — opens the possibility of achieving this goal. However, as scales when these are considered within a territorially limited
already mentioned, market price and income metrics are only available housing market (the municipality here is assumed as the spatial
for municipalities. delimitation of each housing market);
ii) homogeneity of higher resolution spatial partitions (the statis‑ divisions. In this case, and given the detail, it is possible to yield summary
tical units) is presumed, further allowing extrapolation of their results at the most disaggregated level — the civil parish. Thus, we have:
aggregate measurable properties as observables of the explanatory
variables of household income and housing prices. Housing cost burden index — established as the proportion of
households residing in subsections (of the civil parish) where the ratio
The process of estimating housing prices and household income for between the estimated (median) value of housing and the (median)
more detailed spatial partitions (statistical sub‑sections) also presumes value of income is greater than 0.4.
the risk of the characteristics of these partitions going beyond the
sample limits of the data considered in the estimation of the models Recovering the notion that the definition of adequate housing
(carried out at the municipal level). The occurrence of this pheno‑ involves a variety of interdependent aspects, difficult to capture
menon increases the uncertainty of the estimation process, aggravated within a single simplified metric, it is important to emphasize that data
by the constraint of accurately measuring the errors arising from this on housing costs and income also makes it possible to present a risk
specification. To mitigate this potential problem, a data pre‑processing assessment of housing unaffordability.
process was carried out, excluding subsections with values lower than In fact, admitting the achievement of housing affordability through
3 records, for the metrics of number of households and number of the mobility of households in search of housing units compatible
dwellings, from the estimation and analysis of results. with their income as a hypothesis, will only make sense if the search
Phase 3: measure affordability at a local scale and describe its territo‑ is considered within the boundaries of their local community — the
rial patterns. municipality. Thus, it is proposed to measure the:
Inspired by the housing cost burden metric proposed in the statistical • Housing unaffordability risk index — defined as the proportion
system (Statistics Portugal/Eurostat), we propose the development of a of resident households (sum of households residing in subsections
territorial housing cost burden index, using the ratio of 0.4 between the that meet the criterion) for which the (median, subsection) income
estimated median housing costs and the estimated median income as a value estimate is lower than the (observed) median income for
normative reference (similar to what is defined in the official metric). all households in the county and where the (median, subsection)
housing cost estimate is higher than the municipality's observed
From this metric, it is possible to develop a spatial analysis that allows median housing cost.
for the identification of territorial patterns of housing affordability
by summarizing these estimates for the usual political administrative Lastly, to better identify the observable territorial patterns of the
metrics here proposed, we recommend an analysis of the recog‑
nized relations of spatial dependence which emerge from territorial
phenomena. As proposed by Waldo Tobler in what is commonly called Table 9Modelling data for housing costs
the «first law of geography» — «everything is related to everything
Variable Description Year Adm. Level Sources
else, but near things are more related than distant things» (Sui, 2004;
Housing costs 2016 Subsection Statistics Portugal:
Tobler, 1970, 2004) — the interaction relations of individuals and their (Median, €/year) and link
Obtained from a 2011 link
activities (including housing choices) with their surroundings tend to link
HOUSINGCOST17 weighted median
make geographical clusters emerge, which can be identified by techni‑ value per m2 of
(a)
ques developed in spatial econometrics. Global (Moran I) and local dwelling sales (€) and
the median area of
(LISA) (Anselin, 1995) metrics of spatial association are the traditional dwellings (see next
variable)
example of these statistical measures of spatial association which
Dwelling area 2011 Subsection Statistics Portugal:
allow us to identify patterns of territorial connection. (median, m2) link
Obtained as a
It should be noted that, doing justice to Tobler's postulate, spatial AREA (a,c) weighted median
retrieved from the
association metrics thus require the prior definition of what is unders‑ number of dwellings
tood as close, i.e., neighbourhood relationships. These are usually by area.
specified through a weight matrix (W) (Elhorst, 2014), generally estab‑ Dwelling age 2011 Subsection Statistics Portugal:
Obtained from a link
lished in an aprioristic way, based on the adoption of a classical notion AGE (a,b,c) link
weighted median of
of distance or neighbourhood relation, i.e., where the meaning of the buildings’ ages
«near» and neighbour is consistent with topological or geometric rela‑ Proportion of rented 2011 Subsection Statistics Portugal:
RENT (c) dwellings (% of total link
tionships in the conventional Euclidean representation of geographic residential dwellings)
space. It's worth bearing in mind, however, that the notion of space for Population density in 2011 Civil Parish Statistics Portugal:
the purpose of territorial analysis has been a subject of dispute, with 2011 (Pop/Km2) link
DENS
Data from Census
several authors calling for a more flexible approach (Bhattacharjee and 2011
Jensen‑butler, 2006; Paelinck, 2013), which goes beyond the purely Number of available 2016 Civil Parish Turismo de Portugal:
geographical notion — an aspect left here for future development. beds in touristic link
TOURISM (a)
activities (except
campsites)
4.3.3. Data Number of housing 2016 Municipality Statistics Portugal:
TRANS
unit transactions link
Tables 9 and 10 provide a brief description of the variables used in this
work, the maximum spatial disaggregation, and the used data sources.
Variable Description Year Adm. Level Sources Table 10Modelling data for income
Evolution of the 2011 Municipality Statistics Portugal Variable Description Year Adm. Level Sources
number of dwellings and link
in the housing stock, 2016 link Median household income (€/year). Municipality Statistics
2011 and 2016 INC Obtained from the median value of Portugal:
(%), only taking gross reported income per household. link
VARHU (a) into account the
Qualification index. 2011 Subsection
cumulative sum of
new housing units Obtained from the number of years
finished per year QUAL (a,c) of each educational level weighted by
between 2011 and the number of individuals with a given
2016 educational level.
Classification of urban 2014 Civil Parish link Proportion of workers per profession 2011 Civil Parish link
areas as predominantly p(%).
urban area, medium A variable for each p is produced.
TERRITORY_TYPE
urban area and And p is:
predominantly rural 1. Representatives of the legislative
area18 and executive bodies, directors and
EMPp (a,d,e)
NUTSIII19 NUTS III 2016 NUTS III executive managers;
2. Skilled farmers and other skilled
workers in agriculture, fishery and
Notes:
(a) Authors’ simple own calculations from primary data;
forestry;
(b) Source: Statistics Portugal dataset for research and development teams; 3. Skilled workers in industry,
data provided with crossing relations between variables; construction and craftsmen.
(c) Source: Statistics Portugal dataset for research and development teams; data provided
with the higher spatial resolution available (subsection / smallest census track);
Proportion of workers of companies 2011 Civil Parish link
(d) Source: Statistics Portugal dataset for research and development teams; with activities in sector s (%).
data provided with higher disaggregation of variable classes; A variable for each s is produced.
(e) Obtained through statistical pre‑processing (variable reduction techniques; variable selection approach). And s is:
1. Extractive industries (B);
2. Manufacturing industries (C);
3. Electricity, gas, steam, hot and cold
EMPs (a,d,e) water, and cold air (D);
4. Information and communication
activities (J);
5. Real estate activities (L);
6. Education (P);
7. Activities employing domestic staff
and production activities for the
producer's own use (T).
Variable Description Year Adm. Level Sources section thus appears as an effort to integrate these predominantly terri‑
TERRITORY_ Classification of urban areas as 2014 Civil Parish link torial phenomena in the debate of housing affordability in Portugal.
TYPE predominantly urban area, medium
urban area and predominantly rural area.
4.4.1. Housing affordability drivers: an analysis at the
NUTSIII20 NUTS III 2016 NUTS III
municipal level
Notes:
(a) Authors’ simple own calculations from primary data;
4.4.1.1. Explanatory factors of housing prices
(b) Source: Statistics Portugal dataset for research and development teams;
data provided with crossing relations between variables; Of the two key components of a housing affordability metric, the
(c) Source: Statistics Portugal dataset for research and development teams; data provided
with the higher spatial resolution available (subsection / smallest census track); housing price is the most difficult to model. Although it benefits from
(d) Source: Statistics Portugal dataset for research and development teams;
data provided with higher disaggregation of variable classes;
a consolidated econometric approach (hedonic price models), the avai‑
(e) Obtained through statistical pre‑processing (variable reduction techniques; variable selection approach). lable data does not always allow the assurance of the optimal modelling
conditions suggested in the literature. Among various constraints are
4.4. Territorial patterns of housing challenges connected to the availability of information. In fact, most of
affordability in Portugal the available databases are fragmented across different agents and usually
have restricted access applied to all or some of the crucial attributes (e.g.,
The Portuguese territorial reality is admittedly contrasting. Demographic
house prices). The collection of public data from real estate portals is also
aspects have revealed a country in rapid transformation. In this context,
not a reasonable alternative. Not only because these portals limit the
it is expected that the diversity of social, economic, and spatial dynamics
mechanisms that can be implemented to consult the information in the
be associated with distinct patterns of housing affordability.
ads, but also because these data have important limitations, which make
Some of these patterns have been closely observed. For example, the spatial analysis proposed here unfeasible — for example, the absence
housing shortages related to the physical conditions of housing have of property georeferencing and the precise identification of transaction
been mostly identified in their direct relationship with poverty (and prices (or, at least, the records of quotation prices).21
inequality) levels (Fernandes et al., 2016), and their territorial expres‑
An explanatory model of the median housing cost was developed from
sion has been especially associated with the most densely populated
the information available at Statistics Portugal, which was derived
areas, namely the two metropolitan regions.
from the variable provided by Statistics Portugal for the median
As mentioned in section 2.1, profound changes in the demographic price per square metre of traded properties (in 201622) in the various
structure and differentiated population dynamics require us to consider Portuguese municipalities. As mentioned above, the estimation of an
the spreading of housing issues to other regions and social strata. This absolute value is obtained through an estimate of the median housing
area, which in turn is obtained from numbers included in the 2011 The association of housing costs is not only an adaptation of the market
census (for the entire housing stock). to a more dynamic territory regarding the housing function. As it has
been argued, there are interdependencies between the housing market
The set of intrinsic housing attributes, neighbourhood and location and economic activities that intersect with it: the example of tourism is
attributes are also mostly derived from metrics made available through the most well‑known and discussed. Thus, the relevance of the variable
the Statistics Portugal databases (census, typification of urban areas). indicating the supply of tourism beds (TOURISMh) in the model here
For the specific case of the tourism intensity metric, the informa‑ presented, is in line with the evidence that has been documented regar‑
tion obtained derives from georeferenced data made available by the ding the effects of tourism on housing costs — places with a higher
Turismo de Portugal programme23. supply of tourism beds tend to also have higher housing costs.
From the available data set, the final model, adjusted to ensure consis‑
tency of the estimates, presents the variables and their respective
Table 11Main results from the housing costs model
estimated weights, as described in table 11.
Coefficient estimate
Housing area (AREAh) is the main explanatory factor for the median Most statistically (Can be interpreted as elasticity: the % effect of
significant(a) explanation increasing 1% of the explanation variable X in the (annual)
housing cost. As this is a recurrent observation in scientific literature, variables (X) amount of housing costs Y)
it should be noted that it is a result which also derives directly from AREAh 0,31%
the model construction options, or in other words, the estimated AGEh ‑0,13%
variable (PRICEh) resulting from the multiplication of the price per RENTh 0,13%
square metre and the area of dwellings in the unit of analysis. DENSp 0,27%
TERRITORY_classe ‑0,07%
As this model is focused on territorial analysis, it is important to stress (reference: TERRITORY_apu)
that population density is the most relevant variable in explaining TOURISMh 0,17%
housing costs. This is an expected result given the common perception TRANSh 0,16%
that housing prices tend to be higher in more densely populated terri‑ VARh 0,08%
torial areas. This observation is further reinforced by the significant NUTSIIIPT111 to NUTSIIIPT187
(reference: NUTSIIIPT170 — Área Metropolitana de Lisboa)
effect of the variable number of urban buildings traded (TRANSh), and
Global explanatory power of the model (R2): 83%
also by the effect of the variable which translates the growth in the
number of buildings (VARh). These metrics have clear territorial asso‑ Notes:
(a) significance based on α < 0,05.
ciation with the more dynamic and more densely populated areas. (b) region dummy variable coefficients are not presented; but significant coefficients are
obtained for the NUTSIII 112, 119, 11A, 11C, 11D, 11E, 150, 16H, 16J, 181 regions.
Figure 43Housing cost (median) based on Figure 44Housing cost (median) based on Figure 45LISA clusters on the estimated housing
(median) market prices and (median) dwelling (median) market prices and (median) dwelling costs. Civil parishes in 2016
characteristics. Municipalities. 2016 characteristics. Civil parishes. 2016
The mapping of the estimation results (figure 43, by municipality), the rest of the surrounding territory. In fact, these areas correspond to
suggests a territorial pattern where housing costs that are estimated the territories where population growth has been registered in recent
to be higher correspond to municipalities with a higher population years or where population attraction still prevails; these are territories
density, aggregating important political and administrative functions, we can easily identify as overlapping the urban transformation pheno‑
mainly at the regional or national level. Also, there is a recognized social mena (referred to in section 2.1) of urban expansion, suburbanization,
and economic dynamism which is relatively higher in comparison to and urbanization. Thus, the metropolitan areas of Lisbon and Porto,
the northern coast (above Porto) and a large part of the municipalities presentation of these metrics for the national territory, we have opted to
along the territory, where the headquarters of the bodies of intermu‑ show a summary analysis at the civil parish level (figure 44). This mapping
nicipal24 communities are located, emerge as those with the highest allows us to identify the relationship between civil parishes (as muni‑
housing costs. Naturally, these are the municipalities where higher cipality headquarters) and their surroundings as the dominant pattern.
values are recorded for variables such as RENT, DENSp or TOURISMh, In fact, as municipality headquarters are territories that concentrate
which partially capture elements of this socio‑economic dynamism. political administrative functions and relevant public services, this encou‑
rages their socio‑economic dynamism to reinforce urbanization processes
Alongside these territories, Chaves and Sines (and some of their (see also figure 37, figure 38, and figure 41 in section 2). Another terri‑
surrounding municipalities) also appear in the group of territories with torial pattern that becomes visible at the civil parish level is the effect
higher housing costs. Although they do not have such relevant political of activities in the tourism sector: coastal civil parishes, and in particular
and administrative functions as the municipalities mentioned above, those in the Algarve region, emerge with the highest housing costs. This
they stand out due to a few unique features that decisively contribute pattern of territorial association between civil parishes is supported by
to their position. For example, Sines is home to the largest national the recurring measure of spatial dependence (LISA), which allows to
port and a whole set of related activities which qualify its territory as identify clusters of territories where more significant spatial relationships
an urban centre of intermediate size and great socio‑economic dyna‑ are concentrated amongst civil parishes with higher housing costs (in
mism; Chaves, thanks to its geographical positioning (near the border red) similar to those described. The opposite phenomenon (territories
and in the centre of a region where a relevant agricultural and lives‑ that group civil parishes with lower housing costs) occurs in the country‑
tock sector prevails), is also an important centre of socio‑economic side, particularly in the northern and central cross‑border territories.
dynamism. Finally, in the same vein, comes the Algarve region — it is
the recognized central role of the tourism sector in this region that These territorial aspects of the explanatory model of housing costs
supports its leadership amongst territories with the highest housing reveal the expected relationship between higher population concen‑
costs. In this specific case, and regarding the socio‑economic dyna‑ trations and more economically dynamic territories as an explanatory
mism of tourism activities, it is worth highlighting the effect of tourist factor of higher housing costs. In addition to exposing a country with
accommodation on the real estate market, namely through the compe‑ marked territorial contrasts — a conducive factor for inequalities and
tition between the residential and touristic roles of accommodations, inequities — this supports the goal of this paper: the need to confront
and its effects on housing prices, as mentioned above. housing costs with income from a territorial perspective. The desired
sustainment of social and economic dynamism requires that housing
The use of the explanatory model described above (municipal level) to costs be supported by compatible incomes, keeping this relationship at
yield housing cost estimates at the micro level (by subsection) allows us adequate levels to support households’ living conditions.
to create more detailed housing affordability metrics. For an integrated
4.4.1.2. Explanatory factors of household income Table 12Main results from the explanatory factors of income
The development of an explanatory model for the territorial variability Effect from increasing 1% of the Expl.
Explanatory variable Variable in yearly income
of household income is supported by theoretical evidence, especially
QUALf 38%
in regional science, of the triad of socioeconomic factors that deter‑
EMPs_s_1(C) 13%
mine the competitive advantages of different territorial units. In this
EMPs_4(J) 9%
case, the qualifications of the population, the dominant sectors of
EMPs _7(T) ‑6%
activity and the position of firms in these value networks (see, for EMPp _2 ‑26%
example Pavitt, 1984), are indirectly measured here by the structure EMPp _3 ‑29%
of dominant occupations. These key dimensions are further related TERRITORY_amu ‑6%
to evidence from urban studies, which tend to point to these same NUTSIIIPT111 to NUTSIIIPT187
elements as the main factors explaining the segregation of individuals (reference: NUTSIIIPT170 — Área Metropolitana de Lisboa))
at an urban level. Social and cultural factors are also often associated Global explanatory power of the model (R2): 84%
with the latter spatial level of analysis (see, for example Wong et al.,
Notes:
2007). This diversified theoretical framework was integrated in the (a) significance based on α < 0,05;
income estimation model by collecting and incorporating a set of (b) region dummy variable coefficients are not presented; but significant coefficients are obtained
for the NUTSIII 111, 112, 119, 11A, 11B, 11C, 11D, 11E, 150, 16B, 16J, 185 regions
variables (see section 3.3) that cover these different income drivers.
The variable selection procedure, together with the modelling process,
The need to identify, amongst the initial set of available metrics, those
allows us to highlight the role of activity sectors with population
that measure distinct territorial phenomena, required a prior analysis
employed in manufacturing, information and communication activities
of primary data in order to distinguish, in the detail provided on occu‑
(which includes information technology) and family‑based economic
pations (10 types, at level 1) and economic sectors (21 sectors, at level
activities (where most of the labour is family‑based; and traditional
1 of disaggregation), which ones are effectively relevant to explain
agriculture, for example, is particularly relevant). At the occupation
income differences between the territorial units considered in the
level, we emphasize skilled professionals in agriculture, fisheries,
modelling process (municipalities). This procedure is also necessary in
and forestry, and skilled workers in industry and construction. These
econometric terms, strengthening the assumption of the explanatory
results are in line with important clusters of territorial specialization
variables’ independence and, of course, minimizing the occurrence of
in the Portuguese economy: for example, the role of the manufactu‑
undesirable multicollinearity. The relevant variables and model results
ring industry (footwear, clothing, furniture, plastics, metalworking) in
are shown in table 12.
the regions of Vale do Ave and Vale do Sousa, Estarreja‑Aveiro‑Águeda
and Leiria‑Marinha Grande. The variables on workers’ occupations association of qualification levels with areas of greater demographic
reveal the sophistication and specialization of many of these economic and economic dynamism.
activities, with the above‑mentioned clusters coinciding with the terri‑
tories where there is a higher relative proportion of specialized/skilled Similarly to the exercise carried out for the housing costs model, the
workers (namely, the industrial cluster of the Vale do Sousa region, observation of geographical patterns regarding summary values (at
or more modern agricultural clusters, such as the western region or the civil parish level) of income estimates carried out for subsections
Douro). In the same vein, these results show the prevalence of tradi‑ allows us to observe, in figure 46 and figure 47, a strong connec‑
tional (and subsistence) agriculture in the countryside and remote tion between higher incomes and the territorial units where there is
regions, where not only is this sector of activity proportionally rele‑ greater demographic and socioeconomic dynamism — see also figure
vant, but a greater ratio of occupations requiring lower levels of 37 and figure 38. It is also worth noting the repetition of the pattern
qualification stand out. which differentiates civil parishes that are municipality seats from
others, once again revealing many more marked differences in cross
The differentiated generation of income associated with different ‑border and inland regions. These aspects reinforce the possibility of
economic activities is visible in the estimated effects of the varia‑ important negative effects on housing affordability, which will now
bles. As an example, territories where there is a higher ratio of skilled be analysed below.
workers in industry and construction (EMPp_3) have associated lower
incomes (INC), keeping all other effects static. This is explained by the With respect to territorial patterns resulting from the spatial depen‑
fact that the variable is associated with territorial clusters of economic dence of income, the patterns are slightly different from those
activities where the business model is historically very dependent on previously identified for housing costs. In this case, figure 48 shows
low wages (e.g., the clothing and footwear sectors). A similar pheno‑ regions in the north countryside where several lower‑income civil
menon occurs with farmers and skilled professionals in agriculture, parishes tend to be grouped together surrounded by other low‑in‑
fishing and forestry (EMPp_2). come civil parishes. This pattern seems to conflate with mountainous
and particularly hilly areas, where severe phenomena of rural decline
Lastly, the prominent role of the qualifications index (QUAL) in (and depopulation) prevail — territories coinciding with the Freita,
explaining different income levels is explained by the fact that this Montemuro and Alvão mountains, and parts of the Gerês and
metric aggregates several phenomena: i) real income gains, well Montesinho mountains, are identifiable. Effectively, these territorial
recognized in the literature, as a function of higher qualifications areas correspond to remote civil parishes, dominated by traditional
(maintaining other effects static), ii) most qualified and best paid agricultural activities, in clear decline, and as the model suggests,
activities in any sector tend to cluster territorially (hence the favou‑ connected to comparatively very low incomes.
rable numbers in areas around Lisbon and Porto) and iii) strong
Figure 46Estimated income based on (median) Figure 47Estimated income based on (median) Figure 48LISA clusters by estimated income.
income declared to tax authorities and (median) income declared to tax authorities and (median) Civil parishes in 2015
family characteristics. Municipalities. 2016 family characteristics. Civil parishes. 2016
The pattern that brings territorial units with the highest incomes together Also noteworthy for its size, is the large area around Lisbon, which
matches the pattern joining coastal areas and intermediate urban areas almost completely overlaps the limits of its metropolitan area (the
with the greatest dynamism. The latter are typically associated with urba‑ Lisbon Peninsula region and the Setúbal Peninsula region) and also
nization and urban expansion, either in coastal territories (larger patches) includes contiguous areas belonging to other regions, such as the
or inland urban centres surrounded by declining rural areas (polarizing Middle Tagus region.
urban centres), but clearly yielding a dynamism (and respective higher
household income) that distinguishes them within the regional pattern.
4.4.2. Territorial patterns of housing affordability Figure 49 and figure 50 present the territorial patterns of the
in mainland Portugal housing burden index and the housing (un)affordability risk, presen‑
ting numbers of the proposed metrics calculated for civil parishes
The estimated housing costs and household incomes present a terri‑
(the territorial unit of analysis).
torial distribution that conflates with the territorial reality usually
perceived for these phenomena. This aspect reinforces the sustainabi‑
lity and consistency of the methodology proposed here, allowing us to
move on to the analysis of the two metrics proposed in section 3.2.
Figure 49Housing cost burden index Figure 50LISA clusters of housing cost Figure 51Housing unaffordability risk Figure 52Bivariate LISA clusters for costs
burden index and income
4.4.2.1. Housing cost burden On the country's central coast, intermediate urban areas with a strong
presence of the secondary sector in their economic structure, emerge
In figure 49, we can find the mapping of the housing cost burden
as territories where the burden of housing costs is also relevant. This
index, calculated for the civil parishes of mainland Portugal, comple‑
includes the urban continuum between Aveiro, Ílhavo, Vagos, Oliveira
mented by figure 50, which allows us to identify the territorial
do Bairro, and Águeda. Further down, with similar features, an almost
groupings that share one of the four types of association provided by
continuous corridor connects Pombal to Mafra, also including Leiria
the spatial dependence measure.
and Marinha Grande. Lastly, the Alentejo coastline, in particular urban
The data reveals that coastal territories, on the periphery of the most areas around Sines and Santiago do Cacém, where important economic
densely populated urban areas (Lisbon and Porto), are those where the activities of the secondary sector prevail, also stand out as territories
burden of housing costs tends to affect households the most. In these with a higher burden of housing costs.
territories, it is possible to identify three more or less continuous
A central element in this analysis is the strong connection which can
patches, which we can associate to differentiated social, economic,
be visually established between the spots with the highest housing
and territorial characteristics.
cost burden and most noticeable urban transformation processes
One of these territorial groupings, where the burden of housing costs is (of growth, expansion, densification), revealing a housing market that
comparatively more relevant, can be found on the northern coastline, in tends to be less efficient in meeting housing needs.
a territorial continuum covering municipalities such as Viana do Castelo,
The Algarve, which stands out as the territory where the burden of
Ponte de Lima, Barcelos, Vila do Conde, Braga or Guimarães. Going back
housing costs is most intense, is also the one region where explanatory
to what was previously mentioned (section 2), this is a territory that has
factors are most relevantly associated with the weight of tourism acti‑
observed a process of population growth throughout the first decade
vities. The economic specialization of this territory has implications
of the 21st century. However, it is also a territory where several pheno‑
both on the competitiveness this activity exerts on the use of housing
mena of urban transformation are taking place, thus creating a complex
and the fact that this economic sector is often connected with relati‑
pattern by combining phenomena of urbanization, urban decline, rural
vely low wages (which has an impact on household income). Hence,
decline, and counter‑urbanization. The concentration of such diverse
it should be noted that the effects of touristic activity will also have
phenomena, in a densely populated territory with a young demogra‑
partial relevance for some of the above‑mentioned coastal territories,
phic structure (see figure 40), explains the relatively high housing costs
although this is not as clear as in the case of the Algarve coastline.
in a market that, due to its features, tends to be less efficient when
compared to complex urban transformations, which tend to occur in a Lastly, in the Lisbon and Porto metropolitan regions, we find terri‑
more volatile fashion, as this diversity seems to indicate. tories with comparatively moderate levels of housing cost burdens.
Notwithstanding, the fact that the prices registered in the housing quite reasonable degree of preservation. However, households resi‑
market are the highest in the country, it is also in these regions that we ding there tend to be older, from households that have since greatly
find the highest household incomes, contributing decisively to a greater reduced in size and, therefore, have relatively lower incomes.
moderation of the housing cost burden metric. Another aspect that
adds to the moderation of this metric is the preponderance of a housing Thus, these territories tend to be connected to other housing
stock with features which tend to culminate in lower housing costs. In shortages, resulting from the inadequacy of dwellings’ physical
fact, Lisbon and Porto, for example, have a relevant ratio of dwellings in characteristics, either due to size or the prevalence of older and
decay or presenting other characteristics revealing greater inadequacy precarious buildings. These territorial characteristics reinforce the
(for example, small areas in relation to the size of households). This is possible association between housing unaffordability risk metrics
one of the limitations of this type of normative and simplified housing as a complementary mechanism to alert us towards the existence of
affordability metrics, which tend to not consider the effects of composi‑ other dimensions of housing needs, which should be explored and are
tion/adequacy of the housing stock. When these other elements, which strongly interconnected with housing costs, placing the populations in
define housing shortages, tend to be very relevant, the housing cost situations of increased housing risk.
burden metric tends to be more biased. In general, as is to be expected, the risk of housing (un)affordability is
higher in territories where rural decline phenomena, associated with
4.4.2.2. Housing unaffordability risk greater demographic ageing, prevail. These territories tend to house
Since the measurement of the housing cost burden can be biased by an ageing population, on low incomes, with very small family units
heterogeneity effects in the territorial units of analysis, the housing contrasting with the size they assumed in a prior stage of their lives.
(un)affordability risk index makes it possible to identify scenarios Furthermore, as these are territorial areas in decline or surrounded
where the housing cost burden is not an evident element. However, by such processes of decline, the state of conservation of buildings
when analysed independently, the characteristics of dwellings and tends to worsen, and dynamism is practically non‑existent within the
the households residing therein suggest possible risks of housing housing market.
unaffordability.
It should also be noted that figure 16 allows us to scrutinize these
The mapping of housing (un)affordability risk allows us to identify the phenomena a little further — it is the inland territories, recognized
territories where rural decline phenomena prevail as those where these for their serious ageing and depopulation problems, that emerge as
risks are more evident. In fact, within these territories, the housing those where residing households are at greater risk of housing unaf‑
stock features show relatively high‑cost housing — for example, fordability. In these territories, the municipal seat civil parishes are
single‑family dwellings, with unsparing areas, in many cases with a grouped in the same class, reinforcing the idea that they play a central
role as urban centres, where incomes are substantially higher than in evident. Depopulation and lack of economic and social dynamism
their surrounding territories, thus increasing population attraction. have removed these territories from the market valuation logics, thus
However, these apparently more beneficial conditions may also trans‑ contributing to very low housing costs, which is not unrelated to the
late into greater housing unaffordability, especially for the surrounding state of abandonment and degradation of housing units in these terri‑
populations, who in case of need and while maintaining their original tories (see Borges et al., 2012; E. Castro et al., 2013).
income, will not be able to afford housing in these urban centres. This
aspect may thus contribute to reduce the attraction effect of these
urban centres on their surrounding areas, weakening what could repre‑ Figure 53Housing cost burden index
sent an alternative (or complement) for consolidation in face of direct 1.00
«competition» in terms of housing affordability conditions, with more

dynamic territorial areas. In fact, demographic data helps support this
0.75
reading, since these inland urban civil parishes have also registered
Housing cost burden index

significant population decreases, although not as sharp as those in
surroundings territories.
0.50
To support the link between housing affordability and the main 0.35 0.34
urban transformation processes, figure 17 and figure 18 try to provide 0.29 0.29
0.25
some clues. As can be seen, the urban transformation processes
0.16
connected with decline are those in which the housing affordabi‑ 0.13
lity metrics appear to be more favourable to satisfying households’

0
needs. However, in larger cities where this occurs (Lisbon and Porto), Urbanisation Counter- Urban Rural Urban Suburbanisation
-urbanisation expansion decline decline
this phenomenon has been associated with exogenous effects, such
as i) the disturbance of the housing market by economic logics
(tourism and other economic activities), ii) a gentrification process,
iii) processes actively promoted by public bodies (for example, the
attraction of foreign residents with very high incomes) or even iv)
economic financial logics (investment in real estate by large corpo‑
rations, in an international logic which is not adapted to the needs
of most local populations). In rural territories, the rationale is more
Figure 54Housing affordability risk index the civil parish level, allowing us to identify the diversity of territorial
0.80 patterns which characterize housing affordability.
Even after opting to analyse levels of affordability by statistical

0.60
subsection, the available data requires a demanding process of collec‑
Housing unaffordability risk index

tion, transformation, measurement, and modelling. The scarcity of
studies focusing on the relationship between housing affordability and
0.40 0.39 0.37 territorial context makes this effort very relevant, as it allows to signal
0.35 0.36
phenomena at a local level, for which customized solutions can and
0.28
0.24 should be thought out, even though some of their drivers are relevant
0.20 at a macro level and require this global integration.
In the analysis of housing needs, we have tried to emphasize the urban

0 transformation processes as exogenous factors, conditioned to a large
Urbanisation Counter-
-urbanisation
Urban
expansion
Rural
decline
Urban
decline
Suburbanisation extent by the dynamics of demographic change itself, which unavoi‑
dably determine housing shortages, not only in quantitative terms,
but also by directly impacting market mechanisms. This results in the
most visible mismatch between housing supply prices and the finan‑
4.5. Concluding remarks
cial capacity of households seeking housing. Mainland Portugal is
This work presents two major challenges. On one hand, the discus‑ admittedly a coastal territory (with strong contrasts between inland
sion of concepts connected to housing affordability, such as adequacy, and coastal locations), bipolar (through the lenses of the two metro‑
access, availability and need; and, on the other hand, making them politan areas of Lisbon and Porto) and increasingly urbanized. Even
operative at a local level in order to assess the territorial patterns in regions subject to processes of marked population decline and
resulting from the (un)balances between housing prices (translated ageing, there are signs of urban consolidation and expansion in major
into costs) and household income. As it is not possible to work with urban centres. The analysis of the housing (market) phenomenon is
micro data at the housing and household level, it was decided the obviously not, and cannot be, dissociated from these territorial dyna‑
statistical subsections of Statistics Portugal would form the basic unit mics, since housing supply has been a function increasingly reserved
for the preparation of this analysis. For reading and interpretation for the market, but the rules and criteria it establishes are not always
purposes, the results integrated for mainland Portugal are presented at the most appropriate for the territories’ desired social and economic
development.
Moreover, market mechanisms themselves tend to be opaque, and it is capacity and robustness, they performed differently for different
very difficult to monitor both the quantity and type of housing supply, parts of the territory. In this regard, the lower accuracy of the housing
its prices, and drivers. Data on the real estate market is not as trans‑ cost estimates for the main national urban centres (Lisbon and Porto)
parent and accessible as would be desirable, which can be explained stands out. These two cases not only reveal existing information weak‑
by the existence of multiple agents who benefit from this asymmetry nesses, but also suggest that the estimates’ lower level of robustness in
of information. This aspect is outside the scope of this reflection, but these specific cases can be explained by the housing market's features,
it is important to raise an alert concerning the need for regulatory which contain elements that make them autonomous in light of the
intervention in this area. Additionally, there is a lack of culture, rooted country's more global logic.
in public bodies, regarding the collection and supply of information,
as well as the implementation of qualified, effective, and consistent Notwithstanding the limitations that were exposed throughout the
decision‑making processes. Although some efforts are recognized, work, the results are reliable, allowing for the calculation of two
there is still insufficient data to analyse shortages and access to housing affordability metrics: i) the housing expenditure burden
adequate housing with the desirable granularity and precision, not to index and ii) the housing unaffordability risk index. The former
mention the devaluation of prospective analyses that would allow a assesses the ratio between estimated values of housing and household
timely response to these needs. income, with civil parishes as the unit of analysis and accounting for
households living in subsections where the ratio is higher than 0.4;
This chapter presents an innovative approach based on existing statis‑ the latter analyses the proportion of households living in subsections
tical data, seeking to measure and analyse housing affordability in with a median income lower than the median of the municipality, and
Portugal from the perspective of the territorial disparities emerging housing costs higher than the median housing cost in the municipality.
from them. The methodology followed for this research involved the
development of two models, one for income and the other for housing As expected, the results show strong spatial heterogeneity in housing
prices, at a municipal level, allowing for an estimation of prices affordability patterns and significant evidence of spatial dependence,
(costs) and income at the micro level (of the statistical sub‑sections). thus claiming the importance of a territorial outlook in the design
The focus on territorial analysis also led to the adoption of strate‑ of public policies which impact housing. These interventions can be
gies for identifying spatial clusters which emphasize the effects of housing policies specifically, or initiatives from other domains (social
spatial autocorrelation, thus making it possible to group analysis and economic), which, as can be seen here, exert substantial effects on
units (civil parishes) not only by their features, but also by the spatial the housing market and, consequently, can effectively contribute to
relationships they establish with their geographical surroundings. meet housing needs.
Although it was possible to ensure the models’ reasonable explanatory
Lastly, one of the most important messages of this work is the identi‑ Ambrose, B.W. and M. Diop, (2018), «Information asymmetry, regulations
fication of housing affordability challenges in the territories located and equilibrium outcomes: theory and evidence from the housing rental
between the largest cities. On the one hand, regarding the coastline, market», Real Estate Economics.
small and medium‑sized cities, which conflate with urbanization,
Amrhein, C.G., (1995), «Searching for the elusive aggregation effect:
urban expansion and suburbanization processes, and have been serving
evidence from statistical simulations», Environment and Planning A, 27(1),
as alternatives to the larger urban centres (Lisbon and Porto), see the
105–119.
most negative numbers in both metrics, putting them in a challenging
position to achieve reasonable housing affordability levels. On the Anacker, K.B., (2019), «Introduction: housing affordability and affordable
other hand, amongst inland regions, consolidation phenomena in small housing», International Journal of Housing Policy, 19(1), 1–16.
urban and rural centres, which may play a key role in countering depo‑
pulation, also seem to face housing affordability issues, which place Anselin, L., (1995), «Local indicators of spatial association — LISA»,
Geographical Analysis, 27, 93–115.
them in similar positions to those of more socially and economically
dynamic territories. Baker, E., R. Bentley, L. Lester and A. Beer, (2016), «Housing afforda‑
bility and residential mobility as drivers of locational inequality», Applied
References Geography, 72, 65–75.
Aalbers, M. and B. Christophers, (2019), «Centering housing in political

Bhattacharjee, A. and C. Jensen‑Butler, (2006), «Estimation of spa‑
economy», CIDADES, Comunidades e Territórios, 0(38).
tial weights matrix, with an application to diffusion in housing demand»,
in Centre for Research into Industry, Enterprise, Finance and the Firm.
Abramsson, M. and E. Andersson, (2016), «Changing preferences with
ageing — housing choices and housing plans of older people», Housing,
Bitter, C. and D. Plane, (2013), «Housing Markets, the Life Course, and
Theory and Society, 33(2), 217–241.
Migration Up and Down the Urban Hierarchy», In D. F. Clapham, W. A. V
Clark, and K. Gibb (Eds.), The SAGE Handbook of Housing Studies (pp. 295–312),
Allen, J., (2006), «Welfare regimes, welfare systems and housing in Southern
SAGE Publications.
Europe», European Journal of Housing Policy, 6(3), 251–277.
Bogdon, A.S. and A. Can, (1997), «Indicators of local housing affordability:

Alves, S., (2020), «Divergence in planning for affordable housing: a compara‑
comparative and spatial approaches», Real Estate Economics, 25(1), 43–80.
tive analysis of England and Portugal», Progress in Planning, 100536.
Borrowman, L., G. Kazakevitch and L. Frost, (2017), «How long do

households remain in housing affordability stress?», Housing Studies, 32(7),
869–886.
Bourne, L. S., (1981), The geography of housing, Edward Arnold. Cox, W., H. Pavletich and T. Polidi, (2020), 16th Annual demographia inter‑
national housing affordability survey.
Bramley, G., (2007), «The sudden rediscovery of housing supply as a key
policy challenge», Housing Studies, 22(2), 221–241. de Castro, E.A., J.M. Martins and C.J. Silva, (2015), A demografia e o país:
previsões cristalinas sem bola de cristal, Gradiva.
Bramley, G., (2012), «Affordability, poverty and housing need: triangulating
measures and standards», Journal of Housing and the Built Environment, 27(2), Elhorst, J. P., (2014), Spatial econometrics: from cross‑sectional data to spatial
133–151. panels, Springer.
Butler, T. and C. Hamnet, (2013), «Social geographic interpretations of Fernandes, C., N. Crespo and N. Simoes, (2016), «Poverty, richness, and
housing spaces», In D. F. Clapham, W. A. V Clark, and K. Gibb (Eds.), The SAGE inequality: evidence for Portugal using a housing comfort index», In Journal of
Handbook of Housing Studies (pp. 147–162). Economic and Social Measurement (Vol. 41, Issue 4, pp. 371–394), IOS Press.
Can, A., (1992), «Specification and estimation of hedonic housing price Ferrão, J., (2019), «O estado da habitação em Portugal: Uma trajetória
models», Regional Science and Urban Economics, 22(3), 453–474. promissora mas ainda excessivamente em aberto», Cidades, Comunidades e
Territórios, 2019(38), 53–54, link
Carmo, A., J. Ferrão and J. Malheiros, (2014), «Geografias do estado
social: reorganização territorial, habitação e urbanismo», Estado Social: Fingleton, B., (2008), «Housing supply, housing demand, and affordability»,
De Todos Para Todos, 167–200. Urban Studies, 45(8), 1545–1563.
Castro, E., M. Marques, J. Marques and M. Viegas, (2020), «Demografia e Fotheringham, A. S. and D.W.S. Wong, (1991), «The modifiable areal unit
economia: um modelo regional integrado de estimação da população portu‑ problem in multivariate statistical analysis», Environment and Planning A,
guesa», Revista Portuguesa de Estudos Regionais, 55, 18. 23(7), 1025–1044.
Cocola‑Gant, A. and A. Gago, (2019), «Airbnb, buy‑to‑let investment Harvey, D., (2006), «Space as a keyword», In N. Castree and D. Gregory (Eds.),
and tourism‑driven displacement: a case study in Lisbon», Environment and David Harvey: A Critical Reader (pp. 270–293), Blackwell Publishing Ltd.
Planning A: Economy and Space, 0308518X1986901.
Henley, A., (1998), «Residential mobility, housing equity and the labour
Costa‑Font, J., D. Elvira and O. Mascarilla‑Miró, (2009), «“Ageing in market», The Economic Journal, 108(447), 414–427.
place”? exploring elderly people's housing preferences in Spain», Urban
Studies, 46(2), 295–316. Hulchanski, J. D., (1995), «The concept of housing affordability: six con‑
temporary uses of the housing expenditure‑to‑income ratio», Housing Studies,
10(4), 471–491.
IHRU regista aumento das carências habitacionais «na ordem dos 77%.» Marques, J. L., J. Wolf and F. Feitosa, (2021), «Accessibility to primary
(2020, June), TVI24. schools in Portugal: a case of spatial inequity?», Regional Science Policy and
Practice, 1–15.
Kurlat, P. and J. Stroebel, (2015), «Testing for information asymme‑
tries in real estate markets», In Review of Financial Studies (Vol. 28, Issue 8, Marques, J., (2012), «The notion of space in urban housing markets»
pp. 2429–2461), Oxford University Press. [PhD's thesis, Universidade de Aveiro], In Department of Social Sciences,
Policies and Territory: Vol. PhD thesis.
Lefebvre, H., (1991), The production of space.
Marques, J. and P. Batista, (2021), «A noção de espaço e a ciência regional»,
Levantamento Nacional das Necessidades de Realojamento Habitacional, In F. Carbalho‑Cruz and J. Marques (Eds.), 30 anos de ciência regional em perspe‑
(2018), link tiva, Edições Almedina.
Li, J. V., (2015), «Recent trends on housing affordability research: where are Marques, J., P. Batista, E.A. Castro and A. Bhattacharjee, (2020),
we up to?», SSRN Electronic Journal. «Housing consumption», In V. Chkoniya, A. O. Madsen, and P. Bukhrashvili
(Eds.), Anthropological Approaches to Understanding Consumption Patterns and
Linneman, P., (1980), «Some empirical results on the nature of the hedonic
Consumer Behavior (pp. 265–285). IGI Global.
price function for the urban housing market», Journal of Urban Economics,
8(1), 47–68. Marques, J., M. Tufail, J. Wolf and M. Madaleno, (2021), «Population
growth and the local provision of services: the role of primary schools
Lopes, L., (2013), «Autorização de residência para actividade de investi‑
in Portugal», Population Research and Policy Review, 40, 309–335.
mento» (ARI — golden visa), Gaudium Sciendi, 4, 202–208, link
Meen, G. and M. Andrew, (2008), «Planning for housing in the post‑Barker
Lourenço, R. and P.M.M. Rodrigues, (2014), «Dinâmica e contraste dos
era: affordability, household formation, and tenure choice», Oxford Review of
preços da habitação em Portugal e Espanha», Boletim Económico do Banco
Economic Policy, 24(1), 79–98.
de Portugal, 4, 41–54.
Meen, G., (2013), «Housing and the economy», In D. F. Clapham,
Lourenço, R. and P.M.M. Rodrigues, (2017), «Preços da habitação em
W. A. V Clark, and K. Gibb (Eds.), The SAGE Handbook of Housing Studies
Portugal — uma análise pós‑crise», Revista de Estudos Económicos do Banco
(pp. 251–273), SAGE Publications.
de Portugal, III(4), 45–62.
Mulliner, E. and V. Maliene, (2013), «Austerity and reform to affordable
Maclennan, D., (1977), «Some thoughts on the nature and purpose of
housing policy», Journal of Housing and the Built Environment, 28(2), 397–407.
house price studies», Urban Studies, 14(1), 59–71.
Mulliner, E. and V. Maliene, (2015), «An analysis of professional per‑ UN‑Habitat, (2017), The Human Rights‑Based Approach to Housing and Slum
ceptions of criteria contributing to sustainable housing affordability», Upgrading.
Sustainability, 7(1), 248–270.
Tobler, W., (1970), «A computer movie simulating urban growth in the
OECD, (2020), OECD Affordable Housing Database. detroit region», Economic Geography, 46, 234–240.
Openshaw, S., (1984), «Ecological fallacies and the analysis of areal census Tobler, W., (2004), «On the first law of geography: a reply», Annals of the
data», Environment and Planning A, 16(1), 17–31, link Association of American Geographers, 94(2), 304–310.
Openshaw, S., (1981), «The modifiable areal unit problem», Quantitative Waldron, R., (2018), «Capitalizing on the state: the political economy of
Geography: A British View, 60–69. real estate investment trusts and the ‘resolution’ of the crisis», Geoforum, 90,
206–218.
Paelinck, J.H.P., (2013), «Spatial econometrics: a personal overview»,
Studia Ekonomiczne, 152, 106–118. Wetzstein, S., (2017), «The global urban housing affordability crisis»,
Urban Studies, 54(14), 3159–3177.
Pavitt, K., (1984), «Sectoral patterns of technical change: towards a taxo‑
nomy and a theory», Research Policy, 13(6), 343–373. Wijburg, G., M.B. Aalbers and S. Heeg, (2018), «The financialisation of
rental housing 2.0: releasing housing into the privatised mainstream of capi‑
Rodrigues, C. F., R. Figueiras and V. Junqueira, (2016), «Desigualdade tal accumulation», Antipode, 50(4), 1098–1119.
do rendimento e pobreza em Portugal — as consequências sociais do pro‑
grama de ajustamento», Fundação Francisco Manuel dos Santos. Wolf, J., P. Batista and J. Marques, (2020), «Processos de transformação
urbana: uma tipificação baseada na urbanidade, na centralidade e na evolução
Rogers, D. and S.Y. Koh, (2017), «The globalisation of real estate: the poli‑ da população dos territórios», CIDADES, Comunidades e Territórios, 41(41).
tics and practice of foreign real estate investment», International Journal of
Housing Policy, 17(1), 1–14. Wong, D. W., M. Reibel and C.J. Dawkins, (2007), «Introduction —
Segregation and neighborhood change: where are we after more than
Santos, A. C., N. Teles and N. Serra, (2014), Finança e habitação em Portugal. a half‑century of formal analysis», In Urban Geography (Vol. 28, Issue 4,
pp. 305–311). Taylor and Francis Group.
Stone, M. E., T. Burke and L. Ralston, (2011), The residual income approach
to housing affordability: the theory and the practice. Xerez, R., E. Pereira and F. Cardoso, (2019), Habitação própria em Portugal
numa perspetiva intergeracional.
Sui, D. Z., (2004), «Tobler's first law of geography: A big idea for a small
world?», Annals of the Association of American Geographers, 94(2), 269–277.
Chapter 5
Short‑term rentals and real estate
prices in Lisbon: boom and ban 25
Duarte Gonçalves, Susana Peralta and João Pereira dos Santos
5.1. Introduction
According to The Guardian, in June 2019, the mayors of ten European understood that this decision might hamper them from regulating the
cities demanded help from the European Union to tackle the negative presence of Airbnb in their cities through the usual regulation of local
impact of Airbnb and other holiday rental websites on their cities.26 activity, urban planning and housing rules.
These platforms, according to the mayors of Amsterdam, Barcelona,
Berlin, Bordeaux, Brussels, Krakow, Munich, Paris, Valencia and Vienna, The mayor of Lisbon was not amongst the subscribers of this letter,
were changing the nature of the cities and pricing the locals out of possibly because, by June 2019, Lisbon was already into its seventh
city housing. The mayors wanted the explosive growth of short‑stay month of imposing restrictions on short‑term holiday rentals. Indeed,
rentals to be on the agenda of the new European Commission, which in November 2018, the municipality of Lisbon had banned new
would step in one month later. property registries for the short‑term rental market in some of the city
centre's historic neighbourhoods. More recently, he wrote an open
The cities wanted to be able to counter the adverse effects of the letter in the British newspaper The Independent mentioning the social
boom in short‑term rentals, deemed to be the cause behind rising costs of this phenomenon and reassessing post‑pandemic priorities:
rents for full‑time residents and the continuing «touristification» «Essential workers and their families have increasingly been forced out
of neighbourhoods, by «introducing their own regulations depending as Airbnb‑style holiday rentals have taken over a third of Lisbon's city
on the local situation». centre properties, pushing up rental prices, hollowing out communi‑
ties and threatening its unique character.»27 However, the mayor had
The mayors were reacting to a decision by the Advocate General of
not always been in favour of curbing the growth of Airbnb and similar
the European Court of Justice, who issued a non‑binding opinion
platforms. In 2016, at a time when the number of Airbnb's listings in
in April stating that Airbnb should be considered a digital informa‑
the Lisbon metropolitan area had nearly tripled since January 2014,
tion provider rather than a traditional real estate agent. The mayors
he said to Bloomberg: «This is the first time that tourism is allowing
many people to participate in the development process of the city. Difference in differences (diff‑in‑diff) is an econometric strategy used
We shouldn’t be scared of this new dynamic, we shouldn’t be afraid of for policy evaluation. This technique mimics an experimental research
growth. On the contrary, we need to prepare the city to take in even design using observational panel data, thus mitigating the effects of
more tourists. This process creates a series of significant changes to unobserved factors and selection bias.30 Using diff‑in‑diff, we will
the city. Most of them are positive.»28 compute the effect of a treatment (in our case, the ban on short‑term
rentals) on a series of outcomes by comparing the average change over
This study analyses the ban implemented by the municipality of Lisbon time in the outcome variable for the treatment group, compared to
in November 2018 in some neighbourhoods, which provides an ideal the average change over time for the comparison group. The treatment
quasi‑experimental setup to estimate the causal impact of short‑term group comprises the neighbourhoods of Lisbon which were affected
rental regulations on the real estate market. by the short‑term rental ban imposed by the municipality in 2018.
In 2014, Portugal introduced a new law to regulate short‑term rentals, As a comparison group, we choose the neighbourhoods in which new
which consisted in an online simplified registration process to license a registries were prohibited one year later. This ensures that the compa‑
property. Landlords who want to advertise their property on websites rison and the treated neighbourhoods are sufficiently similar, allowing
such as Airbnb are required to acquire the necessary license.29 Four us to make causal inferences. In addition, we will control for 1) time
years later, the government legislated to allow municipalities to regu‑ ‑invariant factors at the local level using civil parish fixed effects, and
late the supply of short‑term rentals. As a consequence, in November 2) quarter fixed effects that capture aggregate economic shocks.
2018, the municipality of Lisbon decided to ban new short‑term rental There are at least three reasons why Lisbon (and Portugal) is an inte‑
registries in pre‑designated neighbourhoods. One year later this was resting laboratory to study the impact of short‑term rentals on the real
extended to some adjacent areas. estate market. The first is methodological: the way in which the reform
This study uses the results in Gonçalves et al. (2020) to quantify the was implemented creates an ideal comparison group of neighbou‑
impact of this ban on registries, Airbnb prices, and the real estate rhoods to be juxtaposed with those treated with the November 2018
market in the city of Lisbon, namely the number of transactions, and ban.31 The next two pertain to the context: Portugal witnessed a very
house prices. The empirical strategy we employ in our working paper rapid real estate price increase in the recent past, and also welcomed a
Gonçalves et al. (2020) takes advantage of the geographical extension touristic boom. We elaborate on these in the next section.
of the ban in 2019 to design a difference‑in‑differences empirical stra‑
tegy which allows for convincing causal estimates.
5.2. Portugal as an interesting case study 5.2.1. The evolution of short‑term holiday rental in Portugal
Decree‑law 128/2014 created a straightforward online registration The simplification of the short‑term rental registration procedure was
process for short‑term rental properties. The Portuguese government accompanied by a sharp increase in the number of tourists. Figure 55
claimed this simplified registration comprised an innovation in the shows the evolution in the number of airport arrivals. In 2019, Portugal
regulation of short‑term holiday rentals, a view confirmed by Andreu hosted more than 16 million foreign tourists, up from 9 million in
Castellano, Airbnb's public relations manager for Spain and Portugal, 2013. Lisbon airport received 29 million passengers in 2019. The World
who in an e‑mail to Bloomberg wrote: «Portugal is one of the leaders in Travel Awards elected Lisbon as the World's Leading City Break
Europe addressing the regulation of the sharing economy».32 Obtaining Destination in 2017, 2018, and 2019.34
a short‑term rental license is necessary to advertise a property on Airbnb
or similar platforms. Failure to comply is punishable by fine and online
Figure 55Airport arrivals in Portugal, Lisbon and Porto
platforms are forbidden to list unregistered properties. The licenses are
made available immediately upon request, after which owners can start 60,000,000
renting their dwellings. The competent authority — Turismo de Portugal

50,000,000
— can conduct visits to the registered properties at any moment to
Number of Airport Arrivals

verify safety regulations. The license belongs to the individual and not 40,000,000
to the house, i.e., it expires when the dwelling is sold.

30,000,000
We now describe the tax treatment of this type of income, in a

20,000,000
nutshell.33 If the annual income from short‑term rentals is above 200
thousand euros, it is treated as business income, i.e., taxes are paid on 10,000,000
revenue net of costs. Below this threshold, the earner has the option
0
of paying tax on a fixed share of 35% of the short‑term rental income. 2010 2011 2012 2013 2014 2015 2016 2017 2018
In both cases, when the owner is an individual, the income is added to
Portugal Lisbon Porto
the owner's total income and thus subject to the individual's marginal
tax rate. As an alternative, the owner may opt to declare the earnings Source: Statistics Portugal
as rental income, in which case it is subject to a flat tax of 28%. Capital

gains from the appreciation of the property during its tenure as a Thanks to the simplified procedure, the number of properties regis‑
short‑term rental are also taxed, but only when the house is sold or tered for short‑term rentals multiplied eightfold in seven years,
withdrawn from the short‑term rental market. from twelve thousand properties in 2013 to more than 94 thousand
properties in 2020, as shown in figure 56. Not surprisingly, given these Figure 57Short‑term rental registries (in Lisbon)
figures, Lisbon tops the ratio of Airbnb listings per inhabitant across 20000
European capitals.35
15000
Number of Registries
Figure 56Short‑term rental registries (in Portugal)
10000
100000
90000
5000
80000
Number of Registries 70000 0

60000 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
50000
New Cumulative
40000
30000 Source: RNAL
20000
By contrast, Porto, which is the second largest city, represents about
10000
one half of the Lisbon market, as depicted in figure 58.
0
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
New Cumulative
Source: RNAL
Lisbon is by far the most sizeable short‑term rental market in the

country; solely responsible for more than 20% of the total number of
registries, as shown in figure 57, which also displays the ban effect
with a sharp decrease in the number of new registries as of 2018.
Given the importance of the capital city in this market, its effect is
also visible on the overall country registries depicted in figure 56.
Figure 58Short‑term rental registries (in Porto) Figure 59Share of Airbnb listings by owners with multiple listings (in Lisbon)
9000 0.7
8000
0.69
7000
0.68
Number of Registries
6000
Share of Listings (%)

5000 0.67
4000 0.66
3000
0.65
2000
0.64
1000
0 0.63
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
9
01
01
01
01
01
01
01
01
01
/2
/2
/2
/2
/2
/2
/2
/2
/2
04
06
08
10
12
02
04
06
08
New Cumulative
Source: RNAL
Source: RNAL
It is interesting to notice that, according to data from the Inside 5.2.2. The boom of the real estate market
Airbnb website, most of the Airbnb listings are owned by individuals The rapid growth in short‑term rentals coincided with the rapid
or firms that advertise several properties on the platform. This is increase in real estate prices. Between 2016 and 2019, the median sale
shown in figure 59. price, per square metre, for Lisbon and Porto, increased by 68.2% and
61.9%, respectively.
The 2019 Property Index report, published by Deloitte, analyses the

evolution of the real estate market in sixteen countries: Spain, The
Netherlands, Hungary, France, Austria, Belgium, Croatia, Poland,
Czech Republic, Italy, Portugal, Denmark, Latvia, UK, Norway, and
Germany. In this analysis, Portugal stands out as one of the countries
with the fastest increase in real estate prices between 2015 and 2018.
In fact, it is amongst the four countries with a price increase of more Figure 61Growth rate of the median sale price in Lisbon — 2016 to 2018
than 30% within this period.
The real estate market in Portugal did not change symmetrically across
the territory, as shown in figure 60.
Figure 60Median real estate price per m2 in Portugal and selected cities
3500
Median Sale Price per m2 (€)

3000
2500
2000
1500 < 5%
25% to 37,5%
1000 37,5% to 50%
> 50%
500
0 Note: This replicates figure 2 from Gonçalves et al. (2020).

Source: Statistics Portugal
1
20 3
20 4
20 1
20 2
20 2
20 3
20 4
20 1
20 2
20 3
20 4
1
Q
Q
16
16
16
16
17
17
17
17
18
18
18
18
18
19
19
19
19
20
20
20
20
20
20
20
20
Figure 61 depicts the city of Lisbon split into its 24 civil parishes, or
Portugal Lisbon Porto Braga Coimbra freguesias in Portuguese. Civil parishes in Portugal have their own
directly elected governments but run a fairly small budget and have
limited competences. The city of Lisbon implemented a decentra‑
The median real estate price per square metre in the city of Lisbon lisation reform in 2014 which transferred competences to its civil
increased from 1886 to 3245 euros, i.e., more than 70% in just 4 years. parishes in the areas of waste management, security and public spaces.
However, they have no zoning powers.
Zooming into the city of Lisbon, we find very heterogeneous growth Figure 63Growth rate of median rent (per m2) for new housing rental
rates of real estate prices in different neighbourhoods, with the agreements in Lisbon, during 2019
historic downtown experiencing the strongest increases. This hetero‑
geneity is also clear if we focus our attention on rent prices per square
metre. However, as shown in figures 62 and 63, 2019 experienced a
slowdown in growth rates when compared with the previous year.
Figure 62Growth rate of the median rent (per m2) for new housing rental
agreements in Lisbon, during 2018
0 to 10%
10% to 15%
15% to 20%
Above 20%
5.2.3. The route to the 2018 reform
The rapid and uneven increase in real estate prices, alongside concerns
0 to 10%
10% to 15%
over local residents being evicted from the city centre due to housing
15% to 20% affordability, prompted a public debate about the merits of the short
Above 20%
‑term rental market and related touristic boom witnessed in Portugal
Source: Statistics Portugal during the decade of the 2010s.
Some Lisbon‑based NGOs and grassroots organizations became very

active in the debate about the excessive number of tourists and the
impact of holiday rentals on housing prices. Some NGO names became
familiar to Lisbon residents for their commitment to the cause, such as The municipality of Lisbon was quick to react, and in November imme‑
Morar em Lisboa (link) or Habita (link). In January 2017, around twenty diately suspended new registries in certain pre‑designated areas, the
organizations promoted a petition calling for public policies to curb so‑called Delimited Homogeneous Turistic Zones, with a new muni‑
the real estate price increase in the capital city, deemed excessive by cipal proposal — Proposta n.º 677/AML/2018. The suspended areas
its organizers. were those where the density of short‑term rentals, measured by
the share of short‑term rental properties in relation to total proper‑
Before the 2017 Portuguese municipal elections, held in October, ties, was considered high, more precisely, above 25%. The number of
the incumbent socialist mayor of Lisbon, Fernando Medina, declared properties in the short‑term rental market was calculated based on
his intention to intervene in the market by imposing limits on the information from RNAL, and the total number of dwellings from the
number of short‑term rental units in some neighbourhoods. Some of 2011 census.
his contenders, mostly on the left, intended to impose similar restric‑
tions.36 Interestingly, right‑wing candidates shared the concern for lack In the following year, there was a public debate about the appropria‑
of housing affordability and the belief that short‑term holiday rentals teness of extending the ban. A new municipal proposal — Proposta
were one of its main determinants. However, they favoured incentives n.º 204/CM/2019 — was put forward in April 2019 and finally
for long‑term rental as a policy solution.37 Medina was re‑elected in approved in November 2019. Figure 10 displays the timeline of
October 2017 with 42% of the votes. the reform. We discuss our datasets below. Our period of analysis
begins in 2015 and ends in the third quarter of 2019. The pre‑treat‑
In August 2018, the Parliament legislated to allow municipalities to ment period corresponds to the quarters before the legal change
regulate new property registries in the Portuguese Registry Office was discussed. The post‑treatment period is divided into the months
for Short‑Term Rentals, the Registo Nacional de Estabelecimentos de during which the reform was discussed, announced, and later imple‑
Alojamento Local, henceforth RNAL. According to Law 62/2018, muni‑ mented. In Gonçalves et al. (2020) we discuss the possible impact of
cipalities can incorporate short‑term rental supply caps into their usual each of these three periods.
zoning regulations. The law became effective in October.
Figure 64Analysis timeline in black; the extension of the prohibition one year later, in November
Treatment Pre-Treatment: 201 5 Q1 - 2019 Q3 Post-Treatment: 2017 Q3 - 2019 Q3
2019, affects the areas filled with dots; finally, we show neighbouring
civil parishes (freguesias) that we used in our analysis in grey.
Discussed 2017 Q3 - 2018 Q3
Approved 2018 Q4 The map clearly shows that all these areas belong to the historic
Implemented 2019 Q1 - 2019 Q3 centre of Lisbon.
Note: This replicates figure 3 from Gonçalves et al. (2020)

5.3. Data sources and empirical strategy
Figure 65Treated areas and control groups We use three sources of data:
• Registo Nacional de Estabelecimentos de Alojamento Local, i.e., the

short‑term rental registry which, as explained above, includes all
properties officially registered to be used as short‑term holiday
rentals.38 Each registry includes the registry date, address, number of
rooms, nationality of the owner, and whether the owner is an indivi‑
dual or a firm. We collected all registries between January 2015 and
November 2018, i.e., a total of 16972 dwellings.
• Our second data source is Inside Airbnb, a website which contains
monthly information about Airbnb listings in Lisbon, scrapped by site
curators from the Airbnb site. It also includes the number of reviews
per listing and the number of listings belonging to the same owner,
between April 2018 and September 2019. According to AirDNA,
Original areas of absolute contention (2018)
Original areas of absolute contention (2019) one of the largest databases on short‑term rental analytics, Airbnb
Neighbour (ENG UK) civil parishes
accounts for 74% of the short‑term rental activity on these type of
Note: This replicates figure 4 from Gonçalves et al. (2020). Neighbour Civil Parishes: platforms in Lisbon. We only use information on entire dwellings, i.e.,
Estrela, Campo de Ourique, Campolide, Avenidas Novas, Areeiro, Beato, Penha de França,
Arroios, São Vicente, Santa Maria Maior, Santo António, and Misericórdia
eliminating the registries on rentals of parts of the house, such as a
guest bedroom.
Figure 65 is a map of the city of Lisbon that highlights the different • The third data source contains quarterly information about the
areas: the ones that were suspended in November 2018 are depicted number of house sales and their average and quartile prices, per
square metre, between the first quarter of 2015 and the third quarter one to estimate the causal impact, as the nature of the policy change
of 2019. Lisbon is the capital of Portugal and Porto its second largest creates a suitable treatment and comparison group. This is an almost
city. They are also the only two metropolitan areas in the country. In ideal experimental setup, in which the treatment is randomized
the historic centres of these two cities, the municipality has a right and allows one to exploit actual reforms that have not been imple‑
of first refusal, according to which it can buy any dwelling for the mented, with the purpose of testing their impact explicitly built into
market transaction price. That is, once the price is agreed between the policy design. In this case, the neighbourhoods with a freeze on
the owner and the buyer, the municipality can exert the option of new short‑term rental registrations in November 2018 are the treated
buying the property for that price. This procedure ensures that the areas. For comparison, we use the neighbourhoods included in the
municipality has individual records of every transaction in the city November 2019 extension, namely Baixa, Almirante Reis, Liberdade,
centre's so‑called urban rehabilitation areas, which include the areas Graça, and Colinas de Santana. This list requires an important clarifi‑
we analyse here. We gained access to this information via a protocol cation. It turns out that the inclusion of Graça and Colinas de Santana
with Confidencial Imobiliário, an independent databank specialized was already being discussed as of April 2019, when the extension
in real estate. Confidencial Imobiliário provides information to credit of the ban started being debated. The neighbourhoods of Baixa,
firms in Portugal, and also to institutions such as Banco de Portugal, Almirante Reis and Liberdade have a lot of property allocated to the
and the European Central Bank, who use the data in their monitoring service sector and were, for that reason, not included in the initial
of the real estate market. We did not have access to the individual discussions.39 In order to avoid the contamination of our results by
records of each dwelling. For anonymity reasons, the quarterly data anticipation effects, we eliminated Graça and Colinas de Santana
has been provided at the «neighbourhood» level, per square metre from our sample in all the empirical analyses that move past 2018. For
and type of dwelling, according to the number of rooms. When the robustness, we also use the neighbouring civil parishes, adjacent to
number of transactions in a neighbourhood during a given quarter the November 2018 and November 2019 freezes, in some parts of our
is too low, price data is eliminated. Fortunately, the city areas in analysis. These areas can be seen in figure 63.
this dataset come from the two municipalities’ information system,
the same one used to design the ban of short‑term rental registries. The fact that our comparison group was also later included in the
Therefore, our real estate transaction data exactly fits the banned ban, ensures that it is very similar to the treatment group and elimi‑
and non‑banned areas in the zoning regulation we use to analyse the nates concerns about unobserved heterogeneity biasing our results
impact of these regulations. (Neumark and Simpson, 2015, page 23). Indeed, treated areas had
short‑term rental to total property ratios of 27% and 29%, while the
We now briefly explain our empirical strategy. The policy change in comparison ones had ratios of 18% and 25%.40
Lisbon has a quasi‑experimental nature. A quasi‑experiment allows
Having defined the treatment and comparison groups, we estimate Figure 66 gives a graphic example of the difference‑in‑differences
difference‑in‑differences specifications. The difference‑in‑differences methodology. It displays the number of registries in the areas banned
approach consists in comparing the outcome variables for the in 2018 and the comparison ones, which were banned one year later.
treatment and comparison group before and after the reform, i.e., The figure shows a very sizeable spike in registries in the treated
November 2018. In principle, one would expect that, absent the areas just before the ban became effective in November 2018. The
reform, the evolution of the two groups would be similar. Therefore, number of registries in treated areas is always above that of compa‑
one can attribute the differential change between the two groups rison ones; which is to be expected, since treated areas were precisely
(if any) to the policy reform. those with a higher density of short‑term rental properties. However,
the important fact for our empirical strategy is that the registries in
We estimate the equations by ordinary least squares. We include civil the two areas are parallel throughout the period of analysis, with a
parish fixed effects to account for unobserved time invariant charac‑ small exception in July 2018, which is, nonetheless, much less sizeable
teristics of civil parishes, such as the existence of amenities or the than the one registered in November. Given the common evolution
proximity to historical sites. We also use quarter dummies to control of the two variables before the ban, we may reasonably attribute the
for the state of the economy and the dynamics of the touristic market spike in October 2018 to the policy change. In other words, absent
in each quarter. As time varying controls, we used civil parishes’ poli‑ the policy change, we would expect the two variables to have a similar
tical alignment with the mayor's party (i.e., whether the civil parish evolution to the pre‑ban one, without the clear spike in the treat‑
is governed by the Socialist Party or not), and the turnout rate in the ment areas. The difference‑in‑differences estimate starts by computing
2013 and 2017 local elections. In addition, our estimations distinguish two average differences between the treated and comparison areas,
three different moments at which the policy may have an impact: the one before the implementation of the ban, and another after the
discussion, the approval and the implementation of the November implementation. It then computes the difference between these two
2019 reform, as detailed in the timeline presented in figure 8. Standard average differences.
errors are clustered at the civil parish level.
We analyse four outcome variables; look at the number of short‑term

rental registries; and also study the evolution of the real estate market
by analysing prices and quantities of dwellings traded in the market.
Lastly, we analyse short‑term rental prices. All our outcome variables
are measured in natural logarithm, allowing us to interpret the results
as percentage changes.
Figure 66Number of short‑term registries lag gave the incumbent owners ample opportunities to register new
1000 properties, which may have curbed the intended effects of the law,
900 at least in the short run.
800
700 We proceed by aggregating registries to the quarterly level. We slightly
Numer of Registries
600 change the definition of quarter, in order to account for the fact that
500 the ban was implemented in November. Therefore, the quarters in our
400 regressions are as follows: December to February, March to May, June
300 to August, September to November.
200
100 The estimates are presented in table 13. Coefficients provide the
0 difference in the change of the natural logarithm of the number of
registries, before and after August 2018. Therefore, they should be
18
18
18
18
18
18
18
18
18
18
18
Q
20
20
20
20
20
20
20
20
20
20
20
18
1/
2/
3/
4/
5/
6/
7/
8/
9/
0/
1/
20
/0
/0
/0
/0
/0
/0
/0
/0
/0
/1
/1
interpreted as the percentage change in the number of registries
01
01
01
01
01
01
01
01
01
01
01
caused by the new law.
Freeze 2019 Freeze 2018
Source: RNAL
Table 13Difference‑in‑Differences results — Ln (registries)

5.4. Results Number of
Sample Estimate Standard error observations
We now turn to the discussion of our main results. We will analyse Overall 0.309 0.04 9440
each of the outcome variables separately. Number of bedrooms: below median 0.296*** 0.06 8704
Number of bedrooms: above median 0.196** 0.03 6544
5.4.1. Registries Nationality of owner: Domestic 0.304*** 0.04 9376
Nationality of owner: Foreign 0.092 0.07 3504
As documented in figure 64, there was a big spike in the number of
registries just before the ban was implemented. It is important to Notes: For more details, please refer to tables 4 and 5 in Gonçalves et al. (2020). The period of analysis ranges
between 2015 Q1 — 2018 Q4. Asterisks depict significance levels of * p < 0.10, ** p < 0.05, *** p < 0.01.
recall the timeline of the legislative change. The Parliament voted the
law which allows municipalities to regulate short‑term rentals in late
Our results indicate the following: overall, the announcement of the
August 2018. The law became effective, i.e., municipalities were able
ban led to a spike of 30% in the number of registries in the treated
to limit the areas, as of late October 2018. Our results show that this
areas; most of the effect was driven by small dwellings, for which the Table 14Difference‑in‑Differences results — Ln (Airbnb prices)
spike is almost 30%, while bigger ones had a spike of just below 20%. Standard Number of
Sample Estimate error observations
It is also very interesting to notice that only domestic incumbent
Overall ‑0.003 0.01 39336
owners reacted to the announcement, possibly because they are more
Number of listings: single ‑0.008** 0.00 10572
aware of the local political debate and were therefore able to incorpo‑
Number of listings: multiple ‑0.001 0.01 28764
rate the anticipated announcement in their decisions.
Number of reviews: above median ‑0.000 0.01 19668
The magnitude of the effect, together with the evidence that it is Number of reviews: below median ‑0.004 0.01 19668
driven by domestic incumbent owners, suggests that the anticipation Notes: For more details, please refer to tables 7 and 8 in Gonçalves et al. (2020). The period of analysis ranges
period suppressed the impact of the ban in the short run.
5.4.2. Airbnb Prices 5.4.3. The real estate market: quantity and price
We will now present the results on rental prices of Airbnb listings. We now rely on the data of the SIR.RU system to analyse the impact
Our analysis is restricted to the prices of entire homes. All prices are of the short‑term rental ban on the real estate market. More preci‑
per night and dwelling, and include dwelling fixed effects. Since fixed sely, we focus on the quantity and price of dwellings traded in the
effects also control for the surface of the property, we run the regres‑ market. For the sake of anonymity, the data is aggregated to the
sion on total rental price. neighbourhood level and type of dwelling, i.e., according to the
number of bedrooms. This explains the smaller number of observa‑
Our results, presented in table 14, indicate that, at least in the short tions, a limitation we tackle by augmenting the comparison group
‑term, the ban had no impact on Airbnb prices. We also provide separate with neighbouring civil parishes, as displayed in figure 63. This leaves
results for the subsample of dwellings owned by individuals or firms us with 9 treated neighbourhoods, which we compare with 54 compa‑
that only have one listing on the platform and those that have several, rison groups.41
and also split the dwellings according to the number of reviews. The
latter is a proxy of quality, as more reviews signal a property that is
rented more often. We do not find significant price impacts in any of
the subsamples. If anything, we identify a minor decrease of 0.08% in
the price of listings owned by single owners, a result that may indicate
temporary excess supply due to the increase in registries.
To have a better idea regarding the dwellings of the geographic areas Figure 68Number of house transactions, by energetic efficiency
considered in our study, we illustrate the evolution of house transac‑ 450
tions across two dimensions: 1) number of bedrooms; and 2) energetic 400
efficiency (where A+ are the most efficient and G are the least effi‑ 350
cient). These are portrayed in figures 67 and 68, respectively. Both 300
Number of Transactions
graphs underline important characteristics of our sample: houses in 250
these areas are typically small (with one or two bedrooms) and display 200
low levels of energetic efficiency.
150
100
50
Figure 67Number of house transactions, by number of bedrooms
0
500
20 1
20 2
20 3
20 4
20 1
20 2
20 3
20 4
20 1
20 2
20 3
20 4
20 1
20 2
20 3
20 4
20 1
20 2
3
Q
Q
450
15
15
15
15
16
16
16
16
17
17
17
17
18
18
18
18
19
19
19
20
400
350 A+ to B- C D E to G
Number of Transactions
300
Source: RNAL
250
200 The possibility to register a dwelling as a short‑term rental creates

150 option value, which is eliminated with the ban. More importantly,
100 we do not expect the surge in the number of registries to eliminate
50 this option value, as the short‑term rental license belongs to the owner
0
and is lost with the property transaction. Therefore, anyone buying a
1
20 2
20 3
20 4
20 1
20 2
20 3
20 4
20 1
20 2
20 3
20 4
20 1
20 2
20 3
20 4
20 1
20 2
3
house does not benefit from the spike in registries depicted in figure
Q
Q
15
15
15
15
16
16
16
16
17
17
17
17
18
18
18
18
19
19
19
20
20
64 and the ban effectively eliminates the option value.

T1 T2 T3
We first analyse the effects on property transactions. The results for
Source: RNAL
the overall sample and for the subsamples of houses with different
numbers of bedrooms are presented in table 15. The ban decreased the
number of houses traded in the market by almost 20%. The evidence
supports the argument that the option to participate in the short‑term
rental market is an important determinant of the housing market Table 16Difference‑in‑Differences results — Ln (prices)
demand in these areas. Moreover, the effect is driven by housing units
Number of
of smaller dimensions, which confirms the results in table 13 that Sample Estimate Standard error observations
the short‑term rental market is segmented according to the size of Overall ‑0.086*** 0.03 462
dwellings. 1 bedroom ‑0.035 0.05 304
2 bedrooms ‑0.197** 0.08 318
3 bedrooms ‑0.019 0.06 232
Table 15Difference‑in‑Differences results — Ln (number of transactions)
Notes: For more details, please refer to tables 11 and 12 in Gonçalves et al. (2020). The period of analysis ranges
Number of between 2017 Q1 — 2019 Q3. Asterisks depict significance levels of * p < 0.10, ** p < 0.05, *** p < 0.01.
Sample Estimate Standard error observations
Overall ‑0.196** 0.08 660 The results are presented in table 16. They show that the suspension
1 bedroom ‑0.219 0.12 584 of short‑term rental registries induced an almost 9% decrease in prices.
2 bedrooms ‑0.259*** 0.06 584 This further confirms the option value regarding the possibility to rent
3 bedrooms ‑0.142* 0.07 557 houses in the short‑term rental market.
Notes: For more details, please refer to tables 9 and 10 in Gonçalves et al. (2020). The period of analysis ranges The results for the subsamples of houses with different number of
rooms confirm that the more «Airbnb‑marketable» properties are
We now turn to real estate prices. The price information is omitted the relatively smaller ones. In this case, dwellings with 2 rooms expe‑
in our data for neighbourhoods with less than 4 transactions in a rienced a price decrease of 20%.
given quarter. The neighbourhoods that are dropped from the sample
due to this restriction are, in principle, different from the remaining 5.5. Conclusion
ones, as their real estate market is probably less dynamic. Therefore,
In this chapter, we discuss the tourism boom in Portugal, with a special
to make our observations as comparable as possible, we focus on
focus on Lisbon and the related spread of short‑term rental proper‑
neighbourhoods with data in all four quarters, i.e., those with a
ties in the city. We analyse the impact of short‑term rentals on the
sufficient number of transactions in a steady flow throughout the
real estate market, taking advantage of a partial ban introduced by
period we analyse. This leaves us with 5 treated and 37 comparison
the municipality of Lisbon in November 2018, which prohibited new
neighbourhoods.42
short‑term rental licenses with immediate effect in some pre‑selected
areas of the capital.
We use a difference‑in‑differences strategy that compares the areas Figure 69The demand driven impact in the real estate market
prohibited in 2018 with the others prohibited one year later, before price
and after the introduction of the ban. This allows us to obtain causal
evidence on the regulation of the short‑term rental market. supply
Our main findings are as follows. We show that the incumbent owners price, demand
before before ban
(mostly, Portuguese ones) rushed to register properties just before the
price, demand
ban was effective. This shows that the anticipation of the policy due after after ban
to its wide public discussion, as we mentioned in Section 2, allowed

incumbent owners to enter the market in the very last weeks during quantity, after quantity, before quantity
which it was allowed. We then analyse the evolution of Airbnb rental

prices, finding no effects, at least in the short‑run.
Our work provides convincing evidence that the surge of short‑term
Regarding the real estate market, we document a sizeable decrease in rental markets does create an upward pressure on real estate prices.
the number and price of traded dwellings. More specifically, we find However, our estimates suggest that, despite their salience, sharing
that the ban led to a 20% decrease in the quantity traded in the market economy platforms are not the full story behind the spectacular
and a 9% decrease in the price. The market for 2‑bedroom flats drives housing market boom in the city of Lisbon.
these effects, showing a segmentation of short‑term rentals in small to
A few final notes are in order. We started this study before the
medium sized apartments. The price decrease for two‑bedroom houses
COVID‑19 pandemic led tourism and the short‑term rental market
is almost 20%. This is consistent with a sharp reduction in the demand
to an almost complete halt. According to Confidencial Imobiliário, the
for this type of flats, confirming that the option to register a property
occupancy rate of short‑term rental properties in Lisbon was down
as short‑term rental is an important driver of the search for new
to 5% in May 2020, a sharp decline compared with the 73% occu‑
properties. We illustrate the demand mechanism in figure 69.
pancy rate registered in May 2019.43 This is a tremendous shock to the
real estate market and the effects will depend on the duration of the
pandemic and the evolution of international travel restrictions. Even if
the impact of this shock is outside the scope of this study, our results
strongly suggest that it may lead to a decrease in real estate prices in
the near future.
References
Angrist J. D. and J.‑S. Pischke, (2009), Mostly harmless econometrics: an empi‑
ricist's companion, Princeton, NJ: Princeton University Press.
Franco, S. F. and J.L. Macdonald, (2018a), «Measurement and valuation

of urban greenness: remote sensing and hedonic applications to Lisbon,
Portugal», Regional Science and Urban Economics 72, 156–180.
Franco, S. F. and J.L. Macdonald, (2018b), «The effects of cultural heritage

on residential property values: evidence from Lisbon, Portugal», Regional
Science and Urban Economics 70, 35–56.
Franco, S. F., C. Santos and R. Longo, (2019), «The Impact of airbnb on

residential property values and rents: evidence from Portugal», Working
paper.
Gonçalves, D., S. Peralta and J.P. Santos, (2020), «Do short‑term rentals

increase housing prices? Quasi‑experimental evidence from Lisbon», Working
paper.
Koster, H., J. van Ommeren and N. Volkhausen, (2018), «Short‑rerm ren‑

tals and the housing market: quasi‑experimental evidence from airbnb in Los
Angeles», Working paper.
Neumark, D. and H. Simpson, (2015), «Place‑based policies», in Handbook of

Regional and Urban Economics, Vol. 5, Elsevier, pp. 1197–1287.
Chapter 6
Urban rents in Portugal: a decline
with no return?
Victor Reis
6.1. Introduction
This chapter analyses the evolution of urban rents in Portugal based on The analysis of the recent evolution of Lisbon and Porto's historic
existing traditional dwellings, as well as their forms and regimes since centres, the most relevant focus of the real estate market, is only
the beginning of the 20th century. It addresses the development of possible after the 2001 census, when lease information disaggregated
traditional housing stock on a national level, and in the cities of Lisbon by civil parish becomes available. The information after the 2011
and Porto, together with their respective historic centres, since the census is scattered and scarce, and it is only in 2017 that data on new
1960 census. We present the main legislative measures taken over the lease agreements are disclosed.
last century in terms of the liberalised rental markets, as well as the
various social housing rental systems related to public housing policies. In this paper, we consider the following assumptions:
From a political point of view, we explore the effects of these • We conduct the analysis only on the so‑called classical family
measures on the rental housing market, especially in terms of its dwellings;
decline and segmentation. However, in order to portray the evolu‑ • No information is presented prior to the 1960 census;
tion of urban rents, we needed to address several obstacles, such as • In censuses from 1960 to 1991, the number of rented dwellings,
the lack of information collected over the decades, especially prior was extrapolated from the proportion of dwellings registering over
to 2001, and the need to perform extrapolations, since until this time occupancy;
data on rented dwellings were presented by households. This entailed • Data on rented dwellings include sub‑leases and exclude so‑called
assessing over‑occupancy and considering changes to concepts which «other situations» and «space lending»;
occurred between censuses. • The data presented by civil parish is aggregated based on changes
occurred after 2013. In this way, we can compare these data with the
most recent statistics for the historic centres of Lisbon and Porto.
6.2. Census information As is well known, the growth rates observed for each decade have
been greatly influenced, both by migration flows and demographic
6.2.1. Portugal: the evolution of conventional dwellings over
changes. In the 1960s, emigration was responsible for the low growth
the last 50 years
of the number of dwellings at 5.5%, while in the 1970s, the decolo‑
As mentioned earlier, the available data presents deficiencies in terms of nization process generated the highest growth ever in housing stock
the evolution of urban rents in Portugal. However, the data that consti‑ (25.2%). Since 2002, with the end of subsidized credit, and after 2007
tutes the statistical base of analysis allows us to visualise a path which with the subprime crisis, there was a slowdown in dwelling construc‑
simultaneously shows a surprising growth in national housing stock and tion, and in the 2011 census growth had declined to 16.7%. The latest
a sharp decline in the number of properties available for rental purposes. data provided by Statistics Portugal on housing stock for 2018, esti‑
mates a growth of only 1.6%, since 2011.
In the five‑decade period between the 1960 and 2011 censuses, the
housing stock of traditional family dwellings more than doubled, rising Note that after the 1991 census, several experts pointed to the exis‑
well above the number of households, and the ratio of dwellings per tence of a housing deficit in Portugal of around half a million homes,
household increased from 1.08 to 1.45 (see next graph). leading to the proposal of «the 500 thousand housing plan»44. At the
time, the total number of dwellings of usual residence was 3.05 million
and there were 3.14 million households, which suggests the need for
Figure 70Evolution of conventional dwellings, households and residents one hundred thousand additional homes. Over the next two decades,
in Portugal (Unit: millions) 1.7 million conventional dwellings were built in Portugal and despite
this 41% increment and only 28.5% increase of households for the
same period, the difference between conventional dwellings and
10,44
10,26
9,79
9,81
households in 2011 was of approximately 52,000 dwelling units.
8,78
8,61
The demographic evolution and the changes in the structure and size
5,86
5,02
of households registered in Portugal by 2011, show a country in which
4,15
4,04
3,65
3,38
3,15
2,92
2,70
2,36
2,56
2,35
the number of families had grown 71.6% since 1960, but the number of
1960 1970 1981 1991 2001 2011 residents had increased only 18.9%. However, the difference between
Census year population growth and housing stock over this period shows divergent
Total dwellings Total households Total resident population
behaviour when analysed from a standpoint of forms and regimes of
dwelling occupancy, as shown in the graph below.
Figure 71Evolution of forms and regimes of dwelling occupancy The weight of vacant dwellings also grew, although less significantly,
92,6% reaching 12.5% in 2011. Despite the rise, it is lower than the 13.8%
83,4% 81,9%
quantified by the 1970 census. However, when comparing the 2011
73,5%
75,5%
73,2% census results with those of the previous decade, an increase of 35.2%
70,7%
can be observed, representing an increase from 543 thousand to
68,1%
56,6 64,7% 735 thousand dwellings in absolute terms;
50,2%
48,0%
45,1% 44,3%
However, it is amongst rented and vacant dwellings that the grea‑
38,8%
28,1% test decline can be observed. In the 1960 census, the weight of these
20,8% 19,9% dwellings was 48%, but by 2011 it had reduced to 19.9%. In absolute
13,8%
7,4% terms, the decline was of around 30%, decreasing from 1.14 million
5,6% 12,5%
10,6% 10,8%
dwellings in 1960, to just over 794 thousand in 2011, for these types of
1960 1970 1981 1991 2001 2011 regimes. It should be noted, however, that in the decades starting from
Census year
1970 and 2001, there was an increase of approximately 7% in rented
Owner-occupied dwellings Rented dwellings
dwellings, for each period.
Dwellings of usual residence Vacant dwellings
Source: Statistics Portugal 6.2.1.1. The case of Lisbon and its historic centre
On the one hand we can observe a continuous and significant drop in When specifically analysing the data for the cities of Lisbon and
weight regarding dwellings of usual residence, which decreased from Porto, which have been subject to a long period (38 years) of legal rent
92.6% in 1960 to 68.1% in 2011. In absolute values, these dwellings freeze, it appears that these patterns present greater deviations than
grew 68.4%, from 2.36 to 3.99 million, while the total number of those registered at a national level.
conventional dwellings more than doubled within the same period.
As can be seen in the following graph, until 1981 Lisbon has more
At the same time, there were also three significant changes in the
households than conventional dwellings, and only in the following
composition of the housing stock:
censuses does the data show a reversal of this reality. The ratio of
Owner‑occupied conventional dwellings grew to 73.2% of the total dwellings per household increased from 0.81 in 1960 to 1.32 in 2011.
number of dwellings of usual residence, when 50 years earlier that
weight was of 45.1%. This corresponds to a nearly tripled increase,
from 1.07 to 2.92 million dwellings;
Figure 72Evolution of conventional dwellings, households and resident Following national trends, the numbers of usual and rented dwellings
population in Lisbon decreased and owner‑occupied and vacant dwellings increased. These
patterns are even more pronounced in the historic centre.
807 167
771 788
705 295
629 830
Figure 73Evolution of forms and regimes of dwelling occupancy in Lisbon
547 733
542 964
218 990
214 670
200 449
322 865
288 481
285 960
270 835
259 191
245 537
243 892
98.2%
236 611
230 060
92.8% 92.5%
84.0%
91.0% 88.3%
1960 1970 1981 1991 2001 2011 76.9%
73.5%
Census year 77.7% 69.4%
61.4% 67.7% 64.1%
Total dwellings Total households Total resident population 61.2%
48.6%
51.8%
Source: Statistics Portugal 47.9%
33.8% 42.3%
31.3%
28.4%
However, while at a national level there was an increase in households 18.3% 23.1%
26.5%
and resident population, in Lisbon there was a sharp decline in these 4.5% 8.0%
15.6%
2.4% 14.0%
values from 1981 to 2001. The 2011 census shows a slight inflection of 1.8% 5.7% 9.2%
1960 1970 1981 1991 2001 2011
this trend. In the 30‑year interval between 1981 and 2011, we observe
Census year
a loss of nearly one‑third of the resident population and 15% of fami‑
Dwellings of usual residence Dwellings of usual residence in historic centre
lies. According to Statistics Portugal's estimates for 2018, the resident
Owner-occupied dwellings Owner-occupied dwellings in historic centre
population decreased 7.4% in comparison to 2011. Rented dwellings Rented dwellings in historic centre
Vacant dwellings Vacant dwellings in historic centre
The next graph shows the evolution of the housing stock in Lisbon
between 1960 and 2011, and the changes that occurred in the forms
and regimes of dwelling occupancy. The data available for 2001 and
The city as a whole witnessed a fall in dwellings of usual residence
2011 also provide information for the three civil parishes within the
to 73.5% of the total number of dwellings. During these 50 years,
city's historic centre (Santa Maria Maior, Misericórdia and S. Vicente).
the total number of dwellings increased by approximately 122,000,
As previously indicated, there is no data available at the municipal
reaching a total of 322,000, and those dwellings of usual residence
level in censuses prior to 2001.
only grew 40,000 in number, totalling 237,000. In 2011, the weight of
dwellings of usual residence in the historic centre was lower (61.2%) 6.2.1.2. The case of Porto and its historic centre
and showed a sharp decrease when compared to 2001.
With regard to the city of Porto, although there are many similarities
In 2011, Lisbon had 15.6% of vacant dwellings, a number higher than with the patterns observed for Lisbon, there are some specificities to
the 12.5% registered at a national level, and 26.5% in its historic centre, be highlighted. The ratio of family dwellings increased from 0.99 in
a percentage higher than that of the city as a whole. In absolute terms, 1960 to 1.36 in 2011.
the number of vacant dwellings in the historic centre amounted to
8,365. If we extend this analysis to the limits of the first ring road
Figure 74Evolution of conventional dwellings, households and resident
which includes the civil parishes of Estrela, Campo de Ourique, Santo
population in Porto
António, Arroios and Penha de França, the number increases to 22,472.
Considering the rest of the civil parishes altogether, a further increase of
27,737 is observed. Thus, Lisbon totalled 50,209 vacant dwellings in 2011.
327 368
302 472
300 095
294 630
263 131
The variation in owner‑occupied dwellings as well as rented dwellings
237 591
presents an almost symmetrical trend, which intersects in the 2001
census, when owner‑occupied dwellings begin to have more signifi‑
137 236
124 494
112 405
100 696
98 416
100 826
99 325
97 582
cance than rented dwellings. Once again, the figures for the historic
81 298
81 910
81 102
80 555
centre deviate from those of the city. In 2011, owner‑occupied
1960 1970 1981 1991 2001 2011
dwellings already have a weight of 51.8% in the city as a whole, while
Census year
in the historic centre the percentage is only 31.3%. The phenomenon
Total dwellings Total households Total resident population
repeats itself in the same year for rented dwellings, which in the city
represent 42.3% and in the historic centre represent 64.1%. Source: Statistics Portugal
In summary, dwellings of usual residence and rented and vacant From the 1981 census onwards, and to a lesser extent in Lisbon, there
dwellings in Lisbon have a greater weight than at country‑level, and is a drop of approximately 30% in resident population, although the
only owner‑occupied dwellings show an opposite behaviour. The number of families does not vary widely. However, in Porto's case, the
historic centre shows a greater weight of rented and vacant dwellings 2011 census does not show stability in the number of resident popu‑
than the city as a whole, while dwellings of usual residence and owner lation, as observed in Lisbon. In contrast, Porto shows a continuous
‑occupied dwellings show a lower rate. decline and according to Statistics Portugal's estimates for 2018, the
resident population suffers a decrease of 9.4% in comparison to 2011.
The following graph shows the evolution of housing stock in Porto, percentage of owner‑occupied dwellings (40.0% to 31.1%) as well as
which is much alike Lisbon's. Similarly to the trend observed, both at vacant dwellings (30.5% to 26.5%) is higher than in Lisbon.
a national level and in Lisbon, Porto displays two trajectories of sharp
decline for dwellings of usual residence and rented dwellings, as well Over the past five decades, dwellings of usual residence in Porto
as increasing trajectories for owner‑occupied dwellings and vacant have fallen to 71.9% of the total number of dwellings. Between the
dwellings. 1960 and the 2011 censuses, this type of dwelling increased by about
56,000, reaching 137,000 units, while dwellings of usual residence grew
22,000 to a total of 98,000 units. Furthermore, in 2011, in the historic
Figure 75Evolution of forms and regimes of dwelling occupancy in Porto centre of Porto, the weight of dwellings of usual residence is much
95.2% lower, only corresponding to 59.7%, and showing a decrease of almost
92.1% 91.7%
84.2%
10% when compared to 2001.
92.0%
84.0% 77.0%
71.9% Porto has a higher percentage of vacant dwellings in 2011 than Lisbon
74.6% 67.4%
59.6% and the country as a whole (18.8% to 15.6% and 12.5%, respectively),
58.8% 59.7%
54.1% while in the historic centre the percentage is even greater than that
50.7%
36.5%
47.5% 43.9% of the city as a whole (30.5%). From the analysis of vacant dwellings
38.7% 40.0% in absolute terms, Porto had a total of 25,833 units according to the
23.2%
23.4%
30.5% last census, of which 9,327 were located in the historic centre, and
14.1%
11.1% 18.8% 16,506 throughout the city. It should be noted that in 2011 the historic
15.1%
3.6% 10.9%
4.8% 7.3% centre of Porto had 30.5% of vacant dwellings, which when compared
1960 1970 1981 1991 2001 2011 to that of the other Portuguese municipalities, is only surpassed by
Census year
Penela (31.8%), and followed by the historic centre of Lisbon (26.5%),
Dwellings of usual residence Dwellings of usual residence in historic centre
Mora (23.9%), Coruche (23.3%) and Tarouca (22.8%). It is important
Owner-occupied dwellings Owner-occupied dwellings in historic centre
to mention that these municipalities represent the country's interior
Rented dwellings Rented dwellings in historic centre
Vacant dwellings Vacant dwellings in historic centre regions, considered to be low density areas and strongly affected by
the depopulation process.
As in Lisbon, the variation in owner‑occupied dwellings and rented

However, in the historic centre of Porto these trajectories show a few
dwellings in Porto has symmetrical trends that intersect in 2001, a time
particularities that must be highlighted: the weight of rented dwellings
when owning a house acquires more weight than renting one. Once
is much lower than in Lisbon (54.1% to 64.1%), while in Porto the
again, the numbers in the historic centre deviate from this pattern. In has been in contraction. In several regions of the country, the decline
2011, owner‑occupied dwellings already weigh 50.7% in the city and in supply has been significant and the price increase very sharp.
40.0% in the historic centre. The relative weight of rented dwellings is
43.9% and 54.1%, respectively. Statistics by Confidencial Imobiliário beginning in 2010 illustrate the
evolution of supply and the contraction of new rental housing agree‑
In summary, as in Lisbon, dwellings of usual residence, as well as rented ments in Portugal, as can be observed in the following graphs.
and vacant dwellings in Porto, have greater weight than at national
level, and only owner‑occupied dwellings show the opposite behaviour.
Furthermore, when focusing on Porto's historic centre, both rented and Figure 76Number of rental housing units supplied per year
vacant dwellings have greater weight than in the city as a whole, while 30 000
27288
25044
dwellings of usual residence and owner‑occupied dwellings have lower 25 000 23686
22143
weights (a behaviour similar to that observed in Lisbon). 19826
Number of dwellings
20 000
16755 17353
15544 16078
15 000 15373 13326 13539
6.2.2. Information available after the 2011 census 10447 10826
8934 9558
10 000 8453 7864
6405 6240
This study was carried out approximately a year before the 2021 4326 4803
3755 4309
5 000 3244 3034 3341
5258 4738
census and, in the nine years since the last census, the information 3130 3386 3128 2842 2400 1679 1961
0 1541 2368 1899 1229 1045 816 694 587 671
available on the evolution of housing stock does not allow for an 2011 2012 2013 2014 2015 2016 2017 2018 2019
in‑depth analysis, as does the information contained in the censuses.
Statistics Portugal produces annual housing stock estimates in terms Portugal Lisbon Metropolitan Area Porto Metropolitan Area
Lisbon Porto
of dwellings, but these do not allow for a disaggregated view of their
forms and occupancy regimes. It also provides housing rent statistics Source: Confidencial Imobiliário
at the local level, with data series starting in 2017 which, being newly
implemented do not allow for comparison with the 2011 census. With the exception of the inflection observed in 2019, the supply of
rental housing has been in continuous decline since 2013 in various
Also available are the statistics produced by Confidencial Imobiliário, areas of the country.
a repository with the most extensive series on the rental market over
the last decade, as well as information from the Tax and Revenues
Authority (AT) which is sporadically disseminated in the media. As we
will see, these statistics show that in recent years, the rental market
Figure 77Number of new housing rental contracts per year Figure 78Number of new rental contracts per year
10 000 90 000
8969
9 000 80 000
84383
8264
Number of new contracts

7965
77723
7825 7774 7709
8 000 70 000
72788
7146
Number of new contracts

7 000 7452 60 000
6201 5973
6 000 6687 50 000
6152 6138 5901 6125
5 000 5821 40 000
5536
4776 30 000
14453
13532
4 000
12453
28305
25916
2838 20 000
6980
24129
6643
6721
3 000 2602 2423
2308
3177
2290
3110
3043
2097 2095 2042 1828 10 000
743
659
653
801
750
702
2 000 1517 1577 1584
994 1827 2004 1873 1197
665
0
1 000 456 488 527 508 478 348 453 Portugal Lisbon Porto Lisbon Porto Lisbon Porto
225 329
0 Metropolitan Metropolitan historic historic
2011 2012 2013 2014 2015 2016 2017 2018 2019 Area Area centre centre
2017 2018 2019

Portugal Lisbon Metropolitan Area Porto Metropolitan Area
Lisbon Porto
Source: Confidencial Imobiliário

These data only refer to new rental agreements, and contain no indi‑
In terms of new rental housing agreements, the number generally grew cation of how many have ceased. Therefore, they do not enable the
until 2013, with the exception of Porto and the metropolitan area of assessment of changes occurred in the rental market as a whole, nor
Porto, where it continued to increase until 2015 and 2016, respectively. do they allow us to understand whether the global trend decline regis‑
Subsequently, there was a drop in these indicators which persisted tered in the censuses continues to be true. In this regard, it is only
until 2018, followed by an inflection and increase in 2019. possible to obtain some indications through the information released
by some media outlets, based on data from the AT. In February 2018,
However, this latest increase is contradicted by housing rental statis‑ Dinheiro Vivo reported that 24,477 fewer annual income returns had
tics at the local level from 2017 to 2019, produced by Statistics been filed, corresponding to 20,474 less landlords45, in conformity with
Portugal and based on data from agreements registered at the AT. data provided by the AT. According to this publication, these figures
With the exception of the city of Lisbon, which in 2019 presented a resulted from the number of filings of Form 44 (Modelo 44), which in
growth of 1.2%, the variations are negative for all other indicators, 2018 had been 130,321 and in the previous year 154,798. A few months
with the two metropolitan areas displaying the greatest reduction: later, on the 24th of October, quoting Correio da Manhã, Dinheiro Vivo
14.8% in Lisbon and 13.8 % in Porto. The national decrease from reported that in the first 10 months of 2018 there was a 3.6% drop
2017 to 2019 was 13.7%, and 3.7% and 4.2% in Lisbon and Porto, in the registration of new housing rental agreements46. The news
respectively.
article added that the termination of agreements had increased from its national weight is 12.5%, the historic centres of Porto and Lisbon
124,000 to 134,000. More detailed information on the recent situa‑ represented 30.5% and 26.5%, respectively.
tion of urban housing rentals was released by Negócios in January
2020, reporting that 754,278 housing rentals were registered through With regard to the indicators of owner‑occupied dwellings and
Portal das Finanças (Portugal's online fiscal platform), based on data rented dwellings, the pattern reverses: in the first case, the weight at a
from the AT47. The comparison of this number with that of the 2011 national level is 73.2%, while in Lisbon and Porto it is 51.8% and 50.7%,
census indicates a reduction of about 5%. The news article refers to respectively; and in their historic centres 31.3% and 40%, respectively.
a relevant aspect of rental housing, especially those with agreements
signed before November 1990, and commonly known as «old agree‑
Figure 79Weight of forms and occupancy regimes in 2011
ments». The 2001 census quantified 440,000 agreements, a number
Portugal
that decreased to 225,000 in the 2011 census. Negócios also added that 73,2%
in 2018 this number dropped to 78,383, and in 2019 to 50,462, corres‑
68,1%
ponding to 6.7% of the total registered agreements.
6.2.3. A profound change in the housing stock Lisbon

Porto historic centre 19,9%
59,7% 54,1% 51,8% 73,5%
42,3%
In the 50‑year interval between censuses, a period during which the 40,0% 12,5%
country more than doubled its housing stock, dwellings of usual resi‑ 30,5% 15,6%
dence lost significant weight, dropping from 92.6% to 68.1% of the

18,8%
total housing stock; the remaining 31.9% are divided between secon‑ 26,5%
31,3%
dary, seasonal or vacant dwellings. In the universe of permanent 43,9%
dwellings, owner occupied dwellings increased from 45.1% to 73.2%, 61,2% 50,7%
64,1%
and rented dwellings fell sharply from 48% to 19.9%. 71,9%
Lisbon historic centre Porto

In Lisbon and Porto, the weight of dwellings of usual residence
presents similar patterns, but when the analysis is restricted to their
Dwellings of usual residence Rented dwellings
respective historic centres the weight is much lower in the 2011
Owner-occupied dwellings Vacant dwellings
census: 61.2% in Lisbon and 59.7% in Porto. Meanwhile, the percen‑
tage of vacant dwellings in both historic centres increased. While Source: Statistics Portugal
As for rental housing, the national weight drops to 19.9% in 2011, but iv) 1948 to 1976
continues to have some significance in Lisbon and Porto, with 42.3%
and 43.9%, respectively; these weights are higher in their respective In 194850 changes to rent prices in Lisbon and Porto were suspended,
historic centres, 64.1% and 54.1%, respectively. a situation which lasted until 1986, and tenancy agreement transmis‑
sions were extended to both antecedents and descendants. In 1966,
These patterns are not repeated in other Portuguese municipalities the Civil Code51 grants unlimited duration to rental agreements. With
and those that come closest to these percentages of housing rentals the April 1974 revolution, rent freeze became widespread throughout
are Guimarães and Amadora (both with 30.5%), Espinho (with 29.9%) the country, evictions were suspended and the illegal occupancy of
and Loures (with 29.5%), which presented higher percentages of vacant dwellings was legitimized.
dwellings of usual residence (approximately 80%) and much lower
vacant dwellings (between 9% and 11%). v) 1976 to 2016
With the first constitutional government being established in 197652

6.3. Rent control in housing markets the reversal of measures from the previous cycle begins. In 1981,
6.3.1. A century of legislative instability discretionary rent agreements are allowed, conditional rent is created,
the annual rent adjustment mechanism is defined and the Rent Law
Urban rental housing in Portugal, in the so‑called free market since is published53 — allowing for exceptional rent updates on agree‑
1910, can be divided into four major cycles, which occurred throu‑ ments previous to 1980. The publication of the Urban Lease Regime
ghout different political regimes: (RAU)54 opened a path of progressive and greater liberalization for
urban rental housing, which was in place until 2016. Limited‑term
iii) 1910 to 1948
tenancy agreements are initiated and changes are made to the rights
Following the Implantation of the Portuguese Republic in 1910, the of tenants regarding the transmission of agreements to antecedents
law «regulating tenancy»48 is published, decreeing the first rent freeze, and descendants. In 199555, this reform was extended to the non‑hou‑
which was supposed to last for one year but lasted for several more. sing rental market. In 2006, with the publication of the New Urban
This legislation would remain in force until the publication of another Lease Regime (NRAU)56, a new process for updating rents is initiated,
decree in 191949, which attempts to provide greater consistency to depending on the building's state of conservation, automatic transfers
rent policy and maintained the transitional and emergency measures of non‑housing leases are over, the form of housing lease transmis‑
enacted during the First World War crisis. During this period, throu‑ sion is restricted and a rent subsidy is created. With Law no. 31/2012
ghout the First Republic and after the establishment of the Estado of 14 August, a transition period is established for the update of
Novo, several attempts were made to update rent prices. rents prior to 1990, according to tenants’ income and age. A special
eviction procedure is created, as well as the National Leasing Office transfer, eviction procedures, security deposits and penalties for
(Balcão Nacional do Arrendamento), and also provisions for the vacancy late rent payments, state social support instruments and the issue of
of property in order to carry out extensive construction work when vacant dwellings.
needed. This law was changed in 201457, with several amendments,
namely relating to the protection of tenants, notification periods and 6.3.2. Indefinetly extended transitional regimes
compensation for improvements made. The rent subsidy regime of
One of the common factors of the various legislative processes within
2012 is regulated through Decree‑law no. 156/2015 of 10 August.
the last 110 years, is the transitional nature of their measures, always
vi) 2016 to the present justified by situations of crisis or emergency. After the first transitional
period, the same measures were generally renewed several times. The
In 2016, the reversal of several approved measures, especially those first rent freeze, enacted on 12 November, 1910, was intended for
from the previous legislature, begins. The notion of «shops with a one year period but remained in effect until 1914, and the mecha‑
history»58 is created, the regime for carrying out maintenance and nism for fixing rents enacted in 1914 was successively changed in
repair work in rented buildings is changed, the right of preference following years and until 1924, maintaining its initial transitional
is modified59 (declared unconstitutional in June 2020), harassment nature. Law No. 1662, published in 1924, set restrictions for rent
and rent penalties are legislatively prohibited60, the transition period updates which would remain in effect until the end of the following
of tenancy agreements to new lease regimes is prolonged (with a year. However, these restrictions were extended until the end of 1927
minimum duration limit of at least three years now imposed on agree‑ and, on the 28th of November, 1927, Decree no. 14630 prolongs the
ment renewals), arrears interest is reduced, and certain types of lease situation once again, «until publication of a new decree». In short,
become lifelong, depending on the specific tenancy situation61 and annual transition becomes definitive. The rent freeze imposed by
including fixed term agreements. Affordable rental programmes for Law No. 2030 in 1948 within Lisbon and Porto makes rent updates
the middle class are created. dependent on the execution of fiscal assessments. In 1966, when the
Civil Code was approved, the need for this update was acknowle‑
Over the 110 years covered by these four cycles there is a dense and
dged, although it still remains to be carried forward to this day. The
vast legislative framework which despite changes in political regimes
correction of rents is only initiated 38 years later with the Rent Law.
has generated a climate of increasing instability and insecurity in the
More recently, the NRAU transitional regime for pre‑1990 agreements
rental market.
set out by Law No. 31/2012, initially for five years, was extended
In the next section, we will examine important legislative interven‑ for another three years in 2017 and for two more years in 2020, thus
tions, such as rent freeze, the duration of rent agreements and their totalling 10 years, twice the term initially established. It would be of
no surprise if it were to be extended until 2022. These are just some
examples portraying the legislative practices that have been in place In short, although there is a rent freeze for existing agreements and,
over decades, common to the First Republic, the Estado Novo and the between 1910 and 1914 there are ordinances which determine that
times of the PREC, and which still exist today. new agreements cannot establish rents higher than the previous
ones, there is still a high increase in the value of rents. Compare,
6.3.3. Rent freeze and price control for example, the limits in Lisbon, which went from 18 escudos to
50 escudos between 1914 and 1919. It should be noted that from
Throughout the aforementioned cycles, the situations in which there
1917 onwards, agreements between landlords and tenants regarding
was an effective rent freeze must be distinguished from those in which
rent increases above the legally established limits were prohibited69.
it was legislated to allow updates, although these did not have any real
It is not clear how long this rule was in force, but it seems that it
practical effects due to procedural factors. In fact, there were only three
ceased to have legal value in 193370.
rent moratoriums: 1) an intended one‑year period in 1910 which lasted
until 1914 and in some situations until 1919; 2) a 38‑year period in the The publication of Law No. 1368 of 21 September, 1922, which remo‑
cities of Lisbon and Porto, which started in 194862; and 3) a rent morato‑ delled the fiscal regime, indexed rents to property income and revoked
rium applied to the entire country in 197463 which lasted for 12 years. the rent freeze of 1919. There was also a permanent deferral of the
revaluation of properties, which present extremely outdated values, indi‑
In 1910, new agreements had to maintain the price of rents from
rectly used to curb the increase of income. This was the plan that froze
previous agreements64. Between 1914 and 1924, a rent control system
rents in Lisbon and Porto from 1948 to 1986, and made several ordi‑
was in force that distinguished maximum rent values between Lisbon,
nances, which were later published to allow rent increases, worthless.
Porto, other cities and the rest of the country. For example, in 1914
rents under 18 escudos in Lisbon, 15 escudos in Porto, 10 escudos in The strong devaluation the Portuguese escudo suffered in the 1920s
other cities and 5 escudos in the rest of the country, were frozen65, contributed towards the impossibility of low increases in rents offse‑
unless the tenant agreed to a higher value. In 1917, the 1914 Decree tting the effects of this devaluation. This situation is not unrelated to
remained in force, prolonging the rent freeze for values that had been the provision published in 1924, stating that «the value of the rents of
previously set, but new maximum values for rents were stipulated: urban buildings must always be set in Portuguese money and currency
25 escudos in Lisbon, 20 escudos in Porto, 13 escudos in other cities at the date of their payment»71, since there were many agreements
and 8 escudos in the rest of the country66. Rents above these limits contemplating payments in foreign currency or even through in‑kind
could be increased in up to 10% of their value67. In turn, the 1919 contributions, as had been approved by the 1867 Civil Code.
Decree determined that in case of the renewal of existing agreements,
rents could not exceed 50 escudos in Lisbon, 40 escudos in Porto, In 1928, rent increases are allowed based on the property's value, but
20 escudos in other cities and 15 escudos in the rest of the country68. up to the respective limits of those values72. In the case of houses that
became vacant or did not serve as permanent housing, landlords were was signed84. In practice, agreements signed in 1984 or 1985 under the
free to set the value of a new rent73. However, in 1948, the Estado free leasing regime, could only begin annual rent updates eight years
Novo imposed a system of rent freezes on the cities of Lisbon and after they had started.
Porto in a «draconian» manner, by blocking land tax assessments. In
1966, the Civil Code74 maintained the suspension of tax assessments With Decree‑law no. 321‑B/90, of 15 October, which publishes the
for the purpose of updating rents in these cities. RAU, a new impetus is given to the process of liberalizing housing rents.
Limited‑term agreements arise with a minimum term of five years, three
After the 1974 decrees75 which extended the rent freeze to the entire years being allowed in the case of real estate management and invest‑
country, it was only in 1981 that a decree was published in an attempt ment companies, and real estate investment funds. Rents are updated
to start reversing this situation, although it did not produce practical annually according to coefficients published by the government.
effects on existing leases. Two rent schemes were created: free leasing,
in which the parties can reach an agreement without an imposed limit Despite the extraordinary rent adjustments initiated in 1986, and the
on the value of the rent but not subject to updates76 and, the condi‑ new fixed term agreements created in 1990, a new update of rent values
tional rent scheme, with its value determined according to a rate and is attempted in 2006 with the publication of the New Urban Lease
formula which took the value of the dwelling into account77. The Regime (NRAU)85. This regime creates a complex rent update system,
conditional rent could be updated annually according to a coefficient which obliges landlords to request a fiscal assessment of the property
published by the government78. However, this decree maintained the and its state of conservation and, in case of unsatisfactory results
suspension of fiscal assessments for the purposes of updating housing impedes any increase to the rent value. It also establishes that rent
rents79, although it revoked the 1974 legislation which froze rents. value cannot exceed 4% of the property's value, and defines a phased
approach to rent updates, with intervals which can reach up to 10 years,
Despite the issuance of legislation which allowed rents to be updated depending on the tenants’ respective income. Adhesion to this system
as a result of maintenance and repair works on buildings, it was only was weak because, besides its complexity and obligatory fiscal asses‑
in 1985, with the publication of what became known as the Rent sments, it also depended on the properties’ state of conservation and,
Law80, that rent freezing ended and an extraordinary revision began as is well known, buildings in Portugal were very degraded after decades
for those prior to 198081. In 1986, conditional rent82 and rent subsidy without proper maintenance, leaving decapitalized landlords unable to
regulations83 were produced, but there were still situations on which conduct the works that would allow them to increase rents.
«liberalization» had no effect, as was the case of rents ulterior to 1980.
In addition, a rather curious provision about this transition of schemes It was under these circumstances, and already under the troika's
allows for the annual update of rents but only from the seventh year memorandum, that Law no. 31/2012, of 14 August, was published: the
onwards, counted from the end of the year in which the agreement most liberalizing impulse attempted to date in terms of urban leasing
in Portugal. This law establishes a transition period of five years for standards remain in 191988. As of 1917, the law penalizes situations in
all lease agreements, with a protection system for people over 65 or which landlords and tenants have agreed on rents above the stipulated
with a disability equal to or greater than 60%, during which the rent is limits, in these cases reducing the agreements to sixty days in order to
updated according to their income. After the transition period, rents force compliance with the new legal limits89. Two years later new legis‑
are updated to the limit of 1/15 of the property's value and a rent lation establishes that agreements are renewed for one year when their
subsidy is created for low‑income families. term is equal to or greater than one year, or for a term equal to the
agreement when it is less than one year90.
Despite the fact that in 2016 a process of reversing the liberaliza‑
tion measures of 2012 was initiated, there were no measures in force From 1924 onwards, the transfer of a lease due to the death of a
to cause a rent freeze, despite regulations imputing obligations to lessee, which was already possible for spouses, also became possible
landlords in terms of repossessions or compensation for tenants, in for any heir who had lived with the lessee for over six months91.
the case of works or agreement terminations. It is also worth mentio‑ Despite efforts towards greater openness in favour of landlords in
ning the creation of the affordable rental programme for the middle 192892, even in cases in which a dwelling was not for permanent use,
class, which will be analysed below, within the framework of the rent as of 1948 the transfer of a lease due to the death of the lessee incor‑
regimes associated with public housing policies. porated antecedents and descendants, who had shared the dwelling
for at least one year93, this way allowing for a second lease transfer94.
6.3.4. The duration and transfer of agreements
With the publication of the Civil Code of 1966, dwelling lease
The duration of agreements and the various forms and provisions used agreements began to have indefinite duration95, and could only be
to extend them, through their renewal, transfer or the right to a new terminated by landlords when a dwelling was by them required for
lease, cannot be separated from this discussion. Prior to 1910, a lease living purposes or reasons of major maintenance or construction.
could have a term of less than one month, and as a rule, written agree‑ These two rules were however suspended in 1974 and 1975, shortly
ments were not required, and rent could be paid on any day of the after the revolution, and repealed in 1977.
month. The 1910 Decree establishes that lease agreements cannot be
made for less than 30 days, rent is paid monthly and leases always start The rule for leases for an indefinite term only ceased to exist in 1990
on the first day of the month86. Longer deadlines are set for communi‑ with the introduction of RAU. Fixed term agreements96, though auto‑
cating opposition to agreement renewals: ninety days for agreements matically renewable at the end of the term for another three years,
with a term of more than one year, fifty days for agreements with terms could be terminated by the landlord with a year's notice97.
between 6 and 12 months, 20 days for agreements with 3 to 6 months
terms and 10 days for agreements with terms under 3 months87. These
In terms of lease transfers, in essence, the rent law maintains the exis‑ In 2019, more changes are introduced to the NRAU, namely,
ting regime with regards to the lessee's death, and includes those who a minimum term of at least one year for lease agreements105. Also,
lived with the lessee for over five years, in conditions similar to those fixed‑term agreements are renewed for at least three years, even if the
of a spouse, in cases in which the lessee was not married98. However, renewal term in the agreement is shorterand the landlord is in opposi‑
the 1990 law considers the possibility for excluding the right to tion106. In other words, the law requires that an agreement terminating
transfer if the tenant has another dwelling in a nearby location99. in 2018, which has been renewed for an additional year, should in fact
be subject to renewal for at least another three years, if the landlord
With the publication of the NRAU in 2006, the regime of fixed term does not timely communicate the intention to terminate the agree‑
agreements remains, but a new figure emerges, associated with the ment, leading to a four year extension. In the case of agreements for
termination of the lease agreement for the demolition or major building an indefinite term, cancellation by the landlord would have to occur at
renovations, regulated by Decree –Law no. 157/2006, of 8 August. This least five years in advance107, whereas before, according to the Law of
consideration gives landlords the possibility to terminate agreements 2012, the requirement was two years. However, the most unexpected
at least five years in advance or requires them to rehouse tenants or changes benefit tenants covered by the NRAU transition regime who
pay compensation for the amount corresponding to two years of rent. have resided in a dwelling for over 15 years and are 65 years of age
In a lease transfer, those who live in non‑marital cohabitation with the or older, or have a degree of disability equal to or greater than 60%.
tenant will have equal treatment and preference100. The reform of the These tenants receive a new life‑long lease entitlement, since landlords
NRAU, carried out in 2012, introduces total flexibility as to the terms cannot oppose the renewal of agreements108, a situation which also
of new lease agreements101, thus allowing agreements with previously applies to similar cases regarding tenants who have lived under fixed
established minimum terms of less than five years. Mechanisms for the term lease agreements for over 20 years 109.
landlord to oppose the renewal of the agreement are facilitated and
respective term periods, previously three years, are reduced102. Decree
6.3.5. Eviction processes
‑law no. 157/2006, of 8 August, is also amended, facilitating solicitations
for demolition, remodelling or major renovations, in order to terminate Eviction is undoubtedly the most sensitive and controversial element
leases. In 2017, the five‑year deadline to shift to NRAU, set in 2012, is of tenancy regimes in any society, and the Portuguese case is no excep‑
extended for another three years, thus totalling eight years103. This rule tion. Given the great legislative complexity and the countless changes
is changed again in 2020, with a new two‑year extension, introduced that have occurred over the last century, we will not go into the same
when the 2020104 Government Budget was approved, increasing the detail as in the two previous topics. The 1907 decree, which had libe‑
NRAU transition period to 10 years. ralized eviction procedures, was revoked in 1910, and several rules
were introduced, clarifying its handling by court and the way in which
tenants opposed it. However, the timeframe for these procedures were more expeditious. First, with Decree no. 22661 of 13 June, 1933, and
quite limited, in some cases with a maximum limit of five days110. later with the Civil Procedure Code published in 1939115.
In 1917, eviction based on lack of convenience for landlords in conti‑ In 1948, Law no. 2030 suppresses the expiry periods for eviction
nuing the lease was also prohibited, regardless of the rent value111. There actions, defined in the 1924 decree, simplifying and accelerating the
were a few exceptions, such as maintenance and repairs approved by respective processes when leases expired. However, there are still
city council, damages resulting from tenants’ negligence, refusal to cases in which eviction cannot be carried out either due to illness or
accept rent increases permitted by law and conduct causing nuisance or absence resulting from official duties, such as public office or the mili‑
disturbance to neighbours, such as «pounding, dragging furniture, and tary, for up to two years.
the like» between one and nine in the morning. Some of these rules,
which already came from the Civil Code of 1867, are still in force today. This decree maintains the provision of appeals of eviction sentences
Nowadays, speaking of placing notices on windows is like remembering a to the Court of Appeal, but forbids the appeal to the Supreme Court,
relic. However, during the first half of the 20th century the law required even when the amount exceeds its sphere of authority, thus revoking
tenants to do so in cases of lease termination or eviction, in order to the 1925 decree.
allow visits to the home, often at predetermined times such as on week‑ With the Civil Code of 1966, restitution of the leased premises could
days from midday to 5 pm. The law also foresaw forced entry into the only be claimed three months after the verification of the facts which
home if tenants failed to comply with displaying their notices112. determined the lease expiry or termination116, and in cases of denun‑
The perception of social repercussions resulting from evictions, and ciation by legal action this verification would be required at least six
the total inability of public authorities to respond to them, contri‑ months in advance, and eviction would only occur three months after
buted to the emergence of new legal provisions to protect tenants in the final decision117.
the following decades. After the 1974 revolution, evictions resulting from the denuncia‑
In the 1918 decree, in cases of proven serious illness of the tenant or tion of leases, for demolition purposes or for occupancy by owners,
family member, evictions would be suspended until the tenant had were suspended with Decree‑law no. 155/75, of 25 March, and
recovered, though the suspension could not exceed two months113. In decrees specifically intended for the urban areas of Porto118. It is
contrast, in the 1924 decree, eviction is determined for cases where the during this period that the turbulent process of illegal occupancy of
tenant has not used the leased premises for over a year114. This rule, still vacant dwellings occurs, with two attempts at legalization in 1975
present in the Civil Code is, however, difficult to prove in court. Over and 1977119. The legal provisions accumulated since 1976 to protect
the following decades, there was an effort to make the eviction process tenants’ rights before an increasingly degraded housing stock, and the
lobbying of landlords using mechanisms provided by the law, such as Code of 1966126 and reinforced in 1990 by article 14 of Decree‑law
demolitions or major construction works, and even the claim of the no. 321‑B/90, of 15 October. This decree establishes landlord‑imposed
leased property for their own housing, created a legal «entanglement» rent deposits exceeding the amount corresponding to one month of
which, together with the increasing inefficiency of the legal system, rent as a speculative crime. With the NRAU, advance payments of
led to delayed eviction actions in the courts, with some cases taking up to three months were allowed127. Despite these legal restrictions
more than 10 years. With the creation of the National Leasing Office placed on individuals, the «Safe Rent Programme» (Renda Segura),
(Balcão Nacional do Arrendamento) and the special eviction procedure launched by the Lisbon City Council in 2020, offers up to 36 months
in 2012120, there is an improvement in the eviction process. Reports of rent anticipation.
from January 2018 indicated that processes have only lingered on
average three to four months121. Also in 2012, landlords were allowed 6.3.7. Penalties for rent arrears
to proceed with the termination of agreements when two rent instal‑
The first reference made to rent arrears was in 1924 and stated that
ments were in arrears122. A situation altered by Law no. 43/2017 of
tenants may avoid eviction if payment of five times the amount of
14 June, which restores the rule that had been in force in 2006, consi‑
arrears is made within eight days128. In 1933, the penalty is reduced
dering a minimum of three instalments. Moreover, in 2012, a rule is
to three times the rent owed129. With the approval of the Civil Code
included which allows landlords to terminate the agreement in case of
in 1966, the penalty is further reduced to double the value of the
four consecutive or non‑consecutive periods of overdue rent within a
rent owed130. In 1977, tenants could simply deposit 50% of the owing
twelve‑month period123. This norm continues in force today.
rent131 as long as payment was made within eight days, in order to
prevent eviction. In 2019, the penalty is reduced to 20% of the rent
6.3.6. Rents paid in advance
owed132. The series of reductions made in terms of compensation are
The issue of advanced rent payment has changed over the past a move towards its growing irrelevance, which may call into question
decades, with legislators seeking to impose limits and landlords defen‑ the obligation of the timely payment of rent.
ding its increase, since it is the main and most immediate form of
guarantee for situations of late rent payment or damages incurred. In 6.3.8. The state's social support instruments
1910, it was forbidden to anticipate rents beyond the amount corres‑
In addition to resettlement actions and the construction of social
ponding to one month, and in these cases, payments were required
neighbourhoods, which began several decades ago, it was only in 1977
to be made at the beginning of the corresponding month124. This rule
the state created its first mechanisms for the protection of evicted
was changed in 1919, allowing advance rent payment on the first
tenants and for the compensation of landlords, in cases where the
business day of the month prior to the corresponding period125, or
eviction of tenants is deferred. The Family and Social Action Institute
guaranteed by legal provisions, an amendment maintained by the Civil
(Instituto da Família e Acção Social) is responsible for managing these With the reform carried out in 2012, the creation of a new income
situations, promoting resettlement solutions for tenants, compensa‑ subsidy was expected to replace the previous one, through Decree‑law
ting landlords for the time they defer evictions and creating a register no. 156/2015, of 10 August. However, this did not occur due to the
of tenants who benefit from these supports133. change of government at the end of 2015.
With the publication of the Rent Law134, a rent subsidy is created, In the following graph it is possible to observe the size and evolution of
regulated by Decree‑law no. 68/86, of 27 March, and aimed at tenants rent subsidies between 1987 and 2019. It should be noted that, in 2008,
whose rents have undergone extraordinary corrections or have been the values of the IAJ and Porta 65 Jovem subsidies should be added.
increased due to conservation works or improvements carried out
by the landlord or the municipality. This allowance was in force from
1987 to 2011. Figure 80Evolution of spending in rent subsidy programmes
1987
1989
In 1990, the RAU attributed the responsibility and means to support 1991
landlords and tenants, in situations where the deferral of housing 1993
1995
eviction occurs, to the Institute for Financial Management of Social 1997
1999
Security (Instituto de Gestão Financeira da Segurança Social), a system 2001
2003
still in place today. 2005
2007
2009
Within the scope of youth support policies, the Rental Incentive for 2011
Young People (IAJ)135 was created in 1992 and operated until 2007, then 2013
2015
being replaced by the Porta 65 Jovem programme136, which introduced 2017
2019
several changes, the last of which in 2017. Initially, the programme was 0 10 20 30 40 50 60 70
aimed at young people up to the age of 30, a limit then extended to Amount in millions of euros
the age of 35.
Subsidies for the Rent Law (1986) Rental Incentive for Young People (1992)
137 Subsidies for Porta 65 Jovem (2007) Subsidies for the NRAU (2006)
As with the Rent Law, a rent subsidy is also created with the publi‑
cation of the NRAU in 2006, aiming to support those most in need, as Source: IGAPHE, IHRU
some become subject to a higher burden rate, according to their income,

as a result of rent increases. However, given landlords’ weak adherence
to rent updates, this subsidy has less effect than that of 1986.
6.3.9. The issue of vacant buildings With the approval of the 2005 Government Budget, a provision is
inserted that alters article 112 of the IMI Code143, allowing munici‑
One of the elements which is always present in the lease legislation,
palities to increase the rate applicable to urban buildings which have
as well as in the political discourse, is the issue of vacant dwellings and
been vacant for at least one year in up to 30%. The absence of ongoing
buildings.
agreements with essential public service providers or billing related to
The 1914 Decree created a mechanism which allows anyone to the consumption of water, electricity, gas and telecommunications.
force the rent of a vacant dwelling, depositing the rent in court in
With the publication of the NRAU in 2006, this change to the IMI
the landlord's name, even if the landlord refused to rent it out138.
Code acquires new wording and, in the case of urban buildings
Exceptions included situations on premises requiring repair work.
which have been vacant for over a year, the respective tax rates are
In 1919, this obligation is extended to tenants and sub‑tenants whose doubled144. At the same time, a tax concept for vacant buildings is
dwellings are unused, further imposing that the new rent value cannot created by Decree‑law no. 159/2006, 8 August, pursuant to which the
exceed that of the last agreement139. signs of vacancy previously defined are maintained.
This matter is only resumed after the 1974 revolution, with owners In 2008, tax rates on buildings vacant for more than a year are doubled
being obliged to accept a lease when, among other conditions, the and, in the case of ruined buildings, tripled145.
previous one has ceased more than 120 days ago140. The rent value
With the approval of the 2012 Government Budget, these rates are
however «would result from free play of supply and demand»141.
once again increased, tripling in both cases146.
With the Rent Law, a rule that increases property tax is approved and
In 2018, the government makes an exception for «unused public real
applied to buildings or parts of vacant buildings. This value results
estate»147, legislating in its own cause, to protect itself from the aggra‑
from the calculation of rent in a conditional rent regime and is attribu‑
vations it applies to the private sector.
table to landlords.
The following year, the increase in the tax rate covers vacant parts
The issue of vacant dwellings was resumed in the year 2000, with
of buildings that are not constituted by horizontal property, thus
respect to the execution of coercive works in private buildings by the
allowing its partial application, which until then was not legally
municipalities, giving them the possibility of maintaining possession
possible148.
of the vacant dwellings and renting them under the conditional rent
regime, for a period which could vary between three and eight years142. Still in 2019, an article is added to the IMI Code as part of the creation
of the urban pressure zone figure, which originates a sixfold increase
in the IMI tax rate and a 10% annual increase, for buildings and auto‑ 6.4.2. The «charitable» income system
nomous fractions located in these areas which have been vacant for
The demountable housing programme, launched in 1938 with the initial
over two years149.
intention of relocating families living in Lisbon's shack neighbourhoods
(Bairro das Minhocas and Bairro da Bélgica), does not include a clearly
6.4. Social lease systems
defined rent regime. Occupancy of houses takes place on a precarious
6.4.1. A wide variety of regimes basis; the rent is fixed and received by the Lisbon City Council153, and
residents may be forced to vacate them within 30 days, or face adminis‑
So called «social» rent schemes took many forms since their first prac‑
trative eviction. This programme was extended to Porto, in 1943154 and
tical application in 1938, with the demountable housing programme150.
to the rehousing of Alta de Coimbra, in order to allow the construction
It is important not to confuse the regime of economic housing, of the Coimbra University, between 1945 and 1949155.
instituted in 1933151 — which took advantage of the first social
It was then followed by the housing programme for poor families
neighbourhood construction policy, launched in 1918152 —, with social
— which sees its execution extended to administrative bodies and
housing leasing systems. In these regimes (which should also not be
Misericórdias (network of charity institutes), establishing the following
confused with low‑income housing), the dwellings were left on the
in its preamble: «implementing very low monthly rents — a few dozen
property of their residents after a period of amortization, during which
escudos — within the limits that are feasible to those for whom they
the state played a role which nowadays belongs to the banking system.
are intended». The dwellings are allocated on a precarious basis and
Over these eight decades, three types of rent systems can be identi‑ rents are fixed, in each case, with the approval of the Minister of
fied. The first, of charitable nature and casuistic analysis, for which Finance, not only taking into account the local conditions but also the
there are no regulatory guidelines established in legal decrees, and residents’ income156. The houses are primarily «assigned to poor fami‑
aimed at «poor» families. The second, with fixed or conditional rent, lies displaced as a result of home demolitions due to urbanization or
determined according to construction costs — but always below other actions of public interest».
market values —, or according to the burden rate or family income.
Through Decree‑law no. 35106 of 6 November 1945, which regulates
And, the third and last, a social income system, with values varying in
the regime for assigning, occupying and transmitting the rights and
accordance to households’ income and composition. Amongst these
obligations of residents, as well as the grounds for home eviction, it
systems, both the first and last allow rents to increase or decrease,
is clear that the decision of establishing rents is the responsibility of
according to families’ changing situation.
the Minister of Finance, who must take «local conditions and the resi‑
dents’ situation» into account.
There is, therefore, a centralized and casuistic decision‑making process, multiplicity of entities, from municipalities, cooperatives, companies
of discretionary nature, which lies in total contrast with what occurs and manufacturing companies to Casas do Povo from 1958 onwards160.
as a result of the economic housing programme, created in 1933. It also applies to housing for public service authorities, such as
the Armed Forces, PSP (Public Security Police) and GNR (National
The programme had a significant impact with the involvement of a Republican Guard). In 1945, basic rents under this regime vary between
wide variety of entities, including the Património dos Pobres (Heritage $300 and $500, for single houses, and between $240 and $40, for apart‑
of the Poor) and the Casas de Pescadores (Fishermen's Houses)157. ments. The base rent is the maximum rent that can be applied, and its
However, as early as 1940, and because there was no legal framework update depends on government authorization, taking into account
for this regime, the Porto Municipality asked the national government the increase in construction or the cost of living. The agreements are
to apply the decree of demountable housing to the real estate deve‑ celebrated for a period of one year and are renewable. Their breach is
lopment being built on Rua Duque de Saldanha. The authorization subject to administrative eviction, namely, in the event of a delay in
was attributed by Decree‑law no. 30750, of 13 September 1940, on the payment of three rents.
the condition that a special regulation for this purpose be submitted Only families registered as partners in regulatory bodies have access
to the government for approval; revealing that, in the absence of this to low‑income dwellings, as long as their income does not exceed an
authorization, the city council would have to apply the general laws of amount equivalent to six times the given rent. If, during the lease,
urban leasing. This situation is repeated in 1957, when the government the total income of the household exceeds 20% of this amount, the
publishes new legislation158, in which it is assumed that there are many family will have to leave the dwelling within six months or face evic‑
dwellings «built by initiative of associations or social assistance insti‑ tion. The tenant must provide annual proof of his household's income.
tutions» and intended for housing poor families, reason why «it was
understood that the occupancy of these dwellings be removed from
6.4.3.2. Limited rent housing
the general legislation on tenancy».
The limited rent housing construction programme was launched in
6.4.3. Fixed or conditional rent schemes 1947161 and was designed to encourage the promotion of «investment
buildings» within the private sector. These «investment buildings»
In this type of rents, four regimes are commonly identified: demand that a maximum rent value be previously fixed for each dwelling
in return for access to cheaper land plots and exemption from several
6.4.3.1. Affordable housing taxes, fees, emoluments, licenses, and also property tax for 12 years.
The affordable housing regime created in 1945159, which lasts until
the creation of the Housing Financing Fund (FFH) in 1969, involves a
The fixing of the rent value aims to counter existent speculative construction is financed by the State, such as cooperatives, housing
processes and is accompanied by several measures to combat cost development contracts and residents’ associations. It is also estab‑
increases in construction materials and reduce tax exemptions for free lished that when there is a right to a new lease167 conditional leasing
rent promotions. is applied, which represents, to date, an increase in the values of the
previously frozen rents. These rules are maintained in the RAU of
In 1958162, the programme was extended for another 10 years, accom‑ 1990168, which extends their application to cases of transfer by death
panied by measures to encourage the promotion of this type of lease, and, also, as a penalty for lease agreements not reduced to writing169.
to the detriment of the free lease. Tenants are allowed to purchase the This rent regime had little impact as there was no inspection to ensure
houses they inhabit, and it is determined that municipalities give prio‑ the application of the obligations determined by law. The registration
rity to these promotions, in public auctions of land163. of any burden of conditional leasing was not mandatory, its formulas
From 1974164 and until this leasing regime is extinguished, the exis‑ were complex and there was lack of tax assessment processes, which
tence of a single housing category is established, and distinction is were oftentimes excessively time‑consuming.
made by typology. Annually, the government publishes an ordinance Since 1990, changes to this regime have been foreseen, depending on
with the update on the criteria applicable for establishing the limit of the publication of the Assessment Code, which only occurred in 2003,
rent values, either taking into account typologies or dwelling areas. with the publication of the IMI Code. Finally, in 2014170, the change is
In 1977165, this decision is decentralized to the municipalities that made with regime being used to determine the maximum value of rent
provide housing services. in the supported rent regime, as well as the value of leases in the Social
Leasing Market, which involved a few thousand housing units belon‑
6.4.3.3. The conditional rent regime ging to the Institute for Housing and Urban Rehabilitation (IHRU) and
This rent regime started in 1981166, as the first attempt to reopen the the several banks.
rental market to new agreements, after the rent freezes of 1948 and
1974. In this regime rent is based on the value of dwellings and, unlike 6.4.3.4. The affordable rent regime
free leasing, which had then been relaunched, sees it value updated on In 2019, the government launches the affordable rent programme171,
an annual basis. Hence, its intention is to attract landlords to a non which has some similarities to the limited rent regime. Its main objec‑
‑speculative rental system that benefits from property tax exemption tive consists in motivating landlords to rent dwellings for values
during the first five years. With the publication of the Rent Law in about 20% below the practiced market value, based on the median of
1985, conditional rent becomes mandatory for dwellings built by the rents calculated by INE172. Lease agreements would continue to have
state and sold to their respective residents, as well as those whose a minimum term of five years, and nine months for student leases.
The rent value must represent a burden of 15% to 35% on household However, the measure has poor results, given that problems with the
income, and in return landlords benefit from property tax exemption, collection of unpaid rents still persists.
applied to IRS or IRC, depending on the case.
When the Special Rehousing Programme was launched in 1993, the
supported rent regime177, which replaced the social rent regime and
6.4.4. Social, support and rent‑support schemes
had been set out since 1990, is also published.
In 1977, and within the scope of the regulation of the Land Law, the
first ordinance to institute the social income regime in Portugal173 There are several issues in question: the standardization of the various
is published. This regime introduces a system which is still in use existing social rent regimes, some of which originated in the 1940s;
today, establishing that the value of rent must be calculated accor‑ the creation of a mechanism for updating the maximum (technical)
ding to household income and composition. This scheme establishes a rent, suspended since 1977 and unsynchronized with the construc‑
minimum rent of 400 escudos and a maximum rent — corresponding tion costs and income of families with greater resources; the annual
to the dwelling's so‑called technical rent174 — which is applied to all and automatic update of the rent value; the end of the ambiguous
households with incomes above three times the national minimum rent subsidy that had never been accounted for; the creation of a
wage (SMN), whereby the rest of the household pay a social rent — progressive system of income burden rates, depending on the income
called individual rent instalment. The difference between social rent of household and the indexation of the minimum rent to the national
and technical rent is considered to be a subsidy in the form of a grant, minimum wage.
and despite the terminology, it has never undergone proper regula‑
The social rent system is resumed, in terms of its main variables, in
tion or been accounted for. Households must provide proof of their
order to follow the variation of increases according to the national
composition and income for the purpose of updating the rent, on an
minimum wage, income and construction costs, without losing
annual basis. The social rent regime is revised in 1983175 but essentially
its original characteristic of evolving in accordance to income and
maintains the previous rules and standards. Despite high inflation, and
household size.
six years having passed since the first ordinance, the minimum rent
remains at 400 escudos. The threshold for the payment of technical These changes, despite having been successfully applied to houses
rent remains the same, and despite the announced intention to explain built after 1993, did not work for older dwellings, due to large diffe‑
the value of income subsidies granted in terms of rent reductions, rences between current and new rents, resulting from their tenants’
for a clearer understanding of the State's effort, this did not happen. increase in income. In many cases there were no rent updates for
A provision allowing all tenants with overdue rent the possibility of 40 years and households with incomes above two thousand euros per
payment without the 50% fines set forth by law is also introduced176. month paid a rent that was below three euros!
In order to correct these issues, the supported leasing178 regime was 6.5. Is it possible to regulate housing
created in 2014, seeking to evolve from a «rent regime» to a «lease supply through legislative action?
regime», by covering all situations of social nature in place with the
In the 50‑year interval between 1960 and 2011, the weight of rented
old systems, some of which were incompatible with the 1993 decree.
dwellings fell from 48% to 19.9% of the housing stock of usual resi‑
The attribution regime was integrated with the management system,
dences and from 44.4% to 13.6% of the total of classic dwellings.
giving it a binding national scope for all entities and creating a system
The latest available figures from 2019 indicate that this decline conti‑
of progressivity in the value of rents, for families with higher incomes.
nued after 2011.
The rent calculation system is maintained, considering household
The legislative measures approved and applied to the field of leasing
income and the possibility for changes, just as before. However, a
since 1910, and promoting the right to housing, usually in circums‑
fundamental change is introduced concerning the powers of adminis‑
tances of crisis, had perverse effects which drastically reduced supply
trative eviction, with housing stock management entities having the
and pushed investment away from this segment. Consequently, a
authority to intervene in order to reduce rents due and illegal occu‑
«home for life» immobility culture develops, as well as a confrontation
pancies, which increased in a concerning way.
of Manichaean nature between owners and tenants, and permanent
This regime is changed a few months after the beginning of the housing is no longer a priority in the budgets and savings of families
following legislature179, with the revocation of innovations which with resources, living in dwellings with frozen rents or social housing.
extended the powers of management entities, immediately eliminating
The censuses show that the decline in rented dwellings occurs when
the process of administrative eviction. Rent update limitation is intro‑
housing stock more than doubled. Leasing becomes the black sheep of
duced in cases where the tenants invoke maintenance needs on the
the real estate market in Portugal.
premises; progressive increase in the rents of higher income families
is removed and the law's national scope is abolished, by allowing each In addition to legislative changes there are also almost caricatural
municipality to approve its own regulation. moments, such as the case of the prohibition of real estate brokerage
activities in matters of leasing, as decreed in 1919181; the idea launched
According to the latest Inquiry into Social Housing Characterisation
in October 2016 — which did not reach the letter of law — of «crea‑
(ICHS), carried out by Statistics Portugal, and referring to 2015180, the
ting a subsidy for poor landlords with old rent agreements»; or, still,
accumulated value of rents owed had already reached 80 million euros.
the passage from the preamble of the 1947 decree which created the
It also estimated that, by the end of 2019, this amount would exceed
programme for the construction of limited rent housing182: «(…) In this
100 million euros. There are many situations in which tenants refuse to
order of ideas, perhaps the radical measure to expressly prohibit the
pay the rent the law requires.
construction of free‑income buildings should be adopted now, thus
leading the respective industry to concentrate all its resources on the The political and economic changes occurred in the last decade are in
construction of houses for low income (…), but, as it is recognized that turn interesting, exposing an initial period during which the real estate
the problem requires a more in‑depth study, in order to coordinate its crisis brings a new glow to the lease market, with increased supply and
multiple aspects, it is not, for the moment, on such a radical path. (…)». reduced prices. The 2011/2012 reforms reinforce this supply, but are
shortly thereafter frustrated by the anathema of the «eviction law»
This text is published two months prior to the announcement of the which, after 2015, opens Pandora's box, with the resumption of real
Marshall Plan for the reconstruction of Europe, which Portugal, as is estate activities, the tourist boom and the succession of announce‑
known, joined. ments, measures and decrees to «regulate» the market, creating the
When we revisit the measures that intend to compel owners to put perfect storm: all landlords who are able to, restrain from renewing
vacant houses and buildings to use, especially in the last two decades, lease agreements and change their dwellings to tourist accommo‑
what we find is a fiscal escalation where, year after year, tax rates dations. There is a sharp drop in the supply of houses to rent and as
increase, without any practical results. the following graph shows, prices skyrocket, especially in Lisbon and
Porto.
Contrary to what happens in most Western European countries,
Portugal did not gain budgetary «muscle» to develop a housing promo‑
tion policy and build a significant public housing stock, capable of Figure 81Index of residential rents (2011=100)
assisting families with greater difficulties in accessing housing. It 160 157.3 158.2
146.7
did not do so during the dictatorship of the Estado Novo, nor under 140
139.6
130.0
democracy, after the April 1974 revolution. Today, this housing stock is 122.4
114.6
120 127.7
made up of about 120 thousand dwellings, which represent only 2% of 102.4
106.5 121.4
96.0 95.0 97.3 109.9
100
the total number of dwellings. 99.5
95.4
90.1 88.4 92.4
95.2
80 88.0 86.5 89.4
85.9
As a result of this incapacity, the problem arises. With rent freezes and
60
the perpetuation of rental agreements, the state pushes its social and
13
14
15
16
17
18
19
11
12
20
20
20
20
20
20
20
20
20
constitutional responsibilities regarding social housing to the private
1
1
Q
Q
Q
sector. Landlords, increasingly decapitalized and in possession of
Portugal Lisbon Porto
progressively degraded buildings, face a real estate market that thrives
in all segments, except leasing. Source: Confidencial Imobiliário
6.5.1. The incentive for owner‑occupied dwellings Figure 82State efforts for housing 1987–2011
Private initiatives have for long abandoned leasing and channelled € 29,2 M; 0,3% € 193,9 M; 2,0%
their efforts to safer and more profitable initiatives. Public authorities, € 1.353,4 M; 14,1%
who are unable to face the controversies and confront the political
€ 166,5 M; 1,7%
costs that would result from the liberalization of urban leasing, found € 13,8 M; 0,1%
a solution in subsidized credit, financing new construction and urban

expansion, and encouraging families to buy a house. It should be noted € 803,9 M; 8,4%
that the same decree which is responsible for freezing rents in the
Former IGAPHE direct promotion programmes
cities of Lisbon and Porto in 1948, has a provision that requires the
Rehousing programmes
government to establish the system of property by floors or horizontal Building rehabilitation programmes
property within 180 days183. This is one of the great drivers of poli‑ Housing development contracts programmes
€ 7.046,7 M; 73,3%
cies promoting self‑owned housing, which become dominant at the Leasing incentives
end of the 20th century. Until then, the so‑called property by floors, Interest subsidies on housing loans
dealt with in small article 2335 of the Civil Code of 1867, was rarely Social security housing grants
applied. It is in 1986, the year the Rent Law is regulated, that the subsi‑ Source: IHRU
dized credit for the acquisition of own housing184, created in 1976185,
is relaunched with great success. The study published by the IHRU in The largest share belongs to interest subsidies on housing loans
2015 on the State's budgetary effort to provide subsidies during the and the acquisition of permanent own housing (73.3%), followed
25 years between 1987 and 2011, is a faithful picture of the period of by rehousing programmes (14.1%). Leasing support initiatives have
greatest brilliance regarding policies and public investment in the area a smaller representation (8.7%). The most empowered, the Rental
of housing, in which leasing shows little impact, as can be seen in the Incentive for Young People (IAJ), presents significant results in 2006,
following graph. with 63.2 million euros of non‑repayable grants. However, it is in 2001
that the largest number of lease agreements is subsidized, totalling
24,676186, that is, only 3.3% of the 740,425 leases existent in Portugal
that year187. The result of these policies is the inversion of the weight
between rented housing and owner‑occupied housing, with the latter
becoming dominant and holding almost three quarters of dwellings of
usual residence.
Is it cheaper to buy or rent a house? 6.6. The segmentation of the lease market
When interest rates started to fall, the perception that it is more favou‑ Whatever remains of the rental market in Portugal is deeply
rable to buy a house than to rent it is created, since the provision of segmented, even more so than the labour market. This segmentation
mortgage loans has similar or lower values in comparison to the value of originates from the three rental regimes in force — subsidised, affor‑
the rents, even after the end of subsidized loans in 2002. The series of dable and free — causing a great disparity in the way families are
housing yields prepared by Confidencial Imobiliário in the last decade188, treated in terms of:
which result from the comparison of transaction prices with rents under
• housing access rules;
new agreements, presents rates that vary on average between 4.1 % and
• duration of lease agreements;
5.3% in the cities of Lisbon and Porto, in the last two years, showing a
• exceptional treatment resulting from the household's socioeconomic
more favourable perspective on purchase than on lease.
condition and the age or disability of any of its members;
There are, however, three factors with regard to property‑related • systems that determine the value of rents.
obligations that have been overlooked. The first results from taxes
Regarding housing access rules, it appears that, in the case of subsi‑
and fees on assets, in particular the IMI. The second corresponds to
dised leasing, there is no minimum or maximum limit on household
insurance costs. In Portugal, fire insurance is only mandatory for hori‑
income to access a home, unless the household comes from the
zontal property, when legislation should have already made multi‑risk
1º Direito (1st Right) programme, which as a rule excludes households
insurance for collective buildings and fire insurance for single‑family
with an average monthly per capita income 1.5 times above the value
buildings mandatory. The third includes expenses associated with
of the social support index (IAS)190. Now, in affordable rent, the
commonholds, regarding the conservation and future maintenance of
average monthly income of the household is required to be compatible
buildings. It is anachronistic that after 26 years the law still states the
with a burden rate between 15% and 35% of the rent value.
rule of keeping at least 10% of the monthly fee regarding commonhold
assessment as a reserve fund for the future maintenance of the buil‑ As for the duration of the agreements, subsidised leasing provides
ding189, when these savings can represent less than one thousand euros lifelong agreements, in practice, while affordable rent provides fixed
after a 10‑year period, which is clearly insufficient to cover the general ‑term agreements with a minimum duration of five years. In the case
maintenance of any building. of free leases, and although the rule points to fixed‑term agreements,
legal changes are made in 2019 to make some of them lifelong: in the
If the legal changes occur to make the set of owner or joint owner
case of tenants who have been on the lease for more than 15 years
obligations consistent with an assessment of all costs, the differences
and are over 65 years old, as well as those with fixed‑term agreements
identified today in housing yields will be reduced or even reversed.
that have lasted for more than 20 years. On the other hand, for both
subsidised leasing and the transition processes operated by the NRAU a common denominator. In addition, the subsidised rental regime has
regarding free leases, mechanisms are created to protect households no interaction with social security, so often beneficiaries of subsidies
taking into account income, age and situations of disability, conditions and support from the social state — pensions, social support, health
which neither included in affordable rent nor, generally, in free leasing. and education — do not pay the rent of the dwellings they lease,
The greatest disparity, however, is registered in the value of rents. In without incurring any penalties for non‑compliance.
subsidised leasing, income is calculated according to the composi‑
tion and income of the household, which does not occur in the other 6.7. Brief conclusion
schemes. But, as subsidised leasing establishes a maximum income for
The problem of urban leasing in Portugal is not short‑term. It is not the
each dwelling, regardless of the household's income, there are many
result of the 2012 reform, nor of the increase in tourist demand which
households with high incomes under this regime who pay much less
has occurred in recent years. The issue is structural and lies in several
in comparison to lower values promoted by affordable rent, which are
bottlenecks that cannot be concealed. From the outset, the state's
often supported by households with one third of the income. As an
inability to create adequate responses to housing shortages. Then, the
example, there may be a household with a five thousand euro monthly
«home for life» culture, which took root and became an anachronism,
income benefitting from subsidised leasing who pays a rent of two
given the evolution of the labour market and the new demands for
hundred euros, while another, under the affordable rent regime, has an
mobility. Finally, the ideological prejudices, which created too many
income of eighteen hundred euros and pays a rent of six hundred euros.
misunderstandings and conflicts, between the right to housing and the
, Some programmes benefit from public support, so in many of these right to property, and today are the main cause for the loss of investor
situations, paradoxically, two equal households with equal income, in confidence in political institutions and the enormous reputational
identical dwellings within the same building, sometimes even on the damage the lease market has suffered with the succession of legislative
same floor, pay very different rents, because they are under different advances made by the latter. In practice, and perversely, public policies
rent regimes. On an even more serious note, it sometimes happens that and legislative measures have been promoting and subsidizing inequa‑
households with lower incomes pay significantly higher rents. And, lity. At the same time, the state was — and still is — unable to meet
just as happens in the labour market, for younger people, although the demand, contributed to contracting supply and discarded its social
there is greater difficulty in accessing homes and greater insecurity in responsibilities to the private sector.
the duration of lease agreements, rents have a higher value. In short,
In short, urban leasing continues to languish and find itself at a dead end.
not only does is intercommunication and harmonization between
public supported schemes inexistent, but each one determines the
support and benefits to be provided in different ways, with income as
6.8. Conclusion especially to Statistics Portugal, instead of watching its surgical and
sporadic dissemination in the media, which invariably raises suspicions
The present study is carried out on the eve of the start of the 2021
about the political purposes of its management.
census, which will reveal, in all dimensions, the changes occurred over
the last decade. After this census, it will be helpful to have complete
Abbreviations
information on the evolution of the housing stock and the rental
market, at least on an annual basis. AT — Tax and Revenues Authority (Autoridade Tributária e Aduaneira)
This will allow a monitoring of the existing reality, and a timely and FFH — Housing Financing Fund (Fundo de Fomento da Habitação)
close assessment of public policies and investments made, which has
IAJ — Rental Incentive for Young People (Incentivo ao Arrendamento por Jovens)
not happened so far. It will also be relevant to carry out an in‑depth
study of the results of the next census regarding vacant dwellings, IAS — Social Support Index (Indexante dos Apoios Sociais)
which will allow an inventory of their location, age, state of conser‑
vation, feasibility of use and the situation of the respective property. ICHS — Inquiry into Social Housing Characterisation (Inquérito
The 2011 census quantified 735 thousand vacant classic dwellings, à Caraterização da Habitação Social)
which accounted for 12.5% of all conventional dwellings. However,
IGAPHE — State Housing Management and Administration Institute
part of these dwellings are either in depopulated territories, just a pile
(Instituto de Gestão e Administração do Património Habitacional do Estado)
of abandoned stones or part of undivided inheritances whose bene‑
ficiaries do not attribute any value to, excusing themselves from the IHRU — Institute for Housing and Urban Rehabilitation (Instituto
task of dividing the property in face of the bureaucratic mazes they da Habitação e da Reabilitação Urbana)
will have to deal with. Therefore, it is important to understand, which
vacant dwellings are inhabitable. IMI —Municipal Property Tax (Imposto Municipal sobre Imóveis)
It is also important to have knowledge, in detail and frequency, of data INE — Statistics Portugal (Instituto Nacional de Estatística)
relating to the management of public or social housing and the third
IRS — Personal Income Tax (Imposto sobre o Rendimento das Pessoas
sector, and the support component for housing integrated in the subsi‑
Singulares)
dies granted by social security.
IRC — Corporate Income Tax (Imposto sobre o Rendimento das Pessoas
Finally, since the Tax and Revenues Authority (AT) has a huge and
Coletivas)
unique repository of data related to real estate and leasing, it will
make sense to go further in making this information available,
NRAU — New Urban Lease Regime (Novo Regime de Arrendamento Urbano) Instituto Nacional de Estatística, (2001), XIV recenseamento geral da população
e iv recenseamento geral da habitação (CENSOS 2001), Lisboa, Gráfica Maia
PREC — Ongoing Revolutionary Process — which designated the period Douro, SA.
between April 1974 and November 1975 (Processo Revolucionário Em Curso)
Instituto Nacional de Estatística, (2012), XV recenseamento geral da população
RAU — Urban Lease Regime (Regime de Arrendamento Urbano) e v recenseamento geral da habitação (CENSOS 2011), Lisbon, INE.
SMN — National Minimum Wage (Salário Mínimo Nacional) Instituto da Habitação e da Reabilitação Urbana, (2015), 25 anos de esforço
do Orçamento de Estado com a habitação — 1987–2011, Lisboa, IHRU
References
Instituto Nacional de Estatística (2016), Inquérito à caraterização da habitação
Instituto Nacional de Estatística, (1964), X Recenseamento geral da população social em Portugal, 2015, Lisbon, Statistics Portugal.
1960, Lisbon, Sociedade Tipográfica, Lda.
Various authors, (2018), Habitação: Cem anos de políticas públicas em Portugal,
Instituto Nacional de Estatística, (1973), 11º Recenseamento da população 1918–2018, Lisbon, Instituto da Habitação e da Reabilitação Urbana, INCM.
1970 — Estimativa a 20%, Lisbon, Sociedade Tipográfica, Lda.
Instituto Nacional de Estatística, (1975), 1º Recenseamento da habitação 1970

— estimativa a 20%, Lisbon, Sociedade Astória, Lda.
Instituto Nacional de Estatística, (1984), XII Recenseamento geral da população

e ii recenseamento geral da habitação 1981, Lisbon, INCM.
Diversos autores, (1993), Livro branco sobre a política de habitação em Portugal,

Santa Maria da Feira, Associações organizadoras do I Encontro Nacional da
Habitação, Raínho and Neves, Lda.
Instituto Nacional de Estatística, (1996), XIII recenseamento geral da popu‑

lação e iii recenseamento geral da habitação (CENSOS 91), Lisbon, Statistics
Portugal.
Conclusion
The contributions in this book provide insightful analyses of the real exuberant behaviour in Lisbon and Porto, seem to have had contagious
estate market in Portugal, both in terms of house prices and rents, up effects on surrounding housing markets.
until the beginning of 2020. The studies put forward analyse the real
The housing price synchronisation analysis in chapter 3, between
estate market from an aggregate country level perspective, as well
Portugal and other eleven European Union member states (Belgium,
as, from a more disaggregate perspective at the municipality and civil
Denmark, Finland, France, Germany, Ireland, Italy, Netherlands, Spain,
parish levels.
Sweden, and the United Kingdom), covers the period of 1988–2019,
The results of chapter 1 show that the real estate market contributes and reveals two primary cycles in the Portuguese House Price Index,
to one‑fifth of Portuguese GDP, and that its cycles have been strongly one about nine years long and the other with a duration of approxi‑
correlated with the Portuguese business cycles over the last two mately 14 years; these two cycles almost run the whole sample and
decades. Housing is a relevant asset in the balance sheet of Portuguese overlap. A shorter cycle, of close to four years, corresponding to the
households, which presents a composition similar to that of the international financial crisis, and the subsequent sovereign debt crisis
European Union's average, especially in 2017. However, there is a great (2008–2012), is also identified.
deal of variation between European countries, and Portugal belongs to
The analysis shows that the British housing market is the one most
a group in which a high percentage of households do not have liquid
synchronized with the Portuguese. Italy, Spain, and France come next.
assets, therefore being more vulnerable to aggregate shocks.
A more detailed analysis revealed that Spain is the country with which
In recent years, broad based exuberant behaviour is evident in Portugal has the largest regions of statistically significant coherency.
Portuguese house prices, although timely differences across municipa‑ Until early the 2000s, there is a region of high coherency between
lities can be observed. Chapter 2 provides an in‑depth analysis of this the 4 and 8 year frequencies; and after that, prominent coherencies at
behaviour and highlights that these recent dynamics can result from higher and lower frequencies are detected. Furthermore, it is found
a range of reasons other than speculative bubbles. For instance, local that Spain is the leader and Portugal the follower.
and foreign investment associated to tourism dwellings, foreign direct
Synchronicity has also been observed between Portuguese cities.
investment and shifts in preferences can explain some, but suppo‑
It is observed that Braga, Santarém, Portalegre, Leiria, and Setúbal are
sedly not all, of the exuberant behaviour observed over recent years in
quite de‑synchronized with the rest of the country (the last two form
the Portuguese real estate market. Moreover, it is also shown that the
a cluster of their own). Although the remaining cities seem relatively
well synchronized between themselves, some subclusters can be iden‑ One of the most important results of this chapter is the identification
tified; one is formed by Lisbon, Porto, and Faro (the three district of housing affordability challenges in territories located between the
capitals with international airports); and the other by Aveiro, Bragança, largest cities. On the one hand, regarding the coastline, we find small
Coimbra, and Guarda. and medium‑sized cities, which conflate urbanization, urban expansion
and suburbanization processes, and have been serving as alternatives
The two most important Portuguese cities, Lisbon and Porto, are to large urban centres (Lisbon and Porto), placing them in a challen‑
synchronized at longer run cycles, with Lisbon's cycle leading Porto's. ging position to achieve reasonable housing affordability levels. On the
However, at higher frequencies, the cycles diverge and the correla‑ other hand, among the inland regions, phenomena of consolidation in
tion is negative. Lisbon and Faro became very synchronized after 2005, small urban and rural centres, which may play a key role in countering
with Lisbon leading. Viana do Castelo and Beja are the most synchro‑ depopulation, also seem to face housing affordability issues that place
nized pair and their cycles are almost simultaneous. Leiria and Viseu them in similar positions to those of more socially and economically
are the least synchronized pair but display a common long‑run cycle. dynamic territories.
The results of chapter 3 suggest that the Portuguese housing market
is segmented and displays regional heterogeneities. The fact that the The analysis of chapter 5 focuses on the impact of short‑term rentals
housing cycles of Portuguese cities may be de‑synchronized, suggests on the real estate market. Taking advantage of a partial ban introduced
that housing policies should be designed locally. by the municipality of Lisbon in November 2018, which prohi‑
bited new short‑term rental licenses with immediate effect in some
Chapter 4 introduces two housing affordability metrics: i) the housing pre‑selected areas of the capital, revealed some interesting conclu‑
expenditure burden index and ii) the housing unaffordability risk sions. More specifically, it is shown that incumbent owners (mostly
index. The former assesses the ratio between estimated values of Portuguese) rushed to register properties just before the ban was
housing and household income, using civil parishes as the unit of effective. This shows that the anticipation of the policy, due to its
analysis, and the households living in subsections where the ratio is wide public discussion, allowed incumbent owners to enter the market
higher than 0.4; the latter analyses the proportion of households living in the very last weeks before licensing was interrupted.
in subsections with median income lower than the median of the
municipality, and housing cost higher than the median housing cost The results in this chapter document a sizeable decrease in the
in the municipality. As expected, results show strong spatial hetero‑ number and price of traded dwellings. More specifically, this chapter
geneity in housing affordability patterns and significant evidence of shows that the ban led to a 20% decrease in the quantity traded on
spatial dependence, thus claiming the importance of a territorial pers‑ the market and a 9% decrease in price. The market for 2‑bedroom
pective in the design of public policies that impact housing (which is in flats drives these effects, showing a segmentation of short‑term
line with the results of chapter 3). rentals in small to medium sized apartments. The price decrease for
two‑bedroom houses is almost 20%, which is consistent with a sharp The COVID‑19 pandemic has had negative impacts beyond the health
reduction in the demand for this type of flat, confirming that the sector and generated a worldwide crisis which is having consequences
option to register a property as short‑term rental is an important on basically all sectors of the economy, including real estate. Moreover,
driver of the search for new properties. This chapter also highli‑ the pandemic crisis has also led to the re‑organisation of activities in
ghts the fact that the surge of short‑term rental markets does create terms of urban spaces and mobility, as well as different usage of house
upward pressure on real estate prices. It also indicates that, despite spaces.
their salience, sharing economy platforms are not the full story behind
the spectacular housing market boom in the city of Lisbon. As was shown in the different chapters of this book, the trends of the
real estate market in the pre‑COVID era are clear, whereas we can only
Finally, chapter 6 addresses the important topic of urban leases in hypothesise about the post‑COVID period, at this moment. The real
Portugal and its complexities, and suggests that this is not a short‑term estate market potentially requires adjustments in order to meet
problem. It is structural and a consequence of several bottlenecks. needs, be useful to a society which is changing in its requirements and
One is the lack of adequate responses to housing shortages; the other guarantee its wellbeing.
is the «home for life» culture, which took root and became an anachro‑
nism, given the evolution of the labour market and the new demands This change in trend may be due to several factors related to the lock‑
for mobility; and finally the ideological prejudices, which created too down, including households’ economic instability and lower incomes,
many misunderstandings and conflicts between the right to housing alongside changes in lifestyle and working conditions. In addition, the
and the right to property, and today are the main cause for the loss of decrease in tourism, and therefore in related rents, has also strongly
investor confidence in political institutions and the enormous reputa‑ affected the real estate trend in Portugal. Overall, this sector is facing
tional damage that the lease market has suffered with the succession important challenges which will likely affect its short and long run
of legislative advances made by the latter. An interesting overview of dynamics.
renting legislation in Portugal is also provided, contributing towards a
better understanding of the dynamics and evolution of this important
dimension of the real estate market in Portugal.
A final note is in order, as this study began before the COVID‑19

pandemic, which originated a significant shock in the real estate
market, whose effects will depend on the duration of the pandemic
and the economy's recovery process.
Notes
1. The analyses, opinions and findings of this paper 10. Carried out by municipal councils, in the context of the 19. This information will be included as dummy
represent the views of the authors, and are not necessarily elaboration of local housing strategies, its goal is to identify and variables in the econometric model.
those of the Banco de Portugal or the Eurosystem. characterise the situation of households that meet the criteria
20. This information will be included as dummy
for access to the various housing solutions recommended by
2. For a more complete overview of different bubble models and variables in the econometric model.
the program; this document is the basis for access to public
how they might be tested econometrically, see Gürkaynak (2008).
(financial) support for resolving situations of housing needs. 21. Real estate supply data from real estate listing portals
3. See Appendix B for details and results of the quantile (e.g. CasaSapo, Imovirtual, Idealista), real estate networks
11. The first local housing strategies (ELH as described in
cointegration test and previously conducted unit root tests. or real estate agencies is difficult to compile. Not only are
footnote 1) were published and approved in 2020, meaning
there no application programmable interfaces (APIs) to
4. For the test statistic and limit distribution that these results are still based on a small sample, therefore
automatically access the data made available on the properties,
of the GSADF test, see Appendix A. the numbers can be expected to be even more significant.
but there is also no centralized registration mechanism
5. Note that Figure 22 only plots the results for the 2013– 12. Even if data on the housing market is exposed that allows the identification, at any given moment, of
2020 period, which does not exclude the possibility of other to many constraints and asymmetries. the property on offer, its data and (quotation) prices.
periods of exuberance having occurred before 2013.
13. More information on this metric can be seen here. 22. The first year for which this metric is presented.
6. We have used (Kruskal 1964a, 1964b)'s stress
14. The annual estimate is based on a weighted average 23. The tourism data comes from the open‑access platform
algorithm to minimize the square differences between the
of the quarterly data, using the number of contracts for available here, which was collected and processed by
distances in the map and the wavelet dissimilarities.
purchase and sale of urban buildings in each quarter as the authors — who cross‑referenced the data with the
7. Local housing strategies («Estratégias Locais de Habitação — a weight — that is, from the data available here. georeferenced database of considered administrative and
ELH») are a mandatory requirement for public and private entities, statistical political partitions (municipalities, civil parishes).
15. Other variable transformation features are the semi
both individual and collective, to access the programme «1.º Direito
‑log and, more generally, the possibilities established by 24. In most cases, these are municipalities that have
— Programa de Apoio ao Acesso à Habitação» (1st Right — Support
the Box‑Cox transformation. The Box‑Cox transformation previously occupied prominent positions in the political
Programme for Housing Access), which focuses on financial support
was introduced by Box G., and Cox D., (1964) as a way to administrative structure, namely with respect to former
for initiatives aiming to solve housing needs, although legislation is
solve the estimation problem in non‑linear regressions. districts/civil governors. They are also municipalities that house
clear that the strategy can and should have a more general nature.
many of the central administration's regional services.
16. The metric in question can be accessed here.
8. This refers to a direct intervention in the context of the housing
25. We thank Vitor Reis for his very careful reading of this chapter
market, which goes beyond the traditional public supply focused on 17. Housing cost € per year is estimated from the housing
and the interesting comments and Alex Coutts for the feedback
population segments that are excluded from the housing market. market value in 2016 (€/m2 total). The yearly value is
on an earlier version. We are grateful to Jacob Macdonald and the
obtained considering a projected payment period of
9. The focus on mainland Portugal arises from the Municipality of Lisbon for help with data sources and management.
30 years and an associated annual charge of 2%.
increasing difficulties in widening this study to the
26. link.
autonomous regions, which due to the insular characteristics 18. Retrieved from the urban area typology («tipologia de área
of their territory, call for a specialized study. urbana» — TIPAU) defined by Statistics Portugal at the civil parish 27. link.
level. For municipalities, this variable is defined considering the most
28. link.
frequent classification in civil parishes within the municipality.
Acesso rápido k Capa | Contents | Introduction | Chapter 1 | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | Chapter 6 | Conclusion | Notes

29. Franco et al. (2019) estimate the quarterly housing prices 43. link. 69. Article 5 of Law no. 828 of 28 September, 1917.
as a function of Airbnb concentration before and after the 2014
44. White book on Housing Policy in Portugal, ENH 1993. 70. Decree no. 22661 of 13 June, 1933.
regulatory reform. They quantify that a 1 p.p. increase in the
share of Airbnb properties increases house prices by 4.5%. 45. link. 71. Articles 1 and 3 of Decree no. 9496 of 14 March, 1924.
30. For more information about difference‑in 46. link. 72. Article 27 of Decree no. 15289 of 30 March, 1928.
‑differences see Angrist and Pischke (2009).
47. link. 73. Article 29 of Decree no. 15289 of 30 March, 1928.
31. To the best of our knowledge, the only paper in the
48. Decree of 12 November, 1910. 74. Article 10 of Decree no. 47344 of 25 November,
literature that exploits quasi‑experimental evidence from short
1966, introducing the Civil Code.
‑term rental restrictions is Koster et al. (2018). The authors 49. Decree no. 5411 of 17 April, 1919.
take advantage of a regulation that prevented landlords from 75. Decrees‑law no. 217/74 of 27 May and 445/74 of 12 September.
50. Law no. 2030 of 22 June, 1948.
short‑term renting any property besides their primary one
76. No. 2 of article 2 of Decree‑law no. 148/81 of 4 June.
in some regions of Los Angeles. They found that, due to this 51. Decree‑law no. 47344 of 25 November, 1966.
law, listings reduced by 50% and house prices by 3%. 77. No. 3 of article 2 and articles 3 and 4 of
52. Decree‑law no. 583/76 of 22 June.
Decree‑law no. 148/81 of 4 June.
32. link.
53. Law no. 46/85 of 20 September.
78. Article 7 of Decree‑law no. 148/81 of 4 June.
33. The thresholds below are from 2019. While the parameters may
54. Decree‑law no. 321‑B/90 of 15 October.
vary from year to year, the overall taxation scheme is the same. 79. Article 14 of Decree‑law no. 148/81 of 4 June.
55. Decree‑law no. 257/95 of 30 September.
34. link. 80. Law no. 46/85 of 20 September.
56. Law no. 6/2006 of 27 February.
35. link. 81. Article 11 of Law no. 46/85 of 20 September.
57. Law no. 79/2014 of 19 December.
36. link. 82. Decree‑law no. 13/86 of 23 January.
58. Law no. 42/2017 of 14 June.
37. link. 83. Decree‑law no. 68/86 of 27 March.
59. Law no. 64/2018 of 29 October.
38. link. 84. No. 3 of article 8 of Law no. 46/85 of 20 September.
39. link. 85. Law no. 6/2006 of 27 February.
40. Empirical evidence for Lisbon suggests that cultural 86. Article 4 of Decree de 12 November, 1910.
heritage and access to green spaces are capitalized in real 62. Law no. 2030 of 22 June, 1948.
87. Article 12 of Decree de 12 November, 1910.
estate prices (Franco and Macdonald 2018 a,b). Gonçalves
et al. (2020) show that differences between treatment and 63. Decrees‑law no. 217/74 of 27 May and 445/74 of 12 September.
88. § 4 of Article 70 of Decree no. 5411 of 17 April, 1919.
comparison areas are also remarkably low in these two regards. 64. Article 9 of Decree of 12 November, 1910.
89. Article 8 of Law no. 828 of 28 September, 1917.
41. Treated neighbourhoods are Madragoa, Bairro 65. Article 1 of Decree no. 1079 of 23 November, 1914.
Alto, Bica, Príncipe Real, Santa Catarina, São Paulo/ 90. Article 30 of Decree no. 5411 of 17 April, 1919.
Boavista/ Conde Barão, Alfama, Mouraria, and Sé. 66. No. 1 of article 2 of Law no. 828 of 28 September, 1917.
91. No. 3 of article 1 of Law no. 1662 of 4 September, 1924.
42. The five treated neighbourhoods are Madragoa, 67. No. 4 of article 2 of Law no. 828 of 28 September, 1917.
92. Decree no. 15289 of 30 March.
Bairro Alto, Santa Catarina, Alfama, and Mouraria. 68. Article 106 of Decree no. 5411 of 17 April, 1919.
93. No. 1 and 2 of article 46 of Law no. 2030,of 22 June, 1948.

94. No. 3 of article 46 of Law no. 2030 of 22 June, 1948. 112. § 7 of article 12 of Decree no. 4499 of 27 June, 1918. 132. No. 1 of article 1041 of the Civil Code
modified by Law no. 13/2019 of 12 February.
95. Article 1095 of the Civil Code. 113. Article 82 of Decree no. 5411 of 17 April, 1919.
133. Articles 10, 15, 16 and 26 of Decree‑law no. 293/77 of 20 July.
96. Article 98 of RAU approved by Decree 114. § 9 of article 5 of Law no. 1662 of 4 September, 1924.
‑law no. 321‑B/90 of 15 October. 134. Law no. 46/85 of 20 September.
115. Article 986 et seq. of Decree no. 29637 of 28 May, 1939.
97. Article 100 of RAU approved by Decree 135. Decree‑law no. 162/92 of 5 August.
116. Article 1053 of the Civil Code approved by
‑law no. 321‑B/90 of 15 October.
Decree‑law no. 47344 of 25 November, 1966. 136. Decree‑law no. 308/2007 of 3 September.
98. Article 40 of Law no. 46/85 of 20 September
117. Article 1097 of the Civil Code approved by 137. Decree‑law no. 158/2006 of 8 August.
amending article 1111 of Civil Code.
Decree‑law no. 47344 of 25 November, 1966.
138. Articles 3 to 5 of Decree no. 1079 of 23 November 1914.
99. Article 86 of RAU approved by Decree
118. Decrees‑law no. 6/75 of 7 January and 232/75 of 16 May.
‑law no. 321‑B/90 of 15 October. 139. Article 108 of Decree no. 5411 of 17 April 1919.
119. Decrees‑law no. 198‑A/75 of 14 April and 294/77 of 20 July.
100. Article 1106 of the Civil Code modified 140. Article 108 of Decree no. 5411 of 17 April 1919.
by Law no. 6/2006 of 27 February. 120. Articles 15 and 15‑A of Law no. 31/2012 of 14 August.
141. Article 17 of Decree‑law no. 445/74 of 12 September.
101. Article 1026 of the Civil Code modified 121. link.
142. No. 7 of article 15 of RAU approved by Decree
by Law no. 31/2012 of 14 August.
122. No. 3 of article 1083 of the Civil Code ‑law no. 321‑B/90 of 15 October, modified by Decree
102. Article 1097 of the Civil Code modified modified by Law no. 31/2012 of 14 August. ‑law no. 329‑B/2000 of 22 December.
by Law no. 31/2012 of 14 August.
123. No. 4 of article 1083 of the Civil Code 143. Article 37 of Law no. 55‑B/2004 of 30 December
103. Article 35 of Law no. 6/2006 of 27 February, modified by Law no. 31/2012 of 14 August. modifying no. 8 and 9 of article 112 of the IMI Code.
modified by Law no. 43/2017 of 14 July.
124. § single of article 3 and § 1 of article 5 144. Article 7 da Law no. 6/2006 of 27 February
104. Article 35 of Law no. 6/2006 of 27 February, of Decree of 12 November, 1910. amending no. 3 of article 112 of the IMI Code.
modified by Law no. 2/2020 of 31 March.
125. § single of article 37 of Decree no. 5411 of 17 April, 1919. 145. Article 93 da Law no. 64‑A/2008 of 31 December
105. No. 2 of article 1095 of the Civil Code, amending no. 3 of article 112 of IMI Code.
126. Article 1091 of the Civil Code approved by
Decree‑law no. 47344 of 25 November, 1966. 146. Article 141 da Law no. 64‑B/2011 of 30 December
106. Articles 1096 and 1097 of the Civil Code, amending no. 3 of article 112 of IMI Code.
127. Article 1076 of the Civil Code modified
by Law no. 6/2006 of 27 February. 147. Article 11 da Law no. 51/2018 of 16 August
107. Paragraph c) of article 1101 of the Civil Code, amending no. 3of article 112 of IMI Code.
128. § 1, a) of article 5 of Law no. 1662 of 4 September, 1924.
148. Article 10 da Law no. 119/2019 of 18 September
129. § single of article 2 of Decree no. 22661 of 13 June, 1933.
108. No. 10 of article 36 of Law no. 6/2006 of 27 February amending no. 3 of article 112 of IMI Code.
modified by Law no. 13/2019 of 12 February. 130. Article 1041 of the Civil Code approved by
149. Article 5 of Decree‑law no. 67/2019 of
Decree no. 47344 of 25 November, 1966.
109. No. 3 of article 14 of Law no. 13/2019 of 12 February. 21 adding article 112‑B to IMI Code.
131. Article 1041 of the Civil Code modified
110. Articles 11, 13, 14 and 18 of Decree of 12 November, 1910. 150. Decree‑law no. 28912 of 12 August 1938.
by Decree‑law no. 293/77 of 20 July.
111. No. 5 of article 2 of Law no. 828 of 28 September, 1917. 151. Decree‑law no. 23052 of 23 September 1933.

152. Decree no. 4137 of 25 April 1918. 174. Technical rate: designates the monthly amount supported
by the household, reflecting the real cost of the housing unit,
153. Articles 11 and 12 of Decree‑law no. 28912 of 12 August 1938.
determined by its typology, year of construction, gross area, age,
154. Decree‑law no. 33278 of 24 November 1943. and state of preservation. The value is also calculated taking the
average cost of construction and the cost of urbanised land into
155. Decree‑law no. 37576 of 10 October 1949.
account, and is only applied in cases where the household has an
156. Article 3 of Decree‑law no. 34486 of 6 June 1945. overall monthly income of more than three times the National
Minimum Wage rate in force (definition adapted from here).
157. Decree‑law no. 35732 of 4 July 1946.
175. Decree order no. 288/83 of 17 March.
158. Decree‑law no. 41470 of 23 December 1957.
176. No. 24 da Decree order no. 288/83 of 17 March.
159. Law no. 2007 of 7 May 1945.
177. Decree‑law no. 166/93, of 7 May.
160. Law no. 2092 of 9 April 1958.
161. Decree‑law no. 36212 of 7 April 1947.
179. Law no. 32/2016 of 24 August.
162. Decree‑law no. 41532 of 18 February 1958.
180. link.
163. Articles 12 and 13 of Decree‑law no. 41532 of 18 February 1958.
181. Article 111 of Decree no. 5411 of 17 April 1919.
164. Decree order no. 759/74 of 23 November.
182. Decree‑law no. 36212 of 7 April 1947.
165. Decree‑law no. 518/77 of 15 December.
183. Article 30 da Law no. 2030 of 22 June 1948.
166. Decree‑law no. 148/81 of 4 June.
184. Decree‑law no. 328‑B/86 of 30 September.
167. Articles 7 and 28 of Law no. 46/85 of 20 September.
185. Resolution of the Council of Ministers of 24
168. Decree‑law no. 321‑B/90 of 15 October.
February 1976, published on 19 March, 1976.
169. No. 3 of article 7 of Decree‑law no. 321‑B/90 of 15 October.
186. Source IGAPHE/IHRU: Program IAJ.
187. Source Statistics Portugal: Census 2001.
171. Decree‑law no. 68/2019 of 22 May.
188. Residential Information System — lease in Lisbon
172. Decree order no. 176/2019 of 6 June. and Porto — quarterly data from 2010 to 2019.
173. Decree order no. 386/77 of 25 June. 189. Article 4 of Decree‑law no. 268/94 of 25 October.
190. Article 8 of Decree‑law no. 37/2018 of 4 June and
article 2 of Decree order no. 311‑D/2011 of 27 December.

Index of figures
11 Figure 1Contribution of the real 32 Figure 16Average value of a residential 41 Figure 28Time varying contagion
estate market to Portugal's GDP dwelling in Lisbon and Porto coefficient from the Porto housing market
12 Figure 2Contribution of the real estate 32 Figure 17Number of total residential to civil parishes in geographic proximity
market to GDP in the European Union dwellings transactions by region 42 Figure 29Real house prices, price‑to‑rent
13 Figure 3GDP and investment cycles in Portugal 33 Figure 18Economic activity and price‑to‑income ratio for Portugal
14 Figure 4Real house prices 33 Figure 19Labour market 43 Figure 30Housing supply
15 Figure 5Real lease prices 33 Figure 20Credit market 44 Figure 31Bank appraisals vs. market prices
16 Figure 6Real lease prices by country 35 Figure 21House prices and estimated 55 Figure 32Time series plot of House Price Index
17 Figure 7Monthly instalment by NUTS III fundamental quantiles (on top). Wavelet power spectrum (at the bottom).
19 Figure 8Monthly instalment 37 Figure 22Evidence of exuberant behaviour 57 Figure 33Cross‑wavelet coherency
20 Figure 9Effort rate in Portuguese districts over time between Portugal and other countries (on
21 Figure 10Amount outstanding 38 Figure 23Evidence of exuberant the left). Phase‑difference (on the right).
23 Figure 11Evolution of ‘hand‑to‑mouth’ behaviour in the Lisbon area over time 59 Figure 34Multidimensional scaling map
households in Portugal and Europe 38 Figure 24Evidence of exuberant 59 Figure 35Hierarchical Tree Clusters
24 Figure 12Net assets and net wealth behaviour in the Porto area over time 60 Figure 36Cross‑Wavelet Coherency
of European households 39 Figure 25Test results for exuberant between Portuguese cities (on the left);
31 Figure 13House prices and bank behaviour in Lisbon Phase‑Difference (on the right).
appraisals in Portugal 39 Figure 26Test results for 76 Figure 37Demographic evolution 2001–2011
32 Figure 14Number of residential exuberant behaviour in Porto 76 Figure 38Demographic evolution
dwelling transactions in Portugal 41 Figure 27Time varying contagion 2011–2015 (municipalities)
32 Figure 15Average value of a coefficient from the Lisbon housing market 76 Figure 39Average age of residential
residential dwelling in Portugal to civil parishes in geographic proximity housing units (2011)
76 Figure 40Elderly dependence figures

78 Figure 41Processes of urban transformation 109 Figure 55Airport arrivals in 126 Figure 70Evolution of conventional dwellings,
in (mainland) Portugal's civil parishes Portugal, Lisbon and Porto households and residents in Portugal (Unit: millions)
79 Figure 42Trajectory of housing prices and income 110 Figure 56Short‑term rental registries (in Portugal) 127 Figure 71Evolution of forms and
91 Figure 43Housing cost (median) based on 110 Figure 57Short‑term rental registries (in Lisbon) regimes of dwelling occupancy
(median) market prices and (median) dwelling 111 Figure 58Short‑term rental registries (in Porto) 128 Figure 72Evolution of conventional dwellings,
characteristics. Municipalities. 2016 111 Figure 59Share of Airbnb listings by households and resident population in Lisbon
91 Figure 44Housing cost (median) based on owners with multiple listings (in Lisbon) 128 Figure 73Evolution of forms and regimes
(median) market prices and (median) dwelling 112 Figure 60Median real estate price per of dwelling occupancy in Lisbon
characteristics. Civil parishes. 2016 m2 in Portugal and selected cities 129 Figure 74Evolution of conventional dwellings,
91 Figure 45LISA clusters on the estimated 112 Figure 61Growth rate of the median households and resident population in Porto
housing costs. Civil parishes in 2016 sale price in Lisbon — 2016 to 2018 130 Figure 75Evolution of forms and regimes
95 Figure 46Estimated income based on (median) 113 Figure 62Growth rate of the median of dwelling occupancy in Porto
income declared to tax authorities and (median) rent (per m2) for new housing rental 131 Figure 76Number of rental housing
family characteristics. Municipalities. 2016 agreements in Lisbon, during 2018 units supplied per year
95 Figure 47Estimated income based on (median) 113 Figure 63Growth rate of median 132 Figure 77Number of new housing
income declared to tax authorities and (median) rent (per m2) for new housing rental rental contracts per year
family characteristics. Civil parishes. 2016 agreements in Lisbon, during 2019 132 Figure 78Number of new
95 Figure 48LISA clusters by estimated income. 115 Figure 64Analysis timeline rental contracts per year
96 Figure 49Housing cost burden index 115 Figure 65Treated areas and control groups 133 Figure 79Weight of forms and
96 Figure 50LISA clusters of 118 Figure 66Number of short‑term registries occupancy regimes in 2011
housing cost burden index 120 Figure 67Number of house transactions, 142 Figure 80Evolution of spending
96 Figure 51Housing unaffordability risk by number of bedrooms in rent subsidy programmes
96 Figure 52Bivariate LISA clusters 120 Figure 68Number of house 149 Figure 81Index of residential rents (2011=100)
for costs and income transactions, by energetic efficiency 150 Figure 82State efforts for housing 1987–2011
99 Figure 53Housing cost burden index 122 Figure 69The demand driven
100 Figure 54Housing affordability risk index impact in the real estate market
Index of tables
14 Table 1Matrix correlation between GDP, 90 Table 11Main results from the
non‑residential and residential investment housing costs model
16 Table 2Additional property and lease 93 Table 12Main results from the
18 Table 3Financing of primary residence explanatory factors of income
22 Table 4Percentage of ‘hand‑to‑mouth’ households 118 Table 13Difference‑in‑Differences
37 Table 5GSADF test results for results — Ln (registries)
Portugal's 18 districts 119 Table 14Difference‑in‑Differences
43 Table 6Number of firms associated to results — Ln (Airbnb prices)
furnished short‑stay dwellings for tourists 121 Table 15Difference‑in‑Differences
48 Table B. 1Test results of ADFGLS results — Ln (number of transactions)
and Phillips‑Perron 121 Table 16Difference‑in‑Differences
56 Table 7De‑synchronisation between results — Ln (prices)
Portugal and other countries
79 Table 8Residential housing and housing
deprivation metrics in Portugal
87 Table 9Modelling data for housing costs
88 Table 10Modelling data for income
Authors
Aguiar‑Conraria, Luis Brinca, Pedro Gonçalves, Duarte
Holds a PhD in Economics from Cornell University, is a Professor Holds a PhD in Economics from Stockholm University. Prior to Holds an MA in Economics from Nova School of Business
of Economics at the University of Minho and dedicates himself joining Nova School of Business and Economics as an Assistant and Economics. He is also a Teaching Assistant at that same
to Macroeconomics and Political Economics. In 2011, he was Professor in 2015, he worked for the European University institution, having collaborated in courses such as Public
awarded the Gulbenkian Prize for the Internationalization of Institute, Stockholm University, the European Central Bank and Economics, Advanced Microeconomics and Industrial Economics.
Social Sciences. He is currently the Director of the Department the University of Minnesota. His research essentially focuses He was an intern economist at the Competition Authority (AdC).
of Economics at University of Minho. He is also a member of on two topics: economic cycles and the interaction between
the Economic Cycle Dating Committee and a columnist for the micro‑heterogeneity and macroeconomic dynamics. He is also an Huget, René
weekly newspaper Expresso. Associate Researcher at the University of Oslo. Currently works as a freelance consultant and is a senior
associate in strategic consulting. His research interests focus
Barros, Vera Gouveia Castro, Eduardo Anselmo on Applied Macroeconomics, Asset pricing and Behavioural
Is an Economist, with a BSc from the Nova School of Business Has been a Professor at the University of Aveiro since 1982, Economics. He holds a BA in Economics from the University
and Economics.She holds an MA in International Economics and having been Vice‑Rector for Cooperation with Society and of Cologne and an MA in Economics from the Nova School of
European Studies from the ISEG Lisbon School of Economics Technology Transfer of the institution. Currently Vice‑President Business and Economics.
& Management, where she is currently completing her PhD in of the Centro Regional Coordination and Development
Economics, with a thesis on Tourism Economics, a field in which Commission (CCDR‑C), he has coordinated dozens of national Lourenço, Rita Fradique
she has developed research, alongside Housing Economics. She and international research projects in domains related to regional Has been an Economist at the Economics and Research
was also a faculty member at the University of Madeira. development. Department of the Banco de Portugal since 2001, having joined
the institution's staff about a decade earlier as a member of
Batista, Paulo Duarte, João B. the Markets Department. She holds a BSc degree in Economics
Holds a PhD in e‑Planning from the University of Aveiro (UA). Holds a PhD in Economics from the University of Illinois from the Nova School of Business and Economics and has
He is a Researcher in the Department of Social, Political and Urbana‑Champaign. Prior to joining Nova School of Business completed the curricular part of the PhD and MA Programme in
Territorial Sciences (UA) and a member of the Research Unit and Economics in 2017 as an Assistant Professor, he worked at Economics from the same University. In recent years, her research
on Governance, Competitiveness and Public Policies. He has Cambridge University. His research has mainly focused on the has included the dynamics of housing markets, with several
developed research and applied work on methods and techniques housing market and monetary policy. He has been published published work in co‑authorship with Paulo M. M. Rodrigues.
for territorial analysis. in the Journal of the European Economic Association and in
the Journal of Applied Econometrics.

Marques, João Lourenço Rodrigues, Paulo M. M.
Is a Professor at the Department of Social, Political and Is a Consultant Researcher at the Economics and Research
Territorial Sciences of the University of Aveiro, teaching and Department of the Banco de Portugal and Adjunct Professor at
researching in the areas of Regional Economics, Regional the Nova School of Business and Economics. He holds a PhD in
Planning and Public Policies, namely, methods and techniques to Econometrics from the University of Manchester and develops
support decision‑making. He is a member of the Research Unit on research in Theoretical and Applied Econometrics to Finance
Governance, Competitiveness and Public Policies, coordinating and Economics. He is a member of the editorial board of several
the Systems for Decision Support Research Group. international scientific journals and an external research fellow
of the Essex Centre for Financial Econometrics at the University
Peralta, Susana of Essex, and the Clive Granger Centre for Time Series Analysis
Is an Associate Professor, with agregação, at the Nova School of at the University of Nottingham.
Business and Economics, and holds a PhD in Economics from the
Université Catholique de Louvain, in Belgium. She is a specialist Santos, João Pereira dos
in Public Economics and has research published in international Holds a PhD in Economics from the Nova School of Business and
journals, including the Journal of Public Economics, Journal of Economics. He was a visiting student at the Harvard Kennedy
Urban Economics, The Economic Journal or Public Choice. School and completed the Advanced Studies Program at the Kiel
Institute for the World Economy. His research focuses on the
Reis, Victor evaluation of public policies. He currently works as a technical
Holds a BA in Architecture and is a Senior Technical Advisor advisor for the cabinet of the Assistant Secretary of State to the
at the Institute for Housing and Urban Rehabilitation (IHRU), Prime Minister.
having assumed the presidency of the institution between
2012 and 2017. During his professional career he was involved Soares, Maria Joana
in the kick‑off of urban rehabilitation operations in Alfama Is an Associate Professor in the Department of Mathematics of
and Mouraria, in the creation of RECRIA, IAJ and PER, in the the University of Minho and a member of NIPE — Centre for
implementation of the 2012 urban lease reform and in the study Research in Economics and Management. Her main research
on the 100 years of housing policies. interests lie in the field of Numerical Analysis and in the
applications of Wavelet Theory to Economics, having developed
a Matlab toolbox for the use of various wavelet tools, in
collaboration with Luís Aguiar‑Conraria.

Francisco Manuel dos Santos Foundation
Published studies
Economics Institutions
O Cadastro e a Propriedade Diversificação e Crescimento Droga e Propinas: Avaliações Justiça Económica em Portugal: Juízes na Europa: Formação, selecção,
Rústica em Portugal da Economia Portuguesa de impacto legislativo Produção de prova promoção e avaliação
Coordination by Rodrigo Coordination by Leonor Sopas; 2018. Coordination by Ricardo Coordination by Mariana Coordination by Carlos
Sarmento de Beires; 2013. Dinâmica Empresarial e Desigualdade Gonçalves; 2012. França Gouveia, Nuno Garoupa, Gómez Ligüerre; 2015.
Custos e Preços na Saúde: Coordination by Rui Baptista; 2018. Pedro Magalhães; 2012. Limitação de Mandatos: O impacto
Justiça Económica em Portugal:
Passado, presente e futuro Encerramento de Multinacionais: A citação do réu no processo civil Justiça Económica em Portugal: nas finanças locais e na participação
Coordination by Carlos Costa; 2013. O capital que fica Coordination by Mariana Recuperação do IVA eleitoral
25 anos de Portugal Europeu: Coordination by Pedro de Faria; 2018. França Gouveia, Nuno Garoupa, Coordination by Mariana Coordination by Francisco
A economia, a sociedade e os fundos Pedro Magalhães; 2012. França Gouveia, Nuno Garoupa, Veiga and Linda Veiga; 2017.
GDP‑linked bonds in the Pedro Magalhães; 2012.
estruturais Portuguese Economy Justiça Económica em Portugal: O Estado por Dentro: Uma etnografia
Coordination by Augusto Coordination by Gonçalo Pina, 2020. Factos e números Justiça Económica em Portugal: do poder e da administração pública
Mateus; 2013. Coordination by Mariana Síntese e propostas em Portugal
Features by Portuguese International França Gouveia, Nuno Garoupa, Coordination by Mariana Coordination by Daniel
Que economia queremos?
Trade: a Firm‑level Perspective Pedro Magalhães; 2012. França Gouveia, Nuno Garoupa, Seabra Lopes; 2017.
Coordination by João Ferrão; 2014.
Coordination by João Amador; 2020. Pedro Magalhães; 2012.
A Economia do Futuro: A visão Justiça Económica em Portugal: O Impacto Económico dos Fundos
Financial Constraints and Business Gestão processual e oralidade Segredo de Justiça Europeus: A experiência
de cidadãos, empresários e autarcas
Dynamics: Lessons from the 2008 Coordination by Mariana Coordination by Fernando dos municípios portugueses
Coordination by João Ferrão; 2014.
‑2013 Recession França Gouveia, Nuno Garoupa, Gascón Inchausti; 2013. Coordination by José Tavares; 2017.
Três Décadas de Portugal Europeu: Coordination by Carlos Carreira, Pedro Magalhães; 2012.
Balanço e perspectivas Feitura das Leis: Portugal e a Europa Orçamento, Economia e Democracia:
Paulino Teixeira, Ernesto Nieto
Coordination by Augusto Justiça Económica em Portugal: Meios Coordination by João Caupers, Uma proposta de arquitetura
‑Carrillo and João Eira; 2021.
Mateus; 2015. de resolução alternativa de litígios Marta Tavares de Almeida institucional
Transport systems in Portugal Analysis Coordination by Mariana and Pierre Guibentif; 2014. Coordination by Abel M. Mateus; 2018.
Empresas Privadas e Municípios: by efficiency and regional impact França Gouveia, Nuno Garoupa,
Dinâmicas e desempenhos Portugal nas Decisões Europeias Instituições e Qualidade
Carlos Oliveira Cruz, Álvaro Pedro Magalhães; 2012.
Coordination by José Tavares; 2016. Coordination by Alexander da Democracia: Cultura política
Costa, Joaquim Miranda Sarmento,
Justiça Económica em Portugal: Trechsel, Richard Rose; 2014. na Europa do Sul
Investimento em Infra‑Estruturas Vítor Faria e Sousa and João
Novo modelo processual Valores, Qualidade Institucional Coordination by Tiago Fernandes; 2019.
em Portugal Fragoso Januário, 2021.
Coordination by Mariana e Desenvolvimento em Portugal Os Tribunais e a Crise Económica
Coordination by Alfredo From made in to created in: A new França Gouveia, Nuno Garoupa,
Marvão Pereira; 2016. Coordination by Alejandro Portes e Financeira: Uma análise ao processo
paradigm for the portuguese economy Pedro Magalhães; 2012. and M. Margarida Marques; 2015. decisório em contexto de crise
Benefícios do Ensino Superior Coordination by Fernando
Justiça Económica em Portugal: O Ministério Público na Europa económico‑financeira
Coordination by Hugo Figueiredo Alexandre; 2021.
O sistema judiciário Coordination by José Martín Patrícia André, Teresa Violante
and Miguel Portela; 2017. The real estate market in Portugal: Coordination by Mariana and Maria Inês Gameiro; 2019.
Pastor, Pedro Garcia Marques
Prices, rents, tourism and accessibility França Gouveia, Nuno Garoupa, and Luís Eloy Azevedo; 2015.
Coordination by Pedro Magalhães; 2012.
Paulo M. M. Rodrigues; 2022.
Society
Como se aprende a ler? Inquérito à Fecundidade 2013 O Multimédia no Ensino das Ciências Porque melhoraram os resultados Os jovens em Portugal, hoje: Quem
Coordination by Isabel Leite; 2010. Statistics Portugal and FFMS; 2014. Coordination by João Paiva; 2015. do PISA em Portugal? são, que hábitos têm, o que pensam
A Ciência na Educação Pré‑Escolar O Quinto Compromisso: Estudo longitudinal e comparado e o que sentem
Fazer contas ensina a pensar?
Coordination by Maria Lúcia Desenvolvimento de um sistema (2000–2015) Coordination by Laura Sagnier
Coordination by António Bivar; 2010.
Santos, Maria Filomena Gaspar, de garantia de desempenho Coordination by Anália Torres; 2018. and Alex Morell; 2021.
Desigualdade Económica em Portugal
Sbyia Saraiva Santos; 2014. educativo em Portugal Igualdade de Género ao Longo da Vida:
Coordination by Carlos Farinha
Dinâmicas Demográficas Coordination by Margaret Portugal no contexto europeu
Rodrigues; 2012.
e Envelhecimento da População E. Raymond; 2015. Coordination by Anália Torres; 2018.
Projecções 2030 e o Futuro
Portuguesa (1950–2011): Desigualdade do Rendimento As mulheres em Portugal, Hoje:
Coordination by Maria
Evolução e perspectivas e Pobreza em Portugal: Quem são, o que pensam e como
Filomena Mendes and Maria
Coordination by Mário As consequências sociais do programa se sentem
João Valente Rosa; 2012.
Leston Bandeira; 2014. de ajustamento Coordination by Laura Sagnier
Envelhecimento Activo em Portugal: Coordination by Carlos and Alex Morell; 2019.
Ensino da Leitura no 1.º Ciclo
Trabalho, reforma, lazer e redes sociais Farinha Rodrigues; 2016.
do Ensino Básico: Crenças, Financial and Social Sustainability
Coordination by Manuel
conhecimentos e formação Determinantes da Fecundidade by the Portuguese Pension System
Villaverde Cabral; 2013.
dos prbyessores em Portugal Coordination by Amílcar
Escolas para o Século xxi: Liberdade Coordination by João A. Lopes; 2014. Coordination by Maria Moreira; 2019.
e autonomia na educação Filomena Mendes; 2016.
Ciência e Tecnologia em Portugal: Identidades Religiosas e Dinâmica
Coordination by Alexandre
Métricas e impacto (1995–2012) Será a repetição de ano benéfica Social na Área Metropolitana
Homem Cristo; 2013.
Coordination by Armando Vieira para os alunos? de Lisboa
Informação e Saúde and Carlos Fiolhais; 2014. Coordination by Luís Coordination by Alfredo
Coordination by Rita Espanha; 2013. Catela Nunes; 2016. Teixeira; 2019.
Mortalidade Infantil em Portugal:
Literatura e Ensino do Português Evolução dos indicadores e factores Justiça entre Gerações: Perspectivas A evolução da ciência em Portugal
Coordination by José associados de 1988 a 2008 interdisciplinares (1987‑2016)
Cardoso Bernardes and Rui Coordination by Xavier Barreto Coordination by Jorge Pereira da Silva Elizabeth Vieira, João Mesquita, Jorge
Afonso Mateus; 2013. and José Pedro Correia; 2014. and Gonçalo Almeida Ribeiro; 2017. Silva, Raquel Vasconcelos, Joana Torres,
Processos de Envelhecimento Os Tempos na Escola: Migrações e Sustentabilidade Sylwia Bugla, Fernando Silva, Ester
em Portugal: Usos do tempo, Estudo comparativo da carga horária Demográfica: Perspectivas Serrão and Nuno Ferrand; 2019.
redes sociais e condições de vida em Portugal e noutros países de evolução da sociedade e economia A pobreza em Portugal:
Coordination by Manuel Coordination by Maria Isabel Festas; portuguesas Trajetos e quotidianos
Villaverde Cabral; 2013. 2014. Coordination by João Peixoto; 2017. Coordination by Fernando
Que ciência se aprende na escola? Cultura Científica em Portugal Mobilidade Social em Portugal Diogo, 2021.
Coordination by Margarida Coordination by António Granado Coordination by Teresa
Afonso; 2013. and José Vítor Malheiros; 2015. Bago d’Uva; 2017.

The Real Estate Market in Portugal: Prices, Rents, Tourism and Accessibility

Uploaded by

Copyright:

Available Formats

The Real Estate Market in Portugal: Prices, Rents, Tourism and Accessibility

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Real Estate Market in Portugal: Prices, Rents, Tourism and Accessibility

Uploaded by

Copyright:

Available Formats

estudos da fundação k 2022

Paulo M. M. Rodrigues, coord.

Paulo M. M. Rodrigues, coordination

Acesso rápido k Capa | Contents | Introduction | Chapter 1 | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | Chapter 6 | Conclusion | Notes /1

Director of publications: António Araújo

© Francisco Manuel dos Santos Foundation

Acesso rápido k Capa | Contents | Introduction | Chapter 1 | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | Chapter 6 | Conclusion | Notes /2

Acesso rápido k Capa | Contents | Introduction | Chapter 1 | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | Chapter 6 | Conclusion | Notes /3

Acesso rápido k Capa | Contents | Introduction | Chapter 1 | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | Chapter 6 | Conclusion | Notes /4

Acesso rápido k Capa | Contents | Introduction | Chapter 1 | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | Chapter 6 | Conclusion | Notes /5

Acesso rápido k Capa | Contents | Introduction | Chapter 1 | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | Chapter 6 | Conclusion | Notes /6

Acesso rápido k Capa | Contents | Introduction | Chapter 1 | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | Chapter 6 | Conclusion | Notes /7

Crowe, C., G. Dell’Ariccia, D. Igan and P. Rabanal, (2013), «How to

Helbling, T. and M. Terrones (2003), «Real and Financial Effects of

Rapach, D. and J.K. Strauss, (2006), «The Long­‑Run Relationship Between

Tsatsaronis, K. and H. Zhu, (2004), «What drives Housing Price Dynamics:

Acesso rápido k Capa | Contents | Introduction | Chapter 1 | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | Chapter 6 | Conclusion | Notes /8

Pedro Brinca and João B. Duarte

Acesso rápido k Capa | Contents | Introduction | Chapter 1 | Chapter 2 | Chapter 3 | Chapter 4 | Chapter 5 | Chapter 6 | Conclusion | Notes /9

This led to a third generation of models that explored ways in which

improved upon the empirical plausibility of marginal propensities to

This is of special relevance for Portugal too as housing is typically 5

and by far the most important asset in Portuguese household port‑

major part of this contribution was made by residential investment,

imputed rents, responsible for roughly 5%. However, by early 2000s,

increase in the contribution of rents, there was also a decrease in the

Yt Yt−1 It It−1 Ht Ht−1 Ht−2 Ht−3 1.15

It 1 0.71 0.57 0.62 0.66 0.67

1.4.1. Price and renting dynamics Source: Eurostat

Regarding the assessment of price developments in the real estate

As we explained in the beginning of this chapter, while comparing the

0 Source: HFCS and authors’ calculations

Looking at the regional segmentation of aggregate data, should we 1500

Prestação média mensal (EUR)

as we can see in figure 7. It is possible to draw a connection line 1000

Alentejo Beiras e Serra da Estrela Autonomous Region of Madeira

1.5. The importance of real estate

We now move on to the description of the importance of real

Europe interchangeably). Amount outstanding 61 792 EUR 101 147 EUR

We start with financing and indebtedness in the acquisition (and

Figure 8Monthly instalment

.004 .004 .004

.003 .003 .003

.001 .001 .001

Portugal European Union Portugal European Union Portugal European Union

Source: HFCS and authors’ calculations

Figure 9Effort rate

Portugal European Union Portugal European Union Portugal European Union

Source: HFCS and authors’ calculations

Figure 10Amount outstanding

1.0e-05 1.0e-05 1.0e-05

8.0e-06 8.0e-06 8.0e-06

4.0e-06 4.0e-06 4.0e-06

2.0e-06 2.0e-06 2.0e-06

Portugal European Union Portugal European Union Portugal European Union

Source: HFCS and authors’ calculations

2010 2013 2017

SI 37.09 6.96 37.26 11 34.56 8.43 25

SK 24.32 3.22 28.73 5.61 30.53 3.92

Rapach, D. and J.K. Strauss, (2006), «The Long‑Run Relationship Between

With around 25% of ‘hand‑to‑mouth’ households, Portugal has not

Table 6Number of firms associated to furnished short‑stay dwellings 3000 12000

Short‑stay accommodation 2112

Poterba, J.M., (1984), «Tax subsidies to owner‑occupied housing:

Vera Gouveia Barros, Luís Aguiar‑Conraria, Maria Joana Soares

Lisbon is by far the most sizeable short‑term rental market in the