Sovereign Default and Tax-smoothing in the Shadow of

Corruption and Institutional Weakness*

Marina Azzimonti† Nirvana Mitra‡
FRB of Richmond CAFRAL & Shiv Nadar IoE

This version: November 2023

Abstract

Emerging countries exhibit volatile fiscal policies and frequent sovereign debt crises, that
significantly diminish the well-being of their citizens. International advisors typically suggest
developed-world solutions as a remedy. We argue that the root of the problem lies in the
institutional environment, which does not incentivize responsible policymaking, particularly
tax-smoothing practices. Focusing on democratic representation and control of corruption,
our dynamic political-economy bargaining model shows that nations with weaker institutions
experience frequent default episodes and greater economic volatility. Our results are in line
with stylized facts from a panel of 58 countries between 1990 and 2022. Through counterfac-
tual experiments, we find that while emerging economy policymakers might favor moderate
reforms to improve democratic representation, achieving the institutional depth seen in devel-
oped countries is politically unfeasible, despite its clear advantages for citizens.

Keywords— Sovereign Debt Crises, Tax Smoothing, Representation and Accountability, Corruption,
Sovereign Default, Fiscal Pro-Cyclicality, Emerging Markets
JEL— D72, E43, F34, E62, F41

* We thank participants of the 2022 Kuznets Mini-Conference: Political Distortions and Economic Development
(Yale), the 2023 AEA Winter Meetings, 2023 SED Meetings, 2023 Spring Midwest Macro Meetings, and seminar
participants at the FRB of Philadelphia and Richmond and the International Monetary Fund. The views expressed
herein are those of the authors and do not necessarily reflect the views of the Federal Reserve Bank of Richmond or
CAFRAL.
†
Corresponding Author. Email: [email protected]
‡
Email: [email protected]

1
1 Introduction
Fiscal policy responses to shocks are widely different across countries. Developed economies smooth taxes,
borrowing in recessions, and paying back during booms1 . This is illustrated in Table 1 (in blue), where a
positive correlation is observed between the primary balance and output.2 This behavior, which dampens
economic cycles, is consistent with optimal debt determination (see Barro (1979)).

Table 1: Business Cycle Moments and Institutions

Moments Developed Emerging

Economic and Fiscal Variables
σ(y) 5.0 7.3
σ(c)/σ(y) 1.1 1.2
σ(g)/σ(y) 0.9 1.3
Fitch Ratings 3.8 2.9
ρ(NX,y) 0.0 -0.1
ρ(PB,y) 0.3 0.0

Representation and Accountability (R&A)

Average 1.2 0.1
Volatility (st dev) 0.1 0.2

Control of Corruption (CC)

Average 1.4 -0.2

Number Countries 30 28

Notes: Yearly sample, 1990-2022, where y =GDP, c =private consumption, and g = public
consumption. All real and log-linearly de-trended. P B is the primary balance (Rev-Exp+rb),
and N X are net exports (X-I), both as % of GDP. Correlations represent regression coef-
ficients. R&A and CC in the range [-2.5,2.5]. See Appendix A.3 for sources and further
computation details.

In emerging economies, borrowing is acyclical on average, and pro-cyclical for certain countries (see
Figure 1). The latter means that borrowing rises during expansions, facilitated by easy access to international
capital markets, and decreases during recessions, often as a result of a sovereign default.3 Evidence of
emerging countries’ reduced capacity to mitigate shocks can be seen in the more pronounced fluctuation in
1
See the studies by Sargent and Velde (1995) or Hall and Sargent (2020)).
2
The negative relationship between net exports and GDP is consistent with this as well, indicating that countries
tend to import goods (e.g. borrow) in recessions.
3
This phenomenon, especially prevalent in Latin America during the 80s and 90s, has been noted by studies includ-
ing Kaminsky et al. (2004), Talvi and Végh (2005), Ilzetzki and Vegh (2008), and more recently Kaas et al. (2020)
and Bianchi et al. (2023). However, our sample period from 1990 to 2022 reveals a significant decrease in the number
of countries showing procyclicality.

2
private and public consumption relative to output, a higher volatility of output itself, as shown in Table 1 and
the higher frequency of sovereign debt crises (see Mitchener and Trebesch (2023)). Even within emerging
economies, some countries are ‘serial defaulters’ (like Argentina), while others (like Chile) have been less
prone to default.

Figure 1: Correlation between Primary Balance and GDP - benchmark sample

1
.5
0
-.5

Egypt
Hungary
Ecuador
Romania
Greece
Uruguay
Portugal
Netherlands
Bulgaria
Malta
Indonesia
South Africa
Mexico
Denmark
Argentina
Latvia
Iceland
Panama
Costa Rica
Poland
Venezuela
Lithuania
Colombia
Slovenia
Kazakhstan
Spain
Thailand
Taiwan
Philippines
Tunisia
Brazil
Peru
Slovakia
Korea
China
Singapore
Luxembourg
Russia
Czech Republic
India
Ireland
Italy
France
Australia
Belgium
Chile
El Salvador
United States
United Kingdom
Austria
Finland
Canada
Sweden
Israel
Switzerland
Japan
Germany
Norway
Developed Emerging

Notes: See Appendix A.3 for sources and computation details.

Observing such diverging outcomes, international organizations typically prescribe policies to serial
defaulters that proved successful in developed countries, such as reducing deficits and the size of the public
sector. For example, IMF loans include conditions emphasizing fiscal austerity and budgetary reforms,
including structural reforms at times 4 . However, policymakers quickly deviate from such prescriptions or
do not implement them at all, as they become politically unfeasible. Unfortunately, it is not enough that
policymakers understand what good policies are; they need to find it optimal to implement them. In this
paper, we argue that the institutional framework in which policies are decided is as important as, if not more
than, the policy prescriptions themselves. We concentrate on two key institutional characteristics, namely the
degree of democratic representation and accountability and the control of corruption. We propose that these
elements play a significant role in guiding fiscal policy and debt management strategies. Once we summarize
how these two factors are measured in the data, we proceed to describe their role in our quantitative model.
4
See https://www.imf.org/en/About/Factsheets/Sheets/2023/IMF-Conditionality

3
 Our empirical proxies are obtained from the World Bank’s Worldwide Governance Indicators. The
degree of representation and accountability (R&A) captures public perceptions about (i) the extent to which
citizens can participate in selecting their government through free elections, freedom of expression, freedom
of association, and a free media (also known as ‘vertical accountability’) and (ii) de-facto checks and balances
(or ‘horizontal accountability’)5 . Positive scores indicate stronger democratic institutions (see additional
details in Appendix A). The second indicator we use is ‘control of corruption’ (CC), capturing the extent to
which public power is exercised for private gain, including both petty and grand forms of corruption, as well
as the capture of the state by elites and private interests. Smaller values indicate that excessive patronage,
nepotism, and ‘favors-for-favors’ are widespread. Table 1 presents summary statistics for these series.

Figure 2: Distribution of R&A (left) and CC (right) across countries

1
1
Kernel Density

Kernel Density
.5
.5 0

-2 -1 0 1 2 -1 0 1 2 3
Representation & Accountability Control of Corruption

Developed Emerging Developed Emerging

Notes: World Bank’s WGI, 58 countries, 1990-2022 (country list in Online Appendix).

Figure 2 shows the distributions of R&A (left) and CC (right) across 58 countries between 1990 and 2022.
On average, developed economies (solid green line) are more representative and exhibit less corruption than
emerging ones (dashed maroon line).6 However, there is some overlap between the CC series, indicating that
some developed economies have levels of corruption closer to those in emerging countries, and vice-versa.7
Figure 3 depicts the average values of R&A plotted against CC for each country, revealing a positive
correlation. In other words, governments with stronger democratic institutions typically exhibit less corrup-
tion (with Singapore being an exception). Different colors are used to emphasize the country’s default risk,
5
The concepts of vertical and horizontal accountability were first introduced by O’Donell (1998). While ‘represen-
tation and accountability’ is a subjective or de-facto measure combining both vertical and horizontal accountability,
there exist de-jure measures that independently capture each dimension. Because they are based on enacted legisla-
tion and power structure (see Lührmann et al. (2020)), they change less frequently and more abruptly than R&A. Our
benchmark measure captures partial changes, as perceived by the population. On average, these measures are highly
correlated with those constructed by the WGI.
6
The sample only includes countries for which we also had economic and fiscal variable observations, as well as
Fitch sovereign debt ratings, as described later. The unrestricted distributions of CC and R&A are shown in the Online
Appendix, Figure 1. Clearly, our sample misses countries with the weakest institutions.
7
The only developed country with negative R&A scores is Singapore. See Tables 1 and 2 in the Online Appendix.

4
 Figure 3: Default risk, representation and accountability and corruption

2
Norway
Sweden
Luxembourg
Netherlands Denmark
Finland
Switzerland
Canada
Iceland
Australia
Belgium AustriaGermany
Ireland
United Kingdom
Portugal
Malta France
Spain United States
Costa Rica Slovenia Uruguay
Italy Japan
1

CzechPoland
Republic Chile
Slovakia Lithuania
Greece Taiwan
Hungary
Latvia
South AfricaKorea Israel
Representation &

Panama
Accountability

Bulgaria
Romania
Brazil
Argentina
India

Mexico
Philippines
El Salvador
Peru
0

Lesotho Singapore
Ecuador
Indonesia Colombia

Thailand

Tunisia

Venezuela
Russia
-1

Kazakhstan Egypt

China
-2

-1 0 1 2 3
Control of Corruption

Low rated Medium rated High rated

Sources: World Bank’s “Worldwide Governance Indicators (WGI)” and Fitch Ratings, 58
countries, 1990-2022. See Appendix A.3 for details.

which is calculated using the long-term average of Fitch ratings (see Appendix A.2 for details). Countries are
categorized into three groups: those with high ratings (marked by green circles), medium ratings (marked
by blue squares), and low ratings (marked by red triangles). Lower ratings indicate a higher probability
of default. Countries with higher representation that keep corruption under control generally have good
sovereign credit ratings, which is predominantly seen in developed economies. It is worth noting that our
sample does not include countries with very low R&A and CC scores, due to either their lack of participation
in international financial markets or insufficient economic data.8
Table 1 presents the standard deviation of representation and accountability for the two groups of coun-
tries, underscoring the volatile nature of R&A in emerging markets. There is also notable disparity within
this group, as illustrated by Figure 4. The government of Chile scores significantly higher than Argentina’s
and India’s, reaching a R&A mean that is much closer to that of developed countries in recent years. Despite
having similar paths until 2015, Argentina and India’s R&A scores diverged soon after, highlighting the
dynamic nature of institutional development.
Taken together, the evidence above indicates that: (i) fiscal policy and macroeconomic outcomes are
more volatile in emerging countries (there is less tax smoothing), (ii) default risk is lower in countries with
8
In the figure, we are only including countries for which we have institutional and economic data, as well as Fitch
ratings’ observations. The scatter plot including all countries with existing R&A and CC scores for this period can be
found in Figure 2 of the Online Appendix.

5
 Figure 4: Representation & Accountability over time (Argentina, Chile, and In-
dia)

1.5
Representation &

1
Accountability
.5 0

1995 2000 2005 2010 2015 2020

Year

Chile Argentina
India

strong representation and accountability and low levels of corruption, and (iii) institutional strength evolves
over time. While developed economies look similar both in terms of their institutional setting and risk of
default, emerging economies are more heterogeneous. In this paper, we build a theory consistent with these
observations, where institutional strength and the pervasive effects of corruption take center stage.
Our study builds on the dynamic political economy bargaining model developed in Azzimonti and Mitra
(2023), to explore how democratic representation and corruption impact policymakers’ behavior towards tax
smoothing and preventing sovereign debt crises.9 The population is composed of symmetric groups repre-
senting diverse regions, industries, castes, and ethnic or religious groups, each with a leader that has political
influence. Group leaders bargain periodically over public goods spending, financing through distortionary
taxes and/or external borrowing, and over whether to repay or default. One group leader is randomly selected
to make a policy proposal, but needs the support of a ‘minimum winning coalition’ (mwc)—formed by a
subset of these groups— to implement policy. In order to secure support, the proposer can promise ‘political
favors.’ These can take the form of targeted public goods, permits for resource exploitation, bribes, sub-
sidies, nepotism, pork spending, favorable exchange rates, or policies affecting non-economic dimensions
such as ethnic or religious in-group favoritism, same-sex marriage and abortion laws, or transgender rights.
A key characteristic is that these favors are costly to provide and can be targeted to benefit a specific group.
The ‘control of corruption’ is an institutional characteristic determining how easy it is to transform public
revenues into favors.
The size of the minimum winning coalition (mwc) in the model plays a crucial role in determining the
number of groups whose preferences are considered when making policy decisions, capturing the notion of
‘representation and accountability.’ A small mwc allows the proposer to pass policies with the support of
9
In Azzimonti and Mitra (2023), we augmented the bargaining model proposed by Battaglini and Coate (2008) by
integrating sovereign default, akin to Arellano (2008) and Cuadra et al. (2010).

6
only a few groups, ignoring the effects of their actions on other groups. Because political favors are financed
with government resources, a common pool problem arises (where all groups contribute, but only a few enjoy
the proceeds of taxation). As a result, the use of favors can be widespread, generating deadweight losses10 .
When the mwc equals the total population, the model collapses to that of a benevolent planner without
commitment. This captures a society where the preferences of all groups are considered, and favors are
never used. International risk-neutral lenders provide loans to the government, and in turn, price sovereign
bonds based on the government’s default risk. This pricing mechanism ensures that the lenders discipline
government borrowing, offering lower prices when the likelihood of default is high. To accurately assess the
risk, lenders must consider, in addition to the state of the economy and the current stock of debt, the current
and future institutional environment as well as the level of corruption11 . Because of this, regime changes
have important implications for debt sustainability.
The latter is illustrated with a quantitative counterfactual experiment in which we evaluate the macroe-
conomic consequences of a permanent regime change that strengthens institutions in two countries (e.g. in-
creases the size of the mwc), but asymmetrically, using Argentina and Chile as case studies. Both countries
were under the rule of military juntas in the 1970s, but transitioned to democracies in 1983 and 1989, re-
spectively. While Argentina and Chile share similar geographic, cultural, and historical characteristics, Chile
was more successful in ensuring well-functioning checks and balances– or horizontal accountability— after
democracy was reinstated. In our model, this is captured by a higher mwc size. We show that our model
can help rationalize the different trajectories in Argentina’s and Chile’s economies in subsequent years. In
particular, we find that Argentina suffers recurrent sovereign debt crises and significant volatility in private
and public consumption, whereas Chile does not, consistently with the data (see Mitchener and Trebesch
(2023)). A key insight is that while both countries can default at any time, institutional strength makes this
option less desirable for Chilean policymakers than for Argentinean ones.
More generally, our analysis reveals that policymakers tend to smooth taxes when R&A is high, as in
developed countries, but less so when R&A is low, as in emerging countries, even when strategic default
is permitted in both cases and policymakers are equally patient. Specifically, we find that policymakers in
countries with strong institutions behave responsibly, reducing debt in good times and avoiding the default
region. In contrast, when the government is less representative (but still democratic), government officials
borrow excessively in good times to give out favors and find it beneficial to default when recessions hit, as
commonly observed in emerging markets. In other words, our model generates ‘political defaults,’ namely
that a politician chooses to default when a benevolent planner would not. As a result, low R&A countries
10
Choi et al. (2021) provides micro evidence of how excess ARRA stimulus grants extended to politically connected
firms in the U.S. lead to fewer jobs.
11
In a model with nominal debt—denominated in domestic currency—policy risk, emanating from less developed
and volatile domestic institutions, would lead to excess returns due to currency risk and a positive violation of UIP, as
documented in Kalemli-Özcan and Varela (2021).

7
exhibit worse tax smoothing (and sometimes even procyclical debt), recurrent sovereign debt crises, and
frequent default events12 .
Addressing corruption can mitigate some impacts of inadequate checks and balances. Yet, our analysis
reveals that reforms enhancing representation and accountability are significantly more effective in stabiliz-
ing economic shocks than similar efforts to curb corruption. While anti-corruption measures could theoreti-
cally improve resource allocation by lowering default risks and economic volatility, our findings suggest that
politicians in environments with extremely weak democratic structures initially prioritize strengthening these
structures. This preference stems from the fact that, although both strategies might lead to comparable out-
comes in terms of taxation, public expenditure, and debt repayment in a political equilibrium, anti-corruption
efforts limit the use of political favors by those in power. However, extensive reforms in representation and
accountability to match those in developed countries are politically unattainable. Politicians are more in-
clined to implement incremental reforms, just large enough to receive favorable treatment from lenders, but
allowing them to maintain their capacity to distribute political favors.
The rest of the paper is organized as follows. Section 2 summarizes how our work fits in the literature.
In Section 3, we describe the economic environment and define the politico-economic equilibrium. Section
4 describes the calibration strategy. We compare the transition to democracy in Argentina and Chile by
simulating a once-and-for-all institutional reform in Section 5 and describe tax smoothing in emerging and
developed economies through the lens of our model in Section 6. In Section 7, we analyze a ‘crackdown
on corruption’ and show how it can be different from a reform that increases representation. Section 8
concludes.

2 Related Literature
The literature on fiscal policy cyclicality has bifurcated into two distinct branches: (i) the traditional public
finance literature and (ii) the more quantitative sovereign default literature. Our paper aims to connect them
by providing a unifying theory of fiscal policy cyclicality and sovereign default decisions.
The traditional public finance literature has predominantly focused on studying developed economies,
assuming the existence of a ‘commitment technology’ for debt repayment. The seminal work of Barro
(1979)13 , followed by the studies of Aiyagari et al. (2002) and Barseghyan et al. (2013), have highlighted
tax smoothing and counter-cyclical debt as key implications of their models (these are summarized in Yared
(2019)). While consistent with the cyclical properties of fiscal policy in developed countries like the US, their
predictions are inconsistent with the empirical behavior observed in emerging economies (as documented by
Gavin and Perotti (1997), Talvi and Végh (2005), and Ilzetzki and Vegh (2008)). An exception is Ilzetzki
12
While autocratic regimes are not the main focus of interest in our applications, we show how the model can be
expanded to analyze them. An important difference between democratic regimes with low R&A and autocratic regimes
is that the latter exhibit less political turnover. We develop this further in Appendix C.2.
13
Also see Lucas and Stokey (1983) and recent work by Hall and Sargent (2020) for a discussion of its empirical
validity.

8
(2011), who generates procyclical fiscal policies in a model subject to political distortions, but assuming
commitment to repay debt (and abstracting from default risk, inconsistent with Figure 3).
At the other extreme, the sovereign default literature—with the pioneering work of Arellano (2008)—
has centered on explaining the behavior of emerging economies, with particular emphasis on Latin America.
An important innovation in these models is allowing the government to strategically default. Within this
environment, Cuadra et al. (2010) shows that debt increases in booms rather than recessions and taxes are
pro-cyclical (see more recent work by Kaas et al. (2020) and Martinez et al. (2022)). The latter is a direct
consequence of counter-cyclical sovereign default risk: When default risk rises the government needs to
raise tax rates to finance public goods14 (see Vegh and Vuletin (2015) for empirical evidence on this chan-
nel). At the end of the day, then, what makes emerging economies different from developed ones in these
papers boils down to one important assumption: whether countries have a commitment technology to repay
debt obligations or not15 . The European sovereign debt crisis, however, demonstrated that even developed
nations, such as Greece, may occasionally opt for strategic default. In this paper, we provide a unifying
theory that can encompass the behavior of countries across different stages of development16 .
We build on the dynamic bargaining political economy model developed in Azzimonti and Mitra (2023)
to study how different degrees of representation and accountability along with poor control of corruption
affect fiscal policy decisions. We relate to the dynamic bargaining literature in political economy, which
typically abstracts from default (see Baron and Ferejohn (1989), Battaglini and Coate (2008), Azzimonti
et al. (2016), or Drazen and Ilzetzki (2023)).17 A recent paper by Cusato Novelli (2020) incorporates de-
fault in a bargaining model, allowing the author to simultaneously match the observed default frequencies
and debt-to-output ratios, something that traditional sovereign default models typically fail to achieve 18 .

Our model can also match these moments. The main difference between our papers is that we consider
a production economy (rather than an endowment one) where taxes are distortionary19 . This generates a
tax-smoothing motive—as in the traditional public finance literature— a key ingredient to our analysis of
the effects of institutional strength on fiscal policy cyclicality and the impact of politics on business cycle
fluctuations. Another important difference between our models is that we allow for democratic represen-
14
Espino et al. (2020) shows, in a model with nominal debt, how governments can use a combination of both inflation
and distortionary taxes to finance debt. The interplay between default risk and inflation is also shown in Mihalache
(2020) in an environment with long-term debt.
15
In addition, it is typically assumed that policymakers in emerging markets are ‘more impatient’ than those in
developed markets, incorporated through differences in their discount factors.
16
While there is heterogeneity across countries on institutional development, we abstract from inequality within
the country. See D’Erasmo and Mendoza (2016) or D’Erasmo and Mendoza (2021) for studies where default has
redistributive effects.
17
Andreasen et al. (2019) considers default but in a voting model and abstracting from productive inefficiencies
associated with the tax system.
18
See a discussion of the difficulties in obtaining this in default models with a benevolent government in the literature
reviews by Tomz and Wright (2007) and Aguiar and Amador (2014).
19
In Mendoza and Yue (2012a), a default amplifies economic fluctuations through a production channel. A default
reduces the ability of producers to obtain the working capital needed to import intermediate goods. An inefficiency
arises after a default since domestic intermediate goods are poor substitutes for imported ones. Taxes are lump-sum in
their environment, so a tax-smoothing motive is not considered.

9
tation to fluctuate over time as seen in the data. This feature of our model can explain why countries with
low democratic representation display more volatile fundamentals than their developed counterparts. Both
democratic representation and corruption, as well as their volatility, amplify the business cycle and increase
the likelihood of sovereign debt crises.
It is important to note that while we incorporate dynamic bargaining, ours is not a model of debt restruc-
turing, where the government renegotiates repayment with foreign creditors (see Yue (2010)). Our study
focuses instead on internal political bargaining, where the key actors are domestic politicians, who often
must agree on tax increases to avoid a default. This internal bargaining process is influenced by the institu-
tional context, specifically the levels of democratic representation and corruption within the government.
We offer three novel quantitative contributions relative to previous work. First, we explain the tax-
smoothing behavior of both developed and emerging countries assuming the same discount factor, which has
not been done in previous quantitative studies (e.g. they need to assume a significantly smaller value for the
discount factor of emerging economies’ government to generate default in equilibrium). There is a body of
work analyzing the role of political distortions on default decisions within the sovereign default literature (see
Hatchondo and Martinez (2010) and Chang (2007)20 ). Hatchondo et al. (2009) study how differences in re-
election probabilities that affect the effective discount factor of a government, can impact default incentives.
Chatterjee and Eyigungor (2019) and Cotoc et al. (2022) study default incentives under endogenous re-
election probabilities, but again assume relatively low discount factors. In our bargaining model, the discount
factor equals the inverse of the risk-free rate as in standard macroeconomic models, allowing us to replicate
the tax-smoothing behavior of developed economies when either institutions are strong (e.g. the mwc equals
the whole population) or corruption levels are low, or both. By weakening institutions (e.g. lowering the size
of the mwc), we can approximate the cyclicality behavior of emerging economies without the need to change
the degree of impatience of policymakers. In the appendix, we illustrate how democratic representation and
political turnover are not equivalent by analyzing an autocracy. That is, a situation where the size of the mwc
is low but the autocrat remains in power forever. We also show that the dynamic behavior of our model is
not equivalent to one where the discount factor is low. Second, we investigate how institutional reform
can improve tax smoothing in an emerging country using Chile and Argentina as case studies. Third, we
conduct a policy experiment to highlight how an improvement in the technology to control corruption differs
from an institutional reform strengthening democratic representation –an exercise that has implications for
international policy-making.

3 The model
This is an infinite-horizon model where time is discrete. The economy has a domestic sector, with competi-
tive firms and workers, a government that decides policy, and international investors that buy debt from the
government.
20
Relatedly, Herrera et al. (2020) models the relationship between popularity (e.g., political booms) and financial
crises.

10
3.1 Economic Environment
The domestic economy is populated by a continuum of infinitely lived agents uniformly distributed across n
groups, representing diverse regions, industries, castes, and ethnic or religious groups, each with a population
normalized to 1. Agents in group i derive utility from private consumption c, labor l, a public good g, and
political favors f˜i . These can be thought of as pork spending (bridges to nowhere), nepotism, bribes, or
targeted public goods. Generically, we think of these as ‘corrupt public practices.’ That is activities that
benefit a subset of the population at the expense of all groups. The instantaneous utility of an agent in group
i satisfies
U (c, l, g, fi ) = u(c, l) + πv(g) + f˜i , (1)

where π denotes the weight of pure public goods relative to the consumption-leisure aggregate. Agents
discount the future at a rate β. In line with small open economy models of sovereign default, we consider
imperfect capital markets that prevent individuals from using assets to smooth aggregate fluctuations. This
is akin to assuming that agents are hand-to-mouth—as the model abstracts from idiosyncratic shocks—
implying that c = wl, with w denoting the economy-wide wage rate.
Firms produce a single non-storable consumption good, y, using the linear technology y = h(z, d)l. The
economy is subject to aggregate TFP shocks, where z ∈ Z follows a first-order Markov process µ(z 0 |z).
We denote the long-run value of TFP by z̄. When z > z̄, the economy is in good times, and when z < z̄ the
economy is in a recession. The function h(z, d), increasing in z, takes different forms in states of repayment
(d = 0) and default (d = 1).
Firms are competitive and maximize profits, subject to a distortionary policy τ ,

max (1 − τ )y − wl.

Policy τ can be interpreted as labor income taxes, revenue taxes, or other costs proportional to the firm’s
output which are mandatory to operate the production technology and generate revenues for the government
(permits, fees, etc.). The behavior of international lenders is modeled following Arellano (2008). Specif-
ically, we assume that an infinite number of identical risk-neutral international lenders are present. These
lenders can borrow and lend at a risk-free rate r from the international capital market and can also buy
government debt at a price q. Because this is a perfectly competitive market, lenders earn no profits in
equilibrium.

3.2 Government Sector

The government collects revenue Rev(τ ) = τ y and uses its proceeds to fund public goods and political
favors. In addition, it can issue one-period non-contingent real bonds that can be bought and sold in inter-
national markets. The net resources raised by the government from capital markets are denoted by qb0 − b,
where a positive value of b0 is the face value of the current debt of the government and q is the price of new
bonds. The government can choose to strategically default on debt b, a decision we denote with d = 1. If

11
the government reneges on its current debt obligations, it is excluded from international credit markets for a
stochastic number of periods determined by probability θ, with credit standing downgraded to Ω = 0. Since
b0 = 0 in such case, it must finance expenditures solely with Rev(τ ) until access to markets is re-gained.
The government budget constraint is
h i X
0
Rev(τ ) − g + (1 − d) qb − b ≥ fi , (2)
i
P
with fi ≥ 0 and i fi corresponds to the total amount of resources devoted to political favors. We assume
that engaging in corrupt practices is costly: For each dollar taken from the budget, only a fraction 1/φ can
be consumed. Hence, when a policymaker spends fi dollars in favors, agents only receive f˜i = fi . The φ
institutional variable φ is our model counterpart for ‘control of corruption.’ Higher values correspond to
better technologies used to crack down on corruption in the public sector.
When the government borrows from abroad, additional resources (measured in consumption goods) are
imported. When the government pays down debt, resources are exported. The aggregate resource constraint
is
X
y =c+g+ fi + nx,
i

where nx = x − i denotes net exports (e.g. exports minus imports). Since there is a single production
good in the world, the exchange rate is equal to 1 at all times. In our model, trade deficits have a one-to-one
mapping with government deficits.

Bargaining process

Each group in society has a leader with political influence, who has a ‘seat’ in the bargaining table where
policy is decided. Policy proposals need the support of enough leaders, m ≤ n, to be implemented. This
process gives each one of them some veto power and opens the possibility for corrupt public practices. We
refer to m as the size of the ‘minimum winning coalition’ or mwc. We allow for m to change stochasti-
cally over time, capturing the evolution of representation and accountability illustrated by Figure 4. More
specifically, we assume that m follows a first-order Markov process with transition probability p(m0 |m).
We define permanent increases or decreases in its long-run average m̄ as “regime changes,” reflecting fun-
damental shifts in the political landscape or institutional structures. On the other hand, temporary shocks to
m are interpreted as fluctuations in the balance of power that may arise due to electoral outcomes (where
one group controls all branches of government) or the fluctuating relative importance of certain groups at
different periods of time.
We use the bargaining protocol from Azzimonti and Mitra (2023), which extends the dynamic legislative
bargaining model (with exogenous status quo) of Battaglini and Coate (2008) to incorporate strategic default.
Group leaders meet at the beginning of a period and one of them is chosen (at random) to make a policy
proposal. Since individuals are identical in all regions, the identity of the leader of each group is irrelevant.

12
A proposal is given by (
{τ, g, b0 , d, f1 , f2 , ..., fn } if Ω = 1
Φ(Ω) = (3)
{τ, g, f1 , f2 , ..., fn } if Ω = 0
When the country is excluded from credit markets, Ω = 0, it is ‘in default’, indexed by d = 1 and unable
to borrow b0 = 0. Otherwise, when Ω = 1, d and b0 are free choices.
If the proposal succeeds in obtaining the support of m leaders, the policy is implemented. If the proposal
fails to obtain enough support, leaders move to the next proposal round in which a new proposer is chosen
at random. If no agreement can be reached in T ≥ 2 proposal rounds, an outsider is appointed to choose a
reference (symmetric) policy.

3.3 Politico-Economic Equilibrium

The aggregate state variables at the outset of any period are the stock of debt b, the TFP shock z, the size m
of the mwc, and the credit standing Ω ∈ {0, 1}. Let s = {z, m, Ω}, and the full state-space be denoted by
Π = {s, b}, We start by describing a competitive equilibrium given government policy Φ(Ω). In equilibrium,
variables depend on Φ(Ω) and Π. Throughout the paper, we write Φ (omitting Ω) to simplify notation.
Firms maximize profits, implying that w(s, Φ) = (1 − τ )h(z, d). To characterize the agents’ problem,
we make further assumptions about the utility functional forms.

Assumption 1. Suppose that u(c, l) is of the GHH family (see Greenwood et al. (1988)),
!1−σ
1 l1+γ g 1−σ
u(c, l) = c− and v(g) = ,
1−σ 1+γ 1−σ

where σ > 0 captures the degree of risk aversion and γ > 0 represents the Frisch elasticity of labor supply.

Because agents are hand-to-mouth, it is easy to show that the individual labor supply l(s, Φ) and optimal
consumption c(s, Φ) are independent of debt decisions,
1
l(s, Φ) = [h(z, d)(1 − τ )] γ and c(s, Φ) = (1 − τ )h(z, d)l(s, Φ).

As a result, aggregate output is also independent of debt, y(s, Φ) = nh(z, d)l(s, Φ).
International lenders make zero profits when Ω = 1 (that is, when the country can actually borrow).
Their break-even bond prices satisfy

1 − d(z 0 , m0 , b0 )
Z  
q(s, Φ) = ∂z 0 ∂m0 |(z, m), (4)
(z 0 ,m0 )∈Ψ(z 0 ,m0 ) 1+r

where Ψ(z 0 , m0 ) = {(z 0 , m0 ) : d(z 0 , m0 , b0 ) = 0} is the repayment set for the government and d(z 0 , m0 , b0 )
is tomorrow’s equilibrium default choice (which only depends on tomorrow’s exogenous states and today’s
borrowing in the political equilibrium, as will become clearer in the next section). Note that q depends on s,
since z and m are Markov processes.

13
 It is useful to define the budget balance excluding favors
 h i
Rev(s, Φ) − g + (1 − d) q(s, Φ)b0 − b if Ω = 1
B(Π, Φ) = (5)
 Rev(s, Φ) − g if Ω = 0
P
with Rev(s, Φ) = τ y(s, Φ). The budget constraint of the government collapses to B(Π, Φ) ≥ i fi .
We focus on a symmetric Markov-perfect equilibrium, implying that any proposer in round k ∈ {1, 2, ..., T }
selects identical policies. The full decision-making problem of the proposer, which is quite heavy in terms
of notation, is relegated to Appendix B.1. Since the other n − 1 legislators are ex-ante identical, the m − 1
coalition members needed to pass the legislation are randomly selected from them. To secure their consent,
the proposer offers favors fi = f to members of the selected mwc and fi = 0 to members outside of it.
This, again, is a result of symmetry. The proposer keeps the remaining fp = B(Π, Φ) − (m − 1)f , which
may be higher than what is given to other mwc members, fp ≥ f .
Following Azzimonti and Mitra (2023), we show in Appendix B.2 that the proposal is accepted imme-
diately, in k = 1, and that the proposer’s problem is equivalent to one where the welfare of the average
member of the mwc is maximized21
  B(Π, Φ)
max U c(s, Φ), l(s, Φ), g + + βEs0 J(Π0 )
Φ φm (6)
s.t. B(Π, Φ) ≥ 0,

with J(Π0 ) denote the expected value of continuation utility for the proposer given that the next period’s
Markov-perfect equilibrium policy is Φ0 (Π0 ),
  B(Π0 , Φ0 (Π0 ))
J(Π0 ) = U c(s0 , Φ0 (Π0 )), l(s0 , Φ0 (Π0 )), g 0 + + β Es00 J(Π00 ). (7)
φn
Note that the left-hand side of the objective function in eq. (6) is not equal to J(Π), from eq. (7). This
happens because the proposer can use the budget to keep fp for certain in the current period. In the future,
the proposer may not even be in the mwc, implying that he or she would receive less in expectation (see
Appendix B.2 for a derivation of the objective function above). This force will create incentives for the
proposer to over-spend and over-borrow relative to a benevolent planner, a feature already highlighted in
Battaglini and Coate (2008). In this paper, we emphasize how this characteristic also affects incentives to
default, and how these incentives become stronger in the shadow of corruption (low φ) and institutional
weakness (low m̄).
The size of the mwc, given by m, is a key parameter of our model: it captures how many votes the pro-
poser needs in choosing a policy. As m grows closer to n, the policymaker must internalize the preferences
of all members of society. When m = n, there is full representation and accountability, and institutions
are strong. On the contrary, when m = 1 there are no independent institutions and the proposer chooses an
21
A key assumption for the proof to go through is the linearity in f . Most dynamic bargaining papers, including
Battaglini and Coate (2008) and Barseghyan et al. (2013), need to make this assumption to make the problem tractable.

14
allocation that devotes significant resources towards her group. Due to the common pool problem, where
all members of society contribute towards government revenues, but only one group can appropriate a large
share of it, this situation corresponds to one with low R&A in the data. Similarly, a government can have
varied degrees of control over corruption. A combination of low R&A and low CC implies maximum in-
centives to divert resources towards favors, as evident from eq. (6). As we will show later, institutional
reform through a change in m̄ or φ will have different implications for the continuation value, and hence
different equilibrium outcomes.

3.4 Characterization and intuition

Suppose that γ, σ ≥ 1 and h(z, d) = z (this last assumption matters for calibration, but not at this point). It
1+ γ1 1
is easy to show that Rev(τ, z) = τhnz(1−τ ) γ . Recall
i that the budget devoted to corrupt public practices
is B = Rev(τ, z) − g + (1 − d) q(z, m, b0 )b0 − b ≥ 0. There are two possibilities: either the proposer
promises political favors in exchange for support, B > 0, or not B = 0 (so the constraint binds).

Case 1: When B > 0,

1
g ∗ (m) = (πφm) σ .

Public good provision is independent of debt and default decisions and increases with the size of the mwc.
This implies that societies with weak institutions, m << n, will tend to under-provide public goods. Taxes
are chosen to equate the marginal cost of taxation in terms of consumption (net of the disutility of labor) to
the marginal increase in the provision of political favors:
 −σ  
γ 1+γ n τ
[(1 − τ )z] γ = 1− => τ ∗ (m, z).
1+γ m γ(1 − τ )

The resulting tax rate is also independent of debt decisions. When m < n, the marginal gain in political
favors associated with an increase in τ is larger than that of a planner. As a result, societies with weak insti-
tutions will tend to over-tax and over-spend on political favors relative to the planner’s preferred allocations.
The expressions above illustrate that whether the country is in default or not does not affect g or τ in this
case. Additional borrowing is used only to increase the budget devoted to corrupt public practices. This also
implies that the incentives to repay debt are unaffected by public good provision and tax distortions.22

Case 2: When B = 0, that is no longer true. Here, borrowing and default decisions affect the provision
of g: h i
g = Rev(τ, z) + (1 − d) q(z, m, b0 )b0 − b ,

22
These results are a direct consequence of the linearity in fi . Relaxing this assumption would make policy functions
smoother, and the results less stark. However, the bargaining model’s level of complexity grows significantly with
non-linearity.

15
and taxes (through g),
 −σ  
γ 1+γ
−σ τ
[(1 − τ )z] γ = πg n 1− .
1+γ γ(1 − τ )

The LHS of the equation above is the same as before, but the RHS is now the marginal gain in additional
public good provision financed by an additional tax dollar. The only effect of m in this case comes through
the endogenous price of debt q(z, m, b0 ). When the country is in default (d = 1), however, g and τ are
independent of institutional strength.

Figure 5: Policy functions for borrowing, taxes, government spending, and po-
litical favors

b’ taxes
default

default
default
b*(mL)
default

τ
b*(mH)
𝑏 0
𝑏 𝑏 b 𝑏 𝑏 b

favors
default

g
default

𝑔
default
default

𝑓 =0
𝑏 𝑏 b 𝑏 𝑏 b

To better understand how the strength of institutions affects fiscal policy in the political equilibrium, we
display two sets of policy functions, for mL < mH , in Figure 5. The figure is constructed assuming that
governments have access to financial markets, Ω = 1. We cannot derive all the theoretical properties of the
model (as the decisions to borrow and default are quite involved analytically), so these functions are stylized
representations of those obtained through numerical simulations.
First, let us focus on the functions in blue, corresponding to a government with a high value of mH .
When b is low, the proposer chooses B > 0, constant taxes τ ∗ (mH , z) and public good provision g ∗ (mH ),
as derived in Case 1. From any b < b1 , debt jumps immediately to b∗ (mH ), reflecting the proposer’s
eagerness to spend fiscal capacity on political favors, even if that means financing them via borrowing. At
b1 , the constraint on B starts to bind and B = 0 thereafter. As the stock of debt rises and, if more debt is
to be used to extend political favors without changing g or taxes, the price of bonds declines steeply (not

16
shown). This makes it more costly to finance expenditures via public debt. As a result, the government is
forced to increase taxes along with higher borrowing, and lower public good provision in order to pay back
debt. At b2 , rolling over the level of debt becomes too costly, so the country defaults (setting d = 1). At
this point, b = b0 = 0, freeing fiscal capacity to reduce taxes to τ d and increase public spending to g d . Note
that the proposer does not waste resources on political favors when in default. This is because policymakers
agree on policy when in autarky in this example23 .
Now compare the blue policy functions (for mH ) with the red ones (for mL ). When debt is low, poli-
cymakers in countries with low mwc requirements spend more on political favors and less on g, as derived
above (see Case 1). In addition, b∗ (m) is larger, implying that countries where institutions are weaker, bor-
row much more when it is cheap to do so. This implies that debt accumulates faster. The price schedule for
debt is also steeper as the same debt implies higher chances for default for mL . Repayment implies cutting
back on favors that are valued more. This can be seen by the fact that the default point moves to the left
(e.g. b2 (mL ) < b2 (mH )). In other words, countries with weaker institutions are unable to sustain as much
debt as those with higher m. This is consistent with the data in Figure 3, where we found that countries with
larger R&A scores had lower default risk. In Appendix 5, we also show that they also have higher levels of
debt/GDP in our sample. Finally, note that these differences arise only when corrupt practices give private
gains to the proposer and other members of the mwc. As φ grows, the gains from providing B > 0 dissipate
and policies converge to those of a benevolent planner. Hence, weak institutions only distort policies and
allocations in the shadow of corruption in this environment.

4 Quantitative Model
This section summarizes the calibration strategy and model fit. In addition, we illustrate how governments
with different mwc sizes respond to a recession. Counterfactual experiments are presented in subsequent
sections.

4.1 Calibration
We calibrate the model to Argentina during the period 1993-2022. A list of the economic and fiscal variables
used, together with data sources, can be found in Appendix A.3. We start by describing the exogenously
determined parameters and then move to the calibrated ones.
A period in the model is one quarter24 . The risk aversion parameter in the utility function is set to
σ = 2, following the literature. The inverse of the Frisch elasticity of labor supply is γ = 2. The risk-
free interest rate, r = 0.55%, equals the value of the real 3-month U.S. T-bill interest rate for the period
23
This also implies that the values of g and τ obtained in our model right after default are identical to the ones that
would arise in a standard default model without politics (e.g. one where choices are made by a benevolent planner).
This is developed further in Appendix C.1, where we contrast our bargaining model to the standard case, but assuming
a lower discount factor as in Arellano (2008).
24
We convert annual data to quarterly series by repeating the annual value for each quarter when quarterly data is
unavailable.

17
under consideration. The discount factor, β, is set to match the inverse of the gross risk-free rate in the
1
model β = 1+r = 0.9945. It is worth noting that the discount factor takes the same value as in standard
macroeconomic models. We do not need to assume an extremely low value for β—as most sovereign default
models do25 —, because the effective degree of impatience in our model depends on m/n. This will be
explained in more detail below. We fix the corruption parameter φ = 1 in the benchmark model, but later
relax this to study the effects of corruption.
The exogenous productivity shock is assumed to follow an AR(1) process of the form:

zt+1 = (1 − ζz )z̄ + ζz zt + zt+1 (8)

where Ezt+1 = 0 and E(zt+1 )2 = σz2 . The income process parameters ζz and σz are chosen by fitting
the above AR(1) process to real GDP per employed person, per year (HP-filtered using a parameter of 100).
The average value of the process z̄ is normalized to 1. The fitted AR(1) process is discretized to 21 possible
realizations of the productivity shock using Tauchen and Hussey (1991). Table 2 reports the resulting values
of ζz and σz used in the simulations.
In terms of the political process, we assume that there are n = 20 groups in the population. This is a
normalization trading off computational accuracy and time26 . The size of the mwc, represented with m, is
also assumed to follow an AR(1) process of the form:

mt+1 = (1 − ζm )m̄ + ζm mt + m
t+1 . (9)

Similar to the productivity shock, we assume Em m 2 2

t+1 = 0 and E(t+1 ) = σm . The values for ζm and σm
are chosen by estimating an AR(1) process on the normalized series of R&A for Argentina for the period
under consideration, displayed in Figure 4. Inspection of this figure suggests that R&A follows a mean
reverting process for most of the samples. We first normalize the data to take values between 1 and 20, and
discretize the fitted AR(1) process to take integer values in the same range. This maps the data to the model,
where a higher value of m would imply more representation and accountability in the fiscal policy decision
process.
Following Chatterjee and Eyigungor (2012), the probability of re-entry into the market following default,
θ, is set to 0.0385. This is approximately equal to 1 ∼
= 26 quarters (6.5 years) of exclusion from financial
θ
markets after a default event. This value is consistent with the estimates by Gelos et al. (2011) and Richmond
and Dias (2009). We also assume that default entails a loss in productivity (see Rose (2005), Bocola (2016)).
While we do not formally model these, we take a reduced-form approach following Chatterjee and Eyigungor
25
For example, Arellano (2008) sets β to 0.8.
26
It is without loss of generality since what matters for computational results is the ratio m/n.

18
 Table 2: Calibration Targets

Parameter Value Target Description

σ 2 CRRA
γ 2 Frisch Elasticity
1
β 0.9945 1+r
FOC
r 0.0055 90 day U.S. Treasury
θ 0.0385 6.5 Years of Exclusion
z̄ 1 Normalized
n 20 Normalized
φ 1 Normalized

ζz 0.925 Persistence Real GDP
AR(1)
σz 0.017 Volatility of Real GDP 
ζm 0.954 Persistence of R&A
AR(1)
σm 0.234 Volatility of R&A 
α0 -0.364 E(Spreads) = 7.05% 

Debt
α1 0.403 = 50%

GDP Jointly Calibrated
Spend
π 1.2 Y
= 0.14 

m̄ 4.77


(2012). In particular, we assume that labor productivity takes the following form27 :
(
z if d = 0
h(z, d) = (10)
z − max{0, α0 z + α1 z 2 }, α1 ≥ 0 if d = 1

We calibrate α0 , α1 , m̄ and π jointly by minimizing a quadratic loss function to match: average spreads,
average of external debt to GDP, and the average value of government spending to GDP in the data. We use
150 equally spaced grid points for the borrowing level ranging from 0 to 100 percent of average GDP. The
model is simulated for one million model periods. The values of the initial and final five thousand periods
are discarded before using the series to compute the relevant moments. The bottom of Table 2 reports the
parameters resulting from the calibration procedure.

4.2 Model Fit

We compute long-run moments from the simulation, and report their values in the second column of Table
3 along with their data counterparts. The numerical algorithm is described in the Section 2 of the Online
Appendix. The moments reported below the horizontal line are the ones targeted in the calibration, and the
ones above the line were not matched by design.
27
The curvature of the default cost function is disciplined by the parameters α0 and α1 . If α0 > 0, and α1 = 0, then
the cost of default is proportional. If α1 > 0, and α0 = 0, then the cost rises more than proportionally to the rise in
productivity. If α0 < 0, and α1 > 0, then for z < − α α1 , the default cost is 0, but the cost rises more than proportionally
0

for higher realizations of z.

19
 The model performs fairly well in matching the data. An important feature of our model is that we
can match the pro-cyclicality of fiscal policies earlier documented by Talvi and Végh (2005). These are
reflected by the strong positive correlation between
 government
 spending and GDP, as well as the negative
NX
correlation between net exports and GDP. A ρ y ,y < 0 means that in recessions (e,g, a decline in y),
N X = x − i increases. Because in our model next exports are just external borrowing, this implies that the
country borrows in booms, not in recessions (which is the opposite of what optimal tax-smoothing behavior
would prescribe). The relative volatility of consumption to output is greater than 1 (as in most emerging
economies) and close to that observed in Argentina. Tax rates in the model are almost acyclical, as in the
data, for the period under consideration28 .

Table 3: Model Fit

Moment Data: Argentina Benchmark

σ(c)
σ(y)
1.26 1.29

ρ(y,c) 0.97 0.81 
ρ(y,g) 0.79 0.70 Pro-cyclical
ρ( NX , y) -0.57 -0.37

y
ρ(y,tax) 0.03 -0.004
ρ(r − r∗ , y) -0.30 0.08
σ(y) 5.0% 3.4%
σ(Spreads)
  2.9% 4.7%
Debt

µ y
50% 50% 
  
µ gy 14% 13% Matched by Construction

µ(r − r∗ ) 7.05% 7.11% 


There are two dimensions where our model does not perform as well. First, the volatility of output is a
bit lower in the model than it is in the data. Second, the correlation of spreads to GDP is smaller and of the
opposite sign to that in the data. The latter might be a consequence of using short-term debt, which results
in a steep bond price schedule. In a recession, debt reduction dominates the effects of increasing spreads
to avoid a default. Another reason could be that the series for m and z are uncorrelated in the model, but
could be correlated in the data. See Mendoza and Yue (2012b) for a theory that can deliver counter-cyclical
spreads.

4.3 Tax smoothing in the shadow of corruption

The canonical models determining optimal fiscal policy state that the government should increase debt rather
than increase taxes in response to a temporary shock (that lowers revenues or increases spending needs). This
results from the assumption that raising tax revenue generates distortions in the economy, whereas selling
28
This calculation also includes periods of default.

20
bonds does not (see Barro (1979), Lucas and Stokey (1983)). The “tax smoothing” hypothesis implies that
government debt should be counter-cyclical (borrow in recessions) and taxes should exhibit low volatility.
Taxes should eventually rise to repay the debt, but should not track the shock process.
In this model, the correlation between the fiscal balance (revenues minus expenditures) and output is
−0.39 instead: the government borrows in booms rather than in recessions, a pro-cyclical policy response.
To understand intuitively why this happens, we simulate the response to a negative TFP shock in period 2,
assuming that aggregate productivity recovers immediately in period 3. In this experiment, we keep the size
of the mwc constant at its long-run average for the calibrated economy, mt = m̄ = 4.77. We analyze the
response for two different initial levels of debt: a low one b0 < b1 and a high one b0 > b1 .

Low Initial Debt

Suppose that the economy starts with no initial debt, b0 = 0. This corresponds to the case where b0 < b1 in
Section 3.4. The evolution of the endogenous variables for our benchmark economy is displayed in Figure 6,
with a dashed-red line. In period 0, before the recession hits, borrowing (defined as qb0 ) jumps immediately
and corrupt public practices skyrocket (favors/GDP go up). The economy enters the recession (period 2) with
high debt, so spreads rises to more than 6 percent. Unable to borrow cheaply, policymakers sharply decrease
1
the provision of public goods and hike taxes. This reduces the labor supply, since l = [h(z, d)(1 − τ )] γ
in equilibrium, exacerbating the effects of the negative shock: GDP not only falls because z goes down but
also because higher taxes reduce the labor supply l. Rather than stabilizing the economy, the government
amplifies the effects of a recession.

Figure 6: Response to a recession, low b0

Borrowing/GDP Spreads Public Cons/GDP Debt/GDP

50 8 14.5 50

40 14 40
6
30 13.5 30
4
20 13 20
2
10 12.5 10

0 0 12 0
0 5 10 15 0 5 10 15 0 5 10 15 0 5 10 15

GDP Private cons/GDP Favors/GDP Tax Rates

18.5 88 50 0.2

40
18 86 0.18
30
17.5 84 0.16
20
17 82 0.14
10

16.5 80 0 0.12
0 5 10 15 0 5 10 15 0 5 10 15 0 5 10 15

It is useful to contrast this to the response of a government with stronger institutions. To that end, we
re-compute the model assuming m̄H = 20 and repeat the experiment (time series displayed in blue). As
described in Section 3.4, a government with higher m behaves more responsibly. First, fewer resources
are devoted to corrupt public practices (favors are zero in this case). Second, borrowing doesn’t increase

21
as much initially. When the recession hits, the government can raise debt levels at a lower cost than in the
benchmark case (i.e., spreads do not jump), so it can afford to lower τ , as tax-smoothing would prescribe.
The decline in GDP is smaller, indicating the government’s role in stabilizing the economy.

High Initial Debt

We conduct the same exercise assuming initial debt to be about 70 percent of GDP, corresponding to b0 ∈
(b1 , b2 ) in Section 3.4. Figure 7 shows that both economies start reducing debt in the first period, but the
benchmark economy (in red) does this at a much faster rate because of the higher spreads it faces. The
benchmark economy imposes higher taxes and has lower GDP and private and public consumption. When
the recession hits in period 2, taxes increase considerably more than in the m̄H economy, even though its
legacy debt in the current period is lower. This happens because markets price a higher default probability
in the economy with weak institutions, despite its lower debt levels.

Figure 7: Response to a recession, high b0

Borrowing/GDP Spreads Public Cons/GDP Debt/GDP

54 10 13.5 60

52 8
55
50 13
6
50
48 4
12.5
46 45
2
44 12
0 40
0 5 10 15 0 5 10 15 0 5 10 15 0 5 10 15

GDP Private cons/GDP Favors/GDP Tax Rates

18.5 88 5 0.22

86 4 0.2
18
84 3 0.18
17.5
82 2 0.16
17
80 1 0.14

16.5 78 0 0.12
0 5 10 15 0 5 10 15 0 5 10 15 0 5 10 15

Once the benchmark economy recovers, taxes drop as the cost of borrowing becomes lower, and debt
stabilizes at a constant level of around 45 percent of GDP. Instead of further reducing the stock of debt, the
government chooses to spend the additional resources on political favors, which increase sharply after the
recession ends. In the case of high debt, there is imperfect tax smoothing in both economies. This shows
up in the decline in both private and public consumption to GDP ratios and increases in taxes during the
recession. However, the negative effect of the shock is stronger in the economy with weak institutions. For
initial debt levels above b2 , the benchmark economy responds to a negative TFP shock by outright defaulting.
The impulse responses are omitted, but available upon request.

22
Political Defaults

We now consider a large recession, where TFP goes down significantly in t = 2 (for only one period). As
in the previous case, countries start the simulation with a high level of debt (of around 70 percent). Figure
8 shows that the country with strong institutions mH (solid blue) faces a large recession with an austerity
measure (e.g. a tax hike), and chooses not to default. Spreads increase marginally.

Figure 8: Response to a “Great Recession,” high b0

Spreads Debt/GDP
8 60

6
40

20
2

0 0
0 5 10 15 0 5 10 15

Borrowing/GDP Tax Rates

60 0.2

40 0.18

20 0.16

0.14
0
0 5 10 15 0 5 10 15

The country with weak institutions mL (dashed maroon), instead, prefers to default in the second period
rather than adopt an austerity measure. This illustrates the heterogeneous response of countries with different
political environments, shedding light on the behavior of European nations following the global financial
crisis. Greece (a country with a R&A of 0.89, close to that of an emerging economy) defaulted in March
2012, whereas Germany (a country with a R&A of 1.38, mH in our example) did not. Our finding is related to
Arellano and Bai (2016), which shows that a country subject to fiscal restrictions, in particular the inability
to increase taxes, will be forced to default when facing a large recession. They refer to those as ‘fiscal
defaults.’ In their paper, taxes are exogenously fixed through a budget rule. In our model, taxes are flexible
but the politician prefers not to increase them when institutions are weak, giving rise to a ‘political default.’
A planner, in this case, would choose an austerity measure instead.

5 Transition to Democracy: Chile vs Argentina

Chile and Argentina were under the rule of military juntas in the late 1970s. Democratic governance was
reinstated in both nations in the 1980s, with Argentina holding its inaugural democratic election in 1983
and Chile following suit in 1989. Free speech, free association, free press, and free elections were fully
restored in both countries soon after. Yet, intriguingly, the trajectories of fiscal and economic indicators
diverged significantly in the post-transition period. Argentina experienced a series of sovereign debt crises

23
and high volatility in output, whereas Chile experienced economic stability and low sovereign debt spreads.
Economists attribute this divergence to economic reforms, particularly market liberalization. However, both
nations embarked on similar economic liberalization journeys. Political scientists, on the other hand, em-
phasize the nuances in institutional reforms following the transition to a free democracy as the main differ-
entiating factor. In this section, we evaluate whether heterogeneity in institutional reform can help explain
the different trajectories of these two countries. More specifically, we evaluate the macroeconomic conse-
quences of a permanent regime change that strengthens institutions in both countries, but asymmetrically.
In particular, we assume that Chile’s m̄ is higher than Argentina’s.

Figure 9: Vertical and Horizontal Accountability

1

1
.9

.9
.8

.8
.7

.7
Horizontal Accountability
Vertical Accountability
.6

.6
.5

.5
.4

.4
.3

.3
.2

.2
.1

.1
0

0 10 20 30 0 10 20 30
Reform Date Reform Date

Chile Argentina Chile Argentina

Notes: VA and HA obtained from the V-Dem dataset.

This choice is informed by reading accounts of how the institutional reform was implemented in these
countries (see O’Donell (1998) and Wigell (2017)), as well as data on the evolution of sub-components of
representation and accountability. In his seminal work, O’Donell (1998) distinguishes between ‘vertical’ and
‘horizontal’ accountability. Vertical accountability (VA) refers to the citizens’ capacity to hold the govern-
ment accountable, evidenced by their freedom to establish political parties and engage in open elections. He
argues that both countries were successful in ensuring VA. Horizontal accountability (HA), instead, refers
to an effective system of separation of powers and checks and balances that impose constraints on how pol-
icy can be used to further politicians’ private and political goals. According to the author, Chile was more
successful than Argentina in securing strong HA, because key veto players could block the incumbent from
governing by decree through legislative and judiciary independence. Figure 9 provides evidence of this
claim, by displaying the evolution of of HA and VA from the V-Dem (Varieties of Democracy) dataset29 .
29
Note that our measure of R&A, voice and accountability, includes both dimensions. Another difference is that
R&A is based on public perceptions (a de facto measure), whereas VA and HA are constructed from practices that are
legally recognized (a de jure measure).

24
Period 0 indicates the year in which the first democratic election was held. While VA is similar after free
elections took place, Argentina’s HA is significantly lower than Chile’s following the transition.

Simulating the transition

In terms of our model, stronger HA is captured by a larger number of groups with veto power, m̄. We
computed our model for two different values of m̄: Argentina’s corresponds to the benchmark of m̄A = 4.77,
whereas Chile’s is higher at m̄C = 8.8 (with n = 20 in both cases). We keep the volatility of m to that of
the benchmark30 . We assume that both countries move from a dictatorship (with m̄i = 1) to a democracy
in period 0, and simulate the two economies for 100 quarters assuming that they face the same sequence of
TFP shocks. We discipline the initial debt-to-output ratios of the two countries using IMF data for the years
in which the institutional reform happened. Argentina’s debt/GDP in the initial period is 55.8 percent (e.g.
the observed value in 1983) and in Chile is 77.1 percent (e.g. the observed value in 1989).

Figure 10: Evolution of policies and allocations to institutional reform

Borrowing/GDP Taxes Spreads Default

80 0.25 20 1
Argentina
0.8 Chile
60 15
0.2
0.6
40 10
0.4
0.15
20 5
0.2

0 0.1 0 0
0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100

Public Cons/GDP Favors/GDP GDP Private cons/GDP

21 90
60
14
20

40 85
13 19

18
12 20 80
17

11 0 16 75
0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100

Figure 10 plots the evolution of fiscal and economic variables. Chile (displayed with a solid blue), has
a higher stock of debt when the reform happens, but starts reducing debt/GDP soon after. Its borrowing,
q 0 B 0 , reaches a smaller value than Argentina’s (red dashed line) in less than 20 quarters. The decrease in
debt obligations is achieved not by increasing taxes in the first period—which are actually smaller than
Argentina’s— or by initially reducing public spending, but instead by choosing not to waste tax dollars on
political favors. Faced with a higher number of veto players and a large stock of debt, the Chilean government
behaves more responsibly: the distortions associated with engaging in corrupt public practices are not worth
the benefit to the proposer, as they need to be distributed among a larger mwc. Note that in this experiment
we did not change φ: the reform did not make it more difficult to engage in corruption, it just made it less
appealing. The elimination of fi is an endogenous choice. Knowing that future policymakers will also not
30
This is also lower for Chile. Lowering the volatility for m will only strengthen our results.

25
engage in corruption, further increases the incentives to reduce indebtedness today (because of better prices
on debt), implying that the proposer is also more patient when m̄ is higher. As a result of this, the Chilean
government can better smooth taxes and consumption when recessions happen (and consistent with Figure
7). Spreads are low for Chile throughout and experiences no default episodes.
Argentina’s post-transition experience is completely different. Rather than reducing borrowing, the gov-
ernment quickly engages in corrupt practices: we see positive values of fi through the simulation. This is
consistent with Wigell (2017), who documents that in the beginning of the 1990s, President Menem (Ar-
gentina) started manipulating targeted benefits as a lynchpin for mobilizing political support31 . Intuitively,
a lower m̄ worsens the common pool problem associated with government expenditures. Fewer members of
the mwc benefit from “priority” projects of the government at the cost of taxes and debt repayment obliga-
tions on the entire population. Favors are financed with higher taxes, and the government under-spends on
public good provision. Markets attribute a higher probability of default to the Argentinean government, so
spreads are much higher. When shocks hit, the economy experiences sovereign debt crises (e.g. jumps in
spreads) and even two default episodes soon after. In other words, Argentina becomes a country where ’to
default is the default.’ The experiment points to the reason for significantly higher volatility in policy and
allocations, as well as lower private and public consumption than in Chile.

6 Tax-smoothing: Emerging vs. Developed countries

The experiment in the previous section suggests that the ability to smooth policy and avert sovereign debt
crises is strongly linked to institutional strength in our model. Is this insight more general or was it simply
an artifact of arbitrary initial conditions or TFP shock realizations? To study this, we simulate the model for
a large number of periods and eliminate the first 5000 quarters before computing statistics. The resulting
long-run business cycle moments for the two countries are shown in Table 4. Argentina’s is displayed in the
first column and Chile’s in the second. These aim to capture generic emerging economies with varying mwc
sizes. In addition, we include a simulation for a hypothetical developed economy with strong institutions
by setting m̄ = n (this also corresponds to the solution of a benevolent planner without commitment to
repay).32
The model predicts monotonicity in fiscal policies and macroeconomic outcomes. Spreads are highest in
Argentina, where institutional quality is the lowest. Government spending relative to GDP and tax rates are
more volatile in Argentina than in Chile, illustrating differences in their ability to smooth taxes. As expected,
Chile exhibits higher volatility of taxes and government spending compared to the developed economy. Our
results explain the heterogeneity observed in the data across the spectrum of emerging economies with vary-
31
In Chile, by contrast, targeted programs remained under firm technocratic control. We can interpret fi as the
use of targeted programs for political support, as we do not have heterogeneity among agents (e.g. there are no other
redistributive programs).
32
To make things comparable, we are only changing m̄ across the three simulations. The TFP shock process and
realizations, as well as other parameters in the model are constant. In other words, we are not re-calibrating these
economies in each column to match spreads. Because Argentina has the worst outcomes of the three, using it as a
benchmark is a lower bound on the ability of economies with stronger institutions to smooth taxes.

26
 Table 4: Business Cycle Moments in the Long Run

Moment Argentina Chile Developed

m̄ = 4.77 m̄ = 8.8 m̄ = 20
Institutions
R&A (mean) 4.77 8.80 20.0
Instability (St Dev) 0.78 0.78 0
Favors/GDP (mean) 4.52% 0.10% 0

Fiscal Policy
Spreads 7.08% 0.06% 0.001%
σ(tax) 0.04 0.01 0.01
σ(g)/σ(y) 1.49 1.25 0.83
µ(b/Y) 50.4% 41.1% 18.3%
ρ(TB,y) -0.37 -0.20 0.55
ρ(FB,y) -0.31 -0.21 0.65
µ(tax) 17.69 13.77 13.63

Macro Outcomes
Sp. Crisis 9.85% 4.12% 2.82%
σ(c)/σ(y) 1.29 1.14 0.84

ing levels of institutional strength. Some of them are serial defaulters, or prone to a spread crisis33 like
Argentina, while others are significantly less prone to such crises like Chile. In our simulations, Argentina
has a 9.85 percent chance of falling into a spread crisis, while Chile is much closer to a developed country
with a 4 percent possibility of the same. Favors as a percentage of GDP are higher on average in Argentina,
implying a significantly higher targeted use of public resources for political gains. This number is negli-
gible in the other two cases with higher m̄. The developed economy exhibits a strong positive correlation
between trade balance and GDP, but this statistic is negative for both Chile and Argentina. This implies that
the developed country can borrow in bad times without having to resort to abrupt tax increases to finance
spending. This is also reflected in the sign and magnitude of the fiscal balance (FB) in all the columns, and
in the magnitude of the ratio of the volatility of consumption to that of GDP. Highly volatile taxes spill over
into private consumption expenditures, making them more volatile than output. That is, policy amplifies the
cycle in countries with weak institutions like Argentina.
Why are spreads so much higher in Argentina? Figure 11 displays the long-run distribution of debt to
GDP in the three economies. The average value of debt to GDP is highest for Argentina (red dashed line),
at around 50 percent. The other two economies have significantly lower values of debt to GDP. Moreover,
the developed economy has a significant mass at zero, indicating a desire to save in good times and use debt
33
Following Trebesch and Zabel (2017), we code a spread crisis when we observe: (i) an increase in spreads com-
pared to the previous period above the 99th percentile of the spread distribution or (ii) an annualized spread of 10
percentage points or higher.

27
 Figure 11: Ergodic debt distribution.

0.09
Developed
Chile
0.08
Argentina

0.07

0.06

0.05
Density

0.04

0.03

0.02

0.01

0
-10 0 10 20 30 40 50 60 70 80 90 100
Debt/GDP

only in recessions34 . Argentina defaults very frequently, as seen at the large mass in -10 (the arbitrary coded
value for debt under default, for this plot’s readability). Chile and the developed economy have no mass
at that point, indicating that they rarely default. After gaining market access (post-default, indicated by the
small mass at 0), Argentina jumps to a higher value of debt-to-GDP (of around 30 percent). This is not the
case for Chile, whose average debt-to-GDP ratio is lower (blue dotted line). A mass at 0 implies that in some
parts of the state space Chile would even like to save. The developed economy with strong institutions has
a distribution of debt/GDP similar to Chile’s, the only difference is that the mass at zero is larger35 .
Figure 12 shows repayment and default regions in the state space for the three economies. The darker-
shaded regions represent combinations of TFP shocks and legacy debt-to-GDP ratios where the government
would optimally repay. The darkest region in the bottom right corner represents the repayment region for
Argentina, while the combination of the two darker regions represents the repayment set for the other two
countries. If the country finds itself in the light yellow region, it defaults. In addition to having higher debt
on average (shown in the previous figure), Argentina defaults at a larger region of the state space, making
it more likely to end up in the default region. The default region for the other two economies is the same.
However, since Chile still borrows more than the developed economy, it is more likely to default.
This result is driven by two forces that change when m̄ increases. On the one hand, stronger institutions
reduce the gains of engaging in corrupt public practices (as political favors must be shared with a larger
number of groups in society). On the other hand, it also increases the likelihood that the proposer is in the
34
See Bianchi and Sosa-Padilla (2023) for an alternative environment where accumulating reserves could help
smooth business cycle fluctuations.
35
Following the tradition in the sovereign default literature, we computed the model restricting debt to be positive
(e.g. ruling out savings). This is not binding in the calibrated benchmark economy (Argentina), but it can become
binding in some states of the word for economies with higher values of m̄ (such as the two examples we presented
above). However, even if we allowed for such behavior, it is unlikely that borrowing will stop completely as in Aiyagari
et al. (2002). With imperfect political institutions (lower representation and corruption), the asset floor is endogenously
determined, which would make borrowing optimal at most points of the state space.

28
 Figure 12: Default regions.

mwc in the future (e.g. receives political favors). This makes the proposer endogenously more patient. In
Appendix C.2, we split these two forces by comparing our benchmark to an autocracy where the proposer
is in power forever. That is a situation in which m = 1, but the autocrat knows it will be in the mwc with
probability 1 (which we interpret as the case of China). We invite an interested reader to visit this extension.

7 Implementing a ‘crackdown on corruption’

In the previous counterfactual experiments, we kept the control of corruption fixed at the normalized value
of φ = 1. In this section, we want to study whether a ‘crackdown of corruption’ (e.g. an increase in φ)
is as effective as a reform that strengthens representation and accountability (e.g. an increase in m̄). An
increase in φ reduces the benefits from engaging in corrupt public practices, since for each dollar taken from
the budget, the proposer and each of the coalition members only get 1/φ of their respective shares. An
increase in m̄ reduces the per-member amount received through corrupt public policies, but also affects the
probability of being part of the mwc in the future. We evaluate which of the two reforms is more palatable
to the proposer at different stages of institutional development, and compare this to the welfare gains of an
average citizen in the population.
We simulate the economy where alternative values of φ or m̄ are increased by {10, 25, 35, 50, 60, 75,
80, 100, 200, 400} percentage points. When φ is increased m̄ is kept constant, and vice-versa. The long-run
response of endogenous variables and their volatilities are shown in Figure 13. The dashed maroon line
depicts crackdowns on corruption (e.g. increases in φ) whereas the blue solid line depicts strengthening the
representation and accountability (e.g. increases in m̄).

When the percentage increase in m̄ is relatively small, debt issuance goes up because the fiscal space
increases as a result of a reduction in spreads. A further increase, however, reduces debt and also the relative

29
 Figure 13: Effect of Increasing m and φ by the Same Magnitude on the Endoge-
nous Variables

Rel. volatility of cons. to GDP Volatility of taxes

1.4 0.04

1.2 0.03

1 0.02

25
50

80

25
50

80
0

0
10

20

40

10

20

40
Debt-to-GDP (%) Spreads (%)

60 6

4
40
2

20 0
25
50

80

25
50

80
0

0
10

20

40

10

20

40
Probability of spread crisis Favors (% of GDP)
10
4 Dem. Rep.(Avg. m)
8 Corruption ( )

6 2

4
0
25
50

80

25
50

80
0

0
10

20

40

10

20

40
Notes: Numbers on the x axis indicate percentage increase in m̄ from 4.77 or φ from 1. When φ is increased m̄ is kept constant and vice-versa.

volatility of consumption to GDP. The volatility of taxes is always decreasing, regardless of whether φ or m̄
are increased, indicating that either reform improves tax smoothing. Corrupt public practices, spreads and
the probability of spread crises also decrease monotonically for all degrees of such reform36 . An increase in
φ leads to qualitatively similar implications, but are much smaller in magnitude compared to that of m̄ when
it comes to reducing the volatility of consumption and taxes. In general, the performance of the economy is
more responsive to increases in m̄ than to increases in φ. Intuitively, volatility in m is equivalent to policy
uncertainty as in Azzimonti (2018). In a case with lower m̄, each member in the coalition carries more
weight, and hence fluctuations in the coalition size lead to larger swings in fiscal policy variables.

7.1 Reform when institutions are weak

The analysis above relates to long-run outcomes under alternative institutional environments. Apart from
affecting these, the m̄ reform also impacts the intertemporal price of fi , and hence the desirability of the
proposer to implement such reform. Figure 14 shows the welfare of the proposer (left panel) and that of an
average citizen (right panel) assuming that either a crackdown on corruption or an m̄ reform is implemented
in the current period. The value at the origin represents welfare if no reform is undertaken. While the average
citizen is always better off when there is less corruption or stronger representation (e.g. welfare is increasing
36
For spreads, debt-to-GDP and favors average values of a one-million-period long simulation are reported.

30
in φ as well as m̄ reforms), the proposer’s welfare is non-monotonic. This happens because the group with
greater political power benefits most from institutional weakness in the short run.

Figure 14: Effect of Increasing m or φ on Welfare

Proposer`s Welfare Average Citizen's Welfare

-375 -375

-380 -380

-385 -385

-390 -390

-395 -395

-400 -400

-405 -405

-410 -410
Dem. Rep. Reform
Corruption ( ) Reform
-415 -415
25
50

10 0

25
50

10 0
0

0
8

8
20

40

20

40
Notes: Numbers on the x axis indicate percentage increase in m̄ from 4.77 (solid-blue) or φ from 1 (dashed-maroon). When φ is increased m̄ is kept constant and vice-versa.

To understand the intuition, it is useful to look at eq.(6): a higher m reduces the weight on favors in the
current period. But it also increases the chances of being chosen as a member of the mwc in the following
period, thereby increasing the chances of receiving political favors in the future37 . Therefore, the relative
price of favors in the current period with respect to the future increases which makes the proposer optimally
reduce favors in the current period and back-load it for the future. This price effect is relevant only because
favors are positive at a lower level of development– when both the parameters are low. Increasing φ however,
increases the price of favors both in the current and the future periods keeping the relative price unchanged.
At lower values of m̄, the effect of a reduction in volatility dominates the relative price effect, and hence the
proposer is better off if m̄ goes up. Increasing m̄ by more than 60 percent leads to a drop in the proposer’s
welfare. Although there is higher fiscal space, borrowing starts to decline as the utility from favors for each
coalition member is lower. Here the relative price effect dominates the volatility effect38 .
Both the rise and the corresponding fall, due to lower favors, in the proposer’s welfare are modest when
a corruption reform is undertaken. Since favors are directly targeted, such a reform never encourages bor-
rowing as much as in the m̄ reform to pay for favors. Consequently this results in lower taxes to finance
favors in the φ-reform regime. An increase in m̄ of 400 percent or more has the same effect on welfare as an
increase in φ of the same magnitude. At that point, moments look similar to a developed country. They are
The unconditional probability of being a member of the mwc in the following period is given by m−1
37
n .
38
After an 80 percent reform, welfare starts to marginally improve again because taxes drop, and consumption and
GDP rise. At this point, fi drops to zero.

31
characterized by relative volatility of consumption to GDP of less than 1, less volatile taxes, lower spreads,
and no spending in political favors, as shown in Figure 13.
From the viewpoint of the average citizen, a φ reform is associated with higher welfare than a m̄ reform
in the initial bursts of reform (up to 70 percent). Lower taxes and lower debt allow for higher consumption
and welfare. However, given a choice, the proposer would rather implement an m̄ reform to attain favorable
treatment from the lenders, move to the peak of the proposer’s welfare (when m̄ is increased by about 60
percent), and stay put. The average citizen would rather have a 60 percent crackdown on corruption, than a
60 percent increase in representation. This shows that political economy frictions may come in the way of
an emerging/developing economy transitioning into a developed country. It is important to emphasize that
foreign lenders price debt by taking into account only the risk of repayment, without any concern about the
use or abuse of the lent resources.

7.2 Reform when institutions are stronger

In this paper, we have reported long-term averages of R&A and CC for all the countries in our sample.
However, over time many of them came out of dictatorships and there had been significant gains in terms of
institutional development.
Argentina, for example, shows a current R&A measure that is 60 percent higher than its long-term
average. Given higher accountability, these countries need to make promises of reform to the electorate
and carry them out to stay in power and get re-elected. The new president-elect of Argentina, Javier Milei,
promises to put an end to corruption during his tenure, along with many other lofty reform promises. In Figure
15 we plot the welfare implications for such promises and evaluate their credibility. More specifically, we
evaluate if a corruption reform is a promise that can be fulfilled.

Figure 15: Effect of Increasing m and φ from peak m̄

Proposer`s Welfare Average Citizen's Welfare

-376 -376 Dem. Rep. Reform

Corruption ( ) Reform

-378 -378

-380 -380

-382 -382

-384 -384

-386 -386

-388 -388

-390 -390

-392 -392
10

20

30

40

50

60

10

20

30

40

50

60

Notes: Numbers on the x axis indicate percentage increase in m̄ from 7.63 or φ from 1. When φ is increased m̄ is kept constant and vice-versa.

32
 We start from m̄ that is already 60 percent higher and also corresponds to the peak for the proposer in
Figure 14. Evidently, the proposer would not want to make any reforms at this point, as they would result
in lower welfare. Increasing φ, however, has an associated welfare loss that is smaller for the proposer. The
average citizen, however, would be better off improving democratic institutions. Intuitively, with already
favorable treatment from the lenders, a φ-reform does not have the price effect that back-loads favors. This
results in higher borrowing and higher taxes than a similar m̄ reform that results in lower aggregate welfare.

8 Conclusion
We introduce a fiscal policy model that explains how countries with different levels of institutional devel-
opment respond to macroeconomic shocks, considering the option of strategic default. The model focuses
on two key institutional factors: representation and accountability (R&A) and control of corruption (CC).
Empirically, we find that countries with strong R&A and CC have higher sovereign bond ratings, less tax
volatility, and more counter-cyclical fiscal policies. In contrast, countries with weaker R&A and CC face
higher economic volatility and more frequent sovereign debt crises.
Our theory highlights the significant impact of political favoritism in shaping these economic outcomes.
When R&A,—represented by m̄ (the size of the mwc) in our model—is low, governments tend to allocate
excessive resources for political favors, exacerbating the common pool problem. A low CC environment
facilitates the misuse of tax revenues for corrupt public practices. Lenders, recognizing these risks, adjust
their lending terms accordingly, often leading to higher borrowing costs and fiscal instability in countries
with weak institutions. We can explain economic patterns in countries with varying institutional strengths
without assuming a lower subjective discount factor, as often done in sovereign default literature. We also
generate ‘prudent,’ counter-cyclical fiscal responses in stronger institutional settings without assuming a
debt repayment commitment. Using the quantitative model, we can explain the divergent economic paths of
Argentina and Chile following the end of the dictatorship rules in the 1980s. A critical observation from our
model is that while both could default at any time, Chile’s stronger institutions make default less appealing
for its policymakers.
We compare the effects of permanent reforms in representation and accountability (R&A) against sus-
tained anti-corruption efforts. We find that in countries with very low institutional development, small im-
provements in R&A can lead to increased political favoritism, at the cost of lower welfare to the average
citizen. This happens as international lenders, encouraged by perceived improvements, offer better loan
terms, but the funds are often used for political gains. Such countries typically cease R&A improvements
once they reach a level that benefits the ruling proposer, with no incentive for further reform. We find that
in emerging countries, the incentives of the average citizen and those of the politicians never align perfectly.
From the perspective of public policy, it is important to keep these considerations in mind. A one-size-fits-all
policy may not be the best solution for economies at every stage of development.

33
References
Aguiar, M. and M. Amador (2014). Sovereign debt. In Handbook of International Economics Vol 4, pp.
647–87. North-Holland.

Aiyagari, S. R., A. Marcet, T. J. Sargent, and J. Seppälä (2002). Optimal taxation without state-contingent
debt. Journal of Political Economy 110(6), 1220–1254.

Andreasen, E., G. Sandleris, and A. Van der Ghote (2019). The political economy of sovereign defaults.
Journal of Monetary Economics 104, 23–36.

Arellano, C. (2008). Default risk and income fluctuations in emerging economies. American Economic
Review 98(3), 690–712.

Arellano, C. and Y. Bai (2016). Fiscal austerity during debt crises. Economic Theory 64.

Azzimonti, M. (2018). Partisan conflict and private investment. Journal of Monetary Economics 93,
114–131.

Azzimonti, M., M. Battaglini, and S. Coate (2016). The costs and benefits of balanced budget rules: Lessons
from a political economy model of fiscal policy. Journal of Public Economics 136, 45–61.

Azzimonti, M. and N. Mitra (2023). Political constraints and sovereign default. Journal of International
Money and Finance.

Baron, D. P. and J. A. Ferejohn (1989). Bargaining in legislatures. American political science review 83(4),
1181–1206.

Barro, R. (1979). On the determination of public debt. Journal of Political Economy 82(6), 1095–1117.

Barseghyan, L., M. Battaglini, and S. Coate (2013). Fiscal policy over the real business cycle: A positive
theory. Journal of Economic Theory 148(6), 2223–2265.

Battaglini, M. and S. Coate (2008). A dynamic theory of public spending, taxation, and debt. American
Economic Review 98(1), 201–36.

Bianchi, J., P. Ottonello, and I. Presno (2023). Fiscal stimulus under sovereign risk. Journal of Political
Economy 0(ja), null.

Bianchi, J. and C. Sosa-Padilla (2023, 7). Reserve Accumulation, Macroeconomic Stabilization, and
Sovereign Risk. The Review of Economic Studies, rdad075.

Bocola, L. (2016). The pass-through of sovereign risk. Journal of Political Economy 124(4), 879–926.

Chang, R. (2007). Financial crises and political crises. Journal of Monetary Economics 54(8), 2409–2420.

34
Chatterjee, S. and B. Eyigungor (2012). Maturity, indebtedness, and default risk. American Economic
Review 102(6), 2674–99.

Chatterjee, S. and B. Eyigungor (2019). Endogenous political turnover and fluctuations in sovereign default
risk. Journal of International Economics 117, 37–50.

Choi, J., V. Penciakova, and F. Saffie (2021). Political connections, allocation of stimulus spending, and the
jobs multiplier.

Cotoc, J., A. Johry, and C. Sosa-Padilla (2022). Sovereign spreads and the political leaning of nations.
Working paper.

Cuadra, G., J. M. Sanchez, and H. Sapriza (2010). Fiscal policy and default risk in emerging markets. Review
of Economic Dynamics 13(2), 452–469.

Cusato Novelli, A. (2020). Sovereign default, political instability and political fragmentation. Review of
International Economics.

Drazen, A. and E. Ilzetzki (2023). Kosher pork. Journal of Public Economics 227, 105001.

D’Erasmo, P. and E. G. Mendoza (2016). Distributional incentives in an equilibrium model of domestic

sovereign default. Journal European Economic Association 14(1), 7–44.

D’Erasmo, P. and E. G. Mendoza (2021). History remembered: Optimal sovereign default on domestic and
external debt. Journal of Monetary Economics 117, 969–989.

Espino, E., J. Kozlowski, F. M. Martin, and J. M. Sanchez (2020). Domestic policies and sovereign default.
FRB St. Louis Working Paper (2020-17).

Gavin, M. and R. Perotti (1997). Fiscal policy in latin america. NBER macroeconomics annual 12, 11–61.

Gelos, R. G., R. Sahay, and G. Sandleris (2011). Sovereign borrowing by developing countries: What
determines market access? Journal of international Economics 83(2), 243–254.

Greenwood, J., Z. Hercowitz, and G. W. Huffman (1988). Investment, capacity utilization, and the real
business cycle. The American Economic Review, 402–417.

Hall, G. J. and T. J. Sargent (2020, May). Debt and taxes in eight u.s. wars and two insurrections. Working
Paper 27115, National Bureau of Economic Research.

Hatchondo, J. C. and L. Martinez (2010). The politics of sovereign defaults. FRB Richmond Economic
Quarterly 96(3), 291–317.

Hatchondo, J. C., L. Martinez, and H. Sapriza (2009). Heterogeneous borrowers in quantitative models of
sovereign default. International Economic Review 50(4), 1129–1151.

35
Henisz, W. J. (2000). The institutional environment for economic growth. Economics & Politics 12(1),
1–31.

Herrera, H., G. Ordonez, and C. Trebesch (2020). Political booms, financial crises. Journal of Political
Economy 128(2).

Ilzetzki, E. (2011). Rent-seeking distortions and fiscal procyclicality. Journal of Development Eco-
nomics 96(1), 30–46.

Ilzetzki, E. and C. A. Vegh (2008, July). Procyclical fiscal policy in developing countries: Truth or fiction?
Working Paper 14191, National Bureau of Economic Research.

Kaas, L., J. Mellert, and A. Scholl (2020). Sovereign and private default risks over the business cycle.
Journal of International Economics 123, 103293.

Kalemli-Özcan, �and L. Varela (2021). Five facts about the uip premium. Technical report, National Bureau
of Economic Research.

Kaminsky, G. L., C. M. Reinhart, and C. A. Végh (2004). When it rains, it pours: Procyclical capital flows
and macroeconomic policies. NBER Macroeconomics Annual 19, 11–53.

Lucas, R. E. and N. L. Stokey (1983). Optimal fiscal and monetary policy in an economy without capital.
Journal of Monetary Economics 12(1), 55–93.

Lührmann, A., K. L. Marquardt, and V. Mechkova (2020). Constraining governments: New indices of
vertical, horizontal, and diagonal accountability. American Political Science Review 114(3), 811–820.

Martinez, L., F. Roch, F. Roldan, and J. Zettelmeyer (2022). Sovereign debt. IMF Working Paper 22/122.

Mendoza, E. G. and V. Z. Yue (2012a, 04). A General Equilibrium Model of Sovereign Default and Business
Cycles*. The Quarterly Journal of Economics 127(2), 889–946.

Mendoza, E. G. and V. Z. Yue (2012b). A general equilibrium model of sovereign default and business
cycles. The Quarterly Journal of Economics 127(2), 889–946.

Mihalache, G. (2020). Sovereign default resolution through maturity extension. Journal of International
Economics, 103326.

Mitchener, K. and C. Trebesch (2023). Sovereign debt in the 21st century. Journal of Economic Literature.

O’Donell, G. (1998). Horizontal accountability in new democracies. Journal of Democracy 9(3).

Richmond, C. and D. A. Dias (2009). Duration of capital market exclusion: An empirical investigation.
Available at SSRN 1027844.

36
Rose, A. K. (2005). One reason countries pay their debts: renegotiation and international trade. Journal of
development economics 77(1), 189–206.

Sargent, T. J. and F. R. Velde (1995). Macroeconomic features of the french revolution. Journal of Political
Economy 103(3), 474–518.

Talvi, E. and C. A. Végh (2005). Tax base variability and procyclical fiscal policy in developing countries.
Journal of Development Economics 78(1), 156–190.

Tauchen, G. and R. Hussey (1991). Quadrature-based methods for obtaining approximate solutions to non-
linear asset pricing models. Econometrica: Journal of the Econometric Society, 371–396.

Tomz, M. and M. Wright (2007). Do countries default in ‘bad times’? Journal of the European Economic
Association 5.

Trebesch, C. and M. Zabel (2017). The output costs of hard and soft sovereign default. European Economic
Review 92, 416–432.

Vegh, C. A. and G. Vuletin (2015). How is tax policy conducted over the business cycle? American
Economic Journal: Economic Policy 7, 327–370.

Wigell, M. (2017). Political effects of welfare pluralism: Comparative evidence from argentina and chile.
World Development 95, 27–42.

Yared, P. (2019). Rising government debt: Causes and solutions for a decades-old trend. Journal of Eco-
nomic Perspectives 33, 115–40.

Yue, V. Z. (2010). Sovereign default and debt renegotiation. Journal of International Economics 80(2),
176–187.

37
A Data Appendix
We collect data on (1) economic and policy variables, (2) institutional quality, and (3) sovereign credit
ratings. Economic and policy data series are annual, ranging from 1990 to 2022, institutional data starts
in 1996 and Fitch ratings in 1995. The three sets of variables, their sources, and detrending methods are
described below. There are 58 countries for which we have data on all three variables which will be referred
to as “benchmark” data, a subset of our full database.

A.1 Institutional Quality

We use two indicators for institutional quality: the strength of Representation and Accountability (R&A) and
the control of corruption (CC). These are part of the World Bank’s “World Governance Indicators” (obtained
from the WGI Website).

• R&A is the degree of Representation and Accountability, proxied by the WGI variable “Voice and
accountability,” which captures perceptions of the extent to which a country’s citizens can participate
in selecting their government, as well as freedom of expression, freedom of association, and a free
media. The variable ranges between [-2.5,2.5], with lower values indicating weaker R&A.

• CC is Control of Corruption (CC), obtained from the WGI variable with the same name, and cap-
tures perceptions of the extent to which public power is exercised for private gain, including both
petty and grand forms of corruption, as well as capture of the state by elites and private interests. The
variable ranges between [-2.5,2.5], with lower values indicating poorer CC.

Alternative measures: The WGI provides alternative measures of institutional quality. These are ‘Gov-
ernment Effectiveness (GE),’ ‘Regulatory Quality (RQ),’ and ‘ Rule of Law (RL)’. Their long run values are
highly correlated with CC: Corr(CC, RL) = 0.96, Corr(CC, GE) = 0.97 and Corr(CC, RQ) = 0.93.
Because of this, we simply use CC in our analysis.
In terms of de-facto measures, there is an alternative measure called ‘political constraints (PolCon)’
developed by Henisz (2000), that we have used in previous work (see Azzimonti and Mitra (2023)). The
series in only available until 2015, and this is why we use R&A instead. But the two are highly correlated:
Corr(CC, P olCon) = 0.85.

A.2 Fitch Ratings

Sovereign ratings are obtained from ‘Fitch Ratings’ (see website and Sovereign Rating Criteria). They are
available since 1995. The Sovereign Issuer Default Ratings (IDs) are evaluations that anticipate the ability
and willingness of a government to fulfill its debt obligations to both private-sector creditors and public debt
securities, regardless of ownership, in a timely and complete manner. Fitch uses a combination of quanti-
tative analysis and qualitative judgments to evaluate sovereign credit risk (e.g., their proprietary Sovereign

38
Rating Model and the Qualitative Overlay). In addition, Fitch considers the impact of the sovereign’s poli-
cies and actions on the country’s overall economic performance, which can have significant effects on the
sovereign’s creditworthiness and can, in turn, be influenced by it. Using a series of factors, Fitch summarizes
the rating at each point in time on the scale AAA to CCC+. We transform these ratings to a 0-4 scale. We
discard all the countries for which we have less than 22 observations and then compute the average for each
country. The third column of Online Appendix Tables 1 and 2 summarizes these values for developed and
emerging economies.
We want to emphasize that structural factors, including ‘political risk’, are incorporated into Fitch’s
ratings. The company develops its measure of political risk incorporating Governance Quality (Government
effectiveness, Rule of law, Control of corruption, Voice and accountability, and Business environment) and
Political Stability & capacity (Political stability and capacity, Legitimacy of regime, Conflict/war risk, Debt
payment record, Risk to economic policy). They also include past default decisions and other economic
indicators when rating a country’s default risk.

Construction of Figure 3: For each year, we we give a value of 4 to the countries with AA or above
ratings, 3.5 to A, 3 to BBB and BB, 2 to B, 1.5 to CCC, 1 to CC, 0.5 to C, and 0 to DDD or below. We
then average these out for all years in which we have observations and eliminate countries with less than 22
observations.

A.3 Fiscal Variables and Economic Aggregates

Variables are obtained from FRED, Eurostat, the World Bank, the IMF, OECD, Fitch Solutions, Statistical
Offices and Financial Ministries.

A.3.1 Variable Definitions

• c: is “Private final consumption, USD (2010 prices).” Computed as the sum of all household spending
on goods and services within the economy. It also includes spending by non-profit institutions serving
households (NPISHs). Adjusted to 2010 prices. Built from NIPA.

• GDP Deflator: Gross Domestic Product: Implicit Price Deflator, Index 2012=100, Annual, Season-
ally Adjusted. Downloaded from FRED, the variable name is GDPDEF.

• F B: is “Budget balance, % of GDP”. It is constructed as fiscal revenue less fiscal expenditure as a

Rev−Exp
percentage of GDP, F B = y , showing the net position or fiscal balance.

• g: corresponds to “Government final consumption, USD (2010 prices)”. The sum of all government
spending on goods and services within the economy. Also includes spending on government com-
munity services. Adjusted to 2010 prices. Built from NIPA.

39
 • P B: is “Primary balance, % of GDP.” Primary balance refers to the budget balance excluding interest
Rev−Exp+rb
payments on government debt as a percentage of GDP, P B = y .

• tax: “Total revenue, % of GDP. ” Fiscal revenue is government revenue or income through taxation,
non-tax revenue (including asset sales), financial resourcing, and capital receipts. This variable is
obtained from the OECD website.

• T otb/y: “Total government debt, % of GDP.” Total government debt refers to the sum of both internal
debt and foreign external debt issued by the government. Series obtained from the IMF.

• T B: is ‘Net exports of goods services, % of GDP.” It is the value of goods and services exported
or sold abroad minus those that are imported or bought abroad. Net exports = exports - imports and
X−I
TB = y . Variable expressed as a percentage of GDP. Built from NIPA.

• y: is “Real GDP, USD (2010 prices).”

Average values of private and public consumption to GDP, total debt to GDP, and tax revenues to GDP
are displayed in Table 5

Table 5: Auxiliary Moments by Economic Group

Moment Developed Emerging

Means
c/Y (%) 53.6 61.5
g/Y (%) 19.7 14.2
b/Y (%) 65.9 60.0
tax (%) 38.7 24.8

Cyclicality
ρ(g,y) 0.37 0.56
ρ(c,y) 0.86 0.94
ρ(PB/y,y) 0.28 0.04

A.3.2 Variable Manipulations

Series are available from 1990 to 2022, although this is an unbalanced panel. We discard all the countries
for which we have less than 22 observations in all variables except taxes (the series tends to have few obser-
vations for this variable). The list of countries in our sample satisfying the criteria can be found in Tables 1
and 2 of the Online Appendix.
We log-linearize: real GDP y, real private consumption c, and real public consumption g. That is, for
each country, we regress ln y on time t using ln yit = α + βt + it . Using the estimated parameters α̂ and β̂,
we compute the residual ˆit = ln yit − α̂ − β̂t and this is our de-tended measure for GDP. The same method

40
is used to de-trend private and public consumption. We use the resulting series to compute: (i) the standard
deviations: σ(g), σ(c), σ(y) and (ii) the cyclicalities ρ(g, Y ), ρ(c, y). The fiscal balance F B, the primary
balance P B, the debt-to-output ratio b/y, and the trade balance T B are neither logged nor de-trended, as
they are expressed as ratios (of output). When calculating their cyclicalities, only the cyclical component of
real GDP is used for consistency. The cyclicality ρ(g, y) corresponds to the estimated coefficient β̂ of the
OLS regression: git = α + βyit + it (where g and y are log-linearized). Cyclicalities for all other variables
are computed analogously.

B Theoretical Details
The policy proposal Φ(Ω) depends on whether the government has access to capital markets Ω = 1 or
it is excluded from them, Ω = 0. More specifically, Φ(1) = {τ, g, b0 , d, f1 , f2 , ..., fn } and Φ(0) =
{τ, g, f1 , f2 , ..., fn }, with b0 = 0 and d = 1 when Ω = 0. As explained in the main text, the proposer
delivers the same amount of political favors f to each member of the mwc and the remaining fp to him or
herself. Policy Φ yields welfare V P (Π, Φ) to the proposer, V I (Π, Φ) to members of the mwc and, V O (Π, Φ)
to those outside of it.

B.1 The proposer problem

In round k, the proposer chooses Φk to maximize
  f
p
max V P k (Π, Φk ) ≡ U c(s, Φk ), l(s, Φk ), g + + β Es0 J(Π0 , Φ0 |Φk )
Φk φ
s.t. V Ik (Π, Φk ) ≥ J k+1 (Π, Φk+1 ) (B.1)
k
fp = B(Π, Φ ) − (m − 1)f ≥ 0
τ, g, f ≥ 0.
The first constraint is an incentive compatibility constraint (IC) stating that the value of accepting the pro-
posal for members of the mwc, V Ik (Π, Φk ), must be at least as large as the value of rejecting it, J k+1 (Π, Φk+1 ).
The second condition ensures the feasibility of government policies, requiring that the budget balance ex-
cluding favors is sufficient to cover the cost of providing political favors to the coalition members. Favors
must be non-negative. The expected future value in the following period is J(Π0 , Φ0 |Φk ), conditional on the
current proposal being accepted.
We define the objects in the IC constraint next. Since members of the mwc obtain f in political favors,
  f
V Ik (Π, Φk ) = U c(s, Φk ), l(s, Φk ), g + + β Es0 J(Π0 , Φ0 |Φk )
φ
when a proposal is accepted. If the proposal is rejected, we move to round k + 1 where a new proposer
is elected with probability n1 . If the member of the current minimum winning coalition mwck becomes a
proposer in k + 1, he/she will receive V P k+1 (Π, Φk+1 ). Note that the states do not change, as all proposal

41
 m−1
rounds happen within a period. With probability n , a member of mwck belongs to mwck+1 , in which
n−m
case they get V Ik+1 (Π, Φk+1 ). With probability n a member of mwck is not in mwck+1 , which delivers
V Ok+1 (Π, Φk+1 ). Recall that anyone outside of the mwc obtains
 
V Ok (Π, Φk ) = U c(s, Φk ), l(s, Φk ), g + β Es0 J(Π0 , Φ0 |Φk ),

as in that case fi = 0. The continuation value of rejecting a proposal to any member of the current mwc,
J k+1 (Π, Φk+1 ), is computed by taking expectations over these three possibilities.

1 P,k+1 m − 1 I,k+1 m − n O,k+1

J k+1 (Π, Φk+1 ) = V (Π, Φk+1 ) + V (Π, Φk+1 ) + V (Π, Φk+1 ). (B.2)
n n n

Re-arranging the expression above,

  B(Π, Φk+1 )
J k+1 (Π, Φk+1 ) = U c(s, Φk+1 ), l(s, Φk+1 ), g k+1 + + β Es0 J(Π0 |Φk+1 ). (B.3)
φn

where Φk+1 is the optimal policy of the next round’s proposer (given any continuation value).

B.2 Simplified problem

The proposer will choose Φk such that the IC constraint holds with equality, V Ik (Π, Φk ) = J k+1 (Π, Φk+1 ).
This implies political favors f satisfy

f  
= J k+1 (Π, Φk+1 ) − U c(s, Φk ), l(s, Φk ), g − β Es0 J(Π0 , Φ0 |Φk ).
φ

Replacing this in the definition of fp (see B.1), we obtain

h   i
fp = B(Π, Φk ) − (m − 1)φ J k+1 (Π, Φk+1 ) − U c(Π, Φk ), l(Π, Φk ), g − β Es0 J(Π0 , Φ0 |Φk )

substituting fp in V P k (Π, Φk ), we can write the objective function as

  B(Π, Φk )
max U c(s, Φk ), l(s, Φk ), g + + β Es0 J(Π0 , Φ0 |Φk ).
Φk φm

Note that we dropped J k+1 (Π, Φk+1 ) from the expression. This is possible because it is unaffected by Φk ,
so it can be considered a constant for the optimization problem above.
On the equilibrium path, proposal rounds 2, ..., T does not occur. This reflects the assumption that any
coalition member accepts a policy mix that ensures that he/she is weakly better off than waiting for the
next round. We can thus write the proposer’s problem omitting k. Moreover, since c(.), l(.), and B(.) are
independent of f , the only policy variables to be determined in the problem above are—with a slight abuse

42
of notation—, Φ(1) = {τ, g, b0 , d} when the government has access to credit markets and Φ(0) = {τ, g}
otherwise. The proposer problem becomes
  B(Π, Φ)
max U c(s, Φ), l(s, Φ), g + + βEs0 J(Π0 )
Φ φm (B.4)
s.t. B(Π, Φ) ≥ 0,
with
  B(Π0 , Φ0 (Π0 ))
J(Π0 ) = U c(s0 , Φ0 (Π0 )), l(s0 , Φ0 (Π0 )), g 0 + + β Es00 J(Π00 ). (B.5)
φn
where Φ0 (Π0 ) denotes next period’s political Markov Perfect equilibrium outcomes. Up to this point,
we have not mentioned anything about the continuation values. The derivation of J(Π0 ) is similar to that of
J k+1 (Π, Φk+1 ). In the next period, the proposer retains proposal power with probability n1 , obtaining welfare
V P (Π0 , Φ0 (Π0 )). With probability m−1 I 0 0 0
n , it will be a member of next period’s mwc, receiving V (Π , Φ (Π )).
With probability n−m
n the current proposer is not in tomorrow’s mwc, which delivers V O (Π0 , Φ0 (Π0 )). The
proposer calculates J(Π0 ) as the expected value of V P (Π0 , Φ0 (Π0 )), V I (Π0 , Φ0 (Π0 )), and V O (Π0 , Φ0 (Π0 )).

C Extensions
C.1 Low β case
The sovereign default literature typically assumes that the government is significantly more impatient than
foreign lenders, imposing β(1 + r) < 1. Without this assumption, the government would not borrow in
the long run and default would never happen in equilibrium. Since r is a fixed risk-free rate, this restrictive
assumption often requires the value of β to be unusually low. In our model, institutional weakness and
corruption make the government endogenously impatient. As a result, we do not need restrictions on β,
which can take a standard value in macro (including β(1 + r) = 1). In this section, we compare our baseline
political-economy model against the standard sovereign default model with an impatient benevolent planner,
which we will refer to as the ‘low β’ case, evaluating their similarities and differences.
The competitive equilibrium given policy is the same in both environments because the domestic private
sector and international capital markets are identical. The government budget constraint is also the same, as
we consider identical policy instruments in both cases. The only difference lies in the maximization problem
of the policymaker. In our benchmark model, the proposer’s problem is given by eq. (6). The maximization
problem of the planner in the ‘low β’ case is, instead,
  B(Π, Φ)
J(Π) = max U c(s, Φ), l(s, Φ), g + + βL Es0 J(Π0 )
Φ φn (C.1)
s.t. B(Π, Φ) ≥ 0,

43
 Figure 16: Policy functions, benchmark (blue) vs ‘low β’ (red)
Low beta

b’ taxes

default

default

default
default
τ
b*(m)
𝑏 0
𝑏 𝑏 b 𝑏 𝑏 b

favors

default
g

default
𝑔

default
default
𝑓 =0
𝑏 𝑏 b 𝑏 𝑏 b

There are two main differences: First, if the planner chooses to give out political favors B(Π, Φ) > 0,
these need to be distributed among all groups (so there is an ‘n’ instead of an ‘m’ in the denominator).
Second, the discount factor of the policymaker is βL , lower than that of the lenders and domestic economic
agents, β. Even though we have included the corruption parameter, this takes the value of φ = 1 in the
benchmark calibration in both cases.
Because in the ‘low β’ case m = n (by construction), fi = 0 for all states of the world: a benevolent
planner does not find it optimal to distort the economy to engage in corrupt public practices, regardless of
how impatient it may be. As a result, B = 0 and we are in Case 2 of Section 3.4. Additionally, when the
country is in default, b = b0 = 0, implying that the choices of τ and g are independent of βL . The first-order
conditions are also independent of m, implying that, in default, taxes and public good provision are the same
regardless of whether we are in the benchmark economy or the ‘low β’ case. Figure 16 illustrates this point.
In that figure, the benchmark case is depicted in blue whereas the ‘low β’ case is shown in red.
Things are different when the government has access to international capital markets. While it is still the
case that there is no revenue devoted to political favors B = 0, the government is more impatient. This will
yield a steeper debt accumulation profile and default at a value of debt smaller than b2 . In equilibrium, this
will result in lower levels of sustainable debt as there is no additional incentive to defer default and enjoy
favors; that is the reason behind the inability to match spreads and debt/GDP ratio in those models. In our
benchmark economy, for low levels of debt, there is a jump in b0 but then a flatter borrowing profile for
intermediate levels of debt. This allows the government to borrow more before a default is triggered. As a
result, our model delivers higher debt/GDP ratios for similar spreads, making it easier to match the data.
To understand this further, we simulated both economies for a finite number of periods. For the ‘low-β’
model (red dashed line) we set the value of β to 0.82 to match the average value of spreads in the benchmark.

44
 Figure 17: Endogenous variables compared to a low-β economy

Spreads Default Borrowing/GDP Taxes

15 1 80 0.25

0.8 0.2
60
10
0.6 0.15
40
0.4 0.1
5
20
0.2 0.05

0 0 0 0
0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100

GDP Private cons/GDP Public Cons/GDP Favors/GDP

20 16 50
18
19 15 40

18 16 14 30

17 13 20
14
16 12 10

15 12 11 0
0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100

The initial debt for both economies is set to 0. We plot the time series of both economies for the same
realizations of the TFP shocks in Figure 17. The one for the baseline model is the blue solid line.
There is a stark difference between the two economies. The baseline model borrowing immediately
jumps to 60% of debt-GDP in period 1. The other one also rises, but at a much slower rate, and never
reaches the level achieved by the benchmark economy. Since spreads are the same on average, higher debt
in the benchmark economy requires higher taxes to repay the debt, as shown in Figure ??. Higher taxes also
imply lower GDP and lower private and public consumption. They are also lower in the benchmark as a
share of GDP because of positive favors paid out. In the ‘low-β’ economy favors are zero by construction.
Although default probabilities are the same in the two models by construction, our simulation generates
two default events for the baseline economy. These are defaults due to political shocks that are absent in a
‘low-β’ economy.
Long-run moments associated with the low β case are displayed in the second column of Table 6 (the
first column shows the results for the benchmark economy). The ‘low-β’ model sustains a lower level of
debt than the baseline economy, even though the average spreads are the same. This implies that at higher
levels of debt, the bond prices for the ‘low-β’ economy are lower, indicating a steeper bond price schedule.
The subjective discount factor, therefore, has a sharper impact on spreads compared to m. This explains why
in models of short-term debt and no possibility of political favors, it is difficult to match debt and spreads
simultaneously to the data. This difference in the slope of the bond price schedule is because, in models with
favors, curvature on the utility function makes favors easier to cut in a recession than increase taxes. If there
are no favors, taxes need to be raised or there has to be a default to smooth consumption.

C.2 Autocracies: the case of China

In our model, the number of groups with veto power m (e.g. in the mwc) affects not only how tempting
it is to engage in corrupt public practices but also the probability of being in the mwc in the future. As m
declines, the proposer becomes naturally more impatient, as this reduces the likelihood of getting fi next

45
 Table 6: Long-run moments: Model Extensions

Moment Benchmark Low β Autocracy

Institutions
R&A (mean) 4.77 20 1
Instability (St Dev) 0.78 0 0
Favors/GDP (mean) 4.58% 0 35.58%
Impatience β 0.99 0.74 0.99

Fiscal Policy
Spreads 7.11% 7.00% 0.004%
µ(tax) 17.69% 13.67% 43.2%
µ(g/Y) 13.1% 13.5% 7.2%
σ(tax) 0.044 0.023 0.018
σ(g)/σ(y) 1.48 1.67 0
µ(b/Y) 50.1% 21.7% 65.2%
ρ(TB,y) -0.37 -0.65 -0.35
ρ(FB,y) -0.31 -0.58 -0.34

Macro Outcomes
Sp. Crisis 9.85% 40.1% 3.59%
σ(y) 3.34% 4.04% 2.70%
σ(c)/σ(y) 1.29 1.48 0.46

period. While we interpret the case of m = 1 as an autocracy, there are examples in which a group remains
in power for long periods (e.g. China). In such a case, the proposer does not need to engage other groups
to implement policies and also knows that it will likely be in power for a long period. To study such an
environment, which is not covered in the main text, we simulated our economy assuming m = 1, but also
that the one group in power will be in power forever. That is, there is full persistence. We understand this is
unrealistic, but it could give us insights into how intermediate but highly persistent cases would work. This
case also explains how the political frictions in this paper are different from turnovers.
The third column on Table 6 displays the results. The first thing to note is that a significant part of the
budget is devoted to political favors, which reach a whooping 35% of GDP (versus 5% in the benchmark
case). Because the autocrat is endogenously more patient (e.g. there is no risk of losing power), default risk
is practically zero. This allows the government to better smooth taxes and public good provision, and sustain
a higher level of debt to GDP. The volatility of output and consumption is also smaller. The autocrat can
implement tax-smoothing policies but at the cost of frequently engaging in corrupt public practices.

46