Global Environmental Change 23 (2013) 1726–1736

Contents lists available at ScienceDirect

Global Environmental Change

journal homepage: www.elsevier.com/locate/gloenvcha

Agricultural risk management policies under climate uncertainty§

Jesús Antón a,*, Andrea Cattaneo b, Shingo Kimura a, Jussi Lankoski a
a
OECD Trade and Agriculture Directorate, Organization for Economic Cooperation and Development, 2 rue André-Pascal, 75016 Paris, France
b
FAO Agricultural Development Economics Division, United Nations Food and Agriculture Organization, Viale delle Terme di Caracalla, 00153 Rome, Italy

A R T I C L E I N F O A B S T R A C T

Article history: Climate change is forecasted to increase the variability of weather conditions and the frequency of
Received 6 February 2013 extreme events. Due to potential adverse impacts on crop yields it will have implications for demand of
Received in revised form 5 August 2013 agricultural risk management instruments and farmers’ adaptation strategies. Evidence on climate
Accepted 14 August 2013
change impacts on crop yield variability and estimates of production risk from farm surveys in Australia,
Canada and Spain, are used to analyse the policy choice between three different types of insurance
Keywords: (individual, area-yield and weather index) and ex post payments. The results are found to be subject to
Climate change
strong uncertainties and depend on the risk profile of different farmers and locations; the paper provides
Risk management
Adaptation
several insights on how to analyse these complexities. In general, area yield performs best more often
Uncertainty across our countries and scenarios, in particular for the baseline and marginal climate change (without
MaxMin increases in extreme events). However, area yield can be very expensive if farmers have limited
Agricultural insurance information on how climate change affects yields (misalignment in expectations), and particularly so
under extreme climate change scenarios. In these more challenging cases, ex post payments perform
well to increase low incomes when the risk is systemic like in Australia; Weather index performs well to
reduce the welfare costs of risks when the correlation between yields and index is increased by the
extreme events. The paper also analyses the robustness of different instruments in the face of limited
knowledge of the probabilities of different climate change scenarios; highlighting that this added layer of
uncertainty could be overcome to provide sound policy advice under uncertainties introduced by climate
change. The role of providing information to farmers on impacts of climate change emerges as a crucial
result of this paper as indicated by the significantly higher budgetary expenditures occurring across all
instruments when farmers’ expectations are misaligned relative to actual impacts of climate change.
ß 2013 Elsevier Ltd. All rights reserved.

1. Introduction distribution for farmers’ expected income. Farmers can adopt

several adaptation strategies in response to these changes.
Climate change affects the mean and variability of weather Adaptation through cropping pattern change can in some cases
conditions and the frequency of extreme events, which to a great ease the exposure of plants to critical higher temperatures
extent determines the variability of production and yields. The risk (Peltonen-Sainio et al., 2011). Also, changing planting time may
management response to these changes is part of farmers’ help avoid heat stress during the critical growth phases (Rötter et
adaptation strategies. Of relevance to risk management, the yield al., 2011). Another means is to introduce more diverse cultivars
of crops is limited to differing degrees by water availability and that differ genetically in their responsiveness to climate conditions
temperature depending on the agro-ecological zone. Examples of (Howden et al., 2007). As regards precipitation changes and water
impacts from extreme events and weather variability are the shortage, farmers can adjust by improving soil water-holding
significantly increased costs resulting from the increased frequen- capacity by adding crop residues or manure, or by adopting
cy of extremely hot days that cause heat stress in crops, or by the conservation tillage (Smith and Olesen, 2010; Känkänen et al.,
timing and amount of rainfall in a specific event. 2011). Altering fertiliser rates to maintain grain or fruit quality
Climate change affects the distribution of yields under a given consistent with the climate is another option.
set of management practices, which in turn affects the probability In a situation where farmers have no insurance, there should be
in principle a strong incentive to adapt to climate change
(Mendelsohn, 2010). Farmer reactions are more nuanced, however,
§
The views expressed in this article are the author’s and not necessarily those of and lack of insurance has shown that there is a lower likelihood of
the OECD, the FAO or its member countries. farmers adopting new technologies (Feder et al., 1985; Antle and
* Corresponding author. Tel.: +33 1 45 24 81 36; fax: +33 1 44 30 61 01.
Crissman, 1990), of lower investments (Skees et al., 1999), but also
E-mail addresses: [email protected] (J. Antón), [email protected]
(A. Cattaneo), [email protected] (S. Kimura), [email protected]
of greater diversification (Skees et al., 1999). Finally, even though
(J. Lankoski). certain practices may decrease risk once they are mastered by the

0959-3780/$ – see front matter ß 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.gloenvcha.2013.08.007
 J. Antón et al. / Global Environmental Change 23 (2013) 1726–1736 1727

farmer, the risk of crop-failure can increase initially because interactions and their policy relevance in three countries with
changing practices can be risky as farmers learn new technologies different characteristics. Of course the reality is even more
(Marra et al., 2003). complex with many more possible scenarios and types of farms
Risk management instruments, such as crop insurance and and with very limited information about their likelihood and
disaster assistance programme, and especially how they are frequency. An alternative strategy for our research would have
designed, will affect incentives to adapt (Collier et al., 2009). For been to simplify the problem by reducing its dimension. However,
example, traditional agricultural insurance (which makes an our analysis is already a simplification and our purpose is also
indemnity payment when the farm incurs a verifiable production using this example to illustrate the difficulties of analysing highly
loss) can help to manage production risk but it is known to be uncertain policy questions such as those related with climate
expensive and will diminish incentives to adapt to climate change. change. One important conclusion of the analysis in this paper is
Weather index insurance or area yield insurance, which do not that scientists and economists need to address the added
require on-farm verification, can help keep administrative costs uncertainty introduced by climate change if they are to give
down as compared to individual yield insurance, and they do not sound policy advice. The results of the analysis help to understand
discourage adaptation since indemnities are paid independently of the dimensions and trade-offs of the policy question, and possible
actual loss incurred by a policyholder. However, they are not a ways to get more robust policies.
means for structural adaptation. Farmers will incorporate any The paper is structured as follows. Section 2 presents crop
insurance subsidies or ex post disaster payments to their insurance instruments and ex post payments analysed in this
production decisions, which may favour insurance over crop paper. In Section 3, the data and the stochastic simulation model
diversification or other risk management and adaptation strate- used for analysis are presented. Results under a climate change
gies. scenario without more extreme events and misalignment (‘‘mar-
Insurance is sometimes used as a disaster assistance tool. It has ginal’’ climate change scenario) are discussed in Section 4 while
the advantage of a formal contract with the financial participation alternative climate change scenarios and robust policy choices
of farmers, the evaluation of damages and a relatively quick under strong uncertainties are, respectively, presented in Sections
payment of indemnities. But support to insurance has also its 5 and 6. Section 7 concludes.
drawbacks; in particular it can prevent the development of other
fully private solutions and it typically does not fully replace ex post 2. Risk management, insurance, and decision making under
assistance. climate uncertainty
Is climate change making insurance and other risk management
policies more needed? How can policy makers take such decisions It is common in the literature to segment risk in a way that
when the information about how different instruments would matches risky outcomes with different tools to transfer, pool or
perform under an uncertain climate is very limited? Building on manage risk. These layers are typically defined in terms of the
previous work examining risk management under climate change probability of occurrence and the magnitude of the losses, and,
(Collier et al., 2009; Heltberg et al., 2009) this paper is the first to therefore, the extent to which risk is catastrophic. The most
address, in an applied context, the risk and the uncertainties efficient instruments to manage risk may differ across layers.
introduced by climate change in the probability of weather events, Following OECD (2009), in the risk retention layer of frequent
and the role of perceptions of this uncertainty in terms of how risk events that cause relatively limited losses (normal risks), farmers
management policies would perform in practice. To investigate are best placed to manage this risk efficiently and smooth their
these issues we provide examples from Australia, Canada, and income; in the market insurance layer, risks are more significant but
Spain, which highlight that the appropriateness of a policy’s design less frequent and there is scope for farmers to use insurance or
depends on how climate change affects the risk structure facing other market options (marketable risks); finally, in the market
farmers. The paper also analyses the robustness of policy failure layer, risks generate very large and systemic (correlated)
instrument relative to current uncertainty on the impact of losses at low frequencies which makes them difficult to pool
climate change on variability of yields. through insurance (catastrophic risks). Government may decide to
The multidimensional, diverse and uncertain nature of the intervene after these catastrophic events, typically with ex post
problem of risk management under climate change makes it payments.
difficult to identify an optimal policy choice. First, there is strong Even though this three-layer approach is conceptually straight-
uncertainty about the quantitative impact of climate change on the forward, it can be challenging to implement in practice. The
variability of yields and production risks. Second there is boundaries between layers are not well drawn and the definition of
uncertainty about farmer’s perceived risks and their degree and catastrophic risks is determined by how government responds to
direction of adaptation response to climate change. Third there is a specific events and manages the demand for assistance. Subsidised
strong farm-specific or idiosyncratic component because different insurance systems are sometimes used to assist farmers after
farms have different risk profiles, are affected differently by disasters. When subsidised insurance becomes a tool to deliver
climate change and have different adaptation responses. Finally disaster assistance the boundaries between catastrophic and
the range of policy options is very large. In this paper we try to marketable risk can become blurred.
tackle each of these dimensions, respectively: analysing three Three types of crop production insurance are investigated and
climate scenarios (one standard or ‘‘marginal’’ climate change compared with ex post assistance: individual yield, area-yield and
scenario, one with higher frequency of extreme events, and a weather index insurance. They have different characteristics in
baseline with no climate change); looking at three different terms of data requirements, administrative costs, distribution of
responses by farmers (adaptation by diversification, structural risk, and its impact on farmers’ incentives to adapt to climate
adaptation and misalignment); characterising three types of farms change. Traditional individual-yield crop insurance makes an
according to their risk profile; and finally comparing four different indemnity payment when the farm incurs a yield loss. To pay
policy options. indemnities, the insurance provider must estimate the value of
This is a highly complex decision making framework where bio- yield loss for each farm and commodity that makes a claim. Hail
physical impacts of climate change interact with the human insurance is the most common peril insurance and is offered in the
response. The strategy followed in this paper is to use science and majority of OECD countries. Multiple-peril crop insurance, which
economic empirical analysis to try to provide insight on these covers losses due to multiple risks, is more complex and rarely
1728 J. Antón et al. / Global Environmental Change 23 (2013) 1726–1736

offered without government subsidies due to the high costs of loss system in Australia. Since the insurance premium is calculated
assessment under asymmetric information (Miranda, 1991; Good- based on systemic precipitation risk, all farmers pay the same
win, 1993). This includes adverse selection because farmers have insurance premium. One of the disadvantages of index insurance
more knowledge about their own distribution of losses than their is, however, the basis risk: the insured farmer could suffer a loss
insurer, and moral hazard because once insured the farmer may and not receive any or enough indemnity. The amount of basis risk
take less care of risks. To avoid moral hazard in loss assessment will depend on how well the chosen index maps individuals’ losses.
arising from asymmetric information, multi-peril crop insurances Large basis risk will diminish its interest because farmers will
usually have high deductibles (typically 30%, which is the requisite perceive index insurance as providing poor protection against risk.
for green box exemption of insurance subsidies under the WTO Assistance after a catastrophe is the typical response in OECD
agreements and is also assumed in the analysis presented here) countries to risks under the last layer covering market failure for
such that a small yield loss is not covered. Since individual yield extreme events. A standard policy response consists of triggering ex
insurance is tailored to the yield risk of farms, in the analysis that post catastrophic/disaster payments. For instance, a flat rate
follows the fair insurance premium is calculated for each payment per unit of land is paid for losses beyond the threshold.
representative farm. The payment is determined by the area of In the modelling exercise, it is assumed that ex post payments are
land that the farmer insures and producers cannot insure more triggered when all systemic yield variables fall below the 40
area than the one they plant. Climate change is likely to affect the percentile (all crops are affected simultaneously) and a lump sum
distribution of yields and the cost of multiple-peril crop insurance. per unit of land, equal to the expected indemnity of area yield
In addition, the complexity of delivering such insurance greatly insurance, is paid. The administrative costs of these ex post payments
increases administrative costs compared to single peril insurance. are presumably low and assumed to be zero in the model.
An alternative crop insurance scheme is area-yield crop Due to severe uncertainties related to climate change impacts
insurance, in which both indemnities and premiums are based and farmers’ adaptation behaviour policy makers must take their
on the aggregate yield of a geographical area. The indemnity equals decisions taking into account this type of ambiguity. Early work by
the difference in value, if positive, between the area yield and some Keynes (1921) and Knight (1921) brought this kind of management
predetermined critical yield level. Participating producers in a uncertainty to the core of economic thinking. Bringing it to the core
given area would receive the same indemnity per insured unit of of policy making has proved more difficult and it is only recently
land, regardless of their own crop yield, and all would pay the same that there have been attempts to seriously tackle this problem
premium rate (Miranda, 1991; Barnett et al., 2005). Area-yield (Gollier, 2011). Ben-Haim and Yakov (2006) provides a very
crop insurance offers advantages over individual-yield crop technical response to cover information gaps in these types of
insurance because it reduces information asymmetries. Adminis- problems. Etner et al. (2010) propose two theoretical alternatives
trative costs are reduced since information regarding the to manage decision making under ambiguity: (1) a standard
distribution of the area-yield is generally available and verification Bayesian treatment consists of assigning objective or subjective
of individual farmers’ production histories would no longer be probabilities to events and then applying preferences using
required. Moreover, because the indemnities would be based on expected values or an expected utility approach and (2) a non-
area yield rather than the producer’s yield, a producer could not Bayesian approach with a formal definition of ambiguity and
significantly increase his indemnity by changing production different degrees of ambiguity aversion. This latter approach may
practices. Thus, under an area-yield insurance programme, moral take several forms depending on the structure of beliefs and
hazard is significantly reduced. For this reason, area-yield priorities about probabilities and the confidence that the decision
insurance usually does not apply a deductible. However, area- maker has in these beliefs. Clarke (2008) used classical decision
yield insurance is less effective if the idiosyncratic or basis risk is theory ideas of min–max and min–max regret to determine robust
high, that is, yield risk of an individual farmer has less correlation decision rules that facilitate adaptation to Climate Change.
with the systemic risk measured by the average production shock Prospect theory (Kahneman and Tversky, 1979) defines individual
across farms in an area. Under changing climate conditions, area- risk behaviour with respect to a reference point and argues risk
yield insurance would have the advantage of maintaining behaviour differs for losses and gains, but this theory has not been
incentives to adapt. applied to policy making. In this paper we apply three different
Weather index insurance is another option that attempts to criteria: a Bayesian approach, and two non-Bayesian: max–min
overcome the moral hazard and adverse selection problems. It and satisficing. The max–min maximises the worse result that can
provides an indemnity based on values obtained from a weather occur under each policy choice, while the satisficing criteria
index that serves as a proxy for losses rather than on the individual establishes a minimum performance defined, for instance, as a
losses of policyholders. The underlying index is based on an minimum proximity to the best performing policy. In this paper we
objective measure, such as rainfall or temperature that exhibits a use a range within 35% from the best performing policy in terms of
strong correlation with the variable of interest, usually crop yields. the cost effectiveness indicator.
A threshold in the proxy variable marks the point at which
payments begin. Once the threshold is reached, the payment 3. A model to analyse insurance demand and diversification
increases incrementally as the value of the index worsens. The
payment rate is independent of the actual loss incurred by a 3.1. A stochastic simulation model
policyholder and, therefore, there is basis risk retained by the
farmer, which is the residual non-systemic component of risk. This section presents the farm-level stochastic simulation
Weather index insurance has some of the same advantages as area- model to analyse how climate change, by affecting the mean,
yield insurance over standard individual yield crop insurance, such variability, and covariance of yields and weather events, affects
as the reduction of moral hazard and adverse selection, and not the appropriateness of different risk management strategies. The
discouraging adaptation to climate change (Collier et al., 2009). usefulness of micro-based stochastic modelling is that it can
Furthermore, it usually requires only that a weather station address the extent to which climate change can have an impact
generates the necessary index. All these advantages translate into on the use of crop insurance and other risk management
lower administrative costs. In this paper weather index insurance strategies (e.g., production diversification by farmers), and the
is calibrated based on regional precipitation risk in Canada and outcome of government’s ex post disaster assistance (see OECD,
Spain, but based on the amount of water inflow into the major river 2009, 2011).
 J. Antón et al. / Global Environmental Change 23 (2013) 1726–1736 1729

The model is tailored to the risk exposure and strategic weather index insurance and ex post payment. In the simulations,
environment revealed by panel data from a small sample of farms individual yield and area yield insurance are made available for
from three countries/locations: 402 non-corporate crop farms in wheat, barley and oilseeds in Canada and Australia, and wheat and
South West Saskatchewan (CAS/Agristability data from ‘‘Where barley in Spain. Only one insurance instrument or ex post payment
Canada delivers’’), 78 broadacre farms in Australia (Broadacre farm is available for each policy scenario.
survey) and 12 crop farms in the province of Valladolid in Spain Since the specification of crop production is neutral to the farm
(National FADN database). The original surveys were not designed size in this model, the representative farm is assumed to cultivate a
to be panels. Therefore a sample of a few farms that remained in fixed area of farmland and allocate land between available
the survey for at least six years were selected from each survey in products in each country. The degree of crop specialisation is
order to estimate individual production variability. The farms in represented by the coefficient of variation of market revenue per
the samples from Canada and Spain are located in the same unit of land. A higher coefficient of variation of crop revenue is used
province and are all within a radius of 100 km, while the farms in as an indicator of less use of crop diversification (adaptation)
Australia are spread across different states. The results from each of strategies and built on a lower diversification index. If the farmer
these countries in this paper should be interpreted as results from adopts a strategy with less diversification and specialises in a
the specific location and sample. specific crop, the diversification index declines because the farmer
The model introduces a set of risk management strategies that allocates more land to crops that generate a higher return with
are relevant in these countries; namely production diversification higher variability. The initial value of so defined production
and three types of insurance: individual crop yield insurance, area diversification index is set at 100 and the change of the
yield insurance and weather index insurance. In addition, the diversification index is expressed as (1) times the percentage
model introduces an ex post payment triggered by a systemic yield change in the coefficient of variation of market return.
loss. The model assesses the producer’s participation in the risk Administrative costs play an important role in demand for
market and its impacts on farm welfare and low income risk. insurance by farmers. Since different insurance instruments have
The representative farms in the samples in the three locations different administrative costs, it is necessary to make assumptions
are assumed to maximise their expected utility, allocating land about their relative costs in order to compare across instruments.
among major products and choosing the degree of insurance These costs estimates need to be comprehensive and are calculated
coverage simultaneously. The model adopts the power utility as the difference between total premiums (paid by farmers or by
function which assumes constant relative risk aversion (CRRA). The government subsidies) and total average indemnities across
advantage of this model is that it treats risk management strategies several years (sum of fair premiums). To this end, the administra-
as endogenous, allowing the interaction between policies and tive costs of insurance are expressed as a percentage on the top of
farmer’s decision to be analysed. the fair premium. These loading factors are: 5% for weather
(rainfall) index insurance, 10% for area-yield insurance, and 30% for
ðp̃  vÞð1rÞ individual yield (2%, 10% and 31% in Spain using information from
Uðp̃ þ vÞ ¼ (1)
1r the Spanish Agricultural Insurance Agency ENESA). This assump-
tion is meant to quantify in an approximate way the impact of loss
where the utility (U) depends on the uncertain farm profit (p̃) and
assessment and payments under different insurance instruments.
initial wealth (v; r stands for the degree of constant relative risk
The government programme to subsidise insurance premiums
aversion (CRRA), which is assumed to be equal to 2 in all our
is modelled as subsidising a fixed percentage of administrative
simulations. Given the expected utility calculated in the optimisa-
costs (95% in Australia and Canada, and 90% in Spain). By definition
tion model, the certainty equivalent farm income is used to
under farmer risk aversion, if insurance were priced with a fair
compute the farmer’s welfare for a given level of risk aversion.
premium and no administrative costs, all land would be insured.
The uncertain profit is defined as the crop revenue less the
This is not observed in reality because farmers face other types of
variable cost for crop production plus the net transfer or benefit
costs related to uncertainty and asymmetric information, and they
from a given risk management strategy. Since the crop specific cost
do not fully insure crop yields even though the administrative costs
is not available in the data, the uncertain variable cost (c̃) is not
are fully covered by the government. The model assumes that
crop specific. The crop specific production cost adjustment factor
farmers perceive that subsidies do not cover any part of the fair
(ci) is calibrated for each crop so that the initial land allocation
premium. This is a reduced form for modelling the additional non-
observed in the data becomes the optimum. The model assumes
observable costs of insurance. The extent to which area-yield
that total land input is fixed and is allocated between crops or
insurance or weather index insurance is attractive to individual
livestock production. Given the Monte-Carlo draw of 1000 price,
farmers depends largely on the correlation between their yield risk
yield, revenue and variable cost combinations, the model
and indices (regional average yield and weather index).
maximises the expected utility with respect to area of land
allocated to each commodity and the level of insurance coverage.
3.3. Calibration of yield and other risks
X
n X
p̃ ¼ ½ð p̃i  q̃i  ci Þ  Li  þ OR  ðL̄  Li Þ  c̃ þ gð p̃i ; q̃i ; lÞ (2) The modelling work in this study is data intensive and based on
i¼1
different sources. Two types of data were collected and used for the
Where P̃i is uncertain output price of crop i; q̃i is uncertain yield of calibration and simulation exercises. First, six to seven years of
crop i; c̃ is uncertain variable cost; ci is the cost adjustment factor of production data were collected from the samples of individual
crop i; Li area of land allocated to crop i; OR is the residual revenue; g farms to calculate individual variability of yields, and the mean
is the insurance net indemnity or government ex post disaster values for calibration. All farms in the samples produce the main
payment; and l is the level of insurance coverage decided by farmer. crops (wheat and barley in the three countries plus rapeseed in
Canada and Australia) in all years; there is a residual composite
3.2. Quantifying adaptation and accounting for transaction product that can differ in composition across countries and years).
costs of policies It is presumed that farmers do not use insurance contracts in the
baseline years. Second, meteorological data, most commonly on
The model introduces and calibrates four government policy rainfall, was used to design location specific weather index
strategies: individual yield insurance, area-yield insurance, insurance (see Annex).
1730 J. Antón et al. / Global Environmental Change 23 (2013) 1726–1736

Table 1
Characteristics of each cluster of farms.

Country Australia Saskatchewan (Canada) Valladolid (Spain)

Risk cluster Low Medium High Low Medium High Low High

Number of farms in cluster 11 42 25 220 144 38 4 8

Area of operation 936 2262 483 380 319 257 87 91
Wheat yield
Mean (tonnes per ha) 3.8 2.0 1.6 1.4 1.9 2.3 3.6 2.6
Coefficient of variation 33.8 36.9 61.8 26.9 31.7 45.3 31.0 31.5

In order to define the profiles of the calibrated farms, the yield risk, while the basis risk will remain the same. This will
hierarchical analysis is applied so as to group farmers according to involve a change in the yield distribution for different crops in the
the similarity of risk. The grouping exercise begins with as many following ways.
clusters as sample farms, but it merges clusters until only one In general, the mean yield decreases with climate change;
cluster remains by applying Ward’s minimum variance criterion. under the model assumption of decreasing absolute risk aversion,
The variables to characterise the cluster are selected according to this would imply greater absolute risk aversion. The variance of
the risk profile of wheat production: the level and variability of yields could increase; but the net effect on income variability also
wheat yield. Table 1 describes the characteristics of three clusters depends on correlations between yields of different crops, and
of farms (low, medium and high risk farm). price yield correlations; for instance, more correlated risks
The model assumes that the yield risk of a given set of (systemic) and lower price/yield correlations (natural hedge) will
commodities in farm ‘‘i’’ can be expressed by a random vector with imply even riskier scenarios. The probability of very extreme
two additive components: events will increase; in statistical terms this could imply an
increase in negative skewness (third central moment) or lower
q̃i ¼ s̃ þ b̃i (3) kurtosis (fourth central moment). For weather index insurance the
The first component s̃ denotes the systemic part of yield risk, correlation of yields with the chosen weather index may change,
affecting all farms in the same area. This variable is built as the also affecting the idiosyncratic yield risk faced by insured farmers.
average production shock for a given farm type. The second This would affect the relative demand and performance of weather
component b̃i denotes basis risk for that particular farm and is the index insurance vis-à-vis other risk management tools.
residual non-systemic or idiosyncratic component of risk. Table 2 outlines the relevance for risk management of the
The matrix of correlations of systemic risk is also constructed as different aspects of climate change relating to agriculture as
an average of correlation across risks. The main characteristic of the outlined above. The hypotheses concerning the quantitative
weather index is the parameter ui expressing the correlation between implications of climate change on the distribution (mean and
the index and the systemic component of yields s̃. Climate change standard deviations) is that only the systemic yield shocks are
will affect s̃ and may also affect ui depending on whether or not the affected by changes in the distribution. It is also assumed that
weather variables capture the limiting factors affecting yields. climate change does not modify the correlation between the
The Monte-Carlo simulation of systemic risk is based on a weather index and the idiosyncratic yield risk; in the case of
truncated normal distribution. The truncated points are selected as an increase in the variance of yields, this implies an increase
maximum and minimum values of the sample data. In Australia in the relative importance of the systemic component of yield
and Spain, the correlation of yields across crops is much higher risk.
than in Canada, implying less scope for crop diversification as a Table 3 summarises the quantitative changes due to climate
risk management strategy. In Australia the correlation between change according to the empirical literature and are the basis for
crop yield and livestock revenue is very small, indicating that the simulations of all the scenarios in this paper. Mean yields tend
diversification between these two operations is a potentially to decrease across the board, but the standard deviations do not
important risk management strategy. The Monte-Carlo simulation always increase and in Canada the decreasing trend dominates.
assumes a normal distribution of idiosyncratic yield risk that is The numbers in this table assume there is structural adaptation of
correlated across crops. farming practices, but three different degrees of adaptations are
explored and discussed in this paper. The numbers for the means in
4. Risk management under climate change Table 3 are adjusted downwards in the non-structural adaptation
scenarios using Howden et al. (2007) estimate of an 11.1% increase
4.1. Calibration of climate change risk in yields due to adaptation.
The numbers in Table 3 quantify the exogenous impact of
Climate change will modify the distribution of risks. It is climate change on the distribution of yields. They are based on
assumed here that it will affect the systemic component of the specific scientific references that are provided at the bottom of the

Table 2
Impact of different forms of climatic changes on the relevant parameters of the model.

Climatic change indicator Impact on yield distribution Impact on link between weather index and yield (u)

Increase in CO2 concentration Will increase the mean for some plants, Correlation should not change (since SD also not changing)
covariance unaffected
Number of days above or below Will generally affect mean yield May decrease correlation
a minimum or maximum temperature
Cumulative rainfall Can affect both mean and variance of yield None (because weather index expressed as cumulative rainfall)
Increase in variability of rainfall Can affect both mean and variance of yield May decrease correlation
Extreme events Will affect skewness of yield distribution, since Depends on whether the index is correlated with extreme event
extreme events tend to lower yields
 J. Antón et al. / Global Environmental Change 23 (2013) 1726–1736 1731

Table 3
Impacts of climate change on the distribution of yields of main crops (% changes excluding adaptation).

Australiaa Canada (Saskatchewan)b Spain (Valladolid)c

Mean Standard deviation Mean Standard deviation Mean Standard deviation

Wheat 7.2 10.3 3.0 2.0 1.8 110.5

Barley 20.0 0.0 10.0 17.0 7.3 89.3
Rapeseed 19.9 6.1 13.0 2.0
Sources: aLuo et al. (2010), van Gool and Vernon (2006), bZhang et al. (2011), and cGuereña et al. (2001). Howden et al. (2007) estimates used to correct for adaptation.

table. In most locations there is a reduction in mean yields, but not increases farmers’ welfare. Ex post payments are not effective
in all locations there is an increase in the variability as measured by in the simulations for Canada because the systemic component
the standard deviations (e.g., rapeseed in Australia, wheat and of risk is relatively small.
barley in Saskatchewan). These estimates are used in the rest of the Climate change in Spain will cause very large increases in the
paper for simulations of ‘‘marginal’’ climate change scenarios that variability of yields and this variance dominates the impact on
do not include adaptation. expected yields. Insurance demand increases significantly for all
instruments. Area yield insurance performs well in reducing
4.2. The impacts of a marginal climate change scenario farmers’ risk. Uptake of area yield insurance can be particularly
high due to high correlation with systemic yield. Climate change
The impacts of climate change on farmers’ response to has a strong effect on non-irrigated farms, with large increases in
insurance can be expressed in terms of the demand for insurance, insurance demand, but this does not translate into improvement of
the diversification in the farm, and the cost effectiveness of the the cost-effectiveness of insurance.
budgetary expenditure to reduce farmers’ risk (Table 4). The
baseline with no climate change in Table 4 is the result of 5. Incorporating climate change uncertainty: combining
simulating farmers’ response to new insurance that was not biophysical and behavioural scenarios
available before. The same insurance simulation is then
undertaken under a new scenario of (marginal) climate change The literature provides consistent information on whether
defined by the changes in yields distribution estimated in Table climate change will increase or decrease average yields for a crop in
3. The impact of climate change on insurance demand is a given region; however, little information is available concerning
represented by the increase in the percentage points of land how variability will be affected. There is a general consensus that
insured. Other impacts of climate change represented in Table 4 in many regions variability of weather conditions will increase, but
are the percentage change in the diversification index and cost there is a lack of information as to how this would affect the
effectiveness. The production diversification index has been probability distribution of crop yields. It is particularly relevant to
used to capture the change in variability of profits due to farmer risk management in agriculture whether the change in variability
choice in the composition of commodities. Two possible is distributed evenly around the mean or whether the probability
indicators of cost effectiveness are used: one measures the of extreme events increases in the form of stronger and more
increase in the lowest 10 percentile income of the farm and it is frequent yield losses.
applied to Australia, and the other is based on the gains on the The difficulty of updating the probability distribution of yields
certainty equivalent (CE) due to reduced variability of income in the presence of climate change may push actors to conjecture
and is applied to Canada and Spain. probability distributions based on historical experience that do not
In Australia, climate change is expected to have a strong impact take into account climate change. This will be referred to here as a
in reducing avearage yields, which dominates the effects on misalignment in expectations when farmers, government or
variability. The highest demand of insurance corresponds to insurance companies have insufficient information to evaluate
individual yield insurance, but it also incurs the highest budgetary the change in systemic risk brought about by climate change and
costs. All of the risk management instruments have a strong behave as if this distribution had not changed.
impact on specialisation in the most profitable (and risky) Table 5 outlines a typology of plausible climate scenarios.
activities in Australia, which increases the net variability of Three scenarios on climate conditions are considered. The
income. Climate change does not alter this crowding out of baseline scenario, used for comparison purposes, assumes
diversification. For the cost-effectiveness performace indicator, business as usual with no climate change. The marginal climate
area yield insurance performs best in terms of increasing low change scenario assumes that the distribution of yields will be
income outcomes of farmers for low and high risk farms, but can modified by climate change according to the most reliable
have negative effects on medium-risk farms. Ex post payments numbers in the empirical literature (see Section 3). Finally, a
have lower crowding out effects than most insurance programmes, more radical climate change scenario assumes that the potential
as can be seen by the values of the diversification index; however, impacts of more frequent extreme climate events affect
they have low cost-effectiveness performance. Higher uptake of agriculture. These two latter climate change scenarios are
area yield insurance and weather insurance by medium risk farms combined with three different sub-scenarios that reflect the
is due to higher yield risk correlation with systemic yield risk and behavioural assumptions on farmers: (i) adaptation by diversifi-
precipitation risk. cation, where farmers simply change the mix of crops in response
In Canada, climate change will not have systematic effects in to climate, (ii) structural adaptation that incorporates changes in
the performance of different instruments because this marginal crop yields from changing cultivars or management practices,
scenario implies a less risky environment in terms of standard and (iii) misalignment in expectations where farmers and other
deviations of yield. Area yield insurance is the best performing agents maintain unchanged their prior about the probability
instrument, in terms of welfare gain, in both the baseline and in distribution of yields.
the marginal climate change scenarios. Demand of this type of The marginal climate change scenario was defined in the
insurance is the highest and the programme generates reduc- previous section according to the best available literature. The
tions of variabiliy expected by the farmers. Individual yield extreme events scenario is developed on the general result that
insurance can be effective in reducing variability which ‘‘extreme events will be more likely to occur under climate
1732 J. Antón et al. / Global Environmental Change 23 (2013) 1726–1736

change’’. However, there is no quantitative information about the

(per $ expenditure)
scope of these extreme events under climate change. This proves
Impact of policy
index of diversification on CE welfare
the structural uncertainty behind any quantitative estimation of
the implications of climate change on the variability of
agricultural production. The extreme events scenarios in this

1.26

0.31
0.02

0.02

0.05

0.01

0.01

0.01
paper are obtained by adding an additional stochastic extreme
systemic shock that affects the lowest 25 percentile of the yield
distributions. The realised Monte-Carlo values of the 25 lowest
(percentage change)

percentile of the original distribution of systemic yields under

Impact of CC on

climate change are multiplied by a random number extracted

from a uniform [0,1] distribution. An empirical distribution is
then calibrated. Its lowest values are halved on average.
1.4

4.3

5.3

1.6

5.6
18.0
7.0

7.0
Concerning the behavioural scenarios, adaptation by diversi-
fication occurs endogenously in the simulations. Conversely the
Impact

structural adaptation scenario is based on Howden et al. (2007),

Valladolid (Spain)a

with no CC of CC

4.7

31.7

48.9

43.6
0.0

81.0
% of land insured

which estimated that, on average, under conditions of a rain

decrease and 2–4 degrees of temperature increase, adaptation
benefits imply an increase of 11.1% in yields. In the absence of
Baseline

82.5

17.7

59.3

55.1
19.0
100.0

better information, it is assumed that both structural and non-

structural adaptation scenarios imply the same shock in the
standard deviation of yields. Finally, the misalignment scenario
(per $ expenditure)

simply assumes that farmers’ make their decisions as if

Impact of policy
on CE welfare

distribution of production had not changed with the climate.

It is difficult to disentangle the interactions between the
impact of insurance on the incentive to adapt and the degree of
0.1
0.1

0.1
0.2
0.1

0.1
0.1
0.0

0.0

0.0
0.0
0.0

adaptation adopted by farmers. In the modelling framework used

for this analysis, insurance will have an effect on the crop mix
(percentage change)

chosen endogenously (adaptation by diversification); however,

Impact of CC on

the yield effects associated with other forms of adaptation

diversification

(structural adaptation) are assumed to be exogenous to the

index of

insurance demand decisions.

In Spain low risk farms are irrigated farms, while high risk are non-irrigated. No medium risk farm was considered.
0.6

0.4
0.7

0.1
5.9
4.8

1.1

1.9

0.0
0.0
5.0

6.0

Why do different scenarios matter? The results for the

Saskatchewan (Canada)

Canadian sample are used to illustrate the impact of misalign-

Impact

ment in expectations and extreme climate change scenarios.

44.1

15.5

13.2
1.9
36.4

1.2

5.1

6.2
0.5
with no CC of CC

Table 6 shows that under marginal Climate Change the

budgetary costs of some programmes increase substantially if
% of land

expectations are misaligned (relative to case with no structural

Baseline
on 10 lowest income insured

adaptation, but with correct expectation of impacts), ranging

60.6
21.5
58.7
31.6

58.3

47.4

23.8
46.5
37.7

from a 55% increase under individual yield insurance to a nearly

seven-fold increase in budgetary costs for area yield insurance. As
expected, extreme events with misalignment imply even larger
percentile (per $
Impact of policy

increases in budgetary costs for all the different programmes.

expenditure)

This is because agents are not able to foresee the large costs
associated with extreme events variability. On average the costs
2.1
2.2

3.7
1.1
0.4
0.3
0.3

0.6

0.4
0.1
0.4
6.0

are multiplied by six. Area yield insurance, the best performing

instrument under marginal climate change in the Canadian
diversification

sample, has the highest increase in terms of budgetary costs

Impact of CC
Change in indicators due to Climate Change (marginal Scenario).

(percentage
on index of

under the extreme events misalignment scenario.

change)

Weather index insurance has the lowest increase in costs

13.5
1.3
4.2

5.8

1.8
3.3
4.3
0.5

0.1
0.9
0.6
3.0

under extreme events and misalignment. This is due to two

factors: lower levels of risk covered by indexes and the
with no CC (change in the

assumptions that the increase in systemic variability affects

Impact of CC

percentage)

directly the area yield more than the aggregate weather index.
The weather index used for Canada is the cumulative rain fall in
1.9
22.9

1.2
1.1
12.3

19.3
19.5
14.3
0.0

April–October, which is not always able to capture the full range

of extreme events compared to other insurance based on yields.
% of land
Australia

In terms of the welfare gains from risk reduction, they are

Baseline
insured

Individual yield insurance

77.1

19.7

54.8
48.7
80.2
100.0
Medium risk farm 100.0

19.0

51.0

increased for all programmes under the extreme events

Weather index insurance

misalignment scenario because they cover additional risks. But

Area yield insurance

these increases do not match the much higher increases in

Medium risk farm

Medium risk farm

Medium risk farm

Ex post payments

budgetary costs. The cost efficiency of weather insurance

High risk farm

High risk farm

High risk farm

High risk farm

Low risk farm

Low risk farm

Low risk farm

Low risk farm

increases significantly, because the cost of this programme does

not explode. Therefore the best policy response to climate can
depend on the scenario, in particular for extreme scenarios on
Table 4

which budgetary costs of measures in place can explode, and

a

there is a need to analyse a set of plausible scenarios.

 J. Antón et al. / Global Environmental Change 23 (2013) 1726–1736 1733

Table 5
Typology of climate change scenarios and farmer response.

Description Climate scenarios

Baseline Marginal Extreme events

(no climate climate
change) change

Behavioural Business-as-usual Expresses how policy instruments would function without Baseline
response climate change
Adaptation by Based on expected impact on yields assuming farmers can only adapt Marginal Extreme
diversification by diversifying among existing varieties
Structural Expected impact on yields based on the literature, assuming farmers Marginal with Extreme with
adaptation can switch to crop varieties that reduce impact of climate change adaptation adaptation
Misalignment Farmers make production decisions based on their historical Marginal with Extreme with
(no adaptation) experience and therefore do not take into account the increase in misalignment misalignment
systemic risk (no adaptation)

Table 6
Budgetary costs of different policies under three scenarios in Saskatchewan, Canada (CAD 1 000).

Baseline Marginal climate change Extreme events

No struct. adapt. With misalignment With misalignment

No policy – – – –
Individual yield insurance 68 179 227 399
Area yield insurance 82 80 630 1070
Weather index insurance 36 32 95 88

Ex post payment 56 41 199 308

Percentage of triggering 4 5 14 17
Budgetary cost when triggered 867 840 1419 1925

In terms of ex post payments, their costs are on average lower (whether they will undertake structural adaptation and whether
than those for area yield insurance. However the budgetary outlays their expectations will be misaligned). It is therefore imperative to
in the years in which extreme events occur can be very high. This take ambiguity into consideration when providing policy advice, as
makes this type of payments less attractive from a perspective of opposed to trying to identify ‘‘optimal’’ as if these probabilities
budgetary stability. were known with certainty.
First, a standard Bayesian approach to ambiguity will be used
6. Risk management policy decisions under severe uncertainty assigning a priori probabilities to the different scenarios: 30% for
the baseline with no climate change, 60% for marginal climate
The marginal climate change scenario with adaptation by change and 10% for extreme events; the probabilities of
diversification that was analysed in Section 4 is only one possibility behavioural response are set at 60% for structural adaptation,
among several scenarios. The seven combinations of scenarios 30% for adaptation by diversification and 10% for Misalignment.
discussed are all plausible outcomes according to current empirical Then two alternative simple non-Bayesian approaches will be
evidence, and are based on the limited literature available on the considered: a satisficing and a MaxiMin criteria. Both represent
impact of climate change on both the mean and variability of different degrees of ambiguity aversion by the decision maker and
yields, and the role of adaptation. The optimal policy choices are they respond to the idea of providing robust choices. A robust
different for different scenarios as shown in Table 7. The available choice is defined as one that performs ‘‘reasonably well’’ under a
knowledge is not sufficient to match each scenario with a variety of different plausible scenarios even though it does not
scientifically estimated probability. These severe uncertainties necessarily provide the most cost effective policy. Other criteria are
are often called ‘‘ambiguity’’ due to lack of information about the also available in the literature (Etner et al., 2010), but all involve
likelihood of the different scenarios occurring (baseline, marginal some a priori beliefs about probabilities, confidence on these
climate change and extreme events), and how farmers will behave probabilities, and/or ambiguity aversion of the government.

Table 7
Optimal policy choice under different scenarios.

Country case Baseline Marginal climate change Extreme events

Adaptation by Adaptation Misalignment Adaptation by Adaptation Misalignment

diversification
diversification

Australia Area yield Ex post payments Area yield Area yield Ex post payments Ex post payments Ex post payments
Lowest income percentile
Canada Area yield Area yield Area yield Weather index Weather index Weather index Weather index
CE gaina
Spain Area yield Area yield Area yield Area yield Area yield Area yield Area yield
CE gaina
Note: Results in this table are shown as ‘‘average’’ across the farm clusters in each country. Aggregation of 10% lowest income results in Australia is made by pooling all farms,
while a weighted average of CE gain is applied in the samples of Canada and Spain.
a
Gains in Farmers’ Certainty Equivalent of welfare.
1734 J. Antón et al. / Global Environmental Change 23 (2013) 1726–1736

Table 8
Robust policy choice across climate change scenarios.

Country case Bayesian optimum Satisficing MaxiMin

Australia Ex post payment – Ex post payment

Lowest income percentile
Canada Area yield Weather index Weather index
CE gain
Spain Area yield Area yield Area yield
CE gain

In Australia, area yield insurance still performs well under some measured by yield variability of the main crops in continental
of the marginal change scenarios, but it becomes too costly under Spain, but that yield variability on the Canadian Prairies will likely
the extreme events scenarios (Table 7). Ex post payments perform be reduced for crops such as wheat and barley. In Australia, the
better under extreme events because their cost have a more evidence varies with some commodities showing increased
moderate increase and are relatively effective for low income production risk and others showing reduced risk.
objectives when there is a big share of risk that is systemic as is the In the case of climate change scenarios that only marginally
case of Australia. They also have smaller crowding out effects on change the risk environment, the demand for insurance increases
diversification. only marginally (except in Spain) and the crowding out of
Area yield performs best under baseline and most marginal adaptation by diversification remains. Individual yield insurance
climate change scenarios in Canada. But its performance is tends to be very costly for governments, while weather index
dramatically reduced under extreme events because the costs of insurance and ex post payments are cheaper on average. Ex post
this programme increases much more than for other programmes. payments are highly variable and can be extremely high in some
On the contrary the costs of weather index do not explode under years. On the whole, however, if climate change only slightly
extreme events, while the correlation between yields and weather modifies yield variability then the new risks associated with
index improve. Weather index becomes the best performing climate change do not seem to be an appropriate justification or
instrument under the extreme events scenarios. basis on which to develop new risk management policies.
Area yield insurance outperforms all other instruments across The analysis in this study goes beyond a standard climate change
all scenarios in Spain. It becomes the most balanced instrument in scenario and investigates policy making under strong uncertainty.
combining reasonable effectiveness and reasonable costs. First, two different climate change scenarios are examined: standard
There are two pertinent caveats that nuance these results. First, climate change versus a situation with numerous extreme events.
insurance types are a continuum from individual to area yield and to Second, three different behavioural responses by farmers are
weather index, with an increasing basis risk. Area yield in the Spanish examined: no response due to limited information on how climate
case is not that different from individual yield because there are very change affects yields (misalignment); adaptation by diversification;
few farms in each area. The boundaries between them are artificially and structural adaptation. The strong uncertainties about the
defined as sharp. Second, ex post payments are likely to have climate change scenarios and behavioural responses (referred to as
transaction costs due to the difficulties of identifying the systemic ‘‘ambiguities’’) are organised in seven scenarios. Estimating the cost-
events when they happen. This could make them less attractive. effectiveness of each measure in each scenario is a complex
The Bayesian solution is obvious in the case of Spain because the quantitative exercise and the results are not always intuitive and
best performing policy is always area yield insurance for all differ across countries and farm types. These results highlight the
scenarios (Table 8). However in the case of Australia, the Bayesian usefulness of scenarios to deal with strong uncertainties.
probability assumptions privilege the instruments that perform best In policy problems subject to high uncertainty and diversity
under extreme events (ex post payments); and in the case of Canada such as risk management in agriculture under climate change,
the Bayesian probability weights favours instruments that perform there is not such a thing as an ‘‘optimal policy’’. In order to
better under marginal climate change (area yield). More robust non- contribute to these areas of decision making that are becoming
Bayesian approaches to decision making, which provide more more relevant over time, scientists have to try to find ways to
conservative (higher) weights to worst cases, such as the MaxMin, analyse this complexity without taking the shortcut of over-
privilege instruments that perform better in the worst scenarios of simplifying the problem. This paper provides new insights on how
extreme events: ex post payments in Australia and weather index to deal with these uncertainties and identify robust policies, which
insurance in Canada. The satisficing criterion does not always will vary depending on the country context and the information
provide a single solution, like in the case of Australia. available to farmers of the impact of climate change.
At least in some countries, a more cautious policy approach Area yield performs best more often across the countries and
would avoid area yield insurance, mainly because it can potentially scenarios presented here, in particular for the baseline and
become very expensive under climate change and misalignment. marginal climate change. But Area yield can be very expensive
However a possible policy response would also include the under extremes scenarios. In this case: ex post payments perform
development of research and information strategies that impede well to increase low incomes when the risk is systemic like in
systematic errors or biases in the perceptions about future risks. Australia; Weather index performs well to reduce the welfare costs
of risks when the correlation between yields and index is increased
7. Conclusions by the extreme events.
The possibility of extreme events and misalignment scenarios
Few insights into the impact of climate change on the variability significantly changes the policy decision environment. The
of crop yields are provided in the available literature, although analysis of government’s best response to this ambiguity is very
there is relatively more empirical information of its impacts on the challenging and requires a significant change in the approach.
level of average yields. The impact of climate change differs Rather than identifying optimal policies, the definition and
depending on the location. For example, the most reliable sources understanding of the plausible scenarios is a core part of the
to date reveal that climate change will increase production risk as analysis. Governments may seek the implementation of ‘‘robust’’
 J. Antón et al. / Global Environmental Change 23 (2013) 1726–1736 1735

policies that are not optimal under any scenario but that may be able Weather index insurance
to respond well to different environments and avoid very bad
outcomes, particularly under extreme events and misalignment. The Weather index insurance is calibrated based on regional
misalignment scenarios are characterised by high budgetary precipitation risk in Canada and Spain, but based on the amount
expenditure and low adaptation practices. Other policy initiatives of water inflow into the major river system in Australia. The
that focus on information and training can help prevent the design follows a standard weather index contract, in which an
misalignment of risk perceptions. This study shows that it is insurance indemnity is triggered at a certain value of the
technically feasible to define plausible scenarios and implement weather index and increases to a full indemnity when the index
robust criteria in response to strong climate change uncertainties. exceeds a certain limit.1 Since the insurance premium is
An additional policy challenge to prevent misalignment is how to calculated based on systemic precipitation risk, all farmers
communicate complex information and uncertainties about climate pay the same insurance premium and there is no upward limit
change and policy implications to policy makers, farmers, insurers for insurance subscription. The yield loss is evaluated based on
and other stakeholders. But this is beyond the scope of this paper. the expected price level of wheat in Australia and Canada, and
barley in Spain.
Acknowledgements

We gratefully acknowledge the valuable help from Agriculture Appendix B. Supplementary data
and Agri-Food Canada (AAFC), Ministerio de Agricultura, Alimen-
tación y Medio Ambiente of Spain, and the Australian Bureau of Supplementary data associated with this article can be found, in
Agricultural and Resource Economics and Sciences (ABARES) to the online version, at doi:10.1016/j.gloenvcha.2013.08.007.
access and use the respective survey data.

References
Appendix A. Calibration of risk management strategies
Antle, J.M., Crissman, C.C., 1990. Risk efficiency, and the adoption of modern crop
varieties: evidence from the Philippines. Econ. Dev. Cult. Change 38 (3) 517–
Individual yield insurance 537.
Barnett, B.J., Black, J.R., Hu, Y., Skees, J.R., 2005. Is area-yield insurance competitive
with farm-yield insurance? J. Agr. Resour. Econ. 30 (August) 285–301.
Individual yield insurance is tailored to the yield risk of individual
Ben-Haim, Yakov, 2006. Info-Gap Decision Theory: Decisions Under Severe Uncer-
farms. The indemnity is paid in the case where the individual crop tainity. Elsevier Ltd..
yield turns out to be below the insured level of yield (30% of Clarke, H., 2008. Classical decision rules and adaptation to climate change. Aust. J.
deductible), which means that the farmer needs to plant these crops Agr. Resour. Econ. 55, 487–505.
Collier, B., Skees, J.R., Barnett, B.J., 2009. Weather Index Insurance and Climate
to insure yield risk. To avoid moral hazard and adverse selection Change: Opportunities and Challenges in Lower Income Countries. Geneva Pap.
effects, the model assumes the perfect insurance market so that risk Risk Ins.: Issues Practice 34 (July) 401–424.
neutral insurance companies offer a crop insurance contract at the Etner, J., Jeleva, M., Tallon, J.-M., 2010. Decision theory under ambiguity. J. Econ.
Surveys, http://dx.doi.org/10.1111/j.1467-6419.2010.00641.x.
price equal to the expected value (fair insurance premium) with no Feder, G., Just, R., Zilberman, D., 1985. Adoption of agricultural innovations in
administrative cost or government subsidy. The fair insurance developing countries: a survey. Econ. Dev. Cult. Change 33 (2)
premium is calculated for each representative farm. The payment is 255–298.
Gollier, C., 2011. Le calcul du risque dans les investissements publics. Centre
determined by the area of land that the farmer insures and producers d’Analyse Stratégique (Rapports et Documents No 36).
cannot insure more area than they plant. The forward price applied Goodwin, B.K., 1993. An empirical analysis of the demand for multiple peril crop
to calculate the insurance premium and indemnity is set at the insurance. Am. J. Agr. Econ. 75, 425–434.
Guereña, A., Ruiz-Ramos, M., Dı́az-Ambrona, C.H., Conde, J.R., Mı́nguez, M.I., 2001.
expected price level. Assessment of climate change and agriculture in Spain using climate models.
X   Agron. J. 93, 237–249.
q̃
g1 ¼ p fi  qhi  Li  Max 0; bqi  i  Heltberg, R., Siegel, P.B., Jorgensen, S.L., 2009. Addressing human vulnerability to
qhi climate change: toward a ‘no-regrets’ approach. Global Environ. Change 19 (1)
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Indeminty receipt 89–99.
   Howden, S.M., Soussana, J.F., Tubiello, F.N., Chhetri, N., Dunlop, M., Meinke, H.M.,
q̃
ð1 þ g Þ  p fi  qhi  Li  E Max 0; bqi  i (4) 2007. Adapting agriculture to climate change. Proc. Natl. Acad. Sci. 104, 19691–
qhi 19696.
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} Keynes, J.M., 1921. A Treatise on Probability. Macmillan, London.
Insurance premium payment Knight, F., 1921. Risk, Uncertainty and Profit. Houghton Mifflin, New York.
Kahneman, D., Tversky, A., 1979. Prospect theory: an analysis of decision under risk.
where pfi is the forward price of commodity i; Li is the area of land Econometrica 47 (2) 263–291.
for commodity i which farmer insures its yield; Qhi is the historical Känkänen, H., Alakukku, L., Salo, Y., Pitkänen, T., 2011. Growth and yield of spring
average yield of commodity i; bqi is the proportion of yield insured cereals during transition to zero tillage on clay soils. Eur. J. Agron. 34, 35.
Luo, Q., Bellotti, W.D., Hayman, P., Williams, M., Devoil, P., 2010. Effects of changes
for commodity i; g is the net of administration cost of insurance in climatic variability on agricultural production. Climate Res. 42 (2)
and subsidy to insurance premium. 111–117.
Marra, M., Pannell, D.J., Abadi Ghadim, A., 2003. The economics of risk, uncertainty
Area-yield insurance and learning in the adoption of new agricultural technologies: where are we on
the learning curve? Agr. Syst. 75 (2–3) 215–234.
Mendelsohn, R., 2010. Agriculture and Economic Adaptation to Agriculture. COM/
Area yield insurance is based on systemic yield risk. Insurance TAD/CA/ENV/EPOC(2010)40/REV1.
premium is calculated by crop from the systemic yield risk
parameter so that all farmers pay the same insurance premium. 1
In Canada, weather index insurance is triggered if the cumulative rainfall index
The model assumes no deductible so that the insured farmer between 1 April and 31 October falls below 250 mm in the region. If the cumulative
receives an indemnity when the systemic yield falls below the precipitation index falls below 150 mm, insurance compensates the full value of
expected level. Unlike individual yield insurance, farmers do not yield loss. The indemnity is linearly reduced between the precipitation index
need to actually plant the insured crops, but the model assumes between 150 and 250 mm. In Australia, the triggering point is set at annual inflow to
Murray system of 9000 Gl. Complete yield loss is assumed below the 1000 Gl level
the insured area cannot exceed total area of land. The forward price of Murray inflow. The weather index insurance in Spain is triggered when the
applied to calculate the insurance premium and indemnity is set at cumulative rainfall index between 1 April and 31 October falls below 200 mm in the
the expected price level. region and assumes complete yield loss below 150 mm of cumulative rain fall.
1736 J. Antón et al. / Global Environmental Change 23 (2013) 1726–1736

Miranda, M.J., 1991. Area yield crop insurance reconsidered. Am. J. Agr. Econ. 73, 223– Skees, J., Hazell, P., Miranda, M., 1999. New approaches to crop yield insurance in
242. developing countries. In: EPTD Discussion Paper No. 55. International Food
OECD, 2009. Managing Risk in Agriculture: A Holistic Approach. OECD, Paris. Policy Research Institute, Washington, DC.
OECD, 2011. Managing Risk in Agriculture: Policy Assessment and Design. OECD, Smith, P., Olesen, J.E., 2010. Synergies between the mitigation of and adaptation to,
Paris. climate change in agriculture. J. Agr. Sci. Cambridge 148, 543–552.
Peltonen-Sainio, P., Hakala, K., Jauhiainen, L., 2011. Climate induced overwintering van Gool, D., Vernon, L., 2006. Potential Impacts of Climate Change on Agricultural
challenges for wheat and rye in northern agriculture. Acta Agr. Scand. Sect. B Land Use Suitability: Barley. Technical Report. Department of Agriculture,
Soil Plant Sci. 61, 75–83. Government of Western Australia, South Perth.
Rötter, R.P., Palosuo, T., Pirttioja, N.K., Dubrovski, M., Salo, T., Fronzek, S., Aikasalo, R., Zhang, A., MacGregor, R., Fuqun, Z., Becker, R., Williams, J., Cihlar, J., 2011. Will
Trnka, M., Ristolainen, A., Carter, T.R., 2011. What would happen to barley projected climate change have adverse impacts on land resources for the
production in Finland if global warming exceeded 4C?. A model based Canadian Prairies – a spatially explicit assessment. (unpublished manuscript).
assessment. Eur. J. Agron. 35, 205–214.