Producer Insurance and Risk Management
    Options for Smallholder Farmers




                                                                                                                          Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
                                             Vincent H. Smith*


Many smallholder families are exceptionally prone to potentially catastrophic decreases in
their incomes and access to food. Over the past decade, therefore, policy makers and econo-
mists have increasingly focused on potential mechanisms for expanding risk management
strategies available to those families. Commercially provided weather-based index insur-
ance products, perhaps partially funded by subsidies, have been of particular interest
because of their apparent potential to provide payments to smallholder families when they
are most in need of help. However, the empirical evidence from a wide range of studies indi-
cates that, absent relatively substantial subsidies, small holder farmers will not purchase
commercially priced index products or even “all risk” products where payments are tied to
the farm’s crop losses. There are three important reasons why this is the case. First, small-
holder farmers already have many ways of managing their risks, including informal com-
munity-based initiatives, on-farm production decisions and off-farm work. Second, index
insurance schemes are subject to considerable basis risk; families often do not receive an
index insurance indemnity when they experience a substantial crop loss on their farms.
Third, the fixed costs of delivering crop insurance to smallholders make such coverage
expensive. The potential market for weather index insurance therefore may be limited to
insuring relatively large groups of farmers, either directly or indirectly though providing
micro finance and other lending institution with coverage against widespread loan defaults
associated with catastrophic events like major droughts. Alternatively, weather indexes
could simply be used to more accurately target emergency aid. Index insurance, Risk
management, Food insecurity, Basis risk. JEL codes: D61, O13, Q18

Introduction
In low-income countries, many families are desperately poor and exceptionally
prone to substantial and potentially catastrophic decreases in their incomes and

The World Bank Research Observer
# The Author 2016. Published by Oxford University Press on behalf of the International Bank for Reconstruction and
Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
doi:10.1093/wbro/lkw002           Advance Access publication March 29, 2016                                  31:271–289
access to food. Within rural communities in those countries, families with small-
holder farms typically face extraordinary production risk challenges in dealing with
income volatility and food insecurity. Those challenges are even more severe for the
very poorest families in those communities, and not simply because as farmers they
face periodically severe production risks. Often those families farm poor quality land
on which crop yields are especially volatile because the household has neither the
financial or other assets, nor the human capital to acquire, maintain, and develop
the productivity of the land. A lack of human capital and, potentially, concerns about
financial risks may also limit those families’ abilities to adopt new crop and livestock
production practices that have the potential to increase yields and farm incomes.




                                                                                                         Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
   Over the past decade, therefore, economists who focus on rural poverty problems
at the World Bank, governmental agencies, and throughout the global development
economics community have increasingly been concerned with understanding how
very poor farm households use risk management strategies to mitigate and cope
with short-term food and income security crises resulting from crop failures and
livestock losses (Barnett et al, 2008; Carter, 2008; Binswanger-Mkhize, 2010). A
natural extension of this focus has been on the potential role of governments and
international agencies in improving and expanding the array of risk management
strategies available to those very poor farmers (for example, Boucher et al, 2005;
Hazell and Hess, 2010; de Janvry et al, 2014).
   The policy and infrastructure-related risk management and risk-coping strategies
mentioned above include the following: smoothing household incomes and food
consumption through a wide range of agricultural insurance schemes; improving
access to finance from formal institutions (most typically microfinance institutions);
facilitating local coping mechanisms (e.g., through informal credit systems at the
village level in rural areas); providing improved technologies (including new crop
varieties, improved access to irrigation, etc.) and targeted emergency disaster aid;
and subsiding crop storage capacity at the household level, as well as providing
savings incentives and emergency cash transfers. These strategies also involve tradi-
tional programs that have long been recognized as central to economic development
and increasing smallholder families’ farm incomes, for example, increasing general
education and health services, developing roads, improving communications and
water management and supply systems, and expanding agricultural R and D invest-
ments and agricultural extension services. In addition, these strategies include pro-
grams that mitigate the adverse effects of domestic tax and other policies on income
volatility and agricultural productivity.
   This paper examines the viability, scope, and availability to poor farmers of alter-
native risk management strategies for addressing short-term food security crises,
with a focus on formal and informal insurance schemes. We then consider the po-
tential costs and benefits of alternative government and international aid policies
intended to expand and enhance the array of risk management strategies that very

272                                     The World Bank Research Observer, vol. 31, no. 2 (August 2016)
poor farmers can use. A central issue in comparing alternative approaches is the op-
portunity cost of government and foreign aid funds, some of which over the past
decade have been allocated to various crop insurance projects (see, e.g., Marenya,
Smith, and Nkonya 2014) and, by straightforward extension, what constitutes effi-
cient as opposed to inefficient policy (Alston and Hurd 1990).



Incentives for Smallholder Farm Household Risk Management
and Coping Mechanisms




                                                                                         Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
Binswanger-Mkhize (2010) has long argued that, as a group, smallholder farmers
understand the income and food security risks they face, and without any help
from policy experts, address those risks in many important and often subtle ways.
Thus, as Binswanger-Mkhize (2010) and Wright and Hewitt (1994) have pointed
out, the costs and benefits of the often informal mechanisms for risk management
and risk coping available to farmers play a major role in determining the real oppor-
tunity costs they face with respect to many of the risk management tools that have
recently been considered by policy analysts and policy makers, such as formal crop
and livestock insurance policies. However, as Binswanger-Mkhize (2010) and
Miranda and Farrin (2012) have recently suggested, many of the analyses that have
presented arguments for the introduction of new risk management strategies have
not adequately considered the portfolio of existing risk management tools available
to farmers and their costs and benefits. The consequence has been an overemphasis
on the argument that, because the production risks that smallholders face have po-
tentially catastrophic consequences, they must be extremely risk averse and, there-
fore, willing to expend considerable resources to access specific risk management
tools such as crop insurance.1
   Both Wright (2014) and Binswanger-Mkhize (2010) have emphasized that what
matters for households, including smallholder households, is their consumption of
goods and services, not just the income or food supply they obtain from any specific
crop, and that the purpose of managing risk for most of those families is to smooth
consumption. In particular, as some economists have also recently begun to reem-
phasize in several contexts, in their risk management and risk coping strategies those
households are likely to place particular emphasis on mitigating the consequences of
potential shocks that could result in “extreme left tail” events (Goodwin 2014).
   These types of catastrophic events have been given considerable emphasis by
some researchers as sources of disincentives for new technology adoption by very
poor smallholders.2 Such catastrophic events may also result in substantial reduc-
tions in asset holdings that undercut the household’s future income stream. Selling
livestock in order to obtain food is one example of what has been described as a
poverty trap syndrome related to asset depletion (Carter 2008). The problem is

Smith                                                                              273
often compounded by the fact that the assets have to be sold in depressed markets.
Livestock, for example, will bring a much lower price in an environment where feed
is scarce because of drought. Various forms of crop insurance have been proposed as
a ( partial) solution to mitigating such poverty trap problems (Skees et al. 1999;
Barnett et al. 2008).
    In standard utility-based models of an economic agent’s insurance decision, the
optimal strategy is for the agent to perfectly smooth consumption (Borch 1990). In
practice, whether the setting is a rich household in a rich country or a desperately
poor family in a poor country, almost no one does that for a variety of reasons. First,
such a goal is often infeasible because insurance companies are unwilling to offer




                                                                                                          Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
contracts that perfectly smooth income (deductibles exist for a reason—often to mit-
igate moral hazard incentives). Second, it is also typically sub-optimal from the per-
spective of households that face liquidity and other constraints on access to
financial capital markets. In addition, de Janvry, Dequiedt, and Sadoulet (2014)
have recently pointed out that investing in individual crop insurance protection can
be welfare-reducing for smallholder families who also rely on informal risk sharing
arrangements within a group of families because of free riding problems. These
authors also note, however, that as a group those families may benefit from jointly
purchasing index-based insurance against adverse events such as drought.



Risk Management and Risk Coping Strategies
Taking actions to mitigate and reduce the frequency and effects of extreme left tail
events is often feasible and all but the most inveterate gamblers (extremely risk loving
individuals) are likely to make non-trivial short-term and, where feasible, longer-
term investments to mitigate or avoid extreme or moderately extreme adverse out-
comes. The extent to which longer-term investments in risk mitigating strategies will
be made by smallholder farmers critically depends on the permanence of the property
rights they believe they possess with respect to the land they farm.
   In relation to strategies for mitigating risk, it may be useful to draw a taxonomic
distinction between the strategies that smallholder farm households use to address
production, income, and food security risks on an ex ante basis, here called risk
management strategies, and what, in a reactive sense, happens ex post, that is, after
a severe adverse event occurs, here called risk coping mechanisms. The distinction
is somewhat arbitrary because the potential for ex post responses at the household,
extended family, and community levels is taken into account by those households
when they make ex ante plans for addressing risk.
   For smallholder households, short-term (within a crop season) risk management
investments include the following: on-farm enterprise diversification, diversification
of household labor between on-farm and off-farm income-generating activities

274                                      The World Bank Research Observer, vol. 31, no. 2 (August 2016)
(including working for other entities and self-employment/entrepreneurial activi-
ties), crop loss mitigation, spatial diversification of individual crop and livestock en-
terprises, and explicit self-insuring strategies such as food storage (Buschena,
Smith, and Di 2005) and holding livestock as potentially marketable assets (Barnett
et al. 2008).
   Crop loss mitigation strategies may include the use of chemical inputs (e.g., pesti-
cides and herbicides) and the adoption of new seed varieties that are more drought
resistant but also more expensive. Such strategies may also include short-term in-
vestments in irrigation, either through access to an existing village canal-based irri-
gation system or well, or the use of family labor to carry water from available




                                                                                           Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
sources to where it is needed. Investments in communication systems (e.g., cell-
phones) can also help smallholder farmers to mitigate price risks associated with
marketing their crops and livestock, enabling them to understand the range of price
offers available in the markets in which they sell their crops and, as a result, sub-
stantially reducing price volatility and the frequency of low price offers for their
crops (Jensen 2007).
   Short-term risk coping strategies (what smallholder farmers may be able to do
after they experience an extreme or relatively extreme adverse event) include bor-
rowing or receiving gifts from extended family members, borrowing from lenders in
the local community who offer informal access to credit (e.g., relatively wealthy
farmers in the village), borrowing from semi-formal organizations such as “iddirs”
(groups of households that, within a local community, jointly provide semi-formal
insurance for burial expenses incurred by members of the group), and borrowing
from formal financial credit institutions (e.g., microfinance agencies or banks).
These strategies also include selling assets such as livestock and whatever farm
equipment they may have to obtain food, often (as noted above) in depressed “fire
sale” markets for those assets.
   Longer term (multi-season) investments in risk management strategies by
smallholder farmers may include improvements in land quality (through soil con-
servation practices such as agroforestry and terracing), development of a well or a
site-specific irrigation system (e.g., through damming a local stream), partici-
pation in developing and maintaining a long term community irrigation system,
investments in livestock herds as a longer-run enterprise diversification strategy,
and resource-sharing arrangements among villagers to obtain risk reducing
inputs like agricultural chemicals at a lower cost. Such investments may also
include longer term social arrangements; for example, through marriages that
create incentives for expanded risk sharing among extended families by creating
family ties with households in other locations (effectively, creating spatial diversifi-
cation of crop yield and income risk within the horizontally extended family), and
maintaining ties with siblings who have established their own nuclear families
elsewhere.

Smith                                                                               275
Informal Insurance Programs
Many family and community-based risk coping and risk sharing schemes are, in
effect, informal insurance programs. In those schemes, implicitly or explicitly,
households agree to help one another in times of trouble, as among the iddirs in
Ethiopia that provide informal burial insurance (Dercon et al. 2014). As a result, in
those informal arrangements, indemnities in the form of “gifts” are likely to be pro-
vided when a potentially catastrophic decrease in an individual household’s income
occurs, but often payment of those indemnities is uncertain because implicit obliga-
tions to help a household in times of trouble may not be fulfilled. So too are the




                                                                                                         Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
analogs of premium payments, and for the same reason: the “premium” is the recip-
rocal promise to provide resources to the other families in the group when they are
in need of help. For example, within an extended family there may well be an expec-
tation that, if one household experiences a catastrophic loss event, then other
households in the extended family who are not affected, or much less severely affect-
ed, will “step up to the plate” and help out.
   Importantly, these schemes tend to have very low financial overhead costs, al-
though non-cash expenditures of social relationship capital may not be so low. For
example, the household, which is likely to face severe liquidity constraints, does not
have to make any cash outlays in the form of pre-loss premium payments to partici-
pate in the informal insurance scheme, as the “premium” is the promise to help out
after an adverse event has occurred if another family is in trouble. The schemes are
also extremely flexible, and do not typically involve binding legal commitments, al-
though the requirement to conform to well-understood customs established by the
community may be just as binding in many circumstances.
   However, these arrangements are perhaps most likely to fail when help is most
needed because the catastrophic event that severely curtails the household-in-
need’s real income (e.g., a region-wide drought) also has similar adverse impacts on
all participants in the informal risk sharing arrangement. Some policy analysts
have argued that such problems provide an obvious rationale for more formal insur-
ance programs in which policies are guaranteed by reinsurance through commer-
cial reinsurance companies and/or the domestic government, and/or aid agencies,
while not using an explicit market failure argument for their subsidization.
   Essentially, the argument is that such policies are complements to the informal
risk management strategies that already exist and will therefore be valuable and
commercially viable. Others have also suggested that formal insurance is more
likely to be taken up by individual farmers when leaders of informal insurance ar-
rangements or local communities are well-educated about formal insurance prod-
ucts (e.g., Dercon et al. 2014, and Gine  ´ et al. 2008). In that sense, commercial or
subsidized index insurance products may be somewhat complementary to informal
schemes, especially if they address risks common to the participants in the group

276                                     The World Bank Research Observer, vol. 31, no. 2 (August 2016)
(e.g., drought) while the informal schemes address risks that are idiosyncratic to in-
dividual households (Mobarak and Rosenzweig 2013).
   However, de Janvry, Dequiedt, and Sadoulet (2014) show that insurance against
a common shock purchased by a member of a common group can and is perhaps
likely to lead to free riding by other members of that group. Thus, for the individual,
such insurance may have a negative value. The authors therefore suggest that
group insurance policies that exclude the feasibility of free riding are likely to be
preferred and, from a commercial perspective, are likely to be offered at a lower per
hectare cost because they spread fixed costs over a larger insured area.




                                                                                          Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
Commercial and Other Formal Insurance Programs
Formal crop and livestock insurance programs offered on a commercial basis have
been proposed as a viable risk management tool for individual farmers by a wide
range of policy makers and analysts (see, e.g., Barnett et al. 2008; Vedenov and
Barnett 2004; Skees et al. 1999). However, as Hazell, Pomerada, and Valdes
pointed out in 1986, absent substantial government subsidies, farm-specific all risk
yield- or revenue-based insurance programs have never been offered on a commer-
cially successful basis. The reason is that, given the premium rates the private insur-
ance sector requires to offer all risk crop insurance policies, smallholder farmers do
not buy them. Binswanger-Mkhize (2010) and Miranda and Farrin (2012) essen-
tially draw similar conclusions for weather index-based insurance programs, even
though the administrative costs associated with index insurance products are lower
and adverse selection problems less extensive.3 As Smith and Glauber (2012) and,
almost twenty years before them, Wright and Hewitt (1994) noted, citing Patrick’s
(1988) seminal study of Mallee Valley wheat farmers in Australia, such is also the
case in developed countries for both index insurance and individual farm yield-
based insurance programs.4

Index Insurance
Over the past ten years, as discussed above, both insurance practitioners and some
economists have argued that some form of crop or livestock insurance products can
be offered successfully on a commercial basis to smallholder farmers because the
required administration and operations loading factors are much lower than for
all risk insurance (see, e.g., Hazell and Hess (2010); Vedenov and Barnett 2004;
and Skees et al. 1999). Further, these advocates have argued that access to such in-
surance products will mitigate the negative impacts of adverse events on both the
families’ degree of food insecurity and their willingness to adopt new technologies
that on average increase their real incomes but also increase the volatility of their
crop yields.

Smith                                                                               277
   The infeasibility of implementing all risk or multiple peril crop insurance con-
tracts that provide indemnities based on a farmer’s actual yields and yield histories
for major subsistence crops to smallholder farmers has been almost universally ac-
knowledged (see, e.g., Binswanger-Mkhize 2010; Miranda and Farrin 2012), just
as they are widely acknowledged to be infeasible in developed countries in the
absence of substantial subsidies (Kramer 1983; Wright and Hewitt 1994; Goodwin
and Smith 1995; Smith and Glauber 2012). The loading factor—the amount in
excess of the premium needed to cover expected indemnities—required by a private
insurance company to cover their administrative and operations costs is widely
viewed as simply far more than almost all farmers are willing to pay. Self-insurance




                                                                                                         Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
and other risk mitigating strategies are less costly and more efficient.
   One reason for the costliness of all-risk crop and livestock insurance contracts is
that monitoring moral hazard behaviors (sometimes called hidden action behav-
iors) is perceived to be expensive. Automobile insurance and property/casualty in-
surance, which on the supply side are typically competitive markets, have been
described as comparable lines in the insurance business with respect to moral
hazard effects (e.g., see Smith and Goodwin 2011). Loading factors for such lines of
business are typically in the range of 40 –50% of expected indemnities. In the
context of smallholder farmers, another factor is that the fixed costs associated with
issuing and managing such policies are relatively high and have to be spread over
an area of crops that is very small.
   However, it should be noted that a similar overhead fixed cost problem exists
for index insurance programs marketed to individual smallholders farming one
or two hectares of land (Boucher, Barham, and Carter 2005). As discussed
above, that is one reason why de Janvry et al. (2013) argue that group-based
index insurance contracts are more viable. In addition, participation in all risk
crop insurance programs is likely to be relatively low because of adverse selection
(hidden information) effects. The reason is that premium rates cannot be tailored
to individual farms’ actual loss experiences because of inadequate data on yields
and ancillary farm-specific information (Goodwin 1993; Smith and Baquet
1996).
   These are not new insights. Halcrow (1949), in a developed country context, laid
out the moral hazard and adverse selection issues associated with “all risk” crop
insurance at the farm level over sixty years ago, and the issues were reexamined by
Miranda (1991) in his seminal analysis of area (county) yield-based index insur-
ance contracts. Both Halcrow and Miranda argued that area yield index contracts
that cover dozens of farmers in a specific region (say a county or a sub-county grid
area that is 20 kilometers by 20 kilometers) rarely create incentives for moral
hazard behaviors and are likely to substantially mitigate adverse selection problems.
In a development context, however, reliable historical data on area yields are typi-
cally not available on an adequate basis to estimate premium rates and develop

278                                     The World Bank Research Observer, vol. 31, no. 2 (August 2016)
actuarially viable contracts. Hence, it is natural to think of using an index based on
a variable (or set of variables) that is closely related to area and farm crop yields in
developing an insurance policy.
   Weather is a major factor in determining crop yields and the availability of
forage. Weather indexes have thus become the focus of much of the work on poten-
tial agricultural insurance products. As rainfall can be measured relatively easily
and inexpensively, it has received considerable attention in pilot projects and theo-
retical and simulation analyses as the basis for, or the sole component of, a weather
index.5 Satellite images of plant growth have also been proposed as the basis for a
vegetation growth index. In the United States, such indexes are used in a heavily




                                                                                            Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
subsidized area-based insurance contract to provide coverage against forage loss in
regions with relatively low levels of annual rainfall. However, vegetation indexes
based on satellite images can only effectively be utilized in areas where thick cloud
cover occurs relatively infrequently.

Basis Risk in Index Insurance
No matter how refined the weather index or any other area-based index may be, as
Miranda (1991) emphasized, crop insurance based on an area index is subject to
what he called basis risk. Basis risk exists in two forms, both of which derive from
the underlying problem that the index on which indemnity payments are based is
not perfectly positively correlated with an individual farm’s actual loss experience.
The first type of basis risk is that the farm may receive an indemnity when it has not
suffered any substantial loss. This sort of basis risk is not especially important in the
context of enabling a smallholder household to cope with a severe adverse change
in their real income and food security from crop and livestock losses. The second
type of basis risk is important; the farm may not receive an indemnity when it does
experience a substantial loss.
   Over the past six years, the issue of basis risk has been given increasingly more
serious attention than it received in some relatively early analyses of index insur-
ance.6 For example, Smith and Watts (2009) examined the extent of basis risk in a
rainfall index insurance instrument that reflected typically estimated correlations
between plant growth and rainfall at the location of the weather station where rain-
fall is being measured (which they report as about 0.7) and the correlation between
plant growth at that location and other locations in a typical area to be covered by
the index (which they also report as about 0.7). Under somewhat restrictive inde-
pendence assumptions, using a Monte Carlo approach and allowing correlations
between the index and on-farm crop yields to vary, these authors’ results indicate
that basis risk is likely to be substantial.
   Smith and Watts’ findings are presented in table 1 for farms that experience fairly
substantial crop losses of between 50 –70% of expected yields and who insure at an

Smith                                                                                279
Table 1. Indemnity Payment Outcomes for Farmers Experiencing Significant Yield Losses
(yields between 50 – 70% of average)
                                             Probability of Indemnity Event

Rainfall Index-Area Yield Correlation        No Indemnity            Small Indemnitya            Large Indemnityb

0.00                                         0.784                   0.114                       0.103
0.20                                         0.723                   0.145                       0.132
0.40                                         0.661                   0.179                       0.159
0.60                                         0.590                   0.219                       0.192
0.80                                         0.479                   0.316                       0.205
0.90                                         0.377                   0.429                       0.194




                                                                                                                      Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
0.95                                         0.278                   0.559                       0.163
1.00                                         0.000                   1.000                       0.000


  Notes: aA small indemnity is an indemnity paid when the rainfall index has a value of between 50–70%; ba large
indemnity is an indemnity paid when the rainfall index has a value of less than 50%.




index strike trigger of 70% (an indemnity payment is made when the rainfall
index’s value falls below 70% of its expected value). These authors report that even
if the correlation coefficient between the farm’s yields and the index is as high as
0.9, there is 37.3% probability that the farm will not receive an indemnity. If the
correlation coefficient is 0.6, which is perhaps a more realistic estimate, then the
probability of no indemnity payment increases to 59% and the likelihood of a sub-
stantial indemnity payment is only about 20%.
    A recent study by Jensen, Mude, and Barrett (2014) examines basis risk in the
context of the demand for a livestock mortality insurance product for Northern
Kenya farmers, which was specifically designed to minimize basis risk and cover
losses of livestock. The product was offered to households for which the livestock en-
terprise provides about 70% of the smallholder household’s income. The index on
which the insurance product was based was derived from a Normalized Difference
Vegetation Index (NDVI), the index first used by the USDA Risk Management Agency
to provide forage index insurance to farmers in similarly arid regions of the United
States.7 The northern Kenya product was priced to cover the costs of providing the in-
surance and, while initially 28% of eligible farmers purchased coverage in the first
period the insurance was available, participation subsequently declined rapidly. This
is a widely observed pattern for pilot index insurance schemes offered to smallholder
farmers at a commercial or near full cost basis (Miranda and Farrin 2012).
    Jensen Mude, and Barrett (2014) estimate correlation coefficients between indi-
vidual farm mortality rates and the index that applies to those farms in the insurance
product. Consistently among the five districts covered by the product, for well over
half of the farmers, the correlation coefficient is less than 70%. For between 9–29%

280                                                  The World Bank Research Observer, vol. 31, no. 2 (August 2016)
of the farms in each of the five districts, these authors report a negative correlation
between on-farm losses and losses indicated by the forage-based index. Not surpris-
ingly, but importantly, Jensen Mude, and Barrett find that basis risk has a major
adverse influence on the household’s willingness to buy insurance coverage and that,
as a household’s understanding of that basis risk improves, the household becomes
less likely to purchase the insurance coverage. Similar results on the adverse effects of
indicators of increased basis risk are reported for smallholder farmers in rural India
by Gine ´ et al. (2008) and by Mobaruk and Rosenzweig (2012, 2013).
   Elabed, Bellermare, Carter, and Guirkinger (2013) have proposed the use of a
double trigger contract to mitigate basis risk. In that contract, village-wide cotton




                                                                                             Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
yields in Mali are used to make payments to individual farmers for losses only if dis-
trict yields are sufficiently low to indicate that low village yields are not the result of
moral hazard behaviors (in other words, the village is effectively viewed as a “collec-
tive” farm operated as a single entity). As such, even though described as an index
product, the product has substantive elements of an all risk contract. The double
trigger concept is not new; for example, Watts and Associates proposed a double
trigger index product to the USDA Risk Management Agency in the mid-1990s
when index insurance products were first being developed as part of the US crop in-
surance program. Certainly, index insurance products based on a double trigger can
be designed to mitigate basis risk. However, by increasing the frequency with which
payments are made (even though a double trigger can also mitigate the frequency
with which farmers receive indemnities when they do not experience losses), such
contracts are often likely to increase premium rates.
   Unambiguously, then, the literature seems to be clear about three aspects of basis
risk in the context of index insurance. First, the theoretical literature indicates that
an increase in basis risk will reduce a farmer’s willingness to pay for an index insur-
ance product (see, e.g., Dercon et al. 2014; Miranda and Farrin 2012; and de
Janvry et al. 2014). Second, even though the potential exists for innovative product
designs to mitigate basis risk to some extent, basis risk is pervasive, extensive, and
substantial in almost all index insurance products. Third, the empirical evidence
consistently shows that basis risk has substantial adverse effects on smallholder
farmers’ participation in crop insurance products. As Jensen, Mude and Barrett
(2014) note, it seems likely that in many pilot index insurance schemes, as farmers
come to realize the extent of that risk and, as a corollary, more accurately assess the
expected net payoff from the insurance policy, they abandon the program.

Pricing Index Insurance Products and Farmers’ Willingness
to Pay for Them
The price of a competitively-offered commercially-provided insurance policy has
two components; the expected indemnity and the operating and administration

Smith                                                                                 281
costs associated with delivering the product. A risk neutral or risk averse economic
agent facing no liquidity constraints will therefore be willing to purchase an actuar-
ially fair insurance policy for which the price is simply the expected indemnity.
However, errors in rate setting associated with differences in the information sets
available to the insured individual and the insurance company often result in
adverse selection problems that are typically extensive in all risk crop insurance
contracts but widely viewed as less pervasive in index insurance contracts. Positive
loading factors mean that only risk averse economic agents will purchase a com-
mercially offered insurance policy. Further, if those risk averse agents face liquidity
constraints, the requirement that, in contrast to many informal insurance schemes,




                                                                                                          Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
premiums be paid when purchasing coverage at planting time rather than after
harvest will further diminish the likelihood of purchasing the insurance.
   Consistently, studies of willingness to pay for either multiple peril insurance prod-
ucts based on the farm’s actual yields or index insurance products have reported that
farmers are not willing to pay very much, if anything at all, for crop insurance cover-
age (Wright and Hewitt 1994). That finding holds whether the farmers being investi-
gated are located in rich countries like the United States and Australia or poor
countries like Morocco and Tanzania (Smith and Watts 2009). Basis risk simply
makes matters worse for index-based insurance products as compared to all-risk in-
surance products that always pay indemnities when a farmer incurs an insured loss.
   In effect, as Smith and Watts (2009) suggested in their review of the then extant
willingness-to-pay studies, once the loading factor for an insurance product exceeds
about 9%, only extremely risk averse farmers are willing to buy coverage. Numerous
studies of risk preferences among smallholder farmers have done little more than rep-
licate Binswanger’s (1981) finding that most are moderately risk averse and in fact,
on average, extreme risk aversion attitudes occur about as frequently as risk loving at-
titudes occur (i.e., relatively infrequently). This finding is consistent with the results
of most studies of farmers’ willingness to pay for agricultural insurance and goes a
long way to explaining why so few smallholder farmers have participated in most of
the pilot index insurance programs that have been offered over the past decade.
   If a 9% loading factor is effectively the demand side choke price for multiple peril
insurance, then the choke price for index insurance is likely to be lower. An impor-
tant question, therefore, concerns what sort of loading factor is required by private
insurers to deliver index insurance to farmers. Very early estimates by some academ-
ics that index insurance could be provided with loads of 2 –5% now appear to be
very optimistic. For example, BASIX in India was initially willing to offer an index
product to smallholder farmers with about a 15% loading factor, but then estimated
that a minimum load of 25% would be required to cover the company’s operations
and administration costs.
   Even in developed countries, where a single policy is likely to cover hundreds or
thousands of hectares over which fixed costs can be spread, insurance companies

282                                      The World Bank Research Observer, vol. 31, no. 2 (August 2016)
appear to require a loading factor well in excess of 15% to offer an index insurance
product (either via subsidies or underwriting gains). For example, two types of cost
that are always incurred by primary insurance companies are often overlooked in
discussions about agricultural insurance. These are the primary companies’ rein-
surance costs and the costs of the financial capital they have to hold in order to
offer the insurance in the first place. Those costs, by themselves, are likely be close
to or exceed 9% of the expected indemnity associated with the policy.8

Using Government and Aid Agency Resources to Support Risk
Management and Risk Coping Strategies




                                                                                            Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
There are clearly important limits to the value of commercially-provided and priced
index insurance as a risk management strategy that would alleviate the food and
income insecurity effects on smallholder households of catastrophic and moderate
adverse crop and livestock production shocks. The question is then whether domestic
government or foreign aid resources should be used to subsidize access to those prod-
ucts.9 Too little attention seems to have been given to this issue. However, the theoret-
ical and empirical evidence that does exist suggests that subsidizing access to such
products may be sub-optimal. For example, as discussed above, de Janvry et al show
that where informal risk sharing occurs, purchasing individual insurance is likely to
have a negative impact on the farm’s welfare because of free riding effects. A better
way to go that, as de Janvry, Dequiedt, and Sadoulet (2014) also show, may have the
potential to increase the group’s economic welfare would be to offer such coverage to
the group. However, if the product is to be subsidized, why not just provide disaster
aid when an area-wide adverse event occurs and avoid many of the costs associated
with private delivery of the insurance product? As Smith and Glauber (2012) note,
these costs are often substantial relative to the amount of aid to be provided.
   One argument for private delivery of foreign aid dollars though crop insurance
schemes, instead of allocating the aid through the government, is that governments
are more likely to divert aid from the communities that need the help than are
private insurance companies.10 However, as Hazell and Hess (2010) suggest, one
potential use for weather and plant growth indexes is to more accurately target
foreign emergency and other aid dollars to the communities who most need them.
If all farmers in a community come to expect aid when widespread adverse out-
comes for the community occur, then informal risk sharing programs are less likely
to fail in those circumstances because on a community-wide basis, resources are en-
hanced. To some extent, some of the anti-poverty trap benefits claimed for individu-
ally purchased index insurance products might then accrue in terms of technology
adoption and reductions in fire sales of livestock assets because smallholder families
become more certain that those informal risk sharing networks will work when
they need help.

Smith                                                                                283
   A recent study by Marenya, Smith, and Nkonya (2014) suggests that index insur-
ance may also not be an optimal policy for encouraging longer-term investments in
risk management practices. These authors carried out stated-choice experiments
with respect to alternative incentives for the adoption of agroforestry yield-enhancing
conservation practices by Malawi smallholder farmers in the Shire River Valley.
Their results indicate that farmers will choose cash payment incentives that are
substantially smaller than the subsidies associated with an ideal index insurance
product (essentially an index product that has no basis risk) as their preferred in-
centive for adopting the same conservation practices that would increase yields and
reduce the frequency of extreme adverse shocks.




                                                                                                         Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
   If index insurance targeted to individual smallholder farmers is not a silver bullet
solution for improving their ability to manage and cope with income and food
security risks, then what would be a more effective and efficient use of domestic gov-
ernment and foreign aid funds? The answer is both discouraging and encouraging
because many of the public policy and aid-driven initiatives that are most likely to
benefit smallholder farmers are programs that have been supported for a long time.
For example, public investments have high returns when they are made in agricul-
tural research targeted towards both enhancing yields and reducing their volatility
(e.g., improving drought resistance) and in extension programs that enable small-
holder farmers to minimize production risks associated with adopting new varieties.
Programs that allow smallholder farmers to cope with climate change-driven reduc-
tions in growing season rainfall and changes in the timing of that rainfall are likely
to substantially mitigate the volatility of their yields and incomes. These programs
clearly include investments in water storage and irrigation systems.
   Finally, two alternative uses of agriculture-related index insurance products have
been extensively discussed. One approach is to offer index insurance to private
credit institutions that would otherwise not offer loans to smallholder farmers. To
the extent that such an approach is a least cost way of ensuring those farmers have
some access to financing for purchasing inputs that are likely to improve and stabi-
lize their crop yields and livestock operations, there may be a genuine economic jus-
tification for such an approach. However, an alternative that accomplishes the same
objective is to establish a farm credit system underwritten by the government and/
or aid agencies. Whether the former or latter approaches would be more effective is
not clear, as in the former case the financial intermediaries’ actions may be subject
to moral hazard effects, and in the latter case there may be several reasons for a
“government failure” problem.
   The other use of index insurance products is for the lending institution to bundle
loans to farmers with an insurance policy; the effect, absent subsidies, is to increase
the cost of the loan to the farmer, which reduces their return from the inputs they
purchase with the loan. In addition, as Miranda and Farrin (2012, p 413) and,
earlier, Goodwin and Smith (1995) note, (assuming that the lending institution

284                                     The World Bank Research Observer, vol. 31, no. 2 (August 2016)
receives the indemnity payment when a farmer defaults on the loan) the incentives
for the lending institution to monitor loans and invest in strong loan recovery
actions may be mitigated. If the insurance product is subsidized, one result could be
that a substantial number of low quality loans are made with potentially serious
adverse consequences for the financial systems and government expenditures.

Summary
Smallholder households in many developing countries have many ways of manag-
ing the income and agricultural production risks they face. These approaches




                                                                                        Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
include production practices such as enterprise diversification, the use of risk re-
ducing resources, storage, and household investments in marketable assets such as
livestock, as well as participation in extended family- and community-based risk
sharing arrangements. Nevertheless, those households are still subject to relatively
frequent crop and household income losses that have potentially catastrophic con-
sequences for the welfare of their members.
   In response, especially over the past decade, economists and policy makers have
searched for innovative ways of improving those households’ resiliency with respect
to such adverse events. A particular focus has been the potential for smallholder
households to use commercially viable weather-based index insurance products to
improve their welfare. However, increasingly, the empirical evidence indicates that,
as is the case for very rich farmers in developed countries, smallholder farmers in
developing countries are unlikely to purchase such products absent substantial sub-
sidies, which their governments probably cannot afford.
   There are several reasons why those farmers are not willing to pay for such in-
surance. Fundamentally, and perhaps most importantly, many smallholder
farmers appear to have less expensive ways of managing risk through practices
that are already available to them. In addition, the overwhelming majority of
those farmers do not appear to be sufficiently risk averse to be willing to pay pre-
miums that cover the loading costs that would be incurred by the insurance com-
panies who would supply the insurance. Further, the extensive degree of basis risk
associated with most index-based agricultural insurance products severely exacer-
bates the problem.
   Index insurance, however, might be helpful in other contexts. For example, if
communities purchase such insurance, perhaps on a subsidized basis, the indemni-
ties provided when the community experiences broad-based crop losses may facili-
tate the performance of informal or semiformal risk sharing agreements within the
community. Alternatively, providing index insurance products to smallholder
households by tying them to operating loans used to purchase improved inputs may
reduce the impacts of adverse events on the household, although such bundling
raises the cost of the loan to the household and may provide a moral hazard

Smith                                                                             285
disincentive for the bank that receives the indemnity in cases of loan default to be
diligent in its loan management practices.
   Finally, using government and/or international aid agency resources to subsidize
access to index insurance may seem like a potentially useful use of such funds, espe-
cially if the consequence is the adoption of more productive but perhaps higher risk
technologies by very poor smallholder households. However, those funds have
many other potential uses and the empirical evidence that very poor farmers who
purchase such insurance are likely to adopt new technologies is weak. Perhaps
more evidence is needed about the relative returns from subsidizing crop insurance
as opposed to the returns from those other uses (such as subsiding the adoption of




                                                                                                                 Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
conservation practices, providing education, and increasing location-specific agri-
cultural research and extension programs). However, much of the evidence current-
ly available indicates that many of those other programs are likely to provide
substantially higher returns.


Notes
* Vincent Smith is a professor of Economics at Montana State University.
   1. Very recently, however, researchers have begun to question the utility of index and other insur-
ance products as vehicles for risk management by small holders. See, for example: Binswinger
Mkhize’s (2010) assessment of the potential value of such products, which he begins by asking
whether their utility has been overhyped: Macours (2013) recent review of the evidence from random-
ized control experiments with respect to index insurance; and the survey by Miranda and Farrin
(2012) of the academic literature and practitioner experiences with respect to index insurance.
   2. While not the primary focus of their study, McIntosh, Sarris, and Papadopoulos (2013) report
that when smallholder farmers in Ethiopia were given a voucher to purchase a weather index insur-
ance product, for the most part the farmers used the voucher to fully subsidize insurance coverage for
a part of the land they farmed, leaving the rest of the land uninsured. At the very least, these authors’
findings suggest that unsubsidized index insurance is not viewed as a viable risk management tool by
those farmers.
   3. See, for example, Binswanger-Mkhize (2010), Carter (2008), Dercon et al. (2014), and Miranda
and Farrin (2012).
   4. To the best of our knowledge, the only commercially sustainable form of “stand alone” agricul-
tural insurance has been insurance against specific perils such as hail or fire (Kramer 1983; Goodwin
and Smith 1995; Smith and Glauber 2012) and markets for such insurance have tended to be small in
scale. All other “successful” insurance programs have either encouraged extensive voluntary partici-
pation through large subsidies, as in the United States, Canada, India, and Brazil (Hazell, Pomerada,
and Valdes 1986; Mahul and Stutley 2010; Smith and Glauber 2012; and Wright 2014), by mandat-
ing participation as a condition for participating in other government programs (Goodwin and Smith
1995), or, perhaps, by bundling the insurance with another commodity. Approaches like Syngenta’s
bundling an insurance policy with the purchase of seed by a smallholder is a scheme that makes the
insurance product a tied good; in effect, it is a way of extracting economic surplus associated with seed
purchases to cover insurance policy costs.
   5. Weather indexes can be relatively simple or complex. For example, a lack of rainfall and excessive
heat during the growing season are both major causes of crop loss for commodities like maize and
corn. A more optimal insurance index for those crops would include measures of both rainfall and
heat. Reliable estimates of daytime temperatures are much more difficult to acquire on a consistent


286                                             The World Bank Research Observer, vol. 31, no. 2 (August 2016)
basis, however, and may be less meaningful for even relatively small areas when measured at a specific
location. Hence, most attention has focused on rainfall or indexes based on satellite measures of
vegetation growth. However, as a reviewer pointed out, index insurance may also cover floods and
hurricanes.
    6. For example, Vedenov and Barnett (2004) suggested that correlations between weather indexes
and crop yields would be sufficiently large to ensure that basis risk would not be a major problem for
farmers.
    7. Uptake for the U.S. NDVI product has been relatively substantial, mainly because US farmers
pay a premium that is less than 50% of the expected indemnity (Glauber 2013; Goodwin and Smith
2013).
    8. Industry sources suggest that reinsurance and financial capital costs each are likely to be 3–5%
of the expected indemnity, and that reinsurance costs will be substantially higher than that for books
of business that are relatively small (less than $100 million of premium).




                                                                                                           Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
    9. A reviewer pointed out that aid agencies typically have development funds and emergency relief
funds and that insurance programs are typically defined as development programs but essentially are
used to provide emergency relief. Thus, for many aid agencies, the internal incentive structure for the
use of funds to support insurance is misaligned.
    10. Private insurance companies cannot always be relied on to deliver aid dollars where they are
most needed. Issues about the integrity of insurance companies abound, even in countries where the
regulatory and legal systems are well developed.



References
Alston, J. M., and B. H. Hurd. 1990. “Some Neglected Social Costs of Government Spending on Farm
   Programs.” American Journal of Agricultural Economics 72 (1): 149–56.
Bardsley, P., A. Abey, and S. Davenport. 1984. “The Economics of Insuring Crops against Drought.”
   Australian Journal of Agricultural Economics 28 (1): 1– 14.
Barnett, B. I., C. J. Barrett, and J. R. Skees. 2008. “Poverty Traps and index-based transfer instru-
   ments.” World Development 36 (10): 450–74.
Borch, K. H. 1990. Economics of Insurance. Amsterdam: North Holland.
Buschena, D., V. H. Smith, and H. Di. 2005. “Policy Reform and Farmers’ Wheat Allocation in Rural
  China.” Australian Journal of Agricultural and Resource Economics 49 (2): 143–58.
Binswanger, H. P . 1981. “Attitudes toward Risk: Theoretical Implications of an Experiment in Rural
   India.” The Economic Journal 91 (364): 867– 90.
Binswanger-Mkhize, H. 2010. “Is There Too Much Hype about Index Based Agricultural Insurance?”
   The Journal of Development Studies 48 (2): 187– 200.
Boucher, S., B. Bradford, and M. Carter. 2005. “The Impact of Market Friendly Reforms on the
  Operation of Credit and Land Markets in Honduras and Nicaragua.” World Development 33 (1):
  107–28.
Carter, M. 2008. “Inducing Innovation: Risk Instruments for Solving the Conundrum of Rural
   Finance.” Working paper, Department of Agricultural and Applied Economics, University of
   Wisconsin.
de Janvry, A., V. Dequiedt, and E. Sadoulet. 2014. “The Demand for Insurance against Common
   Shocks.” Journal of Development Economics 106 (Jan): 227– 38.
Dercon, S., R. Vargas Hill, D. Clarke, I. Outes-Leon, and A. Seyourn Taffesse (2014). “Offering Rainfall
   Insurance to Informal Insurance Groups: Evidence from a Field Experiment in Ethiopia.” Journal of
   Development Economics 106: 132–43.


Smith                                                                                               287
Elabed, G., M. F. Bellemare, M. R. Carter, and C. Guikinger. 2013. “Managing Basis Risk with
   Multiscale Insurance.” Agricultural Insurance 44: 419– 31.
Gine´ , X., R. Townsend, and J. Vickery. 2008. “Symposium on Access to Finance: Patterns of Rainfall
   Insurance Participation in Rural India.” World Bank Economic Review 22 (3): 539 –66.
Glauber, J. W  . 2013. “The Growth of the Federal Crop Insurance Program, 2001–11.” American
   Journal of Agricultural Economics 95 (2): 482–8.
Goodwin, B. K. 1993. “An Empirical Analysis of the Demand for Multiple Peril Crop Insurance.”
  American Journal of Agricultural Economics 75 (3): 425–34.
       . 2014. “Agricultural Policy Analysis: The Good, the Bad, and the Ugly.” Presidential Address
   presented at the 2014 annual meetings of the Agricultural and Applied Economics Association,
   July 28, Minneapolis, MN.




                                                                                                                 Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
Goodwin, B. K., and V. H. Smith. 1995. The Economics of Crop Insurance and Disaster Aid. The AEI Press.
  Washington, D.C.
Goodwin, B. K, and V   . H. Smith. 2013. “What Harm is Done by Subsidizing Crop Insurance?”
  American Journal of Agricultural Economics 95 (2): 489–97.
Halcrow, H. G. 1949. “Actuarial Structures for Crop Insurance.” Journal of Farm Economics 31 (3):
  418 –43.
Hazell, P., and U. Hess. 2010. “Drought Insurance for Agricultural Development and Food Security in
  Dryland Areas.” Food Security 2: 395– 405.
Hazell, P., C. Pomerada, and A. Valdes. 1986. Crop Insurance for Agricultural Development: Issues and
  Experience. Baltimore: Johns Hopkins University Press.
Jensen, R. 2007. “The Digital Provide: Information (Technology), Market Performance and Welfare in
   the South Indian Fisheries Sector.” Quarterly Journal of Economics 122 (3): 879– 924.
Jensen, N. D., A. G. Mude, and C. B. Barrett. 2014. “How Basis Risk and Spatiotemporal Adverse
   Selection Influence Demand for Index Insurance: Evidence from Northern Kenya.” ILRI working
   paper, August.
Kramer, R. A. 1983. Federal Crop Insurance. Agricultural History 97 (1): 181– 200.
Marenya, P., V. H. Smith, and E. Nkonya. 2014. Relative Preferences for Soil Conservation Incentives
  among Smallholder Farmers: Evidence from Malawi. American Journal of Agricultural Economics 96
  (3): 690– 710.
Macours, K. 2013. “Volatility, Agricultural Risk, and Household Poverty: Micro-evidence from
  Randomized Control Trials.” Agricultural Economics 44: 79 –84.
Mahul, O., and C. J. Stutley. 2010. Government Support to Agricultural Insurance: Challenges and
  Opportunities for Developing Countries. Washington, DC: The World Bank.
McIntosh, C., A. Sarris, and F. Papadopoulos. 2013. “Productivity, Credit, Risk and the Demand for
  Weather Insurance in Ethiopia.” Agricultural Economics 44: 399–417.
Miranda, M. J. 1991. “Area-Yield Crop Insurance Reconsidered.” American Journal of Agricultural
   Economics 73 (2): 233–42.
Miranda, M. J., and K. Farrin. 2012. “Index Insurance for Developing Countries.” Applied Economic
   Perspectives and Policy 34 (3): 391–427.
Mobarak, A, M., and M. R. Rosenzweig. 2012. “Selling Formal Insurance to the Informally Insured.”
  Yale University Economics Department Working paper No 97, New Haven, Connecticut.
      . 2013. “Informal Risk Sharing, Index Insurance, and Risk Taking in Developing Countries.”
   American Economic Review 103 (3): 375– 80.
Patrick, G. 1988. “Mallee Wheat Farmers’ Demand for Crop and Rainfall Insurance.” Australian
   Journal of Agricultural Economics 32 (1): 37 –49.


288                                             The World Bank Research Observer, vol. 31, no. 2 (August 2016)
Skees, J. R., P. Hazell, and M. Miranda. 1999. New Approaches to Crop Yield Insurance in Developing
   Countries. International Food Policy Research Institute. EPTD Discussion Paper No. 55.
Smith, V. H., and A. E. Baquet. 1996. “The Demand for Multiple Peril Crop Insurance: Evidence from
  Montana.” American Journal of Agricultural Economics 78 (1): 75– 83.
Smith, V. H., and J. W. Glauber. 2012. “Agricultural Insurance in Developed Countries: Where Have We
  Been and Where Are We Going?” Applied Economic Perspectives and Policy 34 (3): 363 –90.
Smith, V. H., and B. K. Goodwin. 2011. “Private and Public Roles in Providing Agricultural Insurance
  in the United States.” In J. Brown, ed., Private and Public Roles in Insurance. Washington DC:
  AEI Press.
Smith, V. H., and M. A. Watts. 2009. Index Based Agricultural Insurance in Developing Countries. Report
  prepared for the Bill and Melinda Gates Foundation.
Vedenov, D. V., and B. J. Barnett. 2004. “Efficiency of Weather Derivatives as Primary Crop Insurance




                                                                                                             Downloaded from http://wbro.oxfordjournals.org/ at International Monetary Fund on October 27, 2016
   Instruments.” Journal of Agricultural and Resource Economics 29 (3): 387 –403.
Wright, B. D. 2014. “Multiple Peril Crop Insurance.” Choices, August.
Wright, B. D., and J.A. Hewitt. 1994. “All-risk Crop Insurance: Lessons from Theory and Experience.”
  In D. L Hueth, and W   . H. Furtan, eds., Economics of Agricultural Crop Insurance: Theory and Evidence,
  73 –114. Boston: Kluwer Academic Publishers.




Smith                                                                                                 289