untitled

Edited by
Molly E. Hellmuth, Daniel E. Osgood, Ulrich Hess, Anne Moorhead and Haresh Bhojwani

Case studies and authors
Unlocking development potential in Malawi: Erin Bryla and Joanna Syroka
Disaster insurance in Ethiopia: Ulrich Hess and Laura Verlangier
National-level drought risk management in Malawi: Joanna Syroka
A farmer-centric approach in Ethiopia: Marjorie Victor and Stephane de Messieres
Insurance for contract farming in India: Sonu Agrawal, Jamie Anderson, Francesco Rispoli, Ulrich Hess and

Laura Varlangieri

Disaster relief in Mexico: Jesus Scamilla, Marcela Denisse Cruz Rubí, Laura Verlangieri and Ulrich Hess
Catastrophe risk insurance in the Caribbean: Simon Young, Francis Ghesquiere and Olivier Mahul
Insuring development goals in the Millennium Villages Project: Eric Holthaus, Neil Ward, Cheryl Palm and

Daniel Osgood

Vietnam: Flood insurance in the Mekong Delta: Jason Hartell and Jerry Skees
Central America – a diﬀerent approach for launching index insurance: Carlos Arce
Barriers to implementation in the Ukraine: Roman Shynkarenko, Ulrich Hess and Laura Verlangieri
The private sector builds a market in India: Sonu Agrawal, Ulrich Hess and Laura Verlangieri
Laying the foundations for farm-level insurance in Ethiopia: Erin Bryla and Joanna Syroka
Supporting farmers – and a government seed program – in Brazil: Ronaldo Neves, Laura Verlangieri and

Ulrich Hess

Stakeholder communication is key in Thailand: Ornsaran Pomme Manuamorn and William Dick
Scaling up in India: The public sector: Kolli Rao, Ulrich Hess and Laura Verlangieri
Livestock insurance in Mongolia: Jerry Skees and Robin Mearns

Contributing authors
Jamie Anderson, Mirey Atallah, Walter Baethgen, Anthony G. Barnston, Paul Block, Mohammed S. Boulahya,
Molly Brown, Erin Bryla, Sandro Calmanti, Michael Carter, Pietro Ceccato, Bronwyn Cousins, Tufa Dinku,
Chris Funk, Yannick Glemarec, Lisa Goddard, Arthur Greene, David Grimes, James W. Hansen, Amor V.M. Ines,
James W. Jones, Yasir Kaheil, Kristopher Karnauskas, Arun Kashyap, Robert Kelly, Abedalrazq Khalil, Sergey Kirshner,
Pradeep Kurukulasuriya, Upmanu Lall, Alain Lambert, Malgosia Madajewicz, Olivier Mahul, Simon Mason,
Megan Mclaurin, Holger Meinke, Janot-Reine Mendler de Suarez, Vincent Moron, Michael Norton, Eric Patrick,
Anthony Patt, Nicole Peterson, Alexander Pfaﬀ, Gonzalo Pizarro, Francesco Rispoli, Andrew W. Robertson,
Kenneth E. Shirley, Asher Siebert, Mariana Simoes, Jerry Skees, Chris Small, Andreas Spiegel, Pasquale Steduto,
Liqiang Sun, Joanna Syroka, Pablo Suarez, Juerg Trueb, Calum Turvey, Maria Velez, Laura Verlangieri, Marta Vicarelli,
Neil Ward and Koko Warner

Review team
Barry Barnett, Joanne Bayer, Stefan Dercon, Andrew Dlugolecki, Peter Hazell, Shadreck Mapfumo and
Kolli Rao

First published 2009

All rights reserved. The publisher encourages fair use of this material provided proper citation is made. No repro-
duction, copy or transmission of this report may be made without written permission of the publisher.

ISBN 978-0-9729252-5-9

Correct citation
Hellmuth M.E., Osgood D.E., Hess U., Moorhead A. and Bhojwani H. (eds) 2009. Index insurance and
climate risk: Prospects for development and disaster management. Climate and Society No. 2. International Research
Institute for Climate and Society (IRI), Columbia University, New York, USA.

Index insurance, development and disaster management

iii

Foreword

At the ﬁrst annual meeting of the Global
Humanitarian Forum on ‘The Human Face of
Climate Change’, in June 2008, Swiss Re and
the International Research Institute for Climate
and Society hosted a policy roundtable on the
use of index insurance for poverty reduction.
The discussion brought together experts from
ﬁelds as diverse as reinsurance, climate science,
economics and food security to take stock
of their experience and insights on how this
innovative tool can best serve development.
They highlighted how index insurance is being
tested in the context of development, disaster
management and climate change adaptation.
Pilot experience had shown that index insurance
has the potential to reduce conditions of
chronic underdevelopment by both enabling
investment and reducing shocks in agricultural
livelihoods. At the roundtable, important issues
were identiﬁed, such as the role of diﬀerent
stakeholders, the challenges for scaling up to
meet broader development objectives and the
value of index insurance as a development tool.

Outcomes of the inaugural Forum made it

clear that we have the technology, the expertise
and the ﬁnancial resources to take on the
challenge presented by climate change – but that
action to tackle these problems is largely missing.
We need to step out of our narrow conﬁnes and
areas of expertise and share our knowledge – to
pool our resources and act together in order to
have real impact. This publication, born out of

discussions at the Forum and subsequent
meetings, represents the commitment of one
group of experts to pool and share their
experience and knowledge of index insurance.
I applaud the eﬀorts of this community to take
the time and energy to capture current
knowledge and experience and make it available
to the broader global development community.

For several years our eyes have been set on

Copenhagen. As a global community we have
been trying to focus attention on climate change
impacts and to bring innovation to the service
of equity and sustainability. This year the United
Nations International Strategy for Disaster
Reduction (UNISDR) launched the global
assessment report on disaster risk reduction –
which continues to remind us how ill-prepared
we continue to be in the face of climate-related
hazards. As an innovation, index insurance may
hold answers for some of the more obstinate
problems faced by the poor and the vulnerable.
I hope this publication will help us to appreciate
how much has been learned over the last few
years, and show us where we can usefully
concentrate our collective eﬀorts.

Koﬁ A. Annan
President
Global Humanitarian Forum

Partner statement

Before the current economic crisis, the
preceding years of global economic growth
were optimistically reﬂected in the progress
indicators of the Millennium Development
Goals. In sub-Saharan Africa, for example, the
economies of countries grew at more than 6%
in 2007. As the world’s economy has slowed,
these advances are now being threatened. These
trends imply risks of deepened hunger and
reduced aﬀordable access to adequate nutrition.

At the same time, the Intergovernmental

Panel on Climate Change’s fourth assessment
report has warned us that climate change is
likely to reduce food production potential,
especially in some already food-short areas.
It further states that there is now higher
conﬁdence in the projected increases in
droughts, heat waves and ﬂoods, as well as their
adverse impacts – which will be hardest felt
by the most vulnerable, who are often in the
weakest economic position.

Climate has always presented a challenge to

those whose livelihoods depend on the weather.
Even though a drought (or a ﬂood, or a
hurricane) may happen infrequently, the threat
of the disaster is enough to block economic
vitality, growth and wealth generation during
all years – good or bad. The risk of drought and
ﬂooding can keep people in poverty traps, as
risk-adverse behavior limits productivity and
the willingness of creditors to lend to farmers,
for example. Lack of access to ﬁnancial services,

especially in rural areas, in turn restricts access
to agricultural inputs and technologies, such as
improved seeds and fertilizers. At the national
level, when disaster strikes, many developing
countries rely on humanitarian aid, whose delay
can lead to higher human and economic costs.

There is a global recognition of the pressing

need for fresh approaches to confront these
challenges at scale. This type of thinking
is epitomized by the Hyogo Framework,
which advocates for a new approach to
disaster management focusing on disaster
risk reduction, as well as the Bali Action Plan
which advocates for a more comprehensive
consideration of risk sharing and transfer
mechanisms, such as insurance. In this
publication, we discuss one innovative response
to enable poverty reduction through better
climate risk management: index insurance.

The partners and contributors to this

publication are working together in their
networks throughout the world on developing
and testing index insurance as an approach
which, when combined with other ﬁnancial,
governance, structural and policy options, can
enable us to better meet our collective goals of
poverty reduction and economic growth. This
entails working with national governments,
such as the Ethiopian government and those
of the Caribbean region, to enable more
timely and reliable disaster response. It means
working together with farmers and the private

Index insurance, development and disaster management

sector in places such as India, Mongolia, China,
Ethiopia, Nicaragua and Thailand to remove
the barrier of climate risk and enable access
to credit. It involves using innovative climate
science, such as remote sensing, to expand
coverage of index insurance to areas where data
are sparse or limited.

As we move forward towards meeting

international goals such as poverty reduction
or climate change adaptation, we are applying
new thinking to confront old and new
problems. As the publication details, we
are doing this by bringing state-of-the-art

knowledge and practice into new settings: by
applying innovative science and technology,
by enhancing the role of private sector players,
by connecting to international risk pooling,
and by working with countries in developing
the capacity of their people and institutions.
Increasingly, we have recognized that
strengthening capacity – from the community
level to the global scale – is at the centre of
the development challenge. This publication
highlights the relevance of our work and the
critical importance of tackling this agenda
together.

Raj Singh
Chief Risk Oﬃcer, Swiss Re

Josette Sheeran
Director, World Food Programme (WFP)

Raymond Oﬀenheiser
President, Oxfam America

Stephan Zebiak
Director General, International Research
Institute for Climate and Society (IRI)

Kanayo Nwanze
President, International Fund for Agricultural
Development (IFAD)

Olav Kjorven
Assistant Secretary-General,
Assistant Administrator & Director,
Bureau for Development Policy, United Nations
Development Programme (UNDP)

Acknowledgements

Many people contributed to the preparation
of this report. The core team, contributors
and reviewers are listed at the front of the
publication. Editorial services were provided by
Anne Moorhead and Simon Chater of Green
Ink Ltd, UK, and design services by Jason
Rodriguez and Francesco Fiondella of IRI, and
Christel Chater and Paul Philpot of Green Ink
Ltd, UK.

During 2008–09 a series of meetings were

held that explored diﬀerent aspects of scaling
up index insurance. Many of the players who
worked on pilot projects and are now involved
in eﬀorts to scale up attended and contributed
their experiences, insights and technologies.
This publication is largely a distillation of the
outputs of this process. It is not a detailed
collection of the workshop materials; these
are available elsewhere (Barrett et al., 2007;
Bhojwani et al., 2008; technical papers at
http://iri.columbia.edu/csp/issue2/workshop).
Rather, it synthesizes the main points, and
builds on them to reﬂect the ongoing debate on
if and how index insurance could have a role to
play in poverty reduction, disaster risk reduc-
tion and development. The document reﬂects a
wide range of views from diﬀerent stakehold-
ers, some of which may be conﬂicting.

The team would like to acknowledge the

input of many stakeholders from development

partners, relief organizations, universities, research
institutes, the private sector, civil society and
nongovernment organizations, who were present
at the following workshops:

Humanitarian Forum’s Annual General Meeting
in Geneva, 24–25 June 2008

7–8 October 2008

Force on Commodity Risk Management in
Brussels on 22–24 October 2008

in Andhra Pradesh (Hyderabad) India on 18
November 2008

WFP and IFAD in Rome on 30–31 March
2009.
The team would also like to acknowledge the

contributions of the Weather Risk Management
Facility (WRMF), a WFP and IFAD collabora-
tion sponsored by a planning grant from the Bill
& Melinda Gates Foundation. We are indebted to
the World Bank’s Commodity Risk Management
Group for their contributions to the report, which
greatly enhanced the quality and the process.

The team gratefully acknowledges the ﬁnancial

support of Oxfam America, UNDP and National
Oceanic and Atmospheric Administration
(NOAA) in the preparation of this report.

Index insurance, development and disaster management

vii

Contents

Climate, poverty and index insurance

..............................................................................................

Climate and poverty

...................................................................................................................................................

Climate risk management

......................................................................................................................................

A brief introduction to index insurance

.........................................................................................................

Index insurance for development

.....................................................................................................................

Index insurance for disaster management

..................................................................................................

Climate change and index insurance

..............................................................................................................

Next steps – this publication

................................................................................................................................

Case studies I

.....................................................................................................................................................................

Unlocking development potential in Malawi

.........................................................................................

Disaster insurance in Ethiopia

...........................................................................................................................

Scaling up index insurance: The contract

......................................................................................................

Scaling up in context

..............................................................................................................................................

Identifying and quantifying the risk

..............................................................................................................

Measuring index parameters

.............................................................................................................................

Establishing

probabilities

.....................................................................................................................................

Estimating the price

................................................................................................................................................

Understanding

impacts

........................................................................................................................................

Case studies II

....................................................................................................................................................................

National-level drought risk management in Malawi

..........................................................................

A farmer-centric approach in Ethiopia

........................................................................................................

Insurance for contract farming in India

.......................................................................................................

Disaster relief in Mexico

........................................................................................................................................

Catastrophe risk insurance in the Caribbean

...........................................................................................

Insuring development goals in the Millennium Villages Project

................................................

Vietnam: Flood insurance in the Mekong Delta

.....................................................................................

viii

Contents

Scaling up index insurance: Operational issues

........................................................................................

Demand for the product

......................................................................................................................................

Local ownership and capacity

..........................................................................................................................

Legal and regulatory issues

................................................................................................................................

The challenge of subsidies

..................................................................................................................................

Case studies III

...................................................................................................................................................................

Central America – a different approach for launching index insurance

.................................

Barriers to implementation in the Ukraine

................................................................................................

The private sector builds a market in India

...............................................................................................

Laying the foundations for farm-level insurance in Ethiopia

.........................................................

Supporting farmers – and a government seed program – in Brazil

.........................................

Stakeholder communication is key in Thailand

.....................................................................................

Scaling up in India: The public sector

...........................................................................................................

Livestock insurance in Mongolia

.....................................................................................................................

Lessons learned and recommendations

........................................................................................................

What is the potential of index insurance for development and

disaster

management?

...........................................................................................................................

Lessons learned and recommendations

...................................................................................................

Global initiatives on ﬁnancial risk transfer

................................................................................................

102

References

........................................................................................................................................................................

105

Acronyms

...........................................................................................................................................................................

111

Index insurance, development and disaster management

Climate, poverty and index insurance

Climate and poverty

The climate has always presented a challenge
to those whose livelihoods depend on it.
Moving away from such dependence is usually
an early step in economic development, but
many millions have not yet succeeded in
taking that step. As climate variability and
uncertainty increase with climate change,
human development reversals are a distinct
possibility (UNDP, 2007). Climate has thus
become an urgent issue on the development
agenda.

For poor people, a variable and unpredict-

able climate presents a risk that can critically
restrict options and so limit development.

The climate presents a challenge to small-scale farmers and can signiﬁcantly limit their

options; Scott Wallace/World Bank

The risk materializes at two levels: the direct
effects of a weather shock, and the indirect
effects due to the threat of a weather shock
(whether it occurs or not).

When a weather shock occurs, poor people

are vulnerable. Local coping strategies often
break down. Poor people have few assets to
fall back on, and may be forced to sell these in
order to survive so that when the crisis is over
they are in a much worse position than before.
These impacts can last for years in the form of
diminished productive capacity and weakened
livelihoods. And climate change threatens
both more frequent and more severe extreme
events (IPCC, 2007).

Climate, poverty and index insurance

Under the threat of a possible weather

shock, poor people avoid taking risks
(Rosenzweig and Wolpin, 1993). They shun
innovations that could increase productivity,
since these innovations may increase their
vulnerability, for example by exhausting the
assets they would need to survive a crisis or by
requiring them to spend money without being
sure of a return (Dercon, 1996). Creditors
are unlikely to lend to farmers if drought (for
example) might result in widespread defaults,
even if loans can be paid back easily in most
years. This lack of access to credit critically
restricts access to agricultural inputs and
technologies, such as improved seeds and
fertilizers. Even though a drought (or a ﬂood,
or a hurricane) may happen only one year in
ﬁve or six, the threat of the disaster is enough
to block economic vitality, growth and wealth
generation in all years – good or bad.

Poverty limits the capacity of people to

manage weather risks, while these same risks
contribute to keeping people poor. Climate
change will greatly exacerbate this situation;
and developing countries, which are least
responsible for climate change, face its greatest
impacts. New tools are urgently needed to help
vulnerable people deal with climate change,
and the uncertainty that accompanies this.

It is not only the poor who need such

tools. After a climate-related disaster, govern-
ments struggle to ﬁnance relief and recovery
efforts and maintain essential government
services. Disaster response can be delayed for
several months as humanitarian aid trickles

in, which results in even higher human and
economic costs (Goes and Skees, 2003).

Risk transfer approaches such as insurance

have played a role in mitigating climate risk
in many parts of the world. However, they
have generally not been available in develop-
ing countries, where insurance markets are
limited if they exist at all, and are not oriented
towards the poor. A new type of insurance –
index insurance – offers new opportunities for
managing climate risk in developing coun-
tries. If designed and introduced carefully, it
has the potential to contribute signiﬁcantly
to sustainable development, by addressing a
gap in the existing climate risk management
portfolio. However, this potential has yet to
be proven; and there are some signiﬁcant
challenges that must ﬁrst be addressed.

Index insurance can be applied across a

diverse range of weather-related risk prob-
lems, from loss of crops due to drought, to
loss of livestock in harsh winter conditions,
to losses resulting from hurricanes. It can
be purchased at different levels of society
– at ‘micro-level’ by small-scale farmers, at
‘meso-level’ by input suppliers or banks, or at
‘macro-level’ by governments, for example. It
is not a ‘cure-all’ and will be inappropriate in
many situations; but it may be a useful option
in many others. As awareness and knowledge
of this new tool increase, and if the challenges
described in this publication can be overcome,
index insurance could become widely avail-
able as an additional option for those facing a
weather risk.

Index insurance, development and disaster management

The introduction of index insurance

can bring together a new set of actors and
new resources to address some of the more
persistent problems associated with poverty. It
also reﬂects a growing interest in, and a move
towards, market-driven solutions to these
problems. Shifting responsibilities from public
agencies, which ‘provide’ interventions to ‘ben-
eﬁciaries’, to market-based mechanisms where
people choose the services and technologies
they prefer, may offer the poor a more sustain-
able development model. Public–private
partnerships and private-sector development
are key to this approach, which must ulti-
mately deliver what customers demand.

Climate risk management

Climate risk is not a new phenomenon, and
climate risk management (CRM) in the broad
sense has long been practised. Farmers antici-
pate the rains, using various indicators, and
time their planting and inputs based on their
best estimates; they install irrigation systems
if they can; and they reduce risk exposure by
diversifying their livelihoods as far as possible
(Dercon, 1996; Ellis, 2000). Scientists have
also sought ways to help manage the risk
that climate presents. Agricultural research
has developed crop varieties that are drought
tolerant, for example, and soil management
practices that increase soil moisture-holding
capacity. Weather forecasts have been a major
advance in helping people plan appropriately.

In recent years, advances in climate science

have catalyzed the development of new CRM

practices. The improved use of climate infor-
mation in planning and resource management
has contributed to robust advances in disaster
risk reduction and climate change adaptation
(Meza et al., 2008; IFRC, 2008). The ﬁrst
publication in the Climate and Society series
describes and analyses some examples of
CRM in Africa (Hellmuth et al., 2007).

Index insurance is proposed as a new

CRM tool that may help people cope with
current weather-related risks and, if designed
properly, perhaps also future risks associated
with climate change. Depending on their
circumstances, people have a variety of risk
management mechanisms available to them.
Index insurance will not replace these options,
but should ﬁnd a role alongside them. It could
ﬁll the gap that occurs in the current portfolio
of coping mechanisms when these break
down in the face of a weather shock.

A brief introduction to index
insurance

Index insurance is insurance that is linked
to an index, such as rainfall, temperature,
humidity or crop yields, rather than
actual loss. This approach solves some of
the problems that limit the application of
traditional crop insurance in rural parts of
developing countries. One key advantage is
that the transaction costs are lower. In theory
at least, this makes index insurance ﬁnancially
viable for private-sector insurers and afford-
able to small farmers. Another important
advantage is that index insurance is subject to

Climate, poverty and index insurance

less adverse selection and moral hazard than
traditional insurance.

An example of index insurance, and the

most common application in developing
countries so far, is the use of an index of
rainfall totals to insure against drought-
related crop loss. Payouts occur when rainfall
totals over an agreed period are below an
agreed threshold that can be expected to
result in crop loss. Unlike with traditional
crop insurance, the insurance company does
not need to visit farmers’ ﬁelds to assess losses
and determine payouts. Instead, it uses data
from rain gauges near the farmer’s ﬁeld. If
these data show the rainfall amount is below
the threshold, the insurance pays out.

As well as reducing costs, this means that

payouts can be made quickly – a feature that
reduces or avoids distress sales of assets. This
process also removes moral hazards such as
the ‘perverse incentives’ of crop insurance,
where under certain conditions farmers may
actually prefer their crops to fail so that they
receive a payout. With index insurance, the
payout is not linked to the crop’s survival or
failure, so the farmer still has incentives to
make the best decisions. Another feature that
reduces moral hazard is that index insurance

1 Adverse selection occurs when potential borrowers or insurees

have hidden information about their risk exposure that is not

available to the lender or insurer, who then becomes more likely

to erroneously assess the risk of the borrower or insuree. Moral

hazard occurs when individuals engage in hidden activities

that increase their exposure to risk as a result of borrowing or

purchasing insurance. These hidden activities can leave the

lender or insurer exposed to higher levels of risk than had been

anticipated when interest or premium rates were established.

uses objective, publicly available data, so
individuals are unable to distort a situation to
their beneﬁt.

Rapid payouts are the major advantage of

index insurance when this is used as a disaster
management tool. Again, time-consuming
loss assessments are not needed, as payouts are
based on objective data. With index insurance
in place, governments and relief agencies can
plan ahead of crises, knowing that funds will
be available when they need them. Planning is
also facilitated because governments and relief
agencies can track the index and prepare an
early response.

But several critical components need care-

ful attention if index insurance is to be work-
able. Index insurance is new, and can be difﬁ-
cult for stakeholders to understand – time and
resources must be invested in explaining how
it works. It depends on the availability and
reliability of quality data, which is a signiﬁ-
cant challenge in most developing countries.
But perhaps most importantly, index insur-
ance is vulnerable to basis risk. Simply put,
basis risk is when insurance payouts do not
match actual losses – either there are losses
but no payout, or a payout is triggered even
though there are no losses. Obviously, if either
of these situations occurs too frequently, the
insurance scheme will not be viable, and may
even damage livelihoods (Skees, 2008). The
contract design, and in particular the selection
of an appropriate index, is crucially important
in minimizing basis risk. Other factors that
have implications for basis risk are proximity

Index insurance, development and disaster management

of the insured crop to a weather station, and
availability of climate data (Carriquiry and
Osgood, 2008).

The potential of index insurance has been

demonstrated by a number of projects in
various developing countries. A selection of
these projects, representing different regions
of the world and different applications of
index insurance, are presented as case studies
in this publication.

Index insurance is just one of a number

of related index-based ﬁnancial risk transfer
products that work on the same principles (for
details of others, see Skees et al., 2008b). In
this publication, the term index insurance is
used loosely to include the range of products.
Some of the case studies, for example, use
other related products such as weather deriva-
tives, but for ease of reading we call them all
index insurance. The reader should bear in
mind that much of the discussion is relevant
to the broader range of products. Also, the
discussion refers mainly to index insurance for
crop failure due to drought, since this is the
most common application so far; but much
of it is also relevant to applications beyond
drought and crop failure.

Index insurance for development

Index insurance may be able to help people
manage the weather risk that is partially
responsible for keeping them in poverty traps.
Poor people are not only at direct risk from
extreme weather events, but even without bad
weather they are at a disadvantage because the

risk blocks their opportunities. Lenders, for
example, may not extend credit to them. They
are therefore unable to invest in inputs that
would improve productivity in good-weather
years. Evidence suggests that farmers often
sacriﬁce 10–20% of income when using tradi-
tional risk management strategies (Gautam et
al., 1994).

But if they can take out insurance,

either individually or collectively (by farmer
associations, for example), the picture may
change. When lenders know that borrowers
are covered by insurance, they may be more
likely to extend credit to them. Farmers may
then choose to make investments that may
raise their productivity. If the weather is bad
and crops fail, the insurance will pay out and,
as a Malawian farmer put it, “I do not have
to worry about paying back loans in addi-
tion to looking for food to feed my family”
(Hellmuth et al., 2007). This insurance can be
sold either at the micro-level – to individual
farmers or households, or at the meso-level –
to banks or cooperatives for example.

Many of the case studies illustrate this

use of index insurance (Tables 1 and 2).
‘Unlocking development potential in Malawi’,
on page 13, is a good example. Malawi is one
of Africa’s poorest countries, with the majority
of its workforce engaged in smallholder farm-
ing. Crops are mostly rainfed, and drought
is an ever-present risk. These farmers have
little or no access to formal ﬁnancial credit to
buy agricultural inputs, because the chance of
them defaulting on loans is high (although,

Climate, poverty and index insurance

as the case study shows, not always because
of drought). The case study illustrates how
providing index insurance linked to loans may
help them to invest in inputs for cash crops
such as groundnuts and tobacco.

Of course, loan markets face many

challenges, of which climate risk is only one.
The Malawi case illustrates the importance
of these other challenges, and highlights the
need to identify the best role for insurance if
it is to contribute to development.

Index insurance for disaster
management

Several pilot projects have been exploring the
use of index insurance as part of the disaster
risk management portfolios of governments
and relief agencies (Tables 1 and 2). Disaster
risk reduction emphasizes preparedness
ahead of disasters, in order to limit the lives,
livelihoods and assets lost. Governments and
relief agencies, which usually bear the costs of
responding to large-scale disasters, have taken

Table 1. Some features of index insurance for development and for disaster relief

Index insurance for development

Index insurance for disaster relief

Intended
development
uses that are
being actively
explored

Help farmers escape poverty by removing
barriers preventing them from being
more productive, e.g. enabling access
to credit so that farmers can be more
effective in ‘good’ weather years. Direct risk
management. Protecting investments

Save lives and livelihoods through
more cost-effective and timely disaster
response. The timelier response may help
prevent people from falling into poverty
traps

Target group

Smallholder farmers and agricultural
laborers with growth potential, or their
suppliers and ﬁnancers; institutions within
the agricultural supply chain that work
with farmers. Contract is held at meso- or
micro-level by household, farmer group,
cooperative, microﬁnance institution,
NGO, or contract farming corporation

Those vulnerable to disaster, particularly
those in chronic poverty. Contract is held
at macro-level by government or relief
agency

Subsidies?

Hotly debated! Subsidies can severely
distort incentives and encourage the
promotion of ineffective or inappropriate
products. However, if used responsibly,
subsidies may play an important role in
launching products

Yes. Disaster relief programs are by
deﬁnition subsidized. The insurance is a
ﬁnancing mechanism designed to ensure
more effective use of these subsidies

Case studies

Ethiopia, Honduras, India, Jamaica, Malawi,
Mongolia, Nicaragua, Peru, South Africa,
Thailand, Ukraine, Vietnam

Brazil, Caribbean, Ethiopia, India, Malawi,
Millennium Villages Project, Mexico

Index insurance, development and disaster management

out insurance policies linked to weather indi-
ces that will pay out when extreme weather
events precipitate a disaster. The key advan-
tages are the speed of payout, which allows
rapid response; and the ability to plan ahead
of a disaster, in the knowledge that funds will
be available when they are needed.

As in the case of index insurance products

designed for development, integrating index
insurance into disaster management strategies
can help poor people whose livelihoods are
closely linked to the weather and who are at
risk of falling into poverty traps if a weather
shock occurs. Such people have assets –
animals or farming equipment, for example
– but may be forced to sell them to survive
a crisis, and then ﬁnd themselves without
the means to earn a living once the crisis is

over (Baulch and Hoddinott, 2000; Barrett et
al., 2001; McPeak and Barrett, 2001). Here,
insurance is designed to enable prompt disaster
response, allowing people to hold on to their
assets and quickly recover after the crisis.

In Ethiopia, for example, the Government

and the World Food Programme (WFP) have
worked together on a pilot project that uses
a rainfall index-based insurance contract to
provide emergency relief (see ‘Disaster insur-
ance in Ethiopia’ on page 17). Drought drives
agricultural losses, which in turn result in food
and income shortfalls and subsequently an
increase in the number of food aid beneﬁciar-
ies. Rainfall is used as a proxy for economic loss
resulting from drought and serves as a simple,
objective basis for the contract. Payouts are
not dependent on time-consuming and often

Index insurance has potential to help people escape poverty; Eric Holthaus/IRI

Climate, poverty and index insurance

subjective needs assessments, and can therefore
be made much more quickly. The rainfall-based
drought index produced for this pilot project
shows high correlation (around 80% for the
period 1994–2004) between losses that would
have been covered by the index and the number
of food aid beneﬁciaries. As this type of project
proceeds, it will be valuable to substantiate this
correlation with direct measurements of crop
yields and beneﬁciary livelihoods (WFP, 2007).

Climate change and index
insurance

There is much debate about the implica-
tions of climate change for index insurance,
centered around the following three questions.
Can index insurance contribute to adaptation
strategies in developing countries? Can it play
a role in managing the uncertainty associ-
ated with climate change? And does climate
change challenge the viability of index-based
insurance products?

There are at least three ways in which

index insurance might help build adaptive
capacity: as a risk transfer mechanism within a
comprehensive strategy for managing cli-
mate risk in the face of climate change; as a
mechanism to help people access the resources
needed to escape climate-related poverty; and
as a mechanism to incentivize risk reduction.

A comprehensive strategy for adapta-

tion in the agricultural sector, for example,
could include adapted crop varieties, micro-
irrigation, rainwater harvesting and improved
soil conservation practices. However, a certain

amount of risk would remain – and this
remaining risk might be covered by index
insurance.

The second way index insurance can

contribute to adaptation is through building
more resilient livelihoods by enabling access
to increased credit, technology and inputs.
Insured loans allow lenders to recuperate their
money even in a year where the climate causes
production losses. The loans allow people to
invest in more intensive livelihood strategies
which may help them to escape poverty traps.
The increase in wealth and in economic
resilience allows people to buffer themselves
from the direct impacts of the climate.

Beyond this, a key challenge in designing

insurance in the face of climate change is
to incentivize risk reduction through price
signals and risk management stipulations.
For example, contracts could stipulate that
certain risk reduction mechanisms, such
as the adoption of wind-resistant cropping
patterns or drought-tolerant crop varieties,
must be in place if crops are to be covered.
In Ethiopia, a pilot project is testing a scheme
in which cash-constrained farmers pay for
insurance premiums through their labor on
community assets which reduce risk, such as
water-harvesting structures. The cost of the
premiums in this project contains a compo-
nent that is a price signal to reﬂect long-term
trends (Oxfam America, 2009; see ‘A farmer-
centric approach in Ethiopia’ on page 44 and
‘Climate variability and change and index
insurance’ on page 38).

Index insurance, development and disaster management

Climate change has implications for index

insurance design and pricing, but need not
make such insurance non-viable. Contracts
are typically drawn up for a single season or a
few years at most, so they can be adapted as
climate change takes hold. The challenge is
to accommodate the added uncertainty due
to climate change while keeping premiums
affordable. Most of the pricing in insurance
contracts at present reﬂects year-to-year
variation, with only a small element allowing
for longer term trends. If, over time or in some
locations, incremental climate changes were
to lead to prohibitively expensive insurance,
that would signal the need to switch to more
radical forms of adaptation, such as changing
the crops grown or, in marginal areas, giving
up crop production altogether and taking up
livestock production, or migration. Climate
science can play a role in reducing uncertainty
through better understanding of the climate
system and how climate is likely to change
over the coming decades.

The climate variations experienced over

the next few decades will include variations
due to increasing greenhouse gases, but also
variations arising from natural processes
within climate systems. With the world’s cli-
mate scientists largely focusing on the former,
the latter are currently little understood. Yet
these internal components are likely to have at
least as much impact on the climate over the
coming decades as increased emissions. This
realization is leading to increased efforts to
understand these decadal processes.

The box ‘Climate variability and change

and index insurance’ on page 38 illustrates
climate change issues and their implications
for index insurance and the CRM portfolio.

Next steps – this publication

Most projects involving index insurance have
so far been small-scale, with two notable
exceptions in India and Mexico, which have
successfully scaled up. The projects have
tested the use of index insurance in a range
of applications and for different groups,
from supporting poor farmers in their efforts
to protect and enhance their livelihoods
to helping governments or relief agencies
manage climate-related crises. It is still an
open question whether index insurance can
contribute signiﬁcantly to sustainable devel-
opment and to disaster management. If it
is to do so, it will need to scale up to reach
very many more people. This publication
looks at the challenges – both technical and
operational – that accompany this proposed
scale-up.

At the same time it is acknowledged that

scaling up may not be sufﬁcient by itself to
impact on poverty reduction and develop-
ment and that index insurance may not be
appropriate in many places. Impact studies are
urgently needed to better understand under
what conditions index insurance can play
this role. These conditions may prove quite
restricted.

Many of the pioneers of index insur-

ance have contributed case studies to this

Climate, poverty and index insurance

publication, capturing the challenges they
encountered and the lessons they learned.
These case studies are found in three sections
beginning on pages 13, 42 and 72.

The next section focuses on the index

insurance product – the contract, including the
index itself – and highlights the main issues
that need attention if scale-up is to be suc-
cessfully achieved. It draws on the case studies
in explaining and discussing these issues; and
it outlines new approaches that might help
overcome some of the limitations currently
facing index insurance contract design.

In the following section, the operational

challenges encountered by many of the case
studies are examined, particularly those asso-
ciated with the development of new insurance

markets in areas with little prior experience in
weather index insurance.

Key messages and recommendations

resulting from this study are then presented.
This section is particularly aimed at those
who would like to see index insurance realize
its potential as a tool for sustainable devel-
opment. It identiﬁes where investment is
needed to enable scale-up. Many donors have
expressed an interest in index insurance and
are keen to invest, but this must be done with
care if insurance-based interventions are to
achieve development objectives.

The ﬁnal brief section looks at some of

the ongoing and planned global initiatives
that could support the development of index
insurance markets.

Index insurance is a tool that could be added to the existing portfolio of CRM options; Jason Hartell/GlobalAgRisk

Index insurance, development and disaster management

Table 2. Case studies

Case study

Development
or disaster
relief?

Risk

Index

Target user

Year ﬁrst
implemented

No. of beneﬁciaries in
2008 (or most recent
available ﬁgures)

Notes

Case study section I, starts on page 13

Unlocking
development
potential in
Malawi

Development

Drought

Rainfall

Tobacco
farmers

2005–06

2500

Began targeting
groundnut (and maize)
farmers, but due to
supply chain issues
changed to support
the stronger tobacco
system

Disaster
insurance in
Ethiopia

Disaster relief

Drought

Rainfall

WFP
operations in
Ethiopia

2006

316,000 (2006)

No payout in 2006.
Policy not renewed in
following years. US$25
million contingency
ﬁnancing contract
paid out in 2008. Scale
up to US$250 million
planned for 2010

Case study section II, starts on page 42

National-level
drought risk
management
in Malawi

Disaster relief

Drought

Rainfall
(linked
to maize
production)

Government
of Malawi

2008

Currently in ﬁrst trial
year

A farmer-
centric
approach in
Ethiopia

Development

Drought

Rainfall

Teff farmers in
Adi Ha

Under

development

Insurance
for contract
farming in
India

Development

Late blight
disease

Temperature
and humidity

Potato
farmers under
contract to
PepsiCo

2007

4575

Disaster relief
in Mexico

Disaster relief

Natural
disasters
impacting
smallholder
farmers,
primarily
drought and
ﬂood

Rainfall

State
governments

2002

800,000

Catastrophe
risk insurance
in the
Caribbean

Disaster relief

Hurricanes
and
earthquakes

Wind speed
and shaking

Caribbean
country
governments

2007

16 countries

Insuring
development
goals in the
Millennium
Villages
Project

Disaster relief

Drought

Combination
of rainfall and
vegetation
remote
sensing

Inhabitants of
villages in the
MVP

2007

Experimental product
transacted in three
MVP villages in Kenya,
Ethiopia and Mali in
2007. Not repeated in
2008

Case studies I

Table 2. Case studies

Case study

Development
or disaster
relief?

Risk

Index

Target user

Year ﬁrst
implemented

No. of beneﬁciaries in
2008 (or most recent
available ﬁgures)

Notes

Vietnam: Flood
insurance in
the Mekong
Delta

Development

Early ﬂooding
of rice ﬁelds

River level

The state
agricultural
bank

Under

development

Business interruption
insurance contract
currently being
considered by the state
agricultural bank

Case study section III, starts on page 72

Central
America –
a different
approach for
launching
index
insurance

Development

Drought and
excess rainfall

Rainfall Commercial

groundnut
and rice
farmers

2007

16 (400 anticipated

in 2009)

Barriers to
implementation
in the Ukraine

Development

Drought Rainfall Smallholder

farmers of
grain crops

2005

2 (2005)

One year only; project
discontinued

The private
sector builds a
market in India

Development

Drought,
ﬂood,
extreme
temperatures,
weather-
linked crop
diseases, fog,
humidity

Various,
mostly simple
weather
parameters

Farmers
(small,
medium and
large)

2003

Estimated

150,000 in total

Laying the
foundations
for farm-level
insurance in
Ethiopia

Development

Drought Rainfall Smallholder

maize farmers

2006

28 (2006),

13 (2007)

Small pilot as part of
feasibility study

Supporting
farmers – and
a government
seed program
– in Brazil

Disaster relief

Drought

Area yield

Maize
farmers in
government
seed program

2001

c. 15,000

Stakeholder
communication
is key in Thailand

Development

Drought Rainfall Smallholder

maize farmers

2007

388

Scaling up
in India: the
public sector

Disaster relief

Drought,
excess rainfall

Rainfall

Farmers
(small,
medium and
large)

2004

700,000

Livestock
insurance in
Mongolia

Development

Large
livestock
losses due to
severe
weather

Area livestock
mortality rate

Nomadic
herders

2006

4100

Index insurance, development and disaster management

Case studies I

Unlocking development potential in Malawi

A pilot project in Malawi illustrates how
index-based weather insurance can be used
to strengthen supply chain relationships and
support lending to small-scale farmers. This
project has also revealed a number of opera-
tional challenges, with interesting lessons for
others setting out on this path.

In Malawi, contract farming is becoming

more common, but credit to the rural sector,
which would allow small-scale farmers to take

advantage of this opportunity by increasing
their access to inputs, has been relatively
limited. Index insurance can support such
farmers as they become more market-oriented
and take on higher levels of risk by buying
improved seeds and fertilizers.

Managed by a partnership comprising the

World Bank, the pro-poor international insur-
ance group MicroEnsure and local stakeholders,
the pilot project is now in its fourth year. The

The project supports small-scale farmers in Malawi; Janot-Reine Mendler de Suarez/GEF-IW:LEARN

Case studies I

International Research Institute for Climate
and Society (IRI) has contributed technical
support to this project. During the ﬁrst year, the
2005–06 cropping season, 892 farmers located
within 20 km of four weather stations purchased
the insurance, which was bundled with a loan
for groundnut production inputs. The ground-
nut sector was chosen for this pilot operation
because (i) the crop is relatively drought
sensitive, (ii) farmers had been unable to invest
in new groundnut varieties due to their high
cost, and (iii) the farmers’ union had established
a new marketing system for the crop, which
envisaged loan recovery at the point of sale.

For the 2006–07 cropping season the pilot

expanded, with the addition of maize and a
ﬁfth weather station, taking numbers up to
1710 farmers. The marketability of groundnuts
was used to secure insurance and loans for
maize, which is the main food crop in Malawi.
Maize suffers from signiﬁcant price volatility
and fragmented marketing, so farmer loans are
generally not available for maize inputs alone
without some other collateral for the lender.
By combining a loan (and weather insurance)
for maize with a loan for a cash crop covered
by insurance, lenders felt comfortable that the
proﬁts from the cash crop could be used to
repay the loan for maize if necessary.

These pilots stimulated interest among

banks, ﬁnanciers and supply chain participants
such as processing and trading companies and
input suppliers. However, they also demon-
strated that other risks within the supply chain
can have a serious impact on loan recovery rates

and on the sustainability of the supply chain
itself, threatening the viability of a stand-alone
index insurance scheme. For example, during
the 2005–06 season there were problems due
to poor seed quality. In addition, during the
initial pilot operation, banks learned that the
new marketing arrangements in the groundnut
supply chain were not sufﬁciently organized to
ensure loan recovery at the point of sale, leading
to substantial side-selling and related failures
and delays in loan repayments. This showed
that to scale up the groundnut insurance would
require an associated scaling up of the nascent
marketing chain. The banks agreed that the
weather insurance tool would be more useful
in a commodity sector with stronger marketing
arrangements and oversight of farmers, such
as tobacco, and possibly paprika, tea, coffee
and cotton, where similar marketing chains are
developing.

Thus, during the 2007–08 season, the pro-

gram moved to the tobacco sector and ceased
operations in groundnuts. Malawi’s tobacco
sector currently represents the largest pool of
recipients of credit in the country; demand
for credit is high because of the input needs
associated with tobacco production. Because
all tobacco in Malawi is sold through auc-
tion, there is a constriction point that allows
banks to recover loans directly before farmers
receive their sales proceeds. This creates more
certainty for lenders, who have an assured
and trusted mechanism for recovering loans.
Traditionally, maize inputs are also included
in farmer tobacco loan packages.

Index insurance, development and disaster management

In 2007–08, weather insurance covered

a portfolio of loans jointly held by a tobacco
processing/trading company, Alliance One,
and the Opportunity International Bank
of Malawi (OIBM), rather than individual
loans held by farmers. OIBM bought an
index-based weather insurance policy from
the Insurance Association of Malawi (IAM)
in November 2007. The policy covered
Alliance One tobacco farmers within 30 km
of Lilongwe and Kasungu weather stations.
Though the policy was a portfolio policy
designed to cover the OIBM and Alliance
One’s exposure, the contract was based on
individual insurance so that the companies
could easily associate payouts triggered by
speciﬁc stations with speciﬁc farmer groups
and crops. OIBM and Alliance One shared
the cost of the insurance with the farmers,
but did not engage in detailed communica-
tion with them on the nature of the product,
given that this was a pilot. Since the value
of the initial tobacco transaction was greater
than the sums insured under previous pilots,
the IAM was able to establish contracts with
the international reinsurance market for
these products for the ﬁrst time. A portion of
this risk was thus reinsured by IAM on the
international risk markets in late 2007.

All of the companies involved, and several

new players, expanded this program during the
2008–09 season. Both the portfolio approach
pioneered in 2007 and a farmer-level approach
were offered. While Alliance One chose to
pursue a portfolio approach again, Limbe Leaf,

a contract farming company, chose to offer the
product directly to their farmers. In addition
two more banks agreed to provide ﬁnancing for
these loans. The size of the program increased
signiﬁcantly, covering 2500 farmers and a total
value of transactions exceeding US$2 million.
There are plans to expand the program still
further for the 2009–10 season.

Lessons

The groundnut pilot revealed that problems
related to production, marketing and sales can
undermine credit repayment and hence the
effectiveness of the insurance policy. To make
insurance viable for this sector, complementary
investments are necessary to strengthen con-
tractual relationships. These will likely feature
additional ﬂows of resources, improved farmer
advice and oversight, and better links between
input provision and commodity sale. Efforts to
reduce side-marketing will also be important.

The conclusion for Malawi, therefore,

is that insurance should be closely linked to
more formal and better coordinated supply
chains. The aims now are to scale up index
insurance in conjunction with such chains,
and to expand into new areas and crops only
when these meet this criterion.

The low density of automated rainfall

stations in the country is a limitation to scal-
ing up in Malawi. For example, an estimated
110,000 smallholder tobacco growers curently
cultivate close to a reliable weather station. If
53 new rain gauges were installed, an addi-
tional 200,000 farmers could be included in

Case studies I

Food Security, staff of participating commercial
banks, contract farming companies and local
research institutes. The ADP-SP provides
funding for this activity.

For the pilot, nine insurance companies

worked together to underwrite the risk.
National regulatory approval was not required
during the pilot phase, although the regulator
closely observed the program’s development.
If the private sector is interested in expand-
ing the program, it will need to engage the
national regulatory authority so that it gains
a better understanding of the risks being
insured, and can make necessary changes or
additions to existing regulation and policies.
Support for the development of an appropri-
ate legal and regulatory framework is also
included in the ADP-SP.

the program. A government program sup-
ported by the World Bank and Norway, the
Agricultural Development Project – Support
Program (ADP-SP), is currently investing in
new weather stations and infrastructure.

The development of capacity and techni-

cal expertise in insurance will be critical for
continued growth of the sector. The pilot
program includes some training, but more will
be needed. Speciﬁcally, national players will
need to take a larger role in contract design and
underwriting. To date, contract design has been
carried out by the World Bank and its partner,
MicroEnsure, with limited support from local
experts. In the coming years, the IAM will
need to take over responsibility in collaboration
with experts from the Meteorological Services
Department, the Ministry of Agriculture and

Index insurance, development and disaster management

Disaster insurance in Ethiopia

attempted to demonstrate the transfer of
national drought risk to the global insurance
market. Although there was no payout that
year, as rainfall was above average, the project
successfully demonstrated the feasibility of
the concept. The policy was not renewed in
2007 due to lack of donor support; however,
scale-up is now under way, with a US$250
million contract planned for 2010.

The insurance targets a group described

as the ‘transiently food-insecure’ – some
5 million people who are directly at risk when
a drought occurs. The Ethiopian government
has a Productive Safety Net Program (PSNP)
that targets the chronically food-insecure,
that is, the poorest people who face ongoing
food insecurity whatever the weather. The

Ethiopia provides an example of how index
insurance is being used to revolutionize
disaster response. The current response system
waits for evidence that people are hungry
before appealing to donors and eventually
providing relief, which misses the opportunity
to act in time to prevent famine. However,
impending drought-related famine can be
detected, and used to trigger an index insur-
ance payout during the growing season. By
acting quickly, disaster response can be much
more effective at much lower cost.

The country’s ﬁrst national index-based

disaster insurance program was implemented
in 2006. Developed by a partnership of the
World Food Programme (WFP) and the
Government of Ethiopia, the pilot project

Some 5 million people in Ethiopia are directly at risk when a drought occurs; Daniel Osgood/IRI

Case studies I

transiently food-insecure, who are able to feed
themselves in a good season but suffer when
the rains fail, are largely overlooked. The
emergency appeal system that should help
them during extreme drought is usually too
slow in its response to prevent distress sales of
assets, and they then move into the chroni-
cally food-insecure group. Ultimately, this
makes the PSNP unsustainable – unless the
weather risk component can be addressed.

Index insurance provides the missing piece

of the puzzle, allowing timely relief to come
to those in need quickly when rains fail. In
this way, the PSNP program can focus on
its core task of supporting the chronically
food-insecure. Another advantage is that the
insurance product facilitates price discovery
for Ethiopian drought risk in international
ﬁnancial markets. Putting a price on drought
risk allows for better understanding of invest-
ment tradeoffs for mitigation/management of
drought risk.

The reinsurance company AXA Re won

the bid to provide the insurance for the pilot
project. The premium was set at US$930,000,
WFP signed the contract on behalf of the
Government of Ethiopia, and United States
Agency for International Development
(USAID) paid most of the premium. The
contract stipulated a maximum payout of
US$7,100,000 in the event of severe drought.
Although rainfall was above normal that year,
the pilot demonstrated the feasibility of index
insurance at this level. It included capacity
building with government and local partners,

strengthened the national meteorological
reporting network, and demonstrated that
there is sufﬁcient quality meteorological data
in Ethiopia for disaster index insurance.

The contract

The Ethiopia Drought Index (EDI) was
developed using historical rainfall data from
the national meteorological agency, and a
crop–water balance model. The index had an
80% correlation with the number of food aid
beneﬁciaries from 1994 to 2004, indicating
that it is a good proxy for human need when
drought strikes.

As the 2006 agricultural season pro-

gressed, rainfall was monitored daily at 26
weather stations across the country. Extension
ofﬁcers in the ﬁeld reported that the index
effectively tracked rains and crop growth
during the 2006 season; at the same time,
areas where the index and approach could
be improved were noted. Remote sensing of
rainfall was used to ensure that there was no
tampering with station-based rainfall data.

The EDI value at the end of the contract

period, 31 October 2006, was well below the
US$55 million trigger level, so there was no
payout (see ﬁgure).

Some basis risk was evident within the

contract. Standard Famine Early Warning
System Network (FEWS-NET) sowing
periods were used to calibrate the 2006
model, but these did not always correspond
with actual farming practices. For example,
the traditional cereal teff, used to make bread

Index insurance, development and disaster management

and for forage, was sown later than modeled.
In some cases it experienced water stress
because the rains ended early, but accord-
ing to the model the crop had already been
harvested and was not shown to be affected.
Standard growing cycles from FEWS-NET
and Food and Agriculture Organization of
the United Nations (FAO) were also used,
but in Ethiopia growing cycles vary greatly
according to altitude and local crop variety;
in some areas, for example, a 240-day variety
of sorghum is grown, but the model assumed
a standard 150-day variety. This also led to
discrepancies. Information on actual sowing
periods and growing cycles for the main crops
in the different areas was being collected for
future projects, to reduce this source of error.

LEAP

In preparation for the second phase of the
Ethiopia drought insurance project, WFP
and the World Bank have developed a piece
of software called LEAP, which stands for
Livelihoods, Early Assessment and Protection.

Based on the FAO Water Requirement
Satisfaction Index (WRSI), the software allows
users to quantify and index the drought and
excessive rainfall risk in a particular administra-
tive unit in Ethiopia. LEAP can then be used
to monitor this risk, and to guide disburse-
ments for a PSNP scale-up.

LEAP uses ground and satellite rainfall

data to cover the whole of Ethiopia, even
areas where weather stations do not exist,
so that every administrative unit in the
country can be included. It runs localized
models to convert rainfall data into crop or
rangeland production estimates and subse-
quently into livelihood stress indicators for
vulnerable populations. It then estimates the
ﬁnancial magnitude of the livelihood-saving
interventions these people need in the event
of a weather shock. Thus, LEAP provides a
good proxy estimate of the funding needs of
protecting transiently food-insecure people’s
livelihoods at a time of shock through an
independent, objective, veriﬁable and replica-
ble index of livelihood stress.

The Ethiopia Drought Index, 1952–2006.

Harvest year

Trigger

Index

Ethiopia dr

ought index v

alue (US$ millions)

1952

1962

1972

1982

1992

2002

Scaling up index insurance: The contract

Scaling up index insurance presents many
challenges. This section focuses on the issues
surrounding the design of insurance contracts
– issues that must be addressed if large-scale
implementation is to be feasible. Getting
the contract ‘right’ is critical – but there is
no shortcut to doing this. What is largely
an experimental product will require a great
deal of adaptation and validation to meet the
diverse applications and objectives that many
expect of it.

Scaling up in context

Expectations are high for index insurance
– perhaps unrealistically so. To successfully
scale up insurance for development objectives,
contracts are expected to be affordable and
easy to understand; to have minimal basis
risk; to provide extensive coverage in different
areas; and to deliver this performance against
a background of limited or non-existent
markets, limited capacity, weak distribution
and regulatory systems, and limited contract
design skills.

Contracts developed for development-

oriented pilot projects have usually been
highly tailored prototypes, ‘handcrafted’ by
experts and developed for very speciﬁc and
local needs. The main challenge now is to
efﬁciently develop contracts with basic design
principles that are well understood and gen-
eralizable, so that they can be transferred and

adapted to different places and conditions.
And importantly, the task of the initial con-
tract exploration, or ‘seeding’, and designing
must be manageable at the local level, so that
contracts genuinely respond to local demands
and changing needs. Training, and innovative
training technologies, will be vital.

Some of these challenges are less acute at

the disaster relief level, since just one contract
can address the risk for an entire nation.
Nevertheless, improvements in methodology

Scaling up will require capacity building at the local level;

Marjorie Victor/Oxfam America

Index insurance, development and disaster management

are needed, as there are many challenges faced
at this level as well, particularly associated
with integration into national disaster man-
agement programs and reaching the target
beneﬁciaries. Malawi provides an example of
how index insurance is being integrated into
national risk management to support national
food security (see ‘National-level drought risk
management in Malawi’ on page 42).

A considerable problem is the shortage

of historical and real-time data in the very
regions of the world that are most in need
of risk management solutions. Data are the
essential building blocks of index insurance
– they are needed in order to build a loss
history, to link this to an index, and to work
out loss probabilities as a basis for pricing.
They are key to identifying and reducing
basis risk. Solutions to the data problem may
seem straightforward – it is relatively easy to
install new rain gauges in remote areas, for
example. But historical data are also essen-
tial, and these will still be lacking. There are
potential solutions to this problem – histori-
cal datasets can be simulated with the help of
existing weather stations, for example – but
they are not always simple, and some depend
on advancing current science and developing
new technologies.

The beneﬁts that index insurance offers

for both development and disaster manage-
ment have, not surprisingly, attracted the
attention of professionals in these ﬁelds.
Many are advocating scale-up as the next
stage – but this must be accompanied by a

caution. We do not yet fully understand the
behavioral responses to index insurance, or the
impacts it may ultimately have. There is an
urgent need for impact studies, as discussed at
the end of this section.

There is also a danger that eagerness to see

beneﬁts from index insurance may cause pilots
to be scaled up too rapidly, before appropri-
ate methodologies have been developed or
before supporting markets and systems can
be strengthened. While funding is needed to
expand and replicate projects, investments
must be made carefully, allowing products
to scale up at a strong but realistic rate and
to adapt as new knowledge is generated and
shared. In parallel, time and resources must
be invested in capacity building at the local
level. It will be important to strengthen the
innovation system as a whole, so that index
insurance can play its part in a robust portfo-
lio of responses to climate risk that include a
range of management options.

The rest of this section looks at contract

design against the backdrop of these chal-
lenges. It is structured according to the main
tasks faced by the designers of index insurance
contracts, namely:

A ﬁnal section discusses the need for

evaluation studies to answer the many ques-
tions surrounding index insurance and its
impacts.

Scaling up index insurance: The contract

Identifying and quantifying the risk

At the core of contract design is determining
what risks should be addressed and how these
might be addressed through an index. The
contract design question is not how to cover
every risk, but which risks can best be covered
using this ﬁnancial tool, and when this tool
is more cost-effective than other options for
managing a particular risk.

The ‘right’ risk that index insurance can

address is one that is an important livelihood
constraint, is not adequately addressed through
other options, and can be closely correlated
with an index that can be reliably measured. It
is important to remember that there are many
risks that cannot be addressed through index
insurance. Also, the insurance must respond to
customer needs and demands.

To identify the right risk for farmer-level

projects, a unique blend of local knowledge and
science is needed. Interaction between clients
and insurance experts as well as good demand
assessments are key. In Ethiopia, a pilot project
implemented by Oxfam illustrates how the risk
can be identiﬁed through extensive stakeholder
interactions, and particularly by listening to
farmers (see ‘A farmer-centric approach in
Ethiopia’ on page 44).

Determining the right risk to meet client

demands essentially involves analysis of the
existing risk management portfolio – the
coping mechanisms, management practices
and technical or institutional innovations
already available – and the gaps. Agricultural
systems modeling can be useful as a way to

integrate the multiple sources of informa-
tion needed to understand risks, and to
build an index that minimizes basis risk for
crop- or livestock-related applications (see
‘Agricultural systems modeling’) As with any
technology, it is important to use agricultural
systems modeling appropriately, ensuring
that models are well calibrated and validated.
Ultimately, the client is the only person
who possesses a detailed understanding of
the risks they face and of their alternative
risk management options, so communica-
tion efforts that facilitate the sharing of
this knowledge are critical if contracts are
to address clients’ real needs. As projects
prepare to scale up, a key lesson from experi-
ence so far is that a high level of stakeholder
interaction needs to be ongoing to improve
contract design. For example, in India the
contract developed by BASIX and partners
has been substantially altered following
farmer feedback sessions.

For social protection and disaster manage-

ment purposes, identiﬁcation of key risks is
somewhat easier. Major threats to lives and
livelihoods at the local, regional and national
scales are well known and are already the
focus of national disaster management efforts.
Thus identifying the risk falls upon the
disaster management system.

In most of the farmer-level case studies,

the task of identifying the risk and choos-
ing the index has been addressed in part
through a participatory stakeholder process.
Index designers model potential indices

Index insurance, development and disaster management

Agricultural systems modeling

Agricultural systems modeling was ﬁrst developed in the 1960s and 1970s to help understand interactions
between the environment and crop and livestock systems. It has evolved so that today very complex systems
can be modeled. Indeed, systems modeling is not limited to biological processes but can also look at climate and
livelihood systems, for example.

These models can be used to produce indices that better correlate with crop production than simple rainfall

indices. This is because models better reﬂect the ‘real-life’ situation, where yields depend not simply on the
amount of rain but on interactions between the weather, soil–water–nutrient dynamics, crop management and
crop physiology. Better correlation reduces basis risk, making these models attractive for index insurance contract
designers. Accuracy can be improved by incorporating remote sensing of vegetation conditions or soil surface
moisture into modeling. Agricultural systems modeling offers an array of tools of varying complexity, from simple
water balance models such as the FAO Water Requirement Satisfaction Index (WRSI) to sophisticated process-
based models such as the Decision Support System for Agrotechnology Transfer (DSSAT) suite.

There are a few examples of agricultural systems models being used in index insurance. The Agriculture

Insurance Company of India (AIC) used the INFOCROP model developed by the Indian Agricultural Research
Institute (IARI) to model weather–yield relationships and design weather indices for different crops and locations
in 2007–08 (see ‘Scaling up in India: The public sector’ on page 87). WRSI-related products have been used as
indices in the Ethiopian government’s disaster relief contract as well as some of the Millennium Villages Project
(MVP) contracts. WRSI is based on a relatively simple model that predicts water-limited crop yields by looking at
evapotranspiration (actual and potential) against crop sensitivity during the different growth stages. The WRSI is
essentially a rainfall index, but one that also takes account of how crops respond to rainfall. At the other end of
the complexity spectrum, an index insurance product based on DSSAT yield predictions has been used for maize
production in southern Georgia, USA (Deng

et al., 2008).

Agricultural systems modeling can also be useful in modeling the linkages between the index and the risk,

as well as in understanding the role insurance might play among the suite of risk management options. Another
use is to demonstrate the potential beneﬁts of index insurance by modeling how improved access to resources
such as fertilizer might impact production and income. Models can also estimate optimum levels of inputs to
production systems, and hence levels of credit (or inputs) that might be usefully bundled with insurance. Finally,
models can help to understand how seasonal forecasts may affect index insurance and agricultural management
decisions (see box ‘Seasonal forecasts’ on page 35).

As with any tool, agricultural systems modeling is not a panacea, and it will produce unreliable results if used

inappropriately. It is important not only to validate the model with past outcomes but also to physically verify
the predictive capacity of the model during implementation. When considering use of a crop or forage model
as the basis for an index, one should understand its capabilities and limitations, and also understand the system
being modeled, consider the levels of accuracy needed, evaluate model performance for the given application,
and ensure that calibration is performed if needed, ideally by an independent technical institution. Discussion of
model assumptions and outputs with farmers and local experts can be very valuable in this process. For example,
in Tanzania it was found through validation with stakeholders that the most vulnerable period for maize occurred
a few weeks later than had been modeled. If the climate and agricultural knowledge of farmers and local experts
had not been included, this could have led to contracts with a high level of basis risk. Closing the loop through
responsible agricultural systems modeling and communications yielded a much improved product.

Adapted from Baethgen

et al. (2008).

Scaling up index insurance: The contract

with stakeholders, allowing for feedback on
critical issues, particularly trade-offs such as
how much (if any) of the premium should be
dedicated to addressing competing risks, or
between payout size and payout frequency.
Experimental approaches are being explored
as possible methodologies for index selection.
In Kenya, Malawi and Ethiopia, farmers
have played games based on index contracts,
using real money. The farmers select between
the alternative versions of the index contract
being considered by designers, under simu-
lated weather conditions and events. During
the game, the farmers are asked questions
about how similar the game is to their own
experiences, to quantify agreement and
mismatches between farmer experience and
other information sources (Patt et al., 2008).
For scale-up, methodologies will need to be
established to perform this time-intensive
exercise more systematically and efﬁciently.

The tailored approach to identifying and

quantifying the risk has been effective in
pilot projects but was at times very costly (see
‘Participation: Key to farmer uptake?’). It is
critical that efforts continue to streamline
these procedures, relying more on local
champions, to build more replicable indices.
The challenge is to balance efforts to lower
the costs of scaling with the need for reliable
and effective contracts.

Measuring index parameters

Once a contract period has begun, the
parameter indexed needs to be measured,

with data that are reliable and tamper proof.
Measurements need to be made close to
the insured location so that they accurately
reﬂect weather events at that location. With
pilots, these requirements have been relatively
easy to meet; with larger scale projects it is a
different matter.

For a rainfall-based index insurance

contract, data are typically obtained from a
meteorological station. Requirements are that
the station should be managed by the national
meteorological service or a reliable private
provider, should meet international weather
measurement standards such as those set
by the World Meteorological Organization
(WMO), and should be secure, preferably

Collecting rainfall data; Stephane de Messieres/

Oxfam America

Index insurance, development and disaster management

Participation: Key to farmer uptake?

Ultimately, scaling up of index insurance for poverty reduction and development will depend on a large number of
farmers choosing to take up the insurance. This must be an informed choice based on a clear understanding of the
insurance – what it covers, what it does not cover, and what it offers vis-à-vis other risk management options. This
is no simple matter. Index insurance is new and can be complex, and farmers (as well as other stakeholders) have
a difﬁcult time understanding it. Farmer decision making is also not a simple process. The concern is that farmers
may, for these reasons, make a decision that is against their own interests and miss an opportunity to reduce the
risks they face.

Participation may be the key. By involving the potential beneﬁciaries of index insurance projects in design and

implementation, it is hoped that their input and assistance will both improve the design and create greater support
for the project within the target communities. The interactions during these stages should also enhance trust in the
project and among its participants.

All of the pilot projects invested a great deal of time in interaction between stakeholders. This has been a two-

way process, helping implementers to understand the needs of farmers as well as helping farmers to understand
the insurance and its potential beneﬁts. At the same time, alongside some of the pilots, there has been research
into farmer decision making. This has yielded some interesting and sometimes unexpected results, which have
implications for scaling up.

Both economic and non-economic factors play a role in decisions whether to buy insurance. From an economic

perspective, researchers found that farmers did not always make the ‘obvious’ choice – the one that would be most
likely to beneﬁt them economically in the long run. Instead, the decision depended on factors such as the cost of the
premium, the timing of the premium and the payouts, and the bundling of the insurance with other inputs in a loan
package. In Ethiopia, for example, farmers explained that they only had cash available to pay the premium just after
the harvest, so efforts to sell premiums to them at other times would likely fail. Farmers in both Ethiopia and Malawi

Many factors inﬂuence farmer decision-making; Eric Holthaus/IRI

Scaling up index insurance: The contract

automated. To make the insurance product
workable, insured farmers must usually be
located close to such stations. However, it is a
fact that there is a shortage of these stations,
particularly in the regions of the world where
index insurance might prove most useful for
development and disaster relief. This needs to
be addressed as a fundamental constraint to
scale-up.

Projects in India and Mexico came up

against a shortage of weather stations as
they began to scale up (see ‘Insurance for
contract farming in India’ and ‘Disaster relief
in Mexico’, on pages 47 and 49). Both have

addressed this problem by building new
private stations. In India, some 1200 new
stations have so far been set up, by private
and public companies. In Mexico, Fundación
Produce is building some 764 automated
stations. Alongside the urgent need to
expand the network of weather stations,
whether through public or private sector
efforts, capacity building and resources to
manage both the stations and the data are
also required.

The Caribbean Catastrophe Risk

Insurance Facility (CCRIF) is currently
investigating the feasibility of including

conﬁrmed that it was important for them to receive payouts at roughly the same time that cash from crop
sales would have materialized, as they had arranged their ﬁnancial commitments around this period.

But researchers also found that non-economic factors often override economic considerations.

Investigations in Malawi, Ethiopia and India all show trust to be a crucial factor in guiding people’s decisions
over insurance – trust in the product itself, and in the organizations involved in selling and managing it.
People are more likely to buy insurance if they see other members of their community buying it; and their
own prior experience with insurance is most important of all. In particular, this translates into willingness
to buy if there has been experience of a payout.

There are many other factors that inﬂuence decision making, some at the individual or household level,

for example a potential customer’s perception of risk and degree of risk aversion. All of these factors will
be very relevant in efforts to scale up index insurance. Indeed, the researchers who investigated decision
making in the pilot studies strongly recommend that similar levels of interaction are included as projects
are scaled up and out. They believe that participation has a crucial role in developing trust in the institutions
involved, in the product, and also in the clients’ own decisions to purchase the insurance – and without
this trust there may not be sufﬁcient uptake. Of course, a balance must be established between the level
of interaction and the cost of interaction at large scales.

The same researchers also stress the importance of regulation, by the government, international

bodies and the insurance industry itself. Regulation needs to ensure that contracts are fair, that they
are accompanied by transparent information, and that claims will be paid promptly. It is crucial that the
commercial actors in index insurance markets are honest players, and that their actions contribute to trust
in the product, rather than destroying trust. A single bad example could, the researchers say, set index
insurance development back by many years.

Adapted from Patt

et al. (2008).

Index insurance, development and disaster management

rainfall in its catastrophe insurance, following
Hurricane Noel which caused severe damage
due to ﬂooding that was not covered in exist-
ing contracts (see ‘Catastrophe risk insurance
in the Caribbean’ on page 52). However,
adequate rainfall data, both historical and cur-
rent, are limited in the region. As a solution,
a new ‘extreme weather monitoring network’
has been proposed which will record extreme
rainfall and wind events and act as a veriﬁca-
tion mechanism.

There are other ways to address the lack

of data, especially for disaster management
applications. Remote sensing can be useful
for veriﬁcation purposes and perhaps even as
a source of data for the insurance index (see
box, ‘Remote sensing’, page 28). Remotely
sensed data can either be blended with
ground-collected data to improve overall data
quality, or used alone, if they have already
been validated and shown to reﬂect conditions
on the ground accurately. In the Ethiopia
disaster relief project, for example, remotely
sensed rainfall data are used to extend the
index to areas where there are insufﬁcient
ground data. In the Millennium Villages
Project (MVP) remotely sensed greenness is
used to enhance an index based on rainfall
data (see ‘Insuring development goals in the
Millennium Villages Project’ on page 56).

One advantage of remote sensing is

that data cannot be tampered with by the
client. Even when such data are not accurate
enough to be directly used to measure the
index, they can be very useful for validating

ground measurements and conﬁrming that
the latter have not been tampered with.
This validation is particularly important in
national-level projects where an index meas-
ured by government weather stations may
determine large payouts to that government.

Typically, products involving remotely

sensed data have been used by governments
or intermediaries rather than in farmer-level
projects because it is difﬁcult for farmers
to relate the performance of their crops to
such intangible indices. Farmers also may
not trust the insurance company to use such
data responsibly in calculating payouts. These
concerns may change as farmers become
more experienced with index products and
the level of trust builds between the insurance
company and the farmer. The micro-level
Oxfam America-led project in Ethiopia is
using remotely sensed rainfall for farmer-level
contracts until newly installed rain gauges can
be sufﬁciently calibrated; this project is devot-
ing signiﬁcant time and energy to engaging
farmers and building trust and understanding
of products.

Spatial basis risk – the failure of the

index to represent individual losses because
of differences across locations – is a major
challenge to scale-up. It is correlated with
the distance between the weather station
and the insured location, but the acceptable
distance varies according to factors such as
homogeneity of the landscape. An arbitrary
maximum distance of 20 km has often been
used in pilots for rainfall index products,

Scaling up index insurance: The contract

Remote sensing

Data from satellites can be used in index insurance as an alternative or a supplement to ground-collected data.
These remotely sensed data have several advantages. Data are independent and tamper-proof, for example, and
are available across large areas of the Earth and in near real-time. They, and their derived products, are easily
available via the internet. However, the data are not direct measures but proxy measures of, for example, rainfall or
vegetation. This means they come with a level of uncertainty. The majority of data from satellites is relatively ‘coarse’,
i.e. low quality and low resolution; the satellites with highest resolution often do not have global coverage and
are usually very new. And many satellites are deployed for temporary research projects, so collection of long-term
data may not be part of their program. Given these caveats, there is a great deal of promise in remote sensing
if effort is taken to research, validate and improve products. Remotely sensed data may be particularly useful
coupled with other types of information; and offer a valuable fallback option when a ground-based observation
station fails during the contract period.

Rainfall

There are two main ways of estimating rainfall from satellites. One uses thermal infra-red imaging to look at storm
clouds and their height. The assumption is that more rain falls from storm clouds with deeper vertical extent. The
height is related to the temperature of the top of the cloud, which is obtained from the infra-red images. The
colder the temperature, the higher the cloud top. Rainfall is estimated based on the length of time the cloud top
is below a certain temperature.

Passive microwave measurement offers an alternative way of estimating rainfall. The sensors measure actual rainy

areas rather than clouds, making this a more accurate method. However, these sensors are not currently available
on geosynchronous satellites, so fewer data are available (in both time and space). (Geosynchronous satellites have
orbits that match the earth’s rotation, keeping them above the same spot on the earth all the time.)

Data from these two approaches may be merged to combine the better rainfall identiﬁcation of the passive

microwave method with the better sampling of data from the thermal infra-red images. These data can also be
blended with rain gauge data to further improve accuracy.

Remotely sensed rainfall data are potentially very useful where there are limited rain gauge data, because of few

weather stations or uneven distribution of gauges, for example. This is often the case in rural parts of developing
countries – which are precisely the areas that index insurance is targeting with development objectives. However,
the level of uncertainty with these data is high, and they often fail to adequately reﬂect the actual amount of rainfall.
Thermal infra-red sensors do not allow for variation in rainfall intensity under the clouds; rain may fall from lower
clouds, for example near coasts or in mountainous areas; and cirrus clouds may incorrectly indicate rainfall because
of their height, whereas in fact they are not deep enough for rainfall to develop. Passive microwave limitations are
mainly due to background emission from the land surface, which varies for different vegetation and soil water
content, as well as to the relatively few data available, including historical data.

There are ongoing efforts to extend and enhance archived satellite rainfall data to produce useful time series.

Other plans for the future include a satellite mission with advanced passive microwave and radar instruments,
which should greatly improve satellite rainfall estimation (the Global Precipitation Measurement Mission,

http://gpm.gsfc.nasa.gov/).

Adapted from Dinku

et al. (2008).

Index insurance, development and disaster management

Vegetation

The vegetation measure that is most commonly used for index insurance is the Normalized Difference Vegetation
Index (NDVI) (although new, improved vegetation measures are on the way). The NDVI is derived from satellite
measures of ‘greenness’, which reﬂect the level of photosynthesis in the vegetation on the ground. Higher values
of NDVI indicate greater vigor and amounts of vegetation, and index insurance contracts generally insure against
a decline in NDVI over an area, which would usually be due to drought.

The assumption with NDVI is that crop (or forage) production correlates with remotely measured vegetation

vigor over a given area and over time. This usually works well for forage, but is less straightforward with food crops
such as maize, soybean and wheat. Here, ﬁnal crop yield (seed production) is affected by factors such as nutrient
availability or water stress during critical growth stages, which may not be reﬂected in average ‘greenness’.

Nonetheless, NDVI has been piloted for index insurance in different parts of the world, with some success. In

the USA, where crops are grown on a large and uniform scale which enhances the NDVI correlation, the Vegetation
Index Plan of Insurance is being trialed for pasture, rangeland and forage (http://www.rma.usda.gov/policies/
pasturerangeforage). The Millennium Villages Project (MVP) has developed index insurance using NDVI combined
with a rainfall index to insure against maize losses.

Where crops are mixed with natural vegetation, such as over much of Africa, monitoring crop status with the

NDVI is more of a challenge. In this case, NDVI is often not used to measure yields directly, but rather to detect a
drought that would cause losses. The MVP index used the average NDVI over a large region to detect the browning
of natural vegetation during times of year when a lack of rainfall would severely damage crops.

Many NDVI products are now available through the internet. Some datasets extend from the early 1980s,

providing very useful time series. Reprocessing of older data has improved resolution, while sensors with ever
higher spatial resolution are also becoming available – currently down to 10 meters, allowing the identiﬁcation of
very precise areas with pasture or crop problems. The US Department of Agriculture is also developing an NDVI
product with 9-hour latency and 500-meter resolution, which will enable rapid analysis of crop conditions.

Adapted from Ceccato

et al. (2008).

The HARITA project in Ethiopia is using satellite data to complement ground

data (see page 44); Google, LeadDog Consulting, MapData, TerraMetrics;

prepared by Daniel Osgood and Chris Small/IRI

Scaling up index insurance: The contract

and up to 50 km for products based on
temperature and humidity; but analysis is
lacking on how to establish such distances.
Again, the implication is a need not only for
more weather stations but also for improved
understanding of the differences in climate
between locations and across regions.
Remote sensing may have a role here too,
to give a sense of the magnitude of these
differences.

The vulnerability of an index to spatial

basis risk also depends on the index itself
as well as the risk being targeted. Because
of the nature of drought, for example, it is
often relatively easy to develop indices for
this risk that are robust. Since drought is a
phenomenon that builds slowly over time,
differences in daily rainfall between locations
are often not critical; instead it is the longer
term total which is important. The problem
can be much more challenging for excess
rainfall, ﬂood, wind or frost insurance, in
which a single highly localized event may
drive the bulk of the loss. Where these risks
must be dealt with in the course of scaling up,
research is needed to quantify them in more
detail. For a discussion of index insurance
applications and challenges in water resources
applications, see ‘Applications for reservoir and
ﬂood management’ on page 31, and ‘Vietnam:
Flood insurance in the Mekong Delta’ on page
59. These challenges are particularly great
for farmer-level projects. For disaster relief
projects, contracts can be designed to pay
out when large aggregate losses are likely to

occur so that relief efforts can be ﬁnanced
without having to determine exactly where
the losses occur.

Establishing probabilities

Getting the index ‘right’ depends on a solid
understanding of the probabilities associated
with a given risk. This typically depends on
long datasets of acceptable quality, which
enable the likelihood of an extreme event,
the level of vulnerability and exposure, and
the losses incurred to be reliably estimated. A
key scale-up issue is the limited availability of
data, both historical and current; and ways to
circumvent this lack of data.

One common approach, known to insur-

ance designers as ‘historical burning cost
analysis’, is extremely transparent and can be
easily communicated to stakeholders. For this
analysis, the probability distribution of the
indexed parameter is calculated entirely by
observing past measurements. For example, for
a rainfall index, the past several decades of rain
gauge data are used to represent the set of pos-
sible rainfall events. These are then used to cal-
culate possible payouts. Although useful, when
applied without other analyses this approach
has signiﬁcant limitations. For example, one or
two major events can distort the probabilities,
while any event that has not happened in the
historical record is not considered.

Because of the limitations of historical

burn analysis, it is typically complimented
with rainfall modeling and simulation (see
‘Rainfall simulation’, page 34). Using data that

Index insurance, development and disaster management

Applications for reservoir and ﬂood management

Index insurance has so far mostly been used to insure against drought, but it has promise beyond this. Its
application to water resources, particularly reservoir and ﬂood management, is attracting interest as an alternative
or complementary strategy to traditional methods.

Reservoir index insurance can help smooth costs associated with shared demands on limited water resources.

Where urban needs compete with irrigated agriculture, for example, payouts to farmers in dry years would
compensate their losses. Experts envisage a combination of insurance, forecasting and adaptive operation strategies
as the way forward – though they emphasize that success depends on water managers being open to new ways of
working. Progress also depends on the development of effective indices, which is a signiﬁcant challenge. Reservoir
inﬂows have been used, but are susceptible to basis risk where there may be diversions upstream.

After drought, ﬂoods are the next biggest threat to agriculture – does index insurance have a role in managing

ﬂood risk? Work is still experimental, and ﬂood is clearly more problematic than drought to link accurately to an
index, but there is some potential.

Disasters arising from ﬂoods may also be managed using index insurance, with the same advantages – rapid

payouts, not dependent on actual assessments of damage or destruction – as other disasters. Again, the challenge
is to develop an appropriate and effective index.

Southeast Asia has the highest number of people affected by ﬂooding, and studies to investigate the feasibility

of index insurance have been carried out in Thailand and Vietnam. The Thailand project examined farmer-level
insurance, and concluded that it was unlikely to be workable. Frequent, localized ﬂooding in the river valley would
lead to the scheme attracting only high-risk farmers and would also lead to very high premiums. Results from
Vietnam were more promising. Rather than targeting farmers, this project looked at the feasibility of ‘business
interruption insurance’, to be purchased by the Vietnam Bank for Agriculture and Rural Development. The bank
would lend money to rice farmers and would cover itself against losses in the event of ﬂooding early in the
growing season. This may be trialed in the near future, with a contract based on recorded water levels at a main
river gauge station (see ‘Vietnam: Flood insurance in the Mekong Delta’ on page 59).

These two studies provide some useful lessons for those working on the development of ﬂood insurance.

First, farmers do to some extent manage their own risk exposure to ﬂoods where these have some pattern.
Insurance is only feasible for infrequent, unpredictable and widespread ﬂooding. Second, there are signiﬁcant
technical issues that need to be addressed. Each ﬂoodplain is unique, and its speciﬁc characteristics need to be
considered. Upstream development can affect ﬂood events downstream, for example. Deﬁning agricultural ﬂood
events requires a good understanding of various aspects of ﬂooding – extent, duration, depth, etc – and their
relationship to the different stages of crop growth. As a crop grows, the critical thresholds at which a ﬂood event
results in damage also change. This issue of timing makes modeling of ﬂood risk for agriculture very challenging.
However, remote sensing may be a useful tool here, enabling rapid and objective assessment of ﬂood extent and
duration, even at high resolution. Application of these technologies is most feasible for river inundation ﬂood
rather than ﬂash ﬂood events.

Farmer-level ﬂood insurance has additional technical challenges because of highly localized variation in ﬂood

exposure and the potential for anti-selection against the insurer. Flood index insurance aimed at intermediaries such
as banks, or at governments, therefore seems more promising in the current data and technological environment
than schemes directly targeting farmers. Risk management projects need to take a more holistic approach in

Scaling up index insurance: The contract

working with intermediary clients like agricultural banks. It is important to understand the credit risk management
methodology used by these institutions, speciﬁcally related to how natural disaster risks are currently treated and
the broader policy environment in which they operate. A ﬂood insurance product can complement and support
more formal procedures for credit risk management such as interest or principal repayment scheduling.

Flood risk mapping and remote sensing have additional applications. Notably, risk mapping can provide

agricultural credit banks with an objective assessment of risks of default arising from such factors as ﬂood timing
and duration. Better mapping of ﬂood risks, linked to geo-location of borrowing farmers, should enable banks
to reﬁne their lending policies and to plan risk transfer or loan rescheduling policy and pricing. It may support
government departments in advising on farming practices, variety selection, etc. Further, detection of ﬂooded
areas using remote sensing can support objective ex-post compensation systems, where these exist.

Adapted from Block

et al. (2008)

Index insurance for ﬂooding faces technical challenges; V.J. Villafranca/IRIN

Index insurance, development and disaster management

are available, plus an understanding of the
variables that inﬂuence rainfall, the modeling
generates thousands of years of synthetic
rainfall data, which include events that are
possible but have not actually occurred in the
past. This approach can be used to extend
limited datasets, allowing more realistic indi-
ces to be developed. In Mexico, for example,
rainfall and temperature simulations are being
used to replace missing data and produce long
time series for estimating probabilities when
building indices. In 2008 these synthetic
series were used with actual (though limited)
data from 75 weather stations to insure farm-
ers growing crops on some 250,000 hectares
of land (see page 49).

Rainfall modeling has a range of complex-

ity, from simple distributions for rainfall to
sophisticated algorithms that factor in climate
models, multiple stations and remote sens-
ing. At any level of sophistication, modeling
rainfall accurately enough for index insurance
design and pricing is challenging, and research
is necessary to address the many known
problems.

For many locations where index insurance

could be useful, there are few or no histori-
cal rainfall data to ‘train’ a model, and other
approaches are needed to ﬁll the historical data
gap. Satellite-based remote sensing again can
be useful here (see ‘Remote sensing’, page 28).
Satellite imagery can be used to estimate aver-
age rainfall over a pixel as opposed to rainfall
at a particular point, as measured by a station.
Some datasets from satellites extend back

as far as 30 years, and efforts are under way
to enhance these. However, even enhanced
satellite data are relatively crude estimates of
rainfall, and work must be done to determine
whether they can be used to build sufﬁciently
robust indices.

One way of arriving at proxy historical

rainfall and losses and of complementing exist-
ing datasets is to interview farmers and experts.
Currently, most pilot projects hold focus group
interviews even when historical data are good,
to validate the link between client percep-
tions and what is measured at the rain gauge.
Although the data from these interviews
are qualitative rather than quantitative and
may have a great deal of inaccuracy, they do
provide an additional source of information on
historical events. Again, the issue for scaling up
purposes is whether this level of interaction is
necessary on a large scale, and whether it can
be replicated in large-scale projects.

Seasonal climate forecasts are relevant to

index insurance because they may alter the
probability of losses expected on the basis of
other data (see ‘Seasonal forecasts’, page 35).
There is a danger that seasonal forecasts
might be used strategically by clients, under-
mining the ﬁnancial stability of the products.
However, if insurance contracts can integrate
forecasts, the farmer can take advantage of the
forecast while being insured against forecast
inaccuracies. In theory, insurance contract
and project designers could combine the use
of forecasts with that of insurance to provide
price incentives for conservative behavior

Scaling up index insurance: The contract

Rainfall simulation

Models that simulate rainfall are very useful where there are limited actual rainfall data. Using the data that do
exist, and an understanding of the variables that inﬂuence rainfall, the modeling software generates thousands
of years of synthetic rainfall, which includes events that are possible but may not have occurred in the past. Thus,
the use of rainfall simulators can lead to more robust contracts and pricing.

Modeling also helps improve understanding of rainfall correlations between different locations within a region.

This understanding has the potential to help insurers spread risks across areas (see page 36 and ‘Spreading risk’ on
page 37). A third role for rainfall models is to incorporate seasonal forecasts, decadal trends, and climate change
into index insurance contracts (see ‘Seasonal forecasts’ on page 35).

There are different types of rainfall models, and they vary in complexity. Generalized linear models are the

simplest, and have the advantage that they can account for climate change over long timescales. Nonhomogeneous
hidden Markov models are useful because of their ability to reﬂect real scientiﬁc phenomena, such as regional
atmospheric conditions. Nonparametric models have the advantage of more ﬂexibly describing relationships
between rainfall and other variables, because they do not rely on standard probability distributions. Finally,
mechanistic models are a relatively new approach to rainfall modeling in which the occurrence and movement
of rain storms and rain cells are modeled explicitly.

The uncertainty inherent in rainfall simulations is reduced where more actual data are available and used to

‘train’ the model. Also, as understanding improves of the variables affecting rainfall, for example, El Niño–Southern
Oscillation (ENSO) effects, models will become more reliable at predicting future rainfall.

As with any tool, rainfall models and simulations have their limitations, particularly since they were not

developed for index insurance design. They often under-represent the true variation in rainfall, leading to indices
and prices that underestimate risk. They are also bad at representing seasonality, the beginning of a rainy season,
the end of a rainy season, dry spells and other features that are usually of importance in crop losses and index
payouts. Since these tools are often applied in pricing and design, research is needed to improve their performance
for these types of rainfall features.

Adapted from Shirley

et al. (2008).

in years with forecasts of bad weather, and
for productive investments in years where
such investments face low levels of risk
(Osgood et al., 2008). However, a great deal
of methodology must be developed before this
combination can be implemented in practice.
Similarly, long-term climate trends can also
be built into index products – in theory.
However, the current level of technology
available for updating insurance probabilities

and for distinguishing between short-term
processes and long-term change are in their
infancy and a great deal of work remains to
be done to provide workable techniques for
practical application.

Estimating the price

The price of insurance reﬂects the prob-
abilities of payouts, which are in turn driven
by the probabilities of adverse weather events

Index insurance, development and disaster management

Seasonal forecasts

Forecasts of the weather for the season ahead are becoming increasingly skilful, particularly in some tropical
regions, and this has implications for index insurance. On the one hand, forecasts could undermine insurance
schemes as farmers might decide to insure only when the forecast is for a bad season. On the other, if insurance
contracts can be designed to incorporate the seasonal forecast, this could enhance decision making in response
to the forecast. For example, where a seasonal forecast predicts a wetter than normal season ahead, a farmer
might take advantage of the likely good growing conditions by purchasing inputs – but at the same time use
insurance to protect against the lesser likelihood of a drier than normal season. In this way index insurance can
help manage the uncertainty inherent in probabilistic forecasts.

The Malawi case study ‘Unlocking development potential in Malawi’ on page 13 helps to illustrate this. El Niño,

La Niña and neutral years are identiﬁed using historical rainfall data, and insurance cost and size of the input loan
are adjusted as if these had been forecast. Premiums decrease in La Niña years, when there is very little drought
risk and farmers would have been able to cover a signiﬁcantly larger loan. The ﬁgure shows how gross revenue
can greatly increase due to these inputs in La Niña years.

This is a very simple example, and seasonal forecasts have not yet been integrated into index insurance

contracts in practice (although planned contracts on sea surface temperatures in Peru will put some of the key
pieces in place). More work is needed to better understand the effects of a forecast on the insurance, and especially
people’s behavior in the light of a forecast. Tools are also needed to help design and price insurance that integrates
a forecast. Climate forecast products that can be readily applied to insurance would also facilitate integration.

Different strategies might be developed to address the issue of seasonal forecasts depending on the ability

of clients to respond to the information in the forecasts. Where there is no potential to change activities in the
light of a forecast, strategies include closing contract sales before forecasts are available, multiple-year contracts,
or selling options on the right to purchase the insurance. But if farmers have the capacity to make better decisions
based on the forecast it would be appropriate to integrate forecasts into contracts.

Rainfall simulation methods offer a way of integrating forecasts into insurance contracts. Forecasts are treated

as variables, and rainfall is simulated for different forecasts. Contract design optimization and pricing can then be
based on these simulations.

Agricultural systems modeling can be used to better understand how people might change their decisions

in the light of a forecast. The assumption is that farmers would intensify production under anticipated favorable

100

120

Harvest year

oss r

enue

(thousand MK

)

1984

1994

1986

1996

1988

1998

1990

2000

1992

2002

ENSO-based

Standard

Scaling up index insurance: The contract

reﬂected in the index. These probabilities
must be accurately assessed so that prices are
fair to both buyers and sellers and properly
reﬂect the costs of transferring the risk. The
challenges this entails have been discussed in
in the previous section.

Arriving at a workable price is clearly

critical to scale-up, where the aim is a product
that will be bought and sold on the open
market. Yet negotiating a price is particularly
challenging because there is a great deal of
uncertainty over probabilities. There is also
intense pressure to provide products that have
a great deal of coverage at a low price.

Ultimately, prices are determined through

tough negotiation between clients, insurers
and reinsurers. Deciding how uncertainties
should be handled is thus a complex matter
involving the perceptions, trust and negotiat-
ing skills of these actors. If these commu-
nications and trust are not built, large-scale
projects may pose more problems than pilot
projects, where the sums of money at stake
are low and negotiations come under scrutiny
from the international community. But even
small-scale projects have had difﬁculties,
notably in reaching agreement between
insurer and reinsurer. Transparent and

comprehensive documentation and improved
methods for risk characterization for the
reinsurer should reduce uncertainty and keep
prices realistic.

Scaling up can open up options that

reduce prices. From the risk-carrier’s view-
point, errors may cancel out, while aggregat-
ing and pooling risks can reduce variability
for reinsurers, thereby reducing their need for
a safety margin in pricing. Also, in scaling up
it may be possible to access larger pools of
capital, which can help cushion against poor
proﬁtability in the early stages.

In some cases, the climate itself can be

sufﬁciently predictable to allow the price
of insurance to be reduced. Particularly in
the tropics, global processes such as the
El Niño–Southern Oscillation (ENSO)
often lead to strong spatial weather pat-
terns. Drought in some parts of the world is
often associated with ample rainfall in other
regions. Experiences in the case studies show
how this could work – for example, payouts in
Southern and Eastern Africa typically occur
in different years, linked to ENSO status. In
effect, different regions can insure each other
through a central pool, as the world’s climate
provides a natural hedge (see ‘Spreading risk’).

conditions and be more conservative under a forecast of adverse conditions. Models include additional
farm-level factors that inﬂuence decision making, including farmers’ goals, constraints and attitudes
to risk. These models can help develop insurance packages that exploit forecasts to support farm
management.

Adapted from Robertson

et al. (2008).

Index insurance, development and disaster management

Spreading risk

Risk spreading by insurance (and reinsurance) companies makes index insurance more affordable. Scaling up will
open up new risk-spreading opportunities, particularly as climate patterns and relationships across space and
time become better understood.

Risk spreading across space – geographic risk spreading – is feasible where complementary climate patterns

have been identiﬁed in different regions. In Africa, for example, a dry season in the eastern region is often associated
with a wetter season in the southern region, and vice versa. This observation is linked to the ENSO phenomenon – la
Niña events are associated with lower rainfall in eastern Africa and higher rainfall in southern Africa, while during
El Niño the reverse pattern is often seen.

Insurance companies could take advantage of this pattern by developing programs that cover both regions:

drought insurance contracts would be very unlikely to pay out simultaneously in both regions, thus reducing the
company’s risk exposure. Vicarelli (2008) illustrates the potential for premium savings through pooling of contracts
from Malawi and Tanzania, while Hess and Syroka (2005) have estimated premium savings of 23% if the regions
could be covered together in a single insurance portfolio.

Risk spreading may be particularly relevant with climate change ahead. Much of the damage from climate

change is expected to occur through extreme events, but it is unlikely that these events will simultaneously
blanket the earth; rather, they will occur in some places in some years and other places in other years, driven by
processes similar to those we observe today. The global climate can thus self-insure against the risks of extremes
from climate change.

From an institutional point of view, geographic risk spreading might translate into the development of large-

scale (national and multi-national) risk mitigation strategies as well as partnerships between national governments
and the insurance and reinsurance sector.

The Caribbean Catastrophe Risk Insurance Facility (CCRIF) is based on geographic risk spreading. Participating

countries pool their country-speciﬁc risks into one, better diversiﬁed portfolio. Since natural disaster risks in any
given year among the Caribbean islands are randomly distributed, the cost of coverage for the pooled portfolio
is less than the sum of premiums that countries would have to pay individually for the same coverage. In practice,
premiums are reduced by almost half (see ‘Catastrophe risk insurance in the Caribbean’ on page 52).

Adapted from Vicarelli (2008).

Climate processes, including climate

change and decadal processes, have implica-
tions for pricing (see ‘Climate variability
and change and index insurance’, page 38).
For example, when analyzing climate risks,
Swiss Re works to forecast future climate
conditions by observing prices of weather
risks that are traded in liquid markets and

by modeling, such as splitting the time series
into seasonal, climate change and residual
variability. They also use seasonal forecasts to
adjust pricing for conditions in the next few
months. Increased understanding of the links
between different climate systems and proc-
esses should feed into pricing in the coming
decades.

Scaling up index insurance: The contract

Climate variability and change and index insurance

To arrive at index insurance products that make effective adaptation tools – ones that are not undermined by
climate processes – it is essential to remember that climate is more than a long-term trend. Climate is a complex
system of anthropogenic and natural processes operating at yearly, decadal and long-term timescales. If climate
science and index insurance methodologies continue to improve, addressing the complete spectrum of climate
processes, it is likely that index insurance can be robust to climate variability and change and hence a useful part
of societies’ adaptation toolbox.

Today’s insurance need not be built to cover a loss 100 years into the future. Instead, insurance addresses the

coming year, and can therefore be adapted over time, improving as our understanding of climate improves. For a
given year, short-term climate (and market) processes dominate, with long-term trends representing only a small
fraction of the climate variability faced. However, over time, as the cumulative impacts of climate change start to
materialize, they can be reﬂected in insurance in ways that incentivize gradual adaptation.

Because the dynamics of climate vary from region to region, the implications of climate processes for index

insurance will not be uniform over the globe. For example, for the Sahel in West Africa, medium-term climate
processes are particularly strong – more than a quarter of the variation in measured rainfall is on a decadal
timescale (Greene

et al., 2008). In this region, it is especially important to build index insurance that is robust to

decadal processes (Skees

et al., 2008a). For Sahel index insurance contracts in the MVP, substantial work had to be

done to explore decadal scale insurance adaptation algorithms.

For the index insurance contracts in Nicaragua, the picture is different. The past 40 years of data do not suggest

any trend in rainfall (Osgood

et al., 2009). However, climate predictions by the Intergovernmental Panel on Climate

Change (IPCC) suggest that in Nicaragua there could be strong drying trends due to climate change. The ﬁgure
below presents the individual IPCC models (light lines) as well as their average (heavy line) for this region. It can
be seen that, over the next decade or so, the models provide little evidence of a drying trend. However, over the
next 100 years, most (but not all) of the models predict a slow reduction in rainfall, roughly 0.25% per year.

Nicaragua illustrates that the timing of

these processes must be taken into account
when developing insurance contracts. In the
near future, climate change is unlikely to affect
the probability of a drought. For approximately
the next decade, the yearly price of insurance
can therefore be expected to be driven by
year-to-year variability, because that is the
climate risk faced. During this period, farmers
should be able to use insurance to safeguard
investments that increase their productivity.
This should help them build wealth and
acquire the assets needed to allow them to
shift into other livelihoods if this becomes
necessary. In the further future, the probability
of drought increases slightly each year. As this

Precipitation at Managua, Nicaragua

2190

2000

2010

2020

2030

2040

2050

2060

2070

2080

2090

2100

2110

2120

2130

2140

2150

2160

2170

2180

-100%

-75%

-50%

-25%

25%

50%

75%

100%

% Change r

elativ

e t

ﬁrst 20 y

ears

Index insurance, development and disaster management

trend unfolds, the price of the insurance should gradually increase. Over time, the cumulative rise could incentivize
improvements in farming systems or a gradual transition to other livelihood activities.

For the projects in Ethiopia, the climate picture again looks different. IPCC predictions show little evidence

of any trend in precipitation (see ﬁgure). If anything, the models suggest increases in rainfall. In contrast to the
models, over the past decade, there are some sites in Ethiopia where the instrumental record shows decreased
rainfall. This is the case for Adi Ha, where the Oxfam-led pilot is located. Here the observed trend, a decline of about
1% per year, is strong enough to have increased the insurance price.

The difference between observations and

IPCC predictions is being debated by climate
scientists as they work to understand what
is likely to occur in the future, what is due to
climate change and what is driven by other
processes (Funk

et al., 2005). Many scientists

argue that the drying observed is not due to
long-term climate change but is a decadal
process. Research on this issue has fundamental
implications for the design of index insurance
and for the broader policy environment. If the
drying is due to climate change, it could signal
the importance of an eventual transition out
of the insured activities, or of fundamental
changes that enable these activities to be
viable with less rainfall. In addition, if the drying
is driven by greenhouse gas emissions, the
burden faced by the farmers is directly due
to those emitting the gases. However, if the

drying is part of a decadal cycle, it will be more important for contracts to be built into climate risk strategies that
protect against alternating decadal periods of drought and high rainfall.

Temperature is an additional, complicating element. There is a consensus that climate change will lead to

higher temperatures in Ethiopia, as in much of sub-Saharan Africa. This anthropogenic temperature increase
will increase the water needs of crops, leading to more severe and more frequent water stress. However, higher
temperatures could also lead to higher yields in years with large amounts of rainfall. Thus, on top of any impacts
climate change has on rainfall, its temperature impacts are likely to increase the variability faced by farmers and
hence also increase the value of insurance.

Although science does not yet provide us with deﬁnitive answers to these questions, what is currently known

is revealing important issues for index insurance design. Substantial effort needs to be put into advancing climate
science and index insurance methods, so that these can keep pace with climate change.

Adapted from Greene

et al. (2008).

Precipitation at Adi Ila, Ethiopia

2190

2000

2010

2020

2030

2040

2050

2060

2070

2080

2090

2100

2110

2120

2130

2140

2150

2160

2170

2180

-100%

-75%

-50%

-25%

25%

50%

75%

100%

% Change r

elativ

e t

ﬁrst 20 y

ears

Scaling up index insurance: The contract

Understanding impacts

Index insurance appears to be a promising
tool for improving the management of climate
risks. However, some fundamental questions
about it remain to be answered.

What is the impact of index insurance on

the incomes, assets and investment decisions
of those insured? If there is an impact, what
is its distribution in the population? Whom
does index insurance beneﬁt and for whom
does it not work? How does the design of
index insurance affect its impact? In summary,
what role can we expect index insurance to
play in development and disaster relief and
how can we design index insurance for differ-
ent socio-economic and physical conditions in
order to maximize its beneﬁcial impact?

To answer these questions, we need

rigorous impact evaluations of index insurance
programs and projects. Ideally, evaluations
should be integrated into program or project
implementation. Impacts of index insurance
for development and for disaster manage-
ment should both be addressed. Evaluation
methods must be unbiased towards projects at
different scales so that decisions based on the
evaluations are not themselves biased.

Because insurance for development

typically covers events that happen only once
in ﬁve to ten years, it will be important to
study impacts across different geographical
areas in order to increase our understanding
more quickly. While long-term data from
one place will be extremely useful and would
provide the most robust data for evaluation,

even short-term impact evaluations – which
combine analysis of limited data with models
of the physical environment and of behavioral
responses – can help to improve understand-
ing. However, the limitations of impact
evaluation methodology are often tested when
applied to index insurance. Although it may
not always be possible to disentangle the
impacts of insurance from the interventions
it is bundled with, it is important to at least
understand the value of the bundle.

Some of the pilot projects have included

evaluation, but this has usually focused on
adoption rates rather than impacts, and in
some cases the results have led to confusion.
In Malawi, for example, researchers compared
farmer take-up rates for two packages – a
standard insurance and loan package, or a loan
without insurance (Giné and Yang, 2007).
The research project made the latter option
available by itself guaranteeing the loan repay-
ment to the bank. Take-up rates for the pack-
age with insurance were lower than for the
loan only, reﬂecting either the fact that farm-
ers do not expect real repercussions in case of
default, or perhaps that farmers placed little
value on the insurance beyond its originally
intended role of giving access to credit. These
ﬁndings have often been misinterpreted as
showing the product to be unpopular or to
have little impact, whereas in fact demand for
the product was so high that it overwhelmed
the project delivery resources and associated
supply chains. The missing evaluation would
have compared impacts between farmers who

Index insurance, development and disaster management

had access to loans because of the insurance
and farmers who did not have access to either,
so as to determine whether or not their liveli-
hoods had improved.

In India, the International Crops

Research Institute for the Semi-Arid Tropics
(ICRISAT) and the World Bank conducted
two large surveys with Swiss funding in 2004
and 2006, covering 1000 farmers, many of
whom were repeat buyers of insurance who
had started buying in 2003. The results
showed that, while take-up rates had been

high overall, the unsubsidized premiums
to be paid up-front led to low amounts of
insurance bought by the BASIX microﬁ-
nance and livelihoods group, farmers and
self-help groups. Thus, behavior and invest-
ment patterns did not change much. The
survey also found that trust in BASIX as well
as education levels mattered a great deal for
the buying decision, whereas risk aversion
did not matter much (Giné et al., 2006).
Studies of this kind will be essential in the
future, to support scale-up.

Case studies II

Malawi provides an example of how index-
based risk management can be used at the
national level by governments and donors
to manage weather exposure as part of an
integrated and comprehensive approach to
managing risk. As part of the risk management
component of its Agricultural Development
Programme (ADP), the Government of
Malawi, with support from the World Bank
and the UK’s Department for International
Development (DFID), initiated a pilot in
October 2008 that transferred the ﬁnancial risk
of severe and catastrophic national drought to
the international risk markets. This involved

the purchase of a weather derivative contract
from the World Bank Treasury, which simulta-
neously entered into a contract with a leading
reinsurance company.

The ADP aims to strengthen maize

markets in the country so that farmers and
other stakeholders can respond effectively to
price and production shocks. The transaction
will provide predictable and early ﬁnancing,
in the form of a payout to the government,
in the event of a severe national drought.
The contract is based on rainfall measure-
ments and does not depend on actual maize
production. Payments can therefore be

Case studies II

National-level drought risk management in Malawi

The transaction will provide a payout to the Malawi Government if the maize crop in the country fails due to drought;

Curt Carnemark/World Bank

Index insurance, development and disaster management

triggered as soon as the contract ends in
April, rather than waiting for the harvest
assessments in June.

The weather derivative contract is struc-

tured as a put option that is triggered by a
rainfall index

The index links rainfall – its

amount and distribution during the grow-
ing season – to the expected level of maize
production in the country. If the index falls
to 10% below the historical average, the
government will receive a payout at the end
of the season, up to a maximum of US$5
million. For the ﬁrst trial year the sums were
relatively small, but they may be increased
in the future as experience is gained and
the role of this tool within the govern-
ment’s broader risk management portfolio
is better understood. The government plans
to use any payout to buy an option to cap
the import price of maize ahead of actual
imports, potentially saving large sums of
money if major imports are needed later
in the year and prices on the international
market rise. With future grain prices capped,
the private sector is better placed to respond
to domestic shortages by placing orders
for commercial imports. Strengthening
local market responses to production and
price shocks in this way is a critical part of
Malawi’s overall risk management and food
security strategy.

The index uses rainfall data from 23

weather stations throughout the country and

is based on the government’s own national
maize yield assessment model. This is a water
balance crop model, based on an FAO model,
that the government has been using since 1992
to forecast maize production each season. It
appears robust, having clearly picked up previ-
ous droughts in the country.

The World Bank Treasury’s intermedia-

tion facilitated access to the international risk
market.

This arrangement reduced start-up

costs for stakeholders and increased conﬁdence
in the transaction. The World Bank’s inter-
mediation role is temporary, and the goal is to
build capacity in Malawi so as to enable future
contracts to be negotiated directly between the
country and the international ﬁnancial markets.

Strategically, it is important to integrate an

index-based weather risk management activity
into a country’s overall risk management strat-
egy – that of the ADP, in the case of Malawi.
Anchoring risk management activities within
this type of larger investment program is criti-
cal to ensuring local ownership. Additionally, a
connection with the larger investment program
ensures integration with wider policy issues,
while multi-donor coordination can help
identify synergies with other programs and
activities. It will also allow the initiative to
transition from a pilot phase into a long-term,
comprehensive, risk-management strategy for
the government by supporting the develop-
ment of an appropriate institutional framework
for the optimal use of such tools over time.

Case studies II

A farmer-centric approach in Ethiopia

and methods that allow cash-constrained
farmers to pay for premiums with their labor.

Swiss Re’s role is to review and adapt

weather index insurance contracts for com-
mercial viability and conformity to market
standards. One of the world’s leading reinsur-
ance companies, Swiss Re has pioneered
weather risk transfer instruments in develop-
ing countries; in 2007, the company launched
the Climate Adaptation Development
Programme (CADP) to develop and imple-
ment weather risk transfer solutions in
non-OECD countries. OA’s role is to convene

Another index insurance pilot is being devel-
oped in Ethiopia by Oxfam America (OA)
and Swiss Re, in collaboration with IRI, the
Relief Society of Tigray (REST) and other
partners. Still at a relatively early stage, this
project is taking a farmer-centric approach,
and is working to integrate index insurance
with other risk reduction activities such as
improved agronomic practices, conservation
measures, and seasonal and daily weather
forecasting. Project innovations include the
extension of weather insurance to communi-
ties that are technically challenging to serve,

Farmers in Adi Ha; Eric Holthaus/IRI

Index insurance, development and disaster management

the various stakeholders from the local to
the global levels and facilitate a holistic risk
management model. IRI provides technical
expertise.

The project, which is called Horn

of Africa Risk Transfer for Adaptation
(HARITA), is initially targeting teﬀ farm-
ers in the village of Adi Ha in Tigray, with
expansion to other villages and crops planned
after 2009. OA and REST have been work-
ing with farmers in Adi Ha for more than a
decade. HARITA began in late 2007 with
visits by OA, IRI and REST to Adi Ha to
explore the potential for microinsurance and
gauge the community’s interest in a pilot.
OA commissioned an independent demand
study in the community during the dry season
of 2007–08. With farmers’ backing, in early
2008 OA contracted IRI to draft a prototype
weather index microinsurance contract, details
of which are currently under review. The
National Meteorological Agency (NMA)
is collecting meteorological data in Adi Ha,
aided by a new weather station purchased by
OA and installed by NMA in August 2008.

Lack of delivery channels for reaching

remote and inaccessible rural customers is
often a major obstacle to oﬀering microinsur-
ance. To overcome this challenge, the ﬁnancial
institutions involved in the pilot will employ
a partner–agent model. Dedebit Credit and
Savings Institution (DECSI), the second larg-
est microﬁnance institution in Ethiopia, will
act as the insurance agent. DECSI has very
extensive operations throughout Tigray, and

will harness its strong community relation-
ships and reputation to market and deliver
insurance on behalf of Nyala Insurance, the
primary insurance supplier.

A signiﬁcant feature of HARITA is its

eﬀorts to engage farmers as partners in index
insurance design. Many similar projects have
struggled to engage farmers due to the techni-
cal nature of the index and time pressures to
develop a commercial product in the ﬁrst year.
A team of ﬁve community members from Adi
Ha was recruited to join the HARITA project
management team. The project has conducted
workshops with farmers in the village to build
their ﬁnancial literacy. Recently, it carried
out experimental economic risk simulations
(‘games’) with the farmers to understand their
preferences for key parts of the insurance
contract, such as coverage and frequency of
payout.

The project is also working on ways of

overcoming weather data limitations. IRI
has led the exploration of new techniques to
enhance sparse local datasets through a com-
bination of satellite data, rainfall simulators
and statistical tools that interpolate data from
stations nearby. Satellite data will also be used
to improve understanding of the correlation
between rain gauge data and actual losses on
farms. With this information, the project may
be able to reduce basis risk by answering the
diﬃcult question of what is the maximum dis-
tance between farm and rain gauge for which
the rain gauge measurement of precipitation
is valid.

Case studies II

An advantage of this project is that it

builds on established relationships. Oxfam
has had a presence in Ethiopia since the
1960s, and thus has long-standing networks
of trust and knowledge of the country. These
relationships, together with Oxfam’s history of
high-proﬁle successes in other development
projects, paved the way for local partners’
willingness to experiment with the ‘radical’
solution of weather index insurance. Other
partners are also highly respected. REST has
an outstanding track record in high-impact
development, and communities place an
unusually high degree of trust in REST
following the life-saving assistance it provided
to Tigray during and after the Ethiopian civil
war.

The HARITA project complements

Ethiopia’s innovative social protection
scheme, the PSNP. This reaches approximately
8 million vulnerable people, about 11% of

Ethiopia’s total population. The PSNP pro-
vides payments to participating households
in exchange for labor to build community
assets such as water harvesting structures.
Such households tend to be chronically food-
and resource-insecure, and are likely to be
unable or unwilling to pay cash for insurance
premiums, despite ﬁnding risk management
highly relevant to their livelihood strategies.
HARITA is exploring ways to build upon
the PSNP model by enabling farmers to pay
insurance premiums in kind rather than in
cash. Under the scheme, farmers will have the
option of working a few additional days in
exchange for an insurance voucher that pro-
tects them against drought. OA’s focus group
interviews with farmers in communities across
the country suggest that many more people
may be willing to purchase larger amounts of
insurance if premiums can be paid for in labor
rather than cash.

Index insurance, development and disaster management

Insurance for contract farming in India

and oﬀers per kilo increments according to the
quality of potatoes, the use of fertilizers and
pesticides, and the purchase of index insurance.

2007, the insurance pays out based on the
number of consecutive days of average relative
humidity greater than 90% and/or average
temperatures between 10°C and 20°C. The
product is designed to cover yield losses above
40%, with farmers bearing losses up to this
point. Some 4250 farmers purchased the
insurance in 2007, and 4575 in 2008; approxi-
mately 50% were smallholders, owning less
than 5 acres (2.02 hectares) of land.

The ﬁrst season of 2007 revealed a certain

level of basis risk. In some locations the index
showed 85% loss, whereas actual losses were
roughly 50%; other locations received payouts
without suﬀering any loss; while in still others
the loss predicted by the index (45%) was
less than the actual damage (50–60%). In
response, more weather stations were installed
near farms. In the second 2007 season, basis
risk still occurred, but was reduced. Payouts
matched the observed damage in two loca-
tions, though for another two the index
showed lower levels than the actual losses. The
ﬁrst season of 2008 gave satisfactory results,
primarily due to a further increase in the
number of weather stations.

WRMS had installed 250 weather stations

by the end of 2008, and aims to operate a total

In 1995, to secure its supply of potatoes for
potato chips, PepsiCo started a contract
farming program in India. PepsiCo distributes
fertilizer, provides access to pesticides, and
requires contracted farmers to use their potato
seed. It also offers farmers technical advice
on production practices through a network
of agronomists, extension workers and local
facilitators. In 2008, PepsiCo contracted with
approximately 10,000 potato farmers across
the country; it plans to increase that number
to approximately 15,000 by the end of 2009.

Contracted farmers have the option of

buying a weather index-based insurance
product, which is sold by ICICI Lombard
and managed by Weather Risk Management
Services (WRMS). The product is intended
to protect against losses caused by late blight,
a fungal disease linked to temperature and
humidity, and the index incorporates both of
these variables. PepsiCo was motivated to add
index insurance to its contract farming pack-
age in order to limit farmers’ weather-related
risk and establish longer term relationships
with farmers.

Contract farming appears to be an eﬀective

way of involving Indian smallholders in a value
chain. PepsiCo farmers produce 11–14 tonnes
of potatoes per acre, compared with the average
on-farm yield of 8–10 tonnes/acre. Farmers are
oﬀered price incentives: PepsiCo sets a base
buyback price at the beginning of the season

Case studies II

of 400 weather stations by year-end 2009.
(In comparison, the Indian Meteorological
Department operates 600 weather stations
across India.) The cost of installing this new
infrastructure will be recovered through rev-
enue generated by the insurance program and
sales of services to commercial farmers and
to outlets such as newspapers, Reuters and
television stations. WRMS also sends weather
advisory messages and information on how
to prevent crop loss directly to farmers via
mobile phone. To cover all of India, WRMS
estimates that an additional 10,000 weather
stations would be needed, requiring an invest-
ment of approximately US$5–6 million.

The lack of historical data is more diﬃcult

to overcome, presenting a real obstacle to
scaling up the program. Further challenges
include designing products that balance good
coverage with an aﬀordable premium; secur-
ing delivery channels for product distribution;
expanding the limited target market – limited
because farmers who borrow from state banks
are required to take out the government insur-
ance; creating new indices that blend index
insurance and area-based insurance; and using
technology that more accurately captures

weather events and trends, such as NDVI. The
limited participation of reinsurers is another
constraint to scale-up. Reinsurance companies
are reluctant to enter the market due to the
diﬃculty of estimating its size and their risk
exposure.

Nonetheless, if these challenges can be

addressed, the potential to apply this contract
farming model to other crops and value chains
seems high. Farmers integrated into a value
chain have better potential for sustained
growth in their income and improvements
in their quality of life. They are also better
positioned to take advantage of income-
generating options that require up-front
investment. Index insurance can play a key
role in protecting these investments against
weather shocks.

Surveys conducted by WRMS indicated

that farmers trust the program for its capac-
ity to reﬂect actual losses and provide timely
claim settlements. Farmers also seem to have
a good grasp of the likely impact of weather
on yields. In many locations, farmers had
calculated their claims and expected the
forthcoming payouts, which allowed them to
plan their future investments accordingly.

Index insurance, development and disaster management

Disaster relief in Mexico

available to the federal and state governments,
and it eﬀectively transfers the risk from the
government (and the Mexican tax payer) to
the international reinsurance market.

PACC targets poor farmers – most of

those beneﬁting have an income of less than
US$74 per month, with none earning more
than US$222 per month. Table A shows the
levels of support provided through PACC for
such farmers. PACC distributes support for
post-disaster recovery for up to 5 hectares of
land per farmer, to a maximum of US$410 for
farmers growing annual and perennial crops
and US$2275 for small-scale farmers growing
high-value crops.

The index insurance package currently

covers drought and ﬂood (with plans to
include frost soon). Insurance is purchased
exclusively by state governments, or by the
federal government for a speciﬁc state. Federal

Table A. PACC support to farmers for weather-related disasters

Type of crop/farmer

Area covered

Amount of the support

Annual/producers with less
than 20 ha

Up to 5 ha per producer

US$ 82 dollars per ha

Perennial/producers with
less than 5 ha

Fruits, coffee, nopal/producers
with less than 5 ha

US$ 455 dollars per ha

Source: Rules for the Operation of PACC

Mexico’s subsistence farmers are vulnerable
to both excess and lack of rainfall, as 78%
of the country’s arable land is non-irrigated,
depending exclusively on the seasonal rains
from May to November. When disaster
strikes, the rural population is supported in
part by federal assistance funds provided by
the Climatologic Contingency Attention
Programme (Programa de Atención a
Contingencias Climatológicas, PACC). Paid
exclusively from tax revenues, PACC is a
costly and unsustainable attempt to manage
risk after a disaster event.

An index insurance program for cata-

strophic risk was designed by Agroasemex, the
state reinsurance company, with the purpose
of improving PACC by increasing the eﬃ-
ciency, timeliness and distribution of federal
funds to farmers following climate-related
crop failure. The insurance is exclusively

Case studies II

Table B. Average minimum and maximum triggers determined for drought (mm) in Mexico

Year/
Stage

Maize

Bean

Sowing

Flowering

Harvesting

Sowing

Flowering

Harvesting

2003

146

2004

36–58

79–149

27–122

2005

29–66

44–308

14–218

29–58

45–107

24–128

2006

31–66

44–351

14–218

33–58

45–107

25–128

2007

29–66

44–239

14–180

29–54

45–107

25–128

2008

29–66

49–239

26–180

26–58

45–107

24–128

funds are distributed to state governments
according to the level of marginalization of
the insured population: for municipalities
characterized by a high level of marginaliza-
tion, 90% of the insurance premium is covered
by the federal government and only 10% by
the state government; in less marginalized
areas, federal funds pay 70% of the insurance
premium.

On average over six years and across states,

the premium has been 13% of the total sum
insured. The insurance guarantees to pay at
least the amounts promised by PACC, i.e.
at least US$82 per hectare, in the event of
extreme drought or ﬂood. These funds become
available to governments for distribution
to farmers. However, since the insurance is
contracted by the government, farmers do not
participate in the decision to purchase coverage
and, most of the time, are not even aware that
they are insured. Agroasemex manages the
program, and covers itself through reinsur-
ance contracts. The index insurance product

has been registered with Mexico’s National
Insurance Commission, as required by law.

The contract for drought uses a rainfall

index, with triggers based on cumulative rain-
fall for the diﬀerent crop development stages
(sowing, ﬂowering and harvesting). Triggers
are highly speciﬁc to the crop, development
stage and region, and are adjusted each year
(Table B).

Agroasemex has indicated that the level

of basis risk is low and therefore acceptable
to its client, the government. However, since
there have been no studies of the program’s
impact, it is not possible to determine how
the current level of basis risk aﬀects the
smallholders who are ultimately covered by
the program. There were some cases in 2005
where triggers caused payouts when farmers
had not actually experienced crop damage;
while the opposite occurred in 2006, when
some farmers experienced crop losses but
no payments were triggered. In the state of
Guanajato in 2006, only one weather station

Index insurance, development and disaster management

out of 27 in the state triggered inadequate
payouts. In this particular case, Agroasemex
decided not to include this station in the
following years, because it does not report on
a daily basis.

In the event that the index insurance trig-

ger fails to respond to a real loss, the govern-
ment will step in and use contingency funding
to cover the loss.

The program has grown considerably

from 2002 (75,000 hectares and ﬁve weather
stations) to 2008 (1.9 million hectares and
251 weather stations). In 2008, some 800,000
low-income farmers were covered, with a sum
insured of US$132.3 million. In that year, the
purchase of risk transfer instruments repre-
sented 61% of the PACC budget.

Although impressive, scale-up has been

constrained by the limited number of weather
stations producing useable data, and by
limited technical capacity. Only half of the
country’s functioning 1200 weather stations
have good enough real-time or semi-real-time
data to build and monitor eﬀective indices.
A rural producers’ association, Fundación
Produce, is attempting to ﬁll the gap by
building a network of some 764 automated

weather stations. When up and running, these
will provide data to supplement those from
government-run stations. Agroasemex is also
experimenting with data from remote sensing,
and in particular, incorporating the NDVI
into its product.

The lack of historical data is also being

addressed so that the program can continue to
expand. Rainfall and temperature simulations
are used to replace missing data and produce
long time series as a basis for estimating prob-
abilities when building indices. In 2008 these
synthetic series were used with actual (though
limited) data from 75 weather stations to
extend coverage over 250,000 hectares. In 2009
a further 160 government-run stations and 100
Fundación Produce stations will be included,
supported by simulated data series, allowing a
further 2 million hectares to be covered.

Agroasemex is investigating expanding the

program by marketing insurance to individual
farmers, although delivery channels are likely
to be a challenge. As a ﬁrst step, microﬁnance
institutions may serve as an intermediary. So
far, the farmer-level insurance has been a fully
subsidized state-run program, with no involve-
ment of the private sector.

Case studies II

Catastrophe risk insurance in the Caribbean

were to approach the reinsurance market
independently.

Countries are insured against earthquakes

and hurricanes, with rapid payout once the
trigger (a speciﬁed level of shaking or wind
speed respectively) is reached. The aim is to
provide immediate cash to begin recovery
eﬀorts after a major natural disaster. In 2007,
the ﬁrst year of operation, contracts were
designed to cover hurricane or earthquake
events of a magnitude that would be expected
less frequently than once in 20 years, with
the exact level of cover being negotiated with
the member countries individually. Contracts
based on more frequent (less catastrophic)

The Caribbean Catastrophe Risk Insurance
Facility (CCRIF) is unique, being the
world’s ﬁrst multinational index insur-
ance scheme, with 16 member countries.
This spreads the risk (and costs) across the
Caribbean region, making insurance much
more affordable for individual countries (see
ﬁgure). Since natural disaster risks in any
given year are randomly distributed among
the Caribbean islands, the cost of coverage
for the pooled portfolio is less than the
sum of individual coverages. The pooling of
country-speciﬁc risks allows for a reduction
of individual insurance premiums by almost
half, compared with the cost if a government

1-in-200 year PML as a percentage of the aggregate individual country PMLs

Note: The PML is the probable maximum loss for a 1-in-200 year event. It can be interpreted as the risk capital requirements (including reserves and
reinsurance) the Facility should hold if it is to survive such an event. The graph shows how the participation of each additional Caribbean country affects
the level of risk capital needed by the Facility relative to the aggregate of the 200-year PML for each individual country. For example, the relative risk capital
requirement by the CCRIF is reduced by 65% when seven countries participate. It is further reduced by 75% if the CCRIF portfolio includes 17 countries.

Source: CCRIF Background Document, World Bank, 2007

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

Number of participating countries

Index insurance, development and disaster management

disasters would have required higher premi-
ums than they were willing or able to pay.
However, following experiences in the ﬁrst
year, an optional lower deductible (the part
of the loss paid by the insured) was oﬀered in
the second year, taking coverage to an event
that might occur once in 15 years.

The CCRIF is a not-for-proﬁt Cayman

Islands-registered insurance company owned by
a Special Purpose Trust, with the participating
countries as beneﬁciaries of the Trust. Core
funding to set up the CCRIF was provided by
international donors, and on joining each member
country paid a participation fee. These amounts,
together with income from premiums and rein-
surance purchased on the international markets,
provide the core capital for paying claims.

The contract

To develop the indices, catastrophe risk
models were used to estimate probabilities and
associated losses. Using historical hurricane
and earthquake activity records and estimated
losses for these past events, as well as scientiﬁc
inputs, these models project hurricane and
earthquake activity thousands of years into
the future. With these ﬁgures, a country risk
proﬁle is created for each member country,
around which indices are developed. The index
for hurricanes is based on wind speed, while
that for earthquakes uses level of shaking.

Over a season, the indices are tracked

across the region at selected measuring points
(see map). These points are weighted to give
greater value to those representing areas of

Bahamas

1,360

1,020

680

Kilometers

170

340

Bermuda

Haiti

Anguilla

CCRIF 2007 & 2008 covered countries

and measuring points

Dominica

St Lucia

Barbados

Grenada

Jamaica

Belize

Turks & Caicos Islands

St Kitts & Nevis

Antigua &

Barbuda

St Vincent &

the Grenadines

Cayman Islands

Case studies II

greater economic value, so a country’s capital
and major economic centers will have a
greater weight than undeveloped rural areas.
This is so that the index acts as a good proxy
for actual losses to governments.

Measurements are objective and transpar-

ent. For earthquakes, data from the global
seismic data centre of the United States
Geological Survey (USGS) are used to
determine the level of shaking at the measur-
ing points. Wind speed measurements are
calculated using information published by
the US National Hurricane Center (the
WMO’s regional reporting agency for tropical
cyclones), and modeled using simpliﬁed
wind models. When a trigger event occurs a
preliminary calculation is made immediately
after the event, but the ﬁnal calculation is
made 14 days later, to ensure that the best
information is available from the reporting
agencies. The payment for any given event is
made based on a sliding scale relative to the
scale of the loss. The limit of total payments to
a country in a policy year is agreed with each
country individually, although no country can
purchase coverage worth more than US$100
million per hazard.

Experiences

In the ﬁrst year that the CCRIF was opera-
tional, from 1 June 2007 to 31 May 2008,
one earthquake triggered payouts. This
occurred on 29 November 2007 in the eastern
Caribbean, resulting in payouts in Dominica
and St Lucia which were made on 13

December 2007. The earthquake also caused
signiﬁcant damage in Martinique, which is
not a member of the CCRIF.

Six tropical storms reached hurricane

status in the region during this period, but
none triggered payouts. Hurricanes Noel and
Dean caused the most impact in member
countries. Hurricane Noel was at no more
than tropical storm force when it was over
member countries, so wind speeds were not
excessive; however, it was accompanied by
very heavy rains, which caused widespread
ﬂooding and damage in some member coun-
tries. Because the index is based on wind
speed there were no payouts. The CCRIF is
currently investigating adding rainfall to the
product in future years. This would increase
the premium, as premiums are calculated
directly from the amount of risk being
transferred. Obtaining adequate rainfall data,
both historical and current, is a challenge,
and the development of an ‘extreme weather
monitoring network’ is proposed to support
the required modeling and contract develop-
ment.

Hurricane Dean aﬀected the south of

Jamaica. Because of its compact size only
Category 1 winds were felt at the measuring
points; and the measuring points aﬀected
were mostly in rural areas, which have low
weighting. Agriculture was greatly aﬀected,
accounting for more than half the estimated
costs of the hurricane, which totaled US$327
million. However, agricultural losses were
excluded from the country risk proﬁle as they

Index insurance, development and disaster management

do not have a short-term impact on govern-
ment ﬁnances, and cover for such losses was
not factored in when calculating the premium.
Other losses from Dean fell within the
deductible of the policy and so did not qualify
for a payout. In future, the agricultural risk
may be addressed through index insurance
provided directly to farmers.

The lack of payouts following Hurricane

Dean led to unfavorable publicity for the
CCRIF and revealed a lack of understanding
of the contracts and their coverage. The lesson

learned here is that good communications and
public relations are needed to improve aware-
ness of the CCRIF and its purpose.

During the 2008 hurricane season, the

CCRIF paid out nearly US$6.3 million to the
Turks and Caicos Islands following Hurricane
Ike. The payments provided much needed
liquidity to the government to support the
recovery eﬀort.

For additional information see Caribbean

Catastrophe Risk Insurance Facility (2008)
and World Bank (2008).

Flooding in Haiti. The CCRIF may add rainfall to the product in the future, so that damage due to ﬂooding is covered;

Matthew Marek/American Red Cross

Case studies II

Insuring development goals in the Millennium Villages Project

As part of this mission, in 2006 the MVP

surveyed local development professionals in
each village to obtain their opinions on the
predominant risks to project goals. While all
villages mentioned some form of climate risk
as a primary threat to village livelihoods, 10 of
the 12 mentioned drought as the number one
long-term threat.

To manage this climate risk, the MVP

has joined forces with IRI and Swiss Re to
develop an index insurance scheme to insure
the project’s development goals. The aim of

At the Millennium Summit in 2000, the
nations of the world agreed on a comprehen-
sive set of Millennium Development Goals
(MDGs) to reduce extreme poverty and other
forms of deprivation in the developing world. In
2005, the Millennium Villages Project (MVP)
was launched to achieve these goals in 12
‘villages’ across Africa, using an approach that
combines science-based development initiatives
with community participation. The sites were
chosen to be representative of the continent’s
diverse agricultural regions (see map).

Millennium Villages and agroecological zones.

M l

i l

Mill

Vill

i l

Vill

i l

Mil

Mill

Vill

i l

gical zones.

Millennium Villages

agroe

A apted from Dixon et al. 2001. Farming Systems and Povert

r
r y.

y FA

F O

Bonsaaso, Ghana

Ikaram, Nigeria

Pampaida, Nigeria

Tiby,

y Mali

Dertu, Kenya

Mwandama, Malawi

Mbola, Ta

T nzania

Ruhiira, Uganda

Potou, Senegal

No Research Villages:

Maize mixed (1 bimodal) (9 unimodal)

Highland mixed (2)

Highland perennial (8)

Pastoral (11)

Agrosilvopastoral (4)

Cereal-root crops mixed (3 Sudan savanna) (10 Southern Miombo)

Root crops (5 Guinea savanna) (7 M

Miombo)

Tree crops (6)

Coastal artisanal fishing (12)

Irrigated (3b)

Sparse (13)

Paddy rice (14)

Large commercial and small holder (15)

Forest based (16)

Agro-ecological Zones

Gumulira, Malawi

T ya, Mali

Mayange, Rwanda

Sauri, Kenya

Koraro, Ethiopia

es and agroecolog

Index insurance, development and disaster management

the partnership is to mitigate losses during
catastrophic drought years that would oth-
erwise lead to a major redirection of project
resources and put the project in jeopardy.
Payouts are tied to two risk levels – a moder-
ate drought (deﬁned by a 1 in 8 year trigger),
and catastrophic drought (deﬁned by a 1 in 20
year trigger). Indices have been designed for
each of the 12 MVP locations, and in 2007
were transacted for three of these locations
(one each in Kenya, Ethiopia and Mali).

In keeping with the aim to ensure project

sustainability, it was speciﬁed that payouts
from this scheme would be used to sup-
port core development interventions, such
as subsidizing a school feeding program
which provides a market for local farmers
and encourages primary school attendance.
Insurance payouts would also provide
resources to replace agricultural inputs, which
the project subsidizes for the local community.
Thus this product aims to keep people from
falling back into the poverty traps the project
is working to remove.

The result is a unique development inno-

vation. For the ﬁrst time, a major development
project has insured itself against a major
climate risk – drought – which was identiﬁed
by project participants as most likely to aﬀect
the project’s achievement of its objectives.
Equally as innovative, the project is working
with insurance experts at Swiss Re to explore
a combination index based on vegetative
remote sensing and local rain gauges. The
remote sensing component is used to diagnose

large-scale regional drought, while data from
rain gauges identiﬁes more localized drought.

The contract

The index was designed to indicate moderate
to catastrophic drought events, which recur
at intervals of 8 to 20 years. The basis for
the remote sensing portion of the index was
NDVI, with an 8 km pixel resolution. The
NDVI signal was aggregated using a spatial
average of approximately 100 km

× 100 km,

an area roughly 100 times larger than the
village. The index was designed to target the
period in local cropping calendars between
ﬂowering and harvest, which showed the
strongest relationship between local rainfall
and historical yields (taking into account
the lag time of vegetation response to rain).
This large coverage area allowed the index
to capture regional climate trends, which are
more coherent than trends at smaller spatial
scales. Research conducted by IRI showed
that, for many of the MVP locations, when
native vegetation shows signs of stress at this
spatial scale, crop yields are typically greatly
reduced.

To make the index still more robust, local

rain gauge data were added in those locations
which are less spatially coherent (such as
humid forest, mountains and coastal ecosys-
tems). These data were then weighted with
the regional scale vegetation data to provide
a greater sense of local climate variability.
Regional scale vegetation only was consid-
ered suﬃcient to underpin contracts for the

Case studies II

semi-arid regions of Senegal, Mali, Ethiopia,
northern Nigeria and northeastern Kenya,
while an index of vegetation combined with
rainfall was recommended for the wetter (and
less variable) climates of Uganda, Rwanda,
Tanzania, Malawi and western Kenya. No
drought index was recommended for villages
in Ghana and southern Nigeria, as village
surveys there showed a greater risk of ﬂooding
than drought.

The premiums were paid by MVP villages,

through donor support.

Experiences

While none of the three transacted contracts
resulted in a payout in the ﬁrst year, the
process of thinking through CRM strategies
at the local, national and continental scale in
the unique context of a major development
project did provide many opportunities to
reﬁne working concepts of weather insurance
and its interactions with development. In the
villages that had contracts, local project man-
agers were able to operate in a less risk-averse
manner and were better able to take advan-
tage of the good rains that were experienced,

knowing that their input costs for fertilizer
and improved seeds would be covered if
a drought did occur. The experience also
stimulated discussions of risk pooling mecha-
nisms. In theory, for projects like MVP, self-
insurance guided by CRM concepts could be
an effective strategy for reducing the impact
of moderate climate shocks. The project as
a whole could then reinsure for high-impact
climate shocks. Because the MVP is in
transition from the initial donor provision of
resources towards self-ﬁnancing of activities,
insurance strategies are evolving to reﬂect the
changing structure of the project. This is part
of a larger CRM assessment of the project.
Following this transition, it is likely that
the future of insurance in the MVP will be
more targeted towards encouraging farmers’
adoption of inputs.

The challenge of designing an index that

can represent regional scale catastrophic
drought was overcome by combining the
remote sensing of vegetation with local
rain gauge data, with weighting to capture
drought impacts varying according to
climatic zone.

Index insurance, development and disaster management

Vietnam: Flood insurance in the Mekong Delta

An index-based ﬂood insurance product
has been offered to the Vietnam Bank
of Agricultural and Rural Development
(VBARD). The product is designed to pay
for consequential losses that are suffered by
VBARD when ﬂooding creates problems
for farmers in repaying loans. The contract
is being offered by a Vietnamese insurance
company, has support from a global reinsurer,
and has been approved by the Vietnam
regulatory authority. Flood insurance products
are a challenge to develop and as such this

project represents a signiﬁcant contribution
to the application of index-based risk transfer
approaches.

Index insurance market development in

Vietnam is being led by GlobalAgRisk, Inc.,
following a request by the Government of
Vietnam for investigation of index-based
solutions to agricultural risks. The project,
which has received ﬁnancial support from
the Asian Development Bank and the Ford
Foundation, seeks to expand rural ﬁnancial
markets by developing means to transfer

The offered product will protect the Vietnam Bank of Agricultural and Rural Development against loan defaults by rice

farmers, when early ﬂooding damages crops; Ariel Javellana/CPS, International Rice Research Institute

Case studies II

catastrophic production risk out of agriculture.
It comes at an important time of transition
for Vietnam, which joined the World Trade
Organization (WTO) in January 2007 – a
move that is creating pressures for the country
to adopt international standards on ﬁnancial
market behavior and regulation, including
recapitalization of banks by the state.

During the early phases of risk assessment,

GlobalAgRisk and local stakeholders identi-
ﬁed ﬂooding and ﬂood impact on rice produc-
tion as a primary concern in both the Mekong
and Red River deltas. Feasibility and product
design investigations have focused on Dong
Thap province of the Mekong Delta, where
the ﬂood regime is inﬂuenced predominantly
by upstream water ﬂow and where delta-wide
ﬂood modeling has previously conﬁrmed that
downstream and overland ﬂooding in the
province is highly correlated with water levels
measured at the Tan Chau water level gauging
station on the upper Mekong River at the
Vietnam–Cambodia border.

Flood occurs in the delta as part of a

natural annual process that is critical for
maintaining crop productivity. Rice farmers
have adapted their cropping strategies to
accommodate and beneﬁt from the annual
ﬂood. However, the second season (summer–
autumn) rice harvest is put at risk when the
ﬂood cycle begins earlier than normal. Early
ﬂooding leads to yield and quality losses as
well as additional harvest and post-harvest
handling costs. Water levels in excess of 2.5
meters at Tan Chau are known to begin

causing downstream ﬂooding and to interfere
with the harvest when occurring in late June.
Early ﬂooding remains a risk despite infra-
structure improvements, as conﬁrmed through
ﬂood risk modeling and mapping which
incorporates these improvements.

The ﬁnancial strain on farmers from early

ﬂooding often translates into loan repayment
diﬃculties. As a result, VBARD is exposed
to signiﬁcant direct and opportunity costs, or
‘business interruption loss’.

The prototype index insurance contract,

designed to oﬀset these early ﬂood-induced
business interruption losses, is underwritten
against recorded water levels. This is similar
to weather index insurance using weather
measurements at meteorological stations, but
using river gauge data as a proxy for ﬂood
damage. The ﬂood event index is calculated
as the maximum 3-day moving average of
daily water levels at the Tan Chau station
during the period of cover, 20 June to 15 July.
Indemnities are paid for each centimeter of
water once the river level index reaches the
2.8 meter threshold, with maximum payout
occurring when the index reaches 3.5 meters.
Early ﬂooding that exceeds 2.8 meters is
empirically estimated to occur approximately
1 in 7 years, a recurrence rate that represents a
commercially insurable risk.

The product was fully priced and loaded

by a domestic insurance company, Bao Minh
Insurance Corporation, which then sought
and obtained reinsurance against its exposure
from the international company Paris Re.

Index insurance, development and disaster management

No subsidies were involved in the price for
underwriting. A contract for a sum insured
of US$1 million sum was ﬁrst oﬀered to
VBARD in 2008. This represents a maximum
insurable loss that is approximately 14% of
the maximum possible loss. The product was
oﬀered toward the end of the sales closing
period, allowing too little time for the serious
consideration needed before purchasing this
new insurance. It is being oﬀered again in
2009, and VBARD is currently assessing the
option to purchase.

The accomplishments of this project – (i)

designing a unique index-based insurance
against ﬂood in a developing economy; (ii)
accessing global risk capital through reinsur-
ance against the early ﬂood risk; and (iii)
achieving regulatory approval of the product –
represent considerable advances that will ben-
eﬁt index insurance development worldwide.
This project also illustrates the limitations of
developing a product for a single client – in
this instance VBARD, which is still partially
publicly supported and can be recapitalized by
the government in the case of portfolio and
business interruption losses. Thus, the main

motivation for participation hinges on a desire
to demonstrate leadership and to learn how
to assess and manage risk exposure using new
ﬁnancial products, with an eye to the future.

Several additional investments are needed

to continue the development of this insur-
ance product in Vietnam. First, considerable
education and outreach is needed to create
greater stakeholder recognition of residual
ﬂood risk, even in the presence of improved
infrastructure. Second, a next step in the
regulatory process is to investigate whether
these forms of index insurance product
can serve as a type of credit warranty that
reduces the level of required reserves. Finally,
the meso-level product, while a stand-alone
policy, is a ﬁrst step toward ﬁnding workable
solutions at the micro-level, since even if the
bank is protected, very careful thinking is still
needed on how to deal with bad debt. While
rescheduling debt rather than debt forgiveness
improves the commercial viability of VBARD,
rescheduled debt does little to help ﬁnan-
cially stressed producers – the primary target
market of VBARD – recover from a natural
disaster.

Scaling up index insurance: Operational issues

demand for index insurance products, local
ownership of products, legal and regulatory
issues, and the complex question of subsidies.

Demand for the product

Demand for index insurance is the most fun-
damental issue confronting scaling up efforts.
For a product to be successful it must be
worthwhile, address an important risk to the
clients and basis risk must not be too high.

Index insurance is not a viable proposition

in all situations. Also, it is essential that client

This section examines critical operational
issues that could frustrate the scaling up of
index insurance if not addressed. Our discus-
sion is drawn from the pilot projects, whose
community of implementers has acquired
considerable experience of these issues. Here,
we draw attention to the main issues, while
more detail can be obtained directly from
the implementers (World Bank, 2005; Hess,
2007; UNDESA, 2007). We have structured
the discussion according to broad headings
that capture the main recurrent themes:

Groundnut farming in Nicaragua (see case study on page 72); Eric Niu

Index insurance, development and disaster management

expectations are reﬂected in product design.
It is therefore important to perform basic
risk assessment exercises prior to designing
a product, and to undertake demand assess-
ments of product prototypes to determine
if they address demands and expectations,
and how they might be improved to be
more eﬀective (UNDESA, 2007). WFP, for
example, carried out a large-scale demand
assessment including a survey of target client
risk perceptions and needs. Clients must be
active participants in the development and
implementation process.

To ensure that demand is sustained,

products must be adapted and improved as
implementation proceeds. Several of the case
studies modiﬁed products after dry runs or
in response to initial performance gaps. In
the Caribbean CCRIF case, for example,
changes to the index are being considered to
better capture ﬂooding risk associated with
hurricanes (see ‘Catastrophe risk insurance
in the Caribbean’ on page 52). Validation
and robustness investigations begun during
contract design must be continued during
implementation, with improvements made
in response to the issues uncovered. This is
particularly important in places like India,
where there are competing index products
being oﬀered by diﬀerent companies, which
have payouts under diﬀerent conditions.

To be worth buying, the insurance must

represent a value proposition to the client,
that is, the product must be seen to yield a
beneﬁt that exceeds its cost. When subsidies

are being considered for index insurance, it
needs to be established whether the beneﬁts
exceed the unsubsidized cost. If they do not, it
is likely that the product does not address an
important need and that the subsidy money
would be better used in other ways.

When evaluating demand it is important

to consider other interventions that may
be available, for example subsidized crop
insurance, bank credit guarantees or relief
programs. These will often reduce the demand
for index insurance. The case studies in
Nicaragua, Malawi, Ukraine and India all
faced challenges of this kind (see ‘Central
America – a diﬀerent approach for launching
index insurance’ on page 72 and ‘Barriers to
implementation in the Ukraine’ on page 74).

The insurance must be aﬀordable. Index

insurance should not be used to insure against
high frequency risks with big losses, since the
pure risk cost is likely to be too high. The cost
must also be competitive with the alternatives
for managing the risk.

Weather index insurance makes sense only

where weather is one of the major risks con-
fronting households, banks or relief agencies.
In many cases market risks may be greater for
farmers and bankers, while the risk of war or
conﬂict may be greater for relief agencies.

Local ownership and capacity

Insurance markets

When projects seek to address the needs of
the poor, it is essential that local partners
have strong ownership from the outset.

Scaling up index insurance: Operational issues

Implementations that are not championed
by local institutions cannot gain the traction
necessary to have meaningful impacts, scale
or be sustained. At the scaling up stage, it is
vital that local markets are strong enough to
support widespread implementation. However
weather insurance markets are virtually non-
existent throughout the developing world,
and other markets are weak as well. Unless
addressed, this limits the scope for local
partners to participate.

India forms a marked exception, with local

markets well developed in urban areas and
even in some more remote rural areas (see
‘The private sector builds a market in India’
on page 76). Indeed, much of the dramatic
scaling up of index insurance in India is due
to the dynamism of local markets. Highly
active brokers have strong ownership of the
products they oﬀer, competing keenly with
each other to reach the farming community
with aﬀordable and attractive products. There
is a high level of local expertise due to the

government’s historical mandate that micro-
insurance be made available in rural areas.
Thus, the starting point in developing weather
insurance markets is often to lay the knowl-
edge foundations.

Skills

It can be challenging to ﬁnd good contract
designers at the international level – such
people must combine mathematical skills with
a knowledge of climate science and experi-
ence with agro-meteorological models. At
the local level, this challenge is even greater.
In addition to specialized index design skills,
local partners need insurance selling skills.
These need to be especially highly developed
in areas where weather insurance products are
new and have to be explained to the farming
community. All this means that public invest-
ment is vital at project start-up, to ensure
the necessary knowledge and skills are ﬁrmly
established by the time scale-up gets under
way. This has been forthcoming: the World
Bank Climate Risk Management Group and
World Bank Institute have developed training
courses to build capacity in the wide range of
tasks necessary in index insurance; they have
also created online contract development tools
in partnership with IRI and others. Despite
these efforts, more needs to be done if scaling
is to be developed and sustained.

Delivery channels

Closely related to the issue of local ownership
and capacity is that of delivery – the challenge

In India, BASIX take the message to farmers in mobile units;

Sridhar Reddy/BASIX

Index insurance, development and disaster management

of making sure that the farming community
knows about index insurance products and has
access to them. This challenge is the same for
all types of agricultural insurance and other
ﬁnancial products targeting rural farmers.
Many projects have relied on intermediate
institutions to overcome this challenge. When
these institutions are involved, it is important
their staff have a thorough understanding
of the insurance product, are able to market
it with enthusiasm, and are competent at
handling the complex ﬁnancial transactions it
may involve.

In Ethiopia the development of coop-

eratives is being strongly encouraged by the
government as a means to facilitate service
delivery for agricultural inputs, extension
advice, processing and marketing, and to
improve the ability of farmers to market their
products. This avenue was used to reach farm-
ers with index insurance products (see ‘Laying
the foundations for farm-level insurance in
Ethiopia’ on page 79). In Brazil, insurance
providers are taking advantage of a network
of distribution points under a government
seed program (see ‘Supporting farmers – and
a government seed program – in Brazil’ on
page 82).

A new delivery concept for index insur-

ance in disaster management is the early
recovery assistance voucher. These vouchers
are intended to achieve more timely and better
targeted recovery support for those aﬀected
by natural disasters such as droughts, ﬂoods
or hurricanes. Still in the proposal stage, the

concept is that vouchers would be distributed
to people in vulnerable areas ahead of disasters
and would be redeemable after them. A
voucher would give the right to, say, US$200
worth of food or cash at local shops, trading
stations, post oﬃces and other public outlets.
When eligible people obtain vouchers they
would also be able to register their residence
and leave their cell phone number to receive
localized early warnings as well as payment
notiﬁcations (Hess and Portegies, 2009).

Risk takers

Risk takers – local insurance companies will-
ing to underwrite index insurance contracts
or intermediate the risk to the international
market – are essential. Given that index
insurance is a nascent industry, insurance
companies must often build expertise at the
same time as they implement pilot projects,
a factor that increases the level of risk they
face in the short term. Some of the pilots have
taken ‘dry runs’, which allow for the company
to gain experience and trust and for contracts
to be adjusted before the product is launched
‘for real’ on the open market (see ‘Stakeholder
communication is key in Thailand’ on page
85). Government policies, such as those in
Ethiopia and India, which stipulate more
emphasis on microinsurance products or sup-
port a range of agricultural insurance options,
may encourage risk takers to get involved.
However, in some situations they may have
the opposite effect, either by crowding out
index products through other interventions or

Scaling up index insurance: Operational issues

by creating administrative burdens that risk
takers may not be comfortable carrying.

Data

Constraints concerning the measurement and
validation of data, which are key design issues,
must also be actively addressed during the
operational phase. The Agricultural Insurance
Company of India (AIC), for example, conduct-
ed a study of paired automated weather stations
operating on the same premises across a range
of zones, to test the reliability of data. It found
signiﬁcant short-term differences between the
stations. If these differences cause problems for
the insurance products, either the products need
to be improved or the data measurements must
become more accurate (see ‘Scaling up in India:
The public sector’ on page 87).

Measurements must be reliable, timely

and resistant to tampering, with fallback
options formally speciﬁed when there are data
problems. Capacity building may be needed
to ensure that these criteria are met. And
local meteorological stations and the national
meteorological service must be included in the
list of national partners needing to take owner-
ship of index insurance projects. This can be
diﬃcult to achieve where already overstretched
national services do not see the project as a
priority. Data issues are discussed in detail in
the previous section.

Trust

The importance of trust is a theme running
through all the pilots. The glue that binds

partners together, trust is a vital element in
the willingness and ability of local providers
to take projects over and continue them after
the pilot phase. Because the relationships
between players often become strained
during the intense interactions that can take
place over insurance, these relationships must
be underpinned by substantial levels of trust,
which tends to be stronger if established
through collaboration that predates the
insurance project. In the India contract farm-
ing case study, surveys indicated that farmers
trust the program because of its timely
settlement of claims (see ‘Insurance for
contract farming in India’ on page 47). This
is consistent with the experience of BASIX, a
microﬁnance institution that ﬁrst introduced
rainfall index-based insurance to farmers
in India together with the International
Finance Corporation (IFC) of the World
Bank (see ‘The private sector builds a market
in India’ on page 76). Oxfam’s long-standing
presence in Ethiopia, combined with high
levels of respect and trust for the local
partner NGO, the Relief Society of Tigray
(REST), made farmers willing to explore
weather index insurance (see ‘A farmer-
centric approach in Ethiopia’ on page 44).
Feedback from Thai farmers suggested that
the primary motivation for purchasing insur-
ance was trust in the Bank for Agriculture
and Agricultural Cooperatives (BAAC), an
institution that has long-term relationships
with the farmers (see ‘Stakeholder commu-
nication is key in Thailand’ on page 85). The

Index insurance, development and disaster management

lesson is clear: identifying partner organiza-
tions that farmers already trust is critical to
successful scale-up.

Legal and regulatory issues

Often legal and regulatory frameworks
must be developed or improved so that they
address the new issues raised by the introduc-
tion of index insurance. The International
Association of Insurance Supervisors (IAIS)
has yet to produce guidance on how insurance
laws, regulations and practices may need to be
adapted to ensure that index-based insurance
is regulated and supervised in accordance with
international standards. However, a number
of pilot projects have shown that index
insurance can ﬁt within existing national
legal and regulatory frameworks while at the
same time adhering to the basic principles

underlying international standards. Regulators
are generally supportive of efforts to develop
index insurance once properly informed about
its potential social beneﬁts. They are also
much more comfortable when they know that
other countries have successfully developed
contracts for this new class of insurance.

In general, regulators have two primary

responsibilities. First, they must protect the
consumer from any form of misconduct that
can emerge when this new form of insur-
ance is developed and sold. Second, they
must protect the insurance provider from the
ﬁnancial exposure that can follow when offer-
ing insurance against events that have highly
correlated losses, requiring many payouts in
the same year (Skees and Collier, 2008). By
doing the latter, the regulator is also ensuring
that the consumer will be paid when he or

Building trust is critical; Janot-Reine Mendler de Suarez/GEF-IW:LEARN

Scaling up index insurance: Operational issues

she does suffer losses. A central mechanism
for this protection, found in most of the case
studies, is the purchase of reinsurance from
international markets. The Mongolia case
study shows an alternative way of provid-
ing protection, using donor support (see
‘Livestock insurance in Mongolia’ on page 90)

One signiﬁcant regulatory issue is whether

index products should be classiﬁed as deriva-
tives or insurance. Insurance contracts are
intended to cover losses, whereas deriva-
tives are purely ﬁnance market contracts.
Derivatives are ﬁnancial instruments whose
values are derived from the value of some-
thing else (known as the underlying value).
The underlying value can be an asset, an index
(e.g. interest rates or exchange rates), weather
conditions or other items. The option taken
depends on the application.

For programs aiming at development,

insurance is the preferred classiﬁcation as it is
simpler to regulate. This is because the regula-
tory framework for insurance is well suited
to protect the interests of a large number of
smaller clients, with regulations focused not
only on honoring contracts, but also guaran-
teeing protection against losses. It is also a
widely accepted ﬁnancial instrument, often
with existing delivery chains that currently
reach intended clients.

Derivatives are more common as negotiat-

ed deals between two large entities, each with
a substantial capacity for analysis. Because
of this, disaster relief contracts are typically
transacted as derivatives.

The laws of most countries require that,

to be considered insurance, a risk transfer
product must have certain key characteristics.
The two characteristics that oﬀer the greatest
challenge for an index-based product are:
(i) an insurance contract must indemnify or
compensate the insured for loss sustained
due to the occurrence of the insured risk;
and (ii) the insured must have an insurable
interest in the subject insured. If the index is
a suﬃciently good proxy for the loss, there is
a clear link between loss and payout and the
ﬁrst condition can be satisﬁed. Although there
may be some basis risk, this is also the case
for traditional insurance products, even where
the loss is assessed by a loss adjuster (Barnett
et al., 2005). In some cases, framing the
index insurance product as a form of business
interruption insurance also eases the regula-
tor’s concerns about basis risk, by allowing
the insurance to target risks it can cover more
transparently. The second condition is less
of a challenge when products are developed
with exposed users in mind as it is relatively
easy to make the case that the insured has an
insurable interest. In some cases, limits may
be needed on the sum insured, so as to ease
the regulator’s concerns that users might take
on larger ﬁnancial commitments due to the
availability of insurance and thereby increase
their risk exposure rather than reduce it. The
second condition can make it challenging to
oﬀer insurance to laborers or merchants who
do not own cropland but who are impacted by
the crop losses of others.

Index insurance, development and disaster management

Regulators should be involved from

the beginning of the product development
process, as speciﬁc aspects of the contract
design may determine whether or not it meets
regulatory conditions. Likewise, if contingent
capital is required, potential reinsurers should
be involved in contract design to make sure
that the risk can be transferred into reinsur-
ance markets.

The challenge of subsidies

Issues surrounding subsidies are very different
for the two major kinds of application –
disaster relief and development. Since disaster
relief programs are themselves funded by
subsidies, the insurance is simply a ﬁnancing
mechanism to make more effective use of
these subsidies. Responsibility for addressing
the subsidy-related problems of distortions,
delivery issues and perverse incentives falls
to the relief program as opposed to the index
insurance provider.

When the objective is to address pov-

erty and development, the very poor may
be excluded because they cannot aﬀord
risk-based premiums. Some therefore argue
that premiums should be subsidized, so that a
higher proportion of this group will be reached.
However, when insurance is part of a package
that promises signiﬁcant income gains – for
example, a package that includes credit and
inputs – subsidies may be less justiﬁed. When
an inability to pay is not a problem of insuf-
ﬁcient wealth or productivity but instead a cash
constraint at the beginning of the season, loans

that cover the premium as well as the inputs
may be the answer, not subsidies.

It is important that the insurance leads to

societal beneﬁts that exceed its opportunity
costs; that is, the beneﬁts that would have
accrued had the subsidy been used diﬀerently.
Once participating households have achieved
higher levels of productivity, the subsidy
should be withdrawn. Proponents argue that
premium subsidies can thus be used to chan-
nel social beneﬁts to the poor in a structured
and controlled way.

Another argument for subsidies is that

they ‘prime the pump’ for insurance markets
by oﬀsetting high start-up costs until the
market expands, economies of scale are real-
ized, and prices decline. In this approach, the
government or donor guarantees the deﬁcit in
the initial years, thereby instilling conﬁdence
in the insurer and allowing time for the
insurance-buying habit to become established
in the farming community. The premiums can
gradually be brought into line with ‘commer-
cial reality’, turning the deﬁcit into a proﬁt. It
would be foolish to experiment with a product
that is not priced to the risk, but it does make
sense to exclude the development costs and
to predicate the economies of scale and the
beneﬁts of reinsurance that will apply only
once a scheme is fully scaled up.

It is also argued that premium subsidies

can stimulate adoption by encouraging
farmers to use insurance and learn about its
beneﬁts (World Bank, 2007). However, care
should be taken to ensure that subsidies are

Scaling up index insurance: Operational issues

not used to encourage the adoption of ineﬀec-
tive products.

There is a signiﬁcant school of thought

that directly subsidizing premiums distorts
the insurance process and makes it counter-
productive, encouraging people to engage
in overly risky behavior (Skees et al., 1999;
Siamwalla and Valdes, 1986). Opponents of
direct premium subsidies note that farmers
may take out insurance because it is cheap
rather than because it targets a risk they
are facing. Subsidies may directly challenge
the development objectives of index insur-
ance. Because the common mechanism for
direct premium subsidies is to make them a
percentage of the premium, subsidies tend to
be captured by the imprudent or those who
take out large amounts of insurance simply to
obtain the subsidy, eroding the poverty objec-
tives of index insurance. Also, it may become
more diﬃcult to obtain client feedback to
determine if a product is useful if the client
is merely seeking a subsidy. Finally, subsidies
may encourage clients to over-insure, which
can lead to insurance that increases the level
of income variability, making the insurance
payments the source of variability instead of
the crop loss.

As an alternative to direct premium sub-

sidies, governments and/or donors may agree
to pay for the most extreme risk layers, cover-
ing catastrophic losses, while other layers of
risk are covered by the commercial market
(World Bank, 2005; see also ‘Livestock insur-
ance in Mongolia’ on page 90). For example,

for an extreme rainfall insurance contract, the
commercial sector may set an upper limit,
with the government and/or donors covering
the risk above that limit. Using this approach,
a commercial layer of risk can be fully priced
to allow the insurance market to expand.
Should the cost of covering the subsidized
catastrophic layer of risk become too great,
a commercial market would stand a much
greater chance of remaining in place when
governments or donors decide to abandon
the subsidy. The balance between direct relief,
insurance and hybrid approaches is complex.
For example, there are arguments that pro-
viding the poor with post-disaster assistance
directly may be more distorting than subsi-
dized insurance (Bayer and Mechler, 2007).

Governments need to consider carefully

whether subsidies for insurance are the most
cost-eﬀective option for achieving the desired
social objective, compared with such alterna-
tives as food aid, better extension services and
cash transfers (World Bank, 2007). Whenever
possible, it is important to use a subsidy to
remove the cause of a high premium instead
of subsidizing the premium itself.

Instead of direct subsidies, governments

and the donors could also invest in making
the provision of insurance more eﬃcient
and eﬀective, by, for example, creating new
weather stations, setting up weather index
insurance standards, providing technical
assistance to the insurance sector, extend-
ing delivery mechanisms, benchmarking
and evaluating impact, building capacity

Index insurance, development and disaster management

and raising awareness of insurance. There is
a public good problem in establishing the
initial products and infrastructure associated
with insurance. It is expensive to create a new
product and the infrastructure to support the
product. And it is diﬃcult for a private ﬁrm to
recoup these expenses, since competitors may
use both the product and the infrastructure as
a basis for developing and oﬀering their own
products without having to pay the start-up

costs. Because of this free rider problem, there
is a role for subsidies to cover many of the
costs associated with initial product develop-
ment and the introduction of pilots.

If the premium itself must be subsidized,

the subsidy should be used primarily to
reduce excess premiums due to data uncer-
tainty. It should not be so large that it reduces
the premium below the actuarially fair cost of
risk (World Bank, 2005).

Damage following Hurricane Ike; Marko Kokic/ International Federation of Red Cross and Red Crescent Societies

Case studies III

Central America – a different approach for launching index

insurance

from all countries in Central America plus
Mexico. Building on this enthusiasm, the
World Bank formed a partnership with the
Inter-American Development Bank and
the Central American Bank for Economic
Integration and began ﬁnancing activities to
strengthen the agricultural insurance market.
Among these activities were training, work
on regulatory issues, eﬀorts to improve access
to reliable weather data, and the launching of
pilot index insurance projects in Nicaragua
and Honduras. The Nicaraguan pilot began
selling contracts in 2007, while the Honduras
pilot is still at the development stage.

In Nicaragua, the project team worked

from 2005 to 2007, laying the foundations
for the pilots. The team consulted widely,
began building stakeholder interactions, and
started capacity building among the primary
partners, particularly the public Nicaraguan
insurer, the Instituto Nicaragüense de
Seguros (INSER), which agreed to be the
implementing partner. At the same time,
CRMG and IRI worked on technical aspects
of the contracts. Nicaragua has good and
accessible weather data, which facilitated the
progress of the pilots. The national meteoro-
logical service was supportive, providing data
for the indices.

Different strategies have been pursued to
explore the best approaches for launching
development-oriented index. These strate-
gies have included directly approaching
smallholder farmers, or approaching con-
tract farming corporations that work with
smallholder farmers. In Central America a
third strategy has been followed. Instead of
beginning with smallholder farmers, insurers
worked ﬁrst with medium- and large-scale
farmers to quickly build a commercial
product that can later be extended to a wider
clientele, including small-scale farmers. In
addition, this project was a collaborative
effort involving several countries in the
region, encouraging the spread of ideas and
the pooling of efforts. These efforts were
combined with a great deal of capacity
building, with strong leadership from local
insurance companies.

Projects in India and Mexico aroused

interest in index insurance in Central
America, and the Latin American Federation
of Insurers asked the Commodity Risk
Management Group (CRMG) of the
World Bank to explore opportunities for
supporting insurance market development
in the region. A regional workshop held in
Guatemala in May 2005 attracted participants

Index insurance, development and disaster management

Initial design activities were not con-

sciously targeted to large farmers, but these
farmers demanded the products and provid-
ed enough data to tailor the products to their
needs. Because of the scale of the farmers,
the information available was substantially
diﬀerent to that in a pilot focusing solely on
smallholders. Farmers had multiple plots at
diﬀerent distances from the weather sta-
tion. They measured precipitation at each of
their plots, and had computerized records of
rainfall and historical yields, allowing them
to directly validate and provide feedback
on index contract options. Their data and
feedback led to the possibility of addressing
risks that are more challenging to model
than drought, such as excess rainfall. In addi-
tion to the feedback oﬀered by these farmers,
the farm size helped jump-start the index
insurance by providing a larger, more viable
base for the project.

In 2007 INSER sold two contracts to

groundnut farmers, protecting an area of 181
hectares, with average premiums of 4.9% of
the insured value. The contracts were oﬀered
for any combination of three weather risks:
(i) excess rainfall at sowing; (ii) drought
during growth; and (iii) excess rainfall during
harvest. The contracts were designed with
a ﬂexible start date to accommodate areas
with diﬀerent sowing periods. The contracts

triggered payouts and INSER paid indem-
nity losses equal to approximately 32% of the
total premiums.

In 2008 INSER sold 12 contracts for

groundnuts and four contracts for rainfed rice,
protecting a total area of 1,774 hectares, with
premiums averaging 5.5% of the total insured
sum of US$1.7 million. Contracts for both
crops covered any combination of the three
risks above, plus rainfall excess throughout
the season, with a ﬂexible start date. Due to
unusually heavy and prolonged rains during
the country’s 2008 growing season, INSER
received claims equivalent to 86% of the total
collected premiums for groundnuts.

Though insurers’ proﬁts for both years

have been small, INSER believes that the
contracts have provided an excellent dem-
onstration and primed the market for the
next season, when it plans to launch a more
aggressive marketing campaign for the prod-
ucts. The expectation for the 2009 season is
that approximately 400 farmers will partici-
pate and 16,000 hectares will be covered, to a
total value of up to US$10 million.

As anticipated, the project has attracted

interest from the Ministry of Agriculture
and the Fondo de Credito Rural (FCR). In
the coming years, FCR plans to work with
INSER to bring index insurance to some of
its small-scale farmer clients.

Case studies III

Barriers to implementation in the Ukraine

Partners in the project team included the

World Bank’s CRMG, Credo-Classic and
the IFC’s Agribusiness Development Project.
Starting in 2003, the team worked to develop
an index insurance contract. It carried out
extensive consultations with stakeholders,
including farmers, local oﬃcials and scientists.
Approximately 400 producers were questioned
during stakeholder events.

The pilot site was chosen around two

weather stations, at Kherson and Behtery.
Weather data were provided by the Ukrainian
Hydrometeorological Center (UHC), which
maintains 187 weather stations across Ukraine.
Historical weather data were available for the
past 30 years. UHC also provided a vegeta-
tion and risk-sensitivity report for grain crops
(wheat, rye and barley). Computer simulation
models were carried out, and consultations
with farmers revealed that they were most
concerned about the period from early May
until mid-June, when high temperatures and
lack of rain could damage grain crops.

Based on this work, the team developed an

index insurance contract to protect grain crops
from drought. The contract covered the period
15 April to 15 June and captured low rainfall
as a cumulative amount over the period (less
than 70% of the normal 80 mm). Another
contract was developed to capture high
temperatures (+30°C or excessive accumulated
temperatures).

An index insurance pilot project in Ukraine
in 2005 built on some good preparatory
work, but sold just two contracts and has not
been continued. Reﬂecting on the experi-
ence, several key obstacles are evident. The
insurance industry in the country was not
‘ready’ for the new product and was therefore
not sufﬁciently engaged. The index product
was competing with a subsidized crop insur-
ance program. Regulatory hurdles limited
the product to direct coverage of individual
farmers only, eliminating potentially success-
ful options involving the targeting of other
groups, such as agribusiness and ﬁnance
institutions. There may, however, be a role
for index insurance in the future in Ukraine,
particularly as farmers move towards grow-
ing more high-value crops.

Planning and preparation

A role for index insurance was proposed
because of the weather risks to agriculture in
the country. Multi-peril crop insurance was
available, but the loss adjustment procedures
were unclear and payouts were often delayed
for up to 6 months.

The pilot project team consulted six insur-

ance companies, but only one – Credo-Classic
– agreed to join the project; the others cited
lack of funds for the development of new
products and a desire to focus on the govern-
ment’s subsidized crop insurance program.

Index insurance, development and disaster management

Problems

Marketing was in the hands of Credo-Classic,
but unfortunately the company did not
register the index insurance product until
the end of March 2005, which did not allow
sufﬁcient time for effective marketing. The
insurer managed to sell only two cumulative
rainfall contracts for Behtery and none for
Kherson. However, it should be noted that
the company sold only six of the multi-peril
contracts in the same season. Also, regula-
tions in Ukraine mean that only primary
producers of agricultural commodities can
purchase index insurance; the product could
not, therefore, be marketed to input suppliers,
processors or loan providers to insure their
agricultural portfolio.