Analysis of Future Potential of Index Insurance in West Africa Using CMIP5 GCM Data

Extreme climate events have significant adverse impacts on many sectors of society around the world, including the agriculture sector in the developing world. One of the ideas that has been explored to help address these risks and vulnerabilities is the concept of index insurance. Index insurance differs from traditional insurance in that payouts are made on the basis of a geophysical index directly measuring an environmental condition, rather than through a loss claim and verification process. As a consequence of this simpler framework that holds the potential for a more rapid payout, index insurance is perceived as having practical potential in a developing world context where traditional property or agriculture insurance markets tend to be limited or nonexistent. Furthermore, as drought and some instances of river flooding have relatively high levels of spatial coherence (at least over relatively flat terrain), there is the potential that such projects may serve a large area. This work will focus on the exploration of theoretical index insurance contracts in the West African Sahel nations of Niger, Burkina Faso and Mali.

Climate related index insurance in Africa has a number of key challenges. Pilot projects have encountered challenges of establishing trust, communicating the risks and scaling up the number of potential users. In addition, there are a number of technical challenges as well. Among these technical challenges, the changing nature of the climate system itself will induce a change to the underlying basis risk of the insurance contracts by altering the frequency of extreme events.

The broader research behind this study explores the sensitivity and viability of such a model index insurance scheme in the West African Sahel in the context of a changing climate in the 21st century. The theoretical index insurance contracts have two thematic components: a component targeted towards irrigated rice farmers along the River Niger and a component targeted towards subsistence rain-fed farmers with an emphasis on the millet crop. The component targeted towards the irrigated farmer population focuses on flooding risk using streamflow data from the Niger Basin Authority and the component targeted towards subsistence farmers focuses on drought risk using satellite and station rainfall data. Earlier work (AMS presentation, 2014) assessed the viability of several observational datasets (including the GHCN, NOAA PRECL and the GPCP datasets) towards these ends.

This study builds on that prior work by integrating Global Climate Model (GCM) data from a number of CMIP5 models into projections of threshold crossing extreme event frequency and index insurance price and volatility. Specifically, this study uses monthly precipitation and evaporation data from Geophysical Fluid Dynamics Lab (GFDL) CM3, Goddard Institute for Space Studies (GISS) E2H, National Center for Atmospheric Research (NCAR) CCSM4, Centre Nationale de Recherche Meteorologique (CNRM) CM5, and Commonwealth Science and Industry Research Organization (CSIRO) MK 3.6 under the emissions scenarios RCP 4.5 and RCP 8.5.

These five models and two RCPs by no means constitute an exhaustive analysis, but do serve as a representative basis for analysis for understanding how modeled trends in regional rainfall and evaporation might impact the sustainability and pricing of such index insurance contracts. As with earlier analysis in the climate literature on the West African Sahel, there remains uncertainty in the CMIP5 data regarding the projected sign of the projected 21st century precipitation change. In the regional climate literature, there is an acknowledged general positive correlation between Sahel rainfall and the Atlantic Multidecadal Oscillation. Other recent literature (Giannini et al., 2013) suggests that as long as the tropical North Atlantic warms more than the rest of the global tropics, the West African monsoon is likely to be strengthened. It is widely acknowledged that the strong multi-decadal variability signal in the West African Sahel along with the complexity of the Sahelian monsoon dynamics make long term rainfall predictions challenging. The CNRM, NCAR and GFDL models project a wettening trend with respect to 21st century regional precipitation and the GISS and CSIRO models project a drying trend with respect to 21st century regional precipitation.

The methodology for exploring the frequency of extreme events will borrow significantly from two prior publications: (Siebert and Ward, 2011 and Siebert and Ward, 2013). In these prior papers, Monte Carlo simulations have been used to model threshold crossing extreme event frequency under a range of assumptions about the trend and decadal variability of the index distributions. In the current study, there is a methodological expansion of using temporally evolving variance, in addition to exploring a trend in the mean and an autoregressive process to model the persistence. GCM based trends in the mean and variance are fed into the existing Monte Carlo simulation methods to yield extreme event calculations. These prior studies along with a great deal of supporting climate literature (Meehl et al. 2000, 2009) have found that the frequency of extreme events and the range of outcomes are quite sensitive to proposed changes in the mean and to levels of multi-decadal variability consistent with the IPCC and with other regional climate literature. Furthermore, the coefficient of variation (ratio of standard deviation to mean) is found to play a significant role in determining the frequency of threshold crossing events.

Since the frequency of threshold crossing extreme events will correspond, in the index insurance price model, directly to the frequency of payouts, this issue has direct bearing on the pricing and long-term sustainability of such an index insurance contract over time in the context of a changing climate. Models with a drying trend tend to simulate an increase in the index insurance price for the dryland subsistence farmer population and models with a wettening trend tend to simulate an increase in the index insurance price for the irrigated farmer population.