Management of climate risks in agriculture–will weather derivatives permeate?

Abstract

It is a matter of common knowledge that weather represents the major source of uncertainty in crop production. It is to be expected that weather fluctuations will increase in the future due to climate change. Traditionally, farmers tried to protect themselves against weather-related yield variations by buying insurances. More recently, there has been a discussion regarding the use of weather derivatives to safeguard against volumetric risks. Although weather derivatives display advantages over traditional insurances, there is only a relatively small market for these products in agriculture. This is partly attributed to the fact that it is unclear whether and to what extent weather derivatives are a useful instrument of risk management in agriculture. This study applies real yield and weather data from Northeast Germany in order to quantify the risk-reducing effect that can be achieved in wheat production by using precipitation options. To do so stochastic simulation is used. The hedging effectiveness is controlled by the contract design (index, strike level, tick size). However, the local basis risk and the geographical basis risk remain with the farmer. We separate both causes of basis risk and reveal the extent of each. This enables conclusions regarding the design of weather derivatives; thus the question dealt with here is relevant both for farmers and for potential sellers of weather derivatives.

Notes

¹ This definition may appear unusual since the rainfall deficit index will take negative values. However, the definition is convenient for the present application because then the relationship between yield and rainfall deficit index is similar to that between yield and rainfall sum index.

² Several further functional forms for the yield model have been tested; in particular a quadratic and a logarithmic production function. Nonetheless, the linear-limitational production function showed the best fit in terms of R ² for the empirical data and both rainfall indices. It should be noted that this result cannot be generalized. Vedenov and Barnett (2004) point out that a suitable yield-rainfall-model is dependent on type of variety and region.

³ It should be noted that an option cannot be designed in such a way that its payoff is correlated perfectly negatively to the expected revenue from the production, if production function did not display a linear-limitational function form. To insure the production risk for a linear production function, a future can be used. For more complex production functions, several weather derivatives can be combined. In this way, a combination of put and call options could be suitable for a quadratic production function. In the 'left area' ('right area') of the production function, the put option (the call option) insures against volumetric risks.

⁴ Regarding the number of required simulation runs, Haug (1998, p. 40), e.g. stipulates that at least 10 000 runs should be carried out. For technical details describing how to use stochastic simulation to model a wide variety of distributions with established software packages, see, e.g. Winston (1998).

⁵ We are only focussing on the risk-reducing effect of weather derivatives in wheat production, i.e. we abstract from cross effects resulting from the fact that the payoff of a weather derivative is correlated with the yields of several crops.

⁶ The de-correlation function is invariant regarding direction. Thus, topographical differences potentially influencing precipitation are neglected. In Brandenburg, topographical conditions play to the assumption of a correlation independent of location and direction.