Working paper 49 - A Stochastic Production Function Analysis of Maize Hybrids and Yield Variability in Drought-Prone Areas of Kenya

Author(s):  Ashley D. Jones, Timothy J. Dalton, and Melinda Smale


Introduction:

Most of the world's 160 million hectares of maize are rainfed, and an estimated 15 percent of global maize production is lost each year to drought (Edmeades, 2008). Drought stress, which occurs even in better-watered areas, is one of two physical factors that most limits maize productivity (soil quality is the second). Globally, the relative importance of water constraints appears to be highest in Sub-Saharan Africa (Cooper et al. 2008). Over a decade ago, before scientists recognized the potential impact of climate change, Heisey and Edmeades (1999) estimated that in Sub-Saharan Africa, roughly a quarter of the 18 million ha of maize then grown in the lowland and mid-altitude subtropics grew under frequent stress from drought. A study of the economic costs of climate change in Kenya concluded that, on top of the substantial costs of existing climate variability in Kenya, the future additional costs of climate change could be equivalent to a loss of 2.6% of GDP each year by 2030 (SEI 2009).

Variation in precipitation, which can result in yields that are 20 percent lower or higher on an annual basis (Isik, 2003), makes production risk in the drought-prone areas of Kenya a significant element of farmer decision-making process.

Most smallholder farmers in Kenya grow maize, and according to panel data collected since 1997 by Tegemeo Institute, most maize growers plant hybrids (Smale and Olwande, 2011). Historically and today, adoption rates have differed sharply across agro-ecological zones (Hassan 1998; De Groote et al., 2005; Smale and Olwande 2011). Smallholder farmers in Kenya, and particularly farm households that grow maize for food, have limited access to credit and no access to insurance. They have strong incentive to plant seed that reduces the variance of yields and limits their exposure to downside risk. Donors and the Government of Kenya are currently investing in the development of maize varieties that are tolerant to drought and water-use efficient.

To our knowledge, whether hybrid seed reduces or exacerbates yield risk (variability and downside risk) for farmers in drought-prone areas of Kenya has not yet been tested in an economic model of farmer decision-making. This analysis has two objectives. The first is to explore the effects of the maize hybrids currently grown by smallholder farmers in drought-prone areas of Kenya on the mean, variance and skewness of yields. We test these effects in the framework of a stochastic production model, including both full order and partial order moments. The second is to provide "baseline information" to gauge the future impacts of the adoption of water-efficient improved maize by the same population of farmers. This study contributes to an old debate whether the yields of improved seed varieties are higher but also more variable than those of unimproved, farmers' varieties, especially given the use of nitrogenous fertilizer (e.g. Anderson and Hazell 1989).

It also has relevance for a continued debate concerning the efficiency of agricultural research investment to increase domestic maize productivity in high potential areas versus the equity implications of investing in areas with lower yield potential (e.g., Karanja, Renkow and Crawford, 2003). Karanja, Renkow and Crawford (2003) concluded that while technology adoption in high potential areas has a substantially greater, positive impacts on aggregate real incomes, it is likely to have inferior income distributional outcomes compared to technology adoption in marginal regions.

A premise of plant breeding research for drought-prone environments is that improved seed can reduce yield risk for farmers and enhance the food security of smallholders in marginal areas. In Kenya, such a strategy may also reduce the maize import bill, with consequences for the national economy. The methodological framework applied in this study is described next, including the stochastic production framework, specification, functional form and data source. Selected descriptive statistics and econometric findings are presented in Section 4. Conclusions are drawn in the final section.

A Stochastic Production Function Analysis of Maize Hybrids and Yield Variability in Drought-Prone Areas of Kenya