Nonparametric Time Varying Coefficient Panel Model for Climate Change and Rice Yield in Two States in India

There is a scientific consensus that global climate change is already impacting on yields of major staple food crops such as rice, wheat, and soybeans, and this situation will progressively worsen posing severe dangers to global food security. The impact on crop yields of climate change may be varying over time and across states, which is not well addressed in the current literature. This paper proposes a three dimensional (3-D) panel model for rice yields with time-varying coefficients of climatic and non-climatic variables. Moreover, the yield trend is allowed to be an unknown function of time and it varies across the states and districts within the state. The trend and coefficient functions are estimated by the nonparametric local linear method. We apply this approach to two Indian states, one is a coastal state Andhra Pradesh with 11 districts, while the other is an inland state Telangana with 9 districts, for the period 1966-2015. We find that the rainfall effect has been mostly insignificant and then negative since 2005 for the coastal state, while it has been positive and increasing overall for the inland state. The average temperature, on the other hand, has been marginally positive for the coastal districts, whereas it has been mostly negative for inland districts. Both minimum and maximum temperatures have had deeper negative effects on inland districts than the coastal counterparts. The results indicate that farmers coastal state have a better adaptation plan for climate change than those in the inland state and that overlooking heterogeneity in the coefficients can vastly underestimate the climate change impact on the crop yields and food security