Normality of Indian Crop Yields: Application of Panel Analysis

This paper has two-fold contribution. First, normality of the crop yield residuals are examined using panel statistical procedures accounting for trend, autocorrelation and heteroskedasticity. Second, to evaluate the importance of accounting for spatial and temporal variation on the normality of crop yield residuals, the changes in the skewness, kurtosis and D’Agostino-Pearson(K2 ) omnibus normality test across panel and time-series models are examined. These are examined with an empirical application to district level data from 1950- 2002 for 15 crops and 14 states in India composed of a total of 3143 individual reporting districts in India. Results indicate crop yield residuals were normally distributed in 65 percent and 73 percent of districts, respectively based on skewness and kurtosis statistics. Accounting for spatial and temporal variation seems to change the distribution of crop yield residuals in 20 percent, 14 percent and 17 percent of districts based on skewness, kurtosis and omnibus tests respectively.