Panel data models for studying the impact of climate change on certain agricultural crops using some test statistics

Abstract

The issue of climate change is now a global concern and its effects on agriculture in Nigeria have been investigated severally using either time series or cross sectional data approach. The use of time series and cross sectional data approaches led to misspecification errors or omitted variable bias and this affect the efficiency of the model estimation parameters. As a remedy to estimation bias, this study applies the panel data approach. Three panel data models - pooled, fixed effects and random effects models were used to fit a balanced panel data consisting of seven crops for twenty-eight years’ periods. The appropriate model was determined using some test statistics such as the F-Test, Breusch–Pagan Test and the Hausman’s Specification. The analysis of the weather shows significant variation over time indicating the presence of climate change. Using the F- test, the pooled model is rejected which suggests the presence of unobserved heterogeneity. The Breusch-Pagan test also suggests the rejection of the pooled model while the panel effects exist in the panel. The panel model is able to accommodate heterogeneity better than the pooled model. The Hausman test however indicates that the fixed effects model is more appropriate for the study because the assumption that the individual effects are uncorrelated with the other regressors in the model was rejected. The study further reveals higher coefficients value for the pooled model and
random effect model than the fixed effect model indicating that the values may have been overstated. The climatic factors were significant for the pooled and random effect models but not significant using the fixed effect model, thus revealing that the observed climate change have not significantly manifested so much on the selected crops in Nigeria. It was also observed from the study that the intercept coefficients for the different crops were individually highly significant; indicating the presence of individual heterogeneity. This suggests that different crops will be impacted differently. The study therefore affirms the presence of unobserved effects in the data and the fixed effects model accounts for this better than the polled and random effects model when estimating the impact of climate change on agriculture in Nigeria. Since the different crops are impacted differently, the study suggests crop specific and not country wide or national adaptation/mitigation policy