Vector auto-regressive model with exogenous climate conditions for rice varieties in Northern Ghana

Abstract

Multivariate time series analysis is one of the complex technique for studying a consistent time-dependent data on different variables. The data used for the study, which is obtained from the Centre for Scientific and Industrial Research (CSIR), cover the period 1957 to 2016 and involved two sets of variables: one set is made up of yield of four varieties of rice (Mandii, Bazulgu, O.Sativa and Kpukpula) and the second set consist five climate conditions (Annual Rainfall, Temperature, Evapo-tranpiration, Wind speed and Sunshine). As a result of the composition of the data, the classical Vector Auto-Regressive (VAR) model for the yield of rice varieties extended to include two sets of variables. An order-3 lag model is found appropriate for both sets of variables. It is found that inclusion of the climate conditions provides a better model for rice yield than the lagged rice varieties alone. It is also found that temperature and sunshine have a more consistent positive influence on rice yield, particularly for O.Sativa. The study therefore recommends the right varieties that can perform well under shocks of identified best-subset weather conditions.