Determinants of maize production in Tanzania: a comparative analysis of multiple linear and principal component regressions

The study aims to examine the key determinants of maize production using fertilizer, land use, precipitation and temperature as predictors. It specifically seeks to address multicollinearity among predictors and to assess whether principal component regression (PCR) provides more robust results compared to traditional multiple linear regression (MLR). Time-series data covering the period 1961–2022 were used. PCR is applied to transform correlated predictors into orthogonal principal components and its performance is compared with that of the MLR model to evaluate explanatory power and statistical significance of determinants. The findings showed that, PCR model explains 43.2% of the variation in maize production, demonstrating moderate predictive power similar to MLR. PCR identifies land use , temperature and fertilizer as statistically significant at the 5% level, while MLR finds only land use and fertilizer significant. This indicates that PCR effectively uncovers the previously masked effect of temperature due to multicollinearity. The intercept suggests a baseline production level of 13,051 tons when predictors are at their mean values. In modeling under multicollinearity conditions, the PCR demonstrated the methodological advantage in agricultural production. The PCR provides enhanced empirical evidence to inform climate-smart and data-driven agricultural policy decisions by revealing the hidden climatic influences, especially temperature.