Dichotomous dependent variable analysis of farmers decision on land utilization: A comparison between probit and linear probability models

摘要

The study investigated what drives farmers’ decision not to utilise their land for cultivation. The study utilised data from Statistical Office of Kosovo (SOK) with sample size of 4187 agricultural households. To achieve the objective, distance and transaction costs were examined, controlling for socioeconomic characteristics. The dichotomous dependent variable, dummy=1 if land fallow otherwise=0 was regressed over a set of explanatory variables. The model was estimated using both probit regression and linear probability model (LPM). Diagnostics test were performed after estimation. From the results, probit model passed almost all the tests. The results from both probit and LPM have low R2= 0.093. However, most of the explanatory variables showed a consistent sign and significance with useful insights into the determinants of land fallow decisions. The log likelihood ratio (LR) significance test from probit model confirmed the variables of interest are statistically jointly significant at 99% confidence level (p-value<0.001). The study revealed that distance and transaction costs along with socioeconomic factors significantly affect the decision to leave land uncultivated. It was concluded that probit model is better suited than LPM when the dependent variable is dichotomous. A combination of policy measures to reduce possibility of leaving agricultural fallowed was recommended.