Timely repayment of agricultural loans: Evidence from Serbian farmers

Abstract

Our study aimed to determine factors influencing the timely loan repayment of smallholder farmers. We used data from 1735 liquidated loans, collecting a set of 36 feasible determinant variables. The study was two-folded. In the first step, with a 64% accuracy, a Logit model revealed 18 significant predictors of timely repayment. Previously credited clients, special agricultural accounts, average monthly inflow, loan amount, age when applying for a loan, clean credit history, and no credit in the past have a positive influence. In contrast, the number of transactions, profiling, owned farm area, past due records over five days, tax debt status, and livestock negatively influenced timely repayment. In the second step, we used machine learning algorithms to enhance model prediction performance. XGBoost model has envisioned timely repayment with 92% accuracy. As significant predictors, Shapley’s additive explanations identified clean credit history, average monthly inflow, time of owning the account, age when applying for a loan, and horticulture. The study’s findings provide insight into the critical factors in substantially achieving a high repayment rate on borrowed funds.