This study explains and predicts the firm’s dividend payment decision using an unbalanced panel dataset of non-financial firms listed on the Pakistan Stock Exchange covering the period 2009-2020. Several traditional statistical models and modern machine learning algorithms are applied to examine the firm-specific and macroeconomic determinants of dividend decision. Logistic regression and dynamic panel probit models indicate that lagged dividend, firm efficiency and firm size positively influence the firm’s dividend decision whereas leverage and sales growth show negative effects. Lagged dividend is found to be the most influential variable in explaining and predicting the dividend decision while the role of macroeconomic variables is very limited. In the prediction context, the random forests model outperforms the other statistical and machine learning models. The findings provide useful insight for corporate managers, investors, policy makers, and researchers by combining explanatory strength of statistical models and predictive power of machine learning algorithms in a unified framework to understand the dividend payout behavior of Pakistan’s non-financial sector.
Copyrights © 2026