Applying artificial neural network and binary logistic models to predict propensity to pay cash dividend: Evidence from an emerging market

Abstract

This study examines the predictors of the propensity to pay cash dividend including industry structure, natural log of revenue, firm size, big 4 auditors, and financial leverage. The paper draws upon the theory of uncertain binary choice. Pooled unbalanced panel logistic regression and artificial neural network were used to analyze data of 725 firm-year observations obtained from companies’ annual accounts and financial statements from 2012 to 2021. The documented results find that industry structure, natural log of revenue (big 4 auditors, firms’ size and financial leverage) significantly influence the propensity to pay (not to pay) cash dividend. The result on the interaction term shows that industry structure and log revenue has the propensity to significantly predict non-payment of cash dividend. Nagelkerke pseudo R2 indicates that the predictors explain about 36% of variability in payment of cash dividend. The ROC-curves indicate good model fits as areas under the curve are up to .85. We recommend that the management of listed companies and equity stockholders who are interested in dividend payment should consider the history of industry structure and companies’ revenue while those not interested in dividend payment should consider company size, the presence of big 4 auditors and financial leverage. Keywords: Cash dividends, Industry structure, Independent

Keywords

• Cash dividends
• Industry structure
• Independent auditors
• Uncertai binary choice
• Logistic regression
• Artificial neural network