Who Pays Dividend? Statistical Explanation and Machine Learning Prediction

Abstract/Description

Purpose: A firm’s dividend payment decision depends on several factor e.g., liquidity, cash flows, profitability, leverage, sales growth, size, and efficiency of the firms as firm specific variables. The broader set of firm specific and financial or macroeconomic variables contribute to this decision. Thus, dividend payment decision can have complex non-linearities. Earlier research on this issue has focused on modeling the dividend payment decision using statistical or econometric methods e.g., logistic regression.

Study Design/Methodology: This paper explores the factors influencing the dividend-paying behavior of non-financial firms in Pakistan using panel data spanning from 2009 to 2020, encompassing 513 non-financial firms using statistical and machine learning techniques.

Findings: Logistic regression found significant relationships between lagged dividends, firm efficiency, size, and exchange rates with dividend payments, while liquidity, real GDP growth, world GDP, and profitability are found insignificant. The random forests highlight the importance of variables like last year's dividend, current ratio, size, and sales growth. The ANN outperforms other ML techniques with the highest accuracy and the area under the ROC curve for firm-specific models.

Originality/Value: The paper power of machine learning methods in predicting firm’s dividend payment decision using panel structure.

Research limitations/implications: The paper employs only the most popular ML methods, not including deep learning or certain non-parametric methods e.g. KNN.

Practical implication: The paper provides findings that can assist in choosing firms that are more likely to pay dividends.

Social implication: The research can potentially benefit saving investor types e.g. pensioners, retired individuals, or institutional investors.

Track

Finance

Session Number/Theme

1A: Finance

Session Chair

Dr. Adnan Haider; Dr. Aitzaz Ahsan Alias

Start Date/Time

30-5-2024 1:50 PM

End Date/Time

30-5-2024 3:20 PM

Location

MCS – 3 AMAN CED Building

Comments

An earlier version of this paper was presented at the NUST Recent Trends in Statistics and Data Analytics conference Dec 2023

This document is currently not available here.

Share

COinS
 
May 30th, 1:50 PM May 30th, 3:20 PM

Who Pays Dividend? Statistical Explanation and Machine Learning Prediction

MCS – 3 AMAN CED Building

Purpose: A firm’s dividend payment decision depends on several factor e.g., liquidity, cash flows, profitability, leverage, sales growth, size, and efficiency of the firms as firm specific variables. The broader set of firm specific and financial or macroeconomic variables contribute to this decision. Thus, dividend payment decision can have complex non-linearities. Earlier research on this issue has focused on modeling the dividend payment decision using statistical or econometric methods e.g., logistic regression.

Study Design/Methodology: This paper explores the factors influencing the dividend-paying behavior of non-financial firms in Pakistan using panel data spanning from 2009 to 2020, encompassing 513 non-financial firms using statistical and machine learning techniques.

Findings: Logistic regression found significant relationships between lagged dividends, firm efficiency, size, and exchange rates with dividend payments, while liquidity, real GDP growth, world GDP, and profitability are found insignificant. The random forests highlight the importance of variables like last year's dividend, current ratio, size, and sales growth. The ANN outperforms other ML techniques with the highest accuracy and the area under the ROC curve for firm-specific models.

Originality/Value: The paper power of machine learning methods in predicting firm’s dividend payment decision using panel structure.

Research limitations/implications: The paper employs only the most popular ML methods, not including deep learning or certain non-parametric methods e.g. KNN.

Practical implication: The paper provides findings that can assist in choosing firms that are more likely to pay dividends.

Social implication: The research can potentially benefit saving investor types e.g. pensioners, retired individuals, or institutional investors.