Evaluation of Pooled Cross-Sectional Earnings Forecasting Models: An Indian Evidence

Abstract

Purpose : Earnings forecasts are essential for valuation, and a bleak coverage of analysts’ forecasts in emerging economies withholds the valuation research and practices. This study compared the pooled cross-sectional earnings forecasting models in the Indian market to choose alternative sources for earnings forecasts to solve the unavailability of analysts’ earnings forecasts. Specifically, evaluating the theoretical earnings forecasting models of three different propositions: the earnings persistence (Li & Mohanram, 2014, EP) model, the Hou, Van Dijk, and Zhang (2012, HVZ) model, and the Pope and Wang (Harris & Wang, 2019, PW) model. Methodology : This study considered all companies listed on NSE from 1995 – 2022 in an unbalanced panel structure with 36,591 firm years observations. Robust regression was used for the coefficient estimation because of its capability to handle outliers and provide a better model fit. Findings : The results showed that the pooled cross-sectional models are reasonably accurate with the Indian data, restricting average forecast errors between 3% to 10%. The coefficient of earnings greater than one across models signified a high persistence in earnings. The PW model outperformed the other two models in the short run with share prices as predictor variables; whereas, the EP model performed best in the long run. The PW and EP forecast offered incremental information fully encompassing the HVZ forecast. Practical Implications : This study elevated the application of valuation in theories in research and managerial practices where firms’earnings forecasts are an essential input. Originality : This study uniquely compared the earning forecasting models of three proportions in a single setup to validate and suggest sources of earnings forecast for the Indian capital market.