Journal of Contemporary Issues in Business and Government

Abstract

Contradicting evidence on time-series and financial analysts’ forecasting performance calls for further research in financial markets. Motivation to use time-series models rather than analysts’ forecasts stems from recent research that reports time-series predictions to be superior to analysts’ forecasts in predicting earnings for longer periods and for small firms that are hardly followed by financial analysts. The paper aims to explore performance of time series models in forecasting earnings for six firms considering historical data of 11 years from January, 2010 to December, 2020. Monthly average stock data of last 11 years for five firms namely HCL, TCS, Infosys, Reliance, Tech Mahindra and Wipro was considered from NSE site. Every company had 132 values whose graphical plotting and stationarity check was performed. Data series for each of the five companies was found to be non-stationary. After differencing each of them, the series became stationary and graphical plotting was again done. Then best suited ARIMA Model for each stationary time series was determined upon comparison of goodness of fit statistics. After choosing the best suited ARIMA model, residuals were extracted and were found to be random with no external influence whatsoever. Hence forecasting was done based on chosen model for the monthly average stock price of these top six companies of India in 2020. The paper finds that premier ARIMA family models outperform naive time-series models in terms of mean percentage errors, AIC and average ranks. The findings suggest that investors use the selected ARIMA model to form their expectations