STOCK RETURN PREDICTION IN THE NIGERIAN STOCK MARKET
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Accounting Department, Nasarawa State University, Keffi
DOI
Abstract
There have been contending issues on the predictability of stock market returns. The arguments range from best variable to predict returns to methodological appropriateness.
This study, used composite approach to categorise the predictive variables into fundamental, technical, macroeconomic and global variables. The study adopted in-sample and out-of- sample dynamic error component models, Autoregressive Integrated Moving Average (ARIMA) model and Artificial Neural Network (ANN) model to examine the predictability of stock returns in the Nigerian stock market. The data used in this study were for all the listed firms in the market for the 15 year period of 2001 to 2015. The data were grouped into a cross-section of eleven industries. The study shows that all the variable categorisations indicate significant predictive power for in-sample test using dynamic error component model. The out-of-sample tests shows superior predictive power compared with the in-sample tests results. The automatic ARIMA (2,0,4) model also shows significant predictive power with the predictive variables being treated exogenously. The ARIMA model shows better predictive power than the error component model. The ANN model further shows superior predictive power than the dynamic error component and ARIMA models since it has high determinant value. The superiority of the ANN is attributed to its learning ability and capacity to accommodate more predictors with little restrictions. For individual predictive power of the variables, momentum, mean reversion and beta have significant predictive power across all the models. The study recommends the adoption of variables composite approach as used in this study to forecast stock return. By using hybrid of variables, the models will be able to capture varied trends in variables which have varying input implications than relying on individual variable or a category of variable for prediction. This approach has better predictive power as its prediction is likely to be more stable over time.
Description
Keywords
Citation
BEING A THESIS SUBMITTTED TO THE SCHOOL OF POSTGRADUATE STUDIES NASARAWA STATE UNIVERSITY, KEFFI IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE AWARD OF DOCTOR OF PHILOSOPHY (Ph.D.) IN ACCOUNTING AND FINANCE
