An Evaluation of the performance of Logit Models in Stock return prediction

The goal of accounting knowledge is to provide useful information for economic decision making. A major criterion for the usefulness of information is its prediction ability. Predicting stock return is difficult; but always interesting to the investors. Investors and decision makers need information for predicting stock return. Part of this information is covered by financial reporting. Financial reports are the products of accounting, therefore, its usefulness can be considered by the role of financial information in predicting stock returns. One of the most important subjects in such predictions is selecting the appropriate models. This study examines the usefulness of accounting information in an empirical way by using logit models and selecting the stock return's differences. Seventeen financial ratios are included in the study, based on the results of an Iranian research (Rahmani, 2003 & Sheri, 2005) and the results of foreign research (Ou & Penman 1989). Dependent variables are raw return, market-adjusted return and size-adjusted return, as mentioned in three definitions of return. The logit models were estimated based on a pooled data set of a sample of 74 firms listed in Tehran Stock Exchange over a five-year period of 2001 - 2005 and also for each year since 2001-2005. The results demonstrated that financial statements items could predict and explain the stock return changes. The meaningful financial ratios that were significant in estimated models for each single year were not constant over the period, which indicated changes of accounting variables (financial ratios) as an effective factor in the changes of stock return. To compare the performance of logit models with regression in predicting stock return, prediction regression models were estimated over five years (2001-2005). Results demonstrated that the two models performances were not meaningfully different. Finally, on the basis of logit model's prediction over 2001-2003 & 2002-2004, hedge portfolios are formed on the base of two cut off pr plans (0.5, 0.5) & (0.6, 0.4). Results indicated that from 12 portfolios, only 4 portfolios could earn excess returns. This study, by gathering empirical documents about the usefulness of accounting information and investigating the logit model in stock return prediction and implementing the hedged investment strategy, can contribute to the accounting and investment knowledge.