Three Essays on Stock Market Return Predictability:The Role of Average Correlation of Industry Portfolio Returns

Abstract

Stock market return predictability has long been one of the key and unsolved areas of research in finance. Although the stock market has been argued to follow a random walk, researchers have struggled to improve the accuracy of predicting stock market returns through extensively examining forecasting variables such as financial ratios, economic indicators, and behaviour factors. Pollet and Wilson (2010) have recently developed a new predicator and claimed that average correlation reveals the movement of the systematic component of the market return and it predicts the stock market returns. This thesis uses the newly developed predictor, average correlation, to predict stock market returns, both in the US and across a number of developed countries and emerging countries. Three interrelated studies are sequentially undertaken to examine the predictive power of average correlation for future stock market returns. The first study uses the average correlation of the 48 Fama-French industry portfolio returns in the US stock market to predict the US stock market returns. To juxtapose average correlation with conventional predictors, a number of forecasting variables, including term spread, default spared, dividend price ratio, the cyclically adjusted price-to-earnings ratio and investor sentiment, are incorporated in the model. The second study uses 27 non-US financial markets and extends the analysis to the relatively less explored area relating to the predictability of the international stock market returns. The average correlation of industry portfolio returns in each financial market, including more forecasting variables such as industrial production, gross domestic production and financial crisis dummies, is used to predict the stock returns of the financial markets under study. The third study further extends the analysis and uses both the US average correlation from the first study and the local average correlation from the second study as predictors for the stock market returns of each financial market. The US average correlation is posited as capturing the global influence on a particular financial market, while the local average correlation is used to represent the domestic influence within that financial market. The key findings of the thesis are summarised as follows. First, average correlation is a significant predictor for the US stock market returns at a two-month lag and for the returns of other stock markets with a one-month lag. Second, average correlation outperforms all predictors conventionally used in the US stock market, as well as in most other international stock markets. Third, the US and local average correlations predict the local stock market returns, indicating that the global influence has an impact on the local stock market returns and that the US average correlation successfully captures such an influence. The research findings suggest that the average correlation is closely related to stock market returns. The findings of the thesis would be of interest to policymakers as well as stock market practitioners who wish to formulate effective trading strategies.