A Study to Mine Association between Periodical Returns and Business Financial Metrics

Abstract

Over the last five decades, business academics have identified over three hundred ‘return determinants’ that potentially influence stock returns. However, we still do not know if all return-determinants are equally important, or whether a smaller set of return-determinants has a disproportionately larger influence on stock returns. To answer that, a study needs to include all return-determinants of interest in the same study. However, asset pricing research so far has heavily relied on linear regression to study the relationship between stock returns and return determinants. And one of the main constraints with linear regression as a method is its inability to include more than a few potential return-determinants in a single asset pricing study. This constraint is one of the reasons why asset pricing research so far has not been able to answer whether all three hundred plus return-determinants have equal influence on equity returns or whether there is a smaller set that wields a disproportionately higher influence on stock returns. Asset pricing research needs to look beyond linear regression and seek research methods that does not limit a researcher to determine the relative influence of return-determinants on stock returns. In recent times, association mining has been successfully applied to gain powerful insights in multidimensional phenomenon like understanding consumer behavior in a retail store or accurate and early diagnosis of a serious illness. The research problem on hand about the relative influence of return-determinants on stock returns is also multi-dimensional in nature. Hence, the research carried out here applies the technique of ‘association rule mining’. This method allows placing all potential ‘return-determinants’ along with equity returns in a single frame to mine association rules between return-determinants and index beating returns. This approach uncovers strong association rules between a small set of return determinants and index beating stock returns. This small set of ‘key’ return-determinants strongly associated with index beating returns exhibit a disproportionately larger influence on stock returns compared to that of other known return-determinants included in this study. I call this set as the set of ‘key return determinants’. Multiple portfolios are created using the association rules mined in the USA and India. These portfolios provide index-beating performance at a risk lower than the market risk And unlike linear regression, association rule mining does not limit the number of factors that can be included in a single study. Encouraged by the results of mining strong association rules between stock returns and return determinants in a single country data-set, I extended the approach to mine association rules in a multi-country data-set. The objective of doing so was to look for universally applicable strong association rules, which would provide a set of ‘key return-determinants’ applicable to all markets. Such a set of universally applicable ‘key return-determinants’ could become the basis of a universally applicable asset pricing model. However, the association rules mined in a multi-country data set were not as strong as they were in the single country data set (USA or India). This leads me to believe that the strength of association between stock returns and return determinants could be different in different countries, and some of the reasons for those could be: - Regulatory aspects like tax structure, government incentives and subsidies among others. - Differences in capital availability, interest rates and other macro-economic factors like head room for economic growth and inflation could influence stock returns to a great extent Other interesting findings of this study are: - For USA market, using association rule mining, from a pool of forty-four return determinants, we identified a set of five ‘return determinants’ that exhibit a higher influence on stock returns compared to the other return-determinants included in the study. These key return-determinants are seen most frequently in investments with market beating returns. - For India market, using association rule mining, from a pool of thirty-seven return determinants, we identified a set of eight return determinants that exhibit a higher influence on stock returns compared to the other return-determinants included in the study. These key return-determinants are seen most frequently in investments with market beating returns - For both USA and India markets, out-of-sample portfolios created from the association rules have portfolio ‘Beta’ less than one and provide a significant Jensen’s alpha for all holding periods. - For USA markets, the analysis of key return determinants shows that the most important risk to guard against is the risk of ‘overpaying’ for an investment (valuation risk) - For India markets, the analysis of key return determinants shows that companies (stocks) that deploy capital efficiently provide index beating returns. Similarly, companies (stocks) showing consistent sales and profit growth also provide index beating returns. - Associations mined from the multi-country data set point to the possibility of: o Different asset pricing models for different countries o Different return predicting models for different holding periods (short term, medium term and long term) Practicing investors and portfolio managers can leverage the findings of this research in the below manner: - Portfolio managers looking to invest using the 'factor investing' strategy can directly apply the association rules mined in this study to construct portfolios that yield index-beating returns for holding periods ranging from one year to fifteen years. Alternatively, portfolio managers can use the set of 'key return-determinants' identified in this study as factors around which they can develop their factor investment strategies for USA and for Indian markets. - Investment managers looking for techniques to reduce a large, unconventional factor pool to a smaller set, can use the simple technique of ‘association rule mining’ to achieve that. Portfolio managers can use association rule mining to build portfolios without any limits on the number of factors that can be included in the screening process. - This study also provides some interesting insights for equity investment in general. For example, in the US markets, limiting the ‘valuation risk’ seems to be the most important criteria for achieving index beating returns. - Similarly, for a capital constrained, but growing economies like India, companies that deploy capital more efficiently and consistently report growth in sales and book-value provide superior returns compared to the index. My research here looks to answers some of the unanswered questions in asset pricing research and come up with insights that could bridge the wide gap seen between theory and practice of equity investing. The work here tries to answer the 'what' part of the relation between equity returns and return-determinants: ‘What’ return-determinants are materially important in determining equity returns? Are they same or different for different countries? More work is needed to answer the 'why' part of these questions, i.e. 'why' a smaller set of 'key return determinants' seen here wield higher influence on stock returns than the rest. Subsequent research can leverage recent advances in artificial intelligence and large language model capabilities to include important qualitative aspects like corporate governance, investor sentiment, long term economic outlook of a nation, expected growth of company and similar other factors to understand the impact of qualitative aspects on equity returns.