Investor sentiment: From global to local

Abstract

Limits to arbitrage and biases of investors have challenged efficient market hypothesis (EMH) and have given room for investor sentiment in explaining stock market returns. Accordingly, investor sentiment has gained strong traction in the last 20 years and become one of the leading research topics in "Behavioral Finance" literature. Many articles have exceeded 1,000 citation threshold (some have exceeded even 10,000 threshold). This dissertation starts investigating investor sentiment from a global perspective and gravitates towards an emerging market (Turkey) with different hypotheses. This dissertation has three main parts. First, this study focuses on investor sentiment and its explanatory power in anomaly returns in a global context. The findings indicate that investor sentiment's power in explaining anomalies might be universal, however changes based on anomalies. Moreover, the explanatory power of sentiment changes when the size factor is controlled. In other words, the impact of sentiment on returns is related to market capitalization of stocks. For example, sentiment can explain investment anomaly in the US market. However, when separately analyzed, one might notice that this is rather the case for small stocks but not for large stocks. This first part of the dissertation has several implications. First, preference for indirect sentiment indicators over direct ones may improve prediction models both in academia and practice. Secondly, asset managers who follow factor-investing strategies can consider the impact of the size factor. For instance, for book-to-market strategy, selecting larger stocks when sentiment is expected to advance might result in higher returns. On the other hand, the opposite method of trading is valid for investment strategy. The first part of the dissertation offers new ground for anomaly research by fixing the stock characteristics when analyzing the effect of sentiment. Additionally, creating pure anomaly returns from combination portfolios is a new practice that could potentially gain traction. Secondly, this dissertation gravitates toward an emerging market (Turkey) and investigates whether excess stock market returns are related to individual and institutional sentiment. In detail, individual investor sentiment is examined by using fund flows and institutional sentiment is examined by using stock intensity of funds (in other words, asset allocation abilities are tested). Results indicate that flows to mutual funds exhibit no relationship with the stock market returns. Hence, unlike the US market, individual investor sentiment is not related to stock returns in Turkey. In turn, stock intensity of mutual (pension) funds is positively related (unrelated) with excess returns. The predictive power of sentiment indicators is not strong. Some differences between mutual and pension funds can be attributed to the relative asset allocation performance of fund managers and the fee structure in the industry. Third, this dissertation investigates the relationship between asset managers' stock selections and investor sentiment indicators. The findings indicate that stock selections of asset managers are related to sentiment indicators. When sentiment rises, asset managers are inclined to invest in stocks with lower size, lower volatility, lower dividend yield and higher momentum. Yet, characteristics of stocks (size, volatility, price-to-earnings ratio, momentum etc.) in mutual funds are not significantly related to stock returns. Herding and lack of stock selection skills may be the reason behind this result. Another finding is that stocks prone to sentiment changes have higher size, and that strategy of going long with low sentiment prone stocks and going short with high sentiment prone stocks do not yield significant return, hence such a strategy is not an anomaly.