Investor Psychology and Return Seasonalities in the Cross Section

This dissertation studies return seasonalities in the cross section and highlights the link between investor psychology and cross-sectional seasonalities. We document four main categories of return seasonalities in the stock market where predictable return persistence or reversal occurs across different calendar months, weekdays, holidays and firms' earnings announcement days. First, relative performance across stocks during months with high (low) market returns tends to persist during months when aggregate returns are predicted to be high (low), but reverse during months when aggregate returns are predicted to be low (high). Second, relative performance across stocks on weekdays with high (low) market returns tends to persist on weekdays when aggregate returns are predicted to be high (low), but reverse on weekdays when aggregate returns are predicted to be low (high). These two types of seasonalities are robust to placebo tests and don't diminish when controlling for risk and characteristics. Using such months or weekdays, we construct a mood beta with strong predictive power for monthly or daily returns. Third, we document a strong pre-holiday seasonality where stocks with above-average return during the two to three days immediately preceding or on a holiday tend to earn above-average return during the same pre-holiday window for at least 10 years. This pre-holiday seasonality is long-lasting, cannot be explained by a host of firm attributes, is present in foreign equity markets and only among firms with a retail clientele, and tends to reverse in the immediate, post-holiday period. A long-short strategy based on pre-holiday seasonality earns abnormal returns with high Sharpe ratios. Last, we document a strong seasonality during firms' earnings announcement days. We show that a firm's cumulative abnormal return (CAR) in a given fiscal quarter in a given fiscal year exhibits strong persistent patterns at fiscal annual intervals, while SUE doesn't. Such persistence exists regardless of the content of SUE. We also show a strong persistence of firm's earnings response (ER) at fiscal annual intervals. Collectively, we document a broad set of strong and puzzling return seasonalities in the cross section that cannot be explained by risk or characteristics. A model is provided to support our hypothesis that investors' mood, attention and expectation swings are important sources of such seasonalities in the cross section.