Stock return seasonality and the temperature effect

Abstract

Motivated by prior evidence that the relation between temperature and stock returns may be spurious, this study investigates the extent to which accounting for seasonality changes the explanatory power of temperature for stock index returns. Prior research using monthly data indicates that the portion of variability in stock returns that is explained by temperature can be explained equally well by any seasonal variable. Using daily stock market index and temperature data from 42 countries the effect of temperature on both the mean and variance of stock returns is analyzed through the use of GARCH modeling. The results show that a significant portion of the temperature effect is due to seasonal component of raw temperature. Furthermore, deseasonalized temperature has a moderate impact not only on the mean but also on the conditional variance of stock index returns. The results also indicate that the Halloween indicator, which is a seasonal dummy variable, has much less explanatory power using daily rather than monthly data. Its presence does not affect the explanatory power of deseasonalized temperature. Based on the findings, the paper concludes that although the relation between temperature and stock returns is not spurious, it is weaker than indicated by some earlier studies.