A Factor Model of Seasonality in Stock Returns

The Quarterly Review of Economics and Finance 44 (2004) 224–236

A factor model of seasonality in stock returns

Javier Gardeazabal∗ , Marta Regulez

University of the Basque Country, Bilbao 48015, Spain Received 27 June 2002; accepted 13 March 2003

Abstract Most empirical evidence on stock market seasonality is based on the Dummy Variable Approach (DVA). Typically, the DVA leaves too much variability of stock returns unexplained and inference usually leads to weak or null evidence in favor of seasonality. In this paper, we propose an extended DVA (EDVA) which leaves a lower fraction of stock return variability unexplained. We provide empirical evidence on daily seasonality in the Spanish stock market. Inference based on the EDVA finds positive and significant Monday and Friday effects and negative and significant Wednesday and Thursday effects. Extending the analysis to a model with GARCH conditional variances confirms these results and shows heavy daily seasonality in conditional variances. © 2003 Board of Trustees of the University of Illinois. All rights reserved.

JEL classification: C33, G12 Keywords: Stock returns; Daily seasonality; Common risk factors

1. Introduction Most empirical evidence on stock market seasonality is based on the so called “Dummy Variable Approach” (DVA), e.g., French (1980). This approach is based on a regression of stock returns, typically, the return on a market portfolio, on a set of seasonal dummies. The aim of this paper is to illustrate a problem with this approach. Typically, the DVA to seasonality leaves too much variability of stock returns unexplained. Although this is the case for all types of seasonal patterns, the portion of variability unexplained by the DVA increases with sample frequency. In this context, inference on daily seasonality usually leads to weak or null evidence

∗ Corresponding author. Present address: Universidad del Pais Vasco, Dpto. Fundamentos del Analisis Economico II, Avda Lehendakari Aguirre 83,...