• DOI:10.1257/00028280260344786
• Corpus ID: 16002134

Losing Sleep at the Market: Comment

@article{Pinegar2002LosingSA,
  title={Losing Sleep at the Market: Comment},
  author={J. Michael Pinegar},
  journal={The American Economic Review},
  year={2002},
  volume={92},
  pages={1251-1256}
}

• J. M. Pinegar
• Published 1 August 2002
• Economics
• The American Economic Review

In a recent provocative paper in this journal, Mark J. Kamstra et al. (2000) test and reject the hypothesis that the mean weekend return following changes in daylight saving time equals the mean weekend return throughout the rest of the year. The authors report that the average Friday-to-Monday return on daylight-saving weekends is 200–500 percent larger than the average negative return for the other weekends of the year. The Ž nding appears to hold not only in the United States and Canada where daylightsaving date patterns are similar, but also in the United Kingdom, whose patterns ostensibly differ from those in North America. The results also appear robust to alternative statistical methods based on time-varying conditional heteroscedasticity and/or bootstrapping. This paper provides further robustness tests of the results reported by Kamstra et al. I show that the difference between mean weekend returns for daylight-saving and non-daylightsaving weekends is signiŽ cant only for fall changes in daylight saving time and that the fall difference is driven by two outliers associated with international stock market crises. Two separate adjustments for the heteroscedasticity these outliers induce cause the signiŽ cance of the fall difference to vanish. The total sample (spring plus fall) difference remains marginally signiŽ cant for some indexes after heteroscedasticity adjustments with classical Ž xed-level hypothesis tests. However, Bayesian sample-size adjustments produce posterior odds ratios that consistently favor the null hypothesis of no daylight-saving anomaly over the alternative that the anomaly exists. I also fail to reject the hypothesis that daylight-saving and nondaylight-saving weekend returns have equal distributions. For these reasons, I question the robustness of the Ž ndings reported by Kamstra et al. (2000). Section I presents more details of my tests, and Section II summarizes my Ž ndings and discusses their interpretation.