Losing sleep at the market: an empirical note on the daylight saving anomaly in Australia

Economic Papers (Economic Society of Australia), Sept, 2003 by Andrew C. Worthington

1 Overview

In a recent provocative article, Kamstra et al. (2000) found that the avera e Friday-to-Monday stock return on dayli ht savin weekends was 200 to 500 percent lar er than the avera e ne ative return for other weekends (the so-called 'weekend-effect' market anomaly) and thereby associated with a one-day loss of US$31 billion on the NYSE, AMEX and NASDAQ markets alone. Kamstra's et al. (2000) findin s appeared to hold not only in the United States and Canada, where the transition to and from dayli ht savin is broadly similar, but also in the United Kin dom, whose patterns differ from that in North America, and to a lesser extent in Germany. On this basis, Kamstra et al. (2000, p. 1010) su ested that if dayli ht savin was associated with "... the sort of impact investi ated here, an obvious policy implication is to do away with the time chan e alto ether".

The essence of Kamstra's et al. (2000) ar ument is that the 'dayli ht savin effect' is linked with sleep desynchronosis associated with the chan e in the circadian rhythm and its (ne ative) impact on sleep patterns. Every Sprin at 2:00 a.m. on the first Sunday in April US clocks are moved forward one hour, and the followin Fall at 2:00 a.m. on the last Sunday in October clocks are moved back one hour. As with jet la, where chart es in sleep patterns are thou ht to persist up to five days for each one-hour time zone crossed (Waterhouse et al. 1997), the movement to dayli ht savin time (DST) also compresses the day, while the movement from dayli ht savin stretches it, and this also impacts upon sleep patterns.

If, and as hypothesised by Kamstra et al. (2000, p. 1006), "... sleep desynchronosis causes market participants to suffer reater anxiety about a iven situation, ceteris paribus, they may prefer safer investments and shun risk in trades durin the tradin day followin such a disturbance in their sleep patterns ... this could push down stock prices followin dayli ht savin shifts when the desynchronosis is systematic". In fact, the ar ument that shifts to and from dayli ht savin has an impact upon actual behaviour already has parallels elsewhere with Bick and Hannah (1986) and Shapiro et al. (1990) studyin the impact of DST on psychiatric presentation, and Coren (1996a, 1996b) Lambe and Cummin s (2000), Varu hese and Allen (2001) and Sullivan and Flanna an (2002) examinin its role in traffic and pedestrian accidents.

All the same, there are a number of complications associated with Kamstra's et al. (2000) purported dayli ht savin effect, which may not arise in non-financial market contexts. To start with, the dayli ht savin effects exists in parallel with the oft-examined weekend effect, for which a number of competin hypotheses have already been put forward and tested (see, for instance, A rawal and Ikenberry 1994, Chan et al. 1993, 1998 and Wan and Erickson, 997). These include la s in payment and cheque clearin settlements, midweek time pressures on individuals, the tendency for financial advice to be iven after Monday strate y-settin meetin s, and the lar er percenta eof purchases (sales) on Fridays (Mondays) at dealer ask (bid) prices (Kamstra et al. 2000). It may then be possible that the sleep desynchronosis associated with dayli ht savin weekends is just an alternative manifestation of this more usual market anomaly. For example, Pine ar (2002, p. 1256) countered that "the chan e in sleepin patterns from weekdays to weekends occurs with much reater frequency and is very plausibly more pronounced than the chan e in sleepin patterns between dayli ht-savin and non-dayli ht savin weekend. Thus sleep desynchronosis may contribute to the so-called 'day-of-the-week' effect on non-dayli ht savin Mondays also".

Another problem is that dayli ht savin transition weekends are by nature limited in number, and these may be juxtaposed with outliers. Once a ain, Pine ar (2002) questioned Kamstra's et al. (2000) conclusions on the basis that in the last ei hty years three of the lar est percenta e declines in the S&P 500 took place after a fall dayli ht-savin time chan e (most recently Monday 26 October 1987). In response, Kamstra et al. (2002, p. 1263) countered, "... while we do not believe that dayli ht-savin--time chan es cause market crashes, we do believe that dayli ht-savin--time chan es affect the de ree of market fluctuations ... we speculate that severe downturns are more likely followin dayli ht-savin weekends, and we ar ue that the data support this contention" (ori inal emphasis). Lastly, Pine ar (2000) also ar ues that the apparent corroboration offered by Kamstra's et al. (2000) inclusion of Canada, Germany and the United Kin dom in their analysis may be an illusion associated with the normal influence on them by the US market, and that such international evidence should then be treated more cautiously.

The purpose of the present paper is to add to this small but intri uin body of work the results of an analysis of the Australian equity market. To the author's knowled e this is the first of its kind in Australia, and adds si nificantly to the nascent literature concernin the economic benefits and costs of dayli ht savin. The paper itself is divided into four main areas. Section 2 discusses the startin and endin dates of DST in Australia since 1979/80. Section 3 explains the empirical methodolo y and data collection employed in the analysis. Section 4 presents the results. The paper ends with a brief conclusion.