Red sky at night or in the morning, to the equity market neither a delight nor a warning: the weather effect re-examined using intraday stock data

Abstract

Unlike most of the existing literature on the weather effect, we conducted our analysis by employing intraday weather and market data, examining a large set of stocks rather than indices only, including volume and volatility data in the study and inspecting a wide number of weather variables (temperature, humidity, pressure, visibility, wind, cloud, rain and snow). Our analysis covered the Italian stock market for the period August 2005–March 2014 for a total of 2201 trading days. We conclude that no systematic relationship seems to exist between the weather and the Italian stock market. Moreover, our results raise doubts that testing the weather effect by limiting the analysis to indices only can lead to spurious conclusions.

Keywords:

• weather effect
• behavioural finance
• Italian stock market
• market anomalies
• efficient market hypothesis
• trading behaviour
• sunshine effect

JEL Classification:

• G02
• G11

Acknowledgements

We are indebted to Professor Maurizio Fanni, Professor Massimo Bilancia and Professor Caterina Marini for their helpful suggestions in reviewing the preliminary drafts of this paper as well as to the Editor, the Associate editor and an anonymous referee for their insightful comments. All remaining errors are our own. We are also grateful to the Servizio Meteorologico dell'Aeronautica Militare (Italian Air Force Meteorological Service) for having generously provided the meteorological data analysed in this work. Although the paper is the outcome of the continuous collaborations of the two co-authors, V. Roncone in particular edited part of the literature review, organized the database and elaborated part of the data, whilst F. Pizzutilo is responsible for the construction of the methodology, the analysis of the results and the rest of the paper.

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Fabio Pizzutilo http://orcid.org/0000-0001-9828-6374

Valeria Roncone http://orcid.org/0000-0002-1480-331X

Notes

1 Note that some of these weather variables (e.g. atmospheric pressure) are not directly observable by people, thereby possibly making their supposed effect on individuals’ mood more subliminal.

2 A total of 240 companies were listed on the domestic segment of the Milan stock exchange's telematic market (MTA) at the end of the study period.

3 For the ADF test, the general regression equation, which incorporates a constant and a linear trend, was used. The number of lags was made equal to the truncated value of ((n − 1)^(1/3)). For the KPSS test, the Newey–West estimator has been used to approximate σ². The lag parameter has been set at the truncated value of ((3*(n)^(1/2))/13). P-values have been interpolated from Kwiatkowski et al. (1992).

4 Note that, in contrast to all the other weather variables used, barometric pressure is the same indoors and outdoors. Hence, it is the only weather variable to which traders are exposed directly all the time.

5 By their nature, humidity, pressure and visibility assume positive values only. Hence, for models (1) and (3) we inspected if their logarithm or exponential transformation would provide a better fit to the data. Wind speed can assume null or positive values; thus, for this variable we tested the exponential transformation only. A comparison based on the adjusted R-squared, the Akaike information criterion (AIC) and the F-statistic leads to the conclusion that no additional information is added by any of these transformations. To avoid redundancy, the regression results of the transformed weather variables are omitted here, but are available upon request.

6 N stays for north, E for east, S for south and W for west.

7 Okta is the unit of measurement for total sky coverage. It refers to the number of eighths of the sky covered by clouds. The scale of measurement ranges from 0 oktas (completely clear sky) to 8 oktas (completely overcast).

8 We avoided testing for very deep snow cover as this extreme event was limited to a very small number of hourly intervals (72 out of 19,810 hourly intervals with a cover of snow recorded equal to or greater than 10 cm.). As with other explanatory variables (cf. note 3), we tested if an exponential transformation of RAIN and SNOW would provide a better fit to the data for models (10) and (12). No additional information was added by these transformations. The regression results shown in the next section refer to untransformed variables. Those interested in the other set of results are welcome to contact the authors.

9 Those interested in the other set of results are welcome to contact the authors.

10 Those interested in inspecting the results of the de-trended explanatory variables, the logistic regression models or different confidence levels are welcome to contact the authors.

11 The complete data are available to interested readers on request from the authors.

12 The very high correlation between the indices explains the similarities in the relationships which show significance.

13 Moreover, psychological theory (see Forgas 1995, among others) suggests that the impact of mood on the decision-making process is stronger the riskier, the more uncertain and the more abstract the situation. Hence, the weather effect should be stronger at the stock level than in the analysis of indices.