اختلال عاطفی فصلی: زمان شروع و بازیابی

Kamstra et al., 2003, Kamstra et al., 2009 and Kamstra et al., 2012 offer a seasonal affective disorder hypothesis to explain variations in the daily returns of stock indices. We examine Kamstra et al. (2012) new variable called SAD onset/recovery. The analysis reveals concerns for the validity of the SAD hypothesis.

Kamstra et al. (2003) advance the hypothesis that the prevalence of the seasonal affective disorder (SAD) has a systematic influence on stock index prices. There is widespread interest in this hypothesis – a Google Scholar search (21 June 2012 – the winter solstice in the southern hemisphere) reveals that this study is cited in more than 280 papers. We examine the revised empirical version of the Kamstra et al. (2003)SAD hypothesis. The focus is on the new variable called SAD onset/recovery. This variable is used to replace the two variables of their original model. This original model has attracted criticism in the literature. We examine six statistical issues of the new variable – these have not been previously critically evaluated in the literature. Each issue has a profound implication for the SAD hypothesis. The SAD hypothesis is intuitively appealing. As the fall season evolves, there is an increase in the number of people suffering from the depressive illness called the seasonal affective disorder. This increase in depression, which creates a heightening in risk aversion, leads to a reduction in stock index prices. The winter solstice brings about a reversal in the mental processes of sufferers. The depression, and associated risk aversion, weakens and there is a corresponding increase in stock index prices. Asymmetry is a crucial feature of the hypothesis. The length of night is the “instrument most closely tied by clinical research” to the symptoms of the seasonal affective disorder ( Kamstra et al., 2012, p. 943). The original model, based on daily data with the control variables discarded for simplicity, is equation(1) View the MathML source

Following criticisms in the literature, Kamstra et al., 2009 and Kamstra et al., 2012 offer new and improved measures of the seasonal affective disorder. Even a casual reading reveals Kamstra et al., 2009 and Kamstra et al., 2012 deprecate their two-variable 2003 model and at the very same time vigorously defend it. Kamstra et al. (2012, p. 955) conclude with the words “... we believe the best estimate of the influence of SAD on equity returns around the world is the SAD onset/recovery variable based on data from Lam (1998), perhaps scaled by latitude for application to markets from outside of North America”. Our analysis of six issues with the SAD onset/recovery variable does not support this view. These inconvenient issues raise the spectre of profound implications for the validity of the Kamstra, Kramer and Levi SAD hypothesis. Our careful review leaves us unshaken in this depressing conclusion. Kamstra et al. (2012, p. 949) accuse Kelly and Meschke (2010) of attempting to stifle legitimate scientific enquiry. We hope they do not level the same criticism at us – our express goal is to explore the claimed statistical properties of the SAD onset/recovery variable. Lest Kamstra, Kramer and Levi seek to rest their entire case on Lam (1998) and the scaling factor, there is one paramount issue for them to address. The answer is necessary for the sake of any researcher attempting to replicate their SAD incidence variable, or the SAD onset/recovery variable, with clinical evidence from other countries. Why does the brute empiricism of their ‘econometric variable building process’ ( Kamstra et al., 2011, footnote 18, p. 8) create such bizarre results? The process converts the monthly clinical data of Lam (1998) into a daily SAD incidence variable that almost exactly mimics the – currently out of favour – length of night variable SADt! This issue deserves a careful re-examination.