臺大管理論叢 NTU Management Review VOL.30 NO.3

57 NTU Management Review Vol. 30 No. 3 Dec. 2020 5. Discussion 5.1 Robustness Tests The weather effects on the stock returns in Table 2 are based on an annually performed test. In order to ensure that the results are not driven by the annual-test setting, I repeat the tests based on a triennial setting and report the results in Table 4. Because the full sample is 26 years, two settings in which the last and the first subsamples are 2 years were considered. The results for the two settings are consistent with each other and with Table 2. The cumulative responses are significant, and there are no reversals of the impulse responses. Jacobsen and Marquering (2008) find that for 48 national stock markets, temperature and seasonal affective disorder effects disappear once the tests are controlled for other variables with a strong seasonal pattern. I check whether or not the weather effects in this study are spurious due to those other seasonal variables. I add January and Monday dummies as control variables in equation (1). In previous weather studies, the researchers, e.g., Jacobsen and Marquering (2008) and Kang, Jiang, Lee, and Yoon (2010), add these dummies in their regressions. The tests are performed annually. The result is reported in the last rows of Table 4. They are consistent with the result in Table 2. There is no reversal. The permanent weather effects found in this study are not spurious. 5.2 The Weather’s First Principal Component Although the first PC best explains the common variation of the weather variables, I choose the fourth PC in the analysis because it was the fourth PC that affects the return (Khanthavit, 2018). To ensure that the first PC is unable to reveal the time behaviors of weather effects, I re-estimate the model in equation (1) using the first PC for W t . In the full sample test, neither impulse nor cumulative responses are significant. 5.3 Nonsignificant Temporary Weather Effects I found no price reversal, so the significant weather effects are not temporary but permanent. The temporary effects are the net effects from the price-pressure effects against the hold-more effects (Lee et al., 2002). Therefore, the insignificance does not necessarily imply no price-pressure and hold-more effects. The two effects might be significant but would cancel each other out.