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

Time Paths of Weather-Induced Mood Effects on Stock Returns 56 4. Empirical Results Before I proceed with the estimation, I conduct cumulative-sum (CUSUM) and CUSUM-square tests of recursive residuals for parameter stability of the model in equation (1). At the 95% confidence level, the CUSUM tests rejects the stability hypothesis for the fourth PC equation, while the CUSUM-square tests rejects the hypothesis for the stock return and PC equations. The stability-test results support Khanthavit’s (2017) approach for estimating the model using the sample of one year at a time. Table 2 reports the impulse responses of the return to one standard deviation shock of the weather’s fourth PC for 10 days from days 0 to 9. There are no significant price reversals in any year from 1992 to 2017. The cumulative responses over 250 days from days 0 to 249 are significant in 1996, 1999, 2008, and 2016. The average impulse and cumulative responses over the full sample are negative. For the full sample tests, the impulse response on day 0 and cumulative response are significant at the 99% and 95% confidence levels, respectively. These findings lead me to conclude that the weather effects in the Thai stock market exist and are permanent. There are no temporary weather effects on the returns. I conduct further analysis on whether the direct or indirect causes explain the permanent effects. Table 3 reports the impulse and cumulative impulses of the variance σ 2 t to the fourth PC from the model in equation (3) with X' t = [W t σ 2 t r t ]. If the indirect Friedman or create-space effects fully or partly cause the permanent weather effects, first, I must find that Θ j 2,1 for some j and lim J↑∞ Σ J j=0 Θ j 2,1 are significant. The full sample tests show that Θ j=0 2,1 and Σ J=249 j=0 Θ j 2,1 are significant at the 99% confidence level. The absolute correlation |ρ r,σ 2 | due to the weather shock is in the last column of Table 3. For the full sample test, |ρ r,σ 2 | is significant at the 99% confidence level. The statistic 1 – |ρ r,σ 2 | is not different from zero. The direct cause from the price formation process is not significant. I concluded that only the indirect cause explains the permanent weather effects.