ABSTRACT
We examine the impact of the Indian cricket team's performance in one-day international cricket matches on return, realized volatility and jumps of the Indian stock market, based on intraday data covering the period of 30th October, 2006 to 31st March, 2017. Using a nonparametric causality-in-quantiles test, we were able to detect evidence of predictability from wins or losses for primarily volatility and jumps, especially over the lower-quantiles of the conditional distributions, with losses having stronger predictability than wins. However, the impact on the stock return is weak and restricted towards the upper end of the conditional distribution.
Acknowledgement
We would like to thank the Associate Editor and an anonymous referee for many helpful comments. However, any remaining errors are solely ours.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 Michael A. Atherton is a broadcaster, journalist and a former England international first-class cricketer. And the above quote comes from one of his article on the importance of cricket in India, available at: https://www.thetimes.co.uk/article/games-soul-is-not-at-lords-it-is-here-7sq85l0rd.
2 We would like to thank an anonymous referee for pointing this out to us. Given the concerns with the narrowness of the BSE200 index, we conducted some analysis based on the NSE index, with data on it again derived from Datastream. However, since intraday data is not available for the index, we used daily data, and applied the k-th order version of the causality in quantiles test developed by Balcilar et al. [Citation9]. This test allows us to check for causality in returns and volatility (captured by squared returns) within a nonparametric quantiles-based framework. But note that, without intraday data on the NSE, we cannot compute realized volatility and volatility jumps, and are only restricted to returns and squared returns, i.e. volatility. Based on this test, we found similar results to those obtained under the BSE. Specifically, we found that ODI match results affects the returns weakly and primarily at the upper quantiles, while strong impact on squared returns (volatility) is restricted at the lower end of the conditional distribution. Complete details of these results are available upon request from the authors.
3 The win and loss dummies were divided by their full-sample standard deviation, estimated or calculated based on the basic formula of standard deviation of a variable. By doing this, we removed the inherent variability of these dummies, and thus both of them have a standard deviation equal to 1. This allows us to compare the strength of the impact of win or loss on the same dependent variable across the two dummies.
4 By construction the values of is equal zero or almost equal to zero in its lower quantiles i.e. 0.05–0.20. Thus, it is expected that the impact of India's performance in ODI matches on in lower quantiles to be statistically significant equal to zero. This happens due to the fact that there is no asymmetry between downside and upside jumps.
5 When processing with our intraday data, we have applied the common practice of eliminating from the original sample (i.e. the sample before removing any day) those days with infrequent trades (less than 30 percent of the expected observations on operating day at 5-minute frequency) (see [Citation31,Citation25] among others). The procedure applied here yielding a final sample of about 100 observations per day on average.
6 Following Corsi et al. [Citation22], the local variance – which is estimated with a non-parametric filter of length which is iterated in for – has a starting value of equal to . The is the bandwidth parameter which determines the number of adjacent intraday returns included in the estimation of the . In other words, this corresponds to using all available intraday observations in the first step. For the first step, the variable is set to be equal to 3. However, in each step (iteration) large intraday returns are excluded by a condition. Finally, by multiplying each variance estimate with , the threshold for the following step is obtained. The procedure stops after two or three iterations using intraday data (three in our case) (see the Appendix B in Corsi et al. [Citation22] for further discussion).