Abstract
In this note, we use several modern multiple variance ratio tests (VR tests) to investigate whether the financial crisis has an impact on the random walk behaviour of international stock markets. Grouping a pre-crisis- and a crisis-panel in developed, emerging and frontier markets, respectively, and using daily returns of selected Morgan Stanley Capital International (MSCI) International Equity Indices we find that markets classified as developed or emerging mostly follow a random walk whereas we find the opposite for frontier markets. Frontier markets show a higher proportion of countries that experience changes in the random walk behaviour and changes from random walk to nonrandom walk are more frequent. We also find that the choice of multiple VR test does not matter for this kind of analysis.
Acknowledgement
We thank Frank Schuhmacher for helpful suggestions and discussions.
Notes
1 Lo and MacKinlay (Citation1988) showed that compared with Dickey–Fuller unit root and Box–Pierce serial correlation tests, VR tests are more powerful under the heteroscedastic random walk.
2 See Hoque et al. (Citation2007) and Charles and Darné (Citation2009) for a review of individual and multiple VR tests.
3 Because of space considerations the summary statistics and the detailed results of the multiple VR tests are not reported here. They are available from the authors upon request.
4 As advocated by Deo and Richardson (Citation2003), we use relatively short holding periods when testing with daily returns.
5 We also considered the two-point distribution of Mammen (Citation1993) and the Rademacher distribution discussed in Davidson and Flachaire (Citation2008). We confirm that the choice of the specific distributional form does not qualitatively influence the results. The following interpretations are therefore based on the standard normal distribution only.