Re-examining covariance risk dynamics in international stock markets using quantile regression analysis

Research output: Contribution to journalArticlepeer-review

8 Citations (Scopus)

Abstract

This investigation is one of the first to adopt quantile regression (QR) technique to examine covariance risk dynamics in international stock markets. Feasibility of the proposed model is demonstrated in G7 stock markets. Additionally, two conventional random-coefficient frameworks, including time-varying betas derived from GARCH models and state-varying betas implied by Markov-switching models, are employed and subjected to comparative analysis. The empirical findings of this work are consistent with the following notions. First, the beta smile (beta skew) curve for the Italian, U.S. and U.K. (Canadian, French and German) markets. That is, covariance risk among global stock markets in extremely bull and/or bear market states is significantly higher than in stable periods. Additionally, the Japanese market provides a special case, and its beta estimate at extremely bust state is significantly lower, not higher than that at the middle region. Second, the quantile-varying betas are identified as possessing two key advantages. Specifically, the comparison of the system with quantile-varying betas against that with time-varying betas implied by GARCH models provides meaningful implications for correlation-volatility relationship among international stock markets. Furthermore, the quantile-varying beta design in this study relaxes a simple dual beta setting implied by Markov-switching models of Ramchand - Susmel (1998) and can identify dynamics of asymmetry in betas.

Original languageEnglish
Pages (from-to)33-59
Number of pages27
JournalActa Oeconomica
Volume61
Issue number1
DOIs
Publication statusPublished - 2011 Mar 1

All Science Journal Classification (ASJC) codes

  • Economics and Econometrics

Fingerprint Dive into the research topics of 'Re-examining covariance risk dynamics in international stock markets using quantile regression analysis'. Together they form a unique fingerprint.

Cite this