Abstract
Using data from three Central and Eastern European (CEE-3) and two developed stock markets, we present a methodology for validating the existence of shift contagion between these markets. The use of endogenously detected changes in the volatility of stock market returns allows us to define relatively high- and low-volatility regimes for particular stock markets. We verify whether volatility regimes are significantly associated with dynamic conditional correlations (DCCs), thus providing evidence for contagion between stock markets.
Acknowledgement
The paper was prepared within the submitted VEGA project no. 1/0826/11.
Notes
1Such models are naturally estimated for different reasons. For example, Savva et al. (Citation2009) explain the changes in DCCs after the adoption of the Euro. Andersson et al. (Citation2008) analyse DCCs for stocks and bonds with regard to inflation changes.
2We will refer to the test statistics of Inclán and Tiao (Citation1994) as IT test.
3More days were unnecessary; this ensured that there were no missing observations.
4We used the Newey–West lag selection criteria for the Phillips–Perron test, the BIC lag selection criteria for the ADF-GLS test and the procedure described by Hobijn et al. (Citation1998) for the KPSS test. Detailed results are available upon request.
5The methodological details are omitted for the sake of brevity. The entire procedure was conducted using R, source code is available upon request.
6In our case, GARCH (1,1) models were sufficient to capture all ARCH effects; hence, the p and q in EquationEquation 3 are set to 1. The usual GARCH restrictions for non-negativity and stationarity are imposed.
7Matlab code for estimating the DCC MV-GARCH model (and the time-varying correlation test) is provided by Kevin Sheppard at http://www.kevinsheppard.com/wiki/UCSD_GARCH.
8For PX-WIG, it was on 27 October 2008; for PX-BUX, the date was 23 February 2009; and surprisingly, for BUX-S&P 500, the date was 29 July 2002.