Linkages between Indian and US financial markets: impact of global financial crisis and Eurozone debt crisis

Abstract

This paper examines inter-linkages between Indian and US equity, foreign exchange and money markets using the vector autoregressive-multivariate GARCH-BEKK framework. We investigate the impact of global financial crisis (GFC) and Eurozone debt crisis (EZDC) on the conditional volatility and conditional correlation estimates derived from the multivariate GARCH model for Indian and US financial markets. Our results indicate that there is significant bidirectional causality-in-mean between the Indian stock market returns and the Rs./USD market returns, and significant unidirectional causality-in-mean from the US stock market returns to the Indian stock market returns. As regards volatility spillovers, we find that volatility in the Indian stock market rises in response to domestic as well as US financial market shocks but Indian financial market shocks do not impact the US markets. Further, impact of the recent crisis episodes on the covariance matrix is found to be significant. We find that volatility in the Indian and US financial markets significantly amplified during GFC. The conditional correlations across asset markets were significantly accentuated in the wake of the two crisis episodes. The impact of GFC on cross-market conditional correlations is higher for majority of the asset market pairs in comparison to the EZDC.

Keywords:

• Emerging markets
• spillovers
• financial markets
• multivariate GARCH-BEKK
• global financial crisis
• Eurozone debt crisis

Acknowledgement

The authors are grateful to the anonymous reviewer for the useful comments and suggestions that helped in improving the paper.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. See Solnik (1974) and Adler and Dumas (1983).

2. However, in another study, Hakim and McAleer (2010) test for both return and volatility spillovers (i.e. causality-in-mean as well as causality-in-variance) simultaneously across a set of developed country financial markets.

3. Mean spillovers or return spillovers are synonymous with causality-in-mean and volatility spillovers are the same as causality-in-variance.

4. The curse of dimensionality in estimation constrains our ability to analyse beyond a two-country setup.

5. However, it is pertinent to note that since the data analysis is at weekly frequency, we are restricted in our choice of exogenous variables.

6. We use a simplified VAR-multivariate GARCH model without the dummy or the exogenous variables to derive the conditional variances and covariances, while testing for the impact of crises in order to avoid any possible biases from inclusion of these variables.

7. In particular, the endogenous variables have been specified in returns form (log first differences for stock and currency markets and first differences for Treasury bill rates) and, therefore, we conduct the unit root tests for the variables in this form only. However, the exogenous variables, viz. effective federal funds rate, long-term US Treasury yield rate, long-term Indian government securities yield rate, reverse repo rate, US economic uncertainty index and FSI have been first tested in levels and, thereafter, in first differences.

8. Lee and Strazicich (2003) propose the test with breaks in the level and trend under the null hypothesis which conclusively implies trend stationarity under the alternative hypothesis.

9. If a majority of two out of three unit root tests imply a non-stationary series then the series is treated as I(1). Results are available with the authors on request.

10. Further testing of first differences of the exogenous variables (except UNCI) corroborates that all these variables are integrated of order one.

11. Results for the VAR lag selection and exclusion tests have not been reported for the sake of brevity.

12. The p-value was 0.00 and the statistic was 1590.84.

13. The statistic was 23.02 and was significant at 5% level.

14. Exchange rate is defined as the rupees (Rs.) per one dollar (USD) value and, therefore, a rise in the exchange rate signifies a depreciation of the Indian rupee (and an appreciation of the USD).

15. In accordance with the theoretical analysis of financial markets and the existing empirical literature, the series for exchange rate (Rs. vs. USD), S&P 500 and Nifty 50 are modelled as the logarithmic first differences, while those for the Indian and US Treasury bill rates are defined as first differences.

16. A host of empirical specifications with various exogenous variables are attempted and the model with the best-fit is selected. Other exogenous variables that we tried included are policy rates CRR, SLR, repo rate, bank rate and reserve money in the United States and India, economic policy uncertainty index for India, returns on Yen/USD, returns on gold, returns on crude oil, yield spreads in US and India, equity uncertainty index and VIX.

17. Economic Policy Uncertainty Index for the United States has been constructed by Baker, Bloom, and Davis (2013) and is available from the Federal Reserve Board of St. Louis.

18. The Federal Reserve Board of St. Louis’ FSI is constructed using the first principal component of 18 weekly series comprising financial variables pertaining to developed markets (mainly United States), interest rates and yield spreads for developed and emerging markets (J.P. Morgan Bond Index Plus), and other indicators related to global financial markets. It, therefore, captures financial stress in both the stock markets appropriately.

19. The Dynamic Conditional Correlation model proposed by Engle (2002) cannot be utilized to test for causality-in-variance as it focuses on estimating the time-varying conditional correlations.

20. The BEKK model was named after the original contributors Baba et al. (1989).

21. Empirical distributions of financial market returns have been known to have fat tails and so we assume a heavy tailed distribution such as the Student-t distribution, instead of the usual Gaussian distribution.

22. Engle and Kroner (1995) have considered exogenous regressors in their original formulation and deem all the proofs would be the same in the presence of regressors, such as a dummy variable introduced in (2), in the variance–covariance equations.

23. A variable is causal-in-variance (causal-in-mean) for a variable if the conditional volatility (mean) of can be forecasted better by taking into account the present and past information in .

24. Volatility plots and conditional correlation plots are available with the authors on request.

25. According to Engle and Kroner (1995), the unconditional covariance in the case of a BEKK-GARCH model is given by . Contrary to univariate GARCH models, covariance-stationarity in the BEKK model is guaranteed if mod , i.e. the eigenvalues of should be less than 1 in modulus. Some of the eigenvalues that we obtain are above unity and therefore, the unconditional covariance does not exist. However, Nelson (1990), and Lumsdaine (1991) point out that even if a GARCH model is not covariance stationary, it is strictly stationary or ergodic and the standard asymptotically based inference procedures are generally valid.

26. b_ij measures the degree of lagged and cross innovation spillovers from market i to market j.

27. g_ij signifies the persistence of conditional volatility (or volatility persistence) between market i and market j.

28. We did not find significant spillovers in return between the US money market and US stock market. Ehrmann, Fratzscher, and Rigobon (2011) find no spillovers among these markets from their structural model and significant but negligible spillovers from the reduced form estimates. The two parity conditions – Uncovered Equity Parity proposed by Hau and Rey (2006) and Uncovered Interest Parity were not found to hold as well.

29. While inclusion of the variable for capital flows may have enriched the analysis, data for the same is not available at weekly frequency.

30. This has been done to avoid any possible biases in the subsequent tests resulting from inclusion of the exogenous variables in the VAR(1)-multivariate GARCH (1,1)-BEKK specification.

31. According to Pavlova and Rigobon (2007), the bond markets are also likely to be affected since a negative output supply shock in the United States (or EZ) leads to higher bond prices in the domestic economy but a fall in the bond prices of the foreign Economy.

32. In accordance with the timelines provided by the Federal Reserve Board of St. Louis, the GFC is defined from 12 September 2008 marking the filing of bankruptcy by Lehman Brothers on 15 September 2008 to 13 March 2009 since the Federal Reserve Board announced term asset-backed securities loan facility from 19 March 2009. The EZDC is defined as per the timeline provided by the European Central Bank and the Guardian from 30 April 2010 when Greece seeks financial support to 9 July 2010 when Hungary asks for precautionary bailout from the European Union (EU) and International Monetary Fund (IMF), 26 November 2010 as Ireland is bailed out by EU and IMF to 17 December 2010 as European Stability Mechanism is established, and 6 May 2011 when Portugal, Ireland and Greece are bailed out once again to 28 November 2011 when the major Central banks of the world extend coordinated emergency measures.

33. We did not formally test for contagion since Forbes and Rigobon (2002) have suggested that covariance-based models should not be utilized to measure and empirically test for the phenomenon of contagion.

Additional information

Notes on contributors

Pami Dua

Pami Dua is Dean Research (Humanities and Social Sciences), University of Delhi as well as Director and Professor of Economics, Delhi School of Economics. She is currently serving as the President of the Indian Econometric Society. She received her Ph.D. in economics from London School of Economics. Her research interests include time series econometrics, forecasting, macroeconomics and business cycle analysis. She has published widely in various reputed international journals and has authored several books and book chapters.

Divya Tuteja

Divya Tuteja is Assistant Professor (Ad-hoc) at the Delhi School of Economics, University of Delhi since 2013. She received her doctoral degree from the Delhi School of Economics in 2015. Her research interests include macroeconomics, econometrics and international financial markets. She has also worked as a Manager in a reputed Indian multinational bank.