Time-varying financial spillovers from the US to frontier markets

Abstract

We examine US stock index return and volatility spillovers on the mean and volatility of stock index returns of 21 Frontier markets. We entertain potential time variation in spillovers in mean returns by considering a time-varying parameter (TVP) model. Spillovers in volatility are modelled by augmenting a standard GARCH (1, 1) model with current and one-period lagged US conditional volatility effects. The resulting model can be cast in state space form. However, it is not time invariant as the ‘coefficient’ multiplying the state variable (the TVP parameter) is current period US returns. The model is estimated by the Kalman filter. Several important hypotheses of interest are tested using a variety of restricted versions of the general model. An important contribution of the paper is a detailed analysis of the relative contributions from US and own-country lagged effects on both the mean and the volatility of returns in Frontier countries. In summary, our TVP model detects statistically significant time variation in return spillovers, and statistically and quantitatively important volatility spillovers for most Frontier markets. However, the model captures only a small portion of their daily return fluctuations. Most Frontier markets display volatility that is greater both in magnitude and variability relative to the US. Time-varying and quantitatively significant spillovers from the US are important in 13 of the 21 Frontier countries. Quantitative or statistically significant impact of US conditional volatility is found for at least 14 markets. Our results strongly reject the polar null hypotheses of complete market segmentation or complete market integration. Thus, Frontier markets are characterized as neither completely segmented from the US nor completely integrated with it. Results further suggest possible orthogonality in the contributions of current US and lagged own-country returns on Frontier countries’ mean returns.

Keywords:

• frontier markets
• stock index returns
• return and volatility spillovers
• time-varying parameters (TVP)
• time-varying volatility

JEL classification:

• F36
• G15
• C58

Notes

1. The World Bank defines emerging markets as markets for which GDP per capita falls below a certain time-dependent hurdle. Frontier markets are defined as emerging markets that are investable, but have lower capitalization and market liquidity compared to the more developed emerging markets.

2. We also estimated homsoskedastic versions of both models 1 and 2. Parameter estimates for the homoscedastic version of model 1 suggest a time-varying parameter with no persistence. However, as noted above, a GARCH model (model 1) suggests no time variation in this parameter.

3. A test for homoscedasticity with model 1 as the alternative hypothesis was easily rejected. Also, a joint test for time invariance of all parameters in model 1 (joint test for homoscedasticity and no time variation in a^USt) once again was overwhelmingly rejected.

4. Plots in all figures in the paper are shown only for these same selected Frontier countries for the sake of brevity. Complete details are available from the authors on request.

5. Formal variance decompositions are made in studies such as Rockinger and Urga (2001) and Baele (2005).

6. Here, we are using the phrase ‘complete market integration’ in a different sense than commonly used in the literature, as in Bekaert and Harvey (1995). In the latter studies, complete market integration would be tantamount to replacing on the rhs of the first equation in the main text immediately following this footnote with the returns on a world market portfolio consisting of stocks traded in the US and all the Frontier countries.