Investigation of the effect of global EPU spillovers on country-level stock market idiosyncratic volatility

ABSTRACT

Using the multivariate quantile model, this paper develops a global economic policy uncertainty (EPU) spillover measure for each country and investigates the spillover effects on the country-level stock market idiosyncratic volatility across a sample of 23 economies. The regression results show that global EPU spillovers have a positive and significant effect on the country-level stock market idiosyncratic volatility. We find that the effect of developed-market-generated EPU spillovers on the country-level stock market idiosyncratic risk is noticeably larger compared to the effect of emerging-market-generated EPU spillovers. Furthermore, the significant and positive effect of the EPU spillovers on the country-level stock market idiosyncratic volatility continues when we utilize various economic, financial, and political risk factors as controls, as well as when we use alternative measures of stock market idiosyncratic volatility as the dependent variable in our regression analyses.

KEYWORDS:

• EPU spillovers
• country-level stock market idiosyncratic volatility
• multivariate quantile model
• international asset pricing

JEL CLASSIFICATIONS:

• C10
• F30
• G12
• G15

Acknowledgements

We gratefully acknowledge the financial support from the National Natural Science Foundation of China (grant number 71971133), the National Social Science Foundation of China (grant number 21BGL270) and the Shanghai Science and Technology Committee (grant number 23692111400)

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 The GDP weights utilized in this analysis are from quarterly GDP data and time-varying over the sample period. Specifically, we collect the real GDP data of our sample countries from the IMF’s World Economic Outlook Database on a quarterly basis, and the GDP weights utilized in this calculation are adjusted on a quarterly basis.

2 In our analyses we use the first difference of the EPUs in our regression models since the original EPUs are not stationary and do not meet the requirement of the multivariate quantile model (see White, Kim, and Manganelli 2015). In addition, when we look at the distribution of the change in EPUs across 23 countries (see Figure B1 in Online Appendix B), we find that the kurtosis is 7.0546, which is evidence of heavy tails in the distribution. Furthermore, the Jarque–Bera test confirms that the change in EPU is not normally distributed. These findings support the necessity to use the multivariate quantile model in analyzing global EPU spillovers.

3 For each country i and the other N-1 countries, in each rolling procedure, 36 monthly observations are used for different quantiles. As a robustness check, we have also utilized the 48- and 60-month rolling windows and obtained results similar to the ones generated from a 36-month rolling window approach. These results using the 48- and 60-month rolling window regressions are available upon request and suggest that our results are not dependent on the 36-month rolling window technique utilized in this paper.

4 A detailed explanation of the multivariate quantile model (MVMQ) of White, Kim, and Manganelli (2015) is provided in Online Appendix C.

5 Following Caglayan, Xue, and Zhang (2020), we apply the returns of the MSCI ACWI IMI Index minus US one-month T-bill returns to build the global market risk premium R_m-R_f, the returns of the MSCI ACWI Small Cap Index minus the returns of the MSCI ACWI Large Cap Index to build the global SMB factor, and the average returns of the MSCI ACWI Large Cap Value and Small Cap Value Indices minus the average returns of the MSCI ACWI Large Cap Growth and Small Cap Growth Indices to build the global HML factor. The units of the MSCI ACWI IMI Index, the other MSCI ACWI Indices, and the Country IMI Indices are the US dollar, where ACWI stands for All Country World Index.

6 As an alternative, Brandt et al. (2010) and Bekaert, Hodrick, and Zhang (2012) use the individual stocks and the Fama-French three-factor model to obtain the errors and use the standard deviation to calculate the aggregate stock market idiosyncratic volatility. This methodology, however, contains both the firm-level risk and the country-level risk. In contrast, our macro study applies the Fama-French three-factor model to stock market index returns at the country level, which eliminates the firm-level risk and preserves the country-level risk.

7 As explained before, we use the first difference of the EPU in estimating the EPU spillovers with the multivariate quantile model (see White, Kim, and Manganelli 2015). Correspondingly, for consistency, we also use the change in domestic EPU (the first difference of domestic EPUs) in the regression model as well.

8 We also use the law and order and the investment profiles in International Country Risk Guide (ICRG) as alternative political risk factors in our analyses. For all political risk factors tested, we find a significant and negative relation between the political risk factors and the stock market idiosyncratic volatility, which confirms that improvements in political risk significantly decrease the country-specific stock market idiosyncratic risk.

9 We collect the EPUs of all 23 markets from www.policyuncertainty.com. The developed markets include Australia, Canada, France, Germany, Greece, Hong Kong, Ireland, Italy, Japan, the Netherlands, Singapore, Spain, Sweden, the United Kingdom, and the United States. The emerging markets include Brazil, Chile, China, Colombia, India, Mexico, Russia, and South Korea.

10 The detailed methodology of the quantile impulse-response function in the framework of MVMQ of White, Kim, and Manganelli (2015) is provided in Online Appendix C.

11 This result is consistent with Colombo (2013), Mumtaz and Theodoridis (2015), and Biljanovska, Grigoli, and Hengge (2021) who show that foreign-economy-generated EPU spillovers have a larger and stronger impact on the domestic macroeconomic aggregates than the domestic EPU itself.

12 Specifically, in our sample of 23 countries, we have 15 of them as developed economies and eight of them as emerging-market economies. In Panel A, to measure the EPU spillovers generated from developed markets to a country i at time t, we first calculate the GDP weighted first difference of the EPUs coming from the 15 developed markets at time t-1. If country i is a developed economy, however, we calculate the GDP-weighted first difference of the EPUs coming from the other remaining 14 developed markets at time t-1. Then we estimate the bivariate MVMQ in Eq. (1) on a 36-month rolling-window basis and calculate the EPU spillovers generated from the developed economies for each of the 23 countries using the regression in Eq. (2). By repeating this procedure for each country each month, we obtain a separate time-series measure of the developed-markets-generated EPU spillovers for each of the 23 individual countries in our sample.

13 In order to measure the EPU spillovers generated from emerging markets to a country i at time t, we first calculate the GDP-weighted first difference of the EPUs coming from the eight emerging markets at time t-1. If country i is an emerging market economy, however, we calculate the GDP-weighted first difference of the EPUs coming from the other remaining seven emerging markets at time t-1. Then we estimate the bivariate MVMQ in Eq. (1) on a 36-month rolling-window basis and calculate the EPU spillovers generated from the emerging market economies for each of the 23 countries using the regression in Eq. (2). By repeating this procedure for each country each month, we obtain a separate time-series measure of the emerging-market-generated EPU spillovers for each of the 23 individual countries in our sample.

14 In a separate analysis, we use the multivariate quantile model to estimate separately the EPU spillovers generated from individual countries such as the US, the UK, China, and Brazil and test their impact on the stock market idiosyncratic volatility. In line with the findings reported in Table 5, the results show that the EPU spillovers generated from the US, the most developed economy, has the largest and the most significant effect on the stock market idiosyncratic volatility, followed by the EPU spillovers generated by the UK, China, and Brazil. The detailed results from this analysis are tabulated in Table A6 of Online Appendix A.

15 As a robustness test, we apply the multivariate quantile model to estimate the global EPU spillovers at the 50% quantile (as opposed to the 80% quantile) and use this measure in our panel regressions with interactions. We find that results do not change materially compared to the results reported in Table 7. We do not use the global EPU spillovers at the 20% quantile since this variable is not significant. The results from this additional analysis are available upon request.

16 To differentiate between the impact of economic growth on global EPU spillovers and the change in domestic EPU, as an additional test, we run separate regressions of the global EPU spillovers and the change in domestic EPU on economic growth. We obtain the residuals of the global EPU spillovers and the residuals of the change in domestic EPU variables and use them in Eq. (4) in our panel regressions. We once again find a positive and significant relationship between the residuals of the global EPU spillovers and the next-month stock market idiosyncratic volatility. The results from this additional analysis are available upon request.

17 We use daily returns of the MSCI Developed and the MSCI Emerging Markets Index and subtract the daily one-month US T-bill returns to construct the daily developed and emerging-market risk premiums (i.e., R_m-R_f). We use the daily returns of the MSCI Developed and the MSCI Emerging Markets Small Cap Index minus the daily returns of the MSCI Developed and the MSCI Emerging Markets Large Cap Index to generate the daily values of the developed and emerging-market SMB factors. We use the daily returns of the MSCI Developed and the MSCI Emerging Markets Value Index minus the daily returns of the MSCI Developed and the MSCI Emerging Markets Growth Index to generate the daily values of the developed and emerging-market HML factors.

18 Following Caglayan, Xue, and Zhang (2020), we calculate the country-level stock market idiosyncratic tail risk on a monthly basis by applying the CAPM and the Fama and French (1993) three-factor models in a global context by running the daily individual stock market return data on the global CAPM and the global Fama-French three-factor models respectively each month for each country. We then obtain the daily residuals for each of the two models respectively and then estimate the country-level stock market idiosyncratic tail risk by using the 5% VaR (Value at Risk) again for each of the two models. The Value at Risk is a conditional quantile of the return loss distribution.

19 In Table A3 of Online Appendix A, the global EPU spillover variable has statistical significance only at the 10% significance level when the global EPU spillover is estimated by the multivariate quantile model at the 50% quantile.

20 The results of these two additional analyses are available from the authors upon request.

21 The results of these two analyses that add the WUI and the stock market spillover as the control variables to the regression equations can be obtained from the authors upon request.

 

Additional information

Funding

This work was supported by National Natural Science Foundation of China: [Grant Number 71971133]; National Social Science Foundation of China: [Grant Number 21BGL270].

Notes on contributors

Mustafa O. Caglayan

Mustafa O. Caglayan is a Professor of Finance and Knight-Ridder Research Fellow at the College of Business in Florida International University. He holds a Ph.D. in Economics with a concentration in Finance at City University of New York. His research focuses on asset pricing, investments, hedge funds, financial risk management, and portfolio optimization.

Yuting Gong

Yuting Gong is an Associate Professor at the SHU-UTS SILC Business School in Shanghai University. She holds a Ph.D. in Finance at Shanghai Jiao Tong University. Her research interest includes financial econometrics and asset pricing.

Wenjun Xue

Wenjun Xue is an Associate Professor at the SHU-UTS SILC Business School in Shanghai University. He holds a Ph.D. in Economics at Florida International University. His research interest includes financial markets and institutions and financial econometrics.