The Dynamic Extreme Co-Movement between Chinese Stock Market and Global Stock Markets

ABSTRACT

We use time-varying Symmetrized Joe-Clayton Copula model to study the extreme co-movement (boom or crash together) between the Chinese stock market and major stock markets in the world from 2007 to 2017, including developed markets and stock markets on “Belt and Road Initiative” (hereafter B.R.I.). We find that the extreme co-movement probability between Chinese market and “Belt and Road Initiative” markets is higher than developed markets at both tails. Then we study important “real” and “non-fundamental” factors affecting the excess co-movement probability, including bilateral trade openness, financial integration, and economic policy uncertainty. The results of panel regression analysis show that: the bilateral financial integration has significant effects over the lower tail dependence between Chinese and developed markets, but does not affect the extreme co-movement between Chinese and B.R.I. markets. And the bilateral trade openness is an important factor for the extreme co-movement at both tail between Chinese and global markets. The economic policy uncertainty index, especially China’s economic policy uncertainty, plays a key role in the extreme co-movement between Chinese and developed markets at both tails. However, it has sizable effects only at the upper tail co-movement between Chinese and B.R.I. markets.

KEY WORDS:

• economic policy uncertainty
• extreme co-movement
• financial integration
• time-varying copula
• trade openness

JEL CLASSIFICATION:

• C51
• C52
• C53
• Q41
• H71

Notes

1. It was initially called One Belt and One Road, but in mid-2016 the official English name was changed to the Belt and Road Initiative due to misinterpretations of the term one.

2. We consider global stock markets, including: Australia (AORD), France (FCHI), United Kingdom (FTSE100), Germany (DAX), Japan (N225), China (HUSHEN300), India (SENSEX30), Brazil (IBOVESPA), Spain (IBEX35), Ireland (ISEQ), South Korea (KOSPI), Sweden (OMXSPI), Singapore (STI), Italy (ITLMS), Canada (GSPTSE), Russia (RTS), Netherland (AEX), Indonesia (JKSE), Israel (TA100), Malaysia (KLSE), Mexico (MXX), Philippines (PSEI), South Africa (EZA), Thailand (SET), Ukraine (MSCI).

3. The copula framework, as a typical non-linear method measuring the interdependence cross financial markets, reveals a complete description about their tail dependence structure. It allows us to calculate the probability of simultaneous extreme losses or gains directly and has been used extensively in the literature (Hu (2006), Rodriguez (2007), Patton (2006), Dias and Embrechts (2010),Aloui, Aissa, and Nguyen (2013)).

4. In our analysis, the emerging markets include South Korea (KOSPI), India (SENSEX30), Russia (RTS), Indonesia (JKSE), Malaysia (KLSE), Philippines (PSEI), South Africa (EZA), Thailand (SET), Ukraine (MSCI), Singapore (STI), Israel (TA100). All the other countries in our sample are classified as developed markets.

5. $F_{X_1|W}(x_1|w)$ and $F_{X_2|W}(x_2|w)$ are continuous in $x_1$ and $x_2$.

6. An alternative to this approach may be to allow also for time variation in the functional form using a regime switching copula model, as in Rodriguez (2007), for example.

7. Data source: Wind Database. Daily returns are also adjusted for weekends and holidays.

8. We choose AR(2)-EGARCH(1,1) model, which is defined as

(6) $x_t={\mu }_0+{\mu }_1{x}_{t-1}+{\mu }_2{x}_{t-2}+{\sigma }_t{\epsilon }_t$(6)

(7) $ln{\sigma }_t^2=\omega +{\alpha }_{1}ln{\sigma }_{t-1}^2+{\beta }_1\frac{\left|z_{t-1}\right|+{\gamma }_1{z}_{t-1}}{{\sigma }_{t-1}}$(7)

The model is estimated using the Quasi-Maximum Likelihood estimation method.

9. We chose the distribution with the lowest AIC value.

10. Results of the Panel Granger causality tests show that $E\bar{P}{U}_{\mathrm{i}\mathrm{t}}$, $T{O}_{\mathrm{i}\mathrm{t}}$, $F{I}_{\mathrm{i}\mathrm{t}}$ granger cause ${\bar{\tau }}_{\mathrm{i}\mathrm{t}}$, while the inverse does not hold true, which implies that the reverse causality problem will not cause too much bias in our estimation results.

11. ${\tau }_{\mathrm{i}\mathrm{t}}^U=\frac{\sum _{j\mathrm{i}\mathrm{n}\mathrm{m}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{h}t}{\tau }_{i,jt}^U}{\mathrm{N}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{d}\mathrm{a}\mathrm{y}\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{m}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{h}t}$, and ${\tau }_{\mathrm{i}\mathrm{t}}^L=\frac{\sum _{j\mathrm{i}\mathrm{n}\mathrm{m}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{h}t}{\tau }_{i,jt}^L}{\mathrm{N}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{d}\mathrm{a}\mathrm{y}\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{m}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{h}t}$. We also tried other specifications for ${\tau }_{\mathrm{i}\mathrm{t}}^U$ and ${\tau }_{\mathrm{i}\mathrm{t}}^L$ in the robustness check, including the maximum/minimum/mid-day of the daily tail dependence probabilities from the Copula model in that month.

12. Data source: the monthly bilateral trade data between China and other countries is from Comtrade.

13. This measure increases as bilateral trade becomes more important relative to GDP. In other terms, the indicator increases if bilateral trade grows faster than GDP.

14. Data Source: CEIC.

15. We use the monthly EPU data from: www.PolicyUncertainty.com and Baker, Bloom, and Davis (2016).

16. Data source: Crude Oil Prices: West Texas Intermediate (WTI)––Cushing, Oklahoma. We also use the Crude Oil Prices: Brent - Europe as robustness check.

17. $F{I}_{\mathrm{i}\mathrm{t}}$ measures the absolute difference of the credit market returns between country $i$ and China. Therefore, a higher value for $F{I}_{\mathrm{i}\mathrm{t}}$ could imply a lower extent of financial integration between two countries.

18. In our analysis, according to the classification of International Monetary Fund the developed markets include Australia, Canada, France, Germany, Ireland, Italy, Japan, Korea, Netherland, Singapore, Spain, Sweden, UK, USA, Israel.