Modeling Volatility and Risk Spillover Between the Financial Markets of US and China Using GARCH Value-at-Risk Forecasting and Granger Causality

Undergraduate thesis, Department of Economics, Seoul National University (June 2020). Read the thesis · Analysis notebook

Editorially revised 2026: byline, numeric presentation, and prose. Results, methods, figures, and findings are unchanged from the submitted version.

Keywords: VaR(Value at Risk), ARIMA-GARCH model, Risk management

Motivation

Comparative analysis of international economies during two periods of elevated volatility: the Great Recession of 2008 and the Coronavirus Recession. The chosen window captures both against the backdrop of a sustained bull market, allowing the two downturns to be compared directly rather than each against a quiet baseline.

Dataset

Intraday returns (January 2007 - April 2020)

• S&P500
• SSE Composite Index
• Chinese Yuan to USD exchange rate

Source: Yahoo Finance

Libraries

NumPy · Pandas · Statsmodels · SciPy · Seaborn · Matplotlib

Methodology

• Volatility Forecasting:
  • Skewed Student’s t ARIMA-GARCH model
    • Augmented Dickey-Fuller Test for Stationarity
    • Jarque-Bera Test of Normality
    • Box-Ljung Test of Autocorrelation
    • Breusch-Pagan Test for Heteroskedasticity
  • Parametric Value-at-Risk (VaR)
• Risk Spillover: Granger Causality

Results

The ARIMA-GARCH VaR estimates fit the historical series closely, with failure ratios meeting the well-specified threshold at both the 5% and 1% confidence levels across most asset and date-range combinations. The Coronavirus Recession is the exception: the short window and unprecedented volatility together degrade forecast performance.

Risk spillover between the two economies is substantial across the full range, but its predictive power markedly diminishes during both recessions — that is, precisely when a spillover signal would be most useful, it is least reliable.

Repository contents

Path	Contents
Thesis.pdf	The thesis (25pp)
Thesis_code.ipynb	Primary analysis notebook — data prep, model fitting, VaR estimation, Granger tests
Thesis_code.html	Rendered notebook, viewable without Jupyter
jupyter_notebooks/	Working notebooks: ARIMA order selection by AIC, model summaries, raw dataset

References

• Box, G; Jenkins, G. (1970), “Time Series Analysis: Forecasting and Control”, San Francisco: Holden-Day.
• Bollerslev, T. (1986), “Generalized Autoregressive Conditional Heteroskedasticity”, Journal of Econometrics, April, 31:3, pp. 307–27.
• Granger, C. W. J. (1969), “Investigating Causal Relations by Econometric Models and Cross- Spectral Methods,” Econometrica 37, 424-438.
• Granger, C.W.J. (1980), “Testing for Causality: A Personal View,” Journal of Economic Dynamics and Control 2, 329-352.
• Hamilton, J.D. (1994), “Time Series Analysis”, Taylor & Francis US.
• Hansen, B. (1994), “Autoregressive Conditional Density Estimation,” International Economic Review 35, 705-730.
• Lee, S. and B. Hansen (1994), “Asymptotic Theory for the GARCH(1,1) Quasi-maximum Likelihood Estimator,” Econometric Theory.
• Morgan, J.P. (1996), “Risk Metrics–Technical Document”, 4rd Edition, Morgan Guaranty Trust Company: New York.

Citation

Suh, Jongsun (2020). Modelling Volatility and Risk Spillover Between the Financial
Markets of US and China Using GARCH Value-at-Risk Forecasting and Granger Causality.
Undergraduate thesis, Department of Economics, Seoul National University.
https://doi.org/10.5281/zenodo.21507695

ORCID: 0009-0003-5053-4246

The thesis and the analysis code are archived separately: the DOI above cites the thesis; 10.5281/zenodo.21506848 cites this repository.