Previous studies have shown that copulas are an effective means for analyzing dependence structure among financial assets in a constant, and in a time-varying, context. I analyze the dependence structure of daily returns from Korean financial assets using an integrated modeling approach. This approach combines GARCH models with Extreme Value Theory (EVT) and copula methods to investigate single asset and multivariate asset properties. I find that financial returns data are non-normally distributed against the classic assumption of the normal distribution. For these situations GARCH models have been developed to forecast and capture common facts about conditional volatility (e.g., fat tails, volatility persistence and clustering, asymmetry, and leverage effects). I employ GARCH models to estimate the conditional volatility and EVT to estimate the tails of the error distribution of daily returns during 2000-2013. Both leverage effects and long tails are clearly observed in the Korean financial data. By back testing two risk measures (Value-at-Risk and Expected Shortfall), I provide evidence that the GARCH-EVT model produces better estimates than methods which ignore the fat tails of the error distribution. I find that the time-varying dependence structures of multiple asset portfolios are compatible with several shocks (Korean credit crisis in 2002, U.S. home mortgage crisis in 2008, and the EU debt crisis in 2010). The portfolio assets include: Samsung Electronics stock, KOSPI200 Futures, US Dollar Futures, and 3-year Korean Treasury Bonds. I find that the identified domestic and external economic shocks have different impacts on the volatility of returns exhibited by these assets when combined into alternative financial portfolios. Empirical results from the static and time-varying copula models are not significantly different based on tests using information criteria and back testing methods. However, the comparison of these models is useful to demonstrate the connection between the time-varying vector of the dependence structure of each portfolio and the identified shocks. I find that time-varying regular vine copula models provide useful forecast performance and they are highly applicable tools for analyzing risk in financial time series data.