METHOD TO CONSTRUCT A FACTOR COPULA MODEL: AN EMPIRICAL APPLICATION TO THE STI COMPONENT STOCKS.

Abstract

This paper proposes using exploratory factor analysis (EFA) as a key step in constructing the factor copula model introduced by Oh and Patton (2017). It also studies the dependence pattern and latent factor structure of the STI component stocks, using results from dependence tests and factor analysis to arrive at specifications for the factor copula model. The model is applied to 29 STI components. From the factor analysis, I find evidence that support the inclusion factors that go beyond industry classifications. These specific factors are found to have significant influence on stock returns; comparable to the influence of a common market factor and should not be ignored. I also find evidence of tailed dependence and asymmetry of the tailed dependence, indicating that stocks are more correlated during busts than in booms. Lastly, I show that the EFA step provides an improvement in performance in estimating measures of systematic risk.