About the Speaker:

Nicholas Sim is a PhD candidate from Boston College. He received his Masters and Bachelors (first class honours) degrees in Economics from National University of Singapore. His areas of interest are in applied econometrics, macroeconomics, monetary economics and financial economics. He has published in refereed journals such as International Journal of Economic Theory, Economics Bulletin, Applied Economics Letters and B.E. Journal of Macroeconomics. 

Abstract:

Quantile regression is a methodology that examines the influences of a set of regressors on the quantile of the dependent variable. In this paper, we consider a nonlinear quantile regression framework where a regressor may in turn be a conditional quantile itself. Since the true conditional quantile regressor is unknown, feasible estimation entails replacing it with the fitted counterpart. This is a problem of generated regressors and our theoretical contribution is to study the asymptotic implications of generated regressors in nonlinear quantile regressions. As an application, we investigate the dependence between quantiles of international stock returns - the S&P500, FTSE and the Nikkei - using a nonlinear quantile regression model based on the copula function. Estimating from both Gaussian and Student-t copula regression models, we construct correlation surfaces that clearly reflect how the correlations between quantiles of these stock returns behave. For both FTSE-S&P500 and Nikkei-S&P500 pairs, the correlations tend to rise for returns below the 30th percentiles and become very weak for returns beyond the 80th percentiles. Crucially, this is evidence that the correlation asymmetry in international stock returns, where the correlation is larger in bear than in bull markets, is due to changes in the correlation coefficient and hence the joint distribution across different states of the economy.

Reservation:

Admission is free.  Please reply to Christina, e-mail: achristina@ntu.edu.sg or Tel: 6790-5689 to confirm your attendance.