My research interests are time series econometrics, forecasting, financial economics and applied macroeconomics. I am particularly interested in forecasting and forecast evaluation when there is uncertainty regarding predictive model selection and structural breaks.
My research statement can be downloaded here.
Job Market Paper:
“Out-of-Sample Forecast Model Averaging with Parameter Instability”
Forecasting economic variables of interest when structural breaks are possible is highly relevant to many macroeconomic and financial applications. This task becomes even more challenging when the underlying time series data exhibits features such as conditional volatility, as it is difficult to separate the impact caused by parameter instability and volatility. As an alternative to model selection by hypothesis testing, model averaging or forecast combination provides possible predictive gains by better managing the model selection risk. This paper extends Hansen's (2009, Econometric Theory) model averaging theory under uncertainty regarding structural breaks to the out-of-sample forecast setting. We propose new predictive model weights based on the leave-one-out cross-validation criterion (CV), as CV is robust to heteroscedasticity which is relevant to many economic applications. We present the theoretical form that the CV penalty term takes in the setting where both parameter instability and conditional heteroscedasticity are present, then derive the sample optimal predictive model weights based on the cross-validation criterion. To support our theoretical results, we provide Monte Carlo evidence showing that CV weights outperform competing methods (i.e. Mallows' weights, equal weights, and Schwarz-Bayesian weights) in several simulation designs. Last, we apply the CV weights to forecasting the U.S. and Taiwan quarterly GDP growth rates out-of-sample, and demonstrate their better empirical performance compared with other methods.
My job market paper has been accepted for presentation at the following academic conferences:
Work in Progress:
“Forecast Equity Premium with Structural Breaks”
This paper applies the newly developed optimal and robust weighting theory by Pesaran et al. (2013, J Econometrics) to forecasting U.S. market equity premium in the presence of structural breaks. The weights are optimal in the sense of minimizing the expected mean-squared forecast error, or robust to the break dates and size estimation error. Based on break detection and forecast evaluation results, we conclude that parameter instability cannot fully explain the weak predictive ability of most factors considered in Goyal and Welch's paper (2008, RFS).
My research statement can be downloaded here.
Job Market Paper:
“Out-of-Sample Forecast Model Averaging with Parameter Instability”
Forecasting economic variables of interest when structural breaks are possible is highly relevant to many macroeconomic and financial applications. This task becomes even more challenging when the underlying time series data exhibits features such as conditional volatility, as it is difficult to separate the impact caused by parameter instability and volatility. As an alternative to model selection by hypothesis testing, model averaging or forecast combination provides possible predictive gains by better managing the model selection risk. This paper extends Hansen's (2009, Econometric Theory) model averaging theory under uncertainty regarding structural breaks to the out-of-sample forecast setting. We propose new predictive model weights based on the leave-one-out cross-validation criterion (CV), as CV is robust to heteroscedasticity which is relevant to many economic applications. We present the theoretical form that the CV penalty term takes in the setting where both parameter instability and conditional heteroscedasticity are present, then derive the sample optimal predictive model weights based on the cross-validation criterion. To support our theoretical results, we provide Monte Carlo evidence showing that CV weights outperform competing methods (i.e. Mallows' weights, equal weights, and Schwarz-Bayesian weights) in several simulation designs. Last, we apply the CV weights to forecasting the U.S. and Taiwan quarterly GDP growth rates out-of-sample, and demonstrate their better empirical performance compared with other methods.
My job market paper has been accepted for presentation at the following academic conferences:
- Annual Missouri Economics Conference, Columbia, MO, USA, 2014
- Midwest Econometric Group Annual Meeting, Iowa City, IA, USA, 2014
- Canadian Econometric Study Group Annual Meeting, Vancouver, BC, Canada, 2014
- Missouri Valley Economic Association Annual Meeting, St. Louis, MO, USA, 2014
- Southern Economic Association Annual Meeting, Atlanta, GA, USA, 2014
Work in Progress:
“Forecast Equity Premium with Structural Breaks”
This paper applies the newly developed optimal and robust weighting theory by Pesaran et al. (2013, J Econometrics) to forecasting U.S. market equity premium in the presence of structural breaks. The weights are optimal in the sense of minimizing the expected mean-squared forecast error, or robust to the break dates and size estimation error. Based on break detection and forecast evaluation results, we conclude that parameter instability cannot fully explain the weak predictive ability of most factors considered in Goyal and Welch's paper (2008, RFS).
