High-Dimensional Threshold Quantile Regression with an Application to Debt Overhang and Economic Growth
2015/02/03
研討會日期 : 2015-02-03
時間 : 14:30
主講人 : Mr. Tzu-Chi Lin
主持人 : Professor Yu-Chin Hsu
地點 : Conference Room B110
演講者簡介 : Mr. Lin will receive his Ph.D. in Economics from University of Wisconsin-Madison in 2015. His research fields are Econometrics, Macroeconomics, Development Economics, and Forecasting. He is applying for a position of the Institute of Economics, AS now.
演講摘要 : In this paper, we propose a threshold quantile Lasso estimator and investigate the existence of the 90 % debt-to-GDP tipping point found by Reinhart and Rogoff (2010a). We consider a high-dimensional quantile regression model with a potential threshold allowing for time series data. Our contributions include three parts: First, under mixing and sparsity conditions, we derive the consistency of the regression estimates regardless of the identification of the tipping point. Second, we show that the oracle property of adaptive Lasso can distinguish between linear and threshold regression models allowing the number of regressors to be much larger than sample size. Third, we apply the method to check the validity of tipping points between debt and GDP growth for cross-country and country-specific data. We find that the tipping point effects are more common in developing countries vis-á-vis developed countries and the thresholds are heterogeneous across different economies, ranging from 10% to 100%. Channels for the impacts of public debt in developing countries are monetary policies, government spending, and capital investment; whereas demographic factors and the infrastructure investment are prominent transmissions for developed countries.
