Home Address:
  521 Prospect Street, Apt. 3
  New Haven, CT 06511-2129

Telephone: (203) 809-5633 (cell)
                 (203) 776-3879 (home)

Econometric Theory
Applied Econometrics
Financial Economics

Econometrics
Applied Econometrics
Financial Economics

(Oral) Econometrics and Financial Economics, 2003
(Written) Microeconomics and Macroeconomics, 2002

Nonparametric Approaches to Dynamic Panel Modelling and Bias Correction

Professor Peter C. B. Phillips
Professor Donald W. K. Andrews
Professor Yuichi Kitamura

May 2006

M.Phil., Economics, Yale University, 2004
M.A., Economics, Yale University, 2002
M.A., Economics, Seoul National University, 2001
B.A., Economics, summa cum laude, Seoul National University, 1999

Dissertation Fellowship, Yale University, 2006
Carl Arvid Anderson Prize Fellowship, Cowles Foundation, 2005
Cowles Foundation Summer Prize, Cowles Foundation, 2004
Cowles Foundation Graduate Student Fellowship, Cowles Foundation, 2001–2005
University Fellowship, Yale University, 2001–2005
Japan–IMF Scholarship for Advanced Studies, IMF, 2001-2003 (declined)
Ilju Foundation Scholarship for Abroad Studies, Ilju Educational Foundation, 2001–2006 (declined)
U-San Graduate Scholarship, U-San Foundation, 1999–2001
Scholarship for Honors, Seoul National University, 1995–2001

Recipient of Raymond Powell Teaching Prize, Yale University, 2005
Teaching Fellow, Yale University Department of Economics
       Introduction to Probability and Statistics, Fall 2004
       Econometrics II (Graduate), Spring 2004
       Econometrics I (Graduate), Fall 2003
Teaching Assistant, Seoul National University Department of Economics
       Microeconomics, Spring 2001
       Advanced Econometrics (Graduate), Fall 2000
       Topics in Econometrics (Graduate), Spring 2000
       International Finance, Fall 1999
       Time Series Analysis, Spring 1999
Instructor
       Advanced Econometrics (Graduate), Seoul National University, Summer 2001
       Economic Forecasting and Time Series Analysis, Korea Electric Power Corporation, Spring 2001

Head Research Assistant, Korea Fixed Income Research Institute, Spring 2001

Research Assistant for Professor Joon Y. Park, Seoul National Uniersity; Daewoo Securities Co., Ltd., 1999–2001

Co-organizer, Graduate Summer Workshop, Yale University, Summer 2005
Organizer, Econometrics Reading Group, Yale University, 2004–2005

"Dynamic Panels with Spatially Dependent Errors."
"Grouped Mixed Proportional Hazard Models with Spatial Dependence."
"Quantile Cointegrating Regressions."

Professor Peter C. B. Phillips
Cowles Foundation
Yale University
PO Box 208281
New Haven, CT 06520-8281
Phone: (203) 432-3695
Fax: (203) 432-5429
E-mail: peter.phillips@yale.edu

Professor Yuichi Kitamura
Cowles Foundation
Yale University
PO Box 208281
New Haven, CT 06520-8281
Phone: (203) 432-3699
Fax: (203) 432-6167
E-mail: yuichi.kitamura@yale.edu

Professor Hanming Fang (Teaching Reference)
Department of Economics
Yale University
PO Box 208264
New Haven, CT 06520-8264
Phone: (203) 432-3547
Fax: (203) 432-6323
E-mail: hanming.fang@yale.edu

A substantial number of empirical studies involving dynamic panels employ the simplest dynamic specification (first-order linear dynamics) even when there is no obvious justification for doing so. Hence, these models are most likely misspecified and will render invalid any statistical methods based on correct model specification.

My dissertation seeks to tackle this problem by providing more general approaches to econometric specification, estimation, and bias correction in dynamic panel models with fixed effects. The first part of the dissertation considers panel regressions with general linear dynamic specifications and examines explicit asymptotic bias formulae in them. Some proposals for bias correction are suggested, which can be employed under time series misspecification. The second part of the dissertation investigates nonlinear structures in the dynamics of panel models and develops nonparametric estimation techniques suitable for panel system. The asymptotic results extend to partial linear models with exogenous variables.

I. A General Approach to Bias Correction in Dynamic Panels under Time Series Misspecification

Empirical work with dynamic panels has long relied on first-order autoregressive structures, which are unavoidable when the length of time T is small, and where bias correction has been worked out. However, there is little justification for such simple dynamic specification when the lag orders are unknown. First-order models are most likely misspecified, possibly in a serious way. In such cases, attempts to correct the bias using formulae that correct for first-order dynamics would be wrong and may even exacerbate the bias. To address these concerns, this chapter sets up a general framework for analyzing and correcting for biases in dynamic panel regressions under general time series specification.

Specifically, the chapter extends the bias formula in Nickell (1981), which relies on the large cross section sample size N, to the case where the dynamics follow a general autoregressive form. A higher order approximation for the bias is also explored. A limit distribution for estimators is developed that allows for lag order misspecification, when N and T are comparably large.

Since the results manifest additional biases under misspecification, it is argued that model specification should precede any correction for bias in dynamic panel modelling. This approach avoids the potential problem where bias can be exacerbated by corrections that are implied by misspecified models. To assist in the practical implementation of this approach, we suggest a general form for bias correction, in the context of dynamic model specification, which specifically incorporates a new lag order selection method.  An empirical study on habit formation in consumption preferences is presented to illustrate the use of the general bias correction.

II. Nonparametric Estimation for Dynamic Panel Models

In spite of the large and growing literature on nonparametric modelling in econometrics, little attention has been given to nonparametric estimation in dynamic panels. One explanation is the difficulty of treating individual effects and temporal dependence simultaneously in the context of nonparametric estimation. This chapter seeks to overcome this problem by developing series approximations for nonlinear dynamics in a panel system and by extending the standard linear dynamic panel model to a nonparametric form that maintains additive fixed effects.

This approach follows earlier work on cross sectional series estimation by Newey (1997), and generalizes the results to dynamic panels using nonlinear time series techniques. We obtain convergence rates and derive the asymptotic distribution of the series estimator when N and T are comparably large. Just as for pooled estimation in linear dynamic panels, an asymptotic bias is present, which reduces the mean square convergence rate compared with the cross section case. To tackle this problem, bias correction is conducted using a heteroskedasticity and autocorrelation consistent (HAC) type estimator.

Some extensions of this framework are also considered under exogenous variables and partial linear models, which are relevant in applications. The limit theory and bias correction formulae follow by extending earlier results. An empirical study on nonlinearities in growth empirics is presented to illustrate the use of the nonparametric estimation techniques for dynamic panels using the Penn World Table.