Lung-fei Lee '77

Lung-fei LeeLung-fei Lee was born in 1948 in Canton, China. He received a BA degree in mathematics from the Chinese University of Hong Kong in 1971. Prior to coming to Rochester Lung-fei attended the University of Waterloo, Ontario, Canada where he received a MMath. degree in pure mathematics (1972) and a MPhil degree in statistics (1974). Lung-fei received his PhD in economics from the University of Rochester, NY in 1977. His advisors were Professor G.S. Maddala and Professor Sherwin Rosen.

Over the years Lung-fei has served as full professor in the Departments of Economics at the University of Minnesota, Minneapolis (1984-91), University of Michigan, Ann Arbor (1991-96), and Hong Kong University of Science and Technology (1994-2000). Currently, Lung-fei Lee is a University Chaired Professor at Ohio State University, where in 2014 he received a University Distinguished Scholar Award in recognition of his research accomplishments.

Lung-fei Lee is one of the leading researchers in theoretical econometrics. He is ranked third in the Econometriciansí Hall of Fame, which is comprised of the top 100 individuals in theoretical econometrics based on their publications over the period 1989-1995. Lung-fei was elected a fellow of the Econometric Society in 1990. His substantial published works range from break-through articles in theoretical econometrics to applied studies of disequilibrium markets, to topics in mathematical economics. The selected articles below give the idea of Mr. Leeís diverse contribution to the field of economics.

"Generalized Econometric Models with Selectivity", Econometrica, Vol. 51 (1983): 507-512.

Many economics problems, such as immigration and occupational choice, involve multiple choice and censored dependent variables and use econometric models that unify censored-regression and discrete-choice models. In this paper, Lung-fei suggests an approach to formulating such models when investigators have a priori theoretical reasons on the use of specific marginal distributions. Starting with the bivariate normal distribution (or indeed with any flexible bivariate distribution) Lung-fei applies a transformation to construct a bivariate distribution with given margins. The resulting likelihood functions are tractable and allow accurate and computationally simple estimation methods.

"Switching Regression Models with Imperfect Sample Separation Information With an Application on Cartel Stability" (with R.H. Porter), Econometrica, Vol. 52 (1984): 391-418.

This article was motivated by the study of cartel behavior when there are price wars. As firms switch from collusive to non-cooperative behavior, industry supply functions will shift. The authors propose a generalization of the exogenous switching model to the cases when additional imperfect information is available on sample separation. For the problem at hand, a regime classification indicator was available, but the series were inaccurate which evoked the errors-in-variables problem. The proposed efficient estimation method takes into account this additional (albeit imperfect) information, which helps to reduce regime classification error and to derive an optimal regime classification rule.

"Microeconometric Demand Systems with Binding Non-Negativity Constraints: The Dual Approach" (with M.M. Pitt), Econometrica, Vol. 54 (1986): 1237-1242.

This paper considers the estimation of demand systems for samples that contain a significant proportion of observations with zero consumption of one or more goods. The proposed approach uses evirtual pricesi (which can be thought of as reservation or shadow prices) to transform binding non-negativity constraints into non-binding constraints. With the introduction of virtual prices, one can determine the set of goods consumed, i.e., the demand regime, and how switching occurs in response to changes in prices, income or household characteristics. The regime switching conditions are intuitively appealing: goods are not consumed unless their reservation price exceeds their market price.

"Estimation of Linear and Nonlinear Errors-in-Variables Models Using Validation Data" (with J.H. Sepanski), Journal of American Statistical Association, Vol. 90 (1995): 130-140.

Measurement error often occurs in economics data and causes biases in estimation. In the presence of validation data, one might be able to derive consistent estimates from primary data, even without imposing distributional assumptions on regression and measurement error. By validation data, the authors mean samples from an independent assessment of validity study. The authors propose a method for consistently estimating the parameters of nonlinear regression models with measurement errors, in either the explanatory or dependent variables, or both. The estimation method utilizes projection of a nonlinear function onto the space of finite-order polynomials of validation data instead of a linear projection; it is computationally and analytically simpler than nonparametric and semiparametric methods.

"Semiparametric Estimation of Multiple Index Models: Single Equation Estimation" (with H. Ichimura). Ch. 1 in Nonparametric and Semiparametric Methods in Econometrics and Statistics. Ed. W.A. Barnett, J. Powell, and G. Tauchen, pp. 3-39. Cambridge University Press, New York, NY. 1991.

Many econometric models can be regarded as multiple index models with a known number of indices. For instance, the sample selection model of Gronau (1974) and Heckman (1974) is a single index model. Alternatively, consider a situation where a binary decision is a consequence of decisions made by two individuals, but econometrician cannot distinguish between the two. This bivariate choice model with partial observability of Poirier (1980) is a double index model. The authors show that all models that can be represented in a multiple index framework can be estimated by semiparametric least squares methods, if identification conditions are met. The estimator is shown to converge at a parametric n-1/2 - rate to a normal distribution.

"Simulation Estimation of Polychotomous-Choice Sample Selection Models" Forthcoming in Nonlinear Statistical Inference, a Festschrift in honor of Takeshi Amemiya; by Cheng Hsiao, Kimio Morimune, and James Powell, eds, Cambridge U. Press, 2000.

This paper focuses on simulation techniques for estimating sample selection models with polychotomous choices or multiple selection criteria. To dispense with unrealistic assumptions, such as independence across the disturbances in equations, Lung-fei investigates the performance of several simulation estimation methods, including simulated likelihood and two-stage estimation.

Irina Solyanik (June 2000)