科学研究

Prof. Lung-Fei Lee is a professor of economics at the Ohio State University. He is also a fellow of Journal of Econometrics, the Econometric Society, the Spatial Econometrics Association, and the Society for Economic Measurement. He has served as co-editor or associate editor of Journal of Econometrics, Journal of Applied Econometrics, and Regional Science and Urban Economics. His research and publications are in the areas of microeconometrics and theoretical econometrics. His current research is on the development of econometric models of spatial or social interactions. His work has been published in journals including Econometrica, Journal of Econometrics, International Economic Review, Journal of Applied Econometrics.

We introduce a spatial autoregressive model for an origin-destination flow (SARF model). Each flow yn,ij illustrates a signal from an origin j to a destination i. Our model quantifies three channels of spatial influences on yn,ij: (i) effect by outflows from j, (ii) effect by inflows to i, and (iii) effect by flows among third-party units. To accommodate frequent data environments of flows, we introduce SARF Tobit models for a censored flow variable. In the event of no censoring, we present a linear SARF model. To illustrate the formation of zero flows well, SARF hurdle models are developed as an extension, which works better than the standard SARF Tobit specification in our empirical application. We also accommodate two-way fixed effects in the model for origin’s and destination’s innate characteristics. The maximum likelihood (ML) estimation method is employed to estimate the model’s parameters. Asymptotic properties of the MLE are investigated by the spatial near-epoch dependence (NED) concept. Under the fixed-effect specification, the existence of asymptotic bias in the MLE is verified. Hence, we derive the analytic bias correction formula. Test statistics for testing the validity of the normal distribution assumption for the SARF Tobit and SARF hurdle models are provided. Using our models, we estimate the three channels of spatial influences in the U.S. states’ migration flows.