Parameterizing Spatial Weight Matrices in Spatial Econometric Models
SpeakerChang Tan (University of Groningen)
Date and time
November 26, 2020
14:00 - 15:00
The spatial weight matrix generally denoted by the symbol W, is an essential part of each spatial econometric model. To avoid the exogenous pre-specification of the W, using a direct approach with bias correction within a SDM model with both spatial and time-period fixed effects, we estimate the elements of the W together with other common parameters simultaneously. Besides, we allow different parameterization of W of different spatial lags in the SDM model. Further, our model allows different explanatory variables interacting with W whose parameterization is different. The performances of two common forms of W matrices with two general normalization methods are investigated in a Monte Carlo study. Finally, a data set used by Elhorst & Arbia (2010) studying the GDP growth and convergence is applied to evaluate the empirical performance of our approach.