A0490
Title: Multi-dimensional spatial panel data models with fixed effects: Formulation, estimation and inference
Authors: Zhenlin Yang - Singapore Management University (Singapore) [presenting]
Abstract: The formulation is considered, estimation and inference for multi-dimensional (mD) spatial panel data (SPD) models with both observable and unobservable dimension-specific fixed effects (FEs), where the latter is of a growing dimension and may appear in the model additively or interactively of various orders. A general method of formulating the unobservable (mD) FEs so that the observable (mD) FEs can be identified is given. A general M-estimation method is proposed to estimate the common parameters, where the unbiased estimating equations are obtained by adjusting the concentrated quasi-score function with the unobserved FEs being concentrated out. The adjusted quasi-scores (AQS) remove the effects of estimating these incidental FE parameters and thereby lead to M-estimators that are consistent and asymptotically normal. The proposed methods allow (i) the identification and estimation of the effects of space or time-invariant covariates, (ii) the spatial weights and coefficients to vary with time, (iii) the error variance to vary in all dimensions, and (iv) the panels to be nested or unbalanced. Simple inference methods are introduced for each scenario studied. The asymptotic properties of these methods are studied, and finite sample performance is assessed using Monte Carlo simulations.
