A1230
Title: A Bayesian SAR model with endogenous time-varying spatial weight matrices
Authors: Tamas Krisztin - International Institute for Applied Systems Analysis (Austria) [presenting]
Philipp Piribauer - Austrian Institute of Economic Research (WIFO) (Austria)
Christian Glocker - WIFO (Austria)
Matteo Iacopini - Queen Mary University of London (United Kingdom)
Abstract: A Bayesian approach is developed to estimate time-varying weight matrices in spatial autoregressive (or spatial lag) models. The recent approaches are extended for endogenously estimating weight matrices by allowing for a time-varying specification using a finite number of states. A spatial weight matrix, which is binary prior to row standardization, is estimated for each state. State transition matrices are estimated using the forward-filtering backward-sampling algorithm. The virtues of our approach are demonstrated using a dataset of inflation indicators within the Eurozone.
