This paper adopts a spatial probit approach to explain interaction effects among geographical units when the dependent variable takes the form of a binary response variable and state transfers occur at different moments in time. The model has two spatially lagged variables, one for units that are still in state 0 and one for units that already transferred to state 1. The parameters are estimated on observations for those units that are still in state 0 at the start of the different time periods, whereas observations on units after they transferred to state 1, just as in the literature on duration modeling, are discarded. Consequently, neighboring units that did not yet transfer may have a different impact than units that already transferred. We illustrate our approach with an empirical study of the adoption of in ation targeting for a sample of 58 countries over the period 1985-2008.
|Naam||SOM Research Reports|
|Uitgeverij||University of Groningen, SOM Research School|