Heterogeneity in spatial growth clusters

This paper sets forth an approach which allows for dealing with both model uncertainty and threshold effects of unknown form in spatial growth regression models. The estimation of threshold effects designates different spatial growth regimes which account for unknown structural heterogeneities in the parameter estimates. Using stochastic search variable selection priors, the paper deals with the issue of model uncertainty in a flexible and computationally efficient way. The paper uses Bayesian Markov chain Monte Carlo to simultaneously account for threshold effects, model uncertainty, and spatial dependence in regional growth regression models. The approach is illustrated for both identifying model covariates and unveiling growth regimes present in pan-European growth data.