The spatial survival models typically impose frailties, which characterize unobserved heterogeneity, to be spatially correlated. However, the spatial effect may not only exist in the unobserved errors, but it can also be present in the baseline hazards and the dependent variables. A new spatial survival model with these three possible spatial correlation structures is explored and used to investigate the implementation of value‐added tax (VAT) in 99 countries over the period 1970–2009. Estimation is performed by a Bayesian approach through the Markov chain Monte Carlo method. The estimation results suggest the presence of a significant spatial correlation among the VAT introductions of neighbouring countries.