We study a model of tax evasion dynamics with a local enforcement externality due to limited commitment by the state in the allocation of resources to fiscal juridictions. Taxpayers rationally decide their compliance behavior given they do not have precise information about the probability of apprehension, but learn it over time following a simple social learning mechanism, based on their experience.
We analyse the conditions under which the distribution of the perceived probability of apprehension and compliance behavior converge to one or multiple steady states and whether convergence features history dependence. We show that the emergence of steady state multiplicity crucially depends on the level of tax evasion compared to a measure of efficiency in the enforcement process.
The results are used to explore the long run consequences of a fiscal union in cases of equilibrium multiplicity, i.e. when historical initial conditions operate as a selection criterion. As an illustration of the long lasting consequences of initial conditions, we document persistent heterogeneity of tax compliance in Italy, suggesting an interpretation of the so called Mezzogiorno problem in terms of institutional traps, as an alternative to other interpretations based on heterogeneity in the endowment of social capital or other structural characteristics.