Regional Income Convergence in the US: Evidence from a Pair-wise Approach

The neoclassical prediction of long-run convergence is investigated by employing a novel pair-wise econometric procedure. Within this framework, a probabilistic definition of convergence is proposed and forms the basis of the test. We distinguish between the case of strong convergence, where the cointegrating vector is [1, -1] and that of weak convergence, where the latter is relaxed. A pair-wise approach is utilised where all the possible pairs are examined by means of unit roots and cointegration tests. Overall, the evidence is weak with regard to the strong convergence case and stronger for the weak.