This paper proposes a new methodology to evaluate the economic effect of state-specific policy changes, using bankbranching deregulations in the US as an example. The new method compares economic performance of pairs of contiguous counties separated by state borders, where on one side restrictions on statewide branching were removed relatively earlier, to create a natural ''regression discontinuity'' setup. The study uses a total of 285 pairs of contiguous counties along 38 segments of such regulation change borders to estimate treatment effects for 23 separate deregulation events taking place between 1975 and 1990. To distinguish real treatment effects from those created by data-snooping and spatial correlations, fictitious placebo deregulations are randomized (permutated) on another 32 segments of non-event borders to establish empirically a statistical table of critical values for the estimator. The method determines that statistically significant growth accelerations can be established at a 4 90% confidence level in five (and none prior to 1985) out of the 23 deregulation events examined. ''Hinterland counties'' within the still-regulated states, but farther away from the state borders, are used as a second control group to consider and reject the possibility that cross-border spillover of deregulation effects may invalidate the empirical design.