“An early, widely cited finding that the Seattle-Denver experiment increased marital dissolution was challenged by a reanalysis concluding that its negative income tax plans had no effect on the rate of dissolution relative to controls.”