ABSTRACT
Models using level forms of the variables estimated for annual cross sections for both neoclassical and endogenous formulations. For the more elaborate formulations employing ordinary least squares with fixed effects estimation the first year of the relevant period is taken as the base year and the reference metropolitan statistical area (MSA) was arbitrarily selected as it is the first on the data set. The first test was that of absolute beta convergence as proposed by neoclassical growth theory. The estimation proceeded via ordinary least squares and included dummy variables to capture time effects. The sign shifts of the unionization and tax variables are gone, and all hypothesized sources of growth are positively associated with the growth rate of metropolitan income. The time dummy effects appear to a large extent to reflect the macroeconomic events and vagaries of the 1970–1990 period. Similarly, the results of the dummy variables show that in fact, the MSA dummies have a considerable effect.
