Using cross-country variances to evaluate growth theories

Much of the empirical growth literature has attempted to evaluate growth theories by estimating regressions that relate the growth rate of per capita output for a sample of countries to initial per capita output and country characteristics. The resulting inferences are shown to be invalid except under strong conditions. An alternative method that uses cross-country variances is formulated and shown to produce valid inferences under weak conditions. Applying this method to data from thirteen countries over the period 1870-1989 provides no evidence that their per capita outputs have different trend growth rates and much evidence that they revert toward a common trend. This evidence does not support those endogenous growth theories that predict appreciably different trend growth rates across countries.