Keynesian and New Classical Models of Unemployment Revisited.

Several Keynesian and New Classical models of unemployment for the United States are reevaluated. The models are examined for adequacy by testing the cross-equation restrictions (where appropriate) and using diagnostic and nonnested tests that explicitly recognize the problem introduced by generated regressors. The best New Classical model for the 1946-73 period is found to be adequate when it is estimated over the 1946-85 period, whereas the Keynesian model is not. Existing results of tests obtained at the single-equation level ignoring generated regressors are not always supported when the test statistics are calculated using the correct covariance matrix or maximum likelihood. Copyright 1991 by Royal Economic Society.