ESTIMATES OF TECHNOLOGY AND CONVERGENCE: SIMULATION RESULTS

Using a Solow-Swan model with a stochastic saving rate and stochastic productivity we analyse the distributions of parameter estimates that emerge under various choices of technology, and of the dimension of the panel on which cross-section regressions are based. There are distinct asymmetries that characterize these distributions. These asymmetries become more pronounced when the effects of a near-unit root in the productivity shock become magnified over a longer time horizon and when the underlying production function is not Cobb-Douglas. Consequently, relying on traditional econometric transformations of these parameter estimates based on symmetric distributions, such as t-ratios, will be quite misleading if one tries to assess technology parameters and b-convergence.