Do firm sizes and profit rates converge? Evidence on Gibrat's Law and the persistence of profits in the long run

Three panel unit root tests are applied to a 31-year firm size, growth and profit rate data set for 96 large, quoted UK firms. All tests reject the unit root null for log size if the Augmented Dickey Fuller autoregressions exclude a linear time trend. If a linear trend is included, the results are more ambiguous and appear to differ systematically between tests. The view that firm growth is either random or near-random does not receive unequivocal support in all tests. There is however strong and consistent evidence of mean-reversion in profit rates.