Panel tests of Gibrat's Law for Japanese manufacturing

The properties of the standard cross sectional test of the Law of Proportionate Effect (LPE) are compared with those of three alternative panel unit root tests, using Monte Carlo methods. The cross sectional procedure produces biased parameter estimates and the test suffers from a loss of power if there are heterogeneous individual firm effects. Suitably designed panel tests avoid these difficulties. Empirical results for a panel of Japanese manufacturing firms provides some support for the notion that log firm sizes are mean-reverting towards heterogeneous equilibrium values, and that the LPE should be rejected. (C) 2002 Elsevier Science B.V. All rights reserved.