From Gibrat’s legacy to Gibrat’s fallacy. A Bayesian approach to study the growth of firms

In this paper we investigate some properties of the patterns of firms’ growth. Several recent studies about this topic are based on some version of the so-called Gibrat’s Law, which assumes that firms’ growth is erratic. We aimed at testing Gibrat’s Law, as a first step towards a more systematic investigation of the patterns of firms’ growth. Using a Bayesian statistical framework that nests previous approaches, we find that: (i) there seems to be strong evidence against the Gibrat’s law on average, within or across industries; (ii) previous results are probably incorrect because they are based on models that do not exploit all information contained in a panel data set, particularly models that do not control for potential heterogeneity in the convergence rates; (iii) estimated steady states differ across units, and firm sizes and growth rates do not converge within the same industry to a common limiting distribution; (iv) there is only weak evidence of mean reversion, i.e. initial larger firms do not grow relatively slower than smaller firms. Differences in growth rates and in size steady state are firm-specific, rather than size-specific; (v) differences in growth rates do not seems to disappear over time.