On the relationship between firms' size and growth rate

Since the seminal work of Pareto, many empirical analyses suggestedthat the distribution of firms size is characterized by an asymptoticpower like behavior. At the same time, several investigations showthat the distribution of annual growth rates of firms displays aremarkable double-exponential shape. Recently it has been suggestedthat both these statistical properties can be explained by assuming abivariate Marshall-Olkin power-like distribution for the size of firmsin subsequent time steps. Through analytical investigation, I showthat the marginal distribution of growth rates implied by thisassumption does not possess, in general, a Laplace shape and becomesdegenerate when subsequent size levels are perfectly correlated. Assuch, the bivariate Marshall-Olkin distribution is unable to properlyaccount for the observed regularities. The original suggestion isfaulty as it treats firm size levels as stationary stochasticvariables and neglects their integrated nature.