Review of Industrial Organization 23: 217–232, 2003.
© 2004 Kluwer Academic Publishers. Printed in the Netherlands.
Common Properties and Sectoral Speciﬁcities in the
Dynamics of U.S. Manufacturing Companies
GIULIO BOTTAZZI and ANGELO SECCHI
S.Anna School of Advanced Studies, P.zza Martiri della Libertà, 33, 56100, Pisa, Italy
Abstract. The size distribution and growth rate dynamics of U.S. companies have been extensively
studied by many authors. In this paper, using the COMPUSTAT database, we extend the analysis to
disaggregated data, studying 15 sectors of the U.S. manufacturing industry. The sectoral investigation
reveals the presence of general statistical properties that can be considered valid across all the studied
sectors. In particular, the probability density of ﬁrms growth rates invariably displays a characteristic
tent shape and the relation between the size of a ﬁrm and the variance of its rates of growth is
characterized, in different sectors, by very similar scaling relations. The presence of characteristics
that are robust and sectoral invariant hints at the existence of generic statistical properties shaping
the dynamic of ﬁrms across the whole industry.
Keywords: Firm growth, industrial sectors, Laplace distribution, power law.
JEL Classiﬁcations: L1,C1,D2
The statistical analysis of ﬁrm size and its dynamics constitutes one of the tradi-
tional problems in the Industrial Organization literature. Early investigations were
conducted over datasets at a high level of aggregation, typically including large
ﬁrms operating in very different sectors. Hart and Prais (1956), for instance, study
the distribution of the whole U.K. manufacturing industry while Simon and Bonini
(1958) explore the size distribution of the top U.S. companies across the whole
manufacturing industry. These investigations were primarily focused on the stat-
istical characterization of the size distribution of ﬁrms and on the analysis of ﬁrm
growth dynamics in terms of autoregressive stochastic processes (in a large body
of contributions see for instance Dunne et al. (1988), Evans (1987), Hall (1987)).
More recent contributions (Stanley et al., 1996b; Amaral et al., 1997) extend
these studies to the analysis of the growth rates distribution and to the relation
between the size of a company and its rates of growth. In particular they show
that the probability density of growth rates of U.S. ﬁrms aggregated over all the
manufacturing sectors displays, on a log–log scale, a tent-like shape that can be
represented by a Laplace (symmetric exponential) distribution. They also identify