Extreme Fluctuations in U.S. Firm Growth Rates

We describe several important features of annual firm growth rate dynamics. Specifically, we report three major findings, some of which were previously known for certain subsets of U.S. firms and are here shown to obtain for the entire universe of U.S. business firms, and others that are apparently novel. First, the distribution of log growth rates overall are well-characterized by the so-called Laplace distribution, i.e., a two-sided exponential distribution, and the closely-related Subbotin distribution; log growth rates are not normal. This has important implications for overall firm survival since the Laplace distribution has much heavier tails than the normal, meaning that some U.S. firms experience much stronger fluctuations than would be encountered if the economy were more normal (Gaussian). We argue that it is small businesses that bear the brunt of such fluctuations, since they are sufficiently numerous that a certain fraction of them always experience giant annual fluctuations. We suggest that the ostensible dynamism of the small business sector is essentially rationalized by heavy-tailed growth rates. A second finding has to do with the relatively modest differences in growth rate distributions across (two digit) sectors. Specifically, we find that the heavy-tailed character of growth rate distributions is nearly ubiquitous. This is surprising since the number of firms within such sectoral classifications differ by orders of magnitude, age cross-sections vary substantially, technological shocks are presumably quite different, and so on. The only significant deviations from Laplace-distributed growth are toward even heavier-tailed growth, not toward normality. Third, we have discovered that for data on establishments, the variance in growth rates does not depend on establishment size. This is a surprising finding from a number of perspectives. Essentially what it means is that a large establishment is as likely to lay-off 10% of its workers, say, as is a small establishment, or that it is as probable for a small firm of 10 employees to hire 2 new people as it is for a size 100 firm to hire 20 new workers. From these findings we draw a variety of conclusions, some of which may be relevant for policy. Since extreme events/large fluctuations are an inherent part of firm growth, especially for small firms, government policies the goal of which is to shield firms from fluctuations may be difficult to devise. However, policies designed to help firms ride out difficult times may go a long way in promoting firm survival. Similarly, the heavy-tailed character of positive growth fluctuations may be largely responsible for the intrinsic dynamism of small businesses, and any policy aimed at dampening these fluctuations, if effective, may have, as an unwanted side effect, a weakening of the ability of the small business sector to provide growth. An important methodological conclusion that the present investigation highlights is the need for more systematic longitudinal data on firms, primarily multi-establishment ones. For economists a further sharp result is that econometric work utilizing normal specifications of log growth rates (or lognormal specifications of absolute growth rates) is very badly misspecified. The growth of firms is a critical, even foundational, process to modern, industrial economies, involving product, labor and financial markets. A comprehensive, empirically grounded understanding of this process has yet to be fully developed. The present study, highlighting important dynamical aspects of firm growth, contributes to this research enterprise.