This blog is co-authored by: Bob Rijkers at the World Bank and Hassen Arouri at the Institut National de Statistique. Additional contributors include, Caroline Freund and Antonio Nucifora.
SME promotion programs are becoming progressively more popular. While evidence on their effectiveness remains elusive, their policy prominence is predicated on the belief that small firms grow faster and generate the most jobs. Our preliminary analysis of the Tunisian registry of firms, which contains longitudinal information on all formal firms from 1996 until 2010, yields three stylized facts suggesting that large firms are far more important than small firms in generating employment and growth;
Stylized Fact #1: There are very few large firms, but they account for an important share of employment.
To start with, the vast majority of Tunisian firms are small. This may seem a trite observation, but should come as a big surprise to those who believe that small firms grow faster; if they did, one would expect there to be relatively more large firms. As is indicated in Figure 1 below, however, the data show the exact opposite, with 86% of all firms being one-person enterprises and only 0.4% of all firms employing more than 100 workers. Nonetheless, these firms, which we will refer to as large firms, account for more than a third of all jobs, more than all one-person firms combined.
Figure 1: There are very few large firms, but they account for more than a third of all jobs
The fact that most firms are small could also be an artifact of high entry, for example, if firms enter small and then some grow. While the vast majority of entering firms (96%) are one-person firms, entry alone does not account for the observed size distribution. Moreover, small firms are also more likely to exit (more on this below). At the same time, the fact that so many firms start as one-person enterprises is a sign that incumbent firms, which tend to be small, fail to create sufficiently many jobs, as one-person firms are synonymous with self-employment.
Stylized Fact #2: Small firms remain small
Indeed, a second striking feature of the data is that firms stagnate: small firms remain small. This is demonstrated by Table 2, which documents transitions between crude size categories in the short-run, i.e. between 2000 and 2001, and the long run, i.e. between 2000 and 2010. The data clearly show that one—person, micro and small firms hardly ever become large firms, even when we consider decadal, as opposed to annual, growth. In addition, very few micro firms graduate into the small size category. To add insult to injury, the matrices furthermore show that the very smallest firms are also most likely to die.
Figure 2: Limited mobility both in the short- and the long-run
Stylized Fact #3: It’s Age, not Size, that Matters
Of course these transition matrices are rather crude and do not take into account the contributions of entering firms. Graph 1 below presents the results of non-parametric regressions of net job creation, including by entrants, on more refined firm size categories. The results are striking. Firstly, we do not find a strong inverse relationship between firm size and growth – as is demonstrated by the purple line. This contrast sharply with results obtained for the U.S. where small firms contribute significantly to employment growth. Second, once we control for firm age, the relationship between firm size and growth becomes positive; in other words, to the extent that small firms created more jobs, they do so because they are young, not because they are small per se.
Figure 3: Small firms grow because they are young, not because they are small per se
Notes: The dependent variable is the Davis-Haltiwanger-Schuh growth rate, which allows for an integrated treatment of the contributions of entering, continuing and exiting firms. The regressions are weighted and control for industry and year effects; the resulting coefficients are thus interpretable as conditional average net job flows. To minimize the impact of measurement error, we base our size dummies on average size categories. Since we have more than 7 million observations, all size category variables are significant at the 0.01% significance level.
Last but not least, it is important to note that while this note has focused on Tunisia, where the data are most comprehensive in terms of coverage, our analysis of manufacturing firm-level census data in Morocco, Ethiopia and Indonesia yields similarly sobering stylized facts.
Join the Conversation