Firm dynamics and job creation: revisiting the perpetual motion machine

Empirical analyses of firm dynamics, from around the OECD, typically find that:

• most firms are small, including new firms (Bartlesman et al, 2005);
• most firms die young with, e.g. half of all firms “dying” in the first 5 years of life (Anyadike-Danes & Hart, 2018);
• small firms die younger (Evans, 1987) and grow faster than larger firms (Hall, 1987);
• most jobs are created by young firms (Criscuolo et al, 2014);
• a significant majority of firm growth, measured by changes in numbers of employees, can be attributed to a very small proportion of firms (Anyadike-Danes et al, 2014); and
• the size distribution of firms is right skewed – a consequence of the other stylised facts (Cabral & Mata, 2003).

The size distribution of firms have been studied extensively for decades, but it is only in the last decade or so that analyses of firm dynamics using micro-economic data has been able to flesh out the stylised facts listed above.

Prior to that, researchers relied heavily on cross-sectional analyses that revealed stylised laws about the size distribution of firms – such as Gibrat’s (1931) famous observation that firm growth is independent of firm size – but did not shed light on the underlying processes that created these distributions.

Although empirical studies consistently find the same stylised facts about firm behaviour, theory about the fundamental drivers of these dynamics is not so concordant. There are many theories – some partially competing – on what drives the firm dynamics that we observe in the data.

The fact that firms seem to fail at high rates, could be down to firm owners failing to understand their own capabilities and failing to match these abilities to the market (Jovanovic, 1982). Over time, firms can learn more about their capabilities (or relative efficiency) and about the market or industry they compete in. This learning increases their likelihood of growth or survival. Firms that discover they are inefficient, will exit.

Theories that involve firms learning about their own capabilities and the demands of the market can also explain why young firms grow more quickly than older firms. If, early in their life, owners of firms know little about their likelihood of success or about the resources they need to acquire to be successful, they will benefit from learning (as long as they do not fail) and are less likely to fail as their firm ages and they accumulate experience. But, after a time, the rate at which their knowledge improves reduces and as a consequence growth rates decline.

Rossi-Hansberg and Wright (2007) developed a theory that can explain why small firms grow faster than large firms. Firms need to acquire industry-specific human-capital, through learning-by-doing. If there are diminishing returns to human capital, small firms will provide higher returns and will grow faster, as long as they do not fail. This theory can explain differences in firm size distributions across different industries. Industries that rely on industry-specific physical capital provide lower returns to industry-specific human capital and consequently have a greater number of larger firms with higher growth rates.

The number and size of firms in an industry is also likely to depend on the life-cycle of an industry and industry-specific technologies. For example, industries that are relatively new and offer considerable scope for innovation, have: (i) high returns, (ii) large numbers of smaller firms, (iii) high growth rates and (iv) high failure rates. As the industry matures, innovation becomes more difficult and returns decline and the industry consolidates with a smaller number of larger firms (Jovanovic & MacDonald, 1994).

So different theories (and the above are just a sample) provide insights into different phenomena that relate to firm dynamics. They do not coalesce on a single grand theory. Instead, they all provide useful insights into and justifications for empirical observations regarding the birth, death and growth of firms.

[1] References provided typically confirm all of these stylised facts. Here the references are provided alongside facts that are emphasised by each of these studies. Other studies find similar stylised facts. See, for example, Haltiwanger et al (2012).