Search the site by keyword

Firm dynamics and job creation: revisiting the perpetual motion machine

5. Variations across cohorts

Search publication by keyword.

Comparisons of survival rates and job creation across cohorts show conclusions about firm dynamics should not be generalised from a single cohort of firms.

Analysis of the 2001 cohort suggested that firms that are larger at birth have higher survival rates than firms that are smaller at birth. This relationship appeared to be persistent across all firm sizes (monotonic). However, replicating the analysis for other birth years shows that in some years the largest firms are not the ones with highest survival rates (see Figure 5).

For firms born in 2004, the firms that started with 10 to 20 employees have the highest survival rates. But for firms born in 2006, those that started with 20 to 50 employees have the same or higher survival rates than firms with 50 or more employees.

These observations are likely to be influenced by the fact that few firms are born with 50 or more employees (~0.1%). As a result, measurements of the survival rates of large firms will be sensitive to the failure of a small numbers of firms.

A stronger result from comparing growth rates across cohorts is that differences in survival rates, across birth sizes, appear to have narrowed over the time period being analysed. Differences in survival rates, across different birth sizes, were much wider in the 2001 and 2002 cohorts than in the 2004, 2005 and 2006 cohorts. The 2001 and 2002 cohorts show clear differences in survival rates between firms born very small (less than 1 employee) and other firms born small, with less than 10 employees. For later cohorts the survival rates of these smaller firms are almost indistinguishable from each other.

In terms of job creation, the smallest firms in the cohort, at birth, create the most jobs. This effect can be split into a small number of jobs being created by a large number of firms and large numbers of jobs being created by a small number of firms. This is shown in Figure 6 where job creation by firms born with less than 1 employee are large for those that grow to have 1-6 employees and those that grow to have 20 or more employees.

Firms that start out very small are often, but not always, the ones that go on to create the most jobs amongst firms that reach larger sizes. This can be seen in Figure 6 where firms that started with less than one employee went on to contribute the largest increases in employment (creation less destruction) amongst firms with 20+ employees in the 2001, 2002 and 2003 cohorts. In 2004, firms that started with 20+ employees were the ones that created the most jobs. For the 2006 cohort, it was firms that started with 1-6 employees that contributed the most to increased employment.

Figure 5   Survival rates by birth year and size at birth

Survival rates by birth year and size at birth

Figure 6   Net job creation by birth year and size

Cell values are net thousands of jobs created 10 years after birth[1]

Net job creation by birth year and size


For all firms, contributions to national employment growth are dominated by birth effects and early stage growth. That is, firms have their largest effects on overall growth early in their life cycle.[2] This can be seen in Figure 7, which charts contributions to national growth in employee counts by firm year of age and birth size. For each age, the Figure plots the range of contributions of the different cohorts (2001-2006), using ‘box plots’, to summarise the distribution of these contributions to growth. Each of the firm size groupings have their largest effects on overall growth in national employee counts during the first 2 or 3 years of life. Very small firms make their largest contributions during the second and third years of life. For other firms, their effects on growth are largest in the first and second years of life.

The effects shown in Figure 7 are significant relative to typical annual growth in employee counts. From 2001 to 2016, employee growth averaged 2.5% per year nationally. Firms starting with less than one employee contributed around a quarter of this growth, in their second year of existence. Firms that began with 1-6 employees and 50 or more employees each contributed around a quarter of this growth in their first year of life. All told, new firms contributed an average of 2.2% of the average 2.5% increase on employees in private firms – based on contributions to growth in the first year of life. This suggests creation of new firms is a crucially important engine of growth and more important to job creation overall than growth of firms as they age.[3] This result has been observed in studies of firm dynamics in the United States (Haltiwanger et al. 2012) and United Kingdom (Anyadike-Danes & Hart, M., 2018).

Figure 7   Contributions to employment growth by age and birth size

Cohorts born in 2001 to 2006[]

Contributions to employment growth by age and birth size


[1] The box plots, for each age, include a horizontal line indicating the median contribution, boxes showing the inter-quartile range (IQR), a line or “whisker” showing values that are 50% larger (smaller) than the upper (lower) quartile (as long as the maximum (minimum) values are 50% larger (smaller) than the upper (lower) quartiles), and dots that are outliers (values that are more than 50% larger (smaller) than the top (bottom) of the IQR).

[2] At least when growth is measured according to changes in national rolling mean employee counts in the private sector.

[3] This result was also suggested in prior studies for New Zealand, and it has been noted that this result depends on regulation with firm births and deaths a much less important driver of business and employment growth in New Zealand prior to substantial deregulation of the economy in the 1980s (Malcolm, 1993).

[4] Note that values have been suppressed, for confidentiality reasons, for firms born in 2003 and 2004 with 20+ employees that have 1 to 6 or 6 to 10 employees 10 years later. The Figure presents these as zeros because suppressed values create jarring changes in colour hues that hinder interpretation of the Figure.