2. Firm Growth, and High-Growth Firms

Since decades, there has been a keen interest in measuring firm growth. A main reason for this was due to popular interest in job creation. Pioneering work by David Birch (Birch, 1979) investigated the dynamics of employment growth by entrepreneurial firms. An important contribution of David Birch was moving beyond anecdotal cases to focus on large-sample datasets thereby gaining a representative view of job creation by SMEs. In particular, it was his finding that a small number of firms create a disproportionately large number of jobs, that led him to introduce the term “gazelles” to describe high-growth firms. Since Birch (1979), the literature on HGFs has been further developed and standardized such that major statistical offices (e.g. Eurostat and the OECD) collect data on HGF shares in regions and countries over time.

A first challenge for empirical work on firm growth is to decide upon how to measure growth. This task is not necessarily as simple as it sounds. For example, according to an indicator based on a layman’s understanding of growth rates, when a firm with 100 employees that grows by 50% and then declines by 50% will not actually end up back where it started (100 employees) but will end up with 75 employees.¹ Tornqvist et al. (1985) discuss the various ways in which growth rates can be measured, and discuss the desirable statistical properties of growth rates indicators, and conclude by recommending that growth rates should be measured in terms of log-differences:For firm i at time t. Firm size is measured in terms of variable X, which usually corresponds for total sales or employment, although other size indicators are sometimes used (such as value added or total assets).²

Research on firm growth has generally measured growth on an annual basis (Coad, 2009). Measuring growth over a period shorter than one year leads to problems due to within-the-year seasonality and is also difficult given that firm-level data are usually reported at annual intervals in most datasets.

Measuring growth over a period longer than one year has the advantage that random fluctuations might become smoothed out, thus reducing the role of statistical noise and boosting the signal-to-noise ratio. However, measuring growth over a longer period results in fewer observations in the dataset, and many firms that do not survive until the end of the growth period are then not observed. Given this tradeoff, the standardized definition of HGFs focuses on growth over a period of three years (Eurostat-OECD, 2007).

Eurostat-OECD (2007) propose a binary variable for HGFs, which is equal to one for firms that have at least 10 employees in the initial period and a geometric average of at least 20% growth per year over a three-year period.where \(E_{t}\) is the number of employees at time t, and \(S_{t}\) is an indicator of firm size (which can be either sales or employees). This HGF indicator has been applied by many researchers in academic and policy publications (Henrekson and Johansson, 2010; Flachenecker et al., 2020; Grover Goswami et al., 2019; Coad and Srhoj, 2020; Benedetti Fasil et al., 2021), and is generally seen as the standard definition of HGF.

Having a standardized HGF indicator has several benefits. It means that HGF-shares can be compared across countries and industries over different time periods. Econometric results can thus become comparable across studies, leading to a shared understanding between researchers, cumulativeness of knowledge, and meaningful theory-building.

However, the Eurostat-OECD HGF indicator is not without drawbacks, two of which are mentioned here. First, an issue with this HGF indicator is that micro firms (fewer than 10 employees at the start) are excluded from the analysis (Daunfeldt et al., 2015). These firms constitute a lion’s share of all jobs created in the economy. This latter concern can be addressed by using a modified HGF indicator that includes micro firms. One way of doing this is as follows. Recall that a firm starting with 10 employees must grow by at least 8 employees to qualify as an HGF; we can apply this same growth amount to consider that any firm with fewer than 10 employees at start must grow by at least 8 employees if it should be counted as an HGF (Coad and Srhoj, 2020). For example, a firm with 2 employees at start can become an HGF if it grows to 2 + 8 = 10 employees after the three-year period.

Second, requiring that HGFs grow by 20% on average each year can result in a small number of HGFs in the dataset. As such, there has been a recent shift to complement results from a 20% growth threshold with results from a 10% threshold for average annual growth when measuring HGFs (e.g. Flachenecker et al., 2020; OECD, 2021; Benedetti Fasil et al., 2021) to avoid the statistical problems of having too few observations in the HGF = 1 category.³ Interestingly, scale-up authors have sometimes taken a different view, arguing that the 20% threshold is too low. Reuber et al. (2021, p. 1033) explain that “it is not unusual for globally scaling firms to exceed a 40% compound annual growth rate.” Hoffman and Yeh (2018, p. 46) even state that “[d]ropping below even 40% annual growth is a warning sign for investors.”