Identifying high-growth firms in five European countries: micro firms vs. larger firms

Abstract

Purpose

This study uses an extensive sample of firms from Germany, France, UK, Portugal and Spain with the aim of obtaining conclusive results on the determinants that drive a firm to be high-growth firm (HGF). This sample includes micro firms, which are not generally considered in the literature. There are several reasons to take them into account: not excluding an important part of the business fabric, the results can be extrapolated, the study can show if micro firms also present high growth and if there are differences in the factors that determine the probability of being an HGF between both segments of firms.

Design/methodology/approach

A multivariate dynamic model of binary choice is used to analyse the probability of a company being classified as high growth. Then, with the Blinder and Oaxaca decomposition, the differences in the probability of being an HGF between micro firms and non-micro firms are studied.

Findings

The results show that HGFs demonstrate persistence, and younger firms are more likely to be HGFs. Micro firms also register high growth, although they are less dynamic and show a negative differential with respect to larger firms as highlighted by the characteristic component.

Originality/value

In some countries, such as Spain and Portugal, micro firms predominate, and these tend to be less dynamic, so to identify how to improve business dynamics, the factors that limit the growth of this type of company must first be determined. In this paper, in line with Acs and Mueller (2008), we therefore include firms with less than ten employees so as not to exclude an important part of the business fabric and to ascertain whether this type of firm also shows high growth.

研究目的

由於高速增長的公司被發現較其對手享有較大的競爭優勢,故它們成為促使就業機會會異常地淨增的推動器。這些公司在這方面的能力,成為廣泛研究的課題。本文擬就這課題的探討作出一點貢獻。

研究設計/方法/理念

研究的方法是透過把研究焦點集中於一個何謂高速增長公司的更廣泛的定義,而該定義之所以更廣泛,是因為它納入高速增長的微型公司。研究人員以 Birch-Schreyer 指數把公司分類為高速增長。本研究旨在評定高速增長的公司在廣義上是否受賦予可把其區別於低增長公司的獨特特點。為達到這個目標,研究人員使用二元選擇的多元動態模型,去分析一間公司會被分類為高速增長的機率,繼而以 Blinder-Oaxaca 分解方法,去探討微型公司會被分類為高速增長的機率,與非微型公司的機率兩者之間的差異。

研究結果

研究結果顯示、高速增長的公司展示了毅力; 而且，年輕的公司更有可能成為高速增長的公司。研究結果亦顯示、雖然根據特徵成份所強調,微型公司的活力不及較大的公司; 而且,微型公司展示負面的差分,唯它們也躋身高速增長公司的行列。

Citation

Rico-Belda, P. and Cabrer-Borrás, B. (2024), "Identifying high-growth firms in five European countries: micro firms vs. larger firms", European Journal of Management and Business Economics, Vol. ahead-of-print No. ahead-of-print. https://doi.org/10.1108/EJMBE-09-2023-0282

1. Introduction

It is a widely accepted fact in the economic literature that a small number of firms which show high growth are the ones that generate the greatest net employment in the economy (Birch and Medoff, 1994; Schreyer, 2000; Acs and Mueller, 2008). High-growth firms (HGFs) generate a higher employment rate than their competitors, and therefore, HGFs are crucial for boosting growth in economies (Acs and Mueller, 2008; Henrekson and Johansson, 2010; Jansen et al., 2023). This fact explains the great interest shown in HGFs by both policymakers (European Commission, 2010) and researchers.

Birch (1979, 1981) showed that a small number of HGFs generated most of the net employment. Since this seminal work, many studies have focused on identifying the characteristics of HGFs and the factors that determine them. However, as Rocha and Ferreira (2021) indicate, despite the vast academic literature, no definitive conclusions have been reached because of the widely divergent criteria used to identify HGFs. As Cristofaro et al. (2024) highlight, identifying the factors contributing to the high growth of firms provides valuable insights for entrepreneurs, investors and policymakers, ultimately increasing the likelihood of firms’ success. Moreover, the benefits of such identification extend to improved resource allocation, targeted investments and the development of supportive policies that collectively enhance the overall economic landscape.

Mason and Brown (2013) consider that, although public initiative tends to stimulate innovation and entrepreneurship, mainly related to technology, it does not always target those firms that may be more dynamic. According to these authors, “technology policy is not a substitute for high-growth firm policy”. In the same vein, the results of Daunfeldt et al. (2016) for Swedish firms, during the period 1997–2008, “suggest that the relationship between R&D and high growth is, at best, highly complex but is most likely negative”. To implement adequate policies, first the factors that determine the presence of HGFs need to be clearly identified. Similarly, as indicated by Marrocu et al. (2012), the location of a company is an important factor in its development, and it should be given more consideration, since the local endowment of capital, both human and intangible, generates and catalyses externalities, as evidenced by Rico and Cabrer-Borrás (2020a, b).

This study aims to analyse in depth the characteristics of HGFs to obtain conclusive results on the determinants that foster an HGF. In this paper, in line with Acs and Mueller (2008), we include firms with fewer than ten employees to avoid excluding an important part of the business fabric and to ascertain whether this type of firm also shows high growth. Moreover, HGFs are identified using the neutral index proposed by Birch–Schreyer (BS), which allows small firms to be considered without biasing the results towards them, as occurs with the Eurostat-Organisation for Economic Co-operation and Development (OECD) definition.

This paper addresses three research questions: What are the determinants that drive a firm to be HGF? Are there differences in the factors influencing the probability of being an HGF between micro firms and larger firms? Do external factors, related to the location of firms, influence the emergence of HGFs?

To accurately determine the factors of business growth, a dynamic multivariate binary choice model is used, which enables the identification of some of the explanatory factors of the probability of being an HGF. In addition, using the decomposition of Blinder (1973) and Oaxaca (1973), the differences in the probability of micro firms being HGFs are analysed.

This paper makes several contributions to the literature. First, it uses a broader definition of HGF by including micro firms. Second, it considers external factors as determinants of the probability of being an HGF. Finally, it applies a methodology that allows for an analysis of the differences in the probability of being an HGF between micro and non-micro firms.

2. Theoretical background

2.1 Definitions of high-growth firms

Various criteria have been proposed in the literature to classify a firm as HGF (Henrekson and Johansson, 2010). The criteria depend on the indicator used (turnover, employment, profit, etc.), the measure of growth (in absolute or relative terms) and the period considered. Mogos et al. (2021) state that the definition of an HGF is important, since “identification of HGFs and results vary significantly with the specific definition used, as well as with the specific variable used to measure growth”. Delmar et al. (2003) point out that firms classified as HGFs using one definition can cease to be so when using another.

Birch and Medoff (1994) stipulated that an HGF should have a minimum of 20% turnover growth each year over a five-year window, starting from a base-year revenue of at least $100,000. According to Lopez-Garcia and Puente (2012), using turnover to calculate growth leads to comparison problems due to the lack of consensus on how to deflate this variable. For this reason, the growth indicator employment is recorded in the literature as being the more commonly used indicator.

As indicated by Coad et al. (2014) and Aldrich and Ruef (2018), no unanimous agreement exists in the literature on the concept of HGF, although the Eurostat-OECD (2007) indicator is often considered the standard definition for HGF (Grover et al., 2019; Jansen et al., 2023). The Eurostat-OECD definition considers firms with average annualised growth in employees greater than 20% per annum over a three-year period and with ten or more employees at the beginning of the observation period as HGFs. However, as pointed out by Daunfeldt et al. (2015), applying a size threshold of ten employees might exclude many small entrepreneurial firms that show high-growth rates. This creates a sample selection problem, which may lead to incorrect policy decisions since the spread of small HGFs differs across countries, regions and sectors. In our study, in line with Acs and Mueller (2008), firms with less than ten employees are considered; these are referred to as micro firms. Our objective is to compare the behaviour of micro firms with that of firms with more than ten employees and identify whether they present a different behaviour.

The use of relative change in employment as a growth measure is complicated, since this skews towards small firms. However, the absolute change in employment skews towards large firms. For this reason, Birch (1987) and Schreyer (2000) propose an index which combines absolute and relative growth, known in the literature as the BS index:

(1)BS=(Ei,t−Ei,t−3) (Ei,tEi,t−3)

where Ei,t  and Ei,t−3 correspond to employment at the end and beginning of the period, respectively.

HGFs are firms that rank in the top 10% percentile of the BS index. It can be demonstrated that this measure, while still dependent on firm size, always gives rise to a smaller bias towards any size class than either the relative or the absolute measure of growth (Schreyer, 2000). However, Hölzl (2014) considers that BS index gives more attention to absolute job creation and therefore does not provide a definitive solution for measuring growth. In samples containing many firms with less than ten employees, as is the case of Spain and Portugal, the definition using BS index would be the most appropriate.

Birch (1987) used the term gazelle to describe small firms that grow very quickly, as opposed to what he called mice firms, those that are small and with low growth and elephant firms, which are very large but with low growth. The OECD (Ahmad, 2006) proposed using the term gazelle to describe young firms, less than five years old, with an average annualised growth of employees greater than 20% per year over a period of three years and with ten or more employees at the beginning of the observation period.

After presenting the different definitions, this paper considers the following types of firms. (1) HGFs that are in the 10% percentile with the highest value of the BS index [1], and within these, we distinguish between those that have more than ten employees and those that have fewer than ten employees (micro firms); (2) mice firms, those with less than twenty employees (but more than ten employees) and that do not present high growth (Acs and Mueller, 2008); (3) elephant firms, those with more than five hundred employees and not presenting high growth (Acs and Mueller, 2008), and (4) those firms that do not fall within the previous classifications.

2.2 Literature review

Birch (1979, 1981) pointed out that a small number of fast-growing firms were responsible for most of the economy’s job creation. Several empirical studies have corroborated this finding. In addition, many studies have shown that, although HGFs can be of any size, it is young and small firms which predominate. However, Brown et al. (2017) consider that there are many myths around what is an HGF and attempt to break down the information in terms of to what degree young, small and high-technology firms predominate.

Although Gibrat’s law [2] assumes that the growth rate of a firm is independent of its size, this law is inconsistent with the empirical evidence since Sutton (1997) reveals a negative correlation of growth with firm size. Caves (1998) concludes that business growth is independent of size above a certain threshold, while for small businesses the growth rate decreases with size. Likewise, most studies agree that smaller firms tend to grow faster than larger ones (Birch, 1981; Acs and Mueller, 2008; Lotti et al., 2003). However, Acs and Mueller (2008) show that HGFs are present in all sizes.

There is no conclusive evidence in the literature regarding the influence of age. For example, Birch (1981) considers HGFs to be very young firms, and various research studies (Yasuda, 2005; Lopez-Garcia and Puente, 2012; Daunfeldt et al., 2014) show a negative relationship between growth and age. Coad and Karlsson (2022) evidence that the majority of HGFs are small and young. However, Acs et al. (2011) show that HGFs are relatively old, since the average age of these firms was 25 years (Grazzi and Moschella, 2017). As indicated by Delmar et al. (2003) and Mogos et al. (2021), firm age rather than size is often considered a determinant of rapid growth. In this vein, Haltiwanger et al. (2013) show that when controlling for age, the effect of size disappears.

According to Giner et al. (2016), the sector to which a firm belongs has an ambiguous effect on the growth process. However, Henreksson and Johansson (2010) and Satterthwaite and Hamilton (2017) conclude that HGFs have a greater presence in the services sector. Along the same lines, Daunfeldt et al. (2016) find that HGFs are over-represented in knowledge-intensive service industries, although other studies reveal that HGFs tend to be present in all sectors, not necessarily dominating in highly technological sectors (Acs and Mueller, 2008; Henrekson and Johansson, 2010; Lopez-Garcia and Puente, 2012). In fact, with regard to innovation and technology, Kang et al. (2018) show that research and development drive business growth. Likewise, Segarra and Teruel (2014) show that investment in R&D positively affects the probability of being an HGF. However, Daunfeldt et al. (2016) show, in Sweden, that the industries with high research and development (R&D) intensity have a lower share of HGFs than those with a lower R&D intensity.

Benesová et al. (2018) argue that the creativity of human resources and their performance influence the rapid growth of firms. Likewise, Daunfeldt et al. (2015) and Lopez-Garcia and Puente (2012) support the idea that human capital is a fundamental factor in HGFs. The quality of human capital can be measured through labour productivity. However, Moschella et al. (2019) indicate that the results regarding productivity are mixed. Bianchini et al. (2017), Du and Temouri (2015) and Levratto et al. (2010) observe that the higher the productivity, the higher the growth. However, Daunfeldt et al. (2014) find no association between productivity and high growth.

Levratto et al. (2010) and Lopez-Garcia and Puente (2012) find no relation between a firm´s financial situation and high growth. More specifically, Coad et al. (2011) and Bianchini et al. (2017) consider the role of profits, a lesser-researched characteristic, as a determinant of HGFs. They show that HGFs are more profitable than other firms, although, as has been observed, very few studies focus on this variable.

The quality of the business ecosystem and public policies to encourage entrepreneurship are also determining factors in the presence of HGFs. Morris et al. (2015) and Benesová et al. (2018) argue that rapid growth derives from public encouragement of innovation and the creativity of human resources. Likewise, location is a determining factor in the proliferation of HGFs (Bravo-Biosca et al., 2013; Li et al., 2016), since, as indicated by Giner et al. (2016), the main urban and technological spaces are significantly linked to the existence of HGFs. However, for Littunen and Tohmo (2003), the firm´s location is not a determining factor.

Coad et al. (2014) indicate that most of the research on HGFs is static. This has implications for economic policy, which needs assurances that growth at t will remain at t+1; if this cannot be sustained over time, policies that stimulate growth are irrelevant (Daunfeldt and Halvarsson, 2015). However, the empirical evidence in this regard is not conclusive. Hölzl (2014), Mogos et al. (2021) and Erhardt (2021) find that the persistence of high growth depends on the choice of its definition.

Following this review of the literature, we conclude that there are no irrefutable results on the factors that determine a firm to be HGF.

3. Methodology and data

3.1 Methodology

In order to determine the factors that characterise a firm as HGF, a dynamic multivariate binary choice model is proposed, where the probability of being an HGF is a function of the individual characteristics of the firms and the environment in which they operate.

From the utility theory approach, an economic agent, with a rational behaviour, would choose between two exclusive alternatives, maximising the expected utility. The ith agent would choose option 1 if his utility, Ui1, is greater than the utility given by option 0, Ui0. This comparison, from the mathematical point of view, can be expressed through the following probabilistic inequality, Prob(Ui1>Ui0).

The way to quantify utility is by assigning a probability to each of the rational decisions using the following equation:

(2)Pi=Prob(Yi=1)=Prob(Ui1>Ui0)=F(Xijβ)

where F(Xijβ) is the distribution function evaluated for the j characteristics associated with firm i. The vector of the characteristics [3] is denoted by (Xij), while (β) is the vector of coefficients. Thus, the modelling of the choice of firms can be done through the following behaviour equation or binary choice model:

(3)Yi=F(Xijβ)+εi

where Yi is a dichotomous variable that takes the value of 1 if firm i is an HGF and 0 if it is not.

After estimating the discrete choice model, the non-linear decomposition proposed by Yun (2004) is applied to examine the difference in the probability of being an HGF between micro and non-micro firms. The original approach of the method of decomposition, proposed by Blinder (1973) and Oaxaca (1973), makes it possible to determine what part of the difference in probability corresponds to each one of the two components considered by these authors. The first component collects the difference in the explanatory variables observed between the two groups, and the second component collects the difference in the unobservable characteristics, quantified by the discrepancy in the response to the explanatory variables of both groups. The Blinder–Oaxaca method was intended for linear models but was later generalised for non-linear models (Yun, 2004).

If, as indicated, the probability of being an HGF for a firm i, Pi, is given by:

(4)Pi=F(Xijβ)

the breakdown of the difference in the probability of being an HGF between non-micro firms (L) and micro firms (S) can be obtained by:

(5)PL¯−PS¯=(F(XLβˆS)¯−F(XSβˆS)¯)+(F(XLβˆL)¯−F(XLβˆS)¯)

The first addend includes the difference explained by the different characteristics of each group, given the same coefficients. The second addend indicates the unexplained difference, i.e. for the same characteristics, the part corresponding to the different responses of the two groups.

3.2 Database and variables

This paper uses a sample of firms from five European countries: Germany, France, the United Kingdom, Portugal and Spain. From the Orbis database, information on active firms from these five European countries is selected for the three-year window from 2016 to 2019 [4]. There is a sample of 602,604 firms, classified in the following sectors: industry, energy, construction, commerce, transport, hospitality, communications, business services, education, health and other services.

In the literature, HGFs are considered to initially have at least ten employees. In Spain and Portugal, where the size of the firms is small, the selection of the sample can have significant consequences, whereby a particular sector with smaller firms, such as services, is underrepresented. For this reason, this study also considers firms with fewer than ten employees. The restriction imposed by the definition of HGF may lead to sample bias towards medium- and large-sized firms, and the results could not then be extrapolated to the business population. However, to obtain the sample, firms were required to provide information on the number of employees for the years considered, which implies that the firms in the sample are survivors, excluding those less than 4 years old. It should be noted that around 50% of firms disappear within their first 3–4 years (Coad, 2018). Our imposition, together with the inclusion criteria and data verification by Orbis (Moody’s, 2024), mean that young firms are underrepresented in the sample. Despite this, if the criterion of considering only firms with more than 10 employees at the beginning of the sample were established, as required by the OECD-Eurostat definition of HGF, young and small firms would be even further underrepresented.

The dependent variable is a binary variable that takes value 1 if the firm has fast growth and 0 otherwise. The HGFs are those that are in the 10% percentile of the highest value of the BS index. The size of the firm, as an explanatory variable, is quantified by the logarithm of the number of employees, in line with that proposed by Giner et al. (2016) and Moschella et al. (2019). To check whether micro firms have a differential effect on the probability of being an HGF, a dummy variable is generated, which takes the value of 1 if it has less than ten employees and 0 otherwise.

Human capital is measured through a proxy variable such as worker productivity, which is calculated as the ratio between operating income and the number of employees. For the financial conditions of the firms, in this study, the debt ratio is considered, which is the quotient between the total debt and own resources. For Lopez-Garcia and Puente (2012), a high debt ratio could be considered an indicator of fewer financial restrictions. However, excessive debt could pose a risk of presenting solvency and financing problems in the future, which is why Lopez-Garcia and Puente (2012) also include the quadratic term of this variable among the explanatory variables.

The age of the company is quantified by the years that have elapsed since registration of the firm until the end of the sample period. As the literature also suggests, age squared is included as an explanatory variable to test whether age has a non-linear influence on the probability of being an HGF.

In line with the classification made by the National Institute of Statistics (INE) [5] of the sectors according to their technological level, four dummy variables are generated: non-technological, medium-high technology, high technology and cutting-edge technology.

Business profitability is also included as an explanatory factor, which is defined as the quotient between ordinary results before taxes and total assets (ROE). Location is accounted for using country dummies, with a dummy variable being generated for each of the five countries. Location is also considered to be included through the Human Capital Index (HCI) and the Global Competitiveness Index (GCI), published by the World Economic Forum (WEF). The HCI assesses the level of education, skills and employment of the population of each country (WEF, 2016). Likewise, the GCI is an indicator of the quality of public institutions (Schwab, 2016). Finally, as control variables, 11 sectoral dummies are included.

Table 1 shows the distribution of the firms in the sample according to their growth, as well as the descriptive statistics of their characteristics. Figure 1 shows the distribution of firms by country and according to the activity sector to which they belong.

From Table 1 [6], it can be deduced that HGFs are those with the highest profitability. The percentage of HGFs in the micro firm segment is lower than in firms with more than ten employees. Therefore, micro firms appear to be less likely to be HGFs than larger firms. Figure 1 shows that the sectors of business services, manufacturing and commerce predominate in the sample.

4. Results

In the binary choice model, the lagged endogenous variable is included as an explanatory variable, with the aim of checking whether there is persistence in episodes of high growth, the proposed model being a dynamic model. All explanatory variables used in the model are collected at the beginning of the considered period. In this way, the factors can be considered ex ante, avoiding possible sources of endogeneity (Giner et al., 2016; Rodrigues et al., 2021).

The results of the different estimated models are presented in Table 2. All the models are estimated in a manner consistent with the existence of heteroskedasticity in the sample. We consider both the logistic and normal distribution functions, and the results show that the most appropriate specification is that of the Probit model, which is shown in Table 2.

First, it should be noted that the country dummy variables, which collect the location of the firms, and the GCI and HCI variables are highly correlated, so they cannot be included together. For this reason, Model 1 is estimated, including the dummy variables of the countries. Focusing on Model 1, in Table 3, the results indicate a positive and high persistence in the probability that a firm experiences high growth, as evidenced by Lopez-Garcia and Puente (2012), Hölzl (2014) and El-Dardiry and Vogt (2023). Erhardt (2021) also shows persistence in high growth among surviving firms, but when both surviving and exiting firms are considered, high growth does not persist.

In the sample, 65.41% of the firms that present high growth in 2019 also show high growth in 2018. By segment, 74.20% of the firms with more than ten employees had high growth in 2018 and 2019, compared to 43.04% of the micro firms. The coefficient of the micro firm variable indicates that there is a negative differential in the probability of a micro firm becoming HGF compared to larger firms. Likewise, the probability of becoming HGF is determined by the following characteristics of firms: age, profits, productivity, business size and technological level.

The age of the company has a non-linear U-shaped effect, so that the youngest and oldest firms are the ones with the highest probability of being an HGF. The parameter of the age variable is statistically significant, and its negative sign indicates that as age increases, the probability of becoming HGF decreases. Thus, it becomes clear that younger firms are more likely to be HGFs. Likewise, the positive sign of the squared age coefficient indicates that the probability of being an HGF decreases with age but at an ever-smaller rate. Both the size of the firm and the productivity of the employees positively affect the probability of being an HGF. Debt is not statistically significant. Regarding profits, the results indicate that the more profitable the firms are, from an economic point of view, the greater the probability of being an HGF.

The estimated parameters of the variables that reflect the technological level of the firms lead us to conclude that the probability of being an HGF increases with technological level. With the reference category being that of sectors with cutting-edge technology, the non-technological and medium technological sectors show negative differential effects in comparison, while the sectors with high technology do not present differences in the probability of being an HGF in firms with the latest technology. Finally, in relation to the productive sector, all sectors, without exception, present a positive differential effect with respect to the reference category, which is the commerce sector. This result indicates that firms in the commerce sector have a lower probability of being an HGF than any other firm belonging to other sectors. Firms with the greatest differential effect are those in the construction sector.

With Model 1, and for the mean values of the variables, both for micro firms and non-micro firms, a difference in probability of 0.22 points is obtained, which means that micro firms present a lower probability of being an HGF by 22% points than large firms.

In Model 2, the dummy country variables have been replaced by the GCI and HCI variables. As can be seen, the goodness of fit is the same as in Model 1. This model leads us to conclude that external factors, related to the location of the firms, have a positive impact on the probability that the firms are high-growth, corroborating the results obtained in the empirical literature (Acs and Mueller, 2008; Bravo-Biosca et al., 2013; Li et al., 2016).

The Probit model is also estimated for each of the considered firm segments (Table 2, Models 3 and 4) in order to compare their results. From the results obtained, some significant differences are observed in the coefficients of the factors in the two groups of firms. To compare the results, the marginal and differential effects of the factors on the average firm of each of the two segments are calculated. The results are shown in Table 3, and from these, it can be concluded, firstly, that inertia is much greater in large firms than in small ones. Secondly, although age continues to have a non-linear U-shaped effect in both segments, the effect is greater in micro firms, although it decreases at a higher rate than in larger firms. The size variable is not significant in small firms, while in larger ones it positively affects the probability of being an HGF. The marginal effects of both productivity and profitability are quite similar between the two groups of firms. Debt in micro firms has an inverted U-shaped effect. However, in larger firms it has a U-shaped effect, and according to Lopez-García and Puente (2012), “this effect is possibly due to credit not being the main source of external finance of HGFs, but risk capital or, as has been found elsewhere, internal finance from the group to which the high-growth firm belongs”. This difference in the effect is the cause of the debt variable not being significant for the total sample. Regarding the technological level, in both cases, the probability of being an HGF increases with technological level, but in micro firms the differential effect of non-technological ones, with respect to the reference category, is greater than in larger firms.

Finally, regarding sectors, among micro firms, hospitality and construction are the sectors with the highest probability of being an HGF. In firms with more than ten employees, the differential effects with respect to the reference category are smaller and the sectors with the highest probability are construction and business services.

In conclusion, it should be noted that HGFs have high persistence and younger firms are more likely to be HGFs. Likewise, company size, labour productivity, profitability and technological level positively affect the probability of being an HGF. Debt has a positive influence on micro firms, while it is a factor that reduces the probability of larger firms being an HGF. Micro firms show a lower probability of presenting high growth than larger firms. Finally, the retail sector has the least probability of being an HGF, while firms in the construction sector show the highest probability.

To identify the factors that explain the difference in the probability of being an HGF between the two groups of firms, the Blinder–Oaxaca decomposition is applied. The results obtained are shown in Table 4. The difference in the probability of being an HGF between the two segments is 21.0% points, in favour of firms with more than ten employees, a similar figure to that calculated with Model 1. The explained component indicates that if micro firms had the characteristics of larger firms, the probability of being an HGF would increase by 21.5% points. The unexplained component is not statistically significant at 5% significance and, therefore, is null. Thus, the difference in the probability of being an HGF, between the two segments of firms, is explained solely by the characteristic component. When analysing the contribution of quantitative factors to the gap in probability between the two segments of firms, size is seen to be the factor that mainly increases the probability of being an HGF in micro firms. Size is followed by persistence, while the remaining variables either decrease it, such as age, or have no effect at all. Consequently, firm size emerges as the key variable explaining the difference in the probability of being an HGF between the two segments of firms.

5. Conclusions

This study uses an extensive sample of firms from five European countries with the aim of obtaining conclusive results on the determinants that drive a firm to be HGF. Likewise, this sample includes the segment of firms with fewer than ten employees, a category not normally considered in the literature. Exclusion of this segment of firms would mean, with our sample, focusing only on 30% of the data and neglecting the remaining 70%. There are several reasons for using this expanded sample: by not excluding an important part of the business fabric, the results can be extrapolated to the business population; the study can show if micro firms also present high growth and if there are differences in the factors that determine the probability of being an HGF between both segments of firms.

The results lead us to conclude that economic and business policies aimed at fostering HGFs are likely to be sustained over time, given the evident high inertia in the probability of maintaining high growth. Younger firms are more likely to be HGFs; therefore, measures that promote entrepreneurship are essential. Likewise, the size of the firm, labour productivity, profitability and technological level have a positive influence on the probability of being an HGF. Therefore, the appropriate strategies to promote the existence of HGFs are those aimed at improving the productivity of employees, increasing the technological level of firms and fostering growth in business size.

As has been verified, micro firms have a lower probability of being an HGF than larger firms, and the fundamental factor that explains this difference is size. Therefore, policies aimed at increasing the size of a firm would improve the percentage of HGFs in the economy. In general terms, the factors that determine the probability of being an HGF are the same in micro firms as in larger firms, although there are some differences in their effects. For example, inertia is higher in larger firms than in micro firms. Age has a non-linear U-shaped effect in both segments, but the effect is greater and decreases at a faster rate in micro firms than in large firms. Debt in micro firms has an inverted U-shaped effect, while it has a U-shaped effect on large firms. Finally, the probability of being an HGF increases with the level of technology, but its effect is greater in micro firms.

The empirical results show that external factors, related to the location of firms, also determine the existence of HGFs. Therefore, the authorities should also promote the development of capital, both tangible and intangible, in areas where firms are located, as well as guarantee macroeconomic stability and the proper functioning and development of labour and financial markets.

Firms have various business policies that can foster growth. Based on the results, notable strategies include boosting labour productivity through training and incentive schemes, investing in digitalisation and technology, diversifying products or services and implementing marketing campaigns to attract new customers and expand market share. As a future line of research, we envisage breaking down the determinants of HGFs by sector in order to identify effective sectoral strategies.

Figures

Figure 1

Distribution of HGFs by country and productive sector

Table 1

Average values and distribution of firms according to their growth

	Firms		Age			Employees			Productivity			Debt			ROE (%)
	Number	%	Mean	Median	Std. Dev.	Mean	Median	Std. Dev.	Mean	Median	Std. Dev.	Mean	Median	Std. Dev.	Mean	Median	Std. Dev.
1. HGFs
Fewer than 10 employees	18,068	3.00	11	8	10	4	3	3	761.80	106.72	32,626.24	6.13	2.00	29.63	27.64	22.94	83.14
Over 10 employees	46,045	7.64	26	21	22	673	91	7,668	325.98	142.94	2,252.55	4.48	1.47	41.96	20.81	15.74	69.86
2. Mice	53,238	8.83	22	21	13	14	14	3	262.09	88.59	6,365.36	3.44	1.06	25.05	13.43	9.96	61.94
3. Elephant	3,814	0.63	40	29	33	3,847	998	19,676	335.67	150.90	2,120.07	6.19	1.57	82.46	9.75	8.78	84.46
4. Rest of firms
Fewer than 10 employees	402,893	66.86	17	16	11	4	3	3	243.02	77.88	3,882.13	4.23	0.98	38.08	13.18	9.06	73.81
Over 10 employees	78,546	13.03	29	25	19	75	43	81	299.10	127.25	1,431.61	4.39	1.12	67.13	13.30	10.11	69.33
Total of firms	602,604	100.00	20	18	15	89	5	2,658	274.53	88.24	6,803.83	4.27	1.08	42.62	14.22	10.05	72.41

Note(s): Corresponds to data for the year 2016. Turnover in thousands of euros

Source(s): Compiled by the authors from Orbis

Table 2

Probit model estimation

	Total sample				Fewer than 10 employees		Over 10 employees
	Model 1		Model 2		Model 3		Model 4
Variable	Coefficient	p-value	Coefficient	p-value	Coefficient	p-value	Coefficient	p-value
Constant	−2.263	0.000	−3.284	0.000	−1.721	0.000	−2.682	0.000
Persistence (Y in t−1)	1.596	0.000	1.594	0.000	1.707	0.000	1.531	0.000
Firms’ characteristics
Micro firms (1)	−0.100	0.000	−0.097	0.000	–	–	–	–
Age	−0.020	0.000	−0.020	0.000	−0.044	0.000	−0.010	0.000
Age squared	1.15E−04	0.000	1.16E−04	0.000	3.62E−04	0.000	4.90E−05	0.000
Size	0.201	0.000	0.202	0.000	0.007	0.268	0.265	0.000
Productivity	0.127	0.000	0.130	0.000	0.130	0.000	0.114	0.000
Debt	0.004	0.475	0.051	0.398	0.161	0.000	−0.021	0.026
Debt squared	−2.43E−05	0.598	−2.86E−05	0.536	−0.011	0.014	1.54E−04	0.010
ROE	0.069	0.000	0.068	0.000	0.060	0.000	0.058	0.000
Technology (2)
Non-technological	−0.247	0.000	−0.248	0.000	−0.398	0.000	−0.163	0.000
Medium-high	−0.210	0.000	−0.207	0.000	−0.443	0.000	−0.109	0.003
High	−0.053	0.227	−0.050	0.258	−0.250	0.014	0.028	0.582
Productive sector (3)
Manufacturing	0.177	0.000	0.177	0.000	0.296	0.000	0.085	0.000
Energy	0.109	0.000	0.120	0.000	0.212	0.000	0.042	0.169
Construction	0.358	0.000	0.360	0.000	0.438	0.000	0.258	0.000
Transportation	0.288	0.000	0.291	0.000	0.336	0.000	0.218	0.000
Hospitality	0.272	0.000	0.272	0.000	0.458	0.000	0.088	0.000
Communications	0.162	0.000	0.160	0.000	0.060	0.293	0.172	0.000
Business services	0.207	0.000	0.207	0.000	0.142	0.000	0.252	0.000
Education	0.185	0.000	0.172	0.000	0.385	0.000	0.112	0.000
Health	0.203	0.000	0.200	0.000	0.103	0.000	0.229	0.000
Other services	0.234	0.000	0.228	0.000	0.226	0.000	0.228	0.000
Localisation
GCI	–	–	0.049	0.075	–	–	–	–
HCI	–	–	0.011	0.006	–	–	–	–
Control variable
Country dummies	Yes		No		Yes		Yes
Number of observations	602,604		602,604		420,961		181,643
Log likelihood	−1,26,066.20		−1,26,100.80		−55,568.78		−68,095.20
McFadden R-squared	0.38		0.38		0.25		0.34
Prediction R-squared	92.13		92.15		95.70		84.00

Note(s): The endogenous variable is a dichotomous variable that takes the value of 1 if it is a high-growth firm and 0 otherwise

Reference categories: (1) Firms with over ten employees, (2) High technology and (3) Commerce The productivity variable is logarithmic

Source(s): Compiled by the authors

Table 3

Marginal and differential effects in percentage points

		Fewer than 10	Over 10
Variable	Total sample	Employees	Employees
Persistence (Y in t−1)	44.07	44.18	48.56
Firms’ characteristics
Micro firms	3.97	–	–
Age	−0.78	−1.74	−0.39
Age squared	0.00	0.01	0.00
Size	8.03	0.27	10.36
Productivity	5.07	5.20	4.45
Debt	0.17	6.42	−0.83
Debt squared	0.00	−0.45	0.01
ROE	2.74	2.40	2.26
Technology
Non-technological	−7.31	−15.32	−6.40
Medium-high	−6.30	−16.91	−4.29
High	−1.69	−9.81	1.09
Productive sector
Manufacturing	7.05	11.69	3.34
Energy	4.33	8.42	1.63
Construction	14.08	17.08	10.10
Transportation	11.41	13.22	8.52
Hospitality	10.76	17.80	3.44
Communications	6.44	2.40	6.74
Business services	8.22	5.65	9.89
Education	7.37	15.11	4.38
Health	8.05	4.12	8.98
Other services	9.30	8.97	8.91

Source(s): Compiled by the authors from models of Table 2

Table 4

Blinder–Oaxaca decomposition by company size

	Coefficient	p-value
Over 10 employees	0.253	0.000
Fewer than 10 employees	0.043	0.000
Difference	0.210	0.000
Explained component	0.215	0.000
Unexplained component	−0.005	0.053
Persistence (Y in t−1)	0.074	0.000
Firms’ characteristics
Age	−0.018	0.000
Age squared	0.006	0.000
Size	0.145	0.000
Productivity	0.009	0.000
Debt	0.000	0.299
Debt squared	0.000	0.302
ROE	0.000	0.000

Source(s): Compiled by the authors

Notes

1. Within these, the gazelles would be the companies characterised as being less than five years old.

2. Gibrat (1931).

3. In this paper, a dynamic model is proposed, which means the agent’s decision at t−1 is included within the characteristics vector, i.e. the variable Yi at t−1. For simplicity, the time subscript has not been included in the model.

4. For business growth to be calculated over a three-year window, data must be collected for four years. In this case, information is collected on the number of employees in each firm for the years 2016, 2017, 2018 and 2019.

5. See https://www.ine.es/daco/daco43/notaiat.pdf

6. The high dispersion of some variables, such as productivity, makes the median a better measure of position than the mean.