Investigating the Impact of Small Versus Large Firms on Economic Performance of Countries and Industries

Abstract

Following earlier work by Audretsch et al. (2002), we assume that an optimal size-class structure exists, in terms of achieving maximal economic growth rates. Such an optimal structure is likely to exist, as economies need a balance between the core competences of large firms (such as exploitation of economies of scale) and those of smaller firms (such as flexibility and exploration of new ideas). Accordingly, changes in size-class structure (i.e., changes in the relative shares in economic activity accounted for by micro, small, medium-sized and large firms) may affect macro-economic growth. Using a unique database of the EU-27 countries for the period 2002–2008 for five broad sectors of economic activity and four size-classes, we find empirical support which suggests that, on average for these countries over this period, the share of micro and large firms may have been ‘above optimum’ (particularly in lower income EU countries), whereas the share of medium-sized firms may have been ‘below optimum’ (particularly in higher income EU countries). This evidence suggests that the transition from a ‘managed’ to an ‘entrepreneurial’ economy (Audretsch and Thurik 2001) has not yet been completed in all countries of the EU-27.

Appendices

Appendix 1: The Audretsch et al. (2002) Model

In this appendix we show the derivation of the Audretsch et al. (2002) model. The derivation is taken directly from their article (Audretsch et al. 2002, 88–90):

“We test the hypothesis that the extent of the gap between the actual industry structure and the optimal industry structure influences subsequent growth. We start with the assumption that a country’s growth can be decomposed into two components: (i) growth that would have occurred with an optimal industry structure, and (ii) the impact on growth occurring from any actual deviations from that optimal industry structure. This can be represented by

$$ \Delta GN{P}_{cp}=\Delta GN{P}_{cp}^{*}-\gamma \left|SF{P}_{cp-1}-SF{P}_c^{*}\right|, $$

(11)

where the dependent variable is the actual rate of economic growth. ΔGNP ^* _cp is the rate of economic growth in country c in the case where the actual industry structure, summarized by small firm presence (SFP _cp ), is at the optimal level at the start of the period p. For ease of exposition we assume that the optimal industry structure in a country remains constant for the total period under investigation. This is not vital to our analysis. Since we are considering only short-term periods, this may be a reasonable assumption.

Industry structure is multidimensional and spans a broad array of characteristics that defy measurement by a single statistic. However, as explained elsewhere (Audretsch and Thurik 2000, 2001), the most salient characteristic driving the shift in industry structure from the managed to the entrepreneurial economy is that the relative role of small and entrepreneurial firms has increased. Thus, we capture changes in industry structures by changes in the relative importance of small firms.

In Eq. (11) the parameter γ is positive. Deviations of the actual industry structure from the optimal industry structure negatively affect economic growth, both when the industry structure consists of too few or too many small firms. In either case there is a deviation from the optimal industry structure and number of small firms. Taking the first difference of Eq. (11) we obtain

$$ \Delta GN{P}_{cp}=\Delta GN{P}_{cp-1}+\Delta \Delta GN{P}_{cp}^{*}-\gamma \left(\;\left|SF{P}_{cp-1}-SF{P}_c^{*}\right|-\left|SF{P}_{cp-2}-SF{P}_c^{*}\right|\right)\;. $$

(12)

In case both $SF{P}_{cp-1}$ and $SF{P}_{cp-2}$ are above the optimal small-firm share, the expression between brackets reduces to $\Delta SF{P}_{cp-1}$. Indeed, in case the small-firm share is too high, adding small firms to the industry structure reduces economic growth. In case both $SF{P}_{cp-1}$ and $SF{P}_{cp-2}$ are below the optimal small-firm share, the expression between brackets reduces to $-\Delta SF{P}_{cp-1}$. An increase in the small firm share when this presence is below optimal enhances economic performance. Therefore, the sign of the parameter of $\Delta SF{P}_{cp-1}$ reflects whether the small firm presence is below or above the optimal levels for the countries under consideration. In case the parameter is negative, the industry structure consists of too many small firms. In case the parameter is positive, the reverse holds and the industry structure consists of too few small firms.

We will denote the parameter of $\Delta\;SF{P}_{cp-1}$ as κ. Note that this is not the same parameter as γ, since the sign of κ is dependent on whether the actual small-firm share is above or below the optimal one. So, κ can be both positive and negative whereas γ is necessarily positive.

We make some further assumptions to transform Eq. (12) into an equation that can be estimated using the data at hand. First, we approximate $\Delta SF{P}_{cp-1}$ by $\Delta S{F}_{cp-1}-\Delta L{F}_{cp-1}$, the difference between the growth of small firms and large firms in terms of value-of-shipments. Second, we assume that ΔGNP ^* _cp is idiosyncratic with respect to time and country. Therefore country dummies and time dummies (the last to correct for European wide business cycle effects) are included. Thus, ΔΔGNP ^* _cp is approximated by time dummies only because the country dummies drop out when taking first differences. Third, we add an error term e _cp . Summarizing we have

$$ \Delta GN{P}_{cp}=\Delta GN{P}_{cp-1}+{\displaystyle \sum_{p=1}^P{\beta}_p{D}_p+}\kappa \left(\Delta S{F}_{cp-1}-\Delta L{F}_{cp-1}\right)+{e}_{cp}, $$

(13)

where D _p denote dummy variables for periods $p=1,\dots, P$. Factors specific to each time period are reflected by β _p . A high value of this parameter indicates an unexplained increase in the extent of economic growth. In case of a low β _p the reverse holds. The contribution of the shift in the size class distribution of firms to the percentage growth of GNP is represented by κ.”

Note that in the present paper we also have data at sector level. Accordingly, we assume that ΔGNP ^* _cp is idiosyncratic with respect to time, country and sector. However, similar to the country dummies, sectoral dummies drop out when taking first differences of Eq. (11), hence ΔΔGNP ^* _cp is approximated by time dummies only.

Appendix 2: Classification by Economic Development Level

In this appendix we provide a classification of countries based on their GNI per capita in 2005 (Table 3).

Table 3 EU-27 countries, by economic development level, 2005

Appendix 3: Robustness Test: Correcting for (the Possibility of) Reversed Causality

Table 4 presents the results of the robustness test described in Sect. 4.1. Independent variables are cleared from (contemporaneous) business cycle influences.

Table 4 Regression results Eqs. (3) and (5), correcting for reversed causality^a,b,c

Appendix 4: Correlation Matrixes by Economic Development Level

In this appendix we provide the correlation matrixes by level of economic development (Tables 5, 6, 7). The strong significant and negative correlation between the net growth of the share of large firms and the net growth of the share of SMEs is due to the definitions of the variables (see Sect. 2.1). Notice, however, that we include the net growth of the share of Small and Medium-sized Enterprises approximated by the annual percentage growth of real sales by SMEs minus the annual percentage growth of real sales by large firms and the net growth rates of the shares of micro, small, medium-sized and large firms in two different specifications.

Table 5 Correlation matrix for lower developed countries

Table 6 Correlation matrix for higher developed countries

Table 7 Correlation matrix for the general sample