The geography of new business formation in the Czech Republic : a cross-section analysis ( 2003-2014 )

The intent of this paper is to provide empirical evidence of determinants on new business formation in the Czech Republic's regions in the period 2003 to 2014. The results suggest that the regions with strong entrepreneurial culture, measured by a high proportion of entrepreneurs in the population, and with a strong presence of foreigners, had higher rates of new business formation. A positive inuence of agglomeration economies was observed, and the impact of unemployment was negative. However, the results were not consistent when more advanced legal business forms (especially limited liability and joint-stock companies) had been used in the dependent variable. In this case, human capital was the overriding factor determining regional differences in new business formation rates, while entrepreneurial culture and the presence of foreigners were not signicant. Finally, the importance of the industrial structure of regions was conrmed regardless of the dependent variable form.


Introduction
The end of communism in the late 1980s and early 1990s meant that central planning was gradually replaced by market mechanisms.The associated pro-170 Jiří Novosák et al.
The remainder of this paper proceeds as follows.The next section reviews the theoretical background for determinants of new business formation.The methods used in this paper are introduced in the section following this review.Then the empirical results are presented and discussed.In the last section, the authors draw conclusions.

Theoretical framework
The regional determinants of new business formation have been researched using a number of theoretical bases.Delfmann et al. (2014) describe a general theoretical framework that integrates demand-side and supply-side factors, and also the institutional environment, for explaining regional differences in new business formation.The demand-side factors relate to entrepreneurial opportunities, whereas the supply-side factors relate to people's resources, traits and attitudes to entrepreneurship, and the institutional environment relates to sociocultural aspects of entrepreneurship.Moreover, new business formation rates differ regarding industries because entrepreneurial opportunities and barriers are not the same for all industries (Fotopoulos, 2014;Wyrwich, 2012).Altogether, this general theoretical framework may be used for dening the determinants of new business formation.
Agglomeration economies have frequently been examined as a predictive demand-side factor for new business formation.Three types of agglomeration economies are generally considered: (a) the availability of the pooled labour force; (b) non-pecuniary externalities; and (c) knowledge and information The geography of new business 171 spillovers (Lafuente & Vaillant, 2010;Armington & Acs, 2002).New businesses locate themselves in order to take advantage of agglomeration economies (Audretsch & Fritsch, 1994) that have been measured in a number of different ways, including population density (Bishop, 2012;Audretsch & Fritsch, 1994;Lafuente & Vaillant, 2010) and the quality of human capital (Wyrwich, 2012;Lee, Florida & Acs, 2004;Armington & Acs, 2002).Human capital is also regarded to be a supply-side factor of new business formation because it is claimed that better-educated people are more equipped to identify and exploit entrepreneurial opportunities (Fotopoulos, 2014;Lee, Florida & Acs, 2004).
The theory of knowledge spillovers closely relates to demand-side factors of new business formation.The essence of this theory is that new knowledge is not fully exploited by the creators (Bishop, 2012).Subsequently, 'unused' knowledge is the source of entrepreneurial opportunities.It is claimed that the greater the stock of local knowledge, the more entrepreneurial opportunities may arise and be exploited, resulting in high new business formation rates (Bishop, 2012;Lee, Florida & Acs, 2004).In this regard, spatial proximity and the quality of human capital that enhance knowledge spillovers are essential elements of the theory (Fotopoulos, 2014;Bishop, 2012).The stock of local knowledge has been empirically measured in different ways, including patent activities (Lee, Florida & Acs, 2004).
Unemployment has been another frequently examined determinant of new business formation.Two contradictory views are commonly held in existing literature regarding the inuence of unemployment on new business formation.On the one hand, unemployment means less entrepreneurial opportunities because it constrains demand, negatively inuencing new business formation rates (Delfmann et al., 2014;Lafuente & Vaillant, 2010).On the other hand, unemployment increases the supply of potential entrepreneurs, positively inuencing new business formation rates (Bishop, 2012;Armington & Acs, 2002;Audretsch & Fritsch, 1994).Therefore, the nature of unemployment's inuence on new business formation is not apriori clear (Fotopoulos, 2014;Bishop, 2012;Wyrwich, 2012).
The higher propensity of foreigners establishing their businesses is another supply-side factor of new business formation.Lee, Florida & Acs (2004), Delfmann et al. (2014) explain this by noting that: (a) foreigners are disadvantaged in the labour market, owing to low education and language barriers; (b) foreigners are more willing to be risk-takers; and (c) foreigners bring with them their own culture, skills and lifestyle.Consequently, higher proportions of foreigners in the population are expected to positively inuence new business formation rates.
A number of recent studies have shown the persistence of new business formation (Fotopoulos, 2014;Anderson & Koster, 2011;Fritsch & Mueller, 2007).Several mechanisms have been proposed in the research literature for explaining this effect, including agglomeration economies and externalities, selfreinforcing processes, and entrepreneurial culture.Therefore, high rates of new business formation in the past tend to positively inuence high rates of new business formation in the future (Anderson & Koster, 2011;Fotopoulos, 2014).
The regional determinants of new business formation have been mostly examined in 'old' capitalist countries.Research in post-communist contexts has been less frequent due to, among others reasons, the lack of data (Lafuente & Vaillant, 2010).However, the existing studies point out some specic features of the early phase of post-communist transformation.These features include the little effect by agglomeration economies (Wyrwich, 2012;Lafuente & Vaillant, 2010), the strong inuence of the need to 'push' entrepreneurship through unemployment (Fritsch et al., 2014;Earle & Sakova, 2000) and the strong inuence of the quality of human capital (Fritsch et al., 2014;Smallbone & Welter, 2006) on new business formation rates.
This paper adds to the limited literature on the regional determinants of new business formation in post-communist countries by examining their inuence on new business formation rates in the Czech Republic (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014).The period is, apart from partially altered regional determinants, the main continuation of the research carried out by Hájek, Nekolová and Novosák (2015) in which they only dealt with the period 2011-2012.Additionally, new business is dened by considering two forms of businesses -natural persons and business corporations as legally more advanced business entities.

Methodology
The methodology of this paper is based on regression analysis in which the dependent variable is new business formation rates, and the explanatory variables relate to theoretically meaningful regional variables.All the variables refer to the administrative districts of municipalities with extended power (205 units in total), and to the territory of the capital city of Prague (hereafter referred to as regions).The regions are administrative-territorial units, corresponding to the areas between level LAU1 and LAU2.
In this research, a new business is identied from each new identication number registered in the Business Register of the Czech Republic (hereafter referred to as BR CR) in the period 2003-2014, and the end of year state is considered of each year between 2003 and 2014.Generally, the two main legal forms of new businesses in the Czech Republic are natural persons (or 'entrepreneurs'), and also legally more advanced business entities (hereafter referred to as 'business corporations').The latter especially includes limited liability companies and joint-stock companies.Two forms of the dependent variable are employed to reveal the specics of business corporations: (1) all new businesses; and (2) business corporations only.Moreover, the labour market approach is used to standardise the number of new businesses to be independent of regional size (Armington & Acs, 2002;Audretsch & Fritsch, 1994).Hence, the number of new businesses in a region, in either of the forms, is divided by the number of people in the productive age groups.Finally, the dependent variable is logtransformed to approximate the normality assumption of the residuals more closely.
The explanatory variables are derived from the theoretical framework outlined in the preceding section, and dated from 2005 or earlier, minimising endogeneity issues (Bishop, 2012).Three explanatory variables relate to agglomeration economies and the theory of knowledge spillovers: (1) population density, measured as inhabitants per square kilometre in 2003 (DENSITY); (2) The geography of new business 173 the quality of human capital measured as the mean of the years of schooling of the region's population aged 15 and over in 2001 (HUMAN CAPITAL); and (3) the stock of local knowledge measured as the number of patents and utility models per population aged 15-64, and calculated as the mean of the 2002-2005 rates (KNOWLEDGE).The Czech Statistical Ofce (hereafter referred to as CSO) is the source of data for the DENSITY and HUMAN CAPITAL variables, while data about patents is taken from the Industrial Property Ofce (hereafter referred to as IPO).
Two explanatory variables relate to the supply-side factors of new business formation.Firstly, unemployment is the number of registered unemployed people in the region's population aged 15-64, calculated as the mean of the 2003-2005 rates (UNEMPLOYMENT).CSO is the source of data.Secondly, the ratio of foreigners to the number of people in the productive age groups is also calculated as the mean of the 2003-2005 rates (FOREIGNERS).Data on foreigners and the number of people in the productive age groups are taken from the Ministry of Interior (hereafter referred to as MoI) and CSO respectively.
The institutional factors of entrepreneurial culture are proxied by entrepreneurial activity in a particular region at the beginning of the 2003-2014 period.It is measured as the share of employers and self-employed people in the economically active population in 2001 (ENTREPRENEURIAL CULTURE), assuming that high entrepreneurial activity is associated with favourable entrepreneurial culture.CSO is the source of data.
Finally, four explanatory variables are added to check differences in the industrial structure of various regions.These variables are derived by using factor analysis on the CSO data for the proportion of people employed in 9 industries and 206 regions, in 2001.In this regard, three variables are directly the share of employment in the economically active population, in 2001, in the sectors (1) agriculture, forestry and sheries (PRIMARY SECTOR); (2) construction (CONSTRUCTION); and (3) accommodation and food service activities (TOURISM).These variables were excluded from the factor analysis because they had a low value of the measure of sampling adequacy.The remaining six industries were entered into the principal component analysis, extracting an only factor, with positive loadings on nancial services, business services, wholesale and retail trade, public services, transportation, and with negative loadings on manufacturing.Therefore, the factor considers the importance of manufacturing and services in a regional economy.Note that the resulting factor scores (MANUFACTURING-SERVICE) are used in subsequent analyses, and the reliability of the factor is adequate as shown by the coefcient 'alpha'.
Correlation and regression analyses are used to identify regional determinants associated with the rates for new business formation.Firstly, the regression models are estimated by applying linear least squares (OLS) with White's correction for heteroscedasticity.However, the presence of spatial autocorrelation makes the OLS estimates inconsistent and biased.The generalised method of moments (GMM) spatial error models with heteroscedasticity robust standard errors are, therefore, estimated based on the results of the Lagrange Multiplier (LM) tests as suggested by Anselin et al. (1996) and others.Note that the other usual regression assumptions were checked by performing residual diagnostics on the models, and a violation was not detected.

Empirical results
Figures 1 and 2 provide the rst insights into the spatial distribution of new business formation rates.Three ndings are noteworthy here.Firstly, the spatial distribution of new business formation rates is not accidental since both -spatial clusters of high values of new business formation rates and also spatial clusters of low values of new business formation rates -are present.The presence of signicant spatial autocorrelation is also conrmed by Moran's I statistic.This is 0.380 for all businesses and 0.224 specically for business corporations only.Hence, spatial clustering of the regions sharing similar new business formation rates is suggested.Secondly, the level of new business formation is rather high in major urban areas and their surroundings, which indicates the importance of agglomeration economies for new business formation rates.Thirdly, the differences between the two forms of new business formation rates are illustrated in both gures, also suggesting that regional determinants may differ between them.Overall, the ndings substantiate the empirical strategy used in this paper.Table 1 adds to our knowledge base by exploring bivariate correlations between the explanatory variables and the new business formation rates.The three explanatory variables related to agglomeration economies and the theory of knowledge spillovers (DENSITY, HUMAN CAPITAL and KNOWLEDGE) are positively and signicantly correlated with new business formation rates.The positive and signicant correlation is also found between the FOREIGNERS and ENTREPRENEURIAL CULTURE variables, while it is negative for the UNEMPLOYMENT variable.Finally, high shares of employment in (1) construction (CONSTRUCTION); (2) accommodation and food service activities (TOURISM); and (3) service industries (MANUFACTURING-SERVICE) are positively associated with new business formation rates, while high shares of The geography of new business 175 employment in agriculture, forestry and sheries (PRIMARY SECTOR) are negatively associated with new business formation rates.The correlations are as expected according to the theory and the correlations are stable for the two forms of new business formation rates.However, some differences between these two forms may be observed regarding the strength of correlations.The three explanatory variables related to agglomeration economies and also related to the theory of knowledge spillovers (DENSITY, HUMAN CAPITAL and KNOWLEDGE), and the variables related to a more serviceoriented economy (MANUFACTURING-SERVICE, PRIMARY SECTOR) are more strongly associated with business corporations, but the remaining ve variables relate to all businesses.Hence, it seems that new business corporations are more likely to be established in major urban areas.

Discussion
The empirical results presented in the preceding section are further analysed by using regression analysis in this section.Besides the usual OLS cross-section estimation, spatial error models were performed to check for spatial dependence and also to make statistical estimates more robust.Table 2 provides the regression results, which reports standardised coefcients.These were chosen to compare the inuence of the explanatory variables on new business formation rates.The rst two models concern all businesses, while the latter two models concern only business corporations.The presence of foreigners in a region (FOREIGNERS) and their entrepreneurial culture (ENTREPRENEURIAL CULTURE) are the most inuential predictors of new business formation rates in the rst two models.However, the inuence of these two variables is not signicant in the latter two models.This means that the proportion of foreigners to the number of people in the productive age groups, and also the share of employers and self-employed people in the economically active population at the beginning of the 2003-2014 period, both have the strongest impacts on new business formation of all legal forms, but they do not have this strong impact on business corporations.Thus, the results support the theoretical assumptions that: (1) foreigners are more likely to The geography of new business 177 establish their own businesses; and (2) new businesses are more likely to be founded in an entrepreneurially rich environment; nevertheless entrepreneurs are engaged in their business only as natural persons, and they do not establish legally more complicated business forms.Thus, other variables are signicant for explaining spatial variation in new business corporation formation rates.The quality of human capital in a region (HUMAN CAPITAL) is the most inuential predictor of new business corporation formation rates.Better educated people are more equipped to deal with the processes required for establishing a new business corporation.However, this relationship is weakened and even insignicant in the spatial error model when considering the two models calculated with all new businesses in the dependent variable.It means that education is not signicant in determining whether or not a person will conduct business as a natural person (compare this stance with the hypothesis of the importance of human capital in Fritsch et al., 2014;Smallbone & Welter, 2006).
Agglomeration economies, proxied by population density (DENSITY), have positive inuences on new business formation rates, regardless of the dependent variable form.It is noteworthy that the variable is only signicant in the spatial error models at the 0.05 signicance level (compare this with a similar hypothesis of the low effects of agglomeration economies in Wyrwich, 2012;Lafuente & Vaillant, 2010).The store of local knowledge, measured as the number of patents and utility models per population aged 15-64 (KNOWLEDGE), is not signicant in either of the models.Local knowledge may be more important for certain industries than for overall evaluation.The unemployment rate (UNEM-PLOYMENT) has negative impact on new business formation rates, indicating that the demand-side inuence of unemployment is more important than the supply-side inuence (compare this with the hypothesis of the importance of necessity entrepreneurship 'pushed' by unemployment in Fritsch et al., 2014;Earle & Sakova, 2000).Nevertheless, the impact is signicant in the dependent variable only, in the two models for all new businesses.
Finally, table 2 conrms the importance of the industrial structure of regions for explaining the differences in new business formation rates.In this regard, a high share of employment in services (MANUFACTURING-SERVICE) is the strongest positive predictor of new business formation rates in all the models.Similarly, employment in accommodation and food service activities (TOUR-ISM) has a positive and signicant inuence on new business formation rates, regardless of the form of the dependent variable.On the contrary, a relatively large employment in the primary sector (PRIMARY SECTOR) is associated with a lower new business formation rate, but only for new business corporations.The opposite is true for employment in construction (CONSTRUCTION).Hence, a higher tendency is indicated to conduct business as a natural person in these two industries.

Conclusions
The goal of this paper was to identify regional determinants of new business formation in the Czech Republic in the period 2003-2014, contributing to knowledge in this eld in often neglected post-communist country situations.The results brought the following ndings to light.
New business formation varied considerably among the Czech Republic's regions during 2003 to 2014.Both, 'hot spots' and 'cold spots' were identied in this regard.There was a clear tendency for spatial clustering, evoking the question of which regional factors best explain the differences in new business formation rates.
Concerning all new businesses, the strongest predictors of new business formation rates were the proportion of foreigners in the productive age population, and also the proportion of employers and self-employed people in the economically active population at the beginning of the period 2003-2014.Hence, the regions with strong entrepreneurial culture and with a strong presence of foreigners had higher new business formation rates.These rates were further increased in the presence of agglomeration economies and when unemployment was low.Hence, the demand constraints stemming from unemployment were of higher importance than the increasing supply of potential entrepreneurs.Human capital, measured as the years of schooling, was not signicant for explaining regional differences in new business formation rates.
However, this scenario was not repeated when new business corporation formation rates were used as the dependent variable.It was shown that human capital, and not the presence of foreigners and strong entrepreneurial culture, was the decisive factor determining regional differences in new business corporation formation rates.Hence, the presence of foreigners and the number of entrepreneurs in a region led to legally less complicated business forms, whereas human capital was a signicant determinant for conducting business in the form of a limited liability or joint-stock company.The positive inuence of agglomeration economies was also noticed for business corporations.
The importance of regional industrial structure was conrmed, regardless of the dependent variable form.Hence, service-oriented regional economies had higher new business formation rates than manufacturing-oriented regional economies.Moreover, there was clear detection that a strong tourism sector had a positive impact, as also a strong construction sector had a positive impact, and a strong primary sector had a negative impact on new business formation.
Overall, the results support the theoretical principles about the determinants of new business formation, and the results provide the sign of the regression coefcients, as expected.Some specic characteristics of post-communist countries suggested in the literature were also observed.These characteristics particularly concern the effects of agglomeration economies, whose importance was reduced under communist governments.It is anticipated that the inuence of agglomeration economies will increase in future (Ezcurra, Pascual & Rapún, 2007;Nekolová, Novosák & Hájek, 2016), as suggested also by the work of Hájek, Nekolová & Novosák (2015) who found statistically signicant inuence of agglomeration economies on new business formation rates in the time period 2011-2012.On the contrary, the effects of entrepreneurial culture and the presence of foreigners on new business formation rates remain the same as in the work of Hájek, Nekolová & Novosák (2015); however, not for business corporations.Methodologically, these ndings clearly illustrate the importance of different legal business forms for determining the inuence of regional The geography of new business 179 determinants on new business formation rates and this is the main contribution of this paper to the previous work of Hájek, Nekolová & Novosák (2015).

Figure 1 .
Figure 1.New business formation rate -all businesses (2003-2014), Czech regions Source: own elaboration based on BR CR, and CSO

Figure 2 .
Figure 2. New business formation rate -business corporations (2003-2014), Czech regions Source: own elaboration based on BR CR, and CSO

Table 1 .
Pearson correlation coefficients ** significant at the 0.01 significance level; * significant at the 0.05 significance level Source: own elaboration based on BR CR, CSO, IPO, and MoI