Master Thesis Entrepreneurship in the European Regions - A Multi-Level Study -

(1)

Master Thesis

Entrepreneurship in the European Regions

A MultiLevel Study

-Johannes Kleinhempel 1

Supervisors: Prof. S. Beugelsdijk, dr. M.J. Klasing Rijksuniversiteit Groningen

Faculty of Economics and Business

Department of Global Economics and Management

September 7, 2016

(2)

Abstract

This paper empirically examines the influence of regional culture on individuals’ probability of becoming nascent and formally established entrepreneurs. In order to test our hypothesis, we analyse the direct influence of regional Hofstede scores on individuals’ entrepreneurship by estimating a multi-level model using a representative sample of 47456 individuals nested in 123 regions in 15 European countries while controlling for a host of variables such as national-level formal institutions and geography. Our results indicate that individualism and restraint positively influence the share of people that are actively trying to enter into entrepreneurship. However, we do not find evidence of any link between culture and formally established entrepreneurship. We find no evidence of the previously found positive relationship between uncertainty avoidance and entrepreneurship. We interpret our findings as indicative of the influence of cultural pull rather than cultural push factors in increasing entrepreneurship; individuals are more likely to try to become entrepreneurs in environments where the dominant value system is congruent with entrepreneurial behaviours. Remarkably, these pull factors seem to be of materialist rather than post-materialist nature. We conclude that culture influences the revealed desirability of entrepreneurship, the share of people that seek to establish new ventures, but not manifested observable entrepreneurship.

Keywords: Comparative entrepreneurship; regional entrepreneurship; regional culture; Hofstede;

(3)

List of Tables

4.1 Summary statistics . . . 33

4.2 Proportion of variance at individual, regional and country level and LR test vs. logistic model . . . 34

4.3 Two-level random-regional country-fixed effects multi-level model regression results (logit) . . . 35

A.1 Sample breakdown by country . . . 57

A.2 Mean share of entrepreneurs per country . . . 57

A.3 Sample breakdown by regions . . . 58

A.4 Correlation table of individual-level variables: All entrepreneurs . . . 64

A.5 Correlation table of regional-level variables . . . 65

A.6 Two-level random-regional country-fixed effects multi-level model regression results (probit): Main specification . . . 67

A.7 Binomial Logistic Regression . . . 68

A.8 Probit Regression . . . 69

A.9 Robustness test: Two-level random-regional country-fixed effects multi-level model regression results (logit), random draw of max. 622 observations per region . . . 71

A.10 Robustness test: Two-level random-regional country-fixed effects multi-level model regression results (logit), observations where household income was missing ex-cluded . . . 72

A.11 Robustness test: Two-level random-regional country-fixed effects multi-level model regression results (logit), regional quality of governance added . . . 74

A.12 Robustness test: Two-level random-regional country-fixed effects multi-level model regression results (logit), omitting the controls GDP per capita and population density, respectively . . . 75

A.13 EVS items used to calculate individualism . . . 76

A.14 EVS items used to calculate uncertainty avoidance . . . 76

(5)

List of Figures

4.1 Maps of entrepreneurship scores . . . 31

4.2 Maps of Hofstede dimension scores . . . 32

A.1 Plots of mean national and regional entrepreneurship . . . 61

A.2 Plots of mean national and regional culture scores . . . 62

(6)

1 Introduction

S

ince Schumpeter (1912) entrepreneurship has been argued to be of central importance for economic development (Carree and Thurik, 2003, 2010; Kirzner, 1973; Van Stel et al., 2005; Wennekers and Thurik, 1999) and to be "at the heart of national advantage" (Porter, 1990, p. 125). Its pivotal function for economic development stems from the observations that it decreases unemployment (Acs and Mueller, 2008; Birch, 1979, 1987; Thurik et al., 2008) and fosters regional and national competitiveness through increases in productivity (Fritsch and Schroeter, 2011; González-Pernía et al., 2012). Moreover, entrepreneurship generates and contributes to the commercialization of high-quality innovation (Acs and Audretsch, 1988; Audretsch and Keilbach, 2004a; Van Praag and Versloot, 2007) as well as to both the generation and exploitation of knowledge spillovers (Acs et al., 2009; Audretsch and Keilbach, 2004a, 2007; Audretsch and Lehmann, 2005; Van Praag and Versloot, 2007). Moreover, entrepreneurial firms produce eco-nomically relevant knowledge and innovation more efficiently (Love and Ashcroft, 1999). As the comparative advantage of developed nations has shifted towards the production of ideas, knowledge and innovation (Porter, 1990; Porter et al., 2002), governments and supra-national organisations such as the European Union and the Organisation for Economic Co-operation and Development have declared the promotion of entrepreneurship a central objective of their policies (Commission, 1999; OECD, 1998). Beyond economic benefits, entrepreneurship has also been found to contribute to entrepreneurs’ and employees’ job and life satisfaction (Benz and Frey, 2008a,b; Blanchflower, 2000; Blanchflower et al., 2001).

Previous research has identified that there are persistent pronounced differences in start-up rates at the national (Carree et al., 2002; Freytag and Thurik, 2007) and at the regional level (Armington and Acs, 2002; Fritsch and Falck, 2007; Fritsch and Mueller, 2007; Fritsch and Wyrwich, 2014). To explain these differences, the influence exerted by culture on entrepreneurial behaviours is of particular importance since economic, formal institutional, and legislative envi-ronments cannot fully account for the enduring differences in entrepreneurship rates (Bosma and Schutjens, 2011; Freytag and Thurik, 2007; Grilo and Thurik, 2005). A large body of literature has addressed the link between culture and entrepreneurship at the national level (Hayton and Cacciotti, 2013; Hayton et al., 2002; Hofstede et al., 2004; Stephan and Uhlaner, 2010; Wennekers et al., 2007) using mostly nationally aggregated entrepreneurship indicators.1 However, this approach has had four main limitations: (1) Culture directly influences entrepreneurship by

1

(7)

shaping the personal values of individuals, but also indirectly through the emergence of formal institutions, which in turn are grounded in dominant value systems (Licht, 2010; Licht and Siegel, 2006; North, 1991; Roland, 2004; Tabellini, 2010; Williamson, 2000). Thus, it is important to control for the influence of formal institutions and further potentially confounding factors such as geography when assessing cultural antecedents of entrepreneurship. Yet, Bruton et al. (2010, p. 432) observe that in articles assessing the relationship between culture and entrepreneurship "no other institutional factors were examined [...] there was almost a singular focus on culture as the main effect." (2) Sub-national differences in value systems, which are substantial, change very slowly and do not converge (Beugelsdijk et al., 2006; Tabellini, 2010), are omitted by relying on national culture. However, potential entrepreneurs are embedded in regional contexts (Feldman, 2001; Spigel, 2013); regional values systems and norms influence the emergence of entrepreneurial traits and behaviours (Audretsch et al., 2016; Fritsch and Wyrwich, 2016; Obschonka et al., 2013, 2015; Spigel, 2013; Stuetzer et al., 2016). (3) Entrepreneurship has been assessed on aggregate, either at the national or regional level, though both theoretically and empirically it is an individual-level event (Autio et al., 2013; Hundt and Sternberg, 2014; Kirzner, 1973, 1997; Leibenstein, 1968; Schumpeter, 1912). (4) Persistent substantial regional differences in start-up rates cannot be explained (Andersson and Koster, 2010; Bosma and Schutjens, 2011; Fritsch and Wyrwich, 2014).

(8)

and formally established entrepreneurship in estimating the influence of their contextual an-tecedents as they capture different underlying processes and stages of entrepreneurial engagement.

Our results indicate that regional culture influences nascent entrepreneurship, the revealed

desirability of entrepreneurship, but not the formal establishment of new firms, as proposed by

Freytag and Thurik (2007). In particular, we find that regional individualism and restraint positively influence individuals’ propensity of undertaking steps towards becoming entrepreneurs. As opposed to the argumentation of Uhlaner and Thurik (2007) it appears to be materialistic rather than post-materialistic motivators that increase entrepreneurship. Contrary to previ-ous findings, which indicated a positive relationship at the ecological level (Hofstede et al., 2004; Wennekers et al., 2007) we do not find evidence of a relationship between uncertainty avoidance and entrepreneurship. We interpret these findings as indicative that cultural pull rather than cultural push factors increase entrepreneurship. More people have the desire to become entrepreneurs and undertake concrete steps towards that end in environments where the dominant value system is congruent with entrepreneurial behaviours. These results hold after controlling for a host of regional economic-structural factors, and for all national-level influences such as formal institutions and location effects. However, regional culture does not influence the formal establishment of firms. We interpret the absence of a significant link between cultural indicators and manifested entrepreneurship as evidence that though culture influences preferences for entrepreneurship, market conditions and institutional aspects exert strong direct effects on the actual establishment of new firms. We propose that further research should consider the indirect links between culture and entrepreneurship via personal perceptions more explicitly, as in Stuetzer et al. (2014).

This article is structured as follows: in section 2 we review entrepreneurship and cultural theory and establish their relationship. In section 2.4 we derive our hypothesis. Section 3 describes the data and methods used. The results are presented in section 4 and discussed in section 5.

2 Entrepreneurship & culture

(9)

How do these positive effects of entrepreneurship come into existence? One influential theoretical reasoning has been proposed by Leibenstein (1966) who argues that individuals and organizations do not at all times exert the maximum effort to obtain new (market) information or develop new production methods. Consequently, there are slack resources, which constitute opportunities; if potential entrepreneurs identify these opportunities they can capitalize on them (Kirzner, 1973, 1997; Leibenstein, 1968; Yu, 2001). Leibenstein (1968) further stresses that ’gap-filling’, making up for market imperfections by for example performing functions which are non-marketable, and ’input-completing’, the command of all relevant skills needed for the undertaking are central aspects of entrepreneurship. By identifying resources which are not exploited efficiently, entrepreneurs contribute to increases in productivity directly and indirectly, which raises regional and national competitiveness (Audretsch and Peña-Legazkue, 2012; Fritsch and Schroeter, 2011; González-Pernía et al., 2012). On the one hand, the direct effect stems from more efficient goods provision of the new entrant. On the other hand, the indirect effects of entrepreneurship materialize through competitive pressures, which either cause exit, efficiency improvements of existing firms (Fritsch and Schroeter, 2011) or trigger adjustments of the market towards competitive levels (Kirzner, 1973, 1997).

Another crucial channel that has been proposed is a close relationship between entrepreneur-ship and innovation which contributes to economic development (Schumpeter, 1912). Innovation and knowledge spillovers are vital for economic development (Romer, 1990). On the one hand, entrepreneurs innovate through the creation and commercialization of new products and services by developing and utilizing new processes and organizational architectures, which ultimately con-tribute to productivity growth (Van Praag and Versloot, 2007). On the other hand, entrepreneurs benefit from exploiting knowledge spillovers emerging from existing knowledge initially created within universities and R&D departments of large firms (Audretsch, 1995; Audretsch and Keilbach, 2004b; Audretsch and Lehmann, 2005). Furthermore, entrepreneurs contribute to knowledge diffusion by imitation and implementation (Acs and Varga, 2005; Schmitz Jr, 1989). In this view, firms are formed endogenously and one of the main drivers of firm formation and start-ups are ’idle’ innovation which are not being exploited by the incumbent firm or university (Audretsch and Lehmann, 2005). While the (scientific) innovation existed previously, the entrepreneur perceives its full market potential and by (re-)introducing it to the market generates economic knowlegde.2

Innovation is a necessary, but not a sufficient condition for fostering economic development, as it needs to be complemented by entrepreneurship in order for its effects to materialize (Audretsch and Keilbach, 2004a; González-Pernía et al., 2012).The entrepreneur acts as a translator between science or residual unexploited knowledge created by existing organizations and the market. The alertness of individuals determines the probability of discovering valuable opportunities (Kirzner, 1973, 1997; Yu, 2001). It is worth emphasizing that this process of discovery of objective opportunities is a subjective process (Arenius and Minniti, 2005; Kirzner, 1997; Koellinger et al., 2007; van der Zwan et al., 2012; Yu, 2001). Discovery, the subsequent decision to exploit, and

2

(10)

the actual exploitation of opportunities are conditioned by culture, as culture firstly shapes the alertness, values, motivations and traits of individual entrepreneurs (Licht, 2010; Licht and Siegel, 2006). Secondly, culture also shapes the legitimacy and social norms with respect to entrepreneurship (Engle et al., 2010).

2.1 Definitions and operationalizations of entrepreneurship

Given the vast disarray of functions that have been associated with the entrepreneur,3 and the observation that entrepreneurship is an ill-defined concept (Beugelsdijk and Noorderhaven, 2005; Hébert and Link, 1989; OECD, 1998, 2000; Sternberg and Wennekers, 2005), how should entrepreneurship be defined? In their seminal contribution, Shane and Venkataraman (2000, p. 218) define "the field of entrepreneurship as the scholarly examination of how, by whom, and with what effects opportunities to create future goods and services are discovered, evaluated, and exploited", similarly to Venkataraman (1997). The three distinct components of discovery, evaluation and exploitation, as well as the decomposition of entrepreneurship into supply and demand, are of central importance for Shane and Venkataraman (2000). Thornton (1999, p. 20), using a sociological lens, defines entrepreneurship "as the creation of new organizations [..] which occurs as a context-dependent, social and economic process", which underlines the importance of individuals’ embeddedness. An operational definition provided by the GEM consortium is the following: "GEM defines people who are entrepreneurially active as adults in the process of setting

up a business they will (partly) own and or currently owning and managing an operating young business" (Reynolds et al., 2005, p. 209, emphasis in original) or alternatively "Any attempt at

new business or new venture creation, such as self-employment, a new business organization, or the expansion of an existing business, by an individual, a team of individuals, or an established business" (Reynolds, 1999, p. 3). Finally, the OECD (1998, p. 9) notes that "[e]ntrepreneurs are agents of change and growth in a market economy and they can act to accelerate the generation, dissemination and application of innovative ideas [...] Entrepreneurs not only seek out and identify potentially profitable economic opportunities but are willing to take risks to see if their hunches are right" (Audretsch and Keilbach, 2004b). In this study, we define entrepreneurship as a multi-stage engagement process of venture creation which is regionally embedded and influenced by culturally transmitted values and traits. We use this definition as we want to emphasize the influence of culturally induced mental structures on distinct stages of entrepreneurship, and since we aim to emphasize the importance of regional embeddedness.4

2.2 Individual, regional or national entrepreneurship?

Entrepreneurship as a dynamic multi-faceted and multi-staged phenomenon has been analysed at various levels. We now outline why we prefer to treat entrepreneurship as an individual-level event influenced firstly by its regional and secondly by its national context, as we are interested in the question how culture affects entrepreneurship.

3

Classic entrepreneurship perspectives such as the (neo-)Austrian (Kirzner, 1973, 1997), the German (Schum-peter, 1912) and the Chicago school (Knight, 1921) differ markedly in the roles and processes they ascribe to entrepreneurs, though a detailed discussion is beyond the scope of this paper. See for example Wennekers and Thurik (1999).

4

(11)

Individual-level entrepreneurship

Wennekers and Thurik (1999), in summarizing the neo-classical (Knight, 1921; Marshall, 1920), German (Schumpeter, 1912, 1951) and Austrian (Kirzner, 1973, 1997) views on en-trepreneurship conclude: "Enen-trepreneurship has to do with individuals, both with their traits and their actions" (Wennekers and Thurik, 1999, p. 34). Individual factors clearly play a large role for entrepreneurial intentions and actions; socio-demographic variables such as age, gender, education and personal wealth or household income have all been related to entrepreneurship (Caliendo et al., 2014; Davidsson, 2004; Parker, 2009). Moreover, personal traits and values influence entrepreneurship, and a large body of literature has established that entrepreneurs differ from non-entrepreneurs (Beugelsdijk and Noorderhaven, 2005; McGrath and MacMillan, 1992; McGrath et al., 1992; Rauch and Frese, 2007; Zhao et al., 2010).

With reference to classic entrepreneurship theories, it has to be noted that it is the individual who innovates (Schumpeter, 1912) and it is individual’s alertness (Kirzner, 1973) that determines discoveries: Entrepreneurs become entrepreneurs by having a different subjective assessment of objective opportunities than the rest of the population. Moreover, more recent theories of endogenous firm formation also highlight the importance of knowledge embodied in individual agents as sources of knowledge spillovers (Braunerhjelm et al., 2010). Consequently, entry into entrepreneurship should be treated as an individual-level event (Autio et al., 2013; Hundt and Sternberg, 2014; Licht, 2010).

Empirically, this is strongly supported by the relatively rare (Álvarez et al., 2014; Levie et al., 2013; Urbano and Alvarez, 2014) entrepreneurship studies that have employed multi-level models. These studies have consistently found that more than 80% of the variance observed is attributable to the individual level (Autio et al., 2013; Hundt and Sternberg, 2014; Stuetzer et al., 2014; Wennberg et al., 2013). While we note that this number may be overestimated since measurement errors occur at the individual level, both theoretically and empirically entrepreneurship is an

individual-level construct. Consequently, it ought not to be analysed at the ecological level if the

aim is to make individual-level interferences (Alker, 1969; Robinson, 1950).

Entrepreneurship as regional contextual event

(12)

level- in terms of specialization, unemployment rates and income have been observed by Puga (2002), which illustrates the importance of regional analyses.

Theoretically, firstly, it has been suggested that entrepreneurs draw mainly on regional rather than on national resources and networks (Feldman, 2001; Huggins and Thompson, 2015; Spigel, 2013), which is due to the scale of her undertaking. Michelacci and Silva (2007) observe that the share of entrepreneurs starting their venture in their home region is significantly higher than the corresponding share of wage labourers. Spatial inertia arises as people become entrepreneurs where they are born or have spent a large part of their lives which allows them to access regional networks (Stam, 2007; Sternberg and Fritsch, 2011; Wagner and Sternberg, 2004). This provides further evidence of the sub-national dynamics that need to be taken into consideration.

Secondly, knowledge spillovers and knowledge commercialization are regional events (Acs and Plummer, 2005; Audretsch, 1995). Previously, as endogenous theory was developed (Romer, 1990), it was assumed that knowledge -being non-rival and non-excludable- spills over across space and organizations automatically and at no cost (Acs and Plummer, 2005). However, recent evidence indicates that knowledge spillovers occur mostly regionally (Frenken et al., 2007; Jaffe et al., 1993)5. Indeed, Armington and Acs (2002) find indirect evidence for the importance of localized knowledge.

Thirdly, the regional industrial structure exerts a strong influence on entrepreneurial activity through functional specialization (Audretsch and Fritsch, 2002). While some regions specialize in innovative output and are characterized by a highly educated workforce, other regions are focussed on the efficient production of products and are shaped by qualified, but not necessarily highly educated employees, which induces differing development paths (Audretsch et al., 2012; Audretsch and Feldman, 1996; Duranton and Puga, 2001; Fritsch, 2008). Furthermore, pecuniary externalities such as natural resources, the presence of which affects functional specialization, also occur at the regional level. This sub-national functional specialization underlines the rele-vance of regional approaches (Audretsch et al., 2012); regional functional specialization, and thus economic-structural context, rather than national context, influences entrepreneurship (Audretsch et al., 2012).

Fourthly, the observed temporal stability of regional differences in entrepreneurial activity (Andersson and Koster, 2010; Fritsch and Mueller, 2007; Fritsch and Storey, 2014; Fritsch and Wyrwich, 2014) is a further indicator of the relevance of conducting analyses contextualized at that level. Indeed Fritsch and Wyrwich (2014) find evidence of the persistence of regional rates of entrepreneurship despite the passage of 80 years and a World War. By definition, national-level institutions and policies cannot account for the sustained regional differences; we need to adjust the scales of analysis accordingly and go beyond the nation as unit of analysis to uncover the contextual antecedents of entrepreneurship.

5

(13)

Empirical evidence strongly supports the proposition that contextual antecedents of en-trepreneurship should be assessed at the regional level (Armington and Acs, 2002; Feldman, 2001; Fritsch, 2013). As Audretsch et al. (2012, p. 380) illustrate: "The fortunes of regions and entrepreneurs are intertwined: regional endowments provide opportunity and resources for entrepreneurs, while entrepreneurs simultaneously shape the local environment."

Entrepreneurship as national contextual event

As described above, we are primarily interested in regional variation of entrepreneurship. This goes not to say that we disregard the importance of national-level contextual influences; both economic and institutional factors at the national level exert strong influences on entrepreneurship. Amongst the channels are the stage of economic development, macroeconomic stability, openness to trade and integration in world markets, provision of physical infrastructure, provision of credit and equity, protection of property rights, rule of law, contract enforcement, degree of govern-ment regulation and entry regulations, labour laws, taxation, welfare system, entrepreneurship education and policies directed towards fostering entrepreneurship (Blanchflower, 2000; Carree et al., 2002; Freytag and Thurik, 2007; Hofstede et al., 2004; Liñán and Fernandez-Serrano, 2014; Noorderhaven et al., 2004; Sternberg and Wennekers, 2005; Verheul et al., 2002; Wennekers et al., 2005; Wong et al., 2005).

2.3 Regional culture and entrepreneurship

Since culture shapes individual’s values, traits and norms, it has long been argued that cultures are not equally effective in promoting entrepreneurship (Baumol, 1968; Birch, 1987; Hofstede, 1980a; Shapero and Sokol, 1982). The idea that cultures can be classified and that con-sequently it is possible to derive cultural dimensions is based on the assumption that all societies confront similar basic problems, and that these problems are limited in number (Hofstede, 1980a; Kluckhohn, 1962; Kluckhohn and Strodtbeck, 1961). The most prominent definition of culture has been provided by Hofstede (2001, p. 9) who defines culture as "the collective programming of the mind that distinguishes the members of one group or category of people from another" which illustrates the existence of learned differences in value systems between different groups. Gorodnichenko and Roland (2010, p. 1) define culture "as the set of values and beliefs people have about how the world (both nature and society) works as well as the norms of behaviour derived from that set of values". This definition is constructive for this study, since it emphasizes that culturally transmitted values influence behavioural norms and values, which have economic consequences such as entrepreneurship. It is precisely these behavioural consequences which are rooted in culture and their resulting economic effects we are interested in.

(14)

Secondly, the social legitimation perspective builds on social norms as behavioural antecedents Ajzen (1991); higher social status, higher societal encouragement and legitimacy of entrepreneur-ship increase start-up rates (Uhlaner and Thurik, 2007). The congruence of entrepreneurial behaviour with dominant social norms is of central importance for entrepreneurs according to this view. Both perspectives correspond to a pull view, where individuals enter into entrepreneurs due to the expected benefits.

Thirdly, the dissatisfaction perspective explains entrepreneurship by a misfit between individ-uals’ values and ambitions and societies’ predominant structure and value systems. Entrepreneurs are ’pushed’ into starting a venture since they cannot satisfy their achievement motivation or need for autonomy in established organization (Baum et al., 1993). The share of persons who become entrepreneurially active depends on the restrictiveness of the respective value systems. For example, it has been found that a society which scores high on uncertainty avoidance, thus exhibiting a substantial degree of inertia at various levels, may force more people into self-employment (Baum et al., 1993; Hofstede et al., 2004; Noorderhaven et al., 2004; Wennekers et al., 2007), an account that has not been unchallenged by recent findings (Autio et al., 2013) as we discuss in more detail below when deriving the hypothesis.

Based on these perspectives, a number of scholars have tried to derive the ’entrepreneurial spirit’ of regions and countries either indirectly or empirically. Indirectly, regional culture (alter-natively termed entrepreneurial capital) has been identified as a residual effect in explaining the persistent differences in regional entrepreneurship (Audretsch and Keilbach, 2004b; Fritsch and Wyrwich, 2014, 2016; Stuetzer et al., 2016,?).

Empirically, Davidsson (1995) and Davidsson and Wiklund (1997) show that regional en-trepreneurship depends upon regional value systems. Beugelsdijk and Noorderhaven (2004), Beugelsdijk (2007) and Suddle et al. (2010) derive from an individual-level comparison of values of entrepreneurs and non-entrepreneurs, based on European Values Study (EVS) (European Values Study Group, 2015) and World Values Survey (WVS) data (European Values Study Group and World Values Survey Association, 2006) data, a measure of ’entrepreneurial attitude’ (Beugelsdijk and Noorderhaven, 2004) and ’entrepreneurial culture’ (Beugelsdijk, 2007; Suddle et al., 2010).6 They aggregate the individual value profile associated with a high propensity to be self-employed at the regional level to obtain measures of culture and relate this regional cultural measure to economic growth and innovation, for both of which they find evidence. Consequently, they indirectly establish the link between entrepreneurial culture and economic growth (Beugelsdijk, 2007; Beugelsdijk and Noorderhaven, 2004) and between entrepreneurial culture and nascent entrepreneurship (Suddle et al., 2010). Furthermore, Suddle et al. (2010) also test whether the concept of achievement motivation of McClelland (1961), the achievement motivation index developed by Granato et al. (1996), the measure of competitiveness of Lynn (1991) and the GLOBE index of performance orientation (House et al., 2002, 2004; Javidan et al., 2006) influence nascent entrepreneurship. However, they do not find evidence for the hypothesized positive

6

(15)

relationships.

Stuetzer et al. (2014) combine individual level data from the GEM consortium with regional data for western German regions and apply a multilevel model. They do not find a direct link between the regional culture and individuals’ entrepreneurial intentions and actions. However, they propose that there is an indirect link between regional context and (individual-level) en-trepreneurship which runs through individuals’ opportunity perception.

Obschonka et al. (2013, 2015) drawing on an entrepreneurship-prone personality profile as we all as the Big Five personality profile (Costa and McCrae, 1985; McCrae and Costa, 1987; McCrae and Costa Jr, 1999) derive personality-based measures of regional entrepreneurial culture and observe correlations with entrepreneurship activity. Audretsch et al. (2016) use US data from the ’American Nations and Patchwork Community Types’ as measures of regional culture and evaluate how these translate into latent and actual entrepreneurship. We conclude that there is substantial and steadily growing evidence that links regional culture and entrepreneurship. For these reasons, we access culture at the regional level and derive our hypotheses in the next section.

2.4 Hypotheses development

Hofstede’s dimensions of culture are appropriate for and have frequently been employed in entrepreneurship research (Kirkman et al., 2006; Pinillos and Reyes, 2011; Shane, 1992, 1993; Søndergaard, 1994; Thomas and Mueller, 2000; Wennekers et al., 2007). An influential review of the research published on national culture and entrepreneurship is provided by Hayton et al. (2002), the updated version was published as Hayton and Cacciotti (2013). Consequently, in our literature review we do not seek to provide a comprehensive overview of all studies published on the matter but rather address theoretical and methodical aspects as well as the most important findings. Firstly, we briefly describe the dimensions. Afterwards, based on the description, we provide a theoretical link with the supply and demand-side of entrepreneurship.

In order to establish a link between culture and individuals’ behaviour, firstly it needs to be established how culture influences individuals’ values, motivations and traits, and secondly how these factors influence behaviour (Guiso et al., 2006). For this reason, we draw on the entrepreneurial traits literature (Cromie, 2000; Rauch and Frese, 2007; Zhao and Seibert, 2006; Zhao et al., 2010) to derive our hypothesis. We structure our discussion by drawing upon the components of the theory of planned behaviour, attitudes, social norms and perceived behavioural control (Ajzen, 1985, 1991) which has been applied to entrepreneurship research manifold (Engle et al., 2010; Kautonen et al., 2015, 2013; Krueger and Carsrud, 1993; Krueger et al., 2000).

2.4.1 Individualism-collectivism

(16)

to take care only of themselves and their nuclear family. Social networks are loosely-knit and interdependency is low, but the outward orientation of society is high. Moreover, obedience and conformity to social norms are of less central importance in individualist than collectivist societies and social norms are less restrictive. Importantly, status and rewards are granted for individual achievements and others are perceived as individuals rather than classified as in-group or out-group members (Beugelsdijk et al., 2015; Hofstede, 2011; Hofstede et al., 2010; Hofstede, 2001; Pinillos and Reyes, 2011; Triandis, 2001).

Individualistic societies contribute to entrepreneurship through the influence of individualism on individuals’ attitudes, social norms, and perceived behavioural control, as differentiated in the theory of planned behaviour (Ajzen, 1991). Firstly, individualism fosters individuals’ autonomy, since autonomous behaviour is consistent with the values of individualism, most importantly the fact that persons are supposed to take care of themselves, whereas it is less aligned with the values of collectivism which emphasise group loyalty and in-group embeddedness. Research has shown that individuals with higher need for autonomy hold more favourable attitudes towards entrepreneurship, which influences entrepreneurial intentions positively and increases observed manifested entrepreneurship (Cromie, 2000; Rauch and Frese, 2007).

Secondly, personal attitudes as well as social norms with respect to entrepreneurship are favourably influenced by individualism through awarding material and non-material rewards. In individualist societies, recognition and social status are awarded for individual achievements (Hofstede, 1980a). This constitutes a strong motivator for individuals to innovate (Gorodnichenko

and Roland, 2010, 2012) or become entrepreneurially active.

Thirdly, the central importance of personal freedom in individualist societies (Shane, 1992, 1993) at the ecological level furthermore legitimises individuals’ self-directed action and autonomy. Conversely, in collectivist societies, group-interests ought to take precedence over individuals’ interests; individualist actions can be perceived to be alienating and as threat to group interests (Hofstede, 1980a). Furthermore, there is a fear that the acting individual may not share the derived benefits (Hofstede, 1980a). Thus, individuals’ entrepreneurial behaviours are perceived to be less legitimate in collectivist than in individualist societies, while legitimacy is an important antecedent of desirability beliefs (Shapero and Sokol, 1982). Desirability is in turn a central antecedent of behavioural intentions and behaviour (Ajzen, 1985, 1991).

(17)

Moreover, individualistic societies have a stronger outward orientation (Hofstede, 1980a; Shane, 1992, 1993) which enhances access to information, resources and networks and thus increases self-efficacy beliefs (Bandura, 1982) which form a part of perceived behavioural con-trol. Access to information and networks is also vital for entrepreneurial discovery, entry into entrepreneurship and post-entry growth (Autio et al., 2013). Consequently, outward orientation has been argued to be of central importance for entrepreneurship (Shane, 1992, 1993).

On the demand side, most importantly, individualism fosters niche-consumption of highly specialized consumer goods as expression of self. Consumption decisions are expressions of culturally induced values and norms, it is a way of expressing and sharpening the individual iden-tity.Consumption occurs in line with cultural preferences and values (Aaker et al., 2001), personal identity, and it is governed by (group) social norms (Warde, 1994). Since more entrepreneurs engage in niche-production and provide variety (Tiessen, 1997), individualism positively influences entrepreneurship through higher demand for variety and niche products.

Previous research has found that the individualism-collectivism dimension is the most relevant of Hofstede’s framework if one is interested in the economic (Gorodnichenko and Roland, 2010, 2011, 2012) and institutional (Klasing, 2013) consequences of culture. This contention is further supported by cross-cultural psychologists (Heine, 2007; Triandis, 1988).

At the national level, empirical evidence in support of the positive relationship between individualism and entrepreneurship has been provided by McGrath et al. (1992) and Mueller and Thomas (2001). Moreover, Shane (1992, 1993) find that individualism positively influences rates of innovation. Autio et al. (2013) find that while individualism is positively associated with

entry into entrepreneurship, it is negatively related to post-entry growth. Moreover, Pinillos and

Reyes (2011) find that the relationship between entrepreneurship and individualism depends on the stage of economic development of the respective country. For countries in low and medium stages of economic development there is a negative association between individualism and entrepreneurship, while for highly developed countries there is a positive relationship. This finding can be interpreted in line with Gorodnichenko and Roland (2010), who propose that collectivism leads to efficiency gains through facilitated collective action, while individualism is conductive for innovative behaviour. As low and medium income countries are in the factor or efficiency-driven stage of development (Porter, 1990; Porter et al., 2002), efficiency gains from collective action outweigh the benefits of innovativeness.7 Consequently, we hypothesize:

Hypothesis 1 Individualism positively influences individuals’ likelihood of becoming nascent

entrepreneurs.

Hypothesis 2 Individualism positively influences individuals’ likelihood of becoming formally

established entrepreneurs.

(18)

2.4.2 Uncertainty avoidance

Uncertainty is a "basic fact of human life" (Hofstede, 1980a, p. 110) which humans seek to confront by relying on institutions such as technology, laws and religion. Societies cope differently with uncertainty. Hofstede (1980b, p. 45) defines uncertainty avoidance as "the extent to which a society feels threatened by uncertain and ambiguous situations and tries to avoid these situations by providing greater career stability, establishing more formal rules, not tolerating deviant ideas and behaviours, and believing in absolute truths and the attainment of expertise." Consequently, it has to be stressed that it is not merely concerned with risk avoidance, but tolerance of ambiguity. Economically most importantly, deviant and new ideas as well as innovation are rejected and considered to be dangerous (Hofstede, 2011; Hofstede et al., 2010) as they have the potential to induce uncertainty and trigger changes.

Relating uncertainty avoidance to entrepreneurship, we outline three transmission chan-nels: (1) The acceptance of uncertainty and deviant behaviour, (2) the role of risk and (3) dissatisfaction and push hypothesis. Firstly, a high degree of uncertainty avoidance arises from the perception that the world is hostile, which is why laws and social norms are needed to reduce uncertainty (Beugelsdijk et al., 2015; Hofstede, 2001). Both social norms and formal institutions in uncertainty avoidant societies are directed towards maintaining a stable status, and neither deviant behaviour nor new ideas are tolerated (Hofstede, 1980a; Hofstede et al., 2010) as anything novel entails risk and uncertainty. This has supply- and demand-side implications for entrepreneurship. On the supply side, a societal value system characterized by strong uncertainty avoidance influences individuals’ attitudes and diminishes the likelihood that individuals hold a favourable assessment of entry into entrepreneurship, and thus the proportion of individuals who consider entrepreneurship as a viable employment decision and are alert to opportunities (Licht, 2010; Licht and Siegel, 2006; Shapero and Sokol, 1982). Moreover, as deviant and new ideas are rejected and even perceived as threats, social norms discourage entrepreneurship and there is less legitimacy for entrepreneurial behaviours. Lack of legitimacy negatively affects behavioral intentions (Ajzen, 1991) and access to networks and resources of newcomers and thus impedes venture development.

(19)

occupational choice models, risk-aversion is of central importance in determining whether an individual decides to become an entrepreneur or an employee (Davidsson, 2004; Parker, 1997). Empirically, it has been found that entrepreneurs score lower on risk-aversion than managers (Stewart Jr and Roth, 2001, 2004), but that post-entry survival is driven by moderate risk taking (Caliendo et al., 2010). In contrast, cognitive bias studies have revealed that entrepreneurs might not be less risk-averse, but are rather characterized by an overly positive subjective assessment of objective entrepreneurial opportunities (Palich and Bagby, 1995; Simon et al., 2000), which in turn leads them to take more risks unconsciously. This distinction is of high relevance for understanding whether or not there is a link between uncertainty avoidance and entrepreneurship via risk preferences.9 Uncertainty avoidance and risk-aversion are related insofar as that societies characterized by high degree of uncertainty avoidance discourage risky behaviour, and more so uncertain behaviours (in line with Knight (1921)), as they could result in changes, which could trigger further uncertainty. Risky and uncertain behaviours are often risky and uncertain because they are novel and deviant. Entrepreneurs do not seek to mechanically optimize a well defined production function (Baumol, 1968), but rather to conquer the market by introducing a radical novelty (Schumpeter, 1912). Thus, the desirability of risky and uncertain behaviours, determined by both, attitudes and social norms, is reduced in uncertainty avoidant societies. If it holds that entrepreneurs are less risk-averse than the rest of the population, it would be reasonable to assume that societies which exhibit a higher degree of uncertainty avoidance are less conductive to the formation of individual values and traits conductive to entrepreneurship, as risky and uncertain behaviours are generally discouraged. This in turn would reduce entrepreneurial intentions and behaviours. However, if, in line with the cognitive bias view (Palich and Bagby, 1995; Simon et al., 2000) entrepreneurs are overly optimistic and overly confident (Arenius and Minniti, 2005; Koellinger et al., 2007) but do not differ in their preference for risk-taking, there would be no link between uncertainty avoidance and preferences for entrepreneurship via risk-preferences.

Nevertheless, even if entrepreneurs would not differ in their risk assessment, lack of tolerance and legitimacy of deviant behaviours with uncertain outcomes, such as entrepreneurship, would create a negative link between uncertainty avoidance and entrepreneurship. We thus hypothesize: Hypothesis 3 Uncertainty avoidance negatively influences individuals’ likelihood of becoming

nascent entrepreneurs.

Hypothesis 4 Uncertainty avoidance negatively influences individuals’ likelihood of becoming

formally established entrepreneurs.

However, this perspective is not undisputed. Another opposing view taking an entirely different perspective is proposed by Baum et al. (1993) and Hofstede et al. (2004) who argue that societies characterized by a high degree of uncertainty avoidance push enterprising individuals into entrepreneurship as organizations are not promoting innovation and change. Consequently, individuals would start firms as a result of their frustration and/or because they are feeling held back by existing organizations, which corresponds to a "push-view" of entrepreneurship. A theoretical foundation for this view can be taken from the ’cultural self-expression theory’ by

9

(20)

Erez and Earley (1993).10 _{Empirical evidence for the push perspective has been provided by}

Hofstede et al. (2004), Noorderhaven et al. (2004) and Wennekers et al. (2007). Consequently, we alternatively hypothesize:

Hypothesis 5 Uncertainty avoidance positively influences individuals’ likelihood of becoming

nascent entrepreneurs.

Hypothesis 6 Uncertainty avoidance positively influences individuals’ likelihood of becoming

formally established entrepreneurs.

2.4.3 Indulgence vs. restraint

The last dimension, indulgence vs. restraint, was added to the original dimensions of Hofstede (1980a) at a later point by Hofstede et al. (2010) based on the work of Minkov (2007). It is conceptually related and negatively correlated to long-term orientation and inspired by happiness research. Indulgence vs. restraint seeks to capture the extent to which people are free to gratify their needs, wants and desires (Hofstede, 2011; Hofstede et al., 2010). Indulgent societies are characterized by people valuing leisure time, freedom of choice and speech, and generally the freedom to be in control of one’s life. Systemic order is not a central concern. Overall, people remember positive emotions more. Thus, indulgent societies are manifested by a large share of the population declaring that they are very happy (Hofstede, 2011; Hofstede et al., 2010). Most importantly, in indulgent societies, people feel in control of their personal life, and to a certain extent believe that happiness and the circumstances leading to it are the result of choices (Beugelsdijk et al., 2015; Hofstede et al., 2010), which points to a positive relationship between societal indulgence and indviduals’ internal locus of control. Internal locus of control (Rotter, 1966) has been associated with increased propensity to become entrepreneur (Brockhaus, 1982; Caliendo et al., 2014; Cromie, 2000; Shapero and Sokol, 1982; Thomas and Mueller, 2000).

(21)

of entrepreneurship are negatively influenced, and so are attitudes towards entrepreneurship. Consequently, individuals with a high need for autonomy are more likely to be reluctant to actually satisfy this need. Secondly, as people generally are less likely to remember positive emotions (Hofstede, 2011; Hofstede et al., 2010), procedural utility is given less weight in decision making, which also negatively influences attitudes towards entrepreneurship.

In restrained societies, the second component of the theory of planned behaviour, social norms (Ajzen, 1991), reduces entrepreneurial intentions and actions by reducing the legitimacy of entrepreneurial behaviours for the same reasons mentioned above; indulging procedural utility and autonomy is discouraged by the restrictive governance of gratification of needs and wants through peers. This effect is likely accentuated if individuals decide to become entrepreneurs to be more happy while accepting reductions in income (Hamilton, 2000), which reduce legitimacy of entrepreneurship and increases the adverse influence of restraint cultures on entrepreneurship. Consequently, attitudes and social norms determine the desirability of entrepreneurship (Shapero and Sokol, 1982), which is higher in indulgent and lower in restrained societies. There are no obvious links between perceived behavioural control and indulgence vs. restraint; the ex-ternal locus of control of people is concerned primarily with their perceived influence on happiness.

While the supply-side links between this indulgence vs. restraint and entrepreneurship are indirect, the relationship with the demand-side of entrepreneurship is substantial. Restrained cultures by limiting the gratification of needs reduce demand for entrepreneurs’ products and services. This holds particularly true for niche- and lifestyle products which are not essential parts of individuals’ consumption baskets, but which attract a significant amount of entrepreneurship. As more entrepreneurs develop lifestyle and niche products than say new sorts bread -to reduce the argument to a stub- more demand for novel and not existentially required products and services is generated in indulgent cultures. Thus, we hypothesize:

Hypothesis 7 Indulgence positively influences individuals’ likelihood of becoming nascent

en-trepreneurs.

Hypothesis 8 Indulgence positively influences individuals’ likelihood of becoming formally

es-tablished entrepreneurs.

3 Data & methodology

3.1 Data

(22)

the best of our knowledge, we are the first to employ a multi-level model to a multi-country sample of regionally clustered individual-level GEM data (Reynolds et al., 2005) and regional Hofstede cultural indicators based on Beugelsdijk et al. (2015) and Beugelsdijk (year).

3.1.1 Entrepreneurship data

We use regionally clustered individual-level data from the GEM Adult Population Survey (APS) 2012. Detailed information regarding the APS are provided by Reynolds et al. (2005).11 We have chosen these data for a variety of reasons: (1) It allows us to assess entrepreneurship as an event conducted by individuals who are embedded in their respective regional and na-tional context. (2) Using the individual as the unit of analysis is empirically beneficial since it alleviates concerns of endogeneity; individuals cannot be argued to alter their respective cultural environment. Alternative data sources, such as the Eurostat Database, provide only aggregated data at the regional or national level. (3) In a similar vein, it also allows us to control for individuals’ characteristics such as age, gender and education, which have been shown to be important antecedents of entrepreneurship (Davidsson, 2004; Parker, 2004). (4) The indicator for ’nascent entrepreneurship’ is the only measure that can be used to assess entrepreneurial behaviour and intentions at an early stage. (5) Furthermore, the GEM data is the only data source that provides internationally comparable entrepreneurship data for a large number of countries (Acs et al., 2008; Reynolds et al., 2005; Sternberg and Wennekers, 2005).

In the GEM APS, per country, each year at least 2000 individuals are sampled (Reynolds et al., 2005). In some countries, most notably Spain, the national GEM teams sample signif-icantly more people, as shown in appendix table A.1. The regional information which is not publicly accessible has been kindly made available by the respective national teams of the GEM consortium; the regional sample breakdown is presented in appendix table A.3.12 In line with the official GEM methodology (Reynolds et al., 2005) and previous entrepreneurship research, the sample of people who are aged between 18 and 64 years was used (Minola et al., 2016). Thus, we include also students, unemployed and housewives or housemen in our sample, besides part-and full-time employees, since they could become entrepreneurs at a later stage of their lives. However, we exclude people who are retired and/or disabled from the analysis, since they cannot be expected to become entrepreneurs. Excluding persons who are younger than 18, older than 64, retried or disabled reduced our dataset from 64,573 to 56,585 observations.13 Moreover, due to missing data, in particular arising from missing responses to questions which were included in our estimation, our dataset was reduced to 47,456 individual observations.

Recent reviews of GEM research are provided by Álvarez et al. (2014), Amorós et al. (2013), Bergmann et al. (2014) and Bosma (2013). Two caveats regarding the GEM data are due. They relate to firstly its reliance on dichotomous and secondly on single-item measures, which are a

11

Though in Reynolds et al. (2005) the GEM APS 2003 is described, the procedure and data are largely unchanged; for this reason it still valid and frequently cited.

12

As agreed upon in the confidentiality agreement with the GEM national teams, no mean regional scores of entrepreneurship rates are presented.

13

(23)

limitation in particular in the context of attitudinal questions (Bergmann et al., 2014). Regarding the former, Reynolds et al. (2005) argue that the dichotomous nature of items increases the international applicability of the questionnaire as it reduces possible translation problems and cultural response biases.14 _{Regarding single-item measures, on a more practical note, it has been}

proposed that they allow for a faster completion of the survey; Maula et al. (2005) argue that the GEM consortium, in facing the trade-off between sample representativeness and depth of measures, deliberately decided for the former (Bergmann et al., 2014). Due to these limitations we do not include any attitudinal measures such as fear of failure or self-efficacy in our specifications.

3.1.2 Cultural data

We measure culture by using the replicated Hofstede scores (Beugelsdijk et al., 2015). The basis for replicating national Hofstede scores was the WVS (European Values Study Group and World Values Survey Association, 2006). The WVS is a large scale social survey which was started in 1981 and conducted periodically since then. In total, more than 340,000 individuals from almost 100 different countries have been surveyed (European Values Study Group and World Values Survey Association, 2006). As our analysis is centred on European regions, we pursue a similar strategy by generating sub-national cultural indicators from the EVS (European Values Study Group, 2015), the set-up of which is identical to the WVS. The EVS is set up similarly to the WVS; data has been collected from more than 166,000 individuals living in 49 mostly European countries. Most importantly, the EVS data enables us to identify in which region individual respondents are located, which enables us to generate sub-national Hofstede scores. Consequently, this allows us to assess culture at the ecological yet sub-national level and to match EVS data with regionally clustered individual-level GEM data as well as regional Eurostat data to pursue a regional approach as in Tabellini (2010). As shown extensively by Beugelsdijk et al. (2015), the derived indicators are both conceptually and empirically valid and very similar to the original Hofstede dimensions.

3.1.3 Regional clustering of individual-level data

We conduct our sub-national analysis based on the Nomenclature of Territorial Units for Statistics (NUTS) 2015 classification of the European Union.15 _{Contingent upon data availability,}

in some countries, we use the NUTS1 classification while in other we are able to exploit the more fine grained NUTS2 classification. NUTS1 regions are defined to be inhibited by between 3 million and 7 million people. In NUTS2 regions, the lower threshold is set at 800,000 inhabitants and the upper threshold is 3 million. The regions included in this study and the level of NUTS aggregation are also shown in appendix table A.3.

To ensure representativeness of the individual-level data underlying the cultural regional aggregates, a lower bound of 50 individual observations per region was used in deriving the regional Hofstede scores. The same threshold was applied to regionally clustered individual-level observations from GEM. Regions where less GEM or EVS respondents had been sampled were not included in the analysis. Though there exist no established minimum thresholds, Hofstede

14_{To address cultural response bias, a number of correction methods would be available.} 15

(24)

(1981, p. 65) argues that "if a sample is really homogeneous with regard to the criteria under study, there is very little to gain in reliability over an absolute sample size of 50". Since Hofstede has been criticised for this assumption (McSweeney, 2002) and since there is no well-defined minimum threshold, as a robustness check, the lower bound was furthermore set to 75 and to 100 observations per region.16

Due to mis-matches of regional EVS and GEM data, Croatia, Latvia, Lithuania and Slovenia had to be excluded from the analysis. Though there are approximately 2000 individual-level observations in the GEM data, either lack of representativeness of the EVS or GEM data at the regional level or unavailable regional control variables from Eurostat made their inclusion impossible.

3.1.4 Dependent variables

We use three dependent variables to measure entrepreneurship, since the influence of (regional) contextual antecedents on entrepreneurial behaviours differs depending on the stage of the entrepreneurial process (Autio et al., 2013; Hundt and Sternberg, 2014; van der Zwan et al., 2012; Van Der Zwan et al., 2013).

1. An overall measure of entrepreneurial activity, all entrepreneurs, where people who are business owners or have taken concrete steps towards becoming entrepreneurs in the last 12 months and will own a share of the firm are considered to be entrepreneurs. In GEM terminology, in this category, nascent, young and established entrepreneurs are pooled. 2. Nascent entrepreneurs are people who in the last 12 months have been actively trying to

start a new business, will own part of this new business and have not paid wages for more than 3 months.

3. Formally established entrepreneurs are individuals who are owner-managers of existing businesses that have paid wages for more than 3 months. In GEM terminology, this category corresponds to pooling young entrepreneurs and established owner-managers (Bosma, 2013; Reynolds et al., 2005).

Nascent entrepreneurship has been frequently used in empirical research, see, for example Arenius and Minniti (2005), Wennekers et al. (2005), Koellinger (2008), Suddle et al. (2010), Noseleit (2014) and Peroni et al. (2016). A review of studies on nascent entrepreneurship can be found in Davidsson (2006). It is considered to be a dynamic entrepreneurship indicator (Davidsson, 2006; Reynolds, 1997; van der Zwan et al., 2010). It is worth noting that it is conceptually distinct from latent entrepreneurship, which only captures peoples’ general prefer-ences for becoming entrepreneurs in the future, which has also been discussed extensively in the literature (Audretsch et al., 2016; Blanchflower et al., 2001; Grilo and Thurik, 2005; Stuetzer et al., 2016). As nascent entrepreneurship involves not just general preferences for self-employment, but also first concrete steps towards establishing a business, we contend it is a better proxy of entrepreneurial potential as it goes beyond mere desirability beliefs Shapero and Sokol (1982)

16

(25)

but also involves action. Thus, we treat it as a measure of revealed desirability of entrepreneurship.

Formally established entrepreneurship is used as a static measure of entrepreneurship. It measures to which extent people have succeeded in establishing a venture. It can also be termed observable manifested entrepreneurship (Audretsch et al., 2016). We do not use total early-stage entrepreneurial activity (TEA), which has been used frequently in GEM research (Bosma, 2013), as we firmly believe that the formal establishment of a business marks an important transition.17 Similar differentiations of stages have been employed in the literature (Bosma and Schutjens, 2009; Davidsson, 2006; Grilo and Irigoyen, 2006; Hundt and Sternberg, 2014; Stuetzer et al., 2014).

3.1.5 Independent variables

We replicate the steps as described in Beugelsdijk (year) and Beugelsdijk et al. (2015) to derive European sub-national Hofstede scores. In particular, using EVS data we are able to produce scores for the Hofstede dimensions individualism-collectivism, uncertainty avoidance and

indulgence vs. restraint, which we scale from 0 to 100. It is worth noting that as we follow the

empirical replication strategy of Beugelsdijk (year), the last dimension, indulgence vs. restraint is scaled such that 0 indicates indulgence and 100 restraint. The questions of the EVS on which we base the derivation of the Hofstede scores are shown in tables A.13, A.14 and A.15, as in Beugelsdijk (year).18

3.1.6 Individual control variables

Socio-demographic factors explain a large proportion of individuals’ entry into entrepreneur-ship (Davidsson, 2004; Parker, 2004, 2009). Firstly, we control for age and age-squared, as an inverse U-shape has been found for the effect of age on entrepreneurship (Amorós and Bosma, 2014; Bergmann and Sternberg, 2007; Hundt and Sternberg, 2014; Minola et al., 2016). Secondly, women exhibit lower rates of entrepreneurship than men (Arenius and Minniti, 2005; Beugelsdijk and Noorderhaven, 2005; Blanchflower, 2004; Davidsson, 2006; OECD, 1998; van der Zwan et al., 2012; Verheul et al., 2012), thus we control for gender. In addition, we control for household

size, which we operationalize as its the natural logarithm, since we expect a higher propensity

to start a business with increasing household size (Blanchflower, 2000). Research has shown that migrants have a higher propensity to become entrepreneurs than people who have been born in the country where they reside (Davidsson, 2006; Kalantaridis and Bika, 2006; Levie, 2007; Piergiovanni et al., 2012; Rath and Kloosterman, 2000). Consequently, we control for being born abroad. Moreover, we control for education using a categorical variable (0 indicates no formal education; 4 graduate experience, which is the highest level), since a high level of formal education increases the probability of being an entrepreneur by raising human capital and alertness (Davidsson, 2006; Licht, 2010). However, a high level of education also increases the 17_{In robustness checks (not reported; available upon request) it was found that results for nascent entrepreneurs} were confirmed when using TEA. We speculate that this finding is due to the TEA composition, which pools the categories nascent and young entrepreneurs. An overwhelming majority of entrepreneurs in TEA are nascent and very few are young entrepreneurs.

18

(26)

likelihood of having other employment opportunity, and thus raises opportunity costs (Hundt and Sternberg, 2014; Van Der Zwan et al., 2013). Furthermore, we also control for household

income (as a proxy for household savings and wealth for which we do not have data), since on

the one hand access to resources facilitates entry into entrepreneurship. For example, savings can be used as collateral when seeking financing. However, on the other hand, as noted by Hundt and Sternberg (2014), higher income also increases opportunity costs. Household income is operationalized as the tertile into which the respondent falls within his or her country. Since a large share of respondents were not willing to indicate their income, we create a fourth category for ’missing’.19

3.1.7 Regional control variables

As entrepreneurship is a regionally and nationally embedded event, we control for the relevant characteristics at the respective levels; regional control variables were obtained from the Eurostat database.20 _{Previous research has found that the stage of development is one of the main}

determi-nants of entrepreneurship (Blau, 1987; Carree et al., 2002; Kuznets, 1971; Kuznets and Murphy, 1966; Wennekers et al., 2005). Consequently, we control for GDP per capita at the regional level and more generally for economic development at the national level through the country fixed effect.

Furthermore we control for agglomeration economies, and in particular urbanization economies, (Frenken et al., 2007) by employing population density in our estimations. As Frenken et al. (2007, p. 687) describe: "It is the dense presence of [...] organizations (not solely economic in

character, but also social, political and cultural) that supports the production and absorption of know-how, [and] stimulating innovative behaviour." The importance of agglomerations for entrepreneurship and its impact on employment and on growth have been stressed frequently in empirical research (Fritsch and Schroeter, 2011). Moreover, it can be used as a catch-all variable for amongst others quality and accessibility of infrastructure, for land prices and for size of the labour market (Fritsch and Wyrwich, 2016; Stuetzer et al., 2014).

Research has argued that there is a close relationship between unemployment and en-trepreneurship. At the individual-level we have controlled for unemployment to control on the one hand for reduced opportunity costs of engaging in entrepreneurship and conversely for the reduced entrepreneurial alterness due to lack of exposure to a business setting (Hundt and Sternberg, 2014). Beyond individual-level considerations, at the regional level, an elevated unemployment rate may be perceived as an indicator of economic decline and decreased consumer demand (Armington and Acs, 2002; Audretsch et al., 2010; Hundt and Sternberg, 2014). High regional unemployment rates can be perceived by potential entrepreneurs as a signal that the probability of succeeding with start-up activities is lower (Stuetzer et al., 2014). While the results of previous research are inconsistent (Audretsch et al., 2010; Bosma et al., 2008; Sternberg and Fritsch, 2011; Wagner and Sternberg, 2004), consensus is emerging that higher rates of unem-ployment reduce the odds of individuals becoming entrepreneurs irrespectively of whether the

19

In robustness checks, we have also excluded respondents that did not indicate their income, see section A.3.2 and table A.10.

20

(27)

individual itself is employed or unemployed (Carree, 2002; Hundt and Sternberg, 2014; Sternberg and Fritsch, 2011; Stuetzer et al., 2014).21 Consequently, we control for the unemployment rate of people who are older than 25 years.22

Employment in large-scale industries has consistently been associated with lower levels of entrepreneurship (Audretsch et al., 2012; Fritsch and Wyrwich, 2016; Stuetzer et al., 2016; Wagner and Sternberg, 2004). Stuetzer et al. (2016) argue that this effect is due to lower levels of human capital, fewer opportunities to become entrepreneurially active and adverse institutional regimes. Indeed, the region’s industrial history has been found a highly important predictor of regional entrepreneurship (Audretsch et al., 2016; Stuetzer et al., 2016). Consequently, we control for employment in the industrial sector as a proxy for employment in large-scale industries.23

We also control for the share of population with tertiary education, since regional human capital has been shown to affect entrepreneurship through other channels than the entrepreneurs’ human capital (for which we control for at the individual level). Amongst them are firstly knowl-edge spillovers from universities or other companies (Acs et al., 2009; Audretsch, 1995; Audretsch and Keilbach, 2007). Moreover, indirectly, more educated stakeholders, such as customers and employees, positively contribute to entrepreneurship as shown by Millan et al. (2014).

Fritsch and Schroeter (2011) find that the share of R&D employment is an important predictor for the effect of entrepreneurship on regional growth. The regional knowledge stock is an important determinant of potential spillovers and underlies the ’knowledge spillover hypothesis’ and theories of endogenous firm formation (Acs et al., 2009; Audretsch, 1995; Audretsch and Keilbach, 2004a, 2007; Audretsch et al., 2005; Audretsch and Lehmann, 2005). It determines the kind of entrepreneurship in which individuals engage (Fritsch and Schroeter, 2011) as well as entrepreneurship rates by creating entrepreneurial opportunities. Thus, we control for the regional share of employment in R&D.

3.2 Methodology

To establish a meaningful relationship between culture and entrepreneurship, it is vital to account for formal institutions. Formal institutions create and constrain entrepreneurial opportunities and influence the probability of individuals’ trying to pursue them or not (Bosma and Schutjens, 2009). Amongst others, entrepreneurs can be confronted with poorly enforced

21

A more detailed discussion is beyond the scope of this paper. Please refer to Storey (1991) for a review, which despite the fact that it was published two and a half decades ago still contains highly relevant insights, in particular when comparing findings of cross-sectional and panel data studies. Reynolds et al. (2003) argue that conflicting findings in research may be due to pooling necessity and opportunity entrepreneurship. For an excellent regional panel study please refer to Fritsch and Mueller (2004).

22

We have chosen for using the unemployment rate of people older than 25 years rather than 18 years to not confound our control variable by -amongst others- the influence of differing education systems, which have an impact on youth and young adult unemployment. Unreported robustness check revealed that our results are invariant to using unemployment of people aged above 18 years.

23

Master Thesis Entrepreneurship in the European Regions - A Multi-Level Study -

Master Thesis

Entrepreneurship in the European Regions

A MultiLevel Study

Contents

List of Tables

List of Figures

1

Introduction

S

2

Entrepreneurship & culture

2.1

Definitions and operationalizations of entrepreneurship

2.2

Individual, regional or national entrepreneurship?

2.3

Regional culture and entrepreneurship

2.4

Hypotheses development

3

Data & methodology

3.1

Data

3.2

Methodology