Entrepreneurship Dynamics: Entry, Survival and Firm Growth

ENTREPRENEURSHIP DYNAMICS: Entry, Survival, and Firm Growth Vera Catarina Barros Rocha Tese de Doutoramento em Economia Orientada por: Professora Doutora Anabela Carneiro Professora Doutora Celeste Varum Porto, Maio 2014 i Biographical Note Vera Rocha was born on November 30, 1987, in Oliveira de Azeméis, Portugal. She concluded her undergraduate studies in Economics in 2008, at the University of Aveiro, with the final grade of 18/20. In 2010, she concluded her Master studies in Economics in the same university, with the dissertation “Performance and Survival of Foreign and Domestic Firms during crises” and a final grade of 19/20. During her undergraduate and master studies, she received two merit awards from Bank of Portugal and BPI, and four merit scholarships from University of Aveiro for her academic achievement. During her Master studies, Vera developed her research interests on the fields of empirical industrial organization, firm and industry dynamics, entrepreneurship and small businesses. Her Master dissertation resulted in some refereed publications in international journals, as Small Business Economics, Economics Letters and International Business Review. In January of 2010, Vera joined CIPES (Centre for Research in Higher Education Policies) as a research assistant. Since then, she has been also doing research on Higher Education topics. Her research at CIPES has also resulted in some publications in international journals, as Higher Education, Applied Economics, Public Administration Review, Journal of Economic Issues and Cambridge Journal of Regions, Economy and Society. In September of 2010, she started her PhD in Economics at the Faculty of Economics of University of Porto (FEP), under the supervision of Professors Anabela Carneiro and Celeste Varum. Her most recent research on entrepreneurship dynamics has been presented in scientific meetings and international conferences, as the 1st and 2nd International Entrepreneurship Autumn Schools (both in Évora, Portugal), the XXVIII Jornadas de Economía Industrial (Segovia, Spain), the 17th IZA European Summer School in Labor Economics (Buch/Ammersee, Germany) and the 2nd International Conference on the Dynamics of Entrepreneurship (ZEW, Mannheim, Germany). ii Acknowledgements After two long years of advanced Economics lectures, two more years of intensive research on entrepreneurship, hundreds of days of “incubation” in Quadros de Pessoal office, several discussions with my supervisors and subsequent re-re-re-revisions, about ten public presentations in seminars, workshops and conferences, it’s time to take a deep breath and to acknowledge everyone who somehow supported me during this period and contributed to finish this journey. First, I would like to acknowledge my supervisors, Professor Anabela Carneiro and Professor Celeste Varum, for all their help, comments and suggestions during this process, for their continuous encouragement and incentives, which certainly made me to go further and to believe that it is always possible to improve and to grow as a researcher. I also acknowledge GEE–MEE (Gabinete de Estratégia e Estudos – Ministério da Economia e do Emprego) for allowing the use of Quadros de Pessoal dataset, FCT (Fundação para a Ciência e a Tecnologia) for financial support through a doctoral grant (reference SFRH/BD/71556/2010), and CEF.UP for all the financial support provided in international conferences. All the work accomplished during my PhD would not be possible without their valuable support. A special “thank you” goes to CIPES (Centre for Research in Higher Education Policies), in particular to Professor Pedro Teixeira, for all the opportunities of joint work and learning, and for the financial support at particular moments of my PhD, namely for supporting my participation in the Summer School on “Discrete Choice Models for Cross-Section and Panel Data”. Among the colleagues from CIPES, I want to specially acknowledge Ricardo Biscaia, who has been a good friend and co-author in parallel projects. I have also made some good friends during my PhD, who certainly helped me to feel not alone and with whom I have spent very good moments. For all the nice lunches and coffee breaks, I would like to thank António, Joana, Nuno, Rita and Sara. All of those who have read and/or commented some parts of this thesis at conferences and seminars are also really acknowledged. I would like to highlight the valuable suggestions of Professors Francisco Lima, José Varejão, Helena Szrek and Nuno Sousa Pereira. Last, but not least, a huge “thank you” goes to Gonçalo and my parents, for all their love and patience. All my achievements only worth if I can share them with you. iii Resumo Esta tese inclui cinco ensaios sobre empreendedorismo. O primeiro ensaio descreve o modo como os tópicos relacionados com o empreendedorismo ganharam uma relevância crescente na investigação em economia ao longo do último século, e como a ciência económica contribuiu para o desenvolvimento gradual do empreendedorismo enquanto área de investigação. Os restantes quatro ensaios estudam diferentes questões relacionadas com as dinâmicas do empreendedorismo, com recurso a uma grande base de dados que combina informação detalhada do trabalhador e da respectiva empresa para Portugal. No segundo e terceiro ensaios, o indivíduo é a unidade de análise. No segundo ensaio estudam-se as dinâmicas de entrada e saída de mais de 157,000 indivíduos que deixam o seu emprego por conta de outrem e se tornam empregadores durante o período 1992-2007. Destacam-se nomeadamente dois contributos: em primeiro lugar, este estudo avalia como determinadas experiências passadas no mercado de trabalho influenciam as decisões de entrada e saída dos empreendedores nascentes; em segundo lugar, a análise presta particular atenção à natureza heterogénea dos empreendedores e tenta explicar as suas diferentes escolhas no que respeita ao modo de entrada e de saída. O terceiro ensaio analisa os efeitos de aprendizagem e de auto-selecção entre os “empreendedores em série”, utilizando uma estratégia empírica inovadora baseada em modelos de duração em tempo contínuo com selecção. Tendo sido identificados aproximadamente 220,000 indivíduos que deixam a primeira experiência como empreendedores, entre os quais cerca de 35,000 reentram uma segunda vez no empreendedorismo, este estudo avalia se a experiência adquirida no primeiro negócio melhora a sobrevivência dos empreendedores na segunda empresa, tendo em consideração os potenciais efeitos de auto-selecção na amostra. O quarto e quinto ensaios concentram-se nas empresas start-up, e em particular nas spin-offs. O quarto ensaio compara a sobrevivência das spin-offs de tipo pushed e pulled, tendo em conta um conjunto de condições iniciais onde estes dois tipos de empresas podem diferir. A análise cobre 50,656 spin-offs que entram durante o período 1992-2007 e utiliza técnicas de decomposição multivariada aplicadas a modelos de sobrevivência que permitem decompor o diferencial de sobrevivência observado entre as pushed e as pulled spin-offs em diferenças relacionadas com as suas iv dotações/características observadas e diferenças relacionadas com os retornos das mesmas. Por fim, o quinto ensaio estuda de que forma o crescimento do emprego, os fluxos de trabalhadores e a sobrevivências das spin-offs se relacionam com as suas dotações iniciais de capital humano. O estudo concentra-se em três medidas de capital humano à entrada – o nível médio de skills da força de trabalho, a dispersão de skills dos trabalhadores à entrada e a proporção de co-workers na força de trabalho inicial –, e mede os skills dos trabalhadores através de um índice multidimensional que tem em conta características observadas e não observadas do trabalhador. v Abstract This thesis comprises five essays on entrepreneurship. The first essay describes how entrepreneurship topics gained an increasing importance within economic research over the last century and reviews how the gradual development of the entrepreneurship research field is backed in economic science. The remaining four essays study different issues on entrepreneurship dynamics, using a large longitudinal matched employeremployee dataset for Portugal. The second and third essays consider the individual as the unit of analysis. The second paper studies the entry and exit dynamics of over 157,000 individuals who leave paid employment and become business-owners during the period 1992-2007. The contribution of this essay is two-fold: first, it evaluates how particular past experiences in the labor market influence the entry and exit decisions of nascent business-owners; second, it pays attention to the heterogeneous nature of business-owners and tries to explain their different modes of entry and exit. The third essay analyzes learning by doing and self-selection effects among serial entrepreneurs, using a novel empirical strategy based on continuous time duration models with selection. After identifying about 220,000 individuals who have left their first entrepreneurial experience and over 35,000 ex-business-owners who reenter again and become serial entrepreneurs, the study evaluates whether entrepreneurial experience acquired in the previous business improves serial entrepreneurs’ survival, after taking into account self-selection issues. The fourth and fifth essays focus on start-up firms, particularly on spin-offs. The fourth essay compares the survival of pushed and pulled spin-offs, taking into account a set of start-up conditions where they may differ. The analysis covers 50,656 spin-offs entering during the period 1992-2007 and uses novel multivariate decomposition techniques applied to hazard models to decompose the pushed-pulled survival gap into differences in endowments and differences in effects. Finally, the fifth essay investigates how spin-offs’ employment growth, worker flows and survival are associated to their initial human capital endowments. The study focuses on three measures of human capital at entry – workers’ average skills, their skill dispersion and the share of co-workers in the initial workforce –, and measures workers’ skills through a multidimensional skill index that takes into account both observed and unobserved characteristics of the worker. vi Contents Essay 1: Economics and the invisible entrepreneur: Tracing the path towards a new research field ...................................................................................................................... 1 1. Introduction .................................................................................................................. 3 2. Economists founding fathers of entrepreneurship research ......................................... 6 2.1. Joseph Schumpeter.................................................................................................... 7 2.2. Frank Knight ............................................................................................................. 8 2.3. Israel Kirzner ............................................................................................................ 9 2.4. Towards a theory of entrepreneurship – William Baumol ...................................... 11 3. (Re)Discovering the invisible entrepreneur in different fields of economics ............ 12 3.1. Labor Economics – the “occupational approach” ................................................... 12 3.2. Microeconomics and Industrial Organization – the “structural approach” ............ 14 3.3. Macroeconomics: Economic Growth and Development – the “functional approach” ....................................................................................................................... 17 4. Tracing the recent development of entrepreneurship research ................................... 20 4.1. Entrepreneurship research in Economics journals .................................................. 20 4.2. The footprints of economists founding fathers in entrepreneurship research ......... 29 4.3. The increasing institutionalization of entrepreneurship as a research field ............ 38 5. Concluding remarks ................................................................................................... 43 References ........................................................................................................................ 44 Essay 2: Nascent entrepreneurship dynamics: Entry routes, business-owners’ persistence and exit modes .............................................................................................. 53 1. Introduction ................................................................................................................ 54 2. Previous research on entrepreneurial entry and exit .................................................. 56 2.1. Past experiences in the labor market and BOs’ dynamics ...................................... 56 2.2. The heterogeneous nature of nascent BOs: entry routes and exit modes ............... 59 3. Data and methodological issues ................................................................................. 61 3.1. Data ......................................................................................................................... 61 3.2. Identifying transitions from paid employment into business-ownership ................ 62 3.3. Identifying the exits of business-owners and firms ................................................ 64 3.4. Empirical Strategy .................................................................................................. 65 3.4.1. The choice of becoming a BO ........................................................................ 65 vii 3.4.2. The persistence and the exit mode of the BO ................................................ 66 4. Empirical results on entrepreneurial entry ................................................................. 68 4.1. Characterizing the different groups of BOs ............................................................ 68 4.2. Empirical Results .................................................................................................... 70 4.2.1. Multinomial Logit estimation results ............................................................. 70 4.2.2. The effect of recent displacement .................................................................. 76 5. Empirical results on BOs’ exit ................................................................................... 78 5.1. Descriptive statistics and non-parametric analysis ................................................. 78 5.2. Estimation results from cause-specific hazard models ........................................... 84 6. Concluding remarks ................................................................................................... 90 References ......................................................................................................................... 91 Appendix ......................................................................................................................... 100 Essay 3: Serial entrepreneurship, learning by doing and self-selection ................... 107 1. Introduction .............................................................................................................. 108 2. Learning by doing and self-selection: two sources of serial entrepreneurship dynamics ................................................................................................................... 110 2.1. Previous literature on entrepreneurial learning ..................................................... 110 2.2. Research questions and objectives ........................................................................ 113 3. Data and methodological issues ............................................................................... 115 3.1. Data ....................................................................................................................... 115 3.2. Identifying the entry and exit of serial entrepreneurs ........................................... 117 3.3. Econometric model: specification and estimation ................................................ 118 4. Empirical results ....................................................................................................... 122 4.1. Descriptive statistics ............................................................................................. 122 4.2. Naïve Weibull estimation results .......................................................................... 128 4.3. Self-selection and serial entrepreneurs’ persistence ............................................. 132 5. Robustness checks .................................................................................................... 137 5.1. Estimation results for the sub-samples of start-up serial entrepreneurs and young ex-BOs ......................................................................................................................... 137 5.2. Estimating the person-specific effect of ex-BOs .................................................. 140 6. Concluding remarks ................................................................................................. 143 References ....................................................................................................................... 145 Appendix ......................................................................................................................... 152 “(.. .) one can say of the role of the entrepreneur in the mainstream mathematical writings of the …rm much what Mark Twain said of the weather – everyone talks about the subject but no one does anything about it. Every economist surely must be prepared to concede that entrepreneurs are (even if for reasons not fully speci…ed) of great importance. But in standard microtheory they are completely invisible.”Baumol (2003: 57) 1 Introduction Most people who are not economists would probably expect to …nd economics literature full of analyses of entrepreneurship, as economics is, in fact, the social science that deals most directly with contemporary economic reality. However, for long time economics literature had relatively little to say about entrepreneurship. Rewording Baumol (2010: 11), “the really important part of the story of economic well-being.. . was exorcised from the (theoretical) literature”. Despite the early attempts of Richard Cantillon to recognize the role of the entrepreneur …gure to economic phenomena early in the 18th century, the entrepreneur had virtually disappeared from mainstream economics by the end of the 19th century (Baumol, 1993, 2010). The theory’s failure to include the entrepreneur in the mathematical representations of economic reality was a consequence of the extreme simplifying assumptions of neoclassical models, where all agents had perfect information and their economic objectives were clearly stated (Casson, 2003; Bianchi and Henrekson, 2005; Montanye, 2006). These assumptions, which reduced the economic process to clocklike mechanics, overlooked the need for specialized individuals to perform the discovery, coordination, decision-making and riskbearing functions (Schumpeter, 1934; Barreto, 1989; Landström, 2005). As a result, the …rm was assumed to run itself, leaving no room for the entrepreneur in those models (van Praag, 1999). The theoretical …rm has, thus, remained entrepreneurless for long time (Bianchi and Henrekson, 2005), which was compared to a “performance of Hamlet with the Danish prince missing” in the words of Baumol (1968: 66). As a natural consequence – and con…rming that the ideas that cannot be modeled formally tend to be ignored in economics (Bianchi and Henrek3 son, 2005) –the terms “entrepreneur”and “entrepreneurship”are practically nonexistent in the leading graduate textbooks in micro, macro and industrial organization (Johansson, 2004). However, this neglect of the entrepreneur was often alleged to be a source of embarrassment to economists (Cosgel, 1996), as the importance of the entrepreneur in the real world became more and more di¢ cult to ignore (Wennekers and Thurik, 1999). From the early 20th century onwards, there was a “renaissance”of the entrepreneur …gure within economics and academics started to look at entrepreneurship phenomena through di¤erent perspectives. Almost all the branches of economics had something to say about the entrepreneur and his respective importance for some economic problem. Even so, there is, perhaps, no other area of economic analysis where there still is less agreement than on the entrepreneurship de…nition, the identi…cation of the entrepreneur …gure and the nature of the entrepreneurial function. In economic thought literature, we already …nd valuable interpretations of the ideas of particular economists about the entrepreneur …gure (see, for instance, Martin, 1979; Kanbur, 1980; Santarelli and Pesciarelli, 1990) or even about the reasons behind the entrepreneur’s disappearance from mainstream economic analysis (e.g., Barreto, 1989; Cosgel, 1996; Casson, 2003). Nonetheless, a deeper knowledge about how the entrepreneur (re)entered into economics through its several branches, and how the economists’(re)discovery of the invisible entrepreneur may have worked as a platform to build an increasingly autonomous and recognized …eld of research, is still lacking in the literature. This paper, thus, aims at contributing to both economics and entrepreneurship literature, by outlining how the gradual development of the entrepreneurship research …eld is backed in economic science. More than focusing on 4 particular visions of speci…c authors, or confronting similar or opposing views of di¤erent authors, this paper tries to, …rstly, provide a wider vision on the economists’necessity, over the last century, to include the entrepreneur …gure as a potential explanatory agent of several economic phenomena and, secondly, explore how early economists’research on entrepreneurship topics might have contributed to develop a new, and increasingly independent, institutionalized and recognized …eld of research.1 The following sections of the paper are organized as follows. As a starting point, section 2 pays homage to some of the most in‡uential economists who helped to bring the entrepreneur back into economics over the 20th century, summarizing their main ideas and contributions in this regard. Section 3 goes through the main economics …elds where the (re)discovery of the entrepreneur …gure was most remarkable –namely Labor Economics; Microeconomics and Industrial Organization; and Macroeconomics, more precisely Economic Growth and Development –searching for the rationality to include the entrepreneur …gure into the analyses of some economic problems. Section 4 provides a brief bibliometric analysis in order to provide a more quantitative overview of the evolution of entrepreneurship research and to highlight its roots in economics. More than showing the growing relevance of entrepreneurship research in economics journals, we identify the footprints of some key economists in subsequent developments of the …eld, besides uncovering the main signs of increasing institutionalization of entrepreneurship as an academic …eld. Section 5 concludes. 1By an "increasing (...) institutionalized (...) …eld of research" we mean the development of an institutional infrastructure that comprises new institutes and foundations promoting research on entrepreneurship, new journals and outlets attracting and publishing entrepreneurship studies, and emerging mechanisms that recognize and reward individual research on entrepreneurship topics (see also Aldrich, 2012). 5 2 Economists founding fathers of entrepreneurship research There are few issues in economics which are backed up by such a rich historical knowledge base as entrepreneurship (van Praag, 1999). The crucial role of the entrepreneur in economic theory was …rst and foremost recognized in the 18th century by Richard Cantillon (1755, 1931), who became the founding father of the ideas that subsequent economists explored. Cantillon recognized that discrepancies between demand and supply in a market create opportunities for buying cheaply and selling at a higher price, and that this sort of arbitrage would bring equilibrium to the competitive market. People who took advantage of these unrealized pro…t opportunities, even under the lack of perfect foresight of future impacts, were called “entrepreneurs”(Landström, 2005; Hébert and Link, 2006). After Cantillon, throughout the 19th century, a number of economists recognized the merit of entrepreneurial activity and of the entrepreneur …gure, in particular Mill, Say and Marshall (see, for instance, van Praag (1999) for a more detailed survey of these classic views on entrepreneurship). Nevertheless, even if the entrepreneur’s appearance was frequent in the writings of classical economists, he remained a shadowy entity without clearly de…ned form and function, for whom there was no room in economic theories, namely in the theoretical …rm (Baumol, 2010). In consequence, by the end of the 19th century, the entrepreneur had virtually disappeared from the sphere of economic debates and mainstream economics (Baumol, 1968; Swedberg, 2000; Landström et al., 2012). However, the reason for this disregard of the entrepreneur by economists 6 was not a denial of his relevance for economic development, or for the organization of economic activity, but mainly the methodological di¢ culties associated to the lack of analytical tractability of the entrepreneur …gure and his function(s) (Bianchi and Henrekson, 2005). Accordingly, from the early 20th century onwards, a number of valuable e¤orts were made in order to infuse some life to the "invisible entrepreneur" in the several branches of economics. Hébert and Link (1989) suggest that the taxonomy of entrepreneurial theories in economics can be condensed into three major intellectual backgrounds –German, Chicago and Austrian traditions –each one tracing its origin to Richard Cantillon. Within each of them, we emphasize the contributions of Joseph Schumpeter, Frank Knight and Israel Kirzner, respectively. Additionally, we highlight the noteworthy work of William Baumol, who has been struggling over the most recent decades to develop a framework for introducing entrepreneurship into mainstream microeconomic theory. We may suspect that without these contributions and confronts of ideas, entrepreneurship would have not get its deserved space in economics. 2.1 Joseph Schumpeter In the words of Reisman (2004: 3), Schumpeter means “entrepreneurship”. In The Theory of Economic Development (1934), he unveiled his own concept of entrepreneurship by giving a particular role to the entrepreneur …gure in the innovation process. The entrepreneur was, thus, treated as endogeneous for the …rst time. For Schumpeter, the economic system was regarded as a closed circular ‡ow where the stationary equilibrium was attained through the continuous reiteration of the ‡ows between buyers and sellers. Development, in 7 turn, was understood as a dynamic process that would require the disruption of the economic status quo –the so-called “creative destruction”. Basically, Schumpeter realized that economic growth and development resulted not from capital accumulation, but from innovation and “new combinations”(Landström, 2005; Hébert and Link, 2006). This fundamental role of innovating through the introduction of new products, markets or methods of production was given to the entrepreneur, who became responsible for the disturbance of the equilibrium in the economy. Precisely, he defended that “the carrying out of new combinations we call ‘enterprise’; the individual whose function is to carry them out we call ‘entrepreneurs”’(1934: 74). In summary, for Schumpeter, entrepreneurship was the expression of the human impulse to be creative (Khalil, 2007), the prime endogeneous cause of change (or, more precisely, development) in the economic system (van Praag, 1999), and, consequently, the source of permanent disequilibrium (i.e., crises) (Shane, 2003). Shumpeterian entrepreneurial rewards, however, are not permanent, ‡owing from the temporary monopoly rents that eventually arise from the successful introduction of those “new combinations” of ideas and resources. Following the same reasoning, entrepreneurship was understood by Schumpeter as a temporary condition for any person, unless s/he keeps on innovating. 2.2 Frank Knight After Schumpeter, Frank Knight (1921) stimulated one of the pioneer economic approaches of entrepreneurship of the 20th century. With his thesis Risk, Uncertainty and Pro…t, strongly inspired by Cantillon (Hébert and Link, 2006), he stressed the distinction between risk, uncertainty and true uncertainty, 8 defending that entrepreneurship is mainly characterized by action under true uncertainty.2 In few words, the Knightian entrepreneur is the uncertainty-bearer and the judgmental decision-maker that assumes the uninsurable business hazard (van Praag, 1999). Opportunities arise out of the uncertainty related to change and entrepreneurs are assumed to receive a return for making decisions under conditions of true uncertainty (Landström et al., 2012). Knight’s insights, hence, in‡uenced many economists in their analyses of entrepreneurship as an occupational choice problem (Parker, 1996, 2005, 2009; Montanye, 2006). In summary, Knight and Schumpeter had clearly contrasting interpretations of the entrepreneur, as well as of the risk and uncertainty he is exposed to. While the Knightian entrepreneur is essentially de…ned as the uncertaintybearer, the Schumpeterian entrepreneur is the dynamic innovator. For Schumpeter, risk-taking is no case an element of the entrepreneurial function (Schumpeter, 1934: 137), and even though entrepreneurs may risk their reputation, the direct responsibility of failure never falls on them (Martin, 1979; Kanbur, 1980). 2.3 Israel Kirzner The Austrian School also made a notable e¤ort to introduce the entrepreneur in mainstream economics. Restating Baumol (2003), all economists recognized the importance of entrepreneurship, but until the work of the Austrians, little 2According to Knight (1921), risk exists when outcomes are uncertain but can be predicted with some probability, being insurable; uncertainty arises when the probability of outcomes cannot be calculated; true uncertainty, instead, occurs when the future is not only unknown, but also unknowable with unclassi…able instances and a non-existent distribution of outcomes. 9 was done about it. Early inspired by the Knightian uncertainty, the Austrians’view on entrepreneurship (e.g., von Mises, 1949; Menger, 1950; Kirzner, 1973) also defended that the entrepreneur’s success or failure depends on the precision of his anticipation of uncertain events. Therefore, entrepreneurial pro…ts would be the result of entrepreneurs’“ability to anticipate better than other people the future demand of consumers”(von Mises, 1949: 290). Later, Kirzner (1973, 1997), one of von Mises’ students, introduced the key concepts of “spontaneous learning”and “alertness”, two requirements for the “entrepreneurial discovery”to occur. He explored the entrepreneurial element in a Crusoe situation (Kirzner, 1979: 168-169) in order to illustrate both concepts –as Robinson Crusoe (and every entrepreneur) becomes aware of his so-called entrepreneurial vision, he learns. However, this learning is not planned, but subconscious, and the state of mind that enables this “spontaneous learning”about the unrecognized entrepreneurial vision (what Kirzner calls the "subconscious hunch") is alertness. This refers to an attitude of receptiveness and preparedness to recognize unnoticed or unexploited profitable exchange (i.e., arbitrage) opportunities, corroborating the importance of the entrepreneur’s “information-transforming”function already defended by Hayek (1948). In addition, for the Austrian School, the opportunities for entrepreneurial pro…t are only available in disequilibrium, and the attainment of market equilibrium requires entrepreneurial action (Kirzner, 1971; Hébert and Link, 1989; Casson 2005; Endres and Woods, 2006; Khalil, 2007). Hence, the Austrian entrepreneur –understood as the equilibrating force in the economic system –is the antithesis of the Schumpeterian entrepreneur, who instead destroys the equilibrium and moves the economy towards a higher equilibrium position (Shane, 2003). 10 2.4 Towards a theory of entrepreneurship – William Baumol After the insights of Schumpeter, Knight and Kirzner (among others), the importance of the entrepreneur became more and more di¢ cult to ignore in economics. The peak of the discussion was achieved by William Baumol in a highly in‡uential article published in the American Economic Review, in 1968. His often quoted observation that “the theoretical …rm is entrepreneurless – the Prince of Denmark has been expunged from the discussion of Hamlet” (Baumol, 1968: 66) became the classic statement of the gap in economic theory regarding the (in)attention paid to the entrepreneur …gure. Baumol’s view on entrepreneurship pays homage to the insights of Schumpeter, namely on his ideas about the entrepreneur as an innovator and as the potential source of disequilibrium. Throughout his career, Baumol has urged the economists to pay attention to the instrumental role of entrepreneurship in economic renewal and growth (Elliasson and Henrekson, 2004), which may be both positive (productive) and negative (destructive), depending on the social bene…ts of entrepreneurs’innovations (Baumol, 1990, 1993, 2003). This led Baumol to work for years on the incentives under which judgmental decision-making takes place, with special reference to the issue of how far “rent-seeking”dominates entrepreneurial motivation under perverse incentive systems (Casson, 2005). His most recent works have been emphasizing the need for the correct incentives and the right institutional environment to promote productive and creative entrepreneurship –understood as the ultimate determinant of economic growth (Baumol, 2010). Over the most recent decades, Baumol has succeeded where many generations of economists since Richard Cantillon have failed –…nding the entre11 progress, promote growth and improve economic development (e.g., Wennekers and Thurik, 1999). Second, by founding and operating new businesses –even if there is nothing innovative in these acts –the entrepreneur was also expected to create value and new jobs, intensify competition and potentially increase productivity, which in turn may impact positively on the overall economy (e.g., Acs, 2006). However, despite this widespread belief that entrepreneurship was a key factor in economic growth and development, few attempts were made to incorporate entrepreneurship –and the entrepreneurial function in particular – in formal growth and development models until the early 1990s. Entrepreneurship did not …t in theoretical neoclassical growth models not only because perfect competition assumptions implied that there was no pro…t opportunities for entrepreneurs left, but also because the models of general equilibrium did not take into account the dynamics of the Schumpeterian –i.e., innovator –entrepreneur (Schmitz, 1989; Wennekers and Thurik, 1999). Endogenous growth theory has created new possibilities for …tting entrepreneurship (or entrepreneurial activities) into growth models, namely by emphasizing the role of knowledge and innovation for the growth of nations (e.g., Romer, 1986, 1990; Lucas, 1988). Knowledge externalities and increasing returns to scale were two of the key cornerstones of those new endogeneous growth models. These two processes appeared as a black box in the mainstream growth theory, which did not go very far toward illuminating the process by which knowledge externalities produce growth, or by which increasing returns can be manifested in the production process. The discovery of the crucial entrepreneurs’functions in the market process thus …lled this gap (Holcombe, 1998; Acs et al., 2004). More precisely, knowledge and innovation spillovers were recognized to not 18 befall automatically, requiring instead some channel(s) through which they could work and promote growth. Some mechanism was necessary to serve as a conduit for the spillover to occur. Hence, the entrepreneurial function –in particular, entrepreneurs’innovation and start-up activities –started being introduced in endogeneous growth models. Entrepreneurs, by being responsible for the conversion of knowledge into economically relevant knowledge, were thus declared to be the missing link in earlier models (Acs et al., 2004, 2009). Schmitz (1989) and Aghion and Howitt (1992) provided key advances by that time. The former developed a theory where the activity of entrepreneurs –and particularly the activities of imitating, rather than innovating –was shown to drive the growth of nations. The latter, instead, showed that industrial innovations conducted by entrepreneurs, by leading to quality improvements of products, were the key channel to induce progress and growth in the economy. In summary, incorporating entrepreneurship into the framework of economic growth helped to develop endogeneous growth theory mainly by shedding some light on the nature of increasing returns to scale, knowledge externalities and the role of human capital. Knowledge externalities arise when the entrepreneurial insights of some individuals produce entrepreneurial opportunities for others; increasing returns occur because the more entrepreneurial activity an economy displays, the more entrepreneurial opportunities it creates (e.g., Holcombe, 1998). Moreover, the new focus on entrepreneurship pushed the economic growth theory forward, towards the institutional setting within which growth occurs (Baumol, 1990), and away from neoclassical theories that focused on production process’inputs, as labor and capital. Entrepreneurship is already considered one of the key growth components in “new growth theory”(e.g., Audretsch et al., 2006; Henrekson, 2005). More recent concerns on the entrepreneurship-growth relationship have been related 19 to the quality of the entrepreneurship. Some of the latest extensions to existing models have been suggesting the need to encourage high-ability entrepreneurs, as low quality entrepreneurship is argued to retard growth (e.g., Jiang et al., 2010; Jaimovich, 2010). These new results stress the need to provide the right incentives to the most able entrepreneurs, in order to promote productive and growth-enhancing entrepreneurship, and avoid unproductive or even destructive entrepreneurial activities, in line with Baumol (1990, 2010). 4 Tracing the recent development of Entrepreneurship research This section provides a brief quantitative overview of the evolution of entrepreneurship research over the last decades, highlighting its roots in Economics. The analysis is based on thousands of articles from Scopus database published since the early 1970s and explores: i) how entrepreneurship topics have been achieving their space in Economics academic research over the years; ii) how the insights of some of the economists “founding fathers”of the entrepreneur have remained in‡uent in more recent entrepreneurship research; and iii) how fragmented entrepreneurship research currently is as a new research …eld. 4.1 Entrepreneurship research in Economics journals According to Landström et al. (2012), the recent evolution of entrepreneurship research can be described in three phases: a …rst take-o¤ phase during the 80s, a second growth phase after the early 1990s, and a …nal phase mainly 20 characterized by a search for the maturity of the …eld since the early 2000s. In order to illustrate this evolution, we started by performing a search in the Scopus database, by requiring the appearance of, at least, one of the following words or expressions in the publications’title, abstract and/or keywords: “entrepreneur”, “entrepreneurship”, “small …rm”, “start-up”, “self-employment”, “new venture”and “new …rm”.5This allowed us to identify a total of 42,593 articles published between 1970 and 2013. Out of these, 15,701 belong to the Scopus subject area “Economics, Econometrics and Finance” (or 8,444 articles if we exclude those who are also classi…ed in “Business, Management and Accounting”).6 Figure 1.1 illustrates the overall evolution of the total number of articles satisfying the criteria imposed in the search, both in absolute and relative terms. Figure 1.2 provides comparable data for the subject area of “Economics, Econometrics and Finance”. Overall, we con…rm that, despite the early e¤orts of a signi…cant number of economists to claim for more attention to the entrepreneur …gure, research on entrepreneurship topics remained 5Despite this seems to be somewhat restrictive, by imposing these criteria on the search process we are allowing to capture two particular aspects. First, by using these di¤erent combinations of keywords, it is more likely to include a wider number of studies using di¤erent de…nitions of entrepreneurship. As previously discussed, many studies have been linking entrepreneurship either to self-employed individuals (i.e., the entrepreneur in particular), or to the entrepreneurial …rm – which is commonly understood as being a small …rm, a start-up or a new venture, or even to innovation activities conducted by entrepreneurs and their …rms. Besides this, it is possible that di¤erent Economics …elds give a di¤erent relative importance to these di¤erent perceptions of the entrepreneurship phenomenon (for instance, IO may be more concerned with the new/start-up …rm, while Labor Economics may be more focused on the entrepreneur in particular), so by imposing these wider criteria we try to avoid a potential overrepresentation of one of these visions about entrepreneurship. Second, by imposing these keywords to appear in the publications’title, abstract and/or keywords, we are increasing the probability of collecting the publications that are really dealing with entrepreneurship topic, thus minimizing the inclusion of marginal publications (i.e., publications whose main focus is not directly related to entrepreneurship phenomena). 6These results were last accessed in February 2014. 21 relatively scarce until the late 1980s. After the mid-90s, we observe a “boom”in entrepreneurship research, overall and in Economics journals in particular. By 2000, around 1,400 articles ful…lling the criteria described above were published (about 300 in Economics journals). Ten years later, the respective numbers had already doubled. More than 4,000 articles (almost 800 in Economics journals) were published only in 2013. 01.5%0.5% 1% 2% Relative number of publications 01000 2000 3000 4000 Absolute number of publications 1970 1980 1990 2000 2010 Years Absolute number of publications Relative number of publications Fig. 1.1. Evolution of entrepreneurship research in absolute and relative terms - All Subject Areas (42,593 articles) In relative terms, we also con…rm that entrepreneurship topics have been occupying a more relevant space in academic research, and especially in Economics. By the early 80s, the articles focused on entrepreneurship topics 22 accounted for less than 1% of all publications in Economics journals. In most recent years, about 4% of all Economics articles have been concerned with the entrepreneurship phenomena, a much more signi…cant share than that observed in other areas (see Figure 1.1). 01% 2% 3% 4% Relative number of publications 0200 400 600 800 Absolute number of publications 1970 1980 1990 2000 2010 Years Absolute number of publications Relative number of publications Fig 1.2. Evolution of entrepreneurship research in absolute and relative terms - Economics, Econometrics and Finance (8,444 articles) Table 1 reports the journals identi…ed in Scopus database publishing the largest number of articles on entrepreneurship since 1970. The leading Economics journal publishing on entrepreneurship is Small Business Economics (SBE), a journal founded by David Audretsch and Zoltan Acs in the late 1980s. SBE has become one of the outlets of recognized reputation for researchers interested in entrepreneurship topics, currently presenting a broad scope that includes multiple analyses and perspectives of entrepreneurship phenomena. 23 The foundation of SBE is actually recognized as one of the …rst signs of the gradual institutionalization of entrepreneurship as a …eld of research (Landström et al., 2012). Table 1. Economics journals publishing more papers on entrepreneurship Journal Total # Articles Share Small Business Economics 503 6.0% Journal of Banking and Finance 220 2.6% World Development 205 2.4% Industrial and Corporate Change 193 2.3% International Journal of Industrial Organization 145 1.7% Applied Economics 137 1.6% European Economic Review 134 1.6% Economics Letters 124 1.5% Journal of International Economics 116 1.4% Journal of Development Economics 110 1.3% Journal of Public Economics 100 1.2% Applied Economics Letters 97 1.1% Journal of Comparative Economics 79 0.9% Total 2163 25.6% Notes: All articles (8,444) published in 1970-2013, in the Scopus category “Economics, Econometrics and Finance”(excluding “Business, Management and Accounting”), with at least one of the following expressions in their title, abstract or keywords: entrepreneur, entrepreneurship, small …rm, start-up, self-employment, new venture, new …rm. Besides SBE and some general-interest journals (as Applied Economics, EER and Economics Letters), we identify a number of more specialized jour24 nals –particularly in the areas of economic development (e.g., World Development and JDE), industrial organization (e.g., ICC and IJIO) and even …nance (namely the Journal of Banking and Finance) and international economics – accounting for a signi…cant part (over 25%) of the entrepreneurship research published in Economics journals over the last decades. Table 2 complements these data and summarizes the total number of articles on entrepreneurship published, so far, in some highly ranked generalinterest Economics journals, as well as in some relevant …eld journals.7The results con…rm that entrepreneurship questions have deserved signi…cant attention within top academic journals, as American Economic Review, Review of Economics and Statistics, Review of Economic Studies, Journal of Economic Theory and Quarterly Journal of Economics, among others. Even during the last decade, a very signi…cant number of articles dealing with entrepreneurship issues were published in these journals, con…rming that the entrepreneur …gure, the entrepreneurial …rm and/or the entrepreneurial function have, …nally, found its deserved space in mainstream Economics journals. At the same time, since the early 2000s a number of specialized Economics journals started to pay a greater attention to entrepreneurship. Development Economics and IO journals have played a prominent role, with Labor Economics journals somewhat lagging behind. These patterns may actually con- …rm the aforementioned shift in the economists’research interests from the individual entrepreneur towards the entrepreneurial …rm and the aggregate outcomes of entrepreneurial process. 7Among the journals identi…ed in Scopus database, we tried to identify both generalist and specialized highly ranked-journals following some of the most known international rankings, as Thompson Reuters JCR Impact Factor, ISI Web of Science h-index and SJR (SCImago Journal & Country Rank). 25 Table 2. Top Economics journals publishing on Entrepreneurship (generalist and specialized/…eld journals) Selective Top Journals # Papers 1970-2013 # Papers 2000-2013 Journal of Economic Theory 73 45 Economic Journal 52 40 Review of Economics and Statistics 46 39 American Economic Review 42 36 Review of Economic Studies 40 33 Economic Theory 37 31 Quarterly Journal of Economics 31 25 Journal of Political Economy 27 22 Econometrica 18 17 Labor Economics Journals Labour Economics 45 37 Journal of Labor Economics 33 20 Industrial Organization Journals Industrial and Corporate Change 193 152 J. of Econ. Behavior & Organization 155 96 Intern. J. of Industrial Organization 145 72 Review of Industrial Organization 111 68 RAND Journal of Economics 69 52 Journal of Industrial Economics 61 51 Macroeconomics and Economic Development Journals World Development 205 101 Journal of Development Economics 110 79 Journal of Comparative Economics 79 48 Economic Development Quarterly 76 64 Journal of Macroeconomics 38 26 Developing Economies 14 13 Notes: The …rst column reports the total number of articles published in each journal, out of the 8,444 articles included in the Scopus category “Economics, Econometrics and Finance”(excluding “Business, Management and Accounting”). The total number of articles published since 2000 in the same category (last column) was 6,658. 26 Finally, Figure 2 summarizes the main keywords of the entrepreneurship publications identi…ed in Scopus database since the early 70s.8As expected, “Entrepreneurship”and “Entrepreneur”are two of the most frequent keywords of those publications, con…rming that the criteria that we impose in the search is mostly picking up those publications closely related to entrepreneurship topics –either in a more occupational approach, or following a more structural or functional approach. Actually, the three approaches discussed above seem to be clearly identi…ed. 0200 400 600 800 Learning Performance Social Capital International Trade Industrial Development Banking Globalization Export Profitability Labor Market Foreign Direct Investment Competition Venture Capital Economic Growth Economic Development Employment R&D Human Capital/Education Productivity Investment Industrial Performance Firm size Self-employment SMEs Innovation Entrepreneur Entrepreneurship Fig. 2. Top keywords of economics articles on entrepreneurship 8The analysis is based on 12,686 articles classi…ed in the category “Economics, Econometrics and Finance”, published between 2000 and 2013. Almost half of these articles are also classi…ed into the category “Business, Management and Accounting”. The overall pattern of the main topics and keywords covered remains qualitatively unchanged when we exclude them from the database. 27 Table 3. The footprints of key economists in entrepreneurship research Journal # Articles Share Economic Journal 15 1.9% Entrepreneurship and Regional Development 14 1.8% Management Science 13 1.7% Labour Economics 13 1.7% Foundations and Trends in Entrepreneurship 10 1.3% International Journal of Entrepreneurship and Small Busines 10 1.3% International Small Business Journal 10 1.3% Organization Science 10 1.3% Total 231 29.8% A brief analysis of the keywords mostly used in each case con…rm the legacy left –in terms of lines of research –by each of those economists. Most of the articles linking the entrepreneur …gure to Schumpeter deal with topics related to innovation, economic growth and development. Those relying more on the Knightian entrepreneur frequently include “uncertainty”and “decision making” in their keywords. Moreover, more recent research on international entrepreneurship, small businesses’internationalization and globalization also seems to be inspired by Knightian uncertainty –not only due to the profusion of these keywords in the articles referring to the Knightian entrepreneur, but also due to the Knightian footprints identi…ed in journals as Journal of International Entrepreneurship and International Business Review (see Table 3). "Opportunity recognition", "knowledge" and "innovation" are among the keywords of those linking the entrepreneur to Kirzner, while those stimulated by Baumol typically pay attention to innovation, economic development, growth and the importance of the institutional framework. Lucas’followers, in turn, are more concentrated in self-employment and occupational choice 34 problems, besides their attention paid to human capital, education and labor market issues. Last but not least, both Birch’s and Jovanovic’s ideas about the entrepreneur and the entrepreneurial …rm seem to have stimulated new lines of research more concerned with Small and Medium Enterprises, …rm growth and the industrial performance of small businesses. Finally, we identify a number of authors who have been contributing to the development of the entrepreneurship research …eld and who have been relatively more inspired by the insights of those key economists –given the volume of articles where they associate the entrepreneur …gure to particular founding father(s) (see Figure 4). Following the taxonomy of Landström (2005), we can classify these authors as belonging to the Core Group of researchers, as they became over the last decade highly productive researchers in entrepreneurship and whose work has a substantial impact in the …eld.10 Zoltan Acs and David Audretsch are two of those “core researchers”–or “stars”, as Teixeira (2011) designates –in the entrepreneurship …eld. Besides the foundation of Small Business Economics journal, they have been working for several years on small …rms and innovation, as well as on regional policy and on the role of entrepreneurship to economic growth. Their works are frequently based on the Schumpeterian –innovative –entrepreneur, and their analyses of small …rms also follow the insights of more recent in‡uential economists as Lucas and Jovanovic, as Figure 4 illustrates. 10According to Landström (2005: 67), the researchers who have been publishing about entrepreneurship constitute a rather heterogeneous group, being possible to identify Ad-hoc transients, i.e., researchers who appear only once and whose publication within the …eld of entrepreneurship is a one-o¤ event; In‡uential transients, i.e. transient researchers who appear only once, but whose work is in‡uential for entrepreneurship research; Craftsmen, which are the researchers whose names tend to appear more frequently in entrepreneurship articles, meaning that they have stayed within the …eld for a longer period of time; and …nally a Core Group of high-impact researchers. 35 Similarly, Roy Thurik has been mainly focused on small businesses and industrial dynamics, as well as the link between entrepreneurship and the macro economy – thus following the approaches of several economists as Schumpeter, Baumol, Lucas and Jovanovic. Magnus Henrekson’s research interests, in turn, include the relationship between entrepreneurship, economic growth, and structural and technical change, which justi…es his relatively stronger reliance on the ideas of Schumpeter and Baumol. 0 3 6 9 12 15 18 Articles combining "entrepreneur" & "key economist's name" Ac s, Z.J ., Audrets ch, D .B., Henrek s on, M., Mc Mullen, J.S., Minniti, M., Parker , S.C., Sar as vathy, S.D., Shane, S., Shepherd, D.A., Thurik, R., Vivarelli, M., Wr ight, M., Zahra, S.A., Schumpeter Knight Kirzner Baumol Lucas Birch Jovanovic Fig 4. Researchers from the Core Group following the footprints of key economists Some of these “core researchers” have been working on several di¤erent questions at the same time. Maria Minnitti has been concerned with entrepreneur’s entry decision among women and minority groups, as well as with 36 the relationship between entrepreneurship, economic growth and institutions, being thus inspired by di¤erent seminal works. Scott Shane, instead, has been working on theory building and in the conceptualization of the entrepreneurship …eld, virtually covering all major aspects of the entrepreneurship phenomena –the individual(s), the opportunity, the organizational context, the environment and the entrepreneurial process –thus relying on the insights of many in‡uential economists in his works. Simon Parker, despite having been relatively more interested in individual-level analyses of the entrepreneurship phenomena for long time, has been providing diverse contributions to the development of the …eld with strong roots in Economics, as his recent book Economics of Entrepreneurship (Parker, 2009) con…rms. From our analysis, these were among the “core researchers”following more closely the footprints left by some of the key economists discussed throughout the paper. A number of other “stars”in entrepreneurship research also seem to frequently rely on some of those economists’views about the entrepreneur and the entrepreneurship phenomena. From Figure 4, we still identify the works of Mike Wright, Dean Shepherd, Marco Vivarelli, Je¤ery McMullen, Saras Sarasvathy and Shaker Zahra. Wrigth’s research has been concerned with venture capital, buyouts, habitual entrepreneurs and related topics. Shepherd’s research interests include entrepreneurial opportunity, entrepreneurial strategy and the failure of entrepreneurial businesses, while McMullen and Sarasvathy have been more concerned with the cognitive aspects of entrepreneurial action. Vivarelli has been covering several aspects of the dynamics and the innovation of newborn …rms and, …nally, Zhara has been provided important contributions on entrepreneurial knowledge and capability development in emerging 37 global industries, and, more recently, on international entrepreneurship.11 The list of “core researchers” or “stars” in the entrepreneurship …eld is far from being completed, as other names would deserve to be mentioned, as Mirjam van Praag, André van Stel, Per Davidsson, among many others. According to Teixeira (2011), entrepreneurship is an increasingly autonomous, legitimate and cohesive (in)visible college that may encompass from 50 (“stars” and “in‡uential”) up to 99 (reasonably in‡uential, including some “stars”) researchers. Some of them are among the winners of the Global Award for Entrepreneurship Research, others have been emerging over the most recent years, making the entrepreneurship …eld increasingly formalized and anchored in a small set of intellectual bases (Aldrich, 2012). One of them is certainly Economics. 4.3 The increasing institutionalization of entrepreneurship as a research …eld The growth of entrepreneurship as an academic …eld has been furthermore supported by the emergence of institutes and foundations promoting research on entrepreneurship topics, the creation of specialized entrepreneurship journals, the establishment of research awards distinguishing academic “superstars”doing research on entrepreneurship and small businesses, as well as the development of high-pro…le conferences encouraging further research on this subject. 11This information was collected, whenever possible, from the authors’personal webpages and CVs, and/or by inspection of their main publications. 38 Not only scholars, but also governments and policy systems became increasingly interested in entrepreneurs and small …rms over the last years. In 2011, OECD launched “Entrepreneurship at a glance”, a new yearly publication that collects and discusses a number of indicators measuring the state of entrepreneurship around the world. European Commission has also been extremely attentive to entrepreneurship issues, continuously developing new programs and funding mechanisms to support small businesses and to encourage further research on the topic.12 In the United States, Kau¤man Foundation has been playing a crucial role over the last decades, by increasing funding and creating new opportunities for the development of institutional structures supporting entrepreneurship research and collecting high quality data.13 Since 2008, U.S. Small Business Administration has also been conducting and promoting research on entrepreneurship topics. Several international institutions and research centers became aware of the importance of entrepreneurship and started promoting research on the topic. NBER created an Entrepreneurship Working Group in 2003, with the support of Kau¤man Foundation, to conduct several projects related to the so-called “economics of entrepreneurship”. In Europe, the ZEW’s Industrial Economics research group has been strongly involved in the study of start-ups and entrepreneurship dynamics, promoting regular workshops and conferences on the topic. Max Planck Institute of Economics had also a temporary research group focused on entrepreneurship, growth and public policy, where both Zoltan Acs and David Audretsch played key roles as founders and directors, respectively. Some research centers specialized in entrepreneurship research have also been 12See, for example, the Entrepreneurship and Innovation Program, recently launched under the Competitiveness and Innovation Framework Program running since 2007. 13For instance, the recent e¤orts of The World Bank to construct The Entrepreneurship Database have been highly supported by Kau¤man Foundation. 39 emerging around the world, as the Center for Entrepreneurship and Public Policy in US; the LMU Entrepreneurship Center and the TUM Entrepreneurship Research Institute, both in Germany; the Amsterdam Center for Entrepreneurship and the Erasmus Center for Entrepreneurship Research, both in the Netherlands; and the Entrepreneurship and Small Business Research Institute in Sweden, just to name a few. As entrepreneurship research became more institutionalized and sophisticated, new academic journals dedicated to entrepreneurship research started being established from the early 80s onwards, becoming the main outlet for entrepreneurship papers. If, in the past, there were several journals in mainstream economics publishing about entrepreneurship issues, during the 90s and 2000s most of them either disappeared or declined in the rankings as new specialized journals have been founded. As Table 4 shows, at least 20 journals specialized in entrepreneurship and small businesses issues were launched during the last three decades. High-quality works also started being prized. The best-known award is the Global Award for Entrepreneurship Research (early known as the International Award for Entrepreneurship and Small Business Research), granted by the Swedish Entrepreneurship Forum since 1996. Table 5 lists the winners of the 19 prizes bestowed so far. Some of the aforementioned founding fathers and core researchers were already distinguished by this prize, namely Baumol, Kirzner, Birch, Acs, Audretsch, Shane and Zahra. Other core researchers have been contributing to the development of entrepreneurship as an academic …eld, either by highlighting the macro importance of new and small …rms (namely Storey, Reynolds, Beccatini and Sabel, Klepper and Feldman), or for their micro-level analyses of entrepreneurship and small businesses (in particular, Cooper, MacMillan, Aldrich, Gartner, Johannisson, 40 the Diana group, Lerner and Eisenhardt). Despite, overall, they come from an eclectic mix of disciplines (including management, sociology, political science and psychology), the great majority of them have their roots in economics. This con…rms, once more, the signi…cant contribution of economic science in the gradual development of the entrepreneurship research …eld. Table 4. Selected journals specialized in entrepreneurship research Foundation Year Impact Factor Journal of Small Business Managementa1962 1.333 International Small Business Journal 1982 1.469 Journal of Small Business & Entrepreneurship 1983 n.a. Journal of Business Venturing 1986 2.976 Entrepreneurship Theory & Practice 1988b2.242 Small Business Economics 1989 1.130 Entrepreneurship and Regional Development 1989 1.333 Journal of Entrepreneurship 1992 n.a. Academy of Entrepreneurship Journal 1995 n.a. Journal of Development Entrepreneurship 1996 n.a. International Journal of Entrepreneurship 1996 n.a Journal of Entrepreneurship Education 1997 n.a. Int. J. of Entrepren. and Innovation Management 2001 n.a. Journal of International Entrepreneurship 2003 n.a. Int. J. of Entrepreneurship and Small Business 2004 n.a. Int. Entrepren. and Management Journal 2005 5.053 Foundations and Trends in Entrepreneurship 2005 n.a. Strategic Entrepreneurship Journal 2007 1.205 Journal of Social Entrepreneurship 2010 n.a. Journal of Innovation and Entrepreneurship 2012 n.a. Notes: aJournal published on behalf of The International Council for Small Business (ICSB). bIn 1988, the American Journal of Small Business changed its name to Entrepreneurship Theory & Practice. n.a.: not applicable. In summary, entrepreneurship started out as a young – and even mar41 ginalized – …eld, where a mix of economists, psychologists, geographers and also the occasional anthropologist came together to study the wonder and weirdness that is entrepreneurship, in a wide range of fashions and with a few prior assumptions (Rehn et al., 2013), being thus considered to be a mere sub-discipline of management or economics (Teixeira, 2011). Nowadays, despite the …eld still shows strong signs of eclecticism and fragmentation, it has matured and became popular and increasingly institutionalized, revealing a greater legitimacy as a valid academic research area. Table 5. Winners of the Global Award for Entrepreneurship Research Year Winner 1996 David Birch 1997 Arnold Cooper 1998 David Storey 1999 Ian MacMillan 2000 Howard Aldrich 2001 Zoltan Acs and David Audretsch 2002 Giacomo Becattini and Charles Sabel 2003 William Baumol 2004 Paul Reynolds 2005 William Gartner 2006 Israel Kirzner 2007 The Diana Projecta 2008 Bengt Johannisson 2009 Scott Shane 2010 Josh Lerner 2011 Steven Klepper 2012 Kathleen Eisenhardt 2013 Maryann Feldman 2014 Shaker Zahra Notes: aThe Diana Project was composed by Candida Brush, Nancy Carter, Elizabeth Gatewood, Patricia Greene and Myra Hart. 42 5 Concluding Remarks Entrepreneurship is a concept that has gone through many changes, developed greatly during the last decades, and achieved an exceptionally important place both in contemporary academia and in modern public discourse (Rehn et al., 2013). In this paper, we have reviewed the main paths through which the entrepreneur …gure entered into Economics throughout the 20th century, outlining how the gradual development of the entrepreneurship research …eld is backed in economic science. In spite of its rich historical base (van Praag, 1999), entrepreneurship remains the phenomenon which is most emphasized but least understood by economists (Kanbur, 1980; Montanye, 2006). Though the …rst debates about the entrepreneur …gure have emerged in the 18th century, mainstream economics –and microeconomic theory in particular –omitted the entrepreneur …gure for long time, leaving no room for an active entrepreneur in neoclassical models (Baumol, 1993; Johansson, 2004). However, throughout the 20th century, the relevance of the entrepreneur became more di¢ cult to ignore, especially after a number of highly in‡uential economists have recognized his role in the labor market and industry dynamics, innovation, economic development and growth. The entrepreneur, by being endowed with creative talent and innate ability (Lucas, 1978; Jovanovic, 1982), learning capacity and alertness (Kirzner, 1979, 1997) to the pro…table opportunities in the market, started being understood as the responsible for economic progress (Schumpeter, 1934) and job creation (Birch, 1979), though facing true uncertainty and uninsurable risks (Knight, 1921). As a result, the entrepreneur – hitherto treated as an invisible …gure in economic models – has gradually gained a more signi…cant space of analy43 [53] Lucas, R. E. (1988), “On the mechanism of economic development”, Journal of Monetary Economics, 22(1): 3–42. [54] Martin, D. T. (1979), “Alternative views of Mengerian entrepreneurship”, History of Political Economy, 11(2): 271-285. [55] Menger, C. (1950), Principles of Economics. Translation of the …rst edition (1871) and edited by James Dingwall and Burt F. Hoselitz. Glencoe, Illinois: Free Press. [56] Montanye, J. A. (2006), “Entrepreneurship”, The Independent Review, 10(4): 549-571. [57] Parker, S. C. (1996), “A time series model of self-employment under uncertainty”, Economica, 63(251): 459-475. [58] Parker, S. C. (2005), “The economics of entrepreneurship: what we know and what we don’t”, Foundations and Trends in Entrepreneurship, 1(1): 1-54. [59] Parker, S. C. (2009), The Economics of Entrepreneurship. Cambridge: Cambridge University Press. [60] Rehn, A., Brännback, M., Carsrud, A., Lindahl, M. (2013), “Challenging the myths of entrepreneurship?”, Entrepreneurship & Regional Development: An International Journal, 25(7-8), 543-551. [61] Reisman, D. (2004), Schumpeter’s Market: Enterprise and Evolution. Cheltenham: Edward Elgar. [62] Romer, P. M. (1986), “Increasing returns and long run growth”, Journal of Political Economy, 94(5): 1002–1037. [63] Romer, P. M. (1990), “Endogenous technological change”, Journal of Political Economy, 98(5): 71–102. 50 [64] Roy, A. (1951), “Some thoughts on the distribution of earnings”, Oxford Economic Papers, 3(2): 135-146. [65] Santarelli, E., Pesciarelli, E. (1990), “The emergence of a vision: the development of Schumpeter’s theory of entrepreneurship”, History of Political Economy, 22(4): 677-696. [66] Santarelli, E., Vivarelli, M. (2007), “Entrepreneurship and the process of …rms’entry, survival and growth”, Industrial and Corporate Change, 16(3): 455-488. [67] Schmitz, J. A. (1989), “Imitation, entrepreneurship, and long-run growth”, Journal of Political Economy, 97(3): 721-739. [68] Schumpeter, J. A. (1934), The Theory of Economic Development. Cambridge, MA: Harvard University Press (…rst edition in German: 1911). [69] Shane, S. (2003), A general theory of entrepreneurship: the individualopportunity nexus. Cheltenham: Edward Elgar. [70] Stiglitz, J., Weiss, A. (1981), “Credit rationing in markets with imperfect information”, American Economic Review, 71(3): 393–410. [71] Storey, D. B. (1994), Understanding the small business sector. London: Routledge. [72] Storey, D. B., Tether, B. S. (1998), “Public policy measures to support new technology-based …rms in the European Union”, Research Policy, 26(9): 1037-1057. [73] Swedberg, R. (2000), “The social science view of entrepreneurship: introduction and practical applications”. In R. Swedberg (Ed.), Entrepreneurship – The Social Science View (pp. 7-44). Oxford: Oxford Management Readers. 51 [74] Teixeira, A. C. (2011), “Mapping the (in)visible college(s) in the …eld of entrepreneurship”, Scientometrics, 89(1): 1-36. [75] van Praag, M. (1999), “Some classic views on entrepreneurship”, De Economist, 147(3): 311-335. [76] van Stel, A., Storey, D. J., Thurik, R. A. (2007), “The e¤ect of business regulations on nascent and young business entrepreneurship”, Small Business Economics, 28(2-3): 171-186. [77] von Mises, L. V. (1949), Human Action: A Treatise on Economics. London: Hodge. [78] Wennekers, S., Thurik, R. (1999), “Linking entrepreneurship and economic growth”, Small Business Economics, 13(1): 27-55. 52 Nascent Entrepreneurship Dynamics: Entry Routes, Business-Owner’s Persistence and Exit Modes Essay 2 May 29, 2014 Abstract This paper uses a large longitudinal matched employer-employee dataset to conduct a comprehensive study on the dynamics of nascent business-owners. We identify and follow about 157,500 individuals who leave paid employment and become business-owners during the period 1992-2007. The aim of the paper is two-fold. First, we evaluate how particular labor market experiences in the past in‡uence the entry and exit decisions of nascent business-owners. Second, we pay attention to the heterogeneous nature of business-owners and try to explain their di¤erent modes of entry and exit. At entry, we distinguish between start-up entrepreneurs, acquisition entrepreneurs and intrapreneurs. At exit, we use discrete-time hazard models to study two cause-speci…c hazards: dissolutions and ownership transfers. We …nd that start-up entrepreneurs entering alone are more prone to dissolve the business, while ownership transfers are more incident among acquirers and shared ownerships. A recent job loss is found to push individuals into entrepreneurship and to improve their survival in the …rm. A past job in a large or foreign-owned …rm, instead, seems to increase the opportunity cost of entering and persisting in entrepreneurship. Business-owners’unobserved heterogeneity is also found to play a relevant role, particularly in the duration dependence of each exit mode. Keywords: Entrepreneurship, Business Ownership, Entry, Exit JEL Codes: J24, L26, M13 We are grateful to José Varejão and Francisco Lima for their comments and suggestions on previous versions of this paper. 53 1 Introduction The widespread belief that entrepreneurs are the engine of the market economy, by introducing new innovations, fostering economic growth and creating new jobs (e.g., van Praag and Versloot, 2007), has been motivating great waves of government support around the world encouraging entrepreneurship and the creation of new businesses (e.g., Román et al., 2013). This trend became even more evident during the most recent years, especially since the …nancial and economic crisis of 2008, under a general consensus among academics and policy makers that entrepreneurship may be a promising response to the growing unemployment rates, particularly among the youth (Congregado et al., 2010; Fairlie, 2013; Millán et al., 2014a). Nevertheless, many policies have been focused on the necessity to “produce”more entrepreneurs, but not so much on the necessity to preserve the stock of entrepreneurs (European Commission, 2011). Moreover, while there is already widespread scienti…c research on entrepreneurial entry decision (Parker, 2009a), data limitations have forced most of this literature to disregard what happens after entry, thus leaving out the dynamic aspects of entrepreneurship (Parker and Belghitar, 2006; DeTienne, 2010; DeTienne and Cardon, 2012). Besides, there still are two particular gaps in this scarce literature on nascent entrepreneurship dynamics. First, despite seminal theories claim that prior labor market experience may motivate entrepreneurial entry (e.g., Lucas, 1978; Lazear, 2004), little is known about how particular experiences in paid employment may shape nascent entrepreneurs’entry and exit decisions. To date, the literature has been mostly concerned with the e¤ect of past unemployment experiences (see Evans and Leighton, 1990; Carrasco, 1999; Earle and Sakova, 2000; Reize, 2000; Millán et al., 2014b). However, as the entrepreneurial process consists of distinct activities, including opportunity identi- …cation and resource mobilization (Shane and Venkataraman, 2000), we also expect that individuals’path in the labor market might enable them to accumulate speci…c knowledge and resources, to recognize market opportunities 54 and, consequently, to engage in entrepreneurial initiatives. Job shifting and past employment experiences in large-sized and/or foreign-owned …rms are two of the aspects of individuals’career history that may a¤ect entrepreneurship dynamics, and about which we still have limited knowledge from the literature. Second, most of the entrepreneurship literature has been treating nascent entrepreneurs as a homogenous group of individuals, with entrepreneurial entry commonly corresponding to the start-up of a new venture (with or without employees) (see Parker, 2009a; Parker and van Praag, 2012), and considering entrepreneurial exit to be equivalent to …rm exit. However, starting a new …rm is not the only way individuals can become entrepreneurs – they can also take over an existing …rm –, and entrepreneur’s exit does not necessarily correspond to …rm closure, as entrepreneurs can exit their business while the …rm continues operating under the ownership of other entrepreneur(s). Hence, this paper contributes to the existing literature in a number of ways. First, we overcome most of the data limitations faced by previous studies by using Quadros de Pessoal (henceforth, QP), a large longitudinal matched employer-employee administrative dataset that allows us to track 157,587 individuals who have left paid employment and became business-owners (BOs) during the period 1992-2007. Second, we pay particular attention to the role played by individuals’past experiences in the labor market in their entrepreneurial entry and exit decisions. Finally, we analyze both entry and exit dynamics and recognize that nascent BOs are not homogeneous, by allowing their entry and exit to assume di¤erent forms. At entry, we distinguish between BOs entering via start-up and those entering by acquiring an existing business. Among these, owing to the linked employer-employee nature of our data, we also distinguish between entrepreneurs and intrapreneurs. At exit, we employ discrete-time duration models to study cause-speci…c hazard rates –namely dissolutions and ownership transfers. The remaining sections of the paper are structured as follows. Section 2 summarizes prior …ndings of existing literature on entrepreneurs’entry and exit 55 and establishes the objectives of the paper. Section 3 describes the data, the methodological procedures to identify BOs’entry and exit, and the empirical strategy. Empirical results on BOs’entry and exit are presented and discussed in sections 4 and 5, respectively. Section 6 concludes. 2 Previous research on entrepreneurial entry and exit 2.1 Past experiences in the labor market and BOs’dynamics The decision of entering entrepreneurship has been analyzed during the last decades by an extensive literature under the framework of occupational choice models (see Parker, 2009a). More recent studies have been emphasizing the importance of several variables that may a¤ect the decision of running a business instead of receiving a more stable wage in paid employment, including numerous individual-level speci…cities as gender, age, education (e.g., Livanos, 2009; Berglann et al., 2011) or ability (e.g., Joona and Wadensjö, 2013; Poschke, 2013), unemployment episodes (e.g., von Grei¤, 2009), prior employer’s characteristics (e.g., Hyytinen and Maliranta, 2008; Parker, 2009b) and macroeconomic conditions (e.g., Koellinger and Thurik, 2012). Entrepreneurial exit, in turn, was a topic systematically disregarded in many studies for long time, not only due to data limitations, but also because a great part of entrepreneurship literature suggested that the entrepreneurial process is complete as soon as the new venture is created and ready to operate in the market (DeTienne, 2010). However, the entrepreneurial process is more than just the creation (or acquisition) of a business and does not end with entrepreneur’s entry, but rather with entrepreneur’s exit. Over the last years, a number of studies have been trying to …ll this gap by searching for potential explanations on why some entrepreneurs survive longer 56 in the business than others, using the individual as the unit of analysis. Entrepreneurs’age, gender and education (e.g., Block and Sandner, 2009), their past experiences in unemployment (e.g., Carrasco, 1999; Taylor, 1999; Andersson and Wadensjö, 2007), some characteristics of their businesses (Parker and Belghitar, 2006; Stam et al., 2010) and the overall economic environment (e.g., Haapanen and Tervo, 2009; Millán et al., 2012) are some of the determinants that have been found to a¤ect the length of time an individual persists as an entrepreneur. Even so, we still lack substantial knowledge on other types of determinants, as those related with individuals’past experiences in the labor market. Individuals’career history, by allowing the absorption of speci…c knowledge, the accumulation of contacts and networks, and by potentially a¤ecting both future labor market prospects and the identi…cation of business opportunities, may also shape entrepreneurial entry and exit. So far, the literature has paid particular attention to unemployment experiences, arguing that entrepreneurship is frequently regarded as an alternative to uncertain future career prospects or even to escape from unemployment (Evans and Leighton, 1990; Storey, 1991; von Grei¤, 2009; Millán et al., 2014b). However, while there is widespread evidence that unemployment episodes push individuals towards entrepreneurship, many studies have also reported that entrepreneurs with past unemployment periods are more likely to fail (e.g., Carrasco, 1999; Taylor, 1999; Andersson and Wadensjö, 2007; Millán et al., 2012). Thus, in this study, we analyze how a recent job loss (caused by previous employer’s closure or signi…cant downsizing) a¤ects nascent BOs’entry and exit decisions. Additionally, we pay particular attention to the role of job shifts in the past and employment experiences in large-sized or foreign-owned …rms. There is already evidence that the ‡ow of people between organizations and di¤erent contexts works as an important mechanism for knowledge transfers and skill development (e.g., Song et al., 2003; Frederiksen and Wennber, 2011). In view of that, job shifting may endow individuals with a more diversi…ed 57 set of skills, information and social capital, which not only may make them more likely to become entrepreneurs (Lazear, 2004), but also more able to survive longer in the business. Conversely, a larger number of di¤erent employers in the past may, instead, signal an unstable or unsuccessful path in paid employment, possibly associated to low human capital or ability, which may also motivate entrepreneurial entry, but possibly harm post-entry persistence if entrepreneurship is viewed as a last resort solution (Millán et al., 2014b).1 Accordingly, we test whether and how the number of di¤erent jobs/employers in the past in‡uences the individuals’probability of entering into and exiting from entrepreneurship. Finally, we also explore the role of previous jobs in large-sized or foreignowned companies. Though the literature has been suggesting that previous employer’s size matter, with smaller …rms being understood as places of entrepreneurial learning (e.g., Hyytinen and Maliranta, 2008; Parker, 2009b), other studies also argue that entrepreneurial opportunities and resources accrue to incipient entrepreneurs as a function of the structural position and visibility of their prior employers (e.g., Burton et al., 2002). Accordingly, on the one hand, we could expect that an employment experience in a large or foreign …rm, by possibly providing the new entrepreneur more knowledge, reputation and legitimacy, supports entrepreneurial entry and improves post-entry persistence. On the other hand, such type of experiences in the labor market, by frequently being appreciated by subsequent employers (Sørensen, 2007; Sørensen and Phillips, 2011; Balsvik, 2011), may increase the opportunity cost of leaving paid employment, thus reducing the propensity of entering entrepreneurship and/or accelerating nascent BOs’exit. The lack of empirical evidence on these relationships does not allow the formulation of precise expectations on the e¤ect of these variables. 1In this regard, there is increasing evidence that both high-ability and low-ability agents become BOs (e.g., Joona and Wadensjö, 2013; Poschke, 2013). 58 2.2 The heterogeneous nature of nascent BOs: entry routes and exit modes The literature has been largely de…ning entrepreneurship as self-employment or new venture creation. However, starting a new …rm from scratch is not the only way individuals can get into entrepreneurship. Budding entrepreneurs can also take over an existing …rm, though very few studies have been concerned with this issue. Nonetheless, there are good reasons to believe that entrepreneurs entering via start-up di¤er from those entering by acquiring an existing business. Acquisition can be viewed as an easy mode of penetrating a new market, besides allowing the potential entrant to take advantage of existing facilities, customer base and networks. In contrast, those who decide to install a new venture are faced with time-consuming and risk-taking activities, like building plants, learning the market or training employees (Tarola et al., 2011; Tarola, 2013), besides being more exposed to the liability of newness and smallness (Brüderl and Schüssler, 1990). Also, problems of asymmetric information are more acute in new venture start-ups compared with established …rms, which can be acquired by an outside investor or even by one of the …rm’s employees (Parker and van Praag, 2012). Furthermore, we may also expect that di¤erent learning opportunities (Jovanovic, 1982) about the whole entrepreneurial process are associated to each of those entry alternatives. Accordingly, entrepreneurial entry should not be understood as a homogeneous phenomenon, as di¤erent entry routes may signal di¤erent pro…les of BOs, driven by di¤erent motivations and having distinct post-entry behaviors. A few recent studies actually show that the mode of entry into entrepreneurship is in‡uenced by individual characteristics, as human, social or …nancial capital (see Parker and van Praag, 2012; Bastié et al., 2013; Block et al., 2013). Thus, in this paper, we distinguish between new entrepreneurs entering via start-up or acquisition, also paying attention to intrapreneurs, a particular group of BOs more frequently neglected by the literature on entrepreneurial entry (see 59 way depending on the type of BO they become, we estimate a multinomial logit model, where the outcome yfor individual imay be one of four alternatives: 1) Never BO; 2) Start-up entrepreneur; 3) Acquisition entrepreneur; or 4) Intrapreneur. Thus, and assuming independent and identically distributed (i.i.d.) extreme value distributed error terms, the probability that the outcome for individual iis alternative j, conditional on a vector of variables Xiis pij =eX0 ij 4 P l=1 eX0 il ; j = 1; :::; 4:(1) Table A.I in the Appendix provides a detailed description of the variables included in vector Xi. 3.4.2 The persistence and the exit mode of the BO To study BOs’persistence and exit decisions, we rely on duration models to study cause-speci…c hazards (dissolution and OT). Over again, we focus on the e¤ects arising from BOs’previous experiences in the labor market while paid employees, in addition to the BOs’entry mode. Individual-level characteristics and several characteristics of BOs’…rms (namely size, sector, age and location) are also taken into account in our estimations (see Table A.I for a detailed description of these variables). As survival spells are recorded in an annual basis, discrete time duration models were considered. The length of each individual’s spell as BO (Ti) is therefore assumed to be a discrete non-negative random variable. Moreover, we go beyond most of the previous research on entrepreneurial survival using discrete hazard models (e.g., Carrasco, 1999; Taylor, 1999; Nziramasanga and Lee, 2001; Block and Sandner, 2009; Millán et al., 2012) by incorporating the e¤ect of individuals’unobserved heterogeneity, which is known to mainly a¤ect the in‡uence of time dependence on the exit rate (e.g., Heckman and Singer, 1984; Lancaster, 1990; Jenkins, 2005). 66 Formally, we observe BO i’s spell from period j= 1 (the year of BO’s entry) until the end of the jth period, at which BO i’s spell is either complete (ci= 1) or right-censored (ci= 0) (‡ow sample). The overall probability of exit at discrete time tj,j= 1;2; : : :, given survival until time tj, can be de…ned as hij = Pr(Ti=jjTij) = F((t) + X0 i(t)+"i);(2) where hij is the probability of individual ipersisting as BO in the …rm for exactly jyears; (t)describes the pattern of duration dependence (the baseline hazard); Xi(t)is the vector of time dependent and independent variables; is a vector of unknown parameters to be estimated; "iis a disturbance term that includes the time-invariant unobserved heterogeneity (also known as individual “frailty”) and that is assumed to be uncorrelated with the observable variables of vector Xi(t)(Jenkins, 1995; Cameron and Trivedi, 2005: 613); and, …nally, F()denotes the complementary log-logistic distribution function. We do not impose any functional form for (t). We instead estimate a piecewise constant hazard model, where exit rates are assumed to be constant within each interval (year) but di¤erent between intervals. Thus, in order to estimate the full set of ’s, we have added an indicator variable per duration time tto the model. This ‡exible (non-parametric) modeling has been recognized to be preferred in order to avoid serious misspeci…cations. Moreover, such hazard formulation with a ‡exible baseline hazard function makes an attractive model with which to combine a speci…c heterogeneity assumption (Cameron and Trivedi, 2005: 620). Accordingly, following usual conventions (e.g., Hougaard, 1995; Jenkins, 2005), we assume an Inverse Gaussian distribution for the unobserved heterogeneity term, so that "iis normally distributed with zero mean and unitary variance. Summing up, the discrete time hazard function in (2) may be rewritten as follows: 67 hij = 1 expf exp[(t) + Xi(t)0+ log("i)]g:(3) We then extend this model in order to estimate cause-speci…c hazards for dissolutions and OT. Following the procedures of some previous studies (e.g., Carrasco, 1999; Reize, 2000; Georgarakos and Tatsiramos, 2009), the parameters of a given cause-speci…c hazard are estimated by the single-risk methods exposed above, treating durations …nishing in other states as rightcensored at the last year of available information (Jenkins, 1995; 2005).8 4 Empirical results on entrepreneurial entry 4.1 Characterizing the di¤erent groups of BOs Table 1 brie‡y characterizes the di¤erent types of BOs identi…ed in the data, as well as the control group composed by Never BOs. The variables listed in the table correspond to the vector of variables included in the estimation of the multinomial logit model for BOs’entry.9 8Narendranathan and Stewart (1993) show that, if distinct destination states depend upon disjoint subsets of parameters - which are functionally independent (so far as the inference about j(t)and is concerned) - the parameters of a cause-speci…c hazard can be estimated by treating durations …nishing into other states as censored at the time of exit. However, if the unobserved characteristics are common to or correlated across the states, this simpli…cation may have an e¤ect on the overall hazard rate. Even so, this is a minor issue in our analysis, as we are mainly interested in cause-speci…c hazards, rather than on the overall rate. Nevertheless, as a robustness check, we have alternatively estimated the competingrisks model proposed by Fine and Gray (1999), which did not produce qualitatively di¤erent results for the main variables of interest, comparatively to those obtained with the estimation of cause-speci…c hazard functions with unobserved heterogeneity. However, given the lack of available programs allowing the introduction of unobserved heterogeneity –which is shown to have signi…cant e¤ects on the duration dependence of dissolutions and OTs –in Fine and Gray’s model, we decided to focus on the results from cause-speci…c hazard functions. All these additional results are available upon request from the authors. 9Additionally, estimation also includes the variable Lagged Unemployment Rate (with one-year lag), to take into account potential e¤ects of the business cycle. Year dummies are also included in all estimations. 68 Table 1. Descriptive statistics, by groups of nascent BOs (Portugal, 1992-2007) Never Start-up Acquis. IntraBOs Entrepr. Entrepr. preneur (Number of cases) (5,484,866) (59,688) (27,155) (70,744) Prior experiences in the labor market as paid employee Experience in a large …rm (%) 0.343 0.229 0.279 0.118 Experience in a foreign …rm (%) 0.121 0.118 0.127 0.061 Number of di¤erent employers 1.749 2.946 2.863 1.738 Recent displacement (%) 0.024 0.157 0.081 n.a. Previous wage job characteristics Overeducation (%) 0.304 0.386 0.359 0.407 Tenure (months) 93.93 59.31 71.09 62.39 Management position (%) 0.018 0.053 0.064 0.257 Hourly wage (in logs, 2005 euros) 1.375 1.360 1.379 1.321 Foreign …rm (%) 0.092 0.058 0.068 0.014 Micro …rm (%) 0.197 0.407 0.309 0.671 Small …rm (%) 0.274 0.327 0.318 0.248 Medium …rm (%) 0.244 0.156 0.203 0.064 Large …rm (%) 0.285 0.110 0.170 0.017 Urban location (%) 0.528 0.483 0.534 0.448 Primary sector (%) 0.024 0.017 0.022 0.029 Manufacturing (%) 0.338 0.254 0.284 0.235 Energy & Construction sectors (%) 0.118 0.129 0.110 0.122 Services sector (%) 0.520 0.600 0.584 0.614 Individual-level characteristics Male (%) 0.572 0.693 0.649 0.642 Age (years) 36.47 33.01 34.60 37.88 Less than 9 years of schooling (%) 0.628 0.492 0.515 0.537 9 years of schooling (%) 0.135 0.183 0.169 0.161 12 years of schooling (%) 0.162 0.208 0.183 0.173 College education (%) 0.075 0.117 0.133 0.129 Notes: n.a.: Not Applicable. 69 Regarding the key variables of interest, Never BOs had more frequently a past employment experience in a large-sized …rm. In opposition, past experiences in large or foreign-owned …rms were much less common among those becoming intrapreneurs. Both start-up and acquisition entrepreneurs seem to have more diverse past experiences while paid employees, by having worked in a larger number of di¤erent …rms. Recent job losses were also more frequently su¤ered by those who became entrepreneurs, especially among those entering via start-up. Nascent BOs overall come from micro and small …rms with lower participation of foreign capital. This is particularly evident among intrapreneurs. Education-job mismatches, captured by overeducation in the previous job, were also more common among those becoming BOs.10 Notable di¤erences are also found regarding previous management positions, which were more usually occupied by workers becoming intrapreneurs. Data also show a larger proportion of males, as well as a larger share of individuals with higher educational attainment, among those who became BOs. Intrapreneurs are, on average, the oldest group of individuals, while start-up entrepreneurs are the youngest ones. 4.2 Empirical results 4.2.1 Multinomial logit estimation results Table 2 reports the results for the …nal speci…cation of the multinomial logit model, including all variables in Table 1 (except Recent Displacement)11, as 10Using one of the three most conventional ways of measuring overeducation (see, for instance, Kiker et al., 1997), an individual was considered to be overeducated if s/he had an educational attainment higher than the mode of the educational attainment of recently hired workers in the same occupation (according to the 3-digit International Standard Classi…cation of Occupations) in the same year. These comparisons were performed after converting the years of schooling of each individual and the modal years of schooling in the respective occupation in categories of educational attainment (namely into 4, 6, 9, 12 years of schooling, plus college education). 11Given that, by de…nition, intrapreneurs never su¤er a job loss immediately before their transition, we cannot include this variable in this speci…cation. Otherwise, the model would 70 well as Lagged Unemployment Rate. Table A.II in the Appendix reports the estimation results obtained for the sub-sample of individuals aged up to 30 years old, as a robustness check. Never BOs are used as the reference group for the transitions occurring in each year. After the estimation of this speci…cation, we tested whether some of the di¤erent types of BOs under consideration could be pooled together into a common category. A Wald test –under the null hypothesis of equalizing the estimated coe¢ cients associated with any given pair of outcomes or choices – strongly rejects the pooling of any of these categories of BOs. Therefore, these groups of BOs must be analyzed separately.12 Regarding the role of past experiences in the labor market, the results suggest that a past job in a large or in a foreign company reduces the individuals’ propensity to leave paid employment and become entrepreneurs, regardless their mode of entry (start-up or acquisition). As expected, by potentially improving future labor market prospects, such experiences may increase by more the individual’s expected utility of remaining in paid employment than that obtained as BOs. In contrast, an experience in a foreign-owned …rm in the past increases the log-odds of choosing intrapreneurship rather than paid employment by about 0.08, suggesting that workers who have accumulated knowledge from foreign companies may have a better career progress inside subsequent …rms (e.g., Balsvik, 2011). These results remain consistent for the sub-sample of younger individuals. The diversity of experiences in the labor market also seems to matter, as a su¤er from identi…cation problems. We study the e¤ect of recent displacement experiences in section 4.2.2, after excluding intrapreneurs from the estimation of the extended model. 12Additionally, we have also tested the validity of the Independence of Irrelevant Alternatives (IIA), one strong assumption of multinomial logit models. This assumption is less of a problem when the alternatives are reasonably distinct (Amemiya, 1981). The fact that, according to the Wald test, we are unable to combine any pair of choices emphasizes the dissimilar structure of the alternatives under study. Even so, we have performed a series of Hausman tests, by sequentially omitting each of the categories of BOs from the choice set, re-estimating the model, and then comparing the results from the full model and the several restricted models. We do not obtain systematic evidence to reject the IIA assumption. 71 larger number of job shifts in the past is found to increase individuals’propensity to become BOs, especially among the youngest individuals. On the one hand, the mobility of workers –especially at younger ages –across di¤erent …rms may work as a mechanism for knowledge transfers, accumulation of speci…c skills, resources and networks, which may either help them to progress within a subsequent …rm (by becoming intrapreneurs) or give them a more balanced and diversi…ed skill mix that induce them into business-ownership, in line with the Lazear’s (2004) Jack-of-all-trades theory of entrepreneurship. On the other hand, as previously discussed, a larger number of di¤erent jobs may also indicate some instability in the labor market, which may motivate the transition into entrepreneurship as a solution for the lack of (stable) alternatives in paid employment. For prospective intrapreneurs in particular, the e¤ect of such diversity of jobs apparently reverses with individuals’age – the e¤ect of “number of di¤erent employers” is negative and statistically signi…cant for entries into intrapreneurship in the global sample, which may actually suggest that individuals’mobility across di¤erent employers may be favorable at younger ages, though potentially indicating a more negative (i.e., unstable) employee pro…le at older ages. Regarding the remaining variables, the results show that particular speci- …cities of the previous job also in‡uence BOs’entry decisions. Education-job mismatches related to overeducation, by potentially signaling some underutilization of workers’knowledge and skills, increase the propensity to progress in the …rm hierarchy through intrapreneurship, discouraging the exit towards entrepreneurship, especially among the youngest workers. Workers engaged in management positions in the previous job are also more likely to become BOs than those in other occupations. Exits from paid employment towards entrepreneurship become less likely as job tenure gets longer (the estimated e¤ect is inverted U-shaped for transitions into start-up and acquisition entrepreneurship, with the estimated peak occurring after three months in the job), while the reverse e¤ect is found for transitions into intrapreneurship. 72 Table 2. Multinomial logit estimation results (Portugal, 1992-2007) Start-up Acquisition Intrapreneur Entrepreneur Entrepreneur Prior experiences in the labor market as paid employee Experience in a large …rm -0.6146*** -0.5501*** -0.1062*** (0.0122) (0.0163) (0.0158) Experience in a foreign …rm -0.1793*** -0.1605*** 0.0822*** (0.0151) (0.0206) (0.0203) Number of di¤erent employers 0.8373*** 0.8742*** -0.0263*** (0.0034) (0.0045) (0.0053) Previous wage job characteristics Overeducation 0.0081 -0.0575*** 0.3148*** (0.0101) (0.0149) (0.0101) Tenure 0.0094*** 0.0096*** -0.0019*** (0.0002) (0.0003) (0.0002) Tenure squared/100 -0.0018*** -0.0017*** 0.0002*** (0.0001) (0.0001) (0.0001) Management position 0.6946*** 0.7680*** 3.0495*** (0.0243) (0.0341) (0.0154) Hourly wage -0.0057 -0.1108*** -0.1969*** (0.0130) (0.0195) (0.0125) Foreign …rm -0.0075 -0.0192 -0.9288*** (0.0189) (0.0254) (0.0356) Small …rm -0.6005*** -0.3768*** -1.2394*** (0.0107) (0.0168) (0.0106) Medium …rm -1.1634*** -0.6867*** -2.4505*** (0.0139) (0.0196) (0.0181) Large …rm -1.4307*** -0.8183*** -4.0157*** (0.0167) (0.0221) (0.0350) Urban location -0.1200*** -0.0220 -0.0408*** (0.0097) (0.0143) (0.0096) Primary sector -0.4044*** -0.1267*** -0.3801*** (0.0371) (0.0486) (0.0293) (It continues in the next page...) 73 Table 2. Multinomial logit estimation results (Portugal, 1992-2007) Start-up Acquisition Intrapreneur Entrepreneur Entrepreneur Previous wage job characteristics Energy & Construction sectors 0.0115 -0.0560** -0.1467*** (0.0159) (0.0240) (0.0164) Services sector 0.1158*** 0.1048*** -0.1033*** (0.0117) (0.0166) (0.0118) Individual-level characteristics Male 0.5596*** 0.3028*** 0.2897*** (0.0107) (0.0152) (0.0104) Age -0.0278*** -0.0938*** 0.1285*** (0.0037) (0.0048) (0.0030) Age squared/100 -0.0443*** 0.0726*** -0.1376*** (0.0050) (0.0061) (0.0037) 9 years of schooling 0.6765*** 0.5324*** 0.4172*** (0.0142) (0.0208) (0.0143) 12 years of schooling 0.7083*** 0.6277*** 0.4028*** (0.0139) (0.0210) (0.0149) College education 1.2906*** 1.4269*** 0.5918*** (0.0224) (0.0327) (0.0226) Macroeconomic Environment Lagged unemployment rate -0.0311*** -0.2845*** -0.1940*** (0.0074) (0.0096) (0.0053) Constant -6.8725*** -5.2670*** -6.4640*** (0.0801) (0.1021) (0.0633) N 26,449,546 Log Pseudo-likelihood -838,135.9 Pseudo R20.1421 Notes: *, **, and *** denote signi…cant at 10%, 5% and 1%, respectively. Workerclustered standard errors in parentheses. The model also includes time dummies. Reference categories: Micro Firms for …rm size; Manufacturing for sector; "Less than 9 years of schooling" for individual’s education. 74 Higher wages in the previous job seem to discourage entrepreneurial entry, by increasing the opportunity cost of leaving paid employment. For intrapreneurs, this result may rather con…rm that the transitions into intrapreneurship identi…ed in our data mainly correspond to ownership transfers within family …rms, where wages tend to be lower.13 Our results also con…rm that smaller …rms spawn new entrepreneurs among their employees more often than larger …rms do (see also Hyytinen and Maliranta, 2008; Parker, 2009b; Berglann et al., 2011). Large-sized …rms, instead, by o¤ering better opportunities for the development of internal labor markets (Brown and Medo¤, 1989), reduce the workers’incentive to leave paid employment and become BOs. For intrapreneurs in particular, results show that …rm size and foreign ownership both play a strong negative e¤ect on their transition, con…rming that intrapreneurship (as we de…ne it) is more common within very small domestic …rms –over again, the typical family …rm. Individuals working in large urban centers also seem to be less prone to leave paid employment to become BOs, possibly because these regions are both characterized by relatively …erce market competition and better employment opportunities. Regarding the set of individual characteristics that we control for, results con…rm that men are more prone to become BOs –and especially start-up entrepreneurs –than women (e.g., Uusitalo, 2001; Livanos, 2009; Parker, 2009b). Individual’s age, in turn, exerts di¤erent e¤ects according to the entry route chosen –as workers become older, they are more likely to become intrapreneurs and less likely to become entrepreneurs. Education is also associated with a greater likelihood of transiting into business-ownership, in line with the argument that education enhances individuals’“entrepreneurial talent”(Lucas, 1978; Calvo and Wellisz, 1980), improving as well their ability to identify and evaluate business opportunities. Finally, despite our results overall con…rm the so-called “prosperity-pull” 13Additional estimations using an alternative measure of hourly wages that also includes overtime payments (divided by the sum of normal and overtime hours of work) yielded qualitatively similar results. 75 0.1 .2 .3 .4 .5 .6 .7 Cumulative incidence 0 5 10 15 Years sinc e BOs' entry Single ownership Shared ownership Dissolutions 0.1 .2 .3 .4 .5 .6 .7 Cumulative incidence 0 5 10 15 Years sinc e BOs' entry Single ownership Shared ownership Ownership Transfers Fig. 4. Dissolution and OT Cumulative Incidence Functions, by ownership structure Given the potential survival di¤erences between single BOs and those sharing the ownership of their …rm (Figures 2 and 4), we now split the three possible entry modes taking also into account the ownership structure chosen at entry (single or shared).15 In our data, 58% of the individuals becoming BOs share the business with someone else by the time of their transition into entrepreneurship. The relative importance of shared ownerships seems to be even higher among younger BOs (about 61% of them share the …rm with others at entry). Overall, the proportion of shared ownerships is larger in the subgroup of intrapreneurs (64%) and lower among start-up entrepreneurs (51%). 15We did not take into account this disaggregation of BOs in Section 4, when studying entry patterns, because additional estimations showed that no signi…cant di¤erences exist between the entry determinants of BOs entering alone and those sharing the ownership of their business with others. 82 Table 4. Descriptive statistics, by BOs’exit mode (Portugal, 1992-2007) Survivors Exits by Exits by Dissolution OT Entry mode Start-up entrepr. - single ownership (%) 0.228 0.294 0.119 Start-up entrepr. - shared ownership (%) 0.242 0.218 0.157 Acquisition entrepr. - single ownership (%) 0.054 0.079 0.079 Acquisition entrepr. - shared ownership (%) 0.084 0.068 0.139 Intrapreneur single ownership (%) 0.135 0.167 0.166 Intraprenener shared ownership (%) 0.257 0.174 0.340 Prior experiences in the labor market Experience in a large …rm (%) 0.178 0.209 0.194 Experience in a foreign …rm (%) 0.101 0.100 0.092 Number of di¤erent employers 2.584 2.179 2.267 Recent displacement (%) 0.150 0.120 0.087 Years of experience in the (2-digit) industry 3.240 2.179 2.561 Individual-level characteristics Male (%) 0.671 0.658 0.659 Age (years) 36.32 36.33 37.84 Less than 9 years of schooling (%) 0.454 0.514 0.519 9 years of schooling (%) 0.128 0.177 0.161 12 years of schooling (%) 0.248 0.208 0.180 College education (%) 0.170 0.101 0.140 Firm-level characteristics Firm age (years) 10.93 7.745 11.74 Micro …rm (%) 0.777 0.853 0.719 Small …rm (%) 0.205 0.132 0.229 Medium …rm (%) 0.017 0.014 0.043 Large …rm (%) 0.001 0.001 0.009 Urban location (%) 0.400 0.426 0.435 Primary sector (%) 0.018 0.015 0.020 Energy & Construction sectors 0.126 0.139 0.114 Manufacturing (%) 0.203 0.209 0.234 Services sector (%) 0.653 0.637 0.632 N 43,967 35,016 78,604 83 During the period under study, 72% of BOs have exited their business – 22% have dissolved it and 50% have left the business without closing it down, by transferring it to other BOs. In line with the CIFs estimated above, we …nd a higher proportion of start-up entrants among those dissolving the business. In contrast, we …nd a larger proportion of shared ownerships among those leaving by OT (see Table 4). Survivors present a longer experience (from previous job(s) in paid employment) in the sector where they currently operate and seem to have more frequently su¤ered a recent job loss. Higher educational attainments are also more often among those who survive, and less frequent among those leaving by dissolving the …rm. Lastly, the great majority of BOs’…rms are micro-sized, particularly those owned by BOs who end up dissolving the business. These BOs also own, on average, the youngest …rms. 5.2 Estimation results from cause-speci…c hazard models Table 5 reports the results from the estimation of discrete time duration models for each speci…c exit mode. Table A.III, in the Appendix, reports the results obtained for the sub-sample of young BOs. Besides controlling for BOs’unobserved heterogeneity (or frailty), the estimations are also weighted by the number of BOs in each …rm, at entry. Given that the literature has been suggesting that entrepreneurial teams outperform single entrepreneurs (see, for instance, Lechler (2001) for a brief review), non-weighted estimations could produce biased results by over-representing the businesses owned by two or more BOs. Accordingly, we take this aspect into account, in order to avoid giving a more relative importance to some observations over others. By controlling for BOs’frailty, we are supposing that each individual might belong to one of a number of di¤erent types of BOs (for instance, in terms of BOs’entrepreneurial talent or ability), that each BO’s type is unobserved 84 and that some types of BOs are more “frail” (i.e., more likely to exit) than others. Consequently, neglecting individuals’unobserved heterogeneity might have signi…cant implications in our results, mainly in the duration dependence of the two exit modes under analysis. Typically, the non-frailty model tends to overestimate (underestimate) the degree of negative (positive) duration dependence, besides underestimating the magnitude of the coe¢ cients (Jenkins, 1995, 2005). Figure 5 illustrates the estimated duration dependence of dissolutions and OTs, after controlling for BOs’ entry mode, past experiences in the labor market, and a number of individual and …rm characteristics. We compare the results obtained from frailty and non-frailty models, as well as weighted and non-weighted estimations. Figure A.I in the Appendix shows comparable results for the sub-sample of young BOs. Our results con…rm that BOs’unobserved heterogeneity is signi…cant in our data, playing a relevant role in the duration dependence of both exit modes. Exits by dissolution apparently have negative duration dependence – i.e., the risk of dissolving the business decreases as a BO’s spell in the …rm gets longer. However, as expected, neglecting BOs’frailty overestimates this negative duration dependence. Furthermore, for the whole sample, when estimations are weighted by the number of business partners at entry, the dissolution hazard becomes almost ‡at over time. Ownership transfers, in turn, show a U-shaped duration pattern –i.e., BOs are less likely to exit by OT during their …rst years in business, becoming more prone to transfer the business to other BO(s) about …ve years after entering the …rm. Over again, BOs’unobserved heterogeneity is shown to play a role by largely underestimating the positive duration dependence of exits by OT from the fourth/…fth year onwards. Regarding the main variables of interest, the results con…rm that entry mode signi…cantly shapes BOs’post-entry persistence. On the one hand, after controlling for BOs’observed and unobserved characteristics, start-up entrepreneurs entering alone remain signi…cantly more likely to dissolve the business 85 earlier than all other groups of BOs. On the other hand, they also remain the less likely to exit by OT. So, despite BOs normally become more attached to a business started by them than to an acquired business, they face signi…cantly higher failure risks during …rm’s infancy (Freeman et al., 1983; Brüderl and Schüssler, 1990). Overall, the results not only con…rm that acquiring an existing …rm is less risky than establishing a new start-up, but also that sharing the ownership of the …rm with others contributes to share risks and resources, which probably reduce liquidity constraints and, consequently, dissolution hazards. Concerning the e¤ects arising from previous experiences in the labor market, a prior job in a foreign and/or large …rm is found to accelerate BOs’exit, con…rming that individuals with such employment experiences may become less committed to the …rm, as they have higher opportunity costs of staying in entrepreneurship. Also, a larger number of job shifts in the past signi…cantly hastens BOs’exit, whatever their exit mode. In contrast, individuals becoming BOs after losing their job in paid employment persist longer in the business and show lower exit risks. The e¤ect is even larger for exits by dissolution, so our results do not support that individuals coming from unemployment are less able to run a business or more likely to fail as entrepreneurs (Carrasco, 1999; Hinz and Jungbauer-Gans, 1999; Shane, 2009).16 Industry-speci…c knowledge also seems to improve BOs’survival prospects, reducing both exit risks. 16However, we must underline that our analysis is con…ned to recent job losses. The literature often argues that nascent entrepreneurs coming from unemployment are more likely to fail because their human capital, knowledge and skills tend to depreciate during longer unemployment periods, or because they look at entrepreneurship as a last resort solution for their problems in …nding a job. In contrast, individuals losing their job and immediately reacting by becoming BOs may correspond to high-ability unemployed individuals. For this reason, we should not generalize our results, given that we focus on the e¤ect of recent displacement episodes and our data do not allow an accurate identi…cation of all types of unemployed individuals (namely long-term unemployed individuals). 86 Table 5. Estimation results from the cause-speci…c hazard models (Portugal, 1992-2007) Exit by Exit by Dissolution Own. Transfer Entry mode Start-up entrepren. - shared ownership -0.9158*** 0.1625*** (0.0320) (0.0172) Acquisition entrepren. - single ownership -0.3326*** 0.8513*** (0.0444) (0.0272) Acquisition entrepren. - shared ownership -1.2603*** 1.0998*** (0.0414) (0.0228) Intrapreneur single ownership -0.3350*** 0.7025*** (0.0367) (0.0227) Intrapreneur - shared ownership -1.2264*** 0.9135*** (0.0389) (0.0209) Prior experiences in the labor market Experience in a large …rm 0.2770*** 0.1327*** (0.0228) (0.0114) Experience in a foreign …rm -0.0092 0.1109*** (0.0293) (0.0147) Number of di¤erent employers 0.1149*** 0.1460*** (0.0078) (0.0041) Recent displacement -0.3630*** -0.2007*** (0.0270) (0.0146) Years of experience in the industry -0.0434*** -0.0132*** (0.0025) (0.0012) Macroeconomic environment Lagged unemployment rate 0.0098** -0.2119*** (0.0046) (0.0028) Individual-level characteristics Male -0.1807*** -0.1956*** (0.0174) (0.0087) Age -0.0641*** -0.1434*** (0.0053) (0.0029) Age squared/100 0.0770*** 0.1750*** (0.0063) (0.0035) 9 years of schooling -0.1002*** -0.1587*** (0.0197) (0.0105) (It continues in the next page...) 87 Table 5. Estimation results from the cause-speci…c hazard models (Portugal, 1992-2007) Exit by Exit by Dissolution Own. Transfer Individual-level characteristics 12 years of schooling -0.1364*** -0.1338*** (0.0193) (0.0101) College education -0.4849*** 0.0095 (0.0265) (0.0119) Firm-level characteristics Firm age -0.0396*** 0.0152*** (0.0014) (0.0005) Firm age squared/100 0.0085*** -0.0034*** (0.0005) (0.0002) Small …rm -0.7423*** 0.3540*** (0.0229) (0.0091) Medium …rm -0.8524*** 1.1900*** (0.0431) (0.0192) Large …rm -1.9969*** 1.6927*** (0.1218) (0.0350) Urban location 0.3853*** 0.0024 (0.0177) (0.0080) Primary sector -0.1298*** 0.5682*** (0.0594) (0.0267) Energy & Construction sector 0.6954*** -0.0041 (0.0293) (0.0134) Services sector -0.0355* 0.0968*** (0.0208) (0.0098) N 444,497 444,497 Log Likelihood -191,455.1 -367,763.6 u1.4143 1.2550 Rho 0.7799 0.4892 LR test of rho=0 (2) 436.53*** 2305.42*** Notes: *, **, and *** denote signi…cant at 10%, 5% and 1% respectively. Cloglog model with inverse gaussian frailty, weighted by the number of BOs in the …rm at entry. "Start-up entrepren. - single ownership" are the base category for entry mode. "Less than 9 years of schooling" is the base category for BOs’education. Micro Firms are the base category for …rm size. Manufacturing is the base category for sector. Both speci…cations also include 16 duration dummies to study the duration dependence of each exit mode. 88 .02 .04 .06 .08 .1 0 5 10 15 Years elapsed since BOs' entry No frailty & no weights No frailty, with weights With frailty & no weights With frailty & with weights Duration dependence of dissolution hazards .1 .15 .2 .25 .3 0 5 10 15 Years elapsed since BOs' entry No frailty & no weights No frailty, with weights With frailty & no weights With frailty & with weights Duration dependence of OT hazards Fig. 5. The e¤ects of BOs’frailty and weighted estimation on duration dependence Adverse macroeconomic conditions seem to strongly discourage an exit by OT, in line with the evidence that …rms become acquisition targets more frequently during more favorable economic periods (e.g., Bhattacharjee et al., 2009). Though the overall risk of dissolution seems to slightly increase when economic conditions worsen, younger BOs seem to resist to closing down their …rms during periods of higher unemployment, probably because they will not …nd better alternatives in the labor market (see Table A.III). Regarding the remaining variables, men are found to survive longer than women. Both exit risks seem to decrease with BOs’age, starting to increase after the forties. Higher levels of education are associated with lower exit rates, and especially dissolution rates, in line with previous studies that found that BO’s human capital helps to prevent business closure (Bates, 1990; Headd, 89 2003). The smaller and the younger the …rm, the more likely will be an exit by dissolution and the less likely will be an exit by OT. BOs operating in urban areas are also found to face higher risks of dissolution, possibly due to the pressure of competition. 6 Concluding Remarks This paper studies the entry and exit of over 157,587 BOs, focusing on the e¤ect of past experiences in the labor market, and identifying di¤erent entry routes and exit modes for nascent BOs. Concerning entry, our …ndings suggest that the several types of BOs may be driven by di¤erent motivations. Entrepreneurs seem to enter at younger ages and to be signi…cantly pushed by more unstable trajectories in the labor market, namely by recent job losses and by a larger number of job shifts between di¤erent employers. In contrast, a previous job in a large-sized and/or a foreign-owned company apparently discourages transitions from paid employment to entrepreneurship. Intrapreneurs seem to emerge within very small domestic …rms –the typical family …rm –, especially at middle-ages (around the forties). Adverse macroeconomic conditions are found to discourage the entry of all BOs in general, though start-up entrepreneurs seem to be much more reactive, especially at younger ages, by entering counter-cyclically. Regarding BOs’persistence and exit, our results show that di¤erent exit modes can be predicted by BOs’entry route. New BOs entering, alone, via start-up are more likely to dissolve the …rm, but much less likely to leave by transferring the business to others. Industry-speci…c experience seems to signi…cantly increase the persistence of BOs in the …rm, supporting the importance of learning-by-doing and informational advantages gained through the accumulation of speci…c knowledge. Over again, employment experiences in large-sized or foreign …rms apparently increase the opportunity costs of 90 remaining in business-ownership, accelerating BO’s exit decision. BOs’unobserved heterogeneity is also found to play a signi…cant role in our estimations. After controlling for BOs’frailty, our results suggest that ownership transfers have a U-shaped duration dependence. Neglecting BOs’ unobserved heterogeneity leads to an overestimation of the negative duration dependence of dissolution hazards. Finally, our results do not support the widespread belief that nascent entrepreneurs coming from unemployment are more likely to fail and leave their business earlier. We …nd that those who have been displaced immediately before entering entrepreneurship survive longer, being less likely to leave the business, whatever the exit mode. This may open new lines for future research, as individuals entering entrepreneurship almost immediately after losing their job may be a more reactive and high-ability group that becomes more attached to the BO position, when compared to long-term unemployed individuals becoming BOs, who may, instead, be a low-ability group that looks at entrepreneurship as a last resort solution, being possibly less able to run a business. References [1] Amemiya, T. (1981), “Qualitative response models: a survey”, Journal of Economic Literature, 19(4), 1483-1536. [2] Andersson, P., Wadensjö, E. (2007), “Do the unemployed become successful entrepreneurs?”, International Journal of Manpower, 28(7), 604 – 626. [3] Balsvik, R. (2011), “Is labor mobility a channel for spillovers from multinationals? Evidence from Norwegian manufacturing”, The Review of Economics and Statistics, 93(1), 285-297. [4] Bastié, F., Cieply, S., Cussy, P. (2013), “The entrepreneur’s mode of entry: 91 duration analysis of determinants and stability of self-employment among former unemployed”, ZEW Discussion Papers, No. 00-26. [66] Román, C., Congregado, E., Millán, J. M. (2013), “Start-up incentives: Entrepreneurship policy or active labour market programme?”, Journal of Business Venturing, 28(1), 151-175. [67] Shane, S. (2009), “Why encouraging more people to become entrepreneurs is bad public policy”, Small Business Economics, 33(2), 141-149. [68] Shane, S., Venkataraman, S. (2000), “The promise of entrepreneurship as a …eld of research”, Academy of Management Review, 25(1), 217–226. [69] Singer, J., & Willett, J. (1993), “It’s about time: Using discrete-time survival analysis to study duration and the timing of events”, Journal of Educational Statistics, 18(2), 155-195. [70] Song, J., Almeida, P., Wu, G. (2003), “Learning-by-hiring: When is mobility more likely to facilitate inter…rm knowledge transfer?”, Management Science, 49(4), 351-365. [71] Sørensen, J. B. (2007), “Bureaucracy and entrepreneurship: workplace e¤ects on entrepreneurial entry”, Administrative Science Quarterly, 52(3), 387–412. [72] Sørensen, J. B., Phillips, D. J. (2011), “Competence and commitment: employer size and entrepreneurial endurance”, Industrial and Corporate Change, doi:10.1093/icc/dtr025. [73] Stam, E., Thurik, R., van der Zwan, P. (2010), “Entrepreneurial exit in real and imagined markets”, Industrial and Corporate Change, 19(4), 1109-1139. [74] Storey, D. J. (1991), “The birth of new …rms –does unemployment matter? A review of the evidence”, Small Business Economics, 3(3), 167– 178. 98 [75] Tarola, O. (2012), “Acquisition vsus de novo entry: a theoretical appraisal”, Research in Economics, doi: 10.1016/j.rie.2012.12.001. [76] Tarola, O., Gabszewicz, J. J., Laussel, D. (2011), “To acquire, or to compete? An entry dilemma”, Journal of Industry, Competition and Trade, 11(4), 369-383. [77] Taylor, M. P. (1999), “Survival of the …ttest? An analysis of selfemployment duration in Britain”, The Economic Journal, 109(454), 140-155. [78] Uusitalo, R. (2001), “Homo entreprenaurus?”, Applied Economics, 33(13), 1631-1638. [79] van Praag, M., Versloot, P. H. (2007), “What is the value of entrepreneurship? A review of recent research”, Small Business Economics, 29(4), 351-382. [80] von Grei¤, J. (2009), “Displacement and self-employment entry”, Labour Economics, 16(5), 556–565. [81] Wennberg, K., Wiklund, J., DeTienne, D., Cardon, M. (2010), “Reconceptualizing entrepreneurial exit: divergent exit routes and their drivers”, Journal of Business Venturing, 25(4), 361-375. [82] Yusuf, J. (2012), “A tale of two exits: nascent entrepreneur learning activities and disengagement from start-up”, Small Business Economics, 39(3), 783-799. 99 100 APPENDIX Table A.I. Description of variables included in the empirical models Categories of variables Description of variables Prior experiences in the labor market as paid employee Experience in a large firm Dummy=1 if the individual has ever worked in a large firm (250+ employees) in the past. Experience in a foreign firm Dummy=1 if the individual has ever worked in a foreign firm (foreign capital >=50%) in the past. Number of different employers Number of different firms where the individual has already worked as paid employee until period t. Recent displacement Dummy=1 if the individual has exited a previous job in a firm that either closed or suffered a downsizing. Years of experience in the (2digit) industry* Number of years of experience (as paid employee) in the 2-digit industry where the individual has entered as business-owner. Macroeconomic environment Lagged unemployment rate Annual lagged unemployment rate (one year lag). Entry mode* Start-up entrepreneur single ownership Dummy=1 if the individual becomes an Entrepreneur by establishing a new business alone; 0 otherwise. Start-up entrepreneur shared ownership Dummy=1 if the individual becomes an Entrepreneur by establishing a new business with others; 0 otherwise. Acquisition entrepreneur single ownership Dummy=1 if the individual becomes an Entrepreneur by acquiring an existing business alone; 0 otherwise. Acquisition entrepreneur shared ownership Dummy=1 if the individual becomes an Entrepreneur by acquiring an existing business with others; 0 otherwise. Intrapreneur single ownership (employee buyout) Dummy=1 if the individual becomes the only BO of the employer firm; 0 otherwise. Intrapreneur shared ownership (partnership) Dummy=1 if the individual becomes one of the BOs of the employer firm; 0 otherwise. Individual-level characteristics Male Dummy=1 for males, 0 for females. Age Age of the individual in years, in period t. Age squared/100 Squared value of the age of the individual in period t, divided by 100. Less than 9 years of schoolinga Dummy=1 if the individual has less than 9 years of schooling completed in period t, 0 otherwise. 9 years of schooling Dummy=1 if the individual has 9 years of schooling completed in period t, 0 otherwise. 12 years of schooling Dummy=1 if the individual has 12 years of schooling completed in period t, 0 otherwise. College education Dummy=1 if the individual has a college degree (including masters and/or PhD degrees) in period t, 0 otherwise. Previous wage job characteristics Overeducation Dummy=1 if the individual was overeducated in the previous wage job, 0 otherwise. Tenure Tenure of the worker in the previous wage job, in months. Tenure squared/100 Squared value of the individual's tenure in the previous wage job, divided by 100. Management position Dummy=1 if the individual occupied a management position in the previous wage job, 0 otherwise. 101 Table A.I. Description of variables included in the empirical models (cont.) Categories of variables Description of variables Hourly wage Ratio of the base wage and regular benefits over the total number of normal hours worked in the reference month, in logs (wages in 2005 euros). Foreign firm Dummy=1 if the firm where the individual was previously employed had 50% or more of its capital held by foreign investors, 0 otherwise. Micro firm Dummy=1 if the firm where the individual was previously employed had less than 10 employees, 0 otherwise. Small firm Dummy=1 if the firm where the individual was previously employed had between 10 and 49 employees, 0 otherwise. Medium firm Dummy=1 if the firm where the individual was previously employed had between 50 and 249 employees, 0 otherwise. Large firm Dummy=1 if the firm where the individual was previously employed had 250 or more employees, 0 otherwise. Urban location Dummy=1 if the firm where the individual was previously employed was located in an urban center (i.e. districts of Porto and Lisbon), 0 otherwise. Primary sector Dummy=1 if the firm where the individual was previously employed was operating in the primary sector, 0 otherwise. Manufacturing Dummy=1 if the firm where the individual was previously employed was operating in the manufacturing industry, 0 otherwise. Energy & Construction sector Dummy=1 if the firm where the individual was previously employed was operating in the energy or construction sectors, 0 otherwise. Services sector Dummy=1 if the firm where the individual was previously employed was operating in the services sector, 0 otherwise. Firm-level characteristics* Firm age Age of the firm in years. Firm age squared/100 Squared value of the firm age, divided by 100. Micro firm Dummy=1 if the BO's firm is micro-sized (less than 10 employees); 0 otherwise. Small firm Dummy=1 if the BO's firm is small-sized (10-49 employees); 0 otherwise. Medium firm Dummy=1 if the BO's firm is medium-sized (50-249 employees); 0 otherwise. Large firm Dummy=1 if the BO's firm is large-sized (250 or more employees); 0 otherwise. Urban location Dummy=1 if the BO's firm is located in an urban center (districts of Porto or Lisbon); 0 otherwise. Primary sector Dummy=1 if the BO's firm belongs to the Primary sector; 0 otherwise. Manufacturing Dummy=1 if the BO's firm belongs to the Manufacturing industry; 0 otherwise. Energy & Construction sector Dummy=1 if the BO's firm belongs to the Energy or Construction sectors; 0 otherwise. Services sector Dummy=1 if the BO's firm belongs to the Services sector; 0 otherwise. Notes: * denotes variables that are only included in the estimations of duration models to study BO's duration in the firm. Table A.II. Multinomial logit estimation results for the sub-sample of young individuals (30 years old) (Portugal, 1992-2007) Start-up Acquisition Intrapreneur Entrepreneur Entrepreneur Prior experiences in the labor market Experience in a large …rm -0.7319*** -0.6179*** -0.1734*** (0.0187) (0.0257) (0.0326) Experience in a foreign …rm -0.2086*** -0.1518*** 0.0871*** (0.0233) (0.0327) (0.0422) Number of di¤erent employers 1.1652*** 1.1891*** 0.1210*** (0.0060) (0.0079) (0.0114) Previous wage job characteristics Overeducation -0.0328** -0.0656*** 0.1537*** (0.0148) (0.0223) (0.0171) Tenure 0.0191*** 0.0178*** 0.0020*** (0.0009) (0.0007) (0.0005) Tenure squared/100 -0.0044*** -0.0037*** 0.0001 (0.0007) (0.0004) (0.0001) Management position 0.7611*** 0.9100*** 3.3175*** (0.0460) (0.0688) (0.0291) Hourly wage -0.1093*** -0.2103*** -0.2488*** (0.0222) (0.0349) (0.0264) Foreign …rm -0.0138 -0.0593 -1.1820*** (0.0276) (0.0382) (0.0768) Small …rm -0.5740*** -0.3408*** -1.1534*** (0.0159) (0.0258) (0.0190) Medium …rm -1.1409*** -0.6348*** -2.3283*** (0.0212) (0.0309) (0.0334) Large …rm -1.2982*** -0.7154*** -3.7175*** (0.0239) (0.0325) (0.0675) Urban location -0.1147*** 0.0082 -0.0370** (0.0144) (0.0222) (0.0170) Primary sector -0.2277*** 0.0348 -0.0272 (0.0592) (0.0842) (0.0569) (It continues in the next page...) 102 Table A.II. Multinomial logit estimation results for the sub-sample of young individuals (30 years old) (Portugal, 1992-2007) Start-up Acquisition Intrapreneur Entrepreneur Entrepreneur Previous wage job characteristics Energy & Construction sectors 0.1696*** 0.0497 -0.0070 (0.0235) (0.0380) (0.0285) Services sector 0.1769*** 0.1949*** -0.0912*** (0.0178) (0.0269) (0.0215) Individual-level characteristics Male 0.6660*** 0.4038*** 0.4488*** (0.0158) (0.0236) (0.0183) Age 0.4422*** 0.1034*** 0.3367*** (0.0292) (0.0410) (0.0332) Age squared/100 -1.0841*** -0.4459*** -0.5627*** (0.0585) (0.0832) (0.0669) 9 years of schooling 0.6928*** 0.4994*** 0.5136*** (0.0214) (0.0329) (0.0251) 12 years of schooling 0.9152*** 0.7828*** 0.5667*** (0.0204) (0.0318) (0.0249) College education 1.8873*** 1.9649*** 0.7400*** (0.0331) (0.0508) (0.0421) Macroeconomic environment Lagged unemployment rate 0.0990*** -0.1658*** -0.1254*** (0.0116) (0.0148) (0.0097) Constant -13.9241*** -8.5841*** -9.8856*** (0.3640) (0.5003) (0.4088) Time Dummies YES YES YES N 9,392,808 Log Pseudo-likelihood -316,570.5 Pseudo R20.1429 Notes: *, **, and *** denote signi…cant at 10%, 5% and 1%, respectively. Workerclustered standard errors in parentheses. Micro Firms are used as the base category for …rm size. Manufacturing is used as the base category for sector. An indicator variable for “less than 9 years of schooling”is used as the base category for individual’s education. 103 Table A.III. Estimation results from the cause-speci…c hazard models (sub-sample of young individuals, 30 years old, Portugal, 1992-2007) Exit by Exit by Dissolution Own. Transfer Entry mode Start-up entrepren. - shared ownership -0.6451*** 0.1517*** (0.0453) (0.0276) Acquisition entrepren. - single ownership -0.1885*** 0.7491*** (0.0562) (0.0468) Acquisition entrepren. - shared ownership -1.0011*** 1.0370*** (0.0694) (0.0368) Intrapreneur single ownership -0.2232*** 0.8136*** (0.0514) (0.0425) Intrapreneur -shared ownership -0.8775*** 0.9359*** (0.0610) (0.0352) Prior experiences in the labor market Experience in a large …rm 0.1438*** 0.2052*** (0.0316) (0.0223) Experience in a foreign …rm 0.0412 0.1052*** (0.0404) (0.0287) Number of di¤erent employers 0.1084*** 0.1637*** (0.0123) (0.0085) Recent displacement -0.2720*** -0.2281*** (0.0331) (0.0238) Years of experience in the industry -0.0567*** -0.0296*** (0.0049) (0.0030) Macroeconomic environment Lagged unemployment rate -0.0180*** -0.2113*** (0.0065) (0.0048) Individual-level characteristics Male -0.2522*** -0.2960*** (0.0241) (0.0157) Age -0.0377** -0.2073*** (0.0156) (0.0096) Age squared/100 0.0496* 0.2592*** (0.0262) (0.0153) 9 years of schooling -0.0852*** -0.1609*** (0.0264) (0.0177) (It continues in the next page...) 104 Table A.III. Estimation results from the cause-speci…c hazard models (sub-sample of young individuals, 30 years old, Portugal, 1992-2007) Exit by Exit by Dissolution Own. Transfer Individual-level characteristics 12 years of schooling -0.1277*** -0.1432*** (0.0247) (0.0165) College education -0.4011*** -0.0589*** (0.0383) (0.0219) Firm-level characteristics Firm age -0.0366*** 0.0114*** (0.0023) (0.0009) Firm age squared/100 0.0079*** -0.0024*** (0.0007) (0.0004) Small …rm -0.6269*** 0.3200*** (0.0391) (0.0157) Medium …rm -1.1038*** 1.1780*** (0.0812) (0.0320) Large …rm -1.7400*** 1.7175*** (0.2074) (0.0664) Urban location 0.2686*** -0.0015 (0.0254) (0.0138) Primary sector -0.2509*** 0.4931*** (0.0837) (0.0484) Energy & Construction sectors 0.3134*** -0.0309 (0.0380) (0.0228) Services sector -0.0353 0.0826*** (0.0263) (0.0171) N 127,802 127,802 Log Likelihood -59,128.9 -109,477.3 u1.2441 1.1423 Rho 0.4848 0.4423 LR test of rho=0 (2) 65.47*** 676.75*** Notes: *, **, and *** denotes signi…cant at 10%, 5% and 1% respectively. Reference categories: "Start-up entrepreneurs - single ownership" for entry mode; Micro Firms for …rm size; Manufacturing for sector; "Less than 9 years of schooling" for BOs’education. Both speci…cations also include 16 duration dummies. 105 .04 .06 .08 .1 .12 0 5 10 15 Years elapsed since BOs' entry No frailty & no weights No frailty, with weights With frailty & no weights With frailty & with weights Duration dependence of dissolution hazards .1 .15 .2 .25 .3 0 5 10 15 Years elapsed since BOs' entry No frailty & no weights No frailty, with weights With frailty & no weights With frailty & with weights Duration dependence of OT hazards Fig. A.I. The e¤ects of BOs’frailty and weighted estimation on duration dependence - sub-sample of young individuals (30 years old) 106 Serial entrepreneurship, learning by doing and self-selection Essay 3 May 29, 2014 Abstract It remains a question whether serial entrepreneurs typically perform better than their novice counterparts owing to learning by doing e¤ects or mostly because they are a selected sample of higher-thanaverage ability entrepreneurs. This paper tries to unravel these two e¤ects by exploring a novel empirical strategy based on continuous time duration models with selection. We use a large longitudinal matched employer-employee dataset that allows us to identify about 220,000 individuals who have left their …rst entrepreneurial experience, out of which over 35,000 became serial entrepreneurs. We evaluate whether entrepreneurial experience acquired in the previous business improves serial entrepreneurs’survival, after taking into account self-selection issues. Our results show that serial entrepreneurs are not a random sample of ex-business-owners. Robustness tests based on the estimation of the person-speci…c e¤ect, using information on individuals’past histories in paid employment, con…rm that serial entrepreneurs exhibit, on average, a larger person-speci…c e¤ect than non-serial business-owners. Moreover, ignoring serial entrepreneurs’ self-selection overestimates learning by doing e¤ects. Keywords: Serial Entrepreneurship, Entrepreneurial Experience, Learning, Selection JEL Codes: D83, J24, L26 We are grateful to Francisco Lima, Helena Szrek and José Varejão, as well as to participants at the XXVIII Jornadas de Economía Industrial (held in Segovia, in September 2013), in particular to Vicente Salas Fumás, for their valuable comments and suggestions on previous versions of this paper. 107 biases driven by self-selection into serial entrepreneurship. We characterize entrepreneurial experience Eusing three variables: i) the cumulative years the individual has survived as business-owner in the …rst entrepreneurial experience; ii) previous experience as a start-up founder; and iii) industry-speci…c experience. Regarding the performance measure analyzed in this study, we focus on serial entrepreneurs’survival. According to prior literature, we would anticipate positive (negative) and signi…cant e¤ects from each of these variables on serial entrepreneurs’survival (exit). First, by running a business, entrepreneurs acquire unique speci…c resources, knowledge, skills and contacts that can be used to start and/or acquire subsequent businesses. For this reason, the longer an individual has been in a business in the past –successful or not –the more s/he is likely to have learned about being an entrepreneur, and the larger the stock of knowledge that may be accumulated about customers and suppliers, the wider will be the networks of contacts, as well as market-speci…c information (Cope and Watts, 2000; Ucbasaran et al., 2006; Frankish et al., 2012; Parker, 2012), which may constitute important resources when they decide to try again as entrepreneurs. Second, the particular experience of founding a …rm –given that not all individuals become entrepreneurs by creating a start-up venture (see, Parker and van Praag, 2011; Bastié et al., 2013; Rocha et al., 2013) – may also deliver greater opportunities to learn about the overall entrepreneurial process. Starting a …rm from scratch requires a wide range of skills, and prior …rmfounding experience is believed to help an entrepreneur to acquire and enhance such skills (Zhang, 2011). In addition, learning experiences are expected to be mostly relevant during business’infancy, i.e., during the …rst few years of the …rm (van Gelderen et al., 2005). Finally, learning by doing may also arise from industry-speci…c experience (Frankish et al., 2012; Chen, 2013). Individuals becoming serial entrepreneurs by establishing a business in the same industry where they operated in the past may bene…t from informational advantages and su¤er from lower uncertainty, which may also contribute to their resilience in their second entrepreneurial 114 attempt. However, performance (in this case, hazard rates) is only observed for exbusiness-owners reentering and becoming serial entrepreneurs. This would not be a problem if serial entrepreneurs are a random selection of all ex-businessowners (i.e., if Serial =Nonserial). Nevertheless, there may be unobserved factors –related with ability di¤erences among serial and non-serial entrepreneurs –that simultaneously a¤ect the reentry decision (i.e., the selection in the sample) and the post-reentry performance (precisely, the hazard rates). In such case, the previous estimated e¤ects of entrepreneurial experience may be unreliable – as explained above, the eventual negative association between serial entrepreneurs’hazard rates and experience may not be the result of true learning by doing, but the result of a higher-than-average (unobserved) entrepreneurial ability of serial entrepreneurs. For instance, it is possible that some individuals have survived longer in the …rst business, or started a new venture from scratch instead of acquiring an existing …rm, mainly because they are high-quality entrepreneurs. Similarly, those who have chosen to reenter and remain in the same industry may have made this choice because they have perceived to be more able to run a business in that particular industry than in any other one. Accordingly, the main goal of this paper is to evaluate whether learning by doing hypothesis is veri…ed in our data (i.e., whether hE<0, with h(E; ; X) generally denoting serial entrepreneurs’hazard rates), taking into account selfselection issues. 3 Data and Methodological Issues 3.1 Data This study uses data from Quadros de Pessoal (QP), a longitudinal matched employer-employee administrative dataset from the Portuguese Ministry of 115 Employment. QP is an annual mandatory employment survey that all …rms in the private sector employing at least one wage earner are legally obliged to …ll in.3Requested data cover …rms/establishments (e.g., location, employment, industry, sales, ownership, among others) and each of its workers (for instance, professional situation, gender, age, education, occupational category and skill levels). Firms/establishments and individuals (both workers and businessowners) are identi…ed by a unique identi…cation number, so they can be tracked and matched over time, thus providing very rich information on individual’s backgrounds, career paths and transitions across …rms and industries. All these characteristics of the dataset make QP a suitable database for a dynamic analysis of serial entrepreneurship. Raw QP …les are available for the period 1986-2009, though there is a gap for the particular years of 1990 and 2001, for which there is no available information at the individual-level. We restrict our study to serial entrepreneurs reentering into entrepreneurship between 1993 and 2007, excluding reentries occurring in 2001 or 2002.4Data for the period 1986-1992 was only used to characterize individuals’previous experiences as entrepreneurs or paid employees. The entrepreneur de…nition used in this study corresponds to businessowners (BOs) of …rms with at least one wage earner (i.e., employers). We identify serial entrepreneurs in particular as those ex-BOs who become BOs again, in a di¤erent …rm, after leaving their …rst entrepreneurial experience. 3For this reason, self-employed individuals without employees are not covered by QP. 4To avoid measurement errors on the time spent until reentry into entrepreneurship - one of the variables to be included in our estimations - we had to exclude reentries occurring in 2002, as we are not able to ensure whether the reentry occurs in 2001 (for which no data are available at the worker-level) or in 2002. Besides, we exclude reentries occurring after 2007 because, given the criteria adopted to identify business-owner’s exit, we need at least two years of available information after his/her reentry to clearly identify individual’s exit. 116 3.2 Identifying the entry and exit of serial entrepreneurs We started by identifying in QP …les all BOs who left their …rst business ownership experience.5A total of 219,462 ex-BOs, aged between 16 and 50 years old, were identi…ed and tracked over time, in order to …nd out who has reentered and became BO in a second …rm, and who did not.6About 16% of them (precisely 35,202 ex-BOs) have tried a second chance by becoming serial entrepreneurs during the period 1993-2007. These serial entrepreneurs must be understood as small businesses’owners, given that the great majority of them own micro or small …rms. Over 90% of them either run a limited liability company (Sociedade por Quotas) or a single-ownership business (Empresário em Nome Individual), and most of them (about 65%) are established in Services. For each of those 219,462 ex-BOs, we have retained a set of information related to the previous business-ownership experience, namely the time (in years) the individual has survived as BO in the …rst business, the entry mode in 5A BO was considered to have left the previous business if s/he has de…nitely exited the BO status in the previous …rm. To consider that a de…nite exit has taken place, we have imposed an absence of the BO from the …rm larger or equal to two consecutive years. Accordingly, the identi…cation of ex-BOs’exits had to stop in 2007, as data for 2008 and 2009 were only used to check the presence/absence of each BO in the respective business. Even so, as this study covers reentries occurring between 1993 and 2007, we must restrict the analysis to ex-BOs who have left their prior business until 2006. Previous portfolio BOs were also excluded as we are mainly interested in particular characteristics of the …rst business ownership experience (which must be unique) when taking into account the nonrandomness of serial BOs. Portfolio BOs account for less than 1% of all ex-BOs in the dataset, so their exclusion has no signi…cant impact on our results. 6By imposing the upper limit of 50 years old by the time of the exit from the …rst business we are minimizing the reentries of serial entrepreneurs after attaining the retirement age. Additionally, in order to avoid any bias caused by eventual entrepreneurial experiences occurring before 1986 (thus, not identi…ed in the data), we follow Amaral et al. (2011) and conduct some robustness checks for the sub-sample of ex-BOs who were 30 or younger when leaving their initial business. Even so, they also found that the probability of reentering into entrepreneurship is much higher during the years immediately following the exit from the previous business, also using QP data. So, given that we only consider reentries occurring from 1993 onwards, we believe that those potential left-censoring issues do not pose signi…cant problems in our analysis, as those who were BOs prior to 1986 are estimated to be very unlikely to become serial entrepreneurs so many years later. 117 the previous business (start-up versus acquisition) and the respective industry. Additional information regarding the size of the …rm at the moment of exit, the location of the …rm, the ownership structure of the previous business and the exit mode7adopted by the BO was also gathered, as those variables may play a role when explaining the decision of reentering into entrepreneurship (i.e., the selection process of serial entrepreneurs). We have then followed each of those 35,202 serial entrepreneurs over time, since the moment of their reentry until their last record in QP …les, which may either correspond to the moment of their exit from the second business, or to the last year of available information in the dataset – right-censored cases (Lancaster, 1990; Jenkins, 2005). Following the same procedures adopted to identify the exit of the BOs from the …rst business, we have required an absence of each BO from the …rm, or from the BO position, larger or equal to two consecutive years in order to identify serial BOs’exit year. 3.3 Econometric Model: Speci…cation and Estimation As the primary variable of interest is the time spent by serial BOs in their second business, hazard models were considered. A spell starts when an ex-BO becomes a serial entrepreneur. The duration of that spell corresponds to the 7Regarding the exit mode followed by BOs in their …rst experience, QP dataset allows us to distinguish those who have left by closing down the …rm from those who have exited by transferring the business to others. However, despite some studies associate an exit by dissolution to failure and an exit by ownership transfer to a more successful entrepreneurial experience (e.g., Stam et al., 2008; Amaral et al., 2011; Nielsen and Sarasvathy, 2011), we cannot ensure that this was actually the case, as QP does not provide …nancial data at the …rm-level. Actually, as Amaral et al. (2011: 7) also recognize, an unsuccessful entrepreneurial experience may be understood as the failure to attain or exceed a performance threshold required by the entrepreneur to keep the business running (Gimeno et al., 1997; McCann and Folta, 2012), which does not necessarily indicate that the business is economically unviable. Consequently, some businesses may be transferred to other entrepreneurs with a lower performance threshold. Accordingly, we do not associate more (un)successful experiences to any of these particular exit modes and we use the information on BO’s exit mode from the …rst business just as a control variable when studying individuals’reentry decisions. 118 time elapsed until the exit of the individual from this second entrepreneurial experience. Single-spell duration data were hence obtained by ‡ow sampling. The …nal dataset was constructed in a continuous survival time format in order to estimate the continuous time duration models, controlling for individuals’ selection bias, proposed by Boehmke et al. (2006). We started by testing the suitability of semi-parametric and several parametric survival models (see, for instance, Lee and Wang, 2003; Jenkins, 2005; Cleves et al., 2010) accounting as well for individual-level unobserved heterogeneity, which may produce biased results when ignored (Hougaard, 1995; Jenkins, 2005). All the estimated models were evaluated and compared in terms of their Log-Likelihood, Akaike Information Criteria (AIC) and CoxSnell residuals.8According to these initial tests, Weibull proportional hazard model was found to provide a very satisfactory …t to the data. In order to have a …rst idea on potential learning by doing e¤ects, we started by estimating our “Naïve Weibull model”(as in Boehmke et al., 2006), without taking into account self-selection issues. Formally, for each serial BO i, the probability of exit at time tj,j= 1;2; : : :, given survival until then, can be de…ned as h(tijji; Ei; Xi) = ipexp(E0 i1+X0 i2)tp1"ij;(1) where icorresponds to the time invariant individual-level unobserved heterogeneity term (e.g., individual’s ability); pis a shape parameter determining the duration dependence of the hazard rates (being positive (negative) whenever pis higher (lower) than 1); Eiis the vector of entrepreneurial experience measures; Xiis a vector of individual and …rm time-invariant characteristics, measured at reentry; 1and 2are vectors of unknown parameters to be estimated; and "ij is the error term. The parameters of interest related to potential learning by doing e¤ects are those included in 1. 8Conventional wisdom (e.g., Cleves et al., 2010) suggests that the best-…tting model is the one with the largest Log-Likelihood and the smallest AIC value. 119 However, this naïve analysis of serial entrepreneurs’performance may be biased if there is signi…cant self-selection in the sample of serial BOs. As previously exposed, this problem arises because the outcome of interest – h(tijji; Ei; Xi)–is only observed for those who have become BOs for a second time. So, formally, we have the following two-equation model: h(tijji; Ei; Xi) = (ipexp(E0 i1+X0 i2)tp1"ij  if y i>0 if y i0(2) ci=(1if y i>0 0if y i0(3) where y i=0+Z0 i1+uirepresents a latent variable measuring the difference in the utility (or pro…t) between reentering and not reentering into entrepreneurship, and ciis the corresponding observable realization of reentry decision. Self-selection becomes a problem whenever the error terms of both equations are signi…cantly correlated, which means that there might be factors a¤ecting the survival of serial BOs in their second business that also a¤ected their decision of reentering and starting a second entrepreneurial experience. In order to control for these e¤ects, we use the estimator developed by Boehmke et al. (2006) to estimate a Weibull duration model with selection. Following the logic of existing models for non-random sample selection, this method allows us to model simultaneously both processes –the selection of individuals into serial entrepreneurship and their survival while serial entrepreneurs. The errors of both equations are allowed to be correlated according to a bivariate exponential distribution. Right-censored durations are also accommodated by this method. The vector of variables included in the selection equation covers some exBO’s characteristics (gender and education), a set of speci…cities related to the previous business owned by each individual (location, industry, exit mode 120 and ownership structure), as well as the status of each individual in the labor market before transiting.9 Finally, to overcome potential identi…cation problems, we use as exclusion restriction a variable that, to some extent, proxies individual’s risk aversion and entrepreneurial spirit –the age at which each individual became a business-owner for the …rst time. We may expect that individuals entering entrepreneurship at a younger age are less risk-averse and more entrepreneurial by nature.10 Consequently, we may also expect that those individuals will be more likely to reenter into entrepreneurship and become serial entrepreneurs later on. In contrast, the age at which individuals became entrepreneurs for the …rst time, in itself, is not expected to in‡uence the performance while serial entrepreneurs, after controlling for the accumulated experience as BOs and other speci…cities of previous entrepreneurial experience, as well as individuals’ unobserved characteristics. The lack of more detailed information at the individual-level –for instance regarding individuals’wealth or access to …nancial resources –unfortunately prevents us to use other suitable exclusion restrictions. Even so, the variable under consideration may also be a proxy for individuals’income –at younger ages, individuals’wealth is more likely to be lower.11 Cabral and Mata (2003) 9Regarding the individuals’ status in the labor market before reentering as serial entrepreneurs, QP data allow us to identify whether or not the individual has been in paid employment after leaving the …rst entrepreneurial experience. In QP dataset, whenever an individual is temporarily absent from the annual records, we cannot be sure whether s/he is unemployed, self-employed, out of the labor force or whether s/he transited to the public sector. For this reason, instead of controlling for unemployment spells occurring in between the two entrepreneurial experiences –which cannot be accurately identi…ed –we alternatively control for ex-BOs’transitions into paid employment in the selection equation. 10Occupational choice models drawing upon Knight’s (1921) notion that the individual responds to the risk-adjusted relative earnings opportunities in both paid employment and self-employment corroborate this expected e¤ect of individual’s age (see, for instance, Rees and Shah, 1986). More recent studies have been con…rming that a successful entrepreneur needs to be highly adaptable –i.e., one needs to be able to learn new skills, to adjust to new environments and to handle unexpected situations. In general, younger people are more adaptable, and adaptability decreases with age, so younger individuals are normally more likely to become entrepreneurs than older ones (Liang, 2011). 11This is supported by the well-established “lifecycle hypothesis of saving”, developed 121 precisely found that entrepreneurs’age is a very good proxy for liquidity constraints, as entrepreneurs become wealthier as they grow older. In this regard, and from the point of view of the lifecycle theory of entrepreneurship (see Stangler and Spulber, 2013; Spulber, 2014), the decision of becoming an entrepreneur may be understood as a form of asset accumulation that involves an opportunity cost, especially at younger ages when human capital and …nancial assets may be more limited. Accordingly, if those individuals becoming nascent entrepreneurs at younger ages are also more prone to engage in serial entrepreneurship –albeit their assets are, on average, lower –this will con…rm that this variable may be a very satisfactory proxy for individuals’risk aversion and entrepreneurial spirit. We still perform several robustness checks in order to show that our results are consistent and are not a¤ected by this limitation in using alternative exclusion restrictions. 4 Empirical Results 4.1 Descriptive statistics As a starting point, we provide a simple comparison of the performance in the …rst entrepreneurial experience between individuals who decided to reenter and became serial entrepreneurs, and those who did not. Figure 1 depicts the distribution of BOs’survival time in the …rst business for each of these sub-samples. It clearly shows that non-serial BOs had lower survival rates (more than 60% of them only survived in the …rst business for one year), while serial BOs had a relatively better performance in the previous business, by surviving for longer periods on average. This may be a …rst signal to suspect that, actually, serial BOs are not a random sample of ex-BOs. from the seminal theories by Modigliani and Brumberg of consumer expenditure. 122 Fig.1. Comparative distribution of BOs’survival time (in years) in the …rst business (All ex-BOs, Portugal, 1992-2006) Focusing in particular on serial entrepreneurs, we also compare their performance in the …rst and second businesses. Unconditionally, without controlling for any observed or unobserved characteristics of serial BOs and their …rms, Figure 2 compares the estimated survivor function of serial entrepreneurs in the two experiences, using Kaplan-Meier (KM) estimator (Kalb‡eish and Prentice, 1980). In both cases, the unconditional probability of an individual surviving as BO beyond time twas thus computed as follows: b S(tj) = t Y j=t0 (1 dj nj );(4) where djcorresponds to the number of exits in each time interval and njis the total number of BOs at risk of exit. Table 1 provides a more detailed comparison of serial BOs’unconditional survival rates according to the similarity of the industries where both businesses were developed. 123 Table 3. Estimation results from the Weibull proportional hazard model (Portugal, 1993-2007) (1) (2) Firm-level characteristics Primary Sector -0.1778** -0.1772** (0.0800) (0.0868) Energy & Construction 0.0826** 0.0868** (0.0359) (0.0359) Services 0.0415 0.0452 (0.0295) (0.0294) Macroeconomic environment Reenter in a year of crisis 0.0073 0.0119 (0.0349) (0.0348) Constant -0.2348 -0.1159 (0.2298) (0.2295) Number of observations 35,202 35,202 p (duration dependence) 1.8233*** 1.8212*** Log Likelihood -47,926.1 -47,919.4 Theta 5.4457*** 5.4164*** Notes: *, **, and *** denote signi…cant at 10%, 5% and 1% levels, respectively. HuberWhite robust standard errors in parentheses. Both speci…cations include an individuallevel inverse gaussian distributed unobserved heterogeneity term. Theta corresponds to the variance of this term. "Less than 9 years of schooling" is the base category for individuals’ education. Manufacturing is the base category for sector. As a result, not only may it be more di¢ cult to learn under more unstable experiences, as also few business are identical –possibly favoring the accumulation of business-speci…c rather than general entrepreneurial learning (Chen, 2013) –, so that learning possibilities are modest and di¢ cult to be transferred across di¤erent experiences (Frankish et al., 2012), which may explain the results found for start-up experience. In alternative, the choice of the entry mode may rather re‡ect individuals’attitude towards risk, more than an opportunity to learn. Establishing a start-up …rm, instead of acquiring an existing business, may be a sign of low risk aversion. However, while recent 130 research has been supporting a positive correlation between risk attitudes and entrepreneurial entry decision, the e¤ects on survival are not straightforward, as high risk attitudes may increase exit rates (e.g., Caliendo et al., 2010). Regarding the several individual-level characteristics taken into account in our estimations, results con…rm that men survive longer as serial BOs than women, and that serial BOs’age exerts an U-shaped e¤ect on hazard rates. Higher levels of education seem to be associated with greater exit rates, which may be related to the higher opportunity costs that highly educated individuals probably have by remaining in the business, as they may be more likely to …nd more satisfactory alternatives (in the form of less risk-taking and better remunerated options) in the labor market (Gimeno et al., 1997; Georgellis et al., 2007), particularly after their previous experience as BOs (see Baptista et al., 2012). Being located in an urban center is found to increase exit rates, probably due to the greater competition characterizing large urban regions (Stearns et al., 1995). As expected, sharing the ownership of the second business with other BO(s), by reducing the risk and potentially increasing the sources of capital and knowledge, is found to reduce entrepreneurs’ exit rates. Overall, reentries occurring in times of crises seem not to signi…cantly a¤ect the persistence of serial entrepreneurs. Finally, our results show that serial entrepreneurs’exits present positive duration dependence (the estimated value for pis higher than 1), which means that BO’s exit becomes more likely as time goes by. However, this result is mainly capturing the relatively higher and increasing hazard rates su¤ered during the initial years in business, when the liabilities of newness and smallness play a particularly signi…cant, and thus dominant, role. If, instead, the baseline hazard rate was parameterized according to a non-linear distribution, serial BOs’exit rates would rather show an inverted U-shaped dependence, as expected.13 13Alternative estimations of a loglogistic AFT model showed that the estimated hazard 131 4.3 Self-selection and serial entrepreneurs’persistence We now take into account the possibility that some unobserved factors in‡uence the decision of reentering into entrepreneurship, making serial entrepreneurs a non-random group of ex-BOs. Before analyzing this issue, we brie‡y characterize ex-BOs according to their reentry decision (see Table A.II in the Appendix). The data show that those who became serial entrepreneurs correspond to i) those who survived for longer periods in the …rst business; ii) those with more experience as start-up founders; iii) those with higher levels of education on average; and iv) those who owned larger …rms at the time of exit from the …rst business. All these characteristics shown by serial BOs may also be (positively) correlated with their unobserved characteristics, namely their innate ability or entrepreneurial talent. Accordingly, it becomes crucial to understand whether the unobserved factors that may have in‡uenced the decision of reentering into entrepreneurship were also correlated with the performance shown by serial entrepreneurs after their reentry. Table 4 presents the results obtained from the estimation of speci…cations (1) and (2), now using the two-staged Full Information Maximum Likelihood Weibull duration model with selection developed by Boehmke et al. (2006).14 Over again, the estimations are weighted by the number of BOs in the …rm at entry. The results for the estimated selection equation (reported in Table A.III in the Appendix, …rst column) con…rm that those who established a start-up venture before and those who have survived for a longer period in the …rst business are more likely to try again and become serial entrepreneurs. rates would be increasing during the …rst three to four years of the BO in the …rm, starting to decrease thereafter. The remaining results were not signi…cantly di¤erent from those obtained with the Weibull model. 14The estimations were performed with the program DURSEL (version 2.0) for Stata, for right-censored survival time data, written by F. Boehmke, D. Morey and M. Shannon (available at: http://myweb.uiowa.edu/fboehmke/methods.html) (see also Boehmke et al., 2006). 132 Table 4. Estimation results from the Weibull proportional hazard model with selection (Portugal, 1993-2007) (1) (2) Speci…cities of the …rst entrepreneurial experience Cumulative years as BO -0.0003 -0.0156*** (0.0041) (0.0053) Start-up experience 0.0918*** 0.0912*** (0.0172) (0.0172) Experience in the same industry -0.1475*** -0.2489*** (0.0175) (0.0250) Years elapsed between 1st and 2nd experiences 0.0315*** (0.0034) Cumulative years as BO*Years elapsed 0.0058*** (0.0016) Experience in the same industry*Years elapsed 0.0304*** (0.0054) Individual-level characteristics Male -0.0659*** -0.0674*** (0.0194) (0.0194) Age -0.0640*** -0.0636*** (0.0097) (0.0097) Age squared/100 0.0836*** 0.0836*** (0.0121) (0.0121) 9 years of schooling -0.0249 -0.0235 (0.0235) (0.0234) 12 years of schooling 0.1253*** 0.1259*** (0.0228) (0.0228) College education 0.0915*** 0.0935*** (0.0268) (0.0268) Firm-level characteristics Firm size at reentry 0.0036 0.0029 (0.0116) (0.0116) Urban location 0.0585*** 0.0583*** (0.0173) (0.0173) Shared ownership -0.0892*** -0.0910*** (0.0174) (0.0174) It continues in the next page... 133 Table 4. Estimation results from the Weibull proportional hazard model with selection (Portugal, 1993-2007) (1) (2) Firm-level characteristics Primary sector -0.1333** -0.1321** (0.0627) (0.0627) Energy & Construction 0.0606** 0.0641** (0.0288) (0.0288) Services 0.0243 0.0270 (0.0246) (0.0246) Macroeconomic environment Reenter in a year of crisis 0.0083 0.0126 (0.0267) (0.0267) Constant -1.2500*** -1.1553*** (0.1927) (0.1925) No. of observations 219,462 219,462 Uncensored Obervations 35,202 35,202 p (duration dependence) 1.1745*** 1.1733*** Log Likelihood -179,835.4 -179,833.4 Rho (error correlation) -0.1421*** -0.1422*** Notes: *, **, and *** denote signi…cant at 10%, 5% and 1% levels, respectively. HuberWhite robust standard errors in parentheses. "Less than 9 years of schooling" and Manufacturing are the base categories for individuals’education and sector, respectively. The estimations also con…rm that those who became BOs for the …rst time at younger ages are also more likely to become serial business-owners. In addition, higher levels of education and a larger size of the previous business, among other factors, are also associated with a greater likelihood of reentering into entrepreneurship. Those who (re)entered into paid employment after leaving their …rst entrepreneurial experience, in turn, are found to be signi…cantly less likely to reenter into entrepreneurship, as they may have better opportunities in the labor market and, for this reason, higher opportunity costs of reentering entrepreneurship (Baptista et al., 2012). 134 These second results attest that selection should not be overlooked, as a negative and signi…cant correlation is found between the error terms of the two equations (see the estimated values for rho at the bottom of Table 4). In other words, there are unobserved factors that positively a¤ect reentry into entrepreneurship and simultaneously decrease subsequent hazard rates. This …nding is in line with the theories predicting that those involved in serial entrepreneurship correspond to individuals with higher-than-average innate ability and skills (Holmes and Schmitz, 1990; Plehn-Dujowich, 2010). Additionally, accounting for serial entrepreneurs’self-selection has important implications on the conclusions derived from potential learning by doing e¤ects. First, results now show that the cumulative experience acquired in the …rst business does not exert any signi…cant e¤ect on serial BOs’hazards. The signi…cant negative e¤ects previously found are now shown to be irrelevant (…rst speci…cation) or vanishing in a very short period of time (two years after leaving the previous business, according to the second speci…cation). The e¤ects of industry-speci…c experience are also found to be overestimated when self-selection is ignored – those who tried their luck in the same sector have actually around 14% (1 exp(0:1475) = 0:1371) lower hazard rates than those who moved to a di¤erent sector, instead of 20% (1 exp(0:2290) = 0:2047) lower hazard rates as suggested by the “naïve” Weibull model (speci…cation (1) from Table 3). Even so, this comparative advantage seems to vanish about eight years after leaving the …rst business. Figure 3 compares the marginal e¤ects of both measures of entrepreneurial experience, by years elapsed between the …rst and the second business ownership experiences, obtained from both models –i.e., with and without taking into account serial BOs’self-selection. The overestimation of learning by doing e¤ects when self-selection is ignored is clear-cut in both cases. The presence of signi…cant self-selection also changes the magnitude of almost all coe¢ cients, which were considerably overestimated in the “naïve” Weibull model. The same is applicable to the duration dependence –serial BOs’hazards are found to increase over time, but at a much lower rate when 135 accounting for self-selection (the parameter pis now lower, though still higher than 1). The constant term also decreases considerably when correcting selection bias, con…rming that exit rates of serial entrepreneurs were arti…cially increased in previous naïve Weibull models. In sum, once we account for the decision of reentering into entrepreneurship, the estimated exit rates of serial entrepreneurs decrease, since the negative error correlation biases the baseline hazard rates upwards, when ignored. Overall, our results show that neglecting self-selection of serial entrepreneurs may produce biased conclusions about learning by doing e¤ects that can be transmitted from past entrepreneurial experiences to the current ones. The positive association between prior experience and the performance of serial BOs in subsequent entrepreneurial attempts seems to be mainly due to selection on ability, rather than the result of learning by doing. Some learning by doing is found only through industry-speci…c experience (see also Frankish et al., 2012; Chen, 2013). Otherwise, learning e¤ects seem to be really modest. -.04 -.02 0.02 .04 Marginal effect on the hazard rate 0246810 Years elapsed between 1st and 2nd experience Naïve Weibull Model Weibull Model with Selection Marginal effects of cumulative years as BO -.4 -.3 -.2 -.1 0.1 Marginal effect on the hazard rate 0 2 4 6 8 10 Years elapsed between 1st and 2nd experience Naïve Weibull Model Weibull Model with Selection Marginal effects of industry-specific experience Fig. 3. Marginal e¤ects of entrepreneurial and industry-speci…c experience, by years elapsed between experiences (All serial BOs, Portugal, 1993-2007) 136 5 Robustness Checks 5.1 Estimation results for the sub-samples of start-up serial entrepreneurs and young ex-BOs As a …rst robustness test, we estimate both models (the naïve model and the model with selection) for the sub-sample of entrepreneurs entering via start-up. On the one hand, entrepreneurial activity is more often associated to new venture creation, so individuals who have established a start-up …rm from scratch in both business-ownership experiences may be considered to be the most entrepreneurial ones –at least regarding the risk-taking and the comprehensiveness of entrepreneurial steps they were exposed to. On the other hand, start-up entrepreneurs may be driven by di¤erent motivations and may have a di¤erent post-entry behavior than those entering by acquisition (e.g., Rocha et al., 2013). Finally, our previous results showed that those with a past experience as a start-up founder have actually higher hazard rates in the second venture than those without such experience, suggesting that learning from a past founding experience may be harder than expected. For these reasons, we test the consistency of our results by repeating the analysis after excluding ex-BOs who entered the …rst entrepreneurial experience through acquisition, as well as serial entrepreneurs acquiring an existing business in their second attempt. The …nal sample is composed by 81,587 ex-BOs, out of which 9,479 became serial entrepreneurs by establishing, again, a new start-up venture. Additionally, in order to ensure that our results are not biased by potential left-censoring issues related to eventual entrepreneurial experiences prior to 1986, we also re-estimate both models for the sub-sample of ex-BOs who left their prior business at the age of 30 or younger, as in Amaral et al. (2011).15 Table 5 summarizes the results obtained from the estimation of our pre15We report the results obtained from the sub-sample of young start-up serial BOs. The results obtained for all younger ex-BOs, regardless their entry mode in the …rst and second experience, were not qualitatively di¤erent from those here presented. 137 ferred speci…cation for both sub-samples. The correlation between the error terms remains negative and highly signi…cant in both cases. For start-up serial BOs, both sources of learning by doing (i.e., the cumulative experience as BOs and industry-speci…c experience) are con…rmed to be temporary and overestimated when self-selection is ignored. Figure 4 illustrates these results. For younger BOs, industry-speci…c experience seems to be the only signi…cant source of learning by doing. However, over again, such learning e¤ects are overestimated under the naïve model, as Figure 5 makes clear. Table 5. Estimation results for particular sub-samples Naïve Model and Weibull model with selection (Portugal, 1993-2007) Start-up Young Start-up Serial BOs Serial BOs Naïve Selection Naïve Selection Cumulative years as BO -0.0676*** -0.0433*** -0.0164 0.0155 (0.0140) (0.0109) (0.0646) (0.0538) Exper. same industry -0.3677*** -0.2437*** -0.4693*** -0.2936** (0.0592) (0.0500) (0.1294) (0.1068) Cumul. years as BO*Years elap. 0.0108** 0.0093*** 0.0114 0.0096 (0.0047) (0.0034) (0.0149) (0.0113) Exper. same indus.*Years elap. 0.0412*** 0.0277** 0.0858*** 0.0553*** (0.0144) (0.0111) (0.0281) (0.0211) No. Observations 9,479 81,587 2,011 16,404 Uncensored observations - 9,479 - 2,011 p (duration dependence) 1.8539*** 1.1914*** 1.8241*** 1.1806*** Log Likelihood -13,049.8 -51,941.3 -2,860.4 -10,968.3 Rho (error correlation) - -0.1319*** - -0.1397*** Notes: *, **, and *** denote signi…cant at 10%, 5% and 1% levels, respectively. HuberWhite robust standard errors in parentheses. 138 -.1 -.05 0.05 Marginal effect on the hazard rate 0246810 Years elapsed between 1st and 2nd experience Naïve Weibull Model Weibull Model with Selection Marginal effects of cumulative years as BO -.4 -.3 -.2 -.1 0.1 Marginal effect on the hazard rate 0246810 Years elapsed between 1st and 2nd experience Naïve Weibull Model Weibull Model with Selection Marginal effects of industry-specific experience Fig. 4. Marginal e¤ects of cumulative years as BO and industry-speci…c experience, by years elapsed between …rst and second experience (Start-up Serial Entrepreneurs, Portugal, 1993-2007) -.4 -.2 0.2 .4 Marginal effect on the hazard rate 0 2 4 6 8 10 Years elapsed between 1st and 2nd experience Naïve Weibull Model Weibull Model with Selection Fig. 5. Marginal e¤ects of industry-speci…c experience, by years elapsed between …rst and second experience (Young Start-up Serial Entrepreneurs, Portugal, 1993-2007) 139