Educational and Occupational Biographies of Company's Founders – An Analysis of Personal Longitudinal Data with GSOEP by Anna Heimann, Silke Tegtmeier 1 draft version: April 30, 2011

This paper makes a contribution to entrepreneurship research by analyzing the educational and occupational biographies of company’s founders. The central question that motivates this paper is: Are there any patterns in the educational and occupational biographies of company’s founders? A reproduction of the biographies allows us to compare the individual biographical progressions among each other and to cluster them concerning their similarities respectively dissimilarities. For the present analysis, data of the German Socio-Economic Panel Study (GSOEP) are used. The data originated seven types of founders which we named and broadly described as follow: (1) “Experienced University Graduates”, (2) “Inexperienced / Low Experienced University Graduates”, (3) “Inexperienced / Low Experienced Homemakers”, (4) “Experienced Professionals”, (5) “Low Experienced Non- / Professionals”, (6) “Experienced Late Bloomer” and, at last, (7) “Experienced Seniors”. A typology of founders can help policy makers and counselors to find target-group-specific solutions to support founders. The latter themselves can find better ways to make their companies successful.

Introduction Start-ups are of high relevance in research and practice. In particular, they attract attention in economic policy due to their positive impact on the market economy. They are associated with high competitiveness, a potential for innovation, as well as job creation (Schulte 2002). The quality of a start-up can hardly be predicted as many unforeseeable factors have an impact on

Anna Heimann is an academic assistance at the Department for Entrepreneurship and Startup Management, Institute of Corporate Development, Leuphana University of Lueneburg. Dr. Silke Tegtmeier is post doc (habil.) researcher at the Department of Entrepreneurship and Start-up Management, Institute of Corporate Development, Leuphana University of Lueneburg. Address correspondence to: Anna Heimann, Leuphana Universität of Lueneburg, Department of Entrepreneurship and Start-up Management, Lueneburg, Germany. E-mail: [email protected]. 1

the success of any start-up. However, because the entrepreneur him- or herself contributes decisively to the success, personals factors are often assessed as potential drivers of success. Among others, these drivers can be qualifications taken from the founder‟s vita. Take the decision to provide public funding for a young company that can be oriented to the course of education and employment including the qualifications gained, taken from one‟s vita. Hence, a central question that motivates this paper arises: Are there any patterns in the educational and occupational biographies of company‟s founders, and, if so, which demographical determinants cause these different biographical patterns? This paper makes a contribution to entrepreneurship research by analyzing the educational and occupational biographies of company‟s founders. The several reviewed biographical processes are ranged from the age of fifteen to the last foundation of the entrepreneurs, so this view allows us to see which states are indicated in individual life courses and what kind of human capital the persons cumulate before launching their own company. A reproduction of the biographies delivers data to compare the individual biographical progressions among each other and to cluster them concerning their similarities respectively dissimilarities. For this approach, we use the sequence analysis which has its seeds in biological science in order to compare DNA sequences (Brzinsky-Fay and Kohler 2010; Abbott 1983; Abbott and Forrest 1986). Further, we draw upon a representative longitudinal data set of the German Socio-Economic Panel Study (henceforward cited as GSOEP). The biographical occupations are derived from human capital theory as well as from empirical studies based on human capital. The latter investigate human capital as a driver for the likelihood to become self-employed. By combining several educational and occupational incidents and by classifying these into groups the proceeding of this paper differs from previous analyses, where human capital is involved in an inquiry in form of separate aspects (for example school education, professional training, unemployment). The present paper proceeds as follows. The next section develops the theoretical framework of this study including human capital theory and empirical studies based on human capital. In the 2

third section, methodology and data of the present study are described. Subsequently, the results are presented. The paper closes by a discussion including some suggestions for future research.

Theoretical Framework Human Capital Theory Unfortunately, we could not identify any empirical study in entrepreneurship research that has considered the educational and occupational biographies of entrepreneurs in their entire course so far. In the present study, relevant factors for a demonstration and investigation of entrepreneurs‟ biographies are therefore derived from human capital theory. The state of empirical research gives information about which human capital factors influence the likelihood of an entrepreneurial activity. Human capital theory (henceforward cited as HCT) is mostly applied in labor market research (Becker 1975; Mincer 1974). While traditional applications of HCT almost exclusively emphasize employees, entrepreneurship research focuses on self-employed people, in particular on entrepreneurs. Investigations of HCT in labor market research usually differentiate between general and specific human capital (Becker 1964). General human capital summarizes the entire knowledge, capabilities as well as experience that an individual collects by education and work experience during his or her lifetime (accumulation of general skills). In contrast, specific human capital comprises knowledge and capabilities that an individual gains at a certain job or employer and that become lost when changing the employer because they are company-specific. When HCT is transferred to entrepreneurship research general human capital factors are of particular relevance while specific human capital is not considered due to the fact that it becomes lost when changing from employment to self-employment (Wagner 2006). This is why in lieu thereof Preisendörfer, and Voss (1990) propose to distinguish between general human capital and two more forms: on the one hand, industry-specific human capital, that is, work 3

experience in the industry where the start-up takes place; on the other hand, entrepreneurial human capital. The latter contains capabilities and qualifications that can be useful for performing an entrepreneurial activity, say, management experience as well as former start-up experience (Preisendörfer and Voss 1990). Based on an individual‟s certain configuration of human capital, selection effects can occur in terms of an entrepreneurial activity. Selection processes already happen before a start-up and let us assume that people with high human capital are rather capable of launching a company than others and that these persons also have better prospects of success. This can be argued by a better financial equipment of the founders that can be traced back to the higher income in former postitions. Likewise, persons with higher qualifications have a better access to information via promising market niches. Moreover, these individuals could be well grounded in accurately preparing a venture (Brüderl, Preisendörfer, and Ziegler 1998; K. Wagner 2006; Bulmahn 2002). In addition to the selection effects, Bulmahn (2002) points out the importance of formal qualifications in information economy. According to Spence (1973), these qualifications could also be interpreted as signals and allow an inference from those founders‟ characteristics that are not observable (for example intelligence, perseverance). Thus, founders with higher qualifications, for instance, have better chances to obtain loans from banks who in turn are geared to human capital factors among others in the credit approval process (reputation effects) (Bulmahn 2002; Tamasy 2005). Empirical Studies Empirical research assessed different human capital factors in diverse forms. Among these rank schooling, professional trainings, advanced trainings as well as work experience. Work experience comprises start-up experience, management experience respectively leadership as well as industry experience. Schooling and professional trainings are usually assessed either by counting the number of years or as a dummy that lists all the certificates and degrees. Results reveal that schooling and professional trainings have a positive impact on the likelihood to 4

launch a company respectively on being entrepreneur. (Wagner and Sternberg 2004; Müller 2006; Brixy, Sternberg, and Stüber 2008; Wagner 2005; Pfeiffer and Reize 1999; Kohn and Spengler 2008; Sternberg 2000; Merz, and Paic 2004; Blanchflower 2004; Shane 2008; Davidsson and Honig 2003; Kim, Aldrich, and Keister 2006; Schiller and Crewson 2007; for a review see Seitz and Tegtmeier 2007; Tegtmeier 2008). However, after graduating from professional school additional education (doctoral degree) reduces the likelihood to become an entrepreneur (Blanchflower 2004; Shane 2008). The same has been found for advanced trainings (Pfeiffer and Reize 1999). For start-up experience it can be assumed from HCT that serial entrepreneurs, due to their former entrepreneurial experience, have better prerequisites to launch a business than novice founders. These increase the likelihood of a start-up (Kranzusch and Kay 2007; Müller 2006; Wagner 2005; Wagner and Sternberg 2004; Evans and Leighton 1989; Davidsson, and Honig 2003; Reynolds 1997). Brüderl and colleagues (1998) found that entrepreneurs differ significantly from the general population due to their experience in management and leadership positions (Brüderl, Preisendörfer, and Ziegler 1998; Delmar and Davidsson 2000; Kim, Aldrich, and Keister 2006). There is also empirical evidence for the fact that industry experience is relevant for being entrepreneur or nascent entrepreneur (Brüderl, Preisendörfer, and Ziegler 1998; Kranzusch and Kay 2007; Maisberger 1998; Dimov 2010). Taken start-up experience, management experience, and industry experience as entrepreneurial capital, it can be noted that in addition to general human capital the former is also of high relevance for founders (Brüderl, Preisendörfer, and Ziegler 1998). In contrast to the existing literature that can be characterized as including HC in form of separate aspects, this paper seeks to assimilate the factors of HC mentioned above to establish biographical courses of founders. In the following, we describe the methodology and sample that is used to identify types of entrepreneurs. Afterwards, the results are presented in section four. The last chapter contains a short conclusion and implications for future research.

5

Methodology and Data Methodological Approach A reproduction of the biographies allows us to compare the individual biographical progressions among each other or to a comparison group and to cluster them concerning their similarities respectively dissimilarities. In doing so, we try to put in order complex longitudinal data. A comparison of single courses among each other and the following classification of these is a procedure of discovering structures. Before starting the analysis, the researcher has no orientation about existing coherences in the data set (Erzberger 2001; Backhaus, Erichson, Plinke, and Weiber 2006). To obtain a typology of individual life courses, we use the sequence analysis which has its seeds in biological science in order to compare DNA sequences. In the 1980s, the sequence analysis was implemented in sociology with Andrew Abbott‟s work on musicans‟ careers and ritual dances (Brzinsky-Fay and Kohler 2010; Abbott 1983; Abbott and Forrest 1986). For a sequence analysis, biographies have to be shown in sequences. A sequence means the chronology of two or more consecutive migrations in the course of a process time whereupon a migration is the transition between one status and another (Erzberger 2001). To detect the resemblance among the sequences, the technique of optimal matching (henceforward cited as OMT) is used in this paper. OMT is aimed at identifying the distances between any pair of sequences (Brüderl and Scherer 2006; Abbott 1995; Abbott 1983). One of these two sequences to be compared is taken as the “source course”, the other one as the “target course”. To transform the source course to the target course, transformation costs arise. The transformation costs are composed by substitution and indel costs and should be kept minimal (Erzberger 2001). These costs contain─comparing source and target sequence─the options „substitution“, „deletion“, and „insertion“ and can be weighted differently in accordance with theoretical considerations (Abbott and Hrycak 1990; Aisenbrey 2000; Erzberger 2001; 6

Aisenbrey and Fasang 2010). In the present work, the substitution costs based on mean transition` probabilities between every two neighboring elements of the sequence (Kohler 2007), whereas the replacement of states with high transition rates is less costly (see table 1 below). This approach is suitable for exploratory research because the estimation of costs based on the data itself. However, to define the substitution costs as high as the transition probabilities means to assume that there should be no dependency of the costs on the direction of the movement between pairs of states. So, the transition from state A to state B has the same costs as the transition from state B to state A (Aisenbrey and Fasang 2010). In this analysis, we expect the same substitution costs independent of the movement direction due to the fact that we did hitherto not find any theoretical and empirical studies that analyze the whole biographies of entrepreneurs with the result that particular transitions during the life courses have stronger influence on the probability to create a business. For instance, both the transition from the state of employment to the state of unemployment in a biography and vice versa might have an impact to become an entrepreneur. On the one hand, persons who move from unemployment to employment accumulate practical experiences that they can implement at their self-employment. On the other hand, to become often unemployed is a factor that might push persons to selfemployment, so moving to the unemployment also provides a positive effect for business foundation. The same applies for the other states defined in this work; a transition to a certain state can pull or push a person to entrepreneurship (for example, Klandt and Brüning 2002; Galais 1998; Tegtmeier 2008) and we have no information about the strength of the influence for different individual transitions. Therefore, it is not reasonable to weight the direction of the movement between the same pairs of states with different costs.

Insert Table 1 about here

7

In addition to substitution costs, it is essential to define indel costs. High indel costs in relation to substitution costs indicate that the substitution of states will be preferred to transformation through deletion and insertion; accordingly, higher indel costs will put emphasis on similarities in the sequence of events. In contrast, lower indel costs consider similarities in the timing of events (Kneale, Lupton, Obolenskaya, and Wiggins 2010; Erzberger 2001; Aisenbrey 2000). We fixed the indel costs at the level of 2, which is equal to the highest substitution costs. The overall transformation costs of a sequence pair (source and target sequence) cannot be considered independent of the total length of the sequences and therefore have to be standardized (Aisenbrey 2000). To do this, we follow the approach of Aisenbrey (2000) who analyzes the career trajectories of persons before their work at The Frauenhofer Research Organization. In her research, the sequences are of diverse length due to the fact that the participants start their work at the organization of different ages. She filled the absentee states in these short sequences with the last state until the length of the longest life course. This is necessary to avoid resemblance in sequences that are not superficially based on the trajectories patterns, but also on the sequence length (in association with high indel costs) (Aisenbrey 2000; Stovel and Bolan 2004; Aisenbrey and Fasang 2010). In this paper, we also filled the short sequences with the last position. For instance, a sequence of a person who becomes an entrepreneur at the age of 30 will be continued with the last state of her or his sequence until the age of 68 (age of the longest sequence). Finally, the calculated transformation costs have to be standardized by dividing them through the length of now isochronous sequences. The calculated measures of distance are applied to build groups. By means of a cluster analysis, diverse life courses are combined to groups. In doing so, the life courses within one cluster are as similar as possible while the clusters differ as much as possible among one another (Erzberger 2001). The types of life courses that have been calculated can be used for further analyses. Take a regression model that uses the clusters as dependent or independent variables (Brüderl and Scherer 2006). 8

Sample We conducted the analysis with data of the German Socio-economic Panel (henceforward cited as GSOEP). This panel is a longitudinal data collection that has taken place annually in Germany since 1984. This data collection assesses data of representative households and all the persons living there. Data is collected and provided by the German Institute of Economic Research. By means of yearly questioning, a multiplicity of socio-economic variables is collected aimed at observing and analyzing political and societal changes (Wagner, Göbel, Krause, Pischner, and Sieber 2008; Paic 2009). In addition to the basic variables requested annually nearly without any changes, deeper questions about diverse aspects supplement the questioning via yearly changing modules (Haisken-DeNew and Frick 2005; Wagner, Göbel, Krause, Pischner, and Sieber 2008; Uhly 2000). In total, the GSOEP contains eight subsamples that have been integrated into the panel at different starting times (see table 2 below). For a cross-sectional analysis, the division into subsamples is only of marginal importance because these can be analyzed together. However, for longitudinal analyses, since 1984 not for all subsamples data are available (except for retrospective data, for example, with regard to the life course) (Bergmann 2000).

Insert Table 2 about here

The aim of this work is to analyze educational and occupational trajectories of participants who have been entrepreneur at any time in their biography. Here, the entrepreneur is defined as a person who is running her or his start-up as an independent fulltime venture that represents a source of his or her subsistence. Furthermore, by ascertaining the sample size we have to select persons with statements about their educational or occupational positions for each year of the survey. This is essential in order to avoid gaps within sequences which would distort the trajectory patterns. The GSOEP offers two alternatives to generate the data set needed for the 9

analysis on the Internet: a “balanced” and an “unbalanced” data set. The second alternative captures all persons in all eight subsamples A to H. In contrast, the “balanced” data set, we choose here, only includes those persons that have participated in the questioning in all waves relevant for the analysis. Thus, persons that have dropped out of the questioning at any time (for example leaving the panel) are excluded. However, by selecting the cases, a “balanced” data set runs the risk that the data is not representative anymore (Paic 2009). In consideration of the aspects that participants are not from the same birth cohort in the sample and, thus, the older participants begin to answer the annually question about their education and occupation positions in their life trajectories later than young people we also have to use retrospective data that have been asked in the first survey of the participants and that have been continued with the annual questioned data. Due to the fact that the codification of the annual variable with information about the educational and occupational position is more detailed as the codification of the retrospective variable we have to match the codification with that of the retrospective variable. The merge and the use of the retrospective data consequently means that we have no information about previous self-employment of the participants and, furthermore, no retrospective yearly information about, for example, experience in management and leadership as well as industry experiences which could be considered in the investigation within the HCT. Contrariwise, the use of retrospective data is indispensable to reproduce the life courses of the participants of different ages from the age of fifteen. In our work, the sequences are structured in following states on a yearly time scale: school/university, professional training, employment, self-employment, and unemployment. Since human capital could not only have been accumulated during the life time, but have also been devalued through long unemployment positions or “other activities” (for example military draft, alternative service or parental leave) it is essential to assimilate these states in our investigation. In the next chapter, the empirical results of the sequence analysis are illustrated.

10

Empirical Results In this chapter, we initially examine aggregated indicators of observed sequences to get a description of the whole sample, and, amongst others, to compare these with characteristics of respective clusters in the next step. Aggregated indicators imply the length of sequences, the number of status changes in sequences, and the number of the length of the episodes (see table 4 below) (Brzinsky-Fay and Kohler, 2006). In total, we have found 120 sequences, thereby, the duration of the longest one is 54 years, and the duration of the shortest one is eight years beginning at the age of 15 and ending at the latest start-up of the participants. By adding the first 15 years of life, the length of a sequence shows how old a person is when he or she becomes self-employed, so the average age of the latest start-up in a life course is 42 years (see table 3 below). Other empirical studies approve approximately this age when creating a business; there, the average age varies from 31 to 39 years (for example, Schulte 2002; Schwarz and Grieshuber 2003). The distribution of the sexes in our sample (women 28.3 percent; men 71.7 percent) is also similar to other empirical findings (for example, Klandt 1984; Schwarz and Grieshuber 2003.

Insert Table 3 about here

The participants have in average 5.26 different sequence episodes in their life courses; further, people change the episodes 4.26 times. These indicators can be taken as a measure of the flexibility of transition in a sequence (Brzinsky-Fay 2006). In table 4 below, the length of stay in all states is depicted.

Insert Table 4 about here

11

As can be seen from the table, the participants spend most of their time in full or part time employment; in contrast, a minimum period of time, on average less than half a year is spent at unemployment. The longest period of unemployment continues six years, but only for one person; most of the participants were without an occupation for one year. Hence, in our sample there are no biographies which are dominated by long unemployment periods. Additionally, none of the participants is unemployed directly before becoming self-employed; almost all (98.3 percent) persons are employed before their self-employment. This aspect deviates from other studies which analyze transitions from unemployment to self-employment and show that unemployment influences new firm formation (for example, Ristilä and Tervo 2002). There might be two reasons why none of the persons in our sample is unemployed before becoming self-employed: (1) Participants with long and/or frequent periods of unemployment during the time are more averse to answering the question about their occupational position if they are without a job again. In this way, persons who do not give an answer will not be in our sample due to the fact that we only take into consideration participants who have information about the states in their life course over all waves of the GSOEP. (2) Further, participants with short periods of unemployment cannot be seen in our sample in that we examine the states on a yearly scale; hence, we can only analyze periods of unemployment that continue over one year. For an overview of frequencies of states during the time periods we have aggregated separate states (see figure 1 below). In this context, it is important to note that figure 1 does not take individual life courses into account as a whole due to the aggregation of the states (Brückner and Rohwer 1996; Aisenbrey 2000). It should at least be an overview of the entire sample at the macro level. At the age of 15, most of the participants (71.7 percent) are at school. The second largest proportion of persons (25.0 percent) is already at a professional training at the beginning of the sequence. Additionally, it can be seen from the figure that the frequency of the state “employment” is paramount in relation to other states up to the age of 19 years and retains the

12

highest frequency until the end of the age of 68 years. Finally, we can see that the change to self-employment is relevant nearly for any age in the sample starting at the age of 22 years.

Insert Figure 1 about here

Based on the distances calculated with OMT, the biographical clusters are generated in the next step. In order to identify unique groups of common patterns in the life courses, we use the Ward's hierarchical clustering method. We choose this method due to the fact that many previous studies also use this approach within the sequence analysis and the fact that the Ward`s algorithm achieves, in relation to other clustering methods, best classification results (Backhaus, Erichson, Plinke, and Weiber 2006, quoted in Bergs 1981). To define the number of clusters, we examined the dendrogram and use the Caliński and Harabasz‟ stopping rules and „Duda and Hart‟s‟ stopping rules. These stopping rules are singling out as two of the most efficient rules by Milligan and Cooper (1985) who evaluated and compared 30 procedures for determining the number of clusters. A seven-cluster solution seems to provide the most reasonable grouping of the sequences. Figure 3 below presents the results of the cluster analysis. Before analyzing the separate clusters with their individual life courses, we give a short overview about the aggregate frequencies of states for each cluster as well as for the whole sample above (see Figure 2 below). This representation method allows us to see how the aggregate states allocated over the analyzed life time and, consequently, to see which states dominated at which time in clear format. For the first cluster, we can see that until the age of 28 most of the participants are at school or university; the duration changes then in favor of employment. The same applies for the second cluster; there, the largest share of participants has also been redistributed from the state “school/university” to the state “employment” at the age of 28 years. According to the age of the business foundation, we can see that the earliest one was at

13

the age of 40 and the latest one at the age of 58 (in average at the age of 47) for the first group and at the age of 27 and 38 (in average at the age of 33) for the second group.

Insert Figure 2 about here

While most of the participants of the third group are in school at the age of 15 and 16 and then at the age of 17 and 18 in a professional training, most participants in groups number five and number six are at the beginning of their life courses, at the age of 15, instantly in a professional training. Therefore, professional training begins earlier for the two last named clusters. At the age of 19, most of the persons of all of these three clusters have already started their employment. We can find the youngest entrepreneurs of the whole sample in group five, there the age of launching a business varies from 22 to 31 years (in average 27 years). The fourth cluster starts up a business at the age of 32 to 44 years (in average 38 years), and the sixth cluster starts up a business at the age of 45 to 58 years (in average 51 years). The figure of aggregate frequencies of the last cluster shows that most of the persons there started their employment already at the age of 16 and that the state of employment almost exclusively dominated during their life time. This group represents participants with the highest average age of 64 years at business foundation; whereby the youngest entrepreneurs in this group are 59 years old and the oldest ones are 68 years old. Finally, the third cluster shows participants most of which are positioned at the state of “other activities” at the age of 18 and stay there to a greater or lesser extent at a long time. The age of launching a business varies from 33 to 65 years (on average 52 years). In the next step, we will depict each of the seven clusters with the attention on the individual biographies in them (see figure 3 below). Table 4 above provides general characteristics of the clusters that are the basis to describe individual biographies. The life courses in the first cluster are characterized by participants with the highest length of stay (12.3 years) in the state 14

“school/university”, followed by the second group with a length of 11.7 years. In total, the proportion of those who have a university degree amount to 85.2 percent (group 1 with 40.7 percent; group 2 with 44.4 percent) of all university graduates in the sample (table 3 above). The difference between the average duration in employment from university graduates and between the one of the second group is 12.1 years; that is why we decided to name this persons as inexperienced or low experienced university graduates whose human capital based mainly on university knowledge. However, to say this, we should keep in mind that there is no information for both groups about possible part-time employments during the time at university. By speaking about university graduates we mean graduates of universities as well as the graduates of universities of applied sciences. In cluster “Inexperienced / Low Experienced Homemakers” the proportion of female participants are higher than that of male participants; but relatively to the whole sample the proportion women in this cluster (5.9 percent) is low. This group is characterized by very long positions of “other activities” with an average stay in these of 23.6 years. The average duration of stay in “other activities” is more than twice as high as the duration in employment. However, there are no biographies in this cluster in which the state of “self-employment” immediately follows the state of “other activities”; so the participants changed from the episode of “other activities” initially in employment and then in self-employment. On closer inspection of the employment period directly before firm creation, we can see that by most of the people this period lasts only one year.

Insert Figure 3 about here

The fourth group of “Experienced Professionals” has the highest observed sequence flexibility (6.10 different episodes on average) in spite of the fact that life courses of persons in the first, the third, and in both last groups are on average longer, and, thus theoretically, there will be a 15

higher probability for more changes. This group spends on average time at the employment (16.3 years) approximately as long as the participants in the group of “Experienced University Graduates” (17 years). In total, this cluster has the most unemployment episodes in relation to the sequence length, followed by the fifth group of “Low Experienced Non- / Professionals”. By comparing the fifth with the second group, which both have short sequences and, thus, consists of entrepreneurs who start up a business at young age, we can see that the last one has a higher number of different sequence episodes (3.93 and 4.3 episodes), so there is more fluctuation in this cluster. As we can deduce from the cluster names selected in this work, most of the participants in clusters four and five have a professional training degree. The same applies for “Experienced Late Bloomer”; the participants in this group have a professional training degree as well, whereas, in the first three groups, there are only 10 persons with a professional training. Characteristically for “Experienced Late Bloomer” and “Experienced Seniors” are their long durations in employment episodes; while, the former group stays in employment for 32 years on average the second one has about 10 years more experience in employment.

Discussion This paper has presented an empirical study that utilizes a method of analyzing biographical courses in entrepreneurship research, which is used mainly in social sciences, to explore if there are special typologies of entrepreneurs concerning their educational and occupational courses (human capital), beginning at the age of 15 until the latest business creation. The results of this investigation show that there are seven types of accumulated human capital in biographical courses that represent the different ways before becoming self-employed. In this paper, the biographies are differentiated regarding the presence of the following six states: school/university, professional training, occupation, self-employment, and unemployment as 16

well as the order of these states, the length of stay in these states and the total length of the sequence until the state of self-employment. Therefore, we found seven types which are named and broadly described as follows: (1) “Experienced University Graduates” with long employment periods and (2) “Inexperienced / Low Experienced University Graduates” with short employment periods. (3) In addition, there are the “Inexperienced / Low Experienced Homemakers” who spend most of their life time out of work. Characterizing for biographical sequences of (4) “Experienced Professionals”, (5) “Low Experienced Non- / Professionals” and (6) “Experienced Late Bloomer” is the superficial presence of professional training, but there are differences regarding the length of the employment episodes and the frequencies of unemployment states. (7) At last, the cluster of “Experienced Seniors” represents the oldest business creators with the longest duration of occupations. Neither of the participants in the whole sample transits to the state of self-employment from unemployment in spite of numerous investigations of this phenomenon. The main reason for this is an examination of yearly scaled states. So, we can only see periods of unemployment that continue over one year. Further restriction is the aspect that no one of the participants has children. However, previous analyses indicate that about 50 percent of entrepreneurs in Germany have no children (for example, Merz and Paic 2006). In contrast, the distribution of the sexes in our sample as well as the average age of the participants, compared to other studies, verify the representativeness of the data set. Further, the results of this study imply that counseling and coaching should differentiate between the different types of founders. This helps to better define client target groups and to develop specific solutions for them. However, by choosing other course states, for example previous start-up experience, management experience, respectively leadership as well as industry experience, the clients could be described in a more detailed way, but the GSOEP does not provide the relevant variables for each year of life. We implemented a method for analyzing biographical courses from social sciences, and increased the target of better describing business founders based their human capital factors. Our 17

results have to be assigned to descriptive and explorative research. Hence, this study suggests further avenues for future research. In order to analyze the probability of business success, the obtained types of founders could be configured as independent determinants in an estimate model. Thus, the estimations provide the opportunity to analyze how the probability changes for different types of founders depend on additional relevant variables and differ from the proceeding in previous analyses, where human capital is involved in form of separate aspects in an inquiry and not in form of holistic courses. Unfortunately, relevant variables to investigate the success are completely absent or do not permanently exist in all waves of the data set. However, for an investigation of types of founders and business success (or other aspects) it is not essential to collect the biographical data again; researchers might ask the participants to range themselves in one of the types or they might classify the participants by the characteristic features of the different types such as the presence of certain states and the average duration in them. This pilot study has shown that business founders can be subclassified into seven different clusters by their human capital accumulation. Future research needs to explore if these types or clusters have an impact on the success and how these clusters can be addressed most effectively in start-up counseling and funding.

18

References Alexandra, Uhly (2000). Zur neuen Selbständigkeit. Eine segmentationstheoretische Analyse auf der Datenbasis des sozio-ökonomischen Panels, Frankfurt a. M., Berlin. Andrew, Abbott (1983). “Sequences of Social Events: Concepts and Methods for the Analysis of Order in Social Processes”. Historical Methods, 16, 129-147. Andrew, Abbott (1995). “Sequence Analysis: New Methods for Old Ideas”. Annual Review of Sociology. 21, 93-113. Andrew, Abbott, and Alexandra, Hrycak (1990). Measuring Resemblance in Sequence Data. An Optimal Matching Analysis of Musicians` Careers. American Journal of Sociology, 96, 273-303. Andrew, Abbott and John, Forrest (1986). “Optimal Matching Methods for Historical Sequences”. Journal of Interdisciplinary History, 16, 471-494. Bradley R., Schiller, and Philip E., Crewson (1997). Entrepreneurial Origins: A Longitudinal Inquiry. Economic Inquiry, 35(3), 523-531. Christian, Erzberger (2001). Sequenzmusteranalyse als fallorientierte Analysestrategie, in Work, Strukturen des Lebensverlaufs. Ed. R., Sackmann, M., Wingens, M., Weinheim, 135-162. Christian, Brzinsky-Fay and Ulrich, Kohler (2010). “New Developments in Sequence Analysis”, Sociological Methods & Research, 38(3), 359-364. Christine, Tamasy (2005). “Determinanten des regionalen Gründungsgeschehens“, in Work, Wirtschaftsgeographie. Ed. W. Gaebe, R., Grotz, H., Nuhn, E. W., Schamp, and .L., Schätzl. Münster. David G., Blanchflower (2004). Self-Employment: More may not be better. Swedish Economic Policy Review, 11, 15-73. David S., Evans and Linda S., Leighton (1989). Some Empirical Aspects of Entrepreneurship. The American Economic Review, 79(3), 519-535. Dimo, Dimov (2010). Nascent Entrepreneurs and Venture Emergence: Opportunity Confidence, Human Capital, and Early Planning. Journal of Management Studies, 47(6), 1123-1153. Dylan, Kneale, Ruth, Lupton, Polina, Obolenskaya, and Richard D., Wiggins (2010). “A crosscohort description of young people's housing experience in Britain over 30 years: An application of Sequence Analysis”. DoQSS Working Paper, No. 10-17. Erich, Schwarz and Eva, Grieshuber (2003). Vom Gründungs- zum Jungunternehmen: Eine explorative Studie, Wien, New York. Frederic, Delmar and F., Per, Davidsson, (2000). Where do they come from? Prevalence and characteristics of nascent entrepreneurs. Entrepreneurship & Regional Development, 12, 123. Friedhelm, Pfeiffer and Frank, Reize (1999). Berufliche Weiterbildung und Existenzgründung. Zentrum für Europäische Wirtschaftsforschung (ZEW), Diskussion Paper, No. 99-12. Gary S., Becker (1964). Human Capital. New York, London. Gary S., Becker (1975). Human Capital. New York, London. Gert G., Wagner, Jan, Göbel, Peter, Krause, Rainer, Pischner, and Ingo, Sieber (2008). „Das Sozio-oekonomische Panel (SOEP): Multidisziplinäres Haushaltspanel und Kohortenstudie 19

für Deutschland – Eine Einführung (für neue Datennutzer) mit einem Ausblick (für erfahrene Anwender).“ Wirt Sozialstat Archiv, 2, 301-328. Guido, Bulmahn (2002). Determinanten der Gründungsaktivität und des Gründungserfolgs. Eine empirische Analyse für das Verarbeitende Gewerbe Nordrhein-Westfalens, Hannover. Haisken-DeNew, John P. and Frick, Joachim R. (2005). “DTC – Desktop Compation to the German Socio-Economic Panel (SOEP)” . Accessed on Feb. 3, 2011. Heiko, Bergmann (2000).“ Gründungspotenzial und Gründungsmanagement im Spiegel des SOEP.“ Schriften und Materialien zu Handwerk und Mittelstand, 8. Heinz, Klandt, (1984). Aktivität und Erfolg des Unternehmensgründers. Eine empirische Analyse unter Einbeziehung des mikrosozialen Umfeldes, Bergisch Gladbach. Heinz, Klandt and Erdme, Brünning (2002). Das Internationale Gründungsklima. Neun Länder im Vergleich ihrer Rahmenbedingungen für Existenz- und Unternehmensgründungen, Berlin. Helga, Krüger, Katherine, Bird, Claudia, Born, and Christian, Erzberger (1999). “Berufe im weiblichen Lebenslauf und sozialer Wandel. Institutionelle Steuerung und individuelle Handlungs-strategien.“ Arbeits- und Ergebnisbericht 1997-1999, Sfb 186, Bremen, 103128. Holger, Zumholz (2002). Wege in die Selbständigkeit: die Gründungsaktivität als Resultat eines individuellen Entwicklungsprozesses. Wiesbaden. Jacob A., Mincer (1974). Schooling, Experience, and Earnings. New York. Jari, Ritsilä and Hannu, Tervo (2002). “Effects of Unemployment on New Firm Formation: Micro-Level Panel Data Evidence from Finland.” Small Business Economics, 19(1), 31-40. Joachim, Merz and Peter, Paic (2004). “Existenzgründungen von Freiberuflern und Unternehmern – Eine Mikroanalyse mit dem Sozio-ökonomischen Panel.“ Forschungsinstitut Freie Berufe, Diskussionspapier, No. 46, Lüneburg. Joachim, Merz and Peter, Paic, (2006). “Erfolgsfaktoren freiberuflicher Existenzgründung – Neue mikroökonometrische Ergebnisse mit Daten des Sozio-ökonomischen Panels.“ FFB Diskussion Paper, No. 55, Lueneburg. Joachim, Wagner (2005). “Der Noth gehorchend, nicht dem eigenen Trieb Nascent Necessity and Opportunity Entrepreneurs in Germany. Evidence from the Regional Entrepreneurship Monitor (REM).” Working Paper Series in Economics, No. 10, University of Lueneburg. Joachim, Wagner and Rolf, Sternberg (2004). “Start-up activities, individual characteristics, and the regional milieu: Lessons for entrepreneurship support policies from German micro data.” The Annals of Regional Science, 38, 219-240. Josef, Brüderl and Stefani, Scherer (2006). “Methoden zur Analyse von Sequenzdaten“, in Work, Methode der Sozialforschung. Ed. A., Diekmann, Kölner Zeitschrift für Soziologie und Sozialpsychologie, 44/2004, Wiesbaden, 330-347. Josef, Brüderl, Peter, Preisendörfer, and Rolf Ziegler (1998). Der Erfolg neugegründeter Betriebe. Eine empirische Studie zu den Chancen und Risiken von Unternehmensgründungen, Berlin. Karsten, Kohn and Hannes, Spengler (2008). „Gründungen in Deutschland: weniger aber besser – Chancenmotiv rückt in den Vordergrund. Jährliche Analyse von Struktur und Dynamik 20

des Gründungsgeschehens in Deutschland“, in Work, KfW-Gründungsmonitor 2008. Ed. KfW Bankengruppe, Frankfurt am Main. Kerstin, Wagner (2006). Gründungsausbildung in Netzwerken. Eine komparative Analyse in deutschen Hochschulregionen, Wiesbaden. Klaus, Backhaus, Bernd, Erichson, Wulff Plinke, and Rolf, Weiber (2006). Multivariate Analysemethoden: Eine Anwendungsorientierte Einführung. Berlin. Miriam, Seitz and Silke, Tegtmeier (2007). Mythos Existenzgründer. Persönlichkeitseigenschaften von Gründern im Diskurs, Marburg. Nathalie, Galais (1998). „Motive und Beweggründe für die Selbständigkeit und ihre Bedeutung für den Erfolg“, in Work, Erfolgreiche Unternehmensgründer. Ed. M., Frese, Göttingen, 83-98. Pamela, Müller (2006). “Entrepreneurship in the Region: Breeding Ground for Nascent Entrepreneurs?” Small Business Economics, 27, 41-58. Paul D., Reynolds (1997). “Who Starts New Firms? – Preliminary Explorations of Firms-inGestation.” Small Business Economics, 9, 449-462. Paul, Maisberger (1998). Hinterher ist man immer klüger. Erfahrungen und Erlebnisse rund um die Unternehmensgründung. Wege in die Selbständigkeit – Chancen, Motive, Hindernisse und Erfolgsfaktoren, Bielefeld. Per, Davidsson and Benson, Honig (2003). “The role of social and human capital among nascent entrepreneurs.” Journal of Business Venturing, 18(3), 301-331. Peter, Kranzusch and Rosemarie, Kay (2007). ”2. Chance? Hürden und Hemmnisse bei der Umsetzung von Restarts“, in Work, Jahrbuch zur Mittelstandsforschung 1/2007. Ed. Institut für Mittelstandsforschung Bonn, Wiesbaden, 85-130. Peter, Paic (2009). „Gründungsaktivität und Gründungserfolg von Freiberuflern. Eine empirische Mikroanalyse mit dem Sozio-ökonomischen Panel“, in Work, Schriften des Forschungsinstituts Freie Berufe. Ed. J., Merz, Lueneburg. Peter, Preisendörfer and Thomas, Voss (1990). “Organizational Mortality of Small Firms: The effects of Entrepreneurial Age and Human Capital”, Organizational Studies, 11, 107 - 129. Phillip, Kim, Howard, Aldrich, and Lisa, Keister (2006). “Access (Not) Denied: The Impact of Financial, Human, and Cultural Capital on Entrepreneurial Entry in the United States.” Small Business Economics, 27, 5-22. Reihnhard, Schulte and Alice, Kober (2007). „Habitual Founders – Stand und Perspektiven der empirischen Forschung“, in Work Management kleiner und mittlerer Unternehmen. Stand und Perspektiven der KMU-Forschung. Ed. P., Letmathe, J., Eigler, F., Welter, D., Kathan, T., Heupel, Wiesbaden. Reinhard, Schulte (2002). „Finanzierungs- und wachstumstheoretische Aspekte der Frühentwicklung von Unternehmen und deren empirische Analyse“, habilitation thesis Universität Dortmund. Rolf, Sternberg (2000). Entrepreneurship in Deutschland. Das Gründungsgeschehen im internationalen Vergleich. Länderbericht Deutschland 1999 zum Global Entrepreneurship Monitor, Berlin. S., Bergs (1981). „Optimalität bei Clusteranalysen: Experimente zur Bewertung numerischer Klassifikationsverfahren“, dissertation, University of Muenster. 21

Scott A., Shane (2008). The Illusions of Entrepreneurship: The Costly Myths That Entrepreneurs, Investors, and Policy Makers Live By, New Haven, London 2008. Silke, Aisenbrey (2000). “Optimal Matching Analyse. Anwendungen in den Sozialwissenschaften“, in Work, Studien zur Wissenschafts- und Organisationssoziologie. Ed. J., Allmendinger, Opladen. Silke, Tegtmeier (2008). Die Existenzgründungsabsicht. Eine theoretische und empirische Analyse auf Basis der theory of planned Behavior, Marburg. Udo, Brixy, Rolf, Sternberg, and Heiko, Stüber (2008). „From Potential to Real Entrepreneurship.“ Institut für Arbeitsmarkt- und Berufsforschung, IAB-Diskussion Paper, 32. Ulrich, Koher (2007). „Update of package for Sequence Analysis“ . Accessed on Jan. 11, 2011.

22

Table 1 Substitution Costs

School/ University School/ University Professional Training Employment Selfemployment Other Activities Unemployment

Professional Training

Employment

Selfemployment

Other Activities

Unemployment

0 1.9916667

0

1.9922956

1.9847484

0

1.9996855

2

1.9814465

0

1.9968553

1.9985849

1.988522

2

0

1.9993711

1.9988994

1.9878931

2

1.9996855

0

Table 2 GSOEP Samples Sample A Sample B Sample C Sample D Sample E Sample F Sample G Sample H

Residents in the FRG (starting in 1984 with 4528 households) a Foreigners in the FRG (starting in 1984 with 1393 households) b German Residents in the GDR (starting in June 1990 with 2179 households) Immigrants (starting in 1994/95 with 374 households) Refreshment (starting in 1998 with 1056 households) c Innovation (starting in 2000 with 6052 households) c Oversampling of High Income (starting in 2002 with 1224 households) Refreshment (starting in 2006 with 1506 households) c

a In detail: households with no foreign head of household (no Turkish, Italian, Yugoslavian, Greek, and Spanish nationalities) b In detail: households with foreign head of household (only Turkish, Italian, Yugoslavian, Greek, and Spanish nationalities) c Institutional households (not representative) are not included in recent samples. FRG: Federal Republic of Germany (West Germany, including West Berlin) GDR: German Democratic Republic (East Germany, including East Berlin)

Source: Adapted from Wagner, Göbel, Krause, Pischner, and Sieber (2008, p. 310); HaiskenDeNew, and Frick (2005, p. 19).

23

Table 3 Cluster Characterization Frequency ..Percent.. Age at Business Launching lowest

highest

average

Birth Cohort

Gender

University Degree

Professional Training

lowest highest female male

Group 1

13

10.8

40

58

46.69

1942

1961

2

11

11

4

Group 2

14

11.7

27

38

32.93

1952

1966

3

11

12

4

Group 3

7

5.8

33

65

52.00

1931

1957

5

2

0

2

Group 4

41

34.2

32

44

38.12

1947

1966

12

29

3

34

Group 5

15

12.5

22

31

27.20

1956

1966

3

12

0

12

Group 6

24

20.0

45

58

51.29

1931

1957

7

17

0

21

Group 7

6

5.0

59

68

64.00

1926

1943

2

4

1

4

120

100

22

68

41.82

1926

1966

34

86

27

81

Total

Table 4 Aggregated Indicators Average Length No. of No. of of Different Episodes Sequence States

Whole Sample

Group 1

Group 2

Group 3

27.88

32.69

19.50

38.00

4.12

4.00

3.57

3.71

5.26

5.23

3.93

5.43

Average No. of Change of State Episodes

4.26

4.23

2.93

4.43

No. of Episodes

Length of Episodes

school/university

0.89

4.06

professional training

0.82

2.08

employment

1.73

18.00

other activities

0.44

2.14

unemployment

0.38

0.54

school/university

1.46

12.31

professional training

0.23

0.62

employment

1.54

17.00

other activities

0.69

1.31

unemployment

0.31

0.46

school/university

1.29

11.71

professional training

0.36

0.79

employment

0.93

4.93

other activities

0.21

0.29

unemployment

0.14

0.21

school/university

0.57

1.86

professional training

0.43

1.00

employment

2.00

10.57

other activities

1.43

23.57

unemployment

0.00

0.00 24

Group 4

23.98

14.07

Group 5

36.96

Group 6

50.00

Group 7

4.66

3.80

3.96

3.83

6.10

5.10

4.33

3.33

5.00

4.00

5.83

4.83

school/university professional training

1.07 1.07

2.37 2.66

employment

2.00

16.34

other activities

0.44

1.05

unemployment

0.51

0.71

school/university

0.60

1.47

professional training

0.87

2.67

employment

1.40

7.33

other activities

0.20

0.27

unemployment

0.27

0.47

school/university

0.33

0.75

professional training

1.08

2.75

employment

1.83

31.58

other activities

0.33

0.67

unemployment

0.42

0.54

school/university

0.83

2.17

professional training

0.67

1.33

employment

2.33

43.00

other activities

0.33

1.33

unemployment

0.67

1.17

(continuation of TABLE 4)

Figure 1 Aggregate Frequencies of States Over Time for the Entire Sample 120 100

Case No.

80 60 40 20 0 15

20

25

30

35

40

45

50

55

60

65

Age in years

School/University Other Activities

Professional Training Unemployment

Employment Self-employment

25

Figure 2 Aggregate Frequencies

Cluster 2 “Inexperienced/Low experienced University Graduates”

Cluster 1 “Experienced University Graduates” 14

12

12

Case No.

Case No.

10 8 6 4 2

10 8 6 4 2

0

0 15

20

25

30

35

40

45

50

55

60

65

15

20

25

30

Age in years

35

40

45

50

55

60

65

Age in years

School/University

Professional Training

Employment

School/University

Professional Training

Employment

Other Activities

Unemployment

Self-employment

Other Activities

Unemployment

Self-employment

26

Cluster 3 “Inexperienced/Low experienced Homemakers”

Cluster 4 “Experienced Professionals”

7

40 30

5

Case No.

Case No.

6 4 3 2

20

10

1 0

0 15

20

25

30

35

40

45

50

55

60

65

15

20

25

30

35

Age in years

40

45

50

55

60

65

Age in years

School/University

Professional Training

Employment

School/University

Professional Training

Employment

Other Activities

Unemployment

Self-employment

Other Activities

Unemployment

Self-employment

Cluster 6 “Experienced Late Bloomer”

Cluster 5 “Low Experienced Non-/Professionals” 14 20

10

Case No.

Case No.

12 8 6 4

15 10 5

2

0

0 15

20

25

30

35

40

45

50

55

60

65

15

20

25

30

35

40

45

50

55

60

65

Age in years

Age in years School/University

Professional Training

Employment

School/University

Professional Training

Employment

Other Activities

Unemployment

Self-employment

Other Activities

Unemployment

Self-employment

27

Cluster 7 “Experienced Seniors” 6

Case No.

5 4 3 2 1

0 15

20

25

30

35

40

45

50

55

60

65

Age in years School/University

Professional Training

Employment

Other Activities

Unemployment

Self-employment

28

Figure 3 Types of business founders

Cluster 1

Cluster 2

Cluster 3

Cluster 4

Cluster 5

Cluster 6

Case No.

0 10 20 30 40

0 10 20 30 40 15

Cluster 7

35

55

75

15

35

55

75

School / University

0 10 20 30 40

ProfessionalTraining Employment Self-employment Other Activities Unemploymnet

15

35

55

75

Age 29