Paper presented at the 56th Annual ICSB Conference __________________________________________________________________

Founding Team Gender Structure and the Effects of Spouse Teams on New Venture Performance in Sweden* Marcus Box** Centre for Entrepreneurship, School of Business Studies Södertörn University Sweden

Tommy Larsson Segerlind Centre for Entrepreneurship, School of Business Studies Södertörn University Sweden Abstract Team entrepreneurship, gender, and spouse teams receive increasing attention. In earlier research there have been two ways to classify the literature on venture team formation: as a resource-seeking behavior, or as a manifestation of interpersonal attraction, where trust and social relationships are already established, for example in spouse teams. The two perspectives should be seen as complementary. Yet, past research has mainly focused on the resource-seeking perspective. This might be a too narrow view. We study the impact of gender, founding team gender diversity/homogeneity, and the effects of spouse/non-spouse teams on new venture survival. The analysis is based on a unique material, consisting of non-interrupted series of prospective longitudinal data that covers 1,500 start-ups in Sweden. Our results show that the presence of a spouse founding team has a positive effect on venture performance, but that ventures founded by women as well as mixed gender teams in general have a smaller probability to survive.

__________________________________________________________________ * The authors appear in alphabetical order and have contributed equally to the article. ** Corresponding author. Contact: [email protected]

Introduction Team entrepreneurship, gender structures, and spouse teams receive increasing attention in entrepreneurship and small business research. The team as a level of analysis has gained relatively little attention in past research (Davidsson and Wiklund 2001), and leading scholars have recommended future research to use multi-levels of analysis (Low and Macmillan 1988; Davidsson and Wiklund 2001; Chandler and Lyon 2001). In line with them, we assume that each level of analysis provide us with unique insights. The need for making the team more explicit, and to explore and evaluate earlier theoretical contributions and their applicability to studies of teambased entrepreneurial processes, has been considered urgent (Aldrich and Kim 2007). This study explores the performance effects of teams, team gender composition, and spouse teams on new ventures. It employs a unique, prospective longitudinal database on Swedish firms, covering 1,500 Swedish joint-stock company start-ups. This microlevel data covers companies founded during the first half of the 20th century and consists of non-interrupted series in which all firms are traced over their entire life-cycle, from founding until closure.

Background In the effort to move away from the single-entrepreneur perspective, there have been increased interests in venture teams as unit of analysis (Gartner 1989). The attempts in the 1980s and -90s to capture the venture team as a phenomenon resulted in many interesting empirical results. Survey articles indicated that ventures founded by teams performed better than those started by single founders (Cooper and Gimeno 1992; Lechler 2001; Storey 1998). However, there were still a lack of theoretical perspec1

tives and frameworks that could make us understand and explain these results. Lechler (2001) argued that a problem with earlier studies is that they mainly have focused on venture teams from a resource-seeking perspective. During the first decade of 2000, different scholars have tried to discuss and develop more valid concepts and theoretical models in the area of venture teams. In research on venture team formation, Forbes, Borchert, Zellmer-Bruhn, and Sapienza (2006) found two ways to classify the literature: as a resource-seeking behaviour, or as a manifestation of interpersonal attraction, where trust and relationships are already established. They also argued that the two perspectives are not necessarily contradicting, but rather can complement each other. Aldrich and Kim (2007) discuss the circumstances under which the two models of team formation – as a rational process (resource-seeking approach) or as an interpersonal social relationship model – are most applicable. From a social network perspective, they argue that there are three models of social networks – random networks, small world networks, and truncated scale free (or “fat tail distribution”) networks – that can affect venture team formation. The authors hypothesized, and also found in an earlier study (Ruef, Aldrich, and Carter 2003), that a majority of the teams arise in homogeneous small world contexts. Recent studies indicate that the resource-based view is too narrow, and that we need to find models and frameworks that also include factors of interpersonal social relationships and social similarity, for example in spouse teams (Hellerstedt, 2009; Larsson Segerlind 2009; Liao, Li and, Gartner 2009; Ruef 2010 ). Spouse teams have earlier been neglected in research on venture teams, taking for granted that one part (read: the wife) only symbolically takes part as owner, or member of the board. Thus, it is generally assumed that spouses do not have operative or strategic roles in the 2

venture. The role(s) of founders, spouses, gender, and of social relationships have also been frequently discussed in family business research (Karlsson Stider 1999; Svanström 2003; Sharma 2004). Liao, Li and Gartner (2009) found that functional diversity shows a positive impact on venture formation, with the presence of a spouse pair in a start-up team. Liao et al.’s interpretation where that strong bonding of spousal teams overcomes the potential negative effects, as for example barriers of communication and avoidance of affective and task conflicts from team diversity. From these findings, they ask for future research with a longitudinal design, where the impact of social similarity/diversity and interpersonal relationships can be assessed over longer time periods, and to use different venture performance measurements. Especially, they urge for studies that assess the presence of spouse-pair and the correlation with venture performance. This paper goes forward from the discussion above but with a focus on performance in terms of long-term survival by addressing the research question: How does initial team diversity and team social similarity affect new venture performance? This study contributes to this field by studying the impact of founding teams with specific focus on gender structure and spouse teams on new venture performance. Controlling for commonly used variables in entrepreneurship research – i.e., liability of age and size – it specifically tests the impact of teams, gender, founding team gender diversity/homogeneity, and the effects of spouse/non-spouse teams, on new venture performance. Previous research has employed a wide range of performance indicators (Delmar 1997; Carroll and Hannan 2000; Audretsch 2002), of which the most common ones are organizational growth and organizational survival. In our study, a firm’s ability to survive is used for measuring organizational performance. A firm does not necessarily have to maximize profits, or endow its members with pres3

tige, power or security – but it has to survive in order to be a firm. The goal of survival cannot therefore be overestimated (Starbuck 1965).

Literature Review and Hypothesis Development The first attempts to explain the results that indicated that team-founded ventures performed better than solo entrepreneurs, were based from the assumption that higher degree of diversity of resources and capabilities in the team could explain their success (Kamm and Nurick 1993). The main explanations for the advantages of teams were deduced to the ability to combine diverse people with different characteristics and resources (Sandberg 1992). However in the findings, team diversity showed often no direct significant impact on venture performance (Lechler 2001). Instead the focus has shifted to the impact of social similarity and homophily in venture teams (Aldrich and Kim 2007; Ruef et al. 2003). This interpersonal social relationship perspective used the assumption, but also found, that entrepreneurial groups tend to be composed of members with similar ascriptive characteristics (e.g., gender, ethnicity, etc.) and that this homophily seems to plays a major role in the venture process (Kim and Aldrich 2004). Ruef (2010) found a significant impact of gender homophily on survival (after one year), especially male homophily. On the contrary, Liao et al. (2009) found no effect on gender similarity on the likelihood of new venture formation. Based on the above discussion, we want to test the hypothesis: H1: Social similarity in founding team gender is positively related to the likelihood of new venture survival. Another track in the interpersonal social relationship perspective is the presence of already existing relationships, social commitments and trust in the venture team. In this context, spouse teams but also friendship relations have been of interest in earlier 4

studies (Francis and Sandberg 2000). Friendship relations, family, and spouse pairs seem to be a common basis in venture teams (Ruef et al. 2003). Based on a process analysis of in-depth case studies of venture teams, Larsson Segerlind (2009) made the proposition that the foundation for a venture team is the emergence of social commitment and trust. This trust was a prerequisite to collectively manage the high degree of uncertainty, but also ambiguity (sense-making), in a setting as a new company start-up. Liao et al. (2009) had a strong support for the hypothesis that the presence of spouses in a founding team is positively related to the likelihood of new venture formation. Hellerstedt (2009) found that trust and familiarity were important for the venture team process, and especially that spousal relationships matter. She found support for the hypothesis that the existence of spousal pair(s) had an effect on firms profit during the first year in business. Ruef (2010) argue that the high degree of social affiliation and loyalty in spouse teams appears also to keep the partners committed longer to the venture and therefore may affect the performance (survival) of the new venture. We have earlier presented a discussion that the two perspectives on venture team performance as a resource-seeking or manifestation of interpersonal attraction is not necessarily contradicting each other and the two models can complement each other (Forbes et al. 2006; Larsson Segerlind 2009). As with team diversity, it is argued that social similarity do not per se affect performance. Rather, it is underlying aspects in teams with high degrees of social similarity that give this positive impact, i.e. trust and social commitment. Liao et al. (2009) tested the interaction between spousal team and team functional diversity, and found it statistically significant on the probability for venture formation. They made the interpretation of the results that strong bonding of spousal teams amplifies the potential positive effects of team diversity. They also argue that there can be stage-dependent effects. With the initial presence of social 5

affiliation and trust, the positive effect of diversity comes later, as the organization ages and grows. Thus we hypothesize: H2: The presence of spouses in a founding team is positively related to the likelihood of new venture survival.

Research Setting and Method Our empirical material consists of non-interrupted series of prospective longitudinal micro level data, represented by four birth cohorts – generations – of joint stock companies. The cohorts were founded in Stockholm in 1912, 1930, 1942, and 1950. The distinctive characteristics of this database are that it covers all companies of all sizes and in all types of industries that were founded in Stockholm during these years. The entire population comprises 1,497 firms, and the firms are traced from their founding until their closure or, at most, until 1999. Our material is based on unprinted (unpublished) archival public sources (PRV), and the cohorts have previously been analyzed with respect to firm survival and growth (Gratzer and Box 2002; Box 2005; 2008). When the database was built up, which took place periodically between 1997 and 2004, data on founder characteristics was also recorded with respect to founder team (number of founders), the founders’ gender, and spouse-related information. This particular founder-data in the database has, until now, been previously unexploited. 1

A Prospective Longitudinal Research Design By using an approach with prospective longitudinal data, we maintain that several analytical advantages are obtained. In contrast to cross-sectional research designs, longitudinal methods allow for causal analyses in a more ‘true’ sense, in which each individual research unit in the population can be traced over multiple periods of time. 6

The single cross section is signified by survivorship bias, meaning that not all research units from one and the same generation are included: they have exited the sample before the cross section is carried out. This problem can to some extent be overcome with repeated, multiple cross sections, but the basic problem still remains: firms of all ages (generations) exist in any cross section, and the study of processes – e.g., new venturing activity, growth, survival – is hampered since the same individual research units are typically not included in multiple cross sections (Glenn 1977; Hagenaars 1990; Janson 2000). Scholars have therefore argued that longitudinal research designs are needed in entrepreneurship research (Low and Macmillan 1988; Chandler and Lyon 2001; Davidsson 2008). But survivorship bias is also frequently present in retrospective longitudinal designs, which is the most common research strategy in studies that use longitudinal or panel data. By identifying the research units of interest at a certain point in time, tracing them ‘backwards’ (the normal design in retrospective studies), this procedure still runs the risk of sampling firms of different ages, and therefore firms from different generations. Birth cohorts are in our meaning more representative: they embody all births – all foundings – in a specific year or period. Thus, our data avoids survivorship bias, and can control for the chronology in changes in variables such as not only at what age a firm was terminated, but also at what particular year. Furthermore, our data avoids other biases. Since all companies founded during a particular year are counted recorded, this means that firms of all (start-up) sizes, industries, etc. from that particular generation are included – there is no exclusion of firms of a particular size, sector, or gender, in our material. Thereby, outcomes based on certain unique attributes or characteristics of individual firms in the population can be separated from the general or ‘normal’. Since our study specifically wants to explore the effects of founding team gender 7

structure, and of spouse teams, on the performance of new ventures, we claim that our prospective longitudinal approach can accomplish this in a better way compared to retrospective or cross sectional designs.

Data and Measurements Even if our database has several unique advantages, it is not faultless. The data comprises only joint stock companies (Swedish: aktiebolag), while the greater part of firms that started were – and still are today – organized differently as sole proprietorships (enskild firma) and partnerships (handelsbolag) etc. There is a lack of historical sources – both in quantity and quality – on other legal types of firms, such as sole proprietorships. Thus, in our study we measure only firms that started with limited liability. It is very likely that particularly small-scale entrepreneurs, or women, in a greater extent have organized in other legal types of business, and that this was even more common in the past (Svanström 2003). We acknowledge that the results in our study may only be valid for joint-stock companies. Due to the potential work load, it has also not been possible to include all joint-stock companies in Sweden and this is why we have chosen Stockholm as delimitation. It is difficult to determine whether or not (new) companies in Stockholm deviated from other companies, but they do represent a rather large share of all new company formations. 2 Finally, we are at present not able to measure any change in team composition/succession over each firm’s life cycle – in our study we are presently only able to study the effects of founding teams on business performance.

8

Variables in the Study The database is built on primary sources and archival material – more specifically mandatory company files and documents (PRV) when a company was founded. 3 The empirical material is neither built on already existing data or classifications, nor on existing databases; the cohorts in our study have been traced and recorded ‘manually’. In some instances, key variables have been defined by the researchers. The dependent performance variable employed in this article is the failure, or exit, of a firm. This terminating event could take different forms such as voluntary or involuntary liquidation, bankruptcy, or merger, but our study does not separate the different ways of terminating. By employing the more general measure of firm death it enables the measuring of business performance in a broad sense. Earlier studies have often used this approach when studying firms’ ability to survive (Box 2008). In our analysis, we will concentrate on the firms’ first ten years. Several firms in our data obtained a significantly higher age, but it is very likely that the effects of founder characteristics diminish over time. The archival material on Swedish companies provides information on the nominal stock capital of the firm at founding, its intended line of business (but not from any official classification), as well the persons who founded the company. Several variables that would have been useful, such as number of persons employed, are not recorded in the archives; the only size measure that is possible to use is stock capital. The intended line of business (industry) is based on the founders’ stated line of business when the firm was registered. Here, it has been up to the researcher to assess the industry affiliation – in this case the three categories of ‘services’, ‘trade’, and ‘manufacturing’.

9

Table 1. Descriptive data, age, size and industry.

Frequency

N (%)

Age (years)

P25 P50 (median) P75 Mean

Size (cum %)1)

-19 20-49 50-99 100-499 500-999 1000+

Industry (%)

Services Trade Manufacturing

Entire population

Cohort 1912

Cohort 1930

Cohort 1942

Cohort 1950

1,497

134

452

407

504

6 18 44 25

2 7 18 16

3 12 41 24

10 24 52 27

10 20 45 24

68 81 90 98 99 100

18 45 67 92 93 100

50 70 82 96 98 100

77 88 95 99 100 -

95 98 100 -

35 38 37

28 37 35

31 47 22

26 43 31

46 26 28

1) Stock-capital at start, thousands of kr., deflated values (1930 = 100). CPI from Statistics Sweden.

Table 1 reports basic descriptives on the population of firms. Companies establishing as manufacturing firms on average declined over time (Industry). This can to some extent be linked to the gradually lower costs for incorporating a business; over time, new entrepreneurs in less capital-intense, small-scale sectors – e.g., retail services and trade – could start a business with limited liability. Hence, it gradually became more common to start a new business in the form of a joint-stock company since the real cost for incorporating a business decreased significantly over time: the lowest nominal capital limit for joint-stock companies was 5,000 Swedish Kronor (kr.) from the late 1800s up until the mid-1970s (Box 2008). The oldest cohort from 1912 contained 134 firms while Cohort 1950 had nearly four times as many. The age of the entire population in our study shows that firms on average were terminated at a young age: a quarter of the total population did not survive beyond the age of six (P25). But the large deviation in age between cohorts shows that the cohorts had dissimilar survival patterns over time. The firms in Cohort 1912 – 10

and in Cohort 1930 – experienced a more severe selection process compared to the two youngest ones. This opposes earlier research results, since small firms in our study seem to have had better survival chances than large firms (compare age and size). Firms in older cohorts were generally larger but had commonly shorter life spans. This seems illogical; stylized facts show that small firms have higher risks than large. One explanation could be that stock capital is an unsuitable measure of size, even though earlier studies have shown that different size measures generally are correlated (c.f. Agarwal 1979; Box 2005). It is not unlikely that a small new firm in the early 1900s were considered a (fairly) large new firm in the mid-1950s. The conceptions of small, medium and large firms probably vary significantly both over time and across economies. The number of small firms, measured as number of employees, has increased significantly over the past decades, and average firm size has therefore decreased (Traù 2003). In solving this problem each cohort has been ranked according to size. The largest third of firms in each cohort have been defined as large firms, and the remaining two thirds as small. Thus, a basic definition of start-up size is obtained. Naturally, other criteria could be used but his procedure is a compromise and can control for that the founding size – measured as stock capital – of new entrants actually decreased over time. The team/founder data on each of the 1,497 firms also comes from these archival sources, namely (i.) Founding team members (number of team members). In our study, the very first board of a company is defined as its founder team, regardless if the board remained for a very short time, or for a longer time. This variable cannot completely control for ownership of or involvement from other persons that were team members and founders, but had no formal position in the board (or ownership of the firm); (ii.) Gender. This variable is based on the members’ names. From this in11

formation, each firm and its founding team has been classified as all male, mixed gender, or all female founder teams; and finally (iii.) Spouse teams, measures whether a company was founded by spouses. The definition of spouse teams deserves some attention. A spouse team represents a company that was founded by two persons with the same surname, and if so, if they were a man and a woman. It is possible that some spouses had different surnames, which in our case would risk for underestimate the number of spouse teams. This is however quite uncommon in Sweden. Underestimation would also be the case if one man and one woman were not formally married (but had a relationship or were cohabitants), as well as for homosexual relationships/cohabitants. It is also possible that some companies that have been defined as founded by spouse teams in our database in fact were not spouses – for instance if the man and the woman shared the same surname but were father and daughter. Yet, in most cases women are often recorded as ‘Mrs.’ (Fru) or ‘Miss’ (Fröken), while men often were registered by their profession. Therefore, in our view, the risk for any overestimation of spouse teams is rather small.

Empirical Results In Table 2, we report team-related descriptives with respect to team size, gender, and spouse teams. In the entire population, the most common team size was two team members (nearly 50 percent). Founding team size varies over time, i.e. between cohorts, and it is hard to observe any particular trend. In our data, it seems to have become more common over time to start a new company with only one or with two members. New companies were on average smaller, which could explain this tendency. The share of women members (in all types of teams, single and multimember) 12

also became more intense over time, from 2.3 percent in 1912 to nearly 14 percent in 1950. In that respect, it seems that the business activity of women increased. The domination of men in single member firms was nearly total, and with exception for Cohort 1912 – in which no single member firm consisted of a woman – this domination furthermore seems to have increased (from approx. 90 to 98 percent between 1930 and 1950).

Table 2. Founding teams in cohorts 1912, 1930, 1942, and 1950. Descriptive statistics. Entire population

Cohort 1912

Cohort 1930

Cohort 1942

Cohort 1950

Founding team size (%) One member Two members Three members Four members Five or more members N firms

8.8 49.7 20.5 12.3 8.7 1,497

5.2 30.6 52.2 3.7 8.2 134

8.0 43.6 14.8 21.9 11.7 452

11.5 47.2 18.9 11.1 11.3 407

8.3 62.3 18.5 6.9 4.0 504

Gender of member, all teams (%) Male Female N members

88.2 11.8 4,093

97.7 2.3 390

87.6 12.4 1,346

87.5 12.5 1,139

86.4 13.6 1,218

94.7 5.3 132

100 7

88.9 11.1 36

95.7 4.3 47

97.6 2.4 42

71.4 27.7 1.0 1,365

93.7 6.3 127

68.3 30.3 1.4 416

66.9 32.5 0.6 360

71.4 27.5 1.1 462

6.5 98

2.2 3

6.6 30

9.3 38

5.4 27

Gender of member, single member firms (%) Male Female N firms Gender composition of multimember team (%) All male Mixed gender All female N firms Spouse teams (%) N firms

As regards the gender composition of multimember founding teams, Cohort 1912 once again stands out from the younger cohorts with nearly 94 percent all male founder teams and some six percent mixed gender teams (and no all female teams). The three younger cohorts show that around two thirds of the founder teams entirely 13

consisted of men, and around one third of mixed genders. The all female founder teams varied around one percent, the highest in Cohort 1930 and the lowest in Cohort 1942. Spouse teams firms, as defined in our study, represented 6.5 percent of the entire population of firms. The share of spouse teams varies, with its largest share in Cohort 1942 and the lowest in Cohort 1912. In sum, firms were generally started by men, both as single member teams and multimember teams. Mixed gender teams were in relative large minority (around one third in the entire population), but the overall share of women involved in new venturing activity was only a little more than ten percent. All female firms were more or less an exception as regards both single member and multimember teams: overall, they represented only 20 out of 1,497 firms. A somewhat larger fraction (n=98) consisted of spouse teams, but they were also in an absolute minority. How did new venture performance, measured as survival ability, manifest in our data? As could be observed in Table 1, the age data shows that a large proportion in the entire population died at young age. Treating the entire population as a ‘synthetic’ birth cohort in Figure 1, an analysis of the hazard during the firms’ first ten years confirms earlier research results of a liability of age: the risk for termination was on average higher at low age, and diminished over time (‘All firms’). Particularly over the first five years, this risk was also systematically higher for firms that started as small (‘Small firms’), compared to the segment that started on a larger scale (‘Large firms’).

14

Figure 1. Hazard of survival. First ten years. Entire population.

An event history analysis (Cox proportional hazards model) is used to test our hypotheses. This type of analysis makes it possible to model both the timing of an event, as well as the occurrence/nonoccurrence of the event. In this way, the simultaneous prediction of both whether and when an event will occur is allowed. Our specific model tests the probability (risk) for an event at a certain age, specifically if a firm were terminated before or at the age of ten years (0 if no, 1 if yes). Table 3 reports the independent variables in our longitudinal database that are used to test new venture performance. One variable measures a firm’s start-up size, dichotomized into small and large firms. The variable controls for if size affected the firms’ survival ability (liability of smallness). The variable Single member team is a dummy that controls for the effects on performance from single member teams – i.e. ‘solo entrepreneurs’ – compared to multimember teams. Multimember teams can, according to this defini15

tion, consist of two or more members. Another variable measures each firm’s team gender composition (at start), in which three categories are used: all male teams; mixed gender teams, and all female teams. The last independent variable measures whether or not a firm was founded by a spouse team.

Table 3. Definition of variables in the event history analysis. Variable

Measure

Start-up size/firm size

Dummy: Small = 0; Large = 1

Single member team

Dummy: No = 0; Yes = 1

Gender composition of team

Indicator: All male = 0 (reference category); Mixed gender = 1; All female = 2

Spouse team

Dummy: No= 0; Yes = 1

Our event history analysis, with five related models, confirms that venture performance were affected by the presence/absence of spouse teams and team gender composition. Nearly all variables in our test(s) show statistically significant effects. The first model controls for start-up size only. The risk (odds ratio) is below 1.0 and significant, which means that firms that were started on a larger scale performed better than smaller ones. This is in line with ‘stylized facts’ in much previous research. The second model adds the variable that measures whether a firm was founded by a single member team/solo entrepreneur. There is no statistically significant effect from this variable in the second model (nor in the other three following models) – venture performance did not generally seem to be affected by the presence or absence of a multimember team. Thus, it is hard to observe any general team effect. The third model adds the spouse variable. Here, it is clear that firms started by a spouse team had a statistically significant higher probability to survive compared to non-spouse firms.

16

The fourth model introduces the variable Gender composition of team (and excludes the spouse-variable). In comparison to the reference category (‘All male’), the presence of mixed gender teams do not seem to have had any statistically significant effect on business performance, while all female teams showed a significant effect. Here, the hazard was much higher than 1.0, which means that firms founded by all female teams had lower survival probability. The fifth model includes all four independent variables simultaneously. Firstly, as observed earlier, there was no obvious, general team effect. Secondly, it is observable that the gender composition of the start-up teams shows a statistically significant effect on venture performance: mixed gender teams had on average a higher risk for poor performance compared to the reference group (All male). All female-teams showed an even higher risk. Secondly, as in the second model, ventures founded by spouse teams displayed better performance than non-spouse teams.

Table 4. Event history analysis of venture performance. Model 1

Model 2

Model 3

Model 4

Model 5

0.746*

0.737*

0.733*

0.739*

0.730*

0.788

0.766

0.792

0.790

Mixed gender

1.112

1.276*

All female

2.564*

2.562*

Start-up size Single member team 1)

Gender composition of team

Spouse team

0.641*

0.540*

-2 Log Likelihood

7586.22

7573.99

7568.63

7565.40

7556.47

Chi-square

9.72*

11.89*

16.80*

23.40*

32.20*

Degrees of freedom

1

2

3

4

5

1) Reference category: ‘All male’. * = significant on at least the 5% level.

17

Discussion Our empirical results show that liabilities of age and size, regularly observed in previous research, affected venture performance. Our results also show that spouse teams and team gender composition had a clear impact on venture performance, while controlling for age and size. Interestingly, we found no overall team effect on venture performance: the fact that a firm had been founded by a team or a solo-entrepreneur did not affect its survival chances. However, this is in line with earlier research that indicates that even if there are studies that ventures with more than one founder were more likely to grow, they found no indications that they were more likely to survive (Cooper et al. 1994). Yet, some other team effects were obvious in our results and in that respect certain aspects and characteristics of teams in new venture performance are important to consider in the analysis as well as in future research. The empirical results can only partly confirm the first hypothesis (H1) which stated that Social similarity in founding team gender is positively related to the likelihood of new venture survival. Our indicator for social similarity was gender, and it is true that all male teams performed better than mixed gender teams (which were not socially similar). However, ventures led by all male teams performed not only better than mixed gender teams but also better than all female teams. Consequently, social similarity had a positive impact in some cases (men) but a negative impact in others (women). This is in line with previous research (e.g. Ruef 2010) but we believe that principally two related explanations must be used to explain this difference. Firstly, there is a built-in historical dimension in our data. Conditions for women and for women entrepreneurs were quite different from our days. Swedish women received full (formal) civil rights as late as 1921, and both formal and informal institutional constraints on women’s’ career opportunities and possibilities for business activity 18

and ownership actually remained at least into the 1950s. 4 This might explain our empirical results. As a consequence, it is probably necessary to explicitly make use of gender theory and of entrepreneurship research with a gender focus in order to explain this obvious ‘underperformance’ of ventures founded by women. Previous research has often shown that women entrepreneurs face discrimination in various ways systematically across time and place, which affects venture performance outcomes. 5 The second hypothesis (H2) stated that The presence of spouses in founding team is positively related to the likelihood of new venture survival. We can confirm this hypothesis, since there was a clear positive effect on survival amongst ventures that were founded by a spousal team (in this case a man and a woman). Our results are in line with earlier research (c.f. Hellerstedt 2009; Liao et al. 2009). This hypothesis assumed that with an initial presence of social affiliation, commitment and trust (as in marriages), there is a lagged positive effect of diversity on performance. Given the fact that firms founded mixed gender teams in general showed a higher probability for early termination (see below), we believe that this observation further supports our hypothesis. Our findings show different effects between the presence of certain characteristics of team social similarity (spousal teams) and team gender diversity on performance. With our longitudinal design we can add to earlier research that the collective impact of initial social interpersonal relationships and social similarity/diversity have significant effects also over longer time periods (ten years). Spousal venture teams had, in a relative sense, a higher probability for survival compared to other firms in our population. Mixed gender teams (non-spousal) and all female teams had a lower survival probability than all male founder teams. Firms founded by all female teams showed the highest relative risk for exit. Our interpretation is that the underlying aspects of 19

already existing high degrees of trust and social commitment in the founders teams that have a positive amplifying affect on gender team diversity when the venture ages and, perhaps, grows.

Conclusions We believe that these results are an important contribution to research on venture teams, spouse teams and gender. Our study has shown that variables that are identified in previous research as important for organizational performance – founding teams, gender and spouses – should not be analyzed isolated but simultaneously. Naturally, our data and the measurements employed in the study can be both refined and extended. As stated above, since our data is historical – and probably has a built-in historical/institutional dimension – it would be of great value to compare the effects of gender and spouses (i.e. similarity and diversity) on business performance on more modern data in order to generate new hypotheses and models and to test previous assumptions in research that relate to gender and gender theory, and research on teams and spouses. From our general results it is also possible to make more in-depth studies of particular cases or groups of firms and teams. This is possible and will be carried out in future research.

20

References Agarwal, N C. (1979), “Research Notes on the Interchangeability of Size Measures”, Academy of Management Journal, Vol. 22(2). Aldrich, H. E. and P. H. Kim. (2007). “Small Worlds, Infinite Possibilities?” Strategic Entrepreneurship Journal 1: 147-165. Anna, A. N., Chandler, G. N., Jansen, E., and Mero, N. P. (2000). ”Women business owners in traditional and non-traditional industries.” Journal of Business Venturing, 15(3), 279–303. Audretsch, D. B. (2002), “Entrepreneurship: A Survey of the Literature”, Prepared for the European Commission, Enterprise Directorate General. October, 2002. Box, M. (2005), New Venture, Survival, Growth: Continuance, Termination and Growth of Business Firms and Business Populations in Sweden During the 20th Century, Stockholm: Department of Economic History. Box, M. (2008), “The Death of Firms: Exploring the Effects of Environment and Birth Cohort on Firm Survival in Sweden”, Small Business Economics, Vol. 31(4). Box-Steffensmeier, J M. and Jones, B S. (2004), Event history modeling: A guide for social scientists, Cambridge: University Press. Carroll, G R. – Hannan, M T. (2000), The Demography of Corporations and Industries. Princeton: Princeton University Press. Chandler, G. N., and Lyon, D. W. (2001), “Issues of research design and construct measurement in entrepreneurship research: the past decade”, Entrepreneurship Theory and Practice, 25(4), pp. 101–113. Cooper, A. C. and Gimeno-Gascon, F. J. (1992). “Entrepreneurs, process of founding and new firm performance.” In Sexton D. L and Kasarda J. D. (eds.), The state of the art of entrepreneurship. Boston: PWS-Kent, pp.301-40. Cooper, A. C., Gimeno-Gascon, F. J., and Woo, C. Y. (1994). “Initial human and financial capital as predictors of new venture performance.” Journal of Business Venturing, 9(5), 371-395. Davidsson P. and Wiklund J. (2001). “Level of Analysis in Entrepreneurship: Current Research Practice and Suggestions for the future”. Entrepreneurship:Theory and Practice Vol 25 Issue 4 pp. 8199. Davidsson, P. (2008). The entrepreneurship research challenge . Cheltenham: Edward Elgar. Delmar, F. (1997), ”Measuring Growth: Methodological Considerations and Empirical Results”, in Donckels, R. – Miettinen, A. (eds.) Entrepreneurship and SME Research: On its Way to the Next Millennium. Hants: Ashgate. Du Rietz, A. and Henrekson, M. (2000), “Testing the Female Underperformance Hypothesis”, Small Business Economics 14, 1-10. Forbes, D., Borchert P. S., Zellmer-Bruhn M. E., and Sapienza H. J.. (2006). “Entrepreneurial Team Formation: An Exploration of New Member Addition.” Entrepreneurship Theory and Practice, 31(March) pp. 225-248. Francis D. and Sandberg W. (2000), “Friendship within Entrepreneurial Teams and its Association with Team and Venture Performance”, Entrepreneurship, Theory and Practice, 25(2), pp. 5-26. Gartner, W. B. (1989), “’Who is an Entrepreneur?’ is the wrong question”. Entrepreneurship Theory and Practice, 13, pp. 47-68. Glenn, N D. (1977), Cohort Analysis. Beverly Hills: Sage. Gratzer, K and Box, M (2002), ”Causes of Selection Amongst Swedish Firms: A Contribution to the Development of a Business Demography” (2002), Scandinavian Economic History Review, 2002:1. Hagenaars, J A. (1990), Categorical Longitudinal Data: Log-linear Panel, Trend, and Cohort Analysis. Newbury Park: Sage. Hellerstedt, K. (2009), The Composition of New Venture Teams – Its Dynamics and Consequences. Jönköping International Business School (Diss.) Janson, C-G (2000), “The Longitudinal Study”, in Janson, C-G (ed.), Seven Swedish Longitudinal Studies in Behavioral Sciences. Report 2000:8. Stockholm: FRN. Kamm, J. and Nurick, A. (1993). “The Stages of Team Venture Formation: A Decision Making Model”, Entrepreneurship, Theory and Practice, 17(2), p.17-27. Karlsson Stider, A. (1999), ”Hemma hos firmafamiljen”, Kvinnovetenskaplig tidskrift, 20(1). Larsson Segerlind T. (2009). Team Entrepreneurship - A Process Analysis of the Venture Team and the Venture team Roles in relation to the Innovation Process. Stockholm University.

21

Lechler, T. (2001). “Social Interaction: A Determinant of Entrepreneurial Team Venture Succes.” Small Business Economics, 16, pp. 263-278. Liao, J, J. Li and W.B. Gartner. (2009) “The effects of founding team diversity and social similarity on venture formation.” Journal of Business Venturing. (article in press.) Low, M. and I. C. MacMillan. (1988). “Entrepreneurship: Past Research and Future Challenges.” Journal of Management 14: 139-161. Mason, M. A. and Wolfinger, N. H. (2004), “Cohort Analysis”, in International Encyclopedia of the Social & Behavioral Sciences, Elsevier (eds. Smelser, N. J. & Baltes, P. B.). Menard, S., (1991). Longitudinal research. Newbury Park:Sage. Ruef M., H.E. Aldrich and N.M. Carter. (2003). “The structure of founding teams: homophily, strong ties and isolation among US entrepreneurs.” American Sociological Review, 68: 195-222. Ruef, M. (2010) The Entrepreneurial Group : Social Identities, Relations, and Collective Action. Princeton University Press. Sandberg, W.R. (1992), “A strategic managements’s potential contributions to a theory of entrepreneurship.” Entrepreneurship Theory and Practice 16 1, pp. 73–90. Sharma, P. (2004), “An Overview of the Field of Family Business Studies: Current Status and Directions for the Future”, Family Business Review, 17(1): 1-36. Starbuck, W H (1965), Organizational Growth and Development, Harmondsworth. Storey, D. J. (1998). Understanding the small business sector. London: Thomson Business Press. Statistics Sweden (1940, 1951). Statistisk årsbok för Sverige. Stockholm: Statistiska centralbyrån. Sundin, Elisabeth (1995), ”Kvinnors företagsamhet - en avspegling av samhället”, in Pengarna och livet. Stockholm: NUTEK (Närings- och teknikutvecklingsverket. Svanström, Y. (2003), “Kvinnor i styrelser och som chefer – en historisk bakgrund”. In Mansdominans i förändring: om ledningsgrupper och styrelser: betänkande. Stockholm: Fritzes offentliga publikationer. Traù, F. (2003), Structural Macroeconomic Change and the Size Pattern of Manufacturing Firms. Basingstoke: Palgrave Macmillan. Watson, J. (2000), “Comparing the Performance of Male-And Female-Controlled Businesses: Relating Outputs to Inputs” Entrepreneurship: Theory and Practice, Vol. 26.

22

Notes

1

A cohort is an aggregate of individual units, at least in a statistical sense – all units in a cohort do not necessarily interact. The specific cohorts (founding years) are specifically selected because different conditions prevailed in 1912 compared to 1930, 1942, and 1950. These conditions are so-called cohort effects – which affect the attributes and the development of the research units – that originate in the historical periods during the units experienced various stages of life. Differences between cohorts may prevail over longer periods – not only over the first, formative years (Glenn, 1977). See also Mason and Wolfinger (2004). 2

As an example, it is imaginable that Stockholm companies – active in an increasingly growing, urban environment during the first half of the 20th century – deviated from companies in the rest of the country in that that they may have established in service and trade industries to a greater extent. Historical figures (Statistics Sweden, 1940; 1951) however show that our Stockholm cohorts represent a rather large share of the entire Swedish population of joint-stock company start-ups during these years. In the period 1911/15, a yearly average of around 600 new companies was registered in the entire country. This makes the cohort from 1912 (n = 134) to stand for 22 percent of all company start-ups in Sweden. The corresponding figures during the periods in which the other cohorts were founded are even higher: Cohort 1930 represents approximately 44 percent of all new companies (n = 452); Cohort 1942 around 30 percent (n = 407), and Cohort 1950 about 30 percent (n = 504). In that respect, our data covers a large share of all new joint-stock ventures in these years. 3

The archives of the Swedish Patent and Registrations Office, preserved at the Swedish National Archives and at the Swedish Companies Registration Office, are the main source for the empirical material. At founding, each company was assigned with a registration number, which was maintained throughout its active time. In principle, every individual company in Sweden founded 1897 to 1960 can quite easily be found and traced in the archives. Box (2005). 4

For an overview, see Svanström (2003: 53-65). If unmarried, a woman had her father as legal guardian. In 1858, an unmarried woman obtained the right to apply for majority at court (i.e., her right was the right to apply). If a woman with majority would marry, she once again received a legal guardian (the husband). As widow, however, the woman once more had majority. In 1921 women received formal rights but due to both informal and formal institutional conditions as regards rights to control personal assets etc, several restrictions remained well into the 1950s. 5

Previous research has found that firms owned by women generally underperform relative to firms that are owned by males. This would suggest that enterprises run by women will exhibit poorer performance since women are discriminated, or because of systematic or structural factors that deprive women of vital resources such as education, networks, etc. See Cooper, Gimeno-Gascon, and Woo, 1994; Anna, Chandler, Jansen, and Mero, 2000; Du Rietz and Henrekson, 2000; Watson, 2002.

23