Personality Traits and the Likelihood of Self-Employment: A Journey into the Crafts’ Way of Doing Business

2.1 The Crafts Entrepreneur

According to a widely used theoretical typology (e.g. Miner, Smith, and Bracker 1992; Solomon and Mathias 2020; Woo, Cooper, and Dunkelberg 1991), which goes back to Smith (1967) and Smith and Miner (1983), there is a basic entrepreneurial type that can be expected to be closely associated to business activity in the crafts. For the ‘crafts entrepreneur’, the opportunity for self-directed work and personal autonomy constitutes a primary motivation for venturing into self-employment (Binder and Blankenberg 2022; Bridge and O’Neill 2018; Hornaday 1990; Runst and Thomä 2021). According to this concept, these individuals are less motivated by financial gains or the opportunity to build a large business organization. Instead, crafts entrepreneurs can be more succinctly characterized by personal goals such as a desire for independence, autonomy, and self-realization through active participation in the productive and creative processes (Amin and Roberts 2008; Kuhn and Galloway 2015; Smith and Miner 1983; Solomon and Mathias 2020; Woo, Cooper, and Dunkelberg 1991). They value the act of practicing a trade or occupation, and thus the mastery of knowledge and skills, i.e. the achievement of craftsmanship (Binder and Blankenberg 2022; Hoyte 2019; Nooteboom 1994; Sennett 2008; Thomä and Zimmermann 2020). On the other hand, they tend to see administrative and management tasks tend to be perceived as a necessary evil of entrepreneurial activity (Das and Teng 1998; Smith and Miner 1983; Woo, Cooper, and Dunkelberg 1991). Accordingly, people who opt for self-employment in the crafts often strive for personal competency and creative problem-solving in the creation of individualized, non-mass-produced products and services. They often do not have a well-developed marketing strategy, and if they do, it emphasizes the company’s reputation for delivering high quality.

Since crafts entrepreneurs often find personal self-fulfillment in the acquisition of tacit know-how in a specific technical field, innovations conducted by crafts enterprises are usually of a customized nature and often take the form of step-by-step optimization, further development or adaptation of existing products, services and processes (i.e. incremental innovations; Amin and Roberts 2008; Das and Teng 1998; Miner, Smith, and Bracker 1992; Runst and Thomä 2022; Thomä and Zimmermann 2020).

Moreover, crafts entrepreneurs tend to stick to tried and tested routines (Das and Teng 1998; Hoyte 2019; Miner, Smith, and Bracker 1992). In doing so, they perfect traditional ways of working and doing in a conscientious manner. When they make changes, they do so less proactively and are less likely to create novel markets and introduce fundamentally new products, processes and services. Instead, self-employed people of crafts type behave relatively more reactively: they tend to adapt more slowly to changing conditions and only do so to the necessary extent to safeguard their existing business model, thus limiting entrepreneurial risk (Bridge and O’Neill 2018; Miner, Smith, and Bracker 1992). Crafts entrepreneurs therefore often run long-lasting businesses that tend to remain small, preferring stability over growth (Solomon and Mathias 2020).

2.2 The Formal and Informal Institutional Background of the German Crafts Sector

The German Trade and Crafts Code (TCC, Handwerksordnung) defines and regulates more than 90 crafts trades. In many of these trades, the law requires advanced vocational training to be eligible to start a business, i.e. the head of a crafts company must hold a so-called Meister-degree.^[1] In 2004, some trades have been partially (e.g. bakers and butchers) or fully (e.g. brewers, instrument maker) deregulated. However, some of these reforms have been reversed in 2020. According to official statistics (Federal Statistical Office), there were 5.4 million active craftspeople in Germany in 2021. Of these, around 80 per cent are in crafts trades with entry restrictions that require the owner or employed manager to have a master craftsmen’s qualification. It is therefore reasonable to assume that for most craftspeople the path to self-employment is a sequential one, starting with basic vocational training (and therefore wage employment in the crafts sector) before potentially moving on to become a self-employed business owner in the crafts sector. Nevertheless, we must not lose sight of alternative paths. For example, as a result of the 2004 TCC deregulation, a certain share of individuals may move from wage employment in the non-crafts sector to self-employment in the crafts sector. Another more likely potential pathway is from crafts wage employment to self-employment outside the crafts sector.

In addition to the formal institutions of the German crafts sector, the informal institutional background of the crafts must also be taken into account. For example, the results of Runst and Thomä (2021) provide empirical evidence for the assumption that the likelihood of being a ‘crafts entrepreneur’ type is considerably higher among self-employed individuals in the German crafts sector than in non-crafts occupations, as delineated by the TCC. Based on data from the IAB/ZEW Start-up Panel – a representative survey of start-up businesses in Germany – they show that the desire for self-determined work and the implementation of a new business idea are the most frequent driving forces for starting up a business, both in the crafts and in the non-crafts. However, the motive of ‘self-determined work’ is significantly more important among founders in the crafts than in non-crafts occupations. This points to the singular motives and goals of the self-employed in the crafts, in line with the ‘crafts entrepreneur’ described above, which are driven less by the desire for profit and growth than by the desire for personal autonomy, self-realization and expertise in one’s own trade or occupation. In contrast, the implementation of a new business idea is relatively less often the central motive for founding a company in the German crafts sector – which in turn may indicate a lower personal focus on profit and growth.

2.3 Broad and Narrow Personality Traits

The Big Five is a widely used model of personality traits (Digman 1990; John, Donahue, and Kentle 1991; McCrae and Costa 2008). It states that the main facets of human personality can be condensed into five basic dimensions that remain more or less stable across the lifespan: extraversion, conscientiousness, emotional stability, openness to experience and agreeableness. Extraversion indicates a preference for social interactions. It is also increasingly associated with a higher reward sensitivity (see Cohen et al. 2005; Lucas et al. 2000). Agreeable individuals try to avoid interpersonal conflict, forgive others more easily, and are more thoughtful in their interactions with others. They prefer cooperation over competition. In addition, agreeableness is associated with expressing oneself carefully to avoid upsetting others and showing a higher degree of empathy. On the other hand, people with higher scores of conscientiousness act in a controlled and structured way. They plan ahead and tend towards perfectionism. At the same time, they show a higher degree of accuracy, diligence, precision and determination. The trait of emotional stability indicates how well a person is able to deal with setbacks and stress factors. High emotional stability is associated with less severe mood swings, less anxiety, sadness, insecurity or other negative feelings. Openness indicates how willing a person is to make new experiences. Open-minded persons enjoy change and are more likely to be creative, imaginative and experimental, while they also have a strong imagination.

The impact of the Big Five on the probability of self-employment and related entrepreneurial decisions has been extensively studied. Extraversion and openness in particular and – less clear – emotional stability and conscientiousness can have a positive influence on entrepreneurial decisions (e.g. Brandstätter 2011; Caliendo, Fossen, and Kritikos 2014; Runst and Thomä 2023; Zhao, Seibert, and Lumpkin 2010). Agreeableness shows no effects, apart from one paper finding a positive impact on exits from self-employment (see Caliendo, Fossen, and Kritikos 2014). Zhao, Seibert, and Lumpkin (2010) argue that agreeableness should be negatively related with both entrepreneurial performance and start-up intention, although their empirical results cannot support either hypothesis.

While the Big Five constitute broad personality traits, narrow personality traits are also discussed in the literature. The four most important narrow traits associated with entrepreneurship are achievement motivation, need for autonomy, risk tolerance and locus of control (Bridge and O’Neill 2018). Based on the available data (see Section 3), the influence of the latter two of these traits on self-employment can be examined. A person’s self-efficacy belief or locus of control (LOC, Rotter 1966) reflects the conviction that a person has control over their own life. By contrast, a person scoring low on LOC believes that his or her life is shaped by external factors over which he or she has no control. Furthermore, the degree of individual risk tolerance (Chell, Harworth, and Brearley 1991) is a narrow personality dimension that is plausibly related to entrepreneurial decisions, since the outcomes of self-employment are often uncertain. Moreover, humans have different risk preferences: some people prefer smaller, more secure income streams, while others would rather choose larger but more uncertain ones. The effects of LOC and risk tolerance have also been repeatedly investigated, supporting the hypothesis of a positive impact on the probability of self-employment (see e.g. Caliendo, Fossen, and Kritikos 2014).

2.4 Hypotheses

On the one hand, individuals who are open to new experiences are more likely to recognize entrepreneurial opportunities than others, which should also be true for the crafts entrepreneur (Caliendo, Fossen, and Kritikos 2014). For example, in her conceptual paper, Hoyte (2019) describes the openness of artisan entrepreneurs (representing a subsample of craftspeople who are found in the cultural and creative industries) and their willingness to pass on cultural traditions. On the other hand, the degree of openness may be less important for crafts entrepreneurs as they are less focused on growth, profit-making and radical innovation (see Section 2.1). If one’s primary goal is to practice a trade and to generate incremental innovations in response to situational customer needs or circumstances, openness is less relevant than, for example, in highly innovative ventures that need to be constantly on the lookout for new trends and developments. In the case of crafts, openness should therefore be less strongly associated with self-employment than in the non-crafts (Hypothesis H1).

Entrepreneurship is always socially embedded (Sarasvathy 2001, 2008; Storr 2008), as entrepreneurs need to communicate and socialize with employees, customers, suppliers, financial institutions and other external partners. Generally speaking, extroverts are more attracted to self-employment than introverts (Caliendo, Fossen, and Kritikos 2014). This effect should be equally pronounced among crafts entrepreneurs. Overall, extraversion is therefore hypothesized to be a key driver of self-employment within and outside of the crafts sector (Hypothesis H2).

The relationship between the trait of conscientiousness and self-employment is somewhat ambivalent. On the one hand, highly conscientious people who value routines may not be attracted to the uncertainties of running a business. On the other hand, the efficiency orientation, determination and precision of conscientious people may promote business performance in the long run (Brandstätter 2011; Zhao, Seibert, and Lumpkin 2010). Overall, the empirical literature does not provide evidence of a clear link between conscientiousness and self-employment. However, a positive effect of conscientiousness can be expected in case of the crafts. The desire for self-realization through the practice of a trade, craft or occupation – which is strongly anchored in the crafts entrepreneur type (see above) – and the associated high level of acquired professional-technical knowledge and skills suggests that a higher level of conscientiousness should positively affect self-employment in case of the crafts (Hypothesis H3). This hypothesis is in line with the conceptual proposition by Hoyte (2019) on the potential role of conscientiousness for artisan entrepreneurship in the cultural and creative industries. She states that crafts-based work requires a high degree of conscientiousness not only to learn a craft but also to practice it continuously throughout one’s life and pass it on through generations, thus forming a common occupational identity.

Emotional stability – in particular the ability to make well-considered decisions under time pressure and deal with complex stress factors – is likely to be a general advantage for self-employed individuals (Zhao and Seibert 2006; Zhao, Seibert, and Lumpkin 2010). However, it can also be argued that less emotionally stable individuals may prefer owning a business (most likely a small one) to being in paid employment, as they are less likely to cope well as employees in a challenging work environment, e.g. characterized by stressful team interactions. Thus, individuals scoring low on emotional stability might engage in (necessity) entrepreneurship because it offers them a chance to leave wage employment. In summary, the effect of emotional stability on self-employment is ambivalent, and there are no specificities to be expected in the case of crafts entrepreneurship.

The possible role of agreeableness also seems ambiguous: on the one hand, entrepreneurs need a certain amount of willpower in difficult conversations and the ability to tolerate such conflicts, while on the other hand, a very low level of owner agreeableness may resemble stubbornness and has the potential to disrupt social interaction to such an extent that employees may deliberately withhold their opinions and ideas. If owners insist on their own opinion without considering and including the thoughts of others, it can severely disrupt the business culture and hinder the co-creation of valuable new things (Runst and Thomä 2022). The possible influence of agreeableness on self-employment is thus – as in the case of emotional stability – ambivalent, supported by the mixed empirical results on this trait in the empirical literature (for an overview, see Brandstätter 2011). In the case of crafts, however, where practicing a trade or occupation in a small company (often without employees) can be a primary motivation for self-employment (see the discussion above on the ‘craft entrepreneur’), individuals with lower agreeableness scores may be particularly inclined to start a business. In this way, they can earn a living and realise their ambition to accumulate skills and experience-based competence in a particular trade or occupation without the need to interact and coordinate with co-workers or superiors. Thus, agreeableness should be negatively related to self-employment in the crafts (Hypothesis H4).

In addition to the Big Five, the narrow personality traits of LOC and risk tolerance also play an important role in self-employment decisions (Caliendo, Fossen, and Kritikos 2014; Leutner et al. 2014). A lack of belief in the effectiveness of one’s own actions is likely to have a detrimental effect on self-employment in general. A higher LOC should therefore be beneficial for starting and succeeding in business ownership regardless of the sector (Hypothesis H5). A typical business owner must also be prepared to take personal risks, although it can be assumed that the relative importance of risk tolerance differs between self-employed individuals in the crafts than in non-crafts occupations. The ability to bear risk is particularly important for self-employed persons who have a strong focus on high risk-high reward projects. By contrast, it is less relevant in the case of self-employment, when only tried and tested ways are being followed – as tends to be the case in the crafts (see above). While this does not mean that the willingness to take risks is unimportant among crafts entrepreneurs (Das and Teng 1998) – indeed it is important for all forms of entrepreneurship – typically the rates of new business survival are higher in the German crafts compared with the non-crafts case (Runst and Thomä 2021). Overall, risk should therefore have a lower impact on self-employment decisions in the crafts than in the non-crafts case (Hypothesis H6).

3.1 The German Socio-Economic Panel

Our empirical analysis is based on data from the German Socio-Economic Panel (GSOEP) for 2005 to 2019, a large-scale and representative annual household survey of more than 10,000 individuals in Germany.^[2] The SOEP has been used repeatedly in entrepreneurship research on the Big Five (e.g. Caliendo, Fossen, and Kritikos 2014; Caliendo, Fossen, and Kritikos 2022; Runst and Thomä 2023). Since 2005, a fifteen-item Big Five Inventory (BFI) has been included at regular intervals (i.e. in the five survey years of 2005, 2009, 2013, 2017 and 2019). Small item scales such as the BFI-15 have been shown to be reliable and valid compared to longer versions such as the BFI-44 (Rammstedt and John 2007). Information on the narrow personality trait of LOC is available for 2005, 2010 and 2015, while risk tolerance was surveyed in all years. For the missing years, it is assumed that a person’s traits have not changed and can be replaced by the last values available for the respective person. This assumption is plausible because personality traits are heritable and remain relatively stable over longer periods of time, especially in adulthood (Cobb-Clark and Schurer 2012; Rantanen et al. 2007; Wortman, Lucas, and Donnellan 2012).

The Big Five are generated via a factor analysis, with which the fifteen items are condensed into factor scores standardized to a mean of approximately zero and a standard deviation of approximately one for each of the five personality traits. As an example, the factor loadings for 2005 are shown in the Appendix (Table A1), which correspond to expected patterns (e.g. Hahn, Gottschling, and Spinath 2012; Lang et al. 2011).^[3] The LOC variable is measured in the same way based on ten GSOEP questions on an individual’s perceived self-efficacy, with all respondents’ answers loading onto one factor.^[4] The variable risk tolerance is derived from a single GSOEP question, a self-assessment of one’s own willingness to take risks (on a ten-point Likert scale).

The dataset also contains information on a number of other socio-demographic characteristics of the survey respondents (e.g. age, nationality, educational qualification), which are controlled for in the regression analysis. In line with Caliendo, Fossen, and Kritikos (2014), and Runst and Thomä (2023), the dependent variable self-employment is equal to one if an individual is self-employed, and zero otherwise. While fully recognizing that the concept of entrepreneurship and self-employment are not the same, self-employment is one important aspect of entrepreneurship and has been widely used in empirical research as an indicator of entrepreneurship. Table 1 provides a brief description of all variables, and displays variable means and standard deviations.

The empirical analysis refers to individuals aged between 19 and 59 years of age. In addition, in line with previous research (Caliendo, Fossen, and Kritikos 2014; Caliendo, Fossen, and Kritikos 2022) we exclude certain individuals from the analysis. Disability pensioners, students, farmers, family workers, civil servants and military personnel are excluded from the sample because they have a restricted occupational choice set.^[5] Furthermore, observations from the 2016 and 2017 GSOEP “Refugee samples” are not included in the analysis to ensure sample consistency over time and because there are clear personality differences between the specific group of refugees and the general German population (Runst and Thomä 2023).

The crafts case is captured in the data by using an occupation-based approach (for more details, see Runst and Wyrwich 2023). Depending on the empirical model used, it is based on a person’s past or current occupational status and thus on the concrete professional activities. This allows a relatively clear distinction between crafts and non-crafts in the sense of the hypotheses formulated above. The sorting of persons in crafts and non-crafts occupations is undertaken according to the scheme developed by Runst et al. (2019) and Haverkamp, Wesling, and Seibert (2019), based on the official German classification of occupations (KldB 92 and KldB 2010). Due to the long-time horizon of the GSOEP, this study has opted for the KldB92 classification. A complete list of KldB92 occupations and their crafts or non-crafts assignment can be found in Runst et al. (2019). The craft occupations in our sample, including their respective sample shares, are listed in Table A2 in the Appendix. As a robustness test, this list of crafts trades in Germany is further restricted to crafts occupations that belong to cultural and creative industries (artisan trades, see Table A2), i.e. what might be called the traditional core of crafts entrepreneurship (see the conceptual paper of Hoyte 2019).

3.2 Regression Models

In this study, we use a multi-step analysis design. As a first step, we examine the impact of broad Big Five personality traits, along with narrow traits such as LOC and risk tolerance, on the likelihood of individuals falling into one of four employment categories: wage employment or self-employment, either inside or outside the crafts sector. We therefore run a multinomial logistic regression of the form:

where the dependent variable is the probability of falling into a certain employment status relative to the baseline category in period t, i.e. wage employment in the non-crafts.^[6] The variable ‘Trait’ corresponds to either the broad BF traits or the narrow traits LOC or risk. The lagged ‘Crafts’ indicator equals one if a person was economically active in a crafts occupation at time t − 1, t − 2, or t − 3 and zero otherwise. This lagged crafts variable controls for the fact that the choice of occupation usually precedes the decision to become self-employed (Fritsch 2013). It should be noted, however, that the categorial nature of the dependent variable does not assume that occupational choice precedes the decision to become self-employed, but allows them to be determined simultaneously. If an individual is unemployed at t − 1, we cannot determine the crafts status. In this case we have to rely on t − 2 and t − 3 in order to assign the crafts status. If a person is unemployed in three previous periods, we treat them as belonging to the non-crafts sample.^[7]

A number of control variables are used, such as person’s age, education and training qualifications, a dummy for lagged part-time employment, gender, nationality, lagged unemployment status, work experience in years, the disability status, and a binary indicator for whether the father was self-employed, and capital income (see Table 1 for an overview). We also control for year and region fixed effects. The main research question of this first step is therefore: Is the decision to become self-employed in the crafts or the non-crafts related to personality traits, and are there effects that are specific to the crafts sector, in line with the hypotheses developed in Section 2.3?

In a second step, we take an alternative approach to validate and deepen the results of the first step and run a random effects panel regression model. We test whether personality traits, moderated by crafts sector affiliation in the last three years, are associated with the likelihood of being self-employed in the current period. The model is as follows:

The binary self-employment indicator equals one if an individual is currently self-employed and zero otherwise. The lagged crafts indicator equals one if an individual was economically active in a crafts occupation at time t − 1, t − 2, or t − 3 and zero otherwise. The coefficient of the lagged crafts by trait interaction thus captures the effect of a trait on self-employment for a person who has worked in the crafts in the past. In this specification, self-employment is allowed to take place inside or outside the crafts sector Our hypotheses developed in Section 2.3 suggest that the interaction between the binary variable ‘Crafts’ and certain personality traits will be significantly different from zero, which would imply that the relationship between personality traits and the likelihood of self-employment is not stable across sectors.^[8] In addition to the main dependent variable ‘self-employment’, we also use ‘entry’ into self-employment as an additional entrepreneurship indicator using the same model specification.

In this second step of the regression analysis, we can also examine the possible relationship between certain personality traits and self-employment from an additional perspective: On the basis of a split sample design, we divide the dependent variable for self-employment status into groups of self-employed according to the number of employees they have (i.e. self-employed in general, solo self-employed, self-employed with 1–5 employees) in order to examine whether the craft-specific self-employment drivers we have identified depend on firm size.^[9] This question is relevant in the present context, because as discussed in the theory section above, one of the central motivations of the ‘craft entrepreneur’ is not growth and profit, but the pursuit of personal autonomy, self-realization and the acquisition of experiential skills in one’s own trade or occupation.

The third step in our empirical analysis is to examine whether self-employment success depends on personality traits. This serves to investigate whether the crafts-specific determinants of self-employment identified in the first two steps are effective in terms of performance, i.e. whether self-employed persons with corresponding traits do better on various measures. We therefore restrict our sample to the self-employed in crafts and non-crafts occupations and run various regression models of the following form:

Self-employment success is assessed using three different measures: growth (i.e. the number of employees, measured as an ordinal variable with 7 categories), work satisfaction (measured as an ordinal variable with 11 categories) and the probability of exit. Ordered logistic models are used for growth and work satisfaction, and a random-effects panel regression model is used for the dependent variable on exit from self-employment. In contrast to the first and second step regressions, here ‘Crafts’ is defined as an indicator of current economic activity in the crafts sector. The coefficient of the corresponding crafts by trait interaction therefore captures the effect of a personality trait on a measure of self-employment success in the crafts sector.

4.1 Employment vs. Self-Employment in Crafts Compared to Non-crafts

The results of the multinomial logit regression (model 1) are shown in Table 2. The coefficients should be read in relation to the base group of non-crafts wage employment. Most importantly, columns (3) and (4) show the coefficients for self-employment in the non-craft and craft sectors respectively. Extraversion is positively associated with self-employment regardless of the sector. A one unit increase in the extraversion score increases the multinomial log odds of preferring self-employment (over non-craft wage employment) by 0.20–0.22, holding everything else constant.^[10] A test of equality between these two coefficients cannot be rejected (see column 5, Table 2). The effect is statistically significant at the 1 percent level. This is supported by the marginal effects reported in Table 3. A one unit increase in the extraversion score increases the absolute probability of being self-employed by 1.6 and 0.1 % respectively. Given the baseline probabilities of 9.4 % and 0.6 % of falling into the self-employment categories of non-crafts and crafts, the size of the effect should be considered to be large.

Next, conscientiousness appears to affect self-employment in the opposite direction in the non-crafts and crafts occupations. A one unit increase in the conscientiousness score reduces the probability of preferring self-employment to wage employment by 4.6 % in non-crafts but increases it by 16.8 % in the crafts. According to the marginal effects (Table 3), the former effect is rather small, while the latter amounts to a substantial one-sixth increase compared to the baseline probability. Interestingly, conscientiousness has a similarly positive effect on wage employment in the crafts sector (relative to non-craft wage employment, Table 2), so it seems to be an overall crafts effect. The corresponding equality test between these two coefficients cannot therefore be rejected (see column 6, Table 2). However, the marginal effects in Table 3 suggest that the size of this effect is much stronger in crafts self-employment than in crafts paid employment, when compared to the baseline probability.

While the coefficient of emotional stability is negative for self-employment in both sectors, its size is economically significant only in the crafts, where a one unit increase in the respective factor score reduces the baseline probability by about one sixth (see Tables 2 and 3). In contrast, openness has a positive effect on self-employment in both crafts and non-crafts, and the effect size is large in both cases. The final BF characteristic, agreeableness, is only statistically significant and negative in crafts occupations (Table 2). Again, the marginal effects (Table 3) suggest that the corresponding effect size is quantitatively meaningful. Finally, the narrow personality traits are similarly associated with self-employment in all sectors. An increase in risk or LOC increases the likelihood of preferring self-employment in the crafts and non-crafts to wage employment.

Overall, these results suggest that some traits affect self-employment decisions similarly across sectors (extraversion, openness, risk and LOC), but others seem to be craft-specific, most notably conscientiousness, emotional stability and agreeableness. We therefore find support for hypotheses 2, 3, 4 and 5, while hypotheses 1 to 6 are not confirmed. On the basis of these results, it is possible to imagine an entrepreneurial personality that theoretically maximizes the likelihood of becoming self-employed (in crafts or otherwise). A one-unit deviation increase or decrease in the scores of the seven traits examined here can theoretically double the baseline probability of being self-employed in the craft sector and almost double it in the non-craft sector. Thus, we conclude that personality effects are economically significant.

4.2 Drivers of Self-Employment in General

Table 4 shows the differential impact of the personality traits on the likelihood of being currently self-employed (regardless of sector) for those who have been economically active in the crafts sector in the last three years and those who have not (column 1, Table 4). In addition, a distinction is made according to the number of employees of the self-employed in order to obtain more detailed information on the possible origin of the craft-specific effects identified in the previous section (columns 2 to 4, Table 4). As a robustness test, a further specification uses entry as an alternative dependent variable. This measures whether entry into self-employment took place in the current period (column 5, Table 4).

In line with the multinomial logit results, the trait extraversion is generally positively associated with self-employment, however, there is no evidence for a stronger impact in the crafts. In line with the previous set of results, we find a negative association for conscientiousness and self-employment in general but a positive one in the crafts sector. The latter is mainly due to the group of solo self-employed, where there is a positive association with the trait conscientiousness (column 2, Table 4). At the same time, the craft-specific conscientiousness effect is confirmed when entry is used as dependent variable (column 5). At this point, the question could be asked whether the crafts-specific conscientiousness effect might simply compensate for the negative effect in the non-crafts sector, and thus only have a relative effect. However, an additional test to check the total (or absolute) effect by personality traits on self-employment in the crafts sector shows that conscientiousness has a positive total effect on self-employment in the case of solo self-employed and in the case of entry (see Table A3 in the Appendix).

There is no general effect of emotional stability on the dependent variable. In line with the multinomial logit results, there seems to be some negative impact in the crafts sector, but it is not statistically significant. Openness is positively related to self-employment in general. In the case of the crafts there is a weaker relationship only for solo self-employed and for entry, although this is only significant in the latter case.

The last BF trait, agreeableness, generally has no positive effect on self-employment. However, as with the multinomial logit results, there is an additional negative effect in the crafts sector, mainly due to being a solo self-employed. The two narrow traits, risk tolerance and LOC, are generally positively related to self-employment and there are no additional effects due to being in the crafts sector. Overall, the results of the second step of the empirical analysis are similar to the multinomial logit results presented above, despite the slightly different modelling approach and the slightly different research question underlying both models: Once again we see a positive effect of conscientiousness and a negative effect of agreeableness in the crafts case, both particularly for the solo self-employed. Thus, hypotheses 3 and 4 are again supported by the empirical evidence. As with the multinomial logit results, emotional stability shows a negative relationship (again strongest among the solo self-employed), but this is now not statistically significant.

4.3 Self-Employment Success

Table 5 shows the regression results on several indicators of self-employment success. As mentioned above, we examine whether the identified craft-specific personality factors (notably conscientiousness and agreeableness) also affect the performance of craftspeople once they are self-employed. First, we consider the dependent variable ‘number of employees’, which we interpret as an indicator of the growth orientation of a self-employed person. The results of an ordered logistic regression show that extraversion is generally positively related, while openness and agreeableness are negatively related to the number of employees a self-employed person has (column 1, Table 5). In contrast to agreeableness (and emotional stability), conscientiousness is again confirmed as a specific personality factor in crafts entrepreneurship: There is a negative relationship with the growth orientation of the crafts self-employed (Table 5). The marginal effects at each level of the ordinal categories of the dependent variable are presented in Figure A1. It reiterates the role of the solo self-employed: Once a self-employed person starts to hire employees in his or her crafts enterprise, the effect of conscientiousness becomes negatively significant. This effect diminishes as firm size increases, but always remains negatively significant.

In contrast, the results in column 2 of Table 5 suggest that work satisfaction is generally affected by personality characteristics. All BF traits are positively associated with work satisfaction as well as LOC, and to some extent this is also true for risk. In line with our theoretical outline above, we find an additional conscientiousness premium in the crafts, i.e. higher levels are positively related to work satisfaction among self-employed craftspeople (Table 5, see also Figure A2). Thus, in line with the above findings on the probability of self-employment, conscientiousness can be singled out as a trait specific to the crafts.

Finally, we find no evidence of personality effects on the likelihood of exit, except for two traits: Higher levels of LOC generally have a negative effect on exit, highlighting the favorable influence of this trait on self-employment success. There is also a positive effect of agreeableness in the case of the crafts self-employed, which is in line with the above findings on the effect of agreeableness in the crafts sector.

Overall, therefore, the multi-step analysis that has been carried out provides a conclusive picture, particularly with regard to the trait of conscientiousness (Hypothesis H3). In the theoretical discussion of the ‘craft entrepreneur’, we hypothesized that conscientiousness in the craft sector corresponds to the desire of many self-employed to realize themselves on the basis of their experience-based skills and competences and, for this reason, to deliberately keep their own firm small in order to avoid being overburdened with the management and organizational tasks that automatically go hand in hand with increasing firm size (Hypothesis H3). The empirical results support this hypothesis by showing that conscientiousness is a craft-specific driver that has a positive effect on the decision to become self-employed in the crafts sector. Indeed, it is associated with lower growth orientation and, at the same time, higher work satisfaction among crafts self-employed.

4.4 Additional Specifications

To check the robustness of our results, as mentioned in Section 3.1, we repeat the multinomial logit regression with an artisan sample (see Table A4 in the Appendix). The results are mostly similar to those in Section 4.1, with two exceptions. There is no longer a statistically significant positive openness effect for self-employed craftsmen, although the coefficient is still positive. Instead, we find a higher openness for employed artisans. Moreover, while the agreeableness coefficient is still negative in the case of self-employed artisans, but it is no longer statistically significant. Taking into account the significantly reduced number of observations, we therefore conclude that these results are consistent with the above findings.

We also run the random effects panel regression of Section 4.2 with the artisan sub-sample (Table A5 in the Appendix). Again, we find similar results as before, with two notable exceptions in the case of solo self-employed. In contrast to the total sample of craftspeople, openness and LOC have a positive effect on the decision to become a solo self-employed. This seems plausible given the artistic and creative nature of artisans. Other results show the same coefficient signs, for example in the case of agreeableness, but are not statistically significant. Again, the smaller sample size must be taken into account.

Finally, the sensitivity of the above results to the influence of the control variables is analyzed. Table A6 in the Appendix shows the multinomial logit regression for the total sample from Section 4.1 without controls. The positive effect of conscientiousness is again evident for self-employment in the crafts sector. At the same time, the positive effects of openness and risk and the negative influence of agreeableness are confirmed. Extraversion, emotional stability and LOC are no longer significant, although the signs of the coefficients are in the expected direction. With regard to the crafts-specific effects (especially conscientiousness), our results from Section 4.1 are thus confirmed. The same applies to the random effects regression from Section 4.2. The main results are confirmed in the no-control specification (Table A7 in the Appendix): In the case of solo self-employed, belonging to the crafts sector has a positive conscientiousness effect on self-employment, while in the case of agreeableness it has a negative effect.