How life context affects entrepreneurs' passion and performance

Passion is as diverse and reciprocal a phenomenon as entrepreneurship. In contrast to static conceptions of passion as a personality trait, such as passion for work (Baum and Locke 2004), the dynamic approaches proposed by Cardon et al. (2013) and Vallerand et al. (2003) dominate the entrepreneurship literature (Schulte-Holthaus 2019). EP is defined as "consciously accessible, intense positive feelings experienced by engagement in entrepreneurial activities associated with roles that are meaningful and salient to the self-identity of the entrepreneur" (Cardon et al. 2009: 515). At the center of entrepreneurial engagement are the domain activities of founding, inventing, and developing. In contrast to Cardon et al. (2013), Vallerand et al. (2003) described passion for any conceivable activity or domain while focusing on harmonious and obsessive passion depending on whether a passion has been internalized autonomously or in a controlled manner. By using both approaches—separately and in combination with each other—scholars have empirically explored and explained a wide variety of personal (e.g., personality, education, effort, and status) and contextual (e.g., emotional support, organizational climate, location, and culture) antecedents of passion, as well as its effects on attitudes and behaviors, which, in turn, affect outcomes such as venture creation, funding, or performance (Newman et al. 2021). Less attention has been paid to the life context in which passion and entrepreneurship are embedded, although references to context are inherent in both theories. In EP theory, identity is a dynamic and fluid development process determined by interaction between an individual and its proximal and distal contexts (Radu-Lefebvre et al. 2021). Delineating and balancing different micro identities can generate synergies and conflicts that shape entrepreneurial experiences (Shepherd and Haynie 2009). In Vallerand et al.'s (2003) dualistic model, context affects the internalization of a passion, which can be induced by environmental pressures, causing passion to become obsessive and to conflict with other life aspects (Curran et al. 2015). Thus, passion is inextricably linked to the influence of the environment.

However, the conceptualization of context is problematic in entrepreneurship research in two ways. First, there are multiple researched and unobserved contextual variables that determine human and entrepreneurial development. Second, scholars use different means and frameworks to assess context (Zahra et al. 2014). For instance, Gorgievski and Stephan (2016) distinguished between the immediate (i.e., the work and private environment) and the broader context (i.e., the community, industry, and cultural environment). Meoli et al. (2020) considered social influence as the immediate setting and organizational and environmental influences as the wider setting. How narrowly or broadly the context should be described is a matter of critical debate (Kristof-Brown et al. 2002). For example, a focus on distinct environmental variables is less suitable for explaining overall context effects (as is currently the case in EP research), while a very broad contextualization has limited explanatory power. To address this issue pragmatically, we used Zahra et al.'s (2014: 481) "contextualization in the broadest sense of the term, placing our researched enterprises within their natural settings to understand their origins, forms, functioning and diverse outcomes." As passion is connected to persons and their life domain outcomes, such as work, leisure, family, and partnership (Stenseng 2008), we equated this natural setting with the life context in which entrepreneurs are embedded and operate. Unlike van Veldhoven and Beijer (2012), who looked at the private life only, we extended the life context to the vocational environment in which passion-driven entrepreneurship occurs. Following Salmela-Aro and Nurmi (1997), we defined the life context of enterprising individuals as the interrelated constitution of all personal and environmental factors that are subjectively significant in shaping their courses of action.

To explain the effects of the life context on EP and performance, we used the theory of P–E fit, defined by Caplan (1987: 248) as the degree "of how well the characteristics of the person and the environment fit one another." Given that the theory of P–E fit is a ‘pure process theory’ (Edwards et al. 1998), the contents of P–E interactions can be drawn from theories of the phenomena under consideration, and thus to passion in entrepreneurship. In a wide variety of settings, scholars have empirically demonstrated that P–E fit leads to positive results, such as engagement, well-being, and performance (Edwards and Shipp 2007). Fit (or misfit) can occur at the objective and subjective levels (Edwards et al. 1998). Objective fit defines the actual configuration between a person and an environment (i.e., independent of personal bias), which can be improved by coping (i.e., people change either themselves or their environments). Subjective P–E fit is determined by a person's self-assessment and perception of an environment. In management sciences, subjective fit is a widely used approach to explore and measure the consequences of perceived compatibility between people and their environments in the work context (Vleugels et al. 2022). In line with prior research, we focused on subjective fit to evaluate life context effects because one's life (just like passion) is subjectively experienced and perceived. The phenomenology of P–E fit is a theoretical black box whose mechanisms “lie within the subjective realm and therefore require data from a single source, that is, the person whose perceptions are under investigation (Edwards et al. 2006: 808).

Subjective P–E fit is determined by demands and abilities and needs and supplies. The environment makes demands on persons, which the latter meet using their abilities; this match is called the demands–abilities fit. At the same time, an individual has physical and psychological needs that can be met by the supplies of the environment; this balance is called the needs–supplies fit. As P–E interactions are complex, any effect may always have causes other than the focus under consideration. Therefore, we approached LCF as a "comprehensive notion that necessarily includes one's compatibility with multiple systems in the work environment" (Kristof-Brown et al. 2002: 985). Moreover, we assume that the demands–abilities and needs–supplies fits have a common contextual core that can be analyzed at the global, domain, and facet levels (Edwards and Shipp 2007), which we consider in the next section.

With the decision to take the path of self-employment and business venturing, a person's life is catapulted into an entrepreneurial environment. At the global level of the demands–abilities fit, this change process has always been accompanied by the need to adapt one's life, ensure survival, develop one's business, build teams, and foster creativity and innovation (Timmons 1978). The successful management of these demands depends on a person's industrial, managerial, and entrepreneurial skills and experiences (Davidsson 1991). At the domain level, the demands–abilities fit is characterized by passion, which may originate in the hobby and leisure contexts and does not necessarily need to involve an entrepreneurial domain. However, when people regularly and persistently pursue a passionate activity, they organize their lives on the basis of that passion (Curran et al. 2015). Thus, the environmental shift may require people to adapt their passions and life contexts to the economic and organizational demands of the new setting. In short, "Building a business is a way of life" (Timmons 1978: 12). Passion and entrepreneurship can conflict, especially when ideational and economic imperatives drift apart (Schulte-Holthaus 2019). Schellenberg and Bailis (2015) showed that two passions can also complement each other and create new potentials for development. In sum, passion can become weaker or stronger, expand to new areas, or even be replaced by another passion (Vallerand 2015). At the facet level, the EP domains entail the core activities of founding the business, building the organization, and inventing products and services (Cardon et al. 2013). In addition to the extent to which individual capabilities are available to manage these demands, EP development is shaped by a person's learning orientation, which can be defined as an "individual's predisposition to develop competence by acquiring new skills and mastering new situations" (Türk et al. 2020: 227). Stenholm and Nielsen (2019) found evidence that EP is driven by task-related competences interacting with the social environment. Acquired abilities lead to better mastery of core entrepreneurial activities echoed by positive feelings. The more competence is acquired, and the more familiar the requirements become, the more central the entrepreneurial roles will be to a person's identity (Hoang and Gimeno 2010).

Subjective P–E fit is characterized by a perceived match between a person's needs, motives, and goals and an environment's supplies (Edwards et al. 1998). When considering the needs–supplies fit, Ryan and Deci's (2000) self-determination theory is often used to integrate people's basic psychological needs for autonomy, relatedness, and competence and to explain optimal human functioning. The relationship between this fit and positive outcomes is based on the reasoning that a person's environment provides supplies for the fulfillment of needs when the fit is present (Greguras and Diefendorff 2009). Similarly, Vallerand et al.'s (2003) dualistic model is grounded in self-determination and the pursuit of passionate activities based on the need to feel personal initiative (autonomy), to interact effectively with the environment (competence), and to feel connected to others (relatedness).

At the global level of the needs–supplies fit, entrepreneurial work "is uniquely supportive of individuals' basic psychological needs as it allows them to organize their self-motivated behaviors at work, leading to higher levels of agency, competence, and relatedness" (Shir et al. 2019: 1). Regarding the pursuit of passion, the entrepreneurial environment enables transforming wage labor into a meaningful occupation. At the domain level, needs and supplies are determined by unconventional entrepreneurship, which offers the opportunity to live a nonentrepreneurial passion in the industry and milieu of one's choice and to design a suitable life characterized by settings and communities that share and support this passion. The adaptation of passion to the entrepreneurial context is determined by contact with reality and the accuracy of self-assessment (Edwards et al. 1998). This is evident in, for example, successful musician-entrepreneurs, who identified their passion for music clearly and early in life, choosing the entrepreneurial path and making it the center of their existence (Schulte-Holthaus and Kuckertz 2020). The adaptation process is characterized by causation and effectuation logics—that is, by long-term goal-oriented and adaptive situation-related action and decision-making (Rapp 2022). Living out one's passion requires permanent reflection and adaptability to recognize and proactively address conflicts and potentials in both private and professional environments. At the facet level, the entrepreneurial context provides opportunities for need fulfillment through successful mastery of the domain activities of founding, developing, and inventing. The concept of EP draws on Carver und Scheier's (1998) self-regulation theory to explain how passion coordinates cognition and goal-directed behavior to realize desires (Cardon et al. 2009). Put simply, living one's passion and making a living motivate people to make it work. Therefore, mastering entrepreneurial activities involves demands and needs at the same time. De Mol et al. (2018) showed that individuals' job fit perceptions in entrepreneurial settings shape passion development. When founding, developing, and inventing activities succeed over time, venturing activities are accompanied by positive emotions and a gain in identity centrality, which enables passion to increase. "Passions maintain, renew, and continually create emotions that structure a person's social and economic world" (Guercini and Cova 2018: 387). Entrepreneurship offers the necessary resources, such as people and communities, collaboration and exchange, and jobs and projects, that support the accomplishment of activities and the nurturing of EP. Following this line of evidence, we developed the following hypothesis:

Using a sample of full-time employees, Greguras and Diefendorff (2009) provided empirical evidence that basic needs affect the P–E fit, behavior, and performance. Even in difficult environments, individuals pursue the preservation of their entrepreneurial existence and persistently push for business success (Meyer 2022). Interaction with communities and stakeholders fosters co-creation, helps allocate resources, and allows entrepreneurial opportunities to emerge (Pinelli et al. 2022). Accordingly, the better individuals have reconciled their passions with the demands of the life context (i.e., private and entrepreneurial life), the better EP can unfold and become translated into performance. "These relationships imply a cyclical recursive model in which P-E misfit at time 1 affects coping and defense at time 2, which affect P-E fit at time 3, and so on" (Edwards et al 1998: 53). Misfit induces stress and deviations from normal human functioning (Caplan 1987). Therefore, the consistent organization of the life context driven by passion becomes the basis for unleashing passion and performance. When people perceive being supported by their environments, they are much likelier to succeed in pursuing and achieving their interests and goals (Lent et al. 2000). Accordingly, we expected LCF to have a direct positive impact on performance:

EP initiates goal-directed entrepreneurial action. Cardon et al. (2009: 518) described the EP experience "as a complex pattern of psychological, brain, and body responses activated and maintained by an entrepreneur's passion that, when regulated, aid in motivating coherent and coordinated goal pursuit." Accompanied by intense positive feelings, EP stimulates the entrepreneur to invest time and energy into the business, to overcome obstacles, and to achieve desired goals. In their literature review, Newman et al. (2021) summarized the empirical evidence regarding positive EP effects on entrepreneurial attitudes, behaviors, and outcomes. EP drives performance, which, in turn, feeds back into passion (Lex et al. 2020). However, entrepreneurial performance can manifest itself in many ways (Shepherd et al. 2019). At the individual level, performance may include the accomplished step to self-employment, the start of a new enterprise, or personal financial rewards. In addition, other factors, such as workplace relationships, community impact, and personal fulfillment, determine the perception of individual entrepreneurial success (Wach et al. 2016). At the firm level, survival and growth—often measured using the dimensions of sales, employees, and assets––are commonly employed as indicators in management and organizational research (Hamann et al. 2013). Such economic indicators are part of numerous reports on the output of the cultural and creative industries (e.g., Lhermitte et al. 2015). Therefore, we consider an individual's passion-driven entrepreneurial efforts in a creative domain as factors that determine the value and growth of the entrepreneurial endeavor at the individual level, resulting in the founding of one or more businesses, the hiring of employees, and the generation of sales. Regarding key outcome measures in entrepreneurship research, EP has been proven to positively affect business performance (Mueller et al. 2017), firm survival (Stenholm and Renko 2016), venture growth (Drnovsek et al. 2016), and subjective owner performance (Lex et al. 2020). Although the EP for inventing and developing exhibits stronger correlations with performance than the EP for founding, the underlying mechanisms are diverse, and EP effects on performance, instead of revealing a uniform picture, may demonstrate clear positive tendencies. Consistent with prior research, we developed the following hypothesis:

The data for the analysis came from a sample of entrepreneurial individuals in German cultural and creative industries. Passion is a crucial issue among creative professionals, who often develop a passion for a creative, artistic, or cultural domain, which paves the way for freelance work, self-employment, and small business entrepreneurship and shapes their life context (Schulte-Holthaus 2018). With over 258,000 businesses, the submarkets of the culture and creative industries contribute 7.87% of Germany's entire economic turnover (Federal Ministry for Economic Affairs and Energy 2020). These submarkets include music, books, art, film, broadcasting, performing arts, design, architecture, press, advertising, and software/games. As there is no central register that would enable the identification of every entrepreneur in these industries in Germany, we decided to reach out to entrepreneurs in these submarkets by collaborating with nationwide creative industry associations listed on the web pages of the German Ministry of Economics. Cooperating associations forwarded our survey to their members via online and internal newsletters and called for participation. Although this procedure did not allow us to calculate a response rate for our survey, the manifold associations involved could be considered to represent the German creative industry in its entirety.

Data from 25 associations covering all 11 submarkets were collected. After removing incomplete questionnaires, non-entrepreneurial workers (i.e., those who were merely employed), and suspicious response patterns, the final sample consisted of 406 entrepreneurs. They were between 24 and 85 years of age (M = 51.4, SD = 11.4), with 42% being male and 58% female. Of the participants, 3% had finished lower-secondary or middle school, 20% had finished high school, 70% had finished university, and 5% held a doctoral degree, consistent with the national creative industries' socio-demographics (Weißmann and Liersch 2021). As freelancers, self-employed persons, or small and medium-sized business owners, the participants had been entrepreneurially active for up to 60 years (M = 19.6, SD = 13.1). On average, the participants owned 1.7 companies (SD = 2.1) and had 2.2 employees (SD = 7.8). Typologizing the participants according to the number of employees, 93% could be described as micro (0–9 employees), 7% as small (10–49 employees), and two as medium-sized entrepreneurs (50–250 employees), which corresponds to the actual distribution in the German creative industries (Muller et al. 2018).

Data were collected using a questionnaire with LCF as the independent variable, the three EP domains as mediating variables, and performance as the dependent variable. In designing the questionnaire, we followed the procedural strategies of Jordan and Troth (2019) to minimize common method bias. The survey was anonymous to reduce evaluation apprehension—at the expense of collecting dependent and independent variables from different sources or at different times—because the questions concerned highly personal aspects of life. Some associations had agreed to participate only on the condition of anonymity. Various other strategies, such as informing and instructing participants, separating predictor and criterion variables, counterbalancing questions, improving scale items, and using different question formats and scale anchors, were implemented. Based on established and validated scales, a translation into German and a back-translation into English were performed by two qualified bilinguals––one German and one English native speaker––who were not involved in the study (Brislin 1970). Meaning discrepancies were minor and were resolved by the lead author in agreement with the translators, which improved translation accuracy. Follow-up survey pretests did not indicate any semantic ambiguities. In addition, we followed the recommendations of Jordan and Troth (2019) to perform a statistical test for detecting method variance after completing the data collection. Therefore, we opted for the straightforward solution by Kock (2015), who developed a full collinearity assessment approach to evaluate common method bias in PLS modeling. The structural model assessment resulting from a complete collinearity test in SmartPLS3 showed that the inner value inflation factors did not exceed 1.97 and were thus far below the critical threshold of 3.3 as outlined by Kock (2015), indicating that common method bias was not a problem.

LCF. Understood as an impact indicator of the life context in which a passion is embedded, subjective P–E fit was evaluated by means of a personal project analysis, a multi-modular approach that has been used in personality research for more than four decades to investigate personal projects and their content, dynamics, appraisal, and impact (Little and Balsari-Palsule 2021). The concept is based on the concept of psychological specialization and implies that individuals' progressive adaptation to their environments is reflected in the selection and pursuit of personal projects, which are defined as "extended sets of personally salient action in context" (Little 2007: 25). Personal projects are temporally and spatially extended sequences that can span days, months, or even years and are a meaningful part of an individual's life. Data collection was performed using the project elicitation module and the cross-impact matrix as outlined by Little and Coulombe (2015) to query project content and evaluate the interrelationships (fit/misfit) between the projects that people pursue simultaneously in a specific context at a certain point in time. In the questionnaire, after a detailed introduction, the participants were first asked to think about and list five personal projects. In the second step, they evaluated the mutual influences (fit/misfit) of the projects using a 5 × 5 cross-impact matrix—that is, they were asked to decide whether each project had a very positive (+ 2), positive (+ 1), neutral (0), negative (− 1), or very negative (− 2) impact on the other projects. As projects cannot influence themselves, the matrix diagonal was not assessed, as illustrated in Table 1. The values of the 20 relations were summed up, meaning that the LCF measure had a maximum and minimum value of + / − 40.

EP. We measured EP using the 13-item scale developed by Cardon et al. (2013); the items and subscales proved to be reliable and internally consistent (α = 0.85, 0.72, and 0.77). The scale captured two EP dimensions, intense positive feelings (IPF) and identity centrality (IC), across the domains of passion for inventing (EP_inv), passion for founding (EP_fnd), and passion for developing (EP_dev). All items were assessed using a symmetric and equidistant 5-point scale to approximate interval-level measurement in structural equation modeling (Hair et al. 2016). The scale ranged from 1 (strongly disagree) to 5 (strongly agree) to query items such as "I really like finding the right people to market my product/service to" in order to capture the range of intense positive feelings (IPF_dev1) towards EP for developing and "Nurturing and growing companies is an important part of who I am" to measure the identity centrality (IC_dev1) of EP for developing (for the full scale, see Cardon et al. 2013: 394). We retained the labeling and numbering of the items and followed the authors' recommendations to assess the values for the different domains separately and not calculate an "overall average-across-all-domains" value of EP because domain levels can differ noticeably from one another.

Entrepreneurial performance We opted for a subjective operationalization of performance (PRF). First, subjective measurements can be expected to correlate with their objective equivalents (Dess and Robinson 1984). Second, we avoided directly gathering sensitive financial and economic data, which could have been viewed negatively by creative professionals. To capture individual performance, we used the common management dimensions of sales growth, asset growth, and employee growth (Hamann et al. 2013). These dimensions allowed us to conclusively capture performance at the individual level, independent of the respondents' entrepreneurial status (i.e., freelancer, self-employed, hybrid, or entrepreneur). Following Gupta and Govindarajan (1984), we measured the PRF of each dimension via the respondents' fulfillment and the personal importance attached to the dimension, a procedure using which researchers have achieved high inter-item reliability for their chosen dimensions (Covin and Slevin 1989; Lee et al. 2019). The participants used a 5-point Likert scale ranging from 1 (not at all important) to 5 (very important) to indicate how important each dimension was for making entrepreneurial decisions and then rated how satisfied they were with the development of each dimension compared to their competitors, ranging from 1 (not at all satisfied) to 5 (very satisfied). The three values for the importance of sales growth, asset growth, and employee growth were set in relation to one another and normalized to a total value of 1, which minimized the bias of intersubjective assessment. Percentages were then multiplied using the associated fulfillment indices, resulting in weighted performance indices for sales growth, asset growth, and employee growth, which were added to an overall PRF index.

The hypothesized relationships suggested that structural equation modeling was an adequate analytical technique to shed light on our research question. It allowed us to estimate and test the correlative relationships between latent independent and dependent variables and their underlying structures that emerged from the theoretical framework. More specifically, we conducted a PLS analysis. First, this approach uses proxies as approximations of the latent constructs, which are composed of weighted indicators and typically include measurement errors. Integrated ordinary least squares regressions minimize these errors in the target constructs and maximize their explained variance. Second, PLS modeling does not involve any distributional assumptions about the population; moreover, it is robust against inadequacies such as skewness, multicollinearity, and model misspecification (Cassel et al. 1999). To achieve a statistical power of 80% with four independent variables acting on one dependent variable at a significance level of 1% and a minimum R² of 0.1, 158 are cases required (Hair et al. 2016). Our dataset of 406 cases clearly exceeded this requirement, which was needed to capture the heterogeneity of the creative and cultural industries in which, among other actors, artists, culture lovers, and entrepreneurs act at all levels of the value chain. Furthermore, PLS provides a convenient way of assessing unobserved heterogeneity. Consequently, PLS structural equation modeling is particularly suitable for theory building and predicting LCF as an antecedent of EP and performance. Analyses were performed using the SmartPLS3 software package (Ringle et al. 2015).

Setting a maximum of 300 iterations and a stop criterion of 10⁻⁷, the PLS algorithm was used to converge and compute a stable solution based on which the quality criteria of the reflexive measurements were assessed (Ringle et al. 2015), as shown in Table 2. All loadings were well above the recommended threshold of 0.7, and the indicator reliabilities (IRs) were above 0.5, except for EP_dev1 and EP_inv5, which only slightly missed conventional thresholds. Given that the EP scales were well-established and validated constructs, we decided to retain both indicators due to their consistency and better comparability with other EP studies. The average variance extracted (AVE) for all constructs was well above the required minimum of 0.5 and supported the convergence validity of our reflexive measures. Values for composite reliability (CR) and Cronbach's alpha (CA) were between 0.60–0.90, confirming internal consistency reliability. Discriminant validity was supported as well. Indicator loadings on their associated constructs were higher than all cross-loadings. An examination of Fornell and Larcker's criterion (1981) showed that all latent constructs shared, on average, a higher proportion of variance with the respective indicators than with any other latent construct. Finally, the heterotrait–monotrait analyses by Henseler et al. (2015) supported discriminant validity by exhibiting heterotrait–monotrait correlation ratios below 0.85. These results indicate the validity and reliability of the measurement model.

Next, we assessed the structural model. The values of the inner variance inflation factor were within the range of 0.20–5.00, indicating that collinearity between the constructs was not a problem. Then, we conducted PLS bootstrapping using the settings outlined above. Surprisingly, the structural model relations were completely contrary to our expectations. The paths from LCF to the EP domains exhibited p-values at the significance level of 0.01, but their R² values of ≤ 0.054 were marginal. Correlation coefficients of ≤ 0.232 must be considered negligible. Effect sizes f² of LCF on EP_inv (0.06), EP_fnd (0.05), and EP_dev (0.02) indicated only tiny effects. These results did not provide support for Hypotheses 1a, 1b, and 1c. Moreover, the paths from LCF and EP to PRF exhibited correlation coefficients below 0.05 with insignificant p-values. The total effects f² of LCF and all EP domains on PRF were 0, meaning there was no evidence for Hypotheses H2 and H3a, H3b, and H3c. Therefore, we subsequently examined the data for unobserved heterogeneity, which may have caused different effects in the subsegments to be canceled out in the overall sample. Unobserved heterogeneity is a possible source of endogeneity that may arise from omitted variables that induce latent effects on model estimates (Hult et al. 2018). As we could not rule out potential endogeneity, we addressed this issue by checking for unobserved heterogeneity, a major cause of endogeneity.

We followed the two-step methodological approach to PLS modeling outlined by Hair et al. (2017) to segment the data. First, the finite mixture algorithm in SmartPLS3 was used to detect unobserved heterogeneity in structural relationships and determine the number of segments (Sarstedt et al. 2011). Second, as the finite mixture approach cannot account for unobserved heterogeneity in measurement models, prediction-oriented segmentation was subsequently applied to assign observations to the number of segments (Becker et al. 2013). As our data did not contain any missing values in the model variables, no further observations needed to be deleted. Four dependent variables, a significance level of p = 0.01 and a minimum R² of 0.25 require at least 58 cases per segment to achieve a statistical power of 80% (Hair et al. 2016). For a total of 406 observations, this allowed for a maximum of seven segments. As shown in Table 3, Akaike's information criterion (AIC) and the consistent AIC (CAIC) resulted in a clear number. The comparison of the modified AIC with factor 3 (AIC3) and factor 4 (AIC4) and the Bayesian information criterion (BIC) indicated four to six segments, which all display a normed entropy (EN) value of > 0.5. Hair et al. (2017) advocated choosing fewer segments than what the AIC indicates, which left us with four to five possible segments. However, only the four-part solution met the minimum size of 58 cases for all segments; thus, it was the preferable solution.

Afterwards, prediction-oriented segmentation was performed to divide the observations into four groups. To avoid the algorithm converging on a local optimum, the result with the highest value in the change of objective criterion (up to 1.36) was chosen after running 10 repetitions. The assignments generated group sizes of 91 (22%), 90 (22%), 142 (35%), and 83 (20%) observations.

The evaluation of the measurement model was analogous to the description in 4.1. All outer loadings were greater than 0.7, except for IPF_dev1, IPF_inv1, and IC_inv1, which fell slightly below. Only in Segment 2 did the value for IPF_dev1 show a critical indicator reliability of 0.31. CR and CA were well above 0.7; the Fornell-Larcker criterion and cross-loadings indicated unrestricted construct validity and reliability across segments. The heterotrait–monotrait ratio in Segment 3 had a critical index of 0.97 for EP_fnd and EP_dev. Occasionally, ratios approach the value of 0.9. However, this does not indicate a lack of discriminant validity when constructs are conceptually similar (Henseler et al. 2015), which was the case with the EP domains. Accordingly, the evaluation supported the overall validity and reliability of the measurements in the four segments.

Prior to the multigroup analysis of the partitioned models, the measurement invariance of the composite models was assessed to ensure that variances in the segments did not arise from meaning differences in the latent variables. Following Henseler et al. (2016), we first examined the criteria for configural invariance, which showed that all measurements, coding, and calculations were performed consistently. Next, we ran permutation tests in SmartPLS3 (1,000 permutations, two-tailed, p = 0.05) and checked for significant differences in composite scores. With four segments, this involved analyzing the scores in six pairwise comparisons. For the combination of Segments 1 and 2, we found a minor significant difference in the composite scores for EP_dev. For all other combinations, compositional invariance was established. Thus, partial measurement invariance as a precondition for the comparison of the correlation coefficients in the four segments was confirmed. In addition to common method bias and unobserved heterogeneity, measurement model variance was excluded as a source of endogeneity.

We then evaluated the segments' structural models. The inner value inflation factors of the models were all below the threshold of 5, indicating that collinearity among predictor constructs was not a problem. All parameter estimates of the four structural models are presented in Figs. 2 and 3. With correlation coefficients between 0.54 and 0.82, an explained adjusted variance of 0.29 and 0.67, and effect sizes of > 0.35, Segments 3 and 4 indicated a strong positive effect of LCF on EP, which partially supported Hypotheses H1a, H1b, and H1c. Segments 3 and 4 consisted of 225 cases and represented a larger portion of the sample. However, the hypotheses were not supported in Segments 1 and 2. In Segment 2, LCF had a slight negative effect on EP, which indicated only weak effects (f² = 0.02–0.08, R²_adj < 0.09), whereas Segment 1 indicated LCF's strong adverse impact on EP (r >  − 0.71; R²_adj > 0.5; f² > 1.01). The positive effect of LCF on PRF (H2) was not supported in any of the segments. In Segments 3 and 4, the correlations (r ≙ − 0.4) and effect sizes (f² = 0.09) showed a minor negative effect. The negative impact was even stronger in Segment 1 (r =  − 0.77; f² = 0.25).

The empirical results regarding Hypothesis H3 showed that the three EP domains do not comprehensively translate positively into PRF but rather involve diverse combinations. On the one hand, EP_inv can have a moderate positive (Segment 2: r = 0.48; f² = 0.25) or a strong positive effect on PRF (Segment 3: r = 0.71; f² = 0.53). On the other hand, EP_inv can have a small negative impact (Segment 4: r =  − 0.27; f² = 0.10) or even reduce performance (Segment 1: r =  − 0.89; f² = 0.63). This was also the case for EP_dev, which exhibited a substantial positive effect in Segment 4 (r = 0.86; f² = 0.52) and a moderate negative effect in Segment 2 (r =  − 0.58; f² = 0.36). Accordingly, Hypotheses H3a and H3c were partially supported. Evidence for the assumption that EP_fnd drives performance was lacking. Three segments demonstrated no significant relations, and Segment 3 even indicated a strong negative impact (r =  − 0.78; f² = 0.28), meaning that H3b was falsified.

Table 4 summarizes the structural relationships. In Segments 3 and 4, LCF acted positively on all three EP domains as theoretically predicted, but not all EP domains translated positively into performance. In Segment 4, PRF stemmed from EP_dev and in Segment 3 from an EP_inv. The adjusted explained variance in Segments 3 and 4 (0.46 and 0.47) was significantly larger than in Segments 1 and 2 (0.27 and 0.39). Therefore, Segments 3 and 4 accounted for performance more comprehensively. Segments 1 and 2 were entirely characterized by negative life-context effects, which meant that the EP domains could only develop, at most, a weak positive effect on performance (Segment 2: r =  − 0.48; f² = 0.25). It is interesting to note that in no segment did more than one EP domain positively drive PRF and that all segments contained exactly one domain that hindered performance.

To understand the segments in greater detail, we applied multinomial logistic regression to see whether the segments could be explained by socio-demographic (age, gender, and education) and entrepreneurial data (entrepreneurial years, firms, employees, and employment hours). Such variables are commonly used as controls in passion and performance studies that employ structural equation modeling (e.g., Ho and Pollack 2014; Stenholm and Renko 2016; Feng and Chen 2020). The results revealed only one significant difference between Segments 3 and 4. Higher education (β = 0.380; p = 0.03) and more employment hours (β = 0.019; p = 0.04) increased the likelihood of respondents being from Segment 4. Therefore, these variables do not influence and explain the segments and could also be ruled out as sources of endogeneity. Subsequently, we conducted qualitative post hoc analyses to achieve a more detailed description and a better understanding of the segments by combining inductive analysis of additional qualitative data with the search for patterns and types described in the research literature. First, we qualitatively evaluated the individual denotations of passion, which we collected in the questionnaire using the following free-text field: "Type in a keyword that most closely denotes your passion." Following Gioia et al. (2012), we grouped the entries into 31 superordinate themes and aggregated them into 10 dimensions. Frequencies and percentages in the segments were calculated to determine which passion dimensions emerged in each segment. The results can be found in Table 7 in the Appendix. Second, following Little (1983), we assigned the respondents' free-text personal projects into academic (school- and university-related), occupational (job-related), health/body (appearance-, health-, or fitness-related), interpersonal (related to other persons), intrapersonal (self-related), leisure (recreation-related), and maintenance (livelihood- and administration-related) projects. Table 5 displays the percentages of project categories in the segments, enabling a comparison of the project contents. Third, because LCF was the outcome of P–E interactions, which may be co-determined by the QDEP, as outlined by Cardon et al. (2017), we drew on Vallerand's (2015) passion criteria to establish an indicator of for each of these six QDEP, for which currently no validated scales exist.

For example, we paraphrased passion for people as being "passionate about working with family, satisfying customers, and building meaningful relationships with employees, vendors, or affiliates" (Cardon et al. 2017: 29) and treated it as an activity for assessing the passion criteria—for example, "This activity is a passion for me" (Vallerand 2015: 72). We examined the passion criteria analogously to the individually denoted passion using the same scale anchors as for measuring EP. The average scores for the six QDEP, the three EP domains, and the individually denoted passion criteria are contrasted in Table 6.

Bonneville-Roussy et al. (2011) proposed a threshold of ≥ 4 on a 7-point scale for judging people to be passionate about an activity. This corresponds to a threshold of ≥ 3 on a 5-point scale. QDEP means of ≥ 3 suggested that all segments involved a pronounced passion for people, for the product or service, and for a social cause—that is, for domains for which no validated measures exist. The average scores of all six QDEP were ≥ 3. In Segments 1 and 2, the average QDEP scores were higher than those of EP, indicating that QDEP covers EP more comprehensively. In Segments 3 and 4, the opposite was the case: EP was more pronounced than QDEP. Accordingly, QDEP and individual passion may play an essential role in the relationship between LCF, EP and PRF. Fourth, when analyzing the qualitative data, we conducted an iterative comparison with salient patterns found in the existing cultural and creative industries literature to describe the four empirical segments. The "recursive iteration among theoretical patterns […] and observations will help understand processes and develop theories that are contextualized in time" (Bouncken et al. 2021a: 7). Based on the structural models and our post hoc analyses, the segments were assigned to four types of passionate entrepreneurs frequently discussed in the literature: artepreneurs, culturepreneurs, creative entrepreneurs, and lifestyle entrepreneurs.

Segment 1 Artepreneurs are self-employed artists who primarily pursue artistic goals and are characterized by openness to new ideas, creativity, and high emotionality (Arenius et al. 2021). The strong adverse effects of LCF on EP and PRF (up to r = − 0.89) suggested that an artistic life context inhibits passion for business. Individual passion was most frequently associated with art and design (30%), leisure and living (14%), and people and family (8%). LCF was characterized by high relative proportions of health/body (15%), interpersonal (13%), and intrapersonal (13%) projects. Engaging with the self, other people, and the world dominates the lives of artepreneurs. Accordingly, QDEP averages related to people and inventing were the most prominent. Nevertheless, the structural model showed that EP_inv reduced performance, possibly because EP_inv is highly emotion driven. Factor loadings for IPF_inv (0.777–0.864) were considerably higher than for the identity of being an inventor (0.631). The lowest QDEP average was for product/service (3.46), signaling that artepreneurs' passion for their products and services was the least entrepreneurial compared to the other segments. The peak level for individual passion (3.14) supported the conclusion that artistic passion only rudimentarily spills over into the domains of true EP.

Segment 2 Culturepreneurs are hybrid cultural and entrepreneurial agents who mediate between culture and service production in creative milieus and professional scenes (Lange 2011). LCF had only a small negative effect on EP but no direct link to PRF. Individual passion was the least related to art and design (12%) and peaked regarding creativity and culture (19%), literature and language (18%), and leisure and living (12%). Occupational (36%), leisure (14%), and maintenance (9%) projects exhibited considerably higher proportions compared to artepreneurs. Interpersonal projects (10%) had the lowest ranking. Therefore, culturepreneurs lives are much more dominated by profession, life organization, and leisure in the fields of art, culture, and entrepreneurship. The QDEP for products and services was more pronounced (3.57). Paradoxically, the QDEP for inventing (3.11) was the weakest among culturepreneurs, but EP_inv exhibited a moderate positive effect (r = 0.484) on PRF. Factor loadings showed that EP_inv was strongly driven by emotions for developing new products and solutions (0.810–0.928) but not by emotions for strategically searching and commercializing market gaps (0.636). The more EP_dev was pronounced, the more it slowed down the performance of culturepreneurs (r =  − 0.579).

Segment 3 Creative entrepreneurs can be described as individuals whose innovative capacity arises from creativity, experimentation, and collaboration in knowledge and work contexts in which innovation is often not the primary goal but occurs unintentionally during creative processes (Wijngaarden et al. 2021). In addition to creativity and culture (17%) and language and literature (15%), the dimensions of leisure and living (15%), curiosity and learning (10%), and engagement and flow (9%) exhibited the highest percentages. Life was most dominated by occupational (39%), health/body (14%), and maintenance (11%) projects. Direct comparison with artepreneurs indicated that creative entrepreneurs are noticeably less concerned with inter- and intrapersonal and health/body projects. Their life context enables the most pronounced development of EP (3.45). Interestingly, all other values for QDEP and passion criteria were in the middle range and did not involve minimum and maximum levels. Performance stemmed from EP_inv, although the loadings on inventor identity were low (0.634). Emotion-driven creative venturing appeared to be the basis of unintended innovation. On the contrary, the founder identity exhibited a much higher loading (0.876), but EP_fnd slowed down PRF almost to the same extent that EP_inv promoted it.

Segment 4 Lifestyle entrepreneurs are individuals who operate in a hobby or leisure context by balancing personal and professional life aspects such as autonomy, workload, life quality, and profit within a distinct social community and its habitus (Wallis et al. 2020). The substantial positive impact of LCF on all three EP domains (r = 0.589–0.806) supported this attribution. Likewise, the life context had a negligible adverse effect on PRF as lifestyle entrepreneurs often pursue profit and growth only as long as the latter promote lifestyle and life quality. Their individual passions exhibited the second-highest average (3.09), which suggests the strength of overriding individual passions. However, the project categories were broadly distributed so that leisure and living (11%) did not exhibit an elevated percentage. The comparatively highest shares were related to music and instruments (12%), spirituality and nature (8%), and business and challenge (6%). Lifestyle entrepreneurs exhibited the most pronounced QDEP for product or service (3.74) and for growth (2.74). The strong effect of EP_dev on PRF (r = 0.862) showed that developing one's life entrepreneurially is a crucial driver of business performance. At the same time, EP_inv had a minor negative effect on PRF (r =  − 0.274), and its IPF loadings (0.780–0.785) were much higher than for inventor identity centrality (0.622). Therefore, pursuing innovation emotionally and passionately tends to result in failure. Lifestyle entrepreneurs see themselves as founders, as the founder identity had the highest loading (0.826). However, their performance stemmed from EP_dev. Lifestyle-determined contexts empower the development of EP, which most effectively translates into business performance.

Our study makes three contributions to passion research in entrepreneurial contexts. First, we used P–E theory to explain how LCF affects EP and PRF. Our notion of LCF complements the entrepreneurship and psychology literatures on passion. The empirical results support LCF as an antecedent of EP, although Hypothesis H1 was only partially supported due to positive effect of LCF on PRF. Negative outcomes in Segments 1 and 2 show that the higher the LCF, the weaker the EP domains. In these segments, the development of a pure EP according to Cardon et al. (2013) was not compatible with the proximal life context. Moreover, the categories of artepreneurs and culturepreneurs allow us to explain the negative LCF-EP relation more closely. When EP domains are not incorporated into the organization of life, EP cannot unfold. In contrast, for the creative and lifestyle entrepreneurs (Segments 3 and 4), the life context had a full positive effect on EP—that is, the life context was suitably organized and substantially contributed to the partial unfolding of the classic EP domains. For both entrepreneur types, the positive effects of EP_inv and EP_dev were significantly stronger and explained a greater variance in PRF, which supports the assumption that LCF enables EP and drives performance. Surprisingly, the direct effect of LCF on PRF was negative for all segments except for the culturepreneurs. It is possible that the fit measured using personal project analysis involved aspects whose subjective importance we do not know but which hamper performance. This would be the case, for example, if the interaction between family and entrepreneurship were positively assessed but the family had a higher priority than entrepreneurial ambitions (Davis and Shaver 2012).

Second, the effect of EP domains on performance was not uniformly positive. The results showed that EP_inv and EP_dev can promote performance as well as reduce it. This is intriguing because extant research has almost unanimously demonstrated a positive EP impact on performance. One exception was Adomako and colleagues, who also found a combination of positive effects by EP_inv and EP_dev and negative effects by EP_fnd on firm growth (Adomako et al. 2018; Adomako und Ahsan 2022). Nevertheless, the dysfunctional outcomes of passion remain largely unexplored (de Mol et al. 2020). Our study contributes to this gap by providing empirical evidence and an explanation of how different domains can generate such adverse outcomes. In the creative economy, passion often focuses on creative activities, which are hardly represented by the EP domains. Therefore, different QDEP weightings can result in conflicts between creative, life, and economic imperatives and hamper performance (Schulte-Holthaus and Kuckertz 2020). Artepreneurs illustrate this phenomenon most clearly. The presence of a highly artistic passion that includes trying out new ideas and opportunities in an effective entrepreneurial manner clashes with passionate ambitions for business development. Thus, our findings contribute to the exploration of the diversity of passions and provide new evidence that "entrepreneurs likely have multiple, rather than singular, sources of passion" (Cardon et al. 2017: 31).

Third, our study shows both theoretically and empirically how the life context affects entrepreneurial endeavors and contributes to the literature that considers context a necessity for the advancement of entrepreneurship theory (Welter and Baker 2021). LCF is the result of a history of shifts and decisions in an individual's life that are determined by personality and the resources and restrictions of the environment. Using personal project analysis to identify the impact of P–E interactions bridges the dichotomy between passion as a static personality trait and a dynamic phenomenon. We provide a heuristic approach that enables approaching the life context as one independent variable. Thus, our study design represents a fruitful way of contextualizing the initiation, engagement, and performance of entrepreneurial endeavors (Shepherd et al. 2019). The interpretation of the results demonstrates that "context becomes part of the story being told" (Zahra et al. 2014: 494). Without classifying and describing artepreneurs, culturepreneurs, creative entrepreneurs, and lifestyle entrepreneurs, the unanticipated negative effects could not have been conclusively explained.

Passion and entrepreneurship are life-shaping characteristics that need to be aligned with one's private and vocational life so as not to cancel out each other. The notion of LCF enables assessing and thus achieving a better understanding of the individual constellation and the power of the life context. Established entrepreneurs can use our approach to reflect on, unleash, and integrate passion into their entrepreneurial life ambitions. People with a nonentrepreneurial passion may identify and anticipate the potentials and hurdles of business venturing. Living in a pluralized, functionally differentiated society means that people today have many options for choosing and shaping their leisure time and careers; among these options, the combination of passion and entrepreneurship is promising. In academia, interventions using personal project analysis already support students in finding meaning and increasing engagement, well-being, and performance (Coulombe et al. 2020). Therefore, we would like to see study findings such as ours being debated in entrepreneurship education, which would constitute a practical contribution by highlighting the social benefits of the findings of entrepreneurship research in the "discussions of how to engage and empower people by giving them the skills necessary to explore careers as entrepreneurs" (Wiklund et al. 2019: 429).

We see three fundamental limitations that pave the way for future research. First, although our hypotheses were developed against the background of unconventional entrepreneurship, they were tested on a cultural and creative industries sample for which we could not assess the response rate and thus could not establish its representativeness. At the same time, we reached a great variety of unconventional entrepreneurs and elaborated on four types of passionate entrepreneurs based on the creative industries sample. Although the qualitative post hoc results were transparently and comprehensibly generated, which contributes to the internal validity of complex structures (Bouncken et al. 2021a), the generalizability of our findings remains limited. Against this background, the flexible pattern-matching approach (Sinkovics 2018; Bouncken et al. 2021b) offers a promising starting point for further research. The logic of flexible pattern matching allows to develop a theory-based explanation for the negative relations between life context, passion, and performance, to contrast it with qualitative data using a theory-driven sample, and to further elaborate the emergent theory by reconciling empirics with the research literature. Moreover, future studies in other environments and industries—particularly in high-performance environments such as the Silicon Valley—should focus on heterogeneity, similarities, and differences of the life context effects on passion and performance. We still know too little about which constellations of EP domains generate adverse outcomes in particular settings (Newman et al. 2021). Researching the "elements of the environment affecting entrepreneurship" has the potential to explain how individuals construct contexts and how these contexts shape their attitudes and behaviors (Welter and Baker 2021: 1155). Second, we argue that positive and negative interactions between personal projects may also be shaped by their relative weights. Personal project analysis offers additional modules to capture affective and cognitive dimensions of personal projects, such as the importance or progress of projects (Little and Coulombe 2015). Future studies may use the additional modules to explore the effects of passionate action in private and entrepreneurial contexts in greater detail (Laguna et al. 2016). Third, our findings were restricted to the three measurable EP domains. The attempt to examine QDEP proposed by Cardon et al. (2017) using Vallerand's passion criteria (2015) involved a non-valid scale and can only serve as a first indicator. To capture passion in entrepreneurship more comprehensively, we urgently need to develop reliable scales for QDEP to take research on passion in entrepreneurial contexts to the next level.