Journal of Business Research 68 (2015) 639–653

Contents lists available at ScienceDirect

Journal of Business Research

Disentangling the effect of prior entrepreneurial exposure onentrepreneurial intention

Florian B. Zapkau a,⁎, Christian Schwens a, Holger Steinmetz b, Rüdiger Kabst b

a University of Düsseldorf, Faculty of Business Administration and Economics, Universitätsstr. 1, 40225 Düsseldorf, Germanyb University of Paderborn, Faculty of Business Administration and Economics, Warburger Str. 100, 33098 Paderborn, Germany

⁎ Corresponding author. Tel.: +49 2118102994; fax: +E-mail addresses: florian.zapkau@hhu.de (F.B. Zapkau)

(C. Schwens), holger.steinmetz@upb.de (H. Steinmetz), ka

http://dx.doi.org/10.1016/j.jbusres.2014.08.0070148-2963/© 2014 Elsevier Inc. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 4 April 2013Received in revised form 13 August 2014Accepted 21 August 2014Available online 13 September 2014

Keywords:Prior entrepreneurial exposureRole modelsWork experienceEntrepreneurial intentionTheory of planned behavior

The present paper disentangles the effect of prior entrepreneurial exposure on entrepreneurial intention in termsof different types of exposure and their perceived quality. Drawing on the theory of planned behavior, the paperanalyzes whether attitude, subjective norm, and perceived behavioral control mediate the influence ofentrepreneurial role models and work experience in small or newly founded firms on entrepreneurial intention.Testing our hypotheses on data from374 individuals, the studyprovides differentiated support for our theoreticalpredictions. The results contribute to resolving previously inconclusive findings by offering a differentiatedunderstanding of how different types and the perceived quality of prior entrepreneurial exposure influenceindividuals' entrepreneurial intention.

© 2014 Elsevier Inc. All rights reserved.

1. Introduction

Entrepreneurial intention represents the commitment of individualsto start a new business (Krueger & Carsrud, 1993). Several scholarsemphasize the importance of entrepreneurial intentions as a first steptowards entrepreneurial behavior (i.e., starting a business) (Bird,1988; Krueger & Carsrud, 1993). In fact, prior research suggests thatintentions are the single best predictor for planned behaviors, such asstarting a business (Bagozzi, Baumgartner, & Yi, 1989; Kim & Hunter,1993). Analyzing entrepreneurial intentions is of particular importanceas new firms facilitate the transfer from innovations to marketableproducts and services, mitigate inefficiencies within an economy, andcreate new jobs (Zhao, Seibert, & Hills, 2005).

Prior entrepreneurial exposure encompasses an individual's person-al history related to entrepreneurship such as entrepreneurial parentsor prior work experience in a small or newly founded firm (Krueger,1993; Peterman & Kennedy, 2003). Previous research investigating thedirect impact of prior entrepreneurial exposure on entrepreneurialintention displays inconclusive findings (Chlosta, Patzelt, Klein, &Dormann, 2012; Shook, Priem, & McGee, 2003). Some authors findentrepreneurial parents to stimulate children's entrepreneurial inten-tion (e.g., Crant, 1996; Matthews & Moser, 1995), while others do notsupport this view (e.g., Gird & Bagraim, 2008; Kolvereid & Isaksen,2006; Tkachev & Kolvereid, 1999). Research on the influence ofwork experience in small or newly founded firms is comparatively

49 211 8114579., christian.schwens@hhu.debst@upb.de (R. Kabst).

scarce but nonetheless displays rather ambiguous findings as well(e.g., Autio, Keeley, Klofsten, Parker, & Hay, 2001; Kautonen, Luoto, &Tornikoski, 2010; Matthews & Moser, 1995).

The reasons for these inconclusive findings can be twofold: First,prior entrepreneurship literature does not sufficiently account for thefact that starting a business is intentional (Bird, 1988; Krueger &Carsrud, 1993). In this regard, models with direct predictors inade-quately reflect that the influence of exogenous variables (such as priorentrepreneurial exposure) on entrepreneurial intention occurs throughattitudinal variables (such as attitude, subjective norm, and perceivedbehavioral control in the case of Ajzen's (1991) theory of plannedbehavior). Second, differentiated views accounting for different typesof prior entrepreneurial exposure are limited. Most studies analyzethe effects of parental role models and neglect to account for othertypes of prior entrepreneurial exposure such as work experience insmall or newly foundedfirms (Matthews &Moser, 1996). This approachis problematic as both types of exposure may provide individuals withdifferent learning experiences (Chlosta et al., 2012; Fairlie & Robb,2007). Moreover, extant studies also largely neglect to account for thequalitative dimension of prior entrepreneurial exposure (Carr &Sequeira, 2007; Kim, Aldrich, &Keister, 2006). Hence, inconclusive resultsmay stem from the fact that exposure perceived as positive may differ-ently affect individuals' entrepreneurial intention compared to exposureperceived as negative (Krueger, 1993; van Auken, Fry, & Stephens, 2006).

The aim of the present paper is twofold: First, we develop anintention-based framework and investigate the impact of prior entrepre-neurial exposure on entrepreneurial intentionmediated by attitude, sub-jective norm, and perceived behavioral control. In this regard, we linkprior entrepreneurial exposure (i.e., (1) observation of self-employed

640 F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

parents and (2) prior work experience in a small or newly founded firm)with the three attitudinal variables proposed by Ajzen's (1991) theory ofplanned behavior (TPB) (i.e., attitude, subjective norm, and perceived be-havioral control) to explain entrepreneurial intention. Second, we sepa-rately account for the perceived quality of prior entrepreneurialexposure as a determinant of entrepreneurial intention.

We seek to contribute to extant literature byproviding amore differ-entiated understanding of the relation between prior entrepreneurialexposure and entrepreneurial intention. In this regard, our first contri-bution is on the link between different types of prior entrepreneurial ex-posure and the three attitudinal variables of the TPB (i.e., attitude,subjective norm, and perceived behavioral control) explaining entre-preneurial intention. We demonstrate how observational exposure(by means of entrepreneurial role models) and direct exposure (bymeans of work experience in small or newly founded firms) affectentrepreneurial intention differently. As a second contribution, weaccount for the perceived quality of prior entrepreneurial exposure. Bythis means, we demonstrate how prior entrepreneurial exposureperceived as positive differently affects entrepreneurial intentioncompared to exposure perceived as negative. In sum, disentanglingprior entrepreneurial exposure (in terms of type and perceived quality)and linking it with the TPB offers a more detailed understanding of theformation of entrepreneurial intention and contributes to resolvingheterogeneous prior findings regarding the prior entrepreneurialexposure and entrepreneurial intention relation.

The next section presents the background literature. We thendevelop hypotheses, which we test on a dataset consisting of studentsand professionals. The paper closes with a discussion of our findingsand by pointing out implications and limitations.

2. Background Literature

Themajority of earlier literature employs direct effectmodels to investigatehow prior entrepreneurial exposure affects entrepreneurial intention. However,such studies display inconclusive results (Chlosta et al., 2012; Shook et al.,2003). Studying the impact of role models, some studies suggest that childrenwith entrepreneurial parents display higher levels of entrepreneurial intention(e.g., Crant, 1996;Matthews&Moser, 1995).However, other studiesdonot sup-port this view (e.g., Gird&Bagraim, 2008; Kolvereid& Isaksen, 2006; Tkachev&Kolvereid,1999). Studiesanalyzing theeffectofpriorworkexperience insmallornewly founded firms display ambiguous results as well. Some studies(e.g., Kautonen et al., 2010;Matthews &Moser, 1995) find no significant effectof suchexposureon individuals' entrepreneurial intention,whereasotherstudiesreport a positive effect (e.g., Autio et al., 2001;Mueller, 2006).

However, these ambiguous results are not surprising, as priormeta-analyses suggest that exogenous influences such as prior entre-preneurial exposure are only weak direct predictors for behaviorssuch as starting a business (e.g., Sheppard, Hartwick, & Warshaw,1988). In contrast, intentions are the best predictor for plannedbehaviors (Bagozzi et al., 1989; Kim & Hunter, 1993). However,intentions derive from attitudinal variables, which are influenced byexogenous factors such as prior entrepreneurial exposure (Krueger,Reilly, & Carsrud, 2000). That is, prior entrepreneurial exposureindirectly influences entrepreneurial intention mediated throughattitudinal variables (rather than having a direct impact).

The TPB (Ajzen, 1991) is one of the predominant theoretical frame-works to analyze the formation of intentions in various fields (Armitage& Conner, 2001). The TPB claims that three conceptually distinct attitu-dinal variables determine intention: attitude towards the behavior,subjective norm, and perceived behavioral control in regard to thebehavior. Attitudes refer to the degree to which an individual evaluatesa specific behavior as favorable or unfavorable (Ajzen, 1988). Subjectivenorm mirrors individuals' perceived social pressure by attachmentfigures to perform or not to perform a specific behavior (Ajzen, 1988).Perceived behavioral control contains perceptions of the ability to suc-cessfully execute and control the focal behavior (Ajzen, 1991, 2002).

In general, the more favorable the attitude and subjective norm in re-gard to the behavior and the greater the perceived behavioral controlover the behavior, the stronger the individual's intention to performthe focal behavior (Armitage & Conner, 2001).

In addition to Ajzen's (1991) TPB, other empirical researches onindividuals' entrepreneurial intention ground on Shapero's “model ofthe entrepreneurial event” (SEE) (Shapero & Sokol, 1982). The SEEspecifically aims at explaining entrepreneurial intentions, which derivefrom individuals' perceiveddesirability aswell as perceived feasibility ofstarting a business. Additionally, the SEE includes a third predictorlabeled propensity to act, which reflects individuals' willingness to acton one's decisions (Krueger et al., 2000; Shapero & Sokol, 1982).

Both the TPB and the SEEfind broad acceptance in thepresent study'sresearch domain (Schlaegel & Koenig, 2014) and are complementarytheoretical approaches to explain individuals' entrepreneurial intention.To this end, the TPB and the SEE share a considerable conceptual overlap(Krueger, 2009; Krueger & Carsrud, 1993). In particular, bothmodels usea predictor representing the “willingness” (attitude in the TPB, perceiveddesirability in the SEE) stemming from individuals' outcome expecta-tions resulting from the behavior (i.e., starting a business) as well as apredictor representing individuals' perceived “capability” to successfullyperform the focal behavior (perceived behavioral control in the TPB,perceived feasibility in the SEE) (van Gelderen et al., 2008).

However, both models also display differences, which have to betaken into consideration when deciding upon which of the theoriesbest applies to a study's goals and research design. From a conceptualstance, the TPB includes a specific predictor (subjective norm) accountingfor social influences (e.g., from rolemodels) on entrepreneurial intention,whereas the SEE integrates such influences in the perceived desirabilitypredictor (Nabi, Holden, & Walmsley, 2006; van Gelderen et al., 2008).Moreover, Krueger et al. (2000) assert that a theory-consistent integra-tion of the “propensity to act” component in the SEE requires a longitudi-nal research design as triggering events (such as spotting a businessopportunity), which force individuals to act, precede individuals' desir-ability and feasibility perceptions. From a methodological stance, it isimportant to compare both models' explanatory power when decidingupon which of the theories to choose for a study. A recent meta-analysis drawing on broad empirical evidence (123 independent sam-ples, n = 114,007 individuals) by Schlaegel and Koenig (2014) findsthat the TPB explains a larger proportion of variance in entrepreneurialintention compared to the SEE (SEE: R2= .21; TPB: R2= .28). Analyzingthe influence of each model's attitudinal variables on entrepreneurial in-tention, the meta-analysis by Schlaegel and Koenig (2014) finds that allattitudinal variables (attitude, subjective norm, perceived behavioralcontrol) of the TPB positively influence entrepreneurial intention. In con-trast, results for SEE's attitudinal variables aremixed.While perceived de-sirability and perceived feasibility positively impact individuals'entrepreneurial intention, propensity to act has no significant influence.Finally, the TPB also displays high explanatory power in other researchfields than entrepreneurship (Ajzen, 1991; Armitage & Conner, 2001;Sutton, 1998) while comparatively fewer studies employ the SEE inmultivariate empirical studies in entrepreneurship research (Guerrero,Rialp, & Urbano, 2008; Solesvik, Westhead, Kolvereid, & Matlay, 2012).

Acknowledging the important contributions made by studies draw-ing on the SEE to explain entrepreneurial intentions and considering thetwo frameworks as complementary (rather than contradictory), thepresent study draws on TPB rationale as the above conceptual andmethodological issues are particularly pertinent for the present study'sgoals and research design.

In an effort to overcome previously inconclusive findings from directeffect models, some authors employ intention-based frameworksassuming indirect influences of exogenous factors on entrepreneurial in-tention. For example, Krueger (1993) tests the effect of breadth of priorentrepreneurial exposure (an aggregated sum score consisting of severaltypes of prior exposure) on entrepreneurial intention mediated by per-ceived desirability and perceived feasibility of starting a business. While

641F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

breadth of prior entrepreneurial exposure has no effect on the perceiveddesirability, it positively affects the perceived feasibility of starting a busi-ness. Accounting for the perceived quality of prior exposure (again, withan aggregated sum score), the study reveals a positive impact on theperceived desirability of starting a business, whereas no such effect existson the perceived feasibility of starting a business. Krueger's findingssuggest that “more” and “more positive” exposure is not always sufficientto increase individuals' perceived desirability or perceived feasibility ofstarting a business and, in turn, their entrepreneurial intention. However,the study does not answer which specific types of prior entrepreneurialexposure (e.g., role model exposure vs. work experience) raise individ-uals' desirability and feasibility perceptions. However, prior exposuremight not be additive (Peterman & Kennedy, 2003) and, hence, maynot be adequately captured by aggregated sum scores.

In a recent study, Liñán and Chen (2009) analyze the influence of ex-posure to entrepreneurial role models and previous self-employmenton entrepreneurial intention mediated by the attitudinal variables ofthe TPB. The authors find that entrepreneurial role models positivelyaffect individuals' attitude towards starting a business, whereas nosuch influence exists on subjective norm or perceived behavioral con-trol. In contrast, prior founding experience positively affects only TPB'ssubjective norm component. While Liñán and Chen's study contributesto amore differentiated picture of howdifferent types of prior entrepre-neurial exposure affect attitudinal variables and in turn entrepreneurialintention, some issues remain unresolved. First, the study only vaguelyspecifies exposure to entrepreneurial role models as “knowing anentrepreneur personally”. Hence, the specific impact of entrepreneurialparents – the most influential role models on individuals' occupationalpreferences (Pallone, Rickard, & Hurley, 1970) – remains unclear.Second, the study neglects the qualitative dimension of prior entrepre-neurial exposure. Hence, it remains unclear whether results may stemfrom different effects of exposure perceived as positive or negative.

In sum, literature on the effects of prior entrepreneurial exposure onentrepreneurial intention still displays a rather heterogeneous picture.Direct approaches are largely abandoned due to their limited explanato-ry power and predictive validity (Krueger et al., 2000). However,indirect approaches have yet only partly disentangled the prior entre-preneurial exposure construct (in terms of different types and perceivedquality of exposure) and, hence, are not fully able to resolve conflictingresults. In sum, there is still need for a more differentiated understand-ing of the relationship between prior entrepreneurial exposure andentrepreneurial intention.

3. Hypotheses development

Prior exposure stems from two different means: observation and di-rect experience (Bandura, 1977; Latham & Saari, 1979). Individuals'

Atti

Subje

no

Perce

behaviora

PEX: work

experience in small /

newly founded firms

PEX: parental

role models

H1

H2

(a)

(b)

(c)

(+)

(+)

(+)

Fig. 1. Researc

observation of self-employed parents is referred to as entrepreneurialrole model exposure. Individuals may obtain direct entrepreneurialexperience by working in a small or newly founded firm (Kautonenet al., 2010; Krueger, 1993). Linking prior entrepreneurial exposurewith Ajzen's (1991) TPB, the following sections develop a comprehen-sive rationale how entrepreneurial role model exposure and priorwork experience in a small or newly foundedfirm impact entrepreneur-ial intention mediated by attitude, subjective norm, and perceivedbehavioral control. Fig. 1 depicts our research model.

3.1. Entrepreneurial role model exposure

We expect individuals with parents who have previously started abusiness to display higher levels of entrepreneurial intention.Moreover,we expect this influence to be mediated by the three attitudinalvariables of the TPB (Ajzen, 1991) (i.e., attitude, subjective norm, andperceived behavioral control). Generally, social learning theory(Bandura, 1977, 1986) emphasizes the effect of behavior acquisitionthrough the observation of others referred to as role models. Observa-tional learning from role models influences individuals' personalitydevelopment and hence the formation of attitudes which in turn formintentions (Bandura, 1977). Naturally, children are especially exposedto their parents' behaviors. Thus, parental role models are of particularimportance for the development of children's attitudes as role modeleffects are stronger the more relevance and credibility a role modelpossesses (Scherer, Adams, Carley, & Wiebe, 1989; Shapero & Sokol,1982). Children learn by observing their parents and internalize thesetriggers into their mental models. These in turn affect their decisionpolicies, including employment decisions (Bandura, 1986; Schröder &Schmitt-Rodermund, 2006).

Social learning theory suggests the observation of entrepreneurialrole models to be influential on individuals' attitudes towards differentcareer alternatives. Individuals observe occupational behaviors ofvalued role models and at the same time form cognitive evaluations oftheir own actual or future capabilities and interests and, in turn, out-come expectations regarding the observed career field (Krumboltz,Mitchell, & Jones, 1976; Scherer, Adams, & Wiebe, 1989). By creatingan environment strongly influencing the personal characteristics oftheir children, entrepreneurial parents convey the desirability of anentrepreneurial career to their offspring (Matthews & Moser, 1995;Shapero & Sokol, 1982). Thus, being exposed to entrepreneurial parentsshapes children's attitude in regard to self-employment (Carr &Sequeira, 2007).

Parental role models may also affect offspring's subjective norm inregard to starting a business. Role model influence may also occurthrough social persuasion. Conversations and discussions with parentalrole models provide individuals with insights into career alternatives

tude

ctive

rm

ived

l control

Entrepreneurial

intention

(+)

(+)

(+)

h Model.

642 F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

and, hence, exert a strong influence on individuals' career intentions(van Auken et al., 2006). Moreover, children are likely to become apart of their entrepreneurial parents' social networks exerting furthersocial pressure on them to start a business (Kim et al., 2006).

Parental rolemodelsmay also affect offspring's perceived behavioralcontrol in regard to starting a business. Individuals may learn certainskills and behaviors necessary for starting a business by observing rolemodels (Scherer, Adams, & Wiebe, 1989). Children receive an informaltransfer of business knowledge andmethods from their entrepreneurialparents. This human capital strengthens the offspring's conviction tosuccessfully execute the tasks related to starting a business (Dunn &Holtz-Eakin, 2000; Scherer, Brodzinski, & Wiebe, 1991) and, thus,leads to a higher perceived behavioral control. In sum, we hypothesize:

Hypothesis 1. Exposure to entrepreneurial role models positivelyinfluences entrepreneurial intention. This influence is mediated by(a) attitude, (b) subjective norm, and (c) perceived behavioral control.

3.2. Prior work experience in a small or newly founded firm

Weexpect individualswith priorwork experience in a small or newlyfounded firm to display higher levels of entrepreneurial intention. More-over, we expect this influence to be mediated by the three attitudinalvariables of the TPB (Ajzen, 1991) (i.e., attitude, subjective norm, and per-ceived behavioral control). Prior to employment decisions between paidemployment and self-employment, individuals draw decision-relevantinformation in particular from their own memory. Availability and con-tent of such information largely depends on individuals' prior exposure.Besides drawing from similar experiences (such as actual prior self-employment), individuals are also able to utilize comparable experiences(such as prior work experience in a small or newly founded firm) in thecognitive decision-making process (Katz, 1992). Previous researchshows that prior work experience is an important component of humancapital for potential entrepreneurs (Kim et al., 2006). Small and newlyfounded firms provide a work environment ideally suited for sharing,experiencing, and learning the skills beneficial for starting a business(Rotefoss & Kolvereid, 2005; Stuart & Abetti, 1990).

Work experience in small or newly founded firms is likely to influ-ence individuals' attitude in regard to starting a business. The workenvironment in small or newly founded firms is different in terms offlexibility, possibilities to participate, working hours, or job securitycompared to large firms. Hence, such particular experience is likely tofoster the development of entrepreneurial work attitudes (Kautonenet al., 2010; Parker, 2004). This rationale is consistent with Dyer(1994) who suggests that individuals who previously worked for anentrepreneur prefer an entrepreneurial lifestyle. Moreover, suchexperience allows individuals to accurately evaluate the personalconsequences of starting a business (Scherer, Adams, & Wiebe, 1989).

Prior work experience in a small or newly founded firm is also likelyto positively influence individuals' subjective norm in regard to startinga business. First, it seems rather likely that reference people encourageindividuals to engage in vocational activities in which they alreadyhave gathered human and social capital giving them the opportunityto realize greater utility from this capital compared to other vocationalopportunities (Douglas & Shepherd, 2000). Second, it is even possiblethat reference people approach individuals with business opportunitiesbecause they perceive the individual competent enough to executethese opportunities as they have acquired the necessary skills in thecourse of their work experience (MacMillan, 1986).

Prior work experience in a small or newly founded firm also posi-tively influences the perceived behavioral control over the process ofstarting a business (Scherer, Adams, & Wiebe, 1989). Work experienceallows potential entrepreneurs to gain experience and to obtain neces-sary skills relevant for starting their own business (Brenner, Pringle, &Greenhaus, 1991). Despite acquiring general business human capital,

work experience provides potential entrepreneurswith the opportunityto obtain job- or industry-specific business human capital, which allowsidentifying potential customers and competitors. Moreover, potentialentrepreneurs gain access to social networks for market information,capital, or hiring employees. In addition, they develop supplier andcustomer relationships (Fairlie & Robb, 2007; Kim et al., 2006). Lastly,individuals with work experience in small or newly founded firms aremore likely to be generalists rather suited to start their own businesscompared to specialists better suited for specific tasks in larger firms(Gibb, 2002). In sum, we hypothesize:

Hypothesis 2. Prior work experience in a small or newly founded firmpositively influences entrepreneurial intention. This influence ismediatedby (a) attitude, (b) subjective norm, and (c) perceived behavioral control.

3.3. Perceived quality of prior entrepreneurial exposure

We expect prior entrepreneurial exposure perceived as positive topositively influence individuals' entrepreneurial intention. More specif-ic, we expect the influence of entrepreneurial role models and priorwork experience in a small or newly founded firm perceived as positiveon entrepreneurial intention to be mediated by the three attitudinalvariables of the TPB (Ajzen, 1991) (i.e., attitude, subjective norm, andperceived behavioral control).

We expect entrepreneurial rolemodel exposure perceived as positiveto have a more positive influence on the attitudes in regard to starting abusiness opposed to exposure perceived as negative (Krueger, 1993;Matthews & Moser, 1996). Based on prior observational learning fromrolemodels, individuals form cognitive evaluations of career alternativeswhich can either encourage or discourage them from choosing a specificcareer path (Krumboltz et al., 1976;Mitchell &Krumboltz, 1984). This ra-tionale is supported by prior research (e.g., Barling, Dupre, & Hepburn,1998), which found children's perceptions of parental work experiencesinfluential on their own work attitudes. Hence, role model exposureperceived as negative may foster the development of negative attitudestowards self-employment and, in turn, discourage individuals fromfollowing the role models' behavior (Mungai & Velamuri, 2011).

We expect that work experience in small or newly founded firmsperceived as positive has a more positive influence on individuals'attitudes in regard to starting a business opposed to exposure perceivedas negative. By providing a work environment closely related to an en-trepreneurial career, work experience in small or newly founded firmsinfluences individuals' attitudes in regard to starting a business (Dyer,1994; Kautonen et al., 2010). Prior research analyzing individuals' expo-sure to entrepreneurship education programs (which are rather similarto small or newly founded firm work experience (Fayolle, 2005)) sup-ports this rationale. Entrepreneurship education perceived as positivehas a positive impact on individuals' attitudes in regard to starting abusiness (opposed to entrepreneurship education perceived as nega-tive) (Peterman & Kennedy, 2003; Wilson, Kickul, & Marlino, 2007).

In contrast to prior studies (e.g., Scherer, Adams, Carley, et al., 1989;Scherer et al., 1991), we argue that not the role model's or small/newlyfounded firm's objective success (e.g., in terms of profitability)influences individuals' attitudes, but whether the individual him- orherself perceived prior exposure as positive or negative. Even objective-ly negative experiences (e.g., bankruptcy) from which the individuallearns how to avoid errors in the start-up process might be consideredas positive (Krueger, 1993). In contrast, objectively successful exposurecould be considered as negative by an individual, due to parents' longworking hours or economic uncertainties (Kim et al., 2006; van Aukenet al., 2006). In sum, we hypothesize:

Hypothesis 3. Exposure to entrepreneurial role models perceived aspositive positively influences entrepreneurial intention. This influenceis mediated by (a) attitude, (b) subjective norm, and (c) perceivedbehavioral control.

643F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

Hypothesis 4. Prior work experience in a small or newly founded firmperceived as positive positively influences entrepreneurial intention.This influence is mediated by (a) attitude, (b) subjective norm, and(c) perceived behavioral control.

4. Data and methods

4.1. Data

Following TPB reasoning, entrepreneurial intention needs to bestudied prospectively rather than retrospectively (Krueger & Carsrud,1993). In other words, entrepreneurial phenomena need to be studiedbefore they occur (Davidsson & Honig, 2003). Previous entrepreneur-ship research often grounds on samples consisting of existing founders.However, this research suffers from selection bias resulting fromsampling only existent and, hence, successful founders neglectingindividuals who aborted their startup-attempt as well as hindsightbias and memory decay from surveying start-up attempts retrospec-tively (Davidsson & Honig, 2003; Krueger & Carsrud, 1993). In contrast,analyzing entrepreneurial intentions requires samples, which includeindividuals who may or may not intend to start a business (Kruegeret al., 2000). We test our hypotheses on a dataset of individuals (n =421) consisting of students (n = 245) and professionals (n = 176)from Germany.

Data collection took place between June and December 2009. Datawere gathered by respondents filling out either a paper-based or an on-line questionnaire. Due to missing data, we had to eliminate 47 casesfrom the sample. Thus, the final sample consists of 374 cases (students:n= 227; professionals: n = 147). Consistent with numerous researcheson entrepreneurial intention (e.g., Krueger et al., 2000; Liñán & Chen,2009), we collected data from students, who – due to their age andeducational status – face the decision between paid-employment andself-employment in the immediate future (Matthews & Moser, 1995;Scherer, Adams, Carley, et al., 1989). However, previous research indi-cates that student entrepreneurs differ from non-student entrepreneurs,which means that findings may not be universally applicable (Robinson,Huefner, & Hunt, 1991; Shook et al., 2003). Hence, following Chlosta et al.(2012), we also included individuals who already started their profes-sional career in our sample (professionals). This approach allowssampling individuals who are older on average and have a wider rangeof prior work and founding experience (Autio et al., 2001). As the TPBaims at explaining behavior intended in the foreseeable future (Ajzen &Madden, 1986) and the link between intention and subsequent behaviorneeds to be clear and salient to individuals (Sheppard et al., 1988), wecollected data from individuals facing career decisions in the near future(Krueger, 1993). The students in our sample were in their final year atuniversity, the professionals were employed in an IT firm currentlyundergoing restructuring. Thus, starting a business within the next twoyears was a career option for all individuals in our sample.

4.2. Measures

This section describes the measurement of the constructs in ourresearch model. To obtain reliable and valid measures, we draw onestablished measurement scales from prior literature as impropermeasurement leads to questionable findings and potentially unsoundconclusions (Crook, Shook, Morris, & Madden, 2010; Short, Ketchen,Combs, & Ireland, 2010).

4.2.1. Entrepreneurial intentionConsistent with the TPB as our theoretical underpinning, our

measurement approach follows Ajzen's (1991, p. 181) definition ofintention (“indications of how hard people are willing to try, of howmuch of an effort they are planning to exert, in order to perform thebehavior”) and his recommendations on how to measure this latent

construct (Ajzen, 1991, 2006). To this end, we employ a multi-itemmeasure consisting of desires (“I want to perform the behavior”), inten-tions (“I intend to perform the behavior”), and self-predictions (“I willperform the behavior” or “How likely is it that you perform the behav-ior”). Using such a mixedmeasure of intention is widespread in generalintention-based research (see Armitage & Conner 2001 for examples) aswell as in research on entrepreneurial intentions (e.g., Chen, Greene, &Crick, 1998; Davidsson, 1995). Furthermore, we follow Ajzen's (1991)recommendations for measuring intentions that accurately predict thefocal behavior. First, intention measures must accurately correspondto the focal behavior (i.e., starting a business andnot, for example, tryingto start a business). Second, intention measures must include a foresee-able time span (here: two years) duringwhich the focal behavior shouldoccur. In other words, the link between intention and behavior needs tobe clear and salient to individuals (Sheppard et al., 1988). This specifica-tion is necessary, as intentions need to remain stable in the interval be-tween their measurement and the occurrence of the focal behavior.Otherwise intervening events (that potentially change individual's in-tention) may markedly reduce the predictive validity of the intentionmeasure (Ajzen, 1991; Krueger & Carsrud, 1993).

Based on these considerations, we measured entrepreneurial inten-tion by asking respondents whether they (1) intend, (2) expect,(3) want to start a business within the next two years, and (4) howthey rate the likelihood of starting a business within the next twoyears. Each item representing entrepreneurial intention was measuredon a Likert scale ranging from “1” to “7”.

In sum, our way of measuring entrepreneurial intention is also con-sistent with recent recommendations by Thompson (2009) regardingthemeasurement approach. That is, wemeasure entrepreneurial inten-tion on a continuous (7-point Likert) scale as opposed to a categoricalmeasurement approach as categorical measures tend to oversimplifythe distinction between individuals who display / do not display entre-preneurial intention and are unable to express individual's level of in-tention. Moreover, we use reflective rather than formative indicatorsas also recommended by Diamantopoulos and Siguaw (2006).Additionally, we use multi-itemmeasurement to assess individuals' en-trepreneurial intention. Multi-item measurement allows for assessingthe reliability and validity of latent constructs such as entrepreneurialintention. In this regard, our four-item measure of entrepreneurial in-tention displays high internal consistency (Cronbachs α = .965), aswell as sufficient convergent and discriminant validity (see Table 1 fordetailed results). Lastly, we report the wording of each of the fouritems measuring entrepreneurial intention ensuring full replicabilityof our approach.

Despite several similarities, our measurement of entrepreneurialintention differs from the one by Thompson (2009) by including atime span of two years until the focal behavior (starting a business)should occur. In contrast, Thompson states that the “point in the fu-ture might be imminent or indeterminate (…)” (Thompson, 2009,p. 676). To account for whether the time span is problematic, weconduct a robustness check by calculating the correlation of ourfour-item entrepreneurial intention measure with an item (mea-sured on a 7-point Likert scale) from Kolvereid (1996) (“If youcould choose between being self-employed and being an employeein an organization, what would you prefer?”), which does not in-clude such a finite time span. The correlation between bothmeasuresis significantly positive (r= .45; p≤ .001). Additionally, we examinethe robustness of our entrepreneurial intention measure in compar-ison to themeasure developed by Liñán and Chen (2009) in an entre-preneurial context. To this end, we collected additional data fromn = 136 German students. After verifying the reliability and validityof both scales with satisfying results, we computed the bivariate cor-relation between bothmeasures. The correlation coefficient amountsto r = .64 (p ≤ .001) indicating a high positive correlation betweenour entrepreneurial intention measure and the one established byLiñán and Chen (2009).

Table 1Convergent validity.

Factor

Entrepreneurialintention

Attitude Subjectivenorm

Perceived behavioralcontrol

EI 1 97EI 2 .99EI 3 .87EI 4 .91Att. 1 .75Att. 2 .93Att. 3 .82Att. 4 .83SN 1 .79SN 2 .79PBC 1 .67PBC 2 .50PBC 3 .78

Note: Extractionmethod: principal axis factorization; Rotationmethod: Promax; Normal-ization with Kaiser; Rotation converged after six iterations; Loadings below .3 not shown.

644 F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

4.2.2. AttitudeTomeasure the attitude towards starting a businesswe employ a di-

rect overall measure of attitude consistentwith TPB reasoning proposedby Ajzen (1991). To this end, we use a semantic differential with fouritems tapping respondents' attitudes towards starting a business(Cronbachs α= .901). We asked respondents whether starting a busi-ness within the next two years would be foolish/smart, harmful/beneficial, worthless/useful, bad/good for them. Each item wasmeasured on a Likert scale ranging from “1” to “7”.

4.2.3. Subjective normTo measure subjective norm, Ajzen (1991) suggests obtaining an

overall measure by asking respondents to rate the extent to which at-tachment figures would approve or disapprove of them performing aspecific behavior. Our subjective norm scale consists of two items eachranging from “1” = strongly disagree to “7” = strongly agree. Weasked respondents whether people that are of importance to her/himexpect her/him to start a businesswithin the next two years andwheth-er such people think that she/he should start a business within the nexttwo years (Cronbachs α = .787).

4.2.4. Perceived behavioral controlMeasures of perceived behavioral control have to tap respondents'

confidence of being able to successfully perform a specific behavior. Inaccordance with Ajzen (2002), we measure perceived behavioral con-trol as overall measure containing efficacy aswell as controllability per-ceptions. To this end, we used three Likert-scaled items (each rangingfrom “1” to “7”) asking respondents to indicate whether starting a busi-ness within the next two years would be impossible/possible or easy/difficult for them and whether starting a business within the next twoyears would be beyond/within their control (Cronbachs α = .727).

4.2.5. Prior entrepreneurial exposureWe adapted ourmeasures for the two types of prior entrepreneurial

exposure under investigation from Krueger (1993). We asked respon-dents to indicate whether (1) their parents had previously started abusiness and whether (2) they previously worked for a small or newlyfounded firm (“0” = no; “1” = yes).

4.2.6. Perceived quality of prior entrepreneurial exposureFollowing each question asking respondents whether they had a

specific type prior entrepreneurial exposure ((1) parental role models,(2) work experience in a small or newly founded firm), respondents –who claimed prior exposure in a specific field – were asked to ratewhether they perceived this exposure as negative (coded “−1”), nei-ther/nor (coded “0”), or positive (coded “1”). In contrast, respondentswho had no exposure in a specific field were excluded from thefollow-up question. This way of measuring the perceived quality ofprior entrepreneurial exposure grounds on Krueger (1993).

4.3. Assessing reliability and validity

To assess the reliability of our scales we calculate Cronbach's alpha.The values ranging from .727 to .965 indicate high internal consistency(Nunnally, 1978). Next, we employ several procedures to assess the va-lidity of our scales. First, we assess convergent validity by conducting aprincipal axis factor analysis with the items measuring our four latentvariables (i.e., attitude, subjective norm, perceived behavioral control,entrepreneurial intention). The Kaiser–Meyer–Olkin test (.910) andBartlett's sphericity test (p b .001) both yield satisfying results, suggest-ing our data is well suited for conducting a factor analysis. Principal axisfactor analysis extracts three factors with eigenvalues greater than one,whereas the fourth factor displays an eigenvalue of .93. Thus, we takethe scree plot into account, which suggests a four-factor solution.Table 1 displays the rotated factor matrix with four factors illustratingthat all items load on their theoretically assigned factors only.

Second, we assess discriminant validity consistent with the criterionintroduced by Fornell and Larcker (1981). Accordingly, discriminant va-lidity exists whenever the average variance extracted (AVE) for eachconstruct is higher than the squared correlation between the constructs.Hence, we analyzed each pair of latent constructs and found them all todemonstrate sufficient discriminant validity.

4.4. Assessing measurement invariance

As responses to certainmeasurement itemsmay systematically varyacross different groups of individuals (such as students and profes-sionals), conclusions based on non-invariant scales may be ambiguousor at worst erroneous (Adler, 1983; Steenkamp & Baumgartner, 1998).Hence, testing for measurement invariance is necessary to ensure thatscale items measure their underlying constructs equivalently acrossgroups (Singh, 1995). Consistent with Steenkamp and Baumgartner(1998),we apply a sequence ofmultigroup confirmatory factor analysesto test for cross-group measurement invariance. First, we test forconfigural invariance (i.e., equal factor structure across groups) findingthe unconstrained baselinemodel to fit satisfactory (χ2= 225.21, dƒ=154,χ2/dƒ= 1.46, IFI=.98, TLI=.98, CFI=.98, RMSEA= .04). Second,we test for metric invariance (i.e., equal factor loadings across groups)by constraining the measurement weights across groups. Fit indices(χ2 = 238.36, dƒ = 163, χ2/dƒ = 1.46, IFI = .98, TLI = .98, CFI = .98,RMSEA = .04) and the insignificant increase in χ2 between theconstrained and the unconstrained baseline model (Δ χ 2 = 13.15;p = .16) suggest metric invariance. Third, we test for scalar invariance(i.e., equal intercepts across groups) by constraining the measurementintercepts across groups. Fit indices (χ 2 = 257.48, dƒ = 176, χ2/dƒ= 1.46, IFI= .98, TLI= .98, CFI= .98, RMSEA= .04) and the insignif-icant increase in χ2 between the constrained and the unconstrainedbaseline model (Δ χ2 = 32.27; p = .07) suggest scalar invariance.

4.5. Assessing common method variance

As we gathered data from a single respondent using a single meth-odology (paper-based- or online-questionnaire), it may be susceptiveto common method bias (CMB) (Podsakoff, MacKenzie, Lee, &Podsakoff, 2003; Podsakoff & Organ, 1986).We employ two proceduresto evaluate the magnitude of CMB. First, we employ Harman's-One-Factor-Test to assess the extent of CMB (Podsakoff & Organ, 1986;Podsakoff et al., 2003). Principal component factor analysis with six var-iables extracts two factors with eigenvalues greater than 1 (first factor:42.4%, second factor: 17.7%). As no single factor emerges andnone of thefactors accounts for most of the variance, CMB is not a concern. Second,following Podsakoff et al. (2003), we conduct a confirmatory factor

645F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

analysis (CFA) testing three different models: (1) loading all of the itemsonto one commonmethod factor (χ2= 968.70, dƒ=90,χ2/dƒ= 10.76,IFI= .77, TLI= .74, CFI= .77, RMSEA= .16, AIC= 1028.70), (2) loadingall items onto their theoretically assigned and correlated variables (χ2=158.27, dƒ = 77, χ2/dƒ= 2.06, IFI = .98, TLI = .97, CFI = .98, RMSEA=.05, AIC = 244.27), and (3) loading the items onto theirlatent correlated variables as well as onto an additional method factor(χ2 = 116.86, dƒ = 62, χ2/dƒ = 1.89, IFI = .99, TLI = .98, CFI = .99,RMSEA= .05, AIC= 232.86). As models two and three fit similarly wellto the data (i.e., adding a common method factor does not significantlyimprove model fit), no severe threat of CMB exists.

5. Results

Table 2 displays the means, standard deviations, and correlationsamong the focal variables. As none of the correlations exceeds thethreshold of .7, no serious risk of multicollinearity exists (Anderson,Sweeney, & Williams, 2002). However, as the highest correlation is.61, we additionally compute the variance inflation factor (VIF) foreach independent variable. As the highest VIF is 1.39 and thus stayswell below the threshold of 2.5 (Allison, 1999), severe multicollinearityamong the variables is absent.

We apply structural equation modeling (SEM) to test our hypothe-ses. As recommended by Anderson and Gerbing (1988), we follow atwo-step approach. First, we estimate the measurement model usingCFA to assess the reliability and validity of our latent constructs. Second,we estimate the hypothesized structural model.

5.1. Measurement model

We assess the measurement model's fit by drawing on the Chi-square/df ratio, the Tucker Lewis Index (TLI), the Incremental Fit Index(IFI), the Comparative Fit Index (CFI), and the Root Mean Square Errorof Approximation (RMSEA). The Chi-square/df ratio (2.06) for themea-surementmodel is below the threshold of 3.0 (Kline, 1998), whereas TLI(.97), IFI (.98), and CFI (.98) all exceed the threshold of .95. The RMSEA(.05) stays below the cutoff point of .06 (Hu& Bentler, 1999). Hence, ac-cording to all fit indices, ourmeasurementmodel displays a goodmodelfit. Moreover, the standardized factor loadings in the CFA are all above.56 exceeding the recommended minimum of .4 (Ford, MacCallum, &Tait, 1986).

5.2. Structural equation modeling

We test the influence of entrepreneurial role model exposure andwork experience in a small or newly founded firm on entrepreneurialintention mediated through attitude, subjective norm, and perceivedbehavioral control (model 1). To this end, we use the full sampleconsisting of individuals who claimed to have none, one, or both types

Table 2Descriptive statistics and correlation among variables.

Mean SD 1 2 3 4 5 6

1 Entrepreneurialintention

2.45 1.81 1

2 Attitude 4.10 1.33 .61** 13 Subjective norm 2.02 1.50 .53** .34** 14 Perceived behavioral

control3.65 1.52 .59** .48** .34** 1

5 PEX: parental rolemodels

0.30 0.46 .12* .07 .12* .08 1

6 PEX: workexperience

0.48 0.50 .16** .15** .06 .19** .14** 1

Note: n = 374; Pearson correlation (bivariate) with listwise deletion; SD: standarddeviation; PEX: Prior Entrepreneurial Exposure.*: p ≤ .05, **: p ≤ .01.

of prior entrepreneurial exposure and did not report anymissing values(n = 374).

We follow recent recommendations by James, Mulaik, and Brett(2006) to test our mediator Hypotheses 1a–c and 2a–c. To this end,two steps have to be completed. First, it is necessary to determinewhether the hypothesized mediator relationships are full or partial.Given that the TPB theoretically assumes sufficiency (i.e., TPB's attitudi-nal variables fully mediate the influence of external factors on inten-tion) (Ajzen, 1991), we use a fully mediated model as baseline modelfor subsequent model comparisons. Moreover, full mediation modelsare the most parsimonious type of mediation models (in other words,have more degrees of freedom) and, hence, are easier to rejectcompared to partial mediation models. Thus, assuming full mediationis consistent with the basic philosophy of science doctrine (Mulaik,2001). Second, we use SEM techniques to test our mediationhypotheses. To this end, we test paths from the predictor variables(i.e., entrepreneurial role model exposure, work experience in a smallor newly founded firm) to the mediator variables (i.e., attitude, subjec-tive norm, perceived behavioral control) as well as a path from themediator variables to the dependent variable (i.e., entrepreneurialintention). In contrast, direct paths from the predictor variables to thedependent variable are not included as such direct effects are not anecessary condition for establishing mediation (James et al., 2006;Zhao, Lynch, & Chen, 2010).

The fit indices of our hypothesized full mediation model (modelA) suggest goodmodel fit: The χ2/df ratio (2.01) is below the thresholdof 3.0 (Kline, 1998). The TLI (.97), the IFI (.98), and the CFI (.98) allexceed the threshold of .95. The RMSEA (.05) stays below the cutoffpoint of .06 (Hu & Bentler, 1999). Against our baseline model, we testthree nested models (assuming partial mediation) and one alternativenon-mediated model assuming only direct effects. Compared to modelA,model B includes a direct path from entrepreneurial rolemodel expo-sure to entrepreneurial intention. Model C differs from model A byincluding a direct path from work experience in a small or newlyfounded firm to entrepreneurial intention. Model D adds two directpaths from both predictor variables to entrepreneurial intention. Lastly,model E assumes only direct (non-mediated) effects from both predic-tor variables on entrepreneurial intention. As Table 3 suggests, thedifferences between χ2 values are not significant for models B–Ecompared to model A. Model A displays the lowest value for the AkaikeInformation Criterion (AIC). In sum, these results suggest that model Ahas the best model fit. Hence, the attitudinal variables of the TPB fullymediate the relationship between entrepreneurial role model exposureas well as work experience in a small or newly founded firm on entre-preneurial intention.

Despite several similarities, the traditional test for mediation byBaron and Kenny (1986) differs from our approach by assuming partialmediation as baseline model, which is inappropriate for the SEM ap-proach (Wang, 2008). James et al. (2006) as well as Zhao et al. (2010)provide a detailed discussion on the similarities and differencesbetween the traditional test for mediation and the more recent SEMapproach. Moreover, Iacobucci, Saldanha, and Deng (2007) provide ev-idence that the SEM approach as outlined above is superior to tradition-al regression analysis when testing for mediator relationships. In thesame vein, MacKinnon, Lockwood, Hoffman, West, and Sheets (2002)contrast 14 methods to test for mediator relationships. Their analysesreveal that the traditional approach by Baron and Kenny (1986) hasthe lowest statistical power. In contrast, the authors recommend testingfor mediation by analyzing indirect effects applying the SEM approachas outlined above.

Taking the standardized path coefficients of the hypothesized fullmediation model (model A) into account, results suggest that all attitu-dinal variables of the TPB display a highly significant (p ≤ .001) andpositive effect (attitude: .27, subjective norm: .33, perceived behavioralcontrol: .40) on entrepreneurial intention. Exposure to entrepreneurialrole models positively influences the subjective norm in regard to

Table 3Model comparisons for model 1.

Model comparisons - model 1a

Model and structure χ2 df Δχ2d χ2/df TLI IFI CFI RMSEA AIC

Ab PEX → TPB → EI 158.53 79 2.01 .97 .98 .98 .05 240.53B PEX → TPB → EI and PEX: parental role models → EI 158.37 78 .16 2.03 .97 .98 .98 .05 242.37C PEX → TPB → EI and PEX: work experience → EI 158.45 78 .08 2.03 .97 .98 .98 .05 242.45D PEX → TPB → EI and PEX: parental role models → EI and PEX: work experience → EI 158.27 77 .26 2.06 .97 .98 .98 .05 244.27Ec PEX → EI 486.84 86 5.66 .87 .90 .90 .11 554.84

a) n = 374.b) Hypothesized model (full mediation).c) Non-mediated model.d) Significance levels: *: p ≤ .05; **: p ≤ .01; ***: p ≤ .001.Note: df: Degrees of Freedom; TLI: Tucker Lewis Index; IFI: Incremental Fit Index; CFI: Comparative Fit Index; RMSEA: RootMean Square Error of Approximation; AIC: Akaike InformationCriterion; PEX: Prior Entrepreneurial Exposure; TPB: Theory of Planned Behavior; EI: Entrepreneurial Intention.

646 F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

starting a business (.13, p ≤ .05), whereas it displays no significant ef-fect on attitude (.06) and perceived behavioral control (.08). In contrast,prior work experience in a small or newly founded firm has no signifi-cant effect on the subjective norm in regard to starting a business(.05), whereas it positively affects the attitude (.15, p ≤ .01) as well asthe perceived behavioral control in regard to starting a business (.22,p ≤ .001). In sum, these findings support our Hypotheses 1b, 2a,and 2c, whereas we have to reject Hypotheses 1a, 1c, and 2b. Table 4summarizes our results and Fig. 2 displays our results graphically.

Additionally, we test for group-specific differences in resultsbetween students and professionals. To this end, we apply the criticalratio (CR) test to detect significant differences between pairs of path co-efficients for each group. Path coefficients are considered significantlydifferent between groups (at the p ≤ .05 level) if the CR exceeds an ab-solute value of 1.96 (e.g., Eisingerich & Rubera, 2010; Yun-Jeong& Kelly,2013). However, the CR values for all paths in model 1 are lower than1.96, which suggests that no significant group-specific differences inthe relations between the two types of prior entrepreneurial exposure,the threemediating attitudinal variables of the TPB, and entrepreneurialintention exist. Hence, the results of hypotheses testing (based onmodel 1) apply to both groups.

In a second set of models, we analyze the effects of the perceivedquality of prior entrepreneurial exposure on entrepreneurial intentionmediated through attitude, subjective norm, and perceived behavioralcontrol.Model 2a examines this relation for exposure to entrepreneurial

Table 4SEM results for model 1.

Hypotheses

AttitudeSubjective normPerceived behavioral control

H1a PEX: parental role modelsH1b PEX: parental role modelsH1c PEX: parental role modelsH2a PEX: work experienceH2b PEX: work experienceH2c PEX: work experienceFit measures

χ2dfχ2 / dfTLIIFICFIRMSEA

a) n = 374.b) Standardized coefficients with standard errors in parentheses.c) Significance levels: *: p ≤ .05; **: p ≤ .01; ***: p ≤ .001.Note: df: Degrees of Freedom; TLI: Tucker Lewis Index; IFI: Incremental Fit Index; CFI: Comparatial Exposure.

role models, whereas model 2b analyzes the effect of work experiencein a small or newly founded firm.We test the influence of the perceivedquality of these two types of prior entrepreneurial exposure in separatemodels, as we included only individuals in each respective analysis whowere able to rate their prior exposures as negative, neither / nor, or pos-itive. Thus, the sample size is reduced to n = 113 (students: n = 77;professionals: n = 36) in model 2a and n = 173 (students: n = 113;professionals: n = 60) in model 2b.

Again, we assume complete mediation as baseline model for modelcomparison testing. The fit indices suggest good fit for both baselinemodels. Chi-square/df ratios (1.68 for model 2a, 1.39 for model 2b)are below the threshold of 3.0 (Kline, 1998). The TLI (.95 for model 2a,.98 for model 2b), the IFI (.96 for model 2a, .99 for model 2b), and theCFI (.96 for model 2a, .98 for model 2b) all match or exceed the thresh-old of .95. The RMSEA stays below the cutoff point of .06 (Hu & Bentler,1999) in model 2b (.05), whereas model 2a displays a slightly higherRMSEA of .08. However, the RMSEA tends to overreject models due tosmall sample sizes (Hu & Bentler, 1999). As the sample size of ourmodel 2a is only n = 113, we follow Hu and Bentler's (1999) recom-mendation to draw on a combination of IFI and CFI to assess themodel fit (which yield satisfactory results for our model 2a as outlinedabove).

To test our mediator Hypotheses 3a–c and 4a–c, we, again, contrasteach baselinemodel with competingmodels.More specific, we contrastthe baselinemodels (model A) with one nestedmodel assuming partial

Model 1a

Coefficientsb,c

Entrepreneurial intention .27*** (.08)Entrepreneurial intention .33*** (.07)Entrepreneurial intention .40*** (.06)Attitude .06 (.14)Subjective norm .13* (.17)Perceived behavioral control .08 (.23)Attitude .15** (.13)Subjective norm .05 (.16)Perceived behavioral control .22*** (.21)

158.53792.01.97.98.98.05

ive Fit Index; RMSEA: RootMean Square Error of Approximation; PEX: Prior Entrepreneur-

Attitude

Subjective

norm

Perceived

behavioral control

Entrepreneurial

intention

PEX: work

experience in small /

newly founded firms

PEX: parental

role models .27***

.33***

.40***

.13*

.15**

.22***

Fig. 2. Results model 1.

647F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

mediation (model B) and one alternative non-mediated model assum-ing only a direct relation between the predictor and the dependentvariable (model C). Compared to model A, model B includes a directpath from quality of entrepreneurial role model exposure / quality ofwork experience in a small or newly founded firm to entrepreneurial in-tention.Model C assumes only a direct (non-mediated) effect from eachpredictor variable on entrepreneurial intention. Table 5 displays that thedifferences between χ2 values are not significant for models B and Ccompared to the respective baseline model A. Each baseline model Adisplays the lowest AIC value. In sum, these results suggest that eachbaseline model A best fits the data. Hence, the attitudinal variables ofthe TPB fully mediate the relationship between quality of entrepreneur-ial role model exposure as well as quality of work experience in a smallor newly founded firm on entrepreneurial intention.

Next, we report the standardized path coefficients for each completemediationmodel (models 2a and 2b). As Table 6 displays, attitude, sub-jective norm, and perceived behavioral control display a significantlypositive effect on entrepreneurial intention in model 2a (attitude: .34,

Table 5Model comparisons for models 2a and 2b.

Model comparisons — model 2aa

Model and structure

Ab Quality: parental role models → TPB → EIB Quality: parental role models → TPB → EI and Quality: parental role models → EICc Quality: parental role models → EIa) n = 113b) Hypothesized model (full mediation)c) Non-mediated modeld) Significance levels: *: p ≤ .05; **: p ≤ .01; ***: p ≤ .001Note: df: Degrees of Freedom; TLI: Tucker Lewis Index; IFI: Incremental Fit Index; CFI: ComInformation Criterion; TPB: Theory of Planned Behavior; EI: Entrepreneurial Intention

Model comparisons — model 2ba

Model and Structure

Ab Quality: work experience → TPB → EIB Quality: work experience → TPB → EI and Quality: work experience → EICc Quality: work experience → EI

a) n = 173.b) Hypothesized model (full mediation).c) Non-mediated modeld) Significance levels: *: p ≤ .05; **: p ≤ .01; ***: p ≤ .001.Note: df: Degrees of Freedom; TLI: Tucker Lewis Index; IFI: Incremental Fit Index; CFI: ComparaCriterion; TPB: Theory of Planned Behavior; EI: Entrepreneurial Intention.

p ≤ .001; subjective norm .27, p ≤ .01; perceived behavioral control:.41, p ≤ .001) and model 2b (attitude: .32, p ≤ .001; subjective norm.35, p ≤ .001; perceived behavioral control: .34, p ≤ .001). We analyzethe predictors' effects on TPB's attitudinal variables. Only the path coef-ficients of perceived quality of entrepreneurial role model exposure onattitude (.23, p≤ .05) aswell as on subjective norm (.27, p≤ .05) displaya significant relation inmodel 2a. In contrast, the effect of entrepreneur-ial role model exposure perceived as positive on perceived behavioralcontrol is non-significant (.08). Model 2b suggests that the perceivedquality of work experience in a small or newly founded firm has nosignificant effect on any of the attitudinal variables of the TPB (.02 on at-titude, .07 on subjective norm, .04 on perceived behavioral control).Summing up, we find support for Hypotheses 3a and 3b, while wehave to reject Hypotheses 3c and 4a–4c. Fig. 3 displays results formodel 2a, whereas Fig. 4 depicts results for model 2b.

Similar to model 1, we test for group-specific differences in resultsbetween students and professionals in models 2a and 2b. Again, theCR values for all paths in models 2a and 2b are lower than 1.96, which

χ2 df Δχ2d χ2/df TLI IFI CFI RMSEA AIC

116.10 69 1.68 .95 .96 .96 .08 188.10115.90 68 .20 1.70 .95 .96 .96 .08 189.90210.64 74 2.85 .86 .88 .88 .13 272.64

parative Fit Index; RMSEA: Root Mean Square Error of Approximation; AIC: Akaike

χ2 df Δχ2d χ2/df TLI IFI CFI RMSEA AIC

95.60 69 1.39 .98 ,99 ,98 .05 167.6095.44 68 .16 1.40 .98 .98 .98 .05 169.44

243.92 74 3.30 .88 .90 .90 .12 305.92

tive Fit Index; RMSEA: RootMean Square Error of Approximation; AIC: Akaike Information

Table 6SEM results for models 2a and 2b.

Model 2aa Model 2bb

Hypotheses Coefficientsc, d Coefficientsc, d

Attitude Entrepreneurial intention .34*** (.20) .32*** (.12)Subjective norm Entrepreneurial intention .27** (.14) .35*** (.11)Perceived behavioral control Entrepreneurial intention .41*** (.13) .34*** (.10)

H3a Quality: parental role models Attitude .23* (.15)H3b Quality: parental role models Subjective norm .27* (.22)H3c Quality: parental role models Perceived behavioral control .08 (.26)H4a Quality: work experience Attitude .02 (.15)H4b Quality: work experience Subjective norm .07 (.19)H4c Quality: work experience Perceived behavioral control .04 (.24)Fit measures

χ2 116.10 95.60df 69 69χ2 / df 1.68 1.39TLI .95 .98IFI .96 .99CFI .96 .98RMSEA .08 .05

a) n = 113.b) n = 173.c) Standardized coefficients with standard errors in parentheses.d) Significance levels: *: p ≤ .05; **: p ≤ .01; ***: p ≤ .001.Note: df: Degrees of Freedom; TLI: Tucker Lewis Index; IFI: Incremental Fit Index; CFI: Comparative Fit Index; RMSEA: Root Mean Square Error of Approximation.

648 F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

suggests that no significant group-specific differences in the relationsbetween perceived quality of prior entrepreneurial exposure, thethreemediating attitudinal variables of the TPB, and entrepreneurial in-tention exist. Hence, the results of hypotheses testing (based onmodels2a and 2b) apply to both groups.

As hypotheses testing based on large sample standard errors maylead to inaccurate results in our comparatively small samples in models2a and 2b (MacKinnon et al., 2002), we additionally employ thebootstrapping method (Preacher & Hayes, 2004, 2008) to substantiateourfindings. Bootstrapping refers to an empirical estimation of the sam-pling distribution of a statistic based onmultiple resamples drawn fromthe existing data. The resulting bootstrapping sampling distribution isthen used to generate p-values as well as confidence intervals and, inturn, test hypotheses (Efron & Tibshirani, 1993). Following recommen-dations by Shrout and Bolger (2002), we use 1000 bootstrap resamplesfrom the existing data to estimate direct and indirect effects. Resultsconfirm our previous findings as outlined above. Regarding perceivedquality of entrepreneurial role model exposure (model 2a), themean standardized indirect effect on entrepreneurial intention is .17(p ≤ .05), whereas the mean standardized direct effect is non-significant (.04; p = .75). In contrast, neither the indirect (.05; p =

Fig. 3. Results

.45) nor the direct (− .02; p = .68) effect of perceived quality of workexperience in a small or newly founded firm on entrepreneurial inten-tion is significant.

As an additional robustness check, we apply the more conservative(MacKinnon, Warsi, & Dwyer, 1995) Sobel test for mediation (Sobel,1982). The results support our findings as outlined above. The effect ofperceived quality of entrepreneurial role model exposure (model 2a)is mediated through attitude (p ≤ .1) and subjective norm (p ≤ .1)but not through perceived behavioral control (p = .47). The effect ofperceived quality of work experience in a small or newly founded firm(model 2b) is neither mediated through attitude (p = .75), orsubjective norm (p = .40) nor through perceived behavioral control(p = .58). Table 7 summarizes the results from hypotheses testing.

Lastly,we evaluate the predictive validity of all ourmodels as severalauthors (e.g., Armstrong, 2012; Gigerenzer & Brighton, 2009) empha-size that it is not sufficient to rely on fit indices as even good fittingmodels may lead to poor predictions. In order to avoid this potentialfallacy, we follow a recent recommendation by Woodside (2013) andcross-validate our models by randomly splitting each original samplein two samples (calibration and validation sample) using each sampleas a holdout sample to assess the predictive validity of the other sample.

model 2a.

Fig. 4. Results model 2b.

649F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

Consistent with the requirements to assess the cross-validity of struc-tural equation models (Byrne, 2004, 2010), we constrain each modelsstructural paths for each sub-sample. That is, the validation sample'sstructural paths are constrained to the estimated values of the calibra-tion model's structural paths. In order to assess each model's predictivevalidity, χ2 and CFI differences between the unconstrained andconstrained models are taken into account. Results of this procedurelend further support to the predictive validity of all our models, asnone of the χ2 differences between unconstrained and constrainedmodels are statistically significant (p N .1) and the differences in regardto CFI values are remarkably small (Δ CFI ≤ .006) and, hence, wellbelow the critical threshold of Δ CFI ≤ .01 (Cheung & Rensvold, 2002).

6. Discussion

The present study's overall aim is to disentangle the effect of priorentrepreneurial exposure on entrepreneurial intention to gain a betterunderstanding of how prior entrepreneurial exposure influences indi-viduals' intention to become an entrepreneur. To this end, we analyzehow different types of exposure (observational exposure to entrepre-neurial role models, work experience in a small or newly foundedfirm) and the perceivedquality of exposure influence individuals' entre-preneurial intention mediated through the attitudinal variables of theTPB (i.e., attitude, subjective norm, and perceived behavioral control).The stability of our results across groups (students and professionals)suggests a broad generalizability of our findings.

Table 7Summary of results from hypotheses testing.

Model Hypothesis Relationship

Model 1 H1a PEX: parental role modelsH1b PEX: parental role modelsH1c PEX: parental role modelsH2a PEX: work experienceH2b PEX: work experienceH2c PEX: work experience

Model 2a H3a Quality: parental role modelsH3b Quality: parental role modelsH3c Quality: parental role models

Model 2b H4a Quality: work experienceH4b Quality: work experienceH4c Quality: work experience

Note: PEX = Prior Entrepreneurial Exposure.

Our findings strongly confirm prior studies (e.g., Kolvereid, 1996;Tkachev & Kolvereid, 1999), which support the applicability of the TPBto explain entrepreneurial intention (with, however, only implicitly as-suming indirect effects of exogenous variables). Moreover, we find em-pirical support for our theoretical prediction that the three attitudinalvariables (attitude, subjective norm, and perceived behavioral control)of the TPB mediate the influence of prior entrepreneurial exposure onentrepreneurial intention. This finding is consistent with the formula-tion of the TPB (Ajzen, 1991) as exogenous influences (such as prior en-trepreneurial exposure) on intention are mediated through attitude,subjective norm, and perceived behavioral control. Furthermore, thisoutcome may explain why prior research on demographic characteris-tics of entrepreneurs testing direct-effects models instead of applyingindirect intention-based frameworks such as the TPB displays ratherinconclusive results (Shook et al., 2003). Based on our findings, weencourage future research to employ indirect (intention-based)modelsin order to reduce the inconclusive findings from direct effects models.One alternative framework in this regard may be social cognitive careertheory (SCCT) (Lent, Brown, & Hackett, 1994). Rather similar to the TPB,the SCCT regards individual's outcome expectations (which reflect TPB'sattitude) and self-efficacy (which is related to TPB's perceived behavior-al control) influential for individuals' career intentions. However, theSCCT also postulates reciprocal influences (i.e., higher self-efficacyleads to more positive outcome expectations regarding entrepreneurialbehavior), which are largely neglected in the TPB (Segal, Borgia, &Schoenfeld, 2002). As Liñán and Chen (2009) have shown thatindividual's subjective norm influences entrepreneurial intention

Finding

Attitude RejectedSubjective norm ConfirmedPerceived behavioral control RejectedAttitude ConfirmedSubjective norm RejectedPerceived behavioral control ConfirmedAttitude ConfirmedSubjective norm ConfirmedPerceived behavioral control RejectedAttitude RejectedSubjective norm RejectedPerceived behavioral control Rejected

650 F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

through attitude and perceived behavioral control, an additional avenuefor future researchmight be to delve deeper into the reciprocal relationsbetween attitudinal variables (Krueger, 2009).

Prior entrepreneurial exposure can occur through the observation ofrole models as well as through direct experience (Bandura, 1977;Latham & Saari, 1979). Moreover, individuals may perceive prior entre-preneurial exposure as positive or negative (Krueger, 1993). Opposedto conventional entrepreneurshipwisdom,we find hardly any significantconnection between entrepreneurial role model exposure (i.e., exposureto parents who previously started a business) and entrepreneurial inten-tion mediated through the attitudinal variables of the TPB. Exposure toparental role models only positively influences the subjective norm,whereas it has no significant effect on either attitude or perceived behav-ioral control in regard to starting a business. Hence, individuals withentrepreneurial parents perceive social pressure to start a business.However, they neither have a more positive attitude towards starting abusiness, nor do they feel more capable of successfully performing thetasks related to starting a business compared to individuals withoutentrepreneurial parents. Only when individuals perceive parental rolemodel exposure as positive, they develop a more positive attitude to-wards starting a business.We suggest two explanationswhy our analysisdoes not support prior direct studies (e.g., Crant, 1996; Matthews &Moser, 1995), which emphasize a positive influence of entrepreneurialrole models on entrepreneurial intention. First, prior research analyzingdirect effect models often employs samples of existing business owners,which may have simply taken over their parents' business. Hence, futureresearch should further explore how founderswho intend starting a newbusiness differ from successors (Zellweger, Sieger, & Halter, 2011). Sec-ond, prior research mostly neglects to take into account how (positiveor negative) individuals actually perceived the parental role model expo-sure. We extend prior literature by demonstrating that only prior rolemodel exposure perceived as positive positively affects individuals' atti-tude towards starting a business. This finding extends a study by Kimet al. (2006) who suggest that prior founders are strongly discouragedfrom starting another business by negative experiences. Hence, our find-ings at least partly resolve ambiguous results of prior research analyzingthe (direct and unrated) effects of entrepreneurial role models onentrepreneurial intention.

However, in neither case (unrated or perceived as positive) does en-trepreneurial rolemodel exposure lead to a higher perceived behavioralcontrol in regard to starting a business. This finding casts doubts on theeffectiveness of observational learning from role models (as suggestedby Scherer, Adams, Carley et al. (1989), Scherer, Adams and Wiebe(1989)). Only direct experiences such as work experience in a small ornewly founded firm seem to convey the necessary tacit knowledgerelevant for starting a business. In other words, individuals, who wereexposed to entrepreneurial role models, are not able to transfer theobserved business knowledge to their own startup attempt. Future re-search should hence control for industry effects in role model relation-ships, as some knowledge observed from entrepreneurial role modelsmight be industry-specific (Kim et al., 2006) and hence not overly usefulin every startup attempt. Moreover, this finding also contradictsprevious research arguing that entrepreneurial parents provide theirchildren with financial, human, social, and other resources necessaryto successfully start a business (Aldrich & Cliff, 2003; Scott & Twomey,1988).

Besides exposure to entrepreneurial role models, we investigate theeffects of prior work experience in a small or newly founded firm on theattitudinal variables of the TPB. Our results suggest that such direct ex-perience leads to a more positive attitude in regard to starting a busi-ness. Moreover, prior work experience in a small or newly foundedfirm seems to convey tacit knowledge necessary for starting a businessas it elevates individuals' perceived behavioral control over the processof starting a business. Our findings extend an earlier study by Kim et al.(2006) who find that mainly current business ownership and manage-rial experience positively affect entry into nascent entrepreneurial

activities, whereas general work experience has no effect. Our resultsemphasize that small or newly founded firms also provide an environ-ment in which entrepreneurial learning through direct experiencesoccurs. Moreover, working for a small or newly founded firm alsoheightens individuals' attitude towards starting a business. The latterfinding empirically supports Dyer (1994) who emphasizes that priorwork experiences might expose individuals to an entrepreneurial life-style, which in turn leads to higher a level of entrepreneurial intention.In contrast, the perceived quality ofwork experience in a small or newlyfounded firm has no effect on any of the attitudinal variables precedingentrepreneurial intention. Hence, the positive effects on attitude andperceived behavioral control occur through such (unrated) work expe-rience alone— it is not required that individuals perceive the exposureas positive.

7. Implications and limitations

7.1. Implications

Our results confirm that attitude towards starting a business,subjective norm, and perceived behavioral control in regard to startinga business are better predictors of entrepreneurial intention comparedto direct effects of demographic characteristics such as prior entrepre-neurial exposure. These three attitudinal variables depend (amongstothers) on one's upbringing, education, or prior experiences (Douglas& Shepherd, 2000). Hence, they can be altered by policy makers settingthe economic preconditions for entrepreneurship in general and in (en-trepreneurial) education and training programs in particular (Kolvereid& Isaksen, 2006). To have an impact on entrepreneurial intentions,initiatives promoting entrepreneurial activity must render starting abusiness both desirable (by changing behavioral beliefs relating to atti-tudes) and feasible (by changing control beliefs relating to perceivedbehavioral control) in the eyes of potential entrepreneurs (Gird &Bagraim, 2008; Krueger et al., 2000). However, as subjective norm isalso oneof the significant predictors of intention, itmay not be sufficientto change the behavioral and control beliefs of potential entrepreneurs.In fact, the approval of potential entrepreneurs' attachment figuresmust also be taken into consideration, when paving the way for entre-preneurship (Kolvereid & Isaksen, 2006).

For initiatives promoting entrepreneurship such as entrepreneurialeducation and training programs, it may be reasonable to screen andselect potential projects based on the attitude, subjective norm, andperceived behavioral control of potential entrepreneurs. However, asthese attitudinal variables are not directly observable, prior entrepre-neurial exposure may serve as a signal to identify promising entrepre-neurs (Krueger, 1993; Scherer, Adams, & Wiebe, 1989). However, ourresults suggest that one may not regard prior entrepreneurial exposureas unidimensional, but rather account for differences in terms of typeand perceived quality of prior exposure. Themere observation of entre-preneurial role models has only a very limited impact on the threeattitudinal variables preceding entrepreneurial intention per se. Onlyrole model exposure perceived as positive leads to a more positive atti-tude towards starting a business. However, role model exposure alonedoes not convey the necessary skills to potential entrepreneurs. Theperceived behavioral control over the process of starting a business isonly positively affected for individuals with prior direct experience(such as work experience in a small or newly founded firm). Our resultsimply that the integration of rolemodels in entrepreneurship educationand training programs– as recommended by several scholars (e.g., Scott& Twomey, 1988; van Auken et al., 2006) – has only a positive effecton attitude towards starting a business, if trainees perceive the exposureas positive. However, entrepreneurship trainees are only able tolearn relevant skills for starting a business by own start-up experiencesor internships. Learning by doing the necessary competencies gives in-dividuals a greater confidence in regard to starting their own business(Kolvereid & Isaksen, 2006). These findings call for entrepreneurship

651F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

education and training programs with complementary theoretical andactive elements.

7.2. Limitations

Like most empirical studies, our study has limitations. Althoughresearch focusing on entrepreneurial intentions is rather common inentrepreneurship research, employing intentions as dependent variableraises issuesworthmentioning. Themajor limitation of intention-basedresearch is that validity and stability of the link between intention andbehavior can only be directly observed by longitudinal research(Davidsson & Honig, 2003). However, prior studies emphasize thatthe TPB accurately predicts planned behavior in a variety of fields(e.g., Armitage & Conner, 2001). Despite such evidence from otherfields, entrepreneurship research would greatly profit from studies in-vestigating the link between intention and behavior in an entrepreneur-ial context (Fayolle & Liñán, 2014). One important step in this directionis a recent study by Kautonen, VanGelderen, and Tornikoski (2013). Theauthors use longitudinal data to demonstrate that the three attitudinalvariables of the TPB are significant predictors of entrepreneurial inten-tion, whereas entrepreneurial intention in turn is a significant predictorof entrepreneurial behavior. However, more work needs yet to be donein this regard as for example this study's results need to be replicatedwith larger samples or in different cultural contexts. Moreover, theindividuals in our sample were facing immediate career choices andwere asked to indicate their intention to start a business within thenext two years. This rather limited time span advocates for a moreaccurate intention-behavior link (Ajzen & Madden, 1986).

Second, onemay criticize themeasurement of our dependent variablein light of other (established) measures of entrepreneurial intentionexisting in the field. However, we believe that our entrepreneurial inten-tion measure is robust for several reasons: first, we extensively confirmthe validity and reliability of our measure consistent with recent recom-mendations (e.g., Mullen, Budeva, & Doney, 2009). Second, we examinethe robustness of our measure compared to alternative measures ofentrepreneurial intention developed in an entrepreneurial context(i.e., Kolvereid, 1996; Liñán & Chen, 2009) indicating significantly highpositive correlations. Third, prior research emphasizes that the TPB isremarkably robust to different specifications of the intention variable(Krueger, 2009), which is further supported by results from a recentmeta-analysis (Bae, Qian, Miao, & Fiet, 2014) suggesting non-significantdifferences between different entrepreneurial intention measures.Fourth, our measurement approach is consistent with the state-of-the-art recommendations on how to measure entrepreneurial intention byThompson (2009). Fifth, mixed measures of intention (combining inten-tions with desires and behavioral expectations such as our measure)work particularly well in a vocational context when individuals are notyet fully decided between career alternatives as particularly behavioralexpectancies implicitly consider the choice between competing alterna-tive behaviors (Kautonen et al., 2013; van Gelderen et al., 2008).Sixth, our dependent variable uses a rather specific reference point(i.e., “starting a business”) compared to competing measures fromentrepreneurship researchers using “becoming an entrepreneur” asreference point. However, the term “entrepreneur” is rather vague andhas been differently interpreted and operationalized by prior research(Thompson, 2009).

Another potential limitation is that our sample consists mainly ofGerman individuals. Thus, our results are at least partly dependent onthe cultural and economic conditions in Germany and thus may not beuniversally applicable.

References

Adler, Nancy J. (1983). A typology of management studies involving culture. Journal ofInternational Business Studies, 14(2), 29–47.

Ajzen, Icek (1988). Attitudes, personality and behavior. Chicago, Illinois: Dorsey Press.

Ajzen, Icek (1991). The theory of planned behavior. Organizational Behavior & HumanDecision Processes, 50(2), 179–211.

Ajzen, Icek (2002). Perceived behavioral control, self-efficacy, locus of control, and thetheory of planned behavior. Journal of Applied Psychology, 32(4), 665–683.

Ajzen, Icek (2006). Constructing a TpB questionnaire: Conceptual and methodologicalconsiderations. Retrieved 15.09.2013, from. http://www.uni-bielefeld.de/ikg/zick/ajzen%20construction%20a%20tpb%20questionnaire.pdf

Ajzen, Icek, & Madden, Thomas J. (1986). Prediction of goal-directed behavior: Attitudes,intentions, and perceived behavioral control. Journal of Experimental SocialPsychology, 22(5), 453–474.

Aldrich, Howard E., & Cliff, Jennifer E. (2003). The pervasive effects of family on entrepre-neurship: Toward a family embeddedness perspective. Journal of Business Venturing,18(5), 573–596.

Allison, Paul D. (1999). Multiple regression: A primer. Thousand Oaks: Pine Forge Press.Anderson, James C., & Gerbing, David W. (1988). Structural equation modeling in prac-

tice: A review and recommended two-step approach. Psychological Bulletin, 103(3),411–423.

Anderson, David Ray, Sweeney, Dennis J., & Williams, Thomas Arthur (2002). Statistics forbusiness and economics. South-Western: Mason.

Armitage, Christopher J., & Conner, Mark (2001). Efficacy of the theory of plannedbehaviour: A meta-analytic review. British Journal of Social Psychology, 40(4), 471–499.

Armstrong, J. Scott (2012). Illusions in regression analysis. International Journal ofForecasting, 28(3), 689–694.

Autio, Erkko, Keeley, Robert H., Klofsten, Magnus, Parker, George G. C., & Hay, Michael(2001). Entrepreneurial intent among students in Scandinavia and in the USA.Enterprise & Innovation Management Studies, 2(2), 145–160.

Bae, Tae Jun, Qian, Shanshan, Miao, Chao, & Fiet, James O. (2014). The relationship be-tween entrepreneurship education and entrepreneurial intentions: A meta-analyticreview. Entrepreneurship: Theory and Practice, 38(2), 217–254, http://dx.doi.org/10.1111/etap.12095.

Bagozzi, Richard P., Baumgartner, Johann, & Yi, Youjae (1989). An investigation into therole of intentions as mediators of the attitude-behavior relationship. Journal ofEconomic Psychology, 10(1), 35–62.

Bandura, Albert (1977). Social learning theory. Englewood Cliffs, NJ: Prentice-Hall.Bandura, Albert (1986). Social foundations of thought and action: a social cognitive theory.

Engelwood Cliffs, NJ: Prentice-Hall.Barling, Julian, Dupre, Kathryne E., & Hepburn, C. Gail (1998). Effects of parents' job inse-

curity on children's work beliefs and attitudes. Journal of Applied Psychology, 83(1),112–118.

Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in socialpsychological research: Conceptual, strategic, and statistical considerations. Journal ofPersonality and Social Psychology, 51(6), 1173–1182.

Bird, Barbara (1988). Implementing entrepreneurial ideas: The case for intention.Academy of Management Review, 13(3), 442–453.

Brenner, O. C., Pringle, Charles D., & Greenhaus, Jeffrey H. (1991). Perceived fulfillment of or-ganizational employment versus entrepreneurship: Work values and career intentionsof business college graduates. Journal of Small Business Management, 29(3), 62–74.

Byrne, Barbara M. (2004). Testing for multigroup invariance using AMOS graphics: A roadless traveled. Structural Equation Modeling, 11(2), 272–300.

Byrne, Barbara M. (2010). Structural equation modeling with AMOS: Basic concepts,applications, and programming (2nd ed.). New York: Routledge.

Carr, Jon C., & Sequeira, Jennifer M. (2007). Prior family business exposure as intergener-ational influence and entrepreneurial intent: A theory of planned behavior approach.Journal of Business Research, 60(10), 1090–1098.

Chen, Chao C., Greene, Patricia Gene, & Crick, Ann (1998). Does entrepreneurial self-efficacy distinguish entrepreneurs from managers? Journal of Business Venturing,13(4), 295–316.

Cheung, Gordon W., & Rensvold, Roger B. (2002). Evaluating goodness-of-fit indexes fortesting measurement invariance. Structural Equation Modeling, 9(2), 233–255.

Chlosta, Simone, Patzelt, Holger, Klein, Sabine B., & Dormann, Christian (2012). Parentalrole models and the decision to become self-employed: The moderating effect ofpersonality. Small Business Economics, 38(1), 121–138.

Crant, J. Michael (1996). The proactive personality scale as a predictor of entrepreneurialintentions. Journal of Small Business Management, 34(3), 42–49.

Crook, T. Russell, Shook, Christopher L., Morris, M. Lane, & Madden, Timothy M. (2010).Are we there yet? An assessment of research design and construct measurementpractices in entrepreneurship research. Organizational Research Methods, 13(1),192–206.

Davidsson, Per (1995). Determinants of entrepreneurial intentions. Paper presented at theRENT XI Workshop, Piacenza, Italy.

Davidsson, Per, & Honig, Benson (2003). The role of social and human capital amongnascent entrepreneurs. Journal of Business Venturing, 18(3), 301–331.

Diamantopoulos, Adamantios, & Siguaw, Judy A. (2006). Formative versus reflectiveindicators in organizational measure development: A comparison and empiricalillustration. British Journal of Management, 17(4), 263–282.

Douglas, Evan J., & Shepherd, Dean A. (2000). Entrepreneurship as a utility maximizingresponse. Journal of Business Venturing, 15(3), 231–251.

Dunn, Thomas, & Holtz-Eakin, Douglas (2000). Financial capital, human capital, and thetransition to self-employment: Evidence from intergenerational links. Journal ofLabor Economics, 18(2), 282–305.

Dyer, W. Gibb (1994). Toward a theory of entrepreneurial careers. Entrepreneurship:Theory & Practice, 19(2), 7–21.

Efron, Bradley, & Tibshirani, Robert (1993). An introduction to the bootstrap. New York:Chapman & Hall/CRC.

Eisingerich, Andreas B., & Rubera, Gaia (2010). Drivers of brand commitment: A cross-national investigation. Journal of International Marketing, 18(2), 64–79.

652 F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

Fairlie, Robert W., & Robb, Alicia (2007). Families, human capital, and small business: Ev-idence from the characteristics of business owners survey. Industrial & Labor RelationsReview, 60(2), 225–245.

Fayolle, Alain (2005). Evaluation of entrepreneurship education: behaviour performing orintention increasing? International Journal of Entrepreneurship and Small Business,2(1), 89–98.

Fayolle, Alain, & Liñán, Francisco (2014). The future of research on entrepreneurial inten-tions. Journal of Business Research, 67(5), 663–666.

Ford, J. K., MacCallum, R. C., & Tait, M. (1986). The application of exploratory factor anal-ysis in applied psychology: A critical review and analysis. Personnel Psychology, 39(2),291–314.

Fornell, Claes, & Larcker, David F. (1981). Evaluating structural equation models with un-observable variables and measurement error. Journal of Marketing Research, 18(1),39–50.

Gibb, Allan (2002). Creating conducive environments for learning and entrepreneurship:Living with, dealing with, creating and enjoying uncertainty and complexity. Industryand Higher Education, 16(3), 135–148.

Gigerenzer, Gerd, & Brighton, Henry (2009). Homo heuristicus: Why biased minds makebetter inferences. Topics in Cognitive Science, 1(1), 107–143.

Gird, Anthony, & Bagraim, Jeffrey J. (2008). The theory of planned behaviour as predictorof entrepreneurial intent amongst final-year university students. South African Journalof Psychology, 38(4), 711–724.

Guerrero, Maribel, Rialp, Josep, & Urbano, David (2008). The impact of desirability andfeasibility on entrepreneurial intentions: A structural equation model. InternationalEntrepreneurship and Management Journal, 4(1), 35–50.

Hu, Li-tze, & Bentler, PeterM. (1999). Cutoff criteria for fit indexes in covariance structureanalysis: Conventional criteria versus new alternatives. Structural Equation Modeling:A Multidisciplinary Journal, 6(1), 1–55.

Iacobucci, Dawn, Saldanha, Neela, & Deng, Xiaoyan (2007). A meditation on mediation:Evidence that structural equations models perform better than regressions. Journalof Consumer Psychology, 17(2), 139–153.

James, Lawrence R., Mulaik, Stanley A., & Brett, Jeanne M. (2006). A tale of two methods.Organizational Research Methods, 9(2), 233–244.

Katz, Jerome (1992). A psychosocial cognitive model of employment status choice.Entrepreneurship: Theory & Practice, 17(1), 29–37.

Kautonen, Teemu, Luoto, Seppo, & Tornikoski, Erno T. (2010). Influence of work historyon entrepreneurial intentions in ‘prime age’ and ‘third age’: A preliminary study.International Small Business Journal, 28(6), 583–601.

Kautonen, Teemu, Van Gelderen, Marco, & Tornikoski, Erno T. (2013). Predicting entre-preneurial behaviour: A test of the theory of planned behaviour. Applied Economics,45(6), 697–707.

Kim, Phillip H., Aldrich, Howard E., & Keister, Lisa A. (2006). Access (Not) denied: Theimpact of financial, human, and cultural capital on entrepreneurial entry in theUnited States. Small Business Economics, 27(1), 5–22.

Kim, Min-Sun, & Hunter, Kohn E. (1993). Relationships among attitudes, behavioralintentions, and behavior. Communication Research, 20(3), 331–364.

Kline, R. B. (1998). Principles and practice of structural equation modeling. New York:Guilford.

Kolvereid, Lars (1996). Prediction of employment status choice intentions.Entrepreneurship: Theory & Practice, 21(1), 47–57.

Kolvereid, Lars, & Isaksen, Espen (2006). New business start-up and subsequent entryinto self-employment. Journal of Business Venturing, 21(6), 866–885.

Krueger, Norris F. (1993). The impact of prior entrepreneurial exposure on perceptions ofnew venture feasibility and desirability. Entrepreneurship: Theory & Practice, 18(1), 5–21.

Krueger, Norris F. (2009). Entrepreneurial intentions are dead: Long live entrepreneurialintentions. In A. L. Carsrud, & M. Brännback (Eds.), Understanding the entrepreneurialmind (pp. 51–72). Springer.

Krueger, Norris F., & Carsrud, Alan L. (1993). Entrepreneurial intentions: Applying thetheory of planned behaviour. Entrepreneurship & Regional Development, 5(4),315–330.

Krueger, Norris F., Reilly, Michael D., & Carsrud, Alan L. (2000). Competing models ofentrepreneurial intentions. Journal of Business Venturing, 15(5/6), 411–432.

Krumboltz, John D., Mitchell, A.M., & Jones, G. B. (1976). A social learning theory of careerselection. The Counseling Psychologist, 6(1), 71–81.

Latham, Gary P., & Saari, Lise M. (1979). Application of social-learning theory to trainingsupervisors through behavioral modeling. Journal of Applied Psychology, 64(3),239–246.

Lent, RobertW., Brown, Steven D., & Hackett, Gail (1994). Toward a unifying social cogni-tive theory of career and academic interest, choice, and performance. Journal ofVocational Behavior, 45(1), 79–122.

Liñán, Francisco, & Chen, Yi-Wen (2009). Development and cross-cultural application of aspecific instrument to measure entrepreneurial intentions. Entrepreneurship: Theory& Practice, 33(3), 593–617.

MacKinnon, David P., Lockwood, Chondra M., Hoffman, Jeanne M., West, Stephen G., &Sheets, Virgil (2002). A comparison of methods to test mediation and other interven-ing variable effects. Psychological Methods, 7(1), 83–104.

MacKinnon, David P., Warsi, Ghulam, & Dwyer, James H. (1995). A simulation study ofmediated effect measures. Multivariate Behavioral Research, 30(1), 41–62.

MacMillan, I. C. (1986). To really learn about entrepreneurship, let's study habitual entre-preneurs. Journal of Business Venturing, 1(3), 241–243.

Matthews, Charles H., & Moser, Steven B. (1995). Family background and gender:Implications for interest in small firm ownership. Entrepreneurship & RegionalDevelopment, 7(4), 365–377.

Matthews, Charles H., & Moser, Steven B. (1996). A longitudinal investigation of the im-pact of family background and gender on interest in small firm ownership. Journalof Small Business Management, 34(2), 29–43.

Mitchell, Lynda K., & Krumboltz, John D. (1984). Social learning approach to career deci-sion making: Krumboltz's theory. In D. Brown, & L. Brooks (Eds.), Career choice anddevelopment (pp. 235–280). San Francisco: Jossey-Bass.

Mueller, P. (2006). Entrepreneurship in the region: Breeding ground for nascententrepreneurs? Small Business Economics, 27(1), 41–58.

Mulaik, Stanley A. (2001). The curve-fitting problem: An objectivist view. Philosophy ofScience, 68(2), 218–241, http://dx.doi.org/10.2307/3081065.

Mullen, Michael R., Budeva, Desislava G., & Doney, Patricia M. (2009). Research methodsin the leading small business–entrepreneurship journals: A critical review withrecommendations for future research. Journal of Small Business Management, 47(3),287–307, http://dx.doi.org/10.1111/j.1540-627X.2009.00272.x.

Mungai, Edward, & Velamuri, S. Ramakrishna (2011). Parental entrepreneurial rolemodelinfluence on male offspring: Is it always positive and when does it occur?Entrepreneurship: Theory and Practice, 35(2), 337–357.

Nabi, Ghulam, Holden, Rick, & Walmsley, Andreas (2006). Graduate career-making andbusiness start-up: A literature review. Education + Training, 48(5), 373–385.

Nunnally, J. C. (1978). Psychometric theory. New York: McGraw-Hill.Pallone, Nathaniel J., Rickard, Fred S., & Hurley, Robert B. (1970). Key influences of occu-

pational preference among black youth. Journal of Counseling Psychology, 17(6),498–501.

Parker, S.C. (2004). The economics of self-employment and entrepreneurship. Cambridge:Cambridge University Press.

Peterman, Nicole E., & Kennedy, Jessica (2003). Enterprise education: Influencingstudents' perceptions of entrepreneurship. Entrepreneurship: Theory & Practice,28(2), 129–144.

Podsakoff, Philip M., MacKenzie, Scott B., Lee, Jeong-Yeon, & Podsakoff, Nathan P. (2003).Commonmethod biases in behavioral research: A critical review of the literature andrecommended remedies. Journal of Applied Psychology, 88(5), 879–903.

Podsakoff, Philip M., & Organ, Dennis W. (1986). Self-reports in organizational research:Problems and prospects. Journal of Management, 12(4), 531–545.

Preacher, Kristopher J., & Hayes, Andrew F. (2004). SPSS and SAS procedures for estimatingindirect effects in simple mediation models. Behavior Research Methods, Instruments, &Computers, 36(4), 717–731.

Preacher, Kristopher J., & Hayes, Andrew F. (2008). Asymptotic and resampling strategiesfor assessing and comparing indirect effects in multiple mediator models. BehaviorResearch Methods, 40(3), 879–891.

Robinson, Peter B., Huefner, Jonathan C., & Hunt, H. Keith (1991). Entrepreneurial re-search on student subjects does not generalize to real world entrepreneurs. Journalof Small Business Management, 29(2), 42–50.

Rotefoss, Beate, & Kolvereid, Lars (2005). Aspiring, nascent and fledgling entrepreneurs:an investigation of the business start-up process. Entrepreneurship & RegionalDevelopment, 17(2), 109–127.

Scherer, Robert F., Adams, Janet S., Carley, Susan S., & Wiebe, Frank A. (1989). Role modelperformance effects on development of entrepreneurial career preference.Entrepreneurship: Theory & Practice, 13(3), 53–71.

Scherer, Robert F., Adams, Janet S., & Wiebe, Frank A. (1989). Developing entrepreneurialbehaviours: A social learning theory perspective. Journal of Organizational ChangeManagement, 2(3), 16–27.

Scherer, Robert F., Brodzinski, James D., &Wiebe, Frank A. (1991). Examining the relation-ship between personality and entrepreneurial career preference. Entrepreneurship &Regional Development, 3(2), 195–206.

Schlaegel, C., & Koenig, M. (2014). Determinants of entrepreneurial intent: A meta-analytic test and integration of competing models. Entrepreneurship: Theory &Practice, 38(2), 291–332.

Schröder, Elke, & Schmitt-Rodermund, Eva (2006). Crystallizing enterprising interestsamong adolescents through a career development program: The role of personalityand family background. Journal of Vocational Behavior, 69(3), 494–509.

Scott, Michael G., & Twomey, Daniel F. (1988). The long-term supply of entrepreneurs:students' career aspirations in relation to entrepreneurship. Journal of SmallBusiness Management, 26(4), 5–13.

Segal, Gerry, Borgia, Dan, & Schoenfeld, Jerry (2002). Using social cognitive career theoryto predict self-employment goals. New England Journal of Entrepreneurship, 5(2),47–56.

Shapero, Albert, & Sokol, L. (1982). The social dimensions of entrepreneurship. In C. Kent,D. L. Sexton, & K. Vesper (Eds.), The encyclopedia of entrepreneurship (pp. 72–90).Englewood Cliffs: Prentice-Hall.

Sheppard, Blair H., Hartwick, Jon, & Warshaw, Paul R. (1988). The theory of reasoned ac-tion: A meta-analysis of past research with recommendations for modifications andfuture research. Journal of Consumer Research, 15(3), 325–343.

Shook, Christopher L., Priem, Richard L., & McGee, Jeffrey E. (2003). Venture creation andthe enterprising individual: A review and synthesis. Journal of Management, 29(3),379–399.

Short, Jeremy C., Ketchen, David J., Combs, James G., & Ireland, R. Duane (2010). Researchmethods in entrepreneurship: Opportunities and challenges. Organizational ResearchMethods, 13(1), 6–15, http://dx.doi.org/10.1177/1094428109342448.

Shrout, Patrick E., & Bolger, Niall (2002). Mediation in experimental and nonexperi-mental studies: New procedures and recommendations. Psychological Methods,7(4), 422.

Singh, Jagdip (1995). Measurement issues in cross-national research. Journal ofInternational Business Studies, 26(3), 597–619, http://dx.doi.org/10.2307/155563.

Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equa-tion models. In S. Leinhardt (Ed.), Sociological methodology 1982 (pp. 290–312).Washington: American Sociological Association.

Solesvik, Marina Z., Westhead, Paul, Kolvereid, Lars, & Matlay, Harry (2012). Student in-tentions to become self-employed: The Ukrainian context. Journal of Small Businessand Enterprise Development, 19(3), 441–460.

653F.B. Zapkau et al. / Journal of Business Research 68 (2015) 639–653

Steenkamp, Jan-Benedict E. M., & Baumgartner, Hans (1998). Assessing measurement in-variance in cross‐national consumer research. Journal of Consumer Research, 25(1),78–107, http://dx.doi.org/10.1086/209528.

Stuart, Robert W., & Abetti, Pier A. (1990). Impact of entrepreneurial and managementexperience on early performance. Journal of Business Venturing, 5(3), 151–162.

Sutton, Stephen (1998). Predicting and explaining intentions and behavior: How well arewe doing? Journal of Applied Social Psychology, 28(15), 1317–1338.

Thompson, Edmund R. (2009). Individual entrepreneurial intent: Construct clarificationand development of an internationally reliable metric. Entrepreneurship: Theory andPractice, 33(3), 669–694.

Tkachev, Alexei, & Kolvereid, Lars (1999). Self-employment intentions among Russianstudents. Entrepreneurship & Regional Development, 11(3), 269–280.

van Auken, Howard, Fry, Fred L., & Stephens, Paul (2006). The influence of role models onentrepreneurial intentions. Journal of Developmental Entrepreneurship, 11(2), 157–167.

van Gelderen, Marco, Brand, Maryse, van Praag, Mirjam, Bodewes, Wynand, Poutsma,Eric, & van Gils, Anita (2008). Explaining entrepreneurial intentions by means ofthe theory of planned behavior. Career Development International, 13(6), 538–559.

Wang, Catherine L. (2008). Entrepreneurial orientation, learning orientation, and firmperformance. Entrepreneurship: Theory and Practice, 32(4), 635–657.

Wilson, Fiona, Kickul, Jill, & Marlino, Deborah (2007). Gender, entrepreneurialself-efficacy, and entrepreneurial career intentions: Implications for entrepreneur-ship education. Entrepreneurship: Theory and Practice, 31(3), 387–406.

Woodside, Arch G. (2013). Moving beyond multiple regression analysis to algo-rithms: Calling for adoption of a paradigm shift from symmetric to asymmetricthinking in data analysis and crafting theory. Journal of Business Research,66(4), 463–472.

Yun-Jeong, Shin, & Kelly, Kevin R. (2013). Cross-cultural comparison of the effects of op-timism, intrinsic motivation, and family relations on vocational identity. CareerDevelopment Quarterly, 61(2), 141–160.

Zellweger, Thomas, Sieger, Philipp, & Halter, Frank (2011). Should I stay or should I go?Career choice intentions of students with family business background. Journal ofBusiness Venturing, 26(5), 521–536.

Zhao, Xinshu, Lynch, John G., & Chen, Qimei (2010). Reconsidering Baron and Kenny:Myths and truths about mediation analysis. Journal of Consumer Research, 37(2),197–206.

Zhao, Hao, Seibert, Scott E., & Hills, Gerald E. (2005). The mediating role of self-efficacy inthe development of entrepreneurial intentions. Journal of Applied Psychology, 90(6),1265–1272.