The Effects of Entrepreneurship Education

Transcription

1 The Effects of Etrepreeurship Educatio Georg Vo Graeveitz, Dietmar Harhoff, Richard Weber To cite this versio: Georg Vo Graeveitz, Dietmar Harhoff, Richard Weber. The Effects of Etrepreeurship Educatio. Joural of Ecoomic Behavior ad Orgaizatio, Elsevier, 2010, 76 (1), pp.90. < /j.jebo >. <hal > HAL Id: hal Submitted o 2 Sep 2013 HAL is a multi-discipliary ope access archive for the deposit ad dissemiatio of scietific research documets, whether they are published or ot. The documets may come from teachig ad research istitutios i Frace or abroad, or from public or private research ceters. L archive ouverte pluridiscipliaire HAL, est destiée au dépôt et à la diffusio de documets scietifiques de iveau recherche, publiés ou o, émaat des établissemets d eseigemet et de recherche fraçais ou étragers, des laboratoires publics ou privés.

2 Title: The Effects of Etrepreeurship Educatio Authors: Georg vo Graeveitz, Dietmar Harhoff, Richard Weber PII: S (10) DOI: doi: /j.jebo Referece: JEBO 2562 To appear i: Joural of Ecoomic Behavior & Orgaizatio Received date: Revised date: Accepted date: Please cite this article as: vo Graeveitz, G., Harhoff, D., Weber, R., The Effects of Etrepreeurship Educatio, Joural of Ecoomic Behavior ad Orgaizatio (2010), doi: /j.jebo This is a PDF file of a uedited mauscript that has bee accepted for publicatio. As a service to our customers we are providig this early versio of the mauscript. The mauscript will udergo copyeditig, typesettig, ad review of the resultig proof before it is published i its fial form. Please ote that durig the productio process errors may be discovered which could affect the cotet, ad all legal disclaimers that apply to the joural pertai.

3 The Effects of Etrepreeurship Educatio Georg vo Graeveitz a,, Dietmar Harhoff a,c, Richard Weber b a Ludwig-Maximilias-Uiversität, LMU Muich School of Maagemet, INNO-tec, Kaulbachstraße 45, D-80539, Muich, Germay b Ludwig-Maximilias-Uiversität, LMU EC (Etrepreeurship Ceter), Geschwister-Scholl-Platz 1, D-80539, Muich, Germay c Cetre for Ecoomic Policy Research (CEPR), Lodo Abstract Etrepreeurship educatio raks high o policy agedas i Europe ad the US, but little research is available to assess its impact. To help close this gap we ivestigate whether etrepreeurship educatio affects itetios to be etrepreeurial uiformly or whether it leads to greater sortig of studets. The latter ca reduce the average itetio to be etrepreeurial ad yet be socially beeficial. This paper provides a model of learig i which etrepreeurship educatio geerates sigals to studets. Drawig o the sigals, studets evaluate their aptitude for etrepreeurial tasks. The model is tested usig data from a compulsory etrepreeurship course. Usig ex ate ad ex post survey resposes from studets, we fid that itetios to foud declie somewhat although the course has sigificat positive effects o studets self-assessed etrepreeurial skills. The empirical aalysis supports the hypothesis that studets receive iformative sigals ad lear about their etrepreeurial aptitude. We outlie implicatios for educators ad public policy. JEL Classificatio: D83, J24, L26, M13 Keywords: etrepreeurship, etrepreeurship, educatio, Bayes Rule, learig, sigals Ackowledgemets: We would like to thak two aoymous referees for helpful commets. We also thak the participats of the 2009 SFB TR 15 Coferece i Tutzig ad the 2009 Max Plack Aual Summit Coferece o Experimetal Etrepreeurship for commets ad suggestios. Particular thaks go to Lotta Väääe ad Michael Fritsch for detailed suggestios for improvemets. Dietmar Harhoff ad Georg vo Graeveitz ackowledge the support of the SFB Trasregio 15. The usual caveat applies. Correspodig author addresses: graeveitz@lmu.de (Georg vo Graeveitz ), harhoff@lmu.de (Dietmar Harhoff), richard.weber@lmu.de ( Richard Weber) Page 1 of 39

4 1. Itroductio Policy-makers geerally cosider ew veture formatio to be istrumetal for ecoomic growth ad techological progress. The academic literature is less uaimous. While may studies fid positive effects (Reyolds et al., 1994; Sheshiski et al., 2007; Audretsch ad Fritsch, 2002), others cautio that etrepreeurship i geeral may ot create growth i idustrialized coutries. However, may scholars cocede that particular forms of etrepreeurship are associated with sizable positive effects (Sterberg ad Weekers, 2005; va Stel et al., 2005; Wog et al., 2005; Acs ad Szerb, 2007). There appears to be some agreemet o the role of start-ups fouded by uiversity graduates ad faculty. Academic etrepreeurs are likely to employ more people tha their o-academic couterparts (Dietrich, 1999), ad fouders with uiversity educatio apparetly make higher ivestmets i their busiess tha o-academic etrepreeurs (Reyolds et al., 1994). Moreover, their firms perform disproportioately better (Shae, 2004). Uiversity spi-offs also create importat spill-over effects for the regioal ecoomy (Harhoff, 1999; Shae, 2004). I awareess of these fidigs, some policy-makers declare the sesitizatio ad advacemet of potetial fouders at tertiary educatioal istitutios a primary goal of iovatio policies. As a result, a wide rage of etrepreeurship educatio efforts have bee iitiated (Fayolle, 2000; Liñá, 2004; Kuratko, 2005). The effects emaatig from etrepreeurship educatio are still poorly uderstood. Several previous studies fid a positive impact of etrepreeurship educatio courses or programs at uiversities o perceived attractiveess ad feasibility of ew veture iitiatio or eve o actual startup activity (Tkachev ad Kolvereid, 1999; Peterma ad Keedy, 2003; Fayolle et al., 2006; Souitaris et al., 2007). Other studies fid evidece that the effects are egative (Oosterbeek et al., 2010). There may be methodological reasos why the literature has ot geerated cosistet assessmets as of yet. While the studies provide itriguig results, may of them ted to have methodological limitatios. For example, few studies employ a pre-post desig, ad eve fewer ivolve a cotrol group. Most of the studies focus o self-selected participats with some existig predispositio towards etrepreeurship which is likely to bias results i favor of educatioal itervetios (Gorma et al., 1997). But methodological weakesses may ot be the oly problem. Curretly, there is also o agreemet o what would costitute a suitable coceptual model for the aalysis of effects of etrepreeurship educatio. Most studies measure the impact of etrepreeurship educatio merely by searchig for uiform course-iduced chages i etrepreeurial itetios. We argue that this approach may be misleadig because it masks importat sortig effects which ca be socially positive eve if etrepreeurial itetios declie as a cosequece of etrepreeurship traiig. Our paper seeks to make three cotributios. First, we develop a ew coceptual perspective which emphasizes learig ad discovery of oe s ow aptitude for etrepreeurship. I this view, etrepreeurship educatio allows studets to better assess whether they should pursue a etrepreeurial career. This view differs sigificatly from the implicit otio that etrepreeurship educatio somehow ehaces studets willigess to become etrepreeurs. We cast this ew perspective i a formal model where learig follows Bayes Rule. 1 Usig this 1 Recet research by behavioral ecoomists demostrates that people do ot always update their beliefs accordig to Bayes Rule (Rabi ad Schrag, 1999; Charess ad Levi, 2005; Charess et al., 2007). However, the experimets udertake by Charess et al. (2007) demostrate that Bayes Rule describes learig behavior better if subjects update their beliefs after iteractio with people i larger 1 Page 2 of 39

5 model of updatig ad learig, we show that if studets differ i their aptitude for etrepreeurship ad if etrepreeurship educatio helps them ucover these differeces, etrepreeurship educatio may ot always lead to stroger etrepreeurial itetios o average. Thus, the ew coceptual perspective allows us to develop a possible explaatio why the results i the literature have bee somewhat cotradictory. As a secod cotributio, we study the effects of a large-scale compulsory etrepreeurship course at a major Germa uiversity, usig a pre-test post-test desig. We explore whether studets leared about their ow etrepreeurial aptitude from this course. A descriptive aalysis of studets itetios to become etrepreeurs shows that the course iduces sortig. Especially studets who are iitially ucertai about their etrepreeurial aptitude are able to determie more clearly whether or ot they are suited to etrepreeurship after the course. We test three hypotheses derived from our model. The results cofirm that learig about etrepreeurial aptitudes takes place, ad that it occurs as the model predicts. We ca also show that while average measures of attitudes do ot chage sigificatly, these attitudes do chage at the idividual level. These effects have ot foud ay attetio i the extat literature. I a third cotributio we address the welfare implicatios of our results for future evaluatio of etrepreeurship educatio. Give our coceptual model ad the empirical support provided by the data, we argue that a focus o startup propesity aloe is misleadig. I order to come to a comprehesive assessmet of etrepreeurship educatio, the gais from improved matchig betwee studets ad career paths eed to be cosidered. Iformig o-etrepreeurial idividuals that they are ot well-suited for startup activities may be as valuable as cofirmig ad stregtheig etrepreeurial tedecies i other studets. We argue that this is ot oly the coceptually appropriate approach, it is also ethically the right route to take. The remaider of the paper cotais six sectios. Next, we review the literature o etrepreeurship educatio ad its impact o etrepreeurial activity. I Sectio 3, we discuss a formal model of learig based o Bayesia updatig. Sectio 4 describes the settig of our study ad cotais a descriptive aalysis of the data. Sectio 5 provides a empirical test of the predictios from our model. Sectio 6 cocludes ad discusses future research. 2. Literature Review 2.1. Etrepreeurship as Itetioally Plaed Behavior Sice much of the literature relates to the theory of etrepreeurship as itetioal behavior (or variats of that theory), we summarize this theory briefly. Itetioality is a state of mid directig a perso s attetio (ad therefore experiece ad actio) toward a specific object (goal) or a path i order to achieve somethig (meas) (Bird, 1988). Ay plaed behavior is best predicted by observig itetios toward that behavior, ot by attitudes, beliefs, persoality or demographics (Bagozzi et al., 1989). Thus, accordig to the social psychology literature, itetios are the sigle best predictor of plaed behavior. This holds especially whe the target behavior is rare, hard to observe or whe it ivolves upredictable time lags (Ajze, 1991). Whe the target behavior affords a perso groups, which applies to the course settig we ivestigate here. 2 Page 3 of 39

6 complete cotrol over behavioral performace, itetios aloe should be sufficiet to predict behavior, as explaied i the theory of plaed behavior (Ajze, 1991). Itetios have bee foud to be a ubiased predictor of actio, eve where time lags exist, for example i career choices (Let et al., 1994). Hece, itetios predict behavior, while i tur certai specific attitudes predict itetio. Attitudes derive from exogeous iflueces (Ajze, 1987). Thus, itetios are idirectly affected by exogeous iflueces: either they drive attitudes or they moderate the relatioship betwee itetios ad behavior (i.e. facilitate or ihibit the realizatio of itetios). Ad itetios serve as a mediator or catalyst for actio: itetio-based models describe how exogeous iflueces chage itetios ad, i the ed, actual behavior. This is cofirmed by meta-aalytic studies (Kim ad Huter, 1993). Across a wide variety of target behaviors ad related itetios, attitudes explai over 50% of the variace i itetios, itetios i tur explai over 30% of the variace i behavior. This compares to 10% usually explaied by trait measures or attitudes aloe (Ajze, 1987). May researchers see etrepreeurship as a typical example of plaed itetioal behavior (Bird, 1988; Katz ad Garter, 1988; Krueger ad Brazeal, 1994). Havig a etrepreeurial itetio meas that oe is committed to startig a ew busiess (Krueger, 1993). The attitude towards etrepreeurship may be iflueced by educatioal measures. However, despite the recogitio that educatio ad prior etrepreeurial experieces may ifluece people s attitudes towards startig their ow busiess, the impact of etrepreeurship educatio, as distict from geeral educatio, o itetios towards etrepreeurship has remaied largely uexplored (Dockels, 1991; Krueger ad Brazeal, 1994) Research o Effects of Etrepreeurship Educatio Research about the effects of etrepreeurship educatio is still i its ifacy (Gorma et al., 1997). May studies to date simply describe etrepreeurship courses (Vesper ad Garter, 1997), discuss the cotet of good etrepreeurship educatio (Fiet, 2001) or evaluate the ecoomic impact of courses by comparig takers ad otakers (Chrisma, 1997). Some researchers have proposed a positive lik betwee etrepreeurship educatio ad etrepreeurial attitudes, itetio or actio, but the evidece is still ot strog (Gibb Dyer, 1994; Robiso et al., 1991; Krueger ad Brazeal, 1994). Some empirical studies do cofirm that there is a positive impact of etrepreeurship educatio courses or programs at uiversities o perceived attractiveess ad perceived feasibility of ew veture iitiatio (Tkachev ad Kolvereid, 1999; Fayolle et al., 2006). Reviews of literature o eterprise ad etrepreeurship educatio (Daiow, 1986; Gorma et al., 1997) ad of particular etrepreeurship programs (McMulla et al., 2002) give evidece that these programs ecourage etrepreeurs to start a busiess. But usually, there are serious methodological limitatios. For example, studies rarely ivolve cotrol groups or a form of stochastic matchig (Block ad Stumpf, 1992), basic cotrols as pre- ad post-testig are ot employed ad most studies survey participats with a existig predispositio towards etrepreeurship, biasig the results i favor of educatioal itervetios (Gorma et al., 1997). The studies by Peterma ad Keedy (2003), Souitaris et al. (2007) ad Oosterbeek et al. (2010) are three 3 Page 4 of 39

7 remarkable exceptios, usig pre-test-post-test cotrol group desigs. Peterma ad Keedy (2003) fid that exposure to eterprise educatio affected etrepreeurial itetios of high-school studets. Studets from 17 Australia schools participated i this study, with 109 participats i the treatmet ad 111 studets i the cotrol group. Iterestigly, the authors detect a differetial effect: participats with weak ex ate etrepreeurial propesities experieced a stroger positive treatmet effect tha participats with strog ex ate etrepreeurial itetios. Souitaris et al. (2007) fid that sesitizatio through a semester-log (Jauary-May) etrepreeurship program led to stroger etrepreeurial itetios. The authors employed a pre-test-post-test cotrol group-desig ad coducted their survey at two major Europea uiversities surveyig sciece ad egieerig studets. They received 124 matched questioaires from the program group ad 126 from the cotrol group. The studets of the program group took a etrepreeurship course as a elective module withi their curriculum. Hece, the allocatio of studets to the program group was ot fully radom, ad differet classes were taught by differet academic istructors so that the treatmet might have differed across classes. Oosterbeek et al. (2010) study the impact of etrepreeurship educatio i a compulsory course, usig a differece-i-differeces framework. Sice studets may have self-selected ito differet school locatios, locatio choice (ad thus treatmet) is istrumeted. Their results show that the effect o studets self-assessed etrepreeurial skills is isigificat. Moreover, the effect of the course o etrepreeurial itetios is sigificatly egative. While this literature has geerated iterestig isights, the research o effects of etrepreeurship educatio still has huge gaps. Usurprisigly, several researchers have called for more research to aswer the questio if etrepreeurship educatio ca ifluece etrepreeurial perceptios ad itetios (Dockels, 1991; Kator, 1988; Krueger ad Brazeal, 1994; McMulla et al., 2002). Descriptive ad retrospective studies are ot sufficiet to provide covicig evidece for the presumed effects (Alberti, 1999; Gorma et al., 1997; Matthews ad Moser, 1996). Peterma ad Keedy (2003) call for the developmet of credible methods of testig precoceived hypotheses, usig large sample sizes ad cotrol groups, i order to move this youg field of research beyod its exploratory stage (Alberti, 1999). Our paper is meat to cotribute to this ageda. But we ote a coceptual gap too - oe of the studies discussed here has attempted to ivestigate the ature of learig processes that occur i the course of etrepreeurship educatio. Studets are assumed to face o ucertaity regardig their ow skills ad iterests, while the process of learig itself is hardly modeled. 2 This is where our study deviates from its predecessors. We explore the possibility that studets eed to fid out above all whether etrepreeurial activity suits them, i.e., whether they have sufficietly high etrepreeurial aptitude to become etrepreeurs. Depedig o what they lear studets may adjust their etrepreeurial itetios upwards or dowwards Etrepreeurial Aptitude Etrepreeurship educatio may have several distict effects. First, etrepreeurship educatio is likely to ifluece kowledge ad skills. Courses at tertiary istitutios are usually orieted towards teachig of methods, 2 Recet work o learig uder certaity otes that this is geerally true of the literature o skill formatio i ecoomics (Cuha et al., 2006; Cuha ad Heckma, 2007). 4 Page 5 of 39

8 cocepts ad facts. We coceive of this compoet as reducig the cost of becomig a etrepreeur. However, the skills ad the kowledge taught will have geeric compoets (e.g., kowig how to write a busiess pla is also helpful i a established corporatio). Therefore, such educatio may ot shift etrepreeurial itetios by much. Noetheless, where etrepreeurship educatio also affects the attitudes ad perceptios of studets it may affect etrepreeurial itetios, ad thereby actios. Fially ad most importatly for our paper, etrepreeurship courses may allow studets to egage i etrepreeurial activity i a experimetal settig, e.g., by supportig fouders of startups i their actual day-to-day activities. This experiece may also be emulated i the classroom, e.g. i startup simulatios. We argue that this latter compoet will be particularly valuable i helpig studets reduce ucertaity as to how suitable a etrepreeurial career is for them persoally. The most importat effect of etrepreeurship educatio may therefore lie i studets adjustig ad refiig their assessmet of their ow etrepreeurial aptitude. This view implies that some studets lear that etrepreeurial careers are ot well-suited for them while the educatioal measures lead other participats to come to a more positive assessmet of their etrepreeurial aptitude. This perspective is radically differet from the oe that has domiated the literature. I that view, studets face o ucertaity about their ow type ad therefore do ot have to lear about their etrepreeurial aptitude. As a cosequece, a importat aspect of etrepreeurship educatio may have bee eglected. 3. Learig about Aptitude for Etrepreeurial Tasks I this sectio we discuss a formal model of learig about ow aptitude for etrepreeurial tasks. The model itself is set out i the appedix (Appedix A). Our startig poit is the assertio that most studets are ulikely to be very certai about their aptitude for etrepreeurial tasks. We model learig about etrepreeurial aptitude as a process of Bayesia updatig of beliefs based o sigals geerated before ad durig a etrepreeurship course. By geeratig these sigals the course affects studets discovery of their ow aptitude for etrepreeurial tasks. I the course of learig about etrepreeurship the studet may receive coflictig sigals ad differet studets will receive sigals of differet quality. Ay course o etrepreeurship ca therefore be expected to geerate three types of outcome: studets who lear othig ad studets who either lear ad discover that they like or dislike etrepreeurship. I such a cotext it is iterestig to characterize what effects a iformative course o etrepreeurship will have o the distributio of studets beliefs about their ow etrepreeurial aptitude. For istace a course that leaves the average belief about etrepreeurial aptitude uchaged but leads to greater polarizatio of beliefs may be couted a success, because it has more clearly separated those who are suited to etrepreeurship from those who are ot. Our pricipal iterest here is i characterizig the distributio of studets beliefs about ow aptitude for etrepreeurial tasks. We show how iformative sigals about etrepreeurial ability geerated before ad by etrepreeurship educatio determie such beliefs, allowig studets to lear their type. I our surveys we elicited the stregth of the sigals that studets have received. I other words we measure how sure a studet is that they either are or are ot a etrepreeur. A weak sigal is equivalet to a sigal that leaves 5 Page 6 of 39

9 the studet usure about their aptitude. This iformatio allows us to test for learig by exploitig variatio i the stregth of sigals that studets receive. The model we develop focuses o the cosequeces of such variatio i sigal stregth. Below we provide three formal results that ca be tested empirically: I We show that studets beliefs about their etrepreeurial aptitude display higher variace after a course, if ucertaity about etrepreeurial aptitude is sufficietly high before the course ad if the course provides iformatio. II We show that ay course which provides studets with iformative sigals about etrepreeurial ability will have two mai effects: i) it will leave studets who receive cosistet sigals before ad durig the course with stroger beliefs about their etrepreeurial aptitude tha studets who do ot receive cosistet sigals; ad ii) it will leave studets with stroger pre course sigals with stroger beliefs about their etrepreeurial aptitude. III We show that a iformative etrepreeurship course will chage the beliefs of participats less the stroger the pre course sigal, if sigals are cosistet ad sufficietly precise. While oe of these predictios is particularly surprisig the beefit of a formal model is to provide exact coditios uder which these seemigly obvious results obtai. We fid below that these coditios matter for the statistical tests we udertake. The formal setup of the model is relegated to the appedix. For the purposes of what follows here the followig assumptios we make should be oted. We assume studets are ucertai about their etrepreeurial aptitude but kow how this aptitude is distributed i the populatio aroud them. Their pre uiversity life ad a etrepreeurship course at uiversity provide studets with two idepedet sigals of their etrepreeurial ability. For simplicity we distiguish just betwee etrepreeurs ad employees. Beig (truly) a etrepreeur meas that oe s ow utility from beig i a etrepreeurial fuctio is greater tha the utility from beig i a employee fuctio. Coversely, we label employees all studets who are better suited to o-etrepreeurial work. The label employee is ot iteded to be pejorative. A importat fuctio of etrepreeurship educatio is to help studets self-select ito activities which they are most suited to. Our model shows whe this type of sortig is supported by etrepreeurship educatio. Beliefs after Etrepreeurship Educatio We begi with the most obvious implicatio of updatig of beliefs: If there are etrepreeurs ad employees i the populatio of studets, if these all receive iformative sigals, if etrepreeurs pre course sigals that they are etrepreeurs are ot too strog ad if studets update their beliefs accordig to Bayes Rule, the we ca show that: Propositio 1 The distributio of beliefs after the course will have greater variace tha the distributio of beliefs before, if sigals before are ot too strog ad if sigals are iformative. 6 Page 7 of 39

10 I the appedix we derive the expectatio ad the variace of studets beliefs that they are etrepreeurs before ad after the course. A compariso of these variaces shows that the variace of beliefs after the course is greater tha the variace of beliefs before, if the stregth of the pre course sigal is low ad if the course geerates a iformative sigal. A course that provides iformative sigals will raise the umber of studets who have leared somethig about their aptitude for etrepreeurial tasks. If studets also received iformative sigals before a course, the the sigal provided by the course will icrease sortig of studets ito two groups that are icreasigly sure that they are or are ot etrepreeurs. However, this will be less likely if studets already kow a lot about their etrepreeurial ability or if the sigal geerated by the course is very oisy. I Sectio 5 below we test the followig hypothesis: Hypothesis 1 The variace of beliefs after the course is greater tha the variace of beliefs before the course. Now cosider the effects of pre course sigals o the post course beliefs of etrepreeurs ad employees. Cosistecy of sigals will lead to stroger beliefs. Also, greater stregth of sigals received before the course will make beliefs after the course stroger. Propositio 2 If the two sigals received by studets are cosistet, the beliefs after the course will be stroger, tha if sigals are icosistet. Stroger pre course sigals lead to stroger beliefs after the course. We prove this Propositio i Appedix Appedix A.2. While this propositio may look trivial it is importat to ote that beliefs after the course will also be stroger for those studets who receive coflictig sigals. The hypothesis correspodig to Propositio 2 is: Hypothesis 2 i) If sigals are cosistet the beliefs after the course are stroger. ii) Stroger pre course sigals lead to stroger beliefs after the course. To test the hypothesis we regress a measure of strog pre course sigals (S FPS ) ad of cosistet sigals (CS ) o the variace of beliefs after the course ( B) aroud their mea. The depedet variable is defied such that stroger beliefs icrease the level of the depedet variable. It does ot matter whether the belief that oe is a etrepreeur is close to oe or close to zero. I both cases studets have strog beliefs ad i both cases the level of the depedet variable is high. Hypothesis 2 implies that the coefficiets o the measure of extreme sigals, the measure of cosistet sigals ad their iteractio are all positive. Our empirical model is: B = β 0 + β 1 CS + β 2 S FPS + β 3 CS X + β 4 X + ɛ, (1) where B (B[2] µ(b[2])) 2 captures the squared deviatio of studets beliefs after the course (B[2]) from the mea, CS is a measure of cosistet sigals, S FPS is a measure of the stregth of the pre course sigal ad CS X 7 Page 8 of 39

11 is the iteractio of the latter two variables. X represets a vector of cotrol variables. Hypothesis 2 predicts that β 1 > 0, β 2 > 0 ad β 3 > 0. The Chage i Beliefs after Etrepreeurship Educatio Fially, cosider chages i studets beliefs resultig from a etrepreeurship course. We fid that it is quite difficult to characterize the relatioship betwee the size of the chage i studets beliefs about their aptitude for etrepreeurial tasks ad the stregth of pre course sigals they receive. This is due to the fact that for ay give pre course sigal the studet ca receive several differet kids of sigals after the course ad we have to characterize all of these differet cases. If we assume that the sigalig process is iformative ad also reliable, the we ca derive a additioal predictio. We have already assumed that sigals are iformative above. Here we defie a reliable sigal such that studets have a probability greater tha 1/2 of receivig the correct sigal for their type. I such a settig there will be more studets with correct ad cosistet sigals tha studets with misleadig ad cosistet sigals. Give these assumptios we prove the followig additioal result: Propositio 3 If studets receive sufficietly precise ad reliable pre course sigals the those who receive cosistet sigals will chage their beliefs less as pre course sigals become stroger. We prove this result i Appedix Appedix A.2. Propositio 3 is tested with the followig hypothesis: Hypothesis 3 If studets receive cosistet sigals, the those amog them who have received stroger sigals before the course will chage their beliefs less. To test this hypothesis we regress the square of the chage i beliefs due to the course o a measure of the stregth of pre course sigals ad of cosistet sigals. We predict a egative coefficiet o the iteractio of strog ad cosistet sigals. The depedet variable is squared, sice our model makes predictios about the extet of a chage i beliefs, ot about their directio. The empirical model i this case is: = γ 0 + γ 1 CS + γ 2 S FPS + γ 3 CS X + γ 4 X + ɛ, (2) where 2 captures the squared chage i studets beliefs. The remaiig variables are defied as above. Hypothesis 3 predicts that γ 3 < 0. Propositio 3 is weaker tha Propositio 2. It relies o the additioal assumptio that the sigalig process is reliable. Additioally, it is weaker because our model predicts that i the couterfactual case i which studets receive icosistet sigals there are two groups with differet reactios to more precise pre course sigals. Our model predicts that these two groups will be of equal size, i which case these reactios cacel out i aggregate. I smaller populatios we may see deviatios from this predictio. 8 Page 9 of 39

12 4. Data Descriptio Before we test the hypotheses itroduced i the previous sectio, this sectio focuses o the data geerated by our surveys. First, we discuss data collectio i the cotext of a etrepreeurship course. I particular, we focus o the problem of sample selectio. Next, we characterize the effects of the etrepreeurship course o studets attitudes to etrepreeurship. It ca be show that the course we study is comparable i its effects to other courses studied i the literature, e.g. Oosterbeek et al. (2010). Fially, we provide descriptive evidece cosistet with the hypothesis that these studets leared about their etrepreeurial aptitude durig the course Data Collectio ad Sample Selectio The settig for data collectio is the Departmet of Busiess Admiistratio, i the Muich School of Maagemet, at Ludwig-Maximilias-Uiversität (LMU) Muich, oe of Germay s largest uiversities. The Bachelor curriculum at the Muich School of Maagemet is somewhat utypical due to its obligatory etrepreeurship educatio course Busiess Plaig. Every busiess admiistratio studet i the Bachelor of Sciece curriculum at LMU has to eroll i this course i the third semester of their study program. The objectives of the Busiess Plaig course are threefold: i) to teach studets basic capabilities eeded i the plaig ad maagemet of a startup eterprise, i particular to covey the ecessary kowledge ad skills for craftig a complete busiess pla; ii) to sesitize studets for etrepreeurship accordig to the classificatio by Liñá (2004): studets are supposed to acquire kowledge about small eterprises, self-employmet ad etrepreeurship so that they ca make ratioal career decisios; iii) to allow studets to gai practical experiece by iteractio with real-world etrepreeurs. The objectives of this course do ot ecompass ay otio of covicig studets to become etrepreeurs or to describe etrepreeurship as a particularly desirable optio. While the ecoomic importace of etrepreeurship is clearly sigaled, studets are ot meat to be idoctriated. The studets we survey took the course betwee October 2008 ad February They took part i eight lectures coveyig the priciples of busiess plaig. These lectures held by LMU faculty were augmeted with iput from experts o fiacial plaig, etrepreeurial marketig, experieced etrepreeurs ad ivestors. Thus studets obtaied first-had isight ito their busiesses. The studets also atteded small group tutorials. I these tutorials they practiced applicatio of busiess plaig cocepts ad developed their ow busiess pla. Feedback was give by fellow studets, a teachig assistat ad a tutor. Studets were surveyed (either usig a writte or a olie survey) directly before the kickoff sessio of the course ad immediately before the time whe the studets received their grades at the ed of the semester. The survey istrumets used had bee reviewed by three academics ad 12 o-participatig studets to esure clarity of wordig ad face validity of the costructs. Due to ethical cocers, we did ot eforce participatio i the two surveys. The two survey istrumets were largely idetical. However, the secod survey also cotaied items used i the course evaluatio. 3 The survey forms were aoymized i both rouds, ad matchig was achieved by 3 The survey forms are available upo request. 9 Page 10 of 39

13 employig a volutary structured idetificatio code. 4 Participatio i the Surveys ad Possible Selectio Biases. The settig of this course presets a particularly suitable framework for our study sice busiess studies studets do ot self-select ito the Busiess Plaig course. Moreover, give that studets iteract with real-world etrepreeurs we believe that they receive iformative ad importat sigals of their ow ability as etrepreeurs. While this settig allows us to elimiate the sample selectio issues that affect most previous work o etrepreeurship educatio we caot avoid sample selectio due to self selectio ito busiess studies. Therefore, it is desirable to repeat the survey we preset here i etrepreeurship courses that iclude a more diverse set of participats i the future. We proceed to ivestigate whether the patter of resposes ad o-resposes to our survey suggests ay additioal sources of sample selectio bias. I total we received resposes from 357 studets. They represet 97.8 percet of total erollmet i the Busiess Plaig course. 265 studets participated i the first ad 274 i the secod survey. For 196 studets we were able to match two survey resposes. While our research desig has the advatage that studets caot self-select ito the course itself, we may still face selectio issues due to differetial propesities to respod to our surveys. A first idicatio that we do ot face major (or possibly ay) selectio bias due to o-respose ca be take from Table 1. This provides demographic variables for three sets of respodets: those who oly respoded to the first survey, those who respoded to both surveys, ad those who oly respoded to the ex post survey. Participats to both surveys were sigificatly youger tha those who respoded to oly oe survey roud. This may reflect studets behavior - older studets are likely to feel more pressure to focus o their studies ad may therefore be less willig to waste time o surveys. Moreover, studets ot participatig i both surveys were more likely to have self-employed parets (i the pre- ad post-survey group) ad had higher itetios to foud (i the pre- ad post-survey group). However, it is importat to ote that while the itetio to foud declies sigificatly withi the group aswerig the survey ex ate ad ex post, the itetio to foud is ot sigificatly differet, ex ate or ex post, for those who oly respoded oce ad those who respoded both times. Table 1: Demographic Characteristics ad Sample Compositio subgroup age female protestat o- parets itetio itetio Germa self-employed ex ate ex post (years) (%) (%) (%) (%) (%) (%) pre-survey oly (N=69) both surveys (N=196) post-survey oly (N=78) Note: -p < 0.10, -p < 0.05 Differeces sigificat betwee studets who participated i both surveys ad pre- or post-group oly. Post-survey age was corrected by 0.30 years to correct for caledar time of survey. 4 The code cosisted of the first letter of the first ame of the studet s mother, the last letter of the studet s ame, the first digit of the studet s moth of birth, ad the first letter of the studet s place of birth. 10 Page 11 of 39

14 Give that we have some iformatio about o-respodets for both of the two surveys, we ca use a multivariate test to establish whether the likelihood of respodig i the ex ate (ex post) survey is systematically related to characteristics revealed i the secod (first) data collectio. We ra two probit regressios i which we predict respose behavior as a fuctio of sex, age, religio, atioality ad the employmet status of parets ad frieds. Additioally, we icluded scale variables for the studets attitude towards etrepreeurship, the perceived social orms i favor of etrepreeurship, the perceived etrepreeurial self-efficacy, ad the perceived feasibility of a startup project. Both probit regressios cotaied 11 regressors ad were either largely or totally uiformative (p=0.089, =251 i the case of participatio i the post-survey as a fuctio of ex ate data, ad p=0.267, =263 i the case of ex ate participatio as a fuctio of data collected i the secod roud). The margial explaatory power i the ex post survey participatio is due to o-germa participats ad studets with self-employed parets. The o-participatio of these studets is likely to itroduce a coservative (if ay) bias i our results. 5 The subsequet discussio focuses o the matched sample with ex ate ad ex post iformatio from 196 studets Overall Course Assessmet ad Impact o Attitudes ad Skills Previous work o etrepreeurship educatio has show that etrepreeurship courses may reduce the umber of studets who ited to foud (Oosterbeek et al., 2010). We cofirm this fidig here. The followig sectio provides a iterestig perspective o it: etrepreeurship educatio helps studets lear about their etrepreeurial aptitude. Table 2: Attitudial Measures ad Assessmets Ex ate S.E. Ex post S.E. Differece p-value Attitude towards etrepreeurship (0.100) (0.110) p = (scale, 5 items) Risk preferece (0.111) p = (scale, 6 items) Etrepreeurial self-efficacy (0.094) (0.096) p = (scale, 20 items) Feasibility of start-up project (0.060) (0.065) p = (scale, 6 items) Perceived social orms (2.178) (2.057) 1.75 p = (weighted sum of 4 items) Note: N=196. Resposes from matched surveys of LMU studets. First we explore the impact of our etrepreeurship course. Table 2 summarizes evidece o ex ate ad ex post assessmets of several classical attitudial measures. First, we use a scale comprised of five items to measure studets attitude towards etrepreeurship. 6 We test the scale usig iter-item correlatio. Scale reliability is high for both surveys (Crobach s alpha=0.886 ad i the first ad the secod survey, respectively). To 5 The detailed results of these probit regressios are available upo request. 6 These items are take from Gudry ad Welsch (2001) ad have bee used i Kolvereid ad Isakse (2006). 11 Page 12 of 39

15 maitai the scale s iformatio, we do ot stadardize the two measures. We also obtai a scale measure of risk preferece 7 based o 6 items (Crobach s alpha ad 0.798), etrepreeurial self-efficacy 8 based o 20 items (Crobach s alpha ad 0.942), a assessmet of the perceived feasibility 9 of hadlig a startup project (six items, Crobach s alpha ad 0.747) ad fially a measure of perceived social orms. 10 The latter is based o four items askig for a assessmet whether family members, frieds, fellow studets or other importat persos thought that the respodets ought to become etrepreeurs. These were trasformed to yield a symmetric scale, which was the multiplied by a weight obtaied i a survey item i which respodets idicated to which extet they cared about the particular opiio. This measure is best cosidered a formative variable sice the social ifluece of family members, frieds, fellow studets or other importat persos may be additive. Table 2 summarizes the mea values of these measures ad their differeces. Oly the perceived feasibility of hadlig a startup project displays a statistically sigificat chage of about 7 percet of its ex ate value as a result of the course. A larger chage is apparet i a cofidece measure summarized i Table 3 below. Ex post, studets agree less to the statemet I ca always coclude my projects successfully tha ex ate, ad the chage is margially sigificat (p=0.087). The cofrotatio with a real-world problem may have led to a adjustmet of assessmets. A large ad sigificat improvemet is apparet i the respose to the statemet I kow everythig that is eeded to start a ew eterprise. The ex ate average respose to that statemet was betwee do ot agree ad rather ot agree (mea value 2.50) ad shifts to a mea value of 3.87 (betwee rather ot agree ad either agree or disagree ). Table 3: Cofidece Assessmets Ex ate Ex post Differece p-value 1. I ca always coclude my p = projects successfully. 2. I kow everythig that is eeded p < to start a ew eterprise. 3. I am very self-cofidet p = Note: N=196. Resposes measured o ratig scales from 1 to 7 i matched surveys of LMU studets. Moreover, the measure of geeral self-cofidece has rise sigificatly, but much less tha the respose to the etrepreeurship-specific questio. We coclude from these aswers that the course has had a sigificat positive effect o studets skills ad self-cofidece, ad that it may have led to a reduced, ad possibly more realistic assessmet of project success. 7 This scale is based o items first used i the 2004 wave of the Socio-ecoomic Pael (SOEP) ad employed by Dohme et al. (2008) ad Caliedo et al. (2009). 8 This scale is based o 20 items derived from the work of Che et al. (1998); De Noble et al. (1999); Aa et al. (2000). The items are used together by Kolvereid ad Isakse (2006). 9 This scale is derived from Kolvereid (1996a) ad subsequetly used by Tkachev ad Kolvereid (1999); Souitaris et al. (2007). 10 This measure is derived from Kolvereid (1996b) ad has foud use i Tkachev ad Kolvereid (1999); Kolvereid ad Isakse (2006); Souitaris et al. (2007). 12 Page 13 of 39

16 I Appedix Appedix B we provide further evidece that shows that studets gave a positive evaluatio of various aspects of the etrepreeurship course they took. We coclude from this evidece that studets evaluatio of etrepreeurship is ot the result of a poor course Chages i Etrepreeurial Itetios This fial sectio shows that studets etrepreeurial itetios have become more proouced as a result of the course. Etrepreeurial itetios were surveyed with two items i the questioaires. First, we asked a direct questio Would you like to foud your ow eterprise at some poit? requestig a yes or o-respose. Secod, we asked for a idicatio of agreemet regardig the statemet I ited to foud my ow eterprise withi the ext five to te years with resposes o a seve-poit ratig scale. The results are preseted i Tables 4 below ad B.13 i Appedix Appedix B. Table 4: Ex ate ad ex post Etrepreeurial Itetios I ited to start my ow eterprise withi the ext five to te years. Ex post respose Total strogly disagree % disagree % somewhat disagree % Ex ate eutral % respose somewhat agree % agree % strogly agree % Total % 13.3% 15.8% 15.3% 13.3% 17.9% 15.3% 9.2% 100.0% Note: N=196. Resposes from matched surveys of LMU studets. Table 4 provides results for the more detailed measure of etrepreeurial itetios. Cosistet with the less detailed measure, the average score (iterpretig the scale as metric) has decreased from 4.08 to 3.89 (p=0.069 i a twotailed test, N=196). Additioally, the distributio itself is iformative. The share of eutral resposes declied from 19.4% to 13.3%. The eutral overall balace i the ex ate survey (40.4% vs. 40.2% with egative vs. positive assessmets) gave way to a slightly more egative result (44.4% vs. 42.3%). These chages are small, but they idicate that the course helps studets to develop more precise plas for their future. The umber of studets with eutral assessmets declies, opiios grow stroger. This result is also apparet i Table 5 where we cross-tabulate a discrete measure of chages i etrepreeurial itetios with ex ate itetios. The table shows that studets with strog ex ate opiios were less likely to chage their itetios tha idifferet studets. Chages i itetios occur mostly amogst the udecided, as oe would expect i a world with Bayesia updatig durig the course. 13 Page 14 of 39

17 Table 5: Chages i Etrepreeurial Itetio by ex ate Itetio I ited to start my ow eterprise withi the ext five to te years. Chage i ex post respose Chage No chage Total strogly disagree % % 16 disagree % % 27 somewhat disagree % % 36 Ex ate eutral % % 38 respose somewhat agree % % 29 agree % % 26 strogly agree % % 24 Total % % 196 Note: N=196. Resposes from matched surveys of LMU studets. However, it would be misleadig to believe that oly the udecided leared somethig from the course. Table 6 provides evidece to this regard. I the upper pael of the table, we display which percetage of studets who idicate a particular level of etrepreeurial itetios have parets or frieds who are self-employed. For example, while oly 12.5% of those who disagree strogly with the statemet I ited to foud my ow eterprise withi the ext five to te years have self-employed parets, the share of studets with self-employed parets is 58.3% for those i the highest respose category. There is a clear bivariate relatioship betwee paretal self-employmet ad studets itetios. This is eve more clearly visible oce we coditio paretal self-employmet o positive experiece. The relatioship is less proouced for self-employmet of frieds, but agai clearer oce oe requires self-employmet to have bee a positive experiece. Table 6: Ex ate ad ex post Etrepreeurial Itetios Ex ate statemet: I ited to foud my ow eterprise withi the ext five to te years. paret self-... ad positive frieds self-... ad positive Level of agreemet employed experiece employed experiece N strogly disagree 12.5% 6.3% 68.6% 31.3% 16 disagree 29.6% 22.2% 66.7% 55.6% 27 somewhat disagree 36.1% 30.6% 77.8% 58.3% 36 eutral 43.2% 35.1% 83.8% 75.7% 37 somewhat agree 41.4% 34.5% 82.8% 65.5% 29 agree 53.8% 53.8% 69.2% 65.4% 26 strogly agree 58.3% 54.2% 91.7% 87.5% 24 Total 40.5% 54.2% 77.9% 64.6% 195 Ex post statemet: I ited to foud my ow eterprise withi the ext five to te years. paret self-... ad positive frieds self-... ad positive Level of agreemet employed experiece employed experiece N strogly disagree 23.1% 15.4% 69.2% 46.2% 26 disagree 32.3% 29.0% 71.0% 51.6% 31 somewhat disagree 40.0% 36.7% 86.7% 70.0% 30 eutral 53.8% 34.6% 84.6% 73.1% 26 somewhat agree 29.4% 26.5% 76.5% 64.7% 34 agree 50.0% 46.7% 73.3% 66.7% 30 strogly agree 66.7% 66.7% 88.9% 88.9% 18 Total 40.5% 54.2% 77.9% 64.6% 195 Note: paretal ad frieds self-employmet is take from the ex-ate survey i both paels. Note: N=195. Resposes from matched surveys of LMU studets. 14 Page 15 of 39

18 While these results are ot surprisig, the secod pael of Table 6 shows how the course affected studets itetios. Here we fid that the share of studets with self-employed parets or frieds has become much higher i the lower ex post respose categories. The share of studets with self-employed parets has almost doubled ow i the lower respose category, ad it has icreased somewhat i the upper oe. This shows that the course detached some participats from their ex ate etrepreeurial itetios. At the ed of this sectio we are left with a iterestig puzzle. The course has apparetly reduced the etrepreeurial itetios of participatig studets. 11 However, it has also led studets to develop firmer future plas. Studets state that they feel more assured regardig the capabilities eeded to foud a ew eterprise (see Table 3 ad Table B.12). Moreover, we fid iterestig evidece that studets reshape their itetios ad opiios regardig etrepreeurship durig the course. Weak opiios become more defied, ad studets become detached from previous covictios as determied by paretal backgroud ad former persoal eviromet. We cosider these descriptive results to be importat sice they shed ew light o the learig process itself. I the ext sectio we test the predictios of the theoretical model to see whether Bayesia updatig provides a explaatio of what we observe. 5. Testig the Learig Model I this sectio we test the learig model discussed i Sectio 3 above. We begi by presetig the variables which eter our regressios. The we preset results of a differeces of variaces test ad two regressios. We provide evidece to support all three of our hypotheses Descriptio of Variables Table 7 below sets out descriptive statistics. At the top of the table we preset the depedet variables, below that we preset the explaatory variables. The statistics are preseted for the sample of 189 studets whose resposes to pre ad post questioaires we are able to match. We elicit studets beliefs about their ow etrepreeurial aptitude by askig them about their itetio to foud their ow compay. A set of questios related to the feasibility of foudig a compay is used to measure what sigals the studets have received about their etrepreeurial aptitude before ad durig the course. The depedet variables. We test Hypothesis 1 usig detailed measures of studets itetios to foud a eterprise withi the ext five to te years. We aalyze the variace of the itetio to foud before ad after the course. 11 We have to soud a warig here - the setimet of studets regardig etrepreeurship may also have bee affected by the fiacial crises that bega to impact the ecoomy at the ed of exactly the time whe studets erolled i this class. 15 Page 16 of 39

19 Hypothesis 2 focuses o the stregth of studets beliefs that they are (are ot) etrepreeurs after the course. We take the squared deviatio from the mea of studets itetios to foud after the course ( B) as our measure of the stregth of studets beliefs after the course. 12 Hypothesis 3 is based o the chage i studets itetios resultig from etrepreeurship educatio. The depedet measure here is the squared chage of itetios to foud ( ). Here the chage refers to the differece betwee itetios measured before ad after the course. 13 Table 7: Descriptive Statistics Variable Mea Media Std. Deviatio Miimum Maximum Itetio to foud afterwards B Perceived social orms Scale: etrepreeurial self efficacy Scale: risk preferece Scale: feasibility assessmet before course Chage i feasibility assessmet scales Stregth of sigals before Stregth of sigals afterwards Cosistet sigals Parets self-employed Frieds self-employed o- Germa Female Protestat Note N = 189. Resposes from matched surveys of LMU studets respodig to a full set of questios. 12 I ureported results we fid that the precise defiitio of strog beliefs does ot affect our fidigs. Regressios i which we take the squared differece betwee studets itetios ad the media of the seve poit Likert scale o which itetios to foud were reported are qualitatively similar to the results we report below. 13 This variable is clearly skewed. Regressios i which we use its logarithm as the depedet variable produce qualitatively idetical results to those reported below. 14 This variable is defied i Sectio This variable is defied i Sectio Page 17 of 39