Attributes of legitimate venture failure impressions

Article history: Received 18 May 2016 Received in revised form 22 December 2016 Accepted 4 January 2017 Available online xxxx The current research investigates the effectiveness of impression management strategies available to entrepreneurs to foster social legitimacy with stakeholders following venture failure. We use a conjoint experiment to examine howdifferent attributions of causes of failure influence the general public's legitimacy judgments. The most effective strategy proves to be the entrepreneurs distancing themselves from the failure, in that they attribute the failure to external factors that are not under the entrepreneurs' volitional control, and brought about by circumstances that are unlikely to reoccur. Our analysis also considers how the audience members' dispositional agreeableness and general self-efficacy influence judgment formation. © 2017 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Social stigmatization following venture failure can create substantial psychological, economic, and social costs for the entrepreneur (Cardon et al., 2011;Shepherd and Patzelt, 2015;Ucbasaran et al., 2013). Entrepreneurship scholars have increasingly investigated impression management strategies that enable entrepreneurs to create legitimate failure impressions as a means of mitigating the stigma of failure (Shepherd and Haynie, 2011;Singh et al., 2015). Impression management strategies reflect tactical self-presentations that aim to protect the individual's current image or foster a new, desired image (Goffman, 1959;Kowalski and Leary, 1990;Sutton and Callahan, 1987). Successful impression management strategies create socially appropriate failure impressions that build a legitimate base for future career actions (Semadeni et al., 2008;Weisenfeld et al., 2008;Zimmerman and Zeitz, 2002), support emotional recovery (Byrne and Shepherd, 2015;Mantere et al., 2013), and facilitate learning from failure (Cope, 2011;Shepherd, 2003;. The legitimacy gained and social stigma avoided through the successful application of impression management strategies can also boost the individual's motivation and confidence to re-enter entrepreneurship (Simmons et al., 2014). Goffman (1959) distinguishes between back-stage and front-stage impression management. Back-stage settings involve a limited target audience, such as are found in meetings with board members or employees (Elsbach, 2003;Gardner and Martinko, 1988), or when pitching business ideas to investors Parhankangas and Ehrlich, 2014). These settings offer entrepreneurs the opportunity to interact closely with their audience and to address how others evaluate venture failure (Bolino et al., 2008). In contrast, front-stage settings are characterized by a broad and diverse target audience including not only the failed venture's stakeholders but also stakeholders with an important role in the entrepreneur's future, such as potential employers, investors, and customers, who may as yet be unknown to the entrepreneur (Carter, 2006;Leary, 1994;Morrison and Bies, 1991). In such settings, opportunities for interaction are limited, and it is therefore important for entrepreneurs that the first public explanations of their venture failure create a favorable impression suitable for a diverse audience (Elsbach, 1994;Semadeni et al., 2008). Despite recent advances in impression management research in entrepreneurial contexts (e.g., Nagy et al., 2012;Parhankangas and Ehrlich, 2014;Shepherd and Haynie, 2011), we have limited knowledge of how entrepreneurs can present venture failure in a way that enhances their legitimacy with public audiences.
The present study addresses this gap by using a conjoint experiment to investigate how observers from among the general public judge the legitimacy of different impressions of entrepreneurial failure. Presenting a failure impression to the general public represents the challenge of garnering legitimacy judgments from a diverse population of stakeholders who are potentially important yet indefinable at the time. From an audience-centered perspective, evaluative legitimacy judgments of failure impressions represent a public sounding board to review the effectiveness of impression management (Elsbach, 1994;Lounsbury and Glynn, 2001;Morrison and Bies, 1991;Überbacher, 2014). The study was conducted with 601 participants from the German adult population. Following legitimacy judgment theory (Bitektine, 2011), the participants were asked to judge the social appropriateness of eight failure impression scenarios (Zimmerman and Zeitz, 2002), and two participant characteristics measured in a post-experiment survey were included as moderators of the conjoint experiment: agreeableness (interpersonal trait) (Graziano et al., 2007) and self-efficacy belief (personal trait) (Bandura, 2012). The failure impression scenarios varied in terms of three dominant attributions of failure derived from attribution theory (Heider, 1958;Weiner, 2000): whether the failure was caused by factors internal or external to the entrepreneur (the locus of causality); whether the factors leading to the failure were under the entrepreneur's control or not (controllability); and whether the cause of failure could or could not happen again (stability).
Our work generates four contributions. First, we extend the legitimacy judgment theory (Bitektine, 2011;Bitektine and Haack, 2015) with attribution theory (Heider, 1958;Weiner, 2000) to not only explain how observers from the general public make evaluative legitimacy judgments but also what dominant attributions influence those judgments, be it positively or negatively. This framework allows us to explain the effectiveness of impression management strategies for the general public (Goffman, 1959;Morrison and Bies, 1991), by arguing which attributes of a failure impression lead to the evaluator judging a failure event "acceptable" or "normal" (Cardon et al., 2011: 80).
Second, we apply this theory extension to contribute to the literature on entrepreneurial failure, especially the recent research on post-failure impression management (Shepherd and Haynie, 2011;Simmons et al., 2014;Singh et al., 2015). More specifically, we contribute to filling the gap in terms of the lack of empirical evidence on the effectiveness of different impression management strategies in addressing venture failure. The principal finding of our study is that the most effective impression management strategies for garnering legitimacy from the public for the failure event are those where entrepreneurs distance themselves from the failure. These distance-taking impressions ascribe the cause of failure to forces that are external to the entrepreneur/firm, not under the entrepreneur's volitional control, and subject to circumstances that are unlikely to reoccur. These findings correspond to two impression management strategies outlined in Shepherd and Haynie's (2011) conceptual framework: denying responsibility and defining failure in a positive light. Both of these strategies aim to minimize social stigma by fostering a positive self-view that symbolically decouples the failure from the entrepreneur (Überbacher and Jacobs, 2016).
Third, we further complement the research on learning from entrepreneurial failure by pointing out the lack of learning opportunities as a potential drawback of impression management strategies based on distance-taking. By presenting failure as caused by external and uncontrollable circumstances, entrepreneurs can maintain professional legitimacy, but at the same time they do not invite and benefit from constructive feedback from stakeholders (Morrison and Bies, 1991) that would support their learning from failure (Cope, 2011;. We highlight the need for a more fine-grained understanding of how entrepreneurs can manage evaluative legitimacy judgments, while at the same time establishing learning from failure as an important avenue for future research. Fourth, we advance the evaluator perspective on entrepreneurial failure (Shepherd and Patzelt, 2015) by generating new insights into how an observer's characteristics influence the formation of evaluative legitimacy judgments (Bitektine, 2011). We theorize and empirically examine how dispositional agreeableness (an interpersonal trait) and general self-efficacy (a personal trait) moderate the relationships between evaluative legitimacy judgments and attributions of causes of failure. By discussing these new insights in light of existing theory, we advance our knowledge of the understated role of individual characteristics in evaluative legitimacy judgments (Bitektine, 2011) and why observers make more or less harsh evaluations of entrepreneurial failure (Shepherd and Patzelt, 2015).

Theory and hypotheses
2.1. Impression management, legitimacy judgments, and attributions of failure Impression management of entrepreneurial failure concerns the strategies entrepreneurs employ to make their business failure appear legitimate (Shepherd and Haynie, 2011;Singh et al., 2015;Vaara, 2002). A legitimate impression (Elsbach, 2003;Lounsbury and Glynn, 2001) of failure is one that the audience perceives as "desirable, proper, or appropriate" (Suchman, 1995: 574). In order to create such an impression, the entrepreneur regulates information provision by highlighting certain attributes while downplaying, or even hiding, others Parhankangas and Ehrlich, 2014;Überbacher and Jacobs, 2016). By engaging in impression management, "entrepreneurs can (hope to) maintain a favorable impression of themselves in the eyes of important stakeholders [and thus] avoid, or minimize, stigma" (Shepherd and Haynie, 2011: 187). Minimizing stigma can support the entrepreneur's emotional recovery and learning from failure (Byrne and Shepherd, 2015;Jenkins et al., 2014;Shepherd, 2003), and aid in building a legitimate professional image for future career actions (Elsbach, 1994(Elsbach, , 2003Garud et al., 2014;Weisenfeld et al., 2008).
Our focus is on front-stage impression management that aims to generate a legitimate impression of a venture's failure for a diverse public audience. Following an audience-centered perspective on impression management (Überbacher, 2014), we view such public legitimacy as "social judgment of acceptance" (Zimmerman and Zeitz, 2002: 414). We build on legitimacy judgment theory (Bitektine, 2011;Bitektine and Haack, 2015) to understand how public legitimacy judgments of entrepreneurial failure are formed, and complement it with attribution theory (Heider, 1958;Weiner et al., 1988;Weiner, 2000) to specify what features of the failure event influence those judgments.
According to legitimacy judgment theory (Bitektine, 2011), an observer's judgment is based on her or his knowledge and values reflecting the two most general foundations of legitimacy: the cognitive and the evaluative (Aldrich and Fiol, 1994;Suchman, 1995;Überbacher, 2014). In the context of entrepreneurial failure, cognitive legitimacy refers to an evaluator's tacit, taken-for-granted understanding of the characteristics of venture failure (Bitektine, 2011;Suchman, 1995). Thus, judgments based solely on cognitive legitimacy are made passively and not based on active evaluation Pollack et al., 2012). If an evaluator understands the attributes of venture failure, the impression of that failure is cognitively legitimate (Shepherd and Zacharakis, 2003).
Evaluative legitimacy extends the judgment process by including an evaluator's explicit normative assessment of how the business failure happened. In addition to cognitively processing the characteristics of venture failure, the assignment of evaluative legitimacy requires the evaluators to judge whether the attributes of entrepreneurial failure are socially appropriate (Tost, 2011;Überbacher, 2014). An evaluative legitimacy judgment therefore subjects entrepreneurial failure "to further scrutiny and questioning" (Bitektine, 2011: 157). Consistent with the aim of our study, we focus on public evaluative legitimacy judgments of entrepreneurial failure. This is because normative judgments on whether the way the venture failure happened is "deemed acceptable or 'normal'" (Cardon et al., 2011: 80) determine whether the evaluator supports or sanctions the failed entrepreneur (Bitektine, 2011), which in turn influences the experience of post-failure stigma (Singh et al., 2015;Weisenfeld et al., 2008). Furthermore, Bitektine's (2011) legitimacy judgment theory suggests that the evaluators' dispositional social consciousness (an interpersonal trait) and cognitive ability (a personal trait) influence their judgments of social acceptance. Socially conscious evaluators are more willing and able to engage in perspective taking, and consequently demonstrate high levels of agreeableness when judging other people and their behavior (Bitektine, 2011;Graziano et al., 2007;Todd et al., 2012). Therefore, evaluators with higher levels of agreeableness than their peers are likely to show compassion toward failed entrepreneurs and trust that their intentions were not malevolent; attitudes that result in more forgiving judgments of business failures and their causes (Judge and Zapata, 2015;Shepherd and Patzelt, 2015).
Consistent with the social cognition theory of efficacy beliefs (Bandura, 2012), Bitektine (2011) further suggests that an evaluator's cognitive ability influences the formation of evaluative legitimacy judgments. The evaluator's belief in her or his own ability to complete tasks and reach goals influences whether that person generally thinks more pessimistically or optimistically about challenges (Bandura, 1982), which in turn results in a more integrative or compromising approach respectively to making social judgments (Corcoran and Mallinckrodt, 2000). Therefore, evaluators with a high level of self-efficacy belief tend to evaluate behavioral outcomes more positively, than their counterparts with weaker self-efficacy beliefs, and are thus likely to judge failure as an appropriate challenge to be overcome rather than as a threat to be avoided (Wood and Bandura, 1989).
Extending the evaluative judgment approach (Bitektine, 2011) with attribution theory (Heider, 1958), we argue that individuals evaluate entrepreneurial failure impressions based on three fundamental questions: Who was involved? Who is responsible? Will it happen again? (Heider, 1958;Weiner et al., 1988;Weiner, 2000). Hence, judgments of the legitimacy of entrepreneurial failure reflect the evaluation of the appropriateness of three attributional dimensions of failure: (a) the locus of causality, whether the cause of failure was internal (own mistakes) or external (misfortune caused by outside actors or events) to the entrepreneur; (b) controllability, whether the cause was under the entrepreneur's volitional control or not; and (c) stability, whether the cause of failure was transient (could not happen again) or enduring over time (could happen again) (Harvey et al., 2014;Weiner, 2000).
Entrepreneurs can use the three attributional dimensions to justify venture failure (Mantere et al., 2013) and to regulate the information about the failure event with the aim of creating a favorable impression in the eyes of evaluators (Überbacher, 2014). Building on Sutton and Callahan (1987), Shepherd and Haynie's (2011) conceptual analysis identifies three impression management strategies that are applicable to public legitimacy judgments and capture the (combined) use of the three failure attributions. The first strategy-defining failure in a positive light-involves the entrepreneur acknowledging that the venture has ceased operations while at the same time, emphasizing positive outcomes such as critical lessons learned and portraying failure "as a unique event (a perfect storm)" (Shepherd and Haynie, 2011: 186). The attributional emphasis in this strategy is thus on stability: the failure would be unlikely to happen again. The second strategy-denying responsibility-derives from the self-serving assumption in attribution theory (Heider, 1958) and involves entrepreneurs protecting their personal credibility by downplaying their own role in the failure and ascribing it to external causes. The attributional emphasis is on an external locus of causality and uncontrollability. The third and seemingly counterintuitive strategy-promoting a negative self-view-elicits sympathy from the evaluators by emphasizing the entrepreneur's commitment to the venture, and the loss that the failure has caused, without presenting the failure as a positive experience (Shepherd and Haynie, 2011;Sutton and Callahan, 1987). The attributional emphasis is on an internal locus of causality and controllability.
Next, we use the theoretical framework presented above to develop testable hypotheses on how different attributions of causes of failure (locus of causality, controllability, and stability) influence the assignment of evaluative legitimacy to a failure event. Moreover, as theorized above, we investigate the moderating effects of dispositional agreeableness and self-efficacy on the relationships between legitimacy judgments and locus of causality and controllability. Fig. 1 summarizes the hypothesized relationships graphically, and the following section explains the rationale behind each hypothesis.

Locus of causality, controllability, and stability
The locus of causality in venture failure is either internal or external (Cardon et al., 2011;Mantere et al., 2013). An internal attribution of failure communicates the entrepreneur's personal responsibility for failure (mistakes made), whereas an external attribution shifts the blame to actors (e.g., financiers, competitors) or events (e.g., an economic crisis) beyond the boundaries of the organization. The locus of causality in a failure impression is likely to influence observers' judgments of the extent to which they associate the entrepreneur with the failure event (Semadeni et al., 2008). We argue that public audiences evaluate failure impressions that provide a coherent account of external factors causing the failure more positively than impressions that portray the entrepreneur's mistakes as contributing to the failure. If entrepreneurs do not separate themselves from the business failure event, observers have a clearly defined target for the blame, and the outcome may be a negative legitimacy judgment (Cardon et al., 2011;Ucbasaran et al., 2013). Therefore, adopting the impression management strategy of denying personal involvement and ascribing the causes of failure to external factors might deflect the blame from the entrepreneur personally, and result in a higher level of social approval (Shepherd and Haynie, 2011). Based on these arguments, we hypothesize: Hypothesis 1. A failure impression that presents the failure as caused by external factors is positively associated with evaluative legitimacy.
The controllability attribution in entrepreneurial failure impressions relates to the extent that the failure is presented as having been under the entrepreneur's volitional control (Harvey et al., 2014;Weiner et al., 1988). For example, a failure attributable to increased competition could be considered as uncontrollable, whereas insufficient effort made by the entrepreneur or a lack of knowledge are typically viewed as controllable factors (Mantere et al., 2013). When observers perceive the cause of a negative event as controllable, then the person at the center of the event is typically viewed as being responsible for the outcome (Hip-Fabek, 2006). This perception of responsibility might lead to a more negative reaction toward a failed entrepreneur than in a situation where the failure is associated with circumstances beyond the entrepreneur's control (Graham et al., 1993). Although observers are likely to judge entrepreneurs who caused the failure of their venture punitively if they believe the entrepreneur could have avoided the failure, the judgment might be more positive if the entrepreneurs' actions are judged to be impelled by aspects beyond their control (Hamilton, 1980). In other words, if the cause of venture failure is perceived as uncontrollable, the entrepreneur will not be held personally responsible for it (Graham et al., 1997). The lack of controllability can elicit sympathy from and the approval of public audiences, resulting in positive legitimacy judgments of the failure event (Sutton and Callahan, 1987). For these reasons, we hypothesize: Hypothesis 2. A failure impression that presents the failure as caused by forces beyond the entrepreneur's control is positively associated with evaluative legitimacy.
The stability attribution in failure impressions refers to the amount of variability or permanence in the circumstances that led to the failure (Graham et al., 1997;Tomlinson and Mryer, 2009). For example, business failure would be perceived as a unique event if it were presented as unlikely to happen again, whereas it would be ascribed to permanent factors if it were presented as being likely to reoccur in similar circumstances (Shepherd and Haynie, 2011). Causal stability is closely related to an entrepreneur's expectations of future success, that is, hope or hopelessness (Weiner, 2000). Entrepreneurs who present business failure as a reoccurring phenomenon increase the chances that observers will distrust their ability to succeed in the future. Therefore, communicating failure as nonrecurring is likely to set a more legitimate base for an entrepreneur's future career (Garud et al., 2014). Following these arguments, we hypothesize: Hypothesis 3. A failure impression that presents the failure as caused by factors that are unlikely to reoccur is positively associated with evaluative legitimacy.
Furthermore, attribution theory suggests that the stability attribution influences the attributional dimensions of the locus of causality and controllability (Hess, 2008;Tsiros et al., 2004). In line with this argument, we propose that the presentation of causal stability not only affects legitimacy judgments by shaping observers' expectations of the failed entrepreneur's future success, but also sets conditions on how the observer evaluates locus of causality and controllability in a failure impression (Seiders and Berry, 1998;Weiner, 2000). For example, if an entrepreneur admits personal responsibility for a business failure and is at the same time perceived to be likely to fail again in a similar situation, that failure impression is likely to be more negative than if the failure were attributable to internal but unstable causes. Similarly, entrepreneurs who admit that the events leading to the failure were under their control, and are at the same time perceived to be likely to fail again, are likely to be judged more negatively than if the failure is presented as being caused by controllable but nonrecurring factors. The permanence of the cause presented can lead observers in both cases to believe that the entrepreneur will consistently make the same mistakes, which will ultimately result in a negative judgment (Maxham and Netemeyer, 2002). Accordingly, observers who evaluate a business failure as a one-time nonrecurring phenomenon are more likely to view a failure event as a challenge to be overcome, leading to a positive legitimacy judgment compared to that when the cause of failure is ascribed to stable causes. Based on these arguments, we present the following two moderating hypotheses: Hypothesis 4a. A failure impression that ascribes the cause of failure to external factors will be judged more positively in terms of evaluative legitimacy if the failure is presented as a one-time nonrecurring phenomenon than if it is presented as being likely to reoccur in similar circumstances.
Hypothesis 4b. A failure impression that portrays the cause of failure as uncontrollable will be judged more positively in terms of evaluative legitimacy if the failure is presented as a one-time nonrecurring phenomenon than if it is presented as being likely to reoccur in similar circumstances.

Dispositional agreeableness and self-efficacy
The legitimacy judgment approach is rooted in social cognition theories of traits (Bandura, 1982(Bandura, , 2012Fiske, 1993) and suggests that an individual's social consciousness (an interpersonal trait) and cognitive ability (a personal trait) can influence that person's evaluative judgment of others (Bitektine, 2011). Regarding social consciousness, recent social judgment research has emphasized the role of the observer's level of dispositional agreeableness in explaining positive or negative evaluations of other people's attitudes and behaviors (Graziano et al., 2007;Graziano and Tobin, 2013). In the context of judging entrepreneurial failure, one main argument is that evaluators with a high level of dispositional agreeableness are likely to believe that a failed entrepreneur's intentions were not malevolent and that the justification presented of the business closure reflects the true situation. Thus, observers engage in perspective taking (Todd et al., 2012) and feel more sympathetic toward an entrepreneur's explanation of the failure (Shepherd and Patzelt, 2015). As a result, individuals recording high levels of agreeableness make more positive legitimacy judgments of entrepreneurial failure than would individuals with low levels of agreeableness. According to this logic, when entrepreneurs distance themselves from failure by ascribing its cause to external, uncontrollable, and unstable forces, observers with a high level of dispositional agreeableness are even more likely to evaluate a failure justification as appropriate than would individuals with a low level of agreeableness.
However, another possible line of argument is that evaluators with a high level of agreeableness disapprove of a failure impression that ascribes failure solely to external, uncontrollable, and unstable forces. Fiske's approach to social judgments (1993: 163) posits that an individual tends to evaluate the justification of other people's action and events based on "prototypical" personal characteristics, which are assumed to be "stable, long-lasting, and internally caused." The approach suggests that an evaluator will be motivated to identify the entrepreneur's personal characteristics, and thus the individual's role in causing the failure event. Given that people with a high level of agreeableness appreciate personal characteristics such as honesty, modesty, and straightforwardness (Graziano and Tobin, 2013), they are likely to judge a failure impression as less appropriate when entrepreneurs distance themselves from the failure by ascribing it to causes that are external and beyond their control. In contrast, failure impressions that portray a negative self-view of the entrepreneur, by referring to internal and controllable causes, can signal an entrepreneur's commitment to the venture (Shepherd and Haynie, 2011) and thus elicit sympathy and confer credibility, because the audience assumes a strong link between the entrepreneur's actions and performance outcomes (Cardon et al., 2011).
In summary, we find arguments for two competing rationales for an evaluator's level of dispositional agreeableness moderating the relationships between the attributions of failure and evaluative judgments of its legitimacy. Hence, we formulate the following competing hypotheses with respect to entrepreneurs presenting a failure impression that distances them from the failure by ascribing it to external, uncontrollable, and unstable factors: Hypothesis 5a. The higher an observer's level of dispositional agreeableness, the more likely it is that an entrepreneur's distancetaking from business failure will be evaluated as socially appropriate.
Hypothesis 5b. The higher an observer's level of dispositional agreeableness, the less likely it is that an entrepreneur's distancetaking from business failure will be evaluated as socially appropriate.
Furthermore, Bitektine's (2011) legitimacy judgment approach suggests that evaluators' cognitive ability (a personality trait) explains how their evaluative judgments are formed. Consistent with the social cognition theory of efficacy beliefs (Bandura, 1982), individuals with strong self-efficacy beliefs are good at motivating themselves and persevering in the face of adversity. They can maintain their confidence after mistakes and think optimistically about the course of action in challenging situations (Bandura, 2012). Hence, individuals with high levels of self-efficacy are likely to have high levels of tolerance to situations involving unstable and uncontrollable factors. They are thus likely to perceive a venture failure ascribed to external and uncontrollable causes as a challenge to be overcome (Wood and Bandura, 1989). As a result, individuals with high levels of self-efficacy are even more likely to approve of failure justifications involving external, uncontrollable, and unstable causes than would individuals with low levels of self-efficacy.
However, there is an alternative logic for the moderating effect of self-efficacy. That logic posits that individuals with high levels of self-efficacy evaluate an entrepreneur's distance-taking from venture failure more harshly than would people with low self-efficacy levels. The underlying rationale is that evaluators with a strong belief in their own capability search for causal attributions that originate from the person involved in the failure event (Whyte et al., 1997). Accordingly, observers with high selfefficacy levels are likely to adopt an internal, trait-oriented norm when making social judgments (Fiske, 1993) and therefore assume that the failure will largely be explained by the entrepreneur's capability (Bandura, 2012). Fiske (1993): 164 further argues that trait-oriented social judgments are based on the stereotypical assumption that "incapable persons can behave only incapably, whereas a capable person can behave incapably and capably." In other words, entrepreneurs who do not hide a negative self-view (Shepherd and Haynie, 2011) and present failure as an event with an internal cause and that is controllable might even be evaluated as confident and committed, rather than incapable. For such entrepreneurs, a venture failure is just another challenge to be overcome. Accordingly, when entrepreneurial failure is ascribed to external, uncontrollable, and unstable forces, an evaluator with high self-efficacy beliefs may distrust the entrepreneur's capability, and thus disapprove of their distance-taking and denial of responsibility.
In summary, we see two competing rationales for the moderating effect of self-efficacy on the relationships between attributions of failure and evaluative legitimacy judgments. Thus, we propose the following competing hypotheses: Hypothesis 6a. The stronger an observer's efficacy belief, the more likely that observer is to evaluate an entrepreneur's distancetaking from business failure as socially appropriate.
Hypothesis 6b. The stronger an observer's efficacy belief, the less likely that observer is to evaluate an entrepreneur's distancetaking from business failure as socially appropriate.

Research design
We used metric conjoint analysis to examine how the locus of causality, controllability, and stability in a portrayal of entrepreneurial failure influence observers' evaluative judgments of the legitimacy of failure explanations. Conjoint analysis is an established method in marketing research to study how buyers make trade-offs among competing products (Green et al., 2001). It has been applied more recently in entrepreneurship research to study entrepreneurial decision making (Behrens and Patzelt, 2015;Holland and Shepherd, 2013;Lohrke et al., 2010;Monsen et al., 2010;Shepherd et al., 2013). The main advantage of conjoint analysis for our research is that it allows us to reduce the complexity of judging an entrepreneurial failure to the attributes of interest (Weisenfeld et al., 2008), and also to estimate the importance observers place on each of these attributes in their judgment (Shepherd and Patzelt, 2015).
We used a web-based survey to collect data in Germany. To obtain a balanced sample in terms of age, gender, and geographic distribution, we relied on an internet panel provider to provide the participants with access to an online survey. The survey instrument included conventional survey questions on the participants' demographics and attitudes, as well as a conjoint experiment, in which the participants evaluated multiple profiles of entrepreneurial failure. Thus, the resulting data have a hierarchical structure: multiple judgments of failure profiles (level 1) are nested within an individual participant (level 2). Full details of the survey and the survey instrument are available from the authors upon request (in German).
In a conjoint experiment, respondents typically assess specific profiles that are combinations of theoretically derived attributes (Priem and Harrison, 1994). The profiles in our study consist of the three attributes in Hypotheses 1-3 (the locus of causality, controllability, and stability) as well as a performance reference as a control measure (Table 1). The performance reference attribute controls for the circumstances under which the failure happened. It takes on the value global if failure is a common event in the branch of business, and the value specific if only the focal entrepreneur has failed. The latter captures failure being an unexpected outcome that may provide a strong signal of an enterprise's leaders being ineffective and ultimately results in a negative judgment of the failure event (Weisenfeld et al., 2008). Consistent with attribution theory and previous studies using a metric conjoint approach, we describe each attribute on two levels. This results in a total of 16 (4 2 ) different failure profiles. A sample profile comprising all four attributes would read: "I am the only entrepreneur to fail in this type of endeavor. I was personally involved. I could have avoided it. It is unlikely to happen again in the future." Table 1 Profiles of entrepreneurial failure in the conjoint experiment.

Attribution
Level Operationalization

Research variables
Locus of causality Internal "I was personally involved" External "It was related to external circumstances" Controllability Controllable "I could have avoided it" Uncontrollable "I could not have avoided it" Stability Recurrent "It is likely to happen again in the future" Nonrecurring "It is unlikely to happen again in the future" Control variable Performance reference Global "Others have failed in this type of endeavor as well" Specific "I am the only entrepreneur to fail in this type of endeavor" Evaluative legitimacy judgment (on a scale from 1 (completely inappropriate) to 7 (completely appropriate) (Bitektine, 2011) How appropriate do you consider the following statement by the failed entrepreneur concerning the upcoming bankruptcy of their firm?
Note: In half of the scenarios presented, bankruptcy was replaced with voluntary closure to introduce variety in the context of the failure. The context was randomly assigned.
Following the evaluative legitimacy judgment approach (Bitektine, 2011;Bitektine and Haack, 2015), the participants evaluated the legitimacy of these profiles based on the following question: "How appropriate do you consider the following statement by the failed entrepreneur concerning the upcoming bankruptcy (voluntary closure) of their firm?" To control for the type of business failure (Shepherd and Patzelt, 2015;Ucbasaran et al., 2013), the profile was randomly contextualized either as a scenario where the founder had to declare bankruptcy or a second one where the entrepreneur voluntarily ceased trading.
As the judgment task in metric conjoint analysis is generally fully replicated to test for respondent reliability, we would have had to ask each survey participant to evaluate 32 profiles in a full factorial design. To ensure that the time-consuming task of profile judgment remained manageable, while not compromising on the content of the study, we employed an orthogonal fractional factorial design to reduce the number of profiles to eight (Behrens and Patzelt, 2015;Holland and Shepherd, 2013). The orthogonal design results in zero correlation between the four attributes in the experiment, which is consistent with attribution theory because the three attributional dimensions are viewed as independent (Tomlinson and Mryer, 2009;Weiner, 1985). We ensured that the resulting design allowed interacting stability with locus of causality and controllability so as to test Hypotheses 4a and 4b.
The conjoint experiment section in our survey instrument opened with instructions for the participant and a practice profile that served to familiarize participants with the conjoint task. To control for specific characteristics of the failed entrepreneur, we informed the participants that in all scenarios the entrepreneur had substantial experience and expertise in the business. To ensure that participants made objective judgments, we further instructed the participants that they were not acquainted with the failed entrepreneur in any personal or professional capacity. Moreover, we asked the participants to assume that the founder was the CEO of the firm and therefore fully in charge of the operations. Finally, we instructed the participants to carefully read each of the profiles, independent of the previous and subsequent profiles.
Next, the participants were asked to rate each of eight profiles twice for a full replication design. We randomly assigned both the order of the profiles and the order of the attributes. For each type of business failure, this process resulted in four different profile versions: a base version; one where the profiles are presented in the same order as in the base version but the order of the attributes within the profiles varies; one where the order of the attributes remains the same but the order in which the profiles are presented varies; and one where both the order of the profiles and the attributes within them vary. To test for possible order effects, we estimated an additional regression model that included dummies for the eight different versions. Because none of the coefficients was significant (p b 0.10), we can conclude that neither the order of the profiles nor the attributes introduce bias to the results.

Sample
Our online survey targeted 6985 individuals aged 18-69 years old from all 16 federal states in Germany. After screening out respondents who did not meet the eligibility criteria, submitted incomplete responses, or did not pass the attention test (a question that read "if you read this press 'fully agree'"), the survey generated 769 usable responses. We assessed the extent of nonresponse bias by using archival analysis (Rogelberg and Stanton, 2007) that is, comparing the demographic characteristics of the sample to that of the German working age population. Referring to the German Federal Statistical Office database suggested that the respondents in our final sample were reasonably representative of the German working age population (taken to be 18-69 years old) in terms of age, gender, and geographic distribution (Destatis, 2015). Because the respondents rated each failure profile twice, we could analyze the correlation between these two ratings. The initial mean test-retest correlation in the sample of 769 eligible individuals was 0.64. In line with previous conjoint studies (e.g., Shepherd, 2008, Shepherd et al., 2013), we used 0.45 as a threshold for the test-retest reliability to exclude unreliable responses from the analysis (Holland and Shepherd, 2013). Applying this criterion resulted in a final sample of 4808 failure profile judgments from 601 individuals. The mean test-retest correlation in this sample is 0.71.

Dependent variable
Consistent with the evaluative legitimacy judgment approach (Bitektine, 2011) and Suchman's (1995) original definition of social legitimacy, the dependent variable of our study reflects the observer's evaluation of the appropriateness of the failure profile. Accordingly, we asked participants to assess the degree to which they found the failure profile appropriate using a 7-point Likerttype scale anchored with completely inappropriate (1) and completely appropriate (7) (see Table 1). Because each respondent rated each profile twice, we used the mean of the two ratings as the dependent variable. The mean rating is 3.51 with a standard deviation of 1.45. The Cronbach's alpha coefficient for the two items is 0.85.

Variables related to the conjoint experiment
The profiles in our conjoint experiment consist of four attributes with two possible values each ( Table 1): locus of causality, controllability, and stability (Weiner, 1985) are the independent variables, whereas the fourth attribute, a performance reference to other firms in the industry, is included as a control (Weisenfeld et al., 2008). The attributes are contrast coded (e.g., Holland and Shepherd, 2013;Patzelt and Shepherd, 2008) such that the variable representing each attribute takes either the value −0.5 or 0.5, thus giving a mean of zero for each variable (Table 3). A further control variable that originates from the conjoint experiment is the type of business failure, which is part of the framing of the experiment as explained above (see Table 1).

Variables related to the participant
To test Hypotheses 5a and 5b, the survey included a five-item scale for agreeableness adapted from the work of Schallberger and Venetz (1999). The five dimensions in the scale are altruism (i.e., concern for others), trust (i.e., belief in the benevolence of others' intentions), straightforwardness (i.e., appreciation of honesty in dealing with others), modesty (i.e., being humble and other-focused), and compliance (i.e., being meek and mild, preferring cooperative solutions to conflict). To test Hypotheses 6a and 6b, we included the German version of the established ten-item general self-efficacy scale (Luszczynska et al., 2005;Schwarzer et al., 1997). Consistent with our theoretical framework, we chose a general self-efficacy measure instead of a specific entrepreneurial self-efficacy measure because our conceptual focus is on the moderating influence of self-efficacy as a dispositional characteristic of the participant. Compared to general self-efficacy, which is a relatively stable trait-like characteristic, entrepreneurial self-efficacy incorporates environmental influences such that it can be elevated through entrepreneurial experience, training, and education (McGee et al., 2009). Although entrepreneurial self-efficacy is useful for examining the specific role of self-efficacy for entrepreneurial decision-making and behavior, it does not adequately capture the role of dispositional self-efficacy of evaluators from the general public as in our theorizing.
We used confirmatory factor analysis to test the discriminant validity of the two scales. A two-factor model, where both scales load on their intended factors, had a significantly better fit to the data than a one-factor model where the items of both scales load on a single factor (chi-squared test with one degree of freedom: 810.38, p b 0.000). The fit indices for the two-factor model suggested a satisfactory fit between the model and the data (Hu and Bentler, 1999): the comparative fit index (CFI) was 0.96; the root mean squared error of approximation (RMSEA) was 0.06; and the standardized root mean residual (SRMR) was 0.03. Hence, we computed composite indices for agreeableness and self-efficacy by averaging the respective item scores. The Cronbach's alpha coefficients for the scales are 0.95 for self-efficacy and 0.80 for agreeableness, which indicate high internal consistency.
Furthermore, our regression models include a number of control variables related to the participants. Because respondents who know former entrepreneurs who have experienced business failure in their personal or professional environments might judge entrepreneurial failure differently from those who do not know such people, we controlled for such vicarious failure experience with a dummy variable coded 1 if an individual had a personal or a professional relationship with a failed entrepreneur and 0 otherwise. Finally, individuals who are more satisfied with their life might evaluate entrepreneurs differently, therefore we controlled for satisfaction with life using the German version of the five-item Satisfaction with Life Scale developed by Diener et al. (1985). The Cronbach's alpha for this scale was 0.92 and a confirmatory factor analysis showed that it is distinct from self-efficacy and agreeableness (CFI = 0.96; RMSEA = 0.06; SRMR = 0.04; a three-factor model comprising life satisfaction, agreeableness, and self-efficacy with all items loading on their intended factors).
Additionally, we controlled for the respondents' employment status, gender, age, and education based on prior research that has found these factors to influence observers' judgments of failure (Shepherd and Patzelt, 2015;Weisenfeld et al., 2008). The respondents' employment status is reflected in the three categories of employed, self-employed, or not working. Gender is a dichotomous variable coded 1 for males and 0 for females, and age was calculated by subtracting the date of birth from the date of the response. Education was coded as a categorical variable with five levels ranging from early school leavers to college graduates. Moreover, we controlled for the net household income of the respondents, which was measured as a categorical variable ranging from b1000 EUR to N5000 EUR net monthly household income. Finally, we included dummy variables for all 16 German federal states on the grounds that observers' attitudes toward entrepreneurial failure might differ not only internationally but also within a country (Cardon et al., 2011). Table 2 presents the means, standard deviations and correlations for the variables related to the survey participants (excluding, for parsimony, the 16 federal state dummies). Because we treat education as an ordinal variable in the correlation matrix, the correlation coefficients are in the form of Spearman's rho.

Analysis and results
The conjoint experiment and the post-experiment survey generated data that is hierarchically structured. One observation consists of the attributes and a judgment of one failure profile (Table 1). There are eight observations nested within an individual respondent. We also have several variables pertaining to the individual respondent ( Table 2). The data structure has two important consequences for our analysis strategy. First, we cannot assume that the eight failure profile judgments by one individual are independent of each other, which can compromise the reliability of the standard error estimates if not accounted for in the analysis (Angrist and Pischke, 2009). Second, Hypotheses 5a-5b and 6a-6b require an estimate of cross-level interactions between variables related to a single judgment of a failure profile (locus of causality and controllability; level 1) and two variables related to the individual respondent (dispositional agreeableness and self-efficacy; level 2).
Based on these requirements, we chose to estimate hierarchical linear models with random-coefficient specifications to test our hypotheses. This technique solves the problem of non-independence for standard error estimation by distinguishing between observation-level (level 1) and individual-level (level 2) error components. It is also an appropriate method for computing crosslevel interactions in clustered data (Hox, 2010). In our models, each individual has their own intercept that is a linear function of an "average" intercept and an error term. Similarly, the coefficients of the attributes of the failure profiles depend on an "average" coefficient and an error component that is specific to the individual respondent. Table 3 includes estimates of the variance components: the standard deviations of the residual errors of the intercept and the level-1 coefficients and also their respective standard errors. The intraclass correlation coefficient for the intercept-only model is 0.30, suggesting that 30% of the variation in the model is explained by the grouping structure of the sample. This means that the characteristics of the individual respondent substantially affect how that person judges the eight failure profiles. Table 3 displays the estimates of five model specifications to test our hypotheses. Model (1) tests Hypotheses 1-3, which predict direct effects of the locus of causality, controllability, and stability on evaluative legitimacy judgments relating to failure. The model estimates show positive and significant effects for controllability and stability, which suggest that presenting failure as an uncontrollable and nonrecurring event is positively associated with legitimacy judgments. Thus, Model (1) supports Hypotheses 2 and 3, but it does not support Hypothesis 1 because the coefficient for the locus of causality is not significant.
Adding the interaction terms to test Hypotheses 4a and 4b in Model 2 sheds more light on the role of locus of causality in the judgment of failure. Both interaction terms (locus of causality * stability and controllability * stability) in the model are statistically significant. To investigate the interactions further, we computed the simple slopes and their standard errors for locus of causality and controllability at both levels of the moderator (stability) (Aiken and West, 1991). Figs. 2 and 3 present the results graphically, including the exact simple slopes and their p-values in footnotes.
Hypothesis 4a proposed that the effect of ascribing failure to external factors (the locus of causality) on judgment is stronger when stability is nonrecurring. However, the information presented in Fig. 2 does not support this hypothesis: the locus of causality is not significant when stability is nonrecurring, whereas it is positive and marginally tending toward the 5% significance threshold (p = 0.086) when stability is recurrent. Therefore, Hypothesis 4a is not supported. In contrast, Fig. 3 supports Hypothesis 4b, which proposed that attributing the cause of failure to be beyond the entrepreneur's control would be evaluated more positively when stability is described as nonrecurring rather than recurrent. Portraying the cause of failure as uncontrollable has a positive and significant effect on judgment, notwithstanding whether stability is nonrecurring or recurrent. However, this effect is greater in magnitude when stability is nonrecurring than when it is recurrent.
Model 3 adds the variables pertaining to the respondents' characteristics (level 2) to the equation. None of these variables exerts a significant effect on failure judgments. We checked the variance inflation factors for this model and found them well below the generally accepted limit of 10 (Kutner et al., 2004), suggesting multicollinearity is not an issue in our analysis. Furthermore, to control for potential endogeneity due to unobservables at the level of the individual respondent, we estimated Model 3 with a fixed-effects specification. The estimates of the individual coefficients and their standard errors are virtually identical to Model 2.
Models 4 and 5 test Hypotheses 5a-5b and 6a-6b by adding cross-level interactions between agreeableness/self-efficacy (level 2) and the locus of causality/controllability (level 1). The product terms for the interactions between locus of control and agreeableness and also between locus of control and general self-efficacy are significant, whereas the product terms for the interactions involving controllability do not reach the 5% significance threshold. We examined the significant interactions further by computing the simple slope of the independent variable (the locus of control) when the moderator (agreeableness or general self-efficacy) is set at one standard deviation unit above (high) or below (low) its mean, and when the stability attribute is set at recurrent or nonrecurring. Including stability in the analysis of simple slopes accounts for the fact that there is a significant level 1 interaction between the locus of causality and stability, which means that the effect of the locus of causality is not independent of stability. Thus, a failure to account for the two levels of stability in analyzing the interaction between agreeableness/self-efficacy and locus of causality could lead to incorrect results. It should be noted that the inclusion of stability in the computation of simple slopes does not mean that we computed a higher-order three-way interaction. It simply means that when computing, for example, the simple slope of locus of causality when agreeableness is high and low, we computed both simple slopes for when stability is recurrent and when it is nonrecurring. The resulting four simple slopes for each of the two significant interaction effects are displayed in Figs. 4 and 5.  Notes: 4808 observations nested within 601 individuals. Maximum-likelihood estimates. *, **, and *** denote significance at the 5%, 1%, and 0.1% levels. A negative coefficient indicates a more negative judgment and a positive coefficient indicates a more positive judgment.
Hypotheses 5a and 5b concern the moderating effect of dispositional agreeableness on the entrepreneur's distance-taking from a venture failure, which manifests in the cause of the failure being attributed to external and uncontrollable factors. Because there are theoretical arguments for positive and negative moderating effects, we proposed two competing hypotheses: dispositional agreeableness enhancing the positive effect of distance-taking (Hypothesis 5a) or it weakening the positive effect of distance-taking (Hypothesis 5b). The content of Fig. 4 shows that an external locus of causality has a positive effect on judgment when agreeableness is high (and stability is recurrent), whereas the effect is negative when agreeableness is low (and stability is nonrecurring). This finding highlights the central role of stability as a moderator of the relationship between the locus of causality and judgment. Attributing failure to external causes leads to a more positive judgment for observers with a high level of agreeableness but only if stability is described as recurrent. If observers have a low level of agreeableness, attributing the failure to internal causes is a better strategy for those seeking a positive legitimacy judgment, but only if failure is simultaneously described as being caused by nonrecurring factors. In terms of the competing Hypotheses 5a and 5b, the results provide support for either of them depending on whether the cause of failure is attributed to recurrent or nonrecurring circumstances. The moderating effect of general self-efficacy on the relationship between the locus of causality and evaluative legitimacy judgment is also dependent on whether stability is described as recurrent or nonrecurring. Fig. 5 shows that an external locus of causality has a positive effect on judgment when self-efficacy is low (and stability is recurrent), whereas the effect is negative when self-efficacy is high (and stability is nonrecurring). Thus, attributing failure to external causes is likely to be an effective strategy for achieving positive legitimacy judgments if the observers exhibit a low level of self-efficacy and stability is described as recurrent. However, if observers have a high level of self-efficacy, it would be more effective to attribute failure to internal causes and portray the causes of failure as nonrecurring. In terms of the competing Hypotheses 6a and 6b, we do not find support for the argument that high self-efficacy enhances the positive effect of distance-taking, so Hypothesis 6a is not supported. Because  a high level of self-efficacy can reduce the positive effect of distance-taking, we find conditional support for Hypothesis 6b. If the cause of failure is attributed to nonrecurring circumstances and observers have high levels of self-efficacy, not taking distance from the failure by ascribing its causes to one's own mistakes leads to a more positive legitimacy judgment than would attributing the failure to external factors.

Discussion
This analysis adds to the emerging literature on post-failure impression management in the entrepreneurial context (Shepherd and Haynie, 2011;Singh et al., 2015). It does so by examining the effectiveness of different impression management strategies Parhankangas and Ehrlich, 2014). More specifically, we demonstrate how different attributions of the cause of failure affect the general public's evaluative legitimacy judgments (Bitektine, 2011) of the failure in interaction with the evaluator's dispositional agreeableness and general self-efficacy. Below we outline how the above analysis contributes to theory development.
First, we extend evaluative legitimacy judgment theory (Bitektine, 2011;Bitektine and Haack, 2015) with attribution theory (Heider, 1958;Weiner, 2000). In doing so we offer a theoretical framework that explains how public audiences make evaluative legitimacy judgments, and how the way an entrepreneur presents the circumstances of the failure influences those judgments. Thus, we add to the explanation of the effectiveness of impression management for public audiences (Goffman, 1959;Morrison and Bies, 1991) by showing how the dominant attributions of entrepreneurial failure-the locus of causality, controllability, and stability (Weiner, 2000)-influence whether the audience members perceive the failure event as "acceptable" or "normal" (Cardon et al., 2011: 80). Developing this knowledge is important for at least two reasons. First, evaluative legitimacy judgments determine whether the evaluator socially supports or sanctions the failed entrepreneur (Bitektine, 2011), which in turn affects the experience of post-failure stigma (Singh et al., 2015;Weisenfeld et al., 2008). Second, opportunities for interaction are limited in the front-stage setting involving public audiences, and it is therefore important for entrepreneurs that the first public explanations create a legitimate venture failure impression (Elsbach, 1994;Semadeni et al., 2008).
Second, our findings add to the recent impression management research on entrepreneurial failure (Shepherd and Haynie, 2011;Singh et al., 2015) by demonstrating that failure impressions are effective when they ascribe failure to external, unstable, and uncontrollable forces, which we refer to as the entrepreneur's distance-taking from the failure event. These empirically robust insights are consistent with two impression management strategies derived from Shepherd and Haynie's (2011) conceptual work: those being denying responsibility and defining failure in a positive light. Both strategies aim to deliver legitimacy and manage social stigma by creating a positive self-view of the entrepreneur, where failure is symbolically decoupled (Überbacher and Jacobs, 2016) from the entrepreneur and presented as a unique event-a "perfect storm" (Shepherd and Haynie, 2011: 186). On the other hand, we find little empirical support for the effectiveness of impression management strategies that convey a negative self-view (Shepherd and Haynie, 2011). An internal attribution of failure is only effective in very specific situations: when the failure is also presented as being caused by nonrecurring circumstances, and the evaluator either has a low level of agreeableness or a high level of self-efficacy. We presume that in general, such strategies might be more effective when the entrepreneur can convey the impression in private dialogue with a small audience. In such situations, and in contrast to a situation that involves conveying failure to a broad undefined audience, the entrepreneur is more likely to trust that the audience's intentions are benevolent and that they will keep the justification of failure confidential. In conclusion, our findings suggest that distance-taking is a defensive-yet-effective impression management strategy (Morrison and Bies, 1991) to circumvent the common anti-failure bias in the public sphere (Cardon et al., 2011;McGrath, 1999;Semadeni et al., 2008). Third, our study has implications for the research on the relationship between post-failure impression management and entrepreneurial learning. Impression management theory suggests that defensive (in contrast to assertive) strategies imply that individuals do not seek, and are less likely to receive, constructive feedback (Bolino et al., 2008). The underlying assumption is that "individuals will be less likely to ask for feedback when they believe that the elicited feedback message will damage their public image" (Morrison and Bies, 1991: 526). Our findings support the effectiveness of this logic by demonstrating that distance-taking creates legitimate attributions of entrepreneurial failure. In other words, entrepreneurs who ascribe the failure event to external and uncontrollable circumstances are more likely to maintain professional legitimacy in the eyes of the public. However, a drawback of this strategy is the reduced likelihood of receiving constructive feedback from stakeholders who could support the entrepreneur's learning from failure (Cope, 2011;Shepherd, 2003;. Recently, Shepherd and Patzelt (2015): 273 argued that "there is little to learn from an event in which one had no control." We suggest that it is also difficult for public observers to reflect and comment on the entrepreneur's past behavior if venture failure is solely presented as something caused by external actors and events that were beyond that entrepreneur's control. Hence, denying involvement and controllability can reduce social pressure, and through that, support emotional recovery (Byrne and Shepherd, 2015), but possibly at the expense of limited feedback to spur learning from failure. This argument calls for further research on how entrepreneurs might manage evaluative legitimacy judgments while at the same time supporting emotional recovery and securing constructive feedback from stakeholders to support learning.
Fourth, we advance the evaluator perspective on entrepreneurial failure (Shepherd and Patzelt, 2015) by contributing to our understanding of how dispositional attributes of the observer and the attributes of the failure (Mantere et al., 2013) interactively influence legitimacy judgments of venture failure. In particular, our study generates three important results: First, our findings support the previously acknowledged role of perspective taking (Todd et al., 2012) in the evaluation of entrepreneurial failure (Shepherd and Patzelt, 2015) by showing that the higher the level of an observer's agreeableness (an interpersonal trait) (Graziano et al., 2007) the more likely she or he is to approve of a failure impression that ascribes failure to external causes. However, this relationship only holds if the entrepreneur describes failure as a recurring event. Second, impressions that attribute the cause of failure to the entrepreneur's own mistakes are judged more positively than external attributions if the evaluator has a low level of agreeableness, but only if the impression also describes the failure as a one-time, nonrecurring event. Accordingly, we suggest that observers with high levels of agreeableness are pro-socially motivated and thus feel more sympathetic (Graziano and Tobin, 2013) to an entrepreneur blaming stable external causes for the failure event. However, we argue that when failure impressions emphasize nonrecurring internal causes of the failure, observers with lower levels of agreeableness, who are less driven by pro-social motivation, appreciate entrepreneurs explaining their role in causing the failure.
Fifth, we find a significant negative moderating effect of an evaluator's self-efficacy (a personal trait) on the relationship between the locus of causality and the legitimacy judgment. In other words, observers with high self-efficacy levels disapprove of impressions that ascribe business failure to external causes. Therefore, our findings provide novel empirical support for traitbased cognitive abilities affecting the formation of evaluative legitimacy judgments (Bitektine, 2011) in the entrepreneurship context. Combining insights from Fiske's (1993) and Bandura's (2012) social cognition theories, we suggest that evaluators with high levels of self-efficacy adopt an internal, trait-oriented norm when making legitimacy judgments, and therefore those evaluators assume a strong association between the entrepreneur's capability and the failure event. However, considering the role of stability attributions in legitimacy judgments, we further suggest that evaluators with strong self-efficacy beliefs are also likely to trust the entrepreneur's claim that the failure is a challenge to be overcome and will not happen again. Hence, observers from the general public with high levels of self-efficacy are more likely (than those with low self-efficacy levels) to perceive impressions that ascribe business failure to internal nonrecurring causes as more legitimate than failure impressions that blame external factors. This finding points to the need for more detailed examinations of the role of efficacy beliefs in the context of venture failure. A productive starting point for such work could be the elaborate conceptualization of different forms of efficacy in the theory of mixed control (Monsen and Urbig, 2009).
In conclusion, we believe that our findings on the (inter-)personal traits affecting legitimacy judgments point to the need for further research that generates deeper insights into the interplay between evaluators' characteristics, attributes of failure impressions, and the formation of evaluative legitimacy judgments. Such research would contribute to a more detailed understanding of why observers make more or less harsh evaluations of entrepreneurial failure (Shepherd and Patzelt, 2015).

Limitations and future research
Our conjoint experiment offers the first robust evidence on what attributes of entrepreneurial failure impressions contribute to the general public's evaluations of the legitimacy of failure events. However, the study is not without its limitations. One such is our focus on the general public as an audience for failure impressions. The advantage of a focus on the general public is that it represents the average reaction to different impression management strategies of the diverse population of the failed entrepreneur's potential stakeholders. However, this is also its limitation: our results do not provide information on the potentially different reactions of specific groups of stakeholders. The general public may have a different understanding of failure attributions, especially compared to those of stakeholders closer to the failed venture. Those other stakeholders include significant others (e.g., friends, family members), who can strongly influence how an entrepreneur experiences post-failure stigma. In addition, potential employers-or customers, investors, and other business associates in the case of reentry into entrepreneurship-can critically affect the failed entrepreneur's future career prospects.
Furthermore, there are actors that can be influenced with impression management strategies that in turn shape the general public's perceptions of venture failure and its legitimacy judgments. Bitektine (2011) cites the role of the media in shaping public opinion and various regulatory actors that have the power to change the rules (e.g., through registration, licensing, certifications, regulatory sanctions) as influencing the way entrepreneurs in a given domain/industry must operate. Further study of these stakeholders and their role in shaping the entrepreneur's experience of post-failure stigma and post-career legitimacy would add to our understanding of impression management after venture failure.
Second, prior research on institutions and entrepreneurship suggests a strong association between national cultural values and entrepreneurship beliefs (Kibler and Kautonen, 2016;Stenholm et al., 2013). It is possible that our results based on the German culture are not fully generalizable to another cultural context, where failure in general might be perceived differently. Hence, to extend the knowledge gained from our single-country study, further cross-country research explaining potential differences in the macro-level institutional embeddedness of micro-level legitimacy judgments would be merited. Here, Bitektine and Haack's (2015) multi-level theory of legitimacy judgments could offer a useful framework for future studies aiming to explain an embedded evaluator perspective on entrepreneurial failure. That framework would allow an examination of how individual legitimacy judgments are subject to their national cultural and political-economic settings, and offer a more detailed analysis of both institutional stability loops in individual legitimacy judgments and legitimacy judgments of failure that are critical of existing macro-level institutions that, for example, can foster an anti-failure bias in society (Cardon et al., 2011;McGrath, 1999).
Third, although our selection of the factors influencing legitimacy judgments is based on attribution theory, which has been used in prior research on venture failure (Mantere et al., 2013) and is consistent with a number of impression management strategies outlined in Shepherd and Haynie's (2011) conceptual work, the three attributes in our analysis certainly do not form an exhaustive list of factors that characterize venture failure. Hence, more research is needed to address a wider range of attributes of failure that entrepreneurs can use to create legitimate failure impressions for the general public and other stakeholder audiences. One way forward would be to use the work of Bolino et al. (2008), which offers a list of different impression management behaviors, as a foundation for identifying and operationalizing further failure impression characteristics that would add to our current findings.
Fourth, although our study provides new insight into how the evaluators' personal characteristics affect how harshly they judge failure impressions (Shepherd and Patzelt, 2015), our study does not provide information as to how the harshness of the judgment affects the entrepreneur in terms of learning, wellbeing, and future career path (Singh et al., 2015). Future research might develop a more holistic picture of the legitimacy of entrepreneurial failure impressions by examining the failed entrepreneurs' characteristics, and how they affect the entrepreneurs' sensemaking of the business failure (Byrne and Shepherd, 2015;Mandl et al., 2016;Mantere et al., 2013), and also the way those entrepreneurs experience their stakeholders' judgment of failure, and in turn, the level of stigma that results from such judgments. For example, it would be important to develop a more nuanced understanding of how cognitive and emotional states as well as the interplay between core entrepreneurship constructs, such as opportunity confidence (Davidsson, 2016) and fear of failure (Cacciotti et al., 2016), influence the way legitimacy judgments affect an entrepreneur's wellbeing and decision on whether to re-enter entrepreneurship after a business failure (Shepherd and Haynie, 2011;Ucbasaran et al., 2013).

Conclusion
Complementing legitimacy judgment theory with attribution theory, this study presents a conjoint analysis to explain how observers from the general public evaluate the social legitimacy of entrepreneurial failure impressions. By explaining how observers make legitimacy judgments and what dominant failure attributions influence those judgments, we develop an understanding of the effectiveness of diverse impression management strategies for communicating the failure in a way that minimizes social stigma and sets a legitimate base for future career actions. The principal finding of our study is that the most effective impression management strategy available to entrepreneurs seeking post-failure legitimacy with the general public is to distance themselves from the failure. Entrepreneurs might therefore attribute the cause of failure to forces that are external to the entrepreneur/firm, not under the entrepreneur's volitional control, and subject to circumstances that are unlikely to occur again. Moreover, we provide insight into how the relationship between legitimacy judgments and impression management strategies that communicate distance-taking is moderated by the observer's level of dispositional agreeableness (an interpersonal trait) and self-efficacy beliefs (a personal trait).