Mind Your Language: Do Entrepreneurially
Worded Job Ads Deter Female Job Seekers
and Can Feminine Wording Help?
Alexandru Daniel Iosif
Student ID: 11792841
MSc Entrepreneurship
Universiteit van Amsterdam | Vrije Universiteit
Supervisor: Dr. Yuval Engel
31.07.2018
Preface
The copyright rests with the author, Alexandru Daniel Iosif. The author is solely responsible for the content of the thesis, including mistakes. The university cannot be held liable for the content of the author’s thesis.
Abstract
Prior research shows that masculine wording in job ads deters women job seekers while feminine wording encourages them to apply. Because entrepreneurship is associated with masculinity across cultures, entrepreneurial wording is expected to similarly deter women job seekers by activating a masculine stereotype and signaling women’s lack of fit. An experimental design was employed to test the extent to which job seekers’ intention to pursue a job were differently affected by job ad wording. Participants (99 women and 92 men) were randomized to view and evaluate job ads that either contained a large proportion of entrepreneurial or feminine words. Analysis of a moderated mediation model testing the gender conditional effect of job ad wording, via gender diversity perceptions and anticipated belonging, on job pursuit intentions indicated the model to be non-significant, except for the well-documented link between anticipated belongingness and job pursuit intentions. An extensive discussion of possible reasons and confounds that led to these results is offered and a systematic yet patient approach to tackling this complex issue in future research is endorsed.
Acknowledgements
I hereby extend thanks to my tutor, Dr. Yuval Engel, for his unflinching support and timely guidance, his high standards, but also for emboldening me to think and take decisions independently: He provided me with all the elements necessary for genuine and long-lasting learning, and for that I am grateful. I would also like to thank my parents, Zamfira and Vasile Iosif, for supporting me emotionally, as well as financially. Without you, this wouldn’t have been possible!
Table of contents Preface... ii Abstract ... iii Acknowledgements ... iv Table of contents ... v Introduction ... 1 Theoretical Framework ... 5
The Effect of Entrepreneurial and Gendered Wording in Job Ads on Perceived Organizational Diversity ... 5
The Link Between Perceived Organizational Diversity and Anticipated Belongingness ... 7
How Anticipated Belongingness is Positively Associated with Intentions to Pursue ... 10
Methods... 12
Design ... 12
Participants ... 12
Materials ... 13
Measures ... 15
Perceived Gender Diversity ... 15
Anticipated Belongingness ... 15
Job Seeker’s Gender ... 15
Intention to Pursue ... 15
Control Variables ... 15
Procedure ... 16
Analyses ... 17
Results ... 17
Scale Reliability and Validity ... 19 Randomization Check ... 20 Manipulation Checks ... 20 Hypothesis Testing ... 22 Discussion ... 24 Interpretation of Results ... 24 Theoretical Implications ... 29
Further Limitations and Suggestions for Future Research ... 30
Conclusion ... 33
References ... 34
Appendices ... 39
Appendix A. Entrepreneurial Job Advertisement ... 39
Appendix B. Feminine Job Advertisement ... 40
Appendix C. Entrepreneurial Job Advertisement (Word Counts) ... 41
Appendix D. Feminine Job Advertisement (Word Counts) ... 42
Appendix E. Control Variables and Response Options ... 43
Introduction
"But if thought corrupts language, language can also corrupt thought.”
- George Orwell
Research is clear on the effects of gender diversity within organizations: It pays off. It has a positive effect on overall business performance (Hoogendoorn, Oosterbeek, & Van Praag, 2013) as well as revenues, the number of customers, and profits (Herring, 2009). Gender-diverse organizations exhibit higher levels of employee motivation, as well as increased customer understanding and satisfaction (Desvaux, Devillard-Hoellinger, & Baumgarten, 2007; Herring, 2009). Furthermore, they benefit from comparatively better organizational images (Witt & Rode, 2005), as well as superior corporate social responsibility records, thus, better reputations (Bear, Rahman & Post, 2010; Kassinis, Panayotou, Dimou, & Katsifaraki, 2016). A more diverse talent pool can also be attracted by signaling diversity (Rau & Hyland, 2003), which, given the adjacent benefits, could prove highly favorable. Heightened levels of creativity, innovation and problem-solving are also among the advantages of gender heterogeneity (Campbell & Minguez-Vera, 2008; Østergaard, Timmermans, & Kristinsson, 2011), the presence of females in leadership positions being particularly beneficial for companies requiring high degrees of innovation (Dezsö & Ross, 2012).
This is likely relevant in the case of startup companies. These firms start off with limited resources and a lack of a strong reputation and historical data (Kraus, Harms, & Fink, 2009), which they need to outlast. Oftentimes, this is achieved by building favorable and constructive business cultures in order to attract investors and clients essential to their early (and continued) success, as well as personnel which fit their ambitions and values (Cardon & Stevens, 2004; Hoogendorn et al., 2011; Witt & Rode, 2005). Hiring the best staff available can also lend small firms a competitive advantage in a dynamic and aggressive market
(Barrett & Manyson, 2007). Nevertheless, a prominent gender gap still exists in entrepreneurship, this tendency being observed across countries and cultures (e.g., de Bruin, Brush, & Welter, 2006; Mitchell, 2011; Goktan & Gupta, 2015; Kanze, Huang, Conley, & Higgins, 2018; Kuppuswamy & Mollick, 2016), with the latest Global Entrepreneurship Monitor (2017) report emphasizing that “narrowing the gender gap in terms of entrepreneurial activity remains a priority focus for policy makers in all economies”.
The entrepreneurial stereotype is likely an important culprit, via its association with agentic traits such as risk-taking, autonomy, aggressiveness, innovativeness, and individualism, most also being stereotypically masculine characteristics (Eagly & Karau, 2002; Gaucher, Friesen, & Kay, 2011; Gupta & Bhawe, 2007; Gupta, Turban, Wasti & Sikdar, 2009; McDonnel & Morley, 2014; Prentice & Caranza, 2002). As such, whenever men pursue an entrepreneurial occupation, they very much match their societies’ archetypal roles, in consequence feeling empowered and supported (Gupta, Turban, & Bhawe, 2008; Thébaud, 2015). Women, however, are habitually described using communal traits (e.g., interdependent, warm, modest, submissive; Ellemers, 2018, Gaucher et al., 2011), and thus have to overcome social expectations and appraisals effectively rooting for their failure (e.g., Ellemers, 2018; Gupta et al., 2008; Gupta et al., 2009; Ramos et al., 2018). This assumed mismatch creates negative self-assessments, leading to perceptions of inability and feelings of insecurity, as well as increased self-critique, subsequently discouraging action (Ellemers, 2018; Gupta, Turban, & Pareek, 2013; Gupta, Goktan, & Gunay, 2014; Thébaud, 2010).
In the context of organizations, the use of gendered wording is likely a key medium for disseminating stereotypical information: Real-world job ads were found to contain a higher number of masculine words for male-dominated work domains, while feminine wording was relatively constant across both feminine and masculine domains (Gaucher et al., 2011). This is especially relevant, given that masculine wording/grammatical forms were
found to lead women, regardless of objective criteria such as skill-fit, to negatively assess their suitability regarding advertised jobs, while men were not significantly dissuaded (e.g., Gaucher et al., 2011; Horvath & Sczesny, 2016). Furthermore, the addition of feminine wording and reduction of masculine wording within ads positively affected women’s perceptions of the job being advertised (Gaucher et al., 2011). Especially in the case of startups, given their lack of reputation and history (Kraus et al., 2009), elements in the recruitment process, such as the wording, design, and tone of job ads, can be expected to significantly affect job-seekers’ perceptions of the organization (e.g., Chapman, Uggerslev, Carrol, Piasentin, & Jones, 2005; Phillips, Gully, & Castellano, 2014; Reeve & Schultz, 2004).
As aforementioned, there are many practical advantages that young firms may derive from having gender diverse teams (e.g., Campbell & Minguez-Vera, 2008; Dezsö & Ross, 2012; Hoogendorn et al, 2013). Furthermore, the propagation of gender stereotypes can be harmful to women and men alike in numerous ways (Ellemers, 2018). Consequently, the present research deems it beneficial to investigate an understudied mechanism which may be spreading detrimental stereotypes and gender-biasing staffing in startups: the presence of entrepreneurial wording in job ads. Thus, we ask:
To what extent does the presence of entrepreneurial wording in startup job ads negatively impact female job-seekers’ intention to apply, and does the use of feminine wording instead neutralize this negative effect?
The first part of the conceptual model (see Figure 1) under investigation is rooted in insights from signaling theory (Connelly, Certo, Ireland, & Reutzel, 2011; Spence, 1973), as well as past work on the effects of gendered wording on job-seekers (Gaucher et al., 2011) and stereotypes (e.g., Ellemers, 2018; Gupta et al., 2009) which together suggest that, all else kept constant, entrepreneurial words in job ads, given incomplete information, are likely to
signal male dominance, while feminine words, gender diversity. Furthermore, based on self-to-prototype matching theory (e.g., Bian, Lesile, Murphy, & Cimpian, 2018) and insights from stereotype activation theory (e.g., Gupta et al., 2008; Gupta et al., 2013), it is expected that the perceived degree of diversity within the company will affect anticipated belongingness, with gender as a moderator. Given that perceptions of belongingness have been shown to elicit a wide range of positive effects, including higher domain engagement and general well-being (e.g., Baumeister & Leary, 1995; Walton & Cohen, 2007), as well as increased job appeal (Gaucher et al., 2011), it is plausible that anticipated belongingness will, in turn, mediate the relationship between perceived gender diversity and the intention to pursue an opportunity (ITP; Highhouse, Lievens, & Sinnar, 2013).
Data to test this model is obtained using a web-based experiment in which potential job-seekers are presented one of two job ads, which were manipulated to only differ in the percentage of entrepreneurial and feminine wording. Following exposure to the stimuli, participants completed a survey containing the measures which make up the conceptual model, as well as various control items, manipulation checks and demographics.
The current study brings a number of theoretical and practical contributions. First, by utilizing words pertaining to the entrepreneurial (e.g., Short, Broberg, Cogliser, & Brigham, 2010) as well as feminine (Gaucher et al., 2011) stereotypes as stimuli, and aiming to explore their effects on job-seekers via multiple established measures, it serves to refine our understanding of stereotype activation dynamics in the context of entrepreneurship, by, for example, shedding light on the effects of a particular set of stereotypically entrepreneurial words, thus contributing to the relevant literature (e.g., Gupa et al., 2008; 2013; 2014). Second, it aims to uncover perceptions and reactions to organizations, as they are produced in job seekers by stereotyped recruitment material and may ultimately serve to refine hiring practices in startups, while contributing to the emerging body of startup HR research (e.g.,
Moser, Tumasjan, & Welpe, 2017; Phillips, Gully, & Castellano, 2014). Finally, if the proposed model is confirmed, it will provide insight into the psychological phases a job-seeker goes through when reading gender-biased information related to a potential workplace. Thus, for future research, it will indicate multiple avenues for designing experiments and interventions to better understand and control mental phenomena during recruiting, as well as implicitly exploring and confirming links between the several theories involved.
Theoretical Framework
The Effect of Stereotypical Wording in Job Adverts on Perceived Organizational Diversity
Signaling theory was initially developed based on observed knowledge gaps between organizations and prospective employees (Spence, 1973), but, being highly intuitive, was later adapted to a plethora of domains (e.g., HRM, business, financial markets; Connelly et al., 2011). It posits that a “receiver” will interpret a signal sent by the “signaler” in various ways to make sense of their attributes (Connelly et al., 2011; Spence, 1973), and would imply that job-seekers (i.e., receivers) need to rely on the information they have at their disposal, however limited it may be, if they are to gauge an organization’s (i.e., signaler’s) characteristics. This first contact often comes in the form of a job advertisement, the content within serving as the only proxy available for assessment. In the context of startups, as they often lack the reputation and historical data of more established organizations (Kraus et al., 2010), and job ads can only present very limited information, the resulting situation is necessarily defined by a high degree of informational asymmetry. This implies that the various cues in the recruitment material (e.g., wording, tone, layout) will become important signals to potential applicants, which they will promptly make use of to infer facts about the respective organization (e.g., Chapman et al., 2005; Goldberg & Allen, 2008; Phillips et al., 2014). The applicant might further interpret these and other unintentionally exhibited cues so
as to evaluate and deduce the organization’s personality, meaning subjective elements will at least in part affect the perceived suitability between the employee and potential employer (Lievens & Slaughter, 2016). Accordingly, research has shown that wording affects job seeker’s impressions of the organization being advertised (e.g., Born & Taris, 2010; Feldman, Bearden & Hardesty, 2006; Gaucher et al., 2011; Rynes & Cable, 2003), and that signaling via recruitment materials can influence their attitudes, as well as choices (Cable & Turban, 2003; Rynes et al., 1991).
It is well documented in the literature that gendered signaling within job ads might create noteworthy effects in their readers (e.g., Born & Taris, 2010; Gaucher et al, 2013). Gender stereotypes are deeply embedded within language, their existence being acknowledged in various situations (e.g., differential language use of females/males as well as when writing about them; Leaper & Ayres, 2007; Madera, Hebl, & Martin, 2009), a vast amount of evidence converging around the themes of communal attributes (e.g., loyal, empathetic) for women and agentic ones for men (e.g., independence, risk-taking; Eagly & Karau, 2002; Ellemers, 2018). Relevant to the current study, Gaucher and colleagues (2011) found that the percentage of gendered wording in adverts influenced job-seekers’ perceptions of gender diversity within the organization advertising the opening – i.e., an increased amount of masculine wording caused readers of both genders to expect more males to work in the company, while female wording caused job-seekers to perceive relatively more females within the organization. Thus, gender-stereotypical wording can shape in a very concrete and profound manner a reader’s perception regarding an organization’s character (Gaucher et al., 2011). As mentioned before, and germane to the startup ecosystem, the entrepreneurial stereotype, by being associated with agentic traits such as risk-taking, autonomy, independence and aggressiveness, is much akin to the masculine stereotype (Eagly & Karau, 2002; Gupta & Bhawe, 2007; Gupta et al., 2009; Short et al., 2010). In fact, this link with
masculinity is so deeply ingrained within the societal discourse on entrepreneurship (Ahl, 2006; Lewis, 2006), that researchers and the media may effectively end up ignoring most feminine forms of entrepreneuring, or conceptualizing them as outliers, thus giving rise to biased reporting and gendered theory-building, which serve to maintain the gender gap plaguing the entrepreneurial profession (Hamilton, 2013).
Converging these lines of argument, it can be expected that job ads for startups which contain words pertaining to the entrepreneurial stereotype (e.g., aggressive, innovative, revolutionary, risk-taking; see Short et al., 2010 for a comprehensive list of terms), are likely to signal to readers a “masculine” organization, ultimately decreasing the expected proportion of females working within said organization (Gaucher et al., 2011). The opposite is expected to be true for an increased percentage of feminine wording (e.g., Gaucher et al., 2011; Gupta et al., 2013; 2014). As stereotypes are likely stable in content for members of both genders (Eagly & Karau, 2002; Ellemers, 2018), differing mainly based on culture/language (Cuddy et al., 2015), this judgement should remain stable regardless of the job seeker’s gender, instead fluctuating as a function of the proportion of stereotyped wording within the text (Gaucher et al., 2011). Thus, we hypothesize:
H1: The percentage of entrepreneurial words within job ads is negatively related to perceived organizational diversity, such that participants reading a job ad containing a lower number of entrepreneurial words and a higher number of feminine words will perceive a significantly higher percentage of females within the organization than those reading ads containing comparatively more entrepreneurial words.
The Link Between Perceived Organizational Diversity and Anticipated Belongingness
The perceived gender proportions within the organization are expected to affect the sense of anticipated belongingness within the target group, on the basis of gender (e.g., Murphy, Steele, & Gross, 2007). Insights further substantiating this claim stem from
self-to-prototype matching theory, which posits that people rely on their self-concepts to assess suitability in a number of domains, similarities between their characteristics and those of the perceived archetypal member of the target group being desirable (e.g., Bian, Lesile, Murphy, & Cimpian, 2018). Initial evidence for this mechanism came from a study of students’ prospective housing choices, which showed that participants preferred housing in which the prototypical tenant was similar to the image they held about themselves (Niedenthal et al., 1985). Ever since, the same theory was successfully applied to explain, for example, women’s underrepresentation in male-dominated fields such as computer science (Cheryan, Plaut, Davies, & Steele, 2009), as well as men’s underrepresentation in female-dominated fields, such as English (Cheryan & Plaut, 2010). Furthermore, Bian and colleagues (2018) demonstrated that women were led by a perceived mismatch between them and the prototypical member of a professional or educational field, more so than by stereotype threat (i.e., the expectation that they will be potentially negatively stereotyped), to feel anxiety and lack of belongingness within said field, which in turn mediated the interest to pursue the presented opportunity. Importantly, the perceived incompatibility was prompted in women by associating the presented fields with requirements widely held to be analogous to the male, but not the female stereotype (i.e., the trait of “brilliance” being essential to success; Bian et al, 2018; Bian, Leslie, & Cimpian, 2017).
Relevantly, and in line with self-to-prototype matching theory, Gaucher and colleagues (2011) found that job advertisements containing masculine wording were shown to decrease women’s anticipated belongingness within prospective positions, while men, albeit not to a statistically significant degree, displayed a lower level of belongingness in jobs advertised using a greater percentage of feminine wording. Corroborating these lines of evidence, findings from stereotype activation theory showed that implicit activation of a gender stereotype (i.e., via subtle cues, without explicit reference to stereotyped group), leads
to stereotype assimilation (i.e., thinking or acting in line with the presented stereotype; e.g., Gupta et al., 2008; 2013). Following a description of an experienced (i.e., prototypical) entrepreneur using masculine traits (e.g., aggressive; autonomous), a potential business opportunity was evaluated less favorably by women than by men – the opposite was true when the characteristics describing the entrepreneur were feminine (e.g., caring, humble; Gupta et al., 2013). Nonetheless, it is worth noting that other studies only found a gender interaction for male wording, the researchers tentatively remarking that this might be due to the fact that some professions, such as entrepreneurship, are too intensely associated with masculinity to be activated as a feminine profession (e.g., Gupta et al., 2008).
As aforementioned, the commonly held stereotypes about gender define women as communal (e.g., affectionate, helpful, kind) and men as agentic (e.g. ambitious, aggressive, assertive; Eagly & Karau, 2002; Ellemers, 2018; dichotomy also used by Gaucher et al., 2011), the latter sharing attributes with the entrepreneurial stereotype, which emphasizes agentic traits as essential to success (Gupta et al., 2009; Short et al., 2010). Finally, entrepreneurship is a field widely associated with masculinity within society (Ahl, 2006; Hamilton, 2013).
Linking these facts, when an employer’s workforce is perceived to be dominated by men, or instead, as more gender diverse, it can be expected that job seekers’ anticipated belongingness is gaged through the process of comparison between one’s own characteristics and those of the prototypical member of the potential team - characteristics which will likely be inferred based on the stereotype implicitly activated via the stimuli (i.e., words) they are exposed to (Bian et al., 2018; Gaucher et al., 2011; Gupta et al., 2013). Thus, females perceiving a majority of males within the organization will rate their similarity to the prototypical team member (i.e., a man with masculine attributes), and therefore their belongingness, as lower, while the opposite would be true for a more diverse team; the same
should conceivably apply for males perceiving more females within the organization, as a less optimal fit between the self and the prototypical employee will be assumed (e.g., Bian et al., 2018; Gupta et al., 2013). In consequence, we hypothesize:
H2: The relationship between perceived organizational diversity and anticipated belongingness is moderated by job seekers’ gender, such that this relationship is positive for women while negative for men.
How Anticipated Belongingness is Associated with the Intent to Pursue an Advertised Job
The positive effects of heightened belongingness, which has been described as “a fundamental human motivation” (Baumeister & Leary, 1995), are well documented: It was found that feeling they belonged improved achievement among African American college students (i.e., a stigmatized group) to levels similar to those of their white peers (measured via grade point averages and dropout rates; Walton & Cohen 2007), as well as boosting their health and general wellbeing (Walton & Cohen, 2011). Furthermore, a lower sense of belonging induced by gender stereotypes lowered female high school students’ interest to pursue mathematics, or care for their grades (Good, Rattan, & Dweck, 2012), lowered women’s intention to pursue professional or educational opportunities (Bian et al., 2018), as well as their intentions towards a career in computer science (Cheryan et al., 2009). Much in line with the current model, a sense of anticipated belongingness mediated the relationship between job appeal and job ads containing either feminine or masculine stereotypical wording (Gaucher et al., 2011).
In the context of HR, intentions to pursue were shown to be a better predictor of actual behavior (i.e., likeliness to request more information about a potential employer after reading recruitment materials) than plain attractiveness (Highhouse et al., 2003). This was first established in the context of the theory of reasoned action (Ajzen & Fishbein, 1980). Its
principle of correspondence suggested that relations between attitude and behavior are stronger if the attitude and behavior are measured at the same level of specificity, meaning that the closest proxy should always be chosen (e.g., if one were to measure likeliness of people to go to a music festival, their intention to buy a ticket is a better predictor than general attractiveness towards the festival; Ajzen, 1985; Highhouse et al., 2003). Indeed, the proposition that intentions predict behavior better than attitudes (i.e., attractiveness) is well supported within the literature (e.g., in a meta-analysis of 139 studies; Kim & Hunter, 1993).
Together, these lines of argument suggest that, in the context of investigating the effects of wording on job-seekers, anticipated belongingness will be a strong predictor of intentionality, and thus, of the behavior that readers are likely to eventually engage in (Gaucher et al., 2011; Highhouse et al., 2003). In consequence, we hypothesize:
H3: Anticipated belongingness is positively related to job pursuit intentions.
All considered, we expect that a moderated serial mediation between job ad wording and job pursuit intentions will exist, via perceptions of gender diversity and anticipated belongingness, dependent on gender.
H4: The relationship between stereotypically worded job advertisements and intentions to pursue the job is sequentially mediated via (1) perceptions of gender diversity; and (2) anticipated belonginess, while the participant’s gender moderates the path between perceived gender diversity and anticipated belongingness, such that it is stronger for female than for male job-seekers.
Figure 1. Model and Hypotheses.
Methods Design
An experimental design, in which participants were randomly assigned to one of the two groups, was used to test the hypotheses. One group was presented a job advertisement containing a high proportion of entrepreneurial words, the other, a far lower proportion of entrepreneurial and higher proportion of feminine words (See Appendices C and D). Once they read the ads, the participants were asked to fill out a survey, which in turn produced the experimental data.
Participants
The 191 participants were 92 males (48.16%) and 99 females (51.84%). Their age was on average 23.3 (SD = 3.36; Range = 18 - 35), and they were mainly Dutch (124; 65%), the rest being a highly heterogeneous mix of nationalities. The majority of participants were current students or alumni from the Vrije Universiteit (VU) of Amsterdam, who were contacted via email using the openly available address book in the university’s Outlook Webmail, which permitted selection and sending of messages to accounts marked as “Student” within the database. Some others were collected using social media shares or direct
requests from acquaintances fitting the requirements (~15 acquaintances in total). All the distributed links were anonymous; thus, exact sources of respondents nor response rates can be computed.
Initially, there were 258 recorded responses, out of which 67 were excluded according to a number of pre-set criteria: Failing at least one of the two attention checks (4 subjects); The ad timer indicating that the ad (209 words; see Appendices C, D) was read for less than 20 seconds (trained readers were found to read at speeds of 600 words per minute or above; Bell, 2001) or more than 350 seconds (as it was deemed unnatural and likely to impact results via lack of focus; 19 subjects); Answering “no” to all questions related to being a job-seeker (i.e., Looking for a job in the past 12 months, present, or next 12 months), as none being affirmative meant that there was no proximal personal context to draw on when imagining one was actually looking for a job, thus decreasing response realism (45 subjects); Not understanding English at least “Moderately Well” (8 subjects; however, excluded due to data regarding English level missing – all other participants indicated sufficient English understanding); Being under 18 or over 35 years old, as it was decided that this customary young adult age cohort would also be most relevant as startup employees (6 subjects); Not having finished high school (0 subjects). Some subjects met more than one exclusion criterion. Therefore, adding up the individual exclusion cases will result in a number higher than the 67 participants excluded in total (i.e., 82).
Materials
To be able to conduct the experiment, two startup-like job ads were created (see Appendix A and B), designed to the likeness of recruitment materials found on various ad portals (e.g., Monsterboard, Iamsterdam). The two resulting ads differed mainly in terms of wording: The “entrepreneurial” ad (see Appendix C) contained 10.05% entrepreneurial words (e.g., entrepreneurial, innovation, startup; extracted from Short et al., 2010 who conducted a
computerized text analysis to identify commonly used words in relation to entrepreneurship). The “feminine” ad (see Appendix D) contained 2.39% entrepreneurial terms and 5.75% feminine terms (e.g., loyal, compassionate, empathic; extracted from the list compiled by Gaucher et al., 2011 for their study). Finally, both of them contained three masculine words amounting to 1.43% (i.e., aggressive/leader, independent, competitive; Gaucher et al., 2011) - this was done to avoid the possible confounding effect of stereotypically masculine wording.
Content-wise, the two ads were designed to fit as many types of job-seekers as possible. Therefore, the text listed some mandatory soft skills/attitudes (e.g., communication, leadership, team-work, adaptability, versatility etc.), as well as typically advertised perks (e.g., competitive salary, beautiful office), while mentioning that the company is “currently recruiting for a variety of (professional as well as entry-level) positions”. Thus, one would be expected to apply in order to find out more about the actual position he may fill. The only indication regarding what the company does was that they are perfecting their technology and are backed by top investors – nonetheless, it was assumed that a growing technology company can be expected to offer a variety of jobs for people from most backgrounds (e.g., for STEM as well as humanities students).
A short (due to time constraints) pilot test was also run – three people with experience in hiring/HR (i.e., Arjan Pelders, with experience in team development; Roland van den Brand, a CEO who lists “search and selection” among his specialties; Patrick Trikels, another business owner) were asked via email to complete a small survey to assess the credibility of the ads compared to real-world ones and also to provide suggestions for improvement. All averages below were recorded on a 5-point scale (1-5; Strongly Disagree, Somewhat Disagree, Neither Agree nor Disagree, Somewhat Agree, Strongly Agree).
Although the entrepreneurial ad was not perceived as very realistic (M = 2.33), it was seen as quite typical of a startup firm (M = 3.67). When asked what affected their
perceptions, the focus on startups and culture, as well as traits deemed appropriate for a young business (i.e., creativity, flexibility, enterprising tendencies) were pointed out. Recommended improvements were adding more detail (intentionally left out to appeal to a wider audience), as well as focusing more on teamwork, as the ad was too individualistic. The feminine ad was perceived as more realistic (M = 3.33), but not as typical of a startup firm (M = 2.67). Perceptions were affected by the teamwork element and the managerial orientation of the ad. Again, reducing vagueness was a common suggestion for improvement. Finally, when asked to compare the two, answers were “both immature, the first naïve and optimistic, the second more soft skills oriented”; “creative entrepreneurs vs. managers (wanted)”; “the second ad is more professional and triggers more”.
Measures
The measures presented here were essential to testing the experimental hypotheses, as well as control for vital factors which may affect results. Scale reliability/validity, as well as randomization/manipulation checks, are further discussed in the Results section.
Perceived Gender Diversity. Perceived gender diversity was measured using an
11-point Likert-scale item (i.e., “What is the proportion of female employees”: 0% – 100% in 10% increments). Participants were asked to estimate the percentage or take a guess of the answer if they are not sure. This measure was adapted from Gaucher and colleagues (2011).
Anticipated Belongingness. The four-item scale of anticipated belongingness (α =
0.842) was entirely adopted from Gaucher et al. (2011). It was a 7-point Likert scale ranging from 1 (Strongly Disagree) to 7 (Strongly Agree). It contained items such as “I’m similar to the people who would apply for this job” and “I could fit in well at this company”. The scale contained one reverse score item (i.e., “The type of people who would apply for this job are very different from me”) which was recoded in order to permit the calculation of an average belongingness score, needed for the analyses.
Job Seeker’s Gender. This item asked participants to indicate their gender (i.e.,
Male/Female/Prefer not to say).
Intention to Pursue. To assess job pursuit intentions, a scale containing five 7-point
Likert-scale items (α = 0.820) was adapted in its entirety from Highhouse and colleagues (2003). Examples of individual items are “If this company invited me for a job interview, I would go” or “I would accept a job offer from this company”. It was also averaged for the analysis.
Control Variables. Variables included for control or manipulation assessment
purposes took into account demographics (i.e., Age; Gender; Nationality), employment status (i.e., Current Employment; Looked for job in last 12 months; Currently looking for job; Expecting to look for job within next 12 months; General willingness to work for a startup), education (i.e., Highest completed educational level; Study program), attention checks (see Appendix E), main manipulation check items (i.e., assessing how entrepreneurial the job ads were perceived to be and a qualitative manipulation check, prompting a written explanation regarding what affected one’s perception of the ad, based on Gaucher et al., 2011; see Appendix E), ad-related variables/gender-based manipulation check items (i.e., Ad view timer; Ad realism assessment; Extent of imagination effort; Gender expectation estimates based on ad: e.g., Imagining a female while reading ad, Company looking for males rather than females; see Appendix E), as well as an English language understanding item (see Appendix E for a full list of control variables and their respective measurements).
Procedure
Participants were contacted via email using the openly available address book in the VU university’s Outlook Webmail, which permitted selection and sending of messages to accounts marked as “Student” within the database, as well as using social media shares or direct requests from acquaintances fitting the requirements (~15). They were invited to fill
out an online survey constructed on the Qualtrics platform, via a direct, anonymous link. The ones who accessed the link were met with a small consent page which contained the purpose of the study and a statement of confidentiality. Once they consented, they first had to answer questions related to their age, employment status/plans, and willingness to work in a startup, before being presented with a message regarding the importance of reading the ad that was to follow as attentively as possible. Then, one of the ads (Appendix C or D) was randomly presented. Thereafter, the participants were led through all of the remaining items (i.e., in this order; DV: Intention to Pursue; Mediators: Gender Diversity, Anticipated Belongingness; Qualitative manipulation check item; Demographics; Control Variables). Finally, a “Thank You” and debrief page was displayed. The collected data were downloaded for analysis directly in SPSS format via Qualtrics.
Analyses
Most analyses were carried out using SPSS (v24.0 64bit edition for Mac OS). These were, in the order in which they were done: Obtaining the descriptive statistics reported in the “Participants” section, obtaining the descriptive statistics and correlation coefficients reported in Table 1, computing the chi-squared tests, independent samples t-tests and Mann-Whitney U tests needed for the randomization and manipulation checks, as well as running the Univariate ANOVA analysis needed for testing hypothesis 1. Furthermore, Jamovi (v0.9.1.6 for Mac OS) was used to do an Exploratory Factor Analysis to determine whether the three scales in the study were truly measuring separate constructs, as well as to assess the reliability of these scales (by computing Cronbach’s α).
For testing the significance of the moderated serial mediation model displayed in Figure 1 and implicitly, the validity of the 4 hypotheses, the PROCESS macro for SPSS (PROCESS v3.0; Hayes, 2018; Model 91) was used to conduct a conditional process analysis, with 10000 bootstrapped samples so as to account for skewness in the data and
produce the 95% confidence intervals (10K deemed appropriate for most applications; Hayes, 2013).
Results Descriptive Statistics and Intercorrelations
Descriptive statistics (i.e., Means, Standard Deviations), as well as correlations between relevant variables (using Pearson’s r), are summarized in Table 1. As expected, Anticipated Belongingness correlated with Intention to Pursue, r = .588, p < .001. Importantly, Nationality significantly correlated with both Anticipated Belongingness, r = .241, p < .001, p = .003, and Intention to Pursue, r = .286, p < .001, as did Openness to Startups, also with Anticipated Belongingness, r = .304, p < .001, and Intention to Pursue, r = .270, p < .001.
Table 1
Descriptive Statistics and Correlations
N Mean SD 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.
1. Condition 191 1.48 0.50 —
2. Perceived Gen. Div. 191 4.99 1.26 .058 —
3. Anticipated Belonging 191 4.73 1.02 -.024 .052 — 4. Intention to Pursue 191 4.95 0.95 -.102 .085 .588*** — 5. Gender 191 0.52 0.50 -.213** .075 -.030 .053 — 6. Age 191 23.3 3.33 .030 -.047 -.032 -.006 -.114 — 7. Nationality 191 0.35 0.48 -.042 .163* .241*** .286*** .116 .290*** — 8. Educational Level 191 7.53 1.73 -.001 -.115 -.096 -.126 -.039 .474*** .016 — 9. English Level 191 4.72 0.64 -.054 -.069 .217** .104 -.083 .079 .293*** .060 — 10. Employment Status 191 1.55 0.98 -.037 0.00 .040 -.085 -.128 .154* -.058 .146* -.085 — 11. LFJ – Past 12m? 191 0.61 0.49 -.167* -.092 .065 .055 -.035 .024 -.136 .038 .001 .209** — 12. LFJ – Now? 191 0.32 0.47 -.001 .015 .080 .039 -.104 .206** .038 .037 .040 -.021 .245*** — 13. LFJ – Next 12m? 191 0.82 0.38 -.049 .050 .017 .068 .017 .110 .170* .055 -.035 -.100 -.314*** .113 — 14. Open to Startup? 191 0.84 0.37 .050 .008 .304*** .270*** -.083 .016 .086 -.071 .005 -.026 -.117 .027 .166* — 15. Entrepreneurial Scale 191 5.40 0.94 -.342*** .028 .164* .262*** .157* -.071 .048 .001 -.075 .126 -.062 .045 .113 .177* —
Note. See Appendix E for a full list of variable measurements and codes.
* Correlation is significant at the 0.05 level (2-tailed). ** Correlation is significant at the 0.01 level (2-tailed). *** Correlation is significant at the 0.001 level (2-tailed).
Scale Reliability and Validity
An Exploratory Factor Analysis was run to ascertain whether the 12 items pertaining to the three scales used (i.e., Anticipated Belongingness; Intention to Pursue; Manipulation Check) were indeed measuring distinct constructs in the present sample. An oblimin (oblique) rotation was used. The analysis (Appendix F) yielded three factors explaining a total of 56% of the variance in the data. The items in the “Anticipated Belongingness” scale loaded on a distinct factor with values between 0.590 and 0.923 (21.1% of variance); the items in the “Intention to Pursue” scale loaded on a distinct factor with values between 0.523 and 0.761 (19.5% of variance); finally, the factors in the constructed manipulation check scale loaded on a distinct factor with values between 0.670 and 0.871 (15.4% of variance). It is worth noting that the ITP and Belongingness scales were found to correlate highly (r = .681); however, this was expected, as they are conceptually related items. The “Factor Loadings” table in Appendix F shows all factor loadings, hiding values below 0.25. Furthermore, the Kaiser-Meyer-Olkin test for sampling adequacy yielded a very good result (i.e., 0.845; see Appendix F), and Bartlett’s test of sphericity was highly significant, χ2(66) =
991, p < .001, indicating that the data used for the analysis was appropriate and that the obtained factor loadings are valid. In consequence, the analysis confirms that the tree scales employed likely measured separate constructs in the present data set.
The scales were further tested for internal consistency/reliability, using Cronbach’s α. All three scales indicated adequate reliability (“Anticipated Belongingness”: α = 0.842, M = 4.73; SD = 1.02; “Intention to Pursue”: α = 0.820, M = 4.95; SD = 0.95; “Entrepreneurial Manipulation Check” = α = 0.807, M = 5.40; SD = 0.943). Furthermore, the alternate analyses indicated that removing any independent item would lower overall α, so the scales were used unedited, as initially intended.
Randomization Check
Analyses were run to assess whether participants in the two conditions differed on key variables, which may impact the results. Participants in the two groups did not significantly differ in terms of age, t(189) = -.415, p = .679, current employment status, χ2(3) = 1.650, p =
.648, present job-seeker status, χ2(1) = .984, p = .984, future 12 months expected job-seeker
status, χ2(1) = .465, p = .495, openness to working in startups, χ2(1) = .483, p = .487, job ad
viewing time, t(189) = .012, p = .991, nationality, χ2(1) = .340, p = .560, overall education
level, t(189) = .010, p = .992 | U = 4376, p = .621, degree topics pursued, χ2(16) = 8.573, p =
.930, imagination effort while doing the survey, t(189) = .998, p = .319, and English language understanding, t(189) = .745, p = .457 | U = 4336, p = .516.
Participants in the two conditions did significantly differ in terms of past 12 months job-seeker status, χ2(1) = 5.303, p = .021 (Condition 1 – Yes: 69%, No: 31% | Condition 2 –
Yes: 52.7%, No: 47.3%), which was not deemed highly problematic, as they did not also differ in terms of present or near future job-seeker status. They also differed in terms of gender, χ2(1) = 8.691, p = .003 (Condition 1 – Male: 38%, Female: 62% | Condition 2 –
Male: 59.3%, Female: 47.3%), which is more directly relevant for the analysis, yet also unalterable.
Manipulation Checks
Quantitative manipulation check items were included in the survey, to assess whether the wording manipulation had the expected effect. Three 7-point Likert scale items measuring the extent to which the ads were perceived as entrepreneurial (see Appendix E) were averaged due to their high internal consistency (α = .815). An independent samples t-test revealed a highly significant difference, t(189) = 6.113, p < 0.001, between the entrepreneurial condition (M = 5.70, SD = 0.83) and the feminine condition (M = 5.05, SD =
0.94), therefore indicating that the experimental manipulation worked as expected (i.e., the feminine condition was indeed perceived to be less entrepreneurial).
Based on Gaucher and colleagues (2011), a qualitative manipulation check item was also included (see Appendix E), to assess what precisely affected the participant’s perceptions, and whether the manipulation in wording was abundantly evident: Participants described the company/ads in a multiplicity of ways (full list available on request), but none mentioned anything that would indicate that they perceived the ad to be artificial or unrealistic (i.e., obviously manipulated). A few did point out that “wording” affected their perception, however, that can be said of any read text – without further explanations (which were not offered), this cannot be interpreted as a failure of the manipulation. Furthermore, some participants pointed out the lack of specific job requirements (i.e., consciously adopted characteristic to enable as many people as possible to identify with the job ad), however, this was not perceived as a manipulation, but rather as a customary annoyance (i.e. “it was the […] usual job ad with all the buzzwords and nice concepts, but no actual description of a position. 90% of job ads have the same content”). Consequently, no reasons for concern emerged from the responses to the qualitative manipulation item.
To assess whether the manipulation effort affected participants differently in terms of gender-related perceptions, this being deemed relevant due to the hypothesized link between gendered wording and job-seeker perception, three more items were added (see Appendix E): the entrepreneurial and the feminine conditions were not found to differ significantly in terms of picturing a female as the prototypical employee, t(176.39) = -.952, p = .342 (equal variances not assumed; M = 3.33, SD = 1.29 vs. M = 3.53, SD = 1.54), thinking the company is looking for male rather than female employees, t(189) = .952, p = .342 (M = 3.69, SD = 1.42 vs. M = 3.51, SD = 1.23), as well as thinking that the job is more attractive for males than females, t(189) = 1.229, p = .221 (M = 4.14, SD = 1.52 vs. M = 3.88, SD = 1.39).
Finally, the ads were also compared in terms of their perceived realism (7 Point Likert Scale, see Appendix E): Participants in the “Entrepreneurial” condition (M = 5.13; SD = 1.30) were found to differ significantly, t(173.53) = 2.009, p = .046 (equal variances not assumed), from those in the “Feminine” condition (M = 4.70; SD = 1.60) in terms of perceived realism. Nonetheless, as both averages were in the affirmative range of the Likert scale (i.e., above the neutral level “4”), it was concluded that, in line with the results of the pilot test, materials in both conditions were perceived as reasonably realistic. As such, data obtained based on them was deemed acceptable for testing.
Hypothesis Testing
The results of the conditional process analysis are presented below. Figure 2 displays the unstandardized coefficients of the predictors of perceived gender diversity, anticipated belongingness, and intention to pursue, as well as significance levels.
Hypothesis 1 proposed a higher percentage of entrepreneurial wording within a job ad (i.e., condition 1) would lead, regardless of the job seeker’s gender, to a significantly lower perceived level of gender diversity, as opposed to an ad containing a lower percentage of entrepreneurial wording and a higher percentage of feminine wording (i.e., condition 2). The conditional process analysis indicated, quantified in the first model with Perceived Gender Diversity as an outcome variable, F(1, 189) = .633, p = .427, R2 = .003, that wording did not
significantly predict perceived organizational diversity, b = .146, t(189) = .795, p = .427. Furthermore, a 2x2 (Condition: Entrepreneurial, Feminine x Gender: Male, Female) ANOVA revealed no significant effect of condition, F(1, 187) = 1.127, p = .290; gender F(1, 187) = 1.430, p = .233, or of the gender x condition interaction, F(1, 187) = 1.123, p = .291, on perceived organizational diversity, therefore, hypothesis 1 was not supported.
Hypothesis 2 proposed that the relationship between perceived organizational diversity and anticipated belongingness would be moderated by gender, so that higher
perceived organizational diversity will lead females to feel like they would belong more than perceived male dominance, or more generally, a lower level of organizational diversity, favoring males. The results of the conditional process analysis, quantified in the second model with Anticipated Belongingness as an outcome variable, F(4, 186) = .610, p = .655, R2
= .013, indicated that gender diversity did not have a significant effect on anticipated belongingness, b = -.020, t(186) = -.252, p = .801, with a non-significant main effect of gender, b = -.808, t(186) = -1.321, p = .188, and a non-significant interaction (i.e., perceived organizational diversity x gender) effect, b = .145, t(186) = 1.219, p = .224. Furthermore, taken separately, the effects were non-significant for both males, b = -.0016, 95% CI [-.0240, .0179], and females, b = .0099, 95% CI [-.0201, .0560]. Therefore, hypothesis 2 was not supported.
Hypothesis 3 proposed that anticipated belongingness will be positively related to intentions to pursue a job. The conditional process analysis, quantified in the third model with Intention to Pursue as an outcome variable, F(3, 187) = 34.63, p < .001, R2 = .357, revealed
that this was the case, anticipated belongingness significantly predicting intention to pursue, b = .543, t(187) = 9.924, p < .001. In consequence, hypothesis 3 was supported.
Hypothesis 4 proposed that the wording in the two job ads would affect perceptions of organizational diversity in all participants, which would then go on to affect perception of anticipated belongingness in job-seekers, based on their gender, such that e.g., females would feel more welcome in a more diverse team than a male dominated one, heightened belongingness in turn leading to a greater intention to pursue the advertised position for all job-seekers. The results of the conditional process analysis indicated, via the computed “index of moderated mediation”, that the indirect effect of condition on intention to pursue via the two proposed mediators, with gender as a moderator between them, was non-significant, b = .0115, 95% CI [-.0246, .0647]. Furthermore, breaking down the indirect
effects on the basis of separate genders, this trend holds: neither being male, b = -.0016, 95% CI [-.0240, .0179], nor being female, b = .0099, 95% CI [-.0201, .0560], significantly moderated the proposed serial mediation. In consequence, hypothesis 4 was not supported.
Figure 2. Model and Hypotheses with PROCESS conditional analysis results.
* p < .05; ** p < .01; *** p < .001
Discussion
The goal of the study was to test whether a greater percentage of either entrepreneurial or feminine wording in job ads would influence perceived gender diversity, which in turn, as a function of gender, would affect anticipated belongingness, which finally would affect intentions to pursue. Thus, a gender-moderated, serially mediated relationship between wording and intentions to pursue would be confirmed (see Figure 2). The results did not support the hypotheses, with the exception of a significant relationship between belongingness and intentionality. At first glance, the outcome of the study seems counter-intuitive. Nonetheless, when taking a closer look at the various analyses, methodology and some lines of argument or findings in the literature, several tentative explanations of the results can be put forward.
Hypothesis 3 was supported, which confirmed the well-evidenced relationship between belongingness and intentionality (e.g. Cheryan et al., 2009; Bian et al., 2018) in the current sample: Validated measures working as expected are interpreted to indicate that the sample was not vastly atypical. Additionally, the randomization checks showed that the people in the two conditions did not differ significantly on several key variables. In order for gender diversity to be perceived as different in the two conditions, the stimulus material should have exerted a powerful gender-related effect (e.g., via stereotype activation; Gaucher et al., 2013; Gupta et al., 2013). As this did not occur, neither could the difference in effect on anticipated belongingness as a function of gender, from which it follows that the moderated serial mediation could not ensue. Thus, understanding the failure of the model likely lies in uncovering why the stimuli did not work as predicted.
The included manipulation checks showed that the entrepreneurial ad was indeed perceived as significantly (p < .001) more “entrepreneurial” (M = 5.70) than the feminine one (M = 5.05). However, in line with hypothesis 1 results, the differences in gender-related perceptions were not significant, the averages obtained on the three items measuring this dimension also being close to 4 (i.e., neutrality). Therefore, whatever participants considered to be “entrepreneurial” did not necessarily double as “masculine”. As there was no solid literature to draw upon when attempting to understand the nature and dimensions of the so-called “entrepreneurial stereotype”, the terms from Short and colleagues’ (2010) study were hand-picked, somewhat arbitrarily, based mainly on their demonstrated relation to the entrepreneurial orientation and their semantic fit within the stimulus material. Although these lists of terms were thoroughly validated to belong to one of five dimensions associated with entrepreneurship within the relevant literature (Short et al., 2010), they do not in any way inform on the degree to which the individual words are associated with masculinity or femininity, or with some hidden tertiary variables which could affect gender perceptions.
Therefore, the lists of masculine and feminine words used by Gaucher et al. (2011) served, during the construction of the ads, as the filters for what constituted a gendered term. These lists are likely not exhaustive, especially due to their small size and fragmented origins (i.e., manually sourced from multiple studies, some done decades apart; Gaucher et al., 2011).
Essentially, there was no way to systematically assess, based on the employed literature (or any published work, to the author’s best knowledge), the degree to which words that are related to entrepreneurship also carry gender associations. Additionally, it cannot at this time be known with certainty whether any wording related to entrepreneurship can evoke images of masculinity (e.g., via stereotype activation; this being a key assumption in the current study), or instead only terms that are simultaneously entrepreneurial and masculine/agentic (as was demonstrated in e.g., Gaucher et al, 2011; Gupta et al., 2008). Furthermore, it was shown that (although not at all times; e.g., Gupta, 2008) entrepreneurship
can successfully be activated as characteristically feminine, masculine, as well as neutral,
depending on contextual information (Gupta et al., 2013). From this, it can be inferred that if the words in the entrepreneurial ad were not “sufficiently” masculine, or if the percentage of those that might be semantically associated with masculinity was not high enough, the effect they produced could have been diluted.
Whatever mental content entrepreneurial wording might evoke must logically depend on the reader and their culture – therefore, on the local stereotype related to entrepreneurship. Indeed, gender-stereotypical content has been shown to not vary much within culturally homogenous cohorts, but instead as a function of cultural differences (Cuddy et al., 2015). Furthermore, the entrepreneur-as-male stereotype is being upheld by depictions and rhetoric in the media (Ahl, 2006; Hamilton, 2013), the content of which will likely vary by country. In fact, scores on entrepreneurial orientation dimensions were found to differ significantly among genders and nations, warranting the careful consideration of the
unique cultural context in which studies related to entrepreneurship are run (Lim & Envick, 2013). As a majority of participants in the current study were Dutch (65%), the rest having already lived in the Netherlands for some indeterminate number of years (minimum 1, given data collection towards the end of the academic year), it is likely that Dutch culture and its conceptions of entrepreneurship would have transpired in the reactions participants had to the stimuli, at least to some degree.
The Netherlands, in line with its progressive image, has been shown to be among top societies in percentage of people being against gender discrimination regarding professional matters (i.e., 83.8% disagreed with unequal treatment in labor) only being surpassed by the Scandinavian countries, in a cohort of 57 countries (p. 346; Tesch-Römer, Motel-Klingebiel, & Tomasik, 2008). Furthermore, out of 22 innovation-driven economies included in the sample, the Netherlands fared best in a comparison composed of women entrepreneur’s representation in the media, high status awarded to successful female entrepreneurs, as well as entrepreneurship being a good career choice for women (in this last category vastly surpassing the rest; p. 17; Kelley, Brush, Greene, & Litovski, 2011). The same 2010 GEM Womens’s report found that fear of failure among females regarding potential entrepreneurial pursuits is second-lowest in the Netherlands, only slightly outperformed by Sweden (p. 15; Kelley et al., 2011). Therefore, albeit tentatively, one could argue that a predominantly Dutch sample, given its cultural idiosyncrasies, would display reduced sensitivity to the entrepreneur-as-male stereotype, therefore decreasing the difference between conditions in terms of perceived gender diversity by partially neutralising the gendered effect of both entrepreneurial and feminine terms.
A final explanation of the model’s failure stems from research concerning recruitment message specificity, or lack thereof, and its effects. It must be noted that both ads were described by participants in the pilot test and the experiment to be vague, or not
precise/specific enough (i.e., participants mentioned information scarcity regarding job, ideal employee or firm characteristics), this being a feature adopted intentionally so that as many people as possible would feel that they could apply, irrespective of their backgrounds. Based on the literature, this likely had a counterintuitive and detrimental effect on the experiment: Feldman, Bearden and Hardesty (2006) note that concreteness (i.e., the level of specificity and detail regarding contexts, actions, or objects) has many positive consequences in job-seekers, such as making messages more attractive and captivating for attention, as well as more memorable and credible (MacKenzie, 1986), and that decreasing it will, among others, reduce claim believability and foster negative attitudes towards both the message and its source (Snyder, 1989). Furthermore, in their own experiment, Feldman and colleagues (2006) found that giving readers more specific information about the job itself (i.e., the least articulated element in the present study’s materials) had substantial effects, increasing perceptions of ad truthfulness and informativeness, enhancing attitudes towards the ad and company, as well as perceived job appropriateness. Specificity levels of information related to the company itself and the work context similarly displayed significant effects, albeit on fewer dimensions (Feldman et al., 2006). In the same vein, it was shown that more specific information in ads led to improved perceptions of organisational attributes and person-organisation fit (a construct conceptually related to belongingness), the latter mediating the relationship between message specificity and intentions to pursue the advertised job (Roberson, Collins, & Oreg, 2005). Thus, it is unquestionable that the lack of ad specificity affected the current experiment, particularly given its use of measures such as intention to pursue and belongingness, which were hinted to be significantly affected by the degree of specificity (e.g., Roberson et al., 2005), and more generally, due to the plethora of documented psychological effects of vagueness (e.g., Feldman et al., 2006), which could exert hidden influences on participants. Nevertheless, it is impossible to accurately quantify
the effects ad specificity had in the current study, as no formal measures of concreteness/vagueness were included, so that reactions to the stimulus materials cannot be reliably compared as a function thereof.
Theoretical Implications
The link between belongingness and intentionality, demonstrated multiple times in similar contexts (e.g. Cheryan et al., 2009; Bian et al., 2018), has been replicated in the current study, which serves to strengthen support for it. However, as neither the model nor other individual hypotheses were confirmed, the main contributions brought to the literature are hypothetical reflections, which add nuance and should be considered when investigating the entrepreneurial stereotype in the future.
Given that avoiding confounds is a key goal in any experiment, it must be strongly recommended that whenever studying effects produced by job ads on job seekers, the level of specificity be formally measured during the pilot study as well as the experiment, so as to be known and kept constant across conditions: Leaving this unchecked, especially when the stimulus material is leaning toward vagueness, can have a multiplicity of profound consequences on the results, such as destabilizing attention towards or even retention of the read material (e.g., Feldman et al., 2006; Roberson et al., 2005).
Again, for the sake of accuracy, researches must elucidate the impact of the degree of semantic proximity of entrepreneurial terms to gender categories: Will a word such as e.g., “genius” (i.e., a term related to brilliance which is likely strongly associated with being male; Bian et al., 2018) have the same effect on readers as will e.g., “restyle” or “dream”? They are clearly semantically distinct from the former, yet all three are part of the “Innovativeness” dimension of entrepreneurial orientation (Short et al., 2010). If there are differences, what are their magnitudes and natures? A step towards solving this could be employing a similar computer-aided methodology to Short and colleagues (2010) to gendered wording in the
English language, to create more comprehensive filters for what constitutes a feminine, masculine, or gender-neutral entrepreneurial word. Additionally, even using the materials presently employed (i.e., Gaucher et al., 2011; Short et al., 2010), studies could explore the comparative effects of simultaneously entrepreneurial and masculine/agentic words and purely entrepreneurial words in job ads, e.g., while keeping one category constant, varying the other one to multiple degrees and measuring effects.
Finally, culture should be considered paramount, as it dictates the mental definitions people have about things such as entrepreneurship and how it relates to gender, which can vary significantly (e.g., Kelley et al., 2011; Lim & Envick, 2013). Therefore, culturally homogenous samples are deemed ideal – not only will this likely reduce noise in the results, but it will also be useful in the long term as it will enable a more solid foundation for theory building and the subsequent emergence of clearer international trends or national differences regarding entrepreneurship. Exact replications of studies from one culture in different ones would likely serve the literature well.
Further Limitations and Suggestions for Future Research
Further limitations need to be considered when interpreting the results of the current study. First, it employed a sample of university students and even some alumni who were chosen because they were part of the age group and due to the ease of recruitment via their university emails. A more homogenous sample discipline-wise, for example, could create a more focused effect, as participants would display less variability in attitudes (in general, and towards entrepreneurship), albeit perhaps decreasing generalizability of findings. Furthermore, it would be easier to elaborate on the reasons behind the results based on participant characteristics.
Second, openness to working in startups, an important characteristic in the current design, as evidenced by the highly significant positive correlations with anticipated
belongingness and intention to pursue, was not used for exclusions, as the sample would have lost a great number of participants. Exactly because of the large proportions, this likely had a major effect on the results. It would be advantageous for future research to only consider those that are actually interested in startups as a viable sample.
Third, not all accepted participants were active job-seekers, some only having sought a job in the last 12 months. This was again a choice taken so the number of participants would not be too greatly reduced, but it stands to reason that a homogenous sample in terms of job-seeker status (i.e., active only) would be desirable to reduce confounds. Furthermore, the experience people have as job-seekers may also be important – given that many in the present sample were first or second-year students, it is clear that conceptions regarding work, as well as job-seeker strategies and goals differed vastly within the sample. Future studies may also account for this.
Fourth, much in line with the aforementioned importance of culture, nationality was found to correlate significantly with perceived gender diversity, anticipated belongingness, intention to pursue, age and English level. As such, foreigners might have consistently scored higher on these five dimensions. Given differences on these critical measures and the relatively hefty 35% of non-Dutch in the overall sample, this has likely further distorted the results, which serves to lend credence to the previous call for mono-cultural research.
It is worth noting that re-running the conditional PROCESS analysis including only those open to startups (N = 160), only those who are current or near-future job-seekers (N = 164), only Dutch participants (N = 124), or all of these combined (N = 83), did not produce significantly different results. Nonetheless, in the latter two cases, the second model with Anticipated Belongingness as outcome variable did approach significance (i.e., Only Dutch: p = .057; All excluded: p = .054), in both cases due to stronger (yet non-significant) effects of gender diversity and gender. These results lend credibility to the suggestions above, yet the
current study has multiple other probable confounds (e.g., nature of wording, vagueness of stimuli), the effects of which cannot be estimated a posteriori, such that no robust conclusions can be reached.
Fifth, no matter how well designed the ad and how similar it is to real ones, the setting of an online experiment is by no means naturalistic: Real job-seekers have a totally different mindset, regardless of who they are, when actually looking for jobs. The procedure and the knowledge that one is participating in an experiment will certainly detract from the “realness” of the experience. To be applicable to real-life contexts, this should be averted. Future experiments could, for example, use postings on real job sites to collect participants, to then ask for consent to partake in an experiment once the reader is interested to apply (potentially, for a small reward).
Sixth, due to time constraints, the pilot test was run with a very small sample as well as being based on opinions rather than hard numbers – regardless of the level of expertise, three people cannot be representative for a population. Confirming this, one of the experts mentioned that the feminine ad was more professional, and on average, they also perceived it as more realistic, which was in stark contrast to the results in the actual study (i.e., the feminine ad was significantly less realistic). Therefore, a pilot study in which a sample of actual participants are gathered and led through the experimental procedure, so that preliminary data can be obtained, would be beneficial for future studies. This would also enable the creation of better stimuli and solving methodological issues in due time.
Finally, regarding the realism of the feminine ad, it might also be that using a very high percentage of feminine/communal wording markedly decreases ad realism, not only because of the likely weaker association between entrepreneurial pursuits and femininity (e.g., Gupta et al., 2008), but also because of the communal nature of the feminine wording, which is by definition not action-oriented, as business invariably must be. It might simply be
that being “loyal”, “kind”, “nurturing”, “empathic”, “compassionate”, or all of them together, is simply difficult to associate with a real business, not out of discrimination, but because of the economically-oriented, innovation-dependent (so unavoidably competitive) nature of business. Therefore, perhaps using feminine terms that are more akin to business life would benefit future research, which is yet another reason to compile more comprehensive gendered wording lists.
Conclusion
The present study failed to find evidence for the conceptual/theoretical model, only confirming the well-studied relationship between belongingness and intentionality. However, being among the first of its kind, it serves to raise much-needed awareness to the many confounds that can seep into such a complex experimental design and to recommend a more systematic and cautious approach to studying the entrepreneurial stereotype and its supposed gendered nature in the context of startup recruitment. Attempting to obtain a significant effect with a direct social implication is tempting, yet the author contends that much additional research is needed to be able to reach the ideal state in which practical implications can with full confidence be derived from this budding literature. Further investigation of the semantic fields of various entrepreneurial words is deemed necessary for both designing better stimulus materials and fully understanding their biasing effects, and for this, a better grasp of gendered wording might also be required. Additionally, a focus on more culturally and attitudinally homogenous groups is recommended, in spite of lower generalizability, not only to reduce external confounds, but also to be able to maintain the advertisements specific enough to prevent distorting participants’ perceptions and reactions. Finally, although they may greatly increase the amount of time and resources required to run such an experiment, extensive pilot studies are recommended to tackle the many possible methodological pitfalls in a timely manner.
