UNIVERSITY OF AMSTERDAM & VRIJE UNIVERSITEIT AMSTERDAM
Master Thesis
ENTREPRENEURIAL LANGUAGE SIGNALS IN JOB ADVERTISEMENTS Development and Validation of a Context-Specific Computer-Aided Text Analysis Tool
Johanna Kain Student nr. UVA: 11374659 Student nr. VU: 2597131 16-08-2017 Msc. Program: Entrepreneurship joint degree Supervisor: Dr. Yuval Engel
This document is written by Johanna Kain who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
1 Table of Contents
Abstract ... 2
1. Introduction ... 3
2. Theoretical Framework ... 5
2.1. The Construct of Entrepreneurial Orientation in the Context of Job Advertisements .... 5
2.2. Signaling Theory and the Job Market ... 7
3. Method ... 9 3.1. Dictionary development ... 10 3.1.1. Deductive dictionary. ... 10 3.1.2. Inductive Dictionary. ... 11 3.2. External Validity ... 15 3.3. Reliability ... 17 3.4. Dimensionality ... 18 3.5. Predictive Validity ... 19 3.5.1. Industry ... 20 3.5.2. Location. ... 20 3.5.3. Education. ... 21 3.5.4. Gender. ... 22 4. Results ... 23
4.1. Independent Sample t tests ... 23
4.1.1. Industry. ... 23
4.1.2. Location. ... 24
4.1.3. Education. ... 24
4.1.4. Gender. ... 24
5. Discussion and Conclusion ... 25
5.1. Implications, Further Research and Limitations ... 28
6. References ... 29
2 Abstract
Job advertisements are often the first point of contact between employers and applicants, thus representing a primary signaling device that sets the stage for the entire recruitment process. Previous studies have shown that the language used in job advertisements shapes job seekers’ perception of employer attributes and ultimately their intention to apply for a job. The aim of this study is to demonstrate the existence and use of entrepreneurial language in job advertisements. We bring together insights from signaling theory and the employer image literature to argue that entrepreneurial job advertisements signal distinct, and often unobserved, qualities with the aim to attract applicants with similar entrepreneurial characteristics. To that end, we adapt and further develop a context-specific entrepreneurial dictionary for computer-assisted text analysis (CATA) using a unique dataset of startup job advertisements in the Netherlands (n=588). We then further validate this tool by drawing on a broader dataset of job advertisement from the UK (n=224,024) and show that, consistent with our central hypothesis, the percentage of entrepreneurial words in job advertisement meaningfully varies between different industries, geographical locations, required applicant education level, as well as male and female dominated occupations. Beyond the utility of our dictionary for researchers interested in job advertisement CATA, our discussion provides direction for future research about signaling theory and the use of entrepreneurial language to construct a unique employer image.
Keywords: signaling theory, entrepreneurial orientation, computer-assisted text analysis,
recruitment
3 Entrepreneurial Language Signals in Job Advertisements
Development and Validation of a Context-Specific Computer-Aided Text Analysis Tool 1. Introduction
Job seekers’ decisions are shaped by information in job advertisements, which often serve as the first line of communication between employers and potential employees (Phillips et al., 2014; Lievens & Slaughter, 2016). Because internal company operations such as company culture, working environment or innovativeness are neither easily observable from the outside nor directly spelled out in a brief job advertisement, employers who try to improve the quality of their applicant pool, will often signal these characteristics indirectly (Lievens & Slaughter, 2016). Signaling theory (Spence, 1973), which is fundamentally concerned with reducing information asymmetry between two parties (Connelly et al., 2011), predicts that in order to enhance the effectiveness of their recruitment signals and attract fitting candidates, employers will increase “signal fit” – “the extent to which the signal is correlated with unobservable quality” (Connelly et al., 2011, p. 53). For instance, Moser et al., (2017) argue that to attract new talent, startups will often highlight their distinctive attributes, such as flexible work practices, a focus on personal development, and an informal working environment. In addition, prior research has shown that even small variations in the language used in job advertisements can act as powerful signals shaping job seekers’ perception, intention, and behavior (Gaucher et al., 2011; Phillips et al., 2014; Walker & Hinojosa, 2013; Moser et al., 2015). For instance, Gaucher et al. (2011) find that job advertisements for male dominated occupations contained more masculine words and were less attractive for female applicants. Taken together, this means that because they signal employer attributes, subtle wording differences in job advertisements may attract some applicants and discourage others. This positions job advertisement language as an important phenomenon with far reaching implications for research on employee recruitment and job market discrimination (Phillips et al., 2014; Moser et al., 2015).
While there are numerous company characteristics that employers may try to signal (e.g. Ehrhart and Ziegert, 2003; Highhouse et al., 2003; Lievens & Slaughter, 2016), and such signals may be communicated using several linguistic devices (Gaucher et al., 2011; Phillips et al., 2014; Moser et al., 2017), herein we are particularly interested in examining the role of job advertisement language in signaling entrepreneurial orientation. In line with Lumpkin & Dess, (1996), Miller (1983), Covin and Slevin (1998) and Covin and Slevin (1991) we define entrepreneurial companies as those that exhibit innovativeness, proactiveness, risk taking,
4 autonomy and competitive aggressiveness. Entrepreneurial orientation is important because it has been shown to be associated with firm performance (Lumkin and Dess, 1996; Rauch et al., 2009; Short et al., 2010). Therefore, an entrepreneurial firm character accounts for one of the most important factors for economic growth and corporate success (Kraus, 2011). Even more, entrepreneurial orientation is particularly crucial in the context of employee recruitment as it shapes person-organization fit and may therefore determine employee commitment, turnover intention, and effort on the job (Kristof-Brown et al., 2005; De Clercq & Rius, 2007). Therefore, we ask:
How do employers signal their entrepreneurial orientation to job seekers? And how can we, as researchers, be more effective in detecting and studying these important signals?
To answer these questions, we build on recent advancement in Computer-Aided Text Analysis (CATA) and the development of specific entrepreneurial orientation CATA measures (Short et al., 2010; McKenny et al., 2016). As a form of content analysis, CATA enables the measurement of theoretical constructs by processing text into quantitative data based on the word frequency as defined by a “dictionary” - a list of words that stand to represent the construct under investigation (Short et al., 2010). Thus, for example, studying organizational texts such as CEO shareholder letters, annual reports, and mission statements, research did not only establish that entrepreneurial language exists in this kind of documents, but that it can influence firm level outcomes (Short et al., 2010; Moss et al., 2015; McKenny et al., 2016). However, the currently available entrepreneurial CATA dictionary was unfortunately not originally designed to examine job advertisements and may therefore introduce a potential source of error if blindly applied to this context (McKenny et al., 2016). Hence, the primary goal of this thesis is the development of a revised CATA dictionary to capture entrepreneurial language in job advertisements.
To this end, we follow procedures outlined by Short et al. (2010) and McKenny et al. (2016) and go through multiple deductive and inductive steps to generate, adapt, and validate a reliable context-specific CATA dictionary. Throughout this process, we make use of two unique datasets: (1) a learning sample consisting of 588 startup-specific job advertisements in The Netherlands; and (2) a much larger (n=224,024) and more general sample of job advertisements from the UK. Our results show that entrepreneurial language does exist in job advertisements, and expose meaningful patterns in how it varies between job ads in different industries,
5 locations, occupations, and required education level of applicants. We therefore provide several important contributions to the literature and open exciting new avenues for future research.
The rest of the paper is structured as follows: First, the topic is embedded in a theoretical framework about entrepreneurial orientation in the context of job advertisements, employee attraction and signaling theory. Then, the method to develop the dictionary is explained, as well as the establishment of validity and reliability. Further, the predictive validity for this method of capturing entrepreneurial language is examined by testing hypotheses about variation in the relative proportion of entrepreneurial words in job ads across different industry and occupation characteristics. Finally, findings are discussed and possibilities for further research are outlined.
2. Theoretical Framework
2.1. The Construct of Entrepreneurial Orientation in the Context of Job Advertisements The starting point of solid construct measurement is “a sound theoretical definition of the concept of interest” (Short et al., 2010, p. 326). As this thesis aims to capture entrepreneurial language in job ads correctly and holistically, the construct of entrepreneurial orientation (EO) is used to understand the nature of entrepreneurial language in the context of job ads. As job ads usually consist of a description of firm characteristics, but also include a part describing the characteristics of the ideal candidate, EO is firstly described on the firm-level, and secondly on the individual level. Additionally, in a short third section, the distinct characteristics of new ventures are described, as new entry is seen as the central idea behind the term entrepreneurship (Lumkin and Dess, 1996), and entrepreneurial individuals were found to be especially attracted to these kind of organizations (Williamson et al., 2002; Tumasjan et al., 2011).
Firstly, the construct of EO has emerged from the strategic management literature (Lumkin & Dess, 1996) and has been further developed into a firm- and organization-level construct (Covin & Slevin, 1991; Lumkin & Dess, 1996). Lumkin & Dess (1996) stress that “EO refers to the processes, practices and decision-making activities that lead to new entry” (p. 136-137). Thereby, new entry is not solely limited to the act of launching a new venture in the form of a new startup, but also includes internal corporate venturing (Burgelman, 1983; Covin & Slevin, 1991). In this firm-level context, EO has been used to distinguish entrepreneurial firms from conservative ones. According to Covin & Slevin (1998), entrepreneurial firms undertake strategic decisions that reflect the entrepreneurial management styles of the top management, while the top management of conservative firms acts in a more risk-averse fashion, less innovative, and passive or reactive. Within this firm-level definition of EO, there are different perceptions about the dimensionality of the EO construct, ranging from EO as a
6 three-dimensional construct (Miller, 1983; Anderson et al., 2015) and EO as a five-dimensional construct (Lumkin & Dess, 1996). The five-dimensional framework has been widely used and cited in entrepreneurship literature over various contexts such as firm performance (Lumkin & Dess, 1996), funding-success in microlending (Moss et al., 2015) and to show the benefits of computer-aided text analysis (Short et al., 2010). Furthermore, the content adequacy of that five-dimensional scale has been confirmed in a two-stage study (Saha et al., 2017). Therefore, in this thesis, the following five dimensions are used to capture EO: autonomy, competitive aggressiveness, innovativeness, proactiveness and risk-taking (Lumpkin & Dess, 1996). These are defined in the following: autonomy: “the ability and will to be self-directed in the pursuit of opportunities” as well as “the independent action of an individual or a team in bringing forth an idea or a vision and carrying it through to completion” (p. 140); competitive aggressiveness: “a firm’s propensity to challenge its competitors directly and intensely to achieve entry or improve position, that is, to outperform industry rivals in the marketplace” (p. 148);
innovativeness: “a firm’s tendency to engage in and support new ideas, novelty,
experimentation, and creative processes that may result in new products, services, or technological processes” (p. 143); proactiveness: “acting in anticipation of future problems, needs or changes” (p. 146) and risk-taking: “the firm’s proclivity to engage in risky projects and manager’s preferences for bold versus cautious acts to achieve firm objectives” (p. 146). However, while acknowledging the multi-dimensionality of EO, EO has also been used as an overall construct in studies which have the purpose to holistically capture the construct rather than explore the differences between the distinct dimensions. One example for that are Shepherd & Wolfe (2008), who analyze the overall EO content of sport narratives, as well as Kraus (2011) in his study on EO-performance relationship in service firms.
Secondly, even though EO has been developed as a firm construct, the context of this thesis makes it also necessary to look at what it means to show entrepreneurial orientation as an individual. By stating that the success of internal corporate venturing is dependent on the entrepreneurial activity of the management and operational-level participants, Burgelman (1983) acknowledges the role of individuals as drivers of entrepreneurial activity within a firm. But also in the context of recruiting it is important to look at EO on the individual level, as research showed that the personality and individual traits of an applicant have a large influence on how a company is perceived as a potential employer (e.g. Krauss et al., 2005; Frese & Gielnick, 2014). Krauss et al. (2005) developed a model of individual entrepreneurial orientation based on Lumkin & Dess’ (1996) five-dimensional EO construct providing the following definitions: autonomy: “desire to express one’s individuality in the workplace”,
7 “wanting to make own decisions”, “highly motivated to realize own ideas and visions” (p. 320);
competitive aggressiveness: “assert themselves, enjoy competition, and strive for victory” (p.
320); innovativeness: “a positive mind-set towards new ideas with regard to products, services, administration, or technological processes” (p. 320); proactiveness: “A proactive, self-starting, and persistent orientation that attempts to shape environmental conditions” (Frese et al. 1996, as cited in Krauss et al., 2005, p. 321); and taking: “a positive orientation towards risk-taking” (p. 321). In addition to these five original EO dimensions, several other factors that indicate entrepreneurial orientation in individuals have also been mentioned in the literature, such as motivational/affective factors, especially growth goals and visions, entrepreneurial passion and positive and negative affect (Frese & Gielnick, 2014). Krauss et al. (2005) also introduced two additional EO dimensions namely learning orientation (a person’s ability and willingness to learn from positive and negative experiences) and achievement orientation (looking for feedback, comparison with others, striving to continuously improve the own performance).
Lastly, in line with theory on person-organization fit, it has been found that applicants who show entrepreneurial orientation are more attracted to the distinct features of new ventures, being flat hierarchies (Tumasjan et al., 2011), more responsibilities and independence, intellectual challenge (Sauermann, 2017), a caring company image, and a strong–future-oriented–company vision (Moser et al., 2015). Person-organization fit is the compatibility between people and organizations (Kristof-Brown et al., 2005), suggesting that “individuals are most successful in organizations that share their personalities” (Tom, 1971). To capture the characteristics of new ventures in the context of job advertisements, the study of Moser et al. (2015)1 offers a useful collection of employer attributes based on existing literature (Lievens, 2007; Lievens & Highhouse, 2003) and extended by new categories especially suited to new ventures.
2.2. Signaling Theory and the Job Market
When decisions have to be made, Stieglitz (2002) distinguishes between two types of information that are available to the decision-maker: Firstly, freely available public information, and secondly, private information, which is only available to a smaller part of the public and consists of information concerning quality and information concerning the behaviors and intents of the other party. While it has been stated before that job seekers who show entrepreneurial orientation are attracted to organizations that show same characteristics, these
1 For a comprehensive list of employer image attributes found in a new venture context, see table 4 for instrumental employer attributes and table 6 for symbolic employer attributes in Moser et al. (2015, p. 30 and 32 respectively).
8 characteristics are often part of the second group of private information. For job seekers, the main categories of missing private information concern internal company characteristics such as organizational culture and the general character of “how would it be to work there”. In line with theory on person-organization fit, the perceptions of the employer image are a core element of a job applicants’ attraction to a company (Slaughter et al., 2004). Employer image is a term to describe the mental representations and connotations a potential applicant holds about a company (Highhouse et al., 2009; Lievens and Slaughter, 2016). “Employer image helps applicants to distinguish among employers, results in applicant pools that are larger and of higher quality, leads to quicker decision making and a stronger emotional bond, and is associated with higher organizational financial performance” (Lievens and Slaughter, 2016, p. 28). For innovative, entrepreneurial companies, it is therefore crucial to create a company image that attracts applicants that fit into their culture (Moser et al., 2017). For these reasons, most scholars advise new ventures to signal their distinctiveness vis-à-vis larger, more established organizations (e.g. Williamson et al., 2002; Phillips et al., 2014; Moser et al., 2017; Sauermann, 2017).
Therefore, in line with signaling theory (Spence, 1973), companies who wish to attract a more relevant applicant pool, should bridge this information gap and signal their unobservable qualities to job seekers (Connelly et al., 2011). The effectiveness of the signal is moderated by the “signal fit”, which is “the extent to which the signal is correlated with unobservable quality” (Connelly et al., 2011, p. 53). Spence (1973) was one of the first to connect signaling theory to the job market. He calls the hiring process an investment under uncertainty, as the job seekers true qualities and productive capabilities are not visible right away. This element of information asymmetry is true both for employers as well as for applicants (Ehrhart et al., 2015; Lievens & Slaughter, 2016). Generally, different types of signals can be used by companies to signal unobservable qualities, depending on the stakeholders they want to address and on what an organization wants to communicate about themselves (Connelly et al., 2011). This can reach from having a prestigious board of directors (Connelly et al., 2011) to industry certifications (Moss et al., 2015) and educational degrees by job applicants (Spence, 1973).
In this thesis, the focus is on job advertisements as signals to bridge information gaps (Phillips et al., 2014; Gaucher et al., 2011; Moser et al., 2017). The study of Phillips et al. (2014) shows the large impact of job advertisements on job seekers in their job pursue. Job advertisements often shape job seekers organizational identity perceptions and facilitate the matching process between job seekers and organizations (Walker & Hinojosa, 2013). What is more, only slight changes in wording have a signaling function to the potential applicants about
9 the organization’s character (Phillips et al., 2014). Therefore, job advertisements serve as an important signal a company (signaler) can send out to prospective applicants (signal receivers) in order to deliberately close the gap of information asymmetry in a way that suits the organization (signaler) best. Firms should benefit from that approach, as recruitment accounts for enormous costs, that only better applicant fit can diminish. It has been found that organizational identity building und signals that stress a “standing out” strategy can be helpful for new ventures in their quest to attract qualified employees (Moser et al., 2017). The awareness of that fact can lead to a better applicant fit due to targeted recruiting messages (Kristof-Brown et al., 2005). As a central theoretical proposition, we therefore argue that entrepreneurial and innovative companies use more entrepreneurial language in their job advertisements in order to signal their distinct firm attributes to the outside world and enhance their signal-fit by attracting the right applicants with a better person-organization fit.
However, so far, no tool has been developed to capture such entrepreneurial language in job advertisements. We therefore turn to develop such a tool in the next section. The aim of this paper is to show the existence of entrepreneurial language in job ads and test the utility of measuring such language as a unique form of employer signal. Our central prediction is that the more entrepreneurial the organization, the more it would use entrepreneurial language in its job ads. The reason for that is rooted in the assumption that, generally speaking, organizations wish to create congruence between unobservable qualities (e.g., being innovative and prompting risk taking) and their employer image in the eyes of potential candidates, thereby attracting a more relevant applicant pool (e.g., attracting innovative and risk-taking individuals).
3. Method
The goal of this thesis was to build a comprehensive dictionary to analyze entrepreneurial language and signaling in job advertisements. Dictionaries capturing the construct of interest are needed in computer assisted text analysis (CATA), as the algorithm of CATA tools analyzes texts based on predefined word lists (McKenny et al., 2016). This has been done by first looking at existing dictionaries and then adapting them to the specific study context (Short et al., 2010; McKenny et al., 2016). This deductive list was based on theories about relevant constructs such as entrepreneurial orientation, which are provided in the theoretical section of this thesis. Next to this deductive approach, we examined a learning-sample of 588 job ads from a dataset provided by a Dutch recruiting platform specialized in start-up jobs. This more context-specific data was utilized to learn inductively how the language of entrepreneurial job ads looks like and thereupon build a dictionary capturing entrepreneurial language in job ads more generally.
10 The dictionary created in that way was then applied on a large sample of 224,024 job ads from a more general British job platform. Amongst a number of available CATA-tools, LIWC2015 was chosen for the analysis, as it has been widely used in content analysis studies (e.g. Moss et al., 2015; Wolfe & Shepherd, 2008) and offers the possibility to create custom dictionaries, which was important for this study as the initial EO dictionary had to be adapted to be suitable for job ads. Additionally, in the LIWC-output, word counts are given as a percentage-score of the overall ad word count, which has the advantage of standardizing the detected words of the construct measured and therefore making texts of different length comparable (Pennebaker et al., 2001).
3.1. Dictionary development
3.1.1. Deductive dictionary. To build the dictionary to measure entrepreneurial language in job ads, a dictionary initially derived by Short et al. (2010) and adapted by McKenny et al. (2016) was used as a baseline. This initial dictionary had been developed to capture EO in shareholder letters and consists of five word lists capturing the five EO dimensions autonomy, competitive aggressiveness, risk-taking, innovativeness and proactiveness. In addition to these five lists which count for 358 words in total, an additional inductively derived word list was created by both of the authors, counting 41 words (Short et al., 2010) and four words (McKenny et al., 2016). 2 Using an approach and a dictionary that has been used and validated by other researchers before as a foundation enhances content validity (Short et al., 2010). However, both Short et al. (2010) and McKenny et al. (2016) developed and refined the initial word lists for content analysis in shareholder letters, whose function, target group and language are very different from those of job ads. Therefore, to avoid specific factor error, this baseline-dictionary was adapted in the following process.
The initial EO dictionary by McKenny et al. (2016) has been adapted in terms of word forms. As the text of job advertisements is mostly either talking in the third person singular about “the perfect candidate” or addressing the reader directly, the form of 43 words and terms has been changed and/or added accordingly3. Examples for that would be: “on their own” in the autonomy-category to “on your own”, and “emancipated” added to “emancipation” in the autonomy-category. Also, some words from the original word lists were deleted, as they did not fit into the context of EO in job ads (e.g. “accountability”, “accountable”).4 This refinement
2 For a detailed overview of the original dictionary developed by Short et al. (2010) and McKenny et al. (2016), see table A5 in the appendix, column 1
3 For a detailed overview of these adapted words, see the yellow-colored words in table A5 in the appendix 4 For a detailed overview of the words taken from the initial word lists, see the green-colored words in table A5 of the appendix
11 of the word list was important, as the percent of variance due to specific factor error was high (57%) for entrepreneurial orientation without such a refinement in McKenny et al.’s test (2016).
3.1.2. Inductive Dictionary. An additional list of words has been derived from a “learning dataset” through a multi-phase inductive process. This approach is supported by McKenny et al. (2016) who propose to manually code at least 10% of the whole body of documents being analyzed. The sampling frame was chosen from 588 unique job advertisements (10,755 before duplicate removal and 808 before removing all non-English language ads) with a length from 39 to 412 words, posted between the 5th of January 2015 and the 11th of November 2016 on a Dutch job-board app focused specifically on start-ups. In order to provide the raters with a general understanding of the EO construct in job ads and align the ratings with the construct measured, several random samples have been drawn which were manually coded according to the following questions by both the author and an expert rater: (a) The job advertisement and the job/company it represented are very entrepreneurial; and (b) Employees at the company advertising the job are expected to be entrepreneurial. The rating was done on a one to five Likert-scale (1 = Strongly disagree, 2 = Somewhat disagree, 3 = Neither agree nor disagree, 4 = Somewhat agree 5 = Strongly agree). The final sample for manual coding consisted of 173 job ads chosen from the sampling frame after applying several filters to get a sample with the most entrepreneurial ads (location=Amsterdam only, company size < 200 employees, employee-age < than 35 years). This accounts for 29.42% of all ads from the sampling frame. This additional inductive procedure was also important to capture all the entrepreneurial signals of job advertisements correctly. These 173 ads were manually rated by 2 coders in a fully-crossed rating procedure (Hallgreen, 2012).
To assess interrater reliability, a calculation of the weighted Cohen’s kappa was executed, which is “commonly used for categorical data with an ordinal structure such as in a rating system that categorizes high, medium or low presence of a particular attribute” (Hallgreen, 2012, p. 27), such as prevalent in a Likert-scale. The advantage of Cohen’s weighted kappa over the ordinary Cohen’s kappa is that it takes the degree of disagreement into account by associating each category with a weight. Thereby, Cohen’s weighted kappa weights large differences in ratings (such as 1 by rater A and 5 by rater 2) heavier than only small discrepancies (e.g. 3 by rater A and 4 by rater B) (Artstein & Poesio, 2008). The received linear weighted kappa-value for the 173 manually coded ads was 0.36, which would be interpreted as a fair to moderate agreement by Landis & Koch (1977), whose interpretation of weighted kappa-values has been widely acknowledged in the computational linguistics literature (Artstein & Poesio, 2008). Also, a correlation of the mean manual ratings of all raters with the
12 CATA-results for the overall entrepreneurial language dictionary showed a moderate agreement (rs=.41, p<.00, n=173). These might not seem to be especially impressive agreement
scores; however, these ratings were only used to sensitize the raters to the construct, in order to be able to rate the real subjects of the study, which is the final dictionary (Hallgreen, 2012).
In the next step, those words or phrases were added to the dictionary that both raters thought to be most responsible for a high entrepreneurial rating of an ad. Therefore, the definitions of section 2.1. of the EO construct on the firm- and individual level were used as a guideline. In the following, the two raters independently went over the words to assess their overall fit with the construct of job ad entrepreneurial language. Subsequent evaluations were discussed until both raters agreed (Short & Palmer, 2008). As an example, the term “winner mentality” that was found in ads multiple times was believed by both raters to fit within the definitions of EO in the category of competitive aggressiveness. Further refinements were made in multiple rounds: Shortening of terms that were too long or too specific and therefore would probably only appear in the very ad they were taken from (e.g., “awesome Friday drinks” shortened to “awesome”); adding more alternative forms of existing words and looking at the frequencies of words in the texts. An example for that, multiple forms of the term “winner mentality” were added, such as “winner-mentality” and “winner’s-mentality”, as these forms of spelling were frequently found in other ads. Further on, through the LIWC color-coding function, more words such as “extraordinary” were added and those words that were frequently used out of context were removed (McKenny et al., 2016). All in all, 311 new words were added inductively, and 202 of them could be fit into the word lists from Short et al. (2010) and McKenny et al. (2016), as they were found to fit within the definitions of the five EO categories outlined in section 2.1.5 This step was necessary to make the word lists for each dimension more suited to the content and language of job advertisements.
Finally, drawing on the fact that entrepreneurial individuals were found to be attracted to the distinct features of new ventures that are outlined in section 2.1., a separate inductively created word list named “Additional inductively derived words” was created to capture the distinct characteristics of new ventures. Next to that, the additional factors influencing the individual EO next to the original five EO dimensions from section 2.1. were examined for the creation of this inductive list, such as motivational and affective factors, especially growth goals and visions, entrepreneurial passion, and positive and negative affect. Table 1 shows the final dictionary that was created.
13 Table 1.
Complete dictionary for entrepreneurial language in job advertisements
EO Dimension Content Analysis Words
Autonomy at liberty, autonomic, autonomous, autonomy, autonomously, control your own destiny, emancipated, emancipation, empower, empowered, empowering, empowers, free-thinking, liberty, on your own, prerogative, self-directed, self-directing, self-direction, self-rule, self-ruled, self-ruling, take control, takes control, unaffiliated, unattached, unconfined, unfettered, ungoverned, unregulated, confident, confidence, decisive, flawless execution, freedom, make a difference, makes a difference, manage your own time, personal responsibility, powerful, run the company, running the company, runs the company, self-manage, self-organized, self-starter, self-starting, self-sufficient, sense of ownership, strive, strives, striving, strong opinions, take a leading role, take decisions, takes a leading role, takes decisions, think up, thinks up
Competitive Aggressiveness
aggressive, aggressiveness, aggressively, antagonist, antagonistic, aspirant, aspire, aspires, best-in-class, combat, combative, combats, compete, competes, competer, competing, competition, competitive, competitiveness, competitor, competitiors, competitory, conflicting, contend, contends, contender, contentious, contest, contestant, defends, dog-eat-dog, enemies, enemy, fierce, fight, fighter, fighting, fights, jockey for position, jockeys for position, defend, joust, jouster, jousts, lock horns, opponent, oppose, opposes, opposing, opposition, outgrow, outgrows, outgrowing, pacesetter, plays against, play against, preeminence, pre-eminence, preeminent, pre-eminent, ready to fight, rival, rivalry, struggle, struggles, tussle, tussles, unequaled, unmatched, unparalleled, unrivalled, vying, wrestle, wrestles, champion, conquer, big wins, dead, deal closer, disruption, disruptive, disrupts, disrupt, everything is possible, expand, expands, fast changing, changing, fast paced, fastest growing, paced, fastest-growing, growing, fast-paced, feud, game-changing, get to the point, gets to the point, goal commitment, goal committed, goal oriented, goal-oriented, goal orientation, goal-orientation, grow, growing, grown, grows, growth, growth hack, growth hacker, growth hacking, high growth, high-growth, high impact, high-impact, high-performance, hunt, hunter, hunting, hunts, kick-ass, killing, ninja, push, pushes, pushing, quickly grown, rapidly changing, rapidly expanding, rapidly-changing, razor sharp, scale up, target-driven, top notch, top-notch, underdog, unlimited, winner mentality, winner-mentality, winners mentality, razor sharp, scale up, target-driven, top notch, top-notch, underdog, unlimited, winner mentality, winner-mentality, winners mentality
Risk-Taking adventure, adventuresome, adventurous, ambitious, ambitiousness, audacious, bold, boldness, bold-spirited, brash, brave, challenge yourself, challenged yourself, challenging yourself, challenging, chancy, courageous, courage, danger, dangerous, dare, daredevil, dares, dauntless, dicey, fearless, fearlessly, gutsy, hardship, hardships, headlong, incautious, intrepid, jeopardy, mega-ambitious, no place to hide, no safe path, not for the faint-hearted, plunge, plunges, precarious, precariously, rash, risk, reckless, risk taking, riskiest, risk-taking, risky, temerity, venturesome, wager, doesn't give up, don't give up, not afraid, not hesitate, don't hesitate, isn't scared, not scared, on the fly, on-the-fly, uncertainty, unusual, unusually, ups and downs
Innovativeness ad lib, adroit, adroitness, bright idea, bright ideas, clever, cleverness, conceive, conceives, concocts, concoct, concoction, concoctive, conjure up, conjures up, creative, create, creativity, creation, dreams, dream, dream up, dreams up, expert, freethinker, genius, gifted, hit upon, imagination, imaginative, improvise, improvises ingenious, ingenuity, innovate, innovated, innovates, innovating, innovation, innovations, innovative innovativeness, introduced, introducing, introduction, introductions, invent, invented, invention, inventive, inventiveness, inventor, invents, launch, launched, launching, master stroke, mastermind, metamorphose, metamorphosis, new capabilities, new capability, new compounds, new content, new contents, new core areas, new course, new directions, new family, new features, new generation, new generations, new idea, new ideas, new line of business, new medicines, new molecular entities, new pharmaceuticals, new platform, new process, new processes, new product, new products, new solutions, new systems, new technique, new techniques, new thinking, new tools, new ways, new wrinkle,
new-14
generation, new-product, next generation, next-generation, novation, novel, novelty, patent, patented, patents, process development, product development, product launch, product launches, proprietary, prototype, prototyping, pushes the envelope, push the envelope, R&D, radical, radically, re-engineering, re-engineer, re-engineers, reformulate, refresh, reinvent, re-invent, reinvented, reinventing, reinvention, reinvents, released, renewal, renewing, renew, renews, reshape, reshaped, reshapes, reshaping, resourceful, resourcefulness, revolution, revolutionary, revolutionize, revolutionizes, revolutionizing, roll out, rolls out, rolling out, see things, sees things, transform, transformation, transforming, transforms, adapt, adapts, affinity with technology, ambition to develop, breakthrough, business-savvy, change, continuously evolving, curiosity, curious, digital growth, digital savvy, digitally savvy, evangelizing, evangelize, evangelizes, flexibility, flexible, genius, genially, get things done, gets things done, highly adaptable, identifying new, identify new, identifies new, improvising, instinct, instincts, intensive, intense, introduce own ideas, introduce your own ideas, intuition, intuitive, intuitively, keep developing yourself, latest technology, latest technologies, new opportunity, new opportunities, new business opportunities, new goals, newest technologies, newest technology, next level, nextgen, once in a lifetime, once-in-a-lifetime, out of the box, out-of-the-box, own ideas, questioning, questions, question, state-of-the-art, tech savvy, tech-savvy, web-tech-savvy, you know the drill, you know your stuff, your ideas
Proactiveness anticipated, anticipation, anticipate, anticipates, exploratory, explore, explores, exploration, foreglimpse, foreknow, foresee, foretell, formulate, formulates, formulating, impatient, impatience, industry's first, initiate, initiates, initiating, initiated, initiative, initiatives, inquire, inquiry, investigate, investigates, investigation, lead, leads, look ahead, looks ahead, move ahead, moves ahead, opportunistic, opportunities, opportunity, pave the way, paves the way, pioneer, pioneered, pioneering, pioneers, preparations, prepare, preparing, prepares, proactive, proactively, prospects, prospect, roadmap, scrutinization, scrutinize, scrutinizes, take advantage, takes advantage, well-poised, well-positioned, active, actively, agile, bias for action, can do, can-do, close a deal, closer, contagious, contagiously, drive, driven, drives, driving, dynamic, eager, eagerness, energetic, energetically, energized, energy, energizing, engaging, execute, executes, forward, forwards, go getter, go the extra mile, goes the extra mile, go-getter, hands on, hands-on, hand-on, hand on, hustle, hustler, hustles, hustling, implement your ideas, implements their ideas, keen, keep on, not shy, opening, outgoing, out-going, perseverance, perseverant, persistence, persistent, persistency, pro-active, race contentious, ready to explore, self propelling, self-motivated, self-motivation, self-propelling, take initiative, takes initiative, thrive, thrives, thriving
Additional inductively derived words Short/McKenny
emerging, emerge, emerges, entrepreneur, entrepreneurial, exposure, exposures, founding, high-value, out-doing, outthinking, pursuing, pursue, pursues, unique, uniquely, uniqueness, enterprising
Additional inductively derived words
A game, laugh, accelerator, accelerators, A-game, co-founder, cool office, crowdsourcing, crowd-sourcing, crowd-funding, crowdfunding, crowd funding, daily lunch at the office, digital native, diverse tasks, each day flies by, early stage, early-stage, extensive, extensively, flat hierarchical, flat organizational structure, flat structure, flat-hierarchical, great place to work, great teammate, huge, incubator, incubators, informal, informally, lean, low hierarchical, low hierarchy, low-hierarchical, mad, multi-skilled, no formal hierarchy, off-hours, open, pitching, pitch, pitches, pre-accelerators, pre-accelerator, rock-star, small team, start up, startup, start-up, startups, start-ups, stock options, transparent, transparency, young, awesome, desire to, desires to, don't worry, doesn't worry, dream big, dreams big, enthusiasm, enthusiastic, enthusiastically, enthusiast, excitement, exciting, excited, extraordinary, extremely, extreme, fantastic, funny, fun, incredible, incredibly, insane, insanely, inspirational, inspiring, learn what it takes, love, loves, motivated, motivating, motivation, optimism, optimistic, own vision, own visions, passion, passionate, positive, positively, real impact, sense of humor, sincerely passionate, strong belief, strong beliefs, strong opinion, thrill, thrilled, thrilling, upbeat, vision, visions, visionary
15 3.2. External Validity
In this study, job advertisements were used as the documents of interest, in comparison to shareholder letters, that have been used in previous studies using CATA, such as Short et al. (2010) and McKenny et al. (2016). Job ads are an interesting level of analysis, as they are created by companies as a signal to attract those candidates they believe will fit best to their organization and will bring most value (Phillips et al., 2015; Walker & Hinojosa, 2013). To test the dictionary and assess external validity, a second sample of job ads was drawn from a more general British job platform. A large pool of 2,120,770 ads from 2013 to 2016 was analyzed in LIWC using the dictionary developed above. The only adaptation to the dictionary was a change of the spelling from American English to British English. The final sample of 222,024 job ads was chosen by removing all the ads where information on one of the variables of interest was missing. Also, all ads with less than 50 and more than 1,000 words were filtered out, because the explore-function in SPSS identified all the ads with more than 1,000 words as outliers, and 50 was the mean word count minus 1 SD. To ensure the validity of such a limitation in word count, all the tests were also run on samples both without any limitation on word count and on a more strictly reduced sample by only choosing word counts between 50 and 500 (mean WC+-SD), both not showing significantly different results. Finally, job ads that could bias the results as they contain one dictionary word multiple times, such as “risk” in ads for a business analysis positions containing “risk management” were filtered out.
To examine if entrepreneurial language exists in these job ads, a one sample t test compared to a test statistic of zero has been conducted on the Dutch learning sample and the final sample (see also Short et al., 2010). A one-sample t test is an appropriate test for this measurement after checking the pre-assumptions for its usage. The variables were not perfectly normal, but left-skewed due to a 5% zero-values in the data, however, because of the central limit theorem, one sample t tests are relatively robust to non-normality when the sample size is big enough (Field, 2009). A test result significantly different from zero thereby indicates that entrepreneurial language exists in the sample. Table 3 demonstrates that the t test results are significant, showing that entrepreneurial language measured by the variable “Entrepreneurial language” exists in the job ads. This variable contains the overall word score from the different dictionary dimensions. Table 3 includes a visual demonstration of how this final dictionary was used in LIWC to analyze the job ads. The words marked in red are those words from the dictionary that were searched for by the computer program.
16 Table 2.
Evidence of entrepreneurial language both overall and in the distinct dimensions in job advertisements of firms in Dutch learning sample and British sample
*p < .05
**p < .01
Table 3.
Job ads with no entrepreneurial language content vs. job ads with high entrepreneurial language content; highlighted in the LIWC color-code function
Dutch sample British final sample
N Mean SD Min Max t test N Mean SD Min Max t test Entrepreneurial language 588 3.72 2.41 .00 13.73 37.36** 222,024 2.16 1.41 .00 16.92 725.59** Autonomy 588 0.10 0.28 .00 1.61 8.76** 222,024 0.07 0.18 .00 3.23 188.44** Competitive Aggressiveness 588 0.47 0.74 .00 5.63 15.56** 222,024 0.27 0.39 .00 6.41 323.62** Risk Taking 588 0.11 0.30 .00 2.08 9.09** 222,024 0.11 0.35 .00 9.65 154.75** Innovativeness 588 0.81 0.93 .00 5.04 21.14** 222,024 0.33 0.50 .00 8.63 311.77** Proactiveness 588 1.18 1.22 .00 6.78 23.54** 222,024 0.90 0.74 .00 10.62 573.50** Additional inductively derived words Short/McKenny 588 0.10 0.30 .00 3.23 8.37** 222,024 0.05 0.16 .00 4.48 148.50** Additional inductively derived words 588 0.94 0.99 .00 7.79 23.09** 222,024 0.43 0.52 .00 7.41 391.06**
Entrepreneurial language content= zero Entrepreneurial language content= high
Start-up dataset
Company Description: Online treatment platform for mental healthcare
Job Title: QA Software Tester
Required Skills: HTML & CSS; Web Analytics; Javascript; PHP
Job Tasks: Write Tests; Analytics; Documenting Job Requirements: We need a detail oriented tester to help break our product time and time again. A solid understanding of how browser-specific internet traffic works in modern web applications, including HTTP, SSL, and JSON. Experience with JavaScript, CSS and/or PHP is a pro Experience in functional testing of web applications. Experience creating automated test frameworks for web applications or other software. Experience with continuous integration environments Experience writing regression tests in Behat a plus. Experience with issue management and customer ticketing systems such as Assembla, Freshdesk Development and/or testing experience on iOS would be good, but not necessary.
Entrepreneurial language content= 13.73% Company Description: Discover and share new music!
Job Title: UX Designer
Required Skills: UI / UX Design; Photoshop; Sketching; Sketch; Illustrator; Prototyping; iOS Development; Android Development
Job Tasks: UX Design; Visual Design;
Prototyping
Job Requirements: A vision on how you think we should serve music simple. Distinctive creativity
as the core element for designing a refreshingly
intuitive, yet exciting interface. Responsibility for your contribution to our service. Good communicative skills. Positive energy and an
unlimited drive (also after 5pm). Flexibility to deal with start-up life, its ups and downs and
uncertainty. Experience with designing for iOS, Android and / or Web interfaces
British dataset
Purpose of the job: The Treasury Accounting Assistant is responsible for accounting treasury operations and for managing payment runs in the banking system.
Candidate requirements: GSE grade C or above in Maths and English (or equivalent). A good communicator who is a proven hard-worker and is a team player with an eye for detail. Min. ½ years experience working within an accounts position.
Entrepreneurial language content= 13.29% We are currently looking for an enthusiastic and
creative Marketing Executive to join one of the most highly innovative and respected consumer organization based in the sports sector. Due to recent expansion, we are looking to take on a
forward thinking individual who can continue to
drive growth through a variety of creative
17 3.3. Reliability
Generally, the use of computer assisted text analysis (CATA) is a big enhancement of reliability in context analysis, as CATA-tools have been proved to be advantageous over human coders due to “the ability to process large samples with high speeds and reliabilities” (Short et al., 2010, p. 320). CATA facilitates the measurement of constructs as it translates text content into quantitative data, based on the word frequency (McKenny et al., 2016). Along the same lines, the processing of larger data from a bigger variety of contexts is made possible, which enhances both external validity and statistical power to test hypotheses (McKenny et al., 2016). Because of these benefits, CATA has been used repeatedly over the past years to measure various constructs, including organizational psychological capital (McKenny et al, 2013), firmness of resolve (Brett et al., 2007) and entrepreneurial orientation (Short et al., 2010; McKenny et al., 2016).
Consequently, the most common sources of measurement error that are known from traditional survey research are the following: random response error, transient error and specific factor error (McKenny et al., 2016). To reduce the probability of these measurement errors in content analysis, the measures proposed in McKenny et al. (2016) have been undertaken.
Firstly, random response error is not too much of an issue in content analysis using CATA, as usually, texts are analyzed without the authors knowing about it, so the texts should not be influenced by the authors cognitions. Secondly, transient error is described as possible measurement error arising from differences in the language used in texts produced at different points in time (McKenny et at., 2016). While shareholder letters usually talk about past, present and future events in one document, job ads mostly talk about the company as it is at the time being. For that reason, in contradiction to previous studies using shareholder letters as their texts of analysis in CATA, the analysis of job ads is less subject to this kind of error. Therefore, job ads are documents that reflect the entrepreneurial orientation of a company at the time being very well.
Main duties and responsibilities will include: Verifies ESA file information uploaded into FIRST system. Books and reports foreign exchange, loans, deposits and cash pooling entries in the EPR. Administers the treasury involvement in the payment process. Books bonds and guarantees fees. Performs bank reconciliations and cashbook postings. Check that internal systems are up to date for the reporting units, with regard to list of signatories, limits and reconciliation details. Provide cover and support for other members of the team.
Key Responsibilities: Generating new ideas and
leads via the latest creative marketing techniques.
Driving sales with existing customers through both designing and implementing marketing campaigns. Ability to identify new sales
opportunities through marketing research. The successful candidate will have a passion for marketing with a natural creative flair, and a ‘can do’ attitude. This opportunity offers candidates a fantastic opportunity to take the next step in their career, with an attractive package of bonuses the post holder can look to stamp their personal mark within this open cultured organization.
18 Therefore, the most likely source of error that could occur is specific factor error. Specific factor error is prevalent if the context in which a CATA measure was developed for has an influence on the measure itself, e.g. that the language used to express the same thing varies in different contexts. This is a common source of error in content analysis and leads to a threat of validity and reliability (McKenny et al., 2016). While the texts analyzed in content analysis are not produced by the researchers themselves, errors due to idiosyncrasies in the formulation of survey questions can be excluded. However, in the development of CATA measures – such as new dictionaries – new words measuring the underlying phenomenon are identified by the judgement of the ratings of both researchers and experts, which presents further ground for specific factor error in the form of coder bias. In this study, the possibility of specific factor error has been addressed by the dictionary development steps described in section 3.1. By following other scholars example, the initial EO dictionary has been refined and adapted to the specific context of job ads including the development of an additional inductive word list through a multi-phase inductive process and the assessment of interrater reliability (e.g. Short et al., 2010; McKenny et al., 2016).
3.4. Dimensionality
The purpose of this thesis is to capture entrepreneurial language as an overall construct in job ads, and therefore, the overall word score of the created dictionary has to be used for further analysis (Wolfe & Shepherd, 2008; Kraus, 2011). Still, the multi-dimensionality of the EO construct is acknowledged, and the dictionary has been built based on the five-dimensional EO construct (Lumkin & Dess, 1996). This has the advantage of enhancing the content validity of the dictionary, and lays at the same time the basis for future research on the distinct dimensions of the construct in content analysis on job ads.
For constructs to be multidimensional, each dimension should be distinct from but related to the others, meaning that significant but not perfect relationships (correlation of less than 0.5) between the dimension should occur. When constructs of interest are conceptualized as multidimensional, it is recommended to create multiple word lists to capture the whole construct (Short et al., 2010). Before running the correlation, preliminary analyses were executed to check for assumptions of normality, linearity, homoscedasticity and independence. For that, skewness and kurtosis of the data were looked at, and scatterplots for linearity checked, both with help of the explore function in SPSS. These tests showed that most of the data was not perfectly normally distributed, but left-skewed due to a larger amount of zero-values. Also, the assumption of linearity was violated. Therefore, a Spearman’scorrelation was executed, as this non-parametric test shows reliable results also for data that is not perfectly normal (Field,
19 2009). Table 4 shows the Spearman’s correlation coefficients between the overall dictionary of entrepreneurial language and the distinct EO dimensions. The results indicate that the assumption of EO as a multidimensional construct in the context of job ads is supported over both samples. Each dimension is significantly correlated (at the 0.01 level) to at least one of the other dimensions (other than the overall ‘entrepreneurial language’) in both datasets. These results show that the usage of the five-dimensional EO construct (Lumpkin & Dess, 1996) is legitimate for this analysis. Also, the use of the inductive word list developed by Short and McKenny (2010; 2016) as well as the inductive list created in this thesis (section 3.1.1.) was shown to be legitimate.
Table 4.
Intercorrelations of EO Dimensions to Assess Dimensionality
Dutch learning sample Mean SD 1 2 3 4 5 6 7 8
1. Entrepreneurial Language 3.72 2.41 1.00 2. Autonomy 0.10 0.28 .26** 1.00 3. Competitive Aggressiveness 0.47 0.74 .37** .10* 1.00 4. Risk Taking 0.11 0.30 .29** .10* .09* 1.00 5. Innovativeness 0.81 0.93 .50** .02 -.03 .15** 1.00 6. Proactiveness 1.18 1.22 .65** .15** .07 .04 .12** 1.00 7.Additional inductively
derived words Short/McKenny 0.10 0.30 .26
** .08* .06 .15** .01 .12** 1.00
8. Additional inductively
derived words 0.94 0.99 .59
** .07 .14** .14** .12** .18** .13** 1.00
British final sample Mean SD 1 2 3 4 5 6 7 8
1.Entrepreneurial Language 2.16 1.41 1.00 2. Autonomy 0.07 0.18 .23** 1.00 3. Competitive Aggressiveness 0.27 0.39 .45** .08** 1.00 4. Risk Taking 0.11 0.35 .26** .04** .08** 1.00 5. Innovativeness 0.33 0.50 .42** .06** .08** .05** 1.00 6. Proactiveness 0.90 0.74 .71** .09** .19** .09** .09** 1.00 7. Additional inductively
derived words Short/McKenny 0.05 0.16 .22
** .04** .10** .07** .07** .08** 1.00
8. Additional inductively
derived words 0.43 0.52 .58
** .12** .18** .06** .12** .24** .07** 1.00
**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).
3.5. Predictive Validity
“Predictive validity is demonstrated when constructs of interest are linked with others that are theoretically related” (Short et al., 2010). The basis of this section is the central prediction that entrepreneurial firms use more entrepreneurial language in their job ads in order to signal their distinct entrepreneurial nature to potential job seekers. Therefore, based on this prediction, it
20 was examined whether there are differences in the usage of entrepreneurial language within the groups of variables outlined in this section.
3.5.1. Industry. While the connection of EO and firm performance is widely acknowledged in the literature (Rauch et al., 2009; Lumkin & Dess, 1996; Short et al., 2010), recent literature (Kraus, 2011; Rigtering et al., 2014; Choi & Williams, 2016) has started to explore the role of different industries when it comes to EO-performance relations. The most prevalent industry distinction in the literature for that context are service vs. manufacturing industries. Service firms differ from manufacturing firms in the sense that services are intangible, heterogeneous and perishable, and production and consumption often cannot be separated, which also makes their innovation processes different and less structured (Choi & Williams, 2016; Rigtering et al., 2014). It is especially interesting to look at service industries when it comes to entrepreneurial language in job ads, as the service sector is less capital- and more labor-intensive than the manufacturing sector (Kraus, 2011) and therefore, it is expected that attracting entrepreneurial employees is even more crucial for service- than for manufacturing firms. Rigtering et al. (2014) found that technology action has a higher impact on the relationship of EO and firm performance in manufacturing firms, while marketing action has a higher influence on the same relationship in service industries. These findings again support the prediction that service firms should use more entrepreneurial language in their job ads, as it has been stated in section 2.2. that firms use job ads to build and market a unique employer image to attract the best-fitting candidates. Lastly, when using EO as an overall construct, Rigtering et al. (2014) find a statistically significant higher EO in service- than in manufacturing firms. Therefore:
H1. Job advertisements for firms in service industries contain more entrepreneurial language than job advertisements for firms in manufacturing industries.
3.5.2. Location. Interestingly, it has been shown that despite the rise of modern telecommunication technology, growing internationalization and more flexibility in working relationships such as remote work, both people and the economy itself stay concentrated in certain locations. Especially high knowledge-based and creative industries are concentrating in some specific places such as Silicon Valley, Austin and New York (Florida, 2003). According to Glaeser (2000), cities represent centers of idea creation and transmission. Based on these observations, creativity is seen as a key driver of entrepreneurship (Frese & Gielnik, 2014). In that regard, Florida (2003) built his theory of creative capital, seeing creative people as a main driver of regional development. These people form the so-called “creative class”, which is
21 defined as innovative people who engage in creative problem-solving. Thereby, they are not bound to certain industries, but can be found in every industry. They are self-sufficient, combine existing approaches in new ways and their work is creating meaningful new forms, which equals one of the many definitions of entrepreneurship. This group includes in the core “scientists and engineers, university professors, poets and novelists, artists, entertainers, actors, designers, and architects, as well as the ‘thought leadership’ of modern society: nonfiction writers, editors, cultural figures, think-tank researchers, analysts, and other opinion-makers”, as well as “creative professionals” who work in a wide range of knowledge-based occupations in “high-tech sectors, financial services, the legal and health-care professions, and business management” (Florida, 2003, p. 8). Studies on 500 regions in seven European countries show a positive relationship between creative class occupation, the growth of employment, and entrepreneurship in certain regions (Boschma & Fritsch, 2009). Furthermore, locations with more highly talented people grew faster and attracted even more talent over time (Florida, 2003). As a result of these factors, cities themselves develop an entrepreneurial image (Steyaert & Katz, 2003), a phenomenon that reinforces the clustering of knowledgeable, creative and entrepreneurial people in a region. Hence:
H2. Job advertisements for firms located in cities with a creative image contain more entrepreneurial language than job advertisements for firms located in other cities.
3.5.3. Education. Knowledge spillover theory offers a comprehensive explanation for aforementioned phenomena. Knowledge spillover is developing into a key driver of the commercialization of new knowledge and therefore innovation and growth in both industries and regions. Such spillovers can work in many directions, including spillovers benefits from co-location of firms operating in similar industries, as well as spillovers from universities to the economy (Autretsch & Belitski, 2013). Knowledge spillover theory is especially interesting for the emergence of entrepreneurship in our knowledge-driven economy today, as entrepreneurship includes the action upon new knowledge and ideas (Autretsch & Belitski, 2013). According to creative class theory, especially highly educated people prefer to cluster in innovative, diverse, tolerant and open places (Florida, 2013). In that sense, it is not only the clustering of firms and industries that drive the transformation of a region into a high-growth location, but also the role of people as the driving force behind the growth and global competitiveness of a region (Autretsch & Belitski, 2013). According to human capital theory, highly-educated and productive people are the main driver of growth in a region (Jacobs, 1984, as cited in Florida, 2003). Furthermore, Ilhan & Gurel (2011) found that individuals with higher
22 university education are 1,5 times more likely to become entrepreneurs than people with a lower education. For that reason, next to the firm industry and location, the education level that is required for an occupation is believed to impact the amount of entrepreneurial language in job ads. Especially for an analysis of entrepreneurial language in job ads, looking at the requirements of the exact occupation rather than the industry is important, as the occupation describes what people actually do (Markusen et al., 2006). What is more, job ads mostly talk more about the position itself than about the firm in general. Thus:
H3. Job advertisements for professions requiring higher formal education contain more entrepreneurial language than job advertisements for professions that do not require higher formal education.
3.5.4. Gender. Lastly, also the gender distributions for certain professions and their influence on entrepreneurial language in job ads should be considered. Many scholars underline the gender gap in entrepreneurship (e.g. Thébaub, 2016; Gupta et al., 2008). In the UK, women only account for under one third of self-employment, and only for 17% of business owners,
managers and employers (“UK Female Entrepreneurship,” 2017). Stereotypes about typically
male and female characteristics and abilities are an important explanation for that fact. Entrepreneurship, such as business in general and managerial positions in particular, are associated with masculine characteristics (Gupta et al., 2008). Thereby, language plays an important role, as it is an important medium to transport ideas and beliefs, and language shapes how a phenomenon is perceived in society (Gupta et al., 2013). In line with the before-mentioned theory about person-organization fit, potential applicants feel attracted by job advertisements that give them a feeling of belonging (Gaucher et al., 2011). Previous research on language in job ad shows that in job ads for occupations that are male-dominated, more words with typically male stereotypes were used (Gaucher et al., 2011). As entrepreneurship is mainly associated with masculinity, it can be also expected that language reflecting entrepreneurial orientation reflects characteristics associated with male stereotypes. Thus:
H4. Job ads for occupations that are male-dominated contain a higher percentage of entrepreneurial language than other occupations.
23 4. Results
4.1. Independent Sample t tests
To measure whether there is a significant difference between different groups when it comes to entrepreneurial language in job ads and to test the hypotheses above, an independent sample t test has been executed for each group of interest. This approach has been chosen as all the variables of interest were dummy variables (coded as 0 or 1) and t tests are used to analyze the differences in means of only two groups. To avoid unreliable test results due to pre-assumptions for the test not being met, we first checked for normality. To determine whether the data was normal, we looked at skewness, kurtosis and a Kolmogorov-Smirnov as well as a Shapiro-Wilk test. For most of the variables, the significance of a Kolmogorov-Smirnov test showed that the data is not perfectly normally distributed. Nevertheless, as the sample size is large (n= 224,024), these test results are not too meaningful as they tend to show false significance in large samples (Field, 2009). However, the skewness in all samples was fairly close to zero, even more when the square root transformed dependent variable of the entrepreneurial language dictionary was used. The square root transformation improved normality, as the dependent variable was positively skewed due to five percent zero values. This measure led to a reduction of skewness from .97 to -.45. These are values that can be worked with, especially as t tests are relatively robust to moderate violations of normality, and even more when the sample size is quite large, due to the central limit theorem (Field, 2009). Next to that, the Levene statistic was checked in all cases to make sure that the assumption of homogeneity of variance has not been violated. In case of significance of this test, the respective t test result line was reported.
4.1.1. Industry. To test hypothesis 1 about the difference within industries on the percentage of EO language in job ads, the industries were classified in service and manufacturing according to the classification scheme of Rigtering et al. (2014), who defined the following as service firms: “1) the finance and insurance market” 2) the hospitality market (e.g. restaurants and hotels) 3) firms that fall into the category of science or technical services 4) information and communication firms 5) commercial traders or business firms 6) day care centers and firms active in the education market. We also looked at the classification of Choi & Williams (2016) who defined “computer/IT/telecommunications, wholesale and retail trade, transport and financial services (bankings/insurance) as service firms, while “machine/machine parts, automobile/car parts, electricity/electronics and textile/leather are seen as manufacturing firms. The results show that a significant difference t(222,022)= -54.25, p < .00 could be found between entrepreneurial language in service industries (M=1.40, SD=.53, n=194,571) vs.
