A meta-analysis of measuring employability through its effects on life and career satisfaction

(1)

1

A meta-analysis of measuring Employability through its effects on

Life and Career Satisfaction

Msc Thesis

Author - Gian J.V.F. Lie Student no. - 0165387

Faculty - Amsterdam Business School (UvA) Programme - Leadership & Management (Msc) Date - 06-24-2016

Supervisor, 1st_{- Stefan Mol}

(2)

2

Abstract

A comparative meta-analysis for scales of measuring employability is presented. From an exploration of employability literature, it appears that quantitative employability studies are numerous and divergent. To examine likely candidates for optimally measuring employability, two meta-analyses are performed and compared. This involves two employability scales, (self-)perceived employability and the Career Adapt-ability Scale and two outcome effect sizes, life and career satisfaction. Although mean effect sizes and I2

-heterogeneity levels are different in the same way as predicted, the underlying Q-statistics are not, therefore no further conclusions could be based on this meta-analysis. This is assumed primarily due to a lack of studies included in the analysis that causes statistical noise to distort outcomes. Adding more studies would be the first route suggested, but empirical research specific to this analysis is not yet sufficient in number, perhaps as a result of academic novelty. Further, manifold methodological

differences exist in the current meta-analysis and sampling error can be expected to increase when adding more of the same type of studies.

Statement of originality

This document is written by Student Gian Lie, 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.

(3)

3

1. Introduction.

Today’s rapid shifting of the global economy impacts the stability of what employers need from their employees. These needs keep changing and they require from individuals that seek employment a career orientation that goes beyond the current or any single employer. Research that studies what constitutes this career orientation, its antecedents and its outcomes can be summarized under the concept of employability. Employability can be defined as 'the capability to move self-sufficiently within the labor market to realize potential through sustainable employment' (Hillage & Pollard, 1998, p.2). A more recent review by Fugate, Kinicki and Ashforth, defines employability as ‘a multi-dimensional aggregate of career identity, personal adaptability, and social and human capital’ (p.32, 2004). In an attempt to explore the current academic state of employability, the first goal of this study was to systematically collect all empirical studies regarding employability in its various forms in order to create an overview. The

overview confirmed that there is no single quantitative measure of employability. This leads to the risk of drawing inconsistent conclusions across studies and undermining knowledge creation regarding

employability. This is especially true for the different scales that use the word employability in their title, but consist of idiosyncratic questionnaire formulations (De Witte, 1992; Heijde & Van Der Heijden, 2006; Rothwell & Arnold, 2007; De Cuyper, Van der Heijden, & De Witte, 2011). In addition, the overview revealed a substantial diffusion between measures that could be classified as measuring (aspects) of employability. This could be caused by the variety of definitions employability has

accumulated on itself over the years underlying the different empirical scales (Forrier & Sels, 2003). A different explanation for the wide variation in measures is that a certain amount of studies that were included are only interested in (partial) perspectives of employability to serve unique research purposes regarding career adjustment characteristics (Savickas & Porfeli, 2012; Waters, Briscoe, Hall, & Wang, 2014), human and social capital (Forrier, Verbruggen, & De Cuyper, 2015), employability orientation (Van Dam, 2004), career optimism (Gunkel, Schlägel, Langella, Peluchette, & Reshetnyak, 2013), or career decision making self-efficacy (Guan, Zhou, Ye, Jiang, & Zhou, 2015), among others. Without explicating every one of these example constructs, it becomes apparent that the way these constructs are

(5)

5 titled in some way relates to aspects relevant to employability. To approach the issue of empirical

diffusion, the overview enables an uprooting of the common elements and dissimilarities in the quantitative measurement of employability. The nature of the issue is that employability is a simple construct to grasp, yet it encompasses many psychological, social and contextual constructs at the same time. As a result, empirically measuring employability has become incongruent in literature. This study aims to suggest a new direction for future empirical research to become more consistent in its choice of measure, and ultimately, more consistent in its results.

Currently, there may be some level of agreement on what employability is supposed to represent, and that it is a multidimensional construct involving reciprocally related constructs (Fugate et al., 2003). However, there is no current consensus as far as how to actually measure it. Thijssen et al. add to this point, by describing employability as ‘an attractive but confusing buzzword’ on which ‘each definition emphasizes a diversity of characteristics of (potential) employees, for instance, physical and cognitive suitability, learning, flexibility, adaptation, and mobility, to mention but a few, yet all referring to employment as an outcome.’ The term employability sees use in a variety of disciplines such as

economics, vocational guidance for the disabled, military veterans or rehabilitating prisoners, or guiding public policy and education (Hillage & Pollard, 1998; Stevens, 2002; Kronberg, 2013; Schmutte, 2014; Smaliukienė, 2014; Cerda, Stenstrom, & Curtis, 2015). The urgency of the underlying societal need to deal with unemployment may be an explanation of repeated academic attempts to measure some form of employability.

An illustrative example of the current lack of empirical consensus would be the study by Van der Heijde and Van der Heijden (2006). Here, the authors note that empirical research on employability has thus far not been sufficiently integrating the different organizational levels and academic perspectives that it has taken to conceptualize it. In response, the authors propose a multi-dimensional approach that involves five scales to each denote constituent dimensions of employability: occupational expertise,

anticipation and optimization, personal flexibility, corporate sense and balance. The authors base their

(6)

6 work through the optimal use of competences’ (p. 463). Although the findings are found to significantly relate to different outcomes of employability such as career success and job satisfaction, it remains up for debate whether such a scale would be the definitive way to measure employability. Furthermore, this scale includes responses of both employees and employers. Unlike any other employability scale, this provides control for rater bias. All other scales that were retrieved for this study make use of self-reporting questionnaires, and as such are unilaterally subjected to self-rater bias. Despite the

sophistication in its design, this scale has seen little academic use, arguably because of the recency of its conception. Nevertheless, to its credit, it offers a comprehensive scale for measuring employability, presented with support for its validity and reliability.

A less complicated way to operationalize employability would be of the sort provided by the scale of self-perceived employability by Rothwell et al. (2008). This scale is based on the definition of employability as a ‘self-perceived ability to attain sustainable employment appropriate to one’s

qualification level’. Questions included in this scale are straightforward; f.e. ‘I regard my academic work

as top priority’ or ‘my university has an outstanding reputation in my field(s) of study’. This incorporates

internal and external factors at play, referring to either acquired career-related human capital or perceptions about the availability of labor market opportunities respectively. However, this scale is designed to measure employability for university graduates in particular, which poses a limitation for it to be compared to other scales of employability.

A different scale that bears practically the same name is perceived employability devised by De Witte (1992; De Witte et al., 2011). The questions under this scale are phrased as ‘I am optimistic that I

would find another job, if I looked for one’ or ‘I am confident that I could quickly get a similar job’.

These questions are termed in a more general way than those in the scale of Rothwell et al. which are aimed at university graduates (2008). Moreover, the content of questions is similarly about optimistic perspectives towards finding a suitable job. It raises the question of where the distinctions actually lie, because clearly, both scales measure relevant, albeit different, aspects of employability.

(7)

7 Yet another example that measures aspects of employability is the Career Adapt-ability Scale (CAAS) by Savickas and Porfeli (2012). Career adaptability is defined as ‘a psychosocial construct that denotes an individual’s resources for coping with current and anticipated tasks, transitions, traumas in their occupational roles that, to some large or small degree, alter their social integration’ (p.662). CAAS is divided into four scales of abilities; concern, control, curiosity and confidence. These

adapt-abilities are theorized not as part of the core of the individual personality, but as behaviors existing within the conscious control of the individual that he or she assumes to adapt to the work environment. Even though employability is not a term that Savickas and Porfeli use in their article, it is worth of note that the behaviors measured by the CAAS scale partially reflect both definitions of employability that were provided earlier. Furthermore, from what the preliminary collection of empirical research has shown, CAAS is a prominently used empirical scale. Its widespread use is also shown to be consistent across countries (Savickas & Porfeli, 2012). For now, it suffices to mention the academic prevalence of CAAS and its conceptual similarity to other measures of employability. If anything, CAAS shows restraint in its definition of employability by focusing on characteristics of the individual, and by the logic of Occam’s razor, it would seem preferable over other measures because there are less assumptions underlying resulting in less statistical noise (Myung & Pitt, 1997).

One example that could compete over CAAS could be the earlier mentioned perceived

employability by Van der Heijde and Van der Heijden (2006). Despite that it uses a more inclusive

definition of employability than CAAS, as a scale it offers comparable sophistication by segmenting employability in subscales. The subscales can be expected to provide more validity and reliability than measures of employability that utilize questions superficially directed at an abstract and complex

phenomenon. In contrast, subscales are limited and specific and thus likely to facilitate more accuracy in responses, and stronger reflection of the examined constructs, because responses are elicited by more discrete and understandable questions.

The central aim of this study is to uncover whether the type of empirical scale for employability moderates the strength of employability’s relationship to common outcomes and comment on what this

(8)

8 implies for empirical research of employability going forward. Because of the variation in how

employability is measured, in this study, only two of the scales that measure aspects of employability as proxies will be used in a comparative meta-analysis; (self-)perceived employability (SPE) and CAAS. SPE was chosen based on its conceptual proximity to both general employability definitions of Hillage and Pollard, and Fugate et al. (1998; 2003). CAAS was chosen based on its prominence in empirical research of career-oriented behaviors. The comparison is accomplished by examining through meta-analysis how SPE and CAAS are empirically associated to outcomes of well-being, represented by life (LS) and career satisfaction (CS). Performing the meta-analysis twice, once for CS and once for LS will allow for a richer perspective on the heterogeneity levels associated with the mean outcomes. This can be explained in the following way: the expectation is that the divergence in the setup of the SPE scales will lead to greater variance in its correlations to CS and LS, across different studies. This variance is expected to be greater than that of CAAS, a scale that is argued to measure the same underlying concept. The average of the pooled correlations is referred to as the mean effect size, and the variance of mean effect size is referred to as its heterogeneity. In favor of a meta-analysis is that empirical research on employability has not yet been reviewed in a similar quantitative fashion. A meta-analysis will provide an evidence-based approach to examining the optimal way of measuring employability. Accordingly, this meta-analysis is expected to contribute to the academic perspective on measuring employability.

Unemployment is an issue of global society that is deserving of the public attention being paid to it. Belle and Bullock summarize that unemployment is linked to negative consequences at the both the societal and individual level (2009). At the community level, the impact of high unemployment and related poverty can be exacerbated by low neighborhood quality. For individuals, the effects are manifold; unemployment is associated to higher rates of mental and physical health problems, increased mortality rates and decreased well-being of children and spouses. Employability research is one of the avenues through which government can fight the negative effects of unemployment, by learning how

employability can possibly be nurtured in its working citizens. The desired outcomes would then be increased economic strength and well-being for the individual. For this reason, the employability

(9)

9 outcomes LS and CS will be examined simultaneously, from a range of studies that have examined the associations. An added benefit of using LS and CS is that, compared to other associated outcomes, they provide a solid anchor of comparison by their common occurrence. Holding these as constant outcomes across SPE and CAAS studies, then, will expose of differences between the scales through meta-analytic heterogeneity. The conceptual relationships between employability, SPE, CAAS and well-being, LS and CS are further detailed in the theoretical framework in the following paragraph, which also includes the arguments for predicted effects. The theoretical section is followed by a conceptual model that presents arguments based on the theoretical framework and operationalizes the predictions in formal hypotheses. After this a quantitative meta-analysis is presented. Next, the results are interpreted in a discussion section and elaborated on for their theoretical implications and what they suggest for future research. Lastly, a formal conclusion reviews the entire study with summarizing remarks.

2. Theoretical Framework.

2.1 Employability.

Employability as a concept has seen change over the years since its inception. Two of the most recent studies that attempted to define the concept are reviewed here. These studies both base their

conceptualizations on prior research. Thijssen et al. (2008) provide a historical overview of the concept and present a convenient classification for its different conceptualizations. This framework allows for distinguishing between the different concepts of employability that are included in this meta-analysis. Fugate et al. (2003) summarize what is common to all employability definitions and provide their own thorough definition, which is taken as the leading conceptualization of employability in this study.

The theoretical review by Thijssen et al. (2008) provides a historical overview of how employability research has developed since its inception around 1955, and a classification for

employability research. Not until the start of the 1990s did empirical research on employability develop in earnest, which in the preceding decade, has focused its attention on assisting social policy makers in

(10)

10 Europe and the UK (Hillage & Pollard, 1998; McQuaid & Lindsay, 2005). This is distinguished from academic efforts in the US, where the focus has lied on aspects of individual skills and adaptability (Fugate & Ashforth, 2003; Savickas & Porfeli, 2012). Thijssen et al. assert that lifetime employability is taking over lifetime employment as the current norm in society (2008). The change in focus towards employability is caused by organizational needs for flexibility in a changing labor market. This is indicated by increased job mobility rates, or the rate of changing employers and/or occupation, and the growing availability of consultants or other outside workers. Thijssen et al. pose that this may lead to where working on a steady payroll will become an exception. Although maintaining employability is an ongoing challenge for individuals seeking employment, competence and expertise in a specific field will remain its key aspect. Organizations depend on employees' capability to manage their individual

qualifications which are fundamentally required for the organizations' success. At the same time, job requirements are constantly shifting and the acquired competences and skills are more quickly becoming obsolete. Because the changing labor market is hard to predict, flexibility is the goal for employees that wish to compete therein.

Thijssen et al. detail how the content and perspective of employability has changed over the past decades. Up until the 1980s, employability had signified the role of society and the government in getting people jobs. During the 1980s the attention was directed towards the employers, especially large firms, as responsible for the provision of lifetime employment. The guarantee of lifetime employment became a policy principle that rewarded employee loyalty. From the 1990s onward, the attention changed direction towards the employees themselves to take initiative and manage their own careers proactively. Overall, careers have become boundaryless and/or protean in their nature as a result of increased organizational change, job requirements changing and society transitioning into a culture of lifetime employment, where acquiring a job no longer guarantees the same lifetime employer. Protean careers or boundaryless

careers are concepts that metaphorically describe a self-transformative or adaptive career orientation and

literally careers that progress by increasingly crossing organizational boundaries. The increasing non-permanence of employment simultaneously increases the need for more employability research that looks

(11)

11 past the tradition of simply getting the first job and expecting to keep it, towards what drives and enables the individual to find suitable employment independent of a single employer.

To classify these changing perspectives, Thijssen et al. pose three categories of employability that differ in their scope. The first definition is employability at the ground level and the most commonly used, employability radius: the whole of the employees' personal ability in performing various jobs properly. The second, broader definition is posed as the whole of characteristics influencing future labor market success that includes career and learning competencies or employability competencies/skills. Employability skills describe any employees' readiness and ability to secure employment outside of the employment radius. In a third definition the scope is further widened to include contextual conditions. These are factors that operate at the level of society and the organization as employer. Thijssen et al. provide examples of employability research to show that the contextual conditions that this third

definition includes, see little use in academic attempts to define employability. The authors conclude that employability is a concept that is diffuse and has multiple layers. Although the various ways of defining employability lead to different and useful insights, pressure from society and industry will dictate where priorities lie for future research. Given the environmental turbulence in terms of society and industry, future research would likely to explore how employability functions in the external labor market, or the finding of employment alternative to jobs offered by the current employer or to the state of

unemployment (Thijssen et al., 2008).

In an earlier conceptual review of employability, Fugate et al. (2003) emphasize the underlying person-centered variables as key determinants of the construct because individuals are assumed to have little to no control of what employers set as job criteria externally. Even though what employers need is integral to gaining employment, Fugate et al. focus employability on what is internal to and under control of the individual. Employability is then framed around three key components; career identity (self-defining cognitions relating to career, that actively guide career-adaptive behaviors), personal

adaptability (proactive career orientation independent of any employer), and social and human capital

(12)

12 employability can be classified under the second definition, employability competencies/skills, as per Thijssen et al. (2008). Moreover, Fugate et al. (2003) pose that employability is an aggregate

multidimensional construct, which means that employability is caused by its key determinants, instead of the key determinants being reflective of employability. Employability only has meaning when its

determinants are observed collectively. Furthermore, employability is multidimensional by the way its components are reciprocally related. The authors argue for reciprocal determinism of dimensions with theoretical examples. Career identity steers the question of ‘who am I’, f.e. being a high achiever that proactively deals with the challenges presented by a job which reinforces the level of personal

adaptability. Success through heightened personal adaptability in turn reinforces career identity. Likewise, human and social capital are central to defining ‘who am I’, the career identity, which in return influences the acquisition of more human and social capital. Lastly, social and human capital also critically

determine personal adaptability by opening up job opportunities, either through networking for vacancies or through meeting employers’ criteria. The information from the social network also reinforces personal adaptability by supporting continuous learning. Higher personal adaptability is in turn associated with a better ability to cultivate relationships and the pursuit of training and learning opportunities, leading to higher social and human capital respectively. In this way, Fugate et al.’s conceptualization of

employability can serve as a reference for empirical scales that measure (aspects relevant to) employability.

The following two paragraphs will provide justifications for the inclusion of the scales to

represent employability measures. The two groups of scales are (Self-)perceived Employability (SPE) and The Career Adapt-ability Scale (CAAS).

2.2 (Self-)perceived Employability (SPE).

In this study, SPE is a group of questionnaires that measure self-reported employability by assessing the participants own perspective of his or her ability to procure new employment. Out of eight studies in this meta-analysis, three make use of the ‘self-perceived employability’ scale by Rothwell and Arnold (2007;

(13)

13 Dacre Pool & Qualter, 2013; Praskova, Creed & Hood, 2015). Rothwell and Arnold (2007) explain employability as having two components. Internal employability is built on a number of individual characteristics, such as possessing knowledge and skills, capacity for learning, career management and resilience. In addition, external employability reflects factors external to the individual. This includes the expected value of one’s ability to fulfill alternative job requirements in the current and predicted state of the labor market both inside and outside of the current organization. The sum of these factors pertain to the authors’ own definition of employability as ‘the ability to keep the job one has or get the job one desires’ (p. 25). The scale consists of eleven self-reporting items indicating both external and internal dimensions of employability. From this study, which is included in the meta-analysis, the reported overall effect size was chosen over the two sub-dimensional effect sizes. Dacre Pool and Qualter (2013) use this scale without adaptations, yet they only report sub-dimensional effect sizes. For optimal consistency with the other studies in the meta-analysis, the average of the two effect sizes was taken, weighted by the relative number of questionnaire items they reflect. The third included study that uses this scale is one by Praskova et al. (2015), who perform modest adjustments to suit a respondent sample of prospective graduates. An example modification is ‘people who do the same job as me are valued highly’ into ‘people with the same career choice as me are highly valued’. These adjustments are not expected to impact the meta-analysis through their slight inconsistencies.

Two studies that are categorized under the SPE group make use of the ‘perceived employability’ scale developed by De Witte (1992; De Cuyper et al., 2011; Silla, De Cuyper, Gracia, Peiró, & De Witte, 2009). Despite that the scale title lacks the ‘self’-prefix, the question items are also self-reporting. Assuming that this quality is reflected by this prefix, the scale titles are essentially identical. Silla et al. (2009) adopt five items from the original scale, whereas De Cuyper et al. (2011) make use of only four items. Example questions are ‘I could easily switch to another employer, if I wanted to’ and ‘I am confident that I could quickly get a similar job’. Silla et al. define employability as ‘the available alternatives in the labor market’ (2009, p. 2) and De Cuyper et al. pose the definition as ‘the worker’s perception of his or her possibilities to achieve a new and equal job’ (2011, p. 1487). Both articles do not

(14)

14 use the definition of the original study by De Witte (1992). In both articles there are, however, exhaustive acknowledgements of the developments in employability literature since the scale’s inception in 1992. Moreover, in both articles the original developer is a credited author and assumed to have had input in the adoption of the scale. Although the scale itself has not seen much change, and its underlying precise definition is not consistent over time, this is not expected to impact the results of meta-analysis. What can impact the meta-analysis is that De Witte’s scale questions are smaller in number and are worded

differently than items in the other scales in the SPE group. The questions are prone to elicit different response values, because they partially reflect different aspects of the underlying concept employability. When meta-analyzing the mean effect size by grouping the SPE scales and their associations to LS and CS, this is expected to cause heterogeneity.

A single study in the SPE group makes use of a proprietary questionnaire that consists of three self-reporting items (De Vos & Soens, 2008), titled as ‘perceived employability’. De Vos and Soens adopt one question from the study by Eby, Butts and Lockwood (2003): ‘I believe I could easily obtain a comparable job with another employer’. De Vos and Soens add two new questions: ‘I believe I could easily obtain another job that is in line with my level of education and experience’ and ‘I believe I could easily obtain another job that would give me a high level of satisfaction’. These questions are posed to be reflective of the extent to which respondents believed they were employable, and otherwise no formal definition is provided (p. 452). From observing how these questions are stated, they cover

self-perceptions about one’s own value towards job alternatives in terms of the external labor market, acquired human capital, expected satisfaction with alternatives. It is apparent that compared to the other scales in the group, the questionnaire is at best a superficial measure of employability, regardless of sharing the scale title. In terms of content, it does not explicitly facilitate responses about aspects of employability such as the value of the social network, career management or resilience, which have dedicated items in the scale developed by Rothwell and Arnold, for example (2007). These constructs that are constituents of employability may be indirectly reflected in the responses of De Vos and Soens’ scale. An example would be the value of a social network; even though by the item wording it is not brought to mind

(15)

15 directly, it can still be imagined to have an influence on the response value of the respondent’s belief to easily obtain alternative employment. The lack of depth and precision in this scale is therefore expected to affect the associated outcomes LS and CS, contributing to heterogeneity to the mean effect size of SPE as a group.

The fourth and last scale in the SPE group is ‘perceived employability’ developed by Van der Heijde and Van der Heijden (2006). This scale is applied by two studies that are included in the meta-analysis (Van der Heijde & Van der Heijden, 2006; De Vos, De Hauw, & Van der Heijden, 2011). The full scale consists of 47 items in five subscales of employability, and the questions are posed to both employees and their employers. The five subscales are occupational expertise, which includes

professional knowledge and skills and accounts for social recognition by key individuals, anticipation

and optimization, which covers personal creativity towards future work changes, personal flexibility,

which entails the adaptability towards work or labor market changes outside of the current job, corporate

sense, which includes items that reflect performing well as a part of a team and of the organization as a

whole, and lastly, balance, which refers to the ability to compromise between opposing interests of the employer and the individual’s interests at work and in private. The definition of employability used is ‘the continuous fulfilling, acquiring or creating of work through the optimal use of competences’. Overall, compared to the other scales in the SPE group, this scale is arguably more refined. The underlying definition that has steered the questionnaire design emphasizes the continuity of the process of job

acquisition and maintenance and optimal use of competencies towards that. These differences of emphasis between this scale and the other three are expected to cause heterogeneity in mean effect sizes of SPE as a group. To reduce some of this expected heterogeneity, only the self-reporting end of the scale is taken from the study by Van der Heijde and Van der Heijden (2006). This is consistent with the singularly self-evaluative perspective shared by the other scales. Furthermore, because there is no single summarizing variable for the five subscales, the effect size taken from this study is the average of the five subscale effect sizes, weighted to the number of questions per subscale.

(16)

16 The other study to use this scale only makes use of the self-reporting items (De Vos et al., 2011). In addition, De Vos et al. apply only two dimensions of the five, eight items for occupational expertise and three items of personal flexibility, which are summarized in their study by one variable to denote

‘self-perceived employability’ (p. 442, 2011). Effectively, this scale is a limited measure of employability

compared to the other scales. Nonetheless, by its use towards explaining the variable perceived

employability it can be interpreted as meaning to reflect the same concept, albeit in simpler terms and in a different research context. These inconsistencies are expected to add to heterogeneity in the mean effect size of SPE as a group.

To categorize the SPE group, all four of the included scales can be classified as utilizing an

employability competencies/skills definition. The exception is Rothwell and Arnold’s scale (2007), which

is believed to sufficiently incorporate societal and organizational factors in its external employability dimension, such that the scale may be classified as supporting a contextual conditions definition.

2.3 The Career Adapt-ability Scale (CAAS).

CAAS measures a set of personal resources that describe an individual’s ability to effectively adjust to the requirements of finding employment (Savickas, 1997). CAAS was developed based on career

construction theory (CCT), which proposes that career development and adjustment requires of individuals that they continuously adapt to the social environment in order to achieve subjective and objective career success and a fit of person to environment (Savickas, 2002, 2005). CCT involves the process and outcomes of individuals’ work to meet their own expectations and that of others in regards to successful careers. This includes the tasks of preparing for, starting, and participating in a professional role, next to adapting to changes in the work environment and wok demands. Career adaptability, as defined by Savickas (1997), reflects the individual’s resources for managing the current and future work and career challenges that may impact the individual’s fit of person to environment. Savickas adds that these resources are not stable traits in the way personal characteristics are. Career adaptability thus emphasizes self-regulatory capacities that are within the conscious control of the individual and can be

(17)

17 changed over time. These changes can be prompted by factors existing within the person, the environment and the interaction between them. CAAS has an overall scale and four subscales that measure control, or the level of self-discipline, persistence and effort that is used to take personal responsibility towards one’s development and professional environment, confidence, or the solidity of one’s belief to achieve career goals, problem-solving skills, or tackle obstacles, curiosity, involving the exploration of career-related possibilities and opportunities and concern, which refers to the future orientation in terms of upcoming career challenges and due preparation.

The main difference between CAAS and the SPE group is that career adaptability emphasizes the personal resources required to build a career, whereas the SPE group reflects employability, which in most of its definitions is about getting and keeping a job. If a career is an enduring sequence of progress through working life in various job roles, then career adaptability, which describes the way of maintaining that career or sequence of jobs, is part of employability. In Thijssen et al.’s (2008) categorization, CAAS can be classified as an employability competencies/skills definition.

The studies included in this meta-analysis include five adaptations of the CAAS Form 2.0 (International form by Savickas & Porfeli, 2012; Chinese by Guan et al., 2015; Italian by Santilli, Nota, Ginevra & Soresi, 2014; German and French by Maggiori, Johnston, Krings, Massoudi & Rossier, 2013). CAAS was built from the ground up with the input of vocational research teams from 18 different countries. The collection of input includes a discussion of indigenous theoretical models and culture-specific and culture-general ideas of what constitutes adaptability. In this way, the agreement of the teams on the four subscale constructs is based on cross-national perspectives (Savickas & Porfeli, 2012). This method is assumed to contribute to the reliability of the scales, because the international agreement is part of its initial development, instead of a translation after development, which may lead to linguistic or content misdetermination. For this reason, the expected mean effect sizes of CAAS on outcomes CS and LS is expected to show a modest base level of heterogeneity relative to the SPE group.

(18)

18

2.4 Employability and Well-being.

The concept of subjective well-being could describe the group of outcomes that were commonly found in the overview, in different relations to employability. Page et al. (2009) explain that subjective well-being has three core components: (1) high levels of positive affect, (2) low levels of negative affect and (3) a cognitive evaluation of one’s satisfaction with their life as a whole. Subjective well-being summarizes a substantial group of constructs that were observed in the employability overview such as, cynicism,

burnout, exhaustion, purpose in life, life meaning, job satisfaction, or life and career satisfaction. This

common occurrence of well-being related constructs in empirical research of employability would seem likely from the perspective that higher employability would lead to an individual experiencing more satisfaction from work and life. Lounsbury et al. (2004) contend that from the perspective of individual development, personality is a precursor to a career and as a consequence leads to both career and life satisfaction. Being that employability is defined in terms that in large part coincide with personality, through dispositions and behaviors, it is unsurprising that employability is similarly related to satisfaction with both career and life.

Employability is likely to lead to greater well-being because it implies that the individual takes control of his or her employment prospects. Self-determination theory argues that increased self-regulation, when successful, leads to improvement in mental health and self-motivation when three psychological needs are met - competence, autonomy and relatedness (Ryan & Deci, 2000). When the individual is at liberty to, and actually does, successfully takes control of increasing employability by improving human and social capital, these three needs are met. Well-being can thus be enhanced through employment-related skills and knowledge acquisition, career (and possibly job) autonomy, and

professional socialization.

Among others, life satisfaction and career satisfaction are measures indicative of how well-being manifests itself. These two outcomes are chosen as constant variables in the comparative meta-analysis.

(19)

19

2.5 Life Satisfaction (LS).

Life satisfaction involves a judgmental process, wherein individuals evaluate their quality of life based on their own unique criteria (Pavot & Diener, 1993). Schalock et al. explain the term ‘quality’ is related to human values, such as satisfaction, happiness and health, and ‘life’ would reflect aspects important to existence such as work, family and health (2002). An overall sense of satisfaction with life is one of the three components of subjective well-being, next to frequent episodes of positive affect and minimal occurrence of negative affect (Diener et al., 1984; Myers & Diener, 1995; Page et al., 2009).

Measures for LS included in the meta-analysis are the Satisfaction with Life Scale (SWLS) by Diener et al. (1985), the Orientation to Life Questionnaire (OLQ) by Antonovsky (1987), Silla et al.’s scale for Life Satisfaction (2008) and the non-organizational (i.e. life satisfaction) component of Gattiker and Larwood’s scale for subjective career success (1986). These four scales are all self-reporting Likert-scale questionnaires with response ranges of either five or seven points.

The employability studies by De Cuyper et al. (2011) and Silla et al. (2008) use a LS scale to which is an adaptation of Isaksson et al.’s prior work (2003). This scale includes questions about seven domains of life satisfaction, f.e. ‘how do you currently feel about your life in general?’. This scale has respondents report their general sense of satisfaction with life, similar to what is achieved by Diener et al.’s SWLS. As such, it is considered an adequately equal measure of LS.

Van der Heijde and Van der Heijden (2006), in their study of employability, use the life satisfaction component of Gatikker and Larwood’s for subjective career success (1986). It presents statements for self-scoring such as ‘happy with my private life’ or ‘satisfied with my life overall’. As a questionnaire that is adequately similar to the general sense of satisfaction with life as measured by Diener et al.’s SWLS, this study is deemed fit for inclusion in the meta-analysis.

The employability study by Harry and Coetzee (2013) uses the Orientation to Life Questionnaire (OLQ) by Antonovsky (1987) OLQ measures overall ‘sense of coherence’ and consists of three subscales;

(20)

20 ‘global orientation that expresses the extent to which one has a pervasive, enduring though dynamic feeling of confidence that (1) the stimuli deriving from one’s internal and external environments in the course of living are structured, predictable, and explicable; (2) the resources are available to one to meet the demands posed by these stimuli; and (3) these demands are challenges, worthy of investment and engagement’ (1987, p. 19). From this definition it shows that sense of coherence is indicative of an overall feeling of confidence that is similar the overall sense of satisfaction with life. Both SWLS and OLQ scales measure overall contentment and therefore OLQ is considered to have sufficient conceptual overlap with SWLS to be included in this meta-analysis as a measure of LS.

2.6 Career Satisfaction (CS).

To face the current labor market context that is more volatile than ever, careers have become increasingly boundaryless, with a new standard for inter-firm mobility and unpredictability. Accordingly, career research now emphasizes the personal meaning of career success as the primary focus for evaluation. Subjective career success is a construct that comprises feelings of satisfaction and accomplishment regarding the individual’s career (Seibert, Kraimer, & Liden, 2001; Hall, 2002; De Vos & Soens, 2008).

De Vos et al. (2011) distinguish CS from perceived marketability; both constructs are indicative of subjective career success, of which CS refers to past accomplishments, and perceived marketability concerns future prospects. CS describes the individual’s subjective evaluation of overall career progress, independent of any single employer. Tolentino et al. (2014) define career success as ‘an evaluative concept of accrued positive attitudinal and behavioral outcomes resulting from one’s work experiences’ and distinguish the individual’s subjective appraisal of CS as an indicator of these positive outcomes, similar to De Vos et al. (2011).

The studies included in the meta-analysis make use of two types of scales for CS, the Career Satisfaction Scale developed by Greenhaus et al. (1990) is used in six out of eight studies. The other two make use of the scale by Martins et al. (2002). Greenhaus’s scale consists of five items, based on defining CS as an internally determined career outcome. Examples are ‘I am satisfied with the progress I have

(21)

21 made toward meeting my goals for advancement’ and ‘I am satisfied with the success I have achieved in my career’ (1990). Martins et al.’s scale consists of three items, respondents were asked to report on a seven-point Likert scale, whether they were generally satisfied with their career status, their present job and their progress in terms of promotions. Martins et al. provide no further elaboration on what defines the construct CS (2002). Overall, the items in both versions of the scale are considered to be adequately similar and therefore are both included to represent the outcome CS in the meta-analysis.

2.7 Conceptual Model.

In this paragraph the conceptual model presents theoretical arguments operationalized into formal hypotheses by synthesizing the points from the theoretical background.

The SPE group as a whole can be considered as a collection of snapshot measures of

employability, a self-perceived, general evaluation of the current level of skills, knowledge and network that facilitates acquiring alternative employment. The argument here is that within the SPE group, the only notable exception to this is the anticipation and optimization dimension of Van der Heijde & Van Der Heijden's scale (2006). This dimension is considered to be the only explicit measure of future-orientation and proactivity, compared to all the other scales in the SPE group. CAAS on the other hand, with its conceptual focus on the career is considered by definition as future-oriented and reflective of proactivity. Future-orientation and proactivity are attitudes and behaviors that make it possible to constantly realize positive experiences and avoiding negative experiences, which are key to satisfaction with life (Diener, Emmons, Larsen, & Griffin, 1985). Therefore, because of its reflection of future orientation and proactivity, the expected association of CAAS to LS is higher than the association of SPE to LS.

H1a: The mean effect size of employability in relation to LS is moderated by the type of scale,

(22)

22 CAAS is measure of personal resources of adaptivity that are harnessed towards realizing a career, or an enduring sequence of progress in different job roles. Individuals that score high on overall CAAS, are expected to be successful in meeting the challenges of building a career and as a consequence, experience subjective career success. In this manner, high CAAS scores are also expected to associate to high scores career satisfaction, which is an indicator of subjective career success. Moreover, this

association is expected to be higher than that of the SPE group. To clarify, the SPE group consists of measures employability that are not directly indicative of job acquisition over time, which is required for successful career building. The SPE group measures aspects of employability that directly relate to the current status of employment, more so than to building a career. While this measure may be influenced by enduring, future-oriented behaviors and proactive disposition, which indirectly leads to some respondents to score high on SPE, it does not directly reflect it. The prediction is that a high score of SPE measures would not always coincide with a high score on career satisfaction, in such a way that the mean effect size of SPE to CS is expected to be lower than that of CAAS to CS.

H1b: The mean effect size of employability in relation to CS is moderated by the type of scale,

such that the mean effect size is higher for the CAAS type.

The studies are not expected to show perfectly homogeneous effect sizes, because they are performed in different methodological contexts. There is variance to be expected in mean effect sizes for both scales, based on the differences in language, sample characteristics, and organizational and cultural contexts. The prediction is that the reported mean effect sizes from the studies that associate

employability (SPE and CAAS) to LS and CS show a basic level of so-called methodological heterogeneity.

Both scales are representative of employability as an over-arching concept. The prediction here is that the type of scale will be associated with a different level of mean effect size heterogeneity. The scales grouped under SPE are assumed to indicate the same conceptual domain of self-evaluated employability,

(23)

23 based on the synonymity in their titles and commonalities in their underlying definitions of employability, which was shown in the theoretical framework. However, the theoretical framework also illustrated the differences in question item wording and scale design. The scales all measure self-evaluated

employability, but they exclude or include different aspects of employability. The quality and quantity of scale items measure employability to different extents and to different levels of abstraction. This is expected to contribute to heterogeneity associated with the outcome mean effect sizes.

Based on arguments presented in the theoretical framework, CAAS is expected to be a more reliable empirical measure of employability. This reliability results in the correlations to LS and CS to be more stable, and the mean effect size that summarizes these correlations will have less corresponding heterogeneity when compared to SPE. The prediction is that mean effect sizes for employability will show heterogeneity, beyond methodological, that is moderated by the type of scale, such that heterogeneity levels associated with SPE are higher.

H2a: The heterogeneity corresponding to mean effect sizes between employability and LS is

higher for the SPE type scales than the CAAS type.

H2b: The heterogeneity corresponding to mean effect sizes between employability and CS is

higher for the SPE type scales than the CAAS type.

3. Methods.

3.1 Data Collection.

Constructing an overview of empirical employability served two purposes. First was to gain a comprehensive perspective of the variation in this type of research, and second, as an inventory of possible effect size candidates to include in a comparative meta-analysis. The strategy for retrieval was to first chart a conceptual map of all variables associated with employability in research. The map of

(24)

24 involved. This included employability itself, and all the antecedents and outcomes that could be thought of grouped into umbrella concepts (f.e. ‘job security’ includes both job security and job insecurity). The twenty constructs were then divided evenly, among five researchers, with for each the goal to clarify the concepts by gathering academic definitions through internet search. Another goal was exhaustiveness, met by keeping an eye open for references towards concepts that could not be categorized in the tentative map of twenty concepts. New concepts were not added in this way. The map of concepts and definitions was the referential framework to base the search terms used in multiple academic databases. The map also provides the categorization of constructs in the coding process, based on conceptual similarities. Twenty-one databases were deemed suitable candidates of which fourteen were subsequently excluded for being superfluous. The final set of databases for optimal variance in results consisted of PsycINFO, Business Source Premier, ERIC, Medline, Web of Science, Sociological abstracts and Scopus. From these

databases, a total of 12,764 articles were collected, using PRISMA guidelines as minimum requirements for meta-analytic inclusion (Moher, Liberati, Tetzlaff, & Altman, 2009). Removing duplicates resulted in a total of 8,941 articles for selection. Appendix A contains the full list of candidate databases that were considered.

3.2 Selection.

The 8,941 resulting articles were divided into six, with each of the researchers setting out to manually screen articles for inclusion into a number practical for coding. A preliminary round of selection was conducted on a limited set of fifty articles for each of the researchers. This was to ensure inter-rater reliability by determining to what extent the researchers would independently select the same articles for inclusion, based on the abstracts and using a predefined set of inclusion criteria. After this selection pilot, a round for evaluation took place, to discuss and compare selection results and possibly recalibrate the inclusion criteria. The first set of criteria included all published and non-published, published after 1980 and all types of participant samples expect for mean age under 18, and excluded qualitative studies. After evaluation, the first round of selection was performed on approximately 1,500 articles for each of the six

(25)

25 researchers, based on abstract and title. This resulted in a base of 3,981 studies for coding. Time

constraints required that this number would be reduced further, so new criteria were added: studies of sample populations that could be considered not representative of the general professional population were excluded: (veteran) military, intellectually or physically disabled and rehabilitating criminal offenders. Furthermore, (job) mobility was excluded as a quantitative measure of employability for the studies where this reflected objective migration statistics suitable for econometric research (f.e. Li, 2013), instead of individual characteristics underlying employability. The total number of studies for coding as a result was 1,360. Unfortunately, due to an irrecoverable error that occurred while transferring the

database, 159 articles went missing. The final number for coding was 1,201 articles. This would seem inconsequential to the overall study, considering that for the process of coding, not all 1,201 articles would be coded.

3.3 Coding.

The process of coding involves extracting the effect sizes manually, by closely examining each study and registering each type of employability measure and its first-order correlation to every antecedent or outcome reported in each study. For the sake of simplicity, first-order Pearson correlations were taken as effect sizes, because they can be considered the least confounded by other variables. Effect sizes taken from moderation, mediation, structural equation modelling, etc. would reflect in their value the other variables that are part of the complex relationship. Excluding other candidate effect sizes such as

longitudinal or differential, allows for more consistency among the effect sizes. This is primarily because it relieves the requirement of mathematical conversion and its attached consequence of distortion (Morris & DeShon, 2002). The Pearson correlation coefficient represents the strength of association between two inherently continuous variables and is denoted by ‘r’. The complete set information that was extracted in the coding process can be referenced in Appendix B.

From the pool of 1,201 articles, a random representative sample was taken of 50 articles per researcher. The variability in time requirements for coding article, and the time constraints of the project

(26)

26 led to this conservative number. Wherever possible, any of the researchers could volunteer to perform additional coding. To maintain consistency between researchers coding independently, an overlap of 20% was installed. Every member of the coding team had five articles to code in common with one other member, and another five articles with yet a different member. Overall, the pairwise coding results were deemed adequately similar, so in terms of coding quality, no adjustments were necessary. The supervising member oversaw the matching evaluation and in this process, out of all the coded pairs, the articles that were coded for a larger quantity of effect sizes were included. The total number of studies that were coded was 66. For this meta-analysis, there was a manual search by snowballing references and internet database search. This was to add articles pertinent to the specific research question, that may have been put aside during any step in the selection process. This repeat selection added five studies resulting in a total number of sixteen studies that were coded for this meta-analysis. A diagram of the complete selection and coding process in terms of inclusion numbers is provided in Appendix C.

For the meta-analysis to retain independent and identical distribution in the data, one study was left out (Clauset, 2011). This article was assumed to describe the exact same sample examined by another study that was already included (Cuyper, Bernhard-Oettel, Berntson, Witte, & Alarco, 2008; De Cuyper et al., 2011). Both articles report effect sizes relevant for the meta-analysis (SPE to LS), yet describe near identical sample characteristics (p. 495, 2008; p. 1491, 2011). The more recent article was included.

3.4 Analysis.

3.4.1 Corrections for systematic artifacts.

Following the guidelines of Hunter, Schmidt and Jackson, (1982) a reliability attenuation of the

correlation coefficients was performed. Using the reliability coefficients (reported as Cronbach’s alpha) that were retrieved together with each scale that is included in this study, the correlations can be adjusted to mitigate the impact of relatively unreliable scales on the variance of correlation.

Another artifact to impact measurement observations is range restriction. This occurs when one of the variables in correlation has a standard deviation that is lower in the sample than in the population. For

(27)

27 all variables, however, the standard deviation of the population was unknown. Therefore, a correction to account for range restriction was not performed, because it could not be calculated. A third correction for artifacts from dichotomization was also not performed because none of the question items created a dichotomy in their wording statement.

3.4.2 Adjustments of effect sizes and reliability coefficients.

The effect size for SPE from the study by Dacre Pool and Qualter (2013) was reported as separated in Rothwell and Arnold’s (2007) dimensions of internal and external employability. The average was weighted to the number of question items per dimension, resulting in an approximation of overall effect size that would provide a better fit with the other scales in the SPE group. There was, however a reported overall reliability coefficient so no further adjustments were required. Similarly, in Konstam et al. (2015) and Santilli et al. (2014) CAAS and in Van der Heijde and Van der Heijden (2006) SPE were reported in its segregated effect sizes. This was averaged in the same way according to the number of scale items. The reliability coefficient was construed to its average. It was considered more beneficial to the meta-analysis to include these studies than to leave them out for inconsistency.

3.4.3 Fishers z-transformation.

Fisher’s z-transformation allows the effect sizes to be weighted by inverse variance weighting, which is theoretically optimal and preferred over straightforward sample size weighting (Geyskens, Krishnan, Steenkamp, & Cunha, 2009). Inverse variance weighting grants prominence to effect sizes in a simultaneous analysis, according to the standard error of each study. Standard error is a more suitable indicator of precision than sample size (Lipsey & Wilson, 2001). To achieve this, the correlations must be transformed, because the standard error of correlation is dependent on the correlation itself and this is not acceptable for the weighting calculation. By using Fisher’s z-transformation the calculation for inverse variance weighting is made possible, because effect sizes are approximated by a standardized z-value which has a standard error that only depends on sample size.

(28)

28

3.4.4 Publication bias.

To account for publication bias, which is a common issue of meta-analysis, a file-drawer analysis was performed. This bias occurs when articles that report significant results are more likely to be published in academic journals, thus skewing the perception that the true effect is also significant. The true effect in the population is optimally modelled by the combination of all its research, including unpublished reports that may have fallen by the wayside because the evidence was not statistically significant. To this end, Rosenthal’s ‘fail-safe N’ is reported for both the employability-LS and employability-CS meta-analyses (1979). Fail-safe N is an estimate of the number of studies reporting non-significant results that it would take to make the mean effect size under study become not significant (α=0.05). To add, the selection criteria (§3.2) also included unpublished articles to proactively account for the file drawer problem. Nevertheless, the steps of the selection process may have diminished the possibility of retrieving such unpublished reports through dilution (Appendix C).

3.5 Moderation meta-analysis.

The first hypotheses, H1a and H1b, are to establish that the type of scale moderates the outcomes between employability and LS and CS. Moderation analysis occurs in two stages, detection and estimation. (Steel & Kammeyer-Mueller, 2002). First, the meta-analysis will compare the group effect sizes to determine if they are statistically different from each other, which is known as the homogeneity test. The homogeneity test is performed by using the MEANES.SPS macro for SPSS, developed by Wilson (2010). A significant Q-value suggests that variability between effect sizes is at a significant level, such that it warrants search for a moderator (Hedges & Olkin, 2014). To further the support the establishment of moderators to likely exist, the Hunter and Schmidt’s 75% rule is applied (2014). The 75% rule states that if less than 75% of the variability in effect sizes is due to artifacts (e.g. range restriction, sampling error, unreliability), the variance is likely substantive and supports the search for moderators. The rationale for using the 75% rule is that not all of the artifacts that cause statistical noise in data can be corrected for, especially in cases

(29)

29 where the necessary information is missing. Only the attenuation for reliability was performed in this study (§3.4.1), which justifies the use of the 75% rule.

The second stage will estimate the moderation by comparing mean effect sizes (MES) to determine if they significantly differ from each other. In this subgroup analysis, MES are observed for CAAS – LS compared to SPE – LS, and CAAS – CS compared to SPE – CS, using the METAF.SPS macro developed by Wilson (2010). The subgroups are entered as a categorical moderator, and the METAF.SPS macro then produces a meta-analysis analogous to a traditional one-way ANOVA. METAF.SPS presents the overall Q-statistic (QT), in addition to its segregation into Q-statistics for

heterogeneity both between (Qb) and within (Qw) groups. A significant Qb for either of the outcome

groups (LS or CS) will suggest that the type of scale affects the association as a moderator.

The second set of hypotheses, H2a and H2b, predict that the heterogeneity levels associated with subgroup MES are moderated by the type of scale instrument. The scale type is hypothesized to account for a significant part of the heterogeneity within subgroup mean effect sizes, denoted by the Qw-statistic.

If Qw for CAAS and SPE are then converted to I2, this will allow for a comparison of relative

heterogeneity by the standard of Huedo-Medina, Sánchez-Meca, Marín-Martínez and Botella

(I2_{<25%=low, 25%<I}2_{<50%=moderate, I}2_{>75%=high level of heterogeneity) (2006). This is considered}

meaningful if Qw is shown as statistically significant at the α=0.05 level.

In line with the arguments of the conceptual model (§2.7), both scales are assumed to show at least a base level of heterogeneity. In other terms, a part of the heterogeneity denoted by Q is systematic due to sampling error. This is reflected by the application of the random effects model, which likewise assumes that the variability between effect sizes is due to sampling error plus variability in the population of effects.

(30)

30

4. Results.

4.1 Employability and Life Satisfaction.

Homogeneity testing from the MEANES.SPS macro provides Q=94.6574, significant at α=0.05 (p<0.0000; df=7). Hunter and Schmidt’s 75% rule permits (0.01096/0.01914=) 57,26%, which is lower than 75% (2014). On accounts of both indicators, heterogeneity is assumed and the existence of

moderators is suggested. To control for publication bias, Rosenthal’s Fail-safe N for LS (1979) yields 89, suggesting that this meta-analysis needs to include at least 89 more studies that report non-significant findings to disconfirm significance of the total mean effect size. This is considered an acceptably large number to proceed with the current study. The results are presented in Table 1.

From the METAF.SPS analysis for LS, mean effect sizes of the CAAS group and the SPE group are 0.39 (p<0.0000) and 0.25 (p=0.0005), respectively, both significant at the α=0.05 level. This is in line with the prediction of H1a, the mean effect size for the CAAS type scale is higher than the mean effect size of the SPE type. The subgroup mean effect sizes are presented in lines 5 and 10 in Figure 1. The Qb

-statistic is 1.8504, but considered not -statistically significant (p=0.1737). It can therefore not be

confirmed that any observed variation between studies is due the type of scales, which was predicted by H1a.

The second hypothesis H2a, predicted that the heterogeneity levels of observed mean effect sizes would be moderated by the type of scale. For the group of CAAS – LS, Qw; CAAS-LS=0.6925, and

conversion presents I2

CAAS-LS=0.00%, which corresponds to no heterogeneity by Huedo-Medina et al.’s

standard (2006). Qw; CAAS-LS is not statistically significant, however (p=0.8750). As regards the SPE – LS

group of studies, Qw; SPE-LS=5.3338 (p=0.1489), conversion of which grants I2SPE-LS=43.75%, suggesting a

moderate level of heterogeneity. Although the prediction of H2a is suggested by the higher level of I2_for

the SPE group (43.75%>0.00%), the Qw-statistics that underlie both are not significant, therefore

(31)

31

Table 1. MES and confidence intervals are reflected in Figure 1.

Figure 1. Forest plot of effect sizes. Lines 1-4 and 6-9 denote individual effect sizes. Lines 5 and 10 denote subgroup MES. Line 11 denotes the overall MES.

4.2 Employability and Career Satisfaction.

The homogeneity test from the MEANES.SPS macro provides Q=46.4523 (p=0.0000; df=7), significant at the α=0.05 level. By (0.00526/0.00839=) 62.69%, average sampling error accounts for less than 75% of total variability, which passes the ruling of Hunter and Schmidt (2014). On these grounds, heterogeneity is assumed present and due to more than sampling error alone, which suggests the possibility of

moderators. To account for publication bias, Rosenthal’s Fail-safe N is calculated as 471 (1979). This large number of non-significant findings necessary that were not included, is considered high enough to proceed with the meta-analysis. The results for career satisfaction are summarized in Table 2.

(32)

32 The first hypothesis regarding CS studies, H1b, predicted that the outcome of employability on CS would be higher for CAAS than for SPE. From the METAF.SPS macro, mean effect sizes for CAAS – CS is 0.52 (p=0.0000) and for SPE – CS it is 0.42 (p=0.0000). These results are in accordance to the prediction of H1b, suggesting that the association of employability to LS is higher for the CAAS type of measurement. Figure 2 depicts the subgroup MES in lines 5 and 10. The corresponding Qb-statistic is

1.8699 and not significant (p=0.1715). Therefore, no conclusion can be drawn to support that the differences between subgroup mean effect sizes are meaningfully different. Consequentially, H1b is rejected.

The second hypothesis of the CS meta-analysis predicts meaningfully different heterogeneity levels associated with the subgroup mean effect sizes that are due to the type of measurement scale. For the CAAS – CS group, Qw; CAAS-CS=2.5426 is not significant (p=0.4676). This converts to I2CAAS-CS=0.00%,

which signifies low heterogeneity by Huedo-Medina et al.’s standard (2006). Likewise, for the SPE – CS group, Qw; SPE-CS=3.8630 is not significant (p=0.2766). The I2SPE-CS is 22.34%, which would indicate a

modest level of heterogeneity within the subgroup. Although the prediction of H2a is suggested by the difference in I2_{(for SPE heterogeneity levels are higher; 22.34%>0%), the underlying Q}

w-statistic are not

significant. Accordingly, H2b is rejected.

(33)

33

Figure 2. Forest plot of effect sizes. Lines 1-4 and 6-9 denote individual effect sizes. Lines 5 and 10 denote subgroup MES. Line 11 denotes the overall MES.

5. Discussion.

The results for two meta-analyses of employability show mixed, but ultimately inconclusive evidence that the type of scale used to measure employability may moderate its association to LS or CS. The results show a tendency in line with the first hypotheses predicting CAAS to be more strongly associated to CS and LS than SPE. However, for both CAAS and SPE this variance between subgroup mean effect sizes cannot be confirmed by this analysis, because the Qb-statistic in both cases was not significant. The first

explanation for this would be the lack of quantitative studies that were included for each group. Despite this, the presence of heterogeneity in all four subgroups was confirmed, based on significant primary Q’s. A possible reason that every Qb-statistic derived for each group was not significant, is because the sample

size for each subgroup is relatively small (k=4 in all four subgroups). The observation that in both cases of LS and CS the mean effect sizes differed in the direction as was predicted provides at best a marginal indication of the tendency within the current studies. The expectation is that including more studies that report on the same associations could possibly average out some of the statistical imprecision caused by a small sample size. At the time of writing, more studies were not yet available. Moreover, all the studies that were included were relatively recent in terms of their publishing year, the CAAS group ranges 2006 to 2015 and the SPE group ranges 2007 to 2016. This may indicate that the association of employability to CS and LS is a relatively new subject of empirical employability research.

A meta-analysis of measuring employability through its effects on life and career satisfaction