Job security has been a central tenant of civil service since the Pendleton Act of 1883 (Perry, Hondeghem, & Wise, Reference Perry, Hondeghem and Wise2010: 687) in the USA. As Van Riper (Reference Van Riper1958) noted, the job security rules introduced in the USA were not novel. They were rooted in the British civil service, from which US civil service reformers borrowed liberally. For most of the 20th century, public employers granted their employees high levels of job security in the USA, both within the federal government and across levels of government as well as in governments across the European Union (EU) states. However, the job security that public institutions such as governments and educational enterprises grants to employees has been a source of continuing controversy, as prominent in education reform of 2010 as civil service reform of the 1970s (Savas & Ginsburg, Reference Savas and Ginsburg1973; Campbell, Reference Campbell1978). The attack on the job security of government employees continues unabated today, with states, such as Georgia and Florida, and federal agencies, such as the Department of Veteran Affairs, promoting at-will employment as an alternative to traditional job tenure systems (Kellough & Nigro, Reference Kellough and Nigro2006; Perry, Hondeghem, & Wise, Reference Perry, Hondeghem and Wise2010: 687; Kettl, Reference Kettl2015; Hijal-Moghrabi, Sabharwal, & Berman, Reference Hijal-Moghrabi, Sabharwal and Berman2017; Davidson, Reference Davidson2018) in the USA. In addition, governments across the EU have been a significant policy shift toward a more liberal model of weaker job security and increased labor market flexibility by weakening and removing employee protections in postcrisis Europe (Heyes & Hastings, Reference Heyes and Hastings2016; Schaufeli, Reference Schaufeli2016; De Cuyper, Piccoli, Fontinha, & De Witte, Reference De Cuyper, Piccoli, Fontinha and De Witte2018; Hastings & Heyes, Reference Hastings and Heyes2018).
However, this recent policy choice about job security has been based largely on normative and ideological considerations rather than behavioral science evidence. In the absence of better evidence, we are faced with making decisions largely on ideological grounds (viz., federal reforms in the US Department of Homeland Security and Department of Defense during the George W. Bush Administration; state reforms in Georgia and Florida). Given the fact that politicians will not only continue to voice their views but also shape public law, we do not expect that ideology-based policies will disappear. Thus, recent developments indicate that we should emphasize the balance between normative considerations and behavioral science evidence related to the design of job security rules. We believe research accumulated during the last three decades about job security will give us an opportunity to support more debate and administratively-rationale alternatives to current job security system and guide future public and organizational policy making.
The objective of this article is to synthesize public and private sector research regarding the relationship between job security and employee work attitudes (i.e., job satisfaction and organizational commitment) to provide a more evidence-based footing for future public job security policy. Job security is viewed as an important determinant of individual and organizational performance if for no other reason than its influence on an organization's ability to discipline or remove poor performers (Campbell, Reference Campbell1978). However, the strength and form of the relationship remain unclear: studies have found positive (e.g., Davy, Kinicki, & Scheck, Reference Davy, Kinicki and Scheck1997; Preuss & Lautsch, Reference Preuss and Lautsch2002), negative (e.g., Andaleeb, Reference Andaleeb1996; Cavanaugh & Noe, Reference Cavanaugh and Noe1999; Jeon, Reference Jeon2009), and curvilinear (e.g., Brockner, Grover, Reed, & Dewitt, Reference Brockner, Grover, Reed and Dewitt1992) relationships between job security and work attitudes. Quantitative meta-analysis method allows general conclusions and principles to be drawn from a previously accumulated job security research. By means of meta-analysis, we will be able to provide more conclusive answers to current job security rule change issues and phenomena, which, in turn, make research findings more understandable to policy maker and practitioners (Le, Oh, Shaffer, & Schmidt, Reference Le, Oh, Shaffer and Schmidt2007: 7–9).
Apart from the study of the variable relationship between job security and employee work attitudes, the present study will also conduct subgroup analysis. Five significant subgroup analysis will be conducted: named organizational type; employee's nationality; tenure; age; proportion of females. By this, the variability of the findings from the research literature will be explained. They will also be helpful for clarifying the variability noticed among these studies and in this way it will be clarified as to when job security may increase or outcome of work attitude may decrease. Finally, they are relevant as to what extent different (levels of) or job securities help or obstruct the outcome of work attitudes or whether they hamper at all. In this way, we believe current meta-analysis findings offer practical guideposts for future policy that contrast with recent, prominent reforms.
The paper begins with a review of the literature on theoretical formulations of the role of job security and workplace outcomes, and empirical research about job security and employee attitudinal outcomes. Next we present the data and methods used in the study. We then summarize the key results. We conclude with a discussion of the results and suggestions for further research.
Review of Literature
Theoretical understanding of job security
Generally, job security is defined as a legal employment contract between employee and agency that lead to assurances for continued employment (Greenhalgh & Rosenbtatt, Reference Greenhalgh and Rosenbtatt1984: 439; Romzek, Reference Romzek1985: 283). Greenhalgh and Rosenbtatt (Reference Greenhalgh and Rosenbtatt1984) developed a theoretical model to understand the nature, causes, and consequences of job insecurity. They defined job insecurity as powerlessness to maintain desired continuity in a threatened job situation. The general assumption underlying this definition is that job insecurity can be understood by an individual's perceptions of the immediate work environment. Regarding the theoretical model for job insecurity, Greenhalgh and Rosenblatt argued that individual's subjective threat represented by job insecurity is derived from objective threat by means of the individual's perceptual processes. They argue that job insecurity, in turn, is projected to affect a variety of individual organizational behaviors such as productivity, turnover, and resistance to change. Thus, Greenhalgh and Rosenblatt's research can be interpreted that subjective job security (they call it ‘job insecurity’) is influenced by objective job security and it is projected to affect a variety of employee work attitudes and behaviors.
Greenhalgh and Rosenblatt's model of job insecurity is a reasonable approach to a very complex job security issue. Based on their theory, researchers assume that workers in permanent jobs (high level of objective job security) will exhibit lesser perceptions of job insecurity (high level of perceived job security) than workers on temporary contracts (low level of objective job security) and this leads to increased positive employee work attitudes by giving them a surer expectation of employment continuity (Ashford, Lee, & Bobko, Reference Ashford, Lee and Bobko1989; Pearce, Reference Pearce, Cooper and Rousseau1998; De Witte, Reference De Witte1999; Klandermans, Hesselink, & Vuuren, Reference Klandermans, Hesselink and Vuuren2010).
Empirical study on the relationship of job security with work attitudes and behavior
The general premise behind job attitudes is that some employees enjoy their jobs and experience an emotional connection to their work and organization, whereas others do not (Michel & Bowling, Reference Michel and Bowling2013: 96). Job attitudes refer to these work-related evaluations, which include job satisfaction and organizational commitment (Sverke, Hellgren, & Naswall, Reference Sverke, Hellgren and Naswall2002; Michel & Bowling, Reference Michel and Bowling2013: 96). According to attitudinal theory, these work attitudes precede general work behaviors, which include job performance and turnover (Jaramillo, Mulki, & Boles, Reference Jaramillo, Mulki and Boles2011: 343).
While the concept of job security as a contract for continued employment is accepted as leading to desirable consequences, recent research on relationships of job security with employee attitudes and behavior yields mixed and contradictory findings (Wetzel, Soloshy, & Gallagher, Reference Wetzel, Soloshy and Gallagher1990; Connelly & Gallagher, Reference Connelly and Gallagher2004). Some studies that compare permanent and temporary workers report that permanent contracts were associated with higher levels of organizational commitment (Morrow, McElroy, & Elliot, Reference Morrow, McElroy and Elliot1994; Martin & Hafer, Reference Martin and Hafer1995; Van Dyne & Ang, Reference Van Dyne and Ang1998) and higher job satisfaction (Hall & Gordon, Reference Hall and Gordon1973; Miller & Terborg, Reference Miller and Terborg1979; Krausez, Brandwoin, & Fox, Reference Krausez, Brandwoin, Fox and Freese1995).
Other studies find the opposite pattern. Some studies show that employees under short-term contracts show positive job attitudes and work behaviors than permanent counterparts (Roberts, Glick, & Rothchford, Reference Roberts, Glick and Rothchford1982; Eberhardt & Shani, Reference Eberhardt and Shani1984; Jackofsky & Peters, Reference Jackofsky and Peters1987; Fields & Thacker, Reference Fields and Thacker1991; Fenton-O'Creevy, Reference Fenton-O'Creevy1995; Sinclair, Martin, & Michel, Reference Sinclair, Martin and Michel1999). In a survey of 2966 unionized retail employees of a large Midwestern ‘supercenter’ retailer (i.e., grocery and general merchandiser) with 109 locations across five states in the USA, Sinclair, Martin, and Michel (Reference Sinclair, Martin and Michel1999) found that low levels of job security, as represented by a short-term contract, was associated with higher job satisfaction, organizational commitment than permanent workers. Similarly, lower levels of job security in the form of temporary contracts of employment present they had a higher, rather than lower levels of job satisfaction and commitment to the organization (McDonald and Makin, Reference McDonald and Makin2000).
Still others show job security is not associated with work attitudes and behaviors. As cited by Van Dyne and Ang (Reference Van Dyne and Ang1998) and Tansky, Gallagher, and Wetzel (Reference Tansky, Gallagher and Wetzel1995) found that no differences in self-reports of organizational commitment or affective commitment between temporary workers and regular permanent workers in nursing departments. Their findings show that the mental health and work commitment of temporary employees was not different from permanent employees. A large body of empirical research looks at how job security is related to employee work attitudes and behavior. The studies yield mixed and conflicting findings.
Hypothesized relationships between job security and work attitudes
This research uses social exchange theory as a framework for studying the relationships between job security and employee job satisfaction and organizational commitment. In general, organizational researchers have used the social exchange concept to explain how the organization's investments in human resource (HR) practices and the organizational environment will elicit positive work attitudes and behavior (Gould-Williams & Davies, Reference Gould-Williams and Davies2005; Gould-Williams, Reference Gould-Williams2007; Nishii & Mayer, Reference Nishii and Mayer2009; McClean & Collins, Reference McClean and Collins2011; Van de Voorde, Paauwe, & Van Veldhoven, Reference Van de Voorde, Paauwe and Van Veldhoven2012).
According to Blau (Reference Blau1964), social exchange can be defined as ‘voluntary actions of individuals that are motivated by the returns they are expected to bring and typically do in fact bring from others’ (Blau, Reference Blau1964: 93). On this basis, employees who positively value HR practices will reciprocate through showing attitudes and behaviors that are valued by the organization (Eisenberger, Fasolo, & Davis-LaMastro, Reference Eisenberger, Fasolo and Davis-LaMastro1990; Gould-Williams, Reference Gould-Williams2007; Van de Voorde, Paauwe, & Van Veldhoven, Reference Van de Voorde, Paauwe and Van Veldhoven2012).
These employee reactions to HR practices depend on employees’ perceptions of how committed the employing organization is to them (Eisenberger, Fasolo, & Davis-LaMastro, Reference Eisenberger, Fasolo and Davis-LaMastro1990; Romzek, Reference Romzek1990; Wayne, Shore, & Liden, Reference Wayne, Shore and Liden1997; Gould-Williams & Davies, Reference Gould-Williams and Davies2005; Ashikali & Groeneveld, Reference Ashikali and Groeneveld2015). For instance, Eisenberger, Fasolo, and Davis-LaMastro (Reference Eisenberger, Fasolo and Davis-LaMastro1990) describe that ‘positive discretionary actions by the organization that benefited the employee would be taken as evidence that the organization cared about one's well-being’ (Reference Eisenberger, Fasolo and Davis-LaMastro1990: 51). It can thus be argued that when employees perceive that organizations value and deal equitably with them, they will reciprocate these ‘good deeds with positive work attitudes and behaviors’ (Hass & Deseran, Reference Hass and Deseran1981; Aryee, Budhwar, & Chen, Reference Aryee, Budhwar and Chen2002: 268; Gould-Williams & Davies, Reference Gould-Williams and Davies2005: 4).
As we discussed, job security refers to an employee's expectations about the stability and longevity of his or her job in an organization (Greenhalgh & Rosenbtatt, Reference Greenhalgh and Rosenbtatt1984; Davy, Kinicki, & Scheck, Reference Davy, Kinicki and Scheck1997; Kraimer, Wayne, Liden, & Sparrowe, Reference Kraimer, Wayne, Liden and Sparrowe2005). Specifically, when employers fulfill their employees’ expectations and make employees feel that their jobs are secure, the equal exchange relationships between employers and employees are established (Conway & Coyle-Shapiro, Reference Conway and Coyle-Shapiro2012; Colquitt, Baer, Long, & Halvorsen-Ganepola, Reference Colquitt, Baer, Long and Halvorsen-Ganepola2014; Lu, Du, Xu, & Zhang, Reference Lu, Du, Xu and Zhang2017).
Additionally, Greenhalgh and Rosenbtatt (Reference Greenhalgh and Rosenbtatt1984)’s job insecurity model offers another theoretical framework for understanding the relationship of job security and work attitudes. As we discussed, job security is termed as the legal employment done between the agency and the employee so that continued employment will be assured (Greenhalgh & Rosenbtatt, Reference Greenhalgh and Rosenbtatt1984: 439; Romzek, Reference Romzek1985: 283). Keeping in view the theoretical model for job insecurity, Greenhalgh and Rosenblatt claimed that job insecurity, which is a form of the individual's subjective threat, results from objective threat interpreted through the intuitive procedure of the individual. According to their argument, job insecurity is expected to have an impact on different individual employee behaviors like productivity, turnover or resistance to change. Based on their theoretical ground, it is believed by the researchers that the workers who are in permanent jobs (high level of objective job security) will have little impression of job insecurity (high level of apparent job security) than those workers who are employed on short-term contract (low level of objective job security). This increases the positivity of the employees’ work attitude as they are assured of having continued employment (Ashford, Lee, & Bobko, Reference Ashford, Lee and Bobko1989; Pearce, Reference Pearce, Cooper and Rousseau1998; De Witte, Reference De Witte1999; Klandermans, Hesselink, & Vuuren, Reference Klandermans, Hesselink and Vuuren2010). Studies that compare permanent and temporary workers report that permanent contracts were associated with higher levels of organizational commitment (Lee & Johson, Reference Lee and Johson1991; Morrow, McElroy, & Elliot, Reference Morrow, McElroy and Elliot1994; Martin & Hafer, Reference Martin and Hafer1995; Van Dyne & Ang, Reference Van Dyne and Ang1998) and higher job satisfaction (Hall & Gordon, Reference Hall and Gordon1973; Miller & Terborg, Reference Miller and Terborg1979; Krausez, Brandwoin, & Fox, Reference Krausez, Brandwoin, Fox and Freese1995). So the following hypothesis can be suggested:
Thus, employing social exchange theory and job insecurity theory, we expect that employees perceive that organizations and employers treat and make employees feel that their jobs are secure, they will respond with positive attitudes toward the job and organization. So the following hypothesis can be suggested:
Hypothesis 1
There is a positive relationship between job security and job satisfaction and organizational commitment.
Subgroup analysis for the job security and work attitudes
Apart from the study of the association of job security with respect to work-related attitudes (job satisfaction and organizational commitment), another objective of this study is to identify the subgroup analysis of these entire relationships. The variability of the findings in the research literature will be clarified by concentrating on five significant subgroups which have been left uninvestigated.
Type of organization
Organizational type is considered to be the first subgroup variable. Since public and private organizations differ institutionally, relationships with job security must be different. So the generalizable design between job security and work attitudes and behaviors could be affected depending upon the fact whether the type of organization is public or private. There are some studies which show that public employees do not give much importance to job security (Crewson, Reference Crewson1997). Others have the opinion that there is no difference among the employees belonging to the two sectors, regarding the needs of job security (Rainey, Reference Rainey1982; Wittmer, Reference Wittmer1991; Gabris & Simo, Reference Gabris and Simo1995). Moreover, the findings of some other studies show that public employees give higher importance to job security (Baldwin, Reference Baldwin1987; Jurkiewicz, Torn, & Roger, Reference Jurkiewicz, Torn and Roger1998; Houston, Reference Houston2000). In the author's view, the employees of public organizations give much more value to job security than that of the private organization employees. So, the hypothesis will be:
Hypothesis 2a
Job security will have a larger impact on job satisfaction in public organizations than it does in private organization.
Hypothesis 2b
Job security will have a larger impact on organizational commitment in public organizations than it does in private organization.
Country of origin
Differences in country of origin may be in systematic association with the relationship between job security and work-related attitude. Most (more than 60%) of studies are based on the current data from the USA, Canada, and Europe, while the remainder of the studies is from Africa, Middle East, South America, and Asian countries on this study. One of the reliable findings is that cultural values are helpful in studying the differences in variables that are related to work (Schwartz, Reference Schwartz1999). For example, people are inspired by Confucianism in the pursuit of long-term benefits and to withstand short-term loss (King & Bond, Reference King, Bond, Tseng and Wu1985). The role of country of origin is studied herein to better understand the relationship of job security and work-related attitudes. Here the hypothesis will be:
Hypothesis 3a
Job security will have a larger impact on job satisfaction in Asian, Africa, Middle East, and South America samples than it does in North America and Europe.
Hypothesis 3b
Job security will have a larger impact on organizational commitment in Asian, Africa, Middle East, and South America samples than it does in North America and Europe.
Age of employee
Age is related to aspects of career trajectory as well as the role of psychology in a person's life events. So, age has a connection to one's position in the organization as well as with those aspects which are beyond the area of work such as social life or family. According to the researchers’ opinion, older employees with the responsibility of family may be more thoughtful about their job security than younger employees (Kuhnert & Vance, Reference Kuhnert, Vance, Quick, Murphy and Hurrell1992; Finegold, Mohrman, & Spreitzer, Reference Finegold, Mohrman and Spreitzer2002). Moreover, the older employees are more distressed by the fear of losing their jobs than younger employees, since the younger ones have more mobility than the older ones in terms of occupation (Kuhnert & Vance, Reference Kuhnert, Vance, Quick, Murphy and Hurrell1992). Therefore, the hypothesis will be:
Hypothesis 4a
Job security will have a larger impact on job satisfaction in older employees than it does in younger employees.
Hypothesis 4b
Job security will have a larger impact on organizational commitment in older employees than it does in younger employees.
Tenure of employee
An aspect of the moderating impacts of tenure on the organizational commitment is that employees with limited term are not certain whether they will be able to perform up to the mark in a new job and so their level of commitment and motivation toward the organization increases (Wright & Bonett, Reference Wright and Bonett2002: 1184). Still, many studies have shown that there is positive relationship between the tenure of the employee and organizational commitment (Mathieu & Zajac, Reference Mathieu and Zajac1990; Abdullah & Ramay, Reference Abdullah and Ramay2012). The employees who have long terms with the organization are more committed than those employees having short terms (Greenhalgh & Rosenbtatt, Reference Greenhalgh and Rosenbtatt1984; Abdullah & Ramay, Reference Abdullah and Ramay2012). So, the hypothesis will be:
Hypothesis 5a
Job security will have a larger impact on job satisfaction in long tenured employees than it does in short tenured employees.
Hypothesis 5b
Job security will have a larger impact on organizational commitment in long tenured employees than it does in short tenured employees.
Gender
Difference of gender in job security has been found by several studies (De Witte, Reference De Witte1999; Rosenblatt, Talmud, & Ruvio, Reference Rosenblatt, Talmud and Ruvio1999). Some research findings have claimed that with respect to job security, male employees are more cautious than their female counterparts as the female employees remain more concerned with economic insecurity. However, in the opinion of Rosenblatt, Talmud, and Ruvio (Reference Rosenblatt, Talmud and Ruvio1999), as compared with women, men are less distressed by the fear of losing the job than women as generally speaking, men have greater mobility regarding their occupation than women. Specifically, their contention is that female employees are more concerned with their job security than male employees. So, in regards to these data, the hypothesis will be:
Hypothesis 6a
The relationship between job security and job satisfaction is dependent on the extent to which the employees are female and the relationship becomes more positive when more of the study sample is female.
Hypothesis 6b
The relationship between job security and organizational commitment is dependent on the extent to which the employees are female and the relationship becomes more positive when more of the study sample is female.
Methods
Summary of literature searches
A wide literature search has been conducted, which is based on computer-based findings and manual so that studies published (and unpublished studies) from 1980 to January 2014 can be identified. Investigating the literature on job security and employee work attitudes (i.e., job satisfaction and organizational commitment) more thoroughly and in order to avoid any preferences in the inclusion of the studies, a series of search strategies have been undertaken. At first, the Psych Info, Social Sciences Citation Index, and ABI/Inform databases were searched for the identification of studies on the relationship between job security (and job insecurity) and work attitudes (job satisfaction and organizational commitment). For the computer-oriented literature search, many keywords were used (for instance, we searched for the keywords ‘Job Security’ or ‘Job Insecurity’ in combination with the keywords ‘Job Satisfaction’ or ‘Organizational Commitment’).
Second, a manual article-by-article search was conducted throughout several management, organizational behavior, and HR journals. In that search, 32 journals from 1980 to 2014 were examined.
Third, attempts were made with respect to identification as well as for gaining access to those studies conducted in countries where English is not spoken but published in journals using English.
Regarding the three search approaches, a preliminary database consisting of 298 articles was set up for further examination. In the present meta-analysis of job security, studies should be in English and the subjective experiences of job security of the employees should be measured. Additionally, they have to submit reports of zero-order correlations between the job security (insecurity) of individual employees and criteria variables of interest (meaning job satisfaction and organizational commitment). After the application of these decision rules, the final database consisting of 37 articles reported effect sizes for 45 independent studies or samples totaling N of 27,871. A list of all the studies which have been used in the meta-analysis is marked by an asterisk in the Reference section.
Coding of studies
After selection of the studies, each individual study based on the outcome variable has been examined and the size of sample and the effect size coded. The effect size on the basis of zero-order correlations from each sample was also coded. If there was any effect size having an opposite orientation, it was recorded to reflect the proper direction of the effect (Pearson correlation between job insecurity and work attitudes).
Other variables were coded in order to examine the potential subgroups associated with the interest. First, in order to assess the type of organization, all of the studies were categorized according to organization type. Organization types were coded into one of the three sectors (i.e., public-nonprofit, private, and mixed-hybrid). The public-nonprofit organization consisted of central government, state/regional/local governmental bodies, and nonprofit agencies. The private organization consisted of manufacturing and service businesses. The mixed-hybrid organization consisted of education, health organization, or samples from multiple types of organization including both public (government sector) and private ones (industrial or service sector).
Second, studies were coded for the sample of origin. A great portion (more than 60%) of the studies is based on USA, Canada, and Europe, while the remainder of the studies is from Africa, Middle East, South America, and Asian countries. In each of the studies, the sample of origin was coded through their classification into one of three extensive samples of origin categories. So, the North America, Europe, and Asia, Africa, South American, and Middle East are included in the sample of origin category.
Third, Ng and Feldman (Reference Ng and Feldman2008) were followed when the chronological age of each sample was classified into one of the three groups. This was done on the basis of the distribution of mean ages for all study samples included in the meta-analysis. The distribution of mean ages spans: mean age less than 35 years; mean age between 35 and 40 years; and mean age over 40 years.
Fourth, to study the impact of tenure, each study was coded regarding the tenure of mean ranging from: mean tenure of less than 11 years; mean tenure between 11 and 14 years; and for over 14 years.
Lastly, each study was coded regarding the percentage of females in the study sample, and after that, the percentage of the study sample of females was classified into three groups. This was done on the basis of the percentage distribution of the female workers for all the study samples included in the meta-analysis. The distribution was: percentage of female workers less than 25%; percentage of female workers between 25 and 50%; and percentage over 50%.
Two coders independently coded all the relevant articles. To verify coding accuracy, the effect sizes (zero-order correlations of job security with job satisfaction, and commitment), sample sizes, reliability estimates of the outcomes, type of organization, sample of origin, age, tenure, and percentage of females were compared. The interrater agreement rate was high at 94%. We resolved all of the remaining discrepancies (mostly typographical errors, failing to reverse the sign of a correlation coefficient between job security and job satisfaction, and organizational commitment) by thoroughly double-checking the primary studies in question.
Analysis
The meta-analysis was conducted with Hunter and Schmidt (Reference Hunter and Schmidt1990 and Reference Hunter and Schmidt2004)’s formula (see Appendix 1). Following Hunter and Schmidt (Reference Hunter and Schmidt1990 and Reference Hunter and Schmidt2004), we corrected correlations and their variances for sampling error and for measurement errors in predictors (see Appendix 1) (detailed calculation procedure described on Appendix 1 & 2 section). The correlation coefficients between job security and work-related attitudes (job satisfaction and organizational commitment) are the two effect sizes of interest.
The following procedures were employed: (1) information on three distributions (observed correlations, consistency [reliability] of the independent variable and dependent variable) have been assembled by the studies; (2) correlations of work attitudes (job satisfaction and organizational commitment) derived from one study that referred to the same category were aggregated since average correlations do not violate the assumption of independence (Hunter & Schmidt, Reference Hunter and Schmidt2004; Matthijs Bal, De Lange, Jansen, & Van Der Velde, Reference Matthijs Bal, De Lange, Jansen and Van Der Velde2008); (3) each correlation was corrected for the statistical artifact of sampling error (Kooij, Jansen, Dikkers, & Lange, Reference Kooij, Jansen, Dikkers and Lange2010: 1119; Taft, Watkins, Stafford, Street, & Monson, Reference Taft, Watkins, Stafford, Street and Monson2011: 25); (4) this corrected distribution is further corrected for the other available artifacts (e.g., measurement error). Thus the variance due to sampling error and other artifacts is subtracted out, and what is left is an estimate of the population variance (see Taft et al., Reference Taft, Watkins, Stafford, Street and Monson2011: 25). (5) For the interpretation of the validity generalization results, confidence interval (see Cohen, Reference Cohen1993) was used, meaning the confidence interval of a significant mean correlation does not include zero; (6) regarding the guidelines proposed by Cohen (Reference Cohen1992, Reference Cohen1998), the magnitude of the mean correlation was interpreted; with .1 interpreted as having small effect; .3 as medium effect; and .5 as large effect (Kooij et al., Reference Kooij, Jansen, Dikkers and Lange2010: 208).
For assessing the different effect sizes, various approaches were used. First, the percentage of variance in the effect sizes explained by the statistical artifacts (sampling error and criterion unreliability) is reported. The 75% rule is offered as a rule of thumb by Hunter and Schmidt (Reference Hunter and Schmidt2004) for identifying the potential of moderated relationships. That is, if 75% or more of the variance across studies is explained by sampling error and measurement unreliability between samples, it is likely the remaining 25% stems from uncontrolled artifacts. If we account for 75% of the variance or less, or the corrected variance is still large, then the next step is to test for the influence of moderator variables (Hunter & Schmidt, Reference Hunter and Schmidt2004: 145–146). It is obvious that current review corrected only the sampling error and predictor/criterion unreliability. Second, for assessing the accuracy and distribution of effect sizes estimates (Whitener, Reference Whitener1990), confidence interval as well as credibility intervals were calculated. Third, for testing homogeneity of corrected correlations, Q-statistics were calculated. Finally, the subset method proposed by Hunter and Schmidt (p. 293) was followed to assess whether coded variables actually moderated job security–work attitudes (job satisfaction and organizational commitment) relationships. This method entails assigning studies to different subsets and performing a separate meta-analysis within each subset.
Description of the selected studies
Twelve of the selected studies were conducted among the employees of public organizations, whereas employees in private organizations were examined by 19. A sample of North American origin was examined by 14 studies, and 19 studies examined samples having European origin. Nine samples having Middle East, South American, Asian, and African origin were also examined. From the various studies, the mean ages ranged from 27.5 to 48 years. The average tenure of the various studies had a range between 1.72 and 21 years. In Table 1, sample characteristics, sample size, average age, mean tenure, proportion of female employee, reliability of job security, and the relationships between job security and work attitudes (job satisfaction and organizational commitment) and their reliabilities are listed. The correlations between job security and job satisfaction are shown to range from −.08 to .54 and correlation with organizational commitment from −.09 to .97. So, it is expected that specific characteristics of the studies moderate the relationships (Hunter & Schmidt, Reference Hunter and Schmidt2004). The measurement of reliabilities of job security ranged between .68 and .97 and the reliability of job satisfaction ranged between .64 and .96 and organizational commitment is between .6 and .91.
Sample characteristics, organizational type, average age and tenure, percentage of female employee, sample size, reliability of job security, and Pearson correlations of job security with outcomes (job satisfaction and organizational commitment), and reliabilities of the outcomes
Results
Table 2 includes the number of outcomes (k) and participants in each analysis (N), weighted (uncorrected and corrected) mean random-effect sizes (mean r and ρ), the 95% confidence interval (CI), three homogeneity statistics; the percentage of variance accounted for by sampling and measurement error, the 95% credibility interval, and Q-statistic. First, we considered Hypothesis 1, which predicted a positive relationship between job security and work attitudes, by examining the results of overall analysis (Table 2). The results of overall analysis suggest there is a positive relationship between job security and job satisfaction (true score correlation ρ = .32; 95% CI .260–.285) and organizational commitmentFootnote 1 (true score correlation ρ = .25; 95% CI .193–.220). Using Cohen's (Reference Cohen1998) framework, the two effect sizes are in the medium range of magnitude.
Meta-analytic results of relationship between job security and job satisfaction and organizational commitment
Note. k = the number of studies; samples; N = the number of individuals in the k samples; 
$\bar{r}$ = sample-size-weighted uncorrected correlation; ρ = mean true score correlation; SD ρ = standard deviation of true score correlation; 95% confidence Interval = 95% confidence interval for ρ; Var.expl. = percentage of variance in corrected correlations attributable to all the artifacts considered; Credibility Interval = 95% credibility.
*p < .05.
Further, three procedures tested whether unknown moderators exist. We followed Griffeth, Hom, and Gaertner's procedure (Griffeth, Hom, & Gaertner, Reference Griffeth, Hom and Gaertner2000). First, we estimated the degree to which two prime statistical artifacts (sampling error and scale unreliability) can account for between-study variance in observed correlations. In accord with Hunter and Schmidt's rule (Hunter & Schmidt, Reference Hunter and Schmidt1990 & Reference Hunter and Schmidt2004), we concluded that moderators exist if these artifacts explain less than 75% of the observed variance in correlations. Second, to test whether moderators exist, we computed the Q homogeneity statistic (Hunter & Schmidt, Reference Hunter and Schmidt1990: 168). Third, we calculated the 95% credibility interval (Whitener, Reference Whitener1990). Meta-analysts typically infer moderators when the credibility interval is either large or includes zero. Yet a credibility interval may include zero if the actual effect size is nearly zero (i.e., when there is hardly any effect to moderate), while the precise meaning of a ‘large’ credibility interval is unclear (Sage & Koslowsky, Reference Sage and Koslowsky1993). Due to the ambiguity of these interpretations, we interpreted the credibility interval only if the other two tests disagree about a moderator's presence. As such, we followed Sagie and Koslowsky's prescription (Sage & Koslowsky, Reference Sage and Koslowsky1993) to emphasize the 75% rule and the Q-statistic when probing for moderators (Griffeth, Hom, & Gaertner, Reference Griffeth, Hom and Gaertner2000: 477).
Subgroup analyses
On the whole, percentages of variance accounted for by artifacts (e.g., sampling error) were extremely low, ranging between 6.89 and 5.73% (Table 2); and in accord with Hunter and Schmidt's rule (Hunter & Schmidt, Reference Hunter and Schmidt1990 & Reference Hunter and Schmidt2004), we concluded that moderators exist if these artifacts explain less than 75% of the observed variance in correlations. In addition, we also computed the Q homogeneity statistic, which yields a significant χ2 when moderators exist (Hunter & Schmidt, Reference Hunter and Schmidt1990: 168). The value obtained by the χ2 test (Q = 536.6 and Q = 525.4, respectively) are larger than the value required for statistical significance (Table 2). Thus, current test suggests that the unexplained variance is significantly greater than zero, suggesting that the remaining variance is due to additional moderators or statistical artifactsFootnote 2 (Arthur, Bennett, & Huffcutt, Reference Arthur, Bennett and Huffcutt2001: 92). Hence, it was necessary to investigate if the type of organization, sample origin, age, tenure, proportion of women in moderation of relationships between job security and work attitudes (organizational commitment and job satisfaction) were significantFootnote 3.
In the next section, whether certain conditions have an influence on the effect size of job security and work attitudes (i.e., job satisfaction and organizational commitment) was examined through subgroup analyses.
Subgroup analyses for the relationship between job security and job satisfaction
Table 3 shows the results regarding the subgroup analyses for the relationship between job security and job satisfaction. Subgroup analyses showed a significant association between the effect size of job security and job satisfaction and different subgroups (i.e., organizational type, country of origin, labor age, tenure, and female employee percentage).
Meta-analytic results of subgroup analyses for the relationship between job security and job satisfaction
Note. k = the number of studies; samples; N = the number of individuals in the k samples; 
$\bar{r}$ = sample-size-weighted uncorrected correlation; ρ=mean true score correlation; SD ρ = standard deviation of true score correlation; 95% confidence Interval = 95% confidence interval for ρ; Var.expl.= percentage of variance in corrected correlations attributable to all the artifacts considered; Credibility Interval = 95% credibility interval for ρ; Q-statistic = Chi-square test for moderators.
*p < .05.
For the organizational type subgroup analyses, it was predicted in Hypothesis 2a that job security would be a larger impact on job satisfaction in public organizations than it does in private organization. However, there was failure in finding support for Hypothesis 2a. Rather, job security was more positively related to job satisfaction among employees of private sectors (ρ = .372; 95% CI .299–.329) in relation to those of public organizations (ρ = .188; 95% CI .119–.192).
Hypothesis 3a explained that when the origin country is in Asia and Africa, the relation between job security and satisfaction is significantly positive. There is not supportive evidence of Hypothesis 3a as a positive relation between job satisfaction and job security among Europe (ρ = .335; 95% CI .268–.299) larger than North America (ρ = .286; 95% CI .214–.259) and Asian, Middle East, South America, African nationalities (ρ = .209; 95% CI .126–.209).
Subgroup analyses based on the age group of labor also showed significant relationships between the effect size and age group of labor. The results were highly positive especially among the employees within the age group of 35 years less (ρ = .272; 95% CI .197–.255), as compared with older aged employees that showed a lower positivity (ρ = .251; 95% CI .188–.239). Here the relationship between job security and job satisfaction was higher among younger workers (ρ = .272; 95% CI 197–.255) in comparison to samples of workers from the middle age group (ρ = .256; 95% CI .187–.244) versus older workers (ρ = .251; 95% CI .188–.239). Hence, there was no support for Hypothesis 4a.
Tenure-based subgroup analyses also proved that tenure constituted a significant relationship between the effect size and tenure. This proves that effects would be stronger in cases where the tenure ranges 11 years less and between 11 and 14 years. Job security effects were the strongest in short tenure levels (11 years less) (ρ = .371; 95% CI .221–.299). Longer tenure showed weaker effects (ρ = .208; 95% CI .147–.214) and when the tenure was medium level (11–14 years), it was moderate effect (ρ = .218; 95% CI .151–.201). In sum, there was no support for Hypothesis 5a.
When subgroup analyses are based on percentage of women workers, the samples show that female employees also showed significant relationships between the effect size and percentage of women workers. According to Hypothesis 6a, the relation is better if there are more numbers of female workers. It was not found in support of Hypothesis 6a that the percentage of female workers is 25% less, there are more positive relations (ρ = .306; 95% CI .213–.299) as compared with more than 50% of female workers (ρ = .225; 95% CI .168–.213).
Subgroup analyses for relationship between job security and organizational commitment
Table 4 shows the results regarding the subgroup analyses for the relationship between job security and organizational commitment. Subgroup analyses results (Table 4) showed a significant association between the effect size of job security and organizational commitment and different subgroups (i.e., organizational type, country of origin, labor age, tenure, and female employee percentage).
Meta-analytic results of subgroup analyses for the relationship between job security and organizational commitment
Note. k = the number of studies; samples; N = the number of individuals in the k samples; 
$\bar{r}$ = sample-size-weighted uncorrected correlation; ρ = mean true score correlation; SD ρ = standard deviation of true score correlation; 95% confidence Interval = 95% confidence interval for ρ; Var.expl. = percentage of variance in corrected correlations attributable to all the artifacts considered; Credibility Interval = 95% credibility interval for ρ; Q-statistic = Chi-square test for moderators.
*p < .05.
For the organizational type subgroup analyses, according to Hypothesis 2b, the relationship between job security and organizational commitment would be higher in public organization as compared with private organization. As a support of Hypothesis 2b, it was found that job security was more positively related to organizational commitment among employees of public-nonprofit sector (ρ = .264; 95% CI .198–.247) in comparison to private organization employees (ρ = .196; 95% CI .142–.185). Apart from this, those working in mixed organizations (hybrid) were more related to job security and showed the strongest effects (ρ = .319; 95% CI .226–.275).
For the employee's country origin, Hypothesis 3b proved that when the employees’ country of origin is Asia or Africa, the relationship job security is more positively related to organizational commitment. Hence there is supportive evidence of Hypothesis 3b as a more positive relation between job security and organizational commitment among Asia, Middle East, and African sample (ρ = .385; 95% CI .268–.340) as compared with Europe (ρ = .204; 95% CI .152–.186) and the USA (ρ = .327; 95% CI .240–.297).
Age group of labor-based subgroup analyses also showed significant relationships between the effect size and age group of labor. The results were highly positive especially among the employees within the age group of 35 years less (ρ = .277; 95% CI .202–.247), as compared with older aged employees that showed a lower positivity (ρ = .199; 95% CI .144–.200). Here the relationship between job security and organizational commitment was higher among younger workers (ρ = .277; 95% CI .202–.247) in comparison to samples of workers from the middle age group (ρ = .273; 95% CI .190–.251) versus older workers (ρ = .199; 95% CI .144–.200). Hence, there was no support for Hypothesis 4a.
Subgroup analyses that are based on tenure also proved that tenure constituted a significant relationship between the effect size and tenure. But there was no support for Hypothesis 5b. The positive results of organizational commitment and job security were the highest for 11 years less (ρ = .420; 95% CI .322–.390) but was much lower in tenures between 11 and 14 years (ρ = .165; 95% CI .095–.170). The weakest effects were observed in tenures above more than 14 years (ρ = .111; 95% CI .064–.127). In sum, there was no support for Hypothesis 5b.
When subgroup analyses are based on percentage of women workers, the samples show that female employees also showed significant relationships between the effect size and percentage of women workers. Hypothesis 6b proved that if women workers were larger, there is a more positive effect. But support for Hypothesis 6b was not available. In sum, the relationship between job security and organizational commitment becomes weaker when less than 25% of all employees consist of female workers (ρ = .256; 95% CI .185–.238) and when it is more than 50% of women comprising the employee group (ρ = .163; 95% CI .107–.156). When female workers comprise 25–50% of the workforce, the strongest effects are visible on organizational commitment (ρ = .318; 95% CI .233–.288).
Discussion
Taking an overall approach, it was found that the medium-sized associations between job security and each work attitude variables (i.e., job satisfaction and organizational commitment) were found, with true score correlations (ρ) of .327 for job satisfaction, and .253 for organizational commitment. These results highlight the significance of job security at the workplace, in shaping and enhancing attitudes of employee. Job security is perhaps the sole reason why work environment and employee benefits are given so much importance. However, job security issues have different patterns in different situations, which future studies will show.
The present study aims at detecting stabilizing factors (subgroups) of these overall relationships, by focusing on five important subgroups – organizational type, employee nationality, age, tenure, and gender.
We have analyzed the relationship between job security and work attitudes in both organizational types. Interestingly, there were different patterns for the organizational type for both work behavior relationships, i.e., job satisfaction and organizational commitment. Specifically, there were higher associations in private organization for the job satisfaction (relative to public organization sample). It appears as if employees who work in the private organization tend to have stronger reactions to job security. That is, they were more likely to job satisfaction (compared with public organization workers). While the finding that higher associations were obtained from public organization for the organizational commitment (relative to private organization sample), this implies that workers in public organization may be more sensitive to job security than employees in private organization, they were more likely to organizational commitment. The meta-analysis actually shows that both private and public employees might crave lesser amount of job insecurity (high job security) even when jobs are, by rule, highly secure at public sector.
The findings of subgroup analyses that are based on country of origin are interesting. Differential findings were obtained for the country of origin type subgroup analyses. First, Non-Asia-, Africa-, and South America-based employees have a greater dependence on the job security variable for the relationship with job satisfaction. It means that Europe and North America employees show higher associations to work attitudes concerning job security with satisfaction factored in as compared with the Asia, Africa, and South America instances. Second, it appears as if it appears as if employees not based in the USA tend to have stronger reactions to job security for the relationship with organizational commitment. As predicted, for the relationship job security and organizational commitment, higher associations were found for Asia and Africa samples (relative to North America and Europe samples). This could be the result of cultural differences in the whole conception and response toward the phenomenon of job security. It stems from here albeit more thorough and extensive research is required to understand cultural and ethnic connotations from a differential point of view. It would appear that further work is needed to understand cultural factors in this association.
Interestingly, differential patterns were also found in terms of age, tenure, and proportion of female workers subgroup analyses. The meta-analysis results show that age is highly related to the job security and employee work behavior, i.e., job satisfaction and organizational commitment. We had assumed that with older age, more cravings for job security would emerge. But this showed a divergent pattern in terms of employee age for the relationship between job security and job satisfaction and organizational commitment. This correlation was more directly proportional for younger employees, putting forth the fact that older workers go by the responsibility of maintaining a relationship with organization, i.e., organizational commitment, in terms of a relatively inadequate job holding for short period (short tenure), hence they are less affected by job security (Allen & Meyer, Reference Allen and Meyer1990; Löckenhoff & Carstensen, Reference Löckenhoff and Carstensen2004).
Contrary to our hypothesis, the meta-analysis results also show that differential patterns were found in terms of employee tenure subgroup analyses. It is expected that with long tenured employee has a more positive pattern with respect to job security and work attitudes (i.e., job satisfaction and organizational commitment), but this was not always so. It was moderately large for new employees (short tenure) and rapidly diminished with passage of time. We found that correlation between job security and organizational commitment was more positive for new employees and rapidly declines and level off with increasing time. This implies that this correlation was more positive for short-level tenured employee group than for long tenured employee group. This positive moderation of tenure in the job security, job satisfaction, and organizational commitment relations corroborates the expectation that employees with higher levels of tenure who oftentimes feel certain about their job security will decrease their level of job satisfaction and commitment to the organization.
With respect to proportion of female workers, the effect size between the subgroups of proportion of female workers was significant different. However, contrary to our hypothesis, we found that the relationship between job security and work attitude variables (job satisfaction and organizational commitment) is more positive when female works comprise 25–50% of the workforce compared with less than at low (less than 25%) or high levels (more than 50%) of proportion of female workers. It shows that female workers occupy a median position in terms of the relationship between job security and job satisfaction and organizational commitment. Hence, these results show that an inverted U-shaped curve reflects that job satisfaction and organizational commitment slows down at low or high proportions of female workers. In the next section, theoretical and practical implications of these findings for policy and for research will be discussed.
Theoretical and practical implications
The results of current meta-analysis will be helpful for resolving the current theoretical debate concerning job security and work attitudes relationship (Byron, Khazanchi, & Nazarian, Reference Byron, Khazanchi and Nazarian2010). In sum, it has been found that there is a linearly positive relationship between job security and work attitudes (job satisfaction and organizational commitment). Hence, job security does increase employee job satisfaction and organizational commitment, but this is a simplistic approach. Employee age and employee tenure are more compelling decisive factors. In these cases, subgroup analyses for the relationship between job security, job satisfaction, and organizational commitment by employee age and tenure group revealed a diminishing pattern of linear relationship. Job satisfaction increases with youth (and short tenure) and the opposite is true for older workers (long tenure). The same is that what we found for organizational commitment. Our findings also have other broader theoretical implications. We found that job security increased organizational commitment for Asia, Middle East, South American, and Africa samples (relative to North American and Europe samples). This finding here suggests that cultural differences occupy a huge position to analyze these relationship patterns. Job security response and reaction varies widely from nation to nation. Interpretations are quite multifarious. In addition, we found that organizational type (private sector vs. public sector) of the employees included in the sample acted as significant subgroup factors. This finding shows that models of job security and work attitudes also incorporate different type of organization (sectoral difference) that account for how each type of organization (different sector) differentially interpret and react to their job security.
The current meta-analysis findings call attention to several important considerations important for developing effective public job security policy.
Practically, the current meta-analysis shows that the relationship of job security to employee work attitudes is positive and increasing rather than curvilinear. Thus, job security is worth retaining in some form, contrary to the logic of at-will employment. Many organizations aim to increase employee work attitudes, and our findings indicate that possible organizational interventions such as at-will employment policy may decrease employee work attitudes and individual performance. This research results suggest that policy makers seeking to increase employee work attitudes and behavior should ensure that their employees feel a sense of job security in their current job. Policy makers may increase job satisfaction and organizational commitment by enriching the emotive content of a job. Further, in the job security policy case, organizational decision makers may presume that a fully guaranteed tenure would maximize outcomes in terms of employee work attitude. However, the results point out that at-will employment policy might lead to decreased instead of increased employee work attitudes. Instead, job satisfaction and organizational commitment were more positive for new and medium-term employees, but rapidly decline and level off over time. This suggests that granting high job security for the short- and medium-term (but not long-term) may be a mean to maximize employee job satisfaction and organizational commitment. Lesser yet meaningful job tenure rules could be attained via such steps as short- to medium-term employment (e.g., 5–15 years) contracts. Thus, through this study, practitioners in public organization limit the application of fully-guaranteed tenure policy based on understanding of the tenure subgroup analysis, they may avoid such unexpected negative outcomes. Lastly, the results suggest that policy may consider selectively adding some subgroup factors to the organizational environments, such as those moderate level of proportion of female workers, in order to improve their employee work attitudes.
Limitations and future research
Despite the strengths of this meta-analysis, some limitations call for attention. First, future research should focus on the large number of different job security questionnaire type studies in the field. One of the most apparent differences among studies examining the relation between job security and work attitudes (i.e., job satisfaction and organizational commitment) may be in the questionnaires that have been used to assess employee perception of their job security (Schwinger, Wirthwein, Lemmer, & Steinmayr, Reference Schwinger, Wirthwein, Lemmer and Steinmayr2014: 746). There were differences in their operationalization of job security (or insecurity) in the studies included in this analysis. Thus, future researchers can hypothesize that choosing different job security (or insecurity) questionnaire can be responsible for dissimilar correlations between job security and work attitudes. In addition, future research should consider the construct validity of the available questionnaires in order to gain a precise measure of job security (or insecurity) effects on work attitudes using additional meta-analysis. Thus, we advise future researchers to conduct meta-analytic investigations and to examine relevance of potential moderator variables on the relation between job security and employee work attitudes and behavior. Since subgroup analyses are observational, another issue in interpreting the results is the difficulty to detect confounders, which can have a moderated effect, such as the type of job security measurement scale and construct validity of the available job security questionnaire. Thus, future research could perform either between-group Q-statistics or meta-regression to further examine their moderator variables and hypotheses. Second, as with other meta-analysis, there is a mismatching of different methods applied in the different studies undertaken (e.g. there exists considerable heterogeneity in how job security can be enhanced in terms of variables, factors, and conditions). We advise control conditions and a detailed explanation as to how these are manipulated or applied in the future research. Third, industry effects (e.g., health care and education) were not tested in this meta-analysis. Health care and education may produce far different results than general government, if we could assess effects in these industry sectors. This also points to important demands for future research. Future research will bring in more mediators and moderators to determine under what conditions and situations these variables can be applied.
Conclusion
The objective of this article is to synthesize public and private sector accumulated research regarding the relationship between job security and employee work attitudes (i.e., job satisfaction and organizational commitment) to provide a more evidence-based footing for future public job security policy by conducting a quantitative meta-analysis method. A current meta-analysis highlights the significance of job security in enhancing employee attitudes and job performance and the relationship of job security to employee work attitudes is positive. Thus, job security is worth retaining in some form, contrary to the logic of at-will employment. We believe current meta-analysis research findings will give us an opportunity to support more debate and administratively-rationale alternatives to current job security system and guide future public and organizational policy making.
Author ORCIDs
Hyunkang Hur, 0000-0001-8805-1731.
Appendix 1
Following Hunter and Schmidt's (Reference Hunter and Schmidt1990 and Reference Hunter and Schmidt2004) procedure, we corrected correlations (and their variances) for sampling and measurement errors, the largest sources of spurious between-study variation (Hom, Griffeth, & Carson, Reference Hom, Griffeth, Carson, Rabin, Vocino, Hildreth and Miller1995: 532 Hunter & Schmidt, Reference Hunter and Schmidt2004: 81). To correct sampling error, we first averaged correlations between job security and job satisfaction, and organizational commitment, weighting by sample size. If the population correlation is assumed to be constant over studies, then the best estimate of that correlation is not the simple mean across studies but a weighted average in which each correlation is weighted by the number of persons in that study (Hunter & Schmidt, Reference Hunter and Schmidt2004: 81). Thus, sample-size-weighted mean uncorrected correlation is
$$\bar r = \displaystyle{{\mathop \sum \nolimits [ { N_i{^\ast}r_i} ] } \over {\mathop \sum \nolimits^ N_i}}$$where r i is the correlation in study i and N i is the number of persons in study i.
Next, we adjusted this correlation for unreliability by inserting the job security, job satisfaction, and organizational commitment reliability coefficients (averaged across different samples) in to the classic attenuation correction formula (Hunter & Schmidt, Reference Hunter and Schmidt1990: 119; Hunter & Schmidt, Reference Hunter and Schmidt2004: 96).
$$\rho _{TU} = \displaystyle{{\rho _{xy}} \over {\sqrt {r_{xx}} \sqrt {r_{yy}}}} $$Thus, the estimation of adjusting correlation for unreliability in a Hunter and Schmidt meta-analysis involves two steps. To correct for the artifacts, we first compute the mean compound artifact attenuation factor:
$$\bar A = {\rm Ave}( a ) {\rm Ave}( b ) = \bar a\bar b$$where
$$a = \sqrt {R_{xx}} \;, \; \; b = \sqrt {R_{yy}} $$Second, from this we compute the mean actual study correlation (formulas for these procedures are provided in Hunter & Schmidt, Reference Hunter and Schmidt1990: 175–176; Hunter & Schmidt, Reference Hunter and Schmidt2004: 151–152)
$$\rho = {\rm Ave}( {\rho_i} ) = \displaystyle{{{\rm Ave}( r ) } \over {\bar A}}$$ Next, the confidence interval using the standard error for the mean effect size for homogeneous studies would be generated around the sample-size-weighted mean effect size to estimate accuracy of the estimate of the mean effect size (Whitener, Reference Whitener1990: 317). The confidence interval generated using the standard error reflects the effects of sampling error and is therefore applied to sample-size-weighted mean effect sizes (=
$\bar r$) that have not been corrected for research artifact (Whitener, Reference Whitener1990: 316). Thus, the confidence interval is generated using the standard error for the mean correlation and is applied to the mean sample-weighted r before the correction of attenuating artifacts (Arthur, Bennett, & Huffcutt, Reference Arthur, Bennett and Huffcutt2001). The estimation of confidence intervals in a Hunter and Schmidt meta-analysis involves two steps. First, the sample-size-weighted mean effect size that has not been corrected for research artifact and standard error of the mean effect size are calculated. Second, confidence intervals are generated around the uncorrected, sample-size-weighted mean effect size using the standard error of the mean effect size (Whitener, Reference Whitener1990: 316). Schmidt, Hunter, Outerbridge, and Goff (Reference Schmidt, Hunter, Outerbridge and Goff1988: 668) provided, and Hunter and Schmidt (Reference Hunter and Schmidt1990 and Reference Hunter and Schmidt2004) derived, the formula for the standard error of the mean correlation for homogeneous studies (Whitener, Reference Whitener1990: 316)
$${\rm SE} = (1 - \bar r^2)/(N - K)^{1/2}$$where 
$\bar r$ is the sample-size-weighted mean uncorrected correlation, N is the total sample size, and K is the number of studies. Thus, the 95% CI generated using the standard error reflects the effects of sampling error is
$$\bar r \pm 1.96{^\ast}{\rm SE}$$Arthur, Bennett, and Huffcutt (Reference Arthur, Bennett and Huffcutt2001) noted that a relatively small confidence interval band that does not include zero is considered a favorable outcome. It indicated that relatively little sampling error remains in the sample-size-weighted mean correlation (Arthur, Bennett, and Huffcutt, Reference Arthur, Bennett and Huffcutt2001: 73).
Meta-Analytic Results of Organizational Commitment (OC) Questionnaire Type Subgroups Analysis between job security and organizational commitment
In addition, three procedures tested the ‘true’ generality of each job security, job satisfaction, and organizational commitment correlation. These tests estimated nonartifactual variation of this correlation, detecting whether or not (unknown) moderators condition these correlation (job security–job satisfaction, job security–organizational commitment). First moderator test assessed the degree to which statistical artifacts explain variance in observed correlations (Hunter & Schmidt, Reference Hunter and Schmidt1990 and Reference Hunter and Schmidt2004). Because current meta-analysis only corrected for sampling error and unreliability, we defined 75% (or more) artifactual contribution as signifying no moderators. However, if we account for 75% of the variance or less, or the corrected variance is still large, then the next step is to test for the influence of moderator variables (Hom, Griffeth, & Carson, Reference Hom, Griffeth, Carson, Rabin, Vocino, Hildreth and Miller1995: 533; Arthur, Bennett, & Huffcutt, Reference Arthur, Bennett and Huffcutt2001: 59–67; Hunter & Schmidt, Reference Hunter and Schmidt2004: 145–146). Thus, formula for the calculation for determining the total percent of the observed variance accounted for by sampling error and attenuating artifacts is (Hunter & Schmidt, Reference Hunter and Schmidt1990: 176; Arthur, Bennett, & Huffcutt, Reference Arthur, Bennett and Huffcutt2001: 83; Hunter & Schmidt, Reference Hunter and Schmidt2004: 152)
$$100 \times \left[ {\displaystyle{{({\rm VAR}( e ) + {\rm Var}({\rm AV})} \over {{\rm Var}(r)}}} \right]$$where
(1) Var (e) = sampling error variance =
$\sigma _e^2 = (1 - \bar r^2)^2/\bar N - 1$, where $\bar N$ = T/K and K is the number of studies and T = $\mathop \sum \nolimits^ N_i$ is the total sample size(2) Var(AV) = Variance due to the attenuating artifacts =
$\bar \rho ^2\bar A^2V$, where V (sum of the squared coefficients of variation) = $V1 + V2 = \displaystyle{{{\rm Var}(a)} \over {{[ {{\rm Ave}(a)} ] }^2}} + \displaystyle{{{\rm Var}(b)} \over {{[ {{\rm Ave}(b)} ] }^2}}$(3) Var(r) = Variance of study coefficients
$ = \sigma _r^2 = \mathop \sum \nolimits (Ni*(r^2_{i - \bar r})} )/\mathop \sum \nolimits^ N_i$
Second, we computed 95% credibility intervals (using variances fully corrected for attenuating artifacts) around ‘true’ population correlations (Whitener, Reference Whitener1990; Hom, Griffeth, & Carson, Reference Hom, Griffeth, Carson, Rabin, Vocino, Hildreth and Miller1995: 533; Arthur, Bennett, & Huffcutt, Reference Arthur, Bennett and Huffcutt2001: 87–89). If credibility intervals including zero signal moderators and suggest moderators are probably operating and the corrected mean correlation size is really the mean effect of several population (Arthur, Bennett, & Huffcutt, Reference Arthur, Bennett and Huffcutt2001: 89), there may be more than one population involved and possible subpopulations (moderator analyses) should be investigated (Arthur, Bennett, & Huffcutt, Reference Arthur, Bennett and Huffcutt2001: 87 and 89). In addition, unlike confidence intervals, which are generated using the standard error of the mean correlation before statistical artifacts are corrected for, credibility intervals are generated after the mean correlation has been corrected for attenuating artifacts (Arthur, Bennett, & Huffcutt, Reference Arthur, Bennett and Huffcutt2001: 87). The estimation of credibility intervals in a current meta-analysis involves two steps. First, the mean actual study correlation that has been corrected for research artifact and corrected standard deviation around the mean of the distribution of true correlation are calculated. Second, credibility intervals are generated around the corrected actual study correlation using the corrected standard deviation around the mean of the distribution of true correlation (Arthur, Bennett, & Huffcutt, Reference Arthur, Bennett and Huffcutt2001: 87). Hunter and Schmidt (Reference Hunter and Schmidt1990 and Reference Hunter and Schmidt2004) derived the formula for the standard deviation around the mean of the distribution of true correlation (Arthur, Bennett, & Huffcutt, Reference Arthur, Bennett and Huffcutt2001: 87; Hunter & Schmidt, Reference Hunter and Schmidt2004: 151–152)
$${\rm Variance\ of\ actual\ attenuated\ correlations} = {\rm Var}( \rho ) = [{\rm Var}( {\rho_0} ) - \bar \rho ^2\bar A^2V]/\bar A^2$$where
Variance of attenuated study population correlation = Var(ρ 0) = Var(r) − Var(e).
Variance of study coefficients = Var(r) = 
$\sigma _r^2 = \mathop \sum \nolimits (Ni*(r^2_{i - \bar r})) /\mathop \sum \nolimits^ N_i$
Sampling error variance = 
${\rm VAR}( e ) = \sigma _e^2 = (1 - \bar r^2)^2/\bar N - 1$
$\bar \rho ^2\bar A^2V$ = variance due to artifact variation.
Thus, the 95% credibility interval generated by using the corrected standard deviation around the mean of the distribution of true correlation is
$$\rho \pm 1.96{^\ast}\sqrt {Var( \rho ) } $$ Finally, the χ2 tests the null hypothesis that there is no real variance in unattenuated correlations; that all of the observed variance is due to variation in artifacts and to sampling error (Hunter & Schmidt, Reference Hunter and Schmidt1990: 168). If the value obtained by the χ2 test is larger than the value required for statistical significance, the test suggests that the unexplained variance is significantly greater than zero, suggesting that the remaining variance is due to additional moderators or statistical artifacts (Arthur, Bennett, & Huffcutt, Reference Arthur, Bennett and Huffcutt2001: 92). Hunter and Schmidt (Reference Hunter and Schmidt1990: 168) derived the formula test statistic Q is 
$\hat S^2$
$$Q = \left[ {\displaystyle{{K{^\ast}{\rm Var}(r)} \over {{\hat S}^2}}} \right]$$
$$Q = \left[ {\displaystyle{{K{^\ast}{\rm Var}(r)} \over {(\rho^2{^\ast}{\rm A}{\rm A}^2{^\ast}V) + {\rm Var}(e)}}} \right]$$where k = number of studies (correlations); Var(r) = variance of the sample-weighted rs; ρ = rho, AA = compound attenuation factor, V = compound variance factor; Var(e) = sampling error variance. So under the null hypothesis, Q will approximate a χ2 distribution with k−1 degrees of freedom.
Heterogeneity test results of relationship between job security and job satisfaction and organizational commitment
Appendix 2
Hunter and Schmidt method (also called psychometric meta-analysis method) is particularly useful when researcher assume that there are some methodological flaws in the underlying studies (i.e., finite sample size and variations in measurement errors). Some job security research lacked large enough sample sizes to yield reliable results. In addition, since the measuring instruments in job security research are subject to measurement error (imperfect reliability), they produce effect size estimates that are made smaller (attenuated) by this measurement error (Borenstein, Hedges, Higgins, & Rothestein, Reference Borenstein, Hedges, Higgins and Rothestein2009: 341). Using Hunter and Schmidt's psychometric meta-analysis method, researcher can test that the variation in job security predictions across situations for the same job situations could be largely accounted for by study imperfections such as sampling error, and variations in measurement errors (Le et al., Reference Le, Oh, Shaffer and Schmidt2007: 8). Thus, we can estimate what the effect would be if there were no methodological limitations. In addition, this psychometric meta-analysis method corrects for sample limitations, results tend to be larger than using Fisher's Z. This is not a form of bias, per se, unless, researcher should be aware of this limitation.
