Cognitive and Affective Job Insecurity: A Meta-Analysis and a Primary Study

Outcomes of JI

COR theory (Hobfoll, 1989) proposes that individuals are motivated to obtain and conserve resources. According to COR theory, psychological stress occurs under three conditions: (1) resources are lost, (2) resources are threatened with loss, and (3) there is no resource gain after resource investment. COR theory categorizes resources as objects (e.g., housing), personal characteristics (e.g., optimism), conditions (e.g., seniority), or energies (e.g., money) that are valued in their own right or serve as a means for attainment of other resources. Stable employment is considered a condition resource that is valued by employees not only for its own purpose but also for its ability to facilitate the attainment of other resources (e.g., housing, food, clothing, income, social status, and respect; Jahoda, 1981). Therefore, JI implies a threat to employee resources in the form of lost employment and income or lost valuable job features (Hellgren, Sverke, & Isaksson, 1999). As such, we expect that experiencing high JI results in negative work-related and individual outcomes (see Figure 1). Although we use COR theory as the theoretical foundation, it is worth noting that some outcomes of JI (e.g., CWB) may be better explained by other theories instead of COR theory, which we specify below. An overview of the definition for each variable examined in Study 1 is available in the online supplemental material.

Correlates of JI

We examine the relations between JI and the following variables in an exploratory manner because there is mixed evidence about the causal flow between JI and these variables, despite consistent evidence about the direction of the associations (i.e., positive or negative) between JI and these variables.

CJI and AJI

The JI construct has both cognitive (beliefs) and affective (emotional state) components (Huang et al., 2010; Reisel & Banai, 2002). Although the affective component of JI has been proposed to best reflect the JI construct (Reisel & Banai, 2002; Sverke et al., 2002), the literature to date has treated JI as an essentially cognitive phenomenon (Huang et al., 2010). Indeed, most JI studies focus on cognitive rather than affective conceptualizations (Huang et al., 2010). For example, Greenhalgh and Rosenblatt (1984), Hartley, Jacobson, Klandermans, and Van Vuuren (1990), and Boswell, Olson-Buchanan, and Harris (2014) did not include an affective component in their operationalization of JI. Consequently, predominant JI measures tap only into employees’ perceptions of various events influencing the job or its various features (e.g., Ashford, Lee, & Bobko, 1989). As a result, few studies have employed both conceptualizations.

On the other hand, Johnson, Messe, and Crano (1984) conceptualized and measured the affective and cognitive components of JI. Similarly, Reisel and Banai (2002) advocated to combine CJI and AJI as a single construct and aggregated the two into one measure. Subsequently, some JI measures contain both cognitive and affective items (e.g., the four-item Job Insecurity Scale by De Witte, 2000; also see Vander Elst, De Witte, & De Cuyper, 2014), making it difficult to tease out the differential effects of CJI and AJI (Huang, Niu, Lee, & Ashford, 2012; Sverke et al., 2002). Inevitably, this has limited the conceptual development of JI.

However, as early as 1992, Borg and Elizur argued for the importance of distinguishing between the cognitive versus affective nature of the JI construct, finding that CJI was a necessary but not sufficient requirement for AJI to ensue. For example, two employees occupying identical positions with identical levels of CJI may not necessarily have identical AJI, which could be influenced by a number of variables (e.g., financial dependence on the job, employability; Borg & Elizur, 1992; Huang et al., 2010; Probst, 2003). More recently, Mauno and Kinnunen (2002) defined JI as uncertainty about the future of one’s job (i.e., CJI) as well as worry over job loss (i.e., AJI), which were assessed by two separate scales measuring each aspect of JI. Similarly, Probst (2003) developed two separate measures to assess CJI (i.e., the Job Security Index) and AJI (i.e., the Job Security Satisfaction Scale). Nevertheless, it remains an unexplored question whether differentiating between CJI and AJI is empirically meaningful.

In order to examine whether CJI and AJI can be considered distinct, we focus on three pieces of evidence. First, CJI and AJI must not be too highly correlated (Campbell & Fiske, 1959). When the correlation between two constructs approaches values commonly agreed on as acceptable for reliability coefficients (e.g., r = .70 or higher), this calls into question the distinctiveness of the two constructs (also see Berry, Ones, & Sackett, 2007). On this basis, there is empirical evidence for and against the distinction between CJI and AJI as previous research revealed that the correlation coefficient between CJI and AJI ranged from .02 (Staufenbiel & König, 2011) to .86 (Tilakdharee, Ramidial, & Parumasur, 2010). However, none of these correlations were corrected for statistical artifacts. Also, as with any primary study with a moderate sample size, sampling error could have affected the relative magnitudes of these correlations. Thus, a meta-analysis of such correlations, correcting for statistical artifacts, would be most useful. Therefore, we hypothesize that there will be a positive correlation between CJI and AJI but that this correlation will not reach .70 even when corrected for measurement error.

The distinctiveness of two constructs can also be demonstrated by the finding that these two constructs display differential correlations with other constructs. If CJI and AJI show differing patterns of correlations with outcomes/correlates, then CJI and AJI would represent distinct constructs. Because Huang et al. (2010), Huang et al. (2012), and Probst (2003) found that compared to CJI, AJI was more proximally related to employee outcomes, we propose that:

A third avenue to examine the distinctiveness of CJI and AJI is to investigate the incremental effect of AJI above and beyond CJI. To the extent that AJI overlaps with CJI, the potential of AJI to predict unique variance above and beyond CJI is limited. If AJI captures valid, unique variance in outcomes/correlates above and beyond CJI (i.e., incremental effect of AJI), it would provide support for CJI and AJI as distinct constructs. The value of AJI increases if it provides incremental information above and beyond CJI. Thus, we predict that:

As aforementioned, some JI measures, that is, mixed job insecurity (MJI) measures (e.g., Hellgren et al., 1999; Johnson et al., 1984), include items tapping into both CJI (e.g., “I can keep my current job as long as I want it”) and AJI (e.g., “I am afraid of losing my present job”; Francis & Barling, 2005). Construct contamination (i.e., a measure includes content that is not related to the conceptual content domain; Messick, 1995) can distort the predictor-criterion relation. For the present meta-analysis, construct contamination is operationalized as items within a JI measure assessing both CJI and AJI. Construct contamination in a JI measure can create “noise” in the relations between JI and outcomes/correlates and, consequently, attenuate effect sizes. Therefore, we predict that:

The Mediating Role of AJI

We also investigate whether AJI is a mediator linking CJI and its outcomes. At least three theories support the mediating role of AJI. According to cognitive appraisal theory (Lazarus & Folkman, 1984), CJI is the primary appraisal in that perceived JI is an evaluation of the likelihood of the loss of one’s job or attractive job features, which may be perceived as irrelevant, positive, or stressful (Jacobson, 1991). AJI is the result of the secondary appraisal, in which one reappraises a potentially stressful event in the light of one’s available resources to deal with the potential threat. Similarly, based on Weiss and Cropanzano’s (1996) affective event theory, the experience of a work event (e.g., CJI) can elicit affective reactions (e.g., AJI) that contribute to the formation of work attitudes and behaviors (Mignonac & Herrbach, 2004). Additionally, appraisal theories of emotion (Roseman, 1991) emphasize the cognitive determinants of emotions. Specifically, individuals first appraise stressful events (e.g., CJI), and these cognitive processes produce emotional responses (e.g., AJI). In line with these, AJI can be considered as an emotional reaction to CJI, which, in turn, leads to negative outcomes. Indeed, Probst (2003) argued that the effects of CJI on outcomes might be mediated by AJI. Empirically, Huang et al. (2010) found that AJI partially mediates the relations between CJI and job satisfaction, organizational commitment, and somatic well-being in two samples of employees from China. Using two-wave data from three large Chinese organizations, Huang et al. (2012) found that AJI partially mediates the relations between CJI and psychological well-being and job performance measured 6 months later. Taken together, these theories and studies suggest that AJI might be the underlying mechanism linking CJI and its outcomes. Thus, we propose that:

Method

Results

Table 1 summarizes the results for all outcomes/correlates of JI. Relations between JI and its outcomes/correlates were significant (at the p < .05 level) when the 95% confidence interval (CI) did not include zero. With five exceptions (i.e., continuance commitment, presenteeism, accidents, CWB, and musculoskeletal disorders without samples with over 10,000 participants ¹ ), JI was significantly related to the majority of outcomes/correlates that we examined in this study.

OPEN IN VIEWER

Table 1

Results of the Meta-Analysis for Outcomes and Correlates of Job Insecurity

Variables	N	k	r	SDr	ρ	SD ρ	80% CV	95% CI
Work-related outcomes
1. Psychological contract breach	13,352	19	.29	.08	.33	.10	[.22, .45]	[.29, .37]
2. Job satisfaction	152,410(313,395)	223(227)	−.30(–.41)	.14(.20)	−.37(–.51)	.17(.25)	[–.58, –.16]([–.83, –.20])	[–.39, –.35]([–.54, –.48])
3. Work-task satisfaction	11,719(126,128)	16(17)	−.26(–.40)	.11(.05)	−.31(–.49)	.14(.07)	[–.49, –.14]([–.59, –.40])	[–.37, –.25]([–.52, –.46])
4. Pay satisfaction	26,327(140,736)	26(27)	−.28(–.46)	.14(.11)	−.34(–.62)	.18(.18)	[–.57, –.12]([–.85, –.40])	[–.41, –.27]([–.68, –.56])
5. Promotion satisfaction	5,171	9	−.35	.08	−.41	.11	[–.53, –.29]	[–.47, –.35]
6. Coworker satisfaction	14,488	20	−.28	.13	−.35	.20	[–.60, –.10]	[–.42, –.28]
7. Supervisor satisfaction	17,657	25	−.33	.14	−.42	.19	[–.66, –.18]	[–.49, –.35]
8. Organizational commitment	20,002	43	−.25	.16	−.33	.20	[–.57, –.08]	[–.39, –.27]
9. Affective commitment	90,014(105,214)	165(166)	−.24(–.23)	.16(.15)	−.29(–.28)	.19(.18)	[–.53, –.05]([–.51, –.05])	[–.32, –.26]([–.31, –.25])
10. Continuance commitment	9,132	30	−.08	.21	−.08	.31	[–.46, .30]	[–.16, 00]
11. Normative commitment	3,205	15	−.25	.18	−.30	.22	[–.56, –.04]	[–.41, –.19]
12. Work engagement	15,387	26	−.17	.12	−.20	.13	[–.36, –.04]	[–.25, –.15]
13. Vigor	14,242	17	−.21	.06	−.24	.06	[–.30, –.19]	[–.27, –.21]
14. Dedication	2,806	9	−.13	.07	−.16	.08	[–.21, –.11]	[–.22, –.10]
15. Absorption	1,677	7	−.11	.10	−.15	.12	[–.27, –.03]	[–.25, –.05]
16. Burnout	22,840	21	.20	.08	.24	.10	[.12, .36]	[.20, .28]
17. Emotional exhaustion	50,308(71,572)	65(66)	.26(.24)	.10(.09)	.30(.29)	.11(.10)	[.17, .44]([.17, .41])	[.27, .33]([.26, .32])
18. Cynicism/ Depersonalization	17,964	28	.37	.14	.45	.16	[.26, .64]	[.39, .51]
19. Disengagement	418	3	.23	.04	.29	.04	[.29, .29]	[.23, .35]
20. Personal accomplishment	11,737	14	−.22	.12	−.28	.14	[–.45, –.11]	[–.36, –.20]
21. Absenteeism	11,719	11	.09	.04	.10	.05	[.05, .14]	[.07, .13]
22. Presenteesim	5,812	8	.05	.09	.06	.11	[–.07, .19]	[–.01, .13]
23. Turnover intention	63,442(199,115)	131(133)	.25(.32)	.14(.11)	.31(.42)	.17(.15)	[.10, .51]([.23, .60])	[.28, .34]([.40, .45])
24. Job search behavior	7,619	12	.21	.14	.25	.16	[.06, .45]	[.16, .34]
25. Actual turnover	1,862	7	.13	.06	.15	.07	[.15, .15]	[.10, .20]
26. Safety behaviors	8,334	11	−.12	.08	−.14	.09	[–.25, –.04]	[–.20, –.08]
27. Accidents	5,988	11	.10	.09	.11	.09	[.01, .21]	[–.05, .17]
28. Work motivation/effort	13,278	14	−.18	.11	−.24	.15	[–.42, –.05]	[–.32, –.16]
29. Job performance	5,644	25	−.11	.10	−.14	.12	[–.25, –.02]	[–.19, –.09]
30. OCBs	15,353	45	−.15	.10	−.18	.12	[–.32, –.05]	[–.22, –.14]
31. CWBs	3,048	10	.12	.19	.14	.23	[–.15, .42]	[.00, .28]
Individual outcomes
1. Strain	45,467(64,633)	64(65)	.21(.18)	.12(.11)	.24(.21)	.15(.13)	[.06, .42]([.05, .38])	[.21, .27]([.18, .24])
2. Life satisfaction	14,844	24	−.19	.09	−.22	.11	[–.34, –.10]	[–.26, –.18]
3. General health	35,587(56,851)	35(36)	−.15(–.14)	.07(.06)	−.19(–.18)	.09(.07)	[–.29, –.08]([–.26, –.10])	[–.22, –.16]([–.21, –.15])
4. Psychological health	67,119	86	−.24	.10	−.30	.12	[–.44, –.15]	[–.33, –.27]
5. Physical health	91,563	101	−.18	.08	−.23	.10	[–.35, –.11]	[–.25, –.21]
6. Musculoskeletal disorders	2,318(23,528)	5(6)	.09(.12)	.10(.03)	.11(.15)	.12(.04)	[–.03, .25]([.10, .20])	[.00, .22]([.12, .18])
7. Anger	1,644	4	.10	.04	.12	.05	[.12, .12]	[.07, .17]
8. Anxiety	7,994	16	.21	.05	.26	.06	[.22, .30]	[.23, .29]
9. Depression	22,077	27	.25	.10	.30	.13	[.14, .46]	[.25, .35]
10. Work-life conflict	35,071	33	.21	.15	.26	.19	[.02, .50]	[.20, .32]
Correlates
1. Negative affectivity	9,093	31	.20	.13	.24	.16	[.05, .43]	[.19, .30]
2. Positive affectivity	6,020	15	−.30	.14	−.39	.22	[–.65, –.12]	[–.48, –.30]
3. Organizational justice	29,930	21	−.28	.09	−.34	.10	[–.47, –.21]	[–.39, –.29]
4. Distributive justice	7,594	17	−.31	.13	−.38	.19	[–.31, –.14]	[–.46, –.30]
5. Procedural justice	12,615	28	−.29	.11	−.35	.12	[–.49, –.21]	[–.40, –.30]
6. Interpersonal justice	4,644	12	−.22	.10	−.25	.12	[–.38, –.12]	[–.31, –.19]
7. Trust in organization	3,482	10	−.45	.10	−.56	.14	[–.73, –.39]	[–.64, –.48]
8. Trust in management	10,787	31	−.39	.16	−.45	.19	[–.69, –.21]	[–.51, –.39]
9. Support	13,570	16	−.11	.10	−.17	.13	[–.32, –.02]	[–.25, –.09]
10. Organizational support	22,412	27	−.20	.14	−.24	.16	[–.44, –.03]	[–.30, –.18]
11. Supervisor support	29,538(39,650)	34(35)	−.20(–.21)	.09(.08)	−.23(–.25)	.12(.11)	[–.38, –.08]([–.39, –.11])	[–.26, –.20]([–.28, –.22])
12. Peer support	24,484(34,596)	26(27)	−.18(–.18)	.09(.07)	−.22(–.22)	.12(.10)	[–.37, –.08]([–.35, –.10])	[–.26, –.18]([–.25, –.19])
13. Job involvement	19,725	33	−.10	.11	−.13	.13	[–.29, .03]	[–.18, –.08]
14. Hazard exposures	7,155(17,267)	7(8)	.16(.15)	.08(.05)	.20(.19)	.11(.07)	[.07, .33]([.10, .28])	[.13, .27]([.15, .23])

Note: Values in parentheses are estimates including samples with more than 10,000 participants. N = number of participants; k = number of samples; r = sample-size-weighted mean observed correlation; SD_r = standard deviation of r; ρ = r corrected for unreliability; SD_ρ = standard deviation of ρ; CV = credibility interval of ρ; CI = confidence interval; OCB = organizational citizenship behavior; CWB = counterproductive work behavior.

Relationship Between CJI and AJI

Table 2 presents the meta-analytic results for the relation between CJI and AJI. The sample-size-weighted correlation between CJI and AJI was .53, and the reliability-corrected correlation was .65. However, the 95% CI around the corrected correlate ranged from .56 to .74, which included .70. Thus, we failed to find support for Hypothesis 1a.

OPEN IN VIEWER

Table 2

Results of the Meta-Analysis for the Relation Between Cognitive Job Insecurity and Affective Job Insecurity

Variable	N	k	r	SDr	ρ	SD ρ	80% CV	95% CI
Overall analysis
Cognitive job insecurity–Affective job insecurity	10,230	30	.53	.21	.65	.23	[.37, .93]	[.56, .74]

Note: N = number of participants; k = number of samples; r = sample-size-weighted mean observed correlation; SD_r = standard deviation of r; ρ = r corrected for unreliability; SD_ρ = standard deviation of ρ; CV = credibility interval of ρ; CI = confidence interval.

Outcomes and Correlates of CJI and AJI

To examine whether CJI and AJI had different relations with outcomes/correlates, we provide a side-by-side comparison of CJI and AJI corrected correlations with the outcomes/correlates in Table 3. As can be seen in the upper half of Table 3, AJI had significantly stronger relations with 16 out of 27 outcomes than did CJI, including job satisfaction, work-task satisfaction, pay satisfaction, supervisor satisfaction, coworker satisfaction, organizational commitment, affective commitment, burnout, emotional exhaustion, work motivation, strain, life satisfaction, general health, psychological health, physical health, and work-family conflict. The difference in correlations ranged from .02 to .28, averaging .14. However, CJI had significantly stronger relations with cynicism/depersonalization, turnover intentions, and accidents than did AJI. Additionally, there was no significant difference in correlations between AJI and CJI when predicting psychological contract breach, promotion satisfaction, normative commitment, work engagement, personal accomplishment/professional efficacy, job performance, and OCB. Surprisingly, continuance commitment was negatively related to CJI but positively related to AJI. Displayed in the lower half of Table 3, AJI had significantly stronger relations with 5 out of 9 correlates than did CJI, including procedural justice, trust in management, overall support, organizational support, and job involvement. The difference in correlations ranged from .10 to .18, averaging .14. However, CJI had significantly stronger relations with supervisor support and peer support than did AJI. Additionally, when correlating with positive affectivity, negative affectivity, and promotion satisfaction, the cognitive and affective nature of the JI construct did not generate significant differences. Thus, we found that AJI correlated more strongly with the majority of theoretically relevant outcomes/correlates than did CJI (21 out of 36), providing some support for Hypothesis 1b.

OPEN IN VIEWER

Table 3

Meta-Analytic Comparisons Among Corrected Correlations of Affective Job Insecurity, Cognitive Job Insecurity, and Mixed Job Insecurity With Outcomes and Correlates

	AJI (ρaji/k)	CJI (ρcji/k)	MJI (ρmji/k)	ρaji – ρcji	ρaji – ρmji	ρcji – ρmji	ΔR2 (AJI) a	ΔR2 (CJI) a
Outcomes
1. Psychological contract breach	.39/5	.31/6	.32/8	.08	.07	−.01	.06	.01
2. Job satisfaction b	−.43/43	−.37/97	−.33/83	−.06***	−.10***	−.04***	.06	.01
3. Work-task satisfaction b	−.33/9	−.24/6	N/A	−.09***	N/A	N/A	.05	< .01
4. Promotion satisfaction	−.46/5	−.35/4	N/A	−.11	N/A	N/A	.09	< .01
5. Pay satisfaction b	−.45/15	−.19/8	−.40/3	−.26***	−.05	.21***	.18	.02
6. Supervisor satisfaction	−.49/16	−.19/8	N/A	−.20***	N/A	N/A	.23	.03
7. Coworker satisfaction	−.43/12	−.18/7	N/A	−.25***	N/A	N/A	.17	.02
8. Organizational commitment	−.47/7	−.24/17	−.40/19	−.23***	−.07**	.16***	.17	.01
9. Affective commitment b	−.37/32	−.35/70	−.20/63	−.02*	−.17***	−.15***	.04	.02
10. Continuance commitment	.07/7	−.07/8	−.15/15	.14***	.22***	.08***	.02	.02
11. Normative commitment	−.37/4	−.29/6	−.27/5	−.08	−.10	−.02	.06	< .01
12. Work engagement	−.17/5	−.22/9	−.19/12	.05	.02	−.03	< .01	.02
13. Vigor	N/A	−.14/5	−.26/10	N/A	N/A	.12***	N/A	N/A
14. Burnout	.37/9	.28/6	.21/6	.09**	.16***	.07***	.06	< .01
15. Emotional exhaustion b	.36/10	.31/31	.28/24	.05*	.08***	.03***	.04	.01
16. Cynicism/Depersonalization	.41/8	.46/12	.43/8	−.05*	−.02	.03	.02	.06
17. Personal accomplishment	−.31/3	−.27/9	N/A	−.04	N/A	N/A	.03	.01
18. Turnover intention b	.25/20	.33/67	.32/44	−.08***	−.07***	.01	< .01	.05
19. Accidents	.05/3	.18/7	N/A	−.13***	N/A	N/A	.01	.04
20. Work motivation/effort	−.33/4	−.08/7	−.31/3	−.25***	−.02	.23	.13	.03
21. Job performance	−.13/3	−.15/17	−.11/5	.02	−.02	−.04	< .01	.01
22. OCB	−.18/5	−.17/26	−.21/14	−.01	.03	.04	.01	< .01
23. General health b	−.27/4	−.15/12	−.22/19	−.12***	−.05*	.07***	.05	< .01
24. Psychological health	−.34/19	−.29/42	−.29/25	−.05***	−.05**	.00	.04	.01
25. Physical health	−.29/18	−.20/56	−.26/27	−.09***	−.03*	.06***	.04	< .01
26. Strain b	.37/10	.19/32	.27/22	.18***	.10***	−.08***	.11	< .01
27. Life satisfaction	−.29/4	−.22/11	−.17/9	−.07**	−.12***	−.05*	.04	< .01
28. Anxiety	N/A	.24/11	.32/5	N/A	N/A	−.08**	N/A	N/A
29. Depression	N/A	.29/15	.32/10	N/A	N/A	−.03*	N/A	N/A
30. Work-family conflict	.44/9	.16/16	.19/8	.28***	.25***	−.03	.20	.03
Correlates
1. Negative affectivity	.28/5	.23/21	.26/5	.05	.02	−.03	.03	< .01
2. Positive affectivity	−.31/3	−.38/9	−.46/3	.07	−.15	.08	.01	.06
3. Organizational justice	N/A	−.36/12	−.34/9	N/A	N/A	−.02	N/A	N/A
4. Distributive justice	N/A	−.33/8	−.40/9	N/A	N/A	.07	N/A	N/A
5. Procedural justice	−.44/5	−.29/15	−.37/8	−.15***	−.06**	.08***	.11	< .01
6. Interpersonal justice	N/A	−.23/5	−.27/5	N/A	N/A	.04	N/A	N/A
7. Trust in management	−.47/7	−.37/14	−.57/10	−.10***	.10***	.20***	.09	.01
8. Support	−.27/4	−.14/11	N/A	−.13***	N/A	N/A	.06	< .01
9. Organizational support	−.34/6	−.16/7	−.35/14	−.18***	.01	.19***	.10	.01
10. Supervisor support b	−.22/4	−.32/17	−.17/13	.10**	−.05	−.15***	< .01	.05
11. Peer support b	−.09/4	−.29/14	−.17/8	.20***	.08*	−.12***	.02	.09
12. Job involvement	−.25/9	−.12/12	−.09/12	−.13***	−.16***	−.03	.05	< .01

Note: Bolded difference in corrected correlation comparisons represents instances in which affective job insecurity (AJI) has a significant stronger relation with the variable than cognitive job insecurity (CJI) and mixed job insecurity (MJI; a job insecurity measure with both CJI and AJI items). The sample size used in each regression is the harmonic mean of the sample sizes for the three meta-analytic correlations included in each regression. k = number of samples; ρ_aji = corrected correlation between AJI and variable; ρ_cji = corrected correlation between CJI and variable; ρ_mji = corrected correlation between MJI and variable; ΔR² (AJI) = the increase in the squared multiple correlation when regressing each common outcome/correlate on CJI and AJI instead of just CJI; ΔR² (CJI) = the increase in the squared multiple correlation when regressing each common outcome/correlate on CJI and AJI instead of just AJI; N/A = not available; OCB = organizational citizenship behavior.

a

All values are statistically significant due to large sample sizes.

b

Correlates without samples with over 10,000 participants.

*

p < .05.

**

p < .01.

***

p < .001.

Path Analyses Based on the Meta-Analytic Findings

Using the meta-analytic correlation matrices (see Table 4), we carried out separate path analyses for each outcome to test the mediation role of AJI. Table 4 displays the results of path analyses for the full mediation and partial mediation models. Compared to the partial mediation model, the full mediation model removed the direct link between CJI and the outcome of interest. Fit indices for the partial mediation model were not available because it was statistically saturated. However, because the full mediation model was more parsimonious than the partial mediation model, one can make the case that it was a better fitting model as long as the absolute fit was adequate (i.e., nothing was lost by removing the direct link between CJI and the outcome of interest). This was indeed the case because the full mediation model exhibited good model fit (CFI ranging from .96 to 1.00; SRMR ranging from .00 to .08) when predicting the majority of outcomes with 3 exceptions (i.e., cynicism, turnover intentions, accidents). Compared to the partial mediation model, the indirect effect of CJI on the outcome of interest mediated by AJI became stronger in the full mediation model when predicting 15 out of 21 outcomes, including psychological contract breach, job satisfaction, affective commitment, normative commitment, work engagement, burnout, emotional exhaustion, cynicism/depersonalization, personal accomplishment, turnover intention, job performance, OCB, life satisfaction, psychological health, and physical health. Thus, we found support for Hypothesis 3.

OPEN IN VIEWER

Table 4

Results of the Meta-Analytic Path Analyses

	Meta-analytic correlations (SDρ) used in path analyses			Fit indices for path models		Standardized pathways in path models			Indirect effect
Psychological contract breach (PCB)	CJI-AJI	CJI- PCB	AJI-PCB	CFI	SRMR	CJI-AJI	CJI-PCB	AJI-PCB
Overall relationships	.65 (.23)	.31 (.11)	.39 (.07)
Partial mediation model (N = 4,674)				—	—	.65	.10	.33	.21***
Full mediation model (N = 4,674)				.99	.02	.65	—	.39	.25***
Job satisfaction (JS)	CJI-AJI	CJI- JS	AJI-JS	CFI	SRMR	CJI-AJI	CJI- JS	AJI-JS
Overall relationships	.65 (.23)	−.37 (.16)	−.43 (.19)
Partial mediation model (N = 21,244)				—	—	.65	−.16	−.33	−.22***
Full mediation model (N = 21,244)				.98	.04	.65	—	−.43	−.28***
Organizational commitment (OC)	CJI-AJI	CJI-OC	AJI-OC	CFI	SRMR	CJI-AJI	CJI- OC	AJI-OC
Overall relationships	.65 (.23)	−.24 (.12)	−.47 (.27)
Partial mediation model (N = 3,710)				—	—	.65	.11	−.54	−.35***
Full mediation model (N = 3,710)				.99	.03	.65	—	−.47	−.31***
Affective commitment (AC)	CJI-AJI	CJI-AC	AJI-AC	CFI	SRMR	CJI-AJI	CJI-AC	AJI-AC
Overall relationships	.65 (.23)	−.35 (.18)	−.37 (.19)
Partial mediation model (N = 16,776)				—	—	.65	−.19	−.25	−.16***
Full mediation model (N = 16,776)				.97	.05	.65	—	−.37	−.24***
Continuance commitment (CC)	CJI-AJI	CJI-CC	AJI-CC	CFI	SRMR	CJI-AJI	CJI-CC	AJI-CC
Overall relationships	.65 (.23)	−.07 (.25)	.07 (.33)
Partial mediation model (N = 2,045)				—	—	.65	−.20	.20	.13***
Full mediation model (N = 2,045)				.96	.05	.65	—	.07	.05**
Normative commitment (NC)	CJI-AJI	CJI-NC	AJI-NC	CFI	SRMR	CJI-AJI	CJI-NC	AJI-NC
Overall relationships	.65 (.23)	−.29 (.24)	−.37 (.21)
Partial mediation model (N = 1,139)				—	—	.65	−.09	−.31	−.20***
Full mediation model (N = 1,139)				.99	.02	.65	—	−.37	−.24***
Work engagement (WE)	CJI-AJI	CJI-WE	AJI-WE	CFI	SRMR	CJI-AJI	CJI-WE	AJI-WE
Overall relationships	.65 (.23)	−.22 (.21)	−.17 (.09)
Partial mediation model (N = 4,373)				—	—	.65	−.19	−.05	−.03*
Full mediation model (N = 4,373)				.96	.05	.65	—	−.17	−.11***
Burnout (BO)	CJI-AJI	CJI-BO	AJI-BO	CFI	SRMR	CJI-AJI	CJI-BO	AJI-BO
Overall relationships	.65 (.23)	.28 (.16)	.37 (.10)
Partial mediation model (N = 3,188)				—	—	.65	.07	.33	.21***
Full mediation model (N = 3,188)				1.00	.02	.65	—	.37	.24***
Emotional exhaustion (EE)	CJI-AJI	CJI-EE	AJI-EE	CFI	SRMR	CJI-AJI	CJI-EE	AJI-EE
Overall relationships	.65 (.23)	.31 (.12)	.36 (.10)
Partial mediation model (N = 8,394)				—	—	.65	.13	.27	.18***
Full mediation model (N = 8,394)				.98	.03	.65	—	.36	.23***
Cynicism (CY)	CJI-AJI	CJI-CY	AJI-CY	CFI	SRMR	CJI-AJI	CJI- CY	AJI-CY
Overall relationships	.65 (.23)	.46 (.18)	.41 (.07)
Partial mediation model (N = 6,339)				—	—	.65	.34	.19	.13***
Full mediation model (N = 6,339)				.90	.08	.65	—	.41	.27***
Personal achievement (PA)	CJI-AJI	CJI-PA	AJI-PA	CFI	SRMR	CJI-AJI	CJI-PA	AJI-PA
Overall relationships	.65 (.23)	−.27 (.14)	−.31 (.14)
Partial mediation model (N = 2,164)				—	—	.65	−.12	−.23	−.15***
Full mediation model (N = 2,164)				.99	.03	.65	—	−.31	−.20***
Turnover intentions (TI)	CJI-AJI	CJI-TI	AJI-TI	CFI	SRMR	CJI-AJI	CJI-TI	AJI-TI
Overall relationships	.65 (.23)	.33 (.17)	.25 (.13)
Partial mediation model (N = 15,035)				—	—	.65	.29	.06	.04***
Full mediation model (N = 15,035)				.92	.07	.65	—	.25	.16***
Accidents (Acc)	CJI-AJI	CJI-Acc	AJI-Acc	CFI	SRMR	CJI-AJI	CJI-Acc	AJI-Acc
Overall relationships	.65 (.23)	.18 (.09)	.05 (.05)
Partial mediation model (N = 3,591)				—	—	.65	.26	–.12	−.08***
Full mediation model (N = 3,591)				.94	.06	.65	—	.05	.03**
Job performance (JP)	CJI-AJI	CJI-JP	AJI-JP	CFI	SRMR	CJI-AJI	CJI-JP	AJI-JP
Overall relationships	.65 (.23)	−.15 (.12)	−.13 (.13)
Partial mediation model (N = 1,872)				—	—	.65	−.11	−.06	−.04†
Full mediation model (N = 1,872)				.99	.03	.65	—	−.13	−.08***
Organizational citizenship behavior (OCB)	CJI-AJI	CJI-OCB	AJI- OCB	CFI	SRMR	CJI-AJI	CJI-OCB	AJI-OCB
Overall relationships	.65 (.23)	−.17 (.09)	−.18 (.20)
Partial mediation model (N = 4,019)				—	—	.65	−.09	−.12	−.08***
Full mediation model (N = 4,019)				.99	.02	.65	—	−.18	−.12***
Strain (Str)	CJI-AJI	CJI-Str	AJI-Str	CFI	SRMR	CJI-AJI	CJI-Str	AJI-Str
Overall relationships	.65 (.23)	.19 (.13)	.37 (.12)
Partial mediation model (N = 9,553)				—	—	.65	–.09	.43	.28***
Full mediation model (N = 9,553)				.99	.02	.65	—	.37	.24***
Life satisfaction (LS)	CJI-AJI	CJI-LS	AJI-LS	CFI	SRMR	CJI-AJI	CJI-LS	AJI-LS
Overall relationships	.65 (.23)	−.22 (.11)	−.29 (.03)
Partial mediation model (N = 5,176)				—	—	.66	−.06	−.25	−.17***
Full mediation model (N = 5,176)				1.00	.01	.65	—	−.29	−.19***
General health (GH)	CJI-AJI	CJI-GH	AJI-GH	CFI	SRMR	CJI-AJI	CJI-GH	AJI-GH
Overall relationships	.65 (.23)	−.15 (.08)	−.27 (.08)
Partial mediation model (N = 4,959)				—	—	.65	.04	−.30	−.19***
Full mediation model (N = 4,959)				1.00	.01	.65	—	−.27	−.18***
Psychological health (PsyH)	CJI-AJI	CJI-PsyH	AJI-PsyH	CFI	SRMR	CJI-AJI	CJI-PsyH	AJI-PsyH
Overall relationships	.65 (.23)	−.29 (.12)	−.34 (.10)
Partial mediation model (N = 13,962)				—	—	.65	−.12	−.26	−.17***
Full mediation model (N = 13,962)				.99	.03	.65	—	−.34	−.22***
Physical health (PhyH)	CJI-AJI	CJI-PhyH	AJI-PhyH	CFI	SRMR	CJI-AJI	CJI-PhyH	AJI-PhyH
Overall relationships	.65 (.23)	−.20 (.09)	−.29 (.09)
Partial mediation model (N = 14,492)				—	—	.65	−.02	−.28	−.18***
Full mediation model (N = 14,492)				1.00	.01	.65	—	−.29	−.19***
Work-life/family conflict (WLC)	CJI-AJI	CJI-WLC	AJI-WLC	CFI	SRMR	CJI-AJI	CJI-WLC	AJI-WLC
Overall relationships	.65 (.23)	.16 (.13)	.44 (.18)
Partial mediation model (N = 11, 878)				—	—	.65	–.22	.59	.38***
Full mediation model (N = 11,878)				.96	.05	.65	—	.44	.29***

Note: The sample size used in each path analysis is the harmonic mean of the sample sizes for the three meta-analytic correlations included in each path analysis. Italicized coefficients represent instances of “net suppression.” Em dashes represent instances in which that piece of information was unavailable. CJI = cognitive job insecurity; AJI = affective job insecurity; CFI = comparative fit index; SRMR = standardized root mean square residual.

†

p < .10.

*

p < .05.

**

p < .01.

***

p < .001.

Discussion

As predicted by COR theory and other theories, Study 1 demonstrates that JI is significantly related to the vast majority of outcomes examined in this meta-analysis with five exceptions. In addition to comprehensively examining outcomes/correlates of JI, Study 1 is the first to meta-analytically explore the cognitive and affective nature of the JI construct. First, Study 1 indicates that when corrected for unreliability, CJI and AJI are strongly related, with the corrected correlation being .65. Second, compared to CJI, AJI has stronger relations with the majority of outcomes/correlates. Third, results further indicate that AJI explains valid, unique variance in the common set of outcomes/correlates above and beyond CJI, which provides evidence of the incremental contribution of AJI. This highlights the importance of considering AJI and CJI separately in the operationalization and theory development in the JI literature.

Compared to MJI, AJI has stronger relations with the majority of outcomes. However, less conclusive evidence is observed in the relations between AJI and MJI with correlates. Contrary to our prediction, CJI displays stronger relations with only 7 out of 35 available outcomes/correlates compared to MJI. We offer two explanations for this. First, the percentage of AJI-related items in the MJI measures ranges from as low as 20% (e.g., Francis & Barling, 2005) to as high as 67% (e.g., Hellgren et al., 1999). Thus, differential levels of “noise” introduced in these MJI measures may influence the associations between MJI and outcomes/correlates. Second, the bandwidth-fidelity dilemma proposed by Ones and Viswesvaran (1996) indicates that broader global personality measures have stronger criterion-related validity when predicting overall job performance. Similarly, Cronbach (1960) argues that narrow bandwidth measures should be used to predict narrowly defined specific criteria. Thus, the predictive validity of AJI, CJI, and MJI may depend on the outcomes being predicted (e.g., specific or broad). Future research may explore this possibility.

Consistent with three theories (Lazarus & Folkman, 1984; Roseman, 1991; Weiss & Cropanzano, 1996) and the empirical findings of Huang et al. (2010) and Huang et al. (2012), path analyses suggest that AJI mediates the relations between CJI and 15 out of 21 outcomes examined in this study. There are six exceptions. First, AJI may not fully mediate the relation between CJI and continuance commitment, given that the indirect effect of CJI on continuance commitment mediated by AJI becomes stronger in the partial mediation model compared to the full mediation model. There are four cases in which when including the direct link from CJI to outcomes, the coefficient of the relation between CJI and the outcome (i.e., organizational commitment, strain, general health, and work-family conflict) becomes opposite to its corrected zero-order correlation in sign. The relation between AJI and accidents becomes negative while the corrected zero-order correlation between AJI and accidents is positive. These are instances of suppression models (MacKinnon, Krull, & Lockwood, 2000). Thus, the relations between CJI, AJI, and these six variables may be more complicated than can be fully investigated based on the present data.

In sum, Study 1 indicates that CJI and AJI are two distinct concepts demonstrated by the findings that (1) AJI has stronger (and, thus, differing) relations with the majority of outcomes/correlates than does CJI and (2) AJI accounts for appreciable incremental variance in outcomes/correlates over CJI and that AJI mediates the relations between CJI and various outcomes. Upon meta-analytically demonstrating the distinctiveness of CJI and AJI, Study 2 seeks to reveal a moderator in the CJI-AJI linkage.

The Moderating Effect of Work Centrality

As aforementioned, Hobfoll (1989) posits that resources may be valued in their own right. However, the value of resources may vary among individuals, and personal values influence the value placed on resources (Halbesleben et al., 2014; Morelli & Cunningham, 2012). According to Halbesleben et al. (2014), although one’s employment may serve as a valuable resource, individuals may value employment differently. This idea seems promising for specifying the conditions under which CJI may have differential impacts on AJI. Indeed, Halbesleben et al. argue that it is critical to examine the subjective evaluation of the value of resources (e.g., one’s work) when understanding the impact of the threat of resource loss (e.g., CJI) on one’s outcomes (e.g., AJI).

Halbesleben et al. (2014) and Morelli and Cunningham (2012) suggest that the value of the resource is the importance of the said resource to an individual. We use work centrality to represent the importance of work. Work centrality is the extent to which individuals believe that their work plays an important role in their life, irrespective of their current job (Bagger & Li, 2012; Hirschfeld & Feild, 2000; Paullay, Alliger, & Stone-Romero, 1994). People who consider work as a central life interest attach great importance to their work. In other words, individuals who are high in work centrality perceive the work role to be an important and a core part of their lives; work is something to be engaged in for its own sake (Hirschfeld & Feild, 2000).

Thoits proposes identity-relevant stressors that “are psychologically damaging only insofar as an individual values or is committed to the role domains in which those stressors occur” (1992: 249). That is, the negative effects of stressors in a certain role might be exacerbated when that role is more salient or important to the individual. We expect employees who rate their work as highly important (i.e., high work centrality) to fare worse in the face of CJI. Those who highly value work and therefore have more to lose are more psychologically vulnerable and experience more negative outcomes as a result of CJI (Shoss, 2017) because CJI implies a perceived threat to their highly valued resource (i.e., employment; Thoits, 1992). Thus, individuals’ affective reaction to CJI (i.e., AJI) should be exacerbated when they view their work as highly important and experience high CJI. On the basis of the values of resources in COR theory (Halbesleben et al., 2014) and identity-relevant stressors (Thoits, 1992), we propose that:

Method

Results

Means, standard deviations, intercorrelations, and scale reliabilities are presented in Table 5. ² To reduce multicollinearity (Aiken & West, 1991), we centralized CJI and work centrality before entering them into regression equations (see Table 6). As predicted by Hypothesis 1, CJI was positively related to AJI. Although work centrality was not significantly related to AJI, work centrality altered the relation between CJI and AJI. We then conducted simple slopes analyses and plotted the interactions (Aiken & West, 1991). As can be seen in Figure 2, CJI had a stronger relation with AJI (Sample A: b = 0.84, SE = 0.10, p < .001, 95% CI = [0.64, 1.05]; Sample B: b = 0.70, SE = 0.08, p < .001, 95% CI = [0.55, 0.85]) for those with high work centrality, while CJI had a weaker relation with AJI (Sample A: b = 0.48, SE = 0.10, p < .001, 95% CI = [0.28, 0.69]; Sample B: b = 0.49, SE = 0.07 p < .001, 95% CI = [0.36, 0.63]) for those with low work centrality. Thus, the moderating effect was significant in both samples, although the 95% CI around the simple slopes slightly overlapped. Taken together, both samples provided support for Hypothesis 4.

OPEN IN VIEWER

Table 5

Correlations and Descriptive Statistics of Study 2

Variable	1	2	3	α	M	SD
1. Cognitive job insecurity		−.02	.63**	.94 (.93)	1.41 (0.81)	1.14 (1.07)
2. Work centrality	−.18*		−.01	.80 (.84)	2.88 (2.73)	1.30 (1.42)
3. Affective job insecurity	.58**	−.11		.91 (.93)	2.80 (0.86)	1.29 (1.00)

Note: Correlation coefficients below the diagonal are from Sample A (N = 162). Means, standard deviations, and alphas in parentheses and correlation coefficients above the diagonal are from Sample B (N = 212).

*

p < .05.

**

p < .01.

OPEN IN VIEWER

Table 6

Multiple Regression Results of Study 2

	Affective Job Insecurity
Step and Variable	Sample A	Sample B
Step 1
Cognitive Job Insecurity	0.76 (0.09)***	0.62 (0.05)***
Work Centrality	0.00 (0.08)	0.00 (0.05)
R2	.34	.39
Step 2
Cognitive Job Insecurity × Work Centrality	0.20 (0.08)*	0.11 (0.05)*
R2	.37	.40
ΔR2	.03	.01

Note: Sample A (N = 162); Sample B (N = 212). The table shows regression coefficients with standard errors in parentheses.

*

p < .05.

***

p < .001.

OPEN IN VIEWER

Figure 2

Moderating Effect of Work Centrality in the Relationship Between Cognitive Job Insecurity and Affective Job Insecurity in Study 2

Discussion

Halbesleben et al. (2014) argue that overlooking the value of resources is a notable gap in the COR literature. Integrating the value of resources with identity-relevant stressors (Thoits, 1992), we argue that for individuals who highly value a particular resource, the possibility of losing that resource becomes more detrimental. Although the CIs of simple slopes of the CJI-AJI relation for those with high versus low work centrality slightly overlap, Study 2 indicates that the relation between CJI and AJI is contingent upon the extent to which the resource (i.e., work) holds value to the individual (i.e., work centrality). When individuals attach great importance to work, they report higher AJI 3 months later as a result of CJI in two independent samples, suggesting that CJI is more harmful for those with high work centrality.

Theoretical and Practical Implications

According to COR theory (Hobfoll, 1989), when perceiving the threat of resource loss, people have decreased well-being. Consistent with this proposition, Study 1 presents a meta-analytic summary of detrimental outcomes of JI. Exploring the threat of resource loss has been identified as one of the underdeveloped areas in the COR literature (Halbesleben et al., 2014). Thus, we contribute to COR theory by providing meta-analytic support for one of its propositions. That is, when individuals are threatened with resource loss (e.g., JI), they experience a host of negative outcomes. To date, there are only two meta-analyses on outcomes of JI. Replicating Sverke et al. (2002), the most recent meta-analysis on outcomes of JI was by Cheng and Chan (2008) with articles from 1980 to 2006. Given the increasing prevalence of JI and burgeoning research attention, there is a need to update the outcomes of JI and deepen our understanding of the JI construct. As the most ambitious examination of outcomes/correlates of JI, we examine 41 outcomes and 14 correlates of JI based on 535 independent samples using meta-analysis. Thus, compared to 8 outcomes examined in the previous meta-analyses, Study 1 represents a more comprehensive, quantitative summary of JI’s outcomes/correlates.

The estimates of the relation between JI and four outcomes (i.e., organizational commitment, turnover intention, psychological health, and physical health) found in our meta-analysis were largely comparable to those reported by Cheng and Chan (2008). Three notable discrepancies between our results and those of Cheng and Chan were that the estimates of the association between JI and job satisfaction (ρ = –.37), job performance (ρ = –.14), and job involvement (ρ = –.13) found in our meta-analysis were smaller than those reported in Cheng and Chan (ρ_{job satisfaction} = –.43; ρ_{job performance} = –.21; ρ_{job involvement} = –.20). Additionally, Cheng and Chan failed to distinguish the target of trust. Instead, they reported the estimate of the association between JI and trust was –.49. To provide a more precise estimate of the relation between JI and trust, we reported that the corrected correlate between JI and trust in an organization was –.56, while the corrected correlate between JI and trust in management was –.45.

Unfortunately, neither of these two meta-analyses considers different components of the JI construct or their differential impacts on outcomes and relations with correlates. We therefore clarify the distinct conceptual meanings of CJI and AJI and highlight the importance of treating them as two different constructs. Thus, we contribute to the JI literature by examining the relation between CJI and AJI, their differential correlations with outcomes/correlates, and the incremental contribution of AJI over CJI. Our results show that AJI has a stronger relation with the majority of outcomes/correlates than does CJI and that AJI explains valid, unique variance in the outcomes/correlates after controlling for CJI. However, CJI generally accounted for little incremental variance in the common outcomes/correlates beyond AJI. Thus, it may not be worth going to the trouble to collect CJI, especially when one is interested in outcomes of JI but constrained by survey space. Additionally, our results suggest the mediating role of AJI in the relation between CJI and its outcomes. This provides strong justification for considering AJI separately from CJI in the future conceptualization and theory development in the JI literature.

Furthermore, Study 2 provides support for the importance of the value of resources when applying COR theory. Specifically, we reveal that individuals who highly value their work experience more AJI 3 months later as a result of CJI in two independent samples. Although individuals with high work centrality experience many benefits associated with their work (Hirschfeld & Feild, 2000), their suffering may be more intense when the likelihood of job loss is high. Thus, by integrating the values of resources (Halbesleben et al., 2014) with identity-relevant stressors (Thoits, 1992), Study 2 contributes to both COR theory and the JI literature.

Kalleberg (2013) argues that today’s employees are confronted with fundamental shifts in the nature of work characterized by less secure employment systems compared to those faced by earlier generations of employees. Indeed, JI may be one of the stressors inherent in the workplace that employees have to “live with.” Although organizational changes may be unavoidable and therefore organizations may not be able to change employee perceived JI, organizations may help employees to effectively cope with JI. For example, organizations may foster high quality leader–member exchange (Loi, Ngo, Zhang, & Lau, 2011) by building the relation between supervisors and their employees on the basis of respect, mutual obligation, and trust (Graen & Uhl-Bien, 1995). Organizations could also provide employees with participative decision-making opportunities (Probst, 2005) by allowing employees to have a substantial voice in job-related decisions and enhance organizational communication (Jiang & Probst, 2014) by implementing a “realistic merger preview” (Schweiger & DeNisi, 1991) and opening the lines of communication with employees. While country-level and state-level income inequality may exacerbate one’s reaction to JI (Jiang & Probst, 2017), greater employment protection and income security (Carr & Chung, 2014; Debus, Probst, König, & Kleinmann, 2012) may help alleviate effects of JI. Thus, offering societal-level support structures (e.g., longer unemployment assistance, job skills retraining programs) may mitigate some adverse consequences of JI.

Upon revealing the mediating role of AJI in the relations between CJI and its outcomes, organizations may design interventions to target AJI. For example, organizations could utilize interventions that target employees’ emotional reactions to the negative changes to their job (e.g., expressive writing exercises; Spera, Buhrfeind, & Pennebaker, 1994). Indeed, Probst and Jiang (2016) found that an emotion-focused intervention (as opposed to problem-focused coping) is more efficacious in attenuating negative reactions to the threat of layoffs. Notably, Study 2 identifies a group of individuals (high in work centrality) who are highly vulnerable to CJI. As such, organizations should pay particular attention to individuals with high work centrality. For example, during layoffs or restructuring, organizations may ensure that employees with high work centrality are involved in emotion-focused interventions prior to problem-focused ones.

Limitations and Implications for Future Research

Although both studies contribute to COR theory and the JI literature, both studies have some limitations. First, although our results largely support COR theory, Study 1 does not directly investigate the explanatory mechanisms through which JI results in negative consequences. Thus, future research may explore threats to both manifest and latent benefits of work (Jahoda, 1981) as possible mechanisms underlying the relations between JI and its outcomes (see Vander Elst, Näswall, Bernhard-Oettel, De Witte, & Sverke, 2015). The majority of primary studies included in Study 1 are based on cross-sectional self-reported data. Future research on JI will benefit greatly from longitudinal studies that can more appropriately test the causal relation between JI and its outcomes (De Witte, Pienaar, & De Cuyper, 2016) and between CJI and AJI (Huang et al., 2012). As predicted, JI is significantly related to the vast majority of outcomes examined in Study 1. However, JI is not significantly related to continuance commitment, presenteeism, accidents, CWB, and musculoskeletal disorders. The relations between JI and these variables might be more complex than can be fully understood based on the present data. Thus, future research is needed before a definitive conclusion can be reached regarding these relations. Additionally, like any technique, the employed meta-analytic path analysis technique in Study 1 has its limitations (Bergh et al., 2016). Indeed, Yu, Downes, Carter, and O’Boyle (2016) propose to use full-information meta-analytic structural equation modeling to account for the heterogeneity of effect sizes when analyzing a path model. Future research may replicate our results using this advanced technique to examine the generalizability of the proposed path model.

In addition to conceptualizing JI as CJI and AJI, Hellgren et al. (1999; also see Ashford et al., 1989; Greenhalgh & Rosenblatt, 1984) argue that another important theoretical distinction concerns qualitative JI (i.e., threat to the continuity of the important job features) and quantitative JI (i.e., threat to the continuity of the job itself). Thus, future research may use a meta-analytic approach to examine these two types of JI and their differential relations with outcomes and correlates (cf. Reisel & Banai, 2002). Similarly, the results of the recent meta-analysis on predictors of JI (Keim, Landis, Pierce, & Earnest, 2014) may be updated by considering different types of the JI construct (e.g., cognitive and affective; qualitative and quantitative) and their differential predictors.

Study 2 uses two independent samples to consistently demonstrate the moderating effect of work centrality in the relation between CJI and AJI. However, both samples are based on employees from the United States. Thus, future research may examine our moderating hypothesis with a more representative sample from various countries. Although the moderating role of work centrality in the linkage between CJI and AJI was statistically significant in both samples of Study 2, the practical size of the observed effect is relatively small. However, this is not uncommon in field studies as previous research indicates that many effect sizes associated with interactions in field research tend to be moderate (McClelland & Judd, 1993).

Building on results from Study 1 and Study 2, future research should continue to explore potential moderators that might explain why employees with the same level of CJI may report different levels of AJI. For example, the relation between CJI and AJI may be contingent upon one’s optimism, employability, and family responsibility. In addition to individual differences, contextual variables may moderate the relation between CJI and AJI. For example, a strong JI climate at one’s work group (Jiang & Probst, 2016) might exacerbate the CJI-AJI relation. On the other hand, future research may examine the underlying mechanisms linking CJI to AJI. For example, work rumination may explain the association between CJI and AJI.