Microsoft Word ji job Pres Preprint docx

Analytical Approach 
To test Hypotheses 3 and 4, we modeled the focal variables as latent variables in a 
mediation model and included perceived threat controllability and threat proximity as the first-
stage latent moderators. Because there might be other mechanisms linking JI to our outcomes 
(e.g., stress, social exchange; Staufenbiel & König, 2010), we model both an indirect effect 
through job preservation motivation and a direct link from JI to our outcomes. To test the 
moderation effects, we used the XWITH command in Mplus to create latent interaction terms. 
Because case-based bootstrapping is unavailable in Mplus for models with latent interactions, 
Monte Carlo bootstrapped confidence intervals with 10,000 repetitions were used to assess the 
conditional indirect effects and +1/-1 standard deviation (SD) of the moderator (Preacher & 
Selig, 2012). 
STUDY 2: RESULTS 
Table 4 displays the means, SDs, Cronbach’s alphas, and bivariate correlations.
-------------------------- 
Insert Table 4 about here 
-------------------------- 
CFA 

---

JOB INSECURITY AND JOB PRESERVATION 22
CFA models were used to confirm construct distinctiveness. As in Study 1, items were 
used as indicators for all corresponding constructs except for counterproductive work behavior, 
which used four 3-item parcels as indicators. The anticipated 8-factor (i.e., JI, job preservation 
motivation, counterproductive work behavior, self-presentation ingratiatory behavior, knowledge 
hiding, job performance, perceived threat controllability, and threat proximity) model was tested 
against four alternative models. As seen in Table 5, only the 8-factor model showed a 
satisfactory fit: χ
2
(377) = 706.17, CFI = .95, RMSEA = .054, SRMR = .048, and the lowest 
BIC. The 8-factor model fit better than the four alternative models (p < .01).
-------------------------- 
Insert Table 5 about here 
-------------------------- 
To examine the CMV, similar to Study 1, we added the common method factor to the 
eight-factor CFA model and evaluated the change in model fit. The model with common method 
factor had larger BIC, equivalent CFI, RMSEA and SRMR (χ
2
(376) = 705.94, BIC = 17282.80, 
CFI = .95, RMSEA = .054, SRMR = .048), and the factor loadings of items between the model 
with and without the common method factor did not change.
2