Testing a “Push–Pull” theory of work commitment among organizational professionals

Abstract

Writers have suggested that the current trend toward decreased job security requires employees to commit more strongly to newly “professionalized” occupations to compensate for social and resource support no longer received from their employers. And it has sometimes been implied that such a shift toward increased professional commitment will arise naturally as organizational commitment is whittled away by perceived job insecurity. We propose that job insecurity does not automatically push the employee toward professional commitment, but rather that such commitment stems from the pull of perceived occupational professionalization. We construct a nonrecursive model proposing relationships between job insecurity, perceived professionalization, and both organizational and professional commitment. This model is supported (using structural equation modeling) in a study of 622 employees in 3 occupations: corporate law, human resource management, and computer programming, all of which can be considered professions or semiprofessions. Finally, we suggest how occupations can be fashioned better to support employees when faced with job insecurity and job loss.

Notes

1 Tel.: [phone].

2 Tel.: [phone].

1 The χ²/d.f. ratio for Model 1 is 0.713, where anything less than 2.0 is considered acceptable fit. The probability coefficient represents the chance that the model variance/covariance matrix is the same as the data matrix. A score greater than .05 indicates good fit and ours is .544, meaning that the probability that our model is a good fit with the data is sizeable. The root mean square residual (RMR) is the square root of the average squared amount by which the sample variances and covariances differ from their estimates obtained under the assumption the test model is correct. An RMR of 0 indicates perfect fit, and 1 smaller than .05 is considered acceptable. Our RMR measure is .019, well within the bounds of acceptability. The Goodness-of-Fit Index (GFI) indicates the closeness of the data matrix to that of the proposed model. It varies between 0 and 1, where one indicates perfect fit, and anything over .95 is considered to be acceptable. The adjusted Goodness-of-Fit Index (AGFI) adjusts the goodness of fit for degrees of freedom: in essence, reducing the fit measure for adding new paths and variables. In our model the GFI is .999 and the AGFI .991, both quite good scores. The Bentler–Bonnet Normed Fit Index (also called Δ−1; see Arbuckle, 1995, p. 528; Bentler & Bonett, 1980) refers to the degree of closeness to the perfectly fitting model. A fit index of .90 or greater is considered acceptable and our model has a fit of .998. Bollen's (1989b) Incremental Fit Index (or Δ−2) focuses on the discrepancy between the test model and a baseline model; it varies between 0 and 1 where unity represents perfect fit. Our score is a very high 1.00.

2 We calculated the differences between each of the subsample model coefficients (12 for each of 3 models) and the overall, full-sample model coefficients. The average of these 36 difference scores was .07, thus indicating only minor variations in models as tested in the occupational subsamples.