Union Commitment

Factor Controversy

Factor solutions based on confirmatory analyses have produced a lack of consensus about the underlying dimensionality of the union commitment construct. All the following potentially conflicting interpretations of the nature and structure of the construct have been suggested: The factors are orthogonal (independent) and replicable across samples of nonprofessional and professional workers. The factors are parsimoniously defined by two oblique (nonindependent) factors, one that describes union attitudes and opinions (union loyalty and belief in unionism) and one that depicts pro-union behavioral intentions (i.e., responsibility to the union and willingness to work for the union). The factors are identifiable in an oblique four-factor solution and show stability over time.

Indexes for closeness of fit between a hypothesized factor model and an observed model (i.e., fit indexes) have produced some clarity about the dimensionality of the construct. All the following potentially revealing, albeit disparate, results have been reported: Fit indexes for an oblique four-factor solution are significantly better than one-factor, two-factor, or higherorder factor solutions. Belief in unionism is a methodological (artifactual) factor caused by negatively worded items. Belief in unionism is related to work commitment rather than to union commitment. Belief in unionism is the most stable of the commitment factors and influences union loyalty and responsibility to the union. Fit indexes for an oblique three-factor solution are improved with deletion of belief in unionism items. An oblique three-factor solution based on union loyalty, responsibility to the union, and willingness to work for the union shows that stability of items across factors (measurement invariance) can be assumed between men and women members, and to some extent between longtime and new members, but cannot be assumed between rankand-file members and stewards. Fit indexes for an oblique three-factor solution are improved with controls for similarity of scores within local unions (with controls for nonindependent observations), highlighting the need to adjust individual-level results by unitlevel (contextual) variation.

...