A test of the measurement equivalence of the revised Job Diagnostic Survey: Past problems and current solutions.

Abstract

The measurement equivalence of the revised Job Diagnostic Survey (JDS) was studied across samples from five worker populations. Samples included workers at a printing plant, engineers, nurses and nurses' aides, dairy employees, and part-time workers. Data were analyzed according to Jgreskog's model for simultaneous factor analysis in several populations (SlFASP), revealing the five factors contained in Hackman and Oldham's theory of job characteristics. A sixth factor also appeared that apparently resulted from the two different formats used on the instrument. When the data from each group were analyzed separately by principal axes factor analysis, three-, four-, and five-factor solutions appeared. To explain these inconsistencies, a Monte Carlo simulation was conducted. Matrices representing the a priori JDS factor loadings and a hypothetical, lengthened JDS with twice the number of items per factor were used in the simulation with three sample sizes (Ns = 75, 150, and 900). Results suggested that for scales like the JDS, which has only a few items per factor, sample sizes larger than those typically recommended are needed to consistently recover the true underlying structure. The simulation results support our conclusions that the SIFASP solution is preferable to the principal axes solution and that the JDS provides measurement equivalence across worker populations. The multidimensional nature of the job characteristic approach to job design is appealing to both researchers and practitioners because of its potential for diagnosing similarities and differences among jobs. Essentially, the goal of this approach is to provide a job-descriptive language that makes it possible to conceptualize jobs independently of worker skills, abilities, and activities (Fleishman & Quaintance, 1984). Most instruments developed to assess job characteristics are in survey form and are completed by job incumbents (Hackman & Oldham, 1975, 1980; Sims, Szilagyi, & Keller, 1976). For such instruments to be useful and for scale scores to be comparable across workers from different occupations and organizations, the instrument must provide equivalent measurement across diverse worker subpopulations. Equivalence of measurement is obtained when workers performing tasks with the same standing on the latent job characteristic construct give the same average rating (i.e., have the same expected observed score). Currently, no measure of job characteristics has been shown to provide measurement equivalence. For example, one of the most popular measures of job characteristics in the psychology literature is Hackman and Oldham's (1975, 1980) Job Diagnostic Survey (JDS), which was developed to assess five job characteristics across organizations and organizational levels. Despite