Evaluating weighted models of measuring job satisfaction: A cinderella story

Abstract

Data obtained from a national probability sample of 1533 American workers were used to assess the validity of ten methods of weighting job satisfaction ratings by importance ratings. The methods varied systematically in terms of the input they employed, scale treatment, and the types of mathematical operations used for weighting. Each method was evaluated with data from two subsamples of workers who differed in terms of the magnitude of their intraindividual correlations between satisfaction and importance ratings. The four criterion variables used to evaluate the weighting methods were measures of overall job satisfaction, job-related tension, likelihood of leaving one's present job, and mental health. The data not only failed to support the hypothesis that the validity of job satisfaction ratings may be increased by weighting them by importance ratings but indicated, on the contrary, that importance-weighting actually reduced the validity of satisfaction ratings. This reduction was attributed to several statistical problems, principally the failure of satisfaction and importance ratings to meet the demanding scaling assumptions of weighting models. An inconsistency was pointed out between the research activities necessary for providing an ideal test of a weighting model and those necessary for developing a generally useful measure of job satisfaction based on the model.