Reliability of job satisfaction measures

In this paper, we evaluate consistency in answers to subjective questions about job satisfaction and explore the implications of any inconsistencies. We do this by analyzing a cross-country data set for 6 EU countries where respondents were posed the same question about overall job satisfaction twice within the same questionnaire. We find that, on a 0–10 point ordered utility scale, 80% either classify themselves identically or in the immediate adjacent and that the differences in classification are symmetric around zero. Furthermore, we find that highly satisfied workers report most consistently. When job satisfaction is used as an explanatory variable, we show how OLS-parameter estimates provide a lower bound and IV-estimates an upper bound of the true estimate – and that the bounds are fairly tight. When job satisfaction is used as dependent variable, we generally find high consistency when parameters are highly significant in both models, while less significant or insignificant parameter estimates may change considerably. This indicates that higher significance standards may be advisable in analyses with satisfaction measures as dependent variable compared to more traditional models that are not based on subjective data.