General life satisfaction and specific satisfactions Structure, interpretation and comparability across populations

Each year, satisfaction with living conditions is measured through six questions in the French sample of the Statistics on Income and Living Conditions (EU-SILC) survey: satisfaction with housing, leisure, work, and relations with family and friends are each the subject of a specific question, while a sixth question concerns satisfaction with one’s current life, often referred to as "satisfaction with life in general" or "general satisfaction with life". This is very often the only question used in studies of satisfaction: it makes sense a priori insofar as life seems to encompass all the dimensions that one could imagine questioning in detail. In this work, we first show that the "satisfaction" construct underlying the five sub-dimensions studied is twodimensional, opposing a rather material component (housing, work and leisure) to a more relational one (family and friends), the latter explaining a much larger share of the variance than the former. Second, we show that the variance of responses to the question "general life satisfaction" is mainly explained by the material component of the construct and that it is a poor proxy for the two-dimensional satisfaction construct. In other words, using the question on general life satisfaction as a synthetic measure of satisfaction is more or less the same as being interested in satisfaction with material living conditions and very little with social relations. Finally, we show that the two-dimensional satisfaction construct reach metric but not not scalar invariance with respect to age and urban unit size, and only partially metric invariance with respect to standard of living. This makes it possible to compare correlation coefficients involving the latent factors underlying the satisfaction items between groups defined by these criteria of age or place of residence, but not of standard of living. In any case, the means of the latent factors between groups cannot be reliably compared. Strictly speaking, this important limitation implies that it is also difficult to compare the average levels of each of the satisfaction variables between subpopulations defined by these criteria; this limitation also applies to general life satisfaction.