Ask the Child or the Parent? Survey Design and the Measurement of School Satisfaction

In 1965, about one thousand children in the third grade of a midsize Swedish town answered a short questionnaire about school, and their parents answered a parallel questionnaire about the same child (a design repeated in grade six and grade nine). This article uses these paired reports to revisit, in a new construct and country, two sensitivities highlighted by Conti and Pudney (2011): that response-scale format and labeling can reshape reported satisfaction distributions, and that response conditions can induce systematic, gender-differentiated departures from classical measurement error. Parents concentrate on the conventional positive category while children spread across the scale; the distributional distance between informants is largest in grade six, where both answered fully labeled five-point scales. An integrative latent model attributes to the child report extreme-category misclassification probabilities of 0.14–0.34 and to the parent report a systematic threshold shift (context/informant perspective) with noise standard deviation close to one (estimates close to the original British ones). The choice of informant changes econometric conclusions: conditional on achievement, cognitive ability is negatively associated with satisfaction only in the child's own reports, and classroom-context coefficients switch between informants. Measured satisfaction is partly an artifact of who is asked and how.