Stable traits but unstable measures? Identifying panel effects in self-reflective survey questions

Economists and psychologists often measure aspects such as utility, preferences, and personality traits through self-assessment modules in longitudinal household surveys. This paper investigates to what extent such measures are subject to a panel effect or panel conditioning, that is, whether people answer the questions differently the more experience they have answering such questions. First, the paper makes a more general contribution to the literature on panel effects and makes explicit identification issues that arise in different types of empirical strategies. Next, the empirical analysis exploits a design feature of the UK Household Longitudinal Survey that introduces random variation in survey experience within a calendar year. The analysis first confirms the existence of such a panel effect in general life satisfaction, a pattern previously established in other data with a slightly different identification strategy. The data also provide evidence of panel effects in domain satisfactions, although these are less straightforward to interpret. This finding is important if researchers consider repeated measurements of such traits in household surveys to investigate their stability over time for a society or for an individual: the paper illustrates how conclusions on time trends in the subjective data for this case study are influenced if panel effects are ignored.