Measuring Subjective Well-Being: Does survey item wording shape life satisfaction reports?

Measures of subjective well-being are becoming increasingly popular in the field of economics. Most commonly used are single-item life satisfaction scales (questions akin to “how satisfied are you with your life overall?” that respondents answer on a given response scale, often from 0=extreme dissatisfaction to 10=extreme satisfaction). Despite the potential benefits of such a direct welfare measure, the enthusiasm is hampered by concerns about data quality. A major issue is sensitivity of responses to situational factors, which has been shown in several studies. This thesis investigates survey item wording: the phrasing of the question and response scale. The possibility of wording effects is particularly problematic because they could systematically bias results and limit the comparability of data from surveys with dissimilar items. By means of an experiment, differently phrased items were compared, with treatments representing wording variations found among major surveys. In total, 328 people completed a survey with one of the treatment items, where assignment to conditions was random. Life satisfaction averages, dispersion and correlates were compared between items that were equivalent except for one aspect (question tone, specification of a time period or labelling of the response scale). No significant differences were found regarding life satisfaction average and dispersion between the matched versions. Some differences in correlates were found. Taken together, the results suggest that differently phrased survey items lead to similar reports in terms of the overall distribution but that the analysis of well-being determinants could be biased by wording effects. No conclusive evidence was found regarding mechanisms that might explain the divergence in correlate patterns.