Abstract

A continuous and a discrete rating scale were implemented for a single item happiness question in a representative survey. A randomized controlled experiment enables unique analyses on data quality and distributions, which suggest superiority of the continuous scale. Results raise doubts about earlier inferences drawn on correlates of happiness. So far only self-assessed discrete happiness data have been used for research into the determinants of happiness. However, distribution distortions were found for the numerically labeled discrete scale, especially for women. Through this discretization bias, the widely reported gender happiness inequality puzzle can be explained.