The relative utility hypothesis with and without self-reported reference wages

This article uses survey data of 90,000 union employees working in 62 publicly-traded companies in Japan between 1990 and 2004 to study the effect of both own and self-reported reference wages on workers' subjective well-being levels. The availability of self-reported reference wages generates very robust findings that do not depend on questionable identifying assumptions. These findings confirm that higher estimates by workers of their peers' earnings are associated with lower levels of life and job satisfaction. These comparison effects are statistically and economically strong but are smaller in absolute value than the impact of workers' own wages on their own utility. We compare our results with standard tests of the relative utility hypothesis in the literature that recur to alternative proxies for comparison wages, including: (i) Mincer-predicted wages; (ii) cell averages defined over different groups within our dataset; (iii) cell wage averages estimated from an external data source; and (iv) colleagues' average wages. In spite of their potential flaws - that we discuss - these alternative empirical constructs employed in the literature do not introduce a simple classical measurement error problem and the bias attributed to this measurement error issue can go in both directions. We propose a simple IV strategy when the self-reported reference wage is not available that does not eliminate the bias but delivers a lower bound of the true effect. We also address the issue of endogeneity of self-reported reference wages in our subjective well-being regressions by accounting for workers' pessimistic attitudes at the workplace.