The Beauty Premium is Not Robust

A growing literature claims beautiful workers receive higher wages. Using five nationallevel surveys, we find that in multivariate regression models there is no beauty premium. Most beauty coefficients are statistically insignificant, even in replications of published work. Simple model alterations, such as clustering on the primary sampling unit or redefining other covariates, eliminate much of the beauty premia from two surveys. Adding controls for other ascriptive characteristics, such as personality, results in mostly insignificant beauty coefficients. Even in surveys without other ascriptives, the beauty premium is neither consistently significant nor consistently positive. Either beauty’s effect is small, or it is mediated through other variables.