Comment

Abstract

This is a nice paper that uses data from the US Social Security Administration (SSA) to provide empirical evidence on gender differences in earnings. It focuses on differences in gender representation at the top 0.1% and the next 0.9% of the earnings distribution. It examines the persistence of an individual’s presence in these top percentiles, how age and industry composition matter, and gives some feel for life-cycle dynamics, all the while contrasting the presence of women versus men. The authors have access to a 10% representative sample of individual earnings histories from the SSA (constructed by selecting all individuals with the same last digit of a transformation of the social security number). This is a panel data set spanning 32 years: 1981–2012. There is basic demographic information available: age, sex, race, type of work (farm/ nonfarm, employment/self-employment), and earnings. The latter consists of wages and salaries, bonuses, and exercised stock options as reported in Box 1 on a W-2 form. For most of their analysis, they select from the 10% sample all individuals who in that year are between 25 and 60 years old and whose annual earnings exceed a minimum threshold (equivalent to 13 weeks, full-time, at one-half minimum wage). Among the advantages of a panel set are that one can, for example, study earnings over a number of years to smooth out temporary fluctuations, ask questions about persistence, and examine measures of lifetime income. What this data set sacrifices, however, is any rich information about other characteristics of these individuals such as their education, marital status and children, spousal attributes, and occupation other than that captured by broad industry categories. This makes it virtually