William M. Rodgers

THE PERSISTENCE OF GENDER EARNINGS INEQUALITY IN TAIWAN, 1978-1992

During the 1980s, Taiwans industry and export mixes shifted toward higher-skill, technology-intensive products, and lower-skill, labor-intensive industries began moving abroad. Despite improvements in womens skills and educational attainment relative to mens, the mean gender earnings ratio between 1978 and 1992 remained at 65%. The authors analyze household survey data from Taiwans Manpower Utilization Survey to examine why rapid structural change was not accompanied by a narrowing of the gender gap. The results strongly suggest that large losses experienced by women in unmeasured gender-specific factors—which could reflect the effects of labor market intermittency, growing gender differences in unobserved skills, or an increase in wage discrimination against women—offset their relative gains in education and experience. Further evidence provides no support for a widening gender gap in labor force commitment or in unobserved skills, suggesting that wage discrimination against female workers increased over time.

The Impact of Client Nonpayment on the Income of Contingent Workers: Evidence from the Freelancers Union Independent Worker Survey

The authors use the Freelancers Unions annual Independent Worker Survey (IWS) from 2007 to 2011, the unions administrative membership file, and the American Community Survey to estimate the extent to which client nonpayment of independent workers is a problem and the ability of a written contract to reduce the odds of nonpayment. They develop tests for nonresponse and sample selection biases, although due to a lack of instrumental variables and contract endogeneity, they must speculate on the size of the bias. They find that contract use is associated with income that is 13.7% higher (for New York State respondents 21.7% higher) than income produced without a contract. They also find that contract use reduces nonpayment; however, even when a contract is used, 38.8% of respondents still had trouble getting paid: One-third were paid late, and 10% were either paid less than the agreed upon amount or were never paid.

Estimating the bias due to measurement error in the economic returns to schooling: evidence from the 1990 February Current Population Survey

Abstract We use the February 1990 CPS to construct a measurement error corrected estimate of the return to schooling. Our model yields a bias that ranges from 11 to 15 percent. Removing the bias raises the return from 8 to 9 percent. Both estimates are at the bottom of the range found in the literature.