This paper aims to evaluate the variations in the entry and exit of women apprentices in the USA, overall and by race/ethnicity, over the 1995‐2003 period. Also aims to examine how women's representation among new apprentices and their attrition and retention rates vary with individual, training program, and occupational characteristics.

An individual‐level dataset from the US Department of Labor is used to estimate econometrically women's representation in apprenticeship programs and women's odds of completing programs.

Women's representation among new trainees is very low and deteriorating. The results confirm previous findings based on data for the early 1990s that program sponsorship has significant impact on women's representation and retention. Women have better chances of joining the high‐skill construction workforce if they enroll in union‐contractor joint programs. Joint programs feature higher shares of women in the incoming classes and higher odds of graduation in comparison with the unilateral contractor programs. While White women have higher completion rates than Latinas and Black women, the union impact on shares of enrollees is the largest for Black women and the lowest for White women.

The dataset is not nationally representative. It covers 31 states or about 65 percent of all apprentices.

Union sponsorship is necessary but not sufficient to enhance women's integration in the trades. Increasing participation of women in apprenticeship and the trades requires major changes in policies, priorities, and behavior of contactors, unions, and the government to actively recruit women and improve working conditions at the construction site.

This is the first systematic analysis of performance of women apprentices that utilizes the most recent data from the USA.