* Do-files are those used to construct the relevant figures

Sample restrictions

The sample for this paper was constructed from the 1962-2003 March CPS Annual Demographic Survey files, provided by Unicon. The full description of the imposed restrictions can be found in Appendix A of the paper. Summarized below are the main restrictions imposed on the original CPS data.

We restrict the sample to civilian adults between the ages 22 and 65. When constructing the employment and weeks worked samples we make no further restrictions.

In our wage and occupation sample we further restrict our attention to full-time full-year workers, where a worker is considered full-time if she works at least 35 hours a week and full-year if she works at least 40 weeks a year. For these workers we construct the wage sample by taking the annual wage and salary earnings, dividing it by the number of weeks worked and annualizing the resulting weekly wage by multiplying it by 52. Nominal wages are then deflated using the Personal Consumption Expenditure deflator from the NIPA to express all real wages in constant 2002 dollars.

For both the occupation and the wage sample we exclude observations where the worker is working without pay or is self-employed. We assign topcoded observations prior to 1995 a value of the topcode multiplied by 1.75. Finally, we drop observations with very low wages, specifically we use a sixth of the minimum wage as the lower limit below which observations are dropped.

Once constructing our samples, we divide them by gender and into 5 education groups: high school dropouts (HSD), high school graduates (HSG), workers with some college (SC), college graduates (CG) and post-college degree holders (PC). Past 1992 when we have information on an individual's highest degree received the construction of the education variable is straight-forward. Prior to 1991, however, we only have information on the number of grades attended and completed. In order to determine the best educational classification into 5 groups we match observations between 1991 and 1992 to see what the same individual reported as educational attainment under the two classification systems (due to design of the CPS half of the March sample overlaps from one year to the next). Based on this it is clear that the best correspondence is to classify those who have completed less than 12 years of schooling as HSD (with an 89% overlap), those who completed exactly 12 years of schooling as HSG (with an 87% overlap), those who started 13^th grade but did not complete 16 years of schooling as SC (with an 84% overlap), those who completed 16 years of schooling but only started their 17^th year as CG (with an 83% overlap), while those who completed 18 years of schooling as PC (with an 83% overlap). The only difficult decision was how to consider those who completed 17 years of schooling but not 18, as 48.5% of these report having a college degree while 42.5% of them report having a post-college degree. We decided to classify them as college graduates since this gave the smallest break in the composition graphs at the time of the classification change. Also note that compared to the size of college graduates and those who are definitely post-college degree holders (those with 18 years of education or more), the group of those with 17 years of education is small (less than 20% of college grads and less than 30% of post-college degree holders).

There were several changes in the occupational classification system that was used in the CPS. Prior to the 1971 March CPS, the "1960 Census of Population Occupational Classification System" was used, which was sufficiently different from later years to exclude these years from our study of occupations. Starting with the 1971 March CPS the 1970 Census classification was used, starting with the 1983 March CPS the 1980 Census classification was used, while starting with the 1992 March CPS the 1990 Census classification was used, which was very similar to the 1980 classification system. Starting with the 2003 March CPS the "North American Industry Classification System (NAICS)" based on the Census 2000 industry and occupation codes was used. We did not recode this system to be consistent with previous years of data, since this would have given us only one more year of observations.

We constructed the following 3 categories by consistently aggregating the different classification systems:

Professional and managerial occupations: Executives, administrative, and managerial occupations; sales supervisors; engineers and architects; mathematical, natural and computer scientists; computer programmers; social scientists; physicians, dentists and related practitioners; postsecondary teachers; lawyers and judges; writers, artists, entertainers.

White collar occupations: Registered nurses, pharmacists, dietitians, therapists, and physicians' assistants; kindergarten, elementary, and secondary teachers; counselors; librarians, archivists, and curators; social, recreation, and religious workers; technologists and technicians; sales representatives, workers, and related occupations; administrative support occupations; records processing occupations; computer equipment operators; information clerks; material recording, scheduling, and distributing clerks; adjusters and investigators.

Blue collar occupations: Service occupations; billing, posting, and calculating machine operators; farming, forestry, and fishing occupations; precise production, craft, and repair occupations; operators, fabricators, and laborers.

Description of files containing figures

In this section we present all the figures (in Excel format) that were constructed during this project, including the figures that are presented in the paper. The graphs are divided into several subgroups, in accordance with the addressed issues, and the organization of the graphs is similar to that of the paper. For each of these subgroups, we also present the appropriate do-file, which was used to construct the graphs.

Figures from paper

Figures presented in the paper are part of the set of figures that are presented here. Please note that in each Excel file there are hidden pages with data that were used to construct these figures.