We downloaded the IPUMS-CPS data on April 28th 2008 including the years 1963-2006 in the extraction.
Just as for the Census, we constructed an employment sample and a wage sample. We used the first to calculate measures of hours worked and employment and the second to calculate the average weekly wages for U.S.- and foreign-born, males and females, in each skill group and in each census year. The first sample is more inclusive than the second. We construct hours worked, employment and average wage for each of 4 education groups (workers with no high school, high school graduates, workers with some college and college graduates) following as closely as possible the procedure described in Katz and Murphy (1992), page 67-68 and detailed
below.
C.1 Definition of the Samples and Restrictions
C.1.1 EMPLOYMENT SAMPLE
For the EMPLOYMENT SAMPLE our definition aims at including all workers who supplied some hours of work and who were in the education by experience sample as the one used for Census data. The relevant eliminations made in the main sample where:
1) Eliminate people living in group quarters (military or convicts), which are those with the gq variable equal to 0, 3 or 4.
2) Eliminate people younger than 18.
3) Eliminate those who worked 0 weeks last year which corresponds to wkswork2=0 until 1975 and wkswork1=0 after 1975. Eliminate those who worked 0 hours by omitting hrswork=0 until 1975 and uhrswork=0 after 1975.
4) Once we calculate experience as age-(time first worked), where (time first worked) is 17 for workers with no HS degree, 19 for HS graduates, 21 for workers with some college education and 23 for college graduates, we eliminate all those with experience <1 and >40.
Construction of the hours worked and employment by education group and year
We first calculate the total amount of hours worked by sex-education-experience cell in each year by adding up the hours worked by each person in the cell, multiplied by her personal weight (PERWT). Then we calculate the average weekly wage in each of the four education groups (SHS, HSG, SCO, COG) over the 1963-2006 period and we sum up in each education group-year the hours worked by each experience-sex group scaled by the average (1963-2006) weekly wages of the group relative to males with 10-15 years of experience. This converts hours worked to equivalent hours assuming that different groups have different labor effectiveness. We use the same procedure to calculate employment in each education group per year.
C.1.2 WAGE SAMPLE
The construction of the Wage sample follows these steps:
a) Apply the same elimination as for the Employment sample
b) Eliminate those workers who do not report valid salary income (999999) or report 0.
c) Eliminate the self-employed (keep those for whom variable CLASSWKR is between 20 and 28).
Construction of the average weekly wage by education cell in each year.
In each of the sex-education-experience cells in each year we average the weekly wage of individuals, each weighted by the hours worked by the individual. Then we aggregate the average wage of each cell within the
four education groups for each year by weighting the average weekly wage in the sex-education-experience cells by the average (1963-2006) hours worked by the education-experience-sex group relative to the total of the education group. This keeps the composition constant within the education group when calculating average wages.
C.2 Individual Variables Definition and Construction
Education: We defined four education groups: Some High School (SHS), High School Graduates (HSG), Some College (SCO) and College Graduates (COG). All groups in each year are defined using the variable EDUCREC which was built in order to consistently reflect the variables HIGRADE and EDUC99. In particular, we define as SHS those with EDUCREC<=6, then HSG are those with EDUCREC=7, SCO are those with EDUCREC=8 and COG are those with EDUCREC=9.
Experience: Defined as potential experience, assigning to each schooling group a certain age reflecting the start of their working life; in particular, the initial working ages are: 16 years for SHS, 19 years for HSG, 21 years for SCO and 23 years for COG.
Weeks Worked in a Year: For the CPS up to1975 the variable used to define weeks worked in the last year is WKSWORK2, which defines weeks worked in intervals. We choose the median value for each interval so that we impute to individuals weeks worked in the previous year according to the following criteria:
6.5 weeks if wkswork2=1; 20 weeks if wkswork2=2; 33 weeks if wkswork2=3; 43.5 weeks if wkswork2=4;
48.5 weeks if wkswork2=5; 51 weeks if wkswork2==6.
For the CPS after 1975 we use the variable wkswork1 which records the exact number of weeks worked last year.
Hours Worked in a Week: For the CPS up to 1975 we use hrswork, which records the number of hours worked last week, and after 1975 we use uhrswork which records the exact number of hours worked in a usual week by a person.
Hours Worked in a Year: This is the measure of labor supply by an individual and it is obtained multiplying Hours Worked in a Week by Weeks Worked in a Year, as defined above.
Yearly Wages: The yearly wage in current US $ is calculated as the variable INCWAGE
Weekly Wages: The weekly wage for an individual is constructed by dividing the yearly wage as defined above by the number of weeks worked in a year, as defined above.
Hispanic: The Hispanic ethnicity indicator, used to proxy immigrants after 1970, is obtained from the variable HISPAN.