View/Download Dataset

The data come from the Public Use Microdata Sample (PUMS) of the 2013-2017 5 year American Community Survey. PUMS data contains individual responses from the American Community Survey and the 5 year dataset has a large enough sample to produce reliable estimates for geographic areas of any size. The data dictionary  produced by the Census Bureau with detailed descriptions of each variable in the data can be found here.

Since the visualization is related to employment related metrics, we restricted the sample to only those who were employed and those above age 16. Hourly wages were calculated by dividing the total wages or salary income earned in the past 12 months (WAGP) by the number of weeks worked during past 12 months obtained through a recode of the variable (WKW), and then divided by the number of hours people worked in a usual week in that year (WKHP).

Wage gaps were estimated using regression methods. The overall wage gap is the state average difference in hourly wages between the focal group and a comparison group (e.g. white workers relative to Latino workers). To estimate wage gaps, we used 2013-2017 ACS PUMS 5-year estimates, limiting the dataset to civilians engaged in paid labor at the time of the survey and only to those between the ages of 16 and 65. Wages were calculated by dividing annual income in 2017 dollars by the number of hours the respondent worked in the past 12 months. For all calculations, the natural logarithm of wages was used to control for skewness in the wage distribution. To obtain state specific wage gaps, we performed 51 regressions, one for each State plus the District of Columbia. By estimating separate state regressions for each industry, we allow each state to have state specific coefficients.

The overall gap is calculated directly from the American Community Survey. The gaps net of education and experience, are estimated in a statistical model that controls for individual education and experience. The gaps net of occupation and industry further control for the types of firms and job people have been hired into.