Data Source
The estimates included in the state profiles were generated from the American Community Survey (ACS) 5-Year Public Use Microdata Sample, which pools data for years 2017 through 2021. The large number of cases, over 15 million, allows for more detailed analyses of smaller populations, like those of a state or a particular racial/ethnic group.
The ACS has been administered by the U.S. Census Bureau since 2005. It replaced the “long form” version of the decennial census and collects information about Americans’ personal characteristics, family composition, employment, income, and housing that is used to allocate billions of dollars in state and federal funding.
More information on the ACS and the specific data file used for our analysis can be found here.
Analysis
Data were analyzed using Stata (versions 14 and 18). The epctile package was used to generate design-adjusted standard errors for the median earnings estimates.
The ACS allows respondents to report more than one field of college degree. The field of degree categories used in the analysis were not mutually exclusive, meaning that respondents with multiple majors were included in every pertinent field of degree category.
The ACS does not ask respondents about the field of their advanced degree(s). Keep in mind that a respondent’s advanced degree may not be in the same field as their undergraduate degree.
For most of the state profiles, the earnings and unemployment estimates are for workers aged 25-59 (unless otherwise specified). For some less populous states, it was necessary to use a wider age band to generate reliable estimates. Please see the note toward the bottom of the second page of each profile for details.
Earnings estimates are for full-time/year-round workers (those who worked 35 or more hours per week for 50 or more weeks in the previous 12 months).
A roster of the disciplines included in each of the broad field of degree categories used in the analysis is available here.
Reliability of the Estimates
Every piece of quantitative information included on the profiles is an estimate that has a measure of error (also an estimate) associated with it. This measure is called the standard error.
Every estimate included in a state’s profile has a design-adjusted standard error that was less than 30% of the estimate itself. For states with smaller populations, it was necessary to combine field of degree or occupational categories to produce estimates of acceptable reliability. In some cases, it was not possible to publish a reliable estimate (e.g., for Mississippi it was not possible to generate a reliable estimate of the share of college graduates in museum/library jobs who majored in the humanities).
Due to the error associated with the estimates, treat small differences between them with caution. They may not be statistically significant.
The standard error and 95% confidence interval for every estimate included in a state’s profile is available upon request from the Humanities Indicators (please email Program Officer Carolyn Fuqua at cfuqua@amacad.org).