Senior Research Associate
October 27, 2021
How many children are living in poverty? How many families don’t have health insurance coverage? How do these numbers vary across racial and ethnic groups, and for different parts of the country?
In a typical year, PRB staff analyze a range of social, economic, and demographic data from the U.S. Census Bureau’s American Community Survey (ACS) to gain a better understanding of the health and well-being of children and families across states and in local communities. We share these findings with our partners, among them organizations and foundations seeking to reduce poverty or racial and ethnic inequities.
But the news that the Census Bureau won’t be releasing their standard one-year products from the 2020 ACS means that ACS data users like us will have to improvise.
One important resource is the Current Population Survey (CPS) Annual Social and Economic Supplement (ASEC), also known as the March Supplement to reflect the month when most of the data are collected. It is the official source for national poverty data and yields annual data on a variety of national social and economic measures. Although the CPS ASEC can help answer important questions about social and economic well-being, it differs from the ACS in terms of sample size, data collection methodology, and questionnaires.
- The CPS ASEC is based on a much smaller sample than the ACS (about 98,000 households annually compared with 3.5 million households) and doesn’t provide data on local communities. The Census Bureau cautions that one-year CPS ASEC estimates may be unreliable for many states and recommends using multi-year estimates of CPS ASEC data (usually two- or three-year averages) to reduce sampling error.
- Income questions in the CPS ASEC are more detailed and provide more comprehensive coverage of all potential income sources.
- Income reference periods are different. The CPS ASEC asks about all income received in the prior calendar year. So, the 2021 CPS ASEC data provided us with the first detailed information about income received in 2020 during the pandemic. The ACS asks about income in the 12 months prior to the interview.
- Health insurance reference periods also differ. In the CPS ASEC, respondents are asked about health insurance coverage during the previous calendar year, and people are considered to be uninsured if they did not have coverage at any point during the prior calendar year. So, for example, in the 2021 CPS ASEC, a person would be considered uninsured if they did not have coverage at any point during 2020. By comparison, the ACS only asks about health insurance coverage at the time of the interview.
Published Data Tables Provide Results for Many Population Groups
The Census Bureau publishes many detailed poverty tables based on CPS ASEC data, but most are only available at the national level. Table series POV-46 provides state-level estimates, but many state-level estimates based on one-year CPS data may be unreliable. Data users need to examine the provided standard errors and calculate measures of reliability (for instance, 90% confidence interval or coefficient of variation) prior to using these estimates for states.
The Census Bureau also produces health insurance estimates from the CPS ASEC, but published estimates are not available at the state level.
Microdata Can Be Used to Develop Custom Tables and Multi-Year Estimates
Users can access the CPS ASEC microdata either by working directly with the raw CPS ASEC microdata files or using an online-interactive data analysis system. The Census Bureau and IPUMS provide free access to both options. The best place to start for most people is one of the online-interactive systems because statistical programming software (such as SAS, R, and STATA) isn’t necessary to access the data.
The Census Bureau has an online tool called Microdata Access, or MDAT, that enables users to produce custom tables from ACS and CPS data. Although still in beta form, training materials are available.
To produce multi-year CPS estimates with MDAT, you will need to produce single-year estimates with numbers and then manually create a multi-year estimate by summing the numerators, summing the denominators, and then dividing the aggregate numerator by the aggregate denominator (for example, 2019+2020+2021 number in poverty / 2019+2020+2021 number in poverty universe). Note that you will not be able to obtain standard errors with this tool.
Custom tabulations can also be created using IPUMS, although these data are not usually available on the same day that the Census Bureau releases the CPS ASEC data. IPUMS offers the advantage of allowing users to create multi-year estimates in the online data analysis system. But IPUMS does make edits to the raw data for family relationship, income, and poverty estimates, so pay close attention to which variables you are choosing (the original and edited versions of many variables are available). And users will not be able to obtain correct standard errors through this data analysis system. PRB offers short videos on how to use the IPUMS system, using ACS data in the examples.
Raw CPS microdata files are available for download for use with statistical programming software. These CPS data files are also available from IPUMS. An advantage of using the files from IPUMS is that you can tailor the data download to just the years, geographies, and variables of interest, resulting in a much smaller analysis data set.