Home / Population and society / Articles / What low response rates mean for 2020 ACS data — and a potential bright side

The yearly American Communities Survey (ACS) is a vital tool for understanding the US population. State and local governments use it. Businesses use it, and federal entities and mandates like the Department of Housing and Human Services (HUD) and Veterans' Affairs use it to ascertain how Americans live and work — and how to best support them. The data informs Title I funding for students from low-income families, the list goes on and on.

However, low response rates mean the one-year estimates for the 2020 ACS, scheduled for this fall, won't be released. That means it'll be challenging to understand how Americans live — and could affect how US governments distribute billions of dollars.

The ACS is one of the nation's most robust data portraits

The ACS is the premier source of information on the US population, housing, and the workforce. It reveals where people are living, if they rent or own their home, the value of their home, and the kind of conditions in which they live. What's their marital/education/citizenship status? Do they have health insurance? What people spend on gas, if they carpool and how long they commute, their working hours, how many people they live with, the ACS measures all of this and more.

The 3-million-person sample size makes the ACS the most comprehensive data product produced by the US government outside of the decennial census. And it's mandatory, with nonresponses potentially leading to fines up to $5,000.

Low responses and an inability to follow up skewed the 2020 ACS

Normally, ACS data is collected through mailed surveys (though there are also options to fill it out online). Nonrespondents are followed up with in person, including visits to dormitories, prisons, and nursing homes. The pandemic limited opportunities to follow up between March and September 2020.

The Census Bureau found high nonresponse rates last year, particularly among people with lower income, lower educational attainment, and people less likely to own their home. This skewed the reported data so much that it couldn’t be corrected by the Bureau's traditional methods of accounting for population differences in these underlying populations.

After the significant changes brought on in 2020, it would be helpful to have a clearer picture of how those changes impacted people nationwide. How many people lost their employer-based health insurance? Or with eviction moratoriums expiring in some places and municipalities issuing extensions in others, one might want to know the percentage of housing units that are renter-occupied versus owned, the median rent, and how much of people's income goes to rent. The ACS could have provided neighborhood-level insight into this.

Withholding this data reinforces the trustworthiness of the ACS

While the absence of the one-year estimates is a loss, it's heartening to know that the Census Bureau would rather withhold these results rather than release heavily biased, skewed data. The Bureau's choice displays a high bar for quality. They made the hard choice, but this is good data governance.

It's unclear what comes next

The Census Bureau has yet to announce its plan for five-year estimates. Five-year estimates include data on census tracts and even block groups. But if the Bureau doesn't release this five-year estimate, good local data might not be available until 2025.

It's also unclear how this will alter the way governments distribute funds. If tax dollars given to programs are allocated based on old data, will the money serve the right people? If the needs of the population those funds are supposed to help have changed, that won't be apparent when working from old data.

Other surveys, like the Current Population Survey, put out jointly by the Census Bureau and the Bureau of Labor Statistics (BLS), only have data for states or large municipalities. Governments need more granular data (it can get down to the neighborhood level in the ACS) to serve diverse populations of different locales, races/ethnicities, and employment levels.

Plus, other surveys rely on ACS data to design sampling and weight responses. The BLS uses it to update the Consumer Price Index, including improvements to continuously update its housing survey sample.

It's also used to measure the criminal victimization of people with developmental disabilities in conjunction with the National Crime Victimization Survey done by the Bureau of Justice Statistics.

It's not just the low response rates the Bureau has to contend with, but the pattern of who isn’t responding to the survey. These dropping response rates could signal a more significant concern than losing one year of ACS data, with potential ramifications for other data from the Department of Housing and Human Services, the Education Department, the Labor Department, and more. Nonresponse is a growing trend. The government owes its people surveys to be administered in a method so that all required participants can answer, and participants owe timely, accurate responses as a matter of civic duty.