A Random Invitation: The American Community Survey
Earlier this week I opened my mail to find an invitation—not to go to a party, but rather to be a participant in the American Community Survey (ACS). As a sociologist, this was exciting since I have used ACS data in my research and teaching for many years. Now I will get to be a part of the process from the inside.
The ACS is a survey conducted by the U.S. Census Bureau each year to learn more about the American population between the census collections, which take place every ten years. Not only does the survey provide data on population changes, it also provides us with annual data on marital status, housing, education, and income.
While by law the census studies the entire population—or at least attempts to include the entire population—the ACS is a smaller sample of approximately 3.5 million Americans, randomly selected by home address. According to the Census’s population clock site, the U.S. population is just over 314 million, making the ACS survey a sample of just over 1 percent of the overall population.
This might sound miniscule, but using random sampling methods (where everyone in a population theoretically has the same chance at being selected to participate) even a small sample can give us a great deal of useful information. Social scientists who analyze the data also use weighting techniques to match the sample to some of the demographic characteristics of the population to ensure that the sample best reflects the country’s racial and ethnic composition.
Of course, people who have no home address get left out of a survey like this, so there is a built-in selection bias preventing us from learning more about individuals who are homeless. And those who are included may choose not to participate, although technically the law requires those selected to participate. People who are very busy or those who do not understand the instructions may simply fail to complete the survey. If one group or type of respondent is more prone to skip the survey than others, this too could skew the results.
Earlier this year, the House of Representatives voted to eliminate the ACS, arguing that it costs too much money (an estimated $240 million per year out of a budget of about $3.6 trillion) and that the questions about race, age, and income are intrusive. The bill has not passed in the Senate so far, and a petition has been drafted to save the ACS.
Without the ACS, we would lose valuable updates about poverty, about who has health insurance, and about population and transportation shifts that contribute to where the federal government spends money on infrastructure like roads and bridges. Economists expressed concern that businesses would lose valuable customer data as well. (The ACS does not provide individual names, addresses or responses. Disclosure of any individual’s personal information can lead to five years in prison, a $250,000 fine, or both).
One of the nice things about federally collected data is that we all have access to the results, which tell us quite a bit about large-scale patterns (and again, no personal information). I encourage all sociology students—and voters—to actually go to the ACS website and see what kind of data are created from the survey before Congress eliminates it altogether.
Why wouldn’t we want to know how many people live in poverty or lack health care coverage, or be able to track migration patterns? I’ll be filling out my survey when it arrives.