The 2020 count is not the only major program the Census Bureau has had to delay

Best listening experience is on Chrome, Firefox or Safari. Subscribe to Federal Drive’s daily audio interviews on Apple Podcasts or PodcastOne

Now that the 2020 decennial count is more or less in the rearview mirror, the Census Bureau has a new concern. The five-year American Community Survey will also be late, with the pandemic to blame as well. Federal Drive with Tom Temin talked about all this with Census Project co-director Howard Fienberg, who said Congress needs to bolster the Census Bureau with funds to modernize its data infrastructure and enhance the vital American Community Survey.

Interview transcript:

Tom Temin: Mr. Feinberg, good to have you on.

Howard Fienberg: Thanks Tom. It’s great to be here.

Tom Temin: And tell us about the American Community Survey. I said five year but it’s actually an annual event in which they make five year projections of what’s going on. Tell us what the survey is about and why it’s so important to players in the economy.

Howard Fienberg: So the American Community Survey, the ACS, used to be known as the Census long-form, it was the gajillion questions that a small sample of the US population would receive every 10 years. At the same time as we were doing the decennial headcount in the 2000s, that got switched to the ACS so that you could get a rolling sample of the population every year, in hopes of getting more up to date data. Instead of having to rely on this snapshot in time from every 10 years, you’re basically hitting the same amount of the population, about 10% of the population over the course of 10 years. And the reason we are talking about five-year estimates is that as you drill down on small sub populations, groups that are disparate, Native Americans, certain ethnic subgroups, or if you’re talking about remote and rural areas, it’s very difficult to get a good enough sample to be able to have accurate data without averaging across several years of time. So in this case, the Census Bureau focuses on a five average across five years, in order to put together the most accurate sample they can for a lot of those rural areas and smaller populations.

Tom Temin: And businesses, organizations use this data for marketing purposes for planning purposes. I mean, it’s a widely used survey for planning, isn’t it, across the economy?

Howard Fienberg: Yes. So government and the private sector, both are heavy users of ACS data. Most aspects of the federal government funding that comes from the federal government, 1.5 trillion or so every year, most of that is guided by the ACS data. So just there alone, you can see a pretty large impact that includes everything like Medicare and veterans funding, housing funding, all the rest. And that’s just on the government side, on the private sector side, you’re looking at a good portion of the decisions that are made by most companies in this country drawing back to ACS data, because that’s the most accurate and most up to date data that you can draw from. So for example, if you have a member of… my other hat that I wear at the Insights Association, a marketing research company would be drawing on the ACS data on almost daily basis to update how they’re putting together a statistical sample to run a survey. And that’s going to inform how Procter and Gamble is going to be marketing its next cereal, or how ESPN is going to be setting up to advertise for its next new series.

Tom Temin: And it’s almost hard to imagine how they can do sampling. Because I’m thinking of a road that’s about seven or eight miles long, it goes through a couple of different zip codes, a road up in the Northeast that I’ve been visiting and riding on the past few weeks with some frequency, and this road, again it’s a suburban road near a large demographic, statistical metropolitan population area, whatever they call it, and you pass one house that is a little five room shanty from the 1920s. And then 100 yards later, there’s an entrance to a driveway where someone put up a gorgeous 15 room mansion, literally with the three car garage. So what do you sampling if you go on that road? How does the Census Bureau manage that and what do they need to do to enhance the sampling to make sure that they capture all the data they need to?

Howard Fienberg: Big piece of that is increasing the size of that sample, because we are talking about something that goes on every year, so it will help capture that kind of change over time. But the more households you’re able to hit in a given year, the better the quality of the data on the other side. So rapidly increasing the sample size is something that we’ve been hitting on pretty heavily. It wouldn’t actually cost that much money and the Bureau of themselves estimated that it would be about 45 million just to up the sample size by a million households. There are a lot of things that need to be done behind the scenes to bolster how the ACS is conducted. Some of that is just on the side of how they build the sample. Inside the bureau, they have these multiple different things they call frames. And when they’re running any one of their surveys of which they have a whole lot. they set up their sample using these different frames. And they don’t check against them or between them, so they’re not always building on the most up to date data when they’re reaching out to a house. So part of the things they can do behind the scenes would increase just their own efficiency, using their own data to reach out to American households so they can actually reach more people more effectively.

Tom Temin: We’re speaking with Howard Feinberg, he’s co-director of the Census Project. And the project has written to the Senate and House appropriators asking for this funding. And you mentioned that one of the things they should do is modernize the Bureau’s data infrastructure. Is that what you meant by the frames, and these frames are datasets that they use to build samples?

Howard Fienberg: Absolutely. I mean, for any set of any big data operation, you’re looking at overlapping data sets, a lot of duplications. We’re talking about trying to bring the bureau into 21st century data infrastructure. And these are the sorts of things that the ability to draw across multiple databases and integrate them effectively without messing up your data. It’s very difficult. It’s extremely difficult for a government agency to deal with this massive amount of data and do that effectively, while not corrupting the data. It’s a normal practice in the private sector, but even there, it’s not something that you do.

Tom Temin: Got it. And when they do this survey, the American Community Survey, is it households or is it businesses that are being polled.

Howard Fienberg: It is households, but part of the initiative to improve their infrastructure would would be breaking down the barriers between these different frames they use because there’s a huge amount of overlap between their business sample frame and the household frame, because so many businesses at this point are being run out of people’s houses.

Tom Temin: And do the frames correspond with geographical areas?

Howard Fienberg: They do. They’re geographic areas, and a lot of different crosscurrent demographics.

Tom Temin: And they then have to sample more than one household within a frame, because I’m sure even in rural Alaska, they’ll find one house where they have the 15 year old Corolla outside rusting on buckets, and then up the street there’s the guy with the Cadillac. You find this situation pretty much everywhere.

Howard Fienberg: Absolutely. And the more the more you can hit inside of that community, the greater the accuracy of the data that you’re going to be getting out of it. And admittedly, the ACS does not get perfect responses. They’re pretty close. It is the most, the biggest response rate for any survey short of the decennial headcount.

Tom Temin: Sure. What are some of the questions they ask? Give us a sense of what they’re driving at with the survey.

Howard Fienberg: It’s very basic things about income, about transportation and commute, housing. They used to have some very archaic questions about the number of toilets in the house, race and ethnicity. Those questions keep getting more and more complex, huge series of questions. Actually just responded to the ACS myself, for the first time ever. My wife was bowled over by the variety of intrusive questions. And sometimes they’re asking several questions to try to drill down on one point. The point about income is asked across, I think it was about four or five different questions, try to really tease out an exact figure.

Tom Temin: So they know what they’re doing with respect to constructing the survey. It sounds like a complex exercise.

Howard Fienberg: Yes. They spend years before they put a new question on there usually. There’s tons of testing outside of the ACS, just for a question that will ultimately go into the ACS. Then the ACS becomes a testbed for questions for every other survey, eventually.

Tom Temin: I’m disappointed, I’ve never gotten one to my house, which has four toilets by the way, in case anyone wants to know. So basically, you’re arguing for a larger sample within the frames, and that costs money. And then you’re arguing for a modernized data infrastructure such that they can better integrate data sets and also protect them and keep the integrity of the data sets. Is that fundamentally what you’re asking for?

Howard Fienberg: Yes.

Tom Temin: What about the Bureau, is it your sense that they feel the same way?

Howard Fienberg: I can’t speak to everybody’s excitement about increasing the sample size. But certainly, I think the Bureau staff from top to bottom are pretty excited about the frames initiative, and a lot of the efforts internally to improve their data infrastructure. And this does also key in to some of the other initiatives they have going like these pulse surveys that they’re doing of small businesses, and the population, which are really rapidly designed and fielded surveys that are really designed the same way you would run any survey in the private sector. But it’s a ridiculous amount of nimbleness to get out of a government agency. It’s not something I’m used to ever seeing.

Tom Temin: And what’s the survey medium? I mean, they did discover how to enable people safely and uniquely to send in their data for the 2020 count using the internet. That’s how I did it, got the unique number in the mail and so forth. Is that how they do the American Community Survey or is that still paper based?

Howard Fienberg: It is primarily online. You can do it on paper, and I did receive because I did not respond right away, well slow to respond. I was one of the people that was given all sorts of reminders by mail and I received a phone call or two about it. So you’re contacted about it multiple ways. The preference is for you to do it online. And frankly, it is much easier just to respond online if you have the opportunity.

Tom Temin: And of course your release that you sent in which you mentioned going to the appropriators was predicated on the fact that the ACS, the American Survey was delayed this year because of pandemic. How does that tie into the issue?

Howard Fienberg: It’s a big deal. As you said, we talked about the importance of the five year estimates to be able to get reasonable and accurate data for a rural area, remote areas, certain subpopulations, ethnic groups. And a delay in the release of those five year estimates means that people are left wanting and trying to put together, just being able to do proper surveys of those populations themselves in the private sector. Government is not up to date on being able to understand what’s happening in those populations, how best to serve them. Decisions that need to be made in rural areas, a lot of those are probably going to be held off. Investments not going to be made and businesses are not going to open their or businesses are going to leave their for a lack of data.

Tom Temin: Howard Feinberg is co-director of the Census Project. Let’s hope they’re listening. Thanks so much for joining me.

Howard Fienberg: My pleasure, Tom. Thanks for having me.

Related Stories

Comments