Census Bureau scrambling to wrap up 2020 count, verify data

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

The Census Bureau has landed in something of an irresistible force, immovable object situation. Congress did not delay the deadline for delivering population and apportionment information, but the Bureau did get a late start on the 2020 count because of the pandemic. And they were sort of counting on a later deadline, so now it’s scrambling. For an assessment, the Director for Strategic Issues at the Government Accountability Office, Chris Mihm, joined Federal Drive with Tom Temin.

Interview transcript:


Tom Temin: Chris, good to have you back. And so what is the situation here? Are they going to make it in your assessment?

Chris Mihm: The short answer is they will make it, it’s at what cost. And that cost is not just financial, it’s in terms of the quality, the completeness, the accuracy of the count. As you had mentioned is that they had been planning all day, the Census Bureau had been planning for an additional four months, they’ve been expecting to get the relief from Congress. That has not been forthcoming. And so the administration announced at the beginning of August that they were rolling back dates, dates, actually to the original statutory deadline, but that means they have to get out of the field a month earlier. And then they have to do a lot of data processing in a much shorter timeframe than they had expected.

Tom Temin: Right. So the field work takes the time that it normally takes. And they’ve been plagued not only by a shorter period of time in the field as a result of all this, but also I’ve been watching that they have had a higher than they expected attrition rate for the enumerators, which means they have to keep shoveling coal at a fire that they need to keep lit.

Chris Mihm: Yeah, and this is an enormous workload. They’ve got about 56 million households that they were going to be following up on. The good news is, is that at least in the first weeks of the follow up operation, they’re doing better than they had expected in terms of the goals for how those households, the coverage that they would be making. The challenge, of course, is that a lot of the time and follow up is spent on that last several percentage points of the population, and that’s where they’re really going to be very much pressed. You’re right on the turnover. They typically expect about a 10% turnover rate of people coming into training and then afterwards say no this isn’t for me. They’re running over 30% now and that’s putting a lot of stress on the recruitment, on their staffing operations. So they’ve got some real challenges there.

Tom Temin: And probably that attrition is because of the pandemic also.

Chris Mihm: It’s a combination and certainly it’s because of the pandemic people coming in and finding notwithstanding that they’re getting PPE and being told that you don’t go into the house to do the enumeration, you stay six foot apart on the on the front porch as the case may be. People say, no, this just isn’t for me. And in some cases, and this has been traditionally, people come in, take the training and say, I’m just not comfortable period and going and knocking on doors asking people for personal information that they’ve already decided not to provide to the government through answering either in the mail or now the internet.

Tom Temin: And outside of getting those people in and recruited and trained and so on — what are some of the other costs that they would incur to get this done in this new compressed timetable?

Chris Mihm: Well I think that one of the biggest ones, and usually it’s not considered an issue because it flies below the radar screen, and that is the time to do the processing at the back end. They’re going for basically what the original plan was, and traditionally about 150 days, now to about 90 days. And what they do once they get all the Census data in, is they look for duplications, they look for anomalies to tell them that they may be making an error, or there may be a problem in places with the data. Some of that can be done at the same time with various files, some of that has to be done sequentially. So there’s only so much speeding up that you can do. The real risk, of course, is that they see a problem and don’t have the opportunity to step back and really investigate it. Or even worse, that they don’t see a problem and we don’t find out about it for quite a while in future.

Tom Temin: So therefore, they could just make a report based on the data they have. And they do have a big stack of punch cards, that’s for sure. But then it would simply have errors that they could quantify, or would we know, I mean, how would you know it’s bad data?

Chris Mihm: Well, Tom, you and I are both old timers, they don’t use punch cards anymore. But you and I know exactly.

Tom Temin: That was 1920 I guess, not 2020.

Chris Mihm: Every Census has an undercount, it has Census errors in there. As diverse as the population is it’s impossible to capture everyone. The good thing is is that each census so over the last several of them has gotten better. And certainly the differential undercount. The difference in the undercounts among different population groups has also decreased. What we run the risk of this time is that we won’t continue to make historical progress on that and because political representation, and tens of billions of dollars in each year in federal financial aid are dependent upon an accurate Census, getting it wrong, or getting it wrong in a differential sense, has really real consequences for the society.

Tom Temin: Yes, because apportionment takes a pretty big shift to say take away a Congress seat from one state and add it to another. Could the differential in the error rate be such that a district could wrongly lose or gain a member of Congress?

Chris Mihm: I’m not sure. You’re right, it would take a pretty big error rate to get us to that. Probably where it would be more likely to play out as in redistricting data that the states will use in order to redraw their own lines, or in the distribution of federal funds. Both of those have big consequences as well, obviously, for representation.

Tom Temin: And so there are steps they can take, I guess. You have some fresh recommendations. I know you’ve got sort of a rolling set of recommendations over 10 years for census, since it’s something GAO looks at periodically. Are there any fresh recommendations for this, well, I’ll call it emergency situation that they’re in.

Chris Mihm: The biggest thing that we want them to do now is that as they’re streamlining operations, and inevitably, they have to do that, whether it be field operations, or more importantly, some of the data quality and data review operations at the back end. We want them to make sure that those are evidence based decisions that they’re making based on data that they understand the trade offs that they’re making. We’ve been told by the Bureau that they’re certainly doing that. It’s not as though I don’t trust them on this. But I haven’t seen a lot of the documentation that we need to see to give us the full assurance that those trade offs in the low priority areas are the ones that are actually being cut.

Tom Temin: Also the problem areas they have are not evenly divided statistically across the country, they do have some states almost completely wrapped up and ready to go. So it’s a localized problem, isn’t it the lack of full count and the ensuing data processing problems?

Chris Mihm: Absolutely. On the full count aspect of that. And in fact, the Census, as you said, is a national undertaking, but it’s radically local in terms of its implementation. And so, not surprisingly, areas that have the highest workload still to complete are also among those areas that can have the biggest staffing problems. And so it’s a real bad, perfect storm. And then when you put on top of that, the situation with Hurricane Laura, or with the wildfires out in California, they’re not used to taking a census for example, during peak hurricane season. I mean, that’s usually when everything is all kind of done and dusted, and they’re doing the data analysis at that point. They’re actually in the field now trying to find people that had been displaced. And then you put on top of that, obviously COVID, which is a huge kind of national concern on that but is playing out differently in different local areas across the country. It really shows the challenges that they’re up against.

Tom Temin: Could they outsource some of the data processing or could they hire the Energy Department and some of the supercomputers, maybe they do now, I don’t know, to somehow speed that aspect of it up so that they get the quality they need and the error checking they need just faster with parallel processing by several different entities besides the bureau itself.

Chris Mihm: The biggest challenge that we’ve seen with them is not so much the the technical or the computer capacity in order to pull that off, we’re quite comfortable that the bureau can pull that off. Although there are some additional operational tests on key systems that they need to make. It’s more just having the time for their professional staff. And these are long serving career officials that have been involved in multiple Censuses to be able to have the time to step back and look at the data and say, how are we doing here? Do we see any anomalies to have the opportunity to duplicate places where people may have been counted twice? It’s really the time to do that essential kind of analytic work is what they need.

Tom Temin: And for the remaining field work would have helped them to spend the money they need just to keep public awareness up. They have a head a really rigorous advertising campaign. If you surf the web, you see all kinds of pop ups from the Census Bureau. Should they just pour that on to so people are receptive in the last few days they’ve got?

Chris Mihm: Yeah, that has been very helpful to them, as you point out is that they but you know, this was when they came out of the field or didn’t go into the field as a result of COVID back in the spring. They didn’t stop the publicity and outreach campaign, they kept going on that, and now a lot of the focus is on two things. Go ahead and respond and if you don’t respond the person that comes to your door will be safe, it’ll be six feet and they won’t be asking to come into your house and all that sort. They’ve also tried to revitalize a number of the partnerships they have with local communities, especially in areas where there’s undercounts. Having a trusted community voice advocate for the Census and build support is still vital and so they’re still working on those. That’s still a key part. So one important message for everyone is, if you haven’t filled out your form, there’s still time go on the web and do that. That’s how they also get better data is having people actually respond on themselves rather than having to knock on the door.

Tom Temin: Sure. So it sounds like perhaps their biggest challenge is not so much delivering one count that they can support, but one that will be supported by Congress and people perceiving this thing.

Chris Mihm: That is one that is seen and understood and believed to be accurate, not just nationally, but everywhere with equivalent levels of accuracy. At the state level and local levels, that it’s not just the top line count that matters, it’s as you get down into local communities.

Tom Temin: Chris Mihm director for strategic issues at the Government Accountability Office. As always, thanks so much.

Chris Mihm: My great pleasure Tom, take care.