Bureau of Labor Statistics may not have accurate data for feds benefits

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

Several big federal benefits programs depend on the consumer price index. The CPIs are calculated by the Bureau of Labor Statistics. But the indexes may not be entirely accurate. The Government Accountability Office found that the bureau has not evaluated whether the data sources it feeds into the indexes produce accurate ones. For more details, the GAO’s Director of Education, Workforce and Income Security, Charles Jeszeck joined Federal Drive with Tom Temin.

Interview transcript:

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

Charles Jeszeck: Nice to be here. Thanks Tom.

Tom Temin: Tell us first of all the importance of these consumer price indices, I guess there’s more than one. And they apply to a variety of federal benefits programs. What are some of those?

Charles Jeszeck: Well, the Bureau of Labor Statistics in the United States calculates consumer price index, there are a number of different indexes — pretty much every advanced industrialized country does this to get a handle of what’s going on in the economy in sort of the standard of living for workers. The BLS United States, which is actually one of the premier statistical agencies in the world, conducted what they called their headline index, the CPIU, which is the consumer price index for all urban consumers. From that they derive several sub population indexes. And we have a lot of acronyms. We have the CPIW the CPIE, the CPIW in particular is the one that is used by social security to calculate benefit increases for recipients.

Tom Temin: Got it. And what is the issue that you discovered with that particular index?

Charles Jeszeck: It’s not just the CPIW, it’s also the CPIW, which is an experimental index that BLS uses to estimate the cost of living for people over age 62. And what we found here and this is from BLS officials and from documentation is that they’re increasingly unsure whether the data sources they use to calculate these indexes are accurate. And this is in terms of whether the outlets where people shop, whether they’re using the correct ones, whether they’re using the right prices, and weather the mix of goods and services that people use in these sub populations, whether they have accurate estimates there,

Tom Temin: What are the data sources? Do they come from retailers or what are the sources they use?

Charles Jeszeck: Well there’s several sources. The first thing that BLS does is they conduct a survey, the Consumer Expenditure survey, they get this data. And it’s basically what they buy. It’s a 65,000 person sample each quarter over two years. That’s where they get the market baskets. Then what they do with a sample of outlets, store outlets, and also they get rents from a survey, they calculate the prices and they put them together. So you have the weights, what people buy, and then the prices that presumably they’re paying. From that they calculate their headline index, which is the CPIU, and then for the sub population indexes, they for example, for the CPIE they would focus on people over age 62 and they would adjust the weights to figure out the difference in their market basket. So for example, for the CPIE, we know older people tend to use more health insurance or health care, and so they will adjust the weights appropriately.

Tom Temin: So what’s the problem?

Charles Jeszeck: Well the problem is a number of things. First, the sample sizes declining. People aren’t responding. The groups, for example, the CPIW, which focuses on urban wage earners, the number of people in that demographic is declining. So you have fewer people who are filling these things out. The other thing is that the data are often four years old, can be up to four years old, it takes them two years to collect the data, then they have to clean it up, process it. And so that lag increasingly is perceived to be a problem because the world is changing pretty rapidly. So for example, look at what’s happened in the last few months with pandemic, the data for these subpopulations is going to be from four years ago. Are they really picking up the changes and is it really accurate reflecting what people are buying right now? BLS is unsure, it’s not clear whether there is an error here — but there’s more and more reason to suspect that some of the way they’re collecting data, the way that their sample sizes might lead to errors in their calculations.

Tom Temin: So it sounds like the data sources aren’t the problem so much as the way they process it and the fact that it takes them so long to match up what they’ve gathered with usage of it.

Charles Jeszeck: Well, the lag is certainly a problem, but there’s also issues of whether they know for example, where older people shop. We know for example the growth of retail of shopping on the internet, whether they’re taking into account prices in that regard. They’re using surveys where people they’re getting hardcopy. So there’s a question of whether there should be more of an effort to collect prices off the internet, what so called web scraping, whether they should get prices from third parties, presumably somebody like Amazon or Walmart to get prices from them to make things more accurate. A lot of those things that they don’t do currently that we know actually other countries have statistical agencies and other countries are doing right now. The other issue is that there are other sources of data within the federal government, for example, in the Department of Commerce, they collect the national accounts data that could lend itself to again increased accuracy, something which they really haven’t explored fully.

Tom Temin: Got it. So your recommendations then are what and do they have the means to go after them?

Charles Jeszeck: Well, I think there are two recommendations that we made. The first one is that they should explore the use of other data particularly from other government sources, for example, the National Accounts Data, and see whether they can incorporate it into their into their methodologies. Now, they already do get some data from other agencies, for example, they get data from the Department of Energy on electrical prices. They get data from the Center for Medicare and Medicaid Services on adult service prices. So looking at the national accounts and seeing where they could use that, which is calculated from IRS data and a whole bunch of other surveys, could conceivably make their their estimates more accurate. That’s one recommendation. The other one is also to reach out further, they do have an advisory council, but to reach out further to other experts, to other nations, statistical agencies for practices that they’re using, for example, web scraping, use of other government data, third party price data, to see how they can incorporate it more fully into their methodologies. They’re doing some of this now, we think though that they could do more — and frankly, in that regard, working with other institutions, the Bureau of Labor Statistics agreed that they thought that was a good idea.

Tom Temin: I was gonna say they did agree with you and I guess the question becomes, because this was in your report, resources.

Charles Jeszeck: Resources are always a challenge, particularly during these times. I think at some point, policymakers have to decide, look closely at what the appropriate level of resources are. The recommendations that we made, at least simply looking at consulting other stakeholders, exploring the use of other government data, at face value should not be very expensive. Those are some of the things that they already do, it would not really change their current practices to that great extent.

Tom Temin: Charles Jeszeck is Director of Education, Workforce and Income Security issues at the GAO. Thanks so much for joining me.

Charles Jeszeck: Thank you.

Check out the report here.