Some of America's Most Important Economic Data Is Decaying

0:00Your best restaurant location gets 5 star reviews. How do you make every location like your best location? Your best paper mill has been operating at peak productivity. How do you make every mill like your best mill? Your best data center has optimized every drop of water. How do you make every data center like your best data center? The answer is Ecolab. Better performance, better outcomes, better impact. Ecolab. Now every location is your best location. Being a small business owner isn't just a career. it's a calling. Chase for Business knows how much heart and effort go into building something of your own.

0:38Manage all your business finances, from banking to payments to credit cards, all in one place with their digital tools. Plus, access online resources designed to help your business thrive. Learn more at chase.com slash business. Chase for Business. Make more of what's yours. The Chase mobile app is available for select mobile devices. Message and data rates may apply. JPMorgan Chase Bank N.A. Member FDIC. Copyright 2025 JPMorgan Chase and Company.

1:09Bloomberg Audio Studios. Podcasts. Radio. News.

1:25Hello and welcome to another episode of the Odd Lots podcast. I'm Joe Weisenthal. And I'm Tracy Alloway. Tracy, I've said this a bunch of times, and we've sort of danced around this issue on the podcast before, but I've said this a bunch of times. I'm sort of fascinated by the degree to which we sort of take numbers on the screen for granted. Like, you know, we look at a price of a stock and it exists on the screen. It's like, where did that come from? How did that get there? Or like the first Friday of the month, the jobs data flashes on the screen and we just start talking about what the data said, what the unemployment rate was, et cetera.

1:58But we don't really talk enough about like what had to happen behind the scenes to get that number to the screen. Right. I find this really interesting as well, because obviously there's the data collection portion of this. You have to go out and talk to people for certain surveys, certain data points. But also there are so many qualitative and subjective adjustments that you can make to that data. So, for instance, with CPI, I didn't know that CPI weightings are different depending on what city you're in. So, for instance, food at home could matter a lot more in, I don't know, Minneapolis compared to Chicago.

2:37It's really interesting. And there's also qualitative adjustments. So if your fridge gets Wi-Fi connectivity, then that gets incorporated into the price as well. So it's really interesting. It's super interesting. And like I have, you know, this is the issue. We don't talk about data collection, but it's obviously some of the most interesting stuff there is because, of course, you know, things like food at home in one city should be different than another city if you really want to collect a basket. And so it means that what you have is some of the most interesting economics work being done anywhere in a realm that virtually gets no attention.

3:13And when we talk about data collection, you know, you hear it a lot in politics, right? Surveys have gotten, you know, no one answers the phone. It's, you know, it's harder and harder to do high quality surveys. I think you've written about this, like response rates and just the actual cost of collection of all this data is on the rise. Yeah. So I have a bunch of thoughts on this. I'll just say, talking about the economic data might be timely as well, because we know that Trump doesn't really respect, I guess, a lot of official economic data. And he's also trying to cut back on costs or funding of a bunch of different government programs.

3:49So it's clear that, you know, entities like the Bureau of Labor Statistics could potentially be in the crosshairs here. But you're right, response rates, lower response rates have been happening for a long time. And it's really easy to look up like just how bad things have gotten. If you look at the BLS website, for instance, look at response rates on the housing portion of CPI, that's gone from about 70 % back in 2015 to just 57 % now. Response rates on jolts have gone from I think 67 % to just 30%. So that's pretty amazing. And the other interesting thing here is it's not just a US problem. There's no US exceptionalism here.

4:34You see the same pattern in other economies like the UK and New Zealand. And actually in the UK, just last month, the Office of National Statistics said it wasn't going to publish PPI because of a data quality issue. And also trade data had a problem because of errors in the data provided by HM Revenue and Customs. And now there's a task force to look at all of these mistakes and data problems at the ONS. So something that goes beyond the U.S. here. This is a really big deal, especially when you just consider how much economic activity is based on being able to look at high quality data and make decisions from it, obviously in the market, but also in just sort of the quote, real economy, etc.

5:17Anyway, without further ado, we really do have the perfect guest, someone who knows a lot about how the sausage is made and why making the sausage is getting more expensive. We are going to be speaking to Bill Beach. He was most recently the commissioner of the BLS, and he knows all about this. He was the 15th commissioner at the BLS, does research affiliated with the Economic Policy Innovation Center, and he is going to talk to us about all of these issues. Bill, thank you so much for coming on Odd Lots. Oh, man, it's great to be with you and Tracy. Thank you very much. It's a great topic right now.

5:52It's a great topic. I don't know when the process begins, but it's like, I get this number. It says what the jobs report, it says how many jobs were created that month. How did that number get onto my screen or onto bls.gov? And it gets there every month. Yeah, every month. Every month, it's never missed, except maybe once in like a hurricane or something. Anyway, how did it get there? Yeah, the First Friday report, the jobs report, which comes out at 8.30 Eastern time on the first Friday, it consists of two surveys. So let me talk about the first one, which is the one you just mentioned, the number of jobs.

6:28Right. That is a survey of about 400 ,000 firms out of 11.3 million firms in the United States. About 400 ,000 of them have agreed voluntarily to send in a pretty complex survey response every month. They send that survey to electronic collection centers. There's one in Chicago. There's one in Atlanta. And that survey has to contain the 12th day of the month, if it is a workday, or the closest workday to that 12th of November or whatever. And the reason for that is we want to get at least one pay period in the report. Okay, so they submit that data continues to be collected through almost the end of the month.

7:21Not quite, but almost the end of the month. It goes from regional offices, these regional collection centers, to the national headquarters in Washington, where about 40 people, sometimes less, out of the 2 ,000 people that work there, massage the data. Which means that they take the data, which is in the survey. It's a sample. It's just a fragment of the total population. And they multiply the responses by what are called weights. And that gives us the national number. So that number is all done pretty much by Tuesday night preceding the Friday. And on Wednesday, the commissioner gets to hear the data.

8:06The commissioner, by the way, plays no role whatsoever in massaging the data or multiplying the survey results times the weights. Then I would brief the White House. The White House would then brief the Federal Reserve Board Chairman, the Secretary of the Treasury, on Thursday, Thursday night usually. They're sworn to secrecy. And then at 8 o 'clock in the morning, I would walk over to the Department of Labor, because BLS's headquarters is about a 20-minute walk from the Department of Labor. I would brief the Secretary of Labor, and for an hour, from 8.30 to 9.30, the Secretary of Labor and his staff had to remain quiet, and then at 9.30, they could answer questions of the press.

8:54So that basically is the process for the jobs report. There's another survey there. Interrupt me at any time, but that's where we get the unemployment rates called the Household Survey, And that's a survey of 60 ,000 households selected out of a sampling frame. We can talk about that term to be representative of the entire United States by demographic segments. You know, how many people are in a certain age group, male, female, the racial and ethnic compositions and geographic locations. So it's a complex sample. That sample is done usually by Monday of the week prior to the release on Friday.

9:37And then again, I get to see the results. So the commissioner gets to see the results on Wednesday. So that's kind of the calendar. Both the household survey and the employment survey cover the same two-week period in the month. Now, there may be a day or two switch there. Sometimes the household survey has a few more days in it than the establishment survey, but they all have that 12th of the month because we're trying to get the middle of the month. By the way, that's real important because if you have a natural disaster that happens prior to the 12th or after the surveys are closed, that does not affect the results, even though you would think, since it happened in the month prior, that the hurricane might have affected employment results.

10:26Thank you.

10:56The answer is Ecolab. Better performance, better outcomes, better impact. Ecolab. Now every location is your best location. How many vendors does it take to meet all your organization's food needs? Just one. EasyCater, the workplace food platform that lets teams order from a huge variety of restaurants, over 100 ,000 nationwide, all through a single vendor. In addition to all that variety, EasyCater also gives you full visibility of your organization's food spend with invoicing, centralized reporting, and seamless integration with expense management systems, all on one platform. EasyCater, your business tool for food.

11:38To learn more, visit easycater.com slash podcast. So one thing you mentioned is companies, you know, voluntarily responding to these surveys. And one thing I always wondered about is how, I guess, onerous or resource intensive is responding to these surveys. So do I have to have like a person who's dedicated to doing this every month? Does it consume resources? How long is this going to take me? And then also, does the BLS ever try to fact check some of these self-reported data points? Those are great questions. So the survey is not just one or two questions. It takes about 30 minutes to fill it out completely.

12:24We do get responses that are incomplete and we accept those incomplete responses. The survey is usually filled out by someone who can access the company's data on wages, on total employment, on employment by supervisory or non-supervisory job functions. In other words, it's a person usually in the operations office of a company. Now, we ask companies that are very large, like the big box retailers, all the way down to really small corporations or small firms. And it becomes more and more of a difficulty for the smaller firms. And we're very cognizant of that. So BLS has reduced the scope of the survey to get it as small as possible and as free of burdensome compliance as possible.

13:14It's also an electronic survey so they can fill it out on their computer screens and hit a submit button, which greatly improved our response rate, by the way. So that survey is not so hard now to fill out, though I think it still takes about 30 minutes, maybe more if you have to dig out the data if you're not in possession of it. The household survey is a totally different beast. It is a long survey. It consists of over 120 questions. It really is hard to fill out if you kind of don't know the terminology. Let me just give you one. We asked the question on employment. are you working or have you looked for work in the past four weeks?

13:57That's the key question. We've been asking that question for since 19, I think 1943. If they answer I'm working, then they are considered employed. We'll have follow-up questions about that, about their industry, about their tasks and so forth. If they say, no, I'm not working, then we say, well, did you look for work in the past four weeks? And they'll say, well, you know, maybe I thought about it. Well, thinking about it's not enough. You can look for one day in the past four weeks and be considered in the labor force, but unemployed. So sometimes we have to have a verbal follow-up rather than they're just filling out a form.

14:38There has to be a conversation. That's why we don't do more than 60 ,000 households, because it's burdensome for the person who's the questioner. This leads right into, I think, what I was about to ask, you, which is, you know, we're having this conversation April 2025. Your term of office at the BLS ended March 2023. So you're actually under both the two recent presidents under the first Trump administration, then Biden. But let's say, okay, so let's say it's March 2023. How much costlier is that process of communicating with the households than it was, say, in 2013 or 2003? It was more costly, certainly, and that's because of inflation and how inflation affected the wage bill for the survey.

15:29The survey is taken by the Census Bureau under a contract by BLS. So we pay the Census Bureau to go out in the field and conduct the survey. Here's an idea of how much more expensive it was. In 2019, which is the first year of my term, I didn't have to find any additional funding for the household survey. It might have been just a small amount, but nothing significant. By 2023, we were having to find almost$4 million,$5 million more. So the costs of the survey – Is that per year or per month? That would be for the entire year. Okay, okay. The cost of the survey went up significantly. Now, your listeners will be thinking about the federal budget, which is in the trillions of dollars, and say, oh, that's not very much money.

16:17But the Bureau's budget has not really changed for almost a decade. And our costs have gone up progressively, especially during the period of significant inflation, which we had starting in 2021. So finding additional money is really hard. During the pandemic, I found millions of dollars of couch money, unused conference fees, unused travel fees, so I could apply that couch money, if it were, to purposes like buoying up the CPS or building a data. We built a brand new data center out in the suburb of Washington. But once we were back in business and fully using our entire budget, it was very difficult to find those dollars to fill the hole.

17:10And I think you could safely say, I've been arguing this now for over two years, our surveys, but especially the current population survey, the one that gives us the unemployment rate, the labor force data are dying. They're decaying. They're in very serious trouble. And we will have, unless we modernize that survey, we will see a time when we will be like the British, right? Unable to publish portions of it that just don't have sufficient sample for statistical release. Okay, I'm going to ask the obvious question here, but why have costs of conducting these surveys, gathering this data actually gone up?

17:48And I understand, you know, building data centers is probably an expensive process. But on the other hand, you know, going electronic instead of mailing out thousands and thousands of surveys. In theory, I would imagine maybe that saves you some money. So what exactly is causing the increase in expenses here? The problem is not with the electronic surveys. They will eventually be very costly. And there, we're suffering from a different problem, which is just public support for the electronic surveys. So the jobs report, that is that portion of it that's electronically captured by these collection centers in Chicago and Atlanta, I think those are going along all right for the time being.

18:33Our real problem is in the surveys, the data of which is collected by people, survey field survey people. They go out and they talk to the households. And there the cost is the obvious one. it is the wage bill. These people are highly skilled interviewers. They are not high school graduates, and I have nothing against high school graduates, but they are educated, very trained, oftentimes economists who will go out and speak to people and do follow-up interviews. Oh, my gosh. And they'll spend hours and hours doing this. Well, okay, fine. That's great. but it costs a lot to keep those people employed.

19:19They have other opportunities, so we have to have competitive wages. The wage bill is driving that. Collection costs are a little bit higher, that is the processing costs, but the big part of that is the wage bill. And the wage bill is not going to go away. That's just going to continue to go up. So we need to do with the household survey, what we have done with the establishment survey, the firm survey, and that is modernize it so it is more electronically collected. And then we also need to integrate data which can be obtained through the internet on households. Households maybe go onto a platform with a tablet or something and supply, you know, maybe we go to a million households and they supply five or six questions.

20:05And we combine that or blend it with person-to-person collected data. That's what we call modernization. It may not sound so innovative to your listeners, but it is very innovative for the statistical system. Unless we do that, the cost will continue to rise. The response rates, by the way, are continuing to decline. So that's public support. and we will be, we just won't have these surveys in the future. They're too expensive. I want to get to the modernization in a second, but just talk to us about, I mean, I think you said the survey's dying, which is pretty dramatic given the centrality of this.

20:47Between the response rates, the increased wage bill, et cetera, how severe is this crisis? How much time are we talking about? Because I imagine it gets harder and harder to find that couch change to keep it going. Yeah, actually, we're out of time. We have to cut down on the sample. And I'll give you the evidence. My successor, the current commissioner of labor statistics, announced at the end of last year that we would have to reduce the sample in the CPS, that's the household survey, by 5 ,000 households in order to just publish the rest of the sample. Okay, by cutting back on the sample, you cut back on publishing details in the current report.

21:36You may not have enough to publish on teenagers anymore. On certain demographic groups, you may not be able to publish details on certain age intervals, like everyone between the ages of 55 and 60 or 65 and 75. it. Every time you reduce the sample or your response rate falls and it kind of reduces it for you, we can talk about that next. What suffers are the details. Let's go to the CPI, which BLS also does. The CPI is a very big survey, very important survey, but it is oftentimes the case that we don't have details for certain parts of the basket of goods. And so we will average across the months and publish for another month based on averages of prior months.

22:28But then we run out of the integrity of that average and we can't publish the bread price or something like that. That happens more and more because we're getting less and less cooperation by retailers that to allow our surveyors into their stores and to go around grabbing the potato chip bag and counting the number of potato chips in the back. So we've had to innovate on the CPI, on the producer price index, on the import-export price indexes. I hope you haven't noticed the innovations because you're not supposed to, but those innovations have absolutely saved the import and export prices. We had response rates down in the 20 percentiles, completely unpublishable detail.

23:16During the time I was commissioner, we went to using data from the Commerce Department that is collected by the customs officials, and it's working out just great. So we completely left the surveys there. The PPI has problems just like the British are having now, not as bad. And we're using more and more data from private companies who will give us their data from the internet where we collect that data and we combine that with the survey data to keep the things alive. On the CPI, it's very interesting, we get all of our retail gas prices now from a private company that aggregates gas prices for the retail gasoline industry.

24:00And it's a very good source. A lot of our housing prices, now you mentioned housing, we're going to more and more to aggregators, private aggregators. So there are strategies for saving the survey by blending in private data with the survey data. That is not the case with the household survey of the labor force. Nobody really collects that data except BLS. And that's why these demographic surveys, census has a bunch of them, health, the health department, the department of health has a lot, they're in very serious trouble. So I take the point about innovations, but one of the things that's been happening recently is we have been seeing bigger and bigger revisions in a lot of the data.

24:46Is that because of the response issue? So more late responses mean that you're going to get revisions later on as those responses are incorporated. And the reason I ask is I remember speaking to someone at the BLS and here, shout out to the people working at the BLS because you can actually just call them and ask questions. They're really, really responsive. Amazingly, like the only government department that I really know where they'll just pick up the phone and talk to you about methodology. Anyway, I was asking about revisions to PPI for mayonnaise of all things. And the data there had been revised from like 5 % to 10%, which is a pretty big change in the space of a couple months.

25:31And he was talking about how the PPI is subject to revision up to four months after it's published, and it gets updated as new replies come in. So I imagine if response rates are going lower, then maybe that's contributing to some of the revision issue. By and large, our issues have to do with response rate. You're absolutely right. Right. And before I go further, I'm very, very happy you said that about BLS because BLS prides itself in being responsive and transparent, by the way. So that's a good thing. So I think there are two sources of this problem. The first is really hidden and no one talks about it.

26:08So let me just mention when we refresh a survey and you have to do this because people are asked to give their survey responses only for a few months. Right. And then they drop out and you have to find new people. The regional offices, there are six of them in the United States for the BLS, they have people who go out and do what's called initiate a new survey respondent. It's getting more and more difficult now to initiate that respondent. People are saying, we just don't have time. Oh, my gosh, there's so many important things to do. We don't want to give you our data because if we give you our data, then the IRS is going to be after us or we're going to have some kind of OSHA.

26:47You know, we don't, BLS's data is completely protected from the law enforcement aspects of the federal government. It's absolutely only for statistical purposes. That does not prevent people saying, oh, it's going to happen, you know. So if your initiations are dropping and it's harder and harder, no wonder the response rates are dropping. Because people are less enthusiastic. Even if they say, yeah, I'd love to be a part of it. they're not as enthusiastic as they were a generation ago. So the real problem starts at the initiation level. And we're seeing that not only in the household survey and in the price area, but we're also seeing that in another survey we haven't talked about, the National Compensation Survey, which is a wonderful survey on how much people are getting in union houses, union households and non-union households in Northeast and Southwest is terrific about that.

27:46The JOLTS report, the job openings and labor turnover survey that you mentioned at the opening, its response rate has fallen dramatically. And it's largely because people are less enthusiastic generally about participating in government surveys. It's going to be hard to stem that tide. And I think the only way we can do it is by going to these platformed surveys, that is surveys that use the Internet as the main platform, integrating a lot of private data and being highly original in the way we think about preparing a statistic for public use. We're just not going to be facing a different world overnight of really happy people wanting to give all their data to the government.