‘You really can just fire the BLS commissioner’

Good morning. The derivatives exchange CME yesterday launched a product to ease participation in the Treasuries basis trade, in which investors arbitrage the tiny price differences between Treasury bonds and Treasury futures. CME says it will make the trade less risky by increasing transparency and liquidity. But it will also invite a whole lot of smaller investors to engage in a trade that central banks and regulators have warned could threaten financial stability. The next frontier in retail investing? An improvement in market resilience? Daylight madness? Send us your thoughts: [email protected].

Friday interview: Erika McEntarfer

Erika McEntarfer served as commissioner of the Bureau of Labor Statistics from January 2024 until August 2025, when she was fired by President Donald Trump hours after a weak jobs report, a dismissal widely condemned as a politicisation of national statistics. Before the BLS, she spent more than two decades across the US government’s statistical and economic policy institutions, including stints at the White House Council of Economic Advisers, the Treasury and the Census Bureau. She is now a fellow at Stanford’s Institute for Economic Policy Research. She spoke to Unhedged this week about the US labour market, data quality and the integrity of statistics. The text has been lightly edited for clarity and brevity.

Unhedged: How does the labour market look to you right now?

McEntarfer: We had a recent jobs release that was a little below expectations, with some downward revisions to May and April. Payroll growth has been a bit bumpy in 2026, but the average pace of gains has picked up from a lacklustre 2025. The payrolls data is a little hard to interpret right now because there have been changes to labour supply, so I’ve mostly been focusing on the unemployment rate, which has been very steady and stable, moving within a narrow range. Combined with cooling wage growth, this points to a labour market on the loose side of balanced — cautious demand meeting slower labour supply. But I don’t know if I’d want to characterise a very tough labour market for young people as being fully consistent with full employment. I like the term loose side of balanced.

Unhedged: We’ve seen some sizeable revisions to payroll data. What size of revision is normal and how should people think about them?

McEntarfer: It’s important to keep in mind the long-run view: the payroll survey has gotten more accurate over time. Digitisation means much payroll data is now provided electronically, which is improving accuracy. Statistical modelling has gotten better: the birth-death model has gotten more accurate, and seasonal adjustment models have gotten very good. Response rates have fallen over the last 10 years, but right now response rates are not impacting the quality of the data so much as the precision. So it’s not introducing bias, which is a very good thing.

More recently, we saw benchmark revisions that were larger than usual — not historically large, we’ve seen these before — but they were a little weird in that they weren’t driven by birth-death, they were driven by response bias and non-response bias going in the same direction. Usually you’ll get sources of error that cancel each other out, but what we had in the last two years were errors that skewed negative. When I left BLS, we were still trying to figure out why and whether it was going to be persistent. We probably still don’t know the answer, but it looks like we’re bouncing back — right now we’re on a path for positive revisions.

Unhedged: Are the Department of Government Efficiency (Doge) cuts and broader resource constraints affecting data quality?

McEntarfer: They are definitely impacting data quality. Declining response rates are happening everywhere. That’s largely because people and businesses are harder to reach than they used to be. But another factor is that field resources are taking the biggest hit from those budget cuts. Because people have gotten harder to reach, it actually costs more to reach them, but we don’t have the money for those resources — so that’s also impacting response.

In terms of the Doge cuts, this has been well covered, but we had to cut back on price data collection significantly — we had to eliminate a lot of prices from the CPI because we simply didn’t have the resources to collect and process them.

Unhedged: Given those pressures on official statistics, should investors and economists be relying more heavily on private sector data?

McEntarfer: Private sector data is a great complement to public sector data, but it’s not a substitute. The clearest example is the government shutdown last fall. The statistical agencies all went offline for six weeks. There was a delay in the September jobs report, the October jobs report was damaged and there was no October CPI.

There are private sources of payroll data — ADP, Revelio Labs entered the field during the shutdown and there were others, but that data was pretty hard to interpret. It’s an open question how good the signals are from private sector data.

The other reason it was hard to interpret is there was no labour force data. It’s pretty hard to interpret payroll data without knowing what’s happening in the labour force survey — and there’s no private sector version of that data. 

On prices: private sector companies provide prices for goods available in retail, but goods are only about a third of the market basket. The other 60–65 per cent is services, and outside of Zillow for housing, there’s just not a lot of commercial data. So you’ve got a very incomplete picture of what’s happening with prices. The Consumer Expenditure Survey also got delayed during the shutdown — and that survey provides the weights for the basket in inflation indexes. It’s necessary infrastructure to produce inflation data.

Unhedged: At the June Federal Open Market Committee press conference, Fed chair Kevin Warsh said much of the data policymakers rely on comes from “old-fashioned survey methods”. He also pointed to private companies using real-time information that is not subject to much revision. What did you make of those comments?

McEntarfer: I found that statement a bit confusing. It’s absolutely true that many companies use as timely data as they can get their hands on to get the most up-to-date picture of the economy. But I have talked to many business economists, people who work in macro, who work with private sector businesses, who work at the Fed. All of those people rely very heavily on official data for their forecasts and for benchmarking and calibrating the private sector data they have. They do not consider these things substitutes. They are complements.

The other thing I found confusing was the idea that we should get away from data that gets revised — private sector data gets revised too, for the same reasons: it still needs seasonal adjustment, a birth-death forecast and benchmarking against selection bias. Sometimes private sector revisions are larger than the official ones.

Unhedged: Warsh has also argued for looking at trimmed mean or alternative inflation measures. Is there a risk of cherry-picking the data to suit your preferred narrative?

McEntarfer: It’s not clear to me why you would want to focus on trimmed data when you are looking for possible signs that inflation might be moving in the wrong direction because you’re going to miss what’s happening in the tail. That is important information. I don’t know that you want to ignore that as a monetary policymaker.

Unhedged: What would modernisation of US economic data collection look like?

McEntarfer: It would be very nice if the statistical agencies had more resources to allow more companies to provide their data digitally. Right now it’s largely large multi-state firms that provide their data electronically — there are capacity constraints, in part because the agency itself is resource-constrained. Many countries have changed their tax infrastructure to share more data with statistical agencies — to a certain extent, the US government is a giant payroll company and more information from the tax system could be shared with the statistical agency. There’s room for legislators to act here. Right now, US business data is largely an all-volunteer endeavour — firms are not required to report data to BLS. But in many countries the statistical infrastructure is more linked to things that are mandatory. That’s one path for modernisation.

The real constraint here is resources. A lot of this modernisation requires investment, and that requires legislative changes. But where we are right now is that BLS’s budget has declined almost 20 per cent in real terms over 15 years. Doge cut almost 20 per cent of the staff. So it’s gotten harder to innovate, not easier.

Unhedged: An NBER paper estimated that your firing increased the cost of US government borrowing by about $20bn by raising doubts about the integrity of official data . . . 

McEntarfer: The paper in which I’m Figure 1?

Unhedged: Yes. What did you make of that estimate?

McEntarfer: I thought it was a very interesting paper because one of the challenges that we have with estimating the value of all of this data is that it’s really hard to develop counterfactuals and it’s hard to measure the value of something that you give away for free. I’m sorry that I had to be the cause of this information shock or uncertainty shock. But I thought it was a cleverly designed paper.

Unhedged: You have argued that US statistical agencies need stronger independence. What protections are missing under the current system?

McEntarfer: In the US the statistical agencies are not structured as a separate independent agency — we are agencies under cabinet in the executive branch, and that’s just an accident of history. The independence of the statistical institutions is mostly norms-based. We operate in cabinet departments that are political, but there’s a norm that political officials are hands-off on the statistical infrastructure and that the few political appointees in the statistical agencies — I was one, the head of the Census is another — tend to be technocrats, academic, measurement experts, not political people. So a lot of the protections are norms, and they aren’t official. The BLS commissioner doesn’t even have for-cause protection. You can just fire the BLS commissioner because you don’t like the numbers — there’s no law that says you can’t.

Unhedged: After everything that has happened, can US economic data still be trusted?

McEntarfer: The data can still be trusted. The civil servants are still running the agencies. They continue to say that they are not being interfered with. They have said that if that ever stops being true, you will see a mass exodus and a lot of whistleblowing. That is not happening. So I think we can be fairly sure that the data remains trustworthy.

One good read

The loss of open-ended curiosity

FT Unhedged podcast

Can’t get enough of Unhedged? Listen to our new podcast, for a 15-minute dive into the latest markets news and financial headlines, twice a week. Catch up on past editions of the newsletter here.