How not to fire a statistics chief

Shock is reverberating around the US economics and statistical community after Donald Trump on Friday summarily fired the head of the Bureau of Labor Statistics, Erika McEntarfer.

Straight from the authoritarian playbook, the US president did not like the jobs numbers, so took vengeance on the messenger. The large downward revisions announced on Friday had to be wrong, Trump said, because they did not show “The Economy is BOOMING under TRUMP”.

The American Economic Association expressed its “grave concern” that Trump’s actions would threaten the integrity of the information that markets, institutions and the public rely on every day.

It looked pretty bad and it was.

That said, immediate departures of heads of statistical agencies are not just something that happens in dictatorships or the US. Less than three months ago, the head of the UK’s Office for National Statistics, Sir Ian Diamond, departed with immediate effect “due to ongoing health issues”. The UK’s sleepy Royal Economic Society said nothing. The Royal Statistical Society thanked him and spoke of his “outstanding leadership”.

This departure was very British. Diamond resigned between the UK government announcing a review of the ONS’s “performance and culture” in April and its critical assessment when it was finally published in June. At a subsequent parliamentary hearing, officials cast Diamond in such a poor light that the committee chair said they seemed to characterise him as a “sort of hybrid of a Medici prince and Blofeld, [who] seemed unwilling or uninterested in anything anybody had to say”.

There you have it: two statistical heads depart immediately in different ways. Where does it leave UK and US official statistics?

UK flying blind

The UK has no permanent replacement for Diamond and is thinking about changing the role entirely. Meanwhile, the poor Bank of England is still without reliable statistics.

Labour market data is a particular concern. There are now alternative measures of employment, but unhelpfully these are currently showing inconsistent trends. Surveys of companies and households show employment rising, tax records show it falling.

The UK’s labour force survey, which asks households about their jobs market position, still has a terrible response rate of 21 per cent. That is not a typo.

If getting to grips with UK employment is difficult, assessing the participation rate is even harder. The LFS is the only data source that can measure whether people are seeking employment or are inactive. The official figures show a growing then easing inactivity problem after 2020, but it might be largely statistical noise. No one knows.

Officials say Diamond’s departure from the ONS was linked to the statistical agency’s failing on these bread-and-butter issues.

There was no national scandal and no suggestion that the independence of economic institutions was compromised. But there was little transparency in his departure and few signs of improvement. The chair of the ONS board has also resigned, leaving the organisation almost entirely rudderless and the statistical agency reeling.

‘Gold standard’ official US data

When firing McEntarfer, almost everything Trump wrote and said about her was false. Data revisions had not flattered the jobs narrative of his predecessor Joe Biden, and she had definitively not “faked the Jobs Numbers before the Election to try and boost Kamala’s chances of Victory”. There will be every reason not to trust the data once a Trump appointee is in charge of the BLS. He is promising to name his pick to be McEntarfer’s successor this week.

But the fact that Trump lies should not let us get misty-eyed about the success of BLS statistics under McEntarfer. The US story is very familiar for anyone who has followed UK statistical agonies of the past 15 years. Starved of funding, its output is far from gold standard.

There are multiple and growing problems in BLS figures. The inflation data is increasingly guesswork. If the BLS cannot sample the price of a particular item in a particular location — say, loaves of seeded bread — it imputes a figure based on other types of bread in the same location. If that data is also unavailable, it imputes data with the seeded loaf prices in other parts of the US. This second tier of imputation now accounts for 35 per cent of all imputed data in the CPI inflation figures.

This data is regularly mischaracterised as the proportion of “guessed” data in the US CPI. It is not. It is the proportion of imputed data that uses the second tier method. Before McEntarfer was fired, I asked the BLS how much of the total CPI was imputed. It replied with a factual but unhelpful comment. The BLS “does not publish the per cent of all prices that are imputed”, it said. It did not respond to follow-up questions.

The BLS labour market data is also flawed. The US equivalent of the UK’s labour force survey is its current population survey. This is sliding down the same response rate tunnel that plagued the ONS.

The US payroll figures are based on a different survey of establishments, which has its own accuracy problem. My colleague Joel Suss at the FT’s Monetary Policy Radar did a deep statistical dive on the US jobs numbers last year, concluding that “it is time for the Fed not to be too data-dependent in response to the monthly US payroll numbers”.

That was well before last Friday’s jobs numbers. According to Goldman Sachs, the most recent data contained the largest two-month downward revision of jobs growth outside recessions since 1968. It was this revision that resulted in McEntarfer being fired.

That data was extreme. But the payroll figures have been persistently revised lower since 2023, as the chart below shows. And the most recent data for the period April 2024 to March 2025 will almost certainly be revised down further next month when the BLS benchmarks the figures against unemployment insurance records.

It was reasonable, therefore, for Kevin Hassett, director of the US administration’s National Economic Council, to say the BLS should have published a better explanation of its very large revisions on Friday. It was also reasonable for him to warn about a possible slide in US labour market data quality to UK levels.

Trump’s explanation of his decision to fire McEntarfer was ugly and will ensure US data continues to be deemed suspect.

But let’s not fall into the trap of thinking the US is alone in having data that is suspect. The UK generally respects its economic institutions, yet the ONS’s plight is at least as bad as that of the BLS.

But let us also not pretend the BLS was a beacon. It was failing to provide accurate statistics and that was already a huge problem for the Federal Reserve, financial markets and everyone else.

What I’ve been reading and watching

• The European Central Bank seeks to get members of its elected works council to do their day jobs.

• The South African Reserve Bank changed its inflation target from the midpoint of a 3 to 6 per cent range to the bottom end.

• It was a big week for monetary policy decisions that did not move rates, in the US and Japan at least.

• It was tariff week last week, and Jason Furman says the US will suffer most.

A table that matters

Fed board governor Adriana Kugler resigned on Friday with immediate effect.

Kugler’s term was due to end in January next year, as the table below shows. Trump says he will nominate his pick in the next couple of days. New appointees take the remaining term of those they replace.

The eagle-eyed will notice that the nominee is almost certainly also going to be Trump’s pick for Fed chair, since there will be no other certain vacancies on the board before Jay Powell’s term as chair ends in May 2026.

The Fed chair race is therefore hotting up. Trump could, of course, nominate someone to take Kugler’s seat until January, and then nominate someone else. Expect whoever is nominated to be super loyal to Trump in the run-up to January 2026.

It won’t be a peaceful autumn at the central bank.

---

Central Banks is edited by Harvey Nriapia