Over the coming few weeks, alongside the regular coverage of the Clark Center’s polls and other news, On Global Markets will be reporting on the discussions held at the Economic Experts Conference 9/10 October. Most of those sessions were held under the Chatham House Rule, but the fireside chat on the BLS was not.
While the data released by the Bureau of Labor Statistics (BLS) often leads the financial news, the institution itself rarely makes the headlines. That has changed this year. In early August, following unusually large negative revisions to the employment data for June and May, President Trump fired BLS Commissioner McEntarfer and accused the Bureau of displaying a bias against his administration and manipulating the jobs figures to flatter his political opponents and diminish his own record.
Few, if any, serious economists agree with this charge. A poll this summer asked members of the Clark Canter’s US Experts Panel whether ‘there is no evidence to suggest that the employment estimates produced by the Bureau of Labor Statistics are biased so as to favor any particular political party’? 92% of respondents, weighted by confidence, strongly agreed and 8% agreed.
The administration’s initial nominee to replace Dr McEntarfer, Dr EJ Antoni – a conservative economist with no special expertise in statistics – had his nomination dropped at the beginning of the month, reportedly due to a lack of support in Congress.
What comes next is unclear, although concerns have been raised amongst data-watchers about the reliability of BLS numbers in the future. Attendees at the Center’s Economic Experts Conference, held in Chicago on the 9th and 10th of October, heard about the current state of the BLS and what to watch out for in the future from Erica Groshen (a former BLS commissioner) and Michael Horrigan (a former long-standing BLS staffer).
Dr Groshen, who headed the bureau between 2013 and 2017 and is now based at Cornell university, painted a grim picture of the current state of the BLS. Response rates to surveys have declined sharply in recent years, in common with the experience of private sector surveyors and other statistical agencies overseas. That would be hard enough to deal with but, in the case of the BLS, the issue has been compounded by large real term cuts in its funding, partially to the Bureau’s ‘lack of champions’ in Congress.
Staffing numbers, noted Dr Groshen, are down 20% since February and around 30% of senior staff positions are currently unfilled. The administration is meanwhile proposing a further 8% cut in the budget. The budgetary squeeze is having an outsized impact as, Dr Groshen explained, declining response rates to surveys tend to push up the costs of data collection. All of which is hampering the BLS’s ability to invest in modernising itself and its collection methods. While declining survey response rates are an obvious problem, the digitalization of ‘everything else’ in the economy offers an opportunity for more blended data products – taking data produced for other purposes and combining it with survey evidence is likely the future of official governmental statistics but getting to this new system will not be cheap.
Accusations of political bias and the summary firing of the commissioner are hardly likely to have helped an already over-stretched institution.
Michael Horrigan, now of the Upjohn Institute but for many years in charge of both the employment and inflation data series at BLS, put the large revisions of employment data seen in recent months into a longer-term perspective. The real reason for worry, he suggested, was not the charge of political bias but the fact that historically speaking, large revisions in the BLS labor market data have usually been associated with turning points in the jobs market. In September the BLS downgraded the current employment statistics (CES) for the 12 months to March 2025 by 911,000 jobs. That drew further accusations of bias from administration officials although, as Dr Horrigan explained, such revisions usually relate to how the BLS handles the issue of firm births and deaths.
The BLS usually uses the number of firm deaths – which it can roughly track via the number of establishments reporting zero employment – as a proxy for the number of firm births. In normal circumstances the firm births and deaths tend to move roughly inline and this approach leads to only small revisions as fuller data becomes available. But when firm the jobs market is slowing and firm deaths are outpacing firm births by a large margin, this leads to large revisions. The annual benchmark revision to the data for March 2009, for example, just as the Great Recession was starting was a downgrade of 902,000. The recent pattern of revisions may well be a sign that the jobs market has been slowing faster than feared.
Much of the question and answer session focused on what the early signs would be of any politicization of the data output by the BLS. As more than one conference participant put it, this was not exactly anyone’s base case but things which ‘once seemed outside of the likely distribution of events are now within it’.
Attempting to manipulate the numbers would, both speakers agreed, be like ‘taking a wrench to a finely tuned process’. The signs of interference would be straight forward to spot; data would be delayed, tables would be missing, some things would simply not add up. What is more, this could not be done quietly – such moves would almost certainly be accompanied by staff resignations and whistleblowing.
Dr Groschen noted that BLS commissioners usually ‘arrive alone’ and any new appointee arriving with ‘a posse’ of their own could be a worrying sign.
The other thing to watch closely, according to the speakers, is the tone of the commentary in BLS releases. At present, as Dr Groschen put it, ‘liveliness is not a goal. This is a six ounce glass containing three ounces of water, it is not half full or half empty’.
The first warning sign that BLS data was becoming less reliable might be found, not in the numbers themselves, but in the words which accompany their release.
