Donald Trump escalates his war on numbers

Another shocking assault on a non-partisan institution

JUST FOUR months ago Donald Trump was basking in the glow of the jobs report produced by the Bureau of Labour Statistics (BLS). “GREAT JOB NUMBERS, FAR BETTER THAN EXPECTED”, he posted on Truth Social. But on August 1st the BLS announced that employers had added fewer jobs than expected in July; it also revised downward the employment data for the previous two months. This time Mr Trump turned on the messenger. Saying (without evidence) that the figures were “rigged”, he fired Erika McEntarfer, the BLS commissioner who was confirmed last year on a bipartisan basis.

Whatever the explanation, Mr Trump has undermined confidence in the reliability of the agency’s output.

Mr Trump has pledged to appoint a successor to Ms McEntarfer who is “much more competent and qualified” so that BLS statistics are “fair and accurate” and “can’t be manipulated for political purposes”. But the president’s vainglorious sensitivity and norm-breaking intervention will make it difficult for any appointee to re-establish trust. “Unfortunately, as a result of this”, says a BLS staffer, the administration is “going to change the perception of anything that comes out.” The episode also comes at a time when the cash-strapped bureau is struggling to measure an increasingly sprawling workforce and a public reluctant to respond to government surveys.

The numbers produced by the BLS are still some of the most important indicators tracking America’s $30trn economy. Its headline-generating jobs report, usually released on the first Friday of each month before markets open, is read closely by reporters, investors and Fed chairs. The report is drawn from two surveys. The first asks individuals whether they are working or looking for a job; these results determine the unemployment rate. The second survey collects payroll data from businesses and government agencies to track the total number of paid positions in the economy. Together, the measures offer a snapshot of the economy’s momentum.

What specifically caused Mr Trump’s ire was the magnitude of routine technical adjustments known as revisions. The agency releases preliminary estimates based on incomplete data because it is helpful for the Fed and investors to know quickly what is happening to employment. As more data trickles in, the BLS updates its estimates, issuing revisions. These are often relatively modest, typically adjusting employment figures by some 40,000 jobs (or about 20%, on average). But the most recent revisions were three times that. The research unit of Goldman Sachs, a bank, called it the “largest two-month revision since 1968” outside recessions.

Mr Trump alleges malice, claiming the new numbers were manipulated to make him look bad. For that to be true it would require a vast conspiracy co-ordinated by scores of civil servants. BLS staff describe the process of producing the jobs report as designed to prevent meddling. It is intentionally decentralised, with strict digital-security protocols, physical lockdowns in the BLS headquarters during the week preceding public release and siloed workflows that gradually move up the chain of command. If protocol is followed, the commissioner would not collect or analyse any data herself.

It is a puzzle why this month’s revisions were unusually large. Historically, revisions have tended to be larger when the economy is at a turning point—precisely when the imputation methods used by the BLS, which lean heavily on historical data patterns, become less reliable predictors of future conditions. Some data sleuths have pointed out that a disproportionate share of the downward revisions came from state and local government payrolls.

Declining response rates to government surveys may have played a role, too. A decade ago response rates were 20 percentage points higher. Fewer respondents means noisier estimates. Some also point to staffing and budget constraints. Across that same period the BLS budget has fallen by some 12% in real terms, and the White House has proposed cutting it by another 8% this year. The reason could also be more mundane: random chance.

Whatever the explanation, Mr Trump has undermined confidence in the reliability of the agency’s output. Private-sector alternatives cannot replicate what the BLS provides. There is already enormous financial incentive to do so. Trillions of dollars in global assets reprice within moments of a BLS release. If investors could reliably anticipate the results using private-sector proxies they would. But the alternatives that already exist, like ADP’s employment report, are widely regarded as noisy and too narrow.

Private-sector estimates are also benchmarked to government surveys, which serve as a sort of statistical “ground truth”. And even if a private firm managed to reliably estimate headline numbers, they would still fall short on providing granular insights across sectors, geographies and occupations. Decision-making about interest rates and investing is hard enough with BLS numbers that are widely regarded as reliable, within the limits of survey methodology. For a time, at least, those peering at the economy’s horizon will have to cope with dimmer light.