Trump fires the jobs data chief

The commissioner of the Bureau of Labor Statistics was fired on 5 August, breaking a long-standing norm that official economic statistics are kept free of political interference. The dismissal followed a weak July employment report that showed just 73,000 net new payrolls and downward revisions of 258,000 jobs to earlier months.

A bad report, then a sacking

The report was bad by any standard. Beyond the weak July gain, the revisions cut the previous two months' hiring to almost nothing, and the unemployment rate edged up to 4.2 per cent. The president responded by accusing the agency of rigging the figures and dismissed its commissioner, Erika McEntarfer, who had been confirmed by the Senate last year. No evidence of manipulation was offered. Revisions of this size come from late survey responses and from the annual reconciliation with tax records, and the BLS publishes them in the same release for anyone to check.

The agency was already stretched. Its budget has been squeezed for years, survey response rates have fallen, and earlier this summer it cut back price collection for the consumer price index in several cities because of staff shortages. A statistical office short of money and people is more prone to error. Removing its head for publishing inconvenient numbers makes it harder to recruit the replacements it needs. Economists and statisticians with private-sector options have little reason to join an agency whose leaders can be sacked over a single release.

Bond desks lose a reference point

Removing the official in charge of the country's main economic numbers after a bad print puts a new risk into the pricing of bonds. Fixed-income desks around the world depend on BLS data being neutral. They use it to structure interest-rate swaps, calculate inflation accruals on TIPS and price the risk at Treasury auctions. Once traders suspect the jobs figures have been filtered for politics, the releases lose much of their value as information.

The TIPS link is the most direct. The principal of an inflation-protected Treasury is adjusted each month by the consumer price index that the BLS produces. An investor who buys one is buying protection defined by the bureau's own calculation. If there were any doubt about that number, the protection would be worth less, and investors would pay less for it. The same logic runs through inflation swaps and through the many private contracts, from wages to leases, that are indexed to official prices.

The effect on trading is mechanical. The monthly payroll release is one of the busiest moments of the month in Treasury futures and rate swaps, because it is the most comprehensive reading of the economy available at short notice. Dealers quote tight prices around it on the assumption that the number means what it says. If a portion of each release might reflect pressure rather than measurement, dealers need a cushion against being wrong, and that cushion takes the form of wider spreads and smaller trade sizes. Liquidity thins precisely when markets most need to digest new information.

What other countries learned

There are precedents, and none is flattering. Argentina's government interfered with its statistics office from 2007 and published inflation figures far below what private economists measured, and in 2013 the IMF formally censured the country over the quality of its data. Investors responded by ignoring the official numbers and paying a premium on Argentine debt. Greece's under-reporting of its deficit, exposed in 2009, was one of the sparks of the euro-area debt crisis. In both cases the loss of trust outlasted the governments responsible by years.

America is not Argentina, and the BLS is not about to fabricate numbers. The mechanism of damage does not require fabrication, though. It only requires markets to put a small probability on it, and to discount each release accordingly.

Paying for doubt

Bond markets responded by demanding extra yield for the doubt over data integrity. If the executive can lean on official statistics, big investors will have to buy expensive private alternative data instead.

Private sources exist, such as payroll processors' employment counts and online price scrapers, and larger funds already use them. They are partial, however. A payroll processor sees only its own clients, and a price scraper sees only goods sold online. None of them has the legal right to collect data from every employer or the benchmark against tax records that the BLS uses. Replacing a public good with private subscriptions raises costs for everyone and favours the biggest firms, which can afford to buy several sources and triangulate.

The Fed's dilemma

The Federal Reserve leans on these numbers too. It held its policy rate at 4.25 to 4.5 per cent at the end of July, with two governors dissenting in favour of a cut, and markets now expect easing in September. If the central bank cuts after weak jobs data, critics will say it bowed to political pressure; if it waits and the data turn out to be right, it will have kept policy too tight. Any doubt about the figures makes both mistakes more likely.

The case against

The counterargument is that the BLS is a large institution with career staff, published methods and multiple layers of review, and one dismissal does not change how the numbers are produced. That is true for now, and the next few releases will probably be compiled exactly as before. The risk lies in the choice of successor and in whether future commissioners feel free to publish bad news. Watch the nomination for the post, any change to methods or release schedules, and the preliminary annual benchmark revision to payrolls due in September.

Firing the country's chief statistician over a poor jobs report has damaged trust in American economic data, and traders will probably charge wider bid-ask spreads on interest-rate derivatives for a long time to come. The first place to look for the damage is the market reaction to the September payroll benchmark, which the new leadership will oversee and which investors will now read with unusual suspicion.