When the data changes after the decision

Core personal consumption expenditures (PCE) prices rose 0.6 per cent in January, lifting the annual rate to 4.7 per cent and giving policymakers at the Eccles Building a nasty surprise. The upward revisions to earlier months that came with it were worse. The smooth fall in inflation that the Fed thought it had achieved in late 2022 turns out to have been produced by provisional numbers. When the data change after the decisions have been taken, central bankers are left steering by a cracked rear-view mirror.

The Taylor-style rules that shape the Fed's debates lean heavily on real-time estimates of inflation momentum and the output gap. If early readings keep understating how sticky inflation is, the committee will keep making the wrong call.

Why revisions carry weight

In John Taylor's original 1993 formulation, the prescribed policy rate rises by 1.5 percentage points for every extra point of inflation. A revision that adds a few tenths to measured core inflation therefore moves the recommended rate by one and a half times that amount. Estimates of the output gap are revised even more heavily than prices, and they enter the same equation, so a committee reading the rule off first-release data may be reading a number that will not exist a quarter later.

A simple calculation shows what is at stake. Plug January's 4.7 per cent core inflation rate and a closed output gap into Taylor's original rule, with its assumed neutral real rate of 2 per cent, and it prescribes a policy rate of about 8 per cent. The Fed's range is 4.5 to 4.75 per cent. Nobody on the committee applies the formula mechanically, but the gap shows how far the Fed would still be from a conventional benchmark even if it took the latest numbers at face value, and why a few tenths of revision can matter.

Decisions built on sand

The Fed slowed from 75 to 50 basis points in December and then to 25 basis points in February. At the time, the step-down looked justified by what seemed to be a rapid quarterly fall in monthly price pressure.

The revisions show that core inflation in fact stayed flat through the fourth quarter. The committee had loosened financial conditions, and let equity valuations climb, on the strength of figures that did not survive a second look.

The damage arrived in two instalments. Earlier this month the Bureau of Labor Statistics' annual update of seasonal factors turned December's reported fall in consumer prices into a small rise. The PCE release then did the same to the Fed's preferred gauge, erasing the neat downward staircase that officials had cited when they slowed the pace of increases.

Why PCE moves after the fact

The Fed has targeted 2 per cent inflation on the PCE measure since 2012, preferring it to the consumer price index because its coverage is broader and its weights shift as households substitute between goods. That breadth comes at a cost. PCE draws on a wide range of sources, including producer prices, company surveys and administrative data that arrive late, and it inherits whatever revisions those sources undergo. It is the better gauge of inflation in principle, and the less reliable one in the first few months after the fact.

The consequences reach well beyond the Fed. Mortgage rates, which had eased in January as markets bet on an early end to tightening, have climbed again this month, and the housing market's tentative recovery depends on them. Companies that issued floating-rate debt or planned to refinance this spring face higher costs than they expected at the start of the year. Savers in money market funds and Treasury bills, by contrast, stand to earn higher yields for longer.

A lesson from the 1970s

The Fed has been here before, at far greater cost. Research by the economist Athanasios Orphanides found that much of its 1970s failure traced to real-time estimates of spare capacity that later proved far too generous. Policy looked suitably tight on the figures available at the time and loose on the revised ones. Squeezing out the resulting inflation took Paul Volcker's double-digit interest rates and two recessions in the early 1980s.

The case for patience is not trivial. Monetary policy works with long lags, much of the 450 basis points of tightening since last March has yet to bite, and January price data carry their own seasonal quirks as firms reset list prices at the start of the year. A run of benign spring readings would revive that argument. It will carry less weight with a committee that has just been caught out.

There are ways to reduce the problem. The committee can put more weight on measures that are revised less, such as the Employment Cost Index, which tracks wages and benefits quarterly and is closely watched for signs of labour-market inflation. It can also look at trimmed-mean inflation measures, which exclude the most extreme price moves each month and tend to be less noisy. None of these removes the uncertainty, but together they make it harder for one month's quirks to steer policy.

Burned once, more hawkish now

The likely result is a tougher Fed. Officials embarrassed by revisions will stop trusting soft preliminary data or the first signs of a slowdown. Minutes of the February meeting, released this week, showed that a few participants already favoured a half-point increase. Since that meeting, two regional Fed presidents, from St Louis and Cleveland, have said publicly that they saw a case for the larger move. Neither holds a vote this year, but the shift in tone shows where the internal debate is heading.

They will want clear and lasting evidence that demand is weakening before they even discuss ending the tightening. The February jobs and consumer price reports both land before the committee meets on March 21st and 22nd, when officials will also publish fresh projections for the path of rates. Having been misled by early data, the Fed will probably raise rates further than the economy strictly needs, simply to cover the risk that the next set of numbers is wrong too.