The Economist: How should the US Federal Reserve measure inflation?

In the early 1970s Arthur Burns had an inflation problem. As one supply shock followed another, the pipe-smoking Federal Reserve chairman, known as the “Pope of Economics”, responded by excommunicating prices he blamed on “special factors”.

The Arab oil embargo pushing up crude? Look past energy. A collapse in Peru’s anchovy harvest driving up feed costs? Strip out food.

When inflation still looked too high, out went used cars, home purchases and mobile homes.

Even rising jewellery prices were blamed on “gold mania” and waved away, as Stephen Roach, then a young Fed economist, later recalled.

In search of the one true inflation rate, Burns kept narrowing the canon. Inflation, meanwhile, was becoming ever more Catholic.

Half a century later Kevin Warsh has taken up the same challenge. The Fed’s new-ish chairman has made a mission out of pinning down underlying inflation (“underlying” is fast becoming his favourite adjective).

Yet he has dismissed the usual guide, “core” indices excluding food and energy, as a “rough swag” for where inflation is headed.

Mr Warsh instead prefers “trimmed averages”, which exclude the biggest price swings, and measures of inflation’s breadth.

On September 16 he complained that prices were rising by more than 3 per cent in “too many” product categories.

That was one reason why the Fed raised interest rates for the first time in three years. He has even sent a task-force hunting for cleaner, timelier economic signals.

Can Mr Warsh get closer to the underlying trend without stumbling into Burns’s trap?

Underlying inflation is, roughly, the part of today’s inflation most likely to still be there tomorrow.

In a market economy, individual prices are always moving around. A drought makes coffee dearer; a glut makes cars cheaper.

In August prices for mobile-phone services mysteriously jumped by nearly 6 per cent in the space of a month.

Such idiosyncratic movements can move headline inflation without saying much about the broader trend. Central bankers therefore try to look through the one-offs, training their gaze instead on the underlying movement in the general price level.

After the supply shocks of the 1970s, a simple shortcut took hold: strip out volatile food and energy prices.

The resulting “core” measure replaced Burns’s discretion with a fixed rule that usually gave a better steer on where inflation was headed. But it is a crude shortcut.

As Jim Dolmas and Evan Koenig of the Dallas Fed put it, “Not all food and energy items are equally volatile, nor are all of the most volatile items exclusively food and energy.”

Restaurant prices get thrown out despite being fairly stable, whereas more volatile categories such as airfares and hotel rooms stay in. It is, indeed, a “rough swag”.

Economists have long looked for a better swag. In a paper from 1994 Michael Bryan, then of the Cleveland Fed, and Stephen Cecchetti of Brandeis University built a model showing how lopsided movements in individual prices can obscure the common inflation trend.

Their answer was to rank price changes and focus on the middle — either by chopping off the extremes before taking an average, or simply by taking the median.

In their data, the median did a better job of forecasting future inflation than either the headline consumer-price index or the index excluding food and energy. Later research has generally found that such trimmed measures can provide a cleaner signal of durable inflation.

The idea has been put to work. The Cleveland Fed publishes a median and trimmed CPI. The Economist’s predictive inflation tracker downweights extreme price movements rather than discarding them.

Yet pruning can work badly if price patterns shift.

The Dallas Fed releases a trimmed-mean measure of changes in the price index for personal consumption expenditures.

For decades the biggest price falls tended to be larger than the biggest rises. So the Dallas Fed used an uneven cut, dropping the bottom 24 per cent of price changes and the top 31 per cent.

That made sense when inflation was low; less so once more prices began to rise.

During the inflation surge of 2021, the Dallas measure started stripping out too many genuine price increases. The problem persists.

In April researchers at the Dallas Fed warned that their gauge may still be understating inflation. They have since produced an alternative that cuts more evenly from both ends of the distribution.

It would have picked up the surge in 2021 much sooner. For the year to June it put underlying inflation at 2.6 per cent, against 3.3 per cent for core PCE.

But any trimmed measure is calibrated to how prices behaved in the past. If that pattern shifts, as it often can in periods of high inflation, the measure risks trimming away part of the inflation it is meant to capture.

The 500,000,001st price is right

Mr Warsh thinks the Fed can do better. Part of his answer is more timely, granular data.

The Billion Prices Project, pioneered by Alberto Cavallo of Harvard University and Roberto Rigobon of the Massachusetts Institute of Technology, gathered millions of online prices each day to track inflation almost as it happened.

The Fed chair wants similar feeds to supplement official statistics. But more prices do not necessarily reveal the trend. To find it, Mr Warsh prefers to focus on the middle.

Earlier this year he imagined assembling a billion prices and said what really mattered was what was happening to the “500 million and one” price.

That may be an improvement, but the median has a blind spot. A large minority of prices can surge without moving the middle much, as happened in 2021, when median measures understated the underlying trend.

The episode is a reminder that no rule will perfectly capture underlying inflation. The temptation, then, is to compensate by deciding which prices to ignore.

That was the cardinal sin of the Pope of Economics. Mr Warsh would do better to take a leap of faith and embrace an imperfect rule.