VSPRICE-CONSUMER inflation reached 5.4% in America, the highest in 30 years. On November 3, the Federal Reserve announced it would cut bond purchases, a step towards higher interest rates. Most economists say this inflation surge is the result of temporary disruptions caused by covid-19, and that it will pass. But some believe this portends a longer-term trend.
One of the main arguments made by the inflation doves is that only a few items caused much of the total price increases. In the quarter to August, used cars, hotel rooms, and airline tickets made up less than 5% of the U.S. Consumer Price Index, but together accounted for the majority of headline inflation. “This is really extreme,” wrote Paul Krugman, an economist at the time, “and suggests transient bottlenecks rather than widespread inflationary pressure.”
This case is based on two allegations, both of which deserve close scrutiny. One of them stands out clearly: compared to past periods of similar inflation, current price increases are indeed unusually concentrated. The other, that inflation should slow down as a result, is also generally true. However, this effect is too weak for the Fed to breathe easy.
To test these hypotheses, we constructed a dataset of price levels since 1959 for each item – from housing to lottery tickets – in personal consumption expenditure (PCE), one of the Fed’s favorite measures. For each rolling 12-month period, we calculated a measure of the price change between items: their standard deviation. When a few components represent a significant portion of inflation, that number is high. When the prices of most items change by a similar amount, they are low.
In general, standard deviations correlate with inflation: the higher the average price increase, the more the price changes of specific items differ from each other. However, some times have been unusual, with inflation either low but concentrated or high but broad. To identify these outliers, we measured the “excessive” concentration of inflation in each period: the deviation between the actual standard deviation of price changes and what you would expect based on headline inflation. .
This measurement is now abnormally high. In the year to May, inflation was more excessively concentrated than in 97% of the 12-month periods since 1961. It declined slightly as used car prices stabilized, but is still in the 89th percentile.
What does this mean for future inflation? Historically, when the excess of focus has been high, the present has been a poor guide for the future. When inflation is above its 10-year average, as it is now, high excessive concentration makes it more likely to fall. This trend should lead forecasters to lower their inflation forecasts.
The idea that a few big price changes can lead forecasters astray is not new. In the 1970s, economists devised “core” inflation, which excludes food and energy. More recently, “cropped average” measures, which push down items that have fluctuated the most in price, have become fashionable. The Dallas Fed has published articles showing that its version, which excludes the bottom 24% and top 31% of PCE index, predicts inflation better than the core.
However, both of these methods have flaws. Changes in food and energy prices are not necessarily exceptionally large or short-lived. And the cropped means weighting schemes are rife with sheer cliffs. In the version of the Cleveland Fed, which removes the top and bottom 8% of the index, an item from the 93rd percentile, when sorted by price change, is removed entirely, while an item from the 92nd gets all its weight.
With this in mind, we have designed an alternative inflation index. Like the trimmed means, it adjusts the weight of the items based on their recent price changes. But its weights are shaped like a smooth hill rather than a box. Components with near-median inflation are the most prominent, and those whose prices vary the most are least.
Our hill looks a bit like Uluru in Australia: a wide central plateau, flanked by a steep slope on the left and gentler on the right. (The Dallas Fed’s truncated average is also skewed, controlling for the bias caused by the imbalance in the price change distributions.) Most items with negative or low inflation carry significant weight; those with the fastest rising prices account for 25% as much as those in the middle.
When you use data from the previous year to predict PCE inflation over the next year, this method is more accurate than using core inflation or expecting inflation to remain constant. Since 1959, its one-year forecast has also exceeded that of the Dallas Fed’s truncated average.
Part of this apparent advantage stems from the design of our study: The Dallas Fed sought to maximize accuracy for periods and forecast horizons different from ours. However, the errors of its cropped average in the 1970s illustrate the risks of such a deep cropping. In the midst of two oil shocks, some of the fastest rising prices – those for energy-related items – have continued to rise. The less weight an index gave to these goods, the more it predicted inflation in a year.
The current episode of concentrated inflation differs from the 1970s in many ways. Soaring prices for goods like household appliances are unlikely to spill over into other costs like oil prices do. And in general, treating outliers the same as more representative items has been a mistake. But in cases where such price changes foreshadow broader supply constraints, ignoring them altogether can be an even bigger mistake.
To verify the meaning of our “Uluru” method, we have also built a model which provides PCE inflation using only excess concentration and one- and ten-year rolling inflation rates. This approach and the Uluru Index give an inflation forecast of 4.1% over the next 12 months. That’s below the current level of 4.4%, but above the Fed’s target of 2% and the value of 2.3% of the Dallas Fed’s cropped average over the 12 last months. If this forecast comes true, interest rate hikes will almost certainly follow.■
Sources: Bureau of Economic Analysis; The Economist
This article appeared in the Graphic Detail section of the print edition under the title “The Used Car Enigma”