Referenews

Year-over-year measures of inflation are likely to get very screwy, because of price increases a year ago. The average price for a gallon of gasoline went from about $3.50 in February to about $4.20 in March to over $5 by June, according to the Energy Information Administration. Inflation reports for March through June will be calculated from a base that includes those unusually elevated prices.

Base effects can cause analytical mistakes. If gas prices climb from about $3.40 now to $4.50 in June, that would be the second-highest June price on record, but base effects would show it down 10% year over year—perhaps prompting a premature declaration of victory over inflation.

Global food prices followed a similar pattern as gas prices. The benchmark international food-price index produced by the United Nations’ Food and Agriculture Organization soared at the start of the war, then peaked in March. Natural-gas prices soared until August and then began to decline. On a year-over-year basis, these metrics will swing from big increases to big decreases. 

Base effects also crop up in noneconomic fields, for example when changes are being calculated against a very low base. Early in the Covid-19 pandemic, pediatric Covid hospitalizations would sometimes be reported as a percentage change from the previous week. 

If hospitalizations increased from 0.4 to 0.7 it’s a more than 70% increase, which sounds terrifying if the context is omitted that it’s 0.7 out of 100,000.  

The appeal of year-over-year comparisons is that they account for seasonal patterns. Department-store sales, for example, always soar from the third quarter to the fourth quarter then fall in the first. Better, then, to compare one quarter against the same quarter a year earlier, as is standard in corporate earnings reports.

This for a long time was also necessary with key economic data such as consumer prices. The consumer-price index began as a World War I-era initiative to track soaring prices especially in shipbuilding centers, and became a regular Bureau of Labor Statistics publication in 1921. 

The first report touted that the price of 22 essential food items had dropped 10% from December 1919 to December 1920. The price of evaporated-milk prices was down 12%, lard 27% and cabbage 44%. (The early BLS reports are full of amusing time capsules such as butcher charts from different cities, to determine when cuts of meat with different names were in fact the same.)

In 1959, the BLS began working on a system to allow consecutive months to be compared with each other: seasonal adjustment. Seasonal adjustment tries to strip out changes to data attributable to recurring seasonal effects so that, for example, department-store sales in December can be compared with November without the distortion of holiday buying.

Before computers, calculating seasonal adjustment was incredibly complicated. In his autobiography “The Age of Turbulence,”

Alan Greenspan

writes that in 1947 while working at Brown Brothers Harriman, his first job as a paid economist, he had to calculate a seasonal adjustment for weekly department-store sales. Using pencil and paper, slide rules and desktop adding machines, he needed two months to complete the assignment, he wrote.

Seasonal adjustment means you can now compute inflation rates over much shorter periods—a month or a quarter—avoiding misleading base effects. Yet even now the year-over-year interval retains widespread intuitive appeal. 

“Humans everywhere tend to assume the year interval is somehow THE interval. When you ask someone’s age, have you ever gotten an answer in months?” Yale economist

Robert Shiller

said in an email. Mr. Shiller’s Case-Shiller home-price index reports the year-over-year change in home prices as its headline figure, though it’s available seasonally adjusted.  

Seasonal adjustment also has its flaws. It’s based on how data has behaved in recent years. An unusual event, such as the deep 2007-09 recession, the war in Ukraine or the Covid-19 pandemic, could lead to a systematic, and incorrect, seasonal adjustment to remove such effects later on. 

New York Fed research notes, for example, that the worst of the 2007-09 recession struck in early 2009. “As a result, for the subsequent few years, an ‘echo’ of the Great Recession took place as economic data kept exceeding the artificially low expectations for that time of year.”

It’s naturally tempting to distill the economy down to just one or two simple numbers and time frames. But when data gets weird, the only way to make sense of it is with a few more numbers. If you’re only focusing on an annual change, it’s probably not enough to make sense of what’s happening.

“The sausage-making of it is economists and analysts are going to look at a lot of different time spans, different time periods,” said

Ataman Ozyildirim,

senior director of economics at the Conference Board. “You need a lot of different lenses to get a better handle on that momentum in the economy.”

Write to Josh Zumbrun at josh.zumbrun@wsj.com