Early this year, the January jobs report came in hot. The economy added a whopping 353,000 new jobs — at least according to media reports.

But what was actually happening to real people across the country, according to Jonathan Wright, professor of economics at Johns Hopkins University? “I’m going to guess well over a million people lost their jobs,” he said. 

He’s right. January this year saw 2.8 million fewer workers on non-farm payrolls than the month before, according to the Bureau of Labor Statistics. 

“There were a lot of people who were hired for the Christmas holiday season,” said Wright. “Those were temporary jobs, and they’re gone in January.”

The same dynamic happens every year: gift shop retailers, delivery drivers, mall Santas — millions of those jobs disappear. If you just looked at raw payroll data, “you would say that there is a massive recession every January,” said Wright.

Which, there isn’t. So when analysts look at data to figure out how the economy is trending, they often seasonally adjust the data. This involves taking the raw numbers and statistically smoothing them out to remove predictable seasonal swings — like how retail jobs and hard alcohol sales boom each December, or how construction jobs and beer sales spike each summer. 

“What we want is to be able to see through those patterns,” said Betsey Stevenson, professor of economics and public policy at the University of Michigan and former chief economist at the Department of Labor. She said those patterns “can be so big that they would mask and make it hard for us to see real fundamental problems that might be, say, slowing the economy.”

Seasonal adjustment lets us extract meaningful economic signals from the seasonal noise. It’s used in economic datasets ranging from GDP to inflation to employment. Wright said when we read headlines trumpeting a strong jobs report (353,000 new jobs!), “we should remember that deep inside the BLS [Bureau of Labor Statistics], someone has been doing this adjustment.”

One of those someones is Steven Mance, a supervisory mathematical statistician at the BLS. He said his bureau has been publishing employment data since the 1800s — originally just the raw payroll numbers. It was a different part of the government that decided in the 1920s that the non-adjusted data didn’t suffice. 

“The initial seasonal adjustment, a lot of it was done by the Federal Reserve,” said Mance.

The Fed needed a clear picture of how the fundamentals of the economy were shifting year over year; they couldn’t just slash interest rates each January when mall Santas hung up their hats. 

So the Fed got to work creating their own seasonally adjusted employment data based on the BLS’ raw numbers. Making those adjustments each month took time. Fed workers used “mechanical calculators and graph paper and draftsman’s tools,” Mance said.

But others outside the Federal Reserve needed seasonally adjusted data too, including economists, academics and business journalists.

So in the 1960s, the BLS decided to do the adjustment themselves (with the aid of fast and cheap computers) and publish the data for all to enjoy. Because, said Stevenson, “you really don’t need government stats to tell you that fewer people have a job in January coming out of the Christmas holidays.”

What you need, she said, is to know whether getting a job this January is harder than it was last January.