The great economic data crisis

Dion Rabouin, author of Markets


Illustration: Eniola Odetunde/Axios

Economists have long been disparaged for inaccurate predictions, but Friday's jobs report laid bare a new problem for the world's largest economy: questionable data.

Why it matters: Economic data is a crucial element in the movement of asset prices that determine what Americans pay for just about everything.
It's not just the stock market — the yield on U.S. Treasury bonds helps set rates for mortgages, student loans, credit cards and more.
Market moves also determine the value of assets like oil and the dollar, based largely on economic data.

Driving the news: The government's jobs report on Friday wasn't just much better than expected — showing the U.S. added 2.5 million jobs in May, 10 million more than economists predicted — it was full of inexplicable holes and numbers that contradicted other government readings.

To wit, as DRW Trading rates strategist Lou Brien points out, the Labor Department's unemployment insurance report showed that for the week ending May 16 there were 29,965,415 unemployed people receiving unemployment benefits.

The Labor Department's jobs report — which surveys individuals and businesses during the week of May 16 — found there were 20,985,000 unemployed people.

That would mean there were 9 million more people receiving unemployment benefits than there were unemployed people during the exact same survey week.

What they're saying: "Safe to say it is fair to be a bit skeptical of the numbers," Brien said in a note to clients.

Between the lines: The Labor Department also noted that only 35 states reported pandemic unemployment assistance numbers and just 22 reported claims for extended benefits during that week.
The extended benefits data was missing from the nation's second and fourth most populous states — Texas and Florida — suggesting the number of unemployed people is likely higher than the unemployment insurance data show, not lower by 9 million.

The big picture: Economic data is often incorrect or incomplete in its initial iterations, as it is based on human reporting and techniques as simple as making phone calls and filling out questionnaires.
What's different now is that the shock of the coronavirus pandemic is pushing the potential scale of error to previously unimaginable levels.
However, as Friday's trading action showed, the reports can still move markets.

The Labor Department's Bureau of Labor Statistics offered a bit of explanation for some of the irregularities in its numbers, pointing out that data collection for the jobs report was "affected by the coronavirus (COVID-19) pandemic."

How so: "Although [BLS regional data collection centers] were closed, about three-quarters of the interviewers at these centers worked remotely to collect data by telephone," BLS said in its May jobs report, also noting that no in-person surveys were taken during the month.
The pandemic led to a rate of responses to its survey of households that "was about 15 percentage points lower than in months prior to the pandemic."

There's more: The May nonfarm payrolls report included a “misclassification error” that would have made the unemployment rate "3 percentage points higher" than the reported 13.3%.
BLS said it was "investigating why this misclassification error continues to occur" as it's happened in the last three jobs reports.

Go deeper: Unpacking a surprise jobs report