A whole slew of the critical numbers we use to understand this economy — what's happening with jobs, where prices are headed, how much oil supply and demand is likely — all come from the government. Data collecting agencies often generate those numbers by simply asking people and businesses how they’re doing.

The problem is, while the government keeps asking, people and businesses haven’t been answering as much as they used to. Survey response rates have been declining for decades. And when it comes to those jobs numbers, that's become a real problem.

To come up with the total of how many jobs are out there, how many were added, the unemployment rate, and more, the Bureau of Labor Statistics and the Census Bureau survey 120,000 businesses and government agencies and about 60,000 households, said Jon Schwabish, a senior fellow at the Urban Institute.

About two-thirds of households get back to them, and 93% of businesses and governments answer — eventually.

“It can turn out that companies just can’t answer that on time,” said Erica Groshen, former commissioner of the BLS.

Sometimes it takes months, which is why they publish revisions to the data. And a lot of businesses and people do respond, though the number is down a bit. 

The bigger issue is how many of them sign up to be surveyed in the first place. Groshen said those rates have gone down, too.

Out of all the businesses the government asks to work with it on the jobs report, the percentage that agree has gone down from 75% in 2015 to just 35% in April. In the end that means an overall response rate of less than half.

“Once you get response rates dropping below 50%, you start to get very concerned about what that is actually telling you about what’s going on in the economy,” said Paul Donovan, chief economist of UBS Global Wealth Management. “Because if you see a survey with a response rate below 50%, almost by definition the people who are filling in the survey are a bit weird.”

Not like weird-weird, but they are not the majority. They might be motivated by anger at the economy, or something else that just makes them more eager or able to participate, and that can introduce biases.

“BLS can correct by waiting, it can correct for the biases that it knows about, differences between industries, differences in employer size, things like that,” said former BLS chief Erica Groshen.

She said with these corrections the overall accuracy of the jobs numbers hasn’t changed much

“In technical terms, the standard error isn’t that much affected, because the sample is really huge. What has suffered most has been granularity,” Groshen said. 

As in, it’s harder to zoom into the economy see what’s happening in specific locations or specific industries. Basically, the jobs numbers are still good as a Monet — just not so good close up.