Colorado’s employment counts have become so unreliable the U.S. Bureau of Labor Statistics, in an unprecedented move, has stopped issuing reports based on them.

“BLS has observed data quality problems because of ongoing issues with the modernization of Colorado’s unemployment insurance system. As a result, BLS is temporarily suspending publication of Colorado employment, unemployment, and wage data,” the bureau said in a statement issued Dec. 18.

The suspension was made in a program called State and Metro Area Employment, Hours & Earnings. It follows an earlier decision by the Quarterly Census of Employment and Wages, or QCEW program, to suspend metro- and county-level reports, as well as industry-level reports, for Colorado.

“We don’t make these decisions lightly. This is a very rare occurrence. There is a reason why we had to do it,” said Ric Wise, an economist with the QCEW program. Wise said he is unaware of any suspension going back to 2000, although South Carolina did have its reports footnoted in 2018 because of data quality concerns.

Every quarter, employers must file a report and pay unemployment insurance premiums they owe to the state. The reports include a count of workers on the payroll, wages paid, and so on. That data goes into the quarterly census and is used to benchmark information gathered from a monthly survey of around 5,000 worksites.

Because it is so comprehensive, the QCEW is considered the gold standard of employment information. But in Colorado, the BLS says the data coming out of the UI system can’t be trusted because of “unusual movements” that can’t be explained by the economy.

The problems go back to the third quarter of 2023 when the Colorado Department of Labor and Employment rolled out an upgraded unemployment insurance premium filing system called MyUI Employer+. The system promised to offer employers more self-service options and the ability to pay their premiums at the same time they filed their reports, rather than having to do them in separate steps.

However, it required employers to complete a one-time activation process using a PIN sent by the CDLE. The rollout went way off track, to the point that the compliance rate for establishments filing a timely report fell from an average of around 90% in the prior quarters to a dismal 28.3%.

The share of employees captured in the reports went from 95.6% in the prior quarter to 40.6% in the quarter when the new system kicked in. Concerned, experts at the BLS in March started troubleshooting with the state and tried to figure out what was going on.

“We believe it was a combination of learning the new system, a grace period for reporting after go-live and (employers) being unresponsive,” said Labor Market Information’s director Michelle Morelli.

Compliance did bounce back in the following quarters, but not to prior levels, and then it dropped again. In Q2 2024, the reporting rate for establishments was only 80.4%, the second lowest of any state after Michigan. Measured by employees covered, it rebounded to 89.7%, the lowest of any state and below the U.S. rate of 96.3%.

The response rate is an important consideration when it comes to data reliability, but not the only one. Part of the difficulty for the BLS is that it has no say in what vendors states can hire or how they operate their unemployment insurance systems, which are overseen by another Labor Department program.

The BLS, however, can stop publishing data it considers unreliable and that is what it did for Colorado after reviewing the Q2 2024 QCEW report when it was filed in October.

“The Colorado Department of Labor and Employment and its Office of Labor Market Information and Unemployment Insurance Division continue to work with the BLS for resolution, but no timeline has been set to lift the suppression,” Morelli said in an email.

Morelli said the state will continue to produce and publish its 2024 reports, including the Colorado Employment Situation report for December, which is scheduled for release on Jan. 27. What happens in 2025 is up in the air.

“Despite these irregularities, we are starting to see improvement in employer reporting,” Morelli said. She added the state has made “significant progress” in cleaning up the data in the Q2 2024 QCEW file submitted in early October.

Because of timing issues, the BLS won’t receive the cleaned-up file, which is available on the LMI Gateway portal, until after the Q3 report is filed in early January.

The revisions show that Colorado has 6.3% fewer establishments or workplaces than what was submitted to the BLS in October. The employment count is 19.4% higher and wages are 18.1% higher after the cleanup, Morelli said.

Aside from the QCEW problems, Colorado has experienced wild swings month-to-month and big revisions in its monthly survey reports. Even though the initial monthly estimates were all over the place, the QCEW could be counted on to set the counts right. Now the BLS is saying it can’t be counted on.

Counts of unemployment insurance claims are reliable, the BLS said, but not the data sets that provide “critical inputs” into the Current Employment Statistics, which count nonfarm payroll jobs, and the Local Area Unemployment Statistics, which help measure the unemployment rate.

“Without accurate Colorado employment levels, BLS will be unable to complete the 2024 benchmark process or to produce accurate monthly estimates for Colorado,” the BLS said. If employment reports aren’t released next year, economists will be essentially flying blind.

“It was sad to go on the BLS website and see their announcement about the quality of the Colorado data produced. The Colorado data is more unusual than it usually is,” said Broomfield economist Gary Horvath. “I hope BLS and LMI get the data problems figured out soon.”

In a sign of how volatile the counts have become, the CDLE’s Labor Market Information section estimated the state added 9,000 nonfarm jobs in October, a stellar performance versus other states, only to revise that down to 4,100 jobs in a report issued Friday.

It estimates that in November, the state lost 3,900 jobs, a miserable performance, with private employers cutting 5,500 net jobs and government employers adding 1,600. The state’s unemployment rate, calculated from a separate household survey, rose from 4.1% in October to 4.3% in November, marking a three-year high and surpassing the U.S. rate of 4.2%.

Plagued by rampant fraud, long wait times and the questionable denial and delay of benefits since the pandemic, the UI system has at times been a source of frustration for thousands of state residents seeking help.

And now it is for federal labor statisticians concerned that incorrect Colorado counts could undermine confidence in some key economic indicators.

©2024 MediaNews Group, Inc. Visit at denverpost.com. Distributed by Tribune Content Agency, LLC.