The Census Bureau redesigned its questionnaire to improve on past surveys that underreported pension income. According to EBRI, the new questionnaire improved the accuracy of pension-income reporting.

But it also produced numbers on plan participation that deviate substantially from other government data.

According to the Census Bureau, in 2015 only 50 percent of full-time, private-sector wage and salary workers at firms with 500 or more employees participated in a retirement savings plan.

For the same year, data from the Bureau of Labor Statistics’ National Compensation Survey shows 76 percent of workers at firms with 500 or more employees participated in a plan.

“The estimates from the most recent (Census Bureau) surveys could easily be misconstrued as an erosion in coverage, as opposed to an issue with the data,” says EBRI’s report.

The revised estimates show declines in plan participation across all age cohorts. And participation dropped across all income levels. For higher earners making more than $75,000, the percentage of participation dropped nearly 16 percentage points between 2013 and 2015, from about 71 percent to roughly 55 percent, according to the revised survey.

The adjusted results show lower participation in the public sector labor force, and among employers of all sizes. For employers with fewer than 10 workers, participation dropped from about 16 percent to 13 percent between 2013 and 2015.

For employers with 50 to 100 employees, participation dropped from 42 percent to 33.5 percent.

All told, the Census Bureau’s revised survey shows the number of full-time wage and salary workers between the ages 21 and 64 participating in a plan dropped from 51.4 million in 2013 to 42.2 million in 2015.

EBRI’s report says the combined declines between 2013 and 2015 were “far larger than found for any other two-year period, even larger than during periods of high unemployment.”

The revised survey included detailed questions on pension and retirement plan participation used to better determine retirees’ incomes, but that new data was not incorporated in the data used to determine retirement plan participation, EBRI says.

In effect, the new retirement plan questions addressed to retirees are not made available in the Census Bureau’s public data set.

By making that data publicly available, a more accurate estimate of participation rates could be reached by combining the results from the income questions with the plan participation questions, says EBRI.

EBRI says the Census Bureau has no plans to revise its questionnaire.

Unless modifications are made to the Census Bureau’s Current Population Survey, “using the CPS for estimating the participation in pension and other retirement plans will provide misleading and inaccurate estimates and conclusions,” EBRI’s report says.