A new report from CMS said that effectuated enrollment in the Affordable Care Act (ACA) exchanges was lower than 1.9 million because of customers not paying their premium—with some blaming the monthly cost has the reason for ending their coverage.

The exchange enrollment of 12.2 million was already below expectations and below 2016 levels, with the drop partially blamed on the Trump administration’s decision to pull some TV ads in the final days of open enrollment when a number of younger customers tend to pick their plans. The new report updates that figure at 10.3 million people, taking into account 1.9 million customers who either never paid or didn’t continue their premium as of March 15.

A separate report outlined various reasons why customers would let their coverage be cancelled. Cost was cited by 46 percent who terminated coverage before paying their premium. Nearly half (49 percent) who terminated their plans after paying for at least one month’s premium said they had gained coverage elsewhere, most commonly through an employer.

“Consumers are sending a clear message that cost and affordability are major factors in their decision to cancel or terminate coverage,” CMS Administrator Seema Verma, MPH, said in a statement.

This post-open enrollment dropoff has been a trend since the exchanges opened as some consumers pick a plan but then don’t pay their premiums every month. At the end of March 2016, 11.1 million people were enrolled in exchange plans, down from the 12.7 million who had signed up during open enrollment.

A key difference in this report, however, is the time frame used. Previous reports judged effectuated enrollment as of the end of March, the last month where plans purchased in the 2015 and 2016 open enrollment periods would have begun. This first report from the Trump administration, uses March 15 as the cutoff date.

According to independent analyst Charles Gaba, this means data from an estimated 438,000 customers, whose coverage went into effect in March but didn’t pay their first premium until the second half of the month, were not counted in the CMS analysis. By using the March 15 cut-off date, Gaba said the net attrition rate looks worse (15.4 percent vs. his estimate of 11.8 percent).

“The only reason I can think of is … to provide data which is technically accurate but also highly misleading,” Gaba wrote.