To hear the Federal Communications Commission tell it, this is a golden age for broadband access in the United States. According to a newly released report from the agency on the digital divide, the gap between rural and urban internet access has “narrowed substantially, and more Americans than ever before have access to high-speed broadband.” Between the end of 2016 and the end of 2017, the number of Americans without broadband access fell from about 26 million to about 21 million, the report found.

But experts, and even some commissioners at the FCC, say the report is flawed. The data underlying it, they argue, doesn’t truly capture what broadband looks like in rural America — leaving lawmakers and government officials with a warped view of internet access.

The report is an annual release, and started well before the Trump-appointed administration of chairman Ajit Pai. But the 2019 version of the report has come with an unusual amount of political baggage, and is raising hard questions over the quality of the data used in the process.

Internet access advocates and the democratic commissioners at the FCC have already slammed the report. “Regardless of the reporting standard here, the Commission’s mission is to close the digital divide,” Democratic FCC Commissioner Geoffrey Starks said in a statement after the report’s release. “And we must have accurate data about the problem we are trying to solve and the progress we are making toward solving it in order to make effective, data-driven decisions.”

To generate the report, the FCC relies on data from service providers. Using a form, the companies report on the census-derived “blocks” where they serve customers. Questions about the 2019 report started before it was even released: after an FCC press release put out in February trumpeted major gains in access, the nonprofit advocacy group Free Press noticed a major flaw in the figures. A small carrier, called BarrierFree, erroneously reported it served census blocks with nearly 62 million people, which would make it the fourth largest internet service provider in the country.

The FCC rectified the error before the release of the final report, reducing the number of people it believed to have access by about 2 million, but the fact that the flaw was uncovered by Free Press raised questions about how closely the agency was monitoring the data it received. Starks’ statement questioned the figures. “It’s incredible to me that an error this large — approximately 62 million in overstated broadband connections — didn’t materially change the report,” he said.

But questions about the report go beyond any one error. “There’s overstatement already baked into the system,” says Matt Wood, vice president of policy and general counsel at Free Press.

The census block system relies on data that only shows where providers could easily offer service, not where they actually do. If a company provides access to any customer in a census block, the entire block is also added to the count. Starks points out in his statement that some blocks are larger than 250 square miles.

The system is widely seen as imperfect at best, and the FCC has been considering changes to how it counts the data. Even the report itself admits to issues: the data “is not perfect,” it says, but adds that the “report is not an appropriate vehicle” to consider changes.

Without accurate figures, the government can’t take targeted action to improve conditions. The numbers, critics say, also give an inflated view of how hard the industry is working to connect customers who’ve long faced barriers to access.

Republican Commissioner Michael O’Rielly said in a statement approving the report that the data has been “rightfully criticized,” but that the trend was moving ahead positively. Democrat Jessica Rosenworcel said the report “deserves a failing grade.”

What could more accurate broadband access numbers look like? Some have pointed to a Microsoft study that calculated closer to 162 million Americans lacking broadband access — a substantially higher figure than the FCC found. But it’s hard to know what the figures could look like until the FCC requires companies to provide more granular data.

The report is also tasked with determining whether broadband is being deployed in a “reasonable and timely” way. Under the previous democratic administration, the FCC also used the census block data, but despite gains in access, determined it wasn’t moving along quickly enough. That determination changed under the Pai administration, which now says it is. Gigi Sohn — who worked as a counselor for the previous chairman, Tom Wheeler — points to conditions imposed on the merger of AT&T and DirecTV, under the Wheeler administration, as a substantial reason for the gains.

But regardless, she says, the report is wrong to suggest the job of connecting America is moving along quickly and smoothly. “We weren’t the ones going around saying everything’s peachy keen,” she says.