A study conducted for the Commerce Department by the U.S. Census Bureau shows that there has been an increase in households that have mobile data as their only source of broadband. Certainly this huge survey that got 53,000 responses must be considered as statistically valid. One thing the survey showed is that broadband has become a commodity and more people than ever before have some sort of broadband.

The survey also showed that having landline broadband is tied to household income. Nearly one third of households that make under $25,000 per year have mobile data as their only broadband product. And that is up from 16% of these households in 2013. This must be a function of affordability and it’s not hard to see how lower income homes can’t afford both smartphones and landline broadband.

But the survey showed other trends that were a surprise. The survey shows that 18% of homes making between $50,000 per year and $75,000 per year have mobile-only data, a huge jump upward from 8% in 2013. And 17% of homes making between $75,000 and $100,000 are now mobile-only, another big jump up from 8% in 2013. The survey showed that even 15% of homes making more than $100,000 have gone to mobile-only data, another big increase from 6% in 2013.

But there are other industry statistics that don’t seem to jive with these numbers. All of the publicly traded ISPs routinely report customer counts and for 2015, as a whole, the industry showed a net increase of over 3 million new landline broadband connections last year. Industry analysts are expecting almost the same increase for the whole ISP industry in 2016. With about 125 million total households in the country that will mean that nearly 5% more of total households will have added landline broadband in 2015 and 2016. That huge increase in landline data connections seems to contradict the survey numbers.

Certainly there must be many people (particularly in urban areas) that can get by without a landline data connection. In cities there is more and more WiFi available. People have WiFi at work and more and more businesses in the city provide it as a convenience to their customers. And there are plenty of people living in tightly packed apartments that share a WiFi connection with a nearby neighbor.

But it’s hard to think this can account for the huge shift shown by the survey. If the survey is correct, then as many as 10% of all homes have abandoned a landline data connection since 2013 – over 12 million homes – and the industry numbers just don’t back that up.

Certainly any big user of data at home must have a landline connection. Nobody is going to use mobile data for watching video or playing online games. Nobody is even going to use mobile data for any serious amount of school work or for taking work home from the office. The small data caps make that far too expensive.

Our big cellular companies have priced mobile data to be among the most expensive broadband in the world when measured as cost per downloaded gigabit. Nobody who uses any significant data is going to use data that is as much as 100 times more expensive as a landline data connection unless they have no choice. There are certainly plenty of households in rural areas who use their cellular data connection as the only source of broadband. They use it very sparingly and still report getting huge monthly bills for mobile data.

I have no idea how to reconcile these two very different set of facts. We know the number of landline data connections that ISPs sell and that number tells a very different story than this survey. And while it’s anecdotal, I can’t think of one of my hundreds of clients who is not seeing annual increases in landline data customers.

But this survey asked the question to 53,000 respondents, and in the world of statistics this means the information received from the survey should be extremely believable. The survey results would be very suspect if the questions were misleading or if the survey was not administered randomly. But I would think that an organization like the Census would have been careful to have gotten these things right. This difference between these two sets of facts is a puzzle and I am sure that over time that the trend will become clear – whatever it is. But for now I am very skeptical of the survey results until they have been collaborated with other data.