The US government is gearing up to begin the 2020 census which will be administered starting next April 20. For the first time, the census is going to rely heavily on people answering the census questions online. Live census takers will then follow-up with those that don’t submit the online response.

This seems like an odd decision since there are still many people who don’t have home broadband. This seems like a poorly conceived idea by those of us who understand the FCC’s dirty little secret—the FCC has no idea how many homes don’t have broadband.

As a country, we care a lot about an accurate Census. The census data is used for multiple government purposes. The 10-year census is used to redraw both federal and state political boundaries every ten years. The census is used to determine the number of US House Representatives allowed for each state. The government uses the census to allocate the funding for numerous federal programs that allocate funding by population. If an area of the country is undercounted, they lose both political representation and federal funding for a wide variety of purposes.

This all means that there is a significant downside risk for any part of the country that is undercounted in the census. The census is hiring 175,000 fewer door-to-door people nationwide to follow-up on those that don’t answer the first wave of the census, and one has to wonder if they are going to be equipped when vast portions of rural America doesn’t respond to the online census request.

As I said earlier, we have no idea as a country how many people don’t have home broadband. According to the FCC maps, there are still 21 million people in rural America with no access to broadband. However, everybody understands that this number is understated due to the idiotic rules used to count broadband customers by the FCC. We use a self-reporting system where ISPs tell the FCC about their broadband coverage. We know that many ISPs have overstated the speeds they can deliver along with the areas of their coverage. That’s bad enough, but the FCC then compounds this error by assuming that if a census block has at least one broadband customer, that the whole block has broadband. A census block is normally 600-800 homes, and anybody living in rural America understands how large such an area can cover.

We have other people counting broadband that paint a very different picture than the FCC. The one with the widest reach and most credibility is Microsoft. They are able to measure the speed of downloaded software upgrades—a method that tells the real broadband situation at a home. Microsoft estimates that 162 million people in the US don’t have access to broadband that meets the FCC’s definition of 25/3 Mbps. But Microsoft has no way of counting homes with no broadband.

This is not just a rural problem. It’s always been suspected that there are millions of homes in older urban areas that don’t have access to broadband. There are apartments and little pockets of neighborhoods everywhere that were bypassed by the cable companies when they built their networks in the 1970s and 80s. Folks who study this issue estimate that there could be as many as 10 million people in urban areas without broadband access.

Even more importantly, there are millions of people that elect not to buy broadband or who access the Internet only using a cellphone. There are still homes everywhere that either can’t afford the Internet or who refuse to go online. Even among houses with broadband, there are going to be many who don’t have good enough computer skills or the language skills to find and complete the Census questions online.

My guess is that the Census Bureau is going to be totally overwhelmed by the levels of non-response of households that don’t take the census online. There will be huge geographic rural areas where few people respond online. There will be people everywhere who don’t have access to broadband or are unable to navigate the online questionnaire.

In the past, the US Census Bureau believes they got a pretty high response. They got a decently high response from households that completed the paper census forms and had an army of census takers that tracked down houses that didn’t respond. If the completion ratio for the census slips even a few percents, then areas without good broadband are likely to be disadvantaged in the many ways that Census data affects states.

The census was moved online to save money. I think that the decision to go online is probably ten years premature and that the Census Bureau is probably totally unprepared for what’s going to happen next April. I hope I’m wrong.