I first thought through the survey process. I’ve always found a well-designed survey to produce reliable results for questions like quantifying the market share of the major ISPs. I’ve worked with a few cities that had detailed customer penetration data from franchise fee reporting and in those cities our surveys closely matched that data. I’ve also worked in a few cities where we’ve done several surveys in a relatively short period of time and got nearly the same results from multiple surveys. I’ve come to trust survey results – as long as you follow good practices to make sure the survey is conducted randomly the results seem to be reliable.
I then turned to published industry statistics on the number of broadband customers to see what those told me. The two most cited statistics come from USTelecom and Leichtman Research Group (LRG). As of the end of 2017 US Telecom claimed that 79% of homes had a wired broadband connection, defined as any connection that is faster than 200 kbps, which eliminates dial-up. Leichtman Research Group claimed that 84% of homes had a wired broadband connection at the end of 2017 based upon a nationwide survey. Those numbers are significantly different. Luckily both groups also publish counts of national broadband subscribers, providing a second way to compare the two.
In the USTelecom Industry Metrics and Trends report from March 2018, US Telecom said there was 100 million residential broadband ‘connections’ at the end of 2017. They claim total broadband connections of 109 million when adding businesses.
Leichtman Research Group counts broadband ‘subscribers’ every quarter by gathering the statistics from the financial reports from the largest ISPs. LRG includes all of the big ISPs from Comcast down to Cincinnati Bell with 300,000 broadband customers. LRG claims these large companies represent about 95% of the whole broadband market. LRG counted 95.8 million total broadband customers at the end of 2017 – a count that includes businesses. Adjusting to add the remaining 5% of the market, LRG shows 100.8 million total broadband subscribers, including businesses – over 8 million less than what USTelecom counts.
That’s an astounding difference, and it’s obvious the two groups aren’t counting broadband customers the same way. There must be a difference between ‘subscribers’ and ‘connections’.
I’ve only come up with one reason why the counts would be that different. A lot of apartment complexes and business high rises today are served by a big data pipe provided to the landlord, who then provides broadband to tenants. I’m guessing that the LRG numbers considers the big data pipe to be one broadband customer. In most cases the LRG numbers come from quarterly financial reports to shareholders, and my guess is that ISPs consider a subscriber to be an entity that recieves a bill for broadband service.
I further postulate that USTelecom counts the number of tenants in those same buildings as ‘connections’. We know that big ISPs often do that. For example, AT&T agreed with regulators to pass 12.5 million new residences and businesses with fiber as part of their merger with DirecTV. It’s been clear that one of the big components of those new passings comes from units in apartment complexes. If AT&T was to build a fiber past an apartment complex they could count them as passings to satisfy the FCC without having had to get them as a customer.
The other component of the penetration rate equation is the number of US households. That number is just as confusing. I found a lot of different estimates of the number of US households. For example, the US Census says there was 137.4 million total living units at the end of 2017, with 118.8 million occupied living units. Statistica estimates 127.6 million households at the end of 2018. YCharts shows there are 122.6 million households at the end of 2018. That’s a wide range of ways to count potential residential customers in the country.
Finally, when trying to estimate the broadband penetration rates to be expected in cities, you have to back out the rural homes that can’t get broadband from the equation. That’s also a difficult number to pin down and I can find estimates that range from 6 million to 12 million homes with no broadband alternative.
The bottom line is that I don’t really know what I should expect as an urban broadband penetration rate. I can do math that supports a typical urban penetration rate of 92%. Mostly what I learned from this exercise is how careful I need to be when citing national broadband statistics – if you play it loose you can get almost any answer you want.