Even if We Had Zero FCC Data, We Could Estimate Coverage, Take Rates and Quality of Experience

If Federal Communications Commission internet access data were unavailable at all, could we still make educated assessments of broadband coverage, make reasonable estimates of take rates and understand where the shortcomings exist? Of course we could. 

More granular data arguably always is better, but it is mistaken to make a fetish out of FCC data collection. In a filing to the U.S. Federal Communications Commission, Incompass calls for better data collection about internet access, as virtually everybody agrees would be a good idea. 

But we can gain valuable insight into the state of fixed network internet access coverage, usage and performance in many other ways, using what we know about housing stock, occupancy of housing, account data reported by ISPs and surveys of the amount of substitution occurring, where people deliberately choose not to buy fixed network services (or linear video, voice or any other product). 

There are two big and different issues: how well internet service providers perform at supplying access, and how consumers respond to that availability. Take rates alone are not evidence of supply gaps, but often reflect consumer choices, for example. 

Incompass cites Pew Research surveys that  found that 17 percent  of U.S. adults are smartphone-only internet users, meaning that they do not have a traditional high-speed internet connection at home. Using the nationwide average of 2.5 persons per living unit, that suggests seven percent of U.S. homes are “mobile only” for internet access. 

Vacancy rates also matter, as an unoccupied  living unit will not generally be a candidate for purchasing of internet  access. No all living units are occupied at any given time. Vacancy rates can range from more than one percent for owned housing and up to seven percent for rental units. 

Owner-occupied housing units made up 57.9 percent of total housing units, while renter-occupied units made up 30.7 percent of the inventory in the first quarter 2020, according to Census Bureau data. Vacant year round units represented 8.8 percent of total housing units, while 2.6 percent were vacant for seasonal use. 

Approximately 2.2 percent of the total units were vacant for rent, 0.7 percent were vacant for sale only and 0.6 percent were rented or sold but not yet occupied. Vacant units that were held off market comprised 5.3 percent of the total housing stock – 1.5 percent were for occasional use, 1.0 percent were temporarily occupied by persons with usual residence elsewhere (URE) and 2.9 percent were vacant for a variety of other reasons.

Add it all up and 88.6 percent of the housing units in the United States in the first quarter of 2020 were occupied and 11.4 percent were vacant, according to the U.S. Census Bureau.

The Census Bureau also estimates total housing units at 140 million. That implies a potential buyer base of about 124 million units. 

If we deduct the “mobile-only” households (seven percent, or 8.7 million homes, that implies a potential buyer base of about 115 million locations. But not everyone actually uses the internet, which further reduces the addressable base of buyers. 

Though internet usage is virtually universal for adults below the age of 50, only 73 percent of adults over 65 use the internet. About 88 percent of people 50 to 64 use the internet. 

Eventually, adult usage will be virtually universal, but at the moment some percentage of homes might not buy internet access because they do not use the internet. About 15 percent of the U.S. population presently is 65 or older; about 13 percent are in the 50 to 64 age range. 

There are about 34 million households headed by someone 65 or older and perhaps 35 million households headed by someone age 50 to 64. That suggests a potential nine million non-internet homes headed by someone 65 or older, plus four million homes headed by someone 50 to 64. That suggests as many as 13 million households without a need for internet access. 

Subtracting those homes from the base of 115 million potential buyers give us a potential buyer base of 102  million homes. 

Since fixed internet access is sold to locations, not people, all we have to do is compare the current number of fixed network internet access subscriptions with that potential buyer base of 102 million homes to derive an estimate of adoption (take rates). 

Leichtman Research says there were a total of 101 million U.S. internet access accounts held by firms representing 85 percent of the customer base. That suggests total fixed network accounts at 119 million. Granted, some percentage of those accounts are sold to businesses, especially small businesses. 

There are about 30 million U.S. small businesses.  About half of all small businesses are home based and presumably use home internet. So small business probably represents an addressable opportunity of 15 million locations. 

Add that to the consumer addressable opportunity and the universe might be 117 million locations. That is less than the number of accounts already in service. 

That is not a direct measure of quality of experience or speed, but does suggest that most customers who want to buy fixed network internet already do so, whether small businesses or consumers. And there also are ways to assess quality of experience. But that is another exercise and post.

Is Gigabit Speed Really Available to more than 80% of U.S. Housing Units?

Some question statistics that gigabit internet access now is available (can be purchased) by about 84 percent of U.S. residents, especially when based on data reported to the Federal Communications Commission. 

Others might find the claim that gigabit access is not that widely available a bit incongruous, but not for reasons of FCC data reporting. The NCTA says 80 percent of U.S. homes now can buy gigabit speed internet access, up from about 63 percent in 2018.  

And since cable TV operators in the U.S. market have at least 70 percent installed base, looking only at cable TV data provides a non-duplicated view of access speeds. Assume for the moment zero supply of gigabit services by other internet service providers. 

According to the U.S. Census Bureau there are about 137.9 million U.S. housing units. Not all those units are occupied at any particular time, but ignore that for the moment, 

Roughly 8.8 percent of units are not occupied, typically. Vacant year round units represented 8.8 percent of total housing units, while 2.6 percent were vacant for seasonal use. 

Approximately 2.2 percent of the total units were vacant for rent, 0.7 percent were vacant for sale only and 0.6 percent were rented or sold but not yet occupied. Vacant units that were held off market comprised 5.3 percent of the total housing stock – 1.5 percent were for occasional use, 1.0 percent were temporarily occupied by persons with usual residence elsewhere (URE) and 2.9 percent were vacant for a variety of other reasons.

Add it all up and 88.6 percent of the housing units in the United States in the first quarter of 2020 were occupied and 11.4 percent were vacant, according to the U.S. Census Bureau. 

For the moment, ignore that. Retail consumer networks are not built to pass only “occupied” dwellings, but all dwellings in an area. If there are 137.9 dwelling units, with an average of 2.6 persons per household, then coverage of 80 percent of U.S. homes equates to 110.3 million locations. At 2.6 persons per home, that suggests 287 million people are in living units able to buy gigabit internet access from cable operators alone. 

If the U.S. population is 382.2 million, then some 75 percent of the U.S. population can buy gigabit internet access from cable operators alone, assuming no coverage provided by telcos or independent internet service providers. 

Those figures track closely with the FCC figures for “people” able to buy gigabit internet access. If you know anything about the way hybrid fiber coax networks are built, you also know that internet access speeds are designed to be the same at every end user node on the network. 

The architecture uses an optical fiber to node design, with very short electrical repeater segments (generally a few amplifiers) between the optical node and any location. Compared to the archaic all-electrical designs, that means top speeds do not decline with distance to any appreciable extent. 

The point is that if an HFC network is designed and built to support gigabit speeds, it will provide speeds close to that at all locations reached by the network, much as a fiber-to-home network would do. 

The point is that I cannot think of a good reason why the cable claim of passing 80 percent of U.S. home locations with gigabit service available is not believable. 

And that is assuming zero non-overlapping coverage by all other ISPs. After all, all ISPs build gigabit facilities where they believe the demand is greatest. Those also are the places where competition arguably is greatest, such as high-income suburban areas. 

That noted, surveys of rural telcos conducted by the NTCA have found that 25 percent of respondents offer gigabit internet access, while gigabit speeds are offered by a growing number of U.S. ISPs.  

 

Does a Big Expansion of FTTH Make Sense for AT&T?

By some estimates, AT&T passes 55 percent of U.S. homes. Assume total U.S. housing units number 139 million.

Assume a rental vacancy rate of seven percent and a homeowner vacancy rate of 1.4 percent. Assume the percentage of owned housing is 64.2 percent, implying there are 89.2 million owned homes.

In that case, there also are some 35.8 million rental units. That further implies 33.3 million occupied rental units, and some 88 million occupied owned housing units.

Altogether, that implies a total universe of about 121.3 million occupied U.S. housing units. In principle, that means 121.3 milion potential locations a communications service provider might sell services to.

Assume AT&T, passing 55 percent of those locations, therefore could sell fixed service to about 66.7 million locations. That would still be on the high side, as there are some locations--boats, trailers, rented rooms, very-rural locations--that probably are not “sellable” locations. Assume such locations represent one percent of locations, or about 670,000 locations.

So round the addressable base of locations to 66 million.

In its first quarter of 2019, AT&T reported $2.8 billion of internet access revenue, representing about 25 percent of entertainment group revenue of $11.3 billion. AT&T reported 13.8 million broadband accounts in total.

That implies a penetration rate of about 21 percent (so call that the installed base).

AT&T claims 3.1 million “fiber” customers (fiber to the home), with 12.4 million locations passed by FTTH. That implies a take rate of 25 percent, where FTTH is available. Granted, sales should increase over time.

AT&T believes it can boost that take rate to 50 percent over time. Verizon has been able to get a bit more than 40 percent take rates over time, so a 40-percent share target seems reasonable enough.

The issue is what percentage of total passed homes could profitably be upgraded to FTTH.  If AT&T by the end of 2019 has 14 million FTTH homes, that leaves 52 million homes remaining for potential FTTH upgrades.

It is difficult to determine what percentage of those 52 million homes might be amenable for FTTH upgrades. For the sake of argument, assume half those homes actually are in areas dense enough that FTTH is feasible at a cost of about $1,000 per location.

So assume a potential universe of 26 million potential FTTH homes. Assume an actual customer then requires $600 additional cost to activate.

That implies capex of about $26 billion to build the network. Assume AT&T could get 25 percent take rates for the new FTTH services. Here is where it gets tricky. We must assume AT&T already has about 21 percent take rates for internet access already. So what percentage of the new FTTH accounts are incremental, and what percentage are simply upgrades by current customers?

It seems unlikely that AT&T loses many customers by upgrading to FTTH. But that also means a new FTTH network with 25 percent adoption actually represents a net gain of perhaps four percent new accounts.

Even if all the gains at 40-percent adoption are new accounts, AT&T stands to gain about 19 percent new accounts by making the FTTH upgrades. That might represent 4.9 million new accounts.

That implies an investment of nearly $5 billion for customer premises capex. Ignoring time value of money and interest expense, assume capex is at least $31 billion.

Assume “average” internet access revenue of about $130 per quarter, per account, or $43 a month. Assume FTTH boosts average revenue per customer to about $53 a month.

That implies recurring revenue, at 40 percent take rates of about $636 per year, per account. Annual incremental revenue then is about $3.2 billion.

Assume gross margin is about 40 percent. That implies incremental free cash flow of about $1.3 billion annually. So the big question is whether it makes sense to invest $31 billion to earn an additional $1.3 billion in free cash flow.

Where are All the Unserved U.S. Households?

Since the “digital divide” is closing everywhere in the world, it simply stands to reason that the divide ought to be narrowing in the United States as well. That is not to say the divide closes completely, only that clear and steady progress is being made to supply better internet access to citizens who wish to buy it. 

The Federal Communications Commission says “the number of Americans lacking access to fixed terrestrial broadband service at 25/3 Mbps continues to decline, going down by more than 14 percent in 2018 and more than 30 percent between 2016 and 2018.” 

The FCC also notes that the number of Americans without access to 4G Long Term Evolution (LTE) mobile broadband with a median speed of 10/3 Mbps fell approximately 54 percent between 2017 and 2018.

Also, more than 85 percent of U.S. residents now have access (can buy) fixed terrestrial broadband service at 250/25 Mbps, a 47 percent increase since 2017. Over the same period, the number of Americans living in rural areas with access to such service increased by 85 percent, the FCC says. 

 Inevitably, some will lament the existence of differences; decrying a lack of perfection or simply arguing that the numbers are incorrect, arguing that the number of people without broadband access is 42 million or even as high as 162 million. 

It is not clear where those higher figures come from. Looking at connected households is revealing, however. 

The Federal Reserve estimates there are about 140 million housing units., defined as “a house, an apartment, a group of rooms, or a single room occupied or intended for occupancy as separate living quarters.” 

To be more precise, we also would have to account for households that either choose not to buy, or cannot easily buy. Some of those latter cases might be boats that serve as a residence, trailers or rooms rented inside homes where the resident does not buy internet access because the owner or manager of the property supplies the access. 

More than 16 million units are vacant at any particular time, leaving a total of perhaps 124 million units, which accords well with the estimate of 121.6 million households we get if we assume the U.S. population is 304 million persons, with an average household size of 2.5. Then there are 121.6 million households. 

That is the base of total locations fixed networks must reach. But a significant number of households choose not to buy fixed network access. 

Somewhere between 15 percent and 20 percent of U.S. homes are “mobile-only” for internet access, which might represent as much between 18 million and 24 million households. Those customers choose not to buy fixed network internet access, for whatever reason they choose. 

If so, then the number of locations who might buy fixed network internet access is on the order of 97.6 million to 103.6 million sites. 

If take rates for all homes (including the vacant units) are about 80 percent, then we would expect total fixed network accounts to number about 97.3 million locations.

Leichtman Research Group estimates that the largest U.S. telcos and cable companies have about 101.2 million accounts, but that includes business accounts. That matches fairly well the estimate that total fixed network accounts should be about 97.3 million in number. 

The point is that there are very few U.S. locations that do not already buy some form of internet access--mobile or fixed or both. That is difficult to square with claims that huge numbers of peop;le literally cannot buy service at 25 Mbps. 

Consider also that internet access routinely is available from satellite and other wireless and mobile platforms. 

Satellite broadband and fixed wireless operators traditionally have targeted rural homes and small businesses as their primary market, in the past said to include as many as 35 million locations. But estimates vary widely. Some say 80 million people live in rural areas, others say 46 million do, using the U.S. Census Bureau methodology. 

 Satellite broadband providers seem to have three million subscriptions, though some estimates (wrong, in my opinion) suggest that  6.76 percent of U.S. internet subscriptions are provided by satellite. 

Assume there are 139 million U.S. housing units, the high estimate, without adjusting for vacant units or other locations that cannot be wired. That implies nine million U.S. satellite broadband subscribers. No estimate I have seen--ever--suggests there really are nine million U.S. satellite broadband accounts. 

HughesNet believes 18 million homes are its market opportunity. Rental units alone might represent 6.6 million units, although not locations, as some of those units are in multi-family complexes. 

According to Urban.org, 13 million homes are owned by rural residents. Those figures roughly accord with HughesNet estimates of market opportunity. 

A more conservative estimate is that perhaps two percent to three percent of U.S. homes are the primary target for satellite broadband. That would include the most-isolated areas, where there are no terrestrial fixed networks using cabling. In many rural areas that are slightly more dense, wireless ISPs already operate. And, of course, there are many parts of rural areas served by cable operators or telcos. 

The point is that many homes already can buy 25 Mbps service, albeit from a satellite provider. 

A big issue is the presence of fixed wireless ISPs. According to Broadband Now, some 148.4 million U.S. residents are covered by fixed wireless ISPs. Assume an average household size of 2.5. That implies some 59 million rural locations already are reached by fixed wireless ISPs. 

Add all that up and some of us cannot fathom how 42 million to 162 million people actually are not able to buy 25 Mbps internet access.

Unknown "Homes Passed" Data Hampers Revenue Growth Estimates

Some important types of statistics and data are not collected because governments do not force firms or industries to collect it. For example, many governments think it is important to track data on where home broadband exists, where it does not, how fast it operates, who buys and who does not. 

Private firms often have important incentives to track and measure their own revenues, sales, profit margins and growth rates. Financial markets and accounting rules often require measurement of this sort. 

AT&T, for example, reports revenues for mobility, fixed network business revenues and consumer fixed network revenues from internet access, voice and other sources. But those are traditional financial metrics, not operating indices such as penetration or take rates, churn rates and new account gains. 

source: AT&T 

Nobody seemingly believes the same effort should be made to measure the number of home broadband provider locations or dwellings reached by various networks. Better mapping, yes. Metrics on locations passed? No. 

And yet “locations passed” is a basic and essential input to accurately determine take rates (percent of potential customers who actually buy). That input matters quite a lot to observers when evaluating the growth prospects of competitors, even if that figure does not matter much for policymakers, who mainly care about the total degree of home broadband take rates, on an aggregate basis. 

The U.S. Census Bureau, for example, reported some 140.5 million housing units housing units as part of the 2020 census. The estimate for 2021 units is 142.2 million units. Assume 1.5 million additional units added each year, for a 2022 total of about 143.6 million dwelling units. 

Assume vacancy rates of about six percent. That implies about 8.6 million unoccupied units that would not be assumed to be candidates for active home broadband subscriptions. The U.S. Census Bureau, though, estimates there are about 11 million unoccupied units when looking at full-time occupied status. That figure presumably includes vacation homes.

Deducting the unoccupied dwellings gives us a potential home broadband buyer base of about 132.6 million locations. 

More difficult is the degree to which access networks operated by any single contestant actually pass those locations, as firms generally do not report such numbers in quarterly financial or annual reports (they do not have to do so). 

And that is where estimations must be made. AT&T’s 2022 10-K report cites 14.2 million customer locations connected. Assume AT&T has about 20 percent take rates for its home broadband services where it operates. That implies a housing unit coverage of about 71 million dwellings. 

Assume AT&T has a higher take rate of about 39 percent where it operates fixed networks. That implies housing coverage of about 36 million dwellings. 

The estimate of 71 million home passings strikes me as too high, but the estimate of 36 million seems too low. In the past I have used the figure of 62 million homes passed for AT&T. 

Assume Verizon has about 10 million home broadband accounts, with a take rate of 40 percent (a bit high, probably, if we include copper access). That implies housing coverage of some 25.3 million dwellings. 

Leichtman Research Group has estimates of home broadband accounts that vary from company reports. LRG estimates that AT&T has some 15.4 million internet access accounts. The variance might come from business accounts not enumerated. 

Verizon’s consumer accounts might be overstated, as LRG estimates Verizon has about 7.5 million home broadband accounts, not 10 million. Using the LRG account figures, we might estimate Verizon home coverage of about 18.8 million homes, on the high side. 

ISPs

Subscribers at end of 2022

Net Adds in 2022

Cable Companies
Comcast	32,151,000	250,000
Charter	30,433,000	344,000
Cox*	5,560,000	30,000
Altice	4,282,900	(103,300)
Mediacom*	1,468,000	5,000
Cable One**	1,060,400	14,400
Breezeline**	693,781	(22,997)

Total Top Cable

75,649,081

517,103

Wireline Phone Companies
AT&T	15,386,000	(118,000)
Verizon	7,484,000	119,000
Lumen^	3,037,000	(253,000)
Frontier	2,839,000	40,000
Windstream*	1,175,000	10,300
TDS	510,000	19,700
Consolidated**	367,458	724

Total Top Wireline Phone

30,798,458

(181,276)

Fixed Wireless Services
T-Mobile	2,646,000	2,000,000
Verizon	1,452,000	1,171,000

Total Top Fixed Wireless

4,098,000

3,171,000

Total Top Broadband

110,545,539

3,506,827

source: Leichtman Research Group 

Assume Comcast has 31.2 million accounts, with take rates for home broadband of about 52 percent. That implies something on the order of 60 million households. 

Assume Charter Communications has a take rate of about 45.5 percent where it operates fixed networks. Assume Charter has approximately 30.8 million home broadband accounts. That implies a homes-passed figure of about 67.7 million homes. 

If there are 132.6 million U.S. occupied home locations, then Comcast and Charter can reach about 127.7 million of those locations, or about 96 percent of total, as Comcast and Charter essentially have unduplicated networks, not competing in the same geographies. 

That strikes me as unlikely, on the high side. An older rule of thumb is that Comcast and Charter reach about a third of total U.S. locations, each, for a possible reach of up to 66 percent of total U.S. home locations. 

Using different methodologies, I have in the past estimated that Comcast has (can actually sell service to ) about 57 million homes passed, while the Charter Communications network passes about 50 million homes, the number of potential customer locations it can sell to.

Verizon homes passed might number 18.6 to 20 million. To be generous, use the 20 million figure. 

AT&T’s fixed network represents perhaps 62 million U.S. homes passed. CenturyLink never reports its homes passed figures, but likely has 20-million or so consumer locations it can market services to. 

Ignoring the variance in potential customer locations passed, AT&T would seem to have the greatest opportunity in the home broadband space, if it can build optical access connections faster, as has the biggest home footprint and low home broadband market share. 

On the other hand, AT&T revenue is driven by mobility, not the consumer fixed network. So then the question has to be posed as "how much to invest in the consumer fixed network?" compared to other oportunities. A rational person might argue that answer is "not so much."

Capital availability--and financial returns--are always the issue. Even if it dramatically escalated fiber-to-home capital investment, it is not clear AT&T would gain as much new revenue, compared to investing in mobility or business services, for example.

The point of the wider exercise is that we are forced to guess about how many homes each of the major fixed network contestants actually can reach. That, in turn, affects our ability to estimate adoption rates and potential growth opportunities. 

The key point is that the estimates are imprecise. Pinning down the “homes passed” figure, essential as the denominator in any calculation of take rates, requires estimations with variable degrees of uncertainty, especially for the larger networks.