Counting the Numbers

Reader discusses current ISP rankings.

[Response to Top U.S. ISPs by Subscriber: 2001 Year End from February 11, 2002.]

[February 21, 2002]

Dear Editors:

Regarding your US ISP rankings published Feb. 11, 2002, I believe you are mixing data in arriving at the "Other US ISPs" total subscribers of 77.5 million. The footnote explains that you deducted the top 23 listed ISPs' subscriber numbers from the CyberAtlas US online population of about 143 million people (which CyberAtlas got from the NTIA, BTW). The problem is that subscribers and online population are not the same thing as the latter includes people with access at work (where many people may have access, but there is only one subscriber, the employer) as well as multiple members of a household (where one subscription potentially serves several adults and children).

A simple sanity check supports the impossibility of this 77.5 million remainder. If you divide 77.5 million by the total number of ISPs listed at http://thelist.internet.com, or 459, less the 23 listed in your table, you get an average subscriber base for the remaining 436 ISPs of almost 178,000 each. This average would seem to imply that at least a few of these unranked ISPs would significantly exceed the size of several of your ranked ones, which does not seem likely.

Rather than online population, you should be using an estimate of the number of actual subscribers, including households, schools, businesses and others. I have, in fact, been looking for just such estimates, but have not yet found any. Should you know of any source of such data, I hope you will share it with me and your other readers. Until then, I'm guessing the number is somewhere around 70-75 million.

Queried that thelist.internet.com has 7,800 ISPs, not just the 459 national ISPs, Buschman added:

I have also seen higher estimates of the number of ISPs (the 7,800 area), but The List only includes 459 by name in the US (the top portion is, or appears to be, the "all 50 states" ISPs) and some of those have been absorbed or shut down.

But in any case, you are still assuming a greater number of subscribers than there are households and business establishments in all the US. According to the Census Bureau, there are about 105 million households in the US and about 6.3 million business establishments (an "establishment" is a location, BTW). Being very generous, let's assume ALL business establishments and 72 percent of households have internet access (UCLA says 72.3 percent of Americans are online now, including those who only have access at work). That would give you a subscriber number of only 82 million (too high, IMO). Even allowing for multiple accounts at a few households and adding in schools and government offices, one could not reasonably conclude there are 143 million Internet subscribers in the US.

BTW, your article also states that 70 million of the 143 million online population "are reported to be residential users." If the other 73 million are "at work" users, they would have to be online at the 6.3 million business establishments or at schools or government offices. Surely one would not expect a separate subscription for each employee or student. Also, if the 70 million residential users include any number of people in multi-user households (i.e. couples with or without children, single parents, roommates, etc.), then there are less than 70 million residential accounts.

Think about it. If there are 143 million subscribers in the US market, then every single internet user must have his or her own individual subscription. In my household, there are 4 users, but only one subscription. Online population is to subscribers as apples are to oranges.

On further investigation, Buschman added:

Interesting. The ISPs (of the 459) I checked do claim to be national, but if you look closer, you'll find that they have the same access numbers (pick a few at random and check their access numbers in your city). Being "national" is a meaningless claim as it is apparently very easy to hang out a shingle (a virtual shingle, even) and call yourself an ISP. If there really are 7,800 of them in the US, I'd bet that more than 7,500 of them are of absolutely no economic or market significance and that many of them are a single "provider" operating under several different names. A large number are probably people who signed up as resellers in response to some MLM program spam promising them "$10k a month working from home."

Regards,

Robert D. Buschman,
CFO, Vicar Networks, Inc.