The following text is copyright 2003 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.
Reading into the FCC's 'Net access stats
By Scott Bradner
Another year has gone past and the FCC has issued its annual almost useful report on the state of high-speed Internet access deployment in the U.S. Some of the statisticians I used to work with would love this data because it is so easy to twist it to support almost any view, optimistic or pessimistic, about the future of the Internet.
Some of the news coverage of this report (http://www.fcc.gov/Bureaus/Common_Carrier/Reports/FCC-State_Link/IAD/hspd0603.pdf) called it a FCC report on the "status of broadband access" but the FCC is careful to not use the term "broadband." Maybe this is because they have defined "high-speed" to be 200 kbps in at least one direction, which is quite a bit less than most assumptions on what is meant by the term "broadband." The FCC also defines a second term, "advanced services," that means at least 200 kbps in both directions, which is also slower than what I would call "broadband."
At least one news story on the release of the report was headlined "high-speed Net growth slowing" because the growth rate of subscribers to high-speed Internet access was "only" 23% in the second half of 2002 while it had been 27% during the first half - never mind that an additional 1.3 million people subscribed in the second half compared to 1.1 million in the first half. I.e., the rate of the increase in growth slowed somewhat but the growth rate itself was still quite impressive.
The FCC uses an easy to get and easy to misuse measure of the availability of high-speed services - they check to see in what zip codes someone is getting high-speed Internet services. The current report says that someone is receiving high-speed service in 88% of the zip codes in the U.S. and the 99% of the US population lives in these zip codes. But using the provision of high-speed service to at least one subscriber in a zip code as a measure of the overall availability presents, at best, an optimistic view of the real world. It would be far more representative to have some minimum threshold of actual subscribers per zip code. What that threshold should be can not be guessed at from this data because there is no information on what percent of households (or small businesses) are receiving service in each zip code. I expect the FCC actually has this information but feels it can not provide the data because it would give away too much information about individual providers (if you look at the tables the FCC provides you will see that it already blocks a lot of data for this reason).
Another easy to misuse factor in the FCC data is the fact that they look at the speed of the "last few feet" (as the report puts it) when determining if a service is high speed. But they do not include any measure of actual available bandwidth - with the level of oversubscription that occurs in DSL and cable modem access networks, the final link speed can be a meaningless value.
All in all, it is better that the FCC provide these stats, even if it would be better if they would provide more real info.
disclaimer: Harvard has a department dedicated to the use, or misuse, of stats like this but I did not ask them for help in the above.