'Zero provider' zip codes not previously provided ... FC

1
How Useful are Current US Broadband Data?

• Kenneth Flamm
• Technology, Innovation, and Global Security
  Program
• The Robert S. Strauss Center for International
  Security and Law
• University of Texas at Austin
• "Broadband Data What Do They Tell Us and Do We
  Have Enough?"
• Washington, D.C.
• June 28, 2007

2
Are FCC zip code data misleading?

• Break in statistics in 2005
• reported providers doubles from 12/04-6/05
• Previous numbers not comparable
• Zero provider zip codes not previously provided
• Erroneous assumptions used by researchers to
  estimate zero provider areas
• Now known, no obvious fix to public released data
• Zip code areas used by FCC constantly changing
• Do not correspond to Census ZCTA standard
• High turnover, changing areas complicate
  intertemporal comparisons
• Mix of point and geographic zips, mappings
  change frequently
• Changing subset of both point and geo zips
• 41 1999 postal zips used in 6/05 dropped in 12/05
• 20 1999 postal zips not used in 6/05 show up in
  12/05
• Undocumented, proprietary mapping
• Zip codes often big areas
• Service in one corner does not mean available in
  other corner
• Typically about 3700 households per zip code
• Vs. zip 4s about 10 households

3
FCC BB Availability s Clearly Overstated From
Household POV

• Example, me zip code 78731
• FCC says 24 providers competing for my business
  12/05
• But no DSL available in my neighborhood
• Cable only choice
• Advertised satellite unavailable
• Correct number on my block was 1
• Example, Kentucky
• FCC data show 96 of pop in zip codes w/BB as of
  12/04
• Detailed Connect Kentucky survey shows 77 of
  households have BB availability
• Example, fooling with FCC data
• GAO (had access to non-public data) took out
  satellite-only, business only, DSL gt 2.5 miles
  from CO, double-counted cable systems
• Median US provider number per zip code fell from
  8 to 2, pop with no providers at all went from
  1 to 9
• Example, Vermont
• FCC says all zip codes in Vermont had broadband
  in 2005
• Interviewees in September 2006 NY Times article
  say otherwise
• Bottom line FCC providers per zip probably best
  viewed as ordinal indicator of availability
  within zip
• More providers ? higher probability that
  individual household can get service
• Less providers ? lower probability that can get
  service
• No providers ? Pretty darned tough to get service

4
Other USG agencies also not particularly helpful

• BLS uses obfuscating categories in price data
  collected for Internet services PPI
• Dialup and DSL access (!) lumped into one
  category
• Leased lines and DSL in a second category
• Cable and all other forms of Internet access
  lumped into third category
• Not very useful if interested in understanding
  what is going on in broadband
• Fixed weight price indexes also not very useful
  in rapidly changing industry
• Census/Commerce discontinued the one real
  scientific survey that measured Internet access
  modes and prices years ago
• The Internet is growing less important?
• Private consulting surveys have undocumented
  methodologies
• And are, wellprivate!
• Public Pew national survey extremely valuable,
  but very small sample size, cant answer
  geographic/regional questions
• Also a series of one-off surveys, not an
  instrument designed to maintain regular and
  consistent data over time

5
FCC Level 0 Disconnectedness

• Hard core disconnectedness by state
• Absolutely no providers of BB service in FCC
  stats
• 183,000 people, .065 of US pop in 6/05
• .033 US pop in 12/05
• Most Level 0 disconnected states
• Kentucky (gt1 in 6/05)
• Followed by West Virginia, Dakotas, Montana
• Puerto Rico and Vermont among least Level 0
  disconnected states
• Should give us pause!
• pop disconnected generally declines from 6/05
  to 12/05, but some exceptions
• Minnesota, Illinois, Virginia, Washington,
  Indiana
• Very small problem
• Less than .1 pop for US overall
• No state gt 1 pop in 12/05

6
FCC Level 1 Disconnectedness

• lt 4 bb providers in FCC stats
• Significant probability that substantial
  households do not have access to bb
• Very different picture
• Puerto Rico now leads for June 2005, almost 2/3
  of pop (falls to 22 in Dec 2005)
• Right behind, Dakotas, West Virginia, Montana
• Vermont now well above national average in
  disconnectedness
• Share of pop in Level 1 generally declines over
  6-12/2005
• Some exceptions
• Georgia, Illinois, Washington, New York,
  Michigan, Maryland, Massachusetts, Delaware,
  Arizona, California, Connecticut, New Jersey.but
  increases generally very small
• In Illinois, Delaware, Arizona, gt1 pop now more
  disconnected
• Level 1 Disconnected pop coincidentally quite
  close to estimates of pop w/o broadband
• Nationally, GAO says 9 12/04 vs. 7 Level 1 for
  6/05
• ConnectKentucky says 23 lack access to bb spring
  05, vs 19 in Level1 disconnected zips in 6/05
• Conservative rule of thumb?
• Not such a tiny problem
• In 28 US states (including PR), Level 1 gt 5 in
  12/05
• 12 states gt 10
• 4 states (PR, Dakotas, WV) gt 20

7
Questions

• Why not add zip 4 to existing system?
• Records already maintained in this format by all
  major service providers
• If using zips - 4, why not use ZCTA standard used
  by all other USG statistical agencies, enable use
  of other standard data collected by USG agencies?
• Why not break into wired (cable, DSL, fiber, T-1,
  ISDN, etc.) vs. wireless (wi-fi, satellite, etc.)
  classification?
• Hard to understand how truly proprietary info
  revealed
• Why not break into speed bins?
• E.g., .2-1, 1-11, 12-54, 55-110, 111-300,
  301-1K,1K-10K, 10K-100K, 100K mbps
• At least collect on this basis, present in
  aggregate form
• Nominal (peak speed) OK as long as consistent
• Up vs. down

8
Backup
9
Level 0 Disconnectedness
10
Level 1 Disconnectedness
11
(No Transcript)