social graphs of fcc lobbying
TRANSCRIPT
The evolution of lobbying coalitionsCo-filing behavior in FCC docket 01-92 on Inter-carrier
Compensation
Work in progress, 9/30/2009Pierre de Vries, Economic Policy Research Center
University of Washington, Seattle
Acknowledgements
Bill Cline and Elhadji Sy (FCC), for providing the underlying public data in a usable form
Marc Smith and Tony Capone, for developing and supporting the NodeXL visualization tool
Jonathan Banks and Anthony Jones (USTelecom), for help in extracting the “et al.” data, and Risa Pavia (UW) for help in creating the fix list
Conclusions
1. Graph-theory clusters represent real-world alliances
2. Tracking the evolution of clusters can reveal shifts in alliances
3. Improving ECFS interfaces and data quality will improve public knowledge of lobbying activity
Co-filing AnalysisMetadata from FCC ECFS database, FCC docket 01-92, April 2001 –December 2008
– 2,9015 filings, 756 unique filers
Subsidiaries and “doing business as” entities are usually grouped together
– e.g. Cable and Wireless plc, Cable & Wireless, Cable and Wireless USA, Cable & Wireless North America.
– But for some large players, kept parts separate, e.g. Verizon and Verizon Wireless; AT&T and AT&T Wireless
Sub-set of FCC docket 01-92 where two or more entities file together
– Entities that only filed on their own are not shown
– Used either metadata given as multiple entity names, or extracted entities from filed documents where “et al.” or “filed on behalf of” given in metadata
Trade associations and coalitions have not been unpacked into their constituents
– Sometimes distorts data, e.g. AT&T is represented both on its own account and hidden within the “Missoula Plan Supporters” node
– Coalitions unpacked: Oregon Rural LECs, Five State Regulatory Commissions, Coalition for Rational Universal Service and Intercarrier Reform
Company name changes on acquisition/merger not accounted for:
– Frontier Communications Corporation was formerly known as Citizens Communications Corporation
– Don’t distinguish between “old” and “new” AT&T, or Verizon before and after MCI merge
Filing intensity varies over time
NPRM comments (8/21/01)
Reply comments (11/05/01)
Responses (10/18/02) to T-Mo et al Petition for Declaratory Ruling (filed 9/6/02)
FNPRM comments (5/23/05)
FNPRM reply comments (10/20/05)
Responses (10/25/2006) to FCC PN (issued 7/25/06) on Missoula Plan (filed 6/24/06)
FNPRM issued 11/5/08
Links between 01-92 and other dockets
All linked dockets 2001-2008, filtered: – Sub-graphs level 1.5 centered on 01-92 (i.e. all nodes linked to 01-92, and links between them)
– edge weight >40 (dockets on either side of the edge were noted together on a filing more than 40 times)
Edge width and color both indicate edge weight: wider/pinker means more joint mentions
FEDERAL-STATE JOINT BOARD ON UNIVERSAL SERVICE
IMPLEMENTATION OF THE LOCAL COMPETITION PROVISIONS IN THE TELECOMMUNICATIONS ACT OF 1996
ACCESS CHARGE REFORM
In the Matter of Inter-Carrier Compensation for ISP-Bound Traffic
In the Matter of Lifeline and Link-Up
Numbering Resource Optimization
Request Petition for Declaratory Ruling that AT&T's Phone-to-Phone IP Telephony Services are Exempt from Access Charges
In the Matter of IP-Enabled Services
In the Matter of Universal Service Contribution
Methodology Federal-State Joint Board on Universal Service
1998 Biennial Regulatory Review
In the Matter of Establishing Just and Reasonable Rates for Local Exchange Carriers
In the Matter of Federal -State Joint Board on Universal Service High-
Cost Universal Service Support
Petition of AT&T for interim declaratory ruling and limited waivers pleading cycle
established
Companies typically either always file solo, or always jointly
All solo (0%)
(0%, 20%]
(20%, 40%]
(40%, 60%]
(60%, 80%]
(80%, 100%)
All joint (100%)
498
3924 25
8 10
152
Percentage of filings made jointly
Num
ber o
f fier
s 25 entities filed with others in 40%-60% of
cases, e.g. tw telecom, Pac-West
498 entities always filed alone, e.g.
BellSouth, NARUC
152 entities always filed with someone
else, e.g. Broadview, Maine PUC
Solo filers excluded from
co-filing analysis
Links between Filers
If the names of A and filer B both appear on a particular filing…
… they are treated as being linked
A B
The more often they file together… … the darker the line connecting them (think of the lines being
stacked one on another)
A B
Filed on Date 1
A B
Filed on Date 1
A BA B
Filed on Date 3
A BA B
Filed on Date 2
A B
Different co-filing occurrences… … lead to a network structure
Filed on Date 1
A B
So far the graphs looked at all filings simultaneously. Looking at a sequence of dates shows different links at different times:
Date 1 Date 3
A B
Filed on Date 2
A B
C
A B
C
A B
Filed on Date 3
B
C
Date 2
A B
C
A B
C
A B
C
Deriving a Network Structure
The area of the blob is proportional to the total number of filings made over the whole period (solo or joint)
Additional node attributes (1): Size
AT&T filed many times (big blob), whereas PointOne filed seldom (small blob).
However, one can see that they’re roughly equally linked to other filers. That means that AT&T filed more often on its own.
The more pink a blob is, the more important the filer is in the network.
The color represents the “eigenvector centrality”. A filer with high eigenvector centrality is connected to many filers who are themselves connected to many others. This “centrality metric” goes beyond simply counting the number of connections a filer has; connections to filers who are themselves highly connected confer more influence that connections to less well connected filers.
Google’s PageRank algorithm is a variant of this metric; a page is considered important if many other important pages link to it.
Additional node attributes (2): Color
• New Global Telecom and Verizon have the same influence in this graph (same color), even though Verizon filed more often (bigger blob)
• GCI, CompTel, and NCTA filed the same number of times (same size), but CompTel is the more influential (pinker), and GCI less (bluer)
• Even though CTIA filed often (big blob) it’s not very influential/connected in this network (blue color)
Nodes are laid out (by hand) to respect clustering
Clusters calculated using Clauset Newman Moore (2004) algorithm (Wakita & Tsurumi 2007 optimization) to find community structure, gathering vertices into groups such that there is a higher density of connections within groups than between them
A Time Series
Underlying data set has day-by-day granularity; these snapshots are integrated
over much longer periods
2003-2004
Intercarrier Compensation Forum,
filed ICF Plan 5 Oct 2004
“Indep. Wireless Carriers”: T-Mobile, W Wireless, Dobson
“CMRS Petitioners”: T-Mobile, W Wireless, Nextel
CLECs’ “Cost-Based Intercarrier
Compensation Coalition” (CBICC)
2005 – summer 2006
Major CLECs – FNPRM comments & replies
Rural LECs and their associations
CLECs
CLECs, some eventually merging e.g. Lightship, CTC, Conversent; and Xspedius &
tw telecom
Fall 2006 – end 2007Missoula Plan Allies
Missoula Plan Opponents:
Mix of CLECs, ILECs and Indep. Wireless
Oregon Rural LECs, supporting Missoula Plan
Aug/Sep 2008
CLECs opposing Verizon’s September 12
proposal, incl. uniform rate
ILEC/IXC coalition: Ex parte advocating federalizing VOIP, uniform comp rate
for all traffic
Oct 2008Five State regulatory
commissions objecting to “eleventh hour filings”
Mid-size rural LECs opposing flat rate comp, supporting status
quo OPASTCO/WTA Plan
Broadening CLEC coalition opposing change towards flat
rate
Small ILECs trying to slow down process
Nov/Dec 2008
Rural cellular – note they’re closer to the CLECs than the RLECs
Opposition to AT&T/IXC “self-help” from small
LEC and conf-call players
“Coalition for Rational Universal Service and Intercarrier Reform” –
urban & rural CLECs
Summary of Coalition Patterns
Rural LECs and their associations keep to themselves
Opponents are connected: ILECs, CLECs,
and cellular
Top 20 Impact Depends on Chosen Metric
Times Filed Number of Pages Joint Filings Connectedness (Degree)
Influence (Eigenvector)
AT&T Intercarrier Comp. Forum XO NuVox Hypercube
Verizon NTCA NuVox Cavalier Cavalier
NTCA Verizon Cavalier XO iBasis
CTIA Qwest Comm. Broadview Broadview NuVox
Qwest Comm. AT&T Pac-West RCN tw telecom
Verizon Wireless NuVox OPASTCO Pac-West Covad
XO XO RCN One CompTel
NuVox Broadview One tw telecom RCN
Level 3 Cavalier WTA T-Mobile One
T-Mobile Pac-West US LEC CompTel XO
Cavalier Verizon Wireless T-Mobile NCTA 360networks
Pac-West RCN tw telecom US LEC NCTA
USTA Nextel Focal Alltel PAETEC
Core Comm. US LEC Cbeyond Cellular South U.S. TelePacific
Sprint NASUCA Alltel McLeodUSA Citynet
US LEC CTIA McLeodUSA Covad Broadview
ITTA Core Comm. Dobson Hypercube nTelos
CenturyTel tw telecom Xspedius PAETEC R&B
Broadview Sprint Nextel iBasis Cellular South
OPASTCO BellSouth Western Wireless U.S. TelePacific Alpheus
* Filers that appear in three or more columns are color coded
CLECs not only band together, but also file a
lot, and often.NTCA carries the water for
RLECs
ObservationsGraph clusters seem to correspond to real-world interests
A large number of filings form one large connected graph (the blob in the center)– It’s connected in aggregate over the whole time series: shifting alliances
– In the course of the proceeding, one can find a link between opposing parties e.g. a proponent of the “Missoula Plan” like AT&T is linked to an opponent like XO via both of them co-filing with the CTIA at different dates (“Six Degrees of Kevin Bacon”)
There are many parties that only co-file once or twice– Most of them are pairs
– There are a few large groups of co-filers that show up only once in the record, and aren’t seen before or after
Frequent Filers are usually different from Cross-Connectors– Frequent filers like AT&T, CTIA and NTCA don’t often file in coalition
– Connectors that bridge alliances (e.g. Hypercube, iBasis) don’t file all that often
– Some fall in both groups, e.g. XO and Cavalier
Value of the approachInsiders can use graphs to identify:– detailed trends at a glance
– potential collaborators or defectors, e.g. by looking for coalition members who are bridges between groups, or peripheral
Outsiders can grasp the overall structure of a proceeding without having to read the entire record
News organizations can use:– cluster evolution to show changes in coalitions
– network structure to guide understanding of search results
Implications for FCCPoor quality of information input by filers impedes transparency– Endless misspellings of company names
– Not all entities involved in a filing are listed
– Filers make mistakes (e.g. mistyping docket number) but can’t remove mistakes; they simply file again
Require more information in metadata– Require all entities names to appear in the Filed on Behalf Of field, i.e. “et al.” not allowed
– Distinguish between ex parte letters and meetings
Use standard web techniques to facilitate data input and retrieval– Require log-in with company ID when submitting data to ECFS
– Require filer name to be registered; subsequently metadata can only be chosen from pre-registered information, not added de novo
– Offer drop-downs and auto-complete to add co-filers
– Provide web interfaces for downloading data in bulk, and as daily feeds
Improve systems for correcting errors– Allow filers to remove incorrect data (only filer can remove what was filed)
– Penalize errors, e.g. by naming and shaming
– Invest in cleaning up old data: requires public/private effort?
CaveatsGraph depends on metadata entered into ECFS by filers, which can be unreliable– misspellings (e.g. ATT for AT&T)
– inaccuracy (e.g. a filing attributed to AT&T was actually on behalf of a number of companies, and so is not counted as a co-filing)
– ambiguity (e.g. there are many companies whose name includes “Citizens”, and they all seem to be different)
Some Bad Data ExamplesBellSouth:
– BellSouth Coration – BellSouth Corpm – BellSouth Corps – BellSouth D.C. – BellSouth TELECOMMUN – BellSouth TELECOMMUNIC – BellSouth TELECOMMUNICA – BellSouth TELECOMMUNICAT – BellSouth TELECOMMUNICATI – BellSouth Telecommunication – BellSouth Cellular CORPOR
Misspellings of “Communications”:– Comminication– Commnications– Communictions – Commuications – Commuincations – Communcations – COMMUNIATIONS – Communicaitions – COMMUNICAITONS – Communicatiions – Communicatins – Communicationas – Communicationsn – COMMUNICCATIONS – Communictions – Communocations – Comunications – Coommunications – Cummunications
California PUC:– CALIFORNIA PUBLIC UTILIL – CALIFORNIA PUBLIC UTILIT – CALIFORNIA PUBLIC UTILITE – California Public Utilites Commission – CALIFORNIA PUBLIC UTILITI – California Public Utilities – CALIFORNIA PUBLIC UTILITIES COMMISSION – California Public Utilities Commission - 99-204 – California Public Utilities Commission and People of the State of
California – California Public Utiltiies Commission – Calilfornia Public Utilities Commission – Commissioners Lynch and Wood, California Public Utilities
Commission – People for the State of California and Ca. Public Utilities
Commission – People for the State of California and Cal. Public Utilities
Commission – People for the State of California and California Public Utilities
Commission– People of the State of California & Public Utilities Commission – People of the State of California and Cal. Public Utilities
Commission – People of the State of California and California – People of the State of California and California Public Utilities
Commission – People of the State of California and Public Utilities Commission – People of the State of California and the Cal Public Utilities
Commissoin – People of the State of California and the California Public Utilities
Commission – Public Utilities Commission of the State of California – State of California and the California Public Utilities Commission – State of California Public Utilities Commission – The People of the State of California and by proxy for CPUC – The People of the State of California by proxy for CPUC
Caveat (ctd): Coalitions are UnderstatedAnalysis puts a lower bound on connectivity– Some connections are not revealed through co-filing; entities may be
linked, e.g. through participation in a trade association, but not file together explicitly
– Co-filing is undercounted since we rely on the metadata entered into ECFS. Sometimes just one company name given, even though there multiple companies involved, or a list of company names may not include all filers. (This data can be obtained from the underlying document, but at the price of significantly more effort.)
– This analysis deals with only one docket; companies may co-file more frequently on other dockets