the allocation of value for jointly provided services
DESCRIPTION
The Allocation of Value For Jointly Provided Services. By P. Linhart, R. Radner, K. G. Ramkrishnan, R. Steinberg. Telecommunication Systems, Vol. 4, 1995. Presented By :Matulya Bansal. Outline. Introduction The Telephone Carrier Network The Problem of Allocating Values - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/1.jpg)
The Allocation of Value For Jointly Provided Services
ByP. Linhart, R. Radner, K. G. Ramkrishnan, R. Steinberg
Telecommunication Systems, Vol. 4, 1995
Presented By :Matulya Bansal
![Page 2: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/2.jpg)
Outline
Introduction The Telephone Carrier Network The Problem of Allocating Values Co-operative Game Theory Shapley Value Solving the Caller ID Problem Example Conclusion
![Page 3: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/3.jpg)
Introduction
The Caller ID Service Revenue allocation is currently simplistic An allocation mechanism is needed when the
service is provided by more than one carrier
![Page 4: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/4.jpg)
The Telephone Carrier Network
Geographically distributed into Local Access and Transport Areas (LATAs)
Local Exchange Carriers (LECs) operate in LATAs
e.g. Regional Bell Operating Companies (RBOCs)
Long Distance Carriers or InterExchange Carriers (IXCs) provide InterLATA connectivity
e.g. AT&T, MCI, Sprint
![Page 5: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/5.jpg)
The Domestic Telecom Market
Local Calls (involve one LEC) IntraLATA Toll Calls (involve 2 LECs) InterLATA Long Distance Calls (invlove 1 or 2
LECs and 1 IXC)
![Page 6: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/6.jpg)
Basic Problem
How should the Caller ID Service Revenues be divided among the participating companies?
Or equivalently, what should be the payoff of players participating in this collaborative game?
![Page 7: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/7.jpg)
Desirable Properties
Stability : Players have an incentive to participate in the coalition (A solution which is stable is said to be in the core)
Fairness : The allocation should be perceived as in some sense fair
These considerations suggest the use of cooperative game theory
![Page 8: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/8.jpg)
The Core
Let N (1, 2, 3, …, n) be a set of players. Let v(N) be the value generated by the
coalition of all players participating in the Caller ID Game.
Let v(S) be the value generated by any subset S of players where v(S) >= 0 and v() = 0.
![Page 9: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/9.jpg)
The Core
Let N (1, 2, 3, …, n) be a set of players. Let v(N) be the value generated by the
coalition of all players participating in the Caller ID Game.
Let v(S) be the value generated by any subset S of players where v(S) >= 0 and v() = 0.
Let x be an allocation of total value among the players and
),....,,( 21 nxxx
Nii Nvx
)(
![Page 10: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/10.jpg)
The Core
Let N (1, 2, 3, …, n) be a set of players. Let v(N) be the value generated by the
coalition of all players participating in the Caller ID Game.
Let v(S) be the value generated by any subset S of players where v(S) >= 0 and v() = 0.
Let x be an allocation of total value among the players and .
A coalition N shall fall apart unless for every S N.
),....,,( 21 nxxx
Nii Nvx
)(
Nii Svx )(
![Page 11: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/11.jpg)
Shapley Value
One of the most-popular fairness criterion. Introduced by L. S. Shapley in 1953. Has been used for allocation of aircraft landing
fees, cost of public goods & services, water resources costs and depreciation.
![Page 12: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/12.jpg)
Shapley Value : Axioms
Domain Axiom – The allocation depends only on the values that can be earned by all possible combinations of one or more players acting in coalition.
Anonymity Axiom – The allocation does not depend on the players’ labels
Dummy Axiom – A player who adds nothing to the value of the coalition is allocated nothing
Additivity Axiom – If two allocation problems are combined by adding the characteristic function, then for each player the new allocation is just the sum of the earlier ones.
![Page 13: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/13.jpg)
The Shapley Formula
The Shapley Value is given by :
)]()([!
)!()!1(iSvSv
n
snsxi
![Page 14: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/14.jpg)
The Shapley Formula
The Shapley Value is given by :
The equation may be interpreted probabilistically as the expected marginal contribution of player I, assuming that the coalitions form randomly and that each coalition is equiprobable.
)]()([!
)!()!1(iSvSv
n
snsxi
![Page 15: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/15.jpg)
Shapley Value
In general, the Shapley Value need not be in the core. However, if the game is convex, the Shapley Value is in the core.
![Page 16: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/16.jpg)
Shapley Value
In general, the Shapley Value need not be in the core. However, if the game is convex, the Shapley Value is in the core.
Our game is convex : if a RBOC joins in a coalition, it adds value to itself as well as to the coalition.
where
)(}){()(}){( TvaTvSvaSv
TSaNTS },{,
![Page 17: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/17.jpg)
Subtracting the Singletons
The revenue an LEC generates on account of its IntraLATA traffic is really not up for negotiation.
So, we define a new characteristic function w(S) = v(S) – v’(S) where v’(S) is the total value generated due to
intraLATA calls. The additive property ensures that if v(S) is in
the core of game v, then w(S) is in the core of game w.
![Page 18: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/18.jpg)
Solving the Caller ID Problem
We need to calculate the characteristic function to
- determine if the allocation is in the core - calculate the Shapley value To do this in the absence of actual
experiments with all possible coalition structures, we require a demand model for the Caller ID.
![Page 19: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/19.jpg)
The Demand Model
Q : subset of subscribers that have the Caller ID facility (for Q)
: i’s willingness to pay (wtp) for this service
We assume that a subscriber’s wtp is a linear function of the number of calls received.
: number of calls from j to i : total number of calls received by I : average number of calls received The wtp function is assumed to be of the form
)(QLi
ijm
Qj
ijiji mzQL )(
im
)(Qm
![Page 20: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/20.jpg)
The Demand Model (contd …)
Define . Let F(x) be the probability that a subscriber I
drawn at random from Q will have an not exceeding x.
Hence, the total revenue is given by where
We wish to determine the price that maximizes revenue.
This gives the characteristic function to be
)(
)(
Qm
QLX i
iX
)](1[)( uFuQmR
)(Qm
pu
)]([)](1[)( ** SQmuFuSv
![Page 21: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/21.jpg)
Deriving the characteristic function
Assume that there are only 11 players (8 RBOCs and 3 IXCs)
Given : RBOC to RBOC AT&T Traffic
From/To
1 2 3
1 932961 485323
2 368220
3
![Page 22: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/22.jpg)
Approximating InterLATA Traffic
Approximate AT&T market share by RBOC Approximate interLATA traffic for MCI and
Sprint.
![Page 23: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/23.jpg)
Approximating AT&T market share by RBOC
AT&T’s market share in a RBOC in current year is given by the multiplying AT&T’s market share in the RBOC in a recent year with the ratio of AT&T’s US market share in current year to AT&T’s US market share in the recent year.
So, if in 1990 AT&T’s US Market Share was 60 million and it grew to 80 million in 1991 and AT&T’s share in a RBOC was 3 million, it grows to 4 million.
![Page 24: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/24.jpg)
Approximating InterLATA Traffic
The MCI (resp. Sprint) interLATA message volume for calls originating in a given RBOC can be approximated from the AT&T interLATA message volumes by multiplying by a proportionality factor.
1 2 3
MCI .194 .222
Sprint .132
![Page 25: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/25.jpg)
The Zero Normalized Function
X : set of Interexchange Carriers I : set of RBOCs : number of calls originating in l and
terminating in m. A denotes AT&T. : denotes the current market share for
in RBOC , divided by the current market
share for AT&T in that RBOC.
),,( AiiR ml
k
l
XiP kX
li
![Page 26: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/26.jpg)
The Zero Normalized Function
The zero normalized function w can be derived from the characteristic function by subtracting out the singleton coalition values representing IntraLATA calls. It is given by
0),...,( 1 NXXw
0),...,( 1 niiw
)],,([),...,;,...,( '
1111 AiiRPXXiiw
ll
N
k
Xi
n
lNn
k
l
![Page 27: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/27.jpg)
RBOC to RBOC AT&T Traffic
From/To 1 2 3
1 932961 485323
2 368220
3
![Page 28: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/28.jpg)
Proportionality Factors
1 2 3
MCI .194 .222
Sprint .132
![Page 29: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/29.jpg)
Calculating w
w(1,A) = R(1,1,A) = 932961
![Page 30: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/30.jpg)
Calculating w
w(1,A) = R(1,1,A) = 932961 w(1,M) = . R(1,1,A) = .194 * 932961 =
180944
MP1
![Page 31: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/31.jpg)
Calculating w
w(1,A) = R(1,1,A) = 932961 w(1,M) = . R(1,1,A) = .194 * 932961 =
180944 w(1,2,A) = R(1,1,A) + R(1,2,A) + R(2,2,A) +
R(2,1,A)
MP1
![Page 32: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/32.jpg)
Calculating w
w(1,A) = R(1,1,A) = 932961 w(1,M) = . R(1,1,A) = .194 * 932961 =
180944 w(1,2,A) = R(1,1,A) + R(1,2,A) + R(2,2,A) +
R(2,1,A) w(1,2,M) = [R(1,1,A) + R(1,2,A)] + [R(2,2,A) + R(2,1,A)]
MP1
MP1MP2
![Page 33: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/33.jpg)
Calculating Shapley Values
The Shapley values can now be calculated using
)]()([!
)!()!1(iSwSw
n
snsxi
![Page 34: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/34.jpg)
Other Notions of Fairness
Nucleolus- Tries to make the least happy player as happy as possible- Not monotone with respect to value
![Page 35: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/35.jpg)
Other Notions of Fairness
Nucleolus- Tries to make the least happy player as happy as possible- Not monotone with respect to value
Incremental Recording - Allocates points on a per call basis - Simple, but doesn’t guarantee fairness
![Page 36: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/36.jpg)
Conclusion
Two desirable properties for allocation of revenues for jointly provided services are Stability & Fairness
In general, the core contains several solutions Shapley value provides a solution that is stable
and fair. It also ensures marginality and anonymity.
The Caller ID Problem (and in general more allocation problems) can be solved by applying Cooperative Game Theory.
![Page 37: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/37.jpg)
References
“The allocation of value for jointly provided services”, P. Linhart et. al., Telecommunication Systems, Vol. 4, 1995.
“A value for n-person games”, L. S. Shapley, Contributions to the Theory of Games, Vol. 2, 1953.
![Page 38: The Allocation of Value For Jointly Provided Services](https://reader035.vdocuments.net/reader035/viewer/2022081515/56814495550346895db134e9/html5/thumbnails/38.jpg)
Thanks !
Presentation ByMatulya Bansal