Abstract
Congestion models may be studied from either the users’ point of view or the social one. The first perspective examines the incentives of individual users, who are only interested in their own, personal payoff or cost and ignore the negative externalities that their choice of resources creates for the other users. The second perspective concerns social goals such as the minimization of the mean travel time in a transportation network. This paper studies a more general setting, in which individual users attach to the social cost some weight r that is not necessarily 0 or 1. It examines the comparative statics question of whether higher r necessarily means higher social welfare at equilibrium.
Similar content being viewed by others
Notes
In a more general setting, different individuals \(i\) may have different altruism coefficients \({r}_{i}\) (see Chen and Kempe 2008; Hoefer and Skopalik 2013) or attach different weights \({r}_{ij}\) to the personal payoffs of different other persons \(j\) (see Rahn and Schäfer 2013; Anagnostopoulos et al. 2015). The present work does not cover these extensions.
A strategy space is a strategy set that is endowed with a particular topology, with respect to which continuity and related terms are defined.
Obviously, the price of anarchy cannot be strictly increasing for any single game or a finite family of games. With finitely many games, there are only finitely many possible values for the social cost, which means that \({\text{PoA}}^{r}\) can be at most nondecreasing.
Since a route in an undirected network may traverse an edge in either direction, the meaning of this “non-redundancy” condition is slightly different than in the case of a directed network. However, it is shown in “Appendix: Networks” that an undirected network satisfies the condition if and only if it is the undirected version of some directed network as above.
The original meaning of extension-parallel network concerned directed networks. An alternative term for the undirected version of these networks, which is the version considered here, is networks with linearly independent routes. These undirected two-terminal networks are characterized by the property that each route includes at least one edge that does not belong to any other route (Milchtaich 2006b).
These authors actually establish their results for directed networks (and nonnegative costs). However, it is not very difficult to conclude from these results that they hold also for undirected networks, for example, by using Proposition A1 and Corollary A2 in “Appendix: Networks.”
Note that this (asymmetric) game differs from the network congestion game considered in Sect. 3.3 in that edges can be traversed in both directions.
A sufficient condition for Borel measurability of a function is that it is continuous.
A function is finitely-many-to-one if the inverse image of every point is a finite set.
A sufficient condition for absolute continuity is that the function is continuously differentiable.
Technically, this assumption means that there is an extension of each payoff function to an open neighborhood of \(y\) in the \(n\)-dimensional Euclidean space where it has continuous second-order partial derivatives.
A square matrix \(A\) is said to be positive or negative definite if the symmetric matrix \((1/2)(A+{A}^{\mathrm{T}})\) has the same property, equivalently, if the latter’s eigenvalues are all positive or negative, respectively.
Because of a typo, the condition is presented in the paper in a slightly altered, stronger form (Nahum Shimkin, personal communication).
The modified payoff is linear also in \(r\). However, this linearity does not represent a very strong assumption because multiplying the expression on the right-hand side of (1) by any positive-valued function of \(r\) would not affect the preferences expressed by it. In other words, only the ratio \(\left(1-r\right):r\), which gives the marginal rate of substitution of personal payoff for social payoff, matters.
References
Altman, E., Basar, T., Jimenez, T., Shimkin, N.: Competitive routing in networks with polynomial costs. IEEE Trans. Autom. Control 47, 92–96 (2002)
Anagnostopoulos, A., Becchetti, L., de Keijzer, B., Schäfer, G.: Inefficiency of games with social context. Theory Comput. Syst. 57, 782–804 (2015)
Bochnak, J., Coste, M., Roy, M.-F.: Real Algebraic Geometry. Springer, Berlin (1998)
Caragiannis, I., Kaklamanis, C., Kanellopoulos, P., Kyropoulou, M., Papaioannou, E.: The impact of altruism on the efficiency of atomic congestion games. Lect. Notes Comput. Sci. 6084, 172–188 (2010)
Chen, P.-A., Kempe, D.: Altruism, selfishness, and spite in traffic routing. In: Proceedings of the 9th ACM Conference on Electronic Commerce, pp. 140–149 (2008)
Chen, P.-A., de Keijzer, B., Kempe, D., Schäfer, G.: Altruism and its impact on the price of anarchy. ACM Trans. Econ. Comput. 2(4), 17 (2014)
Cominetti, R., Correa, J.R., Stier-Moses, N.E.: The impact of oligopolistic competition in networks. Oper. Res. 57, 1421–1437 (2009)
Edgeworth, F.Y.: Mathematical Psychics. Kegan Paul, London (1881)
Epstein, A., Feldman, M., Mansour, Y.: Efficient graph topologies in network routing games. Games Econ. Behav. 66, 115–125 (2009)
Fotakis, D.: Congestion games with linearly independent paths: Convergence time and price of anarchy. Theory Comput. Syst. 47, 113–136 (2010)
Fotakis, D., Kontogiannis, S., Spirakis, P.: Selfish unsplittable flows. Theoret. Comput. Sci. 348, 226–239 (2005)
Gairing, M.: Malicious Bayesian congestion games. Lect. Notes Comput. Sci. 5426, 119–132 (2009)
Heifetz, A., Shannon, C., Spiegel, Y.: The dynamic evolution of preferences. Econ. Theory 32, 251–286 (2007). https://doi.org/10.1007/s00199-006-0121-7
Hoefer, M., Skopalik, A.: Altruism in atomic congestion games. ACM Trans. Econ. Comput. 1(4), 21 (2013)
Holzman, R., Law-Yone, N.: Strong equilibrium in congestion games. Games Econ. Behav. 21, 85–101 (1997)
Holzman, R., Law-Yone, N.: Network structure and strong equilibrium in route selection games. Math. Soc. Sci. 46, 193–205 (2003)
Karakostas, G., Viglas, A.: Equilibria for networks with malicious users. Math. Program. 110, 591–613 (2007)
Kuniavsky, S., Smorodinsky, R.: Greediness and equilibrium in congestion games. Econ. Lett. 121, 499–503 (2013)
Meir, R., Parkes, D.: Playing the wrong game: Smoothness bounds for congestion games with behavioral biases. ACM SIGMETRICS Perform. Eval. Rev. 43, 67–70 (2015)
Milchtaich, I.: Social optimality and cooperation in nonatomic congestion games. J. Econ. Theory 114, 56–87 (2004). https://doi.org/10.1016/S0022-0531(03)00106-6
Milchtaich, I.: Comparative statics of games between relatives. Theor. Popul. Biol. 69, 203–210 (2006a)
Milchtaich, I.: Network topology and the efficiency of equilibrium. Games Econ. Behav. 57, 321–346 (2006b)
Milchtaich, I.: Comparative statics of altruism and spite. Games Econ. Behav. 75, 809–831 (2012)
Milchtaich, I.: Representation of finite games as network congestion games. Int. J. Game Theory 42, 1085–1096 (2013)
Milchtaich, I.: Network topology and equilibrium existence in weighted network congestion games. Int. J. Game Theory 44, 515–541 (2015)
Milchtaich, I.: Static stability in games. Bar-Ilan University, Research Institute for Econometrics and Economic Theory, Working Paper 1-20, April 2020 (2020)
Monderer, D., Shapley, L.S.: Potential games. Games Econ. Behav. 14, 124–143 (1996)
Rahn, M., Schäfer, G.: Bounding the inefficiency of altruism through social contribution games. Lect. Notes Comput. Sci. 8289, 391–404 (2013)
Rosen, J.B.: Existence and uniqueness of equilibrium points for concave N-person games. Econometrica 33, 520–534 (1965)
Rosenthal, R.W.: A class of games possessing pure-strategy Nash equilibria. Int. J. Game Theory 2, 65–67 (1973)
Singh, C: Marginal cost pricing for atomic network congestion games. Technical Report (2008)
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix: Networks
Appendix: Networks
A directed or undirected two-terminal network is, respectively, a directed or undirected multigraph (where a pair of vertices may be joined by more than one edge) that is endowed with two terminal vertices, \(o\) and \(d\), and satisfies the following “non-redundancy” condition: each edge \(e\) and each vertex \(v\) is included in at least one route in the network. In the undirected case, a route is defined as a (simple) path of the form \(o{e}_{1}{v}_{1}\cdots {v}_{n-1}{e}_{n}d\) (\(n\ge 1\)), that is, one that begins at the origin vertex \(o\) and ends at the destination vertex \(d\). (An alternative, simpler notation is \({e}_{1}{e}_{2}\cdots {e}_{n}\).) In the directed case, a route also has to traverse each of its edges \(e\) in the direction assigned to \(e\), that is, the tail and head vertices of \(e\) must be its immediate predecessor and successor, respectively, in the route. This difference means that the set of routes in a directed two-terminal network is a subset of the set of routes in the network’s undirected version, which is obtained by ignoring the edge directions.
An arbitrary assignment of directions to the edges of an undirected two-terminal network \(G\) does not necessarily give a directed network, as the non-redundancy condition may fail to hold. If the non-redundancy condition does hold, the resulting directed two-terminal network is said to be a directed version of \(G\). Such a version always exists.
Proposition A1
Every undirected two-terminal network has at least one directed version.
Proof
The proof is by induction on the number of edges in the network. If there is only one edge, the assertion is trivial. To establish the inductive step, consider a network \(G\) with more than one edge, and some edge \(e\) incident with the origin \(o\). If no other edge is incident with \(o\), then by the induction hypothesis there exists at least one directed version of the network obtained from \(G\) by contracting \(e\), that is, eliminating the edge and its non-terminal vertex \(v\) and replacing \(v\) with \(o\) as the terminal vertex of all the edges originally incident with \(v\). Any such directed version gives a directed version of \(G\), in which \(e\) is directed from \(o\) to \(v\). Suppose, then, that \(e\) is not the only edge incident with \(o\), and consider the subnetwork \({G}_{e}\) of \(G\) whose edges and vertices are all those that belong to some route in \(G\) where the first edge is \(e.\)
Claim If a route \(r\) in \(G\) includes a non-terminal vertex \(u\) that is in \({G}_{e}\) , then every edge (hence, also every vertex) that follows \(u\) in \(r\) is also in \({G}_{e}\) .
It clearly suffices to consider the edge \({e}^{{\prime}}\) that immediately follows \(u\) in \(r\). Suppose that \({e}^{{\prime}}\) is not in \({G}_{e}\). Let \(v\) the first vertex in \(r\) that follows \({e}^{{\prime}}\) and is in \({G}_{e}\). (Possibly, \(v=d\).) All the edges and vertices in \(r\) between \(u\) and \(v\) are not in \({G}_{e}\). Adding them to it creates a new subnetwork of \(G\), since the addition is equivalent to adding to \({G}_{e}\) a single edge with end vertices \(u\) and \(v\) and then subdividing the edge, if needed, one or more times (Milchtaich 2015, Sect. 2.2). By definition, there exists in the new subnetwork a route \({r}^{{\prime}}\) that includes \({e}^{{\prime}}\). Necessarily, \({r}^{{\prime}}\) also includes \(e\). The conclusion contradicts the assumption that \({e}^{{\prime}}\) is not in \({G}_{e}\), and thus proves the claim.
Consider now the collections of all edges and vertices that are obtained from those in \(G\) by (i) eliminating all edges that are in \({G}_{e}\) and (ii) identifying with \(d\) all non-terminal vertices that are in \({G}_{e}\). It follows from the claim that these collections constitute a network \({G}^{{\prime}}\) (with the terminal vertices \(o\) and \(d\)). To see this, consider any edge or non-terminal vertex in them and a route \(r\) in \(G\) that includes it. It follows from the claim that the inclusion still holds if \(r\) is replaced with its section \({r}_{ou}\), which is the path obtained from \(r\) by deleting all edges and vertices that follow \(u\), where \(u\) is the first vertex in \(r\), other than \(o\), that is in \({G}_{e}\). As \(u\) is one of the vertices identified with \(d\), \({r}_{ou}\) is a route in \({G}^{{\prime}}\).
By the induction hypothesis, both \({G}_{e}\) and \({G}^{{\prime}}\) have directed versions, which together specify directions for all edges in \(G\). It remains to show that these directions define a directed version of \(G\), that is, the non-redundancy condition holds. Any edge or vertex that is or is not in \({G}_{e}\) is included in some route \(r\) in the directed version of \({G}_{e}\) or some route \({r}^{{\prime}}\) in the directed version of \({G}^{{\prime}}\), respectively. By construction, \(r\) is automatically a route also in \(G\) and it honors the directions specified for its edges. Only the second assertion is necessarily true for \({r}^{{\prime}}\), which is a path that starts at \(o\), ends at some vertex \(u\) in \({G}_{e}\), and does not include any other vertex or edge that is in \({G}_{e}\). However, it is possible to extend \({r}^{{\prime}}\) to a route in \(G\) that honors the edges’ directions by replacing the vertex \(u\) with the section \({r}_{ud}^{{\prime\prime} }\) of some route \({r}^{{\prime\prime} }\) in the directed version of \({G}_{e}\) that includes \(u\), where \({r}_{ud}^{{\prime\prime} }\) is obtained from \({r}^{{\prime\prime} }\) by deleting all the edges and vertices that precede \(u\). □
Corollary A1
The collection of the undirected versions of all directed two-terminal networks coincides with the collection of all undirected two-terminal networks.
Clearly, an assignment of directions to the edges of an undirected two-terminal network \(G\) defines a directed version of \(G\) only if it satisfies the condition mentioned in Sect. 3.3, which is that the direction of each edge \(e\) coincides with the direction in which some route in (the undirected network) \(G\) traverses \(e\). However, this necessary condition is not sufficient. For example, it is not difficult to see that an assignment satisfying the condition may render a non-terminal degree-two vertex the head vertex of both edges incident with it, which means that the non-redundancy condition for directed networks is violated. However, it follows from Proposition A1 that if there is only one assignment of directions that satisfies the above condition, then that assignment necessarily does give a directed version of \(G\). Such uniqueness is rather special. Indeed, the following result can easily be deduced from Proposition 1 in Milchtaich (2006b).
Corollary A2
For an undirected two-terminal network \(G\), the following conditions are equivalent:
-
(i)
For every edge \(e\), all routes in \(G\) that include \(e\) traverse it in the same direction.
-
(ii)
For every assignment of allowable directions to the edges in \(G\) such that the direction of each edge \(e\) is that in which some route in \(G\) traverses \(e\), all routes in \(G\) are allowable.
-
(iii)
\(G\) is series-parallel.
Rights and permissions
About this article
Cite this article
Milchtaich, I. Internalization of social cost in congestion games. Econ Theory 71, 717–760 (2021). https://doi.org/10.1007/s00199-020-01274-0
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00199-020-01274-0