Abstract
The classical game theoretic resolutions to Selten’s Chain Store game are unsatisfactory; they either alter the game to avoid the paradox or struggle to organize the existing experimental data. This paper applies co-evolutionary algorithms to the Chain Store game and demonstrates that the resulting system’s dynamics are neither intuitively paradoxical nor contradicted by the existing experimental data. Specifically, some parameterizations of evolutionary algorithms promote genetic drift. Such drift can lead the system to transition among the game’s various Nash Equilibria. This has implications for policy makers as well as for computational modelers.
Similar content being viewed by others
Notes
Selten (1978)
For a precise game theoretic definition of ‘Common Knowledge,’ see Fudenberg and Tirole (1996), p. 542.
Other structures were tested, such as limiting the Entrants’ memory. However, these structures did not produce significantly different results.
The specification was tested for this paper and produced no describable results. Chen and Ni (1996) tried this approach with an eight-stage Chain Store game for which history was truncated to only include the opponent’s behavior. However, they found that this abbreviated version was also intractable.
These observations are made at the phenotypical level. When an Entrant stays out, the Chain Store does not have the opportunity to respond. Therefore, NE outcomes include all games in which all stages end in either enter-acquiesce or stay out.
For example, if a particular generation resulted in more than 70% of all stages ending in enter-acquiesce, the system would be in the SGP state. If more than 70% of stages 1–4 and stages 6–9 ended in enter-acquiesce and stage 5 ended in stay out more than 70% of the time, the system would be in the SGP-1 state. If more than 70% of stages 1–4 and stages 6–9 ended in enter-acquiesce and stage 5 ended in stay out less than 70% of the time, the system would not be in one of the ten states.
When the system transitions from one state’s neighborhood to another, it can, with very low probability, pass through the neighborhood of a third state. The system only approximates a Markov process, because knowledge that the system has been in a state for several generations rules out the (almost trivially small) possibility that the system is passing through. This technically violates the Markov property.
To balance mutation rates across agents with different lengths of chromosome material, mutation was implemented on a per-bit basis. The per-bit probability of mutation was equal for the CS and Entrant agents. However, this led to a difference in the probability that a parent agent was mutated at least once. In the BC version, the per-bit probability of mutation was 2%. The probability of at least one mutation was 16.6 and 30.5, respectively, for the CS and Entrant agents. In the FSM, the probability of mutation was 5% per 16-bit string. The probability of at least one mutation was 14.3 and 18.5, respectively, for the CS and Entrant agents.
Kreps and Wilson (1982b, p. 254).
This variation decreases considerably when the number of generations is increased. For example, the 0.358 SD reported for the BC model in the first column of Table 8 drops to 0.101 when the number of generations is increased to 10,000. This statistic further drops to 0.068 when the number of generations is increased 20,000.
Repeated attempts to increase the value of this statistic through parameter calibration proved ineffective in raising it above 8%.
References
Alkemade, F., La Poutré, J. A., & Amman, H. M. (2006). Robust evolutionary algorithm design for socioeconomic simulation. Computational Economics, 28(4), 355–370. doi:10.1007/s10614-006-9051-5.
Andreoni, J., & Miller, J. (1995). Auctions with artificial adaptive agents. Games and Economic Behavior, 10, 39–64.
Arifovic, J. (2001). Evolutionary dynamics of the transitions across the nash equilibria of a tacit coordination game. European Journal of Economic and Social Systems, 15(3), 109–129.
Axelrod, R. (1984). The evolution of cooperation. New York: Basic Books.
Axelrod, R. (1987). Evolution of strategies in the iterated prisoner’s dilemma. In L. Davis (Ed.), Genetic algorithms and simulated annealing, research notes in artificial intelligence (Vol. 24, pp. 32–41). London: Pitman.
Binmore, K., & Samuelson, L. (1999). Evolutionary drift and equilibrium selection. Review of Economic Studies, 66, 363–393.
Bullard, J., & Duffy, J. (1999). Using genetic algorithms to model the evolution of heterogeneous beliefs. Computation Economics, 13(1), 41–60. doi:10.1023/A:1008610307810.
Casari, M. (2004). Can genetic algorithms explain experimental anomalies? An application to common property resources. Computation Economics, 24(3), 257–275.
Camerer, C., & Weigelt, K. (1988). Experimental tests of a sequential equilibrium reputation model. Econometrica, 56(1), 1–36.
Canning, D. (1992). Average behavior in learning-models. Journal of Economic Theory, 57(2), 442–472.
Chen, S., & Ni, C. (1996). Genetic algorithm learning and the chain-store game. International Conference on Evolutionary Computation, 1996, 480–484.
Dawid, H. (1999). Adaptive learning by genetic algorithms: Analytical results and applications to economic models. Berlin: Springer Press.
Eaton, C., & Morrison, W. (2003). Evolutionary stability in the investment-opportunism-retaliation game. Journal of Bio-Economics, 5(1), 27–45.
Fudenberg, D., & Tirole, J. (1996). Game theory. Cambridge, MA: The MIT Press.
Harrald, P., & Morrison, W. (2001). Sleepers, thresholds, and gateways: Drift and stability in dynamic evolutionary games (working paper). Waterloo: Wilfrid Laurier University.
Haruvy, E., Roth, A., & Ünver, M. (2006). The dynamics of law clerk matching: An experimental and computational investigation of proposals for reform of the market. Journal of Economic Dynamics and Control, 30, 457–486.
Holland, J. (1975). Adaptation in natural and artificial systems. Ann Arbor, MI: The University of Michigan Press.
Holland, J., & Miller, J. (1991). Artificial adaptive agents in economic-theory. American Economic Review, 81(2), 365–370.
Jung, Y., Kagel, J., & Levin, D. (1994). On the existence of predatory pricing: An experimental-study of reputation and entry deterrence in the chain-store game. Rand Journal of Economics, 25(1), 72–93.
Kandori, M., Mailath, G., & Rob, R. (1993). Learning, mutation, and long-run equilibria in games. Econometrica, 61(1), 29–56.
Kreps, D., & Wilson, R. (1982a). Sequential equilibria. Econometrica, 50(4), 863–894.
Kreps, D., & Wilson, R. (1982b). Reputation and imperfect information. Journal of Economic Theory, 27(2), 253–279.
McCarthy, I. (2009). Simulating sequential search models with genetic algorithms: Analysis of price ceilings, taxes, advertising and welfare. Computational Economics, 34(3), 217–241. doi:10.1007/s10614-008-9164-0.
Miller, J. (1988). The evolution of automata in the repeated prisoner’s dilemma. In Two essays on the economics of imperfect information. Ph.D: Dissertation, University of Michigan, Michigan.
Miller, J. (1998). Active nonlinear tests (ANTs) of complex simulation models. Management Science, 44(6), 820–830.
Milgram, P., & Roberts, J. (1982). Predation, reputation, and entry deterrence. Journal of Economic Theory, 27(2), 280–312.
Selten, R. (1978). The chain store paradox. Theory and Decision, 9, 127–159.
Smith, J. (1972). Game theory and the evolution of fighting. In On evolution. Edinburgh: Edinburgh University Press.
Traulsen, A., Semmann, S., Sommerfeld, R., Krambeck, H., & Milinski, M. (2010). Human strategy updating in evolutionary games. In Proceedings of the National Academy of Sciences, PNAS Early Edition. www.pnas.org/cgi/doi/10.1073/pnas.0912515107.
Ünver, M. (2001). Backward unraveling over time: The evolution of strategic behavior in the entry-level British medical labor markets. Journal of Economic Dynamics and Control, 25, 1035–1080.
Young, P. (1993). The evolution of conventions. Econometrica, 61, 57–84.
Young, P. (1996). The economics of convention. Journal of Economic Perspectives, 10(2), 105–22.
Acknowledgments
This work benefited substantially from conversations with John Miller, Bill McKelvey, Scott Carr, and John Holland. Comments from Computational Economics Editor Hans Amman and anonymous reviewers also significantly improved this paper.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Tracy, W.M. Paradox Lost: The Evolution of Strategies in Selten’s Chain Store Game. Comput Econ 43, 83–103 (2014). https://doi.org/10.1007/s10614-013-9364-0
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10614-013-9364-0