Statistical Backwards Induction: A Simple Method for Estimating Recursive Strategic Models

Bas, Muhammet Ali; Signorino, Curtis S.; Walker, Robert

doi:10.1093/pan/mpm029

Cited by 75 publications

(72 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Most noteworthy are the contributions by Signorino (1999Signorino ( , 2002Signorino ( , 2003aSignorino ( , 2003bSignorino ( , 2008) that successfully adapted work by McKelvey and Palfrey (1995, 1998 from experimental settings to observational data. One problem is that these estimators are prone to selection bias.…”

Section: Empirical Models Based On Theoretical Modelsmentioning

confidence: 99%

Strategy and Sample Selection: A Strategic Selection Estimator

Leemann¹

2014

Polit. anal.

View full text Add to dashboard Cite

The development and proliferation of strategic estimators has narrowed the gap between theoretical models and empirical testing. But despite recent contributions that extend the basic strategic estimator, researchers have continued to neglect a classic social science phenomenon: selection.Compared to non-strategic estimators, strategic models are even more prone to selection effects. First, external shocks or omitted variables can lead to correlated errors. Second, because the systematic parts of actors' utilities usually overlap on certain key variables, the two sets of explanatory variables are correlated. As a result, both the systematic and the stochastic components can be correlated. However, given that the estimates for the first mover are computed based on the potentially biased predicted probabilities of the second actor, we also generate biased estimates for the first actor.In applied work researchers neglect the potential shortcomings due to selection bias. This paper presents an alternative strategic estimator that takes selection into account and allows scholars to obtain consistent, unbiased, and efficient estimates in the presence of both selection and strategic action. I present a Monte Carlo analysis as well as a real world application to illustrate the superior performance of this estimator relative to the standard practice. * I would like to thank Noah Buckley, Shigeo Hirano, Jeff Lax, Pierce O'Reilly, Curt Signorino, Neelan Sircar, and Greg Wawro for helpful comments and discussions. Sara Bjerg Moller and Simon Collrad-Wexler provided insights on the IR example. I also thank Thomas Plümper for the feedback he gave me as a discussant at the 2012 EPSA meeting. I am deeply indebted to two anonymous reviewers whose comments -which were thoughtful and very extensive -have significantly improved this manuscript. An earlier version of this paper received the Giancarlo Doria Paper Prize. I am very grateful to CUIT for giving me access to Hotfoot (high performance computing resource) to run the simulations. Companion software (for R) that implements the estimator proposed in this paper will be provided shortly on the author's webpage. All remaining errors are mine.

show abstract

Section: Empirical Models Based On Theoretical Modelsmentioning

confidence: 99%

Strategy and Sample Selection: A Strategic Selection Estimator

Leemann¹

2014

Polit. anal.

View full text Add to dashboard Cite

show abstract

“…4 The bene…t of the proposed method will be fully exploited when an application focuses on sequential interaction in a middle-sized game, which is not overly large but if larger than the two-player binary-choice game. On one hand, the two-player Stackelberg game, which has been widely studied in the theoretical literature, has limited use in empirical research.…”

Section: Relation To the Literature And Applica-bilitymentioning

confidence: 99%

Estimation of Finite Sequential Games

Maruyama

2012

SSRN Journal

View full text Add to dashboard Cite

I propose a new estimation method for …nite sequential games that is e¢ cient, computationally attractive, and applicable to a fairly general class of …nite sequential games that is beyond the scope of existing studies. The major challenge is computation of high-dimensional truncated integration whose domain is complicated by strategic interaction. This complication resolves when unobserved o¤-the-equilibrium-path strategies are controlled for. Separately evaluating the likelihood contribution of each subgame-perfect equilibrium that generates the observed outcome allows the use of the GHK simulator, a widely used importance-sampling probit simulator. Monte Carlo experiments demonstrate the performance and robustness of the proposed method.JEL Codes: C35, C72, C63, C31

show abstract

“…Current statistical models for crisis bargaining entail either traditional game-theoretic assumptions (Lewis and Schultz 2003;Wand 2006;Whang 2010b;Whang, McLean and Kubserski 2013) or "independent agent" assumptions (Signorino 2003;Leblang 2003;Signorino and Tarar 2006;Gent 2007;Bas, Signorino and Walker 2008;McLean and Whang 2010;Carter 2010;Bas 2012). These approaches differ in what they assume about players' uncertainty concerning their own preferences.…”

Section: Introductionmentioning

confidence: 99%

Knowing one’s future preferences: A correlated agent model with Bayesian updating

Bas

Signorino

Whang

2013

Journal of Theoretical Politics

View full text Add to dashboard Cite

We generalize two classes of statistical sequential incomplete information games: (1) those resembling typical signaling games, in which a single agent represents each player, allowing for information to be revealed about future play; and (2) those in which each player is represented by a set of independent agents, where moves do not reveal private information. The generalized model we develop, the Correlated Agent Model, relies on a parameter, ρ, which denotes the correlation between two agents’ private information, i.e. the extent to which a player knows the future private component of her preferences. The independent agent and single agent models are special cases, where ρ=0 and ρ=1, respectively. The model also allows 0 < ρ < 1, a class of games which have not yet been considered. We apply the model to crisis bargaining and demonstrate how to estimate ρ, as well as parameters associated with utilities.

show abstract

Statistical Backwards Induction: A Simple Method for Estimating Recursive Strategic Models

Cited by 75 publications

References 18 publications

Strategy and Sample Selection: A Strategic Selection Estimator

Strategy and Sample Selection: A Strategic Selection Estimator

Estimation of Finite Sequential Games

Knowing one’s future preferences: A correlated agent model with Bayesian updating

Contact Info

Product

Resources

About