Monte Carlo Confidence Sets for Identified Sets

Xiaohong Chen, Timothy Christensen, Elie Tamer

    Research output: Contribution to journalArticle

    Abstract

    It is generally difficult to know whether the parameters in nonlinear econometric models are point-identified. We provide computationally attractive procedures to construct confidence sets (CSs) for identified sets of the full parameter vector and of subvectors in models defined through a likelihood or a vector of moment equalities or inequalities. The CSs are based on level sets of “optimal” criterion functions (such as likelihoods, optimally-weighted or continuously-updated GMM criterions). The level sets are constructed using cutoffs that are computed via Monte Carlo (MC) simulations from the quasi-posterior distribution of the criterion. We establish new Bernstein–von Mises (or Bayesian Wilks) type theorems for the quasi-posterior distributions of the quasi-likelihood ratio (QLR) and profile QLR in partially-identified models. These results imply that our MC CSs have exact asymptotic frequentist coverage for identified sets of full parameters and of subvectors in partially-identified regular models, and have valid but potentially conservative coverage in models whose local tangent spaces are convex cones. Further, our MC CSs for identified sets of subvectors are shown to have exact asymptotic coverage in models with singularities. We provide local power properties and uniform validity of our CSs over classes of DGPs that include point- and partially-identified models. Finally, we present two simulation experiments and two empirical examples: an airline entry game and a model of trade flows.

    Original languageEnglish (US)
    Pages (from-to)1965-2018
    Number of pages54
    JournalEconometrica
    Volume86
    Issue number6
    DOIs
    StatePublished - Nov 1 2018

    Fingerprint

    Confidence set
    Likelihood ratio
    Quasi-likelihood
    Posterior distribution
    Econometric models
    Convex cone
    Trade flows
    Equality
    Local power
    Simulation experiment
    Singularity
    Monte Carlo simulation
    Airlines

    Keywords

    • (profile) quasi-likelihood ratio
    • confidence sets
    • convex cone
    • likelihood
    • moment (in)equality restrictions
    • Partial identification
    • quasi-posterior
    • sequential Monte Carlo
    • subvector

    ASJC Scopus subject areas

    • Economics and Econometrics

    Cite this

    Chen, X., Christensen, T., & Tamer, E. (2018). Monte Carlo Confidence Sets for Identified Sets. Econometrica, 86(6), 1965-2018. https://doi.org/10.3982/ECTA14525

    Monte Carlo Confidence Sets for Identified Sets. / Chen, Xiaohong; Christensen, Timothy; Tamer, Elie.

    In: Econometrica, Vol. 86, No. 6, 01.11.2018, p. 1965-2018.

    Research output: Contribution to journalArticle

    Chen, X, Christensen, T & Tamer, E 2018, 'Monte Carlo Confidence Sets for Identified Sets', Econometrica, vol. 86, no. 6, pp. 1965-2018. https://doi.org/10.3982/ECTA14525
    Chen X, Christensen T, Tamer E. Monte Carlo Confidence Sets for Identified Sets. Econometrica. 2018 Nov 1;86(6):1965-2018. https://doi.org/10.3982/ECTA14525
    Chen, Xiaohong ; Christensen, Timothy ; Tamer, Elie. / Monte Carlo Confidence Sets for Identified Sets. In: Econometrica. 2018 ; Vol. 86, No. 6. pp. 1965-2018.
    @article{8eac8443290844c99a05bfa2de23eeb5,
    title = "Monte Carlo Confidence Sets for Identified Sets",
    abstract = "It is generally difficult to know whether the parameters in nonlinear econometric models are point-identified. We provide computationally attractive procedures to construct confidence sets (CSs) for identified sets of the full parameter vector and of subvectors in models defined through a likelihood or a vector of moment equalities or inequalities. The CSs are based on level sets of “optimal” criterion functions (such as likelihoods, optimally-weighted or continuously-updated GMM criterions). The level sets are constructed using cutoffs that are computed via Monte Carlo (MC) simulations from the quasi-posterior distribution of the criterion. We establish new Bernstein–von Mises (or Bayesian Wilks) type theorems for the quasi-posterior distributions of the quasi-likelihood ratio (QLR) and profile QLR in partially-identified models. These results imply that our MC CSs have exact asymptotic frequentist coverage for identified sets of full parameters and of subvectors in partially-identified regular models, and have valid but potentially conservative coverage in models whose local tangent spaces are convex cones. Further, our MC CSs for identified sets of subvectors are shown to have exact asymptotic coverage in models with singularities. We provide local power properties and uniform validity of our CSs over classes of DGPs that include point- and partially-identified models. Finally, we present two simulation experiments and two empirical examples: an airline entry game and a model of trade flows.",
    keywords = "(profile) quasi-likelihood ratio, confidence sets, convex cone, likelihood, moment (in)equality restrictions, Partial identification, quasi-posterior, sequential Monte Carlo, subvector",
    author = "Xiaohong Chen and Timothy Christensen and Elie Tamer",
    year = "2018",
    month = "11",
    day = "1",
    doi = "10.3982/ECTA14525",
    language = "English (US)",
    volume = "86",
    pages = "1965--2018",
    journal = "Econometrica",
    issn = "0012-9682",
    publisher = "Wiley-Blackwell",
    number = "6",

    }

    TY - JOUR

    T1 - Monte Carlo Confidence Sets for Identified Sets

    AU - Chen, Xiaohong

    AU - Christensen, Timothy

    AU - Tamer, Elie

    PY - 2018/11/1

    Y1 - 2018/11/1

    N2 - It is generally difficult to know whether the parameters in nonlinear econometric models are point-identified. We provide computationally attractive procedures to construct confidence sets (CSs) for identified sets of the full parameter vector and of subvectors in models defined through a likelihood or a vector of moment equalities or inequalities. The CSs are based on level sets of “optimal” criterion functions (such as likelihoods, optimally-weighted or continuously-updated GMM criterions). The level sets are constructed using cutoffs that are computed via Monte Carlo (MC) simulations from the quasi-posterior distribution of the criterion. We establish new Bernstein–von Mises (or Bayesian Wilks) type theorems for the quasi-posterior distributions of the quasi-likelihood ratio (QLR) and profile QLR in partially-identified models. These results imply that our MC CSs have exact asymptotic frequentist coverage for identified sets of full parameters and of subvectors in partially-identified regular models, and have valid but potentially conservative coverage in models whose local tangent spaces are convex cones. Further, our MC CSs for identified sets of subvectors are shown to have exact asymptotic coverage in models with singularities. We provide local power properties and uniform validity of our CSs over classes of DGPs that include point- and partially-identified models. Finally, we present two simulation experiments and two empirical examples: an airline entry game and a model of trade flows.

    AB - It is generally difficult to know whether the parameters in nonlinear econometric models are point-identified. We provide computationally attractive procedures to construct confidence sets (CSs) for identified sets of the full parameter vector and of subvectors in models defined through a likelihood or a vector of moment equalities or inequalities. The CSs are based on level sets of “optimal” criterion functions (such as likelihoods, optimally-weighted or continuously-updated GMM criterions). The level sets are constructed using cutoffs that are computed via Monte Carlo (MC) simulations from the quasi-posterior distribution of the criterion. We establish new Bernstein–von Mises (or Bayesian Wilks) type theorems for the quasi-posterior distributions of the quasi-likelihood ratio (QLR) and profile QLR in partially-identified models. These results imply that our MC CSs have exact asymptotic frequentist coverage for identified sets of full parameters and of subvectors in partially-identified regular models, and have valid but potentially conservative coverage in models whose local tangent spaces are convex cones. Further, our MC CSs for identified sets of subvectors are shown to have exact asymptotic coverage in models with singularities. We provide local power properties and uniform validity of our CSs over classes of DGPs that include point- and partially-identified models. Finally, we present two simulation experiments and two empirical examples: an airline entry game and a model of trade flows.

    KW - (profile) quasi-likelihood ratio

    KW - confidence sets

    KW - convex cone

    KW - likelihood

    KW - moment (in)equality restrictions

    KW - Partial identification

    KW - quasi-posterior

    KW - sequential Monte Carlo

    KW - subvector

    UR - http://www.scopus.com/inward/record.url?scp=85058219465&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=85058219465&partnerID=8YFLogxK

    U2 - 10.3982/ECTA14525

    DO - 10.3982/ECTA14525

    M3 - Article

    VL - 86

    SP - 1965

    EP - 2018

    JO - Econometrica

    JF - Econometrica

    SN - 0012-9682

    IS - 6

    ER -