Copula SMM

Routing Summary

This folder covers the Oh & Patton (2011) research program: SMM estimation for copula-based multivariate models using rank dependence measures as moments. Contains 5 notes.

Concept Map

ConceptNoteTypeDepends OnKey Result
Spearman’s ρ, quantile dependence, tail dependence — pure copula functionalsDependence Measures for CopulasdefinitionCopula EstimationInvariant to marginals; used as SMM moments in Oh & Patton
Oh-Patton DGP, two-stage estimation, rank dependence momentsSMM Estimator for CopulasconceptMethod of Simulated Moments, Dependence Measures for CopulasSMM for copulas when likelihood is unavailable
Assumptions 1–4, Propositions 1–3 (consistency, normality, variance)SMM Copula Asymptotic TheorytheoremSMM Estimator for CopulasFirst-stage estimation error does not affect copula estimator
J-test, over-identifying restrictions, simulated critical valuesSMM Copula Specification TestingtheoremSMM Copula Asymptotic Theory with efficient weight; simulated CVs otherwise
Monte Carlo study + 7-firm financial dependence (2001–2010)SMM Copula Simulation and ApplicationexampleSMM Copula Asymptotic Theory, SMM Copula Specification Testing~20–40% efficiency loss vs. MLE; significant tail dependence in financials

Notes

  • Dependence Measures for Copulas — CONTAINS: Spearman’s rank correlation, quantile dependence, tail dependence coefficients, asymmetry measures, pure copula functionals invariant to marginals
  • SMM Estimator for Copulas — CONTAINS: Oh-Patton DGP, two-stage estimation, rank dependence moments, SMM estimator definition, factor copula model, nesting of GMM/MM
  • SMM Copula Asymptotic Theory — CONTAINS: Assumptions 1–4, Proposition 1 (consistency), Proposition 2 (asymptotic normality with 3 rate cases), Proposition 3 (variance estimation via bootstrap + numerical derivatives)
  • SMM Copula Specification Testing — CONTAINS: Proposition 4 (J-test), chi-squared with efficient weight, simulated critical values for general weight, simulation procedure
  • SMM Copula Simulation and Application — CONTAINS: Monte Carlo for Clayton/Normal/factor copulas, iid and AR-GARCH data, step-size sensitivity, 7 financial firms (2001–2010), tail dependence and asymmetry findings

Sources

See Also