We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

econ.EM

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Economics > Econometrics

Title: Composite Likelihood for Stochastic Migration Model with Unobserved Factor

Abstract: We introduce the Maximum Composite Likelihood (MCL) estimation method for the stochastic factor ordered Probit model of credit rating transitions of firms. This model is recommended to banks and financial institutions as part of internal credit risk assessment procedures under the Basel III regulations. However, its exact likelihood function involves a high-dimensional integral, which can be approximated numerically and next this approximation can be maximized. However, the associated estimates of migration risk and corresponding required capital are generally quite sensitive to the quality of this approximation, leading to statistical regulatory arbitrage. The proposed MCL estimator maximizes the composite log-likelihood of the factor ordered Probit model. We present three MCL estimators of different complexity and prove their consistency and asymptotic normality. The performance of these estimators is examined in a simulation study.
Subjects: Econometrics (econ.EM)
Cite as: arXiv:2109.09043 [econ.EM]
  (or arXiv:2109.09043v1 [econ.EM] for this version)

Submission history

From: Antoine Djogbenou [view email]
[v1] Sun, 19 Sep 2021 02:33:20 GMT (1156kb,D)

Link back to: arXiv, form interface, contact.