Benchmarking default prediction models: pitfalls and remedies in model validation
The DOI matches its Crossref record; 1 of the 1 references it cites were found in a registry or an index.
Abstract
We discuss the components of validating credit default models with a focus on potential challenges to making inferences from validation under realworld conditions. We structure the discussion in terms of: (a) the quantities of interest that may be measured (calibration and power) and how they can result in misleading conclusions if not taken in context; (b) a methodology for measuring these quantities that is robust to non-stationarity both in terms of historical time periods and in terms of sample firm composition; and (c) techniques that aid in the interpretation of the results of such tests. The approaches we advocate provide means for controlling for and understanding sample selection and variability. These effects can in some cases be severe and we present some empirical examples that highlight instances where they are and can thus compromise conclusions drawn from validation tests.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Robust classification for imprecise environments
Provost, F., & Fawcett, T. (2001) · Machine Learning · doi:10.1023/a:1007601015854 · Cited by 1,250 · cited 2×
“8The approach proposed by Provost and Fawcett (2001) requires one to choose probabilistically between two models for a particular loan application, where the probabilities are proportional to the location of an optimal cutoff on the tangent to the two ROC curves of the models.”
BackgroundMethod
Cited by
On the Rise of FinTechs: Credit Scoring Using Digital Footprints
Berg, T., Burg, V., Gombović, A., & Puri, M. (2020) · The Review of Financial Studies · doi:10.1093/rfs/hhz099 · Cited by 919
Comparing the performance of market-based and accounting-based bankruptcy prediction models
Agarwal, V., & Taffler, R. (2008) · Journal of Banking & Finance · doi:10.1016/j.jbankfin.2007.07.014 · Cited by 584
DEMERS, E., & JOOS, P. (2007) · Journal of Accounting Research · doi:10.1111/j.1475-679x.2007.00236.x · Cited by 196
Stein, R. M. (2005) · Journal of Banking & Finance · doi:10.1016/j.jbankfin.2004.04.008 · Cited by 169
“In general and if calibrated correctly, more powerful models produce more accurate probability estimates as well (cf. Stein, 2002). ance line with slope S forms a tangent to the ROC curve for a particular model.”
BackgroundSelf-citationDynamics of firm financial evolution and bankruptcy prediction
du Jardin, P. (2017) · Expert Systems with Applications · doi:10.1016/j.eswa.2017.01.016 · Cited by 103
Fissler, T., Lorentzen, C., & Mayer, M. (2022) · arXiv · doi:10.48550/arxiv.2202.12780 · Cited by 7
“Therefore, choosing a split time such that all training samples are older than the test set is a good strategy, called out-of-time [68] or forward-validation scheme [66].”
Background
Metadata
- Authors
- Stein, R.
- Venue
- The Journal of Risk Model Validation
- Publisher
- Infopro Digital Services Limited
- Published
- 2007-05
- Volume
- 1
- Issue
- 1
- Pages
- 77-113
- ISSN
- 1753-9579 (ISSN-L)
- OpenAlex
- W111251943
- Cited by
- 154
- References
- 1
- FWCI
- 34.03
- Type
- journal article
- Language
- en
- Open access
- closed
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.