Optimizing policymakers' loss functions in crisis prediction: before, within or after?

Peter Sarlina,b, Gregor von Schweinitzc,d,1,∗

a Silo.AI, Helsinki, Finland
b RiskLab Finland at Arcada and Hanken School of Economics
c Halle Institute for Economic Research (IWH), Department of Macroeconomics
d University of Leipzig, Institute for Theoretical Economics


Recurring financial instabilities have led policymakers to rely on early-warning models to signal financial vulnerabilities. These models rely on ex-post optimization of signaling thresholds on crisis probabilities accounting for preferences between forecast errors, but comes with the crucial drawback of unstable thresholds in recursive estimations. We propose two alternatives for threshold setting with similar or better out-of-sample performance: (i) including preferences in the estimation itself and (ii) setting thresholds ex-ante according to preferences only. Given probabilistic model output, it is intuitive that a decision rule is independent of the data or model specication, as thresholds on probabilities represent a willingness to issue a false alarm vis-à-vis missing a crisis. We provide real-world and simulation evidence that this simplification results in stable thresholds, while keeping or improving on out-of-sample performance. Our solution is not restricted to binary-choice models, but directly transferable to the signaling approach and all probabilistic early-warning models.

Keywords: Early-warning models; Loss functions; Threshold setting; Predictive performance
JEL-Classification: C35; C53; G01

Accepted for publication in Macroeconomic Dynamics

Download paper
Latest version (preprint)
ECB Working Paper No. 2025
IWH Discussion Paper No. 6/2015

Online appendix
Online appendix (pdf)

    ☆ Research of Gregor von Schweinitz was partly funded by the European Regional Development Fund through the programme “Investing in your Future” and by the IWH Speed Project 2014/02. Parts of this work have been completed at the Financial Stability Surveillance Division of the ECB DG Macroprudential Policy and Financial Stability. The authors are grateful for the suggestions of two anonymous referees, useful comments from Bernd Amann, Carsten Detken, Makram El-Shagi, Jan-Hannes Lang, Tuomas Peltonen and Peter Welz, and discussion at the following seminars and conferences: 3rd HenU-INFER Workshop on Applied Macroeconomics, IWH Economic Research Seminar, Goethe University Brown Bag Seminar, ECB Financial Stability Seminar, Deutsche Bundesbank Early-Warning Modeling Seminar and the 2015 CEUS Workshop. An online appendix to this paper can be found at http://appliedmacro.org/index.php/special-issue/macroeconomic-dynamics. Data and code are supplied at https://risklab.fi/publications/thresholdoptimization
   ∗ Corresponding author. Telephone: +49 345 7753 744. Email: gsz@iwh-halle.de