top of page

Publications

Published and Forthcoming Articles

Localizing Strictly Proper Scoring Rules

De Punder, R., Diks, C., Laeven, R., and van Dijk, D. (2026).

Journal of the American Statistical Association, 121 (554), 1447-1459

Accepted: October 10, 2025

​

Coverage: Tinbergen, UvA, and NPO Radio 1

We propose censoring as a general mechanism for localizing strictly proper scoring rules. This allows probabilistic forecasts to be evaluated on regions of particular interest while preserving the divergence structure that underlies proper forecast comparison. We illustrate the framework through applications to financial risk management, inflation forecasting, and climate data.

A Sandwich with Water: Bayesian and Frequentist Inference Under Model Misspecification

Vrugt, J., Diks, C., de Punder, R., and Grünwald, P. (2025).
ARC Geophysical Research, 1, Article 16
Published: December 29, 2025

We study Bayesian and frequentist inference under model misspecification, with applications to hydrologic model calibration. The paper shows how sandwich covariance estimation provides robust uncertainty quantification when the likelihood is incorrectly specified. Examples on soil water infiltration, watershed fluxes, and rainfall-discharge modeling show how misspecification affects parameter uncertainty and predictive inference.

 

Working Papers

Tinbergen Discussion Papers

Expected Kullback-Leibler-Based Characterizations of Score-Driven Updates

De Punder, R., Dimitriadis, T., and Lange, R. (2026).
Tinbergen Discussion Paper, 24-051/III

We provide an information-theoretic characterization of score-driven updates based on expected Kullback-Leibler divergence. The paper shows that EKL reductions occur if and only if the expected update direction aligns with the expected score, thereby uniquely identifying score-driven updates, including scaled and clipped variants, as EKL reducing. The results further provide learning-rate bounds and establish EKL as the natural information-theoretic foundation for score-driven models.

Proper and Robust Autoregressive Derivative Adaptive Models

De Punder, R. (2026).
Tinbergen Discussion Paper, 26-022/III

We introduce and characterize PRADA models, a class of observation-driven updates generated by strictly proper scoring rules and strictly consistent scoring functions. By replacing the logarithmic score with strictly locally proper scoring rules, the framework extends score-driven models to a broader local divergence space. The resulting updates are uniquely linked to reductions in the corresponding divergence and allow robust filtering for means, quantiles, expectiles, and jointly elicitable risk measures such as VaR and ES.

Barron-Loss Adaptive Estimation

De Punder, R., Dijkstra, M., and Diks, C. (2026).

Tinbergen Discussion Paper, 26-023/III

We develop BLADE, a Barron-loss-based observation-driven filter for robust time-varying parameter estimation. The paper establishes stationarity of the filter, together with consistency and asymptotic normality for the estimator of the static parameters. As a PRADA update, BLADE reduces the associated expected local divergence, and contractive dominance results show when robustification moves the filtered distribution closer to the truth.

​

Ongoing Projects

Unpublished Working Papers

  • Adaptive Score-Driven Filters. With Dimitriadis, T., and Lange, R.

  • Composite Scores. With Dimitriadis, T., Palumbo, D., and Harvey, A.

  • Focused M-Estimation. With Diks, C., Laeven, R., and van Dijk, D.

  • Flood Tail Risk and Probability Weighting. With Laeven, R., van den Bremer, T., and van der Ploeg, R.

  • A Functional Perspective on Localization. With Resin, J.

  • The Optimal Scaling Matrix. With Dijkstra, M., and Diks, C.

  • Relative Scoring Rules. With Diks, C., Laeven, R., and van Dijk, D.​

​

Conferences

Ramon de Punder and Cees Diks at the Bernoulli-IMS 11th World Congress in Probability and Statistics in Bochum

2026

  • Proper and Robust Autoregressive Derivative Adaptive Models

    • The ​77th European Meeting of the Econometric Society. Dublin, August 17.

    • International Association for Applied Econometrics Annual Conference. Carcavelos, June 25.

    • ​13th Nordic Econometric Meeting. Helsinki, June 1.

  • ​​Barron-Loss Adaptive Estimation

    • The ​77th European Meeting of the Econometric Society (by M. Dijkstra). Dublin, August 18.

    • The International Association for Applied Econometrics Annual Conference (by C. Diks). Carcavelos, June 23.

    • 8th Quantitative Finance and Financial Econometrics International Conference. Marseille, June 5.

    • 13th Nordic Econometric Meeting (by C. Diks). Helsinki, June 1.

    • The Score-Driven and Nonlinear Time Series Models Conference (invited). Venice, May 20.

  • ​Expected Kullback-Leibler-Based Characterizations of Score-Driven Updates

    • The Score-Driven and Nonlinear Time Series Models Conference (by T. Dimitriadis). Venice, May 20.

  • ​​Flood Tail Risk and Probability Weighting: Evidence from The Netherlands

    • ​Second Amsterdam Sustainable Finance Conference. Amsterdam, May 8.

 

2025

 

2024

 

2023

  • ​​Localizing Strictly Proper Scoring Rules

    • International Association for Applied Econometrics Annual Conference. Oslo, June 27.

    • The 12th ECB Conference on Forecasting Techniques. Frankfurt, June 13.

    • The 5th Quantitative Finance and Financial Econometrics International Conference. Marseille, June 9.

    • The 10th International Workshop on Applied Probability. Thessaloniki, June 7.

    • The 16th Meeting of the Netherlands Econometric Study Group. Rotterdam, June 2.

 

2022

bottom of page