pith. sign in

arxiv: 2606.07276 · v1 · pith:CTO4BKKPnew · submitted 2026-06-05 · 🧮 math.ST · q-fin.RM· stat.TH

The Balance Property: The Constrained Case, with a View on Risk Sharing

Mario V. W\"uthrich This is my paper

Pith reviewed 2026-06-27 20:32 UTC · model grok-4.3

classification 🧮 math.ST q-fin.RMstat.TH
keywords balance propertyGLM fittinginsurance pricingrisk sharingconstrained optimizationactuarial models
0
0 comments X

The pith

Constrained GLM fitting enforces the balance property more effectively than prior correction methods while linking it to ex-post risk sharing.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper focuses on the balance property for insurance pricing models, which requires that the sum of fitted actuarial prices equals the total observed losses used in fitting. Maximum likelihood GLMs with canonical links satisfy this automatically, but when it fails, the note reviews two existing correction methods and proposes constrained GLM fitting as a better approach. It also draws a connection between enforcing the balance property and ex-post risk sharing rules among insurance parties.

Core claim

Constrained GLM fitting turns out to be superior to the two previously discussed balance correction methods from Lindholm-Wüthrich, and the balance property connects directly to ex-post risk sharing rules.

What carries the argument

Constrained GLM fitting, which modifies the standard GLM estimation to enforce the balance property as a direct constraint during fitting.

If this is right

  • Fitted models will exactly match the aggregate observed loss by construction.
  • The method preserves the GLM structure while adding the balance constraint.
  • Pricing outputs can be interpreted as outcomes of an ex-post risk sharing arrangement.
  • The approach applies whenever the balance property is required but standard maximum likelihood does not deliver it.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The superiority claim would benefit from explicit derivation or numerical evidence comparing the three methods on the same data sets.
  • Constrained fitting might extend to other model classes beyond GLMs if the balance constraint can be expressed as a linear restriction.
  • Viewing the balance property through risk sharing could suggest new ways to allocate premiums across heterogeneous risks.

Load-bearing premise

That constrained GLM fitting can be shown to be superior to the two prior correction methods through comparison.

What would settle it

A direct numerical or analytical comparison between constrained GLM fitting and the two earlier methods that shows no superiority on relevant metrics such as bias, variance, or out-of-sample performance.

read the original abstract

The balance property is an important property of fitted statistical models deployed for insurance pricing. It guarantees that the total actuarial price in the fitted model is equal to the totally observed loss used to fit the model. This can be seen as an in-sample global unbiasedness property. Maximum likelihood fitted generalized linear models (GLMs) with canonical links automatically fulfill the balance property. Lindholm-W\"uthrich (Scandinavian Actuarial Journal, 2026) discussed two popular balance correction methods in case the balance property fails to hold. This note extends this discussion with a third method, constrained GLM fitting, that turns out to be superior over the two previously discussed ones. Moreover, we highlight the connection between the balance property and ex-post risk sharing rules.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The manuscript extends prior work by Lindholm-Wüthrich (Scandinavian Actuarial Journal, 2026) on balance corrections for GLMs in insurance pricing. It introduces constrained GLM fitting as a third method and asserts that this method is superior to the two previously discussed corrections. The note also draws a connection between the balance property and ex-post risk sharing rules.

Significance. A rigorously supported demonstration that constrained GLM fitting improves upon existing balance corrections would provide a useful addition to the actuarial statistics literature on in-sample unbiasedness, particularly if accompanied by explicit criteria for superiority and links to risk-sharing applications. The current manuscript does not supply the required comparisons or derivations, so the significance cannot yet be assessed.

major comments (1)
  1. Abstract: the central claim that constrained GLM fitting 'turns out to be superior over the two previously discussed ones' is stated without any accompanying definition of the superiority metric, analytic derivation, numerical benchmark, or pointer to a later section containing such evidence. This assertion is load-bearing for the paper's contribution as an extension.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their report and the opportunity to clarify our contribution. We address the single major comment below and commit to revisions that strengthen the presentation of the constrained GLM method without altering the core technical content.

read point-by-point responses

  1. Referee: Abstract: the central claim that constrained GLM fitting 'turns out to be superior over the two previously discussed ones' is stated without any accompanying definition of the superiority metric, analytic derivation, numerical benchmark, or pointer to a later section containing such evidence. This assertion is load-bearing for the paper's contribution as an extension.

    Authors: We agree that the abstract asserts superiority without an explicit metric or supporting derivation visible in the current draft. The intended distinction is that constrained fitting incorporates the balance constraint directly into the optimization, preserving the GLM likelihood structure and avoiding the post-hoc adjustments of the two earlier methods, which can introduce inconsistencies with the original score equations. However, we accept that this argument requires formalization. We will add a new subsection that (i) defines superiority via two explicit criteria (preservation of the canonical-link score equations under the linear constraint and minimal Euclidean deviation from the unconstrained MLE), (ii) supplies the analytic derivation showing that the constrained estimator satisfies both, and (iii) includes a small numerical illustration on synthetic insurance data. The abstract will be revised to reference this subsection. These changes constitute a major revision to the exposition. revision: yes

Circularity Check

0 steps flagged

No significant circularity; extension claim independent of self-referential reduction

full rationale

The paper extends Lindholm-Wüthrich (2026) by introducing constrained GLM fitting as a third balance correction method and asserts its superiority while noting a link to ex-post risk sharing. No quoted step shows the superiority reducing by construction to a fitted parameter, self-defined quantity, or load-bearing self-citation chain; the 2026 reference describes the two prior methods but does not define or force the new method's claimed advantage. The derivation chain for the balance property itself and the risk-sharing connection remains self-contained against external benchmarks and does not collapse into the paper's own inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract, the paper relies on standard domain assumptions from GLM theory and the prior Lindholm-Wüthrich work; no free parameters, new entities, or ad-hoc axioms are visible.

axioms (1)
  • domain assumption Maximum likelihood fitted generalized linear models with canonical links automatically fulfill the balance property.
    Stated directly in the abstract as background fact.

pith-pipeline@v0.9.1-grok · 5656 in / 1186 out tokens · 29006 ms · 2026-06-27T20:32:08.936585+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

19 extracted references · 1 canonical work pages

  1. [1]

    Blier-Wong, C. (2026). A Laplace-based perspective on conditional mean risk sharing. arXiv:2603.01434

    arXiv 2026

  2. [2]

    Brauer, A., Menzel, P., W¨ uthrich, M.V. (2025). Model monitoring: A general framework with an application to non-life insurance pricing.arXiv:2510.04556

    arXiv 2025

  3. [3]

    (2005).A Course in Credibility Theory and its Applications

    B¨ uhlmann, H., Gisler, A. (2005).A Course in Credibility Theory and its Applications. Springer

    2005

  4. [4]

    B¨ uhlmann, H., Straub, E. (1970). Glaubw¨ urdigkeit f¨ ur Schadens¨ atze.Bulletin of the Swiss Associ- ation of Actuaries1970, 111-131

    1970

  5. [5]

    Burguete, J., Gallant, R., Souza, G. (1982). On unification of the asymptotic theory of nonlinear econometric models.Economic Review1/2, 151-190

    1982

  6. [6]

    Chong, W.F., Feng, R., Ng, K.T.H. (2026). Capital-allocation-induced risk sharing. arXiv:2603.26491

    arXiv 2026

  7. [7]

    Delong, L, W¨ uthrich, M.V. (2025). Isotonic regression for variance estimation and its role in mean estimation and model validation.North American Actuarial Journal29/3, 563-591

    2025

  8. [8]

    Denuit, M., Charpentier, A., Trufin, J. (2021). Autocalibration and Tweedie-dominance for insur- ance pricing in machine learning.Insurance: Mathematics & Economics101/B, 485-497

    2021

  9. [9]

    Denuit, M., Dhaene, J. (2012). Convex order and comonotonic conditional risk sharing.Insurance: Mathematics & Economics51, 265-270

    2012

  10. [10]

    Denuit, M., Michaelides, M., Trufin, J. (2026). Balance and fairness through multicalibration in nonlife insurance pricing.arXiv:2603.16317

    arXiv 2026

  11. [11]

    Denuit, M., Robert, C.Y. (2023). From risk reduction to risk elimination by conditional mean risk sharing of independent losses.Insurance: Mathematics & Economics108, 46-59

    2023

  12. [12]

    Gallant, A.R., Holly, A. (1980). Statistical inference in an implicit, nonlinear, simultaneous equation model in the context of maximum likelihood estimation.Econometrica48/3, 697-720

    1980

  13. [13]

    Gourieroux, C., Montfort, A., Trognon, A. (1984). Pseudo maximum likelihood methods: theory. Econometrica52/3, 681-700

    1984

  14. [14]

    Jiao, Z., Kou, S., Liu, Y., Wang, R. (2022). An axiomatic theory of anynomized risk sharing. arXiv:2208.07533

    arXiv 2022

  15. [15]

    Lindholm, M., W¨ uthrich, M.V. (2026). The balance property in insurance pricing.Scandinavian Actuarial Journal2026/5, 506-543

    2026

  16. [16]

    Nelder, J.A., Wedderburn, R.W.M. (1972). Generalized linear models.Journal of the Royal Statis- tical Society, Series A135/3, 370-384. 14

    1972

  17. [17]

    White, H. (1982). Maximum likelihood estimation in misspecified models.Econometrica50/1, 1-25

    1982

  18. [18]

    W¨ uthrich, M.V. (2020). Bias regularization in neural network models for general insurance pricing. European Actuarial Journal10/1, 179-202

    2020

  19. [19]

    (2023).Statistical Foundations of Actuarial Learning and its Applica- tions.Springer Actuarial.https://link.springer.com/book/10.1007/978-3-031-12409-9 15

    W¨ uthrich, M.V., Merz, M. (2023).Statistical Foundations of Actuarial Learning and its Applica- tions.Springer Actuarial.https://link.springer.com/book/10.1007/978-3-031-12409-9 15

    work page doi:10.1007/978-3-031-12409-9 2023