pith. sign in

arxiv: 2512.13562 · v2 · pith:FRNLWBN2new · submitted 2025-12-15 · 💱 q-fin.RM · math.PR

Disability insurance with collective health claims: A mean-field approach

Christian Furrer , Philipp C. Hornung This is my paper

Pith reviewed 2026-05-21 18:04 UTC · model grok-4.3

classification 💱 q-fin.RM math.PR
keywords disability insurancemean-field approximationsemi-Markov modelcollective health claimsintegro-differential equationspricing methodgroup experience rating
0
0 comments X

The pith

Mean-field approximation reduces collective health claims to non-linear equations for disability pricing

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

The paper expands the classic semi-Markov disability model to include both individual and collective health claims, which creates a many-body problem that is hard to solve directly for group insurance. A mean-field approach approximates this many-body system as a non-linear one-body problem. This yields a pricing method based on a lower-dimensional system of non-linear forward integro-differential equations. Simulations show the approximation performs comparably to full Monte Carlo methods. A sympathetic reader would care because the result offers a practical way to price disability coverages that reflect group-level health effects.

Core claim

By expanding the semi-Markov disability model with collective health claims, the pricing task becomes a computationally challenging many-body problem. Adopting a mean-field approach approximates this as a non-linear one-body problem, which produces a transparent pricing method based on a lower-dimensional system of non-linear forward integro-differential equations. A practice-oriented simulation study confirms that the approximation compares favorably to naive Monte Carlo methods.

What carries the argument

Mean-field approximation applied to the expanded semi-Markov model with collective health claims, which converts the many-body problem into a non-linear one-body problem solved by forward integro-differential equations.

If this is right

  • Transparent pricing becomes available for disability coverages that incorporate collective health claims.
  • The computational problem reduces to a lower-dimensional system of non-linear forward integro-differential equations.
  • The method provides a practical alternative to direct Monte Carlo simulation for large portfolios.
  • Improved experience rating is possible in group disability insurance settings.

Where Pith is reading between the lines

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

  • The same mean-field reduction could apply to other multi-state insurance models that involve dependent risks across individuals.
  • Extensions might incorporate time-varying or stochastic collective factors while preserving the lower-dimensional structure.
  • Real-world calibration on large claims datasets could test whether the approximation error remains small outside simulated settings.

Load-bearing premise

The mean-field approximation sufficiently captures the dynamics of collective health claims in the expanded semi-Markov model without introducing unacceptable error.

What would settle it

A simulation in which prices computed from the mean-field equations deviate substantially from prices obtained by full Monte Carlo simulation of the many-body system as group size grows.

Figures

Figures reproduced from arXiv: 2512.13562 by Christian Furrer, Philipp C. Hornung.

Figure 1
Figure 1. Figure 1: State space J “ t1, 2, 3u for classic disability insur￾ance. The arrows represent the possible transitions. Semi-Markov modeling entails the assumption that pZ, Uq is Markov, and smooth semi-Markov modeling adds the assumption that the jump times should be ab￾solutely continuous (with respect to the Lebesgue measure). An alternative, but equivalent, formulation is in terms of the compensators of the multiv… view at source ↗
Figure 2
Figure 2. Figure 2: in [3]. The time complexity given a cut-off KH is of the same order as in the classic semi-Markov disability model, but differs by about a factor of KH. t d Boundary conditions p¯j pη3, 0, hq “ 0 p¯j pη3, η, hq p¯j pη3, η2, hq p¯j pη3, η3, hq 0 η η2 η3 0 η η2 η3 [PITH_FULL_IMAGE:figures/full_fig_p019_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Disability rate (left) and credibility formula (right) for a single realization of the one-individual model (y “ ν 1 ) with the mean-field approximation (y “ v) and the baseline (y “ ζ1). 5 individuals Average present value Density 0 1 2 3 4 5 0.0 0.2 0.4 0.6 0.8 1.0 25 individuals Average present value Density 0 1 2 3 4 5 0.0 0.2 0.4 0.6 0.8 1.0 50 individuals Average present value Density 0 1 2 3 4 5 0.0… view at source ↗
Figure 4
Figure 4. Figure 4: Histograms of average present values for n “ 5, 25, 50, 100; as n increases, a clear bell curve emerges [PITH_FULL_IMAGE:figures/full_fig_p023_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: For n “ 5, 25, 100 disability rates (left) and credibility formulas (right) for a single realization of the n-individual model (y “ ν n) with the mean-field approximation (y “ v) and the base￾line (y “ ζ1). To assess the quality of the convergence, we may additionally study histograms of the average present values 1 n ÿn ℓ“1 ż T 0 e ´ ş t 0 rpsq dsB ℓ,n,mpdtq for m “ 1, . . . , M, where M “ 40, 000. The re… view at source ↗
read the original abstract

The classic semi-Markov disability model is expanded with individual and collective health claims to improve its explanatory and predictive power -- in particular in the context of group experience rating. The inclusion of collective health claims leads to a computationally challenging many-body problem. By adopting a mean-field approach, this many-body problem can be approximated by a non-linear one-body problem, which in turn leads to a transparent pricing method for disability coverages based on a lower-dimensional system of non-linear forward integro-differential equations. In a practice-oriented simulation study, the mean-field approximation clearly stands its ground in comparison to na\"ive Monte Carlo methods.

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

2 major / 2 minor

Summary. The paper extends the classic semi-Markov disability model by adding individual and collective health claims to improve explanatory power for group experience rating. The resulting many-body problem is approximated by a mean-field limit that reduces it to a non-linear one-body semi-Markov process, yielding a pricing method based on a lower-dimensional system of non-linear forward integro-differential equations. The approximation is assessed via a practice-oriented simulation study that compares it to naïve Monte Carlo methods.

Significance. If the mean-field approximation is shown to be accurate with controlled error for finite portfolios, the work would supply a computationally tractable and transparent pricing framework for disability coverages that incorporate collective health effects, addressing a practical gap in group insurance modeling.

major comments (2)
  1. [§5] §5 (Simulation study): The abstract states that the mean-field approximation 'clearly stands its ground' against naïve Monte Carlo methods, yet no details are supplied on the number of simulation runs, the portfolio sizes N examined, the error metrics employed, the specific parameter values, or the range of collective-dependence strengths tested. This information is load-bearing for the central claim that the approximation sufficiently captures the dynamics without unacceptable error.
  2. [§3] §3 (Mean-field derivation): The replacement of collective health claims by their empirical average is introduced without a convergence theorem, explicit error bound, or scaling analysis in N. For the finite N typical of group disability contracts, the O(1/√N) fluctuations in the empirical measure can affect transition intensities and premiums, but this effect is not quantified or bounded, leaving the accuracy of the resulting forward integro-differential equations unestablished beyond the specific cases simulated.
minor comments (2)
  1. [§2] The notation for the collective intensity process could be made more explicit by adding a numbered equation that distinguishes it from the individual intensity.
  2. [Figures] Figure captions should state the exact parameter values and portfolio size N used in each plotted comparison.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments on our manuscript. We respond to each major comment below and indicate the changes planned for the revised version.

read point-by-point responses

  1. Referee: [§5] §5 (Simulation study): The abstract states that the mean-field approximation 'clearly stands its ground' against naïve Monte Carlo methods, yet no details are supplied on the number of simulation runs, the portfolio sizes N examined, the error metrics employed, the specific parameter values, or the range of collective-dependence strengths tested. This information is load-bearing for the central claim that the approximation sufficiently captures the dynamics without unacceptable error.

    Authors: We agree that §5 requires additional detail to substantiate the claims. In the revision we will report the number of Monte Carlo replications, the portfolio sizes N examined, the error metrics (including relative premium error and integrated squared difference on intensities), the full parameter set, and the range of collective-dependence strengths considered. These additions will make the simulation evidence transparent and reproducible. revision: yes

  2. Referee: [§3] §3 (Mean-field derivation): The replacement of collective health claims by their empirical average is introduced without a convergence theorem, explicit error bound, or scaling analysis in N. For the finite N typical of group disability contracts, the O(1/√N) fluctuations in the empirical measure can affect transition intensities and premiums, but this effect is not quantified or bounded, leaving the accuracy of the resulting forward integro-differential equations unestablished beyond the specific cases simulated.

    Authors: The derivation replaces the collective process by its empirical average following the standard mean-field closure for interacting particle systems. We acknowledge that no rigorous convergence theorem or explicit error bound is supplied. In the revision we will add a paragraph discussing the expected O(1/√N) scaling of fluctuations, their likely impact on finite-N premiums, and the practical limitations for very small portfolios, while noting that a full theoretical analysis lies beyond the present scope. revision: partial

Circularity Check

0 steps flagged

No circularity: mean-field limit is an explicit approximation validated externally

full rationale

The derivation begins from an expanded semi-Markov model that includes collective health claims, producing a many-body problem whose state space grows with portfolio size N. The paper then invokes the standard mean-field replacement of the empirical measure by its expectation, yielding a closed non-linear one-body process whose forward integro-differential equations are solved numerically. This step is presented as an approximation whose accuracy is checked by direct Monte Carlo simulation on finite-N portfolios; the simulation constitutes an independent benchmark rather than a tautological re-derivation. No equation is defined in terms of its own output, no fitted parameter is relabeled as a prediction, and no load-bearing uniqueness claim rests on a self-citation. The central pricing method therefore remains self-contained against external numerical verification.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Review based on abstract only; specific free parameters, axioms, and entities not detailed in available text.

axioms (1)
  • domain assumption The classic semi-Markov disability model can be meaningfully expanded with individual and collective health claims.
    Core premise stated in the abstract for improving explanatory power.

pith-pipeline@v0.9.0 · 5626 in / 1132 out tokens · 50181 ms · 2026-05-21T18:04:52.212597+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

16 extracted references · 16 canonical work pages · 1 internal anchor

  1. [1]

    Aalen,Dynamic modelling and causality, Scandinavian Actuarial Journal1987(1987), no

    O.O. Aalen,Dynamic modelling and causality, Scandinavian Actuarial Journal1987(1987), no. 3-4, 177–190

    work page 1987

  2. [2]

    Andersen, Ø

    P.K. Andersen, Ø. Borgan, R.D. Gill, and N. Keiding,Statistical Models Based on Counting Processes, Springer Series in Statistics, Springer, New York, 1993. 26 CHRISTIAN FURRER AND PHILIPP C. HORNUNG

    work page 1993

  3. [3]

    Buchardt, T

    K. Buchardt, T. Møller, and K.B. Schmidt,Cash flows and policyholder behaviour in the semi-Markov life insurance setup, Scandinavian Actuarial Journal2015(2015), no. 8, 660– 688

    work page 2015

  4. [4]

    Christiansen,Multistate models in health insurance, Advances in Statistical Analysis96 (2012), 155–186

    M.C. Christiansen,Multistate models in health insurance, Advances in Statistical Analysis96 (2012), 155–186

    work page 2012

  5. [5]

    Djehiche and B

    B. Djehiche and B. L¨ ofdahl,Nonlinear reserving in life insurance: Aggregation and mean-field approximation, Insurance: Mathematics and Economics69(2016), 1–13

    work page 2016

  6. [6]

    Feinberg, M

    E. Feinberg, M. Mandava, and A.N. Shiryaev,Kolmogorov’s equations for jump Markov pro- cesses with unbounded jump rates, Annals of Operations Research317(2022), 587–604

    work page 2022

  7. [7]

    Feinberg, M

    E.A. Feinberg, M. Mandava, and A.N. Shiryaev,On solutions of Kolmogorov’s equations for nonhomogeneous jump Markov processes, Journal of Mathematical Analysis and Applications 411(2014), 261–270

    work page 2014

  8. [8]

    Furrer,Experience rating in the classic Markov chain life insurance setting: An empirical Bayes and multivariate frailty approach, European Actuarial Journal9(2019), 31–58

    C. Furrer,Experience rating in the classic Markov chain life insurance setting: An empirical Bayes and multivariate frailty approach, European Actuarial Journal9(2019), 31–58

    work page 2019

  9. [9]

    Furrer, J.J

    C. Furrer, J.J. Sørensen, and J. Yslas,Bivariate phase-type distributions for experience rating in disability insurance, 2025, arXiv:2405.19248; to appear in European Actuarial Journal

    work page arXiv 2025

  10. [10]

    Helwich,Durational effects and non-smooth semi-Markov models in life insurance, Ph.D

    M. Helwich,Durational effects and non-smooth semi-Markov models in life insurance, Ph.D. thesis, University of Rostock, 2008

    work page 2008

  11. [11]

    Hoem,Inhomogeneous Semi-Markov Processes, Select Actuarial Tables, and Duration- Dependence in Demography, Population Dynamics (T.N.E

    J.M. Hoem,Inhomogeneous Semi-Markov Processes, Select Actuarial Tables, and Duration- Dependence in Demography, Population Dynamics (T.N.E. Greville, ed.), Academic Press, 1972, pp. 251–296

    work page 1972

  12. [12]

    Mean-field approximations in insurance

    P.C. Hornung,Mean-field approximations in insurance, 2025, arXiv:2511.04198

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  13. [13]

    Jacobsen,Point process theory and applications: Marked point and piecewise deterministic processes, Probability and its Applications, Birkh¨ auser, Boston, 2006

    M. Jacobsen,Point process theory and applications: Marked point and piecewise deterministic processes, Probability and its Applications, Birkh¨ auser, Boston, 2006

    work page 2006

  14. [14]

    Kallenberg,Probabilistic Symmetries and Invariance Principles, Probability and Its Ap- plications, Springer New York, 2005

    O. Kallenberg,Probabilistic Symmetries and Invariance Principles, Probability and Its Ap- plications, Springer New York, 2005

    work page 2005

  15. [15]

    Lewis and G.S

    P.A.W. Lewis and G.S. Shedler,Simulation of nonhomogeneous Poisson processes by thinning, Naval Research Logistics Quarterly26(1979), no. 3, 403–413

    work page 1979

  16. [16]

    Sznitman,Topics in propagation of chaos, Ecole d’Et´ e de Probabilit´ es de Saint-Flour XIX – 1989

    A.-S. Sznitman,Topics in propagation of chaos, Ecole d’Et´ e de Probabilit´ es de Saint-Flour XIX – 1989. Lecture Notes in Mathematics, vol 1464 (P.-L. Hennequin, ed.), Springer Berlin Heidelberg, 1991, pp. 165–251. Department of Mathematical Sciences, University of Copenhagen, Universitetsparken 5, DK-2100 Copenhagen, Denmark Email address:furrer@math.ku...

    work page 1989