Runge--Kutta numerical methods for ruin probabilities in classical risk model

George Kanakoudis; Lazaros Kanellopoulos

arxiv: 2605.21693 · v2 · pith:V4FQOLLNnew · submitted 2026-05-20 · 🧮 math.NA · cs.NA

Runge--Kutta numerical methods for ruin probabilities in classical risk model

George Kanakoudis , Lazaros Kanellopoulos This is my paper

Pith reviewed 2026-05-22 08:51 UTC · model grok-4.3

classification 🧮 math.NA cs.NA

keywords Runge-Kutta methodsruin probabilityclassical risk modelVolterra integro-differential equationnumerical quadratureGamma distributionPareto distributioninsurance mathematics

0 comments

The pith

Fourth-order Runge-Kutta schemes combined with Newton-Cotes and Pareto-adapted Gauss-Jacobi quadrature compute ruin probabilities from the Volterra integro-differential equation for Gamma and Pareto claim sizes.

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

The paper develops fourth-order one-step and two-step Runge-Kutta methods to solve the Volterra integro-differential equation that describes ruin probabilities in the classical risk model. These integrators are paired with Newton-Cotes quadrature and a specialized Gauss-Jacobi rule adapted to Pareto tails in order to approximate the convolution integral that arises from the claim-size distribution. Implementations are given for both Gamma and Pareto claim sizes, together with an alternative reformulation of the integro-differential equation as a first-order system of ordinary differential equations. A sympathetic reader would care because closed-form ruin probabilities are rarely available for realistic claim distributions, so reliable numerical schemes directly support quantitative risk assessment in insurance.

Core claim

The central claim is that fourth-order Runge-Kutta discretizations of the Volterra integro-differential equation, when the convolution term is replaced by Newton-Cotes or Pareto-adapted Gauss-Jacobi quadrature, produce accurate numerical values of the ruin probability for the classical risk model with Gamma or Pareto claim sizes.

What carries the argument

Fourth-order one-step and two-step Runge-Kutta schemes that advance the solution of the Volterra integro-differential equation while the convolution integral is replaced by Newton-Cotes or Gauss-Jacobi quadrature rules.

If this is right

The methods achieve the convergence order expected from fourth-order Runge-Kutta integration for both Gamma and Pareto claim sizes.
Reformulating the Volterra equation as an equivalent first-order system of ordinary differential equations supplies an alternative route to the same ruin probabilities.
Newton-Cotes rules suffice for light-tailed Gamma claims while the Pareto-adapted Gauss-Jacobi quadrature handles the heavy tail of the Pareto distribution.
Numerical results illustrate that the combined schemes remain stable and accurate over a practical range of initial capital values.

Where Pith is reading between the lines

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

The same discretization framework could be applied to other claim-size distributions whose densities admit efficient quadrature, such as lognormal or Weibull laws.
Because the underlying Volterra equation is linear, the computed ruin probabilities can be differentiated with respect to model parameters to obtain sensitivity measures useful for premium calculation.
The approach supplies a template for extending high-order numerical methods to related integro-differential equations that appear in ruin theory with investments or in renewal risk models.

Load-bearing premise

The Volterra integro-differential equation is an exact representation of the ruin probability and the chosen Runge-Kutta and quadrature discretizations converge to its solution at the expected order without introducing uncontrolled truncation or stability errors for the tested distributions.

What would settle it

Direct comparison of the numerical ruin-probability values against independently computed high-precision reference solutions or known exact expressions for small initial surpluses under a Gamma claim-size distribution would show whether the observed error behaves as predicted by the fourth-order theory.

read the original abstract

In this paper, we study Runge--Kutta methods for the computation of ruin probabilities in the classical risk model through the associated Volterra integro-differential equation. The proposed framework combines fourth-order one-step and two-step Runge--Kutta schemes with numerical quadrature formulas to approximate the convolution term. In particular, the convolution term is approximated using Newton--Cotes and Gaussian quadrature formulas, including Simpson's 1/3 rule and Pareto-adapted Gauss--Jacobi quadrature. An equivalent reformulation of the Volterra equation as a system of ordinary differential equations is also considered. Implementations for Gamma and Pareto claim-size distributions are developed. Numerical results are presented to illustrate the effectiveness of the proposed 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.

Desk Editor's Note private letter to a colleague

This paper discretizes the ruin-probability Volterra equation with standard fourth-order Runge-Kutta plus Newton-Cotes or Pareto-adapted Gauss-Jacobi quadrature and shows numerical results for Gamma and Pareto claims, but supplies no convergence analysis or error bounds.

read the letter

The main takeaway is that they take the known Volterra integro-differential equation for ruin probability and apply fourth-order one-step and two-step Runge-Kutta schemes, approximating the convolution with Simpson's rule or a tailored Gauss-Jacobi quadrature. They also rewrite the equation as an ODE system and give implementations for Gamma and Pareto claim sizes, with some numerical outputs to show the approach in practice. That combination for this specific risk-theory setting looks like the only real novelty; the building blocks are established methods applied to an existing equation. The implementations and examples are clear enough that someone needing to compute these probabilities numerically could follow the recipes without much trouble. The soft spot is exactly what the stress-test note flags: no order analysis or global error bound is given, so it is not shown that the quadrature errors preserve the fourth-order accuracy of the RK step, particularly when the Pareto tail limits smoothness after the change of variables. The results are only illustrations, with no baseline comparisons or stability checks, which leaves open whether the observed accuracy holds more generally or just for the chosen parameters. This is the sort of targeted numerical-methods paper that would interest readers working on computational insurance mathematics or Volterra equations in applied probability. It is competent and concrete, so it deserves a serious referee who can ask for the missing convergence work. I would send it to review rather than desk-reject, with the expectation that the authors add error estimates or at least careful numerical order tests before acceptance.

Referee Report

2 major / 1 minor

Summary. The paper develops fourth-order one-step and two-step Runge-Kutta schemes to solve the Volterra integro-differential equation for ruin probabilities in the classical risk model. The convolution term is discretized via Newton-Cotes formulas (including Simpson's 1/3 rule) and Pareto-adapted Gauss-Jacobi quadrature; an equivalent first-order ODE system reformulation is also treated. Implementations are given for Gamma and Pareto claim-size distributions, with numerical experiments presented to illustrate effectiveness.

Significance. If the discretizations are shown to attain their nominal order with controlled global error, the work would supply practical, high-order numerical tools for an important class of Volterra equations arising in insurance mathematics. The Pareto-adapted quadrature is a sensible specialization for heavy-tailed claims and, if validated, would be a useful technical contribution.

major comments (2)

[§3] §3 (numerical methods): No a priori error analysis or stability result is supplied for the combined Runge-Kutta/quadrature discretization of the Volterra IDE. In particular, it is not shown that the O(h^4) quadrature error (Simpson or mapped Gauss-Jacobi) does not reduce the local truncation error of the fourth-order RK step, nor is a global error bound derived for the resulting semi-discrete system.
[Numerical experiments] Numerical experiments section: The reported computations for Gamma and Pareto distributions demonstrate practical performance but supply neither observed convergence rates, a priori error bounds, nor comparisons against exact solutions or established methods; this leaves the central accuracy claim dependent on visual inspection alone.

minor comments (1)

[Abstract] The abstract states that the methods 'accurately compute' ruin probabilities; this phrasing should be softened to 'are used to compute' until convergence is established.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thorough review and valuable suggestions. We address the major comments point by point below, indicating how we plan to revise the manuscript to strengthen the presentation of the numerical methods and their validation.

read point-by-point responses

Referee: [§3] §3 (numerical methods): No a priori error analysis or stability result is supplied for the combined Runge-Kutta/quadrature discretization of the Volterra IDE. In particular, it is not shown that the O(h^4) quadrature error (Simpson or mapped Gauss-Jacobi) does not reduce the local truncation error of the fourth-order RK step, nor is a global error bound derived for the resulting semi-discrete system.

Authors: We agree with the referee that the manuscript lacks a dedicated a priori error analysis for the discretization. Providing a complete global error bound would require significant additional theoretical work on the stability of the Volterra IDE discretization, which we believe is a substantial undertaking. In the revision, we will add a local truncation error analysis in Section 3 to show that the quadrature errors are consistent with the fourth-order accuracy of the Runge-Kutta methods for sufficiently smooth integrands. We will also include a brief discussion on stability. For a full rigorous global error estimate, this may be left for future research. revision: partial
Referee: [Numerical experiments] Numerical experiments section: The reported computations for Gamma and Pareto distributions demonstrate practical performance but supply neither observed convergence rates, a priori error bounds, nor comparisons against exact solutions or established methods; this leaves the central accuracy claim dependent on visual inspection alone.

Authors: The referee is correct that the numerical results rely primarily on visual inspection without quantitative measures. We will revise the numerical experiments section to include tables of computed errors and observed convergence rates, using a reference solution computed on a fine grid. Where possible, for the Gamma distribution, we will compare against known exact expressions or high-precision benchmarks. For the Pareto case, we will provide comparisons with alternative numerical approaches from the literature on ruin probability computation. revision: yes

Circularity Check

0 steps flagged

No circularity: standard numerical discretization of a pre-existing Volterra IDE

full rationale

The paper applies fourth-order Runge-Kutta schemes and standard quadrature (Newton-Cotes, Simpson, Pareto-adapted Gauss-Jacobi) to discretize the known Volterra integro-differential equation for ruin probabilities in the classical risk model. The equation itself is taken from risk theory as an exact representation; the work consists of implementation for Gamma and Pareto claim sizes plus numerical illustrations. No quantity is defined in terms of itself, no fitted parameter is relabeled as a prediction, and no self-citation supplies a load-bearing uniqueness or ansatz. The derivation chain is therefore a direct, non-circular application of existing numerical methods to a standard model equation, with observed accuracy verified by computation rather than by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The approach rests on the standard modeling assumption that ruin probability satisfies the classical Volterra integro-differential equation; no new free parameters, invented entities, or ad-hoc axioms are introduced in the abstract.

axioms (1)

domain assumption Ruin probability in the classical risk model satisfies a Volterra integro-differential equation whose kernel is the claim-size density.
This equation is the starting point for all numerical work described.

pith-pipeline@v0.9.0 · 5648 in / 1240 out tokens · 37224 ms · 2026-05-22T08:51:37.908025+00:00 · methodology

Runge--Kutta numerical methods for ruin probabilities in classical risk model

Core claim

What carries the argument

If this is right

Where Pith is reading between the lines

Load-bearing premise

What would settle it

discussion (0)