Infinite-time ruin probability of a multivariate renewal risk model with Brownian perturbations

Basic Information

• Paper ID: 2510.11229
• Title: Infinite-time ruin probability of a multivariate renewal risk model with Brownian perturbations
• Author: Dimitrios G. Konstantinides
• Classification: math.PR (Probability Theory)
• Publication Date: October 14, 2025
• Paper Link: https://arxiv.org/abs/2510.11229

Abstract

This paper investigates multivariate risk models with a common renewal process and Brownian perturbations. Assuming that the integrated tail distribution of claims follows a multivariate subexponential distribution, asymptotic relations for the infinite-time ruin probability are established. More explicit expressions are provided when the claim distribution belongs to multivariate regular variation. The results demonstrate that the asymptotic behavior of ruin probability is insensitive to Brownian perturbations. Furthermore, it is proven that multivariate distributions with finite expectation belonging to the intersection of multivariate controlled variation and long-tailed distribution classes have integrated tail distributions in the multivariate subexponential class, facilitating the verification of conditions in the theorems.

Research Background and Motivation

1. Problem Background: Modern insurance companies are compelled to maintain multiple risk portfolios and must confront ruin events as well as dependency structures among portfolios. Traditional univariate risk models are insufficient for practical needs.
2. Problem Significance:
   • Insurance company bankruptcy problems under heavy-tailed distributions have attracted considerable attention since the late twentieth century
   • Multivariate risk models better describe real-world multi-line insurance operations
   • Brownian perturbations reflect additional sources of randomness from premiums or claims
3. Limitations of Existing Methods:
   • Previous research primarily focused on cases without Brownian perturbations (δ = 0)
   • Lack of systematic analysis of Brownian perturbation effects
   • Restrictive conditions on multivariate regular variation distributions lead to excessive premium pricing
4. Research Motivation:
   • Extend existing theory to include Brownian perturbations
   • Prove the insensitivity of ruin probability asymptotic behavior to Brownian perturbations
   • Provide more general distribution class conditions for verification

Core Contributions

1. Theoretical Extension: Extends the results of Samorodnitsky and Sun (2016) from the case without Brownian perturbations to the case with Brownian perturbations (δ > 0)
2. Insensitivity Proof: Demonstrates that the asymptotic behavior of ruin probability is insensitive to Brownian perturbations, with identical asymptotic expressions regardless of their presence
3. Condition Simplification: Proves that distributions belonging to the intersection of multivariate controlled variation and long-tailed distribution classes (D ∩ L)_R with finite expectation have integrated tail distributions in the multivariate subexponential class S_R
4. Univariate Case Generalization: Even in the univariate case, the results generalize Veraverbeke's (1993) theorem

Methodology Details

Task Definition

Study the ruin probability of d-dimensional insurance business lines, where:

• Input: Initial capital x, premium rate vector p, claim vector sequence {X^(i)}, Brownian motion perturbations
• Output: Asymptotic expression for infinite-time ruin probability ψ_{b,L}(x)
• Constraints: Integrated tail distribution of claim distribution belongs to multivariate subexponential class

Model Architecture

1. Risk Model Setup The surplus process is defined as:

U(t) = xb + tp - Σ_{i=1}^{N(t)} X^(i) + δ ⊙ B(t)

where:

• x > 0: Initial capital
• b = (b_1,...,b_d): Capital allocation vector with Σb_i = 1
• p = (p_1,...,p_d): Premium rate vector
• {X^(i)}: i.i.d. claim vector sequence
• {N(t)}: Renewal counting process
• δ ≥ 0: Diffusion coefficient vector
• {B(t)}: Multivariate Brownian motion

2. Key Assumptions

• Assumption 1.1: Claim vectors are i.i.d. and independent of arrival times; Brownian motion is independent of other random sources and has non-negative expectation
• Assumption 2.1: Ruin set L is open, L^c is convex, -L is increasing, 0 ∈ ∂L

3. Ruin Probability Expression

ψ_{b,L}(x) = P(Σ_{i=1}^n W^(i) ∈ xA, ∃n ∈ N)

where W^(i) = Z^(i) - δ ⊙ B^(i), Z^(i) = X^(i) - θ_i p, A = (b - L) ∈ R

Technical Innovations

1. Integrated Tail Distribution Method Introduces the integrated tail distribution:

F_I(xA) = (1/Θ) ∫_0^∞ F(xA + vc) dv

2. Brownian Perturbation Treatment By introducing a new net profit condition:

c* = -E[W] = c + δ ⊙ μ > 0

where μ = EB

3. Insensitivity Proof Technique

• Utilizes the increasing properties of set A
• Proves asymptotic equivalence through upper and lower bound estimates
• Key insight: proving ∫_0^∞ P(W ∈ xA + vc*) dv ~ H*(x)

Experimental Setup

Theoretical Verification Framework

This is a purely theoretical study verified through mathematical proofs:

1. Preliminary Lemmas

• Lemma 3.1: Proves asymptotic relations when F_I* ∈ L_A
• Lemma 3.2: Proves H*(x) ~ H(x)

2. Main Theorem Proofs

• Theorem 3.1: Establishes asymptotic relations for ruin probability
• Corollary 3.1: Provides explicit expressions for multivariate regular variation case

3. Distribution Class Verification

• Proposition 4.1: Proves that integrated tails of (D ∩ L)_R class distributions belong to S_R
• Example 4.1: Constructs specific distribution examples

Comparative Methods

• Samorodnitsky & Sun (2016): Multivariate case without Brownian perturbations
• Veraverbeke (1993): Univariate Poisson process case
• Embrechts & Veraverbeke (1982): Classical risk model

Experimental Results

Main Results

Theorem 3.1 (Main Result): Under conditions EX < ∞, F_I ∈ S_A, and net profit condition (2.4):

ψ_{b,L}(x) ~ H(x) = ∫_0^∞ F(xA + vc) dv

Corollary 3.1 (Regular Variation Case): When F ∈ MRV(α,μ), α ∈ (1,∞):

ψ_{b,L}(x) ~ xV(x) ∫_0^∞ μ(b - L + vc) dv

Theoretical Findings

1. Insensitivity: The asymptotic behavior of ruin probability is independent of whether Brownian perturbations exist, with identical asymptotic expressions

2. Condition Simplification: Proposition 4.1 proves (D ∩ L)_R ⊂ S_R, providing more easily verifiable sufficient conditions

3. Univariate Generalization: In the univariate case, results reduce to:

ψ(x) ~ (1/c) ∫_x^∞ F(v) dv

Related Work

Main Research Directions

1. Univariate Heavy-Tailed Risk Models: Classical work by Embrechts & Klüppelberg et al.
2. Multivariate Risk Models: Extensions by Hult & Lindskog, Konstantinides & Li et al.
3. Brownian Perturbation Models: Research by Veraverbeke, Chen et al.

Advantages of This Paper

1. Theoretical Completeness: First systematic study of Brownian perturbation effects in multivariate renewal risk models
2. Relaxed Conditions: Extension from MRV to more general S_R class
3. Methodological Innovation: Develops new techniques for handling Brownian perturbations

Conclusions and Discussion

Main Conclusions

1. In multivariate renewal risk models, Brownian perturbations do not affect the asymptotic behavior of ruin probability
2. Subexponentiality of integrated tail distribution is the key condition
3. The (D ∩ L)_R class provides more practical tools for condition verification

Limitations

1. Assumption Restrictions: Requires Brownian motion to have non-negative expectation
2. Distribution Conditions: Still requires heavy-tail assumptions
3. Practical Application: Lacks numerical experiment verification

Future Directions

1. Extend to time-dependent risk models
2. Consider more general perturbation processes
3. Study finite-time ruin probability

In-Depth Evaluation

Strengths

1. Significant Theoretical Contribution: First proof of multivariate ruin probability insensitivity to Brownian perturbations
2. Mathematical Rigor: Complete proofs with sophisticated technical handling
3. Practical Value: Provides theoretical foundation for insurance practice
4. Clear Writing: Well-structured with precise definitions

Weaknesses

1. Lack of Application Verification: Pure theoretical study without numerical experiments
2. Strong Assumptions: Non-negative expectation assumption for Brownian motion may limit applications
3. Computational Complexity: Practical computation of integrated tail distributions may be difficult

Impact

1. Theoretical Significance: Enriches multivariate risk theory and advances the field
2. Practical Value: Provides tools for multi-line insurance company risk management
3. Extensibility: Establishes foundation for subsequent research

Applicable Scenarios

1. Multi-line Insurance Companies: Risk assessment for correlated business lines
2. Heavy-Tailed Claims Environment: Applicable to insurance markets with frequent extreme events
3. Theoretical Research: Provides new analytical framework for risk theory researchers

References

This paper cites 32 important references, primarily including:

• Samorodnitsky & Sun (2016): Theoretical foundation of multivariate subexponential distributions
• Veraverbeke (1993): Univariate Brownian perturbation models
• Embrechts et al.: Classical works on heavy-tailed distributions and extreme value theory
• Konstantinides series: Systematic research on multivariate risk models