The Classical Cramer-Lundberg Model

Ruin theory studies the probability that an insurer's surplus falls below zero, rendering it unable to pay claims. The Cramer-Lundberg model, developed in the early twentieth century, provides the foundational framework. It represents the surplus process as U(t) = u + ct - S(t), where u is the initial capital, c is the premium rate per unit time, and S(t) is a compound Poisson process representing aggregate claims up to time t.

In this model, claims arrive according to a Poisson process with rate lambda, and each claim has a random size drawn independently from a common distribution. The premium rate c must exceed the expected claim rate (c > lambda E[X], where E[X] is the expected claim size) for the company to be profitable on average. The safety loading theta = (c - lambda E[X]) / (lambda E[X]) measures the percentage by which premiums exceed expected claims.

Exact Solutions and Bounds

The probability of ultimate ruin, psi(u), depends on the initial surplus u, the safety loading, and the claim size distribution. For exponentially distributed claims with mean 1/alpha, an exact closed-form solution exists: psi(u) = [lambda / (alpha c)] exp(-(alpha - lambda/c) u). This formula illustrates the general behavior: ruin probability decreases exponentially with initial surplus, and higher safety loading reduces ruin probability.

For general claim size distributions, exact solutions are rarely available. Two important analytical results provide approximations and bounds:

The Pollaczek-Khinchine formula expresses the ruin probability as a geometric sum involving the integrated tail distribution of the claim size. It shows that psi(u) = (1 - theta/(1+theta)) sum_{n=0}^{infinity} [1/(1+theta)]^n [1 - F_e^{*n}(u)], where F_e is the equilibrium distribution of the claim sizes
The Cramer-Lundberg bound provides an exponential upper bound: psi(u) <= exp(-Ru), where R is the adjustment coefficient (also called the Lundberg exponent). R is the unique positive solution to the equation lambda(M_X(r) - 1) = cr, where M_X is the moment generating function of the claim size distribution

The adjustment coefficient R is a key quantity in ruin theory. A larger R means ruin probability decreases faster with initial surplus, indicating a more stable risk process. R increases with the safety loading and decreases with the variability of claim sizes.

The Sparre Andersen Model

The Sparre Andersen model (also called the renewal risk model) relaxes the Poisson assumption for claim arrivals. Instead of exponentially distributed inter-arrival times, the model allows any distribution for the time between claims. This generalization is important because real claim arrivals may not follow the memoryless pattern assumed by the Poisson process.

When inter-arrival times follow an Erlang distribution (the sum of k independent exponential random variables), the model remains analytically tractable. The ruin probability can be computed using the roots of the generalized Lundberg equation. For general inter-arrival distributions, numerical methods or simulation are typically required.

Perturbed Risk Models

Perturbed risk models add a diffusion (Brownian motion) component to the classical surplus process: U(t) = u + ct - S(t) + sigma W(t). The diffusion term captures the effect of many small, frequent fluctuations in premiums and claims that are not well represented by the compound Poisson process alone. These might include small claims below the reporting threshold, premium collection timing variations, or investment income fluctuations.

The addition of the diffusion component generally increases the ruin probability because the Brownian motion paths can push the surplus below zero even in the absence of large claims. The ruin probability for the perturbed model satisfies a more complex integro-differential equation than the classical model, but analytical solutions exist for certain claim size distributions.

The Gerber-Shiu Discounted Penalty Function

The Gerber-Shiu discounted penalty function, introduced in 1998, unifies the analysis of several ruin-related quantities into a single framework. It is defined as m(u) = E[exp(-delta T) w(U(T-), |U(T)|) I(T < infinity)], where T is the time of ruin, delta is a discount factor, U(T-) is the surplus just before ruin, |U(T)| is the deficit at ruin, and w is a penalty function.

By choosing different penalty functions and discount factors, this single framework yields:

The ruin probability (set delta = 0 and w = 1)
The Laplace transform of the ruin time (set w = 1)
The distribution of the deficit at ruin (set delta = 0 and choose w appropriately)
The joint distribution of the surplus before ruin and the deficit at ruin

The Gerber-Shiu function satisfies a defective renewal equation that can be solved using analytical or numerical methods. This unified approach has become the standard framework for analyzing ruin-related problems in modern actuarial literature.

Practical Relevance and Modern Extensions

While ruin theory originated as a theoretical framework, it has practical applications in regulatory capital calculations, reinsurance program design, and enterprise risk management. The concept of ruin probability is closely related to Value at Risk and expected shortfall, which are standard measures in modern solvency frameworks. The adjustment coefficient and related quantities inform the calibration of capital requirements that ensure a sufficiently low probability of insolvency.

Modern extensions include models with dividend payments (where the company pays dividends to shareholders when surplus is high), models with investment income (where the surplus earns interest), and models with dependent claim sizes and inter-arrival times. Levy process models provide a general framework that encompasses many of these extensions while maintaining some analytical tractability. These developments continue to enrich our understanding of insurance company solvency and inform the design of regulatory frameworks.

Ruin Theory: Classical and Modern Approaches