Calculating the Adjustment Coefficient

The Surplus Process and the Concept of Ruin

Let's begin by recalling the insurer's surplus process, $U(t)$, which represents the capital at time $t$:

Here, $u$ is the initial capital, $c$ is the constant rate of premium income, and $S(t)$ is the aggregate claims process. Ruin occurs if this surplus ever drops below zero.

The following video provides a concise visual and mathematical introduction to ruin theory and the surplus process.

Defining the Adjustment Coefficient

The central question in ruin theory is: what is the probability that $U(t)$ will eventually fall below zero? The adjustment coefficient, typically denoted by $R$, is a key parameter used to answer this question. It acts as an inverse measure of risk—a larger $R$ implies a more stable, less risky portfolio.

The adjustment coefficient is defined in the context of the Cramér-Lundberg model, where we assume:

1. Claims occur according to a Poisson process with rate $\lambda$.
2. Individual claim amounts, $X_i$, are independent and identically distributed random variables, with distribution $X$.
3. The premium rate $c$ is constant.
4. The "net profit condition" holds: the expected premium income per unit of time is greater than the expected claim outgo per unit of time. That is, $c>\lambda E[X]$. This ensures that the surplus is expected to drift upwards over time.

Under these assumptions, the adjustment coefficient $R$ is defined as the unique positive solution to the Lundberg equation:

where $M_X(r)=E[e^{rX}]$ is the moment generating function (MGF) of the individual claim amount distribution $X$.

Let's break down this crucial equation. It creates a balance between the claim process (left side) and the premium income (right side). The left side involves the claim arrival rate $\lambda$ and the MGF of claim severity $M_X(r)$, which captures the entire distribution of claim sizes. The right side simply contains the premium rate $c$. The value $r=R$ at which these two sides are equal tells us something fundamental about the riskiness of the process.

This next resource provides a formal definition of the Lundberg equation and explains why a unique positive solution exists.

Calculating the Adjustment Coefficient: A Closed-Form Example

Now we move to the main task: calculating $R$. For certain claim severity distributions, the Lundberg equation can be solved analytically. The most common case, which you should be very comfortable with, is when claim severities follow an exponential distribution.

Let's assume individual claim amounts $X$ follow an exponential distribution with mean $\mu =1/\alpha$, so $X\sim \text{Exp}(\alpha )$.

1. Find the MGF of X:
   The MGF of an exponential distribution with rate $\alpha$ is $M_X(r)=\frac{\alpha }{\alpha -r}$, for $r<\alpha$.

2. Substitute into the Lundberg Equation:
   Plugging this into $\lambda (M_X(r)-1)=cr$, we get:

   $$\lambda \left(\frac{\alpha }{\alpha -r}-1\right)=cr$$

3. Solve for r:
   We can now solve this equation algebraically.

   $$\lambda \left(\frac{\alpha -(\alpha -r)}{\alpha -r}\right)=cr$$ $$\lambda \left(\frac{r}{\alpha -r}\right)=cr$$

   Since we are looking for the positive root $R$, we know $r\ne 0$, so we can divide both sides by $r$:

   $$\frac{\lambda }{\alpha -r}=c$$

   Rearranging to solve for $r$ gives:

   $$\lambda =c(\alpha -r)⟹\frac{\lambda }{c}=\alpha -r⟹r=\alpha -\frac{\lambda }{c}$$

   Thus, for an exponential claim severity distribution, the adjustment coefficient has a simple closed-form solution:

   $$R=\alpha -\frac{\lambda }{c}$$

The following reading walks through this exact calculation.

Calculation with Other Distributions

While the exponential case is elegant, most other claim severity distributions do not yield such a simple closed-form solution for $R$. In these cases, the adjustment coefficient is still the unique positive root of the Lundberg equation, but we must use numerical methods to find it.

For an exam, you typically won't be asked to perform the numerical root-finding itself, but you should be able to set up the equation that needs to be solved. The process remains the same: find the MGF of the claim severity distribution and substitute it into the Lundberg equation.

Let's look at the setup for a few other common distributions. The following resource provides the necessary MGFs and resulting equations.

Let's summarize the equations you need to be able to construct:

• Degenerate Claims: $X\equiv \beta$ (all claims are a fixed size $\beta$)

  • MGF: $M_X(r)=e^{r\beta }$
  • Equation to solve for $R$: $\lambda (e^{r\beta }-1)=cr$

• Uniform Claims: $X\sim U(a,b)$

  • MGF: $M_X(r)=\frac{e^{rb}-e^{ra}}{r(b-a)}$
  • Equation to solve for $R$: $\lambda \left(\frac{e^{rb}-e^{ra}}{r(b-a)}-1\right)=cr$

• Normal Claims: $X\sim N(\mu ,{\sigma }^2)$

  • MGF: $M_X(r)=e^{\mu r+\frac{1}{2}{\sigma }^2{r}^2}$
  • Equation to solve for $R$: $\lambda (e^{\mu r+\frac{1}{2}{\sigma }^2{r}^2}-1)=cr$

In a practical setting, an actuary would use software to find the root of these equations. For instance, in R, a function like uniroot can solve g(r) = 0 numerically. The paper you just reviewed, "ESTIMATING THE ADJUSTMENT COEFFICIENT IN RISK THEORY", actually contains extensive R code in its annex for this purpose, demonstrating how these theoretical concepts are implemented in practice.

Conclusion

In this lesson, we have formally introduced the adjustment coefficient and established the method for its calculation. This parameter is a compact and powerful measure of an insurer's solvency, integrating premium income with the frequency and severity of claims.

Key Takeaways:

• The adjustment coefficient $R$ is an inverse measure of risk in the Cramér-Lundberg model. A larger $R$ signifies a safer portfolio.
• $R$ is defined as the unique positive root of the Lundberg equation: $\lambda (M_X(r)-1)=cr$.
• The existence of $R>0$ is guaranteed by the net profit condition, $c>\lambda E[X]$.
• For exponentially distributed claims with rate $\alpha$, there is a closed-form solution: $R=\alpha -\lambda /c$.
• For most other claim distributions, $R$ must be found by numerically solving the Lundberg equation after substituting the appropriate MGF for the claim severity.

Preview of the Next Lesson

We have spent this lesson learning how to calculate the adjustment coefficient. In our next session, we will see its primary application. We will introduce Lundberg's inequality, which uses $R$ to place an upper bound on the probability of ultimate ruin. This will connect the computational skill you've learned today directly to the core objective of ruin theory: quantifying and managing the risk of insolvency.