Adjustment Coefficient for Exponential Claims

Why the Exponential Case Has a Closed-Form Solution

The adjustment coefficient is a positive number obtained from an equation involving the claim-size moment-generating function. It is useful in the classical surplus model because it summarizes how the claim distribution and premium rate jointly affect the model’s exponential ruin bound. For exponential claim amounts, the equation can be solved algebraically rather than by numerical root-finding.

The algebra has an important qualification: the equation always has a root at zero. The adjustment coefficient of interest is the positive, nonzero root, and this root exists only under a premium condition. In the exponential case, both the formula and the condition can be stated explicitly.

Model and notation: Assume claims arrive according to a Poisson process with rate $\lambda>0$ per unit time. Claim amounts are independent and identically distributed, independent of the arrival process, and exponentially distributed with rate parameter $\beta>0$. Thus $E[X]=1/\beta$. Premium income is received continuously at rate $c>0$ per unit time. The classical surplus process is $U(t)=u+ct-\sum_{j=1}^{N(t)}X_j$, where $u$ is initial surplus and $N(t)$ is the claim count by time $t$.

For this model, the adjustment coefficient equation is

\lambda\bigl(M_X(r)-1\bigr)=cr,

where $M_X(r)=E[e^{rX}]$ is the moment-generating function of a claim amount. The adjustment coefficient is a positive solution $r$ of this equation. The claim-size moment-generating function must be finite at that value of $r$.

Deriving the Exponential-Claim Formula

For an exponential random variable with rate $\beta$, the density is $f_X(x)=\beta e^{-\beta x}$ for $x\geq 0$. Its moment-generating function is finite only when $r<\beta$. Within that domain,

M_X(r)=\int_0^\infty e^{rx}\beta e^{-\beta x}\,dx =\beta\int_0^\infty e^{-(\beta-r)x}\,dx =\frac{\beta}{\beta-r}.

Substituting this expression into the adjustment coefficient equation gives the equation below. The root $r=0$ is always present because $M_X(0)=1$. To find a positive root, take $r>0$ and divide both sides by $r$; this step is why the zero root must be considered separately.

\lambda\left(\frac{\beta}{\beta-r}-1\right)=cr \quad\Longleftrightarrow\quad \lambda\frac{r}{\beta-r}=cr \quad\Longleftrightarrow\quad \frac{\lambda}{\beta-r}=c.

The last equation yields the only possible nonzero solution. Its validity still needs checking: the solution must be positive and must belong to the moment-generating function’s domain $r<\beta$.

R=\beta-\frac{\lambda}{c}.

Theorem: For Poisson claim arrivals with rate $\lambda>0$, exponential claim amounts with rate $\beta>0$, and constant premium rate $c>0$, a positive adjustment coefficient exists if and only if $c>\lambda/\beta=\lambda E[X]$. When it exists, it is $$ R=\beta-\frac{\lambda}{c}. $$ If $c\leq\lambda/\beta$, there is no positive adjustment coefficient. The equation still has its zero root.

To check the condition, positivity of the candidate requires $\beta-\lambda/c>0$. Since $c>0$, this is equivalent to $c>\lambda/\beta$. Under that same condition, $\lambda/c>0$, so $R=\beta-\lambda/c<\beta$. Thus the positive candidate is inside the moment-generating function’s domain. This also shows why simply obtaining an algebraic expression is not enough: a valid adjustment coefficient must satisfy both the positivity and domain requirements.

Connection with the Safety Loading

The net profit condition for this model is $c>\lambda E[X]$: expected premium income per unit time exceeds expected claim outgo per unit time. If the premium rate is written using a safety loading $\theta$ as $c=(1+\theta)\lambda E[X]$, then $c=(1+\theta)\lambda/\beta$. Substitution gives a convenient equivalent formula when $\theta>0$.

R=\beta-\frac{\lambda}{(1+\theta)\lambda/\beta} =\beta-\frac{\beta}{1+\theta} =\frac{\beta\theta}{1+\theta}.

This form makes the condition transparent. A positive safety loading, $\theta>0$, gives $R>0$. At $\theta=0$, the displayed expression is zero, not a positive adjustment coefficient. If $-1<\theta<0$, the premium rate is positive but below expected claim outgo, and the algebraic nonzero root is negative. A negative value is not an adjustment coefficient.

Specify the model parameters.
Record $\lambda$, the exponential rate $\beta$, and the premium rate $c$, with their time and monetary units.

Check the premium condition.
Compute expected claim outgo per unit time, $\lambda E[X]=\lambda/\beta$. A positive root requires $c>\lambda/\beta$.

Calculate the candidate.
If the condition holds, use $R=\beta-\lambda/c$, or $R=\beta\theta/(1+\theta)$ when the safety loading is specified.

Check the original equation and domain.
Confirm $0<R<\beta$, then substitute into $\lambda(M_X(R)-1)=cR$.

Worked Example: Exponential Claims with Dollar Units

A hypothetical insurer has Poisson claim arrivals at rate $\lambda=1$ claim per year. Claim amounts are exponential with mean $\$800$, and the continuous premium rate is $c=\$1{,}000$ per year. Find the adjustment coefficient, state whether a positive root exists, and verify the result. Report $R$ to six decimal places in inverse dollars.

Method and condition. The exponential rate is $\beta=1/800=0.00125$ per dollar. Expected claim outgo per year is

\lambda E[X]=1\times \$800=\$800\text{ per year}.

Since $c=\$1{,}000$ per year is greater than $\$800$ per year, the net profit condition holds. The implied safety loading is $1000/800-1=0.25$, or $25\%$.

Calculate the root. Using the rate-per-dollar parameterization,

R=\beta-\frac{\lambda}{c} =0.00125-\frac{1}{1000} =0.00125-0.001 =0.00025\text{ per dollar}.

The candidate is positive and $0.00025<0.00125=\beta$, so it is in the domain of the moment-generating function.

Verify in the original equation. At this value,

M_X(R)=\frac{0.00125}{0.00125-0.00025} =\frac{0.00125}{0.001} =1.25.

Therefore the left side is $1(1.25-1)=0.25$ per year, while the right side is $1000(0.00025)=0.25$ per year. They agree. The adjustment coefficient is $0.000250$ per dollar. The calculation also checks the premium: $1.25\times \$800=\$1{,}000$, not $\$800$.

Common mistake avoided. Comparing premium with expected claim outgo requires multiplying the claim mean by the arrival rate. Here that outgo is $\$800$ per year, and the premium is $\$1{,}000$ per year. The safety-loaded premium is $1.25\times \$800=\$1{,}000$; treating it as $\$800$ would erase the loading and lead to the wrong existence conclusion.

Worked Example: Finding the Root from a Safety Loading

For a hypothetical portfolio, claims arrive at rate $\lambda=3$ per year and have exponential mean $\$20{,}000$. Premium is charged at a $25\%$ safety loading over expected claim outgo. Find $c$ and $R$, expressing claim amounts in thousands of dollars. Verify the equation and domain.

Method. In thousands of dollars, the mean claim is $20$, so $\beta=1/20=0.05$ per thousand dollars. Expected claim outgo per year is $3\times20=60$ thousand dollars. Thus

c=(1+0.25)(60)=75\text{ thousand dollars per year}.

Since $75>60$, the net profit condition holds. The closed-form root is

R=0.05-\frac{3}{75} =0.05-0.04 =0.01\text{ per thousand dollars}.

This is positive and less than $\beta=0.05$, as required. The moment-generating function at the candidate is $M_X(R)=0.05/(0.05-0.01)=0.05/0.04=1.25$. The two sides of the defining equation are

\lambda(M_X(R)-1)=3(1.25-1)=0.75, \qquad cR=75(0.01)=0.75.

The equation is satisfied, and the adjustment coefficient is $0.01$ per thousand dollars, equivalent to $0.00001$ per dollar. Keeping the units consistent matters: using $0.05$ per thousand with a premium rate in dollars per year would mix scales.

Worked Example: A Premium Below Expected Claim Outgo

A hypothetical insurer has $\lambda=2$ claims per year, exponential claim amounts with mean $40$ thousand dollars, and premium income $c=70$ thousand dollars per year. Determine whether a positive adjustment coefficient exists. Explain what the algebraic candidate indicates.

Check the condition first. Here $\beta=1/40=0.025$ per thousand dollars, and expected claim outgo per year is $2\times40=80$ thousand dollars. Since $70<80$, the net profit condition fails.

The nonzero-root formula gives the candidate

\beta-\frac{\lambda}{c} =0.025-\frac{2}{70} =0.025-0.0285714\ldots =-0.0035714\ldots

This candidate is negative, so it is not a positive adjustment coefficient. There is no positive root. The equation’s root at zero remains, but it does not satisfy the definition of a positive adjustment coefficient. Reporting the negative candidate as $R$, or taking its absolute value, would be incorrect.

Common Exam Traps and Interpretation

Keeping the zero root. Since $M_X(0)=1$, $r=0$ satisfies the equation. The adjustment coefficient is the positive nonzero root, when one exists.
Skipping the premium condition. The expression $\beta-\lambda/c$ is positive only when $c>\lambda/\beta$. A nonpositive result is not made valid by changing its sign.
Ignoring the moment-generating function domain. For exponential claims, $M_X(r)$ is finite only for $r<\beta$. The positive root meets this condition when the net profit condition holds.
Mixing units. If $\beta$ is measured per thousand dollars, use $c$ in thousands of dollars per unit time. The adjustment coefficient then has units per thousand dollars.
Confusing the root with an exact ruin probability. The adjustment coefficient is a model parameter used in ruin analysis; it is not itself a ruin probability or an exact probability calculation.

The calculation is specific to exponential claim amounts: the simple rational form of their moment-generating function makes the nonzero root linear after cancellation. For other severities, the same adjustment coefficient equation may not have a closed-form solution. Whatever the severity, a proposed root must satisfy the original equation and the relevant moment-generating function’s domain.

Key takeaway: For exponential claims with rate $\beta$, arrival rate $\lambda$, and premium rate $c$, the positive adjustment coefficient is $R=\beta-\lambda/c$ precisely when $c>\lambda E[X]=\lambda/\beta$. With safety loading $c=(1+\theta)\lambda E[X]$, this is $R=\beta\theta/(1+\theta)$ for $\theta>0$.

Check Your Understanding

For each problem, check the premium condition before using the formula. Keep the stated units through the calculation.

A hypothetical portfolio has $\lambda=2$ claims per year, exponential mean claim size $30$ thousand dollars, and premium rate $75$ thousand dollars per year. Find the safety loading and adjustment coefficient. (Answer: $\theta=0.25$; $R=1/150=0.006667$ per thousand dollars.)
Claims arrive at rate $4$ per year and have exponential mean $15$ thousand dollars. Premium is set at a $50\%$ safety loading. Find $c$ and $R$. (Answer: $c=90$ thousand dollars per year; $R=1/45=0.02222$ per thousand dollars.)
A hypothetical insurer has $\lambda=1.5$ claims per year, exponential mean claim size $\$400$, and premium rate $\$600$ per year. Does it have a positive adjustment coefficient? (Answer: No; $c=\lambda E[X]=\$600$ per year, so the nonzero root is zero, not positive.)
Claims have exponential rate $\beta=0.04$ per thousand dollars, arrive at rate $5$ per year, and premium is $150$ thousand dollars per year. Find expected claim outgo, the safety loading, and $R$. (Answer: expected outgo $=125$ thousand dollars per year; $\theta=0.20$; $R=1/150=0.006667$ per thousand dollars.)
For exponential claims with $\beta=0.02$ per thousand dollars and $\lambda=2$ per year, find the minimum condition on $c$ for a positive root. Then calculate $R$ if $c=125$ thousand dollars per year and verify the defining equation. (Answer: $c>100$ thousand dollars per year; at $c=125$, $R=0.004$ per thousand dollars, with $\lambda(M_X(R)-1)=2(1.25-1)=0.5=cR$.)

What You'll Learn