Exponential Severity Increased Limits Factors

Why Exponential Severity Is Useful for Limit Comparisons

A policy limit caps what the insurer pays on an individual claim. Raising the limit increases expected payment per claim, but the increase depends on the severity distribution: a limit change matters more when a substantial share of claims reach large amounts. The increased limits factor (ILF) expresses the expected payment at a higher limit relative to the expected payment at a chosen basic limit.

This tutorial uses an exponential severity model to calculate those expected payments and factors. The model gives a compact formula, but the result is only appropriate when its assumptions are suitable for the hypothetical portfolio being analyzed.

Definition: Let $X$ be a claim severity with an exponential distribution and mean $\theta>0$. Its density and survival function, for $x\geq 0$, are $f_X(x)=\frac{1}{\theta}e^{-x/\theta}$ and $S_X(x)=P(X>x)=e^{-x/\theta}$. If the insurer pays the claim subject to a per-claim limit $u>0$, with no deductible, the payment is $\min(X,u)$. The increased limits factor from basic limit $b>0$ to limit $u$ is the expected payment at $u$ divided by the expected payment at $b$.

The exponential model has a constant hazard: for any $s,t\geq0$, $P(X>s+t\mid X>s)=P(X>t)$. Its only parameter here is the mean $\theta$, measured in the same units as claim severity and the limit. The model is continuous and has no probability mass at zero. Unless stated otherwise, limits apply separately to each claim, the same severity distribution applies at both limits, and there is no deductible, coinsurance, expense loading, or discounting.

Expected Payment at a Per-Claim Limit

For a nonnegative severity $X$, the expected payment capped at $u$ is the area under the survival function from zero to $u$. For exponential severity, integrating that survival function gives a particularly convenient expression.

E[\min(X,u)] =\int_0^u S_X(x)\,dx =\int_0^u e^{-x/\theta}\,dx =\theta\left(1-e^{-u/\theta}\right).

The first equality follows from the tail-integral identity for a nonnegative random variable $Z$: $E[Z]=\int_0^\infty P(Z>x)\,dx$. For $Z=\min(X,u)$, its survival probability is $P(Z>x)=P(X>x)=S_X(x)$ when $0\leq x<u$, and is zero for $x\geq u$. Thus only the interval from zero to $u$ contributes. Integrating $e^{-x/\theta}$ over that interval yields $\theta(1-e^{-u/\theta})$.

This expected payment is not $E[X]$ and is not the conditional mean $E[X\mid X\leq u]$. Claims exceeding the limit still produce the full capped payment $u$. As the limit grows, expected payment increases toward the unlimited mean $\theta$, but remains below it for every finite $u$.

Formula: Under the stated model, for basic limit $b$ and higher limit $u$, the increased limits factor is $$ \operatorname{ILF}(u;b) =\frac{E[\min(X,u)]}{E[\min(X,b)]} =\frac{1-e^{-u/\theta}}{1-e^{-b/\theta}}. $$ The factor is dimensionless. It equals 1 when $u=b$, and is greater than 1 when $u>b$. This formula compares expected loss payments per claim; by itself, it does not account for changes in claim frequency or other rating provisions.

Calculating a Factor from a Basic Limit

The calculation has three parts: evaluate the capped expected payment at the basic limit, evaluate it at the proposed higher limit, and divide the latter by the former. Keep the limit and the exponential mean in the same units. The factor uses the ratio of the two expected payments, not the ratio of the limits.

Worked Example: A Higher Limit Twice the Basic Limit

A hypothetical liability portfolio has exponential claim severity with mean $\$25{,}000$. Calculate the expected capped payment per claim at a basic limit of $\$10{,}000$ and a higher limit of $\$20{,}000$, then find the ILF. Assume no deductible and no discounting.

Method. Use the exponential capped-payment formula at each limit, then take the ratio. The common mistake to avoid is treating the ILF as $20{,}000/10{,}000=2$; the expected payment does not generally double when the limit doubles.

At the basic limit, $b/\theta=10{,}000/25{,}000=0.4$. Therefore,

E[\min(X,10{,}000)] =25{,}000(1-e^{-0.4}) =25{,}000(1-0.670320046) =25{,}000(0.329679954) =\$8{,}242.00.

At the higher limit, $u/\theta=20{,}000/25{,}000=0.8$. Therefore,

E[\min(X,20{,}000)] =25{,}000(1-e^{-0.8}) =25{,}000(1-0.449328964) =25{,}000(0.550671036) =\$13{,}766.78.

The factor is the higher-limit expected payment divided by the basic-limit expected payment:

\operatorname{ILF}(20{,}000;10{,}000) =\frac{0.550671036}{0.329679954} =1.670320\ldots \approx 1.6703.

The check is consistent with the payment amounts: $\$13{,}766.78/\$8{,}242.00$ is approximately 1.6703, and the factor is below 2 even though the limit doubled. Expected payments are rounded to cents and the factor to four decimal places.

Comparing More Than One Limit Choice

An insurer may need to compare several proposed limits against the same basic limit. The common denominator in each ILF is then the expected payment at that basic limit. Comparing expected payments also gives the dollar cost per claim of selecting one option rather than another.

Worked Example: Comparing Two Higher Limits

A hypothetical specialty insurer models claim severity as exponential with mean $\$30{,}000$. Its basic per-claim limit is $\$10{,}000$, and it is considering limits of $\$20{,}000$ and $\$40{,}000$. Calculate both ILFs and the expected payment at each limit. Identify the option with the greater expected payment per claim. Assume no deductible and no discounting.

Method. Calculate all three capped expected payments using the same mean, then divide each higher-limit payment by the basic-limit payment. The factor for the basic limit itself is 1.

For the basic limit, $10{,}000/30{,}000=1/3$, so

E[\min(X,10{,}000)] =30{,}000(1-e^{-1/3}) =30{,}000(1-0.716531311) =\$8{,}504.06.

For the $\$20{,}000$ limit, the exponent is $20{,}000/30{,}000=2/3$:

E[\min(X,20{,}000)] =30{,}000(1-e^{-2/3}) =30{,}000(1-0.513417119) =30{,}000(0.486582881) =\$14{,}597.49.

Its ILF is

\operatorname{ILF}(20{,}000;10{,}000) =\frac{0.486582881}{0.283468689} =1.716531\ldots \approx 1.7165.

For the $\$40{,}000$ limit, the exponent is $40{,}000/30{,}000=4/3$:

E[\min(X,40{,}000)] =30{,}000(1-e^{-4/3}) =30{,}000(1-0.263597138) =30{,}000(0.736402862) =\$22{,}092.09.

Its ILF is

\operatorname{ILF}(40{,}000;10{,}000) =\frac{0.736402862}{0.283468689} =2.597828\ldots \approx 2.5978.

The $\$40{,}000$ limit has the greater expected payment per claim: $\$22{,}092.09$, compared with $\$14{,}597.49$ at $\$20{,}000$. The higher option’s expected payment exceeds the lower option’s by $\$22{,}092.09-\$14{,}597.49=\$7{,}494.60$ per claim. Expected payments are rounded to cents and factors to four decimal places. As a check, both higher-limit factors exceed 1, and the $\$40{,}000$ factor is larger because it has the larger limit.

What the Exponential Mean Implies for Limit Choices

The mean $\theta$ sets the scale of the severity distribution. For a fixed limit, increasing $\theta$ makes that limit smaller relative to the typical claim size and generally increases the share of expected payment affected by raising the limit. The dimensionless ratios $u/\theta$ and $b/\theta$, rather than the dollar amounts alone, determine the ILF.

A useful special case occurs when the higher limit is twice the basic limit. Put $q=e^{-b/\theta}$. Then $e^{-2b/\theta}=q^2$, so the factor simplifies:

\operatorname{ILF}(2b;b) =\frac{1-q^2}{1-q} =\frac{(1-q)(1+q)}{1-q} =1+e^{-b/\theta}.

This is a direct algebraic check on calculations such as the first worked example. It also shows that the factor is less than 2 for any positive $b/\theta$, since $0<e^{-b/\theta}<1$.

Worked Example: Comparing Alternative Limit Choices by Incremental Cost

A hypothetical insurer models claim severity as exponential with mean $\$40{,}000$. The basic per-claim limit is $\$15{,}000$; the two alternatives are $\$30{,}000$ and $\$50{,}000$. Calculate each ILF relative to the basic limit and the additional expected payment per claim when choosing $\$50{,}000$ rather than $\$30{,}000$. Assume no deductible and no discounting.

Method. Compute the capped expected payment at all three limits. Divide by the basic payment for each ILF, and subtract the two alternative-limit payments for the incremental cost. The common mistake to avoid is subtracting the ILFs and interpreting the resulting factor difference as a dollar amount.

At the basic limit, $15{,}000/40{,}000=0.375$, and $e^{-0.375}=0.687289278$. Thus,

E[\min(X,15{,}000)] =40{,}000(1-0.687289278) =40{,}000(0.312710722) =\$12{,}508.43.

At $\$30{,}000$, the exponent is $0.75$, and $e^{-0.75}=0.472366553$:

E[\min(X,30{,}000)] =40{,}000(1-0.472366553) =40{,}000(0.527633447) =\$21{,}105.34.

The ILF is

\operatorname{ILF}(30{,}000;15{,}000) =\frac{0.527633447}{0.312710722} =1.687289\ldots \approx 1.6873.

At $\$50{,}000$, the exponent is $1.25$, and $e^{-1.25}=0.286504797$:

E[\min(X,50{,}000)] =40{,}000(1-0.286504797) =40{,}000(0.713495203) =\$28{,}539.81.

The ILF is

\operatorname{ILF}(50{,}000;15{,}000) =\frac{0.713495203}{0.312710722} =2.281646\ldots \approx 2.2816.

The incremental expected payment per claim for the $\$50{,}000$ option over the $\$30{,}000$ option is $\$28{,}539.81-\$21{,}105.34=\$7{,}434.47$. This subtraction is in dollars per claim; subtracting the factors, $2.2816-1.6873=0.5943$, would instead describe a relative difference in basic-limit units, not a dollar cost. Expected payments are rounded to cents and factors to four decimal places. The larger limit has both the larger expected payment and the larger factor, as expected.

Assumptions, Uses, and Common Traps

An exponential ILF is a model-based estimate. It assumes the exponential distribution adequately represents the relevant claim severities and that the same severity model applies at the basic and higher limits. The formula is not suitable without adjustment if the contract has a deductible, a coinsurance feature, a shared aggregate limit, or a different payment rule. It also does not automatically account for claim expenses, changes in claim reporting or settlement behavior, or limits expressed on a basis different from the modeled severity.

An ILF multiplies the expected capped loss per claim at the basic limit to obtain expected capped loss per claim at the higher limit. If claim frequency is unchanged, it can also be used to scale the corresponding expected loss cost per exposure. If frequency or exposure mix changes, the factor alone is not a complete premium comparison. In practice, actuaries also assess the reasonableness of the fitted severity model and any adjustments needed for the data and coverage being rated.

Using the limit ratio as the ILF. A higher limit twice as large does not mean twice the expected payment. Use the ratio of capped expected payments.
Using the CDF as the expected payment. $1-e^{-u/\theta}$ is the probability that severity is at or below $u$, not the dollar expected payment. Multiply it by $\theta$ to get $E[\min(X,u)]$.
Changing units inconsistently. If limits are in dollars, the mean must also be in dollars. The exponent $u/\theta$ must be dimensionless.
Reversing the ratio. The higher-limit expected payment belongs in the numerator and the basic-limit expected payment in the denominator. For $u>b$, the factor should exceed 1.
Confusing factor differences with dollar differences. Subtract expected payments to get an incremental dollar amount per claim; subtract factors only to compare the changes in basic-limit units.

The central result is simple: under exponential severity, a finite limit $u$ produces expected payment $\theta(1-e^{-u/\theta})$, and an ILF is the ratio of this quantity at the higher and basic limits. Careful interpretation of the payment function and consistent units are essential to using that result correctly.

Check Your Understanding

For each hypothetical problem, assume exponential severity, no deductible, and no discounting unless the question states otherwise. Round expected payments to cents and factors to four decimal places.

A portfolio has exponential severity with mean $\$20{,}000$. Find the ILF from a basic limit of $\$5{,}000$ to a limit of $\$15{,}000$. (Answer: 2.3853)
A hypothetical product has exponential severity with mean $\$50{,}000$. Calculate the expected payment per claim at a $\$25{,}000$ limit. (Answer: $\$19{,}673.47$)
For exponential severity with mean $\$12{,}000$, calculate the ILFs from a $\$6{,}000$ basic limit to limits of $\$12{,}000$ and $\$24{,}000$. (Answer: 1.6065 and 2.1975, respectively)
A hypothetical insurer has exponential severity with mean $\$30{,}000$ and a $\$10{,}000$ basic limit. Find the ILF for a $\$30{,}000$ limit and the expected payment per claim at that limit. (Answer: ILF $2.2299$; expected payment $\$18{,}963.62$)

What You'll Learn