Exponential Severity ILF Calculation

Why Exponential Severity Is Useful for ILFs

An increased limits factor (ILF) compares expected claim payments under two per-claim limits for the same underlying severity distribution. The exponential distribution provides a compact model for this comparison: its survival function has a simple form, so limited expected losses can be calculated directly at any stated limit. The result is useful for understanding how expected payments grow as a policy limit rises, although the model’s assumptions must be checked before using it for a pricing indication.

This tutorial treats $X$ as the ground-up severity of one claim and assumes no deductible. A per-claim limit $u$ means the insurer’s payment on that claim is $\min(X,u)$. All monetary amounts in the worked examples are in thousands of dollars unless otherwise stated.

Definition: A nonnegative loss $X$ has an exponential severity distribution with mean $\theta>0$ if its density and survival function are $$ f_X(x)=\frac{1}{\theta}e^{-x/\theta},\quad x\ge 0, \qquad S_X(x)=P(X>x)=e^{-x/\theta},\quad x\ge 0. $$ The parameter $\theta$ is both the mean and the scale parameter: $E[X]=\theta$. Under a per-claim limit $u\ge 0$, the payment is $Y^L=\min(X,u)$.

The exponential model is memoryless: for $s,t\ge 0$, $P(X>s+t\mid X>s)=P(X>t)$. This property is a characteristic of the distribution, not a general property of insurance severities. The model assumes a continuous, nonnegative severity distribution with a constant hazard rate $1/\theta$. It may be unsuitable when observed severity has a changing hazard, a deductible or other coverage terms not represented in the model, or a tail shape that the exponential survival function does not capture.

Limited Expected Loss from the Survival Function

For a nonnegative loss, a useful identity is that the expectation of a capped payment equals the area under the loss’s survival function up to the cap. For this model, the identity gives an especially simple expression for $E[Y^L]$.

Result: If $X$ is exponential with mean $\theta$ and the per-claim limit is $u$, then $$ E[\min(X,u)] =\int_0^u S_X(x)\,dx =\theta\left(1-e^{-u/\theta}\right). $$

To justify the survival-function identity, write any nonnegative payment $Y$ as the area of its indicator functions: $Y=\int_0^\infty \mathbf{1}_{\{Y>x\}}\,dx$. Taking expectations gives $E[Y]=\int_0^\infty P(Y>x)\,dx$. For $Y=\min(X,u)$, the event $Y>x$ is the same as $X>x$ for $0\le x<u$, and is impossible for $x\ge u$. Therefore $E[Y]=\int_0^u S_X(x)\,dx$. Substituting the exponential survival function and integrating,

\int_0^u e^{-x/\theta}\,dx =\left[-\theta e^{-x/\theta}\right]_0^u =\theta(1-e^{-u/\theta}).

The formula has sensible boundary behavior. At $u=0$, expected payment is zero. As $u$ increases without bound, $e^{-u/\theta}$ approaches zero and limited expected loss approaches the ground-up mean $\theta$. For any finite positive limit, it is less than $\theta$.

Calculating and Comparing ILFs

An ILF is a ratio of expected losses on the same basis. If the base limit is $u_b>0$ and the higher limit is $u_h\ge u_b$, the exponential severity ILF is

\operatorname{ILF}(u_h;u_b) =\frac{E[\min(X,u_h)]}{E[\min(X,u_b)]} =\frac{1-e^{-u_h/\theta}}{1-e^{-u_b/\theta}}.

The factor $\theta$ cancels from the ratio’s numerator and denominator, but the ILF still depends on $\theta$ through the limit-to-mean ratios $u_h/\theta$ and $u_b/\theta$. Thus, an ILF is not determined by the two limits alone. The severity assumption and its parameter matter.

If the same claim frequency applies at both limits, the frequency also cancels. For example, suppose the number of claims per exposure has mean $\lambda$, and claim severities are independent of claim count and have the same exponential distribution at each limit. The expected aggregate payment per exposure at limit $u$ is $\lambda E[\min(X,u)]$; consequently, the ratio of expected aggregate payments is the same ILF. This reasoning concerns expected losses only; other pricing components are not included in the ratio.

Worked Example: Comparing Two Limits

A hypothetical liability portfolio is modeled with exponential claim severity having mean $\theta=40$ thousand dollars. Calculate limited expected losses at limits of 20 and 80 thousand dollars, then find the ILF for the 80-thousand-dollar limit relative to the 20-thousand-dollar base limit.

Method and assumptions. Use $E[\min(X,u)]=\theta(1-e^{-u/\theta})$. This is a per-claim calculation for ground-up losses, with no deductible. Keep full precision until the final rounding; report money to 0.001 thousand dollars and the ILF to four decimal places.

At the 20-thousand-dollar limit, $u/\theta=20/40=0.5$. Since $e^{-0.5}=0.60653066$,

E[\min(X,20)] =40(1-e^{-0.5}) =40(1-0.60653066) =40(0.39346934) =15.7387736.

At the 80-thousand-dollar limit, $u/\theta=80/40=2$. Since $e^{-2}=0.13533528$,

E[\min(X,80)] =40(1-e^{-2}) =40(1-0.13533528) =40(0.86466472) =34.5865888.

The higher-limit ILF relative to the 20-thousand-dollar base is

\operatorname{ILF}(80;20) =\frac{34.5865888}{15.7387736} =\frac{0.86466472}{0.39346934} \approx 2.1975.

Rounded results and check. The limited expected losses are 15.739 and 34.587 thousand dollars, and the ILF is 2.1975. The factor exceeds 1 because the higher limit cannot reduce a claim payment. The expected loss at either finite limit is below the ground-up mean of 40 thousand dollars, as required.

Common mistake avoided. Dividing the two limits gives $80/20=4$, but that is not the ILF. An ILF is a ratio of expected payments, not a ratio of policy limits.

Frequency and the Limited Expected Loss Basis

The severity ILF above uses a per-claim expected payment. A ratemaking calculation may instead express expected losses per exposure. Those quantities agree in ratio only when the frequency basis is consistent. In particular, both limits must use the same expected claim count and the same severity distribution. If the population, claim definition, or frequency differs between limit options, canceling frequency is not justified.

Worked Example: Checking That Frequency Cancels

A hypothetical equipment coverage has a Poisson claim count per policy with mean $\lambda=0.08$. Claim severity is independent of count and exponential with mean $\theta=30$ thousand dollars. Find the expected limited loss per policy at limits of 15 and 45 thousand dollars, and calculate the 45-thousand-dollar ILF relative to 15 thousand dollars.

Method and assumptions. For a Poisson count with mean $\lambda$, the expected sum of per-claim payments is $\lambda$ times the expected payment per claim, provided the severities are independent and identically distributed and independent of the count. First calculate each severity payment, then multiply by $0.08$.

At limit 15, $u/\theta=15/30=0.5$, so the expected payment per claim is

30(1-e^{-0.5}) =30(0.39346934) =11.8040802\text{ thousand dollars}.

The expected loss per policy is $0.08(11.8040802)=0.9443264$ thousand dollars, or $\$944.33$ when rounded to cents.

At limit 45, $u/\theta=45/30=1.5$, and $e^{-1.5}=0.22313016$. Thus the expected payment per claim is

30(1-e^{-1.5}) =30(1-0.22313016) =30(0.77686984) =23.3060952\text{ thousand dollars}.

The expected loss per policy is $0.08(23.3060952)=1.8644876$ thousand dollars, or $\$1,864.49$. The ILF is

\frac{1.8644876}{0.9443264} =\frac{23.3060952}{11.8040802} =\frac{0.77686984}{0.39346934} \approx 1.9744.

Rounded results and check. Expected losses per policy are $\$944.33$ and $\$1,864.49$, and the ILF is 1.9744. The calculation shows explicitly that the common frequency factor $0.08$ cancels. The higher limit produces a larger expected payment, consistent with the result.

Common mistake avoided. Multiplying the limit by claim frequency does not give expected loss. Frequency multiplies the expected limited severity, not the policy limit itself.

Three Limits on One Consistent Basis

Calculating several limits together is a useful check on an ILF schedule. Each limited expected loss should increase as the limit increases, while each ILF should be calculated against the same stated base. The limiting value of the expected loss is the ground-up mean; an ILF based on a finite base limit therefore approaches $\theta/E[\min(X,u_b)]$ as the higher limit grows without bound.

Worked Example: A Three-Limit Comparison

A hypothetical property portfolio is modeled with exponential severity of mean $\theta=50$ thousand dollars. Calculate the limited expected losses at limits of 25, 50, and 100 thousand dollars. Then calculate the 50- and 100-thousand-dollar ILFs relative to the 25-thousand-dollar base.

Method and assumptions. Apply the same ground-up severity model and no-deductible payment definition at all three limits. Round limited expected losses to 0.001 thousand dollars and ILFs to four decimal places, after computing ratios from unrounded amounts.

For the 25-thousand-dollar limit, $u/\theta=0.5$, so the expected payment is $50(1-e^{-0.5})=50(0.39346934)=19.6734670$ thousand dollars.

For the 50-thousand-dollar limit, $u/\theta=1$, and $e^{-1}=0.36787944$. The expected payment is $50(1-0.36787944)=50(0.63212056)=31.6060280$ thousand dollars.

For the 100-thousand-dollar limit, $u/\theta=2$, and $e^{-2}=0.13533528$. The expected payment is $50(1-0.13533528)=50(0.86466472)=43.2332360$ thousand dollars.

Relative to the 25-thousand-dollar base, the ILFs are

\operatorname{ILF}(50;25) =\frac{31.6060280}{19.6734670} =\frac{0.63212056}{0.39346934} \approx 1.6065,

\operatorname{ILF}(100;25) =\frac{43.2332360}{19.6734670} =\frac{0.86466472}{0.39346934} \approx 2.1975.

Rounded results and check. The limited expected losses are 19.673, 31.606, and 43.233 thousand dollars; the ILFs are 1.6065 and 2.1975. The losses increase with the limit and remain below the ground-up mean of 50 thousand dollars. The 100-thousand-dollar ILF agrees with the direct 100-to-25 comparison, confirming that the same base was used.

Common mistake avoided. Using 50 as the denominator for the 100-thousand-dollar ILF would calculate the 100-to-50 factor, not the requested 100-to-25 factor. Always identify the base limit before forming the ratio.

Interpretation and Exam Checks

An exponential ILF is a model-based comparison of limited expected severities. It does not say that every claim payment rises by the factor, nor that a higher limit changes the underlying claim frequency. It summarizes the ratio of expected per-claim payments under the specified limits and severity model. For an expected-loss indication per exposure, a common frequency factor can be included, but only if the frequency assumptions are the same at both limits.

Before finalizing a calculation, check the following:

Is $\theta$ the mean severity, expressed in the same monetary units as the limit?
Does the payment definition match the problem’s limit basis? Here it is ground-up payment $\min(X,u)$, without a deductible.
Is the ILF numerator the expected payment at the higher limit and the denominator the expected payment at the stated base limit?
Are monetary values rounded only after the expected losses and ratios are calculated?
Does each result satisfy $0\le E[\min(X,u)]\le\theta$, and does expected loss increase with $u$?

The exponential formula is aligned with the CAS Exam 8 (Advanced Ratemaking) syllabus as a severity-based ILF calculation. Its simplicity makes it a useful benchmark, but it should not be treated as a universal model for claim severity.

Check Your Understanding

For each problem, use the exponential limited expected loss formula and state the requested results with consistent units.

A hypothetical claim severity is exponential with mean $\$20,000$. Calculate limited expected loss at a $\$10,000$ per-claim limit, rounded to the nearest cent. (Answer: $\$7,869.39$)
A severity model has exponential mean $\$25,000$. Calculate the ILF for a $\$75,000$ limit relative to a $\$25,000$ base limit, rounded to four decimals. (Answer: $(1-e^{-3})/(1-e^{-1})\approx 1.5032$)
A hypothetical policy has Poisson claim frequency with mean $0.12$ per policy and independent exponential claim severity with mean $\$40,000$. At a $\$20,000$ limit, calculate the expected loss per policy to the nearest cent. (Answer: $\$1,888.65$)
For exponential severity with mean $\$30,000$, calculate the ILF for a $\$90,000$ limit relative to a $\$15,000$ base, rounded to four decimals. (Answer: $(1-e^{-3})/(1-e^{-0.5})\approx 2.4150$)
A hypothetical portfolio has exponential severity with mean $\$40,000$. Find the limited expected losses at $\$20,000$ and $\$80,000$, then the higher-limit ILF. Report losses to the nearest cent and the factor to four decimals. (Answer: $\$15,738.77$, $\$34,586.59$, and $2.1975$)

What You'll Learn