Pareto Severity Deductible Calculations

Why Pareto Severity Matters for Deductibles

A deductible removes small losses from the insurer’s payment, but the remaining loss cost depends on how much probability lies in the severity distribution’s upper tail. A Pareto model is useful for examining this relationship because it assigns relatively substantial probability to large losses. The model can therefore show why two portfolios with similar average severities may have quite different expected payments above a high deductible.

This tutorial considers an ordinary per-loss deductible $d$, with insurer payment $Y^L=(X-d)_+=\max(X-d,0)$, where $X$ is ground-up loss. The quantity of interest is $E[Y^L]$: the expected insurer payment per ground-up loss, including losses for which the insurer pays zero. The calculations assume a fixed severity distribution and do not model changes in claim frequency, policyholder behavior, expenses, or reinsurance.

Definition: A Pareto Type I ground-up severity distribution with scale $\theta>0$ and shape $\alpha>0$ has survival function $S_X(x)=P(X>x)=1$ for $0\le x<\theta$, and $S_X(x)=(\theta/x)^\alpha$ for $x\ge\theta$. Its distribution is supported on $X\ge\theta$. For $\alpha>1$, its mean is $E[X]=\alpha\theta/(\alpha-1)$. The parameter $\theta$ sets the minimum possible severity; the shape parameter $\alpha$ controls the tail, with smaller $\alpha$ giving a heavier tail.

The shape condition matters. When $\alpha\le1$, the Pareto distribution has an infinite mean. With an unlimited upper tail and any finite deductible, the expected payment above that deductible is also infinite. The finite expected-payment formulas below therefore require $\alpha>1$.

Expected Payment Above a Deductible

For $d\ge\theta$, the deductible lies in the Pareto tail. A loss exceeds $d$ with probability $S_X(d)=(\theta/d)^\alpha$. Conditional on exceeding $d$, the expected amount by which the loss exceeds the deductible is $d/(\alpha-1)$. Multiplying these two quantities gives the unconditional expected payment per loss:

E[Y^L] = P(X>d)E[X-d\mid X>d] = \left(\frac{\theta}{d}\right)^\alpha\frac{d}{\alpha-1} = \frac{\theta^\alpha d^{1-\alpha}}{\alpha-1}, \qquad d\ge\theta,\ \alpha>1.

The conditional amount $E[X-d\mid X>d]$ is not itself the loss cost per ground-up loss: it applies only to losses that cross the deductible. The exceedance probability must also be included. The conditional result follows from the Pareto survival function:

P(X-d>y\mid X>d) = \frac{S_X(d+y)}{S_X(d)} = \left(\frac{d}{d+y}\right)^\alpha,\qquad y\ge0.

Integrating this conditional survival function over $y\ge0$ gives $E[X-d\mid X>d]=\int_0^\infty (d/(d+y))^\alpha\,dy=d/(\alpha-1)$, which is finite exactly when $\alpha>1$. This justifies the conditional excess amount used in the deductible formula.

If $0\le d<\theta$, every possible loss exceeds the deductible. The insurer payment is then $X-d$ for every loss, so the expected payment is the mean severity less the deductible. Combining the two cases gives:

Result: For a Pareto Type I severity with $\alpha>1$ and an ordinary deductible $d\ge0$,

E[Y^L] = \begin{cases} \dfrac{\alpha\theta}{\alpha-1}-d, & 0\le d<\theta,\\[6pt] \dfrac{\theta^\alpha d^{1-\alpha}}{\alpha-1}, & d\ge\theta. \end{cases}

At $d=\theta$, both expressions equal $\theta/(\alpha-1)$. If $\alpha\le1$, the expected payment is infinite for any finite $d$ under this unlimited-tail model.

Check the model and parameters.
Confirm that severity is Pareto Type I, identify $\theta$ and $\alpha$, and check whether $\alpha>1$.

Compare the deductible with the scale.
Use the mean-minus-deductible expression when $d<\theta$; use the tail expression when $d\ge\theta$.

Keep conditional and unconditional quantities separate.
For a deductible in the tail, multiply the probability of exceeding $d$ by the mean excess conditional on exceeding $d$.

Interpret the result as a loss cost per ground-up loss.
It is not a conditional payment and is not yet a premium rate; frequency, expenses, and other rating provisions are outside this severity calculation.

Worked Calculations

Worked Example: Expected Payment at a High Deductible

A hypothetical equipment insurer models ground-up loss severity as Pareto Type I with $\theta=\$1{,}000$ and $\alpha=3$. The ordinary deductible is $d=\$2{,}000$. Calculate the expected insurer payment per ground-up loss. No interest or mortality assumption is involved; this is a severity calculation under the stated Pareto distribution.

Method and condition. Since $d\ge\theta$ and $\alpha=3>1$, use the Pareto tail formula. First calculate the probability of exceeding the deductible:

P(X>d)=\left(\frac{1{,}000}{2{,}000}\right)^3 =\left(\frac12\right)^3 =\frac18.

The expected payment conditional on exceeding the deductible is:

E[X-d\mid X>d] = \frac{d}{\alpha-1} = \frac{2{,}000}{2} = \$1{,}000.

Multiply the conditional amount by the exceedance probability:

E[Y^L] = \frac18(\$1{,}000) = \$125.00.

Check. Direct substitution gives $1{,}000^3(2{,}000)^{-2}/2=\$125.00$. The expected ground-up severity is $3(1{,}000)/2=\$1{,}500$, so a deductible payment of $\$125.00$ is plausible: most losses do not cross the deductible, and the payment is only the excess for those that do. The common mistake is to report the conditional $\$1{,}000$ as the loss cost per loss; multiplying by $1/8$ is essential.

Worked Example: Comparing Tail Parameters at the Same Mean

Three hypothetical portfolios have Pareto Type I severity distributions with the same mean, $\$2{,}400$, but different shape parameters: $\alpha=2$, $3$, and $4$. For each, calculate the expected insurer payment above an ordinary deductible of $\$4{,}800$. Assume no interest or mortality adjustment; each portfolio uses its stated Pareto severity model.

Step 1: Find each scale parameter. From $E[X]=\alpha\theta/(\alpha-1)=\$2{,}400$, rearrange to $\theta=\$2{,}400(\alpha-1)/\alpha$.

\begin{array}{c|c|c} \alpha & \theta & \text{Check of mean}\\ 2 & \$1{,}200 & 2(1{,}200)/(2-1)=\$2{,}400\\ 3 & \$1{,}600 & 3(1{,}600)/(3-1)=\$2{,}400\\ 4 & \$1{,}800 & 4(1{,}800)/(4-1)=\$2{,}400 \end{array}

For all three portfolios, $d=\$4{,}800$ is at least as large as $\theta$, and all shape parameters exceed one. Apply the tail formula separately:

\begin{aligned} \alpha=2:\quad E[Y^L]&=\frac{1{,}200^2}{4{,}800} =\$300.00,\\ \alpha=3:\quad E[Y^L]&=\frac{1{,}600^3}{2(4{,}800)^2} =\frac{4{,}096{,}000{,}000}{46{,}080{,}000} =\$88.89,\\ \alpha=4:\quad E[Y^L]&=\frac{1{,}800^4}{3(4{,}800)^3} =\frac{10{,}497{,}600{,}000{,}000}{331{,}776{,}000{,}000} =\$31.64. \end{aligned}

Check and interpretation. Each distribution has the same $\$2{,}400$ mean, yet the expected payment above $\$4{,}800$ falls as $\alpha$ rises in this comparison. The smaller shape parameter represents the heavier tail. The calculation uses the stated scale for each shape; holding the mean fixed requires changing $\theta$. A common mistake is to change $\alpha$ while leaving $\theta$ unchanged and then describe the comparison as if the means were equal.

Worked Example: Comparing Two Deductibles

A hypothetical specialty insurer uses a Pareto Type I severity model with $\theta=\$1{,}000$ and $\alpha=3$. Compare expected insurer payments per ground-up loss at deductibles of $\$1{,}000$ and $\$2{,}000$. State the ratio of the higher-deductible payment to the lower-deductible payment. No interest or mortality assumption enters this model.

At $d=\$1{,}000=\theta$, either branch of the formula applies:

E[Y^L\mid d=1{,}000] = \frac{1{,}000}{3-1} = \$500.00.

At $d=\$2{,}000$, use the tail formula:

E[Y^L\mid d=2{,}000] = \frac{1{,}000^3}{2(2{,}000)^2} = \frac{1{,}000{,}000{,}000}{8{,}000{,}000} = \$125.00.

The ratio is $125/500=0.25$. Thus, under this fixed hypothetical severity model, expected payment at the $\$2{,}000$ deductible is one-quarter of expected payment at the $\$1{,}000$ deductible. As a check, the first deductible has exceedance probability $1$ and conditional excess $\$500$; the second has exceedance probability $1/8$ and conditional excess $\$1{,}000$, giving $1(500)=\$500$ and $(1/8)(1{,}000)=\$125$. The ratio compares severity loss costs only; it is not by itself a premium indication.

Worked Example: Deductible Below the Minimum Severity

A hypothetical insurer models ground-up severity as Pareto Type I with $\theta=\$1{,}000$ and $\alpha=3$, and applies an ordinary deductible of $\$400$. Calculate the expected insurer payment per loss. There is no interest or mortality adjustment.

Because $d=\$400<\theta=\$1{,}000$, every loss exceeds the deductible. First calculate the mean severity:

E[X]=\frac{3(1{,}000)}{3-1}=\$1{,}500.

Subtract the deductible from every loss:

E[Y^L]=E[X-d]=E[X]-d=\$1{,}500-\$400=\$1{,}100.00.

Check. The result is positive and is $\$400$ less than the ground-up mean, as it must be when every loss is reduced by $\$400$. The common mistake is to use the tail formula for $d\ge\theta$, even though the deductible is below the support’s minimum. Choosing the branch based on $d$ relative to $\theta$ avoids that error.

Interpreting Tail Effects and Avoiding Errors

For $d\ge\theta$, the formula $E[Y^L]=\theta^\alpha d^{1-\alpha}/(\alpha-1)$ makes the role of the tail visible. The expected payment depends both on the chance that a loss crosses the deductible and on the average amount above it. A deductible can sharply reduce the probability of an insurer payment, while the heavy tail can still leave substantial conditional payments among losses that do cross it.

Comparisons across shape parameters need care. If $\theta$ is held fixed, changing $\alpha$ also changes the mean severity. If the goal is to compare distributions at a fixed mean, first solve $\theta=\mu(\alpha-1)/\alpha$, where $\mu$ is the common mean, as in the second worked example. Even then, a selected loss cost is conditional on the assumed model and portfolio; it does not establish a final rate without considering the other relevant rating provisions.

This Pareto severity treatment is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus. In an exam calculation, identify the Pareto parameterization before substituting values: other Pareto conventions may use a different scale or parameter name, but the survival function stated here fixes the meaning of $\theta$ and $\alpha$.

Common exam trap: The tail formula applies only when $d\ge\theta$, and its result is unconditional per loss. For $d\ge\theta$, calculate $P(X>d)=(\theta/d)^\alpha$ and $E[X-d\mid X>d]=d/(\alpha-1)$, then multiply. Reporting only the conditional excess overstates the per-loss cost. For $d<\theta$, every loss exceeds the deductible, so use $E[X]-d$. Also check $\alpha>1$ before reporting a finite expected payment.

Check Your Understanding

For each hypothetical severity model, state the expected insurer payment per ground-up loss. Round dollar answers to the nearest cent.

A Pareto Type I model has $\theta=\$800$, $\alpha=2$, and ordinary deductible $d=\$1{,}600$. What is the expected insurer payment? (Answer: $\$400.00$)
A Pareto Type I model has $\theta=\$1{,}200$, $\alpha=3$, and $d=\$2{,}400$. Find the exceedance probability, the conditional expected payment given exceedance, and the unconditional expected payment. (Answer: $0.1250$, $\$1{,}200.00$, and $\$150.00$)
A Pareto Type I model has mean severity $\$1{,}800$, shape $\alpha=2$, and deductible $d=\$3{,}600$. Find $\theta$ first, then calculate expected insurer payment. (Answer: $\theta=\$900$; expected payment $=\$225.00$)
A Pareto Type I model has $\theta=\$500$, $\alpha=4$, and ordinary deductible $d=\$250$. Which formula branch applies, and what is the expected insurer payment? (Answer: the $d<\theta$ branch; $\$416.67$)
Two Pareto Type I models have the same mean severity of $\$1{,}200$, with shape parameters $\alpha=2$ and $\alpha=3$. At deductible $d=\$3{,}600$, calculate each expected insurer payment and identify which is larger. (Answer: $\alpha=2$: $\$100.00$; $\alpha=3$: $\$19.75$; the $\alpha=2$ model is larger.)

What You'll Learn