Coverage Modification Boundary Conventions

Why Boundary Wording Matters

Coverage formulas often use phrases such as “at least the deductible,” “exceeds the attachment,” or “reaches the limit.” For a loss distribution with no probability concentrated at the stated boundary, strict and inclusive versions of these events have the same probability. If losses can equal the boundary with positive probability, they do not. The payment formula itself also matters: a loss equal to a deductible usually produces zero excess payment, even when it meets an inclusive threshold condition.

The key is to distinguish three questions: what event does the contract describe, how much is paid at the boundary, and how much probability lies exactly at that boundary? Keeping these questions separate prevents a wording choice from being mistaken for a change in the payment function.

Definition: For a loss $X$, the distribution function is $F_X(x)=P(X\le x)$. Its left limit is $F_X(x-)=P(X<x)$, and the probability of a loss exactly equal to $x$ is the atom $P(X=x)=F_X(x)-F_X(x-)$. Therefore, $P(X\ge x)=1-F_X(x-)$, whereas $P(X>x)=1-F_X(x)$. The inclusive and strict probabilities differ by exactly $P(X=x)$.

The notation $S_X(x)=1-F_X(x)$ uses the strict exceedance probability $P(X>x)$. In particular, $S_X(x)$ is not generally $P(X\ge x)$ when $X$ has an atom at $x$. If $X$ has a continuous distribution at $x$, then $P(X=x)=0$, so the two probabilities agree.

Boundary Events and Payment Functions

For a deductible $d$, the standard excess payment before any payment limit is $Y^L=(X-d)_+=\max(X-d,0)$. A policy with payment limit $u$ has $Y^L=\min((X-d)_+,u)$. Under this usual payment function, a loss equal to $d$ makes no positive payment: $(d-d)_+=0$. Thus the event $X\ge d$ includes losses at the deductible, but the event of a positive payment is $X>d$.

For an excess layer attaching at $a$ with layer limit $L$, the usual payment per loss is $Y^L=\min((X-a)_+,L)$. At $X=a$, the layer is attached but pays zero. The layer makes a positive payment when $X>a$, and it pays its full limit when $X\ge a+L$. A loss exactly equal to $a+L$ is enough to exhaust the layer; “exceeds the exhaustion point” would instead mean $X>a+L$.

\begin{aligned} P(X\ge x)&=1-F_X(x-), & P(X>x)&=1-F_X(x),\\ P(X\le x)&=F_X(x), & P(X<x)&=F_X(x-),\\ P(X=x)&=F_X(x)-F_X(x-). \end{aligned}

These identities follow directly by partitioning the event $\{X\ge x\}$ into the disjoint events $\{X=x\}$ and $\{X>x\}$. Since $P(X<x)=F_X(x-)$, its complement is $\{X\ge x\}$, giving $P(X\ge x)=1-F_X(x-)$. The event $\{X>x\}$ is the complement of $\{X\le x\}$, giving $P(X>x)=1-F_X(x)$. Subtracting the two expressions shows that the difference between the inclusive and strict probabilities is the boundary atom.

Key distinction: A contract’s event wording and its payment function are related but not interchangeable. At a deductible, a loss may satisfy “at least the deductible” while producing zero excess payment. At a payment-limit exhaustion point, a loss equal to the threshold does receive the full limit, even though it does not exceed that threshold.

Applying the Conventions Step by Step

A reliable calculation starts by identifying the exact threshold and the event being requested. Only then should the relevant probability or payment be calculated. The following sequence makes the boundary explicit before arithmetic begins.

Write the threshold event.
Translate “at least $x$” to $X\ge x$, and “greater than $x$” or “exceeds $x$” to $X>x$.

Check for an atom.
Find $P(X=x)=F_X(x)-F_X(x-)$. If it is positive, the two event probabilities differ.

Evaluate the payment at equality.
Substitute $X=x$ into the payment formula. Do not infer the payment from the event wording alone.

Calculate and check.
Use the correct strict or inclusive probability, then verify that the payment and event interpretation agree at the boundary.

Worked Example: A Deductible and a Payment Limit with Discrete Losses

A hypothetical contract pays the amount above a deductible $d=\$1{,}000$, subject to a payment limit $u=\$500$. A loss $X$ has the following distribution: $P(X=\$0)=0.20$, $P(X=\$500)=0.15$, $P(X=\$1{,}000)=0.25$, $P(X=\$1{,}500)=0.25$, and $P(X=\$2{,}000)=0.15$. Find the probability a loss is at least the deductible, the probability it exceeds the deductible, the probability the insurer makes a positive payment, and the expected payment. Also compare reaching and exceeding the payment-limit threshold.

Method and boundary check. The payment per loss is $Y^L=\min((X-\$1{,}000)_+,\$500)$. The deductible event is distinct from a positive payment: equality at $\$1{,}000$ is included in “at least,” but gives zero payment.

At least the deductible. The losses at $\$1{,}000$, $\$1{,}500$, and $\$2{,}000$ qualify. Thus $P(X\ge \$1{,}000)=0.25+0.25+0.15=0.65$.

Exceeds the deductible and positive payment. Only losses at $\$1{,}500$ and $\$2{,}000$ exceed the deductible. Therefore $P(X>\$1{,}000)=0.25+0.15=0.40$. Both produce positive payments, so $P(Y^L>0)=0.40$. The difference between the two deductible probabilities is $0.65-0.40=0.25=P(X=\$1{,}000)$.

Expected payment. The payments at losses $\$0,\$500,\$1{,}000,\$1{,}500,\$2{,}000$ are respectively $\$0,\$0,\$0,\$500,\$500$. Hence $E[Y^L]=0.25(\$500)+0.15(\$500)=\$125+\$75=\$200$.

Reaching the payment limit. The payment limit is reached when $X\ge d+u=\$1{,}500$, giving probability $0.25+0.15=0.40$. The loss exceeds that threshold only at $\$2{,}000$, so $P(X>\$1{,}500)=0.15$. The atom at $\$1{,}500$ reaches the limit exactly. As a check, the probabilities in the loss distribution sum to $0.20+0.15+0.25+0.25+0.15=1.00$, and the expected payment is between $\$0$ and the $\$500$ maximum.

Common mistake avoided: Using $P(X\ge d)$ as the probability of a positive payment would give $0.65$, which incorrectly counts the $\$1{,}000$ losses as paid claims. The correct positive-payment probability is $P(X>d)=0.40$.

Worked Example: Attachment and Layer Exhaustion

A hypothetical excess layer attaches at $a=\$1{,}200$ and has limit $L=\$1{,}000$. Losses have probabilities $0.10$ at $\$1{,}000$, $0.20$ at $\$1{,}200$, $0.30$ at $\$1{,}600$, $0.25$ at $\$2{,}200$, and $0.15$ at $\$2{,}500$. Find the probabilities that a loss is at least the attachment, strictly above the attachment, generates a positive layer payment, reaches the full limit, and exceeds the exhaustion point. Calculate the expected layer payment.

Method. The payment is $Y^L=\min((X-\$1{,}200)_+,\$1{,}000)$. The attachment boundary is $\$1{,}200$, and the exhaustion point is $a+L=\$2{,}200$.

Attachment probabilities. At least the attachment includes losses of $\$1{,}200$ and above, so $P(X\ge \$1{,}200)=0.20+0.30+0.25+0.15=0.90$. Strictly above the attachment excludes the $\$1{,}200$ atom, so $P(X>\$1{,}200)=0.30+0.25+0.15=0.70$. Since a loss exactly at attachment pays zero, the positive-payment probability is also $0.70$.

Exhaustion probabilities. Losses of $\$2{,}200$ and $\$2{,}500$ receive the full $\$1{,}000$ layer payment. Thus the probability of reaching the limit is $P(X\ge \$2{,}200)=0.25+0.15=0.40$. The probability of exceeding the exhaustion point is $P(X>\$2{,}200)=0.15$. Their difference, $0.25$, is exactly the atom at $\$2{,}200$.

Expected payment. The payment is $\$0$ at losses $\$1{,}000$ and $\$1{,}200$, $\$400$ at $\$1{,}600$, and $\$1{,}000$ at each of $\$2{,}200$ and $\$2{,}500$. Therefore $E[Y^L]=0.30(\$400)+0.25(\$1{,}000)+0.15(\$1{,}000)=\$120+\$250+\$150=\$520$. The distribution probabilities sum to $1.00$, and the expected payment lies between zero and the layer limit.

Common mistake avoided: The probability of a full layer payment is not $P(X>\$2{,}200)$. A loss equal to $\$2{,}200$ already produces exactly $\$1{,}000$, so the correct event is $X\ge \$2{,}200$.

Worked Example: A Mixed Distribution with an Atom at the Deductible

A hypothetical loss has a 60% chance of being drawn from a uniform distribution on $[0,\$2{,}000]$ and a 40% chance of being exactly $\$500$. The contract has deductible $d=\$500$ and payment limit $u=\$1{,}000$. Find $P(X\ge \$500)$, $P(X>\$500)$, the probability of positive payment, the probability of reaching the limit, and the expected payment.

Method. The point mass at $\$500$ is $0.40$. The continuous component has no probability at any single point. The payment is $Y^L=\min((X-\$500)_+,\$1{,}000)$, and its limit is reached at $X=\$1{,}500$.

Inclusive and strict deductible events. For the uniform component, the fraction at or above $\$500$ is $(\$2{,}000-\$500)/\$2{,}000=0.75$. Including the point mass gives $P(X\ge \$500)=0.60(0.75)+0.40=0.45+0.40=0.85$. The atom at $\$500$ does not exceed the deductible, so $P(X>\$500)=0.60(0.75)=0.45$. Positive payment also requires $X>\$500$, giving probability $0.45$.

Limit probability. In the uniform component, the fraction at or above $\$1{,}500$ is $(\$2{,}000-\$1{,}500)/\$2{,}000=0.25$. There is no atom at that threshold, so the probability of reaching the limit is $0.60(0.25)=0.15$, the same as the probability of exceeding it.

Expected payment. Conditional on the uniform component, payments are zero from $\$0$ to $\$500$, increase from zero to $\$1{,}000$ between $\$500$ and $\$1{,}500$, and equal $\$1{,}000$ between $\$1{,}500$ and $\$2{,}000$. The integral over the increasing segment is a triangle of area $\frac12(\$1{,}000)(\$1{,}000)=\$500{,}000$; the capped segment has area $(\$500)(\$1{,}000)=\$500{,}000$. Dividing their total area by the uniform interval length, $\$2{,}000$, gives a conditional expected payment of $\$500$. The point mass at $\$500$ pays zero. Thus $E[Y^L]=0.60(\$500)+0.40(\$0)=\$300$. The result is below the $\$1{,}000$ maximum, as required.

Common mistake avoided: Treating the distribution as continuous would omit the 0.40 atom at the deductible and report $P(X\ge \$500)=0.45$. The correct inclusive probability is $0.85$, while the positive-payment probability remains $0.45$.

Common Boundary Errors

Boundary errors usually arise from applying a familiar survival probability without checking its convention, or from assuming that satisfying a threshold means a positive payment. In calculations involving discrete or mixed severity models, write the event in words and symbols before substituting a distribution value.

Using the wrong tail convention. $S_X(x)=P(X>x)$, not $P(X\ge x)$. For an inclusive event, use $1-F_X(x-)$.
Counting deductible equality as a positive payment. Under the usual excess-payment formula, $Y^L=0$ when $X=d$, even though $X\ge d$.
Excluding equality at exhaustion. A payment cap is reached at its threshold. For a deductible $d$ and payment limit $u$, the full limit is paid when $X\ge d+u$.
Assuming all threshold distinctions disappear. Strict and inclusive probabilities agree only if there is no atom at the threshold.

These distinctions matter when a model has fixed claim amounts, rounded observations, minimum charges, or other reasons for probability to concentrate at particular loss values. They also matter when a problem asks for the probability of an event rather than an expected payment: two threshold conventions can produce different event probabilities even when the payment formula at equality is unambiguous.

Check Your Understanding

For each problem, identify the boundary event before calculating. Amounts are hypothetical.

A loss is $\$0$ with probability $0.10$, $\$750$ with probability $0.20$, $\$1{,}000$ with probability $0.30$, and $\$1{,}750$ with probability $0.40$. With deductible $d=\$750$, find $P(X\ge d)$, $P(X>d)$, and the probability of a positive excess payment. (Answer: $0.90,\ 0.70,\ 0.70$, respectively.)
An excess layer attaches at $\$2{,}000$ with limit $\$1{,}500$. Loss probabilities are $0.20$ at $\$1{,}500$, $0.25$ at $\$2{,}000$, $0.35$ at $\$2{,}600$, and $0.20$ at $\$3{,}500$. Find the probabilities of a positive layer payment, reaching the full limit, and exceeding the exhaustion point. (Answer: $0.55,\ 0.20,\ 0.00$.)
A loss is drawn with equal probability from a uniform distribution on $[0,\$2{,}000]$ or is exactly $\$1{,}000$. Find $P(X\ge \$1{,}000)$ and $P(X>\$1{,}000)$. (Answer: $0.75,\ 0.25$.)
A loss is $\$0$, $\$400$, or $\$1{,}000$ with probabilities $0.20,\ 0.30,$ and $0.50$. A policy has deductible $\$400$ and payment limit $\$600$. Find the probability of reaching the limit and the expected payment. (Answer: $0.50$ and $\$300$.)
A discrete loss has probability $0.15$ at $\$900$, $0.35$ at $\$1{,}200$, and $0.50$ at $\$1{,}800$. For a layer attaching at $\$900$ with limit $\$900$, find the probability of a positive payment and the probability the layer pays its full limit. (Answer: $0.85$ and $0.50$.)

What You'll Learn