Introduction
A claim-count question often turns on two small decisions: how many exposures are being counted, and what does the wording require the count to be? Once those are clear, the exact binomial probability follows from a standard formula. The calculation is only valid, however, when the exposure units satisfy the binomial model’s assumptions.
This tutorial focuses on translating claim language into the binomial parameters and event, then calculating the resulting probability. It is aligned with the SOA Exam FAM (Fundamentals of Actuarial Mathematics) syllabus.
The Exact Binomial Probability
Suppose a portfolio contains a fixed number \(n\) of exposure units observed over the same period. For each unit, a claim either occurs or does not occur during that period. Let \(p\) be the probability of a claim for each unit, and let \(N\) be the total number of units with claims.
For a specified integer \(k\) between 0 and \(n\), the probability of exactly \(k\) claims is
where \(\binom{n}{k}=\dfrac{n!}{k!(n-k)!}\). The factors \(p^k\) and \(q^{n-k}\) give the probability of one particular arrangement of \(k\) claims and \(n-k\) nonclaims. The binomial coefficient accounts for all the different arrangements that produce the same total count.
Why the Formula Counts All the Arrangements
Represent the outcome for each exposure unit by an indicator \(I_j\): it is 1 if unit \(j\) has a claim and 0 otherwise. Then \(N=I_1+\cdots+I_n\). For any one specified arrangement with \(k\) claims, independence gives probability \(p^kq^{n-k}\). For example, when \(n=4\), the arrangement claim, no claim, claim, no claim has probability \(pqpq=p^2q^2\).
There are \(\binom{n}{k}\) ways to select which \(k\) of the \(n\) units have claims. These arrangements are mutually exclusive, and each has probability \(p^kq^{n-k}\). Adding their equal probabilities gives the formula. This reasoning also shows why the formula does not apply if claim probabilities differ across units or if their outcomes are dependent.
Identify which units are being counted and how many there are. A policy observed for one specified year is one trial, not one trial per month.
Confirm that each unit has the same claim probability and that the outcomes are independent. Then set \(q=1-p\).
“Exactly \(k\)” means one value of \(N\); “at least \(k\)” includes \(k,k+1,\ldots,n\); “no claims” means \(N=0\).
Use the binomial formula for each required count, add disjoint probabilities when needed, and confirm that the requested count is within \(0,\ldots,n\).
Worked Applications
Worked Example: Exactly Two Claims in a Small Portfolio
A hypothetical insurer observes five independent, comparable equipment policies during one calendar year. Each policy has probability \(0.20\) of generating at least one claim during that year. Find the probability that exactly two policies generate claims.
Method and translation. The trial is one policy’s claim/no-claim outcome for the year. Thus \(n=5\), \(p=0.20\), \(q=1-0.20=0.80\), and “exactly two” means \(k=2\). The assumptions in the problem allow the binomial model.
The number of ways to select two claim policies from five is \(\binom{5}{2}=\dfrac{5!}{2!3!}=10\). Therefore,
The probability is \(0.2048\), or \(20.48\%\). Check: the result is between 0 and 1, and includes the factor 10 for the possible pairs of claim policies. A common mistake is to report only \((0.20)^2(0.80)^3=0.02048\), which is the probability of one particular arrangement, not of exactly two claims in any arrangement.
Worked Example: Translating “At Least Three”
A hypothetical portfolio has four independent policies exposed for one year. Each has probability \(0.25\) of at least one claim during the year. Find the probability that at least three policies have claims.
Method and translation. Here \(n=4\), \(p=0.25\), and \(q=0.75\). “At least three” means \(N=3\) or \(N=4\), so calculate both exact probabilities and add them. The events are disjoint because the total count cannot be both 3 and 4.
Add the two probabilities:
Rounded to four decimal places, \(P(N\geq 3)=0.0508\). Check: the answer is small, as three or four claims among four policies is an uncommon event when each policy’s claim probability is \(0.25\). A common mistake is to calculate only \(P(N=3)\), leaving out the possibility of four claims.
Worked Example: A Claim Condition for Each of Two Groups
A hypothetical service plan covers seven independent devices for one year. Each device has probability \(0.20\) of a claim. Find the probability that at least one of the first three devices has a claim and none of the other four devices has a claim.
Method and translation. The requirement specifies separate conditions on two disjoint groups. For the first three devices, “at least one” is the complement of zero claims. For the remaining four, “none” means four nonclaims. The two group outcomes are independent because the device outcomes are independent.
For the first group, let \(N_1\) be its claim count. Then \(N_1\sim\operatorname{Bin}(3,0.20)\), and
For the other four devices, the probability of no claims is
Multiply the independent group probabilities:
Rounded to four decimal places, the probability is \(0.1999\). Check: the result is the probability of both group conditions, not merely the probability of at least one claim among the first three. Multiplying by the probability that all four remaining devices have no claim enforces the second condition.
Reading Count Language Correctly
The formula gives one exact count at a time. Verbal conditions determine whether one term, several terms, or a complement is required. For \(N\sim\operatorname{Bin}(n,p)\), the following translations are useful:
| Wording | Count event | Calculation |
|---|---|---|
| Exactly \(k\) claims | \(N=k\) | \(\binom{n}{k}p^kq^{n-k}\) |
| No claims | \(N=0\) | \(q^n\) |
| At least one claim | \(N\geq 1\) | \(1-P(N=0)=1-q^n\) |
| At least \(r\) claims | \(N=r,\ldots,n\) | Add the exact probabilities for those counts |
| At most \(r\) claims | \(N=0,\ldots,r\) | Add the exact probabilities for those counts |
The complement is often the shortest route for “at least one,” because its opposite event has a single count, zero. For other conditions, check the endpoint carefully: “more than two” starts at \(N=3\), while “at least two” starts at \(N=2\). Counts outside \(0,\ldots,n\) are impossible and have probability zero.
Common Mistakes and Model Checks
- Omitting the binomial coefficient. The product \(p^kq^{n-k}\) describes one specified arrangement. Multiply by \(\binom{n}{k}\) to include every arrangement with \(k\) claims.
- Using the wrong event. “Exactly two” is one count; “at least two” includes two and every larger feasible count.
- Mixing up \(p\) and \(q\). For exactly \(k\) claims, the powers are \(p^k\) and \(q^{n-k}\), because the other \(n-k\) units have no claims.
- Counting the wrong exposure period. The probability \(p\) must refer to the same period used to define each trial. Do not silently treat monthly probabilities as annual probabilities, or vice versa.
- Applying the model when assumptions fail. A fixed \(n\), equal claim probabilities, and independence are required. If these conditions do not hold, the binomial formula may not describe the claim count.
Check Your Understanding
For each hypothetical scenario, identify the event from its wording and apply the binomial model. Probabilities are rounded to four decimal places unless an exact value is shown.
- Six independent policies each have claim probability \(0.20\) over one year. What is the probability of exactly two policies having claims? (Answer: \(0.2458\))
- Five independent insured devices each have claim probability \(0.20\). What is the probability of no claims? (Answer: \(0.3277\))
- Five independent contracts each have claim probability \(0.20\). What is the probability of at least four claims? (Answer: \(0.0067\))
- Eight independent units each have claim probability \(0.25\). What is the probability of exactly three claims? (Answer: \(0.2076\))
- Five independent units each have claim probability \(0.10\). What is the probability that at least one of the first two units has a claim and neither of the remaining three does? (Answer: \(0.1385\))