Why the Denominator Matters
A claim count by itself does not show how frequently claims occur. Forty claims could represent a high rate in a small book of business or a low rate in a much larger one. To compare experience meaningfully, divide a clearly defined number of claims by the exposure units that were subject to those claims.
The result depends on both parts of the calculation. The numerator must specify what is being counted—for example, reported claims or claims meeting a stated coverage definition. The denominator must measure the corresponding amount of risk exposure over the relevant period. A frequency based on policy-years answers a different question from one based on vehicle-years.
An exposure unit represents a defined amount of insured risk over time. One policy in force for one year may contribute one policy-year of exposure; a policy in force for half a year may contribute 0.5 policy-year. A vehicle-year similarly represents one insured vehicle for one year. Other lines can use other bases. The essential requirement is not that one particular unit be used everywhere, but that the unit be appropriate to the risk and consistently defined.
Frequency is a descriptive rate for the experience being measured. It is not automatically the probability that a policy will have a claim. For example, a frequency of 0.08 claims per policy-year means eight claims per 100 policy-years on that basis. It does not, without additional assumptions, state the proportion of policies with at least one claim: one policy can have more than one claim.
Define the Count and the Exposure
Before calculating, write down what qualifies for the numerator and what contributes to the denominator. Possible count definitions include reported claims during a period, claims assigned to an accident period, or claims for a specified coverage. Those definitions may produce different counts. A count of paid claims, for instance, is not necessarily the same as a count of reported claims. Use the count definition required by the analysis and apply it consistently.
The exposure should match the counted risk as closely as the available data allow. If the numerator counts vehicle-specific claims, vehicle-years may be the relevant denominator. If it counts claims at the policy level, policy-years may be appropriate. If only a particular coverage can generate the counted claims, the denominator should reflect exposure to that coverage rather than unrelated coverage.
State which claims are included and the period or coverage to which they belong.
Identify the insured unit, such as a policy or vehicle, and measure its time exposed during the relevant period.
Calculate claims divided by exposure, then report a convenient scale such as claims per 100 exposure units.
Confirm that the claims and exposures refer to compatible risks, coverage, and time periods.
Exposure is not interchangeable with premium. Premium reflects the amount charged for insurance and may vary with rates and rating factors. Exposure measures the insured risk units and time at risk. Earlier material on matching premium to coverage and exposure helps distinguish these measures; a frequency calculation should use its stated exposure base, not substitute premium for exposure.
Worked Example: Reported Claims per Policy-Year
A hypothetical insurer observes 27 reported claims during a calendar year. The policies in the study contribute 1,800 earned policy-years during that same year. Calculate the reported-claim frequency per policy-year and per 100 policy-years.
Method and assumptions. The numerator is the 27 claims meeting the stated reported-claim definition. The denominator is 1,800 earned policy-years for the same study and period. This is a direct calculation from observed counts and exposure; no probability distribution, mortality assumption, or interest rate is used.
First calculate the frequency per policy-year:
To express this per 100 policy-years, multiply by 100:
Check. At 1.5 claims per 100 policy-years, 1,800 policy-years represent 18 groups of 100, and \(18 \times 1.5=27\) claims. The two reported forms describe the same frequency.
Common mistake avoided: Reporting “1.5 claims” without saying “per 100 policy-years” leaves out the scale and denominator that give the number meaning.
Choosing and Converting Exposure Bases
A policy can cover more than one insured object, and policies can be in force for different lengths of time. Consequently, policy-years and object-years need not be equal. If the analysis counts claims by insured object, changing the denominator from object-years to policy-years can substantially change the numerical rate. The arithmetic may be correct under either denominator, but the two rates answer different questions.
To convert a frequency per exposure unit to a frequency per \(k\) exposure units, multiply by \(k\). This changes only the reporting scale, not the underlying experience. For example, \(0.04\) claims per exposure unit equals 4 claims per 100 exposure units. Always carry the unit through the calculation.
Worked Example: Policy-Years Versus Vehicle-Years
A hypothetical portfolio has 600 policies. During the study period, each policy contributes an average of 0.8 policy-year, and each policy covers an average of 1.5 vehicles. The portfolio has 36 vehicle-specific claims. Calculate the frequency per 100 policy-years and per 100 vehicle-years.
Method and assumptions. Count only the 36 vehicle-specific claims. Treat the stated average exposure figures as applicable to the portfolio for the study period. No probability distribution, mortality assumption, or interest rate is needed; this is an observed-rate calculation.
First calculate policy-year exposure:
Next calculate vehicle-year exposure. Each policy-year corresponds to an average of 1.5 covered vehicles:
The frequency per 100 policy-years is:
The frequency per 100 vehicle-years is:
Check. Both rates reconstruct the same count: \(7.5 \times 480/100=36\), and \(5 \times 720/100=36\). The numerical rates differ because their denominators measure different exposure units, not because one calculation is inconsistent.
Common mistake avoided: Dividing vehicle-specific claims by vehicle-years and labeling the result “per policy-year,” or comparing it directly with a policy-year rate as if the bases were identical.
Match Claim Eligibility to Exposure Eligibility
A denominator can be too broad as well as too narrow. Suppose a portfolio includes policies with different coverages, but the numerator counts claims that could arise only under one coverage. Dividing by all policy exposure implicitly treats policies without that coverage as exposed to the counted risk. The resulting rate is a valid arithmetic ratio, but it does not represent frequency per unit of exposure to that coverage.
The count and denominator also need compatible time assignments. Claims assigned to one period should be compared with exposure for that period on the same basis. A mismatch—such as a full-year claim count divided by exposure from only part of the year—distorts the rate. Before interpreting the result, check that the claim definition, coverage eligibility, and exposure period line up.
Worked Example: Restricting Exposure to Covered Policies
A hypothetical insurer has 700 earned policy-years in a study year. Of these, 500 policy-years have a specified equipment-breakdown coverage. The insurer counts 30 claims arising under that coverage. Calculate the claim frequency per 100 policy-years using (a) all policy exposure and (b) only exposure with the relevant coverage. Identify the rate suited to describing frequency among covered exposure.
Method and assumptions. The numerator contains the 30 claims arising under equipment-breakdown coverage. Exposure is stated in earned policy-years for the same study year. No distribution, mortality assumption, or interest rate is assumed.
Using all policy exposure gives:
Using only policy-years with the relevant coverage gives:
The second rate describes frequency per covered policy-year because its denominator includes only exposure eligible for the counted claims. The first rate can be calculated, but it describes the 30 coverage-specific claims relative to all policy exposure, including 200 policy-years without that coverage. It answers a different question.
Check. The covered exposure is \(500\) of the \(700\) total policy-years, so it cannot exceed total exposure. The rate of 6 per 100 covered policy-years reconstructs \(6 \times 500/100=30\) claims.
Common mistake avoided: Using the total book denominator automatically, without checking whether every included policy was eligible for the claims being counted.
Combining Segments Correctly
When combining groups with different exposure amounts, the overall frequency is total claims divided by total exposure. It is not generally the simple arithmetic average of the group frequencies. The larger-exposure group contributes more to the overall rate because it represents more of the total risk exposure.
Thus the overall frequency is an exposure-weighted average of segment frequencies. A simple average gives each segment equal weight regardless of its exposure, so it generally does not reproduce the total claim count divided by total exposure.
Worked Example: Aggregating Two Exposure Groups
Two hypothetical rating groups use the same claim definition and exposure base. Group A has 12 claims over 120 policy-years; Group B has 18 claims over 360 policy-years. Calculate each group’s frequency and the combined frequency per 100 policy-years. Explain why the unweighted average is not the combined result.
Method and assumptions. Both claim counts refer to the same period and definition, and both denominators are policy-years. The data are observed counts and exposures; no probability distribution, mortality assumption, or interest rate is used.
Group A’s frequency is \(12/120=0.10\) claims per policy-year, or 10 per 100 policy-years. Group B’s frequency is \(18/360=0.05\) claims per policy-year, or 5 per 100 policy-years.
For the combined rate, add counts and exposures before dividing:
The unweighted average of the two segment rates is \((10+5)/2=7.5\) claims per 100 policy-years, which is not the combined frequency. Group B has three times the exposure of Group A, so it must receive three times the weight. Indeed, the exposure-weighted calculation is \((120\times10+360\times5)/480=6.25\).
Check. The combined rate corresponds to \(6.25 \times 480/100=30\) claims, the observed total.
Common mistake avoided: Averaging subgroup frequencies equally when subgroup exposures are unequal.
Common Frequency Errors
- Leaving the unit unstated. “Frequency is 0.06” is incomplete. State whether this means claims per policy-year, vehicle-year, or another exposure unit.
- Mixing claim definitions. Counts based on different statuses, coverages, or time assignments are not automatically comparable.
- Using the wrong time basis. A half-year of coverage contributes half a policy-year, not a full policy-year, when exposure is measured continuously by time in force.
- Confusing frequency with the proportion of policies having a claim. Frequency counts claims; one policy may contribute multiple claims.
- Averaging segment rates without exposure weights. Calculate total claims divided by total exposure, or use the equivalent exposure-weighted average.
These checks matter whenever claim experience is compared across periods or groups. A difference in calculated frequency may reflect a genuine difference in claim occurrence, but it may also arise from a changed claim definition, exposure base, coverage mix, or period assignment. Establishing a consistent basis is therefore part of the calculation, not merely a reporting detail. This foundational measure is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.
Check Your Understanding
For each problem, identify the relevant count and exposure basis before calculating.
- A hypothetical portfolio records 24 claims under a stated definition and 1,600 earned policy-years in the same period. What is the frequency per 100 policy-years? (Answer: 1.5 claims per 100 policy-years.)
- A book has 400 policies, each contributing an average of 0.75 policy-year, and 24 claims. Calculate the frequency per 100 policy-years. (Answer: \(24/(400\times0.75)\times100=8\) claims per 100 policy-years.)
- A hypothetical fleet has 300 policy-years and averages 2.4 covered vehicles per policy-year. It has 36 vehicle-specific claims. What is the frequency per 100 vehicle-years? (Answer: \(36/(300\times2.4)\times100=5\) claims per 100 vehicle-years.)
- A coverage-specific count is 21 claims. The portfolio has 600 total policy-years, of which 350 are eligible for that coverage. Calculate the rate per 100 eligible policy-years and explain why the total exposure denominator gives a different measure. (Answer: \(21/350\times100=6\) claims per 100 eligible policy-years; using all exposure gives 3.5 per 100 total policy-years.)
- Group A has 8 claims over 80 exposure units, and Group B has 24 claims over 320 exposure units, with the same claim definition and exposure base. Find the combined frequency per 100 exposure units and compare it with the unweighted average of group rates. (Answer: \(32/400\times100=8\) per 100 exposure units; the unweighted average is \((10+7.5)/2=8.75\), which is not the combined rate.)