Expected Losses by Coverage and Accident Period

Why Coverage and Accident Period Must Match

An expected loss estimate is only meaningful when its inputs describe the same risks. A premium-based estimate uses premium for a particular coverage and accident period with an expected loss ratio appropriate to that same coverage and period. An exposure-based estimate uses the corresponding exposure units with a loss cost stated per matching unit. Mixing coverage or period labels can produce a mathematically tidy calculation that estimates the wrong losses.

The accident period identifies when the covered event occurred. For example, an accident-year estimate concerns losses from accidents occurring during that year, even if some claims are reported or paid later. A coverage identifies which insured loss obligations are included. If a policy provides both property and liability coverage, those components may have different premiums, expected loss ratios, exposure bases, and loss costs. They should be assigned separately before their expected losses are combined.

Definition: For coverage $c$ and accident period $a$, let $L_{c,a}$ be the aggregate ultimate loss for that coverage and accident period. Let $\Pi_{c,a}$ be its earned premium, $r_{c,a}$ its expected loss ratio, $N_{c,a}$ its exposure count in a specified unit, and $k_{c,a}$ its expected loss cost per unit. The premium-based estimate is $\mathbb{E}[L_{c,a}]=\Pi_{c,a}r_{c,a}$; the exposure-based estimate is $\mathbb{E}[L_{c,a}]=N_{c,a}k_{c,a}$. Use the method for which the inputs and assumptions are defined.

These equations estimate an ultimate loss amount, not the amount already paid or reported. The expected loss ratio is a proportion of premium expected to emerge as loss under the selected definition of loss. The expected loss cost is an expected loss amount per stated exposure unit, such as dollars per vehicle-year or dollars per policy-year. The unit is part of the assumption: a loss cost per vehicle-year cannot be multiplied by a count of policies unless the data establish that one policy is exactly one vehicle-year.

The premium or exposure must also be assigned to the relevant accident period. In an accident-year analysis, earned premium and earned exposures are commonly organized by the period in which the risk was covered, while losses are classified by accident date. The estimate should follow the period definitions of the data and assumptions. Do not substitute calendar-year paid losses or written premium merely because those figures are readily available.

The Two Expected-Loss Calculations

The premium-based method multiplies earned premium by the expected loss ratio selected for that coverage and accident period. The exposure-based method multiplies the number of matching exposure units by the expected loss cost for that coverage and period. In either case, the method requires a consistent scope: the loss assumption must describe the same coverage, loss components, and time period as the premium or exposure measure.

\widehat{U}_{c,a}^{\,\mathrm{premium}}=\Pi_{c,a}r_{c,a} \qquad\text{and}\qquad \widehat{U}_{c,a}^{\,\mathrm{exposure}}=N_{c,a}k_{c,a}

Here $\widehat{U}_{c,a}$ denotes an estimate of expected ultimate loss. The hat emphasizes that the quantity is an estimate based on selected assumptions and observed premium or exposure, not a known final outcome. When multiple coverage-period cells are being summarized, calculate each cell using its own matched inputs first. Then add the resulting loss amounts if the cells are non-overlapping and have compatible loss definitions and valuation bases.

Specify the target.
Name the coverage and accident period, and state what losses are included in the estimate.

Select the matching basis.
Use earned premium with an expected loss ratio, or exposure units with a loss cost per the same unit.

Pair like with like.
Check the coverage label, accident-period label, loss scope, and units on every input.

Calculate each cell and aggregate.
Compute each expected loss separately; add amounts only after verifying that the cells do not overlap.

Why the Formula Holds

The premium formula follows from the definition of a loss ratio. For one coverage-period cell, the loss ratio is loss divided by premium. If the expected loss ratio is $r_{c,a}$, applying that ratio to the matching premium gives the expected loss amount:

r_{c,a}=\frac{\mathbb{E}[L_{c,a}]}{\Pi_{c,a}} \quad\Longrightarrow\quad \mathbb{E}[L_{c,a}]=\Pi_{c,a}r_{c,a}

This step treats $r_{c,a}$ as the selected expected ratio for the cell, not as a ratio calculated from an unrelated coverage or period. The exposure formula follows in the same way. If expected loss per exposure unit is $k_{c,a}$, then the total expected loss across $N_{c,a}$ units is the sum of the expected losses for those units. By linearity of expectation, that sum equals $N_{c,a}k_{c,a}$; independence among units is not required for this expected-value identity.

\mathbb{E}\left[\sum_{j=1}^{N_{c,a}}L_j\right] =\sum_{j=1}^{N_{c,a}}\mathbb{E}[L_j] =N_{c,a}k_{c,a}

For the exposure derivation, each unit must have the same expected loss cost represented by $k_{c,a}$, or the cost must be an appropriate average for the units being counted. If that condition is not met, multiplying one selected cost by the entire count does not represent the stated expected value. Also, neither formula corrects a mismatched period or coverage automatically; matching is a data and assumption task that must occur before multiplication.

Worked Example: Premium-Based Estimates by Coverage and Accident Year

A hypothetical insurer estimates ultimate losses for two coverages in accident years 2024 and 2025. The selected earned premiums and expected loss ratios are shown below. For this calculation, there is no discounting ($i=0\%$); mortality is not applicable. No parametric loss distribution is assumed: the method uses the selected expected loss ratios and premiums only.

Accident year	Coverage	Earned premium	Expected loss ratio
2024	Physical damage	$1,200,000	0.62
2024	Liability	$800,000	0.75
2025	Physical damage	$1,000,000	0.65
2025	Liability	$1,100,000	0.72

Method. Use the premium-based expected loss formula separately for each coverage and accident year. The common trap is to apply one portfolio-wide ratio to all premium, despite the table providing different coverage- and period-specific assumptions.

For 2024 physical damage:

\$1{,}200{,}000(0.62)=\$744{,}000

For 2024 liability:

\$800{,}000(0.75)=\$600{,}000

The 2024 total expected ultimate loss is $\$744{,}000+\$600{,}000=\$1{,}344{,}000$. For 2025 physical damage, $\$1{,}000{,}000(0.65)=\$650{,}000$. For 2025 liability, $\$1{,}100{,}000(0.72)=\$792{,}000$. Thus, the 2025 total is $\$650{,}000+\$792{,}000=\$1{,}442{,}000$, and the two-year total is $\$1{,}344{,}000+\$1{,}442{,}000=\$2{,}786{,}000$.

Check. The four cell estimates sum to $\$744{,}000+\$600{,}000+\$650{,}000+\$792{,}000=\$2{,}786{,}000$, agreeing with the year-by-year total. Each premium was paired with its own coverage-period ratio.

Worked Example: Exposure-Based Estimates With Different Units

A hypothetical specialty carrier estimates losses for two coverages in accident year 2025. Glass coverage has 18,000 earned vehicle-years and a selected expected loss cost of $\$42$ per vehicle-year. Roadside assistance has 12,500 earned dispatch exposures and a selected expected loss cost of $\$31$ per dispatch exposure. Assume no discounting ($i=0\%$); mortality is not applicable. No parametric loss distribution is assumed; only the expected loss costs and exposure counts are used.

Method. Apply each cost only to its stated exposure unit and coverage. The common mistake is to add the two exposure counts first and multiply by a single average cost, even though the counts and costs refer to different units.

For glass coverage:

18{,}000\text{ vehicle-years}\times \$42\text{ per vehicle-year} =\$756{,}000

For roadside assistance:

12{,}500\text{ dispatch exposures}\times \$31\text{ per dispatch exposure} =\$387{,}500

The total expected ultimate loss for the two non-overlapping coverages is $\$756{,}000+\$387{,}500=\$1{,}143{,}500$. The units cancel within each multiplication, leaving dollars in each result.

Check. The glass result is based only on vehicle-years and the roadside result only on dispatch exposures. Adding the dollar estimates is valid here because the two coverage amounts are separate and their loss scopes are assumed compatible.

Worked Example: Combining Premium and Exposure Bases Across Periods

A hypothetical insurer needs an expected ultimate loss estimate for two coverages in accident years 2023 and 2024. For cyber coverage, the selected basis is policies with earned coverage during the year: 1,600 policy-years in 2023 and 1,900 policy-years in 2024. The selected loss costs are $\$225$ and $\$240$ per policy-year, respectively. For equipment liability, the selected earned premiums are $\$640{,}000$ in 2023 and $\$710{,}000$ in 2024; the respective expected loss ratios are 0.55 and 0.58. Assume no discounting ($i=0\%$); mortality is not applicable. No parametric loss distribution is assumed; the calculations use the given mean assumptions.

Method. Use exposure times loss cost for cyber and premium times loss ratio for equipment liability. The common mistake is to apply the cyber loss cost to equipment liability exposures, or to carry one year’s assumption into the other year without support.

Cyber, accident year 2023:

1{,}600\text{ policy-years}\times \$225\text{ per policy-year} =\$360{,}000

Cyber, accident year 2024:

1{,}900\text{ policy-years}\times \$240\text{ per policy-year} =\$456{,}000

Equipment liability, accident year 2023:

\$640{,}000(0.55)=\$352{,}000

Equipment liability, accident year 2024:

\$710{,}000(0.58)=\$411{,}800

The 2023 total is $\$360{,}000+\$352{,}000=\$712{,}000$. The 2024 total is $\$456{,}000+\$411{,}800=\$867{,}800$. The combined estimate is $\$712{,}000+\$867{,}800=\$1{,}579{,}800$.

Check. The four cell amounts sum to $\$360{,}000+\$456{,}000+\$352{,}000+\$411{,}800=\$1{,}579{,}800$. Each amount uses the basis and assumption specified for its coverage and accident year.

Common Assignment Errors

A calculation can be arithmetically correct and still answer the wrong question. Before accepting an expected ultimate estimate, check the assignments, not just the multiplication.

Using a ratio for the wrong coverage. A liability expected loss ratio should not be applied to physical damage premium simply because both premiums are on the same policy.
Using a ratio or loss cost for the wrong accident period. If the selected assumption varies by year, carry each period’s value only into its matching cell.
Confusing accident period with evaluation period. Losses observed at a valuation date may include payments from multiple accident periods. They are not automatically the expected ultimate losses for one accident year.
Mixing written and earned measures. Use the premium or exposure basis specified for the estimate. A written premium total does not become earned premium just because it is available by policy year.
Ignoring exposure units. A count without its unit is incomplete. Confirm whether the loss cost is per policy, vehicle, payroll unit, transaction, or another defined exposure.
Averaging assumptions before calculating. If premiums or exposures differ by coverage or period, calculate the separate expected amounts first. An unweighted average of ratios or loss costs will generally not preserve the correct total.

Key takeaway: Define the coverage-period cell first. Pair it with the corresponding earned premium and expected loss ratio, or the corresponding exposure count and expected loss cost. Calculate cell estimates before adding them; do not let a shared portfolio label substitute for a matched basis.

This approach is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus. The practical discipline is to make every input auditable: a reviewer should be able to see which coverage, accident period, premium or exposure unit, and expected assumption produced each amount.

Check Your Understanding

For each problem, identify the matching basis before calculating. Dollar answers are not rounded unless stated.

A hypothetical insurer has $\$420{,}000$ of earned premium for a specified coverage in accident year 2026 and selects an expected loss ratio of 0.64 for that same cell. What is the expected ultimate loss? (Answer: $\$268{,}800$)
A hypothetical crop policy portfolio has 7,500 earned acre-units for one coverage and accident year. The selected expected loss cost is $\$28.40$ per acre-unit. Calculate the expected ultimate loss. (Answer: $\$213{,}000$)
For accident year 2025, coverage A has $\$300{,}000$ of earned premium and an expected loss ratio of 0.50. Coverage B has $\$450{,}000$ of earned premium and an expected loss ratio of 0.70. Calculate each estimate and the combined expected ultimate loss. (Answer: Coverage A $\$150{,}000$; coverage B $\$315{,}000$; total $\$465{,}000$)
A hypothetical warranty portfolio has 2,400 earned contract-years in 2023 at an expected loss cost of $\$90$ per contract-year, and 2,600 earned contract-years in 2024 at $\$96$ per contract-year. Calculate each accident-year estimate and their combined total. (Answer: 2023 $\$216{,}000$; 2024 $\$249{,}600$; total $\$465{,}600$)
For one accident year, a hypothetical insurer estimates property losses from 4,000 earned property-policy-years at $\$125$ per policy-year. It estimates a separate liability coverage using $\$800{,}000$ earned premium and an expected loss ratio of 0.58. What is the combined expected ultimate loss, and which basis applies to each coverage? (Answer: Property $\$500{,}000$ using exposure; liability $\$464{,}000$ using premium; combined $\$964{,}000$)

What You'll Learn