Building Expected Loss Rates by Classification

Why Classification-Specific Rates Need a Clear Basis

An expected loss rate is meaningful only when its coverage and exposure unit are clear. A rate of $72 per vehicle-year is not interchangeable with $72 per $100 of payroll, even though the dollar amounts match. The same is true of rates for different coverages: a property loss rate and a liability loss rate describe different expected payments.

For an experience rating plan, a class-specific expected loss rate provides a benchmark for the covered losses associated with that class. The rate may be supplied as a prospective loss cost, calculated from prospective loss and exposure inputs, or derived from an indicated gross rate when the loss portion can be identified. In either case, the rate must refer to the coverage and exposure basis used for the plan.

Definition: A classification-specific expected loss rate (also called a prospective loss cost) is the expected covered loss amount for one exposure unit in a specified classification, on a stated coverage and prospective basis. If class $c$ has expected covered losses $L_c$ and exposure $E_c$, both measured on matching bases, its expected loss rate is $r_c=L_c/E_c$. Its units are dollars of covered loss per exposure unit.

The loss amount and exposure denominator must match. For example, if the exposure unit is one vehicle-year, the numerator must represent losses for that same coverage and the appropriate vehicle-year basis. If the exposure is payroll measured in units of $100, a rate is dollars of loss per $100 of payroll—not per dollar of payroll. A loss cost is not a complete premium rate: it excludes expense and other premium provisions unless those are specifically included in the stated measure.

Two Routes to a Class Loss Rate

A prospective loss cost can be provided directly, or it can be built from prospective expected losses divided by matching exposure. Inputs used in this calculation must already be on the intended prospective basis. The calculation itself does not correct mismatched periods, coverages, or exposure units.

A second route starts with an indicated gross rate. A gross rate includes more than expected losses, so it cannot ordinarily be treated as the loss rate. If a consistent expected loss ratio is available, the expected loss rate is the gross rate multiplied by that ratio. The ratio must describe the same class, coverage, and premium basis; it must also define which loss components are included.

Conversion rule: If $g_c$ is a class's indicated gross rate per exposure unit and $q_c$ is the expected loss ratio defined as expected covered losses divided by premium on that same basis, then the expected loss rate is $r_c=q_c g_c$. This conversion does not apply when the ratio or rate uses a different coverage, exposure unit, or loss definition. Without a suitable loss ratio or an equivalent expense and premium provision breakdown, a gross rate alone does not determine its loss component.

Some rating structures express class differences as relativities to a base loss cost. If $b$ is the base loss cost and $a_c$ is class $c$'s relativity, the unnormalized class loss cost is $b a_c$. When a target overall average is specified, a common scaling factor can bring the class rates to that average while retaining their relative differences. The appropriate weights are the class exposure shares on the same prospective basis.

Worked Example: Calculating Prospective Loss Costs by Class

A hypothetical commercial policy program supplies prospective covered-loss estimates for two classes. Class A has expected covered losses of $864,000 and 12,000 exposure units, where one unit is $100 of payroll. Class B has expected covered losses of $1,050,000 and 10,000 units on the same basis. The loss estimates are already prospective and include the same coverage for both classes. Calculate each expected loss rate.

Method. Divide each class's prospective expected losses by its matching exposure units. The result is a loss cost per $100 of payroll.

For Class A:

r_A=\frac{\$864{,}000}{12{,}000\text{ units}} =\$72.00\text{ per unit}

For Class B:

r_B=\frac{\$1{,}050{,}000}{10{,}000\text{ units}} =\$105.00\text{ per unit}

Since one unit is $100 of payroll, the rates are $72.00 and $105.00 per $100 of payroll, respectively. As a check, multiplying each rate by its exposure returns the supplied loss amount: $12{,}000\times \$72=\$864{,}000$, and $10{,}000\times \$105=\$1{,}050{,}000$. The calculation avoids treating the rates as dollars per payroll dollar or adding premium expenses to the loss costs.

Worked Example: Extracting Loss Rates From Indicated Gross Rates

A hypothetical insurer has indicated gross rates of $145 per vehicle-year for Class C and $190 per vehicle-year for Class D. The corresponding expected loss ratios are 0.62 and 0.58. Each ratio is defined as expected covered losses, including allocated loss adjustment expenses, divided by gross premium for the same class and coverage. Calculate the expected loss rates.

Method. Multiply each indicated gross rate by its matching expected loss ratio. The common mistake is to use the gross rate as though it were all loss cost, or to divide the rate by the loss ratio. The ratio is the loss portion of premium, so it is multiplied by the gross rate.

For Class C:

r_C=0.62(\$145)=\$89.90\text{ per vehicle-year}

For Class D:

r_D=0.58(\$190)=\$110.20\text{ per vehicle-year}

The rates are $89.90 and $110.20 per vehicle-year, with the covered-loss definition including allocated loss adjustment expenses as stated. A check is that the loss rates are below their respective gross rates, as expected when the loss ratios are less than 1. Dividing instead would give $\$145/0.62\approx \$233.87$ and $\$190/0.58\approx \$327.59$, which incorrectly makes the loss portion exceed the gross rate.

Relativities and a Target Overall Average

A set of class relativities preserves a pattern of differences, but a chosen base loss cost and those relativities do not necessarily produce a desired portfolio average. The average depends on the exposure mix. For example, a high loss-cost class has more influence on the overall rate when it accounts for a larger share of exposure.

The following relationship explains the correct weights. Suppose each class has exposure $E_c$ and loss rate $r_c$, all on the same basis. Its expected losses are $r_cE_c$. The combined loss rate is total expected losses divided by total exposure:

r_{\text{combined}} =\frac{\sum_c r_cE_c}{\sum_c E_c} =\sum_c w_c r_c, \qquad w_c=\frac{E_c}{\sum_j E_j}.

Thus the combined rate is an exposure-weighted average, not a simple average of class rates unless the classes have equal exposures. If the relativities produce an unnormalized exposure-weighted average $r_{\text{current}}$, and the desired overall rate is $r_{\text{target}}$, multiply every class rate by the common factor $k=r_{\text{target}}/r_{\text{current}}$. Because every rate is scaled equally, the ratio between any two class rates remains unchanged.

Worked Example: Normalizing Class Relativities

A hypothetical property program uses a base prospective loss cost of $78 per $1,000 of insured value. Three class relativities are 0.80, 1.00, and 1.35. Their projected exposure shares are 40%, 35%, and 25%, respectively. The desired overall expected loss rate is $84 per $1,000 of insured value. Find normalized class rates to the nearest cent.

Method. First apply each relativity to the base loss cost, calculate the exposure-weighted average, and then apply one common normalization factor. This method avoids taking an unweighted average or changing the class relativities separately.

The unnormalized rates are:

\$78(0.80)=\$62.40,\qquad \$78(1.00)=\$78.00,\qquad \$78(1.35)=\$105.30.

The exposure-weighted average is:

r_{\text{current}} =0.40(\$62.40)+0.35(\$78.00)+0.25(\$105.30) =\$24.96+\$27.30+\$26.325 =\$78.585.

The common factor required to reach $84 is:

k=\frac{\$84}{\$78.585}\approx 1.06890627.

Apply it to each unnormalized rate:

r_1=\$62.40(1.06890627)\approx \$66.70,\quad r_2=\$78.00(1.06890627)\approx \$83.37,\quad r_3=\$105.30(1.06890627)\approx \$112.56.

The rates are stated to the nearest cent. Using the unrounded rates, their exposure-weighted average is $1.06890627(\$78.585)\approx \$84.00$. Using the displayed rounded rates gives $0.40(\$66.70)+0.35(\$83.37)+0.25(\$112.56)=\$84.00$. The target is met, subject only to rounding.

A Reliable Construction Sequence

Before accepting a class expected loss rate, check the inputs in a fixed order. This prevents a plausible-looking dollar amount from being attached to the wrong coverage or denominator.

Specify the coverage and loss measure.
State which losses are included, such as covered indemnity alone or indemnity plus allocated loss adjustment expenses. Keep the definition consistent across classes.

Name the exposure unit.
Write the unit explicitly, such as one vehicle-year, one policy-year, or $100 of payroll. Confirm that numerator and denominator refer to the same class basis.

Identify the source of the rate.
For prospective losses and exposure, divide loss by exposure. For an indicated gross rate, use a matching expected loss ratio or another documented method that isolates the loss component.

Check classification factors and averaging.
Apply relativities consistently. If an overall target is imposed, use exposure shares for the stated basis and check the weighted average after normalization.

Common Errors and Practical Significance

A recurring unit error is dividing by the wrong scale. In the first example, the exposure is 12,000 units of $100 payroll, or $1,200,000 of payroll. Dividing $864,000 by $1,200,000 produces 0.72 dollars per payroll dollar. That is not the requested rate per $100 of payroll. Converting units gives $0.72\times100=\$72$ per $100, in agreement with the direct calculation.

Using the gross rate as the expected loss rate. Gross rates include premium components beyond expected losses. Isolate the loss portion using a matching ratio or documented rate structure.
Applying a ratio backwards. If the expected loss ratio is losses divided by premium, multiply the gross rate by the ratio to obtain the loss rate; do not divide by it.
Comparing unlike units. A rate per vehicle-year cannot be averaged directly with a rate per $1,000 of insured value. First establish a common meaningful basis, if one is available.
Using the wrong class weights. When matching an overall rate to class rates, weight by exposure on the stated basis, not by the number of classes or by an unrelated premium distribution.
Mixing loss definitions. If one input includes allocated loss adjustment expenses and another does not, they are not consistent expected loss rates until the definitions are reconciled.

These checks matter because a class rate is both a pricing input and a benchmark. A unit or coverage mismatch can make expected losses for a class appear too high or too low, distorting the experience comparison even when the arithmetic is flawless. Clear definitions also make rates interpretable to reviewers and usable across classifications. The methods here are aligned with the CAS Exam 8 (Advanced Ratemaking) syllabus.

Key takeaway: Every expected loss rate needs a class, a coverage definition, a prospective loss basis, and an exposure unit. Calculate or convert the rate only after confirming those elements match; when class relativities must meet an overall target, normalize with exposure weights.

Check Your Understanding

For each problem, retain the stated coverage and exposure basis in your answer.

A hypothetical class has prospective covered losses of $384,000 and 8,000 policy-years of exposure. What is its expected loss rate? (Answer: $48.00 per policy-year)
A class has prospective losses of $525,000 and 7,500 exposure units, each representing $100 of payroll. Calculate its expected loss rate. (Answer: $70.00 per $100 of payroll)
A hypothetical class has an indicated gross rate of $160 per vehicle-year and an expected loss ratio of 0.55, defined for the same coverage and premium basis. Calculate the expected loss rate. (Answer: $88.00 per vehicle-year)
A base loss cost is $50 per exposure unit. Two class relativities are 0.90 and 1.20, and their projected exposure shares are 60% and 40%. Normalize the class rates to an overall target of $60 per unit; give rates to the nearest cent. (Answer: $52.94 and $70.59 per exposure unit, respectively)
A hypothetical program reports a gross rate of $120 per vehicle-year with an expected loss ratio of 0.60 for one class, and a gross rate of $150 per $1,000 of insured value with an expected loss ratio of 0.64 for another. Calculate each expected loss rate and state whether the two results can be directly averaged. (Answer: $72.00 per vehicle-year and $96.00 per $1,000 of insured value; they cannot be directly averaged because their exposure units differ)

What You'll Learn