Standardizing Exposure Mix for Fair Comparisons

Why Exposure Mix Changes the Comparison

Two experience-rated plans can have different aggregate loss rates even when one performs better within every comparable risk segment. The reason may be that the plans cover different proportions of lower- and higher-cost risks. Similarly, a portfolio’s aggregate rate can change from one period to another even if rates within segments do not change, simply because the exposure mix changed.

Standardizing exposure mix addresses this problem by applying the same segment weights to each plan or period. The resulting comparison answers a specific question: what would each set of segment rates produce if every plan or period had the same reference mix? It does not adjust for differences that are not captured by the chosen segments, and it does not by itself establish that a plan caused a change in loss experience.

Definition: For plan or period $j$ and risk segment $k$, let $E_{jk}$ be exposure and $L_{jk}$ be the corresponding loss amount, measured on a consistent basis. The segment rate is $r_{jk}=L_{jk}/E_{jk}$. Given common reference weights $w_k$, where $w_k\geq 0$ and $\sum_k w_k=1$, the standardized rate for $j$ is $\sum_k w_k r_{jk}$.

The unit of a rate depends on the exposure definition. For example, if exposure is policy-years, the rate is dollars per policy-year; if exposure is thousands of payroll dollars, it is dollars per thousand payroll dollars. Keep that basis fixed across segments and periods. Before comparing, also use consistent definitions of loss, valuation dates, and any adjustments applied to the data.

From Segment Rates to a Standardized Rate

The aggregate rate for plan $j$ is $R_j=\sum_k (E_{jk}/\sum_h E_{jh})r_{jk}$. Its weights are the plan’s own exposure shares. Those shares can differ across plans. A standardized rate instead uses a selected reference distribution $w_k$ for all plans. The reference can represent a target portfolio, a fixed baseline period, or another stated population relevant to the decision.

R_j^{\mathrm{std}}=\sum_k w_k r_{jk}, \qquad \sum_k w_k=1

The formula follows from constructing a hypothetical portfolio with total exposure $E^*$ and reference exposure $E^*w_k$ in segment $k$. If the plan’s rate in that segment remains $r_{jk}$, its hypothetical losses are $E^*w_k r_{jk}$. Summing over segments gives total hypothetical losses $E^*\sum_k w_k r_{jk}$. Dividing by $E^*$ gives the standardized rate. Thus, the common weights remove differences in the recorded segment proportions from the aggregate comparison.

Standardization is only as informative as the segmentation and data support. The segments should be defined consistently and be relevant to expected loss differences. Each plan must have enough exposure to estimate the rates in segments receiving positive reference weight. If a plan has no observations in a required segment, its rate there cannot be estimated directly from that plan’s data; do not silently assign a rate or omit the segment. A suitable modeling or external-data method would need to be stated separately.

Worked Comparisons

Worked Example: A Reversal After Standardizing Two Plans

A hypothetical insurer compares two experience plans using policy-years as exposure and incurred losses in dollars. The segment definitions and loss basis are consistent. Plan A has 900 low-risk policy-years and 100 high-risk policy-years, with losses of $9,000 and $4,000, respectively. Plan B has 200 low-risk policy-years and 800 high-risk policy-years, with losses of $2,400 and $24,000. Calculate each aggregate rate, then compare standardized rates using equal reference weights. State which plan has the lower rate on each basis.

Method. First calculate each segment rate as losses divided by policy-years. Then calculate aggregate rates using each plan’s own exposure shares and standardized rates using the common 0.5, 0.5 weights. This method avoids mistaking a different exposure mix for a difference in within-segment rates.

Plan A segment rates. The low-risk rate is $9,000 / 900 = $10 per policy-year. The high-risk rate is $4,000 / 100 = $40 per policy-year. Plan A has 1,000 policy-years and $13,000 of total losses, so its aggregate rate is $13,000 / 1,000 = $13.

Plan B segment rates. The low-risk rate is $2,400 / 200 = $12 per policy-year. The high-risk rate is $24,000 / 800 = $30 per policy-year. Plan B also has 1,000 policy-years and $26,400 of total losses, so its aggregate rate is $26,400 / 1,000 = $26.40.

Standardize. Plan A’s rate is $0.5(10)+0.5(40)=5+20=\$25$. Plan B’s rate is $0.5(12)+0.5(30)=6+15=\$21$. The standardized comparison favors Plan B by $4 per policy-year, even though its aggregate rate is higher by $13.40.

Check. Under equal reference weights, the standardized rates are the simple averages of the two segment rates: $(10+40)/2=25$ and $(12+30)/2=21$. The calculations agree. The reversal occurs because Plan B has much more exposure in the high-risk segment, whose rates are higher for both plans.

Worked Example: Constructing Reference Weights From a Target Portfolio

A hypothetical organization compares two periods across three mutually exclusive risk segments. In Period A, exposures are 100, 300, and 100, with losses of $800, $6,000, and $5,000. In Period B, exposures are 300, 100, and 200, with losses of $3,000, $1,800, and $9,000. Use the combined exposure distribution of the two periods as the reference mix. Calculate that mix, both standardized rates, and the direction of the standardized comparison.

Method. Calculate the rate in each segment for each period. Then form the reference weights from the combined exposures by segment. This gives both periods identical weights and makes clear exactly which population the comparison represents.

Segment rates. Period A’s rates are $800 / 100 = $8, $6,000 / 300 = $20, and $5,000 / 100 = $50 per exposure unit. Period B’s rates are $3,000 / 300 = $10, $1,800 / 100 = $18, and $9,000 / 200 = $45 per exposure unit.

Reference weights. Combined exposure by segment is $100+300=400$, $300+100=400$, and $100+200=300$. Total combined exposure is $400+400+300=1{,}100$. Therefore, the weights are $400/1{,}100=4/11$, $400/1{,}100=4/11$, and $300/1{,}100=3/11$. They sum to $11/11=1$.

Standardized rates. Period A’s standardized rate is $(4/11)(8)+(4/11)(20)+(3/11)(50)=(32+80+150)/11=262/11=\$23.82$ per exposure unit, rounded to cents. Period B’s is $(4/11)(10)+(4/11)(18)+(3/11)(45)=(40+72+135)/11=247/11=\$22.45$ per exposure unit, rounded to cents. The standardized rate is lower in Period B by $15/11=\$1.36$ per exposure unit, rounded to cents.

Check. Period A’s total losses are $800+6{,}000+5{,}000=\$11{,}800$ over 500 units, giving an aggregate rate of $23.60. Period B’s losses total $3{,}000+1{,}800+9{,}000=\$13{,}800$ over 600 units, or $23.00 per unit. These aggregate rates are not the standardized rates: each uses its own period’s segment shares. The reference weights sum to one, and the standardized calculations use those same weights in both periods.

Worked Example: Checking Sensitivity to the Reference Mix

Two hypothetical plans are compared across three segments. Plan C has exposures of 500, 300, and 200 units and losses of $3,000, $4,500, and $7,200. Plan D has exposures of 100, 200, and 700 units and losses of $800, $2,400, and $21,000. A target book has reference weights 0.60, 0.30, and 0.10. Calculate the standardized rates and the hypothetical losses for each plan on a common total exposure of 2,000 units. Then check what equal reference weights would imply.

Method. Calculate segment rates first, apply the stated target weights, and multiply the standardized rates by the common exposure volume. A second, explicitly labeled set of weights shows why the chosen target mix matters; it must not be substituted without a reason.

Segment rates. Plan C’s rates are $3,000 / 500 = $6, $4,500 / 300 = $15, and $7,200 / 200 = $36 per unit. Plan D’s are $800 / 100 = $8, $2,400 / 200 = $12, and $21,000 / 700 = $30 per unit.

Target-mix standardization. Plan C’s standardized rate is $0.60(6)+0.30(15)+0.10(36)=3.60+4.50+3.60=\$11.70$. Plan D’s is $0.60(8)+0.30(12)+0.10(30)=4.80+3.60+3.00=\$11.40$. On 2,000 units, the corresponding hypothetical losses are $2{,}000(11.70)=\$23{,}400$ and $2{,}000(11.40)=\$22{,}800$. Plan D is lower by $600 on that common volume.

Equal-weight sensitivity check. With weights $1/3,1/3,1/3$, Plan C’s rate would be $(6+15+36)/3=\$19.00$, while Plan D’s would be $(8+12+30)/3=\$16.67$, rounded to cents. The ordering is unchanged, but the size of the difference changes from $0.30 to approximately $2.33 per unit. The target mix should therefore be chosen to answer the business question, not selected after seeing which weighting produces a preferred result.

Check. The target weights sum to $0.60+0.30+0.10=1$. Multiplying each standardized rate by the same 2,000 units preserves the rate comparison and avoids confusing rate differences with volume differences.

Interpreting the Results and Avoiding Traps

A crude rate and a standardized rate answer different questions. The crude rate describes the observed portfolio as it exists, including its actual exposure mix. The standardized rate compares segment experience under a common, stated mix. Both can be useful: the crude rate may be relevant to observed total cost, while the standardized comparison is useful when the goal is to compare experience apart from measured mix differences.

A useful exact accounting identity makes that distinction explicit. Let $C_j$ be the crude rate for plan $j$, $S_j$ its standardized rate using the same reference weights, and $M_j=C_j-S_j$. Then for plans A and B:

$$ C_A-C_B=(S_A-S_B)+(M_A-M_B) $$

This follows by adding and subtracting the standardized rates: $(C_A-C_B)=(S_A-S_B)+[(C_A-S_A)-(C_B-S_B)]$. The first term is the standardized rate difference; the second is the difference between each plan’s crude-to-standardized gap. That second term reflects how each plan’s actual mix interacts with its segment rates relative to the reference mix. It is an accounting decomposition, not proof that mix alone caused the observed difference.

A frequent exam trap is to compare only aggregate losses or aggregate loss rates and then attribute the difference to plan performance. In Worked Example 1, the wrong conclusion would be, “Plan A is better because $13 is less than $26.40.” Those are valid crude rates, but they do not use a common mix. The corrected standardized calculations are $25 for A and $21 for B, so B has the lower rate under equal segment weights. State which comparison is being made before interpreting the ranking.

Other checks help prevent errors:

Use rates, not raw segment losses, in the weighted average. The segment weights already account for the reference exposure shares. Weighting total losses directly does not produce a rate.
Keep weights common. If each plan’s own shares are used, the result is its crude rate, not a standardized comparison.
Verify the weights. They must be nonnegative and sum to one. If they do not, the result is not an average under the intended reference mix.
Keep units and loss bases consistent. A rate per policy-year cannot be combined directly with a rate per thousand dollars of payroll. Likewise, incompatible loss definitions or valuation bases weaken the comparison.
Check segment support. A reference weight on a segment with no credible or available rate for one plan makes the standardized result incomplete unless a defensible estimation method is supplied.

Standardization adjusts for the segment variables included in the calculation, not every difference between portfolios. A broad segment can still contain different risk distributions within it. The result can also depend on the reference weights, as the sensitivity calculation illustrates. For a fair and interpretable analysis, specify the target population, segment definitions, exposure basis, loss basis, and weight source before drawing a conclusion.

Key takeaway: Calculate comparable segment rates, choose one relevant reference mix, and apply its weights to every plan or period. Report standardized rates alongside observed crude rates when both the common-mix comparison and actual portfolio experience matter.

Check Your Understanding

For each problem, calculate rates on the stated exposure basis and use the same reference weights for both plans or periods.

A hypothetical plan comparison has low- and high-risk segment rates of $8 and $32 for Plan A, and $10 and $26 for Plan B. Plan A has exposures 90 and 10; Plan B has exposures 20 and 80. Find both crude rates and both standardized rates using equal weights. Which plan has the lower standardized rate? (Answer: Crude rates: A $10.40 and B $22.80; standardized rates: A $20.00 and B $18.00, so B is lower by $2.00 per exposure unit.)
In a hypothetical two-period study, Period 1 has exposures 200 and 800 with losses $2,000 and $24,000. Period 2 has exposures 600 and 400 with losses $6,600 and $14,400. Use reference weights 0.40 and 0.60. Find segment rates and standardized rates. (Answer: Period 1 rates are $10 and $30, standardized rate $22; Period 2 rates are $11 and $36, standardized rate $26.)
Two plans have segment rates, in order, of $5, $20, and $45 for Plan X and $8, $16, and $35 for Plan Y. A target book has exposures of 300, 500, and 200 units. Derive its weights and calculate each standardized rate. (Answer: Weights are 0.30, 0.50, and 0.20; X is $20.50 per unit and Y is $17.70 per unit.)
Plan P has exposures 400 and 100 with losses $4,000 and $5,000. Plan Q has exposures 100 and 400 with losses $1,200 and $16,000. Calculate each crude rate and each standardized rate using weights 0.75 and 0.25. State the standardized difference. (Answer: Segment rates are P $10 and $50, Q $12 and $40; crude rates are P $18 and Q $34.40; standardized rates are P $20 and Q $19, so Q is lower by $1 per unit.)
Plan R has segment rates $20 and $30, while Plan S has rates $22 and $24. Their actual exposure shares are (0.25, 0.75) and (0.75, 0.25), respectively. Calculate crude rates and standardized rates using the target weights (0.40, 0.60). (Answer: Crude rates are R $27.50 and S $22.50; standardized rates are R $26.00 and S $23.20, so S is lower by $2.80 per unit under the target mix.)

What You'll Learn