Why Multiple-Coverage Experience Needs Careful Design
An account may have experience under several coverages, such as property and liability. A plan can compare actual losses with expected losses separately for each coverage, or combine the experience into one account-level measure. The choice affects what the measure says: separate results preserve coverage-specific signals, while a combined result can smooth differences between coverages.
Combining is not simply adding whatever loss and exposure figures are available. The losses and benchmarks must refer to comparable periods and coverage scope, and each loss must be assigned consistently. Exposure bases may differ across coverages, so raw exposure counts are not generally additive. A useful combined measure is built from losses and expected losses expressed in compatible dollars, not from an unexamined sum of unlike units.
The expected-loss amounts are the plan benchmarks: they represent the losses expected for the covered exposures under the selected basis. The earlier material on expected losses as the plan benchmark establishes the importance of matching the benchmark to the experience. Here, the focus is how those coverage-level amounts are used when experience is combined.
Separate Ratios and the Combined Ratio
The combined ratio is an expected-loss-weighted average of the individual coverage ratios. This follows directly from substituting \(A_c=E_cR_c\) into the combined formula:
Thus, a coverage with larger expected losses has more influence on the combined result. The combined ratio is not generally the simple arithmetic average of the coverage ratios. The arithmetic average gives every coverage equal influence regardless of its expected-loss amount.
Worked Example: Finding a Combined Ratio Without Losing the Coverage Detail
A hypothetical small-business account has three coverages. The experience data, all valued at the same date, are: property actual losses of $40,000 against $50,000 expected; general liability actual losses of $30,000 against $20,000 expected; and equipment actual losses of $30,000 against $30,000 expected. Calculate each coverage ratio and the combined ratio. Explain why taking an unweighted average gives the wrong combined measure.
Method. Divide actual losses by expected losses for each coverage, then add the dollar amounts before dividing to obtain the combined ratio.
Property: \(R_{\mathrm{property}}=\$40{,}000/\$50{,}000=0.80\). General liability: \(R_{\mathrm{liability}}=\$30{,}000/\$20{,}000=1.50\). Equipment: \(R_{\mathrm{equipment}}=\$30{,}000/\$30{,}000=1.00\).
The actual total is \(\$40{,}000+\$30{,}000+\$30{,}000=\$100{,}000\). The expected total is \(\$50{,}000+\$20{,}000+\$30{,}000=\$100{,}000\). Therefore, \(R_{\mathrm{comb}}=\$100{,}000/\$100{,}000=1.00\).
The check is direct: actual total equals expected total, so the combined ratio must equal 1.00. By contrast, the unweighted average of the coverage ratios is \((0.80+1.50+1.00)/3=1.10\). That answer incorrectly gives the $20,000 expected-loss coverage the same weight as the $50,000 expected-loss coverage. The common mistake is to average the ratios rather than divide the summed losses by the summed benchmarks.
What a Combined Result Can and Cannot Say
The combined ratio answers a specific question: how do total actual losses compare with total expected losses over the selected set of coverages? It does not show whether every coverage performed similarly. In the example, the combined ratio is 1.00, while the liability ratio is 1.50 and the property ratio is 0.80. Offsetting results can make an account-level measure look neutral even when the coverage-level experience differs substantially.
If a plan applies the same credibility weight \(Z\) to a coverage ratio and uses the indicated factor \(M_c=1+Z(R_c-1)\), the separate adjustments retain that distinction. A combined factor \(M_{\mathrm{comb}}=1+Z(R_{\mathrm{comb}}-1)\) instead uses the aggregate result. This is a simplified comparison, not a claim that every plan must use this particular factor formula. Plan rules may set different weights, boundaries, or adjustable premium components.
Even when the same \(Z\) is used, separate and combined dollar premium adjustments need not agree. The result also depends on how prospective premium is distributed across coverages. If premium shares do not match expected-loss shares, applying coverage-specific factors to each premium can produce a different total adjustment from applying one factor to the total premium.
Worked Example: Different Exposure Units, One Dollar Benchmark
A hypothetical commercial account has property exposure measured in building-years and liability exposure measured in site-years. It has 150 building-years with an expected loss rate of $80 per building-year, and 30 site-years with an expected loss rate of $600 per site-year. Actual property losses are $15,000 and actual liability losses are $12,000. Calculate the expected losses and experience ratios. Explain why adding the exposure counts would not give a meaningful combined exposure.
Method. Convert each exposure to its dollar expected-loss benchmark using its own rate. Then combine the dollar losses and benchmarks.
Property expected losses are \(150\times \$80=\$12{,}000\). Its ratio is \(\$15{,}000/\$12{,}000=1.25\). Liability expected losses are \(30\times \$600=\$18{,}000\). Its ratio is \(\$12{,}000/\$18{,}000=0.6667\), rounded to four decimal places.
The combined actual losses are \(\$15{,}000+\$12{,}000=\$27{,}000\); the combined expected losses are \(\$12{,}000+\$18{,}000=\$30{,}000\). Hence, \(R_{\mathrm{comb}}=\$27{,}000/\$30{,}000=0.90\).
The same result follows from expected-loss weights: \((\$12{,}000\times1.25+\$18{,}000\times0.6667)/\$30{,}000\), using the unrounded liability ratio, equals \(0.90\). The building-year and site-year counts describe different units; \(150+30=180\) has no exposure interpretation. The common mistake is to pool unlike exposure units or their rates directly. Converting each coverage to a consistent dollar comparison first avoids that error.
Claim Attribution Is Part of the Rating Design
A coverage-specific result is meaningful only if losses are assigned to coverages using a consistent rule. A claim may involve multiple insured interests, coverages, or causes. The plan must specify how such a loss is attributed, including whether allocated amounts are split, assigned to one coverage, or treated under another stated rule. The selected rule should be applied consistently to the experience losses and align with the coverage definitions used for the benchmark.
A combined total can be less sensitive to reassignment between coverages if the total dollars remain unchanged. Separate ratios, however, can change materially. This matters when the plan applies different coverage factors, or when a coverage-level result is used to explain or review the account indication. Attribution should not be adjusted merely to make the ratios look more consistent.
Worked Example: Testing the Effect of a Disputed Allocation
A hypothetical account has two coverages with expected losses of $40,000 for Coverage A and $20,000 for Coverage B. A $20,000 claim is under review. Other assigned claims are $25,000 for A and $15,000 for B. Under the initial allocation rule, the disputed claim is assigned $20,000 to A and $0 to B. Under an alternative allocation supported by a revised investigation, it is assigned $10,000 to each coverage. Calculate the coverage ratios and combined ratio under both allocations. Assume no other claim amount changes.
Method. Keep the coverage benchmarks fixed, recalculate actual losses under each allocation, and compare the results. This isolates the effect of attribution.
Under the initial allocation, A actual losses are \(\$25{,}000+\$20{,}000=\$45{,}000\), so \(R_A=\$45{,}000/\$40{,}000=1.125\). B actual losses are \(\$15{,}000\), so \(R_B=\$15{,}000/\$20{,}000=0.75\). Total actual losses are \(\$45{,}000+\$15{,}000=\$60{,}000\). Total expected losses are \(\$40{,}000+\$20{,}000=\$60{,}000\), giving a combined ratio of 1.00.
Under the revised allocation, A actual losses are \(\$25{,}000+\$10{,}000=\$35{,}000\), so \(R_A=\$35{,}000/\$40{,}000=0.875\). B actual losses are \(\$15{,}000+\$10{,}000=\$25{,}000\), so \(R_B=\$25{,}000/\$20{,}000=1.25\). The total actual remains \(\$35{,}000+\$25{,}000=\$60{,}000\), so the combined ratio remains \( \$60{,}000/\$60{,}000=1.00\).
The check is that the disputed $20,000 is fully included once in either allocation, and the total actual losses remain $60,000. The combined result is unchanged, but the coverage results reverse direction. The common mistake is to infer that an unchanged combined ratio means attribution does not matter; it matters whenever the plan uses coverage-specific experience.
A Practical Comparison of Rating Approaches
Before choosing between separate and combined treatment, examine how the coverages relate to the plan’s purpose. Separation is more informative when coverages have distinct expected-loss patterns, experience credibility, or premium treatment. Combining can be suitable when the plan is intended to measure account-level performance and the coverage benchmarks and attribution rules support that interpretation.
| Design question | Separate coverage measures | Combined measure |
|---|---|---|
| What is compared? | Actual and expected losses within each coverage | Summed actual and expected losses across included coverages |
| How are coverage differences shown? | Directly, through individual ratios | Only indirectly; high and low results can offset |
| How are exposure units handled? | Each coverage retains its own basis | Coverage exposures must first be translated to a compatible benchmark, commonly expected-loss dollars |
| What is the main design risk? | Inconsistent assignment or incomparable coverage benchmarks | Unintended weighting, hidden differences, or an invalid aggregation |
A useful design check is to calculate both the coverage-level and combined ratios during testing, even if the final plan uses only one. The side-by-side view reveals whether the combined measure is dominated by a large expected-loss coverage, whether offsetting experience is material, and whether the chosen attribution rule changes the result in a way the plan should recognize.
Verify that actual and expected losses use the same experience period, coverage definitions, valuation basis, and included exposures.
Apply a stated, consistent rule to losses involving multiple coverages, and confirm that each included dollar is counted once.
Compute \(A_c/E_c\) by coverage and \(\sum_c A_c/\sum_c E_c\) for the proposed combined scope.
Identify which expected-loss amounts drive the combined result and check how the selected factor will apply to the prospective premium base.
Common Exam Trap: Treating Ratios as Equally Weighted
A frequent error is to average coverage ratios without considering their expected-loss amounts. In the first worked example, the arithmetic average is \((0.80+1.50+1.00)/3=1.10\), but the correct combined ratio is \(1.00\). The correct formula is the ratio of summed actual losses to summed expected losses, equivalently an expected-loss-weighted average. Another basic check is to add the actual dollar amounts carefully: \(\$40{,}000+\$30{,}000+\$30{,}000=\$100{,}000\), not $130,000. A mistaken total would contaminate every result built from it.
Combining measures is therefore a design choice, not just an arithmetic convenience. A combined measure can give a clear account-level comparison, but it can conceal coverage differences and depends on sound benchmarks, compatible periods, and defensible attribution. Separate measures preserve detail but require careful interpretation and consistent treatment of each coverage.
Check Your Understanding
For each problem, compute the requested measures and check whether the result is consistent with the stated totals.
- A hypothetical account has Coverage A actual losses of $28,800 and expected losses of $32,000; Coverage B actual losses of $57,600 and expected losses of $48,000. Find each coverage ratio and the combined ratio. (Answer: \(R_A=0.90\), \(R_B=1.20\), \(R_{\mathrm{comb}}=1.08\).)
- A hypothetical insurer measures equipment coverage in machine-years and premises liability in site-years. There are 500 machine-years at an expected loss rate of $24 per machine-year, and 80 site-years at $250 per site-year. Actual losses are $18,000 and $10,000, respectively. Find the separate and combined ratios. (Answer: Expected losses are $12,000 and $20,000; ratios are 1.50 and 0.50; combined ratio is \(28{,}000/32{,}000=0.875\).)
- Two coverages have expected losses of $60,000 and $30,000, with actual losses of $54,000 and $18,000. A plan uses credibility \(Z=0.40\) and \(M_c=1+Z(R_c-1)\). Prospective premiums are $120,000 and $80,000. Calculate each adjusted premium and the total. Then calculate the total if one combined factor is applied to the $200,000 total premium. (Answer: Ratios are 0.90 and 0.60; separate factors are 0.96 and 0.84, for adjusted premiums of $115,200 and $67,200, totaling $182,400. The combined ratio is 0.80, factor 0.92, and combined-premium result is $184,000.)
- A coverage’s actual loss total is $40,000, including a disputed $12,000 claim. Its expected losses are $32,000. The claim is moved entirely to another coverage, with no change in total account losses. What happens to the first coverage’s ratio, and what happens to the account combined ratio? (Answer: The first coverage ratio falls from \(40{,}000/32{,}000=1.25\) to \(28{,}000/32{,}000=0.875\); the account combined ratio is unchanged if total actual and expected losses remain unchanged.)
- Coverage X has expected losses of $15,000 and a ratio of 0.80; Coverage Y has expected losses of $45,000 and a ratio of 1.20. Find the combined ratio and compare it with the unweighted average. (Answer: Combined ratio is \((15{,}000(0.80)+45{,}000(1.20))/60{,}000=1.10\); the unweighted average is 1.00.)