Why External Indicators Need to Match the Claim
A medical-cost index, wage index, or vehicle-repair index can provide evidence about how a category of costs has changed. But a change in an index is not automatically the change in an insurer’s claim severity. The useful question is not simply, “Which index rose fastest?” It is, “Which index measures costs sufficiently similar to the costs represented in these claims, over the relevant period?”
For example, a bodily injury claim may include medical treatment and wage-replacement benefits, while a vehicle physical-damage claim may largely reflect parts and labor. A broad medical index may be a reasonable indicator for one component and a poor indicator for another. Even a closely related index can mislead if its geographic scope, covered services, or observation dates differ substantially from those of the claims.
This tutorial is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus. It focuses on evaluating medical-cost, wage, and repair indicators as evidence about severity trend; it does not treat an external index as a completed adjustment to insurance loss costs.
Reading Index Changes and Comparing Periods
Let \(I_0\) and \(I_1\) be positive levels from the same consistently defined index, observed \(t\) years apart. The total index change is measured by the ratio \(I_1/I_0\). If a constant compounded annual rate is used to summarize that change, call the rate \(g\). The model assumes that the index changes at the same compounded rate in each year over the interval:
The equation follows by dividing both sides of \(I_1=I_0(1+g)^t\) by \(I_0\), then taking the \(t\)th root. The resulting annual rate is a compact summary of the endpoints; it does not establish that the index actually increased at that same rate in every year. It is inappropriate to compare annualized rates calculated from different series or substantially different periods as if they were directly comparable evidence.
Index base levels do not need to match to compare percentage changes. For instance, an index rising from 200 to 220 has the same proportional increase as one rising from 100 to 110. The ratio, not the numerical difference in index points, is the meaningful measure of change.
Worked Example: Selecting an Indicator for Medical Severity
A hypothetical insurer is reviewing severity for a group of outpatient medical reimbursement claims. The relevant claims arise in 2021 and 2023. Three hypothetical indices have the following levels:
| Indicator | 2021 level | 2023 level | What it measures |
|---|---|---|---|
| Outpatient medical-cost index | 100 | 121 | Prices for a specified basket of outpatient services |
| Wage index | 100 | 110.25 | Average wages for a defined worker population |
| Vehicle-repair index | 100 | 144 | Prices for a defined basket of repair services and parts |
Step 1: Calculate the total changes. The medical index ratio is \(121/100=1.21\), or a 21% increase over two years. The wage index ratio is \(110.25/100=1.1025\), or a 10.25% increase. The repair index ratio is \(144/100=1.44\), or a 44% increase.
Step 2: Express each change as an annualized rate. The medical rate is \(\sqrt{1.21}-1=1.10-1=0.10\), or 10% per year. The wage rate is \(\sqrt{1.1025}-1=1.05-1=0.05\), or 5% per year. The repair rate is \(\sqrt{1.44}-1=1.20-1=0.20\), or 20% per year.
Step 3: Assess relevance. The outpatient medical-cost index is the closest match to the claims’ medical-service costs. The wage index could be relevant to wage-linked benefits, but it does not directly describe outpatient treatment prices. The repair index describes a different cost category and is not a suitable medical severity indicator merely because it rose more quickly.
Conclusion and check. The medical index indicates 21% total growth, equivalent to a 10% annualized rate for this two-year interval. The index levels are internally consistent: multiplying 100 by \(1.10^2\) gives 121. Method: compare the same index over matching dates, annualize, then judge its cost coverage. Common mistake avoided: selecting the fastest-growing index without checking what costs it represents.
Three Dimensions of Indicator Relevance
An indicator should be evaluated along at least three dimensions: coverage, timing, and relevance to the insured loss mechanism. These dimensions are related, but none substitutes for the others.
Identify which goods, services, or workers enter the index. Check whether those costs are actually represented in the claim severity under review. A general medical index may include services that the insured population rarely uses.
Identify the index observation dates and compare them with the dates relevant to the claims. A calendar-year index may not line up with the accident, policy, or other period being analyzed. An index published with a delay may also describe an earlier cost period than its release date suggests.
Ask whether changes in the indexed costs are likely to move with the costs driving the insured losses. A wage measure may be useful for a wage-linked benefit but weak for a fixed medical reimbursement. A repair index may be relevant to physical damage, but not to bodily injury treatment.
Coverage can be narrower or broader than the claims. For instance, an index of hospital charges may omit outpatient services that dominate a particular claim group. A vehicle-repair index might include both parts and labor, but its mix of vehicle types could differ from the insured vehicles. The name of an index is not enough to establish its fit; its definitions and underlying population matter.
Timing requires attention to the period represented, not just the date an index was published or downloaded. If loss observations cover a period from one quarter to another, an annual index may conceal within-year changes. In addition, two indicators should not be compared by annualized rate without considering whether they use the same start and end dates and comparable update conventions.
Worked Example: Matching a Wage Indicator to Benefits
A hypothetical group of disability claims includes a weekly benefit explicitly defined as a percentage of a worker’s covered wages, subject to a benefit cap. For a two-year review period, a wage index rises from 200 to 242. A medical-cost index rises from 150 to 181.5. Determine which indicator is more relevant to the wage-linked part of the benefit and calculate both annualized rates.
Step 1: Calculate each two-year ratio. For wages, \(242/200=1.21\), a 21% total increase. For medical costs, \(181.5/150=1.21\), also a 21% total increase.
Step 2: Annualize. Each annualized rate is \(\sqrt{1.21}-1=0.10\), or 10% per year. The equal rates do not make the two indices interchangeable; they describe equal proportional changes in different cost measures.
Step 3: Match the benefit mechanism. The wage index is more directly relevant to the part of the benefit tied to covered wages. The cap matters: once a worker’s calculated benefit reaches the cap, further wage increases may not produce the same increase in the payable benefit. The medical index does not directly measure this wage-linked benefit, even though it has the same observed growth rate.
Conclusion and check. Both series increased by a factor of 1.21 over two years, and \(200(1.10)^2=242\), while \(150(1.10)^2=181.5\). The wage indicator better matches the wage-linked benefit mechanism, subject to the cap. Method: distinguish numerical agreement from conceptual fit. Common mistake avoided: assuming equal rates mean equal relevance, or assuming an uncapped wage measure predicts capped payments exactly.
Repair Indicators and the Costs Behind Physical Damage
Vehicle-repair indicators are candidates for physical-damage severity because repair costs may depend on labor rates, replacement parts, and the mix of repair operations. Their usefulness depends on whether the index’s basket resembles the repairs represented in the insured claims. A measure dominated by older vehicles, for example, may not represent claims involving newer vehicles with different parts and repair requirements.
A repair indicator can also be relevant to only part of a claim. A physical-damage payment may include repair costs, while a claim settlement can involve other amounts as well. Therefore, the indicator’s movement should be interpreted as evidence about the cost component it measures, not as a guaranteed description of every payment in the claim data.
Worked Example: Evaluating a Repair-Cost Index
A hypothetical insurer reviews collision severity for a book concentrated in newer passenger vehicles. A repair-cost index designed for newer vehicles records 160 in 2022 and 193.6 in 2024. A broad general-price index records 100 and 106.09 over the same dates. Assess the repair indicator and calculate the annualized changes.
Step 1: Calculate the repair-index ratio. \(193.6/160=1.21\), so the repair index increased 21% over two years. Its annualized rate is \(\sqrt{1.21}-1=10\%\).
Step 2: Calculate the general-price ratio. \(106.09/100=1.0609\), a 6.09% total increase. Its annualized rate is \(\sqrt{1.0609}-1=1.03-1=3\%\).
Step 3: Evaluate coverage and timing. The repair index covers costs more closely connected with the stated collision severity and the newer-vehicle mix. Both series cover the same two-year observation period, making their calculated rates comparable as measures of index change over that interval. The general-price index is broader and provides less specific evidence about repair costs.
Conclusion and check. The repair index increased 21% in total, or 10% annualized; the general-price index increased 6.09% in total, or 3% annualized. The arithmetic checks because \(160(1.10)^2=193.6\) and \(100(1.03)^2=106.09\). Method: compare matching periods, then prefer the indicator with closer cost coverage while noting its limits. Common mistake avoided: treating a broad price index as the best measure simply because it is familiar or widely available.
Common Errors in Interpreting External Indicators
- Using index-point differences instead of ratios. A rise from 100 to 110 and a rise from 200 to 210 are both 10 index points, but they represent different percentage changes: 10% and 5%, respectively.
- Comparing mismatched periods. An index change from January to January should not be treated as directly comparable with a claim change from April to October without considering the different interval.
- Confusing an index with the insured loss. The index measures its defined basket, not necessarily the costs, benefit design, or claim mix in the insured population.
- Using a relevant name as proof of relevance. A medical, wage, or repair label does not establish that the index’s services, workers, vehicles, or geography match the claims.
- Reading an annualized rate as a guarantee about each year. The endpoint calculation summarizes compounded change over the interval. It does not show that the index followed a smooth path between observations.
These checks matter because an external indicator is evidence to evaluate, not a substitute for understanding the claims. The earlier material on trend periods provides the framework for identifying relevant dates; indicator selection then asks whether the outside measure tracks costs plausibly connected to those claims. A close match in coverage and timing strengthens its usefulness, but does not erase differences in benefit terms, caps, populations, or claim mix.
Check Your Understanding
For each problem, calculate the requested change and assess the indicator’s fit to the hypothetical claims.
- A hypothetical outpatient-service index rises from 125 in 2022 to 151.25 in 2024. Calculate its total and annualized change, and state whether it is a plausible indicator for outpatient medical claim severity. (Answer: 21% total; 10% annualized; plausible if its service basket and population reasonably match the claims.)
- A hypothetical wage index rises from 240 to 264 over two years. Calculate the annualized rate. Would it be the more direct indicator for a benefit defined as a percentage of covered wages, or for a fixed medical reimbursement? (Answer: \(\sqrt{1.10}-1=4.88\%\) annually; more direct for the wage-defined benefit.)
- A repair index rises from 180 to 217.8 over three years. Calculate the total change and annualized rate, and identify one coverage question to check before using it for collision claims. (Answer: 21% total; \(\sqrt[3]{1.21}-1\approx 6.56\%\) annually; check whether its vehicle and repair mix matches the insured collision claims.)
- A medical index rises from 150 in January 2022 to 181.5 in January 2024. Claim observations instead run from April 2022 through October 2023. What is the index’s annualized rate over its observed interval, and why should it not automatically be assigned to the claim interval? (Answer: 10% annually over January 2022 to January 2024; the dates differ, so it does not directly measure change over the claims’ April 2022–October 2023 interval.)
- For hypothetical bodily injury claims, a wage index and a medical index both rise 10% annually over the same two years. The claims include medical payments and capped wage-replacement benefits. Explain why neither index alone necessarily represents total claim severity. (Answer: They measure different components, and the benefit cap can limit the effect of wage growth; the mix and amounts of medical and wage benefits also determine total severity.)