From Preference to Observation
A qualitative risk preference communicates what an organization values or wishes to avoid. It may say that the organization intends to limit reliance on a single distribution channel, contact claimants promptly, or preserve its ability to meet cash needs under stress. These statements can guide decisions, but they do not by themselves establish how anyone will determine whether actual experience is consistent with the preference.
A measurable indicator makes the preference observable. It defines what is counted or calculated, for which part of the organization, over what period, in what units, and in which direction a result is considered better. The aim is not to make every preference a complicated model. It is to specify an observation clearly enough that different users would calculate and interpret it in the same way.
An indicator is not the preference itself, nor is it automatically a decision rule. A preference explains what the organization is willing to accept or seeks to avoid. An indicator supplies evidence relevant to that preference. A separately approved criterion may then identify a result requiring attention or action. A numerical criterion cannot repair an indicator whose meaning is ambiguous.
For example, “we want to avoid over-reliance on one intermediary” does not say whether reliance is measured by premium, policy count, or another exposure basis. It also leaves open which entities, products, and time period are included. Until those choices are made, two sound-looking reports could produce different numbers and contradictory conclusions.
The Elements of a Useful Indicator
A practical definition starts with the preference and makes its observable meaning explicit. The following elements provide a disciplined check. Not every indicator needs a ratio or a statistical model, but each needs enough detail for consistent measurement.
State the organizational concern in plain language, such as limiting concentration or avoiding excessive delay.
Specify what will be counted, measured, or calculated. Avoid words such as “high,” “prompt,” or “adequate” unless the indicator defines them.
State whether the result is dollars, days, a count, a percentage, or a ratio. Define every numerator, denominator, adjustment, and exclusion.
Identify the entities, products, activities, or systems included, and specify the observation period or measurement date.
Say whether a lower or higher result is more consistent with the preference, and identify any approved criterion used to interpret the result.
The direction deserves explicit attention. For a concentration percentage, lower will often be more consistent with a preference to diversify. For a coverage ratio, higher may be more consistent with a preference to maintain a buffer. “More is better” is not a safe assumption: it depends on what is measured and how the indicator is defined.
Why Units and Denominators Matter
A count alone can be misleading when the amount of activity changes. If one period has twice as many policies or transactions as another, a higher incident count need not mean the underlying rate of incidents has worsened. A rate puts the count in relation to a clearly defined exposure base.
Suppose \(N\) eligible exposures produce \(C\) recorded events. The event proportion is \(C/N\), and its percentage form is \(100(C/N)\%\). This follows directly from the meaning of a proportion: each of the \(N\) exposures contributes one equally weighted opportunity to be in the event group, of which \(C\) are observed. Multiplying by 100 expresses that fraction per hundred exposures. The rate is meaningful only if the event definition and eligible exposure base are consistent; if exposure periods differ, the denominator may need a time unit, such as policy-years.
A denominator is not a cosmetic detail. A percentage of policies, a percentage of premium, and a percentage of claims answer different questions. Nor does a ratio remain interpretable if its numerator and denominator use different organizational scopes or periods. The indicator definition should make these choices visible rather than leaving them to the person preparing a report.
Worked Example: Measuring Reliance on an Intermediary
A hypothetical insurer has stated a preference to avoid excessive reliance on any one intermediary. For a defined domestic commercial portfolio, it chooses the share of gross written premium attributable to its largest intermediary during each calendar quarter. In one quarter, the largest intermediary generated $24 million of gross written premium, while the portfolio generated $160 million in total. The insurer’s approved decision criterion is that the indicator should not exceed 14%. Determine the indicator and whether it meets that criterion.
Method and assumptions. This is a percentage indicator, not a probability model. The supplied premium amounts are the complete hypothetical quarterly data for the stated scope. No interest rate, mortality assumption, or probability distribution is needed. Both amounts use the same gross-written-premium basis, portfolio, and calendar quarter.
Step 1: Identify the numerator and denominator. The numerator is the largest intermediary’s gross written premium, $24 million. The denominator is total gross written premium for the defined portfolio, $160 million.
Step 2: Calculate the proportion.
Step 3: Express it as a percentage. \(0.15\times100\%=15\%\). The indicator is therefore 15%.
Step 4: Compare with the approved criterion. Since \(15\%>14\%\), the result exceeds the stated criterion by 1 percentage point. Under this definition, a lower result is more consistent with the preference to avoid reliance on a single intermediary.
Check. Fifteen percent of $160 million is \(0.15\times\$160\text{ million}=\$24\text{ million}\), matching the numerator. The common mistake is to divide by a different base, such as premium from only the largest few intermediaries, or to compare a quarterly numerator with an annual denominator. Either changes the indicator rather than merely changing its presentation.
Keep the Definition Stable Across Reports
An indicator can become unusable even when its broad idea is sensible. If one reporting team counts calendar days and another counts business days, “days to contact” does not have a stable meaning. If one report includes reopened claims and another excludes them, the resulting percentages may not be comparable. The definition should state material rules before results are reported.
The same discipline applies to time. A quarterly observation should not silently be compared with an annual criterion unless the conversion or comparison method is defined. For an event that takes time to observe, the data window must also be complete enough to avoid counting eligible cases as successful simply because the observation period has not yet elapsed.
Worked Example: Defining a Timely-Contact Indicator
A hypothetical insurer expresses a preference that claimants receive prompt initial contact. It defines the indicator as the percentage of eligible claims for which first substantive contact occurs more than two business days after the claim is reported. In a quarter, 120 claims have a complete two-business-day observation window; 6 of those claims were first contacted after the two-day limit. The approved criterion is no more than 5%. Calculate and interpret the indicator.
Method and assumptions. This is a proportion of eligible claims. The 120 claims are all claims in the stated scope with a complete observation window, and the 6 late contacts are counted using the stated “more than two business days” rule. No interest rate, mortality assumption, or probability distribution is used.
Step 1: Use the defined event. The event is contact occurring more than two business days after reporting. Contact exactly two business days after reporting is not a late contact under this definition.
Step 2: Divide late claims by eligible claims.
Step 3: Convert to a percentage. \(0.05\times100\%=5\%\). Thus, 5% of eligible claims were contacted after the specified period.
Step 4: Interpret the result. The indicator equals the criterion of no more than 5%, so it meets that criterion. A lower percentage is more consistent with the stated preference.
Check. The other \(120-6=114\) claims were not late under the defined event; \(114/120=0.95=95\%\). This complements the 5% late-contact proportion. A common mistake is to count claims that lack a complete two-day observation window in the denominator. That would treat cases not yet fully observable as if they had been shown to meet the standard.
Ratios Need a Defined Stress and a Usable Denominator
Some preferences concern a buffer rather than a count or percentage. A ratio can express how much of a defined resource is available relative to a defined need. Such a ratio is interpretable only when the resource adjustments, stress assumptions, horizon, and denominator are all specified. If a denominator can be zero or negative, the organization also needs a rule for handling that case; a conventional coverage ratio is not meaningful when there is no positive net need to cover.
Worked Example: Measuring a Hypothetical 30-Day Liquidity Buffer
A hypothetical insurer states a preference to maintain sufficient liquid resources to meet cash needs under a defined 30-day stress. It measures adjusted liquid assets divided by stressed net cash outflows over 30 days. At a month-end measurement date, it has $50 million of liquid assets before a 10% haircut, stressed cash outflows of $42 million, and eligible stressed inflows of $12 million. Its approved criterion is a coverage ratio of at least 1.25. Calculate the ratio and the adjusted-asset headroom above that criterion.
Method and assumptions. The stress inputs are fixed hypothetical amounts over the same 30-day horizon and organizational scope. The 10% haircut is applied to liquid assets. Eligible inflows reduce stressed outflows dollar for dollar. No interest rate, mortality assumption, or probability distribution is assumed; the calculation is conditional on the specified stress.
Step 1: Calculate adjusted liquid assets. A 10% haircut leaves 90% of the $50 million amount:
Step 2: Calculate net stressed cash outflows.
Step 3: Calculate coverage. The denominator is positive, so the defined ratio is meaningful:
The coverage ratio is 1.50, meaning adjusted liquid assets equal 1.50 times the defined net stressed outflows. Because \(1.50\geq1.25\), the result meets the stated criterion.
Step 4: Calculate headroom in adjusted-asset dollars. At a ratio of 1.25, the required adjusted assets for $30 million of net outflows are \(1.25\times\$30\text{ million}=\$37.5\text{ million}\). Actual adjusted assets exceed this amount by \(\$45\text{ million}-\$37.5\text{ million}=\$7.5\text{ million}\).
Check. The ratio is above 1.25 by \(1.50-1.25=0.25\); \(0.25\times\$30\text{ million}=\$7.5\text{ million}\), agreeing with the dollar headroom. The common mistake is to divide the unadjusted $50 million by $30 million, giving 1.67 after rounding, and thereby ignore the specified haircut. The indicator’s calculation rule must be followed, not approximated by a more favorable measure.
Common Measurement Traps
- Leaving the scope implicit. “Company-wide” may not clarify whether subsidiaries, closed products, or outsourced activity are included. State the boundary relevant to the preference.
- Mixing units or periods. A dollar numerator and denominator must be on compatible bases; a quarterly result should not be compared with an annual criterion without a defined method.
- Reversing the direction. A larger number is not automatically worse. State whether lower or higher is more consistent with the preference and why.
- Changing the denominator. A percentage of policies and a percentage of premium can move differently. Use the denominator that matches the stated concern and keep it stable.
- Treating a proxy as the preference itself. An indicator is evidence about a preference, not proof that every dimension of the preference has been captured. Its limitations should be understood when the result is interpreted.
- Using a threshold without an approved basis. A measured value and an action criterion are different things. Do not present an illustrative or analyst-selected number as though it were an approved organizational boundary.
A well-defined indicator supports consistent reporting and meaningful discussion. Its usefulness depends less on numerical complexity than on whether its unit, scope, horizon, calculation, and direction faithfully represent the preference. Translating preferences this way is aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.
Check Your Understanding
For each situation, identify the essential calculation or definition issue and check your result against the answer.
- A hypothetical insurer monitors reliance on its largest distribution partner. During a quarter, that partner accounts for $18 million of gross written premium out of $120 million for the defined portfolio. Calculate the percentage indicator and state its direction if the preference is to limit concentration. (Answer: 15%; lower is more consistent with the preference.)
- A claims preference is translated into the percentage of eligible claims first contacted more than three business days after reporting. Of 80 claims with a complete observation window, 4 meet that event definition. Calculate the indicator and state whether it is no more than a stated 5% criterion. (Answer: \(4/80=5\%\); yes, it equals and therefore meets the no-more-than-5% criterion.)
- Under a hypothetical 30-day stress, liquid assets of $36 million receive a 20% haircut; stressed outflows are $24 million and eligible inflows are $8 million. Calculate adjusted assets, net outflows, and the coverage ratio. (Answer: adjusted assets are $28.8 million; net outflows are $16 million; coverage is \(28.8/16=1.80\).)
- A hypothetical portfolio records 15 defined incidents over 3,000 policy-years. Express the incident rate per 1,000 policy-years and state its direction if the preference is to reduce incidents. (Answer: \(15/3{,}000=0.005\) per policy-year, or 5 incidents per 1,000 policy-years; lower is more consistent with the preference.)
- A preference says that unauthorized access to a specified claims platform should be limited. Write a concise indicator definition that includes a unit, scope, and horizon, and indicate its direction. (Answer: Count unauthorized access incidents meeting a defined incident rule on the specified claims platform during each fiscal year; unit is incidents per fiscal year, scope is that platform, and lower is more consistent with the preference.)