Why Partial Credibility Has a Cap
A limited-fluctuation standard specifies how much exposure is needed for an estimate to meet a stated precision and probability requirement. When the experience has less than that exposure, the square-root rule assigns a partial credibility factor. When the required exposure has been reached, the experience qualifies for full credibility under that standard. Additional exposure does not make the factor greater than full credibility, so the factor is capped at one.
The calculation is short, but two decisions matter: use exposure measured on the same basis as the full standard, and compare actual exposure with the standard before reporting the credibility factor. This tutorial focuses on those decisions and on the interpretation of the cap. It is aligned with the SOA Exam ASTAM (Advanced Short-Term Actuarial Mathematics) syllabus.
The cap is part of the rule, not an optional adjustment. If exposure is below the standard, \(Z\) lies between zero and one. If exposure is equal to or above the standard, \(Z=1\). At exactly zero exposure, the formula gives \(Z=0\), provided the full standard is positive.
Why the Square Root Appears
The square-root relationship is based on how sampling variability changes as exposure increases. Under the standard exposure-scaling assumption, an estimate based on \(E\) units has a standard error proportional to \(1/\sqrt{E}\). This applies when the units are suitably comparable and the relevant variance scales inversely with exposure; it is not a universal rule for every data structure or every estimator.
To see the relative scale, suppose the standard error can be written as \(c/\sqrt{E}\), where \(c\) is constant for the model and measure under consideration. At the full standard, the standard error is \(c/\sqrt{E_F}\). The ratio of the full-standard standard error to the standard error at actual exposure is therefore
Thus, below the full standard, the square root of the exposure ratio gives the relative partial factor. Once \(E\) reaches \(E_F\), the full standard has been met, so the factor is set to one. The rule does not assert that \(Z\) is a probability, nor that it measures every feature of the estimate’s uncertainty.
Write down what one exposure unit represents, such as one policy-year or one machine-hour.
Convert actual exposure or the full standard if they are expressed on different bases.
If \(E\geq E_F\), the capped factor is one. Otherwise, calculate \(\sqrt{E/E_F}\).
Keep the exact ratio or square root during the calculation and round the final factor as requested.
Worked Applications
Worked Example: Exposure Below the Full Standard
A hypothetical portfolio has 144 policy-years of exposure. Its full-credibility standard is 256 policy-years. Calculate the capped partial credibility factor. State whether the cap changes the result.
Method. Compare actual exposure with the full standard, then use the square-root rule because the exposure is below the standard.
First, check the comparison:
The exposure ratio is
Taking the square root gives
Since \(0.75<1\), the cap does not change the result. The partial credibility factor is \(Z=0.7500\), rounded to four decimal places. As a check, the factor is below one because the actual exposure is below the full standard.
Common mistake avoided: Do not use the exposure ratio itself, \(144/256=0.5625\), as the factor. The rule takes the square root of that ratio.
Worked Example: Exposure Above the Full Standard
A hypothetical equipment coverage group has 5,400 covered machine-hours. The applicable full-credibility standard is 3,600 machine-hours. Calculate the capped factor and show what would happen without the cap.
Method. Calculate the uncapped square-root value to see the effect of exceeding the standard, then apply the cap.
The ratio of actual exposure to the full standard is
The uncapped square-root value is
This value exceeds one, but the capped rule gives
The capped factor is \(Z=1.0000\), rounded to four decimal places. The group exceeds the full standard by \(5{,}400-3{,}600=1{,}800\) machine-hours, so full credibility is achieved under this standard.
Common mistake avoided: Reporting \(1.2247\) as the credibility factor ignores the cap. A factor above one is the uncapped calculation, not the capped partial credibility result.
Worked Example: Converting Exposure Units Before Applying the Cap
A hypothetical book contains 30 policies, each exposed for 9 months. The full-credibility standard is 25 policy-years. Calculate the capped partial credibility factor. Use 12 policy-months as one policy-year.
Method. Convert the policy-month exposure to policy-years, compare with the standard in policy-years, and then calculate the factor.
The total exposure in policy-months is
Convert to policy-years:
Now compare the quantities in matching units:
The exposure ratio is
Therefore,
The capped factor is \(Z=0.9487\), rounded to four decimal places. It is below one, so the cap does not change it. As a check, 22.5 policy-years is 90% of the full standard; the factor, about 94.87%, is the square root of that exposure proportion.
Common mistake avoided: Comparing 270 policy-months directly with 25 policy-years is invalid because the units differ. Converting first shows that the exposure is below, not above, the standard.
Interpreting the Cap Correctly
The cap divides the calculation into two cases. The square-root relationship determines the factor only below the full standard; the cap controls the result at and beyond that threshold.
| Exposure comparison | Uncapped square-root value | Capped factor |
|---|---|---|
| \(0\leq E<E_F\) | \(\sqrt{E/E_F}<1\) | \(\sqrt{E/E_F}\) |
| \(E=E_F\) | 1 | 1 |
| \(E>E_F\) | \(\sqrt{E/E_F}>1\) | 1 |
The factor itself is not the full credibility standard: \(E_F\) is an exposure quantity, while \(Z\) is a dimensionless multiplier. They should not be compared as though they were the same kind of number. In particular, having twice the full-standard exposure does not give \(Z=2\); it gives an uncapped value of \(\sqrt{2}\), which is then capped at one.
The square-root rule also has a scope condition. Use it when the applicable limited-fluctuation standard and exposure measure support the assumed inverse relationship between variance and exposure. If exposure units are not comparable, or the variability does not scale in that way, the simple ratio may not apply without an appropriate exposure definition or model. A cap does not correct a mismatched standard or an unsuitable exposure measure.
Common Exam Traps
- Forgetting the cap. If \(E>E_F\), the uncapped square root is above one, but the reported factor is one.
- Using the ratio without its square root. Below the standard, \(E/E_F\) is not the partial factor; calculate \(\sqrt{E/E_F}\).
- Reversing the ratio. Use actual exposure divided by the full standard, not the other way around.
- Mixing exposure units. Convert both quantities to the same basis before dividing.
- Treating the factor as a probability. It is an exposure-based credibility multiplier under the stated rule; it is not a probability of correctness.
A reliable final check is to ask whether the result is between zero and one, whether it equals one whenever \(E\geq E_F\), and whether a below-standard factor is the square root of the correctly oriented exposure ratio. For a below-standard exposure, increasing \(E\) increases \(Z\); for an exposure at or above the standard, the capped value remains one.
Check Your Understanding
For each problem, identify whether the full standard has been met, and apply the cap where needed.
- A hypothetical portfolio has 81 exposure units and a full standard of 144 units. Find the capped factor, rounded to four decimal places. (Answer: \(Z=0.7500\).)
- A hypothetical risk group has 500 exposure units and a full standard of 400 units. Find the uncapped square-root value and the capped factor. (Answer: Uncapped value \(=\sqrt{1.25}\approx1.1180\); \(Z=1.0000\).)
- A hypothetical group has 18 policy-years of exposure and a full standard of 32 policy-years. Find \(Z\), rounded to four decimal places. (Answer: \(Z=\sqrt{18/32}=0.7500\).)
- Twenty contracts are each exposed for 15 months. The full standard is 20 contract-years, and 12 contract-months equal one contract-year. Find the capped factor. (Answer: Exposure \(=20(15)/12=25\) contract-years, so \(Z=1.0000\).)
- A hypothetical book has 250 exposure units against a full standard of 625 units. Calculate the capped factor, rounded to four decimal places, and state whether the cap changes it. (Answer: \(Z=\sqrt{250/625}=\sqrt{0.4}\approx0.6325\); the cap does not change it.)