Why Exposure Volume Changes the Plan Weight
A credibility weight depends on the expected loss volume to which a specified plan formula is applied. If payroll, policy-years, insured units, or another relevant exposure measure changes, the expected loss volume may change too. The weight then needs to be recalculated from the new volume. The important distinction is that expected loss volume may change in direct proportion to exposure, while a square-root plan weight does not.
This tutorial uses a particular plan formula: the capped square-root plan. A different plan may define its weight differently, so always identify the formula the plan actually specifies before calculating. The computations below concern expected losses and plan weights; they do not treat changes in actual claims as changes in exposure.
The exposure unit and expected loss rate must match. For example, a rate stated as dollars of expected loss per $100 of payroll must be multiplied by payroll measured in $100 units—not by payroll dollars directly. If the expected loss rate, exposure definition, or risk mix changes, recalculate \(E\) using the appropriate inputs before applying the plan formula.
How the Weight Responds to a Change in Exposure
Suppose the expected loss rate \(e\) and threshold \(K\) remain unchanged, while exposure changes from \(Q_0\) to \(Q_1\). Then \(E_0=eQ_0\) and \(E_1=eQ_1\), so \(E_1/E_0=Q_1/Q_0\). Before applying the cap, the new weight is the old weight multiplied by the square root of the exposure ratio, provided the starting weight is not already affected by the cap.
If neither the old nor new weight is capped, this simplifies to \(Z_1=Z_0\sqrt{Q_1/Q_0}\). A 25% increase in exposure therefore multiplies an uncapped weight by \(\sqrt{1.25}\), not by 1.25. The simplified relation should not be used blindly when the old or new weight reaches the cap; calculate the new weight from the capped formula.
Why the Square-Root Scaling Appears
The square-root relationship is a feature of this specified plan formula. It also has a statistical motivation under a simplified model: assume exposure units are comparable, their losses are independent, and their mean and variance per unit are stable. If aggregate expected loss is proportional to exposure, aggregate variance is also proportional to exposure. The standard deviation therefore grows in proportion to the square root of exposure, while the relative standard deviation falls in proportion to the reciprocal of the square root of exposure.
More explicitly, if exposure is \(Q\), write expected aggregate loss as \(E=cQ\) and aggregate variance as \(V=s^2Q\), for fixed constants \(c>0\) and \(s^2>0\). The coefficient of variation is then \(\sqrt{V}/E=s\sqrt{Q}/(cQ)=s/(c\sqrt{Q})\). Increasing exposure by a factor \(r\) reduces this relative fluctuation by a factor \(1/\sqrt{r}\). A square-root weight consequently increases with exposure at a square-root rate. The particular threshold \(K\) and cap of 1 remain plan specifications; this argument motivates the scaling but does not establish that every credibility method must use this formula.
Use the measure specified by the plan, such as payroll in $100 units, policy-years, or insured-unit years.
Multiply the exposure amount by its corresponding expected loss rate. If the rate or risk mix changed, use the appropriate updated rate.
For the capped square-root plan, calculate \(\sqrt{E/K}\) and cap the result at 1.
With \(K\) fixed, higher expected loss volume cannot reduce the square-root weight; a proportional exposure change does not produce the same proportional weight change.
Worked Calculations
Worked Example: Payroll Grows at a Constant Expected Loss Rate
A hypothetical commercial property program uses a capped square-root plan with \(K=\$120{,}000\). Its expected loss rate is \(\$3.20\) per \(\$100\) of payroll. Payroll increases from \(\$2{,}400{,}000\) to \(\$3{,}000{,}000\), with no change in the rate or mix. Find both weights and explain how the change compares with payroll growth.
Method. Convert payroll to units of \(\$100\), calculate expected losses at each exposure level, and then apply \(Z=\min(1,\sqrt{E/K})\).
Initial payroll units are \(\$2{,}400{,}000/\$100=24{,}000\). Initial expected loss volume is \(24{,}000\times \$3.20=\$76{,}800\). Therefore, \(E_0/K=\$76{,}800/\$120{,}000=0.64\), and \(Z_0=\sqrt{0.64}=0.8000\).
New payroll units are \(\$3{,}000{,}000/\$100=30{,}000\). New expected loss volume is \(30{,}000\times \$3.20=\$96{,}000\). Thus \(E_1/K=\$96{,}000/\$120{,}000=0.80\), and \(Z_1=\sqrt{0.80}=0.8944\), rounded to four decimals.
Payroll and expected loss volume each rose by \((30{,}000-24{,}000)/24{,}000=25\%\). The weight rose by \(0.8944-0.8000=0.0944\), or 9.44 percentage points. The check is that \(Z_1/Z_0=0.8944/0.8000\approx1.1180=\sqrt{1.25}\), consistent with square-root scaling. Common mistake: multiplying the initial weight by 1.25 gives 1.0000, which incorrectly applies the exposure growth rate directly to the weight.
Worked Example: Policy Count Is Not the Same as Policy-Year Exposure
A hypothetical equipment plan has an expected loss rate of \(\$520\) per policy-year and a full-credibility threshold of \(K=\$500{,}000\). The original book has 600 policy-years. A later book has 900 policies, but each is exposed for an average of 0.75 of a year. Find the old and new weights.
Method. Use policy-years as the exposure unit, rather than treating each policy as a full year of exposure.
Original expected losses are \(600\times \$520=\$312{,}000\). The ratio to the threshold is \(\$312{,}000/\$500{,}000=0.624\), so the original weight is \(Z_0=\sqrt{0.624}=0.7899\), rounded to four decimals.
The new exposure is \(900\times0.75=675\) policy-years. New expected losses are \(675\times \$520=\$351{,}000\), giving \(E_1/K=\$351{,}000/\$500{,}000=0.702\). The new weight is \(Z_1=\sqrt{0.702}=0.8379\), rounded to four decimals.
As a check, exposure rose by \(675/600=1.125\), so the uncapped weight should rise by a factor of \(\sqrt{1.125}=1.0607\). Indeed, \(0.7899\times1.0607\approx0.8379\). Common mistake: counting 900 full policy-years would give \(\$468{,}000\) of expected losses and a weight of \(\sqrt{0.936}=0.9675\). That overstates exposure because the stated policies were not each in force for a full year.
Worked Example: Exposure and Expected Loss Rate Both Change
A hypothetical fleet program has current expected losses of \(\$200{,}000\) and \(K=\$250{,}000\). Next year, payroll is expected to fall by 30%, while the expected loss rate per payroll unit rises by 10%. Find the revised expected loss volume and weight. Assume the rate change applies to the same exposure basis and there are no other changes.
Method. Apply the exposure factor and the expected loss rate factor separately. Since expected loss volume is rate multiplied by exposure, both factors affect the result.
The exposure factor is \(1-0.30=0.70\), and the rate factor is \(1+0.10=1.10\). Thus the new expected loss volume is \(\$200{,}000\times0.70\times1.10=\$154{,}000\). The two changes combine to a volume factor of \(0.70\times1.10=0.77\), a 23% decrease, not a 30% decrease.
The revised weight is \(Z_1=\sqrt{\$154{,}000/\$250{,}000}=\sqrt{0.616}=0.7849\), rounded to four decimals. The old weight was \(Z_0=\sqrt{\$200{,}000/\$250{,}000}=\sqrt{0.80}=0.8944\). The new weight is lower, as expected, because total expected loss volume fell.
As a check, the weight ratio is approximately \(0.7849/0.8944=0.8777\), matching \(\sqrt{0.77}\approx0.8775\) up to rounding. Common mistake: applying only the 0.70 exposure factor gives \(\$140{,}000\) and ignores the higher expected loss rate. The correct calculation includes both changes.
Caps, Comparisons, and Practical Checks
The cap matters when expected loss volume reaches or exceeds \(K\). For example, if \(E=1.21K\), the uncapped square root is \(\sqrt{1.21}=1.10\), but the plan weight is 1, not 1.10. Once a weight is capped, further exposure growth does not increase it under this formula. A calculation based only on the ratio of two weights can therefore mislead if either weight is capped.
When comparing two exposure scenarios, recalculate expected loss volume for each scenario and then apply the plan formula separately. This is especially important when the loss rate changes with the prospective period or when the exposure mix changes. Do not assume a constant expected loss rate unless the problem states it or provides information supporting it.
| Change or condition | Effect on expected loss volume | Effect under the capped square-root plan |
|---|---|---|
| Exposure increases; rate and mix stay fixed | Increases in the same proportion as exposure | Increases at the square root rate, until capped |
| Exposure decreases; rate and mix stay fixed | Decreases in the same proportion as exposure | Decreases at the square root rate, unless the cap changes the comparison |
| Exposure and expected loss rate both change | Changes by the product of the exposure and rate factors | Calculate from the revised expected loss volume and apply the cap |
| Expected loss volume is at least \(K\) | May continue to increase | Weight remains 1 under this plan |
A useful audit is to write down the exposure unit, the expected loss rate per unit, the resulting expected loss dollars, the threshold, and the uncapped square root. This makes unit errors visible and prevents confusing a change in exposure with a change in the weight. In particular, policy count alone may not represent exposure when policies have different terms or exposure amounts; use the exposure measure the plan specifies.
Check Your Understanding
For each situation, calculate expected loss volume first, then apply the specified capped square-root plan. Round weights to four decimals.
- A hypothetical plan has a rate of \(\$2.80\) per \(\$100\) of payroll and \(K=\$70{,}000\). Payroll grows from \(\$1{,}500{,}000\) to \(\$2{,}400{,}000\). What are the old and new weights? (Answer: \(0.7746\) and \(0.9798\))
- A hypothetical book has 420 full policy-years and 180 policies exposed for half a year. Expected losses are \(\$600\) per policy-year, and \(K=\$340{,}000\). Find the weight. (Answer: Exposure \(=510\) policy-years; expected losses \(=\$306{,}000\); weight \(=0.9487\))
- A portfolio has expected losses of \(\$250{,}000\) and \(K=\$400{,}000\). Exposure falls by 15%, while expected losses per exposure unit rise by 20%. Find the revised expected losses and weight. (Answer: \(\$255{,}000\); \(0.7984\))
- A plan has \(K=\$100{,}000\) and current expected losses of \(\$90{,}000\). Exposure doubles with a constant expected loss rate. What is the new capped weight, and what uncapped value would be incorrect to use as the plan weight? (Answer: Capped weight \(=1.0000\); uncapped value \(=\sqrt{1.8}=1.3416\))
- A portfolio has expected losses of \(\$180{,}000\) and \(K=\$250{,}000\). Its exposure increases by 44%, with the expected loss rate unchanged. Find the old and new weights, accounting for the cap. (Answer: Old weight \(=0.8485\); new expected losses \(=\$259{,}200\); new capped weight \(=1.0000\))