Sensitivity to Trend and Development Selections

Why Test Trend and Development Selections?

A loss ratio indication depends on estimates as well as observed experience. Trend selections project losses to the period relevant to the proposed rates; development selections estimate ultimate losses from claims that are not fully developed. If either selection changes, the adjusted losses and indicated rate change can change too. Sensitivity analysis makes the size of that effect visible.

The purpose is not to declare one alternative correct merely because it produces a preferred result. Instead, calculate indications consistently under plausible alternatives, show how much the result varies, and identify which selection accounts for the larger variation. This analysis is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.

Definition: A sensitivity analysis recalculates an indication after changing one or more selected assumptions while holding the other inputs fixed. A one-factor sensitivity changes only one selection at a time. A scenario range is the difference between the largest and smallest indications among the specified scenarios; it is not, by itself, a confidence interval or a probability statement.

Set Up the Indication Before Varying Selections

For one experience period, let $L$ be selected ultimate losses before trend, $D$ the selected loss-development factor, $T$ the selected trend factor to the target period, $P^{OL}$ the premium on-level to the applicable rate level, and $q$ the permissible loss ratio. The loss ratio indication method compares adjusted losses with the premium base and permissible loss ratio.

\text{Adjusted loss ratio}=\frac{LDT}{P^{OL}}, \qquad \text{Indicated rate factor}= \frac{LDT/P^{OL}}{q}, \qquad \text{Indicated rate change}= \frac{LDT}{P^{OL}q}-1.

The rate change is stated as a decimal in the formula: for example, $0.08$ means an indicated increase of 8%. These calculations assume that the selected losses, premium, permissible loss ratio, and factors are on consistent coverage and accounting bases. The scenarios below also assume that the selections being compared are plausible alternatives, that premium and $q$ stay fixed, and that trend and development apply to the stated losses as multiplicative factors. If a factor is already incorporated in a selected loss amount, it must not be applied again.

For multiple experience periods, apply each period’s own development and trend selections to its losses, then add the adjusted losses. Divide by the corresponding total on-level premium and permissible loss ratio. This preserves the weighting created by each period’s premium rather than averaging period indications without weights.

\text{Indicated rate change} = \frac{\sum_i L_iD_iT_i} {q\sum_i P_i^{OL}}-1.

Worked Examples: Measuring Sensitivity

Worked Example: Change Trend and Development One at a Time

A hypothetical specialty property program has selected ultimate losses of $4,800,000 and on-level premium of $8,000,000. The permissible loss ratio is 0.60. The baseline development factor is 1.10 and the baseline trend factor is 1.04. The actuary wants to compare trend factors of 1.02 and 1.06, and development factors of 1.08 and 1.12. Calculate the baseline indication, each one-factor alternative, and each selection’s rate-change range.

Method. Use the loss ratio indication formula. For each alternative, change only the factor under study; retain the baseline value of the other factor. Premium and the permissible loss ratio also remain fixed.

Baseline. The adjusted losses are $4{,}800{,}000 \times 1.10 \times 1.04 = \$5{,}491{,}200$. The adjusted loss ratio is $5{,}491{,}200/8{,}000{,}000=0.6864$. The indicated rate factor is $0.6864/0.60=1.144$, so the indicated rate change is a 14.40% increase.

Trend alternatives. With trend 1.02, adjusted losses are $4{,}800{,}000 \times 1.10 \times 1.02=\$5{,}385{,}600$. The adjusted loss ratio is $5{,}385{,}600/8{,}000{,}000=0.6732$; the rate factor is $0.6732/0.60=1.122$, giving a 12.20% increase. With trend 1.06, adjusted losses are $4{,}800{,}000 \times 1.10 \times 1.06=\$5{,}596{,}800$. The adjusted loss ratio is $0.6996$; the rate factor is $0.6996/0.60=1.166$, giving a 16.60% increase. The trend-selection range is $16.60\%-12.20\%=4.40$ percentage points.

Development alternatives. With development 1.08, adjusted losses are $4{,}800{,}000 \times 1.08 \times 1.04=\$5{,}391{,}360$. The adjusted loss ratio is $5{,}391{,}360/8{,}000{,}000=0.67392$; the rate factor is $0.67392/0.60=1.1232$, giving a 12.32% increase. With development 1.12, adjusted losses are $4{,}800{,}000 \times 1.12 \times 1.04=\$5{,}591{,}040$. The adjusted loss ratio is $0.69888$; the rate factor is $0.69888/0.60=1.1648$, giving a 16.48% increase. The development-selection range is $16.48\%-12.32\%=4.16$ percentage points.

Conclusion and check. In these specified one-factor tests, trend creates the wider range: 4.40 percentage points versus 4.16 percentage points for development. Each alternative changes only its designated factor, and the baseline rate change is between the low and high alternatives for both tests. The common mistake avoided here is changing trend and development together and then attributing the combined effect to only one selection.

Worked Example: Combine Alternatives Across Experience Periods

A hypothetical equipment-insurance portfolio has two experience periods. Period A has selected ultimate losses of $3,000,000 and on-level premium of $5,000,000; period B has losses of $2,000,000 and premium of $4,000,000. The permissible loss ratio is 0.65. The baseline selections are $D_A=1.20,\ T_A=1.05,\ D_B=1.10,\ T_B=1.02$. Consider a low scenario with $D_A=1.16,\ T_A=1.03,\ D_B=1.07,\ T_B=1.00$, and a high scenario with $D_A=1.24,\ T_A=1.07,\ D_B=1.13,\ T_B=1.04$. Find all three rate changes and the scenario range.

Method. Adjust the losses in each period using that period’s factors, sum the adjusted losses, and use total on-level premium of $5{,}000{,}000+4{,}000{,}000=\$9{,}000{,}000$. Do not average the two period loss ratios.

Baseline. Period A adjusted losses are $3{,}000{,}000 \times 1.20 \times 1.05=\$3{,}780{,}000$. Period B adjusted losses are $2{,}000{,}000 \times 1.10 \times 1.02=\$2{,}244{,}000$. The total is $6,024,000, so the adjusted loss ratio is $6{,}024{,}000/9{,}000{,}000=0.6693333$. The rate factor is $0.6693333/0.65=1.0297436$, and the indicated change is a 2.97% increase.

Low scenario. Period A adjusted losses are $3{,}000{,}000 \times 1.16 \times 1.03=\$3{,}584{,}400$. Period B adjusted losses are $2{,}000{,}000 \times 1.07 \times 1.00=\$2{,}140{,}000$. Total adjusted losses are $5,724,400. The adjusted loss ratio is $5{,}724{,}400/9{,}000{,}000=0.6360444$. Dividing by 0.65 gives a rate factor of $0.97853$, or a 2.15% decrease.

High scenario. Period A adjusted losses are $3{,}000{,}000 \times 1.24 \times 1.07=\$3{,}980{,}400$. Period B adjusted losses are $2{,}000{,}000 \times 1.13 \times 1.04=\$2{,}350{,}400$. Total adjusted losses are $6,330,800. The adjusted loss ratio is $6{,}330{,}800/9{,}000{,}000=0.7034222$. Dividing by 0.65 gives a rate factor of $1.082188$, or an 8.22% increase.

Conclusion and check. The rate-change range is from a 2.15% decrease to an 8.22% increase, a width of $8.22-(-2.15)=10.37$ percentage points using the unrounded results. The baseline indication, 2.97%, lies within that range. This scenario range reflects simultaneous changes to both periods’ selections; it does not identify which individual factor drives the range.

Worked Example: Identify the More Sensitive Selection

A hypothetical commercial vehicle book has selected ultimate losses of $3,600,000, on-level premium of $6,400,000, and a permissible loss ratio of 0.625. Baseline selections are development 1.15 and trend 1.06. The plausible development range is 1.12 to 1.18, while the plausible trend range is 1.03 to 1.09. Using one-factor tests, which selection produces the wider rate-change range?

Baseline. Adjusted losses are $3{,}600{,}000 \times 1.15 \times 1.06=\$4{,}388{,}400$. The adjusted loss ratio is $4{,}388{,}400/6{,}400{,}000=0.6856875$. The rate factor is $0.6856875/0.625=1.0971$, giving a 9.71% increase.

Development range. Keep trend at 1.06. At development 1.12, adjusted losses are $3{,}600{,}000 \times 1.12 \times 1.06=\$4{,}273{,}920$; the loss ratio is $0.6678$, and the rate factor is $0.6678/0.625=1.06848$, or a 6.848% increase. At development 1.18, adjusted losses are $3{,}600{,}000 \times 1.18 \times 1.06=\$4{,}502{,}880$; the loss ratio is $0.703575$, and the rate factor is $1.12572$, or a 12.572% increase. The range is $12.572\%-6.848\%=5.724$ percentage points.

Trend range. Keep development at 1.15. At trend 1.03, adjusted losses are $3{,}600{,}000 \times 1.15 \times 1.03=\$4{,}264{,}200$; the loss ratio is $0.66628125$, and the rate factor is $1.06605$, or a 6.605% increase. At trend 1.09, adjusted losses are $3{,}600{,}000 \times 1.15 \times 1.09=\$4{,}512{,}600$; the loss ratio is $0.70509375$, and the rate factor is $1.12815$, or a 12.815% increase. The range is $12.815\%-6.605\%=6.210$ percentage points.

Conclusion and check. Trend is the more sensitive selection over the specified ranges: its range is 6.210 percentage points, compared with 5.724 percentage points for development. The answer depends on the widths and locations of the tested ranges, not simply on whether trend or development is generally more important.

Why the Factors Multiply

The multiplicative structure gives a useful check on the calculations and explains why a selection’s impact can be compared directly. Hold $L$, $P^{OL}$, and $q$ fixed, and write $R$ for the indicated rate change. Then

1+R=\frac{LDT}{P^{OL}q}.

If trend changes from $T_1$ to $T_2$, while development remains fixed, the new indicated factor relative to the old one is

\frac{1+R_2}{1+R_1} = \frac{LDT_2/(P^{OL}q)}{LDT_1/(P^{OL}q)} = \frac{T_2}{T_1}.

The same cancellation shows that changing development alone from $D_1$ to $D_2$ multiplies the indicated factor by $D_2/D_1$. Thus, for a one-factor change, the relative change in $1+R$ equals the relative change in that factor. It is not generally equal to the same number of percentage points in $R$, because $R$ is the indicated factor minus one. If both selections change, the relative change in $1+R$ is $(D_2/D_1)(T_2/T_1)$, not the sum of their relative changes.

Exam Traps and Interpretation

Adding factors instead of multiplying them. If $D=1.10$ and $T=1.04$, the combined multiplier is $1.10 \times 1.04=1.144$, a 14.4% increase in adjusted losses relative to the unadjusted amount. Treating the factors as a 10% and 4% adjustment and adding them gives 14.0%, which is incorrect because it omits the interaction $0.10 \times 0.04=0.004$.
Changing multiple inputs in a one-factor test. If trend and development both change, the result measures their combined effect. It cannot isolate either selection’s individual sensitivity.
Comparing loss ratios with different denominators. Keep on-level premium and the permissible loss ratio consistent across scenarios unless the question explicitly changes them. Otherwise, the measured effect combines different sources.
Confusing percentage change with percentage-point range. A range from a 2% decrease to an 8% increase is 10 percentage points wide, not a 6% range.
Over-interpreting a scenario range. A range based on selected alternatives does not establish probabilities or statistical confidence. It describes only the outcomes under those specified assumptions.

Sensitivity analysis matters because a single indication can conceal how dependent the result is on selected assumptions. A wider range signals that the indicated rate change is more responsive to that selection over the alternatives tested. It does not, on its own, prove that the selection is wrong or determine which alternative should be chosen. Selections still require actuarial judgment and appropriate support.

Check Your Understanding

For each problem, use the specified assumptions consistently. Report rate changes to the nearest 0.01 percentage point where appropriate.

A hypothetical coverage has losses of $2,400,000, on-level premium of $5,000,000, development factor 1.25, trend factor 1.04, and permissible loss ratio 0.60. What is the indicated rate change? (Answer: 4.00% increase)
A hypothetical program has losses of $4,000,000, premium of $7,500,000, development factor 1.10, and permissible loss ratio 0.64. Compare trend factors 1.03 and 1.07, changing no other input. What is the rate-change range, from low to high, and its width? (Answer: from a 5.58% decrease to a 1.92% decrease; range width 3.67 percentage points)
A hypothetical segment has losses of $1,800,000, premium of $3,000,000, permissible loss ratio 0.60, and fixed trend factor 1.05. Compare development factors 1.08 and 1.12. What rate-change range results? (Answer: from a 13.40% increase to a 17.60% increase; range width 4.20 percentage points)
A hypothetical book has two periods. Period 1 has losses of $2,000,000, premium of $4,000,000, $D_1=1.20$, and $T_1=1.05$. Period 2 has losses of $1,500,000, premium of $3,000,000, $D_2=1.10$, and $T_2=1.02$. With permissible loss ratio 0.65, calculate the aggregate indicated rate change. (Answer: 7.63% decrease)
A hypothetical account has losses of $5,000,000, premium of $10,000,000, and permissible loss ratio 0.60. Hold all other inputs fixed at development 1.10 and trend 1.05. Development varies from 1.08 to 1.12; trend varies from 1.03 to 1.07. Which selection has the wider one-factor rate-change range, and by how much? (Answer: Trend, by 0.17 percentage point; trend range 3.67 percentage points versus development range 3.50 percentage points)

What You'll Learn