Measure Tail Impact on Projected Ultimate Losses

Why Measure Tail Impact?

A tail factor summarizes development beyond the latest observed age. Even when two selections look close, applying them to a large cumulative loss can produce a material difference in projected ultimate losses. The useful comparison is therefore not only between the factors themselves, but also between the resulting dollar estimates.

To isolate the effect of the selection, keep the latest-age cumulative loss fixed and apply each alternative factor to that same amount. This is a controlled comparison: any difference in projected ultimate losses then comes from the selected tail factor, not from a changed loss base, valuation date, or development age. The calculations below use hypothetical loss data.

Definition: Let $C_{i,k}$ be the cumulative loss for accident year $i$, evaluated at its latest observed development age $k$. Let $T_k$ be a selected tail-to-ultimate factor from age $k$. The tail-adjusted projected ultimate loss is $U_i=C_{i,k}T_k$. The tail uplift in dollars is $U_i-C_{i,k}=C_{i,k}(T_k-1)$.

This is a deterministic multiplication, not a probability model. It assumes that the selected factor is appropriate for the loss measure and age being projected, and that the cumulative loss and factor use consistent definitions and valuation bases. No interest adjustment, mortality assumption, or probability distribution is used: the examples compare nominal loss amounts at a common valuation date. A factor such as 1.025 includes the original loss base; it represents a 2.5% increase, not a 102.5% increase.

Calculate the Impact on One Loss Base

When the latest-age cumulative loss is the same under each alternative, the comparison is direct. If $T_A$ and $T_B$ are two candidate tail factors, then

U_A=C_kT_A,\qquad U_B=C_kT_B,\qquad U_B-U_A=C_k(T_B-T_A).

The final equality follows by subtracting the two projected amounts: $C_kT_B-C_kT_A=C_k(T_B-T_A)$. It shows why the dollar impact grows in proportion to the loss base. The factor difference alone does not tell the full financial story.

Worked Example: Compare Three Tail Selections

A hypothetical book has a latest-age cumulative loss of $2,400,000. Three proposed tail-to-ultimate factors are 1.018, 1.026, and 1.041. Calculate the projected ultimate and dollar tail uplift under each selection, then compare the highest and lowest projected ultimates. Treat the amounts as nominal dollars at one valuation date; no interest or probability distribution is assumed.

Method. Apply each factor to the same $2,400,000 base. The common base ensures that the comparison isolates the tail selection.

For factor 1.018:

U_1=\$2{,}400{,}000(1.018)=\$2{,}443{,}200.

The uplift is $\$2{,}443{,}200-\$2{,}400{,}000=\$43{,}200$, or 1.8% of the latest-age loss.

For factor 1.026:

U_2=\$2{,}400{,}000(1.026)=\$2{,}462{,}400.

The uplift is $\$2{,}462{,}400-\$2{,}400{,}000=\$62{,}400$, or 2.6% of the latest-age loss.

For factor 1.041:

U_3=\$2{,}400{,}000(1.041)=\$2{,}498{,}400.

The uplift is $\$2{,}498{,}400-\$2{,}400{,}000=\$98{,}400$, or 4.1% of the latest-age loss. The difference between the highest and lowest projected ultimates is $\$2{,}498{,}400-\$2{,}443{,}200=\$55{,}200$. This also equals $\$2{,}400{,}000(1.041-1.018)=\$55{,}200$, which checks the calculation.

Common mistake avoided. The factor 1.041 is not a 104.1% uplift. Its uplift is $1.041-1=0.041$, or 4.1%. The projected ultimate includes the original $2,400,000 as well as the $98,400 tail uplift.

Measure Impact Across Multiple Accident Years

Different accident years may have different latest observed cumulative losses and different selected tail factors. In that case, calculate each accident year's projected ultimate with its own factor, then sum the amounts. If every year uses one common factor, that factor can be applied to the total cumulative loss. If factors vary by year, a total estimate based on an average factor must give appropriate weight to each year's loss base.

For $n$ accident years, the aggregate projected ultimate is $\sum_{i=1}^{n} C_{i,k}T_{i,k}$. If a single factor $T_{\mathrm{agg}}$ is required to reproduce that same aggregate, it is

T_{\mathrm{agg}}= \frac{\sum_{i=1}^{n}C_{i,k}T_{i,k}} {\sum_{i=1}^{n}C_{i,k}}.

This is a cumulative-loss-weighted average. It follows by setting $T_{\mathrm{agg}}\sum_i C_{i,k}$ equal to the aggregate projected ultimate and dividing by the total latest-age cumulative loss. A simple arithmetic average gives equal weight to each accident year, regardless of the size of its loss base, so it generally will not reproduce the aggregate projection.

Worked Example: Compare Low, Central, and High Selections by Year

A hypothetical portfolio has latest-age cumulative losses of $800,000, $1,250,000, and $1,600,000 for three accident years. A review considers the following alternative tail factors. Calculate the aggregate projected ultimate under each set and compare the low and high sets with the central set. Assume all amounts share the same valuation date and loss definition.

Accident year	Latest-age loss	Low factor	Central factor	High factor
Year A	$800,000	1.015	1.025	1.040
Year B	$1,250,000	1.020	1.035	1.055
Year C	$1,600,000	1.025	1.045	1.070

Method. Multiply each year's loss by the factor in the selected column, then add the three projected amounts. No common factor is imposed on the accident years.

Under the low selection, the accident-year projections are:

\$800{,}000(1.015)=\$812{,}000,\quad \$1{,}250{,}000(1.020)=\$1{,}275{,}000,\quad \$1{,}600{,}000(1.025)=\$1{,}640{,}000.

The low aggregate is $\$812{,}000+\$1{,}275{,}000+\$1{,}640{,}000=\$3{,}727{,}000$.

Under the central selection:

\$800{,}000(1.025)=\$820{,}000,\quad \$1{,}250{,}000(1.035)=\$1{,}293{,}750,\quad \$1{,}600{,}000(1.045)=\$1{,}672{,}000.

The central aggregate is $\$820{,}000+\$1{,}293{,}750+\$1{,}672{,}000=\$3{,}785{,}750$.

Under the high selection:

\$800{,}000(1.040)=\$832{,}000,\quad \$1{,}250{,}000(1.055)=\$1{,}318{,}750,\quad \$1{,}600{,}000(1.070)=\$1{,}712{,}000.

The high aggregate is $\$832{,}000+\$1{,}318{,}750+\$1{,}712{,}000=\$3{,}862{,}750$. It exceeds the central aggregate by $\$3{,}862{,}750-\$3{,}785{,}750=\$77{,}000$, or approximately 2.034% of the central aggregate. The low aggregate is $\$58{,}750$ below the central aggregate, approximately 1.552% of the central amount.

As a check, the central latest-age total is $\$800{,}000+\$1{,}250{,}000+\$1{,}600{,}000=\$3{,}650{,}000$. Its loss-weighted aggregate factor is $\$3{,}785{,}750/\$3{,}650{,}000=1.037192$ to six decimal places; multiplying that factor by $3,650,000 reproduces the central aggregate, subject to the displayed rounding.

Common mistake avoided. Averaging the three central factors arithmetically would give 1.035. Multiplying the total loss by 1.035 would produce $3,777,750, not $3,785,750. The years have different loss bases, so the arithmetic average does not preserve the year-by-year projection.

Apply a Factor Built From Tail Link Ratios

When a tail factor is formed as the product of successive annual tail link ratios, apply that product to the latest-age cumulative loss. Do not add the ratios or apply each selected factor as though it were an additional percentage on top of the full ultimate. The product represents cumulative multiplicative development beyond the observed age.

Worked Example: Compare Two Products of Annual Tail Ratios

A hypothetical latest-age cumulative loss is $2,000,000. One tail selection uses annual link ratios 1.010 and 1.020; an alternative uses 1.020 and 1.030. Calculate each combined tail factor and projected ultimate, then measure the impact of choosing the alternative. Assume the annual ratios apply sequentially beyond the same latest age; there is no discounting or stochastic distribution.

Method. Multiply the annual ratios within each selection to obtain the cumulative tail-to-ultimate factor.

For the first selection, $1.010(1.020)=1.0302$. The projected ultimate is $\$2{,}000{,}000(1.0302)=\$2{,}060{,}400$, an uplift of $60,400.

For the alternative, $1.020(1.030)=1.0506$. The projected ultimate is $\$2{,}000{,}000(1.0506)=\$2{,}101{,}200$, an uplift of $101,200.

The alternative increases the projection by $\$2{,}101{,}200-\$2{,}060{,}400=\$40{,}800$. Checking with the combined factor difference gives $\$2{,}000{,}000(1.0506-1.0302)=\$40{,}800$. Relative to the first projected ultimate, the increase is $\$40{,}800/\$2{,}060{,}400$, or approximately 1.9802%. Relative to the latest-age loss, the uplift differs by 2.04 percentage points: 5.06% rather than 3.02%.

Common mistake avoided. The second annual ratio is applied to the development produced by the first, so the ratios are multiplied. Adding them would not give the cumulative factor.

Interpret the Comparison Carefully

A tail-impact calculation answers a focused question: how much does a chosen tail factor change the projected ultimate, holding the latest-age loss fixed? It does not by itself establish which factor is most appropriate. Selection still depends on the reasonableness of the underlying tail estimate and its consistency with the loss measure, development age, and data being projected. The dollar comparison makes the consequence of a selection visible; it is not a measure of selection uncertainty.

The following distinctions help keep the calculation interpretable:

Tail factor versus tail uplift. A factor of 1.030 means a 3.0% uplift to the latest-age cumulative loss. The uplift amount is $C_k(1.030-1)$.
Dollar difference versus percentage difference. The dollar difference between two projections is $C_k(T_B-T_A)$. A percentage difference must name its denominator, such as the first projected ultimate or the latest-age cumulative loss.
Year-level versus aggregate comparison. If factors vary by year, calculate each year separately before adding. A single aggregate factor must be weighted by the respective cumulative losses.
Consistent comparison base. If the loss base changes between alternatives, the resulting difference cannot be attributed solely to the tail selection.

A frequent exam trap is to interpret a factor greater than one as the percentage added, or to compare factors without applying them to the same loss base. For example, the change from 1.020 to 1.030 is 0.010 of the base, so the dollar impact is $0.010C_k$; it is not 1.0% of the first projected ultimate unless that denominator is explicitly used. State the base, the factor, and the resulting dollar amount to make the comparison auditable. This material is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.

Key takeaway: To measure tail impact, apply every candidate tail-to-ultimate factor to the same latest-age cumulative loss. For multiple accident years, project year by year and sum; use a loss-weighted factor if an aggregate factor is needed.

Check Your Understanding

For each hypothetical situation, calculate the requested amount using the stated latest-age loss base. Amounts are nominal at a common valuation date.

A latest-age cumulative loss is $950,000 and the selected tail factor is 1.032. Find the projected ultimate and dollar tail uplift. (Answer: projected ultimate $980,400; uplift $30,400)
A $1,750,000 loss base is projected using either 1.018 or 1.029. Find both projected ultimates and the dollar difference. (Answer: $1,781,500 and $1,800,750; difference $19,250)
Two accident years have latest-age losses of $600,000 and $900,000, with respective factors 1.040 and 1.020. Find the aggregate projected ultimate, the aggregate uplift, and the loss-weighted factor. (Answer: ultimate $1,542,000; uplift $42,000; factor 1.028)
A $3,200,000 latest-age loss has two annual tail ratios of 1.012 and 1.008. An alternative selection has a combined factor of 1.025. Find the first selection's projected ultimate and the difference under the alternative. (Answer: first ultimate $3,264,307.20; alternative exceeds it by $15,692.80)
Two segments have latest-age losses of $1,100,000 and $2,400,000, with factors 1.030 and 1.015. Compare the correctly calculated aggregate ultimate with the result from applying the simple arithmetic average factor to the total loss. (Answer: correct aggregate $3,569,000; simple-average result $3,578,750, an overstatement of $9,750)

What You'll Learn