Developing Losses for Ratemaking

Why Develop Historical Losses?

Historical losses are often evaluated before every claim from an origin period has reached its final amount. To use those losses in a prospective analysis, an actuary may estimate what they will ultimately be by applying selected loss development factors. The purpose is to account for expected future claim emergence, not to change the historical losses for inflation or other trends.

A factor is useful only when it is applied to the loss measure for which it was selected. Paid losses and incurred losses develop differently: paid losses reflect amounts already paid, while incurred losses generally include paid amounts plus case estimates for amounts still owed. Applying a paid factor to incurred losses, or vice versa, can produce a materially misleading estimate.

Definition: A loss development factor (LDF) is a ratio used to project losses from one development age to a later age. An age-to-age factor relates cumulative losses at two successive ages. An age-to-ultimate factor relates cumulative losses at a given age to estimated ultimate losses. The age-to-ultimate factor may include a tail factor for development beyond the last observed age.

From Age-to-Age Factors to Ultimate

Let $C_{a,j}$ denote cumulative losses for origin period $a$, evaluated at development age $j$. An origin period could be an accident year or policy year; the chosen definition must be consistent throughout the analysis. If $f_j$ is the selected age-to-age factor from age $j$ to the next age, then the projected cumulative losses at that next age are $C_{a,j}f_j$.

To project from age $j$ through several later ages to ultimate, multiply the intervening age-to-age factors. If the last observed age is $m$ and the selected tail factor is $t_m$, the age-to-ultimate factor is:

F_{j\to U}=\left(\prod_{k=j}^{m-1} f_k\right)t_m

Here, $F_{j\to U}$ is the age-to-ultimate factor for losses currently evaluated at age $j$. If no additional tail development is assumed, the tail factor is 1. The developed loss estimate is:

\widehat{U}_a=C_{a,j}F_{j\to U}

This multiplication follows from the meaning of each factor. For example, if losses at age $j$ are projected to age $j+1$ by multiplying by $f_j$, and losses at age $j+1$ are projected to age $j+2$ by multiplying by $f_{j+1}$, then:

C_{a,j+2}=(C_{a,j}f_j)f_{j+1} =C_{a,j}(f_jf_{j+1})

Continuing the same reasoning through the final age, and then multiplying by the tail factor if one is selected, gives the product in the age-to-ultimate formula. The product applies to age-to-age factors. If a table already gives age-to-ultimate factors, use the applicable factor directly; multiplying those cumulative factors together would count development more than once.

Identify the loss measure and evaluation age.
Establish whether the data are paid or incurred, the origin period, the valuation date, and the development age represented by each loss amount.

Select matching factors.
Use factors developed for the same loss measure and a sufficiently consistent basis, such as gross or net of reinsurance where relevant.

Build the age-to-ultimate factor.
Multiply the applicable age-to-age factors and any selected tail factor. Do not multiply again if the factor provided is already age-to-ultimate.

Develop and check the losses.
Multiply each current cumulative loss by its age-to-ultimate factor, then check the arithmetic, units, and basis before aggregating.

Basis and Assumptions

A selected factor is an estimate based on an observed or assumed pattern of loss emergence. Applying it assumes that the subject losses are reasonably represented by that pattern. In practice, the actuary considers whether the data and selected factors are comparable in matters such as paid versus incurred basis, gross versus net basis, coverage, claim handling, and the definition of the origin period. A factor may be inappropriate if important differences make the historical development pattern unrepresentative.

Cumulative losses are amounts accumulated through a development age. An incremental amount for a single interval is not the same thing: it must not be treated as though it were cumulative. Similarly, an incurred amount typically includes paid loss and case reserves, while a paid amount does not include unpaid case estimates. The factors and loss amounts must refer to matching definitions.

Development is also distinct from trending. Development projects an observed loss amount to ultimate for its origin period. Trend, addressed separately in prospective ratemaking, adjusts losses for changes over time in claim costs or other relevant conditions. A developed loss is not automatically a trended loss.

Worked Example: Developing Incurred Losses from 12 Months

Method: Multiply cumulative incurred losses by the product of the selected incurred age-to-age factors. This example uses hypothetical data. At 12 months, an insurer has \$480,000 of cumulative incurred losses for an accident year. Selected incurred factors are 1.20 from 12 to 24 months and 1.05 from 24 months to ultimate. Estimate ultimate incurred losses.

Step 1: Check the basis. The observed amount is cumulative incurred loss, and the selected factors are incurred factors. The bases match.

Step 2: Calculate the age-to-ultimate factor.

F_{12\to U}=1.20(1.05)=1.26

Step 3: Develop the loss.

\widehat{U}=\$480{,}000(1.26)=\$604{,}800

The estimated ultimate incurred loss is \$604,800. As a check, the first projection is $\$480{,}000(1.20)=\$576{,}000$ at 24 months; applying 1.05 gives $\$576{,}000(1.05)=\$604{,}800$. The common mistake avoided is applying only the 12-to-24-month factor and stopping before ultimate.

Different Maturities and Loss Bases

Origin periods are often evaluated at different ages. Each period needs the factor for its own age, even when the loss amounts are combined later. This is especially important when comparing recent and mature periods: a recent period usually has more future development remaining.

The following example also illustrates why the loss basis must match. The numerical difference between paid and incurred factors is not an arithmetic issue; it reflects different patterns of emergence in the two measures.

Worked Example: Avoiding a Paid-versus-Incurred Mismatch

Method: Develop each loss measure with factors selected for that measure. For a hypothetical accident year, cumulative paid losses at 12 months are \$300,000, and cumulative incurred losses at the same age are \$420,000. The selected paid factors are 1.50 from 12 to 24 months and 1.20 from 24 months to ultimate. The selected incurred age-to-ultimate factor from 12 months is 1.25. Estimate ultimate paid and ultimate incurred losses, and show the result of a basis mismatch.

Ultimate paid estimate. The paid age-to-ultimate factor is:

F^{\mathrm{paid}}_{12\to U}=1.50(1.20)=1.80

Thus:

\widehat{U}^{\mathrm{paid}}=\$300{,}000(1.80)=\$540{,}000

Ultimate incurred estimate. Use the incurred factor, not the paid factor:

\widehat{U}^{\mathrm{incurred}}=\$420{,}000(1.25)=\$525{,}000

Mismatch check. Applying the paid factor to incurred losses would give:

\$420{,}000(1.80)=\$756{,}000

That \$756,000 result is not supported by the selected incurred factor. The matching-basis estimates are \$540,000 paid and \$525,000 incurred. These are different measures, so they should not be treated as two competing estimates of the same quantity without further analysis. The common mistake avoided is choosing a factor solely because its development ages appear to match, while overlooking that it was selected for a different loss basis.

Worked Example: Combining Accident Years at Different Ages

Method: Apply an incurred age-to-ultimate factor separately to each accident year, then sum the developed estimates. Suppose three hypothetical accident years have cumulative incurred losses and ages as follows: \$900,000 at 36 months for Year A, \$800,000 at 24 months for Year B, and \$700,000 at 12 months for Year C. Selected incurred factors are 1.25 from 12 to 24 months, 1.10 from 24 to 36 months, and 1.04 from 36 months to ultimate. Assume these factors apply to all three years.

Year A, evaluated at 36 months:

F_{36\to U}=1.04,\qquad \widehat{U}_A=\$900{,}000(1.04)=\$936{,}000

Year B, evaluated at 24 months:

F_{24\to U}=1.10(1.04)=1.144

Therefore, $\widehat{U}_B=\$800{,}000(1.144)=\$915{,}200$.

Year C, evaluated at 12 months:

F_{12\to U}=1.25(1.10)(1.04)=1.43

Therefore, $\widehat{U}_C=\$700{,}000(1.43)=\$1{,}001{,}000$.

The aggregate developed estimate is:

\$936{,}000+\$915{,}200+\$1{,}001{,}000 =\$2{,}852{,}200

The current cumulative incurred total is $\$900{,}000+\$800{,}000+\$700{,}000=\$2{,}400{,}000$, so the estimated additional development is $\$2{,}852{,}200-\$2{,}400{,}000=\$452{,}200$. The check confirms that every year was developed from its own age and the total exceeds the current total, consistent with all selected factors being greater than 1. The common mistake avoided is applying the same factor to every year despite their different evaluation ages.

Common Exam Traps

Confusing age-to-age with age-to-ultimate. Multiply the successive age-to-age factors to reach ultimate. Apply a provided age-to-ultimate factor directly.
Using the wrong loss basis. Confirm whether the data are paid or incurred and whether the selected factors use the same basis.
Developing incremental losses as if they were cumulative. Verify what each reported amount represents before applying a factor.
Ignoring the evaluation age. A factor for 12 months is not interchangeable with one for 24 or 36 months.
Calling development trend. Development estimates ultimate losses for an origin period; it does not by itself adjust losses to a future cost level.
Mixing monetary units. If losses are stated in thousands, keep the factors dimensionless and report the resulting amount in thousands unless converting explicitly.

Key takeaway: Develop each historical loss amount from its actual evaluation age using factors selected for the same loss basis. Check whether factors are age-to-age or age-to-ultimate, and keep development separate from trend.

Check Your Understanding

For each problem, identify the applicable factor path before calculating. Amounts are hypothetical.

Cumulative incurred losses are \$250,000 at 24 months. The selected incurred factors are 1.08 from 24 to 36 months and 1.05 from 36 months to ultimate. What is the estimated ultimate incurred loss? (Answer: \$283,500)
Cumulative paid losses are \$180,000 at 12 months. The selected paid factors are 1.40 from 12 to 24 months, 1.10 from 24 to 36 months, and 1.02 from 36 months to ultimate. What is the estimated ultimate paid loss? (Answer: \$282,744)
Cumulative incurred losses are \$600,000 at 24 months. The incurred factor from 24 to 36 months is 1.12, and the 36-month-to-ultimate factor is 1.06. What is the ultimate estimate, and why would a paid factor not be substituted without justification? (Answer: \$712,320; a paid factor is selected for a different loss measure and may have a different development pattern.)
Year P has \$500,000 of cumulative incurred losses at 36 months; Year Q has \$400,000 at 24 months. Using incurred factors of 1.10 from 24 to 36 months and 1.05 from 36 months to ultimate, calculate the aggregate ultimate estimate. (Answer: \$987,000)
A cumulative paid amount of \$200,000 at 12 months is paired with a provided 12-month-to-ultimate paid factor of 1.62. What is the ultimate estimate, and what error would result from multiplying the 1.62 by itself? (Answer: \$324,000; squaring it would incorrectly give \$524,880.)

What You'll Learn