Trend Selections and Projection Periods

Why Trend and Projection Period Matter

Historical experience does not automatically represent the costs expected during a proposed rate period. Changes in claim frequency, claim severity, or both can shift expected losses over time. A trend selection translates an historical measure to a stated future date; the projection period specifies how long that adjustment applies. Even a defensible annual trend can produce a misleading result if the dates are misaligned or the number of years is wrong.

In a filing analysis, trend is distinct from loss development and premium on-leveling. Development adjusts losses for claims not yet fully reported or settled; on-leveling adjusts historical premiums to a consistent rate level. Once those adjustments have been made as appropriate, trend can project the selected experience from its average date to the target date. Applying the same adjustment twice, or treating trend as a substitute for another adjustment, can distort an indication.

Definition: A trend selection is an assumed rate or pattern of change in a measure, such as claim frequency, severity, pure premium, or premium, over time. A projection period is the elapsed time between the measure’s date (often an average date for a group of experience) and its target date. A projection applies the selected trend over that elapsed time.

Compound Trend and Compatible Measures

For a measure that changes by a constant annual compound rate $t$, the projection from date $a$ to later date $b$ is

$$ M_b=M_a(1+t)^n $$

Here $M_a$ is the measure at the starting date, $M_b$ is its projected value, and $n$ is the elapsed time in years. A 4% annual trend has $t=0.04$. If dates are not an integer number of years apart, use the actual fractional period when the selected convention is annual effective compounding: for example, 18 months is $n=1.5$ years. The annual factor is applied once per year, with a fractional power for a partial year.

When frequency and severity are modeled separately, pure premium per exposure is their product: frequency is claims per exposure, severity is dollars per claim, and their product is dollars per exposure. Under compatible trend assumptions, project both components and multiply them. If the annual compound trends are $t_f$ for frequency and $t_s$ for severity, the implied annual pure-premium factor is $(1+t_f)(1+t_s)$, not $1+t_f+t_s$ exactly. The sum is only an approximation that omits the product term $t_ft_s$.

A direct pure-premium trend may instead be selected and applied to pure premium as a whole. Do not apply a direct pure-premium trend on top of frequency and severity trends for the same change; that would count the change twice. Trend choices should match the measure being projected and the experience basis used in the analysis.

Assumption: The compound formula assumes the selected annual rate applies consistently throughout the projection period, and that the quantity being projected is measured on a consistent basis. If the selected trend changes over time, project through each interval using that interval’s factor. If a different convention, such as a linear dollar increment, is specified, use that convention rather than the compound formula.

Choosing Dates and Applying the Trend

The starting date is not automatically the beginning or end of an experience period. It is often an average date intended to represent when the losses or premiums in that period arose. The target date should likewise match the purpose of the projection, such as the average date of the proposed policy period. State both dates and calculate the elapsed time between them before applying a factor.

For experience from several periods, project each amount from its own average date to the same target date, then add the projected amounts. Applying one factor to a total is valid only if the amounts share the same starting date and applicable trend. This date-by-date approach preserves the differing lengths of projection.

Specify the measure.
Identify whether the selected trend applies to frequency, severity, pure premium, losses, or premiums. Check that the data have been adjusted for other relevant changes on a consistent basis.

Set the dates.
Name the average date of the experience and the target date. Convert the elapsed time to years, including fractional years where appropriate.

Apply the selected pattern.
Use the compound factor for a constant annual compound trend, or multiply the interval-specific factors when the selected rates vary over time.

Check the result.
Confirm the factor’s direction and magnitude, verify that each amount reaches the intended target date, and assess the indication’s sensitivity to reasonable alternatives.

Worked Applications

Worked Example: Combining Frequency and Severity Trends

A hypothetical insurer has a historical frequency of 0.08 claims per exposure and an average severity of $5,000 per claim. Project both measures two years forward using annual compound frequency trend of 3% and severity trend of 2%. Find the projected frequency, severity, and pure premium. Assume the trends remain constant for the two-year period. This is a deterministic projection; no interest, mortality, or stochastic loss-distribution assumption is needed.

Method and timing. Apply two years of compound trend to each measure. The common mistake to avoid is adding the trend percentages and applying a single 5% trend to pure premium; the question specifies separate frequency and severity trends.

Frequency. The two-year factor is $1.03^2=1.0609$. Therefore,

$$ 0.08(1.03)^2=0.08(1.0609)=0.084872 $$

The projected frequency is 0.084872 claims per exposure.

Severity. The two-year factor is $1.02^2=1.0404$, so

\$5{,}000(1.02)^2=\$5{,}000(1.0404)=\$5{,}202.00

The projected severity is $5,202.00 per claim.

Pure premium. Multiply projected claims per exposure by projected dollars per claim:

0.084872(\$5{,}202)=\$441.504144

Rounded to the nearest cent, the projected pure premium is $441.50 per exposure. As a check, the implied combined annual factor is $1.03(1.02)=1.0506$, and the two-year pure-premium factor is $1.0506^2=1.10376036$. The original pure premium was $0.08(\$5{,}000)=\$400$, and $\$400(1.10376036)=\$441.504144$, matching the component calculation.

Worked Example: Projecting Over a Fractional Period

A hypothetical pure premium of $750 is stated at October 1, 2024. Project it to April 1, 2026 using a constant annual compound trend of 6%. Assume the dates are exactly 18 months apart and use annual effective compounding with a fractional-year exponent. No interest, mortality, or loss-distribution assumption is involved; the stated trend is the projection assumption.

Method and timing. October 1, 2024 to October 1, 2025 is one year; October 1, 2025 to April 1, 2026 is six months. Thus $n=1.5$. A common error is to use one year because the dates cross only one anniversary, or to use two years by rounding up.

The factor is

(1.06)^{1.5}=1.091336\ldots

Using the unrounded factor for the dollar calculation,

\$750(1.06)^{1.5}=\$818.502\ldots

Rounded to the nearest cent, the projected pure premium is $818.50. The result exceeds $750, as expected under a positive trend, and the factor is greater than one year’s 1.06 factor but less than two years’ $1.06^2=1.1236$ factor. This provides a timing and reasonableness check.

Worked Example: Projecting Multiple Experience Periods and Testing Sensitivity

A hypothetical filing analysis has losses of $500,000 at an average date of July 1, 2023 and $700,000 at an average date of July 1, 2024. Project both amounts to July 1, 2026 at a selected annual compound loss trend of 4%. For a sensitivity check, repeat using 5%. Assume a target-period premium of $1,600,000 is held fixed to isolate the effect of loss trend. These are deterministic calculations; no interest, mortality, or stochastic loss-distribution assumption is needed.

Method and timing. Project each amount from its own average date. The first amount has three years to the target date; the second has two. The common mistake is to apply a single two-year factor to both amounts or to add the unprojected losses and use one arbitrary period.

At 4%, the first projected amount is

\$500{,}000(1.04)^3 =\$500{,}000(1.124864) =\$562{,}432.00

The second projected amount is

\$700{,}000(1.04)^2 =\$700{,}000(1.0816) =\$757{,}120.00

The total projected losses are $\$562{,}432+\$757{,}120=\$1{,}319{,}552.00$. Dividing by the stated target-period premium gives the projected loss ratio:

\frac{\$1{,}319{,}552}{\$1{,}600{,}000}=0.82472=82.4720\%

For the 5% sensitivity, the projection periods are unchanged:

\$500{,}000(1.05)^3=\$500{,}000(1.157625)=\$578{,}812.50 $$ $$ \$700{,}000(1.05)^2=\$700{,}000(1.1025)=\$771{,}750.00

The 5% total is $\$578{,}812.50+\$771{,}750.00=\$1{,}350{,}562.50$. Its projected loss ratio is $\$1{,}350{,}562.50/\$1{,}600{,}000=0.8441015625$, or 84.4102% to four decimal places. The alternative trend increases projected losses by $\$1{,}350{,}562.50-\$1{,}319{,}552.00=\$31{,}010.50$, and increases the projected loss ratio by about 1.9382 percentage points. The fixed premium assumption isolates loss-trend sensitivity; it does not by itself produce a full indicated rate change.

Why the Compound Formula Holds

The compound projection follows by applying the selected annual change repeatedly. After one year, $M_1=M_0(1+t)$. After a second year, the new amount is $M_2=M_1(1+t)=M_0(1+t)^2$. Repeating this argument gives $M_n=M_0(1+t)^n$ for whole years. For a fractional period under the annual-effective compounding convention, the factor $(1+t)^n$ defines the corresponding fractional-year accumulation; it is consistent with the whole-year rule because factors multiply across adjacent periods:

(1+t)^{n_1}(1+t)^{n_2}=(1+t)^{n_1+n_2}

This consistency is why the elapsed time must be measured in years under the same convention as the selected annual rate. A monthly rate, a continuously compounded trend, or a linear change would require its own matching calculation.

Interpreting Selections and Avoiding Traps

A trend selection is an assumption about the future pattern, not a mechanical fact established by the historical average alone. The actuary should make the selected measure, rate, time basis, and target date clear, and consider whether the trend is plausible for the experience and period being projected. When experience periods have different dates, projection to a common date makes the totals comparable. Sensitivity analysis helps show how much the result depends on the selection rather than implying that one uncertain estimate is exact.

Premium trends require particular care in a filing context. A premium amount may change because of rate level, exposure mix, or other effects. Historical premium on-leveling and projection of premium are separate operations with separate purposes. The analysis should state which changes are represented in each adjustment so that a rate-level change is not inadvertently counted again as trend.

Wrong period: Using whole years when the dates are 18 months apart omits six months of trend. Calculate elapsed time from the stated dates.
Wrong order: Adding frequency and severity percentages is not exactly equivalent to multiplying their factors. Use $(1+t_f)(1+t_s)$ for the combined annual factor.
Wrong starting date: Applying one factor to losses with different average dates misstates at least some projection periods. Project each amount to the target date separately.
Double counting: Do not apply a direct pure-premium trend in addition to frequency and severity trends that represent the same change.
Premature rounding: Keep full precision in factors and intermediate projected amounts, then round money or percentages at the requested final step.

Key takeaway: A trend projection is defined by the measure, selected pattern, starting date, and target date together. Verify all four, project each experience amount over its own elapsed period, and test whether a reasonable alternative selection materially changes the indication.

Check Your Understanding

For each hypothetical problem, calculate using the stated annual compound convention and round only at the final step.

A pure premium of $400 is stated at an average date and is projected two years at 5% annual compound trend. Find the projected amount. (Answer: $441.00 per exposure.)
A hypothetical book has frequency 0.12 claims per exposure and severity $8,000 per claim. Apply one year of 2% frequency trend and 3% severity trend. Find projected frequency, severity, and pure premium. (Answer: 0.1224 claims per exposure; $8,240.00 per claim; $1,008.58 per exposure.)
Project $250,000 of losses from January 1, 2022 to January 1, 2025 at 4% annual compound trend. (Answer: $281,216.00.)
A pure premium of $1,000 is projected 18 months at 4% annual compound trend. Find the factor to four decimals and the projected premium to the nearest cent. (Answer: factor 1.0606; $1,060.60 per exposure.)
A $900 amount is projected two years using either 3% or 5% annual compound trend. Find both projected amounts and the difference. Explain why the larger trend produces a larger projection. (Answer: $954.81 at 3%; $992.25 at 5%; difference $37.44. The 5% trend has the larger factor, $1.05^2>1.03^2$.)

What You'll Learn