Trend as Change in Expected Loss Cost

Why Loss Costs Change Over Time

Historical losses are an important input to ratemaking, but the cost of covering a comparable exposure can change between the period when losses occurred and the period for which rates are being developed. Repair costs, medical costs, claim settlement patterns, or other conditions may change. A historical dollar of loss therefore may not represent the expected cost of a future dollar of coverage.

The basic measure for separating the cost level from the amount of business is loss cost per exposure. Trend describes a change in that expected cost per exposure over time. It is not simply the change in total losses, and it is not a premium rate change. Keeping these quantities distinct is essential when projecting historical experience for ratemaking.

Definition: Let $L_t$ be aggregate losses associated with a defined group of exposures in period $t$, and let $E_t$ be the corresponding number of exposure units. The expected loss cost per exposure is $C_t=E[L_t]/E_t$, when the expected aggregate loss and exposure basis are defined consistently. An observed estimate is $L_t/E_t$. A loss trend from period $s$ to period $t$ is a change in the expected loss cost per exposure, measured on a comparable basis. The cumulative trend factor is $T_{s,t}=C_t/C_s$, and the corresponding cumulative percentage change is $T_{s,t}-1$.

An exposure unit is the chosen measure of insurance coverage, such as one vehicle-year or one building-year. The unit must be consistent across the periods being compared. A loss cost of $800 per vehicle-year, for example, is not directly comparable with $800 per policy if policies can cover different numbers of vehicles.

The expectation in $E[L_t]$ refers to the expected loss for the defined exposures, not necessarily the observed losses in a particular period. In practice, actuaries analyze observed experience to estimate expected costs. An observed change in loss cost can be evidence about trend, but it is not automatically a pure measure of trend: changes in coverage, exposure mix, claim valuation, or other conditions can also affect the comparison.

Loss Trend, Exposure Volume, and Premium Rate Level

Three quantities can move independently. Aggregate losses depend on both the number of exposures and the cost per exposure. Premiums also depend on the rates charged. A change in one quantity should not be mistaken for a change in another.

Quantity	What it measures	What can change it
Loss cost per exposure	Losses divided by a consistent exposure measure	Changes in expected claim costs over time, or differences in the risks and terms being compared
Exposure volume	The amount of insured business	More or fewer policies, vehicles, buildings, or other exposure units
Premium rate level	The rates charged for coverage	Rate changes and the distribution of business among rates

If a portfolio has more exposures, its total expected losses can increase even when expected loss cost per exposure is unchanged. Conversely, loss cost per exposure can rise while the number of exposures falls. Premium on-leveling addresses the effect of rate changes on historical premium; it does not convert historical losses to a future loss-cost level. Loss trend and premium on-leveling therefore answer different questions and should be handled separately.

A comparison intended to measure trend should use a consistent basis: the same exposure unit, comparable coverage and limits, a consistent treatment of deductibles, and losses valued on a comparable basis. If these conditions do not hold, the difference in observed loss cost may reflect more than change over time. The estimate may need adjustments or a more carefully defined comparison before it can be interpreted as trend.

Projecting Historical Loss Cost

Ratemaking requires an estimate appropriate to the future period in which the proposed rates are expected to apply. Historical loss experience is projected because the expected cost of settling claims can change between the experience period and that future period. This is a projection of loss cost, not a projection of how many exposures the insurer will write.

When a constant annual multiplicative trend $r$ is supplied as an assumption, one year of change multiplies the cost by $1+r$. For $n$ annual intervals, the compounded projection factor is $(1+r)^n$. This calculation applies a stated assumption; it does not estimate or validate the trend rate. Choosing and estimating trend rates require analysis of appropriate data and are separate tasks.

C_{t+1}=C_t(1+r) \qquad\Longrightarrow\qquad C_{t+n}=C_t(1+r)^n

The compounding follows by repeated application of the one-period relationship. After one interval, $C_{t+1}=C_t(1+r)$. After a second interval, $C_{t+2}=C_{t+1}(1+r)=C_t(1+r)^2$. Repeating the same step $n$ times gives $C_{t+n}=C_t(1+r)^n$. The factor changes the cost per exposure; it does not change the number of exposures.

Worked Example: Separating Loss Cost Change from Portfolio Growth

A hypothetical insurer covers a comparable class of small commercial vehicles on unchanged terms. In one year, it records $468,000 of losses for 600 earned vehicle-years. In a later year, it records $720,000 for 800 earned vehicle-years. Calculate the observed loss cost per vehicle-year in each year and compare the costs. No interest rate, mortality assumption, or probability distribution is used; this is a deterministic calculation from the stated figures.

Method. Divide losses by earned exposures in each period. This converts aggregate losses to a cost per exposure, so the comparison is not driven merely by a different portfolio size.

First period:

\frac{\$468{,}000}{600\text{ vehicle-years}} =\$780\text{ per vehicle-year}.

Later period:

\frac{\$720{,}000}{800\text{ vehicle-years}} =\$900\text{ per vehicle-year}.

The observed change in cost per exposure is $(\$900-\$780)/\$780=\$120/\$780=0.153846$, or about 15.38%. The total loss amount rose by $(\$720{,}000-\$468{,}000)/\$468{,}000=53.85\%$, but that percentage includes the increase from 600 to 800 vehicle-years. The per-exposure comparison is the relevant one for describing the observed change in average cost.

Check. Multiplying each calculated cost by its exposure count returns the stated totals: $\$780(600)=\$468{,}000$ and $\$900(800)=\$720{,}000$. The calculation describes an observed change; interpreting all of it as underlying trend requires the periods to be comparable and other influences to be considered.

Common mistake avoided: Comparing total losses alone would suggest a 53.85% increase in cost. That is incorrect because the later period has more exposures. Divide by exposures before comparing loss costs.

Worked Example: Projecting a Historical Cost to a Future Period

A hypothetical insurer estimates historical loss cost at $1,250 per exposure. For planning, it assumes a constant annual loss trend of 4% and needs a cost level three annual intervals after the historical period. If the future portfolio is expected to contain 500 exposures, calculate the projected cost per exposure and the corresponding aggregate losses. No interest rate, mortality assumption, or probability distribution is used; the 4% is a stated loss-trend assumption.

Method. Compound the given annual trend for three intervals, then multiply the projected cost per exposure by the future exposure count. Keep the cost projection and the exposure volume calculation separate.

The trend factor is:

$$ (1+0.04)^3=1.04^3=1.124864. $$

The projected cost per exposure is:

\$1{,}250(1.124864)=\$1{,}406.08.

The corresponding aggregate expected losses for 500 exposures are:

\$1{,}406.08(500)=\$703{,}040.

Check. The projected cost is greater than $1,250, as expected under a positive trend assumption. Dividing $703,040 by 500 returns $1,406.08 per exposure. Money is shown to the cent; no intermediate rounding was needed.

Common mistake avoided: Applying 4% just once would give $1,300 per exposure. The assumption is annual for three intervals, so the factor is $1.04^3$, not $1.04$.

Keep Loss Trending Separate from On-Leveling

Historical premium may need an on-level adjustment to reflect a different rate level. That adjustment concerns premium. Loss trend concerns expected loss cost per exposure. Applying a premium rate-level factor to losses would mix two distinct calculations.

Worked Example: Applying Loss Trend Without Trending Premium

A hypothetical insurance portfolio has $960,000 of historical losses for 1,200 exposures. The historical loss cost is to be projected two annual intervals forward using a stated 3% annual loss trend. Separately, the premium rate level has increased by 8%; this change is relevant to on-leveling historical premium, not to the loss-cost projection. The future portfolio is expected to have 1,500 exposures. Calculate the projected loss cost per exposure and aggregate losses. No interest rate, mortality assumption, or probability distribution is used.

Method. First calculate historical losses per exposure, then apply the loss-trend factor for two intervals. Finally, multiply by the future exposure count. Do not apply the premium rate-level change to the losses.

Historical loss cost:

\frac{\$960{,}000}{1{,}200}=\$800\text{ per exposure}.

Two-interval loss-trend factor:

$$ 1.03^2=1.0609. $$

Projected loss cost and aggregate losses:

\$800(1.0609)=\$848.72\text{ per exposure}, \qquad \$848.72(1{,}500)=\$1{,}273{,}080.

Check. Dividing the projected aggregate by 1,500 gives $\$1{,}273{,}080/1{,}500=\$848.72$ per exposure. The 8% premium rate-level change does not enter either loss calculation.

Common mistake avoided: Multiplying $848.72 by 1.08 would produce $916.6176 per exposure, but that would incorrectly apply a premium rate-level factor to losses. On-level the historical premium as needed; project loss cost with the loss-trend assumption.

Assumptions and Interpretation

A trend factor is meaningful only in relation to a defined cost measure and time interval. Before using one, identify the historical cost level, the target cost level, the exposure unit, and the assumed trend. If the cost measure or the periods are not comparable, a precise calculation can still give a misleading answer.

A constant trend assumption is a simplifying projection model: it assumes the same multiplicative change in each interval. Actual loss costs need not change smoothly or at a constant rate. A calculated projection is therefore conditional on the stated assumption, rather than a guarantee of future experience.

For ratemaking, the purpose is to express historical expected losses at a cost level appropriate to the future rating period. This makes losses more relevant to the period whose rates are being considered. Exposure volume can then be handled on its own basis, while premium rate-level effects are addressed separately. This separation helps prevent a change in the amount of business or in charged rates from being misread as a change in expected loss cost.

Key takeaway: Loss trend is change in expected loss cost per exposure over time. Calculate or project the cost per exposure on a comparable basis, apply only the relevant loss-trend assumption, and keep exposure volume and premium rate-level adjustments distinct.

Check Your Understanding

For each problem, distinguish loss cost per exposure from aggregate loss and apply only the stated assumptions.

A hypothetical class has $315,000 of losses for 450 exposures in one period and $520,000 for 650 exposures in a later period. Calculate each observed loss cost per exposure and the percentage change in that cost. (Answer: $700 and $800 per exposure; an increase of 14.29%.)
A historical loss cost is $920 per exposure. A constant annual trend of 3% is assumed for two annual intervals, and the future portfolio has 900 exposures. Find the projected cost per exposure and aggregate expected losses. (Answer: $976.03 per exposure; $878,425.20 aggregate.)
A portfolio has historical losses of $1,400,000 for 2,000 exposures. Apply a stated 4% annual loss trend for two intervals, then project to 2,500 future exposures. Find the projected loss cost and aggregate losses. (Answer: $757.12 per exposure; $1,892,800 aggregate.)
A hypothetical insurer records $500,000 of losses for 1,000 exposures in one period and $650,000 for 1,300 exposures in another. Is the change in aggregate losses evidence by itself of an increase in observed loss cost per exposure? Calculate both costs and explain. (Answer: No; both are $500 per exposure, so observed loss cost per exposure is unchanged.)
Historical loss cost is $800 per exposure. A stated 5% annual loss trend applies for three annual intervals. The premium rate level has separately increased by 7%. Find the projected loss cost and state whether the 7% belongs in that calculation. (Answer: $926.10 per exposure; no, the premium rate-level change does not apply to losses.)

What You'll Learn