Estimating Pure Premium Trends

Why Estimate Pure Premium Trends?

An insurer can experience changes in expected losses even when its premium rate has not changed. A useful way to measure those changes is to compare losses with the amount of insurance exposure that produced them. This comparison helps distinguish a change in loss cost per exposure from a change in the number of exposures in the portfolio.

A pure premium trend is an estimated rate of change in losses per exposure over time. It is an aggregate measure: it reflects changes in both claim frequency and average claim severity. The trend describes the observed change in the measured pure premium; by itself, it does not establish why the change occurred or guarantee that the same rate will continue.

Definition: For a defined period and a consistent group of risks, the pure premium is losses divided by the corresponding earned exposures. It is expressed in dollars per exposure unit. A pure premium trend is the estimated rate of change in that measure per unit of time.

The calculation is meaningful only when losses and exposures refer to compatible periods and comparable coverage and risk populations. Losses should be measured on a consistent basis across periods—for example, consistently developed losses if development is relevant—and the exposure unit should retain the same meaning. If the coverage, limits, claim treatment, or composition of the risks changes materially, the observed ratio may reflect those differences as well as underlying loss-cost change. The analyst should identify such limitations rather than interpret every difference as a pure time trend.

For the calculations below, losses and exposures are assumed to be consistently defined and comparable. Dollar amounts are hypothetical. Unless a problem states otherwise, losses are treated as ultimate losses on the same coverage basis, and each period’s exposures are earned exposures for that period.

Calculating Pure Premium and Annual Trend

For period $j$, let $L_j$ denote losses and $E_j$ denote earned exposures. The observed pure premium is $L_j/E_j$. When comparing two observations, the ratio of the later pure premium to the earlier pure premium is the total change factor over the interval. If the observation dates are $t_1$ and $t_2$ years apart, the equivalent annual compound trend is:

PP_j=\frac{L_j}{E_j}, \qquad r=\left(\frac{PP_2}{PP_1}\right)^{1/(t_2-t_1)}-1

Here $r$ is the annual trend rate, and the corresponding annual trend factor is $1+r$. The formula assumes that a constant compound annual rate is an appropriate summary of the change between the two observations. It does not mean the pure premium actually changed at that constant rate in every intervening year. When the interval is one year, the annual rate is simply $PP_2/PP_1-1$. When it is longer or shorter than one year, annualization is necessary to make rates comparable.

Worked Example: Annualizing a Change Across Three Years

A hypothetical insurer observes $720,000 of losses on 1,200 earned exposures in 2020 and $900,000 on 1,200 earned exposures in 2023. Assume the observations represent points exactly three years apart and all losses and exposures are comparable. Calculate each pure premium and the annualized trend. Method: calculate the pure premiums first, then annualize their ratio. This avoids treating a three-year total change as an annual rate.

2020 pure premium:

PP_{2020}=\frac{\$720{,}000}{1{,}200}=\$600 \text{ per exposure}

2023 pure premium:

PP_{2023}=\frac{\$900{,}000}{1{,}200}=\$750 \text{ per exposure}

The total change factor over three years is $750/600=1.25$, a 25% increase over the full interval. The annualized trend is:

r=(1.25)^{1/3}-1 =1.077217\ldots-1 =0.077217\ldots

Rounded to four decimal places as a rate, the annual trend is 7.7217%. As a check, applying the annual factor for three years gives $\$600(1.077217)^3$, approximately $\$750$ per exposure. The annual rate is less than 25% because it is a compound annual rate, not the total three-year increase.

With more than two observations, first calculate the pure premium separately for each period. Then estimate a trend using a method appropriate to the data, such as comparing selected endpoints or fitting a time trend to the series. A straight average of annual percentage changes and a compound rate between endpoints answer different questions; the method should match the intended summary. The periods must also be associated with appropriate dates, such as period midpoints when the observations represent full-year experience.

Worked Example: Pure Premiums Across Three Consecutive Periods

A hypothetical specialty insurer records the following annual experience. Each period is assumed to be comparable to the others, and the observations are one year apart. Calculate the pure premium for each year and the compound annual trend from the first period to the last. Method: calculate each period’s loss per exposure, then use the two-year endpoint interval. This avoids dividing a total change by the number of years as if it were a simple arithmetic change.

Year	Losses	Earned exposures
2021	$400,000	800
2022	$550,000	1,000
2023	$726,000	1,200

The three pure premiums are:

PP_{2021}=\frac{\$400{,}000}{800}=\$500, \qquad PP_{2022}=\frac{\$550{,}000}{1{,}000}=\$550, \qquad PP_{2023}=\frac{\$726{,}000}{1{,}200}=\$605.

There are two years between the 2021 and 2023 observations. The endpoint change factor is $605/500=1.21$, so:

r=(1.21)^{1/2}-1=1.10-1=0.10.

The annualized endpoint trend is 10.00%. The 2022 pure premium is $550, which is 10% above $500; the 2023 pure premium is $605, which is also 10% above $550. Thus, in this particular example, each observed annual change agrees with the endpoint compound trend. That agreement is not guaranteed in other data. The endpoint calculation summarizes the first-to-last change; it does not replace checking the intervening observations.

Why Pure Premium Combines Frequency and Severity

Claim frequency is the number of claims per exposure, and average claim severity is the average loss amount per claim. For a period with $N$ claims, positive claim count, total losses $L$, and exposures $E$, the pure premium can be expressed in terms of these two quantities. This identity clarifies what the aggregate measure contains; it does not require estimating frequency and severity trends separately.

\frac{L}{E} = \frac{N}{E}\times\frac{L}{N} = \text{claim frequency per exposure} \times \text{average severity}

The justification follows by cancellation: multiplying $N/E$ by $L/N$ cancels the claim count $N$, leaving $L/E$. Thus, a change in the number of claims per exposure, a change in average loss per claim, or both can change pure premium. If there are no claims in a period, $L/N$ is undefined; the identity in this form requires $N>0$, although $L/E$ remains defined when exposures are positive.

Worked Example: Interpreting a Trend With Frequency and Severity

A hypothetical fleet program has 1,000 exposures, 200 claims, and $500,000 of losses in 2020. In 2023 it has 1,200 exposures, 360 claims, and $864,000 of losses. Assume comparable exposures and loss definitions, and that the observation dates are three years apart. Calculate the pure premiums, verify the frequency-severity identity in each period, and estimate the annualized trend. Method: calculate frequency and severity as a check on the aggregate ratio. This avoids adding percentage changes when their effects combine multiplicatively.

In 2020, frequency is $200/1{,}000=0.20$ claims per exposure, and average severity is $\$500{,}000/200=\$2{,}500$ per claim. Their product is $0.20\times \$2{,}500=\$500$ per exposure. Directly, the pure premium is $\$500{,}000/1{,}000=\$500$.

In 2023, frequency is $360/1{,}200=0.30$ claims per exposure, and average severity is $\$864{,}000/360=\$2{,}400$ per claim. Their product is $0.30\times \$2{,}400=\$720$ per exposure. Directly, the pure premium is $\$864{,}000/1{,}200=\$720$.

The endpoint pure premium factor is $720/500=1.44$. Annualizing over three years gives:

r=(1.44)^{1/3}-1 =1.129244\ldots-1 =0.129244\ldots

Rounded to four decimal places as a rate, the annualized trend is 12.9244%. The frequency factor is $0.30/0.20=1.50$, while the severity factor is $\$2{,}400/\$2{,}500=0.96$. Their product is $1.50\times0.96=1.44$, the observed pure premium factor. The frequency increase and severity decrease therefore do not add: adding 50% and negative 4% would incorrectly suggest a 46% increase, while the actual increase is 44%.

Pooling Experience and Avoiding Misinterpretation

When several comparable segments or periods are pooled, the resulting pure premium is total losses divided by total exposures. It is an exposure-weighted average of the individual pure premiums, not generally their simple arithmetic average. To see why, write each segment’s losses as its pure premium times its exposures. Adding those losses and dividing by total exposures yields:

\frac{\sum_j L_j}{\sum_j E_j} = \frac{\sum_j E_j PP_j}{\sum_j E_j}.

This pooled value describes losses per exposure across the combined observations. It is not, by itself, a trend estimate, because it does not retain the ordering of the observations over time. A trend estimate needs time information as well as pure premium values.

Key takeaway: Divide each period’s losses by its corresponding exposures before comparing periods. Annualize the ratio over the actual time interval, and interpret the result as a change in aggregate loss cost per exposure—not as a frequency-only or severity-only trend.

Common Mistakes

Using losses alone. If the exposure count changes, comparing loss totals confounds loss-cost change with the change in the amount of business. Calculate losses per exposure first.
Using the wrong interval. A ratio from observations three years apart is a three-year factor. Convert it to an annual rate with a cube root, not by reporting the full percentage change as annual.
Averaging period pure premiums without weights. A simple average gives equal importance to periods of very different exposure volume. For a pooled pure premium, use total losses divided by total exposures.
Assuming the aggregate trend identifies its cause. Pure premium reflects both claim frequency and severity. The aggregate change alone does not reveal which component changed.
Ignoring comparability. Changes in coverage, exposure definitions, loss development, or risk mix can affect the observed ratio. A calculated trend is not automatically a like-for-like estimate of underlying change.

Check Your Understanding

For each hypothetical problem, calculate the requested measure before checking the answer.

A program has $510,000 of losses on 850 exposures in 2021 and $840,000 on 1,200 exposures in 2024. The observations are three years apart and comparable. Find both pure premiums and the annualized endpoint trend, rounded to two decimal places. (Answer: $600 and $700 per exposure; 5.27% per year.)
A hypothetical insurer reports $336,000 of losses on 560 exposures in 2022 and $504,000 on 720 exposures in 2023. Calculate both pure premiums and the one-year trend. (Answer: $600 and $700 per exposure; 16.67%.)
Pure premiums are $400 in 2020 and $450 in 2023. Assuming comparable observations three years apart, calculate the annualized trend, rounded to two decimal places. (Answer: $(450/400)^{1/3}-1$, or 4.00% per year.)
A hypothetical book has 240 claims on 1,200 exposures in 2021, with average severity of $1,800. In 2024 it has 300 claims on 1,250 exposures, with average severity of $1,650. Calculate each pure premium using frequency times severity, then estimate the annualized trend over three years. (Answer: $360 and $396 per exposure; 3.23% per year.)
Two comparable segments in one period have losses of $360,000 on 600 exposures and $540,000 on 1,200 exposures. Find the pooled pure premium and compare it with the unweighted average of the two segment pure premiums. (Answer: pooled pure premium $500 per exposure; unweighted average $525 per exposure.)

What You'll Learn