Projecting Exposure Measures to a Future Period

Why Project Exposure Measures?

A future period can contain more or fewer units of insurance than the historical period used to estimate loss cost. A growing portfolio may have more earned vehicle-years, payroll, or insured value; a shrinking portfolio may have less. Projecting the quantity of exposure makes the expected aggregate loss estimate consistent with the period being priced or analyzed.

The exposure measure is not necessarily the number of policies in force on one date. For a time-based measure, a policy contributes exposure while coverage is active. For a value-based measure, each policy may contribute an amount that changes with its insured value. The units must be stated and used consistently: for example, earned vehicle-years, thousands of dollars of payroll, or units of $1,000 insured value.

Definition: An exposure measure, denoted here by $E_t$ for period $t$, is the quantity of insured risk measured on a specified basis over that period. A projected exposure measure is the estimate of that quantity for a future period. The measure and its units must be specified; a policy count at a point in time is not automatically the same as earned exposure over a period.

Projection of exposure is separate from projection of expected loss cost per exposure. If expected cost per exposure is unchanged and the risk mix remains comparable, more exposure increases expected aggregate losses but does not, by itself, increase expected loss cost per exposure. If the exposure units or the composition of risks change, the cost per exposure may also change and must be considered separately.

Basic Projection Mechanics

Suppose exposure in a base period is $E_0$, and the exposure measure is expected to change at a constant effective annual rate $g_E$. For a projection $n$ years ahead, with annual compounding,

$$ E_n=E_0(1+g_E)^n. $$

The rate $g_E$ must describe change in the stated exposure measure, not automatically change in policy count. If policy count and exposure per policy both change, their effects combine. For example, if the number of policies changes by a factor $1+g_N$ per year and the average exposure per policy changes by a factor $1+g_V$, then the total exposure factor is $(1+g_N)(1+g_V)$, provided the factors apply consistently to the same portfolio and period.

Projection rule: First identify the exposure basis and the period being measured. Then estimate the change in that measure for each projection interval and apply the changes in sequence. Use a compound factor when each period’s percentage change applies to the exposure resulting from the prior period.

For an exposure base made up of distinct segments, project each segment using its own exposure change and add the projected amounts. Applying one unweighted rate to the total can be wrong when segments grow at different rates. A weighted average rate can be used only if its weights and time basis correctly reproduce the segment-level projection.

Worked Projections

Worked Example: Earned Policy Exposure With Midyear Changes

A hypothetical insurer begins a year with 9,000 policies in force. During the year, 1,200 new policies are added evenly through the year, and 600 policies from the opening group terminate evenly through the year. Each policy contributes one policy-year if in force all year. Estimate this year’s earned policy exposure, then project next year’s exposure assuming annual earned exposure grows by 5%.

Method. Calculate time in force for each group, then apply the stated annual growth rate to project the next year. This is a deterministic calculation; no interest rate, mortality assumption, or probability distribution is needed. The 5% is an exposure-growth assumption, not an interest rate.

The 9,000 opening policies would contribute 9,000 policy-years if all remained in force for the whole year. New policies enter evenly, so their average time in force is one-half year. Their contribution is:

1{,}200(0.5)=600\text{ policy-years}.

The 600 terminations are from the opening group and occur evenly. Each terminated policy is absent for an average of one-half year compared with the full-year contribution already counted in the opening total. Subtract:

600(0.5)=300\text{ policy-years}.

Thus this year’s earned exposure is:

E_0=9{,}000+600-300=9{,}300\text{ policy-years}.

Next year’s projected exposure is $5\%$ greater:

E_1=9{,}300(1.05)=9{,}765\text{ policy-years}.

Check. The additions contribute 600 policy-years, while terminations reduce full-year exposure by 300; net earned exposure is therefore 300 above the opening count. Applying 5% to 9,300 gives 465 additional policy-years, for 9,765. The common mistake avoided is treating the 1,200 additions and 600 terminations as if each contributed or removed a full policy-year.

Worked Example: Projecting Total Insured-Value Exposure

A hypothetical property portfolio begins with 6,000 policies, each representing 180 units of $1,000 insured value. Over each of the next three years, policy count declines by 1% and insured value per policy grows by 3%. Project total insured-value exposure at the end of three years, in units of $1,000 insured value.

Method. Project policy count and insured value per policy separately, or multiply their annual factors. This example uses annual compounding; the percentages are exposure changes, not interest. No mortality assumption or probability distribution is required.

Initial total exposure is:

E_0=6{,}000(180)=1{,}080{,}000\text{ units}.

Each year, the policy-count factor is $1-0.01=0.99$, while the insured-value-per-policy factor is $1+0.03=1.03$. Their combined annual factor is:

$$ 0.99(1.03)=1.0197. $$

After three years, total exposure is:

\begin{aligned} E_3 &=1{,}080{,}000(1.0197)^3\\ &=1{,}080{,}000(1.0602719154)\\ &=1{,}145{,}093.6686\text{ units}. \end{aligned}

Rounded to the nearest whole unit, projected total insured-value exposure is 1,145,094 units of $1,000 insured value.

Check. The annual combined factor exceeds 1 because the 3% increase in value per policy is larger than the 1% decline in policy count. Compounding that net increase for three years gives the displayed result. A common mistake is to use only the policy-count trend or only the insured-value trend. Another is to add the percentages and use 1.02; the exact annual combined factor is 1.0197 because both changes apply multiplicatively.

From Exposure Projection to Expected Aggregate Loss

Let $L$ be aggregate loss in a period. If the portfolio is sufficiently homogeneous on the chosen exposure basis and expected loss cost per exposure unit is $c$, then the expected aggregate loss is:

$$ E[L]=E c. $$

Here $E$ is the exposure measure and $c$ is the expected loss amount per unit of that measure. This relationship is a modeling assumption about the portfolio: the selected $c$ must be appropriate for the exposure units being counted. If risk characteristics or the mix of business change, the average cost per unit may change as well.

The identity follows by dividing the covered risk into exposure units. If unit $j$ has expected loss $c_j$, linearity of expectation gives $E[L]=\sum_j c_j$. When every unit has the same expected cost $c$, this becomes $E[L]=\sum_j c=Ec$. For fractional exposure, the same formula applies when expected loss scales proportionally with the fraction of exposure. Dependence among losses does not invalidate this expectation identity, although it can affect the variance of aggregate loss.

Worked Example: Projecting Aggregate Loss With Separate Exposure and Cost Changes

A hypothetical portfolio has 12,000 earned exposure units and an expected loss cost of $240 per unit in the base period. For each of the next two years, exposure is expected to grow by 4% and expected loss cost per exposure unit by 6%. Estimate the expected aggregate loss in the second future year.

Method. Project exposure and loss cost separately, then multiply them. Annual changes are compounded, and no interest rate, mortality assumption, or probability distribution is part of this deterministic projection.

Projected exposure is:

E_2=12{,}000(1.04)^2=12{,}000(1.0816)=12{,}979.2\text{ units}.

Projected expected cost per unit is:

c_2=240(1.06)^2=240(1.1236)=\$269.664\text{ per unit}.

Therefore, the projected expected aggregate loss is:

E[L_2]=12{,}979.2(269.664)=\$3{,}500{,}022.9888.

Rounded to the nearest cent, the estimate is $3,500,022.99. As a check, base expected aggregate loss is $12{,}000(\$240)=\$2{,}880{,}000$. The combined two-year factor is $(1.04)^2(1.06)^2=1.21528576$, and $\$2{,}880{,}000(1.21528576)=\$3{,}500{,}022.9888$, matching the direct calculation.

The common mistake avoided is applying the loss-cost change but leaving exposure at 12,000 units. Exposure growth affects aggregate loss even though it does not, by itself, change the expected cost per exposure unit. Conversely, multiplying by exposure growth does not replace the separate projection of cost per unit.

When the Simple Projection Needs Care

Exposure projection is most straightforward when the exposure basis is stable and the projected changes are well-defined. In practice, the analyst should be explicit about the following:

Measure and units. State whether exposure means earned policy-years, vehicle-years, payroll units, insured-value units, or another basis. Do not combine amounts measured on different bases.
Time basis. Distinguish an in-force count at a date from exposure earned over a year. A point-in-time count does not reveal how much time policies were actually covered.
Compounding interval. Apply each percentage change for the period it describes. A three-year projection at a constant annual rate uses the annual factor three times, not once.
Changes within a unit. If insured value per policy or payroll per worker changes, project the relevant total exposure measure rather than assuming policy count alone captures the change.
Representativeness of cost per unit. A change in the number of exposure units need not change expected cost per unit. A change in risk mix, limits, or the meaning of one unit can change it, so the cost estimate may require separate adjustment.

A useful algebraic check is to write total exposure as the product of its components. For example, total insured value equals policy count times average insured value per policy. If either component changes, the total exposure change follows from the product of both factors. This also makes clear why an exposure trend is not automatically a loss-cost trend: the two quantities have different roles in the aggregate-loss calculation.

Key takeaway: Project the quantity of the stated exposure measure to the target period. Then combine it with expected loss cost per exposure unit appropriate to that period and portfolio. Exposure growth changes projected aggregate expected loss through the number or size of exposure units; it changes expected cost per unit only when the underlying per-unit risk or composition changes.

Check Your Understanding

For each problem, state the projected exposure or expected aggregate loss with the units shown. Annual rates are effective exposure or loss-cost changes, compounded annually; no interest, mortality, or probability-distribution assumption is needed.

A portfolio has 18,000 earned exposure units. The measure is expected to grow by 2.5% per year for two years. Project the exposure. (Answer: 18,911.25 exposure units)
A book begins with 4,400 units of exposure and grows by 5% annually for three years. Calculate the projected exposure, retaining two decimals. (Answer: 5,093.55 exposure units)
A hypothetical workers’ compensation portfolio has 2,500 payroll exposure units. The number of workers falls by 2% per year, while payroll per worker rises by 4% per year. Assuming the two changes apply uniformly for two years, project total payroll exposure. (Answer: 2,596.92 payroll exposure units)
A portfolio has 20,000 exposure units and expected loss cost of $175 per unit. Exposure grows by 3% and expected loss cost per unit by 5% for one year. Project expected aggregate loss. (Answer: $3,785,250.00)
A hypothetical portfolio has two segments. Segment A has 6,000 units and grows by 2% annually; segment B has 4,000 units and declines by 1% annually. Project total exposure after two years by projecting each segment separately. (Answer: 10,162.8 exposure units)

What You'll Learn