Why Inflation Measures Matter in Loss Trend Analysis
A claim that costs more dollars today than a similar claim did several years ago may reflect higher prices for goods and services, changes in the insurance costs being measured, or both. The distinction matters: a broad measure of rising prices is not automatically a measure of how much claim severity has changed. If the two are confused, a trend analysis can understate or overstate the future dollar cost of losses.
This tutorial separates general price inflation, measured by a broad price index, from insurance loss-cost inflation, which concerns changes in claim costs on a defined insurance basis. It also shows how to compare nominal severity across periods and how severity and frequency changes combine in pure premium.
General Inflation and Insurance Loss Costs Are Different Measures
A general price index summarizes changes in the prices of a broad basket of goods and services. If \(G_0\) and \(G_1\) are index values for two dates, the index's price-level change over that interval is \(G_1/G_0\), and its inflation rate is \(G_1/G_0-1\). This describes the index's basket, not the cost of every product or service.
Insurance severity is a different quantity. For a specified period, average severity is total losses divided by the corresponding number of claims, provided the loss and claim counts use consistent definitions. Its observed change is a nominal claim severity trend. That trend may reflect changes in prices relevant to claims, but it can also reflect changes in claim characteristics, coverage, or measurement. Accordingly, it should not automatically be labeled general inflation.
To express a later nominal severity in the price level of an earlier base period, use the ratio of the base-period index to the later-period index. This is appropriate only when the index dates match the severity periods and the index is being used as the stated broad inflation adjustment. It does not establish that the index is the correct predictor of insurance costs.
The resulting amount answers a specific question: how large would the later severity be after removing the price-level change measured by this index? The remaining difference from earlier severity is a relative change not explained by that broad index. It is not necessarily a pure change in the price of insurance-related services.
Worked Example: Separating General Inflation From Severity Growth
In a hypothetical liability portfolio, comparable claims have average nominal severity of $12,000 in Period 0 and $13,200 in Period 1. A broad price index rises from 250 to 265 over the same dates. Calculate general inflation, nominal severity growth, and the change in severity after adjustment for the index. Treat the periods as directly comparable, with no change in coverage or claim measurement.
Method and assumptions. Compare the index ratio with the severity ratio, then convert Period 1 severity to Period 0 dollars. These are deterministic calculations; no probability distribution or interest assumption is needed. The common mistake to avoid is subtracting percentage rates when the required comparison is a ratio of growth factors.
1. General price inflation. The index factor is \(265/250=1.06\). Thus the index inflation rate is \(1.06-1=0.06\), or 6%.
2. Nominal severity growth. The severity factor is \(13{,}200/12{,}000=1.10\), so nominal severity grew by \(1.10-1=0.10\), or 10%.
3. Convert Period 1 severity to Period 0 dollars.
Rounded to the nearest cent, Period 1 severity is $12,452.83 in Period 0 dollars. Compared with $12,000, the change is \(12{,}452.8302\ldots/12{,}000-1=0.0377358\ldots\), or 3.77% to two decimal places.
Check. The relative severity factor after removing the broad index change is \(1.10/1.06=1.0377358\ldots\), which gives the same 3.77% increase. The result is not the claim that the index caused exactly 6% of severity growth; it is a comparison using that index as the adjustment.
Relating Severity and Frequency to Pure Premium
On a consistent basis, pure premium is expected loss per exposure. If frequency is claims per exposure and severity is loss per claim, then the product of frequency and severity gives loss per exposure. The identity requires aligned exposure, claim, and loss definitions; for example, the severity used must correspond to the claims counted in the frequency.
For two periods, the ratio of pure premiums is the product of the frequency ratio and the severity ratio. This is an exact multiplicative relationship. It explains why a pure premium trend cannot be inferred from severity alone: frequency may also change. Likewise, general price inflation is not itself a pure premium trend because it does not specify the frequency change and need not equal the severity change.
Why the Pure Premium Relationship Holds
The relationship follows directly from the definitions. Let \(N\) be the number of claims, \(E\) the number of exposures, and \(L\) total losses. For positive claim counts, frequency is \(N/E\), while severity is \(L/N\). Multiplying cancels the claim count:
The right-hand side is loss per exposure, or pure premium. Applying the same identity to each period and dividing Period 1 by Period 0 gives the exact trend-factor relationship below, assuming all denominators are positive and definitions remain consistent.
If frequency and severity change by rates \(g_f\) and \(g_s\), respectively, then the pure premium change rate is \((1+g_f)(1+g_s)-1\). Adding the two rates is only an approximation when both rates are small; it omits their product \(g_fg_s\).
Worked Example: Comparing General Inflation With Pure Premium Trend
In a hypothetical property program, frequency rises from 0.080 to 0.084 claims per exposure. Average severity rises from $5,000 to $5,350. Over those same dates, a broad price index rises by 4%. Calculate the nominal pure premiums and their trend, then separate the frequency and severity factors. Assume consistent exposure and claim definitions.
Method and assumptions. Calculate pure premium from frequency times severity, and compare the results. Separately divide the severity factor by the broad price factor to express severity growth relative to that index. The common mistake to avoid is calling the 4% index increase the pure premium trend.
1. Period 0 pure premium. \(0.080\times \$5{,}000=\$400.00\) per exposure.
2. Period 1 pure premium. \(0.084\times \$5{,}350=\$449.40\) per exposure.
3. Overall nominal pure premium trend. The factor is \(\$449.40/\$400.00=1.1235\), so the increase is 12.35%.
4. Check by components. The frequency factor is \(0.084/0.080=1.05\), a 5% increase. The severity factor is \(\$5{,}350/\$5{,}000=1.07\), a 7% increase. Their product is \(1.05\times1.07=1.1235\), matching the pure premium factor.
5. Compare severity with general inflation. Relative to the broad index, the severity factor is \(1.07/1.04=1.028846\ldots\), or 2.88% growth after index adjustment, rounded to two decimal places.
Check. The 12.35% pure premium increase reflects both the 5% frequency increase and the 7% nominal severity increase. The 4% broad index is a separate comparison, not a replacement for either component.
Projecting With an Explicit Inflation Assumption
A projection should state which rate it uses and what that rate represents. If the projection is based on an assumed insurance severity trend, a broad inflation rate should not be substituted merely because it is available. Conversely, using a general index as a proxy can be a deliberate assumption, but it should be identified as such. The two rates need not be equal.
Worked Example: Comparing Two Severity Projections
A hypothetical portfolio has a current average severity of $8,000. For a two-year projection, an analyst is considering either an annual insurance severity trend assumption of 6% or an annual general price inflation assumption of 3%. Calculate both projected severities and their dollar difference. Assume each annual rate compounds once per year and remains constant for both years.
Method and assumptions. Apply each annual rate as a compound factor over two years. These are alternative deterministic assumptions, not estimates derived from a probability model. The common mistake to avoid is applying a one-year rate only once across a two-year projection.
1. Insurance severity assumption. The two-year factor is \(1.06^2=1.1236\). Projected severity is \(\$8{,}000\times1.1236=\$8{,}988.80\).
2. General inflation assumption. The two-year factor is \(1.03^2=1.0609\). Projected severity is \(\$8{,}000\times1.0609=\$8{,}487.20\).
3. Difference. \(\$8{,}988.80-\$8{,}487.20=\$501.60\). The insurance severity assumption produces the higher projected severity by $501.60.
Check. Each projected amount equals the current severity multiplied by its stated two-year factor. The comparison does not show which assumption is appropriate; it makes the consequence of the two different assumptions explicit.
Common Mistakes and Interpretation
Inflation measures are useful only when their meaning, period, and application are clear. A broad index can help put nominal amounts on a common price-level basis, but it cannot on its own explain every change in insurance losses or determine a suitable prospective trend.
- Equating index inflation with loss-cost inflation. A general index measures its own basket. Claim severity is a separate insurance measure and may change at a different rate.
- Subtracting rates instead of comparing factors. If nominal severity grows by 10% and an index grows by 6%, the index-adjusted factor is \(1.10/1.06\), not a claim that the exact residual is always obtained by subtracting rates. The subtraction gives 4 percentage points, while the exact relative increase is about 3.77%.
- Calling severity trend pure premium trend. Pure premium also depends on frequency. Use the component factors when both frequency and severity are available and consistently defined.
- Ignoring the period basis. Index and severity dates must correspond to the periods being compared. An annual inflation rate cannot be applied as though it covers a different interval without adjusting the time period.
- Overinterpreting a deflated amount. Removing the change in a broad index gives an index-adjusted comparison. It does not prove that the remaining difference is caused by one particular insurance cost or by a single underlying mechanism.
Check Your Understanding
For each problem, distinguish the stated broad price measure from the insurance loss measure rather than treating them as interchangeable.
- A hypothetical portfolio's average severity rises from $9,600 to $10,080 while a broad index rises from 310 to 322.4. Calculate the index inflation rate and the severity change after adjustment for that index. (Answer: general inflation is 4.00%; index-adjusted severity increases by approximately 0.96%.)
- In a hypothetical program, frequency changes from 0.060 to 0.063 claims per exposure and severity changes from $4,200 to $4,410. Calculate the pure premium in each period and its percentage change. (Answer: $252.00 and $277.83 per exposure; pure premium increases by 10.25%.)
- A hypothetical claim portfolio has current average severity of $12,500. If the assumed annual insurance severity trend is 4% for three years, what is projected severity? (Answer: $14,060.80.)
- A hypothetical broad index and average claim severity both rise by 6% between matched periods. What is the index-adjusted severity change, and what can you conclude about the equality of the observed rates? (Answer: the index-adjusted change is 0%; the rates are equal over these periods, but this alone does not establish that the index caused or will predict the severity change.)
- A hypothetical portfolio has frequency growth of 3% and severity growth of 8% over a period. Calculate the exact pure premium growth rate and compare it with the sum of the component rates. (Answer: \(1.03\times1.08-1=11.24\%\); the simple sum is 11.00%, which is 0.24 percentage points lower.)