Standardizing Segment Frequencies To A Common Mix

Why Standardize Segment Frequencies?

An aggregate claim frequency can change for two different reasons: the frequency within one or more segments may change, or the portfolio may contain a different proportion of exposure in each segment. For example, a hypothetical insurer’s overall frequency could rise because each territory has more claims per exposure unit, because more exposure is now concentrated in a higher-frequency territory, or because both things happened.

To compare the underlying segment experience without letting the changing portfolio mix determine the comparison, apply the same set of exposure weights to each period’s segment frequencies. This is called standardizing to a common mix. It is a comparison of frequencies at a chosen mix, not a prediction that the portfolio actually has that mix.

Definition: For segment \(i\) in period \(t\), let \(E_{it}\) be exposure and \(C_{it}\) be claim count. The segment frequency is \(f_{it}=C_{it}/E_{it}\). Its observed exposure share is \(w_{it}=E_{it}/\sum_j E_{jt}\). The observed aggregate frequency is \(\sum_i w_{it}f_{it}\). Given fixed standard weights \(w_i^*\), where \(w_i^*\geq 0\) and \(\sum_i w_i^*=1\), the standardized frequency in period \(t\) is \(F_t^*=\sum_i w_i^* f_{it}\).

The exposure unit and the definition of a claim must be consistent across segments and periods. If frequencies are reported per 1,000 exposure units, for example, calculate each segment rate as \(1{,}000C_{it}/E_{it}\), then use those rates in the weighted average. The weights are proportions of exposure, not proportions of claims.

Observed and Standardized Frequencies

The observed aggregate frequency is the total claim count divided by total exposure. Since \(w_{it}=E_{it}/\sum_j E_{jt}\), weighting each segment frequency by its observed exposure share gives:

\sum_i w_{it}f_{it} = \sum_i \frac{E_{it}}{\sum_j E_{jt}}\frac{C_{it}}{E_{it}} = \frac{\sum_i C_{it}}{\sum_j E_{jt}}.

This identity explains why the observed-mix weighted average agrees with the pooled count-over-exposure calculation. Standardization replaces the period’s actual weights \(w_{it}\) with fixed weights \(w_i^*\), while retaining that period’s segment frequencies. The resulting difference between periods therefore reflects changes in segment frequencies at the selected common mix.

Interpretation: The standardized change from period 0 to period 1 is \(F_1^*-F_0^*=\sum_i w_i^*(f_{i1}-f_{i0})\). A positive value means frequency increased on the fixed mix; a negative value means it decreased. This result does not by itself explain the cause of the segment-level changes.

Worked Calculations

Worked Example: Comparing Two Periods With a Baseline Mix

A hypothetical portfolio has two segments. In period 0, segment A has 8,000 exposure units and 80 claims; segment B has 2,000 exposure units and 60 claims. In period 1, A has 4,000 exposure units and 48 claims; B has 6,000 exposure units and 150 claims. Use period 0’s exposure mix as the standard. Calculate both observed aggregate frequencies and period 1’s standardized frequency, all per 1,000 exposure units.

Method. Calculate the segment frequencies first, using claims divided by exposure and multiplying by 1,000. Then calculate observed totals and apply period 0’s exposure shares to period 1’s segment rates.

In period 0, the segment rates are \(1{,}000(80/8{,}000)=10\) for A and \(1{,}000(60/2{,}000)=30\) for B. Total claims are \(80+60=140\), and total exposure is \(8{,}000+2{,}000=10{,}000\). Thus the observed frequency is \(1{,}000(140/10{,}000)=14.0\).

In period 1, the segment rates are \(1{,}000(48/4{,}000)=12\) for A and \(1{,}000(150/6{,}000)=25\) for B. Total claims are \(48+150=198\), and total exposure is \(4{,}000+6{,}000=10{,}000\). Thus the observed frequency is \(1{,}000(198/10{,}000)=19.8\).

The period 0 standard weights are \(8{,}000/10{,}000=0.8\) for A and \(2{,}000/10{,}000=0.2\) for B. Period 1’s standardized frequency is therefore \(0.8(12)+0.2(25)=9.6+5.0=14.6\) claims per 1,000.

Check. The observed frequencies are 14.0 and 19.8, while the period 1 standardized frequency is 14.6. Period 0’s standardized rate under its own mix is \(0.8(10)+0.2(30)=14.0\), so the standardized within-segment change is \(14.6-14.0=0.6\) claims per 1,000.

Common mistake avoided. Taking the unweighted average of period 1’s segment rates gives \((12+25)/2=18.5\), but that incorrectly assigns half the standard exposure to each segment. The stated standard mix is 80% A and 20% B, so the correct result is 14.6.

Worked Example: Normalizing Proposed Standard Weights

A pricing analyst proposes standard weights of 0.2, 0.3, and 0.6 for three segments with frequencies of 5, 15, and 25 claims per 1,000 exposure units. Find the normalized weights and the standardized frequency.

Method. Standard weights must sum to one. Here the proposed weights sum to \(0.2+0.3+0.6=1.1\), so divide each by 1.1:

w_1^*=\frac{0.2}{1.1}=\frac{2}{11},\qquad w_2^*=\frac{0.3}{1.1}=\frac{3}{11},\qquad w_3^*=\frac{0.6}{1.1}=\frac{6}{11}.

The normalized weights sum to \(2/11+3/11+6/11=1\). The standardized frequency is \[ \frac{2}{11}(5)+\frac{3}{11}(15)+\frac{6}{11}(25) =\frac{10+45+150}{11} =\frac{205}{11} =18.6364 \] claims per 1,000, rounded to four decimal places.

Check. The weighted numerator is 205, and dividing by the total weight 11 gives \(205/11\), approximately 18.6364. Normalizing the weights leaves their relative proportions intact.

Common mistake avoided. Using the unnormalized values directly gives \(0.2(5)+0.3(15)+0.6(25)=20.5\), which is not a weighted average because the weights sum to 1.1 rather than 1. Dividing by 1.1, equivalently normalizing each weight, gives the correct result.

Worked Example: Tracking a Standardized Change Across Three Segments

A hypothetical insurer wants to compare two periods using a prospective common mix of 50%, 35%, and 15% across three segments. The respective segment frequencies are 6, 16, and 36 in period 0, and 7, 14, and 30 in period 1, all per 1,000 exposure units. Calculate both standardized frequencies and their change.

Method. The proposed weights are already valid because \(0.50+0.35+0.15=1.00\). Apply the same weights to the segment rates in each period.

For period 0: \[ F_0^*=0.50(6)+0.35(16)+0.15(36)=3.0+5.6+5.4=14.0. \] For period 1: \[ F_1^*=0.50(7)+0.35(14)+0.15(30)=3.5+4.9+4.5=12.9. \] The standardized change is \(12.9-14.0=-1.1\) claims per 1,000, a decrease of 1.1.

Check. The segment changes are \(+1\), \(-2\), and \(-6\). Weighting those changes directly gives \(0.50(1)+0.35(-2)+0.15(-6)=0.5-0.7-0.9=-1.1\), matching the difference between standardized frequencies.

Common mistake avoided. The correct comparison uses the identical weights in both periods. Substituting each period’s actual exposure shares would instead compare observed frequencies, which combine within-segment changes with any change in portfolio mix.

Separating Frequency Change From Mix Change

Standardization answers a specific question: what would each period’s aggregate frequency be if the portfolio had the same segment distribution? It does not claim that the selected mix is uniquely correct. The standard might be a baseline period’s exposure mix, a prospective mix, or another clearly specified distribution that fits the comparison.

For a baseline-mix comparison, let the standard weights equal period 0’s observed shares, \(w_i^*=w_{i0}\). The standardized change is then \(\sum_i w_{i0}(f_{i1}-f_{i0})\). The observed change can be separated exactly into this within-segment change and a mix effect:

\begin{aligned} F_1^{\mathrm{obs}}-F_0^{\mathrm{obs}} &=\sum_i w_{i1}f_{i1}-\sum_i w_{i0}f_{i0}\\ &=\sum_i w_{i0}(f_{i1}-f_{i0}) +\sum_i(w_{i1}-w_{i0})f_{i1}. \end{aligned}

To verify the identity, expand the right side: the terms \(\sum_i w_{i0}f_{i1}\) cancel, leaving \(\sum_i w_{i1}f_{i1}-\sum_i w_{i0}f_{i0}\). The first term is the change at period 0’s mix; the second is the effect of changing from period 0’s weights to period 1’s weights, evaluated at period 1’s segment frequencies. This allocation depends on the chosen reference convention; using a different reference can allocate the mix effect differently.

In the first worked example, the observed change is \(19.8-14.0=5.8\). The within-segment change at the baseline mix is \(14.6-14.0=0.6\). The mix effect is \(19.8-14.6=5.2\), and \(0.6+5.2=5.8\). The portfolio’s observed increase is therefore much larger than its increase on the fixed baseline mix.

Practical Checks and Limitations

Before interpreting a standardized result, check that segment definitions, claim-count definitions, exposure units, and observation periods are comparable. A change in how exposures or claims are classified can look like a within-segment frequency change even when it reflects a data-definition shift. Confirm also that standard weights are nonnegative and sum to one; if proposed weights do not, normalize them before calculating the weighted average.

Use exposure weights. Claim shares answer a different question and generally cannot be substituted for exposure shares.
Use one common set of weights. A separate set in each period reproduces the actual changing mix rather than holding mix fixed.
Keep units consistent. If segment rates are per 1,000 exposures, the standardized rate is also per 1,000 exposures.
Interpret the reference mix. Standardized results depend on the chosen weights; state whether they represent a baseline, prospective, or other mix.
Do not overinterpret the result. Standardization isolates the arithmetic effect of mix from the segment rates, but it does not remove random variation or explain why segment rates changed.

This method is useful when an aggregate frequency comparison could be distorted by a changing portfolio composition. It supports a fairer period-to-period comparison and provides a transparent way to distinguish mix effects from changes in segment experience. The calculation is aligned with the CAS Exam 5 (Basic Ratemaking and Estimated Claim Liabilities) syllabus.

Check Your Understanding

For each problem, calculate the requested frequency using the stated weights and units.

A two-segment portfolio has frequencies of 8 and 22 claims per 1,000 exposure units. The standard weights are 70% and 30%. Find the standardized frequency. (Answer: 12.2 claims per 1,000.)
Two segments have frequencies of 9 and 21 in period 0, and 12 and 18 in period 1. Use common weights of 40% and 60%. Find both standardized rates and the change from period 0 to period 1. (Answer: 16.2 and 15.6 claims per 1,000; decrease of 0.6.)
Proposed weights for three segments are 0.25, 0.25, and 0.40. Their frequencies are 4, 10, and 28 per 1,000. Normalize the weights and find the standardized rate, rounded to four decimal places. (Answer: weights \(5/18, 5/18, 8/18\); 16.3333 claims per 1,000.)
In period 0, two segments have exposure shares of 60% and 40%, and frequencies of 10 and 30. In period 1, their exposure shares are 30% and 70%, and frequencies are 14 and 26. Find period 1’s observed and baseline-mix standardized frequencies. (Answer: observed 22.4; standardized 18.8 claims per 1,000.)
Use common weights of 20%, 50%, and 30%. Segment frequencies are 3, 12, and 40 in period 0, and 5, 10, and 32 in period 1, all per 1,000. Find both standardized frequencies and the change. (Answer: 18.6 and 15.6 claims per 1,000; decrease of 3.0.)

What You'll Learn