Blending Experience with a Manual Rate

Why Blend an Experience Rate with a Manual Rate?

An observed rate reflects the experience of a particular group, but its reliability depends on how much experience is available. A manual rate provides a benchmark that is not specific to that group’s observed experience. Limited-fluctuation credibility combines the two: the observed rate receives weight \(Z\), and the manual rate receives the remaining weight \(1-Z\). With less exposure, the blend generally stays closer to the manual rate; with enough exposure for full credibility, the experience rate receives all the weight.

The calculation is direct, but it is easy to reverse the weights or to use a credibility factor greater than one. The key is to identify the experience rate, the manual rate, and the credibility factor before substituting numbers.

Both input rates must be expressed on the same basis: for example, both in dollars per exposure unit for the same coverage and time period. Credibility does not fix a mismatch in units, coverage, or rating basis. The value of \(Z\) may be supplied in a problem or calculated from an exposure standard. Do not infer a credibility rule that the problem has not specified.

Finding the Credibility Factor

When the problem calls for the square-root rule, let \(N\) denote the observed number of exposure units and \(N_F\) the number required for full credibility under the stated standard. The partial credibility factor is the square root of the ratio of actual exposure to the full-credibility standard, capped at one:

For \(N<N_F\), the square-root rule gives \(Z=\sqrt{N/N_F}<1\). At or above the full-credibility standard, the cap gives \(Z=1\). In that case the indicated rate equals the observed experience rate, because the manual rate has weight zero. This rule is appropriate only when the problem specifies or calls for the square-root partial-credibility rule and provides a compatible exposure standard.

This identity is also a useful reasonableness check. If the experience rate is above the manual rate, the indicated rate cannot be below both of them; if it is below the manual rate, the indicated rate cannot be above both. A result outside the interval between the inputs usually signals an arithmetic error, an invalid weight, or a mismatch in the setup.

Applying the Blend

Reading and Checking an Indicated Rate

The blend is not simply an average of the two rates unless the weights happen to be equal. The experience rate receives the credibility weight, while the manual rate receives its complement. A reliable calculation follows this order:

A rate blend communicates how much the observed experience changes the manual benchmark. The identity \(R_{\text{ind}}-R_{\text{man}}=Z(R_{\text{exp}}-R_{\text{man}})\) makes that interpretation explicit: the experience-to-manual difference is reduced by the factor \(Z\). This also explains why a low-credibility result remains close to the manual rate, even when the observed rate differs substantially.

A common exam error is to calculate a correct credibility factor and then attach it to the wrong rate. Another is to compute \(Z\) from an exposure ratio without taking its square root when the square-root rule is specified. Keeping the weights visible, verifying they add to one, and checking the final rate against both inputs helps catch these errors. Do not round the credibility factor prematurely when an exact fraction or sufficiently precise decimal is available; round the final monetary rate as requested.