The Concept of Credibility

A fleet of 12 trucks has 3 years of data and 9 claims averaging $4,200 per truck. The class rate is $6,800. Price at $4,200 and risk inadequate reserves. Price at $6,800 and lose the account. Credibility theory gives the answer: blend in proportion to how much data you have.

HIGH-FREQUENCY: The credibility formula and interpretation of $Z$ appear on nearly every FAM credibility question.

$$\hat{R}=Z\cdot \bar{X}+(1-Z)\cdot M$$

• $Z\in [0,1]$ is the credibility factor
• $\bar{X}$ is individual experience
• $M$ is the class estimate (manual rate, prior, complement of credibility)

When $Z=1$: full credibility, individual data stands alone. When $Z=0$: class only.

KEY: $Z$ always multiplies the individual experience $\bar{X}$, and $(1-Z)$ always multiplies the class estimate $M$. Reversing them is the most common credibility mistake.

Common mistakes

• Applying credibility backwards. Computing $Z\cdot M+(1-Z)\cdot \bar{X}$ weights the class by $Z$ instead of the individual. With $Z=0.40$, individual = 84,000, class = 95,000: wrong formula gives 91,600 instead of 90,600. Trap: $91,600.
• Using $Z=\sqrt{n/n_0}$ when $n>n_0$. If $n=1,500$ and $n_0=1,082$, the raw formula gives $Z=1.177$, but credibility is capped at 1. Using 1.177 overweights individual data. Trap: an estimate that overshoots individual experience.
• Confusing the credibility factor with a confidence level. $Z=0.60$ does not mean "60% confident." It means "assign 60% weight to individual data." Credibility is a blending weight, not a probability.

Bottom line

• Credibility formula: $\hat{R}=Z\cdot \bar{X}+(1-Z)\cdot M$, where $Z$ = credibility, $\bar{X}$ = individual, $M$ = class
• $Z\in [0,1]$ is a blending weight measuring statistical trust in the data, not a confidence level or probability
• Full credibility ($Z=1$): data meets the tolerance standard, class gets zero weight (but this is not a guarantee of accuracy)
• Partial credibility: $Z=\sqrt{n/n_0}$, capped at 1

Exam shortcut

Credibility questions are fast (the formula is simple. The exam tests setup: which value is $\bar{X}$ (individual) and which is $M$ (class)? Always cap $Z$ at 1. Decision rule: if the problem gives a full credibility standard and asks for $Z$, use the square root formula. If it gives process variance and variance of hypothetical means, use Buhlmann ($Z=n/(n+k)$). Remember: "ZIP code") Z times Individual plus (1-Z) times Prior.

The full lesson (about 1,724 words, 11 min read) adds 2 worked examples, all 5 common mistakes, a self-check, free in the app.

Learning objectives

• 4a

Browse all free Exam FAM lessons or jump into free Exam FAM practice questions.