credibility factor Z
/ zee or zed /
The credibility factor Z is the single number at the centre of all credibility theory: it is the weight, between 0 and 1, that you place on a risk's own experience. Think of it as a dial. Turn it to 1 and you trust your own data completely; turn it to 0 and you ignore your data and use the benchmark alone; leave it at, say, 0.4 and you take a 40-percent blend of your own data with a 60-percent blend of the benchmark.
Whatever credibility method you use, the final estimate is the same blend: estimate = Z times (own data) plus (1 minus Z) times (benchmark). The methods differ only in how they compute Z. Classical credibility sets Z by the square-root rule, Z = sqrt(n/N). Buhlmann credibility sets Z = n/(n + k), where k is a ratio of two variances. In every case Z behaves the same way at the extremes: it rises toward 1 as you accumulate more relevant data, and it sits near 0 when you have almost none.
Z is where judgement, theory and regulation all meet. A regulator may dictate the full-credibility standard that pins down Z; a Buhlmann analysis may estimate it from data; a pricing actuary may sanity-check it against experience. The key honesty is that Z is never exactly knowable — it depends on quantities (true variances, the true standard) that we can only estimate — so a sensible actuary treats the chosen Z as a reasoned approximation, not a precise truth, and documents the assumptions behind it.
Three small schools each get a Z by the data they bring: a tiny school with 50 exposures might earn Z = 0.2, a mid-size one Z = 0.55, a large one Z = 0.9. The larger the school's own data, the more its own claims drive its rate.
Z is a per-risk dial: more own data, higher Z, more self-determination of the rate.
Z = 1 does not mean the estimate is exactly right — it only means you are choosing to ignore the benchmark, having judged your own data stable enough. Z near 1 still carries the ordinary sampling error of the data.