Experience Rating¶

Retrospective repricing method — instantiates Moral Hazard Mitigation

Adjusts an actor's future price or terms up or down from their own realized loss history, so today's care or carelessness follows them into tomorrow's cost.

Experience Rating is a feedback loop over time. It watches the actor's own realized losses and reprices the next period accordingly, so the consequence of hidden action arrives not as an immediate loss share but as a changed premium — up for a bad record, down for a clean one. Its distinctive move is that it is retrospective and cumulative: it prices from what actually happened to this actor, not from their attributes at signup (which is risk-adjusted contracting) and not by carving a slice off each loss as it lands (which is a deductible). That backward-looking design is also its fairness problem in miniature — a single unlucky year should not brand a careful actor — so experience rating lives or dies on credibility weighting: how much of the actor's own record to trust versus the class average, given how thin the record is.

Example¶

A manufacturer carries workers' compensation insurance, which pays injured employees' claims. Because the pool bears those costs, safety investment — guarding, training, ergonomics — is discretionary, and a firm can under-invest and let claims fall on the insurer. Experience rating closes that loop. Each year the insurer computes an experience modification factor from the firm's own claim history relative to what a firm of its size and class is expected to incur, and multiplies the manual premium by it. A firm with fewer and smaller claims than expected earns a factor below 1.0 and pays less; an injury-prone firm gets a factor above 1.0 and pays more. This year's safety decisions thereby set next year's premium, and the effect compounds across renewals. Crucially, the formula weights a firm's own record by its credibility — a small employer's handful of claims is treated as partly noise — so random misfortune doesn't swing a careful firm's rate.^[1]

How it works¶

Accumulate the actor's record. Gather this actor's realized losses over a lookback window from the claims or incident channel.
Compare to expected. Measure the record against the expected loss for the actor's class, so the signal is relative, not absolute.
Compute an adjustment. Convert the gap into a repricing factor applied to next period's premium, terms, or access.
Weight by credibility. Scale how far the factor moves by how much history exists, so thin records lean on the class average and only substantial records move the price hard.

Tuning parameters¶

Lookback window — how many past periods feed the rating. Longer smooths noise but responds slowly to genuine change in the actor's behavior.
Credibility weight — how heavily the actor's own record counts versus the class average; the core fairness dial against penalizing bad luck.
Responsiveness — how fast and far the factor moves per unit of experience, trading incentive sharpness against volatility.
Swing limits — floors and ceilings on the factor, bounding how much a single period can help or hurt.
Class definition — how finely actors are grouped for the "expected" baseline; finer groups rate more precisely but thin each class's data.

When it helps, and when it misleads¶

Its strength is that it makes sustained care pay off and sustained neglect expensive, self-updating each period without needing to observe behavior directly — realized losses are the signal. It suits recurring relationships where a track record accumulates.

Its weakness is noise and thin data. A rare, random catastrophe can penalize a careful actor, and small actors never generate enough experience to be rated fairly — which is exactly what credibility weighting exists to temper, and what a naïve implementation gets wrong by over-trusting a short record. The classic misuse is reading noise as signal, repricing hard off one anomalous year. The discipline is credibility weighting, swing limits, and holding the line between losses the actor controlled and those they merely suffered.

How it implements the components¶

Experience Rating realizes the retrospective-repricing side of the archetype:

observability_or_monitoring_channel — it consumes the loss-history record that makes the actor's past behavior visible through its outcomes.
accountability_consequence — the premium or term adjustment is the consequence, arriving as next period's price.
calibration_and_fairness_review — credibility weighting and swing limits keep the repricing fair against noise and thin data.

It does not collect or verify the raw behavior itself — that is Monitoring Requirement — nor set terms up front from risk attributes, which is Risk-Adjusted Contract, nor impose an immediate share of the loss, which is Deductible.

Instantiates: Moral Hazard Mitigation — it ties the actor's future cost to their own realized record, so care and neglect carry forward.
Consumes: Monitoring Requirement supplies the loss and incident records the rating is computed from.
Sibling mechanisms: Risk-Adjusted Contract · Deductible · Monitoring Requirement · Behavior-Conditioned Warranty · Clawback Clause · Collateral Requirement · Copay · Malpractice or Professional Liability · Performance Bond · Shared Liability Clause · Usage Cap or Throttle

Notes¶

Experience rating has a look-alike in the adverse-selection world, where realized-claims data is used to recalibrate a pool's base rates against hidden types entering it. The moral-hazard version here is aimed at behavior after coverage: it reprices a specific actor's own conduct so that staying careful is worth it. The data source can be shared; the purpose — governing hidden action, not hidden type — is what fixes it to this archetype.

References¶

[1] Credibility weighting is the actuarial method of blending an actor's own loss experience with the class average in proportion to how much data the actor's record contains. Workers' compensation experience modification factors are computed this way, so that small employers' random claims move their rate less than a large employer's fuller record moves theirs. ↩