Suite Actuarial
Menu

Reserves

Claims development, reserve estimation, and uncertainty.

Project future claims development with Chain Ladder, Bornhuetter-Ferguson, and the ODP bootstrap. The example's triangles, premiums, and results are in MXN millions.

Read the decision, cash flows, assumptions, method, interpretation, and limits as one continuous case.

Actuarial case explained

Which claims cost has not yet emerged in observed payments?

Purpose

Decision informed

Select a central estimate and separately disclose development, tail, and uncertainty.

Benefits and cash flows

  1. 1Observed paid or incurred claims by origin and development.
  2. 2Future cells projected to ultimate.
  3. 3Reserve equals ultimate less observed.

Assumptions

AssumptionValue / unitSource and status
TriangleCumulative, MXN millionsIllustrative
Reproducible 2019–2024 case
PatternStable across origin yearsConvention
Chain Ladder assumption
TailDeclared or ShermanConvention
Explicit extrapolation

Method

Volume-weighted Chain Ladder

Factors develop each diagonal; Mack and ODP bootstrap quantify error conditional on the model.

Reserva = Σ(Ultimateᵢ − Observadoᵢ)

Results and interpretation

The tail factor can materially move the reserve even though no observed cell identifies it by itself.

Applied example

The reserve and the tail factor

Claims from a given year generate payments over several years. To estimate the required reserve, the actuary projects a triangle of historical payments (2019-2024, MXN millions) with Chain Ladder. The tail factor represents development expected after the last observed period.

Adjust the parameters to observe how the result changes.

Estimated reserve

Estimated ultimate cost

Actuarial interpretation

A 3% tail factor changes the reserve by several million pesos. Reserves are estimates that are sensitive to assumptions; regulation requires documenting them, and actuarial practice quantifies their uncertainty. The lab implements Mack's (1993) model, which estimates a variance per development period and composes the prediction error by origin year; it measures error conditional on Chain Ladder, not model risk.

Validation and limits

What is checked

  • Mack reproduces the Taylor and Ashe case to the currency unit.
  • The ODP bootstrap reproduces the published φ and reconciles with Chain Ladder.

What it does not prove

  • Mack and ODP do not cover model risk, mix change, or unobserved inflation.
  • Every tail is extrapolation and requires external justification.