The Chain Ladder and Tweedie Distributed Claims Data
By Greg Taylor
Abstract
The paper considers a model with multiplicative accident period and development period effects, and derives the ML equations for parameter estimation in the case that the distribution of each cell of the claims triangle is a general member of the Tweedie family.
This yields someknown special cases, e.g., over-dispersed
Poisson (ODP) distribution (Tweedie parameter p = 1), for
which the chain ladder algorithm is known to provide maximum likelihood (ML) parameter estimates, and gamma
distribution (p = 2). The intermediate cases (1 < p < 2)
represent compound Poisson cell distributions with gamma
severity distributions.
While ML estimates are not chain ladder for Tweedie
distributions other than ODP, the paper investigates why
they will be close to chain ladder under certain circumstances.
It is also demonstrated that the ML estimates for
the general Tweedie case can be obtained by application
of the chain ladder algorithm to transformed data. This is
illustrated numerically.
KEYWORDS: Chain ladder, maximum likelihood, separation method, Tweedie distribution