claim reserve prediction: stochastic chain ladder method and bootstrapping

Upload: amira-putri-rahmanida

Post on 10-Mar-2016

16 views

Category:

Documents


1 download

DESCRIPTION

In this final project, the prediction of the outstanding claims liability (claim reserve) is determined by using a stochastic model. The stochastic model discussed in this final project is a Generalized Linear Model (GLM) which produced the same prediction of the outstanding claims liability of that produced by the standard Chain Ladder method.

TRANSCRIPT

  • ABSTRAK

    Prediksi Cadangan Klaim: Metode Chain Ladder Secara Stokastik

    dan Bootstrapping

    Pada tugas akhir ini, prediksi outstanding claims liability (cadangan klaim)

    ditentukan menggunakan suatu model stokastik. Model stokastik yang dibahas pada

    tugas akhir ini adalah suatu Generalized Linear Model (GLM) yang menghasilkan

    prediksi cadangan klaim yang sama dengan prediksi cadangan klaim yang

    dihasilkan metode Chain Ladder standard. Apabila pendekatan stokastik digunakan

    maka prediction error dari estimator cadangan klaim dapat ditentukan. Selain itu,

    pada tugas akhir ini juga dilakukan bootstrapping untuk menentukan prediction

    error dari prediksi cadangan klaim.

    Data yang digunakan sebagai studi kasus dalam laporan tugas akhir ini

    adalah data segitiga run-off incremental claims (Taylor dan Ashe 1983) seperti

    terdapat pada England dan Verrall (1999). Incremental claims pada segitiga run-off

    tersebut diasumsikan berdistribusi Over-Dispersed Poisson (ODP) dengan fungsi

    link logaritma natural (). Prediksi cadangan klaim yang diperoleh menggunakan

    metode Chain Ladder secara stokastik (GLM) adalah 18.680.855,6131 dengan

    prediction error 2.945.695,0585. Pada metode bootstrap, dilakukan simulasi

    sebanyak 30 kali dengan masing-masing simulasi menghasilkan 1000 pseudo run-

    off. Diperoleh rataan prediksi cadangan klaim sebesar 18.873.379,3202 dan rataan

    prediction error sebesar 3.009.170,2321.

    Kata kunci: segitiga run-off, chain ladder, Generalized Linear Model (GLM),

    distribusi Over-Dispersed Poisson (ODP), metode Bootstrap.

  • ABSTRACT

    Claim Reserve Prediction: Stochastic Chain Ladder Method and

    Bootstrapping

    In this final project (skripsi), the prediction of the outstanding claims

    liability (claim reserve) is determined by using a stochastic model. The stochastic

    model discussed in this final project is a Generalized Linear Model (GLM) which

    produced the same prediction of the outstanding claims liability of that produced

    by the standard Chain Ladder method. By applying a stochastic model, it is possible

    to determine the prediction error of the estimator of the outstanding claims liability.

    Furthermore, in this final project, a bootstrap method is applied to determine the

    prediction error of the estimator of the outstanding claims liability.

    The data used as a case study in this final project is a run-off triangle of

    incremental claims (Taylor and Ashe 1983) as found in England and Verrall (1999).

    The incremental claims is assumed to follow an Over-Dispersed Poisson (ODP)

    distribution with natural logarithm () link function . It is obtained that the

    prediction of the outstanding claims liability produced by the stochastic chain

    ladder method (the GLM) is 18,680,855.6131 with its prediction error of

    2,945,695.0585. Applying a bootstrap method, 30 simulations were run and in each

    simulation, 1000 pseudo run-offs were generated. From the bootstrapping

    procedure, it is obtained that the mean of the outstanding claims liability is

    18,873,379.3202 with its mean prediction error of 3,009,170.2321.

    Keywords: run-off triangle, chain ladder, Generalized Linear Model (GLM), Over-

    Dispersed Poisson (ODP) Distribution, Bootstrapping.