Bootstrapping the Poisson log-bilinear model for mortality forecasting. (English) Zbl 1092.91038

Brouhns, Natacha; Denuit, Michel; van Keilegom, Ingrid

Scand. Actuar. J. 2005, No. 3, 212-224 (2005).

This paper presents the Poisson log-bilinear mortality projection model proposed by Brouhns and goes further into the use of bootstrap procedures for the calculation of confidence intervals. The authors compare on the basis of Belgian mortality statistics a nonparametric bootstrap procedure with the parametric bootstrap procedure proposed by Brouhns. The nonparametric approach consists in generating pseudo-data from the Poisson distribution (with the observed death counts as expected values). The derivation of confidence intervals for expected remaining lifetimes and annuity single premiums with the help of the bootstrap are presented. Belgian mortality statistics are investigated using these techniques.

Reviewer: A. D. Borisenko (Kyïv)

MSC:

91B30

Risk theory, insurance (MSC2010)

Keywords:

Poisson regression; age-sex-specific mortality; projected life tables; mortality forecasting; confidence intervals; bootstrap

References:

[1]	DOI: 10.2307/2290201
[2]	Lee RD, North American Actuarial Journal 4 pp 80– (2000)
[3]	DOI: 10.1016/S0167-6687(00)00054-8 · Zbl 1055.62555
[4]	DOI: 10.1111/1467-9876.00393 · Zbl 1111.62359
[5]	DOI: 10.1016/S0167-6687(03)00138-0 · Zbl 1103.91371
[6]	DOI: 10.1016/S0167-6687(02)00185-3 · Zbl 1074.62524
[7]	Brouhns N, Bulletin of the Swiss Association of Actuaries pp 105– (2002)
[8]	Alho JM, North American Actuarial Journal 4 pp 91– (2000)
[9]	DOI: 10.2307/2530689 · Zbl 0611.62136
[10]	DOI: 10.1017/S1357321700003470
[11]	Bell WR, Journal of Official Statistics 13 pp 279– (1997)