<Mortality Projection using Linear and Bilinear Models>
Abstract
In this paper, we compare two types of stochastic mortality forecasting models,
log-bilinear and log-linear, for data from six different countries. With calculating
associated statistics (such as Pearson, Deviance, AIC, and BIC) and comparing the
horizontal mortality projections, we demonstrate the necessity of cohort parame-
ters and the validity and advantages of employing a log-linear model for mortality
forecasting. In doing so, we conclude that the log-linear model provides a viable
alternative to log-bilinear model in a wider range of applications.
Contents
Declaration of Authorship i
Abstract ii
Acknowledgements iii
Contents iv
List of Figures vi
List of Tables viii
Abbreviations ix
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Research Focus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Individual Research Objectives . . . . . . . . . . . . . . . . . . . . 3
1.4 Outline Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Literature Review 5
2.1 Main Methodological Developments . . . . . . . . . . . . . . . . . . 5
2.1.1 Explanation . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Expectation . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.1.3 Extrapolation . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Characteristics of Original Population Data . . . . . . . . . . . . . 9
2.3 Mortality Modelling and Forecasting . . . . . . . . . . . . . . . . . 11
2.3.1 Fitting Data and Parameter Estimating . . . . . . . . . . . 12
2.3.2 Parameter Smoothing and Mortality Forecasting . . . . . . . 16
2.3.2.1 Parameter Smoothing . . . . . . . . . . . . . . . . 16
2.3.2.2 Mortality Forecasting . . . . . . . . . . . . . . . . 17
2.3.3 Comparisons among Models . . . . . . . . . . . . . . . . . . 19
3 Theoretical Analysis of Models 22
3.1 Mortality Modelling and Forecasting Approaches . . . . . . . . . . 22
3.1.1 Log-Bilinear Models: LC and LCC . . . . . . . . . . . . . . 22
3.1.1.1 The Lee-Carter Approach . . . . . . . . . . . . . . 22
3.1.1.2 LCC: A Special Case of RH Model . . . . . . . . . 26
3.1.2 Log-Linear Models: LL and LLC . . . . . . . . . . . . . . . 30
3.2 Goodness-Of-Fit Tests . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2.1 Residuals . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.2.2 Model selection criteria . . . . . . . . . . . . . . . . . . . . . 38
4 Empirical Analysis and Results 40
4.1 Description of the Data . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.2 Model Fitting and Parameter Estimating . . . . . . . . . . . . . . . 43
4.2.1 LC & LCC . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.2.2 LL & LLC . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.3 Model Comparisons . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.3.1 Goodness-of-t Statistics . . . . . . . . . . . . . . . . . . . . 48
4.3.2 Selection Criteria: AIC & BIC . . . . . . . . . . . . . . . . . 49
5 Mortality Projections and Discussion 52
5.1 Mortality forecasting under LCC model . . . . . . . . . . . . . . . . 52
5.2 Mortality forecasting under LLC model . . . . . . . . . . . . . . . . 57
5.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6 Conclusion ................................................................... 64
Appendix A i
Bibliography vi
|