Methodology, Dynamic microsimulation

Research notes · Dec 31, 2012

On weights in dynamic-ageing microsimulation models

Gijs Dekkers, Richard Cumpston

Federal Planning Bureau, Belgium; CESO, Katholieke Universiteit Leuven, Belgium; CEPS/INSTEAD, Luxembourg; Research School of Finance, Actuarial Studies & Applied Statistics, Australian National University, Australia

Volume 5(2) Autumn 2012

Cite as: G. Dekkers, R. Cumpston; 2012; On weights in dynamic-ageing microsimulation models; International Journal of Microsimulation; 5(2); 59–65. doi: 10.34196/ijm.00072

Abstract

A dynamic model with cross-sectional dynamic ageing builds up complete synthetic life histories for each individual, starting from a survey dataset or an administrative dataset. Many of these datasets include weights. This is a problem for dynamic microsimulation models, since the most obvious solution, expanding the dataset, is not always advisable or even possible. This short paper presents an alternative method to use weights in dynamic microsimulation models with dynamic ageing. This approach treats the weighting variable as just another variable in the model, and the weights are only used after simulation to derive weighted simulation results. Tests using Australian data confirm that weighted variables give sensible results, with reductions in runtimes and memory requirements.

This article contains 2 figures and 1 table.

Figures

Figure 1. Example of the alternative strategy.
Figure 2. Weighted and unweighted person projections.

Tables

Table 1. 50-year projections using unweighted and weighted data.
Weighted Data Alignment Moves Runtime (seconds) Runtime as % of unweighted Average number of records as % of unweighted
No No Yes 39.07 100.0% 100.0%
Yes No Yes 31.35 80.2% 81.6%
Yes Strategy 1 Yes 32.60 83.4% 83.0%
Yes Strategy 3 Yes 32.49 83.2% 82.3%
No No No 31.58 100.0% 100.0%
Yes No No 19.72 62.4% 57.9%
Yes Strategy 1 No 20.52 65.0% 60.5%
Yes Strategy 3 No 20.12 63.7% 58.5%