Journal of Population Sciences
From: Evaluation of simple methods for regional mortality forecasts
Approach | Examples |
---|---|
1. Use national or State, or region-type, mortality forecasts | Simply use national or State mortality forecasts for all regions (e.g. Pittenger 1976; CSO 2013), or apply forecasts for broad region types to all regions of that type |
2. Trend regional mortality age schedules towards a long-run target | Trend regional mortality age profiles from the base period converging towards (or diverging away from) a very long-run set of target mortality rates (e.g. Van Hoorn and Broekman 1999 and Lanzieri 2016) |
3. Apply models developed for single national populations to every region | For every region: • Fit a parameterised model schedule and extrapolate parameters • Use functional data models (e.g. Hyndman and Ullah 2007) • Fit new versions of extrapolative models (e.g. Ediev 2008) • Apply the Lee-Carter model (Lee and Carter 1992) or one its many extensions (e.g. Li et al. 2013) |
4. Use multi-population models | • Lee-Carter extension for multiple populations (e.g. Li and Lee 2005) • Two-population age-period-cohort model (Cairns et al. 2011) • Product ratio method of Hyndman et al. (2013) • Bayesian approaches (e.g. Gongaza and Schmertmann 2016) • Multilinear component approach of Bergeron-Boucher et al. (2018) |
5. Use a national or State mortality forecasts and create regional mortality forecasts via simple relationships | • Brass-type relational models (e.g. Brass 1971, Ewbank et al. 1983, and Murray et al. 2003) • National or State mortality age schedules scaled up or down according to the SMR (e.g. Giannakouris 2010 and NRS 2016) • TOPALS relational method using smoothed age profiles of rate ratios (de Beer 2012) • Assume regional death rates decline by the same proportional amount as forecast national death rates (Smith et al. 2013) • Mortality surface method (e.g. Wilson 2014, 2017) • National or State mortality age schedules scaled up or down according to regional/national mortality rate ratios (e.g. ONS 2016) |