Volume 46 - Article 32 | Pages 919–956  

Preparing local area population forecasts using a bi-regional cohort-component model without the need for local migration data

By Tom Wilson


Background: Cohort-component models incorporating directional migration are conceptually robust demographic models which are widely employed to forecast the populations of large subnational regions. However, they are difficult to apply at the local area scale. Simpler models, such as the Hamilton–Perry model, have modest input data requirements and are much quicker, cheaper, and easier to implement, but they offer less output detail, suffer from some conceptual and practical limitations, and can be less accurate.

Objective: The aim of this paper is to describe and evaluate the synthetic migration cohort-component model – an approach to implementing the bi-regional model for local area population forecasts without the need for any locally specific migration data.

Methods: The new approach is evaluated by creating several sets of ‘forecasts’ for local areas of Australia over past periods. For comparison, forecasts from two types of Hamilton–Perry model are also evaluated. Error is measured primarily with an alternative Absolute Percentage Error measure for total population which takes into account how well or poorly the population age–sex structure is forecast.

Results: In the evaluation for Australian local areas, the synthetic migration model generated more accurate forecasts that the two Hamilton–Perry models in terms of median, mean, and 90th percentile Absolute Percentage Errors.

Contribution: The synthetic migration model combines the conceptual and practical advantages of the bi-regional cohort-component model with the light data requirements and ease of calculation of simpler cohort models. It allows the bi-regional model to be applied in circumstances where local area migration data are unavailable or unreliable.

Author's Affiliation

Other articles by the same author/authors in Demographic Research

Projecting the sexual minority population: Methods, data, and illustrative projections for Australia
Volume 45 - Article 12

The geographical patterns of birth seasonality in Australia
Volume 43 - Article 40

Subnational population forecasts: Do users want to know about uncertainty?
Volume 41 - Article 13

Visualising the demographic factors which shape population age structure
Volume 35 - Article 29

The sequential propensity household projection model
Volume 28 - Article 24

Model migration schedules incorporating student migration peaks
Volume 23 - Article 8

Australia's uncertain demographic future
Volume 11 - Article 8

Similar articles in Demographic Research

Probabilistic forecasting of maximum human lifespan by 2100 using Bayesian population projections
Volume 44 - Article 52    | Keywords: Bayesian population projections, human lifespan, population forecasting, supercentenarians

Subnational population forecasts: Do users want to know about uncertainty?
Volume 41 - Article 13    | Keywords: Australia, communication, forecast errors, forecast interval, population forecasting, science, uncertainty

Integrating uncertainty in time series population forecasts: An illustration using a simple projection model
Volume 29 - Article 43    | Keywords: Bayesian approach, England, model averaging, population forecasting, Wales

A new perspective on population aging
Volume 16 - Article 2    | Keywords: aging, historical demography, life expectancy, median age, population forecasting, prospective age

Australia's uncertain demographic future
Volume 11 - Article 8    | Keywords: Australia, migration, migration forecasts, population forecasting, uncertainty

Cited References: 45

Download to Citation Manager


Google Scholar

Article ID