Modelling a flexible platform for forecasting populations over time

CPC research just published in Demography provides a new, flexible platform for forecasting populations.

The article, entitled 'Bayesian Population Forecasting: Extending the Lee-Carter Method' by CPC members, Arkadiusz Wisniowski, Peter Smith, Jakub Bijak, James Raymer and Jonathan Forster demonstrates the advantages of adopting the Bayesian approach for population forecasting and highlights areas where this work could be extended.

Most statistical offices rely on projections based on simple, so called 'cohort component' calculations which take the population in any given year, add on the projected number of births, takeaway the expected number of deaths and add net migration. However these can be unreliable due to differences in the underlying assumptions of these calculations, especially combining emigration and immigration in one category of net migration. This new forecasting approach overcomes some of these problems allowing the inclusion of fertility, mortality, emigration and immigration by age and sex as well as providing measures of accuracy for the results.

In this paper, the authors illustrate their method using the case of the United Kingdom and forecast the components of population change to the year 2024. They compare their forecasts with the official deterministic projections for 2024 as prepared by the Office for National Statistics in 2011.

Key findings include differences in projections with regard to fertility and migration. The authors forecast that the number of people aged 20-35 will be significantly larger than the ONS projections, which assume that net migration stays at a constant rate.

They also forecast a closing in the gender gap between men and women, indicating that there will only be 49,000 fewer females than males in 2024. This compares to a difference of 1 million in 2009. The authors state that the reason for this switch is most likely due to a larger proportion of male migrants and a gradual closing of the gap in life expectancy between the sexes.

Additionally, the authors find that the population predicted by the Office for National Statistics seems to be growing more slowly than in their own forecast. They note that this difference could be due to a more conservative prediction of international net migration.

To read the full article, please visit the Springer Link website.

Posted 18/05/2015 13:36