Background

1. The purpose of this report is to assess the suitability and quality assurance of the data and methods used in the Dynamic Population Model (DPM) to produce admin-based population estimates (ABPEs) in England and Wales. This report covers aspects of data used as inputs to the model, the modelling framework and its strengths and limitations in producing population statistics, as well as the transparency and sustainability of the methods.

2. The official statistics on population size in England and Wales by age, sex and administrative areas (Local Authorities – LAs) are produced based on the population balancing equation, where the starting point (i.e. a population stock on a given day) is defined by the Census, which is then updated (or rolled forward) to 30 June (or mid-year) for each year, with information on births, deaths and movements into and out of the area; special populations such as members of armed forces and prisoners are also accounted for (e.g., ONS 05/09/2022, 23/11/2023b). In England and Wales, we distinguish movements to/from abroad (international migration), to/from constituent countries (cross-border migration), and between areas within England and Wales (internal migration).

3. These current official statistics are referred to as Mid-Year Estimates (MYE). The common issue with these estimates based on censuses is that they are precise in the census year and their quality deteriorates the further we depart from the census, which is known as an intercensal drift (Blackwell et al. 2022, ONS 05/09/2022), also called an unattributable population change (UPC). This drift or UPC can be caused, in principle, by an error in the population baseline from the preceding census (census base) and by the error in the net migration estimates used in creating mid-year population for 2002-2011 (ONS 12/07/2012)*.  Similar concerns have been raised about post-2021 Census MYE, especially at a local authority level (Champion 2024).

*The 2011 Census estimates were around 476,000 higher than the rolled-forward MYEs (ONS 12/07/2012).

4. Since the above issues are well-known to the Office for National Statistics, they have been researching, since before Census 2011 (Blackwell et al. 2022) alternative methods for producing population estimates based on administrative data sources. Currently, these new admin-based population estimates (ABPEs) are considered official statistics in development, previously known as “experimental” (e.g., ONS 23/11/2023a). The ABPEs are produced by using a Dynamic Population Model, which is a statistical model that enables estimation of population stocks and flows and accommodates the accounting identity* at a national as well as sub-population level, e.g. in a particular area. The model takes as input a variety of administrative data on population stocks and flows. The DPM utilises Bayesian inference to estimate model parameters and produce population estimates with measures of uncertainty – credible intervals (Bayesian counterparts of frequentist confidence intervals).

*The accounting identity is satisfied approximately because of the use of open ended last age group.