Error in official age-specific population estimates over place and time

Article type: Research Article

Authors: Marshall, Alan^{a; *} | Christison, Sarah^a | Simpson, Ludi^b

Affiliations: [a] Department of Geography and Sustainable Development, University of St Andrews, UK | [b] Cathie Marsh Institute for Social Research, University of Manchester, Manchester, UK

Correspondence: [*] Corresponding author: Alan Marshall, Department of Geography and Sustainable Development, Irvine Building, North Street, University of St Andrews, KY16 9AL, UK. Tel.: +44 0 1334 463951; E-mail: adm26@st-andrews.ac.uk.

Abstract: Population estimates for sub-national areas underpin resource targeting for public and private expenditure. We quantify the success of the Office for National Statistics Small Area Population Estimates (ONS SAPE) in England (2011) using census-based population estimates as a comparative gold standard. We model the accuracy (% absolute error) of the ONS SAPE for Lower Super Output areas according to place characteristics and broad age groups. We compare the modelled accuracy to similar small area population estimates developed by local planners in 1991 and also to simple methods (2011) that might be used with less investment in estimation. Our results show that the ONS SAPE is of comparable accuracy to locally conducted censuses that provided the most accurate results in 1991. We find no combination of area characteristic or age group in which simpler methods of population estimation (in 1991 or 2001) outperform the ONS SAPE. The ONS SAPE is least accurate for young adult ages and areas that are experiencing high unemployment or in-migration. For such areas and groups local censuses may be used to resolve disputes over population estimates and are where attention might be focussed in order to improve the accuracy of small area population estimates.

Keywords: Small area population estimates, accuracy, error, area type

DOI: 10.3233/SJI-160342

Journal: Statistical Journal of the IAOS, vol. 33, no. 3, pp. 683-700, 2017