Beyond 2011 Administrative data sources and low-level aggregate models for producing population estimates
Overview Options using administrative data sources Description of administrative data sources Comparison of population estimates Aggregate administrative data based models
List of statistical options Administrative data-based options Aggregate model Record Level model Intermediate model Census options Traditional census – long form to everyone Short-form/Long form census Rolling census Short form census and continuous survey Survey options Address register + survey Continuous survey
List of statistical options Administrative data-based options Aggregate model Record Level model Intermediate model Census options Traditional census – long form to everyone Short-form/Long form census Rolling census Short form census and continuous survey Survey options Address register + survey Continuous survey
Administrative Sources Broad Coverage Sources - Patient Register Database (PRD) - Customer Information System (CIS) Sources covering sub-populations - School Census - Higher Education Statistics Authority (HESA) Student Data - Migrant Worker Scan (MWS)
Aggregate Model Initial population estimates e.g. weighted ‘average’ of sources stratified by - age - sex - geography Coverage assessment e.g survey
Record Level model Link admin sources to produce initial population estimates Coverage assessment e.g. survey Estimation to produce weights Revised population estimates
Intermediate model Initial population estimates – aggregate model Match and link in coverage check survey areas - coverage survey - admin sources - address register Estimation to produce weights Adjusted population estimates
Comparison of estimates Population estimates generated from counts of individuals on broad coverage administrative sources These estimates compared with each other and ONS Mid Year Population Estimates National difference in total population between administrative sources ~1%
Percentage difference of total population of broad coverage sources from mid-year estimates
Percentage difference by age group in university towns and cities CIS Males Females Patient Register
Low level aggregate models Basic model Basic model plus weighting Basic model plus coverage survey Statistical model based approaches Hybrid approach
Challenges Reliance on administrative data Variations in definitions and coverage Future-proofing Provision of socio-demographic data Measuring the accuracy of the estimates