The census is the largest collection of socio-demographic statistics on the nation’s population. It occurs every ten year and provides not only a population count of England and Wales, but also detailed characteristics and attributes at small geography levels that government and non-government organisations utilise to underpin their decision making process.
In a move to bring the census into the 21st century the Office for National Statistics’ (ONS) Census Transformation Programme (CTP) is conducting exploratory research on the feasibility of an administrative data census. For people unfamiliar with the concept, it is the utilisation of administrative data for the production of population statistics. In essence an individual will no longer have to fill out a census form as data will be populated from available linked administrative data.
Administrative data census
This transition from the traditional census to a modernised administrative data census is a global movement that is gaining traction as technological advances and information become available and accessible. Other leading countries exploring this transformation process include Australian Bureau of Statistics, Statistics Canada and Statistics New Zealand.
Countries that have successfully transformed the census into an administrative data census include Sweden and Denmark. However the basis of this relies on the presence of a national population register which the UK does not have and is not a current debate topic.
Several benefits have been highlighted by ONS and other countries’ statistical bodies for this new administrative data approach including:
- A reduction on the running cost of a traditional census (the 2011 census cost approximately £480 million)
- Combat the trend for falling response rates
- An increase in the frequency of census and currency of data (e.g. potential to produce outputs every year)
- An increase in the number of data sources available in this digital age
- Integration with other databases available will increase the number of insights available (e.g. currently investigating income)
- Standardisation of government administrative data across departments
Several challenges have also been identified by ONS and other countries’ statistical bodies including:
- Quality of the data (overcounts and undercounts will have a severe impact on policies, planning and access to funding)
- Restrictions in the range of information available (may not be able to produce estimates for all topics on the current census)
- May need sample population surveys and coverage surveys for quality assurance (incurs a cost)
- Long term implementation (2-3 census rounds)
- Public concern over privacy and confidentiality
Statistical Population Database (SPD) V2.0
With census data underpinning many of our policies, City Hall has a strong interest in new exploratory research. In November 2016 ONS released their second publication of Statistical Population Database (SPD) V2.0. Four administrative databases were used in the production of SPD V2.0: School Census, Higher Education Statistics Agency, Department for Work and Pensions and NHS Patient Register. ONS have explicitly made clear that the administrative data research outputs produced are a means of understanding and assessing the feasibility of this new methodology. You can access the SPD V2.0 data here.
Analysis of SPD V2.0 against 2015 ONS mid-year estimates for London at Lower Layer Super Output Area levels reveals considerable gaps in the data. Inner London is particularly problematic with large undercounts of the population for Westminster, Kensington and Chelsea, Camden and Tower Hamlets, likely owing to the high churn rate. Overcount and undercount of the population is present throughout London. A big question mark remains around the data quality of SPD V2.0 and the implications this can have on decision making processes including planning, policy and funding.
Figure. Percentage difference between SPD V2.0 and Mid Year Estimates (MYE) at LSOA for Males 2015. ONS P1 quality standard range: ± 3.8%, P3 quality standard range +3.8% to 8.5% (orange) and -3.8% to -8.5% (light blue).
While the idea of an administrative data census sounds promising and at the very least will drive the government into further realising the potential of administrative data, there are many issues that need to be resolved. These are early days for this new exploratory research and there is much anticipation around the improvement of the data quality of future releases.