Plans for the use of administrative data in 2021 have been criticised as unambitious by the Public Administration Select Committee. Pete Benton, director of 'Beyond 2011', was keen to show that the application of administrative data had not been forgotten and was central to their research programme. This programme now has several strands: to develop methods to support the e-census in 2021; to continue research on conducting a fully administrative census post 2021; and to consider the concept of census further into the future (do we need a ‘census day’?).
Some indication of how the 2021 census may proceed was given by presentations from countries from the international review panel also meeting last week. The US Census Bureau explained how they intended to use daily updates to plan field work after discovering that field staff mileage in 2010 was sufficient to travel the entire road network more than 50 times. New Zealand and Canada also explained their plans to deal with estimating very small population groups in the future.
One of the ideas many countries are considering is having field staff use handheld devices for real time updates on returns. But instead of procuring these at considerable cost as a couple of countries tried yet found extremely difficult previously, they plan to let field operatives bring their own device. It is expected that sufficiently enabled and devices will be owned by most workers and that operational data will not prevent a security problem.
Statistics Australia outlined some plans to show a household summary before submission of the e-census form to ask if anyone had been missed (a surprisingly common problem with children under five). They also indicated some experiments to advance the detail of data collected such as by offering prompts of specialisations in generic job categories like 'nurse'. Their strategy is to design around an e-census rather than just putting the paper form online as has happened in the past.
Prospects for a census based on administrative data are based around two types of outputs: population estimates, and attributes data. Plans to estimate population sizes by age and sex in even small areas from linked, anonymised admin data are well advanced. Li-Chun Zhang from Southampton University outlined new methods to deal with overcount which may even improve on standard census estimates.
Attribute data, the questions which make up most of a census form, are proving more challenging. The administrative alternative put forward in the autumn consultation intended to rely on a large sample survey to get these data but users had felt this would not get the detail they relied on the census to provide. The range of sources which would be needed to garner all the data currently collected is vast, and permission to collect and use it something which would require future legislation.
Discussion amongst the assembled delegates was wide ranging and generated more questions than answers. While there are many sources of administrative data, they are collected for particular purposes, not with population statistics in mind. It will take a long time to work through all of the issues outstanding but there were hints at some possible solutions or partnerships. Most encouraging was the enthusiasm for the research programme expressed by all of the international delegates.
The presentations from the conference are available on StatsUserNet, and the conference was confirmed as an annual event. Plans are afoot to fund more research by developing a partnership with the ESRC, subject to settling the funds to be made available. This could look further into the future, to scan the horizon, and to think more internationally about census methodology. It is easy to say that we have all the data already and it should be easy, but the vast scale of the census and the genuine novelty of methods already emerging make it clear that view is naive.