The results from the UK’s ten-yearly population census, conducted in March 2011, were published last week by the ONS. The results were widely written up over the following days noting the trends over a decade or century and all seemed broadly fine. But a closer inspection shows – as is often the case with statistics – that there are some remarkable numbers hidden away. In particular, the official estimate of population for some local authorities has changed beyond all recognition.
The main points we note are:
- The official population estimates of some local authorities (LAs) changed very significantly as a result of last week’s census.
- Newham, for example, an extreme case, has had an increase of nearly one-third in its popualtion in the latest year when it was previously told that it had a gently declining population year on year.
- The scale of the changes leave some of local government areas unable to trust the figures for planning purposes.
- For some LAs it is clear that either the census figures are wrong or recent years’ ONS mid-year estimates (MYEs) have been wrong.
- Financial allocations from central government to some LAs have been too great or too low over a good number of years if the new numbers are to be believed.
- DCLG will need to do some hard thinking about how it can fairly allocate funds given the uncertainty surrounding the data. Some local authorities seem sure to challenge the latest census numbers or the levels of past funding.
- The greatest problem is in London. But several authorities outside London would also have good cause to complain.
- The ONS will do all it can to avoid the highly embarrassing process of census data revision that occurred after the 2001 census figures were published. It says it has done better data quality assurance this time – and challenges to the numbers look set to test that QA process.
To get a sense of the scale of change the census implies consider these examples:
- Newham (London) would have been told last summer that its population was 240,000 (MYE 2010). That was down about 1,000 over the year from the previous estimate of 241,000 (MYE 2009). The census has now published a figure of 308,000, a rise of 68,000 or 28%. A story suggesting a gently declining population trend is now no longer the best guess – the latest estimate is at a much higher level.
- Brent saw the second largest percentage rise in official estimates in the census – 21%, up 55,000 to 311,000. Waltham Forest, Haringey, Hackney and Greenwich (all in London) all saw double-digit percentage increases. Corby, Boston and Bournemouth all saw rises of 9%+.
- In contrast, Westminster would have been told last summer that its population was 253,000 (MYE 2010). That was up about 4,000 over the year from the previous estimate of 249,000 (MYE 2009). The census then published a figure of 219,000, a fall of 34,000 or 13% – a gently rising trend is apparently no longer the best guess.
- The City of London saw an even more dramatic fall in population from 11,700 to 7,400, but it’s a bit of a special case (and small). But nonetheless an error (or change) in population estimate of 37% is noteworthy.
- The ONS talks about the census population for E&W being accurate – and just 500,000 above what was expected. But in the nine LAs that have seen the largest rise in population (see bullet point above), the census has found 275,000 people who were not there in the MYE 2010. The country total might be about right but the people were not where they were thought to be.
- If the census data is roughly right, this is a story perhaps of massive unmeasured immigration in some areas and the rise of the second home in Westminster. But until more detailed data on, say, country of birth and empty homes are published we will not know.
We are hoping to publish our usual charts to show these trends and await the data from ONS.
Counting population is not easy. The ten yearly census is a major exercise to benchmark the annual estimates used for planning purposes. Inevitably a sporadic exercise (every ten years with independent methodology) will deliver estimates astray from the annual figures updated yearly. Most LAs have census estimates not too far from the latest annually produced so-called mid year estimates. But some are very different. Getting to the bottom of what has gone wrong with population estimates in the past and what is good about the census will be vital if the country is to have accurate data in the future, especially if the 2011 turns out to be the last.
The ONS has provided a “high level” description of the differences between the census estimates of population and what the estimates might have been had the MYE 2010 figures been rolled forward. Even if that analysis provides something of a quality check for the statisticians producing the numbers it underestimates the impact of the new numbers on the users of them. (Look at section 7 of the “Explaining the Difference between the 2011 Census Estimates and the Rolled-Forward Population Estimates” referred to on page 8 of the main census press release.)
Apart from a few dozen LAs, the census looks to have published broadly acceptable numbers – at least they don’t look instantly wrong as they did in 2001. But we’d note two other points. First, we might grumble that Scotland’s data was not published last week. Accordingly, we had the odd and always slightly amusing appearance of the unusual geography of England, Wales and NIreland in many reports of the data, reminding us that the status of the UK took a hit with devolution. Second, we were disappointed by the lack of detailed data in last week’s first release (compared to what was published in 2001). That means we know nothing about the local areas (more detailed than local authorities) and the responses to many other questions on education, ethnicity, travel to work etc. For that – and a full understanding of the radically neLA figures – we have to wait till later in the year.
Note: Until we get the full data from the ONS and chart it, here are the sources:
Article updated to reflect comments from ONS on 2 August.