The 2001 census was meant to be the most accurate picture ever of Britain's population. Then it found that there were nearly 1m fewer of us than expected.by Graham Bowley / November 20, 2003 / Leave a comment
Shortly after graduating from university I went to work at the treasury, where one of my jobs was collecting the latest economic numbers from the central statistical office, or office for national statistics (ONS) as it is now known. Every few weeks I would descend into the bowels of Whitehall where statisticians described the dips and squiggles of the trade gap, or some other data series. Occasionally there was a discreet splutter, and a statistician admitted that the figures I was given last month were slightly inaccurate and had now been revised. We quickly moved on. I never doubted the numbers. My job was to return to the treasury and work out a line for the chancellor (then Norman Lamont) to use in his defence – “We are still in recession but so is half of Europe” – when the figures were released to the press a few days later.
Four or five years on, when I was working at the Financial Times, I had to report on the same economic statistics. The statisticians that dealt with the media were a slightly more outgoing lot than the pure number-crunchers (they would show up at the FT’s Christmas party and get drunk). Once, one hot summer, they said the reason inflation had picked up slightly was because pigs had gone off sex so bacon prices were climbing. They had a twinkle in their eyes. But we reporters wrote it down and got front-page stories. We didn’t doubt what we were told. Why should we? These were Whitehall statisticians.
I hadn’t really thought a lot about these scenes in recent years, until events at the ONS over the past few months brought them back to me. Since last autumn, when the ONS published the long-awaited findings of its 2001 census, it has faced a steady barrage of complaints that it has got its numbers wrong.
It is hard to exaggerate the importance of the census to the ONS. It is a count of everyone who lives in the UK. It takes place once every decade, involves years of preparation, and in 2001 cost ?255m. Its value lies in counting every person more or less at the same time. It asks the same questions in every street in every city, north, south, east and west, so its findings can be compared everywhere. In the swirling existential murk of our modern lives, it is a once in a decade stab at working out who we are.
It also has far-reaching implications. According to what it says about where we live, how old we are, and what we need, companies build houses, and put up supermarkets. The government plans schools, roads, and hospitals, and decides where new housing will be needed. In the great divvying up of ?50bn of local government funds each year, it is the ONS’s annual population estimates – based on the underlying ten-year count – that along with hospitals’ and councils’ special needs dictates where the money flows. As Len Cook, the New Zealander who has run the ONS since 2000, told BBC radio: “Without the census it would probably cost the country billions in poorly allocated decisions and investment misspent.”
But in the 2001 census, something went wrong. Several local councils, including Westminster and Manchester, are complaining bitterly that the ONS has significantly underestimated their populations. They have demanded a recount. At first, Len Cook batted their criticisms away imperiously. But the chasm between his numbers and the figures the ONS had predicted only a year earlier was so large, and Cook’s manner so prickly, that he came across like a man who had something to hide. His position became even less comfortable when a wider group of critics began to question his numbers. Then, this spring, the statistics commission, the body set up to monitor the ONS, announced a census inquiry and even held out the possibility that it might recommend a recount.
This state of uncertainty and confusion contrasts with the confidence with which the census was launched two and a half years ago. Census 2001 was to be an exercise worthy of the 200th anniversary of the first census. In 1800, amid Malthusian anxiety about population growth, parliament passed the Census Act, and the following year overseers of the poor, constables, tithingmen, headboroughs and other officers of the peace collected information from households. Great Britain’s population turned out to be around 9m. Following 1801, there was a census every ten years, except in 1941, but the big expansion in its size and its importance for policy came in the postwar period, the heyday of interventionism and the welfare state.
Then in 1991, the census ran into problems. This was the first sign of the more serious difficulties that would emerge ten years later. In 1991, it became clear that the task of getting census forms to people, and then persuading them to fill them out and return them, was harder than in previous decades. More people were living behind entryphones and security systems, were seldom at home, or just did not feel a sufficient sense of civic duty to fill in yet another form. There were more immigrants, and many of them couldn’t read English, while some of those who could were reluctant to give information for fear of losing benefits. The poorest returns were in the inner cities, where the new lifestyle trends were strongest and the populations the most transient. Memories of the poll tax in the late 1980s were also fresh; confirming your name and address could then attract a large bill.
In the end, the returns were so low – 96 per cent overall, but much lower in some trouble spots – that statisticians by and large chose to ignore the 1991 census. They reverted instead to the 1981 figure, updating it year by year using separate annual statistics on births, deaths and net migration, right up to 2000. When 2001 came round, the statisticians wanted to avoid the problems of 1991. So in the run-up to C-Day – Sunday 29th April 2001 – the ONS and local councils launched a big publicity campaign, called “Count Me In,” to remind Britons of their duty to be counted. In the following days, an army of 63,000 census workers spread out across England and Wales to deliver forms. The country was divided into 115,000 localities, each containing on average 200 households, with every worker responsible for one or more area.
The pink forms had been redesigned to make them easier to understand and fill out. Nevertheless, each form had 40 questions (41 in Wales), compared to around 30 in 1991. It asked the predictable questions about age and marital status. But there were other inquires about ethnic group, academic and professional qualifications, health, how you travelled to work, and for the first time a question about religion. People were asked to fill out the answers on census night, and were then legally required to post the forms back soon after. (They faced a ?1,000 fine if they did not, but this was enforced in only a very small number of cases.)
This so-called postback was new. In earlier censuses, field staff had collected forms by hand, but this was costly and Len Cook wanted to save money. He needed to pay for a second big innovation, the ONS’s pi?ce de r?sistance, which was meant to solve all the deficiencies of 1991: an independent follow-up survey called the census coverage survey, or CCS (later the joke was that CCS stood for census cover-up survey). “It will be innovative and leading edge,” an ONS internal research paper boasted.
The CCS followed hard on the heels of the census proper. In four weeks in May and June 2001, over 4,000 professional interviewers conducted 320,000 10-minute interviews on doorsteps in all regions of the country, although they concentrated their efforts on the inner-city districts likely to have had the worst return on census night. Apart from the census itself, it was the biggest household survey ever carried out in England and Wales, and it provided the ONS with a profile of the people it had missed in the earlier count. In Cook’s words on the Today programme last September, it gave us “rich detail about people who should have been in the census but didn’t complete forms.”
When they had got their CCS data, Cook’s statisticians combined the findings with other records, like the child benefit count. From this information they created “synthetic” households and imputed their details across the country. The Guardian said the ONS was “making it up.” But it was the chief reason why Len Cook felt confident enough to call his count a true “one number” census: a single indisputable figure, he believed, that unlike the patchy 1991 operation covered the whole population.
On 30th September 2002 – 17 months after C-Day – Len Cook announced his one number to the public, along with the wealth of other information the census had gathered. What these details said about Britain and the British was in the most part predictable enough. We were an ageing population (for the first time there were more over 60s than under 16s.) There were more women than men. And increasing numbers were choosing to live in the southeast and southwest (where population rose over the last 20 years), rather than the northeast and northwest (where population fell).
What the ONS said about the total size of the population was less predictable. The number of people living in the UK in 2001 was, Cook announced, 58,789,194. In England and Wales, this meant the population was around 900,000 less than the ONS had calculated in its mid-year estimate for 2000. The problem was, Cook said, that the earlier estimate was wrong. It was too high, as was every population estimate for the past 20 years. This was quite an admission by our national statistician: for 20 years the ONS had been getting its figures wrong.
The reason, Cook said, was simple: emigration. People had simply been leaving the UK in greater numbers than had previously been thought. In fact, this phenomenon was detected in the 1991 census, he now said. But at that time the statisticians had refused to believe it – they put it down to the poll tax effect rather than a new wanderlust – and ignored it. But a decade later, when they saw the same thing happening again, they realised something new was indeed at work. It wasn’t the poll tax, so it had to be emigration.
Len Cook explained that it was young men in their twenties and thirties who were chiefly responsible for the trend. Their numbers were 800,000 lower than the previous year’s forecast. And didn’t this fit with anecdotal evidence? Didn’t more young people now go abroad, on gap years, or to study, or to enjoy the rave culture in Greece? They were flocking to destinations traditional for expats – Australia, Canada – but also Europe. It seemed to make sense. In the 1990s Britons could, for the first time, work without a permit across the EU, and the arrival of low-cost air travel in the late 1990s made it easier still.
Cook called for efforts to improve the quality of Britain’s emigration statistics. Glad to have cleared up the mess, was his message. He acted as if no one could doubt the judgement of our national statisticians. “This census has proved both a landmark and a benchmark,” he said in a joint statement with John Randall and Norman Caven, registrars general for Scotland and Northern Ireland. “It has been a world-class exercise that has given us the most accurate picture possible of our population.”
But over the following weeks, some voices began to raise serious questions about the census’s accuracy. Several regions had their population figures revised upwards. The population of the City of London, for example, rose by 13 per cent. Brent, in north London, which suffered a big drop in 1991, saw its population rise in 2001 after a noisy campaign to get people to fill out their forms. But the new figures also meant some regions of the country experienced large downward revisions to their populations. For local councils, whose annual grants are calculated according to the size of their populations, this raised the prospect of budget cuts. In Manchester, the difference between the mid-year 2000 estimate and the actual 2001 census figure was 46,700 people, or one tenth of the population. Oxford, Cambridge, and Kensington and Chelsea also saw cuts of one tenth or more.
But the biggest fall was in Westminster, where the population was revised down by a quarter, from 244,600 to 181,300. Westminster’s irate officials said this meant they stood to lose ?60m in government grants. This, as they made clear, was a lot of street cleaners and teachers. “We were in a budget meeting when the note [detailing Westminster’s result] was passed to me,” Kit Malthouse, deputy leader and councillor responsible for finance, told me when I visited him at Westminster city hall. “It felt like a car crash. We thought we had little or no choice but to challenge the figures, to go on the attack.”
In the months following the ONS announcement, most councils across the UK accepted the census with full confidence and began to feed the numbers into their planning procedures (as did many of the private sector market research and opinion poll companies that use ONS numbers in their databases). But a group of about 20 rebel councils – those with the biggest population declines – started to build a case that, at least in their regions, the census was flawed. They pointed to practical problems during census gathering – acknowledged by the ONS – such as delays by the Royal Mail in delivering posted-back forms, which in turn hampered the CCS. Westminster found people, including a local MP, who hadn’t even received a form in the first place. The councils also pounced on other supposed errors, such as at Forest Heath in Suffolk, where the ONS reported a 22 per cent population decline but the council, according to the FT, suspected the ONS had simply undercounted two US army bases.
Westminster compiled its own population count, from sources such as the number of people paying council tax, or who registered to vote, or who used its hospitals. For example, it found that between 1991 and 2001, its electoral register rose by 26 per cent, and the primary school rolls by 28 per cent. In the end, its count came up with a figure close to the ONS’s pre-census estimate. “We have a very mobile population, a high proportion of young people, asylum seekers, students, hostels,” Malthouse said. “Twenty-five per cent of our population turns over every year… There were obviously problems in getting forms to the people… They say that our population fell by 6,000 over ten years, but during that period we have built 8,500 homes.”
The councils’ argument was clear. For all its statistical sophistication, the ONS simply hadn’t solved the problems that had dogged the 1991 survey. The fact that councils like Brent had apparently been able to manipulate the count by mobilising their populations around C-Day only supported their argument.
Backed by its own evidence, Westminster threatened legal action against Cook. If the ONS did not admit it was wrong, the council would also apply to the treasury, as the department directly responsible for the ONS, to force a re-run. But Cook continued to defend himself. The councils were wrong, he said, when I went to see him this summer, at the ONS’s shiny headquarters near the Thames in Pimlico.
Cook is a small, bouncy man with wispy hair, moustache, and tinted glasses. He held the same position in New Zealand before coming to London. He attended Henley management school and Insead, enjoys fly fishing and speaks French and Maori. Academics in Britain’s small but fractious statistics community say he rubs some ONS people up the wrong way by forever telling them how much better statisticians do things in New Zealand.
In 2000, he arrived just as the ONS had been restructured to improve the independence and trustworthiness of government data following previous lapses in accuracy. In the late 1980s the CSO – the former ONS – was lambasted for failing to predict the Lawson boom. A decade later Cook’s predecessor lost his job after earnings figures pointed to rapid wage inflation but were then revised down showing few inflationary dangers. Now, in 2003, as the ONS stood accused over the census, Cook emphatically defended his position to me, saying that the councils’ data were overstated: students, for example, registered at more than one address, or in more than one authority. “More people go to school in Westminster than there are young people who live in Westminster. Lots of people live in Westminster in second homes. Are these people resident in Westminster? It has heaps and heaps of maternity hospitals.”
No one should question the census, he said, getting agitated, because it undermines the public’s confidence in it. The missing thousands hadn’t disappeared overnight: they hadn’t been there in the first place. The councils were protesting because the ONS had revealed an uncomfortable truth. Their behaviour was “smash the thermometer if you don’t like the temperature. I am not afraid to revise or change any statistic in Britain. Look at GDP, I can upset chancellors by revising statistics but we do it… Yet I find that the bullying of some odd soul in Westminster…” he broke off, gesticulating. His public relations assistant, fumbling with a tape recorder, looked across anxiously.
After the census was published, the treasury said it backed Cook, but as complaints persisted other parts of central government started to undermine confidence in his figures. The office of the deputy prime minister, the department responsible for council finance, was barraged by complaints from local authorities. Eventually, it said that no matter what the census showed, all councils would get at least a 3 per cent increase in their budgets.
Last spring Cook finally agreed to closer scrutiny of his data. He agreed to release the ONS’s address data to a company, Manchester Geomatics, which was charged with comparing them with similar lists provided by Westminster and Manchester councils. The results of the exercise are due later this autumn.
But a wider constellation of academics and population experts were also raising questions. In July, Philip Redfern, formerly of the office of population censuses and surveys, described some of the figures as “Alice in Wonderland.” The chief doubt centred on the CCS. Was it undermined by the same problems that dogged the census itself? People just didn’t want to open their doors and answer questions, certainly not the sort of people who had failed to return their forms in the first place. It may have helped in areas where the missing forms made up only a few percentage points of households, but not in districts where a quarter of the population had ignored the census.
“If you get a response rate of 95-98 per cent and then you have the coverage survey it is very clear it will work,” Gill Eastabrook, the then chief executive of the statistics commission, told me in May. “What happened in Westminster is that they did not get anything like 90 per cent. It was in the 70s…The problem is in the inner cities. But it is not that simple. Oxford and Cambridge are quite high up the list. It might have something to do with students. This is not about undermining the census as a whole. It is about specific bits.” The commission’s inquiry into the census, conducted at Westminster’s request, is due in the second half of October.
A significant change in census data starts to unpick things way beyond central and local government planning. For example, the population shortfall has implications for European grants. The EU statistics body, Eurostat, bases its GDP per head figures on ONS figures. The latter imply that the UK is richer than we thought – its population is lower and its GDP unchanged, therefore its GDP per head is higher. But this may mean fewer areas qualifying for EU grants. Domestically, the shortfall has also led to some bizarre consequences. Because Westminster’s population was growing less than had been thought, for example, its mortality rate had to be revised upwards. Life expectancy in this part of London was suddenly three years lower. Immigration trends were also redrawn. As well as raising its figures for outward migration, the ONS cut its estimate for people coming into the country to bring it into line with the census. Critics said this was unrealistic and this summer the ONS had a change of heart. It revised the immigration figures back up closer to pre-census levels. The turnaround was the first fruit of a series of internal reviews launched by Cook following the furore. But the uncertainty it created about migration exposed the ONS’s biggest weakness – that the missing thousands could be accounted for by the belief that young men had gone raving in Ibiza.
“To suggest that 800,000 white British males had left these shores unannounced over the last decade was beggaring belief, especially as there was no evidence of them cavorting on Bondi beach,” said David Coleman, professor of demographics at Oxford University. “The influx of asylum seekers and ethnic minorities – many of whom are known from past surveys to be undercounted, especially in major urban areas – would a priori be a more plausible explanation for the shortfall on the census figures.” Illegal immigrants, who would avoid direct, doorstep measures like a census, could show up on other records, like doctors’ lists or housing records – thus possibly accounting for the difference between Cook’s count and councils’ estimates. This inflow of unauthorised immigrants – estimated at up to 50,000 a year – is one of Britain’s most sensitive policy areas; before the closure of Sangatte refugee camp near Calais it was providing almost nightly television news drama. The home office recognises the phenomenon but, by definition, has no clear measure of it. David Blunkett, home secretary, says this gap in our knowledge supports the argument for identity cards. ONS critics think Cook should devote more resources to defining this twilight population group more accurately rather than defending his census shortfall.
When I met Cook he conceded some of these broad points. But if there were errors, he said, they arose because measuring large populations in the modern world is difficult – no one could have done it better. “It is a lot harder to measure population than it was,” he said. “Britain, the US, France, Italy are very big countries with internationally mobile populations… I come from an easier wee world. In New Zealand, 95 per cent of flights come into one airport. When you go out of the country, we know. Here, 20m come in a year.”
He appeared to concede more ground by suggesting that the root of the problem might be that the ONS and the dissident councils were measuring different things. In 1991, the census takers had measured everyone. But in 2001, they measured people only where they were “usually resident” in a household. “Sometimes a large number of people in an area are not usually resident. Perhaps the census has not captured the populations of interest for public policy in Britain. Perhaps it never could.”
Cook’s challenge now is to shore up confidence in the ten-yearly census. His task is to prevent its problems leading to a broader slippage of trust in ONS statistics. For it is not only on the census that Cook is facing pressure. Recently the ONS had to make big revisions to some important economic data. First, when it was discovered that there had been massive VAT avoidance by criminals smuggling in mobile phones and computer parts, the ONS had to increase the UK’s trade gap. Then early this autumn it had to double GDP growth in the second quarter of this year because of problems with construction industry figures. Both revisions have broader implications. They mean that the Bank of England monetary policy committee has set interest rates this year based on a false picture of the economy (the committee cut rates in February and July). Such data recalculations may simply be part of the necessary process of gathering and improving statistics in a large and forever changing economy. (And Mervyn King, Bank of England governor, recently went out of his way to praise the ONS.) But they do reawaken memories of the doubts over accuracy and trust that troubled the ONS during the 1980s and 1990s. The statistics commission, the ONS watchdog, has announced an official inquiry into the GDP figures, in addition to its census inquiry.
In September, Cook announced a second set of results from his census review. After studying data like sex ratios and mortality rates, he found 190,000 people, mostly young men aged 25-34, that the census had missed. The revision effectively means that the ONS’s claim of a “one number census” is dead. Now there are two rival figures for 2001. And the statistic is likely to change again once the address-matching exercise involving Manchester and Westminster councils is complete. Cook says the revisions are simply improving on the data he had available at the census. But they have done little to lift doubts about the value of the exercise in its present form. Cook himself recently raised the prospect that in future the link between the census and mid-year population estimates, and hence public funding, might be weakened. Instead more emphasis might be placed on local authority administrative records or tax and benefit details, especially in inner-city areas. Cook suggested the census could even be augmented by “another major innovation,” like the population register ideas being discussed in the home office and the ONS.
Ironically, the enduring problems with the 2001 count may mean that after 200 years it could be the last one in its present form. It may have demonstrated that measuring population in our more mobile society is just too difficult. The census supplies an amazing array of information, down to the smallest local area, and that information, widely available via the internet, remains tremendously useful. Yet given the level of accuracy, is it worth the cost? It is certainly not safe to use this rough headcount as the benchmark for directing billions of pounds of public sector money around the country. In the end, something stricter, more tangible may have to complement it, or take its place.
“This might sound the death knell of the census in its present form,” said Coleman. “It will strengthen the hands of those who support registration-based schemes and identity cards, used in some other European countries, to keep a tag on the population.”