Numbers games

A new bill proposes independence for the Office for National Statistics—but it doesn't mean what it says
March 22, 2007

Independence for the Bank of England has been one of Gordon Brown's most successful policy initiatives. Independence for the Office for National Statistics seems an equally admirable proposal. But independence for the Bank of England meant, broadly, what was said. Independence for the Office for National Statistics means, broadly, the opposite of what is said. What a difference ten years in office makes.

The current statistics and registration service bill was published at the end of last year, and had its second reading in parliament in January. It promises "a new independent statistics board, outside of ministerial control" for the ONS, which will include several non-government board members.

This legislation comes at the end of a 25 year period in which the credibility of all government information has been eroded. Episodes like BSE and the treatment of intelligence reports on Iraq mean that even ostensibly factual official statements can no longer be considered objective accounts of events.

article body image

Official statistics have also been through several organisational upheavals. The Rayner report of the early Thatcher years, drawing on a mistaken analogy between official statistics and management information systems, argued that reduced resources for government statistical services should be targeted on the needs of government departments. But official statistics are not the same as government statistics. As well as meeting the requirements of policymakers, they feed the process of democratic accountability.

The operation of bias and political influence is subtle. Government statisticians are honest people. They don't make numbers up. They do not accept instructions from politicians as to what numbers to report. But independence can be compromised in more complex ways.

It sounds easy, for example, to count the unemployed. But the spectrum of people who could be considered unemployed ranges from those who would quite like a job if a decent offer came along to people who devote themselves full time to the search for work. The two main ways of measuring unemployment—reviewing benefit claimants and surveying households to ask whether anyone in them is actively seeking work—give different answers. The question "what is the level of unemployment?" has many different answers, none clearly right or wrong. Each of the many adjustments made to unemployment statistics in the 1980s was, on its own, individually defensible, but the overall result was to present a steadily less gloomy picture and to make comparison over time increasingly difficult.

More recently, the treasury has twice shifted the goalposts for assessing the "golden rule"—for measuring the tightness of fiscal policy. Again, the alternative measures are defensible. But no one can overlook the fact that the guidelines would have been breached if these amendments had not been made.

Similar stories can be told across the range of government statistics—the confusing plethora of inflation indices, the exclusion of PFI and other public liabilities from government borrowing, the measurement of public sector output and efficiency. Most statistical questions can be answered in more than one way. If some answers are greeted with approbation and others are not, then very little knowledge of human nature is needed to realise what will happen.

What is collected matters as much as how it is collected. We live in a world in which service industries predominate, most economic growth takes the form of quality improvement and environmental concerns multiply. National statistics have adapted too slowly to such concerns.

For example, the absence of good information on immigration from the accession states of eastern Europe creates a large gap in our understanding of labour market trends. The EU and the British government understated the potential impact in the years before accession, and have subsequently failed to monitor it. Both policy and politics require the collection of relevant data, not simply the data that has always been collected, or the data that politicians want collected.

These issues are, for most people, boring. The difference between arithmetic and geometric means of price relatives will never make the Today programme. I have struggled for years to interest people in the impact that hedonic pricing of computers has on the difference between US and European GDP, without success, although the issue probably accounts for more of the difference in reported productivity growth rate than any difference in economic policies.

The only way of resolving these questions objectively is through skilled, independent professional judgement. The Bank of England and the judiciary act independently primarily because prestige and preferment depend on the esteem of colleagues, not the approval of politicians. Independence is a state of mind, and the state of mind is the product of institutional arrangements, but is not just a matter of institutional arrangements. If a judge had to notify the home office of a decision a week before delivering a verdict in court, or the monetary policy committee had to give the chancellor of the exchequer 48 hours' notice of a change in interest rates before announcing it, then the relationship between politicians and these public servants would be changed fundamentally, and it would be changed whether or not the politicians enjoyed the final power to require that the decision be modified. That is why the issue of pre-release of data to ministers—a practice almost unique to Britain—is so important.

The leading individuals in the Bank and judiciary—the governor, the lord chief justice—enjoy high status which does not depend on public approval. Career progression in these bodies is outside the normal channels and criteria of the civil service. Neither body is engaged in a battle for resources which can be damaged by unhelpful public comments.

These things were once true of national statistics. They are not true any more, and the government's proposals for so-called independence will keep it that way. The role of the national statistician is to be diminished rather than enhanced. Statisticians will continue to shift between ONS and government departments, and responsibility for the greater part of government information will remain within departments. Pre-release will continue, and the treasury will continue to hold the purse strings. The main existing source of professional oversight of ONS work—the statistics commission—is to be abolished, having made the mistake of actually being independent.

The commission will be replaced by a non-executive board. If that board were appointed by and accountable to parliament rather than government, if its responsibilities were primarily professional rather than managerial, if the board were composed mainly of distinguished academics and professional statisticians, if its members devoted sufficient time to engage with questions of detail and if the board had proper oversight of the statistical work of departments as well as of the ONS itself, then it would be credible that its supervision could give junior statisticians the confidence needed to insist that professional judgements were made and presented in a detached and objective way. But none of this is intended, or likely.

The blunt truth is that the proposals to give greater independence to national statistics consist of little more than repetition of the phrase independence for national statistics. Since almost everyone is—publicly—in favour of "independence for national statistics," respondents to the bill's consultation almost unanimously began by welcoming the principle of independence while criticising the specifics. The government has claimed this shows wide support for the proposals but, since there is an almost complete disjunction between the stated objectives of the bill and the substantive content, this response is disingenuous. These proposals, and the manner of their presentation, only compound the mistrust they claim to alleviate.