It does matter. As the Prime Minister said in 2002: “It is important for our statistics to be good. But it is just as important for the statistics to be trusted by all those — Government, the Opposition parties, pressure groups, citizens — involved in debating policy. The alternative is that debate degenerates into disputes about the facts, with the risk that the process of making decisions and monitoring delivery may be compromised.”
Such problems are not unique to the UK. The Statistics Commission has noted similar public concern in countries as diverse as Australia, China, New Zealand, South Africa and the US. India has announced that it is setting up a statutory commission to oversee its statistical work. Nevertheless — for the reasons given by Tony Blair — we still need to ensure that a high measure of public trust is both deserved and achieved.
Before outlining our proposals in the Statistics Commission, it is essential to sketch how statistics are produced, made available and interpreted in Britain. Since 2000, we have had two types of statistics — “National Statistics”, which should be assembled and published in accordance with a Code of Practice introduced by the National Statistician, and other official statistics, which can be produced and issued in any manner decided by the producer department. Departmental ministers decide which statistics are to be treated as National Statistics.
There are also, in effect, two kinds of producers: the ONS produces many of the most important economic series (eg, GDP), but produces only about a third of the total of National Statistics series. The rest, including key series on health, education and crime, come from Whitehall departments or the devolved administrations.
The National Statistician, who answers to the Chancellor of the Exchequer, is expected to act simultaneously as the Permanent Secretary head of the ONS — and hence a key producer of statistics — and as an independent guarantor of quality for all official statistics, whether managing them or not.
There are misconceptions that statistics are like football scores. In reality, many are based on estimates that involve professional judgments. There is often a trade-off between speed of production and accuracy; many so-called revisions are actually updates as new information becomes available.
How they are presented and interpreted can destroy faith in statistics and the system producing them. Newspaper columnists and politicians are the main channel by which the public hears the statistical message. Bad news is always much more newsworthy than good. Elected leaders of any administration will wish to use statistics to make the best possible case for their achievements; opposition leaders will seek opportunities to use the statistics to denigrate the administration. These are immutable facts of life.
Given all this, what can be done to improve public trust in British official statistics? Our research tells us that no single organisation or individual is responsible for the malaise. So no “silver bullet” exists; multiple actions need to be taken to improve the situation by multiple players. The Statistics Commission is developing a plan whose details will be debated with the National Statistician and other parties. We have so far concluded that actions are needed in five areas.
- Independent production and publication
We must ensure that politicians cannot influence statistics and that this is widely understood and accepted. Some greater independence for the National Statistician and giving her greater authority over the actions of statisticians in all government departments and agencies would help. But there also needs to be a separation of the National Statistician role from that of “main producer” of official statistics. We believe that ministers should no longer make decisions on which statistics should be National Statistics and that they should not have earlier access to statistics than anyone else. Publication of official statistics should always be pre- announced as part of a National Statistics timetable. Finally, there should be a new, tougher (and statutory) code of practice, binding on departments and their ministers and enforceable through a new Statistics Commission reporting to Parliament.
- Producing what is really required
We need a much more effective consultation mechanism, involving surveys to establish user needs and satisfaction with statistical outputs. These must cover all user groups, not just central government. All official statistics must be produced using publicly specified methods. We need independent audits of quality and publication of their results. Benchmarking of outputs and statistical processes against best practice overseas, again with publication of results, is highly desirable.
- Independent interpretation
It is essential, though not easy, to ensure that reasonable messages are consistently extracted from the statistics. Clearly no one can control the use of statistics in political debate. But statisticians should provide publicly more interpretation about what data mean and how far they can be taken. In the Statistics Commission’s view, all statistical series should be published with a frank commentary written in plain English, and not in statistical parlance. This should include some indications of the data’s limitations — eg, whether a small change is really a meaningful one or likely just to be a “blip”. Any ministerial press statement on the figures should be released only after the “basic facts” and these interpretations are published.
- Effective planning
Resources will always be in short supply. New information is needed in some areas, while existing statistics may no longer be essential in others. So, effective and efficient use of public resources to create official statistics can be achieved only through co-ordinated planning across all producers of official statistics, rather than simply being a matter for each individual department to address as if its statistical decisions had no bearing on anyone else. This could be done under the umbrella of a new National Statistics Strategic Plan agreed across government. Such planning should also ensure that the National Statistician has access to administrative information collected in government bodies. This would maximise data quality and minimise cost and the burden on respondents to surveys.
- Educating the users
Some commentators have only rudimentary understanding of basic statistics, probability and statistical error. It is no surprise that they misinterpret published figures. Quantitative skills need to be enhanced, not only in schools, but across society — which is inevitably a long-term objective.
Can all this be done? Some of it would be enormously simplified if we had an over-arching statistical law, as have more than 100 other countries. The Statistics Commission has made the case for such an Act, which can be implemented within British conventions and with minimum disruption. The Government is about to review the Framework for National Statistics after five years of operation. The Cabinet Secretary has pledged support for “anything that would entrench traditional Civil Service values of integrity and impartiality”. This is, therefore, a good time to address the desirability of legislation and the host of additional ways in which we can justify and bolster public trust in official statistics. As the Prime Minister said, such trust is essential to our democracy.
David Rhind is chairman of the Statistics Commission
