Radical
Statistics

The Journal

The Subjects

The Books

News

Links

About

Home

RESPONSES TO AND COMMENTS ON GOVERNMENT PAPERS

This response to the Government's Green Paper, Statistics: A Matter of Trust (Cm 3882, February 1998) was sent to the Office of National Statistics in May 1998

(see: http://www.official-documents.co.uk/document/ons/govstat/report.htm)

Radical Statistics Group

Background

Radical Statistics was founded in 1975 by statisticians and research workers with a common concern about the political assumptions and implications of much of their work, and an awareness of the actual an potential misuse of statistics and its techniques. The group is independent from any other organisation.

The interpretation of statistics on a wide range of subjects, including countering misleading uses, is one of our major interests and a frequent subject of articles in our journal Radical statistics. Over the years, we have given particular attention to statistics about health, education, race and nuclear warfare and have published a number of books, pamphlets and articles on these subjects.

We therefore welcome the opportunity to respond to this consultation. Many of us have also taken part in consultation meetings and in compiling responses from other organisations. More generally, we are glad to see the increasing spirit of openness and interest in widening the range of users of official statistics that has developed over the past year.

The need for change

During the 1980s, we commented on many occasions about the misleading uses of statistics by government ministers. We also pointed to the declining quality of some data and the reluctance to collect data or publish analyses that might prove politically embarrassing. A major problem was the lack of any independent body to whom we could officially address such concerns. Although we wrote in 1985 to clinical royal colleges and the Royal Statistical Society to point out problems with NHS statistics, no action was taken at the time. It was not until misuse of statistics had become the subject of high profile articles in the national press and of Cooking the books, a Channel Four Dispatches programme, which drew heavily on our work, that the Royal Statistical Society and others took the problems seriously. Although necessary at that time, this publicity had the unfortunate side effect of heightening public suspicions of statistical activity as a whole. Better mechanisms are therefore needed for the future.

We therefore see the need for an independent body to oversee official statistics. In addition, arrangements are needed to limit the potential ability of future governments to undermine official statistics by curbing their scope and quality and restricting their agenda through political control and major budget cuts. There is little hope of achieving this without legislation.

The scope of national statistics

We should like to see a wide-ranging approach to defining national statistics, in a similar spirit to that expressed in the early aims and current charter of the Royal Statistical Society. These are 'to collect, arrange, digest and publish facts illustrating the conditions and prospects of society in its material, social and moral relations'. Re-interpreting this for the twenty first century raises the question of what should be included, rather than what should be excluded.

In addition to data collected by or commissioned by official bodies, we suggest that the following should be considered for inclusion:

  • Data collected outside government with public funding from bodies such as research councils.
  • Data collected about and by public services, notably the National Health Service.
  • Data about services provided by the private sector with public funds, such as health and social care commissioned from private companies.
  • Data about activities which affect the state of the population, such as private health care, however paid for.
  • Data collection systems with national coverage of at least one of the four countries of the United Kingdom.

Taking a wide-ranging view, it follows that the people involved in the production of national statistics will range far outside the ranks of those formally employed in any national statistical service. This raises issues of accountability and accreditation of data sources. The people who are employed in the statistical service itself should be public servants with public accountability, rather than civil servants accountable primarily to the ministers of the day.

We support the view that criteria are needed to ascertain the quality of data in national statistical systems and the purposes for which they are fit. In doing this, we feel that a distinction should be drawn between primary sources of data such as the Labour Force Survey and the Hospital Episode Statistics and data derived from them, such as unemployment statistics and statistics used as measures of NHS performance.

The key need is to ensure that primary sources are fully and transparently documented. As well as the more obvious issues of completeness of individual data items and response rates, documentation should cover the ways in which data are collected, the questions which are asked and the extent to which the methods used comply with what is intended. The work already done by the Office for National Statistics to develop Statbase should provide a good framework on which to build.

The documentation should also record the primary purposes for which data are collected or the administrative processes of which they are a by-product. In addition, it should record other purposes for which they might be appropriate and purposes for which it would not be appropriate to use them. This information should inform the interpretation of data derived from the source, while recognising the extent to which uses cannot be anticipated in advance. We think that the idea of trying to 'kitemark' each data item for each possible purpose to which it might be put is naive and also underestimates the amount of work that would be involved.

As well as describing existing systems, there is also a need to decide what further data are needed. The consultation document does not discuss how this might be done. As well as identifying new activities or problems for which national statistics are needed, there are also existing areas for which current statistics are inadequate.

We welcomed the demise of the 'Rayner' doctrine but are concerned about its replacement by a market model, in which statistics are described as 'products' to be sold to 'customers'. While recognising the need to recover costs, especially where additional analyses are needed by more specialised users, we also think that more emphasis needs to be put on the role of national statistics as 'public statistics' and 'The people's statistics'. It should be recognised that the basic data are collected at public expense and the public contributes by filling in forms and replying to surveys. This has implications for presenting statistics for a wider public as well as for specialists. It needs to be considered both in the context of responsibilities to the electorate and spending constraints on the public library services.

Accountability and governance

Whatever model is chosen for national statistics (see Thomas, page 86), we think that legislation is needed to ensure sufficient independence for the process of collection, analysis, and publication of national statistics and the availability of documentation which describes how they are collected. Such legislation should enshrine the principle of collecting data about the state of the nations of the UK and about the differences between geographical areas and sectors of the populations within them.

On the other hand, Model D, in envisaging a single statistical office, is too restrictive and accountability solely to a parliamentary committee would be limiting. We do not think models A or B offer sufficient safeguards. We would favour the establishment of a Independent Statistical Commission along the lines set out in Model C, but underpinned by legislation. It should have responsibility for in setting the agenda, taking into account the needs and views of various users and for auditing the quality of national statistics and the documentation of the data. The Commission should not be large, but it should have funds to employ its own staff and commission reviews and other work from outside people with specialised knowledge.

Location of responsibilities

The idea of centralising all national statistical activity within a single department is inconsistent with what we have said earlier about widening the scope of official statistics and government statistics into national statistics. Where statistical work is done within government it is often best for the people doing it to work in the same location as the people responsible for policy work and to have access to data from administrative systems. This does not of course imply the same lines of accountability and this would need to be changed to ensure independence.

Maintenance of professional standards across all statistical work in government The work required of the Head of National Statistics both in strategic development in consultation with the statistical commission and in co-ordinating a wider range of activities would be considerable. It would seem therefore that the work required is too extensive to be combined with that of the Director of ONS, in the same way as the current role of Head of the Government Statistical Service. We therefore recommend separating the two roles.

Statistics and devolution

In many fields the issues posed by devolution have already arisen. For example, NHS statistics are already collected in a different way in each of the countries of the UK. It would be important for both the Head of National Statistics and the Statistical Commission to have a relationship to the devolved bodies in Wales, Scotland and Northern Ireland as well to the UK parliament. In this spirit, harmonisation of data collection should be a specific part of the brief of the Head of National Statistics. This should also cover a relation with cross-border bodies in Ireland and collaboration with relevant organisations in the Irish Republic to produce harmonised all-Ireland statistics as well as continuing involvement with European and international statistical organisations.

This response to Statistics: A Matter of Trust, was sent to Nigel Edison (ONS)

Bristol University academics

As academics at the University of Bristol and users of official statistics, we are pleased that the issue of the integrity of these statistics is being addressed, and we welcome the opportunity to comment on Statistics: A Matter of Trust.

The Green Paper identifies a dual problem. It is concerned both with the substantive quality of statistics, and with the question of public confidence. It is our view that measures must be strong enough to ensure the that effective protection and enhancement of the statistical base, and the question of public and professional trust in official statistics should rest upon their actual trustworthiness.

The overriding need is to establish an independent and accountable National Statistics Office. In our view, no single model set out in section 5 of the Green Paper would be adequate on its own. The model suggested by the Royal Statistical Society's 1991 report Counting with Confidence should be considered as an option. The RSS model combined centralisation and accountability to Parliament with legally-enshrined autonomy and an Independent Statistical Commission with an advisory role. Models D and C in the Green Paper separate the question of centralisation and accountability to Parliament from that of an Independent Statistical Commission, and the document as a whole shows a reluctance to legislate to ensure the autonomy of the Statistical Services. Such legislation is in place in most western countries. We are in favour of such legislation, and believe that parliamentary time for establishing a suitable institutional basis should be regarded as a priority.

Model D proposes that a centralised Independent Statistical Office should be accountable to Parliament through a Parliamentary Committee. If the autonomy of the statistical services is enshrined in law, such accountability could also be effected by reporting through an individual minister. In either case, it is arguable that both the appearance and reality of political independence is best supported by a Statistical Commission.

The form of Commission proposed in Model C in the Green Paper should be strengthened. It should advise through a public report on the maintenance and improvement of the quality and range of official statistics. The composition of the Commission, the means of appointment to it, and the protection of individuals against removal for saying uncomfortable things will all impact on its effectiveness. Composition is crucial. Further discussion with different user groups, which may in turn affect the range of statistics considered important by the Commission, is also essential. At present, the attention given to government and business as primary users of statistics overemphasises statistics oriented to business in comparison with social statistics. This is exacerbated by dependence on the Treasury, which currently has too great an influence over statistical indicators. The United Nations statement appended to the Green Paper gives much greater weight to the production of statistics necessary to broad democratic debate than does the Green Paper itself. Among these might be, for example, the re-introduction of appropriate measures of poverty, and confirmation of the continuation of the General Household Survey.

We are also concerned at the frequent references to the additional costs implied by a revised institutional structure for official statistics. We think it is important that the decision on the final structure is determined on its merits, rather than on a narrow accounting basis. Inadequate statistics waste money - as, for example, in the impact of poor coverage in the 1991 Census on the accuracy and efficiency of Standard Spending Assessments. They also, as the Green Paper generally recognises, undermine the conditions required for an informed democracy.

Dr. Ruth Levitas, Department of Sociology
Dr. Danny Dorling, School of Geographical Sciences
Dr. Will Guy, Department of Sociology
Ms. Christina Pantazis, School for Policy Studies
Professor Martin Boddy, School for Policy Studies
Professor Tariq Modood, Department of Sociology
Dr. Paul Burton, School for Policy Studies
Dr. Mary Shaw, School of Geographical Sciences
Dr. David Gordon, School for Policy Studies
Dr. Brian Caddick, School for Policy Studies
Professor David Quinton, School for Policy Studies
Professor Theo Nichols, Department of Sociology
Dr. Yoav Ben-Schlomo, Department of Social Medicine
Dr. Paddy Hillyard, School for Policy Studies
Dr. David Green, Department of Mathematics
Professor Ron Johnston, School of Geographical Sciences
Dr. Harriet Bradley, Department of Sociology
Dr. Michael Slater, Department of Mathematics
Dr. Sally Barnes, Graduate School of Education
Ms. Frances Heywood, School for Policy Studies
Dr. Rohit Barot, Department of Sociology
Professor George Davey-Smith, Department of Social Medicine
Dr. Thomas Osborne, Department of Sociology
Ms Jackie West, Department of Sociology

Integrity for statistics or for statisticians? A critique of Statistics: A Matter of Trust

Ray Thomas

The Green Paper Statistics: A Matter of Trust invites views on four frameworks, which would give the Government Statistical Service (GSS) varying degrees of independence from government. Model A would strengthen existing arrangements. Model B would establish a governing board with a non-executive chair. Model C would create an Independent Statistical Commission, and Model D would have direct accountability to Parliament. In the meetings about the paper held so far Model C has been favoured.

Helen Liddell, Economic Secretary to the Treasury and Minister responsible for the Office for National Statistics (ONS), states that the Green Paper presents 'the Government's ideas for enhancing integrity, both actual and perceived, through improvements to their overall framework'. But there is no clear explanation of the relationship between integrity and the different frameworks. The main new idea in the Green Paper is the suggestion that the GSS should be responsible for producing National Statistics - which would 'be intended for public use'.

The idea of National Statistics is promising. It can usefully be seen as defining the responsibilities, which the GSS would acquire in return for a degree of independence from the government of the day. Thus the Green Paper puts a new question on the political and social agenda: What goals should the GSS serve?

But the Green Paper fudges the issue of integrity. The Green Paper uses the phrase integrity in contexts such as 'safeguarding the professional integrity of those who collect and present statistics' (Para. 1.7), independence of political influence, 'openness', 'advance publication of release dates' (Para. 3.5) and quality assurance (Para. 3.6), but there is no attempt at definition. Most points relate to the integrity of statisticians rather than the integrity of official statistics.

Government statisticians are civil servants responsible to government. But a degree of independence from the government of the day implies an increase in GSS responsibilities, and in individual government statisticians' responsibilities to the public and to society. Members of the GSS would become public rather than civil servants, and this status should be reflected in their terms and conditions of service. Statisticians would need to be able to speak up independently of the government of the day in order uphold their integrity. This point is not considered in the Green Paper, which takes the integrity of statisticians for granted and does not discuss how independence would affect the status of government statisticians under the different frameworks put forward. Rather, does the Green Paper, appear to allow, if not encourage, government statisticians to claim integrity by retreating into statistical methodology and conformity to international standards?

The leading case study - unemployment statistics

The Green Paper refers to employment statistics four times - more than any other kind of statistics. Unemployment was the enduring problem which, more than any other, led to the Green Paper. But the problems which have occurred in the past with unemployment statistics have little to do with integrity of statistics. Rather they demonstrate a lack of ability of government statisticians to speak independently of government.

With computerisation in 1982 the unemployment statistics became wholly administrative in character. Those registered as unemployed who were not entitled to benefit were excluded. The series for what was known as registered unemployment changed to what became known as the Count of Claimants. Over the next decade users complained of 30 changes in the definition of unemployment, which made the series unreliable. The term 'fiddling the unemployment statistics' came into common use. No one disputed that it is the Government's responsibility to decide who is entitled to be a claimant. The usefulness of Count of Claimants was severely limited by changes in rules and the ways they are applied. But that was a problem of consistency of coverage, not statistical integrity. It was, and is, difficult to dispute that the Count of Claimants validly and reliably covers the numbers receiving unemployment benefit.

The criticism came from two directions. On the one side those who were concerned with forecasting trends in the economy were irritated by the loss of value of the most important indicator of trends in the labour market. On the other side many concerned with social justice were angered by the power the government displayed to reduce the apparent level of unemployment by changes in the regulations which also made it more difficult for those out of work to get unemployment benefit.

The GSS failed to find a defence to these criticisms. The GSS were economical with the truth. They did not say in a clear, loud voice that the government, not statisticians, decides who is entitled to Unemployment Benefit - though members of their staff did make this point softly at Royal Statistical Society meetings. Perhaps members of the GSS were afraid that they would have been sacked had they said this in press releases or in answers to press questions? Perhaps the fear of sacking or demotion for pinning responsibility for changes in coverage on the government of the day was justified? The status of statisticians was threatened. Peter Stibbard, Director of Statistics at the Department of Employment at the time, reported "abuse levelled at my colleagues in the House of Commons, where it was said that if (his statisticians) were a football team they would be banned for bringing their profession into disrepute" (Working Party, 1995, p 405). It is ironic that Stibbard should quote a Member of Parliament. It underlines the failure of the Department of Employment statisticians to point their finger at the Front Bench figures in the House of Commons responsible for the changes in coverage.

The Green Paper implicitly and misleadingly treats this episode as if it was a matter of integrity of statistics. Members of the GSS were treated as Whipping Boys. But the episode points to the importance of the integrity of statisticians and of their terms and conditions of service and job security, not the integrity of statistics. The lack of protection of staff is not a problem peculiar to the GSS. A lot of other people felt in the dark days of Thatcherism that they had to say 'Yes, Sir. Yes, Sir. Three bags full, Sir' in order to avoid being demoted or made redundant, and today even more people suffer from job insecurity.

The Moser and Hibbert meetings

This issue is not limited to statistics of unemployment. Questions about the integrity of government statistics and the responsibilities of government statisticians were the subject of Sir Claus Moser's Presidential address to the RSS in 1980, and the subject of an RSS debate in 1990. Sir Claus was one of the first figures to put forward the idea of a National Statistical Council. On the responsibilities of the government statistician he wrote:

The government statistician commands a vast range of national information and it is his duty to deploy this to the benefit of the entire community. He/she (sic) must serve regional and local authorities; the business community; trade unions; the market research and advertising world; academic scholars and researchers; the media which are a crucial user of government statistics; and last but not least the people and their representatives in Parliament (Moser, 1980: 4).

The Rayner Doctrine, that information should not be collected for publication but primarily for the needs of government, stifled debate for the next decade. But the use of statistics became the subject of a Dispatches TV programme in 1989 called Cooking the Books. This was the first, and probably the only, mainline TV programme to be devoted to that essentially non-telegenic subject of statistics.

The programme got up the nose of those statisticians who like to think of themselves as above political conflict. The GSS called non-government statisticians to their rescue with a meeting of the Royal Statistics Society. Hibbert, Director of the GSS, at that time gave a short paper entitled Public Confidence in the Integrity and Validity of Official Statistics, leaving most of the meeting to discussion.

That discussion is wide-ranging and difficult to summarise (Hibbert, 1990, pp. 126-149). Integrity of statistics did not emerge as a significant theme, but Professor Holt of Southampton University made a key point in saying that it was very difficult to separate the integrity of statistics from the integrity of statisticians. The theme which stands out is the role which statisticians should play in commenting on, and interpreting, the statistics they produce.

Hibbert asserted that it would be wrong for government statisticians 'to pass judgement' on interpretations by ministers. But many speakers disagreed, some explicitly - including MacFarlane (p 131), Derbyshire (p 134), Wishart (p 135), Benjamin (p 137), Redfern (p 138), and Burnhill (p 139), and many more disagreed by implication. The freedom of speech for members of the GSS and the responsibility of the GSS to exercise that freedom is the crux of the matter.

How can members of the GSS fulfil responsibilities to the public, if the comments they are allowed to make on the statistics they produce are controlled by ministers? There is no point in statisticians behaving like "knights in shining armour", as Sir Claus Moser suggested they should (Moser, 1980, p 130), if the knights have to wear visors all the time so that they cannot communicate. GSS job descriptions should specify quite clearly that members of the GSS should interpret the statistics as part of their responsibility to society.

That kind of matter is not raised in the Green Paper. The Green Paper refers to "a critical RSS Report" (Para 2.5) published in 1990. But that reference is not the Hibbert meeting, but is a misstated reference to the paper Counting with Confidence published in 1991- which might be more accurately referred to as an uncritical report! The Green Paper, by giving main attention to the integrity of statistics, side-steps the more important area of the integrity of statisticians and their responsibilities to society.

The idea of national statistics

The idea of National Statistics put forward in the Green Paper has the hallmarks of being a last minute inspiration. The discussion of National Statistics is brief and GSS-centric - but open-ended. The Green Paper rejects the 'Rayner doctrine' that 'statistical information should primarily be collected to satisfy the needs of government'. The Paper states that 'priorities should be driven by the requirements of all users - Parliament, government and the wider community', and asks 'Are there activities or outputs which should be specifically included or excluded from the definition of National Statistics?' (Para 4.7).

One remarkable feature of this question is that it has never been asked before. This seems to be the first time that there has been opportunity for public discussion of the responsibilities of the GSS or the relationship between official statistics and the public interest. The other remarkable feature is that there is not much discussion of the idea in the Green Paper. The Green Paper does not present alternative models for what might be included in National Statistics as it does for the frameworks for the governance of statistics.

The paucity of the discussion of National Statistics probably reflects the grinding down of the Government Statistical Service by eighteen years of Conservative governments, which began with the Rayner Reviews of the GSS in the early 1980s. The substantial but willy-nilly growth in the range and variety of official statistics, which has occurred since then has not been underpinned by any clear strategy. There has been no statistical priority-setting.

Then, after 18 years of distrust and oppression under the Tories, the GSS was suddenly 'liberated' by the new Labour Government. We won't interfere' said Helen Liddell, 'The ONS will take on board the needs of users...the Government is committed to ensuring the widest possible acceptance of these statistics' (letter of 4th July 1997 in response to suggestions about labour market statistics). Anyone attending a meeting with ONS statisticians since May 1997 is aware of the change. Most members of GSS are now genuinely interested in users' views.

The GSS is not fully prepared for the new responsibilities which it has been given. Users' views alone do not provide an adequate basis for development of a strategy for statistics. Eighteen years of distrust and oppression does not provide a good basis for the GSS to deal with the real responsibilities which would be the counterpart of a degree of independence from the government of the day. The GSS does not have the ideas, and has not developed the strategies, needed to put flesh on the bare-bones concept of National Statistics.

It would be a very positive development if the Green Paper led to firm commitments by the GSS to provide statistics which approached the original aims of the Royal Statistical Society, which, according to its Charter, are 'to collect, arrange, digest and publish facts, illustrating the condition and prospect of society in its material, social, and moral relations.' Such broad aims would probably prove to be as over-ambitious for the GSS as they were for the RSS. But within such a broad canvas it is possible to pick out areas which deserve special kinds of attention. I would argue, for example, that the GSS should have responsibilities for producing statistics on inequalities in society.

Statistics of inequality as an example

I would argue that one of the functions of government should be to maintain a degree of equality in society. This function cannot be properly exercised unless it is supported by adequate statistics. So I would also argue that GSS responsibilities should include the production of statistics of inequality in incomes, wealth, access to employment, and pension provision.

The GSS of course already produces many statistics relevant to these inequalities. But the GSS has never seen the production of such statistics as a responsibility - which it might pursue even against the wishes of the government of the day. In recent years it has been left to the Rowntree Foundation to support the production of statistical reports such as Income and Wealth: the latest evidence (Hills, John, 1998), and to initiate a programme such as that recently launched for the establishment of key indicators of poverty and social exclusion (Howarth and Kenway, 1998). The responsibilities to go with a more independent GSS should include the production of statistics of inequality and poverty, and publication in an appropriate form such as the annual Social Trends. Organisations such as Rowntree would then be free, in accordance with their charitable aims, to support developments of more direct value to policy makers and practitioners.

A Royal Commission was set up nearly a quarter of a century ago to make 'a thorough and comprehensive enquiry into the distribution of income and wealth' (Diamond, 1975). The work of the Diamond Commission was buried by Thatcherism, and is long overdue for revival. Statistics of wealth pose particular difficulties. The problem, as pointed out by Diamond (p 69) is that we do not have good statistics on wealth because wealth is not taxed - so there are few administrative statistics. But it is also true that we are unlikely to have a tax on wealth unless there are statistics on the distribution of wealth, which provide a basis for planning such a tax and forecasting yields. A serious attempt to measure the distribution of wealth would be a necessary first step.

The GSS already produces many statistics which indicate the degree of inequalities in incomes. But though poverty is defined in relative terms, the Households with Below Average Incomes Series, as pointed out by Peter Townsend, are not really comparative (Townsend, 1996, p 37). The statistics of unemployment - the Count of Claimants and the LFS Series - use definitions which do not cover joblessness (as the Government has fully acknowledged in its welfare to work programmes). If the purpose is to show inequalities, there is plenty of room for new statistics and new forms of presentation on the distribution of incomes and on employment/unemployment/inactivity - especially showing geographical variations.

The provision of pensions is currently high on the political agenda. A Green Paper is expected shortly The crisis that has put this item on the political agenda is the cost of state pensions and the failure of many private arrangements, as well as equity between members of different income groups. This debate is also likely to raise questions about equity between different generations - where Thomson finds the statistics available from published sources in Britain inadequate as compared with those for New Zealand (Thomson, 1995, p 8, 35, 37, 124). The area of provision for pensions is one which cries out for support with good statistical evidence - both for point-of-time and intergenerational comparisons.

If it were accepted that GSS responsibilities should include the aim of production of statistics on inequalities of these kinds, there would need to be substantial support from outside government. This would probably favour some kind of National Statistics Council as envisaged in model C of the Green Paper, including representatives of bodies such as the Economic and Social Research Council - for its knowledge of the problems of trying to measure inequalities, and the Trade Union Congress - for its interest in the extent of such inequalities.

Administrative statistics

One promising area opened up by the Green Paper is the matter of access to administrative records. The GSS has special responsibility for the production of statistics based on administrative records. But this responsibility cannot be fully realised without access to the administrative records of government for the production of statistics beyond the immediate administrative needs. In the Green Paper, this issue is seen to relate to the power of the GSS within the machinery of government. The Green Paper points to the need for the GSS to 'require access' to administrative sources, and the need for members of GSS to 'participate in the development of administrative systems and research programmes' (Para 4.4).

One of the areas where this advantage is well used is for the production of statistics for Claimant Unemployment for post-code areas, and for the Joint Unemployment and Vacancies Operating statistics database (JUVOS). But in general the GSS does not fully exploit administrative data for the production of statistics. Nearly twenty years ago Sir Claus Moser pointed to the wealth of data potentially available from administrative sources, and to the advantages of exploitation of this source (Moser, 1980, p 21). But it is doubtful whether there has been major improvement over the past two decades.

There is still a high degree of departmental decentralisation within the GSS. Neither the Department of Inland Revenue - the major earning department, nor the Department of Social Security - one of the main spending departments, is part of the ONS. Such departmentalisation is associated with a lack of free flow of statistical information within the governmental machine, and would be a problem in the production of more comprehensive statistics on inequalities. There is little evidence that Moser's call for data links between departments has been widely heeded. The muted plea in the Green Paper for access by the GSS to administrative records should therefore be strongly supported.

Centralisation and survey statistics.

The Green Paper devotes a chapter to the "Location of Responsibilities" which discusses in very general terms the pros and cons of centralisation of the organisation of statistics within government. The advantages of centralisation are much more apparent for administrative statistics than they are for surveys. One of the questions which the Green Paper puts on the political agenda is whether the ONS or even the GSS itself should continue to have direct responsibility for the conduct of social surveys and the census of population.

The Office of Population Censuses and Surveys (OPCS) sets very high standards for the conduct of the census and a wide range of social surveys. But many people do not have faith in the pledges of confidentiality given by interviewers working in central government uniform. Under-enumeration in the 1991 Census may well be symptomatic of lack of faith in the pledge of confidentiality, which was supposed to be given by enumerators. The merging of the OPCS with the Central Statistical Office to create the ONS in 1995 may well have become a handicap in the conduct of surveys. The growing number of record linkage studies cannot contribute to confidence in these pledges.

The integrity of survey and census statistics are matters of research method and management and do not depend upon whether or not they belong directly to central government. It may be that the response rates and the quality of responses in government surveys would be enhanced if the surveying agency were an independent entity within the GSS rather than centralised within the ONS.

The general charges made against official statistics, especially by sociologists, are not to do with integrity, but are more closely related to matters of categorisation and coverage. A recent study of official statistics (and the only study published for many years) characterises official statistics as representing the interests of the state not the citizen (Levitas and Guy, 1996: 6). A standard dictionary of sociology says that the categorisations of official statistics reflect the interests of government and that they may not be meaningful to sociologists (Abercrombie et al., 1994: 290).

It may be difficult to meet the criticisms of the sociologists in production of administrative statistics, which are necessary for the conduct of government. But social surveys are a different matter. The ONS programme of harmonisation of survey questions (ONS, 1995) seems to be being conducted independently of the sceptical sociologists, but there are no good reasons why the design and conduct of social surveys should bow to administrative necessity.

An independent organisation for the conduct of governmental social surveys was recommended by the Heyworth Committee more than thirty years ago (Heyworth, 1965, pp. 43-44). There is now a need to consider whether the work of the OPCS should be part of such an independent organisation, so that the conduct and output of surveys can be judged by academic standards, as well as those which prevail in the GSS.

Dangers associated with independence

The history of the Royal Statistical Society (RSS) itself gives a warning of a danger with the granting of a degree of independence to the GSS. The expectations raised by the idea of National Statistics and the association of the Green Paper with the idea of freedom of information are high. But the history of the RSS suggests that the emphasis on integrity might have unintended consequences which might lead to development in quite different directions.

The RSS has almost completely abandoned its original aims concerned with the production and publication of facts about society. This abandonment is in part a logical consequence of the increased responsibility taken by the GSS for the production of statistics about society. But the growth of interest in statistical methodology has also reduced the interest in facts about society in the RSS.

The danger with the emphasis on statistical integrity in the Green Paper is that granting independence to the GSS may also lead to a move away from collecting facts about society - toward the production of statistics which are prized for their integrity. The GSS can claim integrity on the grounds that the accuracy of a statistic can be stated with precision, or on the grounds that a statistic conforms to an international standard. Such claims can be made independently of the contribution the statistics makes to an understanding of conditions of society.

There have already been several worrying moves of this kind which have occurred in unemployment statistics in the past year since the new government was elected. One example is the increased emphasis being placed on the Labour Force Survey unemployment series as against the Count of Claimants. The LFS measure has important advantages (see Working Party, 1995 and ONS, 1998), but it also has serious limitations (see Thomas, 1997a, 1997b, 1998a, 1998b). The ONS has not made any serious evaluation of the LFS series as a measure of unemployment and has not acknowledged any limitations in the series. The LFS series is promulgated, not because it can be shown to provide useful information about society, but because it is believed to conform to internationally accepted criteria established by the International Labour Office.

One problem with the Labour Force Survey is that does not produce accurate information about claimant unemployment. Governmental and public knowledge of unemployment problems is therefore fragmented because the statistics from the LFS cannot be reconciled with those of the Count itself. A recent record linkage study (Pease, 1997) has shown that the discrepancy was larger than previously thought. But an early response of the ONS to this new evidence of the magnitude of the problem has not been to try to improve the quality of the LFS, but to model the discrepancy (see Haworth and Bowman, 1998).

It is difficult to argue that the production of statistics based on models which make inconsistent statistical series appear to be consistent with each other is in the public interest. There are many unknowns in the relationship between the claimant system, and other components of the social security system, and labour market behaviour. The ONS should focus its work on the provision of statistics which contribute to an explanation of such relationships.

Another problem which the LFS shares with the Count of Claimants is that neither Series takes account of the influence of changes in the social security system which affect changes in both Series. The current coverage of both the Count of Claimants and the LFS unemployment series is quite different from the coverage in the 1980s because more than a million people were switched from the Count to Sickness Benefits. This movement was investigated and estimated by the Sheffield Hallam team (Beatty et al., 1997). But the movement has not been acknowledged by the GSS.

The interaction of the social security system and the unemployment system will continue with the Government's welfare to work programme. The government has come to use the term joblessness to include those whom it regards as unemployed but who are not included in either series. People classified as economically inactive by the LFS Series are being encouraged to claim Job Seekers Allowance. It will become more difficult to interpret patterns shown by the Claimant Count and by the LFS Unemployment Series.

Data on movements of this kind between the different claimant statuses could be provided from administrative data held by the Department of Social Security databases. But the Labour Force Survey in its present form does not aim to provide information on the relationship between the social security system and the labour market, and so has little to contribute to the development of government policies in this area. The LFS continues to be geared to meeting ILO criteria, and if integrity is interpreted by the GSS as conformity to international standards, as presently appears, then the LFS will be of little direct help to the government in its current welfare to work programme.

The promulgation of the LFS unemployment series also militates against the use of the Count of Claimants statistics for the investigation of unemployment at a local level. The ONS provides statistics for the number of claimants in fine geographical detail - down to postcode level. But there is a problem in finding a suitable denominator in order to calculate an appropriate unemployment rate. The traditional solution has been to use workforce rates in which the main item of the denominator is the number of people employed in the area.

The use of workforce rates for local authority areas gives a systematically biased picture of the location of concentrations of unemployment through understatement of unemployment rates for cities and overstatement for suburban and rural areas (see Webster 1998). The GSS has avoided this problems in the past through the use of relatively self-contained Travel to Work areas. The use of TTWAs avoids systematic large scale bias, but only at the expense of the concealment of the growth of concentrations of unemployment in inner city areas (see Webster, 1997, Webster and Turok, 1997, Thomas, 1998a)

Government statisticians have been compliant with this concealment on the ground that TTWAs are consistent with the European standard. The danger with the Green Paper is that the concept of integrity would be used to defend the continued use of TTWAs for the measurement of unemployment, in spite of its weaknesses, and that the demand for more detailed information is met by publication of the systematically biased workforce rates for local authorities.

Next stages

The Green Paper gives a quite unprecedented opportunity for debate and the expression of points of view. Nigel Edison, who was nominated in the Green Paper to deal with comments (see Para 1.8) pointed out that:

'The Green Paper offers the opportunity to make radical changes in official statistics. Nothing is ruled out at this stage. A Green Paper on statistics only comes round every 50 years or, so we are definitely not going to waste the opportunity' (email dated 13th March).

It may be ineffective to complain in the year 2001 about the gaps in coverage of official statistics. The reply will be 'People had their opportunity to make proposals in 1998, and they did not ask us to fill that gap'.

In my view the Green Paper is disappointing because it does not address the important issues. The arguments in favour of a GSS with some independence from the government of the day are very strong. But the proper exercise of such independence requires that Government statisticians have responsibilities to the public and to society. Just as doctors used to swear the Hippocratic oath to focus attention on the health and well being of their patients, so should government statisticians have the responsibility of focusing attention on the health of society. The GSS should have sufficient independence from government to allow it to fulfil this responsibility.

The idea of a National Statistics Commission as described in Model C would probably provide an effective governmental framework for the exercise of such responsibilities by the GSS. But such a reform by itself will achieve little, and there is a real danger that the GSS moves away from providing statistics which give information of value in understanding society, towards the production of statistics which are methodologically defensible or which conform to international standards. The government must articulate clear responsibilities for the GSS, and provide terms and conditions of service for its individual members which will enable them to act as public servants rather than civil servants. This paper has drawn attention to a need for the provision of statistics of inequalities in society to be a part of the responsibilities of an independent GSS. The paper has focused on the importance of the public servant theme with evidence from the area of unemployment statistics.

I imagine that there are parallel situations in other statistical areas. I hope that this paper will provide a stimulus for others to make a contribution.

REFERENCES

Abercrombie, N., S. Hill and B. Turner (1994), The Penguin Dictionary of Sociology, Penguin Books.

Beatty, C., S. Fothergill, T. and A. Herrington (1997) The Real Level of Unemployment, Centre for Regional Economic and Social Research, Sheffield Hallam University.

Diamond, Lord (Chairman) (1975) Royal Commission on the Distribution of Income and Wealth, Report No. 1, Cmnd 6171, HMSO.

Haworth, M. & L. Bowman (1998) Improved ONS Labour Market Statistics, paper given at the BURISA conference, London, April 1988, typescript.

Heyworth (Lord) (Chairman) (1965) Report of the Committee on Social Studies, Cmnd. 2660, London: HMSO.

Hibbert, J. (1990) 'Public Confidence in the Integrity and Validity of Official Statistics' J. R. Statist. Soc. Series A, Vol. 153, Part 2, pp. 123-150.

Hills, J. (1998) Income and Wealth: the latest evidence, York Publishing.

Howarth, C. and Kenway, P. (1998) Monitoring Poverty and Social Exclusion: Why Britain Needs a Key Indicators Report, London, New Policy Institute.

Levitas, R. and W. Guy (1996) Interpreting Official Statistics, Routledge.

Moser, C. (1980) 'Statistics and Public Policy: The Address of the President', Journal of the Royal Statistical Society, Series A (General), Vol 143(1): 1-31.

ONS (1996) Harmonised Concepts and Questions for Government Social Surveys, Office for National Statistics (Update published Nov 1997).

ONS (1998) How Exactly is Unemployment Measured? (Third Edition) Office for National Statistics.

Pease, P. (1997) 'LFS estimates of unemployment-related benefits: the results of an ONS record linkage study', Labour Market Trends, 105 (Nov): 455-460.

Thomas, R. (1997a) 'The Labour Force Survey in the Dock' Radical Statistics, 64/65, Spring & Summer, pp. 4-12.

Thomas, R. (1997b) 'An Integrated Unemployment Series and Beyond', Radical Statistics, 66, Autumn, pp. 6-16.

Thomas, R. (1998a) 'How enlarged Travel to Work Areas conceal inner city unemployment' Radical Statistics, 67, Spring 1998, p 35-45.

Thomas, R. (1998b) 'ILO unemployment and registered unemployment - a case study of Britain', Bulletin Methodologique de Sociologie, Vol. 55 (May).

Thomas, R. (1998c) 'The Politics and Reform of Unemployment and Employment Statistics', Chapter to be published in Dorling and Simpson (Eds) (forthcoming) Statistics and Society, Arnold.

Thomson, D. (1996) Selfish Generations? How Welfare States Grow Old, White Horse Press.

Townsend, P. (1996) 'The struggle for independent statistics on poverty', in Levitas & Guy, Interpreting Official Statistics, Routledge, pp. 26-44.

Treasury (1998) Statistics: A Matter of Trust. A Consultation Document. The Stationery Office.

Webster, D. (1998) 'Through a Glass Darkly: Local Unemployment Statistics and the Diagnosis of Britain's Unemployment Problem', paper given at BURISA Conference, London, 24 April, typescript.

Webster, D. (1997) 'Travel-to-Work Areas and Local Unemployment Statistics: a Glasgow View', in Turok, I. (Ed) Travel-to-Work Areas and the Measurement of

Unemployment, Conference Proceedings, University of Glasgow Department of Urban Studies Occasional Paper 38.

Webster, D. and I. Turok (1997) 'The Future of Local Unemployment Statistics: The Case for replacing TTWAs', Quarterly Economic Commentary, Vol. 22, No 2, March, pp. 36-40.

Working Party on the Measurement of Unemployment in the UK (1995) >The Measurement of unemployment in the UK (with discussion)=, J. R. Statist. Soc. Series A, 158. Part 3, pp. 363-418.

Ray Thomas
The Open University, Walton Hall
Milton Keynes, MK7 6AA
email: r.thomas@open.ac.uk

Journal 069 Index Top of page