Radical
Statistics
REVIEWS
Statistics and policy making
by Ray Thomas
The publication of Dorling and Simpson's (Eds) Statistics in Society is notable in at least three ways. First, the book marks and celebrates the remarkable expansion over the past two decades in the availability of statistics as facts about society. Even under the Thatcher era there was steady growth in the coverage of government social surveys. There has been expansion in the number of longitudinal studies often based on administrative data (see GSS(SPH) Secretariat, 1996). Information technology developments have made the production and dissemination of statistical data easier and less costly than it used to be. In any earlier era it would not have been possible to produce a book with forty-six statistically based chapters covering such a range and variety of topics. The range and variety of statistics that are available in these closing years of the millennium has never previously existed.
Publication of Statistics in Society also showed that the Radstats Group has in effect taken over from the Royal Statistical Society responsibility for reporting on 'statistics' as facts about society. Most readers of this journal will know that the original objectives of the RSS were:
to collect, arrange, digest and publish facts, illustrating the condition and prospects of society in its material and social, and moral relations.
Developments in statistical theory have long since given the word statistics a new meaning as a singular noun. The interests of the statistician have moved away from statistics as facts about society and towards method. The journals of the Royal Statistical Society are as dominated by methodological discussion as is the Statistics in Society book dominated by its concern with society.
A third notable feature of Statistics in Society is not obvious. The book shows in many ways that the relationship between statistics and policy making is far from straightforward. Gordon and MacFarlane write in the introduction, 'the members of the Radstats Group believe that statistics can be used as part radical campaigns for social change' (p xxv, my emphasis). Gordon and MacFarlane are right to indicate limitations in the role played by statistics in social change.
Policy making or support of policy?
It is very common for statistic to be linked to policy making. The Green Paper on official statistics published last year, for example, baldly states that statistics 'help government in the formulation and evaluation of policies' (Economic Secretary to the Treasury, 1998, Chapter 1, para 1). The Green paper does not offer any argument or evidence to support of the link between statistics and policy, and later asserts that National Statistics should 'describe the state of the nation and measure the performance of government' (Chapter 4, para 1) which seem rather different matters.
This paper aims to reduce the confusion about the role of statistics. The paper argues that statistics usually make policy making difficult and that statistics are often an enemy of social reform. This argument may appear extreme or paradoxical, but, as Professor Joad famously used to say, it all depends on what you mean. In this case it depends on what is meant by policy making.
When I have been in the audience for a paper whose title includes the words statistics and policy making I have usually asked if the speaker could give an example of when the statistics have influenced policy. I haven't often had a clear answer. When there has been an answer to this question it has usually been of the form 'Yes it was suggested that we should do X, but the statistics confirmed that our existing policy Y was the best way to deal with the problem raised.
This experience indicates that use of the term policy making commonly covers both support of policies and influence of policies. The functions of statistics makes it important to emphasise the difference between these concepts. Statistics usually support existing policies. The administration of policies typically provides the rationale for the production of statistics. But this actually makes it difficult to use statistics to change policies. Use of the term policy making to cover the support of existing policies is at best imprecise use of language, and disguises the part usually played by statistics.
We usually talk of government's economic policies, for example, as being about such matters as encouraging growth of the economy and controlling inflation. Governments continually use statistics to support such policies. There is political debate on the relative priority to be given to growth and controlling inflation. But there is no evidence that statistics influence the debate between the relative priority to be given to these objectives, or to other objectives like the reduction of unemployment that used to be thought of as part of economic policy.
A section of the Statistics in Society book shows that in recent years statistics have played a prominent role in public debates about school education. But Brimblecombe, Plewis and Goldstein, who wrote chapters in this section, indicate that the production of statistics followed rather than originated the debate. The assessment of pupil performance at the ages of 7, 11, and 14, the production of performance tables for schools and the publication of schools inspection reports were the product of policy making. The subsequent debates have been structured by the statistics produced by these systems. Many doubts have been expressed about the validity and appropriateness of these statistics. But it would be difficult to maintain the argument that this debate has led to modification of the policies which led to their creation. Rather have the statistics been used to reinforce the power of those charged the execution of policies.
Statistics of crime provide another kind of example. The victory of the Conservative Party in the 1979 was in part attributed the weight given to 'law and order' issues. But crime rates grew faster in the 1980s under the Conservative government than they did in the 1970s. To some extent statistics for recorded crime reflected the emphasis given to law and order. Crime rates record policy activity. Toughening policies, increasing the numbers of policemen, and the introduction of performance measures all increase 'police activity', and that activity includes recording crime (see Reiner, 1996). Governments have to conduct household surveys in order to establish whether or not crime rates are really changing. As far as 'crimes known to the police' are concerned, it is difficult to conceive of statistics that could be independent of the policies pursued.
The evidence of participants
Let us look at some relevant historical evidence. It might be thought that Harold Wilson - the only Prime Minister of the United Kingdom who was also a statistician - would provide evidence of the influence of statistics on policy making. Wilson appointed Claus Moser as head of the Government Statistical Service in 1966 - which led to substantial developments in government statistics over the period 1966-1970.
In fact Wilson's performance as a witness is subdued and seems almost out of character - more like the performance of Bill than that of Monica. In his book The Governance of Britain Wilson gives only a footnote to statistics (Wilson, 1976, p 68). What Wilson does not say and the reference itself both give some valuable insights.
The footnote refers to Wilson's presidential address to the Royal Statistical Society in 1973. Wilson's address, Statistics and decision-making in government, gave an account of the expansion of the GSS under his leadership and gave many examples of how statistics had illuminated economic problems faced by his government. But Wilson did not say anything about policy making and referring to his earlier days as a civil servant, said:
.. I learned ... from the immortal words of John Jewkes that you can always get someone to find the answers to the questions, what you need in Government is the man who knows the questions to the answers. (Wilson, 1973, p 2).
Perhaps Wilson made this point elliptically because he did not want to denigrate the function of statistics in front of an audience that had elected him as their president. But Wilson seems to be saying that government is more a matter of articulating issues rather than using statistics. Just as in his book on governance Wilson failed to talk about statistics, so did he fail to talk about policy making in his address on statistics.
The Rayner Report on the Government Statistical Service asserted that statistics are important for 'policy formulation' (Rayner, 1980) and this assertion is consistent with some of the wording in all twenty-two departmental reports. I studied these twenty-two reports as part of a research project supported by what was then called the Social Science Research Council. These reports do not actually include any evidence that statistics influence policy making (reported in Thomas 1984a and 1984b).
Rayner's use of the term 'policy formulation' does not make the distinction emphasised here between the support of existing policies and the development of new policies. Rayner thought of government as having a prescribed set of doctrines rather than being concerned with policy making. That belief may apply well to the Government led by Mrs Thatcher at that time. But Rayner's twenty-two rapporteurs also failed to identify statistics that had influenced previous governments.
The Report on the Office of Population Censuses and Surveys provides a useful example because the putative purpose of a social survey is to produce information about society rather than information about the machinery of government. It might be expected therefore that statistics from governmental surveys were more likely to make a contribution to policy making than administrative statistics that tend to dominate the other reports.
In fact the Report on the OPCS explicitly denies a connection with policy making. The information OPCS provides is "so basic that the need for it is not greatly affected by changes in Government economic or social policy" (Witzenfeld and Craig, 1980, par. 8.14). The Report goes on to say that the need for data is a function of the pace of social change.
.. thus regionalism, abortion law reform, divorce law reform, immigration legislation and so on have all led to work being required of OPCS (par. 8.15).
Policy making and legislation is thus seen as the precursor of the production of statistics. The production of statistics on social conditions is not seen as a contribution to policy development.
Interpretative studies from the U.S
The major American study of relevance, The Politics of Numbers emphasises the reality-creating quality of statistics. Starr characterised the use of statistics as 'automatic pilots', arguing that official statistics become 'cognitive commitments' leading to collective thinking of social phenomena 'in the way that the statistical agencies and commissions have settled on' (Alonso and Starr, 1987, p 53). Prewitt, writing in the same volume, suggests that the production of statistics transforms the subject matter from matters of societal concern to those of administrative or governmental concern. '... arguments about numerical quotas, availability pools and demographic imbalance become a substitute to democratic discussion of the principles of equity and justice' (Prewitt, 1987, p 272). One of the implications of this transformation is that the problems that are not measured tend to get less attention, or are not attended to at all. 'The assumption (is made) that what is not counted is not there' (p 274).
Towards radical use
There is little that is new in this kind of analysis. These points were made in the predecessor of Statistics in Society - Demystifying Social Statistics which was published twenty years earlier (Irvine et al 1979). The pseudonymous Government Statisticians' Collective, wrote about conditions in the British civil service under the title 'How official statistics are produced: views from the inside. In the areas of distribution of wealth, unemployment, and homelessness, the Collective observed that:
What such statistics tend to monitor is not so much the social conditions of wealth, unemployment, or homelessness, but rather the operations of the state agencies responsible for dealing with the matter (Government Statisticians' Collective, in Irvine et al, 1979, p 140)
And the Collective summarised by saying:
It is extremely difficult, if not impossible, to make a really radical criticism of society using available statistical sources, which imprison us in the concepts and concerns that dominate official political and economic life (p 138).
The question that remains unanswered is why statisticians and politicians so commonly link statistics and policy making. Audrey Wise, a back bench Member of Parliament, in her keynote address at the 1999 Radstats Conference, gave one answer. Government ministers like to have statistics because they can be used to drive policies, to refute alternatives, and to dazzle and blind opposition. Ministers, with the support of the Government Statistical Service, have a natural advantage in the use of statistics. But it is not easy for government ministers to admit, or even to be aware, that they are intellectually imprisoned by those selfsame statistics.
Radical uses of statistics depend on interpretations which escape from the intellectual imprisonment that seem to easily enslave the minds of professional statisticians as well as those of government ministers. It cannot be said too often that statistics do not speak for themselves. Prevailing policies usually provide statistics giving measures of a problem area or a social condition, but statistics do not give even a hint of the policies that might solve the problem or change the social condition. Radical use of statistics depends upon finding social theories that help to explain the assumptions embedded in prevailing policies, and that help to escape from those assumptions.
Radical uses of statistics also depend upon new kinds of statistics. Audrey Wise gave the Radstats Group a wonderful example in citing the survey of attitudes to the environment conducted among four-year olds. If you are only three feet tall you have a different view of the world from those who are five or six feet tall. The value of such evidence from an otherwise disenfranchised group highlights the importance for the radical use of statistics of getting information on other groups who are missing or underepresented in official statistics. Ludi Simpson spoke on such missing statistics later in the conference. I look forward to further such reports.
REFERENCES
Alonso, W. and Starr, P. (eds), (1987), The politics of numbers, Russell Sage Foundation: New York
Dorling, Danny and Ludi Simpson (1998), Statistics in Society, London: Arnold
Economic Secretary to the Treasury (1998), Statistics - a matter of trust, Stationery Office
GSS(SPH) Secretariat (1996), Longitudinal Social Statistics: a Guide to Official Sources, London: Central Statistical Office
Irvine, J. et al (1979), Demystifying social statistics, Pluto Press
Prewett, Kenneth (1987), 'Public Statistics and Democratic Politics' in Alonso & Starr (eds.), The Politics of Numbers, New York: Russel Sage Foundation, pp. 261-274
Rayner, D. (1980), Review of government statistical services: report to the Prime Minister, Central Statistical Office, Typescript, pp 71 plus Appendix
Reiner, Robert (1996,) 'The case of the missing crimes', in Levitas & Guy (Eds.), Interpreting Official Statistics, Routledge, pp 185-205
Thomas, Ray (1984a), 'Why Have Government Statistics? (And How to Cut their Cost)', Journal of Public Policy, 4, 2, pp 85-102
Thomas, Ray (1984b), 'A critique of the Rayner Review of the Government Statistical Service', Public Administration, Summer 1984, pp 224-229
Wilson, H. (1973), Statistics and decision-making in Government: Bradshaw revisited. J.R.Statist.Soc. A, 136, 1-19
Wilson, H. (1976), The Governance of Britain. Weidenfeld and Nicholson: London
Witzenfeld, S. and Craig, J. (1980), The review of the government statistical services: Initial study of the Office of Population Censuses and Surveys, London: OPCS, Typescript
Ray Thomas
Tel: 01908 679081
E-mail: r.thomas@open.ac.uk
