Radical
Statistics
Compromises of the Fourth Kind with Statistical" Radicalism"
Roy Carr-Hill
Brought up in a middle class Tory household, I was a 'natural' for the so called 'excesses' of the sixties: arrested for playing bridge on a parking meter with the Provo's in 1962 in Amsterdam; anarchist elected to first ever student union at Cambridge (you see the contradictions started early); Berkeley 1966-1967 including Monterey with Janice Joplin, draft card burning and blocking of Oakland recruiting office; 'writing a thesis' in Paris Jan-August 1968 (someone had to be the other foreign agitator); bit of realism when on holiday in Prague as the tanks rolled in; and so 'retired' to do D.Phil in Penology at Nuffield College, Oxford in September 1968 for three years.
Kept head down - apart from persuading college chaplain to baptise half a dozen collective pink bicycles in college pond (to delight of local journalists) and demonstrating that variations in police practice were mainly responsible for variation both geographically and over time in recorded crime statistics (Carr-Hill and Stern 1978) - until heard rumour that the 'movement' was looking for a 'radical statistician' (clearly a contradiction in terms) to help defend a group accused of a conspiracy to plant bombs. An important part of the evidence against them was a statistical argument along the lines that, among all bomb incidents that had been recorded during 1968-1971 (some 1000 apparently), the 25 which were the subject of the conspiracy charge were "statistically different" (using a chi square test) on a number of criteria (composition, modus operandi, size, type of target, etc); so that if each one of the group being charged could be linked (with circumstantial evidence) to just one of the incidents, they were 'therefore' guilty of the conspiracy charge of the whole group of incidents. Apparently the same nonsense had already been deployed, successfully, against another person (who had receive a sentence of 25 years), and several gangs of safe-breakers. I agreed to give evidence and, in fact, helped their defence in court, in what turned out to be the longest running criminal proceedings in 1972, not only in respect of the statistical 'argument' but also in combating the scientific (mostly chemical) and pseudo-scientific (graphological) prosecution evidence.
The main outcome measures (as we would say these days) is that 4 of the 8 were found not guilty and the other four received 'only' ten year sentences (which led to the reduction of the previous 25 year sentence handed down). However, as a sideline, together with a recently recruited programmer, I explored whether or not there was actually any clustering at all in the 1,000 incidents and submitted a paper to the JRSS Applied Statistics Conference held in the Summer of 1973 at Hull (Carr-Hill and Hopkins 1973). I learnt at this meeting (27.02.99) that the submission of the paper had caused some consternation (which I knew nothing about); the first I knew that anything was amiss was half way through my presentation when I was explaining that the Home Office Statistician giving evidence for the prosecution was either so obviously incompetent that he shouldn't have the job, or being distinctly economical with the truth (though I doubt if I was that polite). I was asked by the Chairman to 'cease and desist'. To qualify what ensued as uproar would be a gross exaggeration: statisticians of all kinds are still far too polite to raise their voice. But there were audible mutterings led, especially, by John Bibby to the effect that we should devise a code of good practice for statisticians and especially for those employed on public service; and this was one of the triggers for the founding meetings of Radical Statistics, the following year.
Personally, however, it became clear that I was very unlikely to be granted access in the future to any sensible data (I had by now finished the data collection for my D.Phil showing very significant variations in sentencing between judges at Quarter Sessions) and so I abandoned both criminology and penology as a career. Equally my probationary period at the University of Sussex (a very respectable establishment Marxist University at the time) was not renewed, partly I suspect because of the above, partly because I organised the opposition to a talk being given by Huntington (architect of the forced draft urbanisation policy in Vietnam), but partly because I had attempted to organise the compulsory Arts - wide Elementary Statistics course around the useful skills of interpretation rather than desperately trying to inculcate t tests etc. to students, many of whom had 'chosen' to do social sciences' because they didn't like or couldn't do mathematics. My 'compromise' had been decided for me.
Amidst all this mayhem, I was invited to what looked like an interview with the OECD in Paris, was offered a job, and accepted to work on the development of Social Indicators of Well Being (seen as a counterpoint to the then - discredited GNP measure). It later transpired this was a version of being kicked upstairs; but for three years I speculated with many like-minded young civil servants from around OECD as to what a useful statistical system would look like, writing in the process one of the first Rad Stats pamphlets Social Indicators for Individual Well Being or for Social Control (Anon 1978) perforce anonymously given our position. The Rad Stats pseudo-liberals of the time - not having yet encountered the cutting edge - thought this was awfully conspiratorial.
Got sacked from that job in 1977 - almost certainly not because of the above, but for other political adventures - and became an unemployed housebuilder in the South of France. Together with French wife decided to volunteer for Mozambique and on the strength of her paediatric qualification, we were accepted for a 2 year contract. In Mozambique found myself teaching statistics in Portuguese, across a very wide range of faculties; but my attempts to report on the oppressive nature of the Marxist Leninism regime was dismissed as Right Wing anarchist nonsense by the politically correct bureaucrats then holding sway on the so-called left in England.
Returned in 1981 to provide 'proper' skool education for children, to controlled statistical environment at the MRC Medical Sociology Unit in Aberdeen. Began to understand the complexities of the inequalities in health debate through the demonstration that a woman's height was more predictive of her outcomes than her partner's social class (Carr-Hill and Pritchard 1989). This meant that politically attractive notions of a class divide in health being reproduced through an easily manipulable social process - such as education - would not do.
But that was only temporary and so moved to what I saw as the eye of the monetarist storms (in health) - the Centre for Health Economics in York on contract from 1984-91. Meanwhile, began to develop interest in planning for basic education and adult literacy in Africa and so, for example, got involved in designing a survey of adult literacy in Tanzania where the charismatic Nyerere had generated a policy of self-reliance and a corresponding education policy aim of terminal education for all. I found that the donors had been conned by the Department of Adult Education in Tanzania, as there was no programme. Yet they were clearly the deserving poor. What would you do? I told the truth and funding was stopped.
During that same period realised that there needed to be an update of a very successful Rad Stats publication of the 1970s, viz Britain's Black Population (Runnymede Trust and Radical Statistics 1980) and set out to edit a revised edition with two non Rad Stats members (Ashok Bhat and Sushel Ohri). This proved to involve innumerable compromises between political correctness and statistical accuracy.
Because I refused to be a tame epidemiologist for the health care cost accountants, I was cast completely loose from secure funding in 1991 to be totally self financing and survived by winning contracts on what was seen as almost undoable tasks: e.g. identifying a relationship between the mix of nursing skills in hospitals and the outcomes of care (and later similarly in primary care). Nevertheless despite winning that kind of contract, the extent of my salary cover was being steadily eroded so it looked as though I would have to prostitute any statistical integrity to dabble in pharmaco-cost-accounting.
However, at the last minute, won tender to review resource allocation formula for Hospital and Community Health Services, which set me off on a trail of simultaneous equation estimation combined with multi level modelling - all in the apparent interest of accuracy but more mystifying than transparent. The outcome, however, was a substantial shift of resources (c£350 million) from the shires to the northern inner cities, even under the Tories. As a result of a subsequent similar exercise with the SSA for Children's Services, proposed a shift of £100m from London to the Northern cities which was fought tooth and nail, but eventually the technical case was conceded. Both of these illustrate the potential impact of statistical analyses - someone loses.
In 1997 celebrated proud records of 10th election not voting for the government and, at the same time, improving my pension fund through ante-post, statistically based, betting. And just now have had final agreement to publish obscure anarchist tract on why mass unemployment is a Very Good Thing.
Lessons
- A monolith - whether left or right - is a bad idea. It stifles debate - leads to Third Way nonsense.
- Whilst telling unpleasant truths often gets you into hot water, it is healthier for the soul.
- What matters is who is measuring, in whose interest and for what purpose.
REFERENCES
Anon (1978), 'Social Indicators: for Individual Well Being or for Social Control', Radical Statistics, BSSRS
Bhat, A., Carr-Hill, R. and Ohris, S. (1988), Britain's Black Population, Second Edition, Avebury: Gower.
Carr-Hill, R. and Hopkins, M. (1973), 'British Bombs', paper presented to JRSS Applied Statistics Conference in Hull
Carr-Hill, R. and Pritchard C. (1992), Women's Social Standing, London: Macmillan
Carr-Hill, R. and Stern, N. (1979), Crime, Police and Criminal Statistics, London and New York Academic Press
Runnymede Trust and Radical Statistics (1980), Britain's Black Population, London, Heineman
Roy Carr-Hill
Centre Health Economics
University of York
Heslington
York YO1 5DD
Tel: (01904) 430000
