Statistics, argues Yasmin Alibhai-Brown, reveal less than researchers would have us believe, so the government and social services departments would learn more by introducing more human context to their indicators

My husband is a social scientist who has published a number of highly respected reports using empirical data to map out and understand some of the most intractable of our social problems.

He wrote Black and White Britain, the Policy Studies Institute's third survey, which contained some indisputable evidence of how much direct and indirect discrimination still existed in this country in the 1980s. The findings were then used for nearly a decade by agencies, the non-governmental sector and public bodies to argue for anti-racist and equal opportunity policies.

And yet we often quarrel about the use and misuse of statistics and how, in this strangely mechanistic world, everybody feels the need to have some numbers work to back up what they are saying as politicians, journalists, and policy pundits.

Like many researchers, he feels passionately that research that is well done and statistically valid is the best way of knowing a problem and finding solutions. I am uneasy about this and am more of an agnostic.

Personally, I am tired of being bombarded by information which claims to be the truth but which can be disproved by alternative figures. Again and again we see examples of how two completely different beliefs can be validated using statistics - even in the same piece of research.

A couple of years ago, for instance, we had a near-farcical national debate about children and divorce when a complex and important report was published on the subject by the Joseph Rowntree Foundation. Pro-divorce pundits said the report proved that children were less damaged by divorce than by a bad relationship between parents. The pro-family lobby concluded exactly the opposite, thus proving that statistics and research were simply weapons with no neutral life of their own.

That said, it is true that some research findings can completely thwart expectations and a set agenda. Two years ago, I wrote my first ever research-based book called True Colours. It was a report on public attitudes to multiculturalism and the role of government in creating a more integrated society. I commissioned the polling body NOP to examine attitudes, using quantifiable questions and answers.

The results were extraordinary: for example, only 15 per cent of people associated African-Caribbeans with the crime of mugging, in spite of the much publicised comments by Paul Condon, the erstwhile head of the Metropolitan Police, who had produced his own statistics "proving" that 60 per cent of muggings in London were carried out by young black men.

The survey also showed that 70 per cent of the white people interviewed thought we should help asylum seekers, and both Asians and whites thought there was too much immigration from South Asia and Africa.

The value of this material was that it was almost all counter-intuitive. But I still felt that we needed another parallel qualitative study, which I conducted, to get people to explain their thoughts on this issue - something which is impossible in quantitive research, even with the best questions in the world.

The qualitative study was not statistically valid but did provide some fascinating insights into the way black, Asian, Jewish, and white Britons feel about this country and each other. Each study on its own would have been inadequate. Together they worked convincingly.

The lesson here is that numbers alone, although very useful, can hide as much as they reveal. I have seen further education and training and enterprise councils falling headlong into a culture of ceaseless form-filling and information gathering. I have also watched the way people in authority become proficient at providing written evidence which has little bearing on reality.

With the government and other institutions getting into figures and league tables, which I think are no bad thing, great care needs to be taken that the package of information includes in-depth interviews which can add flesh to the figures and that all such information is based on some knowledge of the conditions and environment.

It is important to compare like with like - the inspections on children in care that compared Brent with Ealing were enormously useful. It made it clear that in very similar boroughs, the social services department was performing much better in Brent than in Ealing. But what is the point of comparing, say, Ealing with Norfolk?

And why is nobody talking of carrying out frequent focus group studies of client groups and social services staff in every area? The idea is not all that different from the nationwide people's panel, which currently "converses" with Labour and responds to policies and speeches on a regular basis.

This would enable us to contextualise the bald figures and to appreciate that we are talking about living, individual human beings, instead of mere statistics.

The work carried out by social workers is not simple and should never be simplistically measured. But nobody should believe that what they do is so precious that it should never be called to account. No job in the world is so complex that it cannot be broken down into assessable sections.

But this cannot be the final word. Local authorities must find ways of gathering and providing information beyond that called for within the remit of performance indicators, perhaps by issuing companion publications containing qualitative research.

That way the public will get the transparency and accountability it demands and deserves.