Evidence-based policy sounds like a classic case of the absurd opposite. Anyone for a random stab in the dark about what to do? In his interview with The Times today Professor Jonathan Shepherd, of Cardiff University, who is, improbably, both a surgeon and a prominent criminologist, makes three telling points.

First, Professor Shepherd accuses the education and criminal justice systems of falling short by habitually introducing schemes for which the evidence is, at best, scanty. Second, he says that all public services need to embed the culture of research that is found in medicine in which the use of randomised controlled trials is common. Third, Professor Shepherd quite rightly says that policies are never abandoned, even when the evidence showing it to be ineffective is overwhelming. The amour propre of the minister who set the scheme up and the grinding inertia of the bureaucracy are far more powerful than a body of evidence.

Public policy would be all the better if these points were taken up. It is also a good idea to accompany all pilot schemes with a rigorous evaluation, to improve the institutional memory of the relevant departments and to integrate research and practice more thoroughly. That all said, these are testing ambitions. The current Government made a great deal of its desire to root policy in evidence and, if it has failed, it is not wholly for the want of trying. President Obama has made the same claim and he too is finding it difficult to live up to. The evidence about evidence-based policy is mixed.

When a policy is genuinely innovative there is, by definition, no evidence to base it on. Not every policy grounded on a hunch is necessarily wrong. The utilities in Britain had an undistinguished history as private businesses until the 1980s. There was, of course, a lot of evidence that private capital, in general, was a greater source of innovation than public ownership but privatisation was really inspired by the belief that it would work, which indeed it did.

Sometimes the evidence exists only in another country. The Labour Government introduced the Sure Start programme because of the evidence that Head Start had been a success in the United States. The Conservative Party has radical plans for this country’s schools. The evidence comes from Sweden. It is easy for a lot to get lost in translation. Even if the evidence does pertain to this country that is not, in itself, enough. A policy may work on a small scale but not necessarily if it is enacted at scale. A successful scheme may depend crucially on the charismatic individuals involved who are, alas, in short supply.

The lesson is that evidence does not entail policy or implementation. The evidence may be clear but it will be carried out to varying success depending on the design. That said, serious monitoring of progress is an intelligent way of reducing the scope of the state. For example, the academy has pointed out for decades, to no avail, that lots of training schemes do nothing to make young people more employable.

Perhaps the most frustrating failure of centrally commanded public services is that good practice in one place spreads so slowly around the system. If a more thorough integration of research and practice helps the dispersal of good policy it will have proved its worth.