The evidence is clear. Statistics from the United States, Europe, Asia and Africa all point in the same direction: male drivers are more likely to crash their cars than females. Aggressive behaviour, rule-breaking tendencies and a greater willingness to take risks are all thought to contribute. Taken together, male drivers are a riskier bet, and face higher premiums for car insurance as a result.

Last week, the Court of Justice of the European Communities took a wrecking ball to this seemingly evidence-based policy. From December next year, insurance companies will no longer be able to discriminate on the basis of sex: men should see their premiums fall and women will pay more. The decision was greeted with howls of outrage in some quarters. From newspaper headlines declaring it to be “madness” to radio phone-ins that bemoaned a “lack of common sense”, the underlying message from critics was that discrimination against men by insurance companies was fair because it was based on evidence.

The court's decision affects more than car insurance — life assurance premiums paid by men could rise to match those paid by women, where previously they were discounted because men on average die earlier.

Nature is a vocal and staunch supporter of evidence-based policy-making. Yet it is important to distinguish between policies that ignore the evidence and those that consider it but do not give it the deciding vote. Although the European ruling flies in the face of available research on accident and death rates, policy-makers have to consider other relevant factors as well as the scientific data. Europe introduced a laudable law in 2004 to ensure equality between men and women in access to and provision of goods and services. (Last week's ruling was prompted by a challenge under human-rights legislation to a get-out clause in the 2004 law for the insurance industry.) In this case, the drive in Europe to reverse practices that survive as a legacy of centuries of inequality and discrimination against women is right to trump the cold logic of the statistics. Besides, it is far from clear how insurance firms translate the sex differences to quantitative hikes in premiums for young men. The best evidence-based decision-making is also transparent.

Science, and so the data and evidence it gathers, serves society best when it is viewed as part of a wider assessment of risk — not an all-powerful framework within which policy must be placed. A report last week from the UK select committee on science and technology complained that the British government had dropped disruption to air travel by natural disasters from its national risk assessments in 2009, yet saw the eruption of Eyjafjallajökull in Iceland in April 2010 close airports, cancel flights and strand thousands of passengers around the world. The Geological Society in London told the committee that Earth scientists had warned the government of the potential for major disruption for a number of years, but felt that the policy-makers had ignored them. Or perhaps, the policy-makers had heard the warnings, weighed up the costs and benefits, but decided the risk was small enough to do nothing except cross their fingers. It is the job of scientists to raise such concerns, but it is the job of politicians to decide when to ignore them. Although the decisions may be judged with hindsight as being poor policy, they remain evidence-based. Politicians too, however, should not shy away from revealing their reasons.

Overstating the role that scientific evidence should have is most inappropriate when the stakes are highest. For global warming, for instance, the evidence is as clear as the statistical difference between male and female drivers. Yet the policy response must also take into account social, economic and political factors. It is legitimate, if short-sighted, to acknowledge that man-made global warming is real but argue that policies to cut emissions are too expensive to pursue. Dismissing those who take such a position as climate sceptics alongside those who deny the evidence is wrong. When setting policies, there are limits to the role that evidence can have.