Significant digits

We live in an age when good policies are assumed to be evidence-based. And that evidential base is assumed to be at its best when expressed in numbers. The digital information may be derived from quantitative data organised in statistics, or from qualitative data organised in indicators. Either way, evidence in digital form provides the accepted foundation of policy arguments over a very broad range of issues. In the policy realm there are frequent debates over particular policy issues and their associated evidence. But only rarely is the nature of the evidence called into question. Such a faith in numbers can be dangerous. Policies in economic and financial policy, based on numbers whose significance was less than assumed, recently turned out to be quite disastrously wrong. Other examples can easily be cited. The decades-long period of blaming dietary fats for heart disease, rather than sugar, is a notable recent case. We are concerned here with the systemic problem: whether we are regularly placing too much of an evidentiary burden on quantitative sciences whose strength and maturity are inherently inadequate. The harm that has been done to those sciences, as well as to the policy process, should be recognised. Only in that way can future errors be avoided.