Statistics and the referendum

One of my favourite books as a social sciences student was Darrell Huff's slim volume, 'How to Lie with Statistics'. It gave me a healthy scepticism of all statistics, particularly when used in support of a particular campaign.

My favourite example was Harvard University claiming huge numbers of millionaire alumni based on an annual survey. As Huff points out, it may well be that those who hadn't made a fortune might be somewhat less enthusiastic about completing the survey!

The UK Statistics Authority has published a compendium of statistics for use in the independence referendum campaign. Putting my scepticism to one side, the intelligent use of statistics is an essential element of modern debate. While the UK Statistics Authority is neutral on the substantive question of independence, they do care about the safe and effective use of official statistics in public debate. As they put it, "If those statistics are misrepresented or misinterpreted, not only does that lessen the integrity of the debate, it undermines confidence in the statistical system; and that in turn undermines confidence in the vast range of economic and social policies shaped by the statistics."

The paper does give some useful health warnings over the use of comparative statistics. While statistical comparisons between administrations can be valuable, they can also be problematic and uncertain. This is particularly true since devolution as public policy in Scotland increasingly diverts from that in England. Official statistics are only best estimates and are not always available on a consistent or comparable basis.

Despite these qualifications, this is a very useful compendium. Not just for the referendum debate, but also for wider public policy purposes. However, it would be very optimistic to assume that it will necessarily persuade either side not to follow some the statistical tricks highlighted in Huff's book. Peter Jones in the Scotsman and Magnus Gardham in the Herald are perhaps both being optimistic in their analysis today. Do unlike political hacks!

What is more certain is that Andrew Dilnott and the Statistics Authority may well feel free to intervene in future debate, having given this warning shot to both sides.