We should all know never to take a statistic quoted by a politician on trust. But there is a huge difference between the ways in which politicians can (mis)use statistics.
Take, for example, when Labour before the 2015 election kept saying people were £1,600 worse off than they were 5 years earlier. As Tim Harford notes, there are a lot of issues in making any general claim based on earnings data. But as Geoff Tily points out, the £1,600 is hardly a wild exaggeration or gross distortion. It pointed to a key fact, which was an unprecedented decline in real earnings which no one seriously disputes. It would have been incredible if Labour had not kept talking about this.
Tim writes that it represents “a political use of statistics conducted with little interest in understanding or describing reality.” Of course it does not describe the complexity of reality, the differences between the median wage and the experience of the median worker, etc etc. Those complexities need to be set out and Tim does so brilliantly. But politicians in speeches will never do that, and it would be unrealistic to expect them to do so. There is also no evidence presented which justifies the claim that this statistic was used with little interest in understanding or describing reality.
Take another example from a recent post of mine. George Osborne had derived the cost of Brexit by taking the GDP loss and dividing it by the number of UK households. Fraser Nelson, and subsequently Anthony Reuben at the BBC, objected that this was dishonesty (Nelson) or confusion (Reuben) because only about two thirds of GDP was household income. Typical you might think for this Chancellor to misuse statistics to exaggerate. Yet as I explained in the post, what the Chancellor had done was standard practice by economists, because less government spending or investment are also in an important sense costs to households. In that case too, a politician was using a summary statistic in a reasonable way.
Magnitudes are often crucial. It is true, for example, that fiscal policy before the financial crisis was a little on the lax side. But the magnitudes involved could have been corrected by any new Chancellor in one budget with hardly anyone noticing. They are a world away from the magnitudes required to claim Labour were profligate before the crisis, and that austerity was required to clear up the mess that Labour had created. Given the importance (to the result) of that claim before the 2015 general election, it is odd indeed to focus instead on Labour’s claims about real earnings losses.
The other examples Tim discusses in his article - Trump’s crime statistics and Jeremy Hunt’s figures for excess weekend deaths - are indeed totally or highly dubious, for reasons Tim makes very clear. Or an example that is close to my heart: the Prime Minister claiming that they had not cut spending on flood defences, which could be made to be true but hardly describes reality. These are all examples where the politician wants to mislead people. It is this misuse of statistics that we should focus on.