Bullshit and X-ray specs

It’s perhaps to early to call it, but I think this article on bullshit, more politely titled “How politicians poisoned statistics”, by Tim Harford for the FT Magazine, may just be the best thing I read in 2016. “Statistical bullshit”, to Tim’s definition, is espoused by those who do not actually care about the truthfulness of the statistic, but seek to use it in much the same way as they might a colourful adjective or choice of neck-tie.


I spent the early part of my analytical career in the naive belief that what was right, was right; that – through careful analysis and presentation – my job was to uncover truth, present it to the decision-maker and sit back whilst they used it faithfully to help decide the right course of action. Be that a decision about what car to buy, which capital projects to invest in, whether to purchase a company, or implement some public policy, or not.

It is a common delusion amongst fresh graduates of STEM subjects moving into business. Of course it doesn’t take long in the real world to have one’s sharp optimistic corners worn down to smooth cynical bevelled edges. But some of that faith in evidence, and its best use, always remains. That’s why you’ll hear phrases like “evidence-based policy” still whispered, wistfully, by the “pointy-heads” in Whitehall.

Nice analysis. So what?

I’ve realised for a long time that most decision-makers do not share that intuitive faith in, and understanding of, data. Like most human beings, they work better with narratives than they do with numbers. You can perform the cleverest piece of analysis in the world, but it is worth precisely nothing if it has no useful context and is not effectively communicated. In other words if it does not fit a narrative. Nice analysis: so what? Unfortunately many analysts are uncomfortably with narratives. It’s not that they don’t get the importance – though they do often underestimate this – it’s that they associate narrative with ‘spin’. Something subjective. The good analyst has to overcome this aversion, because narratives have the power not just to communicate, but to persuade. And we need to persuade.

Of course, decision-makers are not just people to be persuaded, they are also themselves persuaders. Not only have they to decide the right course of action – something which typically occurs unobserved, in private – they have to persuade others that it is the right course of action. Their boss, their colleagues, the board, the electorate. The success of the decision-maker who is great at the persuading but rubbish at the actual decision-making is something that should concern us all. In business these people can bankrupt you. In politics… well, just look around.

On electoral time-scales, it is much easier to tell whether someone is an effective persuader than it is to tell whether they are an effective decision-maker.

The tendency for modern-day politicians to use statistics in selective, bullshit ways is not just a problem because it upsets my sensibilities. Or even those of Tim Harford. It is a problem because, at best, we the public cannot tell whether the actual decisions – the one’s taken in private – are being taken rationally or not. It is a problem because, at worst, we elect our politician purely on their ability to persuade, not on their ability to take good decision in the public interest.


I mean, I enjoy the Boris Johnson buffoonery show as much as anyone. But he clearly and regularly just makes stuff up. Using statistics like kindergarten crayons to cartoonishly illustrate, rather than evidence, his arguments. I actually have no idea whether, in private, he takes and uses evidence in a more sensible fashion. (This is not a party political point, all political colours and creeds are equally guilty. Boris is just more amusing with it.)

Somehow we all need to be able to don analytical x-ray specs. So we can see the truth behind the spin. So we can see how the actual decisions are made.

If anyone knows where I can get a pair, do please write in.

Leave a Reply