Being wrong about Corbyn

Expanding here briefly on some things I said on twitter, in light of Labour’s very impressive showing in the 2017 UK general election.

There are a number of different ways in which a political analyst – an academic, pollster or pundit – can be wrong.

• • You can make a wrong prediction. This is incredibly easy to do – we all make wrong predictions all the time. Social reality is enormously complex, and it’s basically impossible to make strongly reliable predictions about it.
• • You can be wrong about the probability distribution of possible outcomes. It’s obviously difficult to check whether somebody is wrong in this sense – unlikely outcomes often happen, and likely outcomes often don’t happen. Still, it’s another way of being wrong.
• • You can be wrong about the range of possible outcomes. That is, you can incorrectly suggest that some events are outside – or inside – the space of the feasible. (This is a special case of the previous probability distribution point.) This can, sometimes, be checked – if you say an outcome isn’t possible, and it happens, you were clearly wrong.
• • You can have a poor ‘model’ of social reality, generating your sense of the space of probabilities. This can be a model in a formal sense, as in some polling models. Or it can be a model in an informal sense, meaning one’s view of the important forces and dynamics of the relevant social reality.

In relation to Corbyn’s Labour’s impressive electoral performance, most (though by no means all) of us were wrong in one sense or another. I didn’t venture a prediction, because I thought the uncertainty was too high for a prediction to be made with any useful confidence. But if I had been obliged to make a prediction – professionally, say – I would certainly have predicted a much poorer electoral showing than Labour in fact achieved.

It is, of course, impossible to know whether one’s probability distribution is accurate (and, arguably, what that even means, epistemologically), so I’ll put that aside. In relation to the special case of possible outcomes, however, my range of possible outcomes certainly did include the electoral gains that actually occurred – so I was not wrong in that respect.

Finally, in relation to one’s ‘model’ of social reality: I wrote up my view of the electoral feasibility of Corbyn’s project shortly after he won the leadership, in September 2015 – you can read it here. Reasonable people can of course differ on these issues – a ‘model’ can never be definitively proven or refuted – but to my mind, the analysis in that post has stood up well, in light of subsequent events.

Now, the professional UK pundit class has also been wrong about Corbyn. But I would argue that most of them have been wrong in a different, stronger sense. Not only did many pundits wrongly predict electoral disaster for Corbyn’s Labour, they also often suggested that a strong electoral showing from Labour was somewhere in the probability range between extremely unlikely and actively impossible.

Most prominently, Matthew Goodwin, the political scientist, has now literally eaten his most recent book (‘Brexit: why Britain voted to leave the European Union’) on live television, after tweeting that he would do so if Corbyn’s Labour polled 38% or higher (in fact Labour polled 40%). This demonstrates good grace – but the existence of the tweet in the first place implies not just that Goodwin called the election wrong, but that he also called the space of the feasible wrong. And Goodwin is far from alone in this. The professional pundit class, as a whole, regarded the prospect of Corbyn’s Labour polling at ~40% not just as unlikely, but, for the most part, as absurd.

This in turn speaks to the ‘model’ of social and political reality that informs pundits’ analysis. I think there are a range of different pundit models out there, and surveying them would take a much longer post than this one. But it seems clear enough to me that the overwhelming majority of UK political pundits have badly flawed models of the political and social reality they are paid to analyse and interpret. This – rather than pundits’ bad predictions – is the big analytic problem with recent UK political commentary.

Finally, there are problems with the UK pundit sphere beyond the simply analytic. Most obviously from a ‘pro-Corbyn’ perspective, many pundits were not just badly wrong, but (to be blunt) were arseholes about it. Without wanting to get into an unproductive slanging match on this issue – and recognising that there are pundits to whom this critique does not apply – one of the negative consequences of many pundits’ belligerence towards ‘pro-Corbyn’ voices was epistemic. Pundits’ willingness to treat pro-Corbyn advocacy and analysis with contempt restricted the range of positions and perspectives that pundits treated as worthy of attention – and this in turn prevented pundits from appropriately updating their opinions in light of relevant arguments and evidence. This is one of the major reasons, I think, for the dramatic failure of the pundit class to see Corbyn’s Labour’s electoral success coming.

Now, there is an unfair imbalance in my criticism of the UK pundit sphere. I am not professionally obliged to produce analysis every week (or day!) – if I were, then over the last few years I would have been wrong about countless things. Nevertheless, as a consumer of UK punditry, I can still evaluate and criticise it. Moreover, in evaluating punditry, I’ve argued, whether pundits are wrong matters less than how they are wrong. For the most part, the UK commentariat were not just wrong about Corbyn – they were wrong in the wrong way. That is a bad problem for the UK public sphere.