Is political prediction dead?
Nigel Gould-Davies
The first shots have been fired in the French presidential election campaign. Stakes are high: the result could determine the future of the European Union. But is it worth trying to predict? Forecasters have had a terrible year. Their consensus on the Brexit referendum and US election was badly wrong. In the latter, 60 out of 61 regular tracking polls forecast a Clinton victory. Prediction seems an expensive parlour game that makes pundits and pollsters richer but no-one wiser.
It gets worse. The unpredicted events usher in fresh uncertainty. No-one knows what the consequences of Brexit or the Trump presidency will be. Official forecasts of an immediate adverse impact of a ‘Leave’ vote have proven unfounded — a ‘Michael Fish moment’, as the Bank of England’s chief economist ruefully admits. Meanwhile, pundits’ warnings of the dire consequences of Trumpian trade policy sit uneasily at odds with the ‘Trump bump’ that has driven the US stock market to record highs. Cast the net more widely: 9/11, the occupation of Iraq, Crimea, ISIS… All momentous, all unforeseen. Is political prediction dead?
In one sense, prediction will never die. It is something we cannot not do. Any deliberate action requires assumptions about effects, responses and consequences. To navigate the present we must try to see ahead. Our real concern is that we are flying blind — that good prediction is dead.
But was political prediction ever good? There is no golden age to look back to. Shocking events, accidents of fate and tectonic shifts are strewn through history. The Cold War, often now contrasted with present disorder, seems stable only in retrospect. Those four decades were beset by tensions and crises. Almost no-one foresaw the break-up of the Soviet Union.
So uncertainty has always been with us. But there does seem to be more of it around now: more surprises and bigger consequences. This may be inevitable in an age of extraordinary social and technological change. But it should worry us for deeper reasons, for good prediction is not only valuable in itself: it is the best test of understanding. Accounting for the past, or appearing to do so, is easy. Explaining the future requires real expertise. Most of us listen to experts to hear about the future, not the past. We hope they will tell us what lies ahead. The recent record of poor prediction casts Michael Gove’s famous remark that ‘people in this country have had enough of experts’ in a new light. This is another way of saying that they don’t make better predictions than we can.
Big data, the great disruptive and progressive force of our time, does not yet offer a solution. Paradoxically, big data seems better at helping to win elections than to forecast them. It was a game-changer in Obama’s re-election in 2012, while Trump’s ‘Operation Alamo’ helped secure an improbable victory. Despite the success in ‘nudging’ mass behaviour in other areas, political life has proved more manipulatable than predictable. Data helps, but it is not a panacea.
Should we despair of prediction and, as some advise, focus instead on building a generic resilience against the unknowable future? That would be to go too far. The best recent books on political prediction, which I reviewed recently in International Affairs, offer three lessons:
- Professionalize prediction. Turn it from pundit entertainment into a serious discipline. Demand precision. If something is ‘unlikely’, does that mean a 45%, a 30%, a 1% chance, or what? The implications of each are very different. Get rid of weasel formulations like ‘definitely possible’, pin down forecasts in a form that makes them clear and comparable, then monitor prediction performance. We are not all flying equally blind: some forecasters and their methods prove themselves better than others. So identify best practice and spread it.
- Understand the probabilities and prepare accordingly. A 16% chance of an event sounds trivially low, but it is the probability of rolling a six. So don’t ignore events, especially ones with a big impact, just because they seem unlikely. Brexit will be more disruptive because Whitehall was told not to plan for the contingency of a ‘Leave’ vote. Frankfurt had a Brexit plan but Britain did not. Planning for the improbable — ‘earthquake insurance’ — is unspectacular and can go unnoticed. But you feel its absence when you most need it.
- However partisan your preferences, foster the systems and culture that value and encourage rigour and objectivity in prediction and planning. These activities should be evidence-led, not gut-led. We usually overstate the likelihood of things we want to see happen. Don’t ignore, but scrupulously assess, any data that disconfirm your prior view. You cannot improve on the truth.
As Article 50 is triggered and Britain begins the most intense, complex and significant negotiation it has ever undertaken, Brexiteers should not waft away legitimate but discomfiting concerns about the risks of a poor outcome, while Remainers should not wish the revenge of a disaster on those who defeated them in the referendum. The only certain forecast is that newspapers, airwaves and twitter-feeds will be filled with forecasts that are confident, wrong and quickly forgotten. Political prediction has had a bad run. But there are ways of doing it better if we want to.
Nigel Gould-Davies teaches at the Mahidol University International College in Thailand, and is an Associate Fellow with the Asia and Russia & Eurasia programmes at Chatham House.
His recent article in International Affairs is titled ‘Seeing the future: power, prediction and organization in an age of uncertainty.’ Click here to read it.
For more articles in our March 2017 issue, click here.
For information about how to subscribe to International Affairs, click here.
