The random side of riots

We talk as if we understand why civil disorder happens, rather than recognising the unpredictable processes at play

Around this time last year, I stood at the threshold of my home in Hackney, with a week-old baby asleep inside and two helicopters overhead tracking the looters outside. As far as I could figure out there was trouble about 100m to the south, and more trouble about 300m to the north. I didn’t venture far, I’m afraid; my paternal instincts were stronger than my journalistic ones.

A neighbour, holidaying in Scotland, called to advise me to get my family out of London. He was concerned that civilisation was about to collapse. I wasn’t, but I admit that during those bizarre few days it didn’t seem absurd to entertain the possibility.

Why did the riots happen? Every pundit had an explanation, from government cuts to soft policing. But there’s a very different way to look at last summer.

Consider the following simple model of a potential riot, based on an idea published in 1978 by the sociologist Mark Granovetter. There are 1,000 people in a crowd of protesters, and all of them have some underlying tendency to embark on a looting spree. We might reckon that an outbreak of rioting might be triggered by insensitive policing, or by the poverty of the crowd, or the opportunities for theft or for violent protest. But for simplicity let’s assume that the only thing everyone in the crowd cares about is what everyone else in the crowd is doing. Some people will start looting without much company. Others will hang back until the riot is well under way.

Let’s put a number on this riotous tendency. One anarchic lunatic has a threshold of zero: he requires nobody else’s encouragement to start throwing bricks at the police. Another person has a threshold of one: he needs someone else to get things kicked off, but then he’s happy to join in. Then there’s a person with a threshold of two and another with a threshold of three, all the way up to the wallflower of the crowd who has a threshold of 999 – he’ll join in only when there’s literally nobody else standing back.

As you no doubt appreciate, this crowd will display a domino-like tendency to riot: the first person encourages the second; that pair draws in a third; then a fourth and a fifth, until everyone is on the tear.

An interesting lesson quickly emerges from this simplistic model. Imagine that, say, the fourth person in an otherwise identical crowd doesn’t have a threshold of three but of four. This second crowd will behave itself: after the first two troublemakers start acting up, there is no third person willing to join them. The outcomes could hardly be more different – and certainly there would have been no national and indeed international media frenzy if the 2011 riots had consisted of two yobs causing trouble on the fringes of a protest in Tottenham, and nobody else joining in.

Yet we know, because we constructed the examples, that these utterly different results emerged from all but indistinguishable initial conditions. One person out of 1,000 had a fractionally different inclination to riot (by one-tenth of 1 per cent of the observed range). As Duncan Watts points out in his book Everything Is Obvious Once You Know the Answer, the two crowds would seem identical to any survey or statistical test you might care to deploy. The same tendency for apparently identical conditions to produce utterly different outcomes also appears in field experiments carried out by Duncan Watts, and in recent models based on more realistic networks of influence.

Nevertheless, we persist in talking as though we understand why riots happen, rather than recognising the random self-reinforcing processes at play. Social influence can work that way. Last year it was arson and assault across English cities. This year it is buying Fifty Shades of Grey.

Also published at ft.com.

4th of August, 2012 • Undercover Economist • Comments off