Revolution timing

Why do revolutions so often break out after conditions have started to improve?

The obvious model of revolution says that misery causes revolt: press people hard enough and eventually they break. It has the appeal of arithmetic, and it makes a prediction we can check. The check goes badly. The most wretched years are usually the quietest ones, and the rising comes later — often after a decade in which taxes eased, harvests improved, or a government began, cautiously, to reform itself.

Tocqueville noticed this about the France he was writing the history of: the country was more prosperous in 1780 than in 1750, the crown had begun relieving old burdens, and the provinces where conditions had improved most were the ones where discontent ran hottest. His conclusion was that the most dangerous moment for a bad government is when it starts to mend its ways. Why should relief be dangerous?

Try separating two things that the misery model runs together: how badly off people are, and how badly off they think they should be. Only the first is measured in bread. The second is a moving expectation, and expectations are formed by recent experience — three good years teach you to expect a fourth.

Now suppose conditions improve steadily for a while and then reverse: a bad harvest, a war, a fiscal crisis. What happens to the two lines? Actual conditions fall back, perhaps only to where they stood a decade before. But expectations do not fall back with them; they carry the momentum of the good years. The gap that opens between what people have and what they had come to count on is wider than anything the whole preceding period of steady poverty produced. That is James Davies's J-curve — a sustained rise followed by a sharp reversal — and the underlying idea is relative deprivation: people revolt against a shortfall from an expected standard, not against an absolute level.

That is a satisfying story, and I want to be careful about how much weight it bears. It is very hard to falsify after the fact. Look back far enough from any revolution and you will find both improvement and a downturn somewhere; the curve can be fitted to almost anything. And it explains the fuel without explaining the spark. Grievance of this kind is not rare. Revolution is.

So ask the harder question: what does it actually take for a grievance to become a rising? Not for one person to be furious — for thousands to act at the same moment. And here is the difficulty each of them faces. Under a repressive government, opposing it openly is dangerous unless enough others do it too. So whether you should protest depends on how many others will, which depends on what each of them believes about you.

But nobody knows what anyone believes, because in a repressive setting people misrepresent their views — Timur Kuran's term is preference falsification. Everyone displays loyalty they do not feel, so everyone systematically overestimates support for the regime. Suppose each person has a private threshold: the number of others already in the street that would make them join. The distribution of those thresholds determines whether a cascade is possible, and that distribution is invisible to everyone — to the regime, to the opposition, and to the participants themselves.

Now the reform connection reappears in a second form. Easing repression lowers everybody's threshold at once. A society that looked stable can, without any change in how people privately feel, cross into a state where a small first move — one crowd that does not disperse — pulls in the people whose threshold was two hundred, whose arrival pulls in those at two thousand, and onward. This is why 1989 in Eastern Europe surprised the dissidents as much as the governments. Not because the observers were incompetent, but because the information required was hidden by the very repression under study.

The two accounts are not rivals, and it is worth saying which question each answers. The J-curve is about the level of grievance: why a population is combustible in one decade and not another. The cascade account is about timing and surprise: why combustible material ignites on a Tuesday rather than the previous Tuesday. Neither alone is enough, and the second is the one that explains the phenomenon people find strange.

It also dissolves the temptation to read the surprise as a failure of analysis. If preferences are concealed, then the crucial variable is not merely unmeasured but unmeasurable in advance, and the resulting unpredictability is a structural feature rather than a gap in the data.

One further caution. That revolutions are unpredictable in their timing does not mean they are causeless, and none of this explains what a revolution then does — who ends up governing after a regime falls is a separate question with a discouragingly poor relationship to who filled the streets.

• Why revolutions so often end in a regime more repressive than the one they replaced.
• How threshold models of collective behaviour apply to bank runs, strikes and fashions.

Every term the collection defines is gathered in the glossary.