Crystal Clear

In "Dark Crystal" (July/August 2011), Foreign Policy managing editor Blake Hounshell discusses social scientists’ apparent failure to predict the momentous revolutions of 2011 and concludes that "looking at rows of numbers is a lousy, or at least insufficient, way to predict" the occurrence of nonviolent popular uprisings. Hounshell is probably right about the "insufficient" part, but I think he’s wrong about the "lousy" part. In fact, my own recent analysis gives us grounds for cautious optimism about the usefulness of quantitative methods for forecasting these kinds of events.

From 2001 until the end of 2010, I served as research director for the Political Instability Task Force, a U.S. government-funded research program that develops statistical models to help forecast and explain various forms of political crisis and change in countries worldwide. In early 2011, in preparation for a workshop organized by the Council on Foreign Relations, I used a statistical technique called Bayesian model averaging (BMA) to develop an algorithm that can use data on a number of social, economic, and political conditions to estimate the probability that a given country will experience an onset of nonviolent rebellion in the coming year. Leaning on prior research by Erica Chenoweth and Maria Stephan, a nonviolent rebellion was defined for this project as a "campaign of purposive, nonviolent mass events in pursuit of a political objective." Using this definition, occasional demonstrations or strikes do not count as nonviolent rebellion; for protest activity to qualify, it must be both purposeful and sustained. For example, the large-scale and repeated anti-government demonstrations in Syria in early 2011 would certainly qualify as nonviolent rebellion, while the sporadic protests so far in Algeria would not.

These kinds of uprisings are historically rare events with complex causes, and as such, they are extremely difficult to predict with precision. Nevertheless, it turns out that a statistical model — or, in this instance, a weighted average of several models — could have done a good job of identifying the Middle Eastern and North African countries that have experienced popular uprisings this spring as among the most susceptible to these events in 2011. When all countries worldwide are listed in descending order according to their estimated likelihood of an onset of nonviolent rebellion this year, Egypt ranks 4th; Syria stands 6th; Libya is 25th; Tunisia is 28th; and Yemen is 33rd. With more than 160 countries on the full list, these rankings land almost all of these countries in the top 20 percent of the global rankings, an admittedly arbitrary but nonetheless useful threshold for identifying most-likely cases from forecasts like these. Crucially, the historical data used in developing this algorithm only covered the period 1972-2009, so the statistical analysis did not get to learn from the supposedly surprising events we’re trying to forecast in 2011. In other words, this is not a case of the kind of "hindsight bias" sociologist Charles Kurzman appropriately bemoans in Hounshell’s article.

Given how reliable these forecasts have proved to be so far, I think it’s worth noting that the country pegged by this algorithm as the most likely candidate for nonviolent rebellion in 2011 is China. China reportedly experiences tens of thousands of scattered protests, riots, and strikes each year, but many observers of that country’s politics dismiss those events as background noise in an otherwise well-managed political system. For example, in a March 12, 2011, op-ed in the Wall Street Journal, Francis Fukuyama points to the "higher quality" of authoritarianism in China and the cohesiveness and nationalism of its military to conclude that "China will not catch the Middle Eastern contagion anytime soon." Without putting too much weight on one number, I think it’s fair to say my analysis shows that the country may be riper for nonviolent rebellion than many China watchers believe. Although neither the recent riots in Guangdong province nor the protests in Inner Mongolia would qualify as nonviolent rebellions under this project’s definition, they would seem to support the view that the potential for social unrest in that country is substantial.