The Inescapable Uncertainty of Popular Uprisings

On Tuesday, hundreds of thousands of people turned out in the streets of Ouagadougou to protest a plan to remove terms limits ahead of next year’s presidential election in Burkina Faso. Blaise Compaore has held that country’s top office for 27 years by way of a 1987 coup and four subsequent elections that have not been fair, and his party dominates the legislature for the same reason. Tuesday’s protests are part of a wider and ongoing wave of actions that includes a general strike and stay-aways from schools and universities. A similar wave of protests occurred over several months in 2011. The state’s efforts to repress those challenges killed several people on at least two occasions, and virtually nothing changed in their wake.

Protesters in Ouagadougou on 28 October 2014 (Photo credit: Issouf Sanogo/AFP)

So, will the latest protests in Burkina Faso coalesce into a sustained campaign, or will they soon peter out? If they do coalesce, will that campaign spur significant reform or even revolution, or will it dissipate against repression, redirection, and resistance from incumbent power-holders?

The truth is, no one really knows, and this uncertainty is not specific to Burkina Faso. After decades of thoughtful research, social scientists still can’t reliably predict which bouts of unrest will blow up into revolutions and which won’t.

We can say some useful things about which structural conditions are more conducive, and thus which cases are more susceptible, to sustained popular challenges. A study I co-piloted with Erica Chenoweth (details forthcoming) found several features that can help assess where nonviolent campaigns are more likely to emerge, but the forecasting power of models based on those features is not stellar. Efforts to develop predictive models of civil-war onset have achieved similar results.

Once unrest starts to burble, though, we still don’t understand and can’t model the ensuing process well enough to reliably predict which way it will tip. Across many cases, a simple base-rate forecast will produce very accurate results. Keep betting on the persistence of the status quo, and you’ll almost always be right. If you’re trying to predict what will happen in a specific case at a specific juncture, however, it’s still hard to improve much on that crude baseline.

This persistent uncertainty can be maddening. Lots of smart people have spent a lot of time studying and thinking about these processes, and it feels like all that effort should have yielded bigger gains in predictive power by now.

That failure is also enlightening. If we believe that our efforts to date have been thoughtful and thorough, then the lack of progress on predicting the dynamics of these situations is telling something important about the nature of the underlying process. Uncertainty isn’t just a consequence of these political behaviors; it’s a prerequisite for them. As Phil Arena said on Twitter:

.@dtchimp There's good reason to think we *can't* get much past that, b/c if everyone knew who was willing to resort to violence & under +

— Phil Arena (@filarena) October 28, 2014

.@dtchimp what conditions, the violence would be a whole lot less likely to occur. Uncertainty abt the phenomenon is a key cause thereof.

— Phil Arena (@filarena) October 28, 2014

And it’s not just uncertainty about the potential for harsh state repression, which is what I took Phil to mean by “violence.” Uncertainty about who else will turn out under what conditions, what forms that violence will take and exactly whom it will directly affect, how challengers will organize and adapt in response to those events, what changes in policy or institutions those actions will produce, and who will benefit or suffer how much from those changes are all relevant, too.

In short, the rare political “events” we wish to predict are shorthand for myriad interactions over time among large numbers of heterogeneous individuals who plan and learn and screw up in a changing environment in which information is inevitably incomplete and imperfect. The results are not random, but they are complex, in both the conventional and scientific sense of that term. If we could reliably foresee how things were going to go, then we would adapt our behavior accordingly, and the whole thing would unravel before it even started.

Under these conditions, central tendencies can and do still emerge. A small but growing body of work in political science shows that we can use structural patterns and observations of leading-edge activities to smudge base-rate forecasts a bit in either direction and achieve marginal gains in accuracy. Systems that properly elicit and combine forecasts from thoughtful crowds also turn out to have real predictive power, especially on short time horizons.

Still, the future trajectories of individual cases of incipient revolution will remain hard to foresee with accuracy much beyond the banal prediction that tomorrow will most likely resemble today. That persistent fuzziness is not always what politicians, activists, investors, and other interested or just curious observers want to hear, but on this class of events, it’s probably as clairvoyant as we’re going to get.