I was recently asked by a journalist in North Dakota to comment upon the ongoing protests at Standing Rock, and to predict how it all would end. Like a good social scientist, I hedged a lot, but like a naive one, I obliged his request. Perhaps luckily for me, my words were never printed. Two days after our interview, the Army Corps of Engineers declared that the Energy Transfer Partners Corporation would not be granted an easement, the news editor demanded a different version of the story, and my predictions (right or wrong, I may never tell!) remain in a newsman’s notebook.
But I have been turning over the journalist’s question wondering how soon will come the day when we can predict within a narrow credibility interval, the actions of police or protesters engaged in some contentious struggle.
The capacity to predict, of course, is the sine qua non of mature, policy-applicable scientific theory. But I don’t think many of us would say that we are quite there yet. We have many excellent small-N studies cataloguing mechanisms of contentious politics and repertoires of protest policing (too many to cite in a blogpost). And we have large-N studies giving us some understanding of general patterns or trends in protest activity distributed across various political regimes. But we haven’t brought granular, nuanced, rich, big, and comparable data to questions asking who does what to whom, when, and under what (often dynamic) circumstances. I predict that will change very soon.
The Dynamics of Collective Action (DCA) database represents the closest we have lately come to big, rich, comparable data on protest and the policing thereof. Analyzing 22 variables describing over 24,000 events spanning three decades in New York state, the DCA database has launched a number of articles and careers. By comparing across so many (stand-alone) events motivated by different claims and using a range of tactics, authors have advanced our knowledge of how protests unfold differently against different targets and how movements’ activities depend on their contextualization in SMO fields, markets, and broader national policy processes. (For a listing of all publications based upon the DCA database, click here.)
A few articles using DCA data have also attempted to explain police behavior during who have collected and reported findings from DCA explain, the dataset is somewhat limited in its utility for understanding protest policing. DCA only collects very impressionistic data about police activity: whether police were present or not and whether they engaged in violence or arrest. Thus Soule and Davenport (2009) counsel future researchers to “move away from [police] presence/absence formulations of repression and toward more theoretically and methodologically sensitive conceptualizations of police action.”
The authors suggest, too, that to better understand protest policing behavior – well enough, perhaps, to make public predictions about what police will do during some ongoing movement – we will also need to build models of protest policing that take into account police (and protester) activities at events occurring throughout a protest ‘campaign.’ DCA, however, provides no accounting of campaigns (defined as a series of thematically and operationally linked protest events), instead conceptualizing each event in its dataset as a one-off.
Finally, Soule and Davenport (2009) counsel researchers to “examine the effects of various exogenous factors, such as the overall structure of political opportunities on police use of force and/or violence and arrests.” Here, again, the DCA comes up short even as it provides arguably the best quantitative data available on protest and its policing.
There are a number of reasons, though, to predict a brighter future for protest policing studies. First, the small-N studies of protest policing – while they are fundamentally incapable of marshaling enough data for comparative analyses – have been discovering, elucidating, and confirming the importance of a number of “control performances” (my riff off of Tilly’s contentious performances) in the protest policing repertoire. We have a better idea than ever about what we should be looking for as we take the advice to move beyond “police absence/presence formulations” of protest policing. Second, since the DCA was compiled, the Bureau of Justice Statistics (of the Department of Justice) has released powerful survey results describing the capacities and cultures of police departments across the United States. These data, especially when combined with data on US city political opportunities (housed in census databases and on city government websites describing upcoming elections, government types, and the political bent of their populations) will allow us to understand the “various exogenous factors” affecting protest policing.
Third, the Occupy movement provides us with excellent comparative campaign data. Nearly 200 US cities and towns had an Occupy campaign. Moreover, each of these campaigns was motivated by similar claims, drew on a similar protest performance repertoire, and occurred at the same time. The comparative leverage afforded by such data could hardly be better if social movements scholars had planned and organized the Occupy movement for their own selfish research purposes! Fourth, and finally, our capacity to extract nuanced data about events from news reports – a necessary and painstaking aspect of almost all quantitative protest event analysis research – is improving rapidly. With citizen science (crowdsourcing) approaches, the clever use of natural language processing algorithms, and hybrids between the two, we will soon find that we can parse thousands of news accounts by hundreds of variables of interest. And instead of requiring a decade of effort, a data-gathering and processing project comparable in size/scope to DCA might only take a year or two.
Improved by these four factors, next generation social movements databases will support complex analyses explaining how interactions between police and protesters at multiple levels – within and across events – not only result from political opportunities and police capacities and culture, but also feed back into later interactions. We know, based on our qualitative experience, that an on-the-ground clash fueled by adrenaline can shift the mood and outcome of an entire protest event, an entire campaign, and even (sometimes) the course of history. And we know that police strategies, often based on contextual political opportunities, can increase or decrease the likelihood of on-the-ground clashes and other behaviors. Soon, finally, we will be able to house fine-grained data on all these behaviors and factors in a single place, linking all these levels of analysis through dynamic probabilistic models allowing us to measure the flow of causality through such complex systems.
With such complete, well-operationalized data, we will be able to identify sequences of interaction leading to violent escalations, negotiations, and other outcomes, and the contextual factors influencing them. And then, I predict, we will be able to respond confidently and competently when reporters call asking us to divine what will happen next.