The Diffusion of Social Unrest

Picture of the Swing Riots

Guest post by Gabriel Leon-Ablan, King’s College London.

Social unrest often begins suddenly and spreads quickly.

How does social unrest diffuse? Toke Aidt and I examine this question in two forthcoming articles in the Journal of Politics and in the British Journal of Political Science. Our starting point is that riots diffuse when information about them spreads to new locations and reaches new individuals. This raises a number of important research questions: what are the networks that link these individuals together and allow the information to travel? Who responds to this information? And what role does context, in the form of local conditions and structural factors, play in the process of diffusion?

Following in the research tradition of historical political economy, we address these questions by collecting a detailed dataset that we analyze using a carefully laid out identification strategy. Our work in this project also draws from network analysis and spatial econometrics.

The Swing Riots

Our articles focus on the case of the Swing riots, a wave of unrest that swept through rural England in the winter of 1830-31. We have data on every English parish (there were over 10,000 of them) and the location, date and type of incident for just under 3,000 Swing-related incidents. Although historically known as riots, these incidents were quite varied in nature and are best thought of as belonging to a “repertoire of contention”.

But why look at the Swing riots? Our reasons coincide with two of the key arguments often given to motivate work in historical political economy, as discussed in Volha Charnysh’s blog post: the Swing riots represent a good setting in which to consider more general theoretical propositions, and they are an interesting historical case in their own right.

The main challenge in identifying diffusion empirically is what Buhaug and Gleditsch refer to as the reverse Galton problem: what appears to be diffusion may instead be the result of unit attributes that are spatially clustered. The Swing riots offer a way around this econometric problem: they happened in slow motion, with the riots spreading over the course of a few weeks so that the diffusion is observable, yet they spread fast enough for everything else to remain largely unchanged during the period of diffusion. Econometrically, this means that the only variable that changes in these few weeks is the number of riots in the neighborhood of a location (i.e. the spatial lag), allowing us to use a fixed effects specification to estimate diffusion. A second advantage is that unlike recent riots, where participants have often traveled to a small number of riot locations, in the 18th century participants rioted where they lived. This means that we have hundreds of different riot locations and know the values of structural factors relevant to each of them.


Our second reason for studying the Swing riots is that they were an important historical event: recent work has shown that these riots put pressure on parliament to vote in favor of the Great Reform Act of 1832, the first of the 19th century franchise extensions in Britain. And because of this, the Swing riots were the topic of Hobsbawm and Rudé’s classic social history book on civil unrest.

Networks of diffusion

In the article “The Social Dynamics of Collective Action” we study how the Swing riots spread across England. Unlike recent work that has looked at the protest decision of educated individuals, the Swing riots were a movement of poor and uneducated farm workers. How did they learn about the riots? The historical context, coupled with the poverty and relative isolation of the participants, means that information could only travel in a small number of ways.

We find that local face-to-face interactions drove much of the diffusion. Some of these interactions appear to have occurred during the many fairs that took place in the English countryside. We also look at the role of the transport network and of local newspapers, but neither affected diffusion. This is perhaps not surprising given that this was a movement of mostly illiterate farm workers who rarely traveled far from home.

Organized or spontaneous?

While some studies have shown that riots can be organized by political leaders (Wilkinson), others assume that riots are spontaneous and spread without the need for organizers (Epstein). Were the Swing riots organized, or did they happen spontaneously? It is difficult to observe local organization, but for 1830s England we have a good proxy: at the time, local organizations were active in sending petitions to parliament asking for slavery to be abolished and for Catholics to be given full rights. Using the number of petitions sent from a location as a proxy for that area’s political organization, we find that local organizers played an important part in the diffusion of the Swing riots.

The structural factors

An alternative explanation for the occurrence of social unrest is that it is the direct consequence of local structural conditions. For example, Caprettini and Voth have shown that locations that adopted the newly developed threshing machines experienced more Swing riots. But in general, each specific factor can only explain a small fraction of the overall variation, and so the list of potentially relevant structural factors can be quite long. Furthermore, it is likely that structural factors and diffusion are both important, and they may interact in interesting ways.

In our article “The Interaction of Structural Factors and Diffusion in Social Unrest” we examine how structural factors interacted with the diffusion process. There were two types of interaction. The first was in how local factors facilitated or hindered diffusion: we find that economic conditions and the presence of potential leaders facilitated diffusion. The second was in how diffusion magnified the effect of local structural factors: if a (hypothetical) change in a factor increased local rioting, and this rioting then diffused, the increase in that local factor had in fact led to an increase in rioting across a wider area. Interestingly, the size of this total effect varied with the location of the parish, with the local factors of parishes with a high centrality measure having a much larger effect. This information can be summarized in a frequency distribution of multiplier sizes.

The Swing riots were of great significance in British history, but they also represent an excellent setting in which to test general theories and hypotheses about the diffusion of social unrest.


  • Gabriel Leon-Ablan

    Gabriel Leon-Ablan (@gabrieljleon) is an associate professor in the Department of Political Economy at King’s College London. He studies collective action, social order and political development, with a recent focus on how non-democratic regimes stay in power and what can be done to challenge them. His articles have been published in a number of political science journals, including the Journal of Politics, the British Journal of Political Science and the Journal of Conflict Resolution.

One thought on “The Diffusion of Social Unrest

  1. Nice post. I look forward to reading your papers in greater detail and incorporating the material in my lectures on riot models.

    Just curious, while you cite Epstein’s paper in the post, the other well-known models are missing in this post. (You do cite one of them in your papers)

    Granovetter, M. (1978). Threshold Models of Collective Behavior. American Journal of Sociology, 83(6), 1420–1443.

    Bikhchandani, S., Hirshleifer, D., & Welch, I. (1998). Learning from the Behavior of Others: Conformity, Fads, and Informational Cascades. The Journal of Economic Perspectives, 12(3), 151–170.

    Is this because you find these above models less central to the questions you are trying to answer, than for example, Epstein and Axtell (2002)? Or is it because you prefer the non-epidemiological mechanism over epidemiological mechanism implicit in the Granovetter type models? Or just it was edited out of the final draft?

Leave a Reply