Does repression work?: Measuring repression’s effect on protest using an instrumental variable model

ABSTRACT

Protest and repression are reciprocally related. Governments respond with repression when faced with challenges to their rule. Dissidents choose their strategies, turning out to protest or staying home, based on the state’s behaviour. But what effect does repression have on protestors’ decisions? The existing literature is of two minds on this issue. One school of thought argues that repression suppresses protest. A second school of thought argues that repression increases protest by inducing public backlash against the regime. Efforts to adjudicate these claims are complicated by the endogeneity between protest and repression. We use US economic development assistance as an instrument for government repression. Governments seeking US development assistance eschew the repression of protestors. An instrumental variable analysis of the MAROB Middle East dataset shows that repression (instrumented on US development aid commitments) discourages protest by a dissident group. The likelihood of protest decreases by roughly 20% in a given year if the group is repressed. The need for an instrumental variable model is highlighted by the fact that uninstrumented regressions show the opposite effect, giving the spurious appearance of backlash. Unfortunately for protestors, the appearance is deceiving. Repression works.

KEYWORDS:

• Protest
• Social movements
• Repression
• Ethnopolitical organizations
• Middle East and North Africa

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. The closest approximation we could find was a 2002 episode in which the Egyptian government, religious organizations, and civil society groups verbally condemned a US government announcement that aid programmes on top of existing ODA would be made contingent on improved human rights practices. (See incidents 14494–14502.) This controversy was unrelated to the existing US ODA commitment (our instrument) which did not change. Moreover, there was no evidence of street demonstrations taking place over the issue. The only other NAVCO incident related to US aid is a 2010 statement by the Pakistani Taliban urging the Pakistani government to turn down disaster assistance from the US after an earthquake. (See incident 58437.) Again, the statement did not concern ODA, which is distinct from disaster relief, and there was no evidence of street protest taking place.

2. The dataset does not code umbrella coalitions of organizations as organizations in their own right.

3. The MAROB variables domorgviolence (indicating violence against domestic targets) and transvioltarg (indicating violence against transnational targets) are equal to zero in all group years.

4. The MAROB variable ORGST7 has positive values in one or more group-years, indicating “terrorist activities” – violence strategically targeting non-police, non-military personnel.

5. For further explanation of the original MAROB variables see the dataset’s codebook: http://www.mar.umd.edu/data/marob/me_marob_sept08_codebook.pdf.

6. The data series, “Aid (ODA) commitments to countries and regions [DAC3a],” may be found here: https://stats.oecd.org/.

7. Negative Polity2 values characterize predominantly autocratic regimes and positive values characterize more democratic regimes.

8. The ideology variables are not mutually exclusive. Hezbollah, for instance, is both nationalist and religious in orientation. Although there is no omitted base category, mathematically speaking, four groups in the analysis are coded zero for all five of our ideological variables (Nationalist, Religious, Leftist, Rightist, Pro-democratic). Substantively, these groups tend to be secular, centrist, not advocating for democratic reforms, advocating for their ethnic group without advancing a claim to autonomy or independence as nationalist groups do.

9. All analysis was conducted using Stata version 14.0.

10. The effect is calculated assuming the variable’s mean value of roughly two.

11. The calculation is made assuming the variable’s mean value of 0.4537.

Additional information

Funding

This material is based upon work supported by the Science and Technology directorate of the US Department of Homeland Security under Grant Awards N00140510629 and 2008-ST-061-ST0004, made to the National Consortium for the Study of Terrorism and Responses to Terrorism (START). The views contained in this document are those of the authors and should not be interpreted as representing the views or policies of the Department of Homeland Security or START.

Notes on contributors

Victor Asal

Victor Asal is a Professor of Political Science and Director of the Center for Policy Research at the Rockefeller College, University at Albany, State University of New York. His research interests include comparative politics, the choice of violence by nonstate organizational actors and the causes of political discrimination by states against marginalized groups.

Joseph M. Brown

Joseph M. Brown is an Assistant Professor of Political Science at the University of Massachusetts Boston. His research interests include terrorism, state repression, and protest movements.