Political Violence in Nigeria: A Field Experiment around the 2007 Elections

RESEARCH QUESTION

Can a NGO-sponsored campaign decrease election-time violence in Nigeria? Do social networks have a role in ‘grassroot’ campaigning? Does political violence affect election outcomes or behaviour?
 

PROJECT

Various sources point to the frequency of violence during election time in Nigeria. Humans Rights Watch, for instance, released a full report on Political Violence before the 2003 round of elections - it underlined the potential important role of violence in determining political outcomes in many specific Nigerian states. The same setting is anticipated by most observers to happen during the 2007 elections.

We proposed a deeper understanding of the consequences of election-time violence in terms of electoral behavior, while simultaneously exploring potential avenues to counteract it on policy grounds.

We prepared a randomized intervention research design with the support of a network of local NGOs headed by ActionAid International Nigeria. This network conducted a randomized campaign against political violence, before the 2007 elections, in states with a history of violence – since this treatment was administered to a randomized set of locations, there was a similar set of locations serving as control group.
This campaign, if effective, would create exogenous variation in political motivations for violence. This may materialize in diminishing the will of local politicians to support violent actions by their supporters. Even if not affecting the will of politicians directly, it would induce voters to react differently to violence (e.g. by turning up to vote or by voting differently).
This variation would enable deriving the effects of the campaign itself on violence, studying political violence as an exogenous determinant of voting behavior, and analyzing effects of connectedness between subjects exposed to the campaign (social networks). The latter is a fundamental exercise for the knowledge of how ‘grassroot’ campaigns may be effective.

We intended to derive the effects of this campaign by conducting a suitably designed before-after panel survey structure in both treatment and control locations. These surveys took place immediately before the campaign (which took place before the elections) and immediately after the elections. They elicited information on perceived and experienced crime and violence, as well as on voting behavior, and demographics (including social network proxies).

Similar work on vote buying with comparable designs has taken place in Sao Tome.