Can institutions successfully reduce ethnic conflict? Institutions such as autonomy and federalism are often advocated as a means to prevent ethnic conflict, however empirical evidence is largely mixed with regards to their effectiveness. In a similar manner, political parties have begun to receive more scholarly attention in determining their relationship with ethnic conflict, but their evidence is also mixed. In this research I examine autonomy, federalism, and the banning of political parties within ongoing ethnic group self-determination movements. While I do not find evidence for a relationship between autonomy and conflict, I do find that federalism increases the likelihood of ethnic conflict. Additionally, the banning of ethnic political parties indicates a strong increase the likelihood of ethnic conflict, while the banning of regional political parties significantly reduces the likelihood of ethnic conflict.

Abstract: What explains the use of violence in extra-institutional political campaigns? Domestic groups challenge host states using both nonviolent and violent tactics. While Gandhi’s struggle for India’s independence is perhaps the most famous example of nonviolence, many of today’s bloody civil wars also started out as nonviolent movements. In a world eager to support the self-determination of marginalized groups, both nonviolent and violent groups receive support from foreign actors. Despite this, theories on the use of violence by these groups remain untested empirically.  This paper uses panel data to quantitatively investigate the proposition that external support of extra-institutional political movements causes an increase in the use of violence. A logistic regression model finds no statistically significant relationship between the provision of external support and an increased use of violence in primarily nonviolent campaigns. An additional test on a sample of violent non-state groups finds that battle-related deaths increased when external support was provided in the previous year, a result significant at 99% confidence.