WikiSummary, the Social Science Summary Database

Wilkinson: Votes and violence

From WikiSummary, the Free Social Science Summary Database


Wilkinson. 2004. Votes and violence: Electoral competition and ethnic riots in India. Cambridge: Cambridge University Press.


A political theory of ethnic (religious, racial, linguistic) violence: When politicians need minority support, they prevent violence. When they don't, they don't. And if they need to incite ethnic polarization (e.g. in order to bring more of their ethnic group into the majority party), then they might just promote ethnic violence.

Chapter 2: Town-level factors

Although Wilkinson's theory is state-level, he uses this chapter to examine the town-level causes of ethnic violence. Why a state-level theory? Because although local factors may influence whether violence breaks out, it is the state-level authorities who must order in police and reinforcements to contain it. Even federal troops can't be deployed unless the state government orders it. Thus, they key argument in this book concerns the state-level decision of whether to prevent or encourage violence. However, local can also matter. Hence, this chapter.

Electoral benefits of ethnic violence

The Hindu nationalist parties tend to represent primarily the upper castes. These upper castes cannot attract lower-caste Hindus with promises of redistribution for two reasons: The promise wouldn't be credible, and it may alienate the upper caste supporters. So if the election is going to be close, what can an upper-caste party do to win more votes? Incite Hindu-Muslim riots. This will pull more lower-caste Hindus into the upper-caste party.

Inciting these riots is costly. Often, the mechanism for doing so is to schedule a religious processional (innocuous enough, right?) to go through a Muslim neighborhood, or to raise a Hindu flag over some disputed piece of land. But organizing these processionals and whatnot takes time, money, and so on. In addition, riots can have economic costs. So we should expect these mechanisms to be used primarily when the extra electoral support is most needed, namely:

  • When elections are coming up
  • When the margin in previous state elections was close

Other explanations from lit

  • Economic competition: Violence is most common when the ethnic groups compete economically. Wilkinson finds that economic variables do not predict the outcome of violence, but they do explain its severity. Thus, once groups realize that the police aren't going to stop the violence, they might loot the other group's stuff, but looting won't be the cause of the riot.
  • Violence begets violence: Expect violence where it has occurred before. Why? When faced with uncertainty about the prospect of violence, people think about how the other group has responded in the past. Thus, a history of violence makes violence more likely.
  • Ethnic parity begets violence. As the number of Muslims approaches 50%, violence is more frequent.

Chapter 5: Electoral incentives for Hindu-Muslim violence

Primary argument (see Fig 5.1): In states with high partisan fractionalization (i.e. high partisan competitiveness), parties have greater incentives to compete for the Muslim/minority vote (under certain conditions), therefore there should be less violence. The argument has three elements:

  1. Politics In India are multidimensional, though this varies. As politics becomes more multidimensional, we can expect more overtures to minorities. For example, Northern white industrialists (in the US) had incentives to court the Southern minority vote to gain political power over Southern white planters (in the 1920s-50s). As the number of cleavages rises, the potential for such overtures increases.
  2. It is worthwhile to compete for the minority vote if the electoral benefits of winning the minority vote exceed the political costs of giving the minority what it wants. Since India's Muslims tend to be poor and populous, they tend to have a single demand: security. This might change as Muslims become wealthier and better educated (they might start demanding bgovernment employment, economic privileges, and so on). But for now, there are many Muslim vote that can be bought with security.
  3. Providing security is not costly. It might be if providing minority security threatens the majority; for example, if it involves putting substantial numbers of Muslims in the police and military forces. But most Indian police and military forces have very few minorities, so providing security isn't costly.

How Indian states fit this model

First, there is multidimensionality. Though Muslims favor security, most Hindus are more concerned with economic redistribution and other issues than with holding Muslims down. Second, since Muslims demand less than most Hindu voting blocs (they only want security), they are a low-cost constituency, so they are attractive to parties that need more votes. Finally, less than 1% of the armed forces is Muslim, so granting security doesn't scare Hindus.

Test #1: Statistical relationship btw ENPP and violence

  • Y: Number of riots, and number killed in riots (by state)
  • X: ENPV: Effective number of parties. As it rises, violence goes down (Fig 5.2).
  • Controls:
    • Surprisingly, increased disparity (GINI) and decreased literacy lead to less violence. Though Wilkinson doesn't discuss it, this seems to fit his theory: More poverty and less education make make [Muslim] voters more concerned with security.

Test #2: Do A, Bi, and Bii states fit the model (see Fig 5.1)

As shown in Table 5.4, Gujarat is the only Bii state (see Fig 5.1). It's also the only state that allowed violence.

  • [Concern: Is there really enough variance in X and Y, then, to be confident in the statistical tests?]
  • [Concern: Why does percentage of Muslims supporting the governing party matter? What should matter is the percentage of the governing party's supporters that is Muslim.]

Test #3: Evidence of politically strategic considerations?

Yes. Looks at a couple case studies. People apparently are thinking the way he says they are.


In addition to those mentioned in the summary, this is major: What about reverse causality? In light of Chapter 2, shouldn't we expect ethnic provocations (riots) to lead to less partisan fractionalization, since lower-caste Muslims join the upper-caste party? Violence makes ethnicity the primary cleavage, potentially reducing ENPP.

Chapter 7: Applying the theory comparatively

General idea

This chapter applies the book's theories to historic cases. Basically, it argues that an increase in electoral competition can lead politicians to pay more or less attention to preventing violence. "Governments ... decide whether to prevent antiminority violence by calculating whether doing so will help or hurt them politically." If the ruling party needs the minority's support to win, it will suppress violence. But if it doesn't, it might allow violence. And if it needs a few more votes from the majority, it might encourage violence, since ethnic violence will polarize the ethnic groups and (possibly) bring more of the majority ethnic group into that group's party. Examples:

  • Ireland. At first, Catholics and some protestants were going liberal, and the Tories were primarily Episcopalian. However, they began to lose to the Liberals, so they turned their party into the anti-Catholic party. Thus, they managed to pull some protestant Liberals (Methodists, Presbyterians) into the Tories. Violence was limited, however, since the colonial administration in London did not face these ethnic incentives; rather, London's incentives were for peace and stability, so they promoted an ethnically balanced police force that generally suppressed violence.
  • Malaysia: Independence came through a joint Chinese-Malay struggle. However, soon after independence, the independence party realized that it might not do so well electorally. By provoking ethnic violence against the Chinese, it established itself as "the" Malay party and won comfortably in elections.
  • Romania: Under Ceaucescu, Hungarians [living mostly in Transylvania] were somewhat "Romanized." After independence, the first government sought to allow the Hungarians more freedom. Soon, the Romanians in Transylvania began to be upset that the Hungarians were becoming strong. Fearing that pro-Romanian parties would rise and defefat it, the ruling government allowed significant anti-Hungarian riots to continue for several days before intervening and stopping them. With its pro-Romanian credentials firmly established, the party won comfortably.
  • US: For years, the northerners stayed out of the South, allowing Southern Democrats to use racial issues to remain dominant. By around the 1940s, however, winning the presidency required either party to win key swing states in the North, including Illinois and Michigan. To win in these states, the Democrats played to their strength: urban voters. But with the urban areas in these states becoming increasingly black, the Democrats realized that they would need to win the black vote to remain competitive. Thus, the Democratic leadership ceased allowing blatant intimidation and ethnic attacks in the South.


  • Can you really attribute ethnic conflict primarily to political incentives?
    • What about hate, religiosity, and cultural explanations for the tension?
  • What is the solution to ethnic conflict? An end to democracy?
    • After all, it was the (less democratic) colonial influence of London that minimized ethnic conflict in Ireland.
  • Is this theory presenting only a necessary condition for violence (i.e. violence occurs when politicians allow it to)? What is the sufficient condition? (Ch 2 makes an attempt at this)
  • Where's the logic against reverse causality (for chapter 5)? It seems that high party fractionalization presents a strong incentive for violence--especially since this violence (according to Wilkinson's own logic) can build support for the inciting party.
  • Is this really a theory of violence, or only of electoral heavy-handedness? It seems that the conditions Wilkinson identifies might lead governing parties to engage in massive fraud, not necessarily to incite major violence. Why violence instead of fraud?