• Example: Muzaffarnagar Riot
• Strategic ethnic violence
• Wilkinson (2004)

What factors seem important in explaining why this violence took place?

Fearon and Laitin (2000) identify different different strategic logics:

• “elite manipulation”
  • political elites encourage/foment ethnic violence for strategic reasons
• “on-the-ground”
  • ordinary people have incentives to participate in ethnic violence

Political elites may encourage violence for several reasons:

• Win Elections:
  • increase attractiveness of ethnic vs other parties
  • suppress/displace voters for rival parties
• Consolidate power:
  • shift balance of power within ethnic group
  • establish political dominance of one ethnic group

Violence as a means to accomplish these goals (directly or indirectly)

Theorizes two strategic logics for ethnic violence:

1. Electoral incentives create a motive to perpetrate violence

2. Electoral incentives shape the opportunity for violence (the use of state/police forces to limit violence.)

Why would political elites encourage ethnic violence?

• ethnic parties use violence to mobilize and capture the votes of co-ethnics

  • ethnic groups may not be united behind ethnic parties
  • ethnic parties have incentives to increase the salience of ethnic identity
  • provocative/controversial actions, processions
  • violence links physical safety to ethnic membership \(\to\) salience.

• Violence may unify ethnic voters behind ethnic party \(\to\) winning elections.

If this is true, there are empirical implications:

• ethnic parties must have capability to foment violence
• ethnic parties stand to gain from violence, should encourage it
• violence is more likely near elections, and when elections are competitive (close)
• violence should actually affect voting
• non-ethnic parties stand to lose, should stop violence

What enables elites to foment/stop violence? In India…

• Brass (2004): “institutionalized riot networks”
  • patronage networks connecting low-level party officials, Hindu nationalist organizations, criminals, police officials
  • can be used even when not holding office

Berenschot (2011):

• “riot networks” are really just everyday political networks
  • Political influence over capacities for violence (police/crime), e.g. Delhi Riots
  • political parties use these networks differently

Berenschot (2011):

In Ahmedabad (Gujarat) during 2002 riots:

• Mixed Dalit-Muslim neighborhood with BJP (Hindu party) patronage network saw mobilization for violence
• Mixed Dalit-Muslim neighborhood with Congress (multi-ethnic party) patronage network saw mobilization to defuse violence

Key empirical implications:

• ethnic parties must have capability to foment violence (yes)
• ethnic parties stand to gain from violence, should encourage it (anectdotally)
• violence is more likely near elections, and when elections are competitive (close)
• violence should actually affect voting
• non-ethnic parties stand to lose, should stop violence

Wilkinson (2004) looks at riots in towns in Uttar Pradesh

• towns are more likely to experience riots when elections are less than 6 months away
• towns in which the last MLA election was won by \(<5\%\) more likely to experience riots

• Is this evidence that parties use violence strategically? (Are there other explanations for this pattern?)

Heightened ethnic tensions for other reasons (e.g., nationalist propaganda, prior violence, etc.)

• may cause elections to be competitive
• may make additional riots more likely

Hard to know that electoral competition causes riots to be more likely.

Key empirical implications:

• ethnic parties must have capability to foment violence (yes)
• ethnic parties stand to gain from violence, should encourage it (anectdotally)
• violence is more likely near elections, and when elections are competitive (close) (maybe)
• violence should actually affect voting
• non-ethnic parties stand to lose, should stop violence

How would we know whether riots benefit ethnic / hurt non-ethnic parties at the polls?

• comparing places with riots to places without?
• comparing election results in a place before/after riots?
• comparing changes in election results in places with riots against places without

Nellis et al (2016) find that riot in the year prior to an election…

• decreases votes for Congress Party (multi-ethnic) at the next election, compared to places without riot
• increases votes for BJP (Hindu ethnic party) at the next election, compared to places without riot

Iyer and Shrivastava (2018) exploit “as-if random” riots to find the effect of riots on BJP (ethnic party) voteshare:

• look at riots occurring when Hindu festivals fall on Friday (Muslim holy day)
• these occur due to arbitrary overlap in religious calendars, inducing precipitating events
• find that riots increase BJP voteshare

Key empirical implications:

• ethnic parties must have capability to foment violence (yes)
• ethnic parties stand to gain from violence, should encourage it (anectdotally)
• violence is more likely near elections, and when elections are competitive (close) (maybe)
• violence should actually affect voting (yes)
• non-ethnic parties stand to lose, should stop violence

People with control over state/police/military forces have much stronger capacity to stop violence.

• e.g., arresting/shooting rioters likely to bring a halt to violence.
• Are there strategic incentives for governments to stop or permit violence to continue?

In late February 2002, a train carrying Hindu nationalists home to Gujarat from Ayodhya\(^*\) caught fire, 58 people died

• the cause of fire is disputed, but allegations that Muslims had set fire to the train
• Hindu nationalists across India held demonstrations, processions, or attacks against minorities during this period, explicitly linked to the violence in Gujarat.

Confrontations across India (see squares)

But… major riots (circles) limited to Gujarat

Government in Gujarat did not stop the riots. BJP (Narendra Modi) government:

• transferred officials who prevented riots/arrested Hindu militants
• delayed calling the army
• punished people filing police reports
• instructed officials to not take action to prevent violence

in other Indian states:

• state governments moved swiftly to quash conflict
• few riots; places with deaths entirely due to police firing on rioters.

Why did Gujarat permit riots to occur while other states did not?

Government strategy dictated by elections: will only stop violence if they directly or indirectly depend on votes of people targeted by the riots

This can happen under two sets of conditions:

1. When many parties compete successfully, minority group voters can determine who wins. Permitting riots that target this group may cost any ruling party victory at the next election.

2. When only a few parties are competitive, only parties that need the support of minority voters will stop riots against that group (non-ethnic parties stop violence; ethnic parties do not).

• States with many competitive parties protected Muslims
  • even in Orissa, Kerala, BJP led coalition stopped riots.
• States with few competitive parties yet had ruling coalitions dependent on Muslim voters stopped riots.
• Only in Gujarat was there few competitive parties, ruling party did not need Muslim votes

This logic appears to hold more generally.

Examining Hindu-Muslim riots by month in Indian states between 1961-1995…

• when an Indian state has greater electoral competition (more competitive parties), making Muslims more pivotal, they have fewer riots
• this pattern holds, even when other attributes of the state/month are held constant

Nellis et al (2016)

Indian National Congress party historically depended on Muslim voters. As a non-ethnic party, Congress stood to lose from riots.

Does electing a Congress MLA cause a constituency to have fewer riots?

• when the non-ethnic party has both motive and opportunity, does it work to inhibit ethnic violence?

How would we know whether Congress MLAs stop riots?

• compare places with/without Congress MLAs?
• could there be bias in this comparison?

Nellis et al (2016) focus on places with random exposure to Congress MLA:

• Compare constituencies where Congress barely won (by \(<1\%\)) to where Congress barely lost (by \(<1\%\))
• like an experiment, with treatment (Congress MLA) and control (Non-Congress MLA)
• No significant differences between constituencies where Congress won vs lost close elections.

Key empirical implications:

• ethnic parties must have capability to foment violence (yes)
• ethnic parties stand to gain from violence, should encourage it (anectdotally)
• violence is more likely near elections, and when elections are competitive (close) (maybe)
• violence should actually affect voting (yes)
• non-ethnic parties stand to lose, should stop violence (yes)

Ethnic political parties have electoral incentives to foment ethnic violence or permit it to happen: