Using Text Messages against Politicians with Criminal Charges in Elections in India

Researchers:
Sarika Ringwala
Siddharth George
Location:
Uttar Pradesh, India
Sample:
4,131 villages across 38 assembly constituencies
Timeline:
2014 - 2017
Initiative(s):
Governance Initiative
Target group:
  • Politicians and electoral candidates
  • Voters
Outcome of interest:
  • Electoral participation
  • Voter behavior
Intervention type:
  • Information
Partners:

Voters may not be able to make informed decisions about who to vote for if they lack information about candidates’ qualifications and background. Politicians who have criminal charges, for example, can be detrimental for economic growth, poverty, and crime in their constituencies; yet voters may not have enough information to screen out such candidates during elections. Researchers studied whether sending voters voice and text messages containing information about candidates’ criminal charges affected candidates’ electoral support during the 2017 Uttar Pradesh state assembly elections in India. They found that it reduced total votes for candidates with serious criminal charges by 7.7 percent and increased votes for candidates with no charges by 6.7 percent, and increased overall voter turnout by 1.6 percent. Messages were most effective when they included additional information that prompted coordination among voters.

Policy issue

Voters may not be able to make informed decisions about who to vote for if they lack information about candidates’ qualifications and background. One example is the case of criminal politicians: recent work suggests that politicians with criminal charges reduce economic growth and increase poverty and crime in their constituencies.1 Despite the negative consequences of electing politicians with criminal charges, voters may not have the information required to screen out such candidates and to coordinate in favor of better alternatives. Likewise, lack of information about candidates’ qualifications or policy positions may lead voters to choose candidates based on other factors, like ethnicity or religion, that don’t always align with voters’ own policy preferences.

There is a strong body of research on the impacts of voter information campaigns, but key questions remain: What types of information are most useful for voters facing many choices of different quality? What platforms and methods are most effective for communicating information to voters, and why? In the context of the 2017 Uttar Pradesh elections in India, researchers evaluated the effect of voice and text messages containing information about candidates’ criminal charges on voters’ support for those candidates.

Context of the evaluation

Criminal politicians are a serious concern in Indian politics. In India in 2014, 34 percent of members of parliament (MPs) were facing criminal charges, and 9 percent of legislators faced charges for murder, kidnapping, rape, or armed robbery.

In the state of Uttar Pradesh, the setting of this study and India’s most populous state, nearly 22 percent of members of the legislative assembly (MLAs) elected in 2012 faced serious criminal charges like murder, rape, kidnapping, and extortion. Out of the 119 major party candidates in this study, 35 percent faced at least one criminal charge. 40 percent of those candidates faced charges for violent crimes.

The election of crime-associated politicians has repercussions: MLAs play a significant role in resource allocation within their state, and data shows that MLAs facing criminal charges generate less economic activity than their crime-free counterparts.2 Ethnic politics underline the electoral process, and evidence suggests that votes along ethnic preferences may result in lower quality candidates.3

Mobile phone usage rates have increased in India in recent years; on average, the combined coverage rates of three major telecommunications companies (Idea, Airtel, and Vodaphone) in Uttar Pradesh was 44 percent at the time of this study. This suggests that mobile technology may be a feasible way to share information, even in rural areas.

Details of the intervention

Researchers partnered with the Center for Governance and Development, the UK Department for International Development, and telecom companies Idea, Airtel, and Vodafone to measure the effects of voice and text messages containing information about candidates’ criminal charges on electoral outcomes during the 2017 Uttar Pradesh state assembly elections.

Researchers randomly assigned 4,131 villages across 38 assembly constituencies to one of four groups, sending voice and text messages to over 450,000 individuals:

  1. Basic information: All mobile users with registered numbers in this group of villages received a text and voice message urging them to get to know their candidates and think carefully about their vote. The message also contained information about the number and types of criminal charges, if any, facing all of the major party candidates in their constituency, differentiating between violent and non-violent charges.
  2. Information + coordination: Mobile users in these villages received the contents of the basic information message and content emphasizing that many other citizens in the area received the same message. Researchers sought to determine whether this message would encourage voters to coordinate their response.
  3. Information + ethnic voting: Mobile users in these villages received the contents of the basic information message and an additional appeal to break the habit of voting along caste lines.
  4. Comparison group: Individuals in these villages received no information-related text messages.

Individuals in groups 1 through 3 received one voice message and one text message two days before the election.

In order to obtain information about each candidate’s criminal charges, researchers used publicly accessible sworn statements that all candidates are required to submit to the Electoral Commission of India. Researchers used publicly available electoral data from polling stations, as well as demographic information about villages from the 2011 Indian national census, to determine the effects of the messages on electoral results and turnout.

Results and policy lessons

Researchers are still analyzing data—allresultsreported below arepreliminary.

Upon receiving information about candidates’ criminal charges, voters reduced their votes for candidates with severe charges and increased their votes for candidates with no charges. These effects were strongest among those who received the information + coordination message.

In villages where at least one candidate had a murder-related charge, informational messages increased total votes across all candidates without any criminal charges by an average of 6.7 percent per candidate. While total votes increased, overall vote share across these candidates did not significantly change. Votes for candidates with murder-related charges (murder or attempted murder) dropped by an average of 7.7, reducing their vote share by 2.2 percentage points. However, there was no change in total votes (or the vote share) for candidates with criminal charges not related to murder.

Candidates without criminal charges did not see a boost in total votes on average in villages where no other candidate had a murder-related charge. This suggests that having an opposing candidate with a serious criminal charge may be needed to prompt voters toward alternatives without such charges.

Overall voter turnout increased by 1.6 percent on average in villages receiving any informational message.

Impacts of basic information: This message did not have a discernable impact on vote shares for any candidates.

Impacts of information + coordination: This message increased vote shares for candidates with no criminal charges by an average of around 0.6 percentage points per candidate and reduced vote shares for candidates with murder-related charges by an average of 2.5 percentage points. It did not have a discernable impact on vote shares for candidates with other criminal charges not related to murder.

Impacts of information + ethnic voting: This message did not have a discernable impact on vote shares for any candidates.

These results suggest that voters are more responsive to information about candidates’ criminal charges when they receive additional information that prompts coordination with other voters. This may be because a signal that others have received the same type of information may lead to explicit discussion about, and agreement on, voting for a certain type of candidate – or induce a tacit understanding that other voters will respond similarly.4

Overall, providing this information to voters influenced their voting decisions when transmitted via easily scalable, mobile-based platforms. This suggests that information delivery campaigns can reach voters in remote locations with mobile coverage. Furthermore, spreading information about criminal politicians can elicit a response against the worst offenders (in this case, those with murder-related charges) without creating a strong backlash against potentially better candidates who face petty or spurious charges.

Citation: Siddharth George, Sarika Gupta, Manoj Kumar, and Yusuf Neggers. 2018. “Coordinating Voters against Criminal Politicians: Evidence from a Mobile Experiment in India”. Working Paper.

1.
Nishith Prakash, Marc Rockmore, Yogesh Uppal, et al. Do criminally accused politicians affect economic outcomes? evidence from india. Households in Conflict Network (HiCN), The Institute of Development Studies, University of Sussex, 2015.
2.
Nishith Prakash, Marc Rockmore, Yogesh Uppal, et al. Do criminally accused politicians affect economic outcomes? evidence from india. Households in Conflict Network (HiCN), The Institute of Development Studies, University of Sussex, 2015.
3.
Abhijit Banerjee and Rohini Pande. Parochial politics: Ethnic preferences and politician corruption. 2009.
4.
Arias, Eric, Pablo Balán, Horacio Larreguy, John Marshall, and Pablo Querubín. "Information Provision, Voter Coordination, and Electoral Accountability: Evidence from Mexican Social Networks". American Political Science Review 113 (2): 475-498.