Across much of the Global South, in-person surveys remain a primary means by which researchers collect data on political opinions. Often, the survey enumerator reads out questions and records responses—a process which is regularly overseen by the respondents’ neighbors, friends, and family members.
The presence of these third parties is often unavoidable; consider, for example, that nearby 40% of the housing units in Pakistan – the site of our study – are single-room structures. Even in larger homes, affording a family member full privacy during an interview is not always the norm. Respondents may be nervous about letting strangers into their homes and choose to conduct the interviews at their doorsteps, which in turn can draw the attention of neighbors.
Despite this, little attention has been paid by political scientists to the presence of such known third parties at the time of an interview. In our study, we examine whether third party presence shaped responses to a range of questions about ethnic voting, ethnic discrimination, and interethnic relations. We find that respondents are significantly more likely to express ethnocentric views in the presence of others – a result with ramifications for the study of ethnic politics and survey methodology in the developing world.
Data Collection in Karachi, Pakistan
Our data comes from an original, representative survey of 1805 respondents from the megacity of Karachi, Pakistan. Karachi is home to over 17 million people from all of Pakistan’s ethnic groups; the city is fairly ethnically segregated and has been the site of ethnic political contestation and violence for many decades. Around half of our interviews (48%) were conducted in the presence of known third parties, while the remainder were conducted in privacy.
Results: More Support for Ethnic Politics in the Presence of Others
Our results, based on descriptive analyses – and quite robust to sensitivity analyses for unobserved confounding – are striking. In the presence of known third parties, respondents were significantly more likely to express ethnocentric views across a number of different measures. This included, for example, being more likely to agree that people have a responsibility to vote for coethnics, that the economic conditions of their ethnic group were worse than those of other groups, and that they feared violence on the basis of their ethnicity.
The result of having other present was consistent with the effect of the survey being conducted by a co-ethnic enumerator; in fact, the substantive effect size for third party presence was generally larger than that for enumerator coethnicity (see Figure 1).
Why did respondents express more ethnocentric political views in the presence of known others? Due to endogamy and ethnic segregation, we expect that any individuals present during the interview likely belonged to the same ethnicity as the respondent. As such, we suggest that the presence of others activated a norm of in-group solidarity, resulting in increased stated support for ethnic politics and group rights.
This norm may have been activated for one of a number of possible reasons. First, individuals may overstate support due to a fear of social sanctioning or a desire to reap the social or economic benefits of conforming to such norms. Second, respondents may feel as if they are speaking for their household or the wider community when in the presence of others.
Whatever the specific underlying reason, these interpretations all fall under the umbrella of social influence. Research suggests that individuals place high value on maintaining their closest social relationships, which are difficult to sustain when significant disagreements occur. To probe the validity of this social influence mechanism, we conduct four further tests. First, we test whether third party presence is associated with other variables that should be affected by a desire for conformity. We find that respondents interviewed in the presence of others are significantly more likely to say that they vote for the same party as their neighbors and friends and are also more likely to demonstrate socially desirable traits (such as an interest in politics) and less likely to admit to potentially undesirable habits (such as frequently watching soap operas on television). Second, we find that third-party presence does not affect items not susceptible to social desirability bias in this context (such as a question about electricity provision by a nonpolitical entity). Third, extant research shows that women may be more susceptible to having their opinions shaped (or censored) by existing social norms. Interacting our dependent variable with gender demonstrates that women are more likely than men to offer group-centric views in the presence of others along many of the outcome measures.
Implications for Future Research
Our results have implications both for our understanding of the drivers of ethnic politics and for survey research more broadly. Dominant explanations for ethnic politics in the Global South remain heavily centered around instrumental, clientelism-focused approaches; however, our study underscores the importance of social dynamics and norms in understanding how ethnicity operates in the political realm.
It also underscores the importance of carefully measuring this aspect of the survey environment in future studies. Indeed, researchers should figure the inevitability of third-party presence into the design, implementation, and analysis of surveys. However, the goal should not be to eliminate or artificially limit this common feature of the survey environment. Rather, it should be carefully leveraged to further existing research on the impact of social forces on political opinions and behavior.
Figure 1: Impact of third party presence (A) and enumerator coethnicity (B) on attitudes
This blog piece is based on a forthcoming article “Third Party Presence and the Political Salience of Ethnicity in Survey Data” by Mashail Malik andNiloufer Siddiqui.
The empirical analysis has been successfully replicated by the JOP and the replication files are available in the JOP Dataverse.
About the Authors
Mashail Malik is an assistant professor at the Department of Government, Harvard University. She specializes in the political psychology of identity, with a focus on ethnicity, immigration, and internal migration. For more information visit her website.
Niloufer Siddiqui is an assistant professor of political science at the Rockefeller College of Public Affairs and Policy, University at Albany-State University of New York (SUNY). Her research interests include political violence, political behavior, the politics of religion and ethnicity, voters and foreign policy, and the politics of South Asia. For more information visit her website.