Monthly Archives: April 2016

Where is Fatima Bi? Reflections on conducting surveys in India

The political and social context of a research site may influence the availability, character and quality of statistical information, especially those studies reliant on electoral lists for randomisation. Using my quest for Fatima Bi for an interview as a metaphor, I contemplate the local political factors that influence the construction of electoral lists and voting, and intermediate the ‘availability’ of data. I argue that these factors accentuate or attenuate local trends, and thereby impeding the interpretation of studies.

-X-

After several months of ethnographic work in Area X, I plunged into preparing a survey – making questionnaires, poring over maps and seeking electoral lists from which to sample respondents. I intended to use a strategy frequently used in randomized surveys in India – I selected respondents for interviewing randomly from official voter lists. I marched out on enthusiastically on a clear winter day through the narrow gullies of Area X, printed questionnaires in hand, ready to interview my first respondent, Fatima Bi.

But where was she? Fatima Bi simply could not be found. Frustrated, I sought the counsel of a local NGO worker who informed me that some recently prepared voter IDs had wrong addresses. Together, we asked several individuals in the neighbourhood and, a few hours later, we finally found Fatima Bi.

Here, I use my struggles in implementing a probabilistic sampling strategy and my quest for Fatima Bi as metaphors to reflect on the political realities that distort quantitative work. The promise of neutrality has been at the basis of the surge in quantification in political research, which in turn relies on the ability to randomise. Random samples represent the larger population from within which they are drawn, and thus, sample characteristics can be said to be true of the population. In the coming sections, by describing the machinations in the preparation of voter lists and instances of vote-buying in Area X, I will argue that statistical findings are constituted by context within which they are instantiated and animated, and a statistic, therefore, is a creature of the space it is born in.

Preparation of voter lists

In the few days prior to elections, all parties put ‘leaders’ to work to boost enrollments from potential voters.* These efforts are usually undertaken in the last weeks before enrollments close. Support from political aides during these pre-election enrollment drives involves filling up and collecting forms, following up at the office of the Electoral Registration Officers responsible for electoral lists, and securing and delivering the Election Photo Identity card (“EPIC” or “ID card”).

Those involved in such work claim to be unconcerned about voters’ loyalties and perceive the process of procuring IDs as “social work”** rather than as a partisan effort. There is a debate amongst scholars as to whether parties target marginal voters or loyalists with clientelistic benefits. In Area X, for the delivery of ID cards, parties indeed target loyalists. P, a local Congress leader tells me, “Every party must help their people, is it not?”. Moreover, though the majority of Area X voters have IDs due to government enrollment efforts in the early 2000s, newer voters, mainly migrants, report that they go to their party leader for IDs.

Arguably, in the face of equal effort by all parties, roughly equal proportions of loyalists would be enrolled. Thus, the proportion of voters in the electoral list that support any party may be unaffected, not troubling the randomization process. Here’s the rub – in fact, leaders’ personal ambitions and party interactions with these leaders combine to influence the rate of enrollment.

Leaders ambitions determine the level of investment in the enrollment process. D, a Congress party leader prides himself in his long-term service to the poor. His years of service have earned him the post of nominated member of the municipal corporation , a senior position that gives him status within Area X. Several posters stand testimony to this. A crucial requirement in order to complete enrollments is ‘boys’ to do the required oDaTa (running around). In the few nights before lists close, D leverages his power, generated from his formal position and years of service to engage several local youth to enroll voters. D is politically ambitious and has contested the municipal elections in the past. In contrast, P has lesser goodwill, and is unable to marshal resources.

Additionally, political parties use different approaches in managing ground level networks, also affecting enrollment rates. In Area X, the Congress party is reliant on leaders such as P and D, and other lower level leaders managed by them, who have been nurtured over time. In contrast, the BJP, which is newer in Karnataka is reliant on leaders it engages during election. These workers are otherwise inactive and, unlike P an D, are not entrenched in the bureaucratic systems for enrollment. BJP in-charge for Area X does not even live in Area X.

A peculiar factor is that Area X has several residents from Area Y, which is in a neighbouring assembly constituency and was recently evicted. I found that 7.5\% households in Area X had formerly lived in the evicted area and this proportion represents close to 200 households . Many of these voters continue to vote in Area Y, though some have been enrolled in Area X. While this is a temporal occurrence, studies relying on random sampling in Area Y will be unable to find these voters as they now reside here. Dynamism of neighborhoods due to migration and movement is a feature of urban life and contributes to a significant discrepancies in voting lists.

Thus, election lists in Area X likely do not represent the population that ultimately votes. Temporal factors, like these shifts combine with voter list preparation practices to bring in or exclude voters whose preferences may not have been included in pre-election survey research.

Vote buying and turnout buying

A focus within the clientelism (very simply, the exchange of goods, public or private, for votes) literature has been to determine the motivations of the actors (brokers, voters, party bosses) in the process. While clientelistic practices may undermine the fairness of a democratic process, deeply rooted systems of reciprocity are unlikely to complicate probabilistic research as sampled voters may indicate their preference for clientelistic parties even when surveyed. However, some practices of vote-buying, such cash for vote exchanges have the potential to seriously undermine sample-based survey research as they are concealed and furtively implemented in the days prior to elections. In this section, I describe particular practices of vote-buying in Area X that I have encountered.

In a perverse form of vote-buying in Area X, parties are able to magnify the proportion of votes they actually win by preventing turnout of opposing partisans. Local workers of both major parties allege that parties “purchase” voter IDs of impoverished loyalists*** of the opposing party prior to elections to ensure that they do not cast their vote. Setting aside the normative and legal implications, this practice reduces the monitoring costs associated with targeting goods to voters (leaving only the cost of the good and distribution costs), and is an effective tactic in first past the post systems. The moral turpitude involved makes the precise extent hard to discuss, investigate and verify. However, in accordance with the literature, if the poor are particularly vulnerable, such practices may result in lesser proportions of the poor on voting day than the electoral lists suggest.

Other strategies are also used to prevent or encourage voters to support particular parties. In Area X, voters are made to swear on religious texts that they will vote for a party, others have stated that they were threatened with water disconnections and denial of social entitlements (such as cancellations of ration cards) if they do not vote for a particular party. The instrument varies with the individual and the community, as does the target, and one interviewee told me, “these leaders know our vices!”.

In the case of Area X, these practices lead to a distortion of voters’ presence on election day. The particular vulnerability of the poor to practices of vote-buying may lead to their absence at the pooling booth. This presents a serious challenge to the ability of random samples to represent the views that find utterance (or not) on voting day.

Conclusion

The combined effects of practices of voter list preparation and vote buying in Area X presents a challenge to the expectation in probabilistic research that proportions within the survey sample roughly represent the proportions among those who will vote. Pre-election spurts of enrollment may drive increased presence of certain types of voters; similarly, vote buying by political parties influences who appears to vote as well as the actual vote choice itself. You see, even if Fatima Bi is found, it is not certain that her actions on voting day are in accordance with her survey responses.

Footnotes

* The preparation of voter lists, though overseen by the Election Commission and administered by the Bruhat Bengaluru Mahanagara Palike in Bangalore, is in often controlled by local party workers and leaders. The current system of enrolment allows voters to apply for an election photo identity card and submit relevant proofs. Many voters are unable to cope with the cumbersome process and consequently rely on external help.

** “Social work” is a term freely used in Area X for services rendered where the beneficiary does not directly compensate the service provider.

*** It is generally expected that the poor are most vulnerable to offers of targeted goods as the marginal utility of a particularistic good (usually cash) is higher for the poor.