You are using an older browser version. Please use a supported version for the best MSN experience.

How accurate are exit polls?

LiveMint logoLiveMint 13-05-2014 Manisha Priyam

With results for the Lok Sabha elections being announced on Friday, the anxiety to “know” is being filled by exit polls. There is also the caution that in the past, specially in 2004 and 2009, pollsters have been unable to get the results right, and there has been a bias in the stylized presentations of poll results inclined towards the National Democratic Alliance (NDA). Flying blind, therefore, are debates about adequate sample size and methodology, randomization strategies and, most importantly, the leap taken from vote share to seats.

As new alliances form, voting behaviours shift and there is increasing fragmentation of the political system, placing a bigger challenge on forecasting models. State political systems are the new reality—and these are far from conforming with Duverger’s law of a two-party system emerging with maturity in the political system—the hope is that tweaking the quantitative methods may just be enough to get the figures right.

The salient features of this debate miss out critical elements of the electoral process—time-cycle and voter decision parameters. These are critical to understanding the electoral strategies on offer, high voter turnouts and, ultimately, why the individual votes.

Political ethnography is increasingly being used as a research strategy to understand these issues. Apart from a meaningful understanding of the process, this type of qualitative research allows for a meaningful dialogue with quantitative methods. There is also the possibility of “bridging the gap”, in terms of highlighting specifically in what ways eliciting voter responses or building vote-to-seat projection models may be off the mark. In this article, I build on insights from electoral ethnography in Delhi and Bihar, and present some insights on understanding the vote, and also on the need for abundant caution.

One important lesson underscored by the field-work is that voters are simply not willing to reveal who they will vote for—not to questions popped up by a field data administrator, nor even after building community trust with regular interactions over a period of time. In Delhi’s Sangam Vihar for example, Muslim women voters vented their anger as they spoke of the acute shortage of water in the area, and the backflow of drains in their houses, but never let out who they would vote for in visits over two electoral cycles—for the assembly elections last December and the Lok Sabha elections now.

Men were different—they signalled they would not vote for the Congress during the assembly elections last year, telling the researcher “ab dekhiye, Peace Party bhi to hai” (There’s also the Peace Party to consider). However, at the time of the Lok Sabha elections, they clearly articulated their intention to vote for the Aam Aadmi Party.

Haleem Saab, running a tea shop at the entrance of L-Block, said after much prodding that if the Aam Aadmi Party comes back with some seats, it can form an effective opposition. “Congress to hamaari majboori thi. Magar hum Bhaajpa ko vote nahin de sakte” (We voted for the Congress out of necessity, but we cannot vote for the Bharatiya Janata Party). In both instances, voters of this minority community were forthright in identifying the issues they considered important—inflation and joblessness—but did not state first up who they would vote for.

Similar was the case in Jehanabad, Bihar. Here Dalit voters, once under the influence of the ultra-Left Mazdoor Kisan Sangram Samiti, vented their anger on inflation. “We have bought bhindi at `80 per kg, and even potatoes at over `25 per kg. We celebrated Holi this March buying mutton at over `200 per kg.” These complaints were similar to what I had noted in Sangam Vihar, but the voter choices seemed differently inclined. “Our boys find jobs in Surat—not here. What do we vote for.” A woman cook at a school in Silaunjha, Gaya, told me that the votes would go to “Har, har Modi”. Here, and elsewhere in Chapra, Maharajganj, Siwan and Hajipur, the issues were stated upfront, and voters’ choices given a hint about. Those belonging to atipichhda (extremely backward) caste groups such as the Musahars were scared they would be beaten up by the village dominant if they stated upfront whom they would vote for.

In view of these robust findings across six Lok Sabha constituencies, it is important to ask the question: how could exit pollsters make people speak explicitly about where they cast their vote?

Another challenge refers to data-smoothing from large-scale surveys, and developing a political analysis from it. Specifically, this entails the following:

Making sense of multi-party competition and fragmented political scenarios by calculating the effective number of political parties as in the Taagepaara-Schugart index. This can be calculated on the basis of both the effective number of seats and votes. The difficulties arise, however, when parties fight as alliances. Besides, there are also new parties such as Upendra Kushwaha’s Rashtriya Lok Samata Party in Bihar that do not have a prior electoral history.

Second, the vote share-to-seat projection model uses assumptions about past voting behaviour.

The field work done in Bihar gives another note of caution—the BJP and Janata Dal (United) were electoral allies in the 2009 Lok Sabha and 2010 assembly elections. As allies, they had a share of the Muslim votes too. This time, however, the Muslim voter has steered clear of the BJP. Another issue is the atipichhda votes. The erstwhile allies had votes of the Mahadalits—all extremely backward castes, excluding the Paswans. This time, Ram Vilas Paswan is an ally of the BJP, and the other dalit votes are fragmented between the BJP and the JD(U).

More open debates on modeling strategies, and cross-questioning from ethnographic data can provide useful correctives. We have to wait to see the results on 16 May because a science of the “psephos” is yet to emerge.

Manisha Priyam is ICSSR Fellow at the Nehru Memorial Museum and Library.

More From LiveMint

image beaconimage beaconimage beacon