Exit polls are often wrong, but don’t blame silent voters for it

In Lok Sabha elections, varied types of silent votes frequently cancel each other out and the net result is close to exit poll forecasts.

Representational image | PTI File photo

The exit poll predictions for the number of seats the Bharatiya Janata Party-led National Democratic Alliance is likely to win on 23 May show wide variation. But all the exit polls converge on a few points.

First, all point to a clear second term for Prime Minister Narendra Modi, with a slight chance of the BJP getting a majority on its own. Second, the gap between the top two parties, in terms of seats, is again likely to be huge and the possibility of the Congress party getting close to three digits is extremely low. Third, the opposition alliance in many states have not clicked. On the contrary, it seems that the anti-BJP coalitions in many states are witnessing a miscoordination effect and getting lower vote shares than their combined strength.

Infographic by Arindam Mukherjee | ThePrint

The silent voter

The exit poll findings are at odds with various commentaries that appeared during the campaign period in Indian media. It is natural, then, that we are seeing very sharp reactions in social media on exit poll findings. We were told that there is deep-seated resentment against the BJP-led NDA government and the alliance will lose almost half of its current strength in the house. We were also told that on many occasions that reporters met voters who, on the first instance, showed a preference for the BJP, but upon deep probing and prodding revealed their true preference— voting against the BJP.

The underlying premise behind this phenomenon – the silent voters – comes back every election. The claim is simple and intuitive: a significant chunk of the electorate is reluctant to publicly admit their displeasure with the ruling dispensation (the incumbent party or the dominant caste party in the region) due to fear of physical threat or loss in beneficiary status, but silently expresses it at the polling booth by voting against them. We have been warned many times that these silent voters have often upset exit poll findings.

Who are the silent voters in India? And can they upset even the median numbers coming out in these exit polls? In our view, we need a fresh perspective on the theory of the silent voter. Analytically, we have been misled on this question for too long now.

Not necessarily strategic silence

Why do some voters fail to articulate their vote preference publicly? While fear of backlash is an important factor, we think that for most cases, observers have misconstrued the inability to articulate a clear vote choice as strategic restraint. The reason is rather obvious: a substantial section of the Indian electorate makes up their mind about whom to vote for just a couple of days before or on the polling day. This undecided voter would be naturally caught unaware if someone visits her a week before polling day and asks about her voting decision. Similarly, low-information voters may not actively discuss their vote choice, even in personal interactions. This reluctance may be attributed to self-consciousness about being able to coherently articulate why they voted for whom they voted. Commentators have been inaccurately overestimating the extent of silent voting by ignoring the possibility of late decision-making. And, as research has shown, late deciders are more likely to follow the bandwagon effect, i.e., vote for the party perceived to be winning.

A silent voter could also vote for BJP

Do silent voters always vote against the BJP? The general belief among observers is that the silent vote is most often against parties like the BJP, which is not the most preferred choice among the marginalised sections of India.

Research from other parts of the world suggests that there is silent support for parties that represent the dominant ethnic group. For example, the Bradley effect is a very common term in American election literature that describes a scenario where pollsters consistently fail to accurately predict races between white and non-white candidates.

The research shows that due to a social desirability bias, white respondents would hide the fact that they did not vote for a non-white candidate. Similarly, the phenomenon of ‘shy Tories’ is also prevalent in the United Kingdom. According to it, a lot of conservative party voters do not want to publicly admit their voting preferences because it may not be socially acceptable among friends and neighbours who have liberal leanings.

The possibility of a silent BJP voter in the post-2014 scenario cannot be ruled out. Since 2014, the BJP has been successful in attracting a large segment of Dalit votes). Further, according to National Election Studies (NES) surveys conducted by Lokniti-CSDS, a section of Muslim voters – approximately 8-10 per cent – have voted for the BJP. Due to a general distrust of the BJP among these communities and incidences of both caste and communal violence over the last five years, some voters from these social groups may be averse to publicly accepting that they vote for the BJP. Social pressure from within their own community may force some to refrain from publicly endorsing the BJP.

Silent voter and exit polls

Do exit polls always fail to account for silent voters? These polls do go wrong, and that too very often, but the reason is certainly not the silent voter. Election surveys that follow methodological protocols account for factors that can bias their estimates. They use in-built mechanisms in their survey instrument to reassure respondents about maintaining the confidentiality of the response. Further, many surveys purposely avoid orally eliciting the response on vote choice and employ dummy ballot papers and boxes or a replica of an EVM interface on tablets.

Rigorously conducted surveys also take into account the exact scenario under which an interview was conducted. For instance, the Lokniti-CSDS surveys record whether the respondents were alone while the interview was being conducted and whether the respondent was hesitant in answering some questions. This helps in controlling for the element of fear, social desirability bias, peer pressure, among other such biases.

In conclusion, the silent vote is a much more complex phenomenon than observers have claimed it to be. And pollsters are not unaware of it. It operates in a variety of ways and is highly context dependent. Very often, in a Lok Sabha election, these varied types of silent votes cancel the effect of each other and the net result is very close to the median of exit poll forecasts.

Rahul Verma is a Fellow at the Centre for Policy Research (CPR), Delhi. Pranav Gupta is a PhD student at the University of California, Berkeley.