Assembly Elections 2022 | Why you should take Exit Polls with a pinch of salt

Scientific predictions and modelling are prone to error and sometimes ought to give unexpected outcomes than predicted. (Photo by Pavel Danilyuk from Pexels)

SHAMIK VATSA & NISHANT RANJAN

Election forecasting is a difficult task because it involves an extremely complicated process.

Numerous agencies in recent years have tried their hand at them, some claim to have succeeded while several failed. These complications are not just limited to the logistics and finances involved in conducting these large-scale surveys but also the survey research method adopted by these agencies that form the core of the problem.

This is not to say that some of these methodological problems can’t be solved, the point is that the process of making any forecast in social sciences, in general, is a difficult task to accomplish.

Read also: Special coverage of Assembly Elections 2022

Nevertheless, polling agencies and posters try to do it with whatever resources and expertise that they have got and it takes a lot of hard work on their part to do it. Therefore, some hits and misses become inevitable.

Let us look at Uttar Pradesh’s exit polls of 2022, for instance. If we take the median seats projected for various parties in UP given by all polling agencies and find the mean for each party, the following picture comes up:

The incumbent Bharatiya Janata Party appears to be maintaining a huge upper hand over all its contenders. We have reasons to believe that this might not be the case when the final verdict is announced.

In all fairness, all polling agencies warn their readers to take these forecasts with a pinch of salt because they know that a forecast is different from a verdict.

A comparison of seat share with vote share data doesn’t show such a rosy picture.

Let us understand this with the help of an example. For this, let’s take Axis My India’s exit poll data released for the 2022 UP elections. The median vote share and seat share data for all parties in the state according to Axis My India is the following:

Again, a clear mandate for the incumbent but the devil lies in the vote share data, especially when we segregate it to region-wide distribution. Compare the Vote Share of parties in the Rohilkhand (52 Seats) and Bundelkhand (19 Seats) regions for instance.

The picture looks complicated because while the Samajwadi Party alliance has a greater vote share in this region, in terms of seats, the BJP is way ahead of all the other parties. A similar contradiction can be seen in the case of Bundelkhand too:

What’s striking in the two charts above is that while SP and BJP seem to be contesting on a similar footing in terms of vote share, the BJP is getting a large number of seats.

The only possible explanation for this is that survey data might show an extremely narrow margin of victory for the BJP over other parties (particularly the SP). Now, this creates a problem because while this might just be the result of the upcoming elections the overall sample size is so restrictive (only 0.6 per cent of the population) that by extrapolating the data obtained from here with such narrow margins of victory may end up reversed in the actual verdict.

We don’t mean to argue that these hurdles are only possible with the Axis My India’s poll data. This problem is present across the board and perhaps our only humble suggestion is that polling agencies should be cautious before chest thumping their numbers on television.

The Problems of Sampling

As Neelanjan Sircar points out, one of the major problems with doing election surveys in a country like ours is that the intent of representative sampling is difficult to actuate and this is not the fault of polling agencies.

The other hurdle that election surveys present is what Timur Kuran calls ‘preference falsification’ which is an act of misrepresenting or concealing what one thinks under the perceived social pressures. Kuran’s theory and its relation to the elucidation of social and political change are that the occurrence of preference falsification means that there are always people who may privately disagree with public choices, but may not voice their differences.

No doubt these things are ubiquitous and carry huge social and political consequences. Not to suggest that things like these have happened in these exit polls but the possibility of such happenings, provided we know nothing about the training process the researchers and enumerators go through.

The Silver Lining

Recently, reams and reams of pages are being spent to ascertain how women as voters are shaping electoral outcomes in different elections. West Bengal, Bihar, and Delhi state assembly elections are a case in point.

Notwithstanding the social prejudices and domestic constraints, they are voting in large numbers and their vote share is increasing significantly during some of the last elections. However, women as voters are not as vocal as their male counterparts and are known for bringing the change silently. And there are pieces of evidence to show that women voters are agents of change and they vote differently from men.

In the regions of the Hindi heartland, which is still dominated by feudal and patriarchal norms, their silence is taken for granted. But we believe that their silence makes it difficult for political analysts to gauge the reality on the ground and predict the wind of change. And indeed, the silence of women voters demands more gender-sensitive electoral frameworks. These are the invisible variables that can affect poll forecasting.

Predicting poll outcomes is more of a science than simply setting the social calculus of caste, religion, and charisma of leaders right. Scientific predictions and modelling are prone to error and sometimes ought to give unexpected outcomes than predicted.

We are neither denying the validity of poll predictions nor discrediting their helpfulness in shaping our understanding of electoral and democratic processes. However, we are pointing out the limitations of these poll predictions and making our point not to take them for granted exactly what they predict.

(The authors are associated with the Centre for Policy Research, Delhi. Views expressed are personal)