Bihar Elections: 2 Opinion Polls Predict NDA Win. Can It Go Wrong?

According to CVoter, in 57 constituencies, the leading party’s margin is just 3 percent. NDA leads in 31 of these.

CVoter’s tracker reveals a massive vote difference between BJP & JD(U). This could complicate seat predictions. 

(Kamran Akhter/The Quint)

With the first phase of polling in Bihar due to take place on 28 October, two important opinion polls have released their predictions in the past one week.

CVoter released the data from its latest survey on 24 October and Lokniti CSDS did the same on 20 October. Both surveys have predicted a comfortable victory for the Nitish Kumar-led National Democratic Alliance. Naturally, the opinion polls provoked an angry reaction from Mahagathbandhan supporters, who pointed out the numerous occasions in the past in which opinion and exit polls have gone wrong.

Without going into the merits of the two polls, this article will try and answer the following questions:

• Can polls go wrong?
• What do the opinion polls predict in the Bihar election?
• What is the big picture emerging from these polls?
• What are factors to watch out for in the coming weeks?

Can Polls Go Wrong?

The short answer to this is: Yes, polls go wrong all the time. But there’s need for nuance on this.

There are three ways polls can go wrong:

1. Predicting the overall winner and loser wrong
2. Getting the vote share wrong, beyond margin of error.
3. Getting the overall seat prediction wrong.

The first – getting the overall result in the election wrong – is an absolute disaster and almost unforgivable for a pollster. So there’s no need for elaboration here. It’s the other two aspects that need to be scrutinised more.

Vote Share vs Seat Prediction: Which is More Important?

Much of the public attention is focused on the third aspect – the seat tally prediction.

Predicting seat tally is a very subjective exercise and often goes wrong. Getting this wrong is in fact the most forgivable mistake for a pollster. 

Converting projected vote share into seat prediction is actually a very subjective and inexact exercise and often goes wrong. This is partly because every election is also an aggregation of several local elections.

However, the problem is that since most polls are judged on getting the seat share right, a few pollsters put the horse before the cart. Here’s what they do:

• They invest a great deal of resources in figuring out who is winning each seat and then add them up to try and figure out the final seat tally. Sounds fine, but there’s a catch.
• Sometimes they get the final tally right but this is often the result of predicting a few seats wrongly in favour of one side, a few wrongly in favour of the other but getting a decent result overall due to the two sets of errors cancelling each other out.
• Another problem that this kind of polling faces is that often the sampling done isn't representative in terms of caste, community and gender.

Getting the vote share right is actually more important. If it goes wrong by a large margin, then it would mean that the sampling done in that election was terribly flawed.

Getting the vote share right is actually more important. If it goes wrong, then it would mean that the sampling may have been flawed.

To understand this difference, let's look at one poll – the 2019 Haryana Assembly elections.

• From the point of view of pollsters, this election was intensely discussed because all the exit polls predicted a BJP sweep but India Today-Axis withheld their prediction for a day. They came back the next day and predicted a hung Assembly. When the results came in, they were vindicated as no party won a majority. But let's scratch the surface a bit.
• Axis' predicted vote share was 33 percent for BJP, 32 percent for Congress and 14 percent for JJP.
• Let’s compare it to another agency – CVoter. CVoter's prediction was 42 percent for BJP 26 percent for Congress and 19 percent for JJP.
• The final results were: BJP 37 percent, Congress 28 percent and JJP 15 percent. So Axis got JJP right, CVoter got Congress right and neither of them got BJP right. So Axis may have gotten the tally right, in terms of vote share the two polls were more or less at par.

What Do the CVoter & CSDS Opinion Polls Say About Bihar?

• In terms of vote share, the CVoter survey gives an eight percentage point lead to the NDA while the CSDS survey gives a six percentage point lead. Therefore the gap projected by both is broadly on the same lines.
• The projected vote share for the Mahagathbandhan by the two surveys is in the same range: 35 percent in the CVoter survey and 32 percent in the CSDS one.
• But NDA's projected vote share is quite different: 43 percent in the CVoter survey and 36 percent in the CSDS one.
• The big difference in the two surveys is in the projected vote share for ‘others’. While the CVoter survey gives 23 percent votes to non-NDA, non-Mahagathbandhan parties, the CSDS survey gives 30 percent.
• CVoter gives LJP 4 percent votes and CSDS gives it 6 percent. Since LJP is tapping into the NDA vote mostly, this may explain part of the difference between the two surveys regarding the NDA’s vote share.
• The two polls basically differ in the amount of leakage from NDA to LJP and from Mahagathbandhan to smaller fronts like GDSF and PDA.

CVoter and CSDS predict the same range of votes for the Mahagathbandhan. The main difference is in the predicted vote share of NDA and others. 

• In terms of seat prediction, CVoter gives around 147 seats to the NDA, 87 to Mahagathbandhan, 4 to LJP and 6 to others.
• CSDS gives 138 to NDA, 93 to Mahagathbandhan, 6 to LJP and 9 to others.
• Therefore, CVoter gives 10 seats more to NDA, six seats less to Mahagathbandhan and two seats less to LJP and five less to smaller parties compared to the CSDS poll.
• In terms of CM choice, the gap between NDA’s Nitish Kumar and Mahagathbandhan’s Tejashwi Yadav is wider in the CVoter poll than the CSDS one. According to CVoter, 29 percent respondents picked Nitish and 19 percent picked Tejashwi while in CSDS the data was 31 percent and 27 percent respectively. But if one adds the people who chose Lalu Prasad in the CVoter survey, an option not given in the CSDS survey, Nitish’s lead would be similar.
• Chirag Paswan and Sushil Modi also scored much better in the CVoter poll – 14 percent and 9 percent, compared to 5 percent and 4 percent in the CSDS one respectively.
• Another big difference is that in the CVoter poll 54 percent people picked unemployment as the main poll issue as against 20 percent in CSDS. ‘Development’ is the biggest issue for voters in the CSDS poll, at 29 percent.

What’s the Big Picture?

Despite the differences, few broad trends seem to be common in the two polls:

• NDA has more or less retained the advantage it has due to favourable alliance arithmetic.
• The RJD-led alliance has gained immensely since the Lok Sabha poll debacle. Tejashwi Yadav has managed to consolidate the RJD’s core vote of Yadav and Muslim voters, perhaps even expand a little bit beyond that. However, it hasn’t won over enough shifting voters to upset the NDA’s advantage.
• Despite anti-incumbency, Nitish Kumar’s core support among Kurmis, EBCs and Mahadalits has remained intact, with some leakage.
• The LJP is harming the JD(U), which is evident from the CVoter poll that gives it 14 seats and 3 percentage point votes less than the BJP. But the transfer of BJP votes in JD(U) seats hasn’t collapsed completely despite the LJP’s efforts.
• Barring the LJP, smaller parties could be harming the Mahagathbandhan more than the NDA.

This is what the two surveys seem to be indicating. However there are some X-factors that could still alter equations.

X-Factors to Watch Out For

• CSDS’ poll and an earlier poll by CVoter did indicate that over 20 percent voters are still undecided. This is not unusual and can flip the result. A lot would depend on how parties mobilise in the last bit of campaigning and on the poll day itself.
• According to CVoter 57 out of Bihar’s 243 seats are still within the margin of error – that is where the difference between the leading and second party is three percentage points or less. As of now, NDA leads in 31 out of these 57 seats and Mahagathbandhan in 20. A small swing of 1.5 percent can make these seats turn. If there is such a swing against the NDA then it would fall below the half-way mark.

CVoter says that the margin is less than 3 percent in 57 seats. Even a small swing could turn things in these seats. 

• A swing of 2-3 percent against it could completely reverse the equations. A swing in favour of NDA could lead to an even bigger victory than what’s being predicted.
• Turnout is another major X-factor, especially of BJP voters in JD(U) seats. How COVID-19 impacts the turnout also remains to be seen.
• The LJP factor is still difficult to grasp in surveys because it varies from seat to seat. The party is contesting only about half the seats in the state. It hasn’t put up candidates in most of the seats where BJP is contesting. Therefore, the BJP will have the advantage of committed LJP voters on its side, which its allies like JD(U), HAM and VIP won’t.
• In some seats it is clear that even BJP voters will be voting for LJP candidates and not JD(U) ones – Dinara in Rohtas district is a case in point.
• PM Narendra Modi not criticising the LJP and many of the ads with the PM’s face not featuring Nitish Kumar, has added to this ambiguity about BJP and JD(U)‘s relations. A big X-factor is to what extent BJP is able to transfer its votes.
• Young voters are a big X-factor. A large chunk of young voters are clearly dissatisfied with Nitish Kumar. For many of them, NDA’s “Jungle Raj” barbs against the alleged misgovernance in RJD’s previous tenures, doesn’t resonate as they have no memory of that.