The electoral battle in Haryana was said to be done and dusted with every exit poll on 21 October predicting a clear win for the Manohar Khattar-led Bharatiya Janata Party in the state.
Three of the polls – by Times Now, ABP-CVoter and News18-Ipsos – even went to the extent of predicting a three-fourth majority for the BJP.
However, polling agency Axis and India Today withheld their exit poll for a day and released their findings on 22 October. They have predicted a hung Assembly in the state with the BJP and Congress’ seat tally roughly in the same range.
What added to the suspense is the fact that Axis My India chief Pradeep Gupta chose not to release his predictions on 21 October and instead decided to go back to the field to verify his data.
Gupta is known to stick his neck out with brave predictions. In the 2019 Lok Sabha elections, he was one of the few pollsters to predict BJP crossing 300 seats. He was proved to be correct.
Now, the question is: Why has Haryana thrown up such hugely different predictions?
When Exit Polls Go Wrong
There are a number of ways exit polls can go wrong:
1. The polls get the seat share right but not the vote share, or vice versa.
This is perhaps the most forgivable error for a poll, especially if it roughly gets the vote share right. Conversion to seats is always a complicated exercise and could go wrong.
2. The polls rightly predict who is winning but get the extent of the win completely wrong.
This is the most common mistake polls make. If polls follow a correct methodology and have a representative sample, then it is highly unlikely they would get the drift of the result wrong. Often polls underestimate the extent of a party’s victory. These kind of errors could be seen in most exit polls for the 2019 Lok Sabha elections, the 2015 Delhi elections and 2018 Chhattisgarh elections.
3. The polls go wrong even on who is winning.
This is the biggest blunder for any exit or opinion poll and could reflect that there was something wrong with its methodology or sampling. This could be seen in the 2004 Lok Sabha elections, when most polls predicted an edge for the NDA but in the end the UPA gained a sizable lead.
Similarly, in the 2015 Bihar elections and 2017 Uttar Pradesh elections, many polls predicted a hung assembly but the Mahagathbandhan and BJP respectively won both states comfortably.
The wide variance in the Haryana predictions means that either India Today-Axis My India or the four other polls might end up falling in the third category.
Why the Two Contrasting Predictions?
To understand why the exit polls in Haryana gave such contrasting results and get a sense of what the eventual results might be, we need to look at the predictions with respect to the 2019 Lok Sabha elections and the 2014 Assembly elections.
Projected Change from 2019 Lok Sabha Vote Share
The gap between Lok Sabha and Assembly elections in Haryana has been around 5-6 months, except in 2005, when the Assembly polls were held 10 months after the Lok Sabha polls mainly due to the fact that Prime Minister Atal Bihari Vajpayee advanced the 2004 general elections.
All the Assembly polls since 2000 have shown a few clear patterns:
- The vote share of BJP and Congress has always fallen in Assembly polls, compared to the Lok Sabha polls held a few months before.
- The BSP, like BJP and Congress, tended to do better in national elections than Assembly polls. For instance, its vote share fell by nine percentage points between the Lok Sabha and Assembly polls in 2009.
- Almost always, it has been the INLD and regional parties and independents that have gained at the expense of national parties.
Now, if we look at the vote projections by two of the more reliable polling agencies – CVoter and Axis – and compare them to the parties’ vote share in the 2019 Lok Sabha polls, here’s what we find:
- Both CVoter and Axis predict a major fall in BJP’s vote share compared to the Lok Sabha polls. CVoter predicts a fall of 15.8 percentage points, while Axis predicts a fall of 25.2 percentage points.
- The two agencies differ on the Congress’ fortunes. CVoter predicts a fall of 2.4 percentage points while Axis predicts an increase of 3.6 percentage points.
- Both agencies predict an increase in vote share of “Others”: 18.2 percentage points by CVoter and 21.6 percentage points by Axis.
- There is also a difference on who is gaining within others. CVoter predicts that Dushyant Chautala’s JJP will go from 4.9 percent votes in the Lok Sabha polls to 19 percent in the Assembly polls. Axis predicts a vote share of 14 percent for JJP.
While their seat predictions might be completely different, the two polls aren’t very different in their predictions of which way the votes are swinging compared to the 2019 Lok Sabha elections.
Both polls predict huge losses for BJP in terms of vote share and gains for others, particularly the JJP.
Projected Change from 2014 Assembly Polls
However, the two polls differ on which way votes are swinging compared to the 2014 Assembly polls.
- CVoter predicts a 9.2 percentage points increase in BJP’s vote share and a rise of 5.4 percentage points for the Congress compared to the 2014 Assembly polls. On the other hand, Axis predicts a 11.4 percentage points increase for the Congress, with the BJP remaining where it was.
- The main difference seems to be regarding what is happening to the 46 percent votes (24 percent INLD and 22 percent others) polled by regional parties and independents.
- CVoter’s prediction of 19 percent votes for JJP would mean that Dushyant Chautala’s party is retaining almost three-fourth of the INLD’s erstwhile base (predominantly among Jats). While the remaining appears to be split between the BJP, Congress, independents and what remains of the INLD.
- Axis’ prediction is not very different from the 2014 Assembly poll result, with the BJP and others remaining where they were and the INLD’s 24 percent vote share being split mainly between JJP and Congress. This means that the ‘Modi factor’ might be completely missing.
- Therefore, the main difference between the two surveys is that while CVoter projects BJP eating into both the INLD and others’ vote bank, Axis predicts the Congress eating into these to a much greater extent.
Caste and Community Factor
According to the India Today-Axis My India poll, the BJP’s projected losses compared to the Lok Sabha election are due to “tactical voting by Jats, Dalits and Muslims”.
Speaking to India Today on 22 October, Pradeep Gupta had said that the BJP was incurring losses in around 50 seats where Jats, Muslims and Dalits together are in sizable numbers.
In the 2019 Lok Sabha elections, BJP got the support of almost all communities in Haryana except Muslims.
According to the Lokniti-CSDS survey, its support among Upper Castes and OBCs was above 70 percent. Among Jats and Dalits it was 50 percent and 58 percent respectively.
Its biggest gains compared to 2014 were among these two communities – 39 percentage points among Dalits and 31 points among Jats.
If Axis’ exit poll is correct, Jats and Dalits may have voted for the BJP in the Lok Sabha elections mainly because of Prime Minister Narendra Modi’s popularity as well as national security factor after the Balakot strike and they have possibly drifted away from the BJP as it isn’t Modi but CM Manohar Lal Khattar and local candidates who are on the ballot.
The poll indicates that the ‘Modi factor’ is virtually non-existent and the BJP’s vote share is pretty much where it was in 2019.
The poll also indicates that the Congress has eaten into the INLD’s erstwhile base of Jats and Meo Muslims and also regained many of its own traditional Dalit votes.
The party’s decision to make senior Jat leader Bhupinder Singh Hooda its legislative party leader and senior Dalit leader Kumari Selja its state unit president may have worked well, if the Axis poll results turn out to be true.
Also, the Jat resentment on the reservation issue and Dalit anger on due to rising atrocities as well as the demolition of the Ravidas Temple in Delhi may well have played a role in turning these communities against the BJP.
In addition to the drift of these communities, there might also be the natural shift away from BJP among all communities that is only to be expected between a national and a state election.
(At The Quint, we question everything. Play an active role in shaping our journalism by becoming a member today.)