The trends have solidified and they’ve blown past what almost every exit pollster dared to predict in Bihar. The NDA is cruising to an easy majority, tracking 201 seats at the time of writing this report. The Mahagathbandhan is marooned at 36. Exit polls broadly called an NDA win, but they lowballed the scale of the sweep. And some, especially Axis, the most trusted name in the business, got it spectacularly wrong.

Axis My India delivered another off-target forecast in years. It projected 121–141 seats for the NDA and 98–118 for the MGB, nowhere near the real gap. Its party-wise projections weren’t much better, barring a few exceptions, even though it did correctly anticipate the broad vote-share pattern. 

The actual numbers tell the story: BJP leads in 90 seats (20.24 percent vote share), JD(U) in 84 (19.18 percent), RJD in 25 (22.87 percent), LJP (Ram Vilas) in 19 (5.04 percent), AIMIM in 5 (1.9 percent), and Congress in 6 (8.76 percent). 

Axis’s breakdown had the BJP at 50–56 seats with 18 percent vote share, JD(U) at 56–62 with 18 percent, RJD at 67–76 with 24 percent, and the Congress at 17–21 with 10 percent. It did catch the LJP (RV) surge, predicting 11–16 seats. But with the NDA crossing 200, the numbers simply don’t add up. It's instructive to wait for granular caste, gender, and age demographic data to do a full accounting of their analysis.

Projecting a landslide for the NDA, the little-known Poll Diary is the only polling agency to get close to the overall numbers on the ground. It projected 184-209 for the NDA, followed by 32-49 for the MGB, and 1-5 for others. It also came reasonably close to getting the vote share right. It projected a vote share of 19.83 percent for the BJP, 20.37 percent for the JD(U), 21.87 percent for the RJD, 6.42 percent for the Congress, and 8.68 percent for Jan Suraaj. On X, they even posted a seat-by-seat projection of the results after both phases of polling. The agency doesn’t seem to have a functional website, and there is little publicly available information about it.  

Today’s Chanakya was a little more conservative, but closer than most, projecting a comfortable majority for the NDA. They projected 160 ± 12 seats for the NDA, 77 ± 12 seats for the MGB, and 6 ± 3 for others. 

In terms of vote share, it projected 44 percent ± 3 percent for the NDA, 38 percent ± 3 percent for the MGB, and 18 percent ± 3 percent for others. Chanakya maintains a 3 per cent margin of error in its estimates. Meanwhile, the likes of Matrize, Dainik Bhaskar and TIF Research projected majorities for the NDA, but were conservative in their seat projections. 

What all pollsters got right was that the Jan Suraaj was a non-entity seat-wise. 

Vote share matters, not seat count

A key problem emerges when matching the final results and exit poll data in these elections. Although the likes of Axis hit the vote-share sweet spot, their vote share didn’t translate into seat predictions. 

Similarly, Chanakya, Matrize, and the rest also didn’t hit the mark seats-wise. But Matrize did hit the mark while projecting vote share for individual parties.     

It’s smarter to prioritise the accuracy of vote-share estimates over seat projections. If a pollster gets the vote share right, it has done at least one essential part of the job, even if its seat projections are off. But a seat projection based on an inaccurate vote share is unreliable, because it reflects weak sampling rather than meaningful insight. 

In this regard, Axis, Poll Diary, and Matrize got their vote share projections right within a reasonable margin of error. Chanakya, meanwhile, didn’t publish their party-wise projections for seats and vote share.

What did the exit polls project in 2020 and 2015?

Although exit polls have largely projected the results of the Bihar assembly elections in favour of the NDA, the state’s electoral history suggests these projections have often been inaccurate. In 2020, nearly every major pollster predicted a slight advantage for the RJD-led MGB. 

At the time, Axis My India projected a landslide 139-161 seats for the MGB, barely 69-91 for the NDA, and 3-5 seats for the LJP, which contested the elections alone. C-Voter projected 120 seats for the MGB, 116 seats for the NDA, and 1 for the LJP. Chanakya, meanwhile, projected 180 seats for the MGB, 55 for the NDA, and 0 for the LJP. 

The final results, however, told a different story. The NDA won 125 seats, the MGB finished second with 110 seats, and the LJP with 1 seat. The poll agencies that projected well within the margin of error include: Dainik Bhaskar (120 for the NDA, 116 for the MGB, and 2 for the LJP), P-Marq (123-135 for the NDA, 104-115 for the MGB, and 0-1 for the LJP), and C-Voter (104-128 for the NDA, 108-131 for the MGB, and 1-3 for the LJP).

Rewind to 2015, and the difference between exit polls and the final results was even more glaring. C-Voter projected 112-132 for the MGB, 101-121 for the NDA, and 6-4 for others. Today's Chanakya projected 83 for the MGB, 155 for the NDA, and 5 for others. Axis projected 176 seats for the MGB, 64 for the NDA, and 3 for others. The final results were: 178 for the MGB, 58 for the NDA, and 7 for others. Axis was the only pollster to come close, while the rest were way off the mark.        

After the Lok Sabha poll results last year,  Axis My India MD Pradeep Gupta broke down after being grilled during a live show on India Today. That was a day after he had boasted about Axis My India’s track record. Of the 69 elections his agency polled over a decade, he had got just four wrong – that score is now higher.

You can watch our panel’s analysis of the Bihar elections here.

Support our Bihar election coverage here.