- NL Sena
Some hits, many misses.
Assembly elections concluded in five Indian states yesterday. Counting finished in Rajasthan, Madhya Pradesh, Chhattisgarh, Telangana and Mizoram after a long day of boisterous spokespersons, hyperactive anchors, and keyboard-shattering twitterati.
But even before panellists and politicians had started their war of words, it was those with numbers who had their final say. Last week, polling companies had paired with news channels to put out their exit poll data on the five states. The most visible of these pairs were Republic TV-C-Voter, Republic TV-Jan Ki Baat, Times Now-CNX, ABP News-CSDS, India Today-Axis My India and India TV-CNX.
Exit polls usually involve two important steps of calculation. The pollsters have to first come up with a figure for the vote share among the electoral players. The second and the more difficult step is to convert this share into a given number of seats using complex quantitative models which account for a diverse range of factors.
In the latest round of state elections, most of the exit polls obediently lived up to their reputation of being inaccurate. While some like Republic TV had the luxury of producing two exit polls with opposite projections (a rare show of balance on the channel), others like Times Now-CNX, ABP-CSDS, and India TV-CNX predicted figures that considerably missed the magic numbers, even when enclosed in ranges. A decent exception was the India Today-Axis My India exit poll, whose scoresheet boasts of more hits than misses.
None of the exit polls captured the final tally in Rajasthan with great success. Republic TV-Jan Ki Baat, Times Now-CNX and India TV-CNX overstated their predictions, claiming that BJP would secure 80 or more seats in the elections. Numbers put out by Republic TV-C Voter and India Today-Axis My India polls came close to BJP’s actual figures of 73, but missed the mark by putting Indian National Congress’s (INC) seats above 120.
In the North-eastern state of Mizoram, the Mizo National Front (MNF) won the race with a simple majority of 26 seats, leaving Congress and BJP far behind with 5 and 1 seat respectively. Both Times Now-CNX and Republic-C Voter exit polls broadly predicted this, but they significantly underestimated the MNF’s strength in the state. Whereas Times Now-CNX predicted 18 seats for MNF, Republic TV-CNX put a range of 16-20.
Predictions were way off the mark in Telangana. Republic TV’s C Voter and Jan Ki Baat produced an aggregate range of 50-60 seats for the Telangana Rashtra Samithi (TRS) and 40-60 for the INC. Times Now-CNX gave crisp numbers — 66 for TRS and 37 for INC. An impressive tally, however, was put forward by the Secunderabad-based Centre for Psephological Studies (CPS), whose numbers were telecasted on the regional Telugu channel TV9. The CPS poll correctly predicted 84-89 seats for TRS and 19-21 for the Congress-Chandrababu Naidu combine. Interestingly, on December 4, INC’s Uttar Kumar Reddy had blasted TV9 for broadcasting a range of 16-21 seats for his party, calling it ‘baseless’.
Perhaps the biggest upset for the exit polls was the state of Chhattisgarh. In a state where the BJP pocketed no more than 15 seats, Republic TV’s Jan Ki Baat and C Voter polls announced 40-48 and 35-43 seats, respectively. At par with these were numbers put out by Times Now-CNX and ABP-CSDS exit polls, both fixing their predictions above 45. India Today-Axis My India exit poll came a little close by projecting 21-31 seats for the BJP and 55-65 for the INC.
Madhya Pradesh was a saving grace for the polling gurus. Acutely aware that it would be a tough call between the two national parties, most polls presented no radical gaps between their BJP and Congress numbers. Republic TV played it safe by letting one of its exit polls predict a majority for BJP (Jan ki Baat) and the other for the Congress (C Voter). Times Now-CNX proposed a distribution of 126 and 89 seats for the BJP and the Congress respectively, whereas ABP-CSDS predicted the opposite — 126 for Congress and 94 for the BJP. The India Today-Axis My India exit poll captured the numbers well, putting out an expansive range of 102-120 for the BJP and 104-122 for the INC. The final numbers came out this morning with Congress emerging as the single-largest party (114) and BJP bagging 109 seats.
This was not the first time that the pollsters overshot the mark and backed the wrong horses. In 2015, Assembly election results in Bihar caught most pollsters off guard when a non-BJP coalition swept the state, contrary to their predictions. In Delhi in the same year, AAP secured an absolute majority by winning 67 out of 70 seats — a bolt out of the blue for many, especially the professional forecasters. In an electorate that is as huge and diverse as India’s, methodology and data sampling is key to a successful forecasting model. One such model is perhaps yet to arrive.