Nation wants to know
Psephology. Also referred to as Fekology, and for a good reason. For months, a psephologist will talk about trends and swings, about Jats and Sikhs, about vote shares and caste factors and they will tell you who will win which election with what margin. And for a few months after the elections, more often than not, they will tell you why they got the numbers wrong. In rare cases, they get to boast about getting it right. One thing consistent about all Psephologists is that they are high on aesthetics and creativity. They create pretty charts and graphics coloured saffron and green. The animated charts speak to us eloquently. So much so, that we get totally engrossed and do not care if the predictions are going to go wrong again. Modi is going to win anyway, so why should we bother if his defeat is imagined by pretty images quoting an OBC voter or a past prehistoric trend.
When Shahrukh Khan sings ” Chaand Taare Tor Laoon”, we don’t believe him, right? We just enjoy the song. So also with this new age entertainment industry called Psephology. Watch, listen and enjoy without taking anything seriously. You can never go wrong watching someone else go wrong.
Meet Jai Mrug. He’s an IITB alum, specialises in data analytics and more importantly, he’s a Psephologist. You would have seen and heard him on the Times Now channel’s election coverage debates. Yes, in debates “moderated” by an extreme Arnab Goswami of the “nation wants to know” fame. Jai’s singular distinction is that he is perhaps the only person who has not been interrupted by Arnab who is loud enough to be heard in your neighbours’ house when their TV is on the blink. Because Jai gets his numbers quite right and narrates them with the help of something called the Spectrum graph. Speaks mellifluously and knowledgeably. Looks cute, fair and chubby. Smiles even if his numbers go a bit wrong. In this piece, Jai expounds on the perils of Psephology and shares tricks on making reasonably accurate predictions, never mind the fact that you like them going wrong. We must confess that this insightful piece was procured after a lot of coaxing, cajoling and arm twisting. What worked finally is the threat that we would lock him up with Arnab Goswami if he didn’t turn in his piece.
It all started in the year 1997. A leading Psephologist of India featured on the cover story of a now defunct magazine called Sunday. That triggered off the Mr. Curious in me and I thought Psephology was a journey worth exploring.
To most Indians the success or failure of Psephology would be judged simply by whether you got the final outcome right or wrong. In a multi-party democracy and a first-past-the-post system like ours, there are several factors that can modify the final outcome and these define how uniquely psephology should operate in India.
In part this is about getting the sample right. Your sample needs to be random stratified, with sample sizes and outcomes prepared by those who have their feet on the ground.
In another part this is about getting the forecasting model right, many call this the conversion model or also the smoothening model.
The success of psephological agencies has swung like a pendulum, often erring completely and often being frighteningly close to the real outcome. There are two elections in particular that are interesting to note in this regard. In the 2004 elections practically every agency got it wrong. Almost all predicted an NDA win and the final outcome was UPA going ahead of NDA in terms of both votes and seats.
2014 was a contrast. Most polling agencies were conservative about the performance of BJP. One agency however stuck its neck out, saying that BJP alone would win 291 seats. It turned out that the agency was the closest with BJP winning 284 seats. Most others gave NDA, inclusive of BJP, a forecast close to 272.
To most Indians the success or failure of Psephology would be judged simply by whether you got the final outcome right or wrong.
To understand how Psephology plays out in the Indian scenario, it is important to understand the differences between the two elections and how the mandates were defined. Two factors contributed extensively to the NDA debacle of 2004. One was the phenomenon of the IOU (Index of Opposition Unity), the other was the phenomenon of a splintered polity in the first-past-the-post system. An example of how IOU could benefit a party/ alliance is how the former UPA managed to rope in many smaller allies who helped UPA make it a one-vs-one contest and that lead to a greater seat share for UPA in that election.
Several states bear testimony to this performance of UPA. Bihar, Andhra Pradesh, Jharkhand and Tamil Nadu in particular. Here the UPA alliance polarized almost all the non NDA vote in its favour. These four states gave a massive lead to the UPA, 36 in Andhra Pradesh, 29 in Bihar, 13 in Jharkhand, and all 39 in Tamil Nadu. More than half the tally of UPA came from these states in 2004. In these very states in 1999, NDA had won 103 seats.
A united alliance or a group of parties has the ability to swing last minute votes in its favour and as the marginal vote across seats gets added, it crosses a threshold where the vote share of the opponent ceases to matter. This is precisely what happened in the year 2004. In the states mentioned above, the vote share of BJP/NDA ceased to matter and the opposition had a cake walk. The IOU and last-minute swings have a great potential to upset the apple cart, and this is precisely what happened.
Another phenomenon that is worth noting is a quirk of fate of the first-past-the-post phenomenon which rewards in excess the performer and penalises in excess the under performer in a splintered polity. The vote to seat ratio can be used as a very effective indicator to understand how the vote distribution impacts the seat share of various parties in such a scenario. One of the most important indicators of this distribution is the vote to seat ratio of each of the parties.
The success of psephological agencies has swung like a pendulum, often erring completely and often being frighteningly close to the real outcome.
Uttar Pradesh is a classic example of what can happen when a party fails to cross a particular threshold in terms of its vote share. In the mid and late nineties, BJP benefitted the most from the first-past-the-post system, winning a higher seat share for relatively lesser number of votes, thus the parties vote to seat ratio was relatively lower. Gradually towards 2004, the parties vote share decreased, but the first-past-the-post system heavily penalised the party, reducing it to a marginal player.
In 1998 BJP won one seat for every 3,65,000 votes it polled. This was its lowest vote to seat ratio meaning that it was its most effective conversion, with the party winning 52 of the 80 seats at stake. On the other hand, SP won one seat for every 6,61,000 votes it polled. BSP won one seat for every 30,50,661 votes it polled. In 1999,BJP won one seat for every 5,17,000 votes while SP played catch up winning one seat for every 5,03,000 votes and BSP won one seat for every 8,57,000 votes.
This led to a more level playing field with no party posting a single digit seat tally and each posting a respectable performance. In 2004 the situation completely reversed. BJP won one seat for every 11,81,000 votes while SP won one seat for every 4,06,000 votes and BSP won one seat for every 6,91,000 votes. In 2004 the first-past-the-post system penalised BJP heavily and the party won just 10 seats in the state in this election.
Psephologically one of the biggest challenges is to be able to find out confidently how the vote share differential of each of the parties is spread across the state, and develop an ability to convert that differential into seats. The 2014 election saw BJP swing the state of Uttar Pradesh massively winning 71 out of 80 seats. Again it was a very favourable vote to seat conversion that gave the party the massive mandate.
A third factor that is becoming increasingly clear in Indian politics is the gubernatorial nature of elections in the country, where the winner takes almost all the marginal vote and leaves the opposition with bare nothings. Rajasthan and Madhya Pradesh in 2004 and 2014, and Gujarat in 2014 displayed the gubernatorial trend, where in a two-party contest the loser polled close to a zero or a single digit tally.
Psephologists must clearly understand how to translate a vote share into a decisive seat share, especially understand how a threshold vote share in a two-party system gives a one-way mandate to the winning party. It is important to understand the social spread of the mandate as much as the political spread. Only such an understanding will help comprehend landslides.
The mandate in the February 2015 Delhi elections where AAP took all is an indicator of how effectively the social spread of a party can impact its political fortunes, and produce a one-way result.
An understanding of the new aspirations of a fast-urbanising India, the differing political landscape of every state, along with the vagaries of the first-past-the-post system can help develop a nuanced model that may help forecast outcomes of elections in India.