How I got the 2019 election all wrong

Confessions of an economist who thought he can’t get it wrong.

WrittenBy:Vivek Kaul
Article image

It was close to midnight, sometime during the first week of May. I and a childhood friend were sitting in the AC upper deck of a restaurant near the Metro cinema in South Mumbai and having mutton seekh kebabs.

Or rather, I was having mutton seeks kebabs and he was watching his weight.

“So how many seats do you think the BJP will win?” he asked nonchalantly.

“Around 210,” I replied, digging into a kebab.

“How have you come up with such an exact number?”

“Well, it’s not an exact number, which is why I said around 210.”

“What if you turn out to be wrong?” he persisted.

“I can’t go wrong,” I replied with total confidence.

This was one of the many conversations I had in the run up to the election. I maintained this confidence, the few times I appeared on TV as well.

Of course, a few weeks later when the results came out, I had got it wrong big time, and so had many others. Some of them happen to be veteran political observers, and who are now busy writing pieces headlined 12 Reasons Why Modi Won.

So, what was the logic behind the 210 number? The logic, as was the case with me and many others who predicted that the BJP won’t get a majority on its own, was fairly simple.

We looked at the numbers of the last Lok Sabha elections. The BJP had won a bulk of its 282 seats in Northern, Western and parts of Eastern India. Take a look at this table, which lists the total number of seats won by the BJP in 10 states and the National Capital Territory (NCT) of Delhi in the 2014 Lok Sabha elections.

imageby :

Source: Election Commission of India.

The BJP won 247 seats in 10 states and the NCT of Delhi. In fact, in many states, it won a bulk of the seats. In Gujarat, Rajasthan and the NCT of Delhi, it won all the seats. In Madhya Pradesh, it won 27 out of the 29 seats that the state elects to the Lok Sabha. In Chhattisgarh, it won 10 out of 11 seats. In Jharkhand, it won 12 out of 14 seats. Of course, the big one was Uttar Pradesh, where the party won 71 out of the 80 seats that the state sends to the Lok Sabha. Its ally, the Apna Dal, won two seats as well. In Maharashtra, the BJP was in alliance with primarily the Shiv Sena.

Hence, in that sense, there was no scope for improvement in most of these states and the performance could only go down from here. Mathematicians call this “regression to the mean”, which basically states that extreme outcomes are followed by moderate ones. 2014 was an extreme outcome with Modi and the BJP more or less sweeping 10 states and the NCT. Repeating this would be next to impossible. The party could only go down from here, or so I had come to believe.

There were other points which I had taken into consideration as well. Let’s take a look at them point-wise.

1) The BJP had won 71 seats in Uttar Pradesh, the last time around. In 2014, the Samajwadi Party and the Bahujan Samaj Party, the BJP’s two major opponents in the state, had fought elections separately. This had divided the votes, leading to the SP winning five seats in the state and the BSP drawing a blank.
The first past the post system had been the undoing of both the parties. In this system, the candidate who wins the most votes, despite a bulk of the votes being against him or her, turns out to be the winner.

The BJP benefited tremendously from this in Uttar Pradesh in 2014. It won 42.6 per cent of the votes polled in the state whereas the SP and the BSP got 22.3 per cent and 19.8 per cent of the votes polled, respectively. Hence, the total votes polled in favour of the SP and the BSP stood at 42.1 per cent. There wasn’t much of a difference between the vote shares of the BJP and the SP plus BSP. Given that the SP’s vote share was more concentrated in certain parts of the state, it ended up winning five seats. The BSP’s vote share was spread out throughout the state and hence, it ended up without any representation in the Lok Sabha, from the state.

Of course, both the parties understood this as well. They put their differences aside and entered into an alliance. The question was would the support base of the two parties, the Yadavs for the SP and the Jatavs for the BSP, come together on the ground, or will the caste differences make the alliance ineffective.

2) In fact, the support base of the Yadavs and the Jatavs had come together in bye-elections for two Lok Sabha seats, Gorakhpur and Phulpur, which had happened in March 2018. The SP candidates supported by the BSP won 49.31 per cent and 47.12 per cent of the votes polled and won the elections. There was evidence of the alliance having worked on the ground as well.

In fact, along with this data, I had also happened to read Prannoy Roy and Dorab Sopariwala’s book The Verdict: Decoding India’s Elections, which was released before the Lok Sabha elections. In this book, the veteran election observers write: “Our tentative [emphasis added] finding always has been that alliances do indeed work: the votes of the two (or more parties) in the alliance are indeed additive when they form an alliance. Indeed, we have found that alliances seem to create an added momentum—which means that often the total votes of the alliance are greater than the sum of the parts.”

This is something that had already played out in Gorakhpur and Phulpur bye-elections. Given this, there was no way the BJP could repeat its performance of 71 seats in the 2019 Lok Sabha elections. In fact, over and above the SP and BJP, Ajit Singh’s Rashtriya Lok Dal also joined the alliance. In this situation, it was more than likely that the BJP would end up with 30-35 seats in the state. Given the data available, it was a reasonable assumption to make. Or so it seemed at that point of time.

3) The logic of Uttar Pradesh would apply to Karnataka as well, where the Congress and the Janata Dal (Secular) came together to form an alliance to fight the BJP. In the 2014 Lok Sabha election, the vote share of the BJP was 43.4 per cent, and that of the Congress and the JD(S) was 41.2 per cent and 11.1 per cent, respectively. Clearly, the Congress and JD(S) had polled more votes than the BJP, but given that they were not in alliance, the first past the post system had benefitted the BJP. So, clearly, the BJP couldn’t repeat its performance of winning 17 seats in the state.

4) Over and above all this, the Congress had won three state polls in Rajasthan, Madhya Pradesh and Chhattisgarh. In Rajasthan and Madhya Pradesh, the Congress and the BJP had had a neck-to-neck fight. In fact, in Madhya Pradesh, the vote share of the BJP was slightly higher at 41 per cent against 40.9 per cent for the Congress. In Rajasthan, Congress had got 39.3 per cent of the vote against BJP’s 38.8 per cent. In Chhattisgarh, the Congress had got 43 per cent of the votes polled against BJP’s 33 per cent.

These elections were held in November-December 2018, very close to the Lok Sabha elections. The BJP had swept these states in 2014. Hence, it was a reasonable assumption to make that the performance of the Congress will be better than it was the last time around. Even if the Congress won one-third of the seats, it would make a substantial dent into the BJP’s tally.

5) Along with all this, I was also following many well-respected journalists who were travelling through the large states, particularly Uttar Pradesh and Bihar. I am normally a little sceptical of this exercise, primarily because of the way it is conducted. Going to a place and talking to people in order to figure out which way the hawa is blowing is not a statistically robust exercise. The sample of people a journalist talks to depends on luck on that day and whether that same sample represents the underlying population—in order to catch the right hawa—is totally debatable. Also, whether people are answering honestly is not known. During the course of the elections I got talking to a Muslim cab driver in Mumbai, where I live. He gave me one answer, until he realised that a friend who was also with me in the cab was a Muslim. He then gave me the opposite answer.

The journalists themselves put in a lot of hard work and feel good about it. Over and above it, this makes for good viewing on TV and good reading in a newspaper.

Anyway, the journalists who were traveling through states in order to get an idea of the hawa were largely of the opinion that there was no hawa in favour of Narendra Modi, though some of them admitted to there being an undercurrent (whatever that meant) in favour of Modi.

Basically, all these factors went into my calculation, and I concluded that the BJP will win anywhere between two-thirds and three-fourths of the seats that it had last won. Three-fourth of 282 worked out to 212 seats, which I rounded off to 210, when I answered my friend’s question.

Of course, when the exit polls came and second, when the actual results came, all this logic turned out to be wrong. In fact, the BJP won 242 seats in the 10 states and the NCT that I had considered earlier, more or less maintaining its stupendous performance. It was down by five seats primarily because it had fought on fewer seats in Bihar than last time. It also improved its performance in West Bengal by winning 18 seats against two seats it had won in 2014. In Odisha, it won eight seats against the one seat it had won in 2014. And in Karnataka, it won 25 seats against the 17 seats last time around (again, something the journalists did not catch on to). The regression to the mean continued to remain a theory.

Nearly 10 days later, I have come to the conclusion I became a victim of what behavioural economists call the confirmation bias. As Richard Thaler, who won the Nobel Prize in economics in 2017, writes in Misbehaving: The Making of Behavioural Economics: “People have a natural tendency to search for confirming rather than disconfirming evidence … This tendency is called the confirmation bias … People become overconfident because … they only look for evidence that confirms their preconceived hypotheses.”

This is precisely what happened in my case. I had a hypothesis that the BJP will win around 210 seats and I went looking for evidence that would be in line with my hypothesis. A classic example of the confirmation bias that Thaler talks about.

One of the things that I did despite my scepticism was follow journalists doing ground-level journalism, both on Twitter and on TV. This became a very important input into my calculation because they were saying things I wanted to hear. And it turned out all wrong. This leads to the question where these journalists also looking for information which would be in line with their bias, which was that the BJP and Modi were on a weaker wicket this time around, in comparison to 2014.

The Election Commission does not allow opinion polls during the time when the election is on. It believes that the polls influence voters. Hence, during the time the election is on, and given that it is a long period of time, the only way to figure out what voters are thinking is to follow journalists and what they are saying.

The other question is what other input should I have looked at. Some of my friends who are close to the BJP told me that after the results were out, all that I was saying would turn out to be wrong. But then how could I consider their views as inputs, given that they were close to the party?

Of course, there were media houses and journalists who had maintained from day one that Modi and the BJP are going to win. The trouble here was that these media houses and journalists batted for Modi 24/7. So, how seriously could one take their views?

The flip side to this that has clearly emerged is that the journalists claiming to do neutral, ground-level journalism have their own set of biases, which clearly which does not help. Also, I want to ask them: if this was an undercurrent, what would a current look like?

To conclude, the moral of the story for me is that in 2024, I will stay away from making any projections, as far as possible. As far as my friend is concerned, he did send me a message after the election results came out, reminding me of my theory, and all I could do was send a sheepish smiley, in reply.


We take comments from subscribers only!  Subscribe now to post comments! 
Already a subscriber?  Login

You may also like