Veteran journalist and psephologist Prannoy Roy recently co-authored a book, ‘The Verdict: Decoding India’s Elections’, with Dorab R. Sopariwala. In this interview, Mr. Roy talks about opinion polls, landslide victories and the problem of missing women voters in India.
You and your team have combined academic rigour with journalistic functions and skills, such as bearing witness to events and investigations, to bring about your work on elections. You are an economist, a chartered accountant, a psephologist and also a journalist. What do you think these roles bring to the table for your journalism?
It is a very important question. We generally tend to mix up these roles. There is a qualitative and quantitative aspect to these roles and these shouldn’t be bridged. For example, when I am doing opinion polls, that’s quantitative work, and as journalists we need to do qualitative work. Many journalists tend to try to forecast the elections, which is not their job, especially considering the fact that ours is a first-past-the-post system where a small percentage change in votes will have a huge change in seats — a 3% change in votes could mean a 100 seats changing hands. For a journalist, it will be difficult to have a gut feel for 3%.
I feel the job of journalism is qualitative; it’s to talk about stories, issues. These cannot be simply translated into numbers by an opinion poll. We saw how the farmers of Uttar Pradesh were suffering, and how their lot was different from what the situation was five years ago, and we can’t put that into numbers. The job of a journalist is to tell a story about an election and to avoid forecasting. And pollsters try to be qualitative too, by asking people to list issues of priority, but this is not as simple. They need to talk to people, understand them, and it takes time as a journalist to understand issues.
Journalists tend to hedge their bets. One of the findings in your book is that most often pollsters get the winner right, but they invariably underestimate the scale of victory in terms of seats for the winners. Why?
Pollsters tend to try to play it safe. It is important for them to get the winner right rather than the seats. They predict, so-and-so is the largest party, but they try to project seats that are a standard deviation lower. For example, if their poll shows someone leading in 280-290 seats, they might lower that number to 250. So, if a party wins 250 seats and is the single largest one, pollsters could claim that they got the winner right. If they get the largest party wrong, it is very tough for them to play it down. It is also because respondents, which is particularly true in this election, try to play it safe. They tend to reaffirm that they support their ruling party as they do not always trust the pollster asking the question. How much is the fear factor among the respondents... this is very tough for pollsters to assess.
Is this lack of trust among respondents or misdirection measurable?
At the moment, it is not. We try to ask, who did you vote for last time? And then we try to see if the respondent is exaggerating about the ‘last time’ question and try to adjust for that. But respondents also tend to say that they voted for the ruling party the last time. There is a huge exaggeration about who they voted for the last time. This is a phenomenon the world over. So, I don’t think there is any technique to estimate this fear factor. However, we are trying to use Artificial Intelligence and Deep Learning to understand this behaviour.
You spoke about the qualitative and quantitative aspects of opinion polls. Now, opinion polls are useful in the sense that they give a break-down of the overall mandate as it happened — how certain categories of people voted, for instance. But there is an explosion in opinion polls now. They have become less of a means to understand voting patterns and more of an end in themselves, right?
We get good quantitative information and bases on which journalists can work to understand whether these translate well in the field. But opinion polls should not go any further than the numbers. While journalists tell the real motivation of the electorate, opinion polls may not do that. They are good enough to measure support, but the substantive ‘why’ question is to be answered by journalists using qualitative methods. This is not something that pollsters can do.
One of the most interesting findings of your book is that of landslides. Can you explain this phenomenon at the State and national levels?
We don’t have a fully national election any more, which used to be the phenomenon in the 1950s when we were newly independent and people trusted their politicians and leaders. The Lok Sabha elections are a federation of State elections. And invariably in Lok Sabha elections, each State votes differently and the final result in an election is the combination of ‘landslides’ and results at the State level that could counteract each other. So, we may have a landslide one way in Tamil Nadu and another way in Maharashtra, and so on. We find that 77% of the Lok Sabha elections at the State level have been landslides.
Tamil Nadu is a classic case?
It’s 94% in Tamil Nadu, the top of the list among landslides across States. It is always one way — huge victories for one party or coalitions. Landslides happen because in our first-past-the-post (FPTP) system, a small change in vote share gives a huge number of seats. The combination of the FPTP system and a fragmented Opposition results in landslides.
There is a talk in the media about narratives – say, BJP on populist nationalism, Congress on something else in the past. Does this make sense? Does this work nationwide or is it a myth?
It is a bit of a myth. For example, in Andhra Pradesh, we didn’t hear of Prime Minister Narendra Modi, or of majoritarianism or nationalism at all. These were completely irrelevant. But in some States such as Uttar Pradesh, these narratives do exist to some extent. But as journalists we tend to overemphasise that. We find over time that voters vote based on their life conditions. We went to a village where voters said they are going to vote against the government because a bridge had not been fixed. A respondent did mention Pulwama, but he rated it lower than livelihood in deciding his vote.
Maybe there is no one narrative that dominates across States, but we have had a government that has been in campaigning mode since it got elected in 2014. Has there been any such government that is so keen on getting its message across in the past before? Does this matter in setting the narrative for the elections now?
It does matter a lot. This government and the BJP are extremely efficient in booth management and turnout management. Globally this is seen as the focussed way to win elections. The BJP is superb at that — they have panna pramukhs , booth prabharis, and so on. Even the Left parties have good election management systems, but they don’t have the advantage of using social media. The BJP has social media apps in its election management, which the top leaders in the party use to reach out to panna pramukhs, who get messages in seconds and then everybody gets it. All this translates in good management of turnout. This turnout management plays a very important role in winning elections. We have seen that lower turnouts tend to help cadre-based parties like the BJP because they make sure their voters turn out. Non-cadre parties hope that voluntarily people turn out. The committed voter is expected to turn out more in lower turnout elections.
There is a worry. I am not saying it is going to happen this time in India — a turnout management ploy, like what we have seen in other parts of world, also includes rumour-mongering that makes people worry about going to vote in elections fearing violence. In America, voter suppression prevents certain categories of voters, such as African Americans, from voting. Suppression methods make it complicated for them to even be registered as voters. We have a similar problem.
Do right-wing parties create fear that women don’t turn out to vote?
They do that sometimes. But women in India have had increasing turnouts over the years.
Long time ago, Ashok Lahiri and you developed this index of opposition unity. It was an original contribution to this field. Can you tell us briefly about the significance of this? You speak about David Butler’s work on uniform swings and how this does not apply to all-India equations and how this work inspired yours in your book, right? And that index of opposition unity is a big determinant in India.
Yes. We learnt a lot from the work of Butler, who originated the uniform swings theory, but this worked largely in a two-party system. When he came to India, we found that it doesn’t quite apply here because we have so many parties here. So we had to work out an equation, which is a definition of what determines the change in margin. A margin of victory due to change in votes and a change in Opposition unity is what determines the winner. If there is a perfect two-party system, the index of opposition unity is 100/100. The more fragmented it is, it goes down to 70, 60, 50, and so on, and that determines the margin of victory as much as swings.
So people often say, is this a Modi wave? Was the 2014 election all about a Modi wave? We say, that is a misnomer because he won with 31% of the vote. He won due to the divided Opposition vote. In India, therefore, we must ask, how divided is the Opposition vote this time? For example, the Congress is fighting separately in U.P. How important is that? That is probably more important than waves and swings. The index of opposition unity is not 100 here. It’s 50, 60 or 70.
So how good is the index of opposition unity this time for the BJP State-wide and nationwide?
It is a crucial question. The actual number on index of opposition unity can only be assessed by opinion polls. For example, we found in our surveys that Yadavs are going to vote with the Dalits together in U.P. This is not only additive, but there is a boost beyond the arithmetic too. Because voters tend to believe this could be a winning coalition and it has momentum. So, if two parties bring 20% each, they could get a boost of 5% more votes due to this factor. We find that the average boost/momentum is about 8%.
Do you have data on vote transfer? People say some parties transfer less of their votes than others in a coalition, right?
It is conventional wisdom among journalists, and not just true. We find that vote transfer is almost 100% plus a boost. Journalists say that Yadavs might not vote for [Bahujan Samaj Party chief] Mayawati while Dalits might vote for the Samajwadi Party. That is just not true. Yadavs are voting for Mayawati, we find.
We hear that Muslims vote tactically. But the Muslim vote does split. It is splitting 80%-20% to the BSP-SP alliance and the Congress. In fact, no section is voting 100% for one party. Brahmins don’t vote en masse for the BJP. It could be around 65%. Yadavs don’t vote 100% for the SP. It could be around 80%. Journalists tend to look at quantitative data in extremes.
Clearly the index of opposition unity is higher compared to 2014. Isn’t that so?
Yes, much higher. It is going to make a huge difference, especially in U.P. I reiterate that is the question to be asked is, ‘How much is the index of opposition unity?’ rather than ‘Is there a wave again?’ In fact, the impact of this is much more. In U.P., even if the vote shares are exactly as in 2014, just a combination of these two parties, the BSP and SP, could reduce the BJP’s tally from 73 to half of that. If the Congress was part of the coalition, the seat tally could have dropped to 20. The fact that Congress, with only 6% of votes in U.P., is contesting the elections separately is giving the BJP an extra 14 seats only based on 2014 numbers. Which is 20% of the total 80 seats in the State. The BJP is much smarter in understanding the importance of the marginal vote share — 3-4% resulting in higher number of seats. They are much better in negotiating and even compromising with allies. They give seats to allies to get the extra vote share — 4-5% vote share in their favour to get 10% of the seats.
The Congress has misread the situation in U.P. They could have gained more seats by giving up a few seats to contest in an alliance with the BSP and SP.
Let’s look at the other States. How does the index work in Maharashtra?
We think it will be pretty close to a two-way fight. The poll will show this, but an index of 80 will result in a close one. In Kerala, for example, there are many parties but they realise the need for alliances, with the index of opposition unity being the highest in the country.
Kerala is ahead of the rest of the country in understanding the need for alliances?
Yes, they are way ahead! To do an opinion poll in Kerala is the most wonderful exercise and also the most time-consuming. We do pre-poll pilots to check whether the questions are right. And in Kerala, respondents tell us, your questions are wrong, and they give us a 20 minute-lecture on asking the right question to people! Every interview lasts an hour and a half. Everywhere else, people want us out of there, but people in Kerala want a discussion. Political awareness is superb in Kerala. It is the biggest learning experience for pollsters while doing pilot surveys in Kerala.
The index of opposition unity in Karnataka?
It will make a huge difference. The alliance between the Congress and the Janata Dal (Secular) will not only be additive in terms of votes. Unfortunately, their record after their coming together following the elections... they have been a bit fractious and this has chipped at their popularity, but the alliance itself will make a big difference.
You say in your book that beyond the index of opposition unity, swings in homogeneous Assembly constituencies (across regions) matter. In Karnataka, the JD(S) and the Congress are much stronger in the southern region. But the JD(S) is not so strong in the north.
Very true. JD(S) with its Vokkaliga base is much stronger in the south. But even in the north, a 7% addition makes a bring difference to the alliance. Yes, the impact of the alliance will be stronger in the south than in the north due to the composition of the Vokkaligas and the Lingayats.
It is a very heartening sign in some respects, but also disheartening. You talk about 21 million missing women voters. It came down from 25 million in 2014. But first, the positive thing. You are predicting in this book that in this election women voters may actually outnumber male voters at the all-India level in the Lok Sabha.
Yes. Out-participate in the sense that the turnout figures may be higher but because of the missing numbers, the absolute figures may be less. But in terms of the percentage turnout, it could be higher. And it is not a very difficult thing to forecast because in the Assembly elections, women’s turnout is now higher than men already.
The rate of increase is dramatically different. Women are coming out to vote more than men in every State, and more so in the south of India, of course, where the women are much more proactive. [As pollsetrs] we talk to them... I am telling you, right across the south, the woman inside the house will see you and she will come out and say, what questions are you asking? In U.P., they will be standing at the door and they will rush inside. They don’t want to interact. Although that is changing in U.P. too. In the south, the husband also comes and we ask the woman, do you vote independently or do you listen to what he says? And they say, listen to him, who is he? Sometimes we ask him, do you listen to her? Which they don’t do either. Men and women make up their own mind.
Is there a correlation between turnout of women and representation in terms of candidacy?
Unfortunately, so far there has not been. The percentage of women candidates the parties have nominated has been appallingly low - nowhere near the 50% that they deserve. But because of this increasing turnout of women overtaking men, now the policies of parties are becoming women-focussed.
Actually, the very clever and effective policy of gas cylinders [Ujjwala scheme] has really worked well. And that was just for women. Unfortunately, that has lost a lot of steam because now they have to pay for a second cylinder. So we met many women where the cylinder is just lying there. They got it free. But they are now using the stove again because they can’t afford the ₹800 for replacing that cylinder. Some are using it for fast cooking. In the morning, they quickly use a bit of the cylinder and then in the evening, when they have a little bit of time, they use the stove because the stoves are cheaper. So, that is just one area where parties are focussing on women but you look at the manifestos and look at what they do at election time, you see a lot of leaders saying, will all the women please come to the front? They are talking to women because women now matter much more than before. And that is a heartening sign.
And rural women are doing better than urban women.
Literally, a 5% higher turnout, which is a lot. Village women turnout is the highest in any category in this country and it is growing the fastest. And it used to be more than 20% behind men. Now, rural women voters turn out at a higher percentage than urban voters in towns, and certainly above men as well.
But you know, it is sad. Again, the qualitative side. You see these beautiful, long queues of women, and they are all waiting to vote, and then two-three go in to vote and come out. And one will say, I couldn’t vote. And I say, why not? They say, because my name was not on the electoral rolls. And it is 21 million women like that, which is over 35,000 women per constituency who are over 18, Indian and cannot vote. In the northern States, it is much worse; in the south, women’s registration is nearly 100%. It’s wonderful. And turnout is the highest. In the north, women’s registration on electoral rolls is appalling. The highest percentage of women not registered is in U.P. And the number of women missing in U.P. works out to 85,000 women missing in every constituency. Now, 85,000... margins of victory are much lower than that. So, it is just a tragedy.
You talk about three States – U.P., Maharashtra and Rajasthan — doing particularly badly in the book.
Yes, particularly bad.
A lot of people have written on this before, of various aspects of missing women. And not enough work has been done on why they are missing, and in what categories. And the sense we got — just again casual empiricism and surveys — is that the poorest are not there. The Dalits, the Scheduled Castes, the Muslim voters, the women are not registered. So, there is not only 21 million missing, but a bias in the missing. And the ‘why’ is the next step we are going to see. Why should these 21 million be missing? And we had hoped to go to the Supreme Court to say, in this one election, we cannot accept this. Please ask the Election Commission that any woman who comes and has got an ID that shows that she lives in that area and is above 18, let her vote. Then we found that, surprisingly, the right to vote is not a fundamental right in India. It is a statutory right. By the way, NOTA is a fundamental right. But the right to vote is a statutory right. So, you can only go to the Supreme Court with an SLP, given the time constraint, if it is a fundamental right.
When they prepared the electoral rolls, the first time, a number of women were left out because they couldn’t reveal their names. They were, wife of so-and-so. I think the first Election Commission looked at it and they had to decide to leave them out. So, the number of women missing. Now this is an interesting story.
There is a correlation between empowerment and registration. Where they are more empowered...
May be. But a lot of work needs to be done on why this is happening. I mean, the Election Commission has got a lot of schemes to get women involved. They say, look, it is not our fault because there are communities who don’t want them photographed. There are communities who don’t want the age of the woman to be revealed. But I don’t think enough work has been done because, if it is a biased sample, like it is only Scheduled Caste, Scheduled Tribe and Muslim women not being registered, it has a huge impact on the election.
It is said that the BJP scores more heavily among the men than the women. The Congress does better. At the State level, we have observed, based on the exit polls and the other polls, that the AIADMK under M.G. Ramachandran and Jayalalithaa had been doing distinctly better among women than men. Had only men voted, the DMK could not have been defeated. At the national level, does this hold?
Yes, generally the BJP has higher support amongst men than women, traditionally over the years. They are fairly a male-dominated party. The Rashtriya Swayamsevak Sangh has no women. But I think they are trying to change that. But at the moment, I think, yes, they do much better amongst men. So, the rising women turnout is a bit of a worry for the BJP. Although, again, through the LPG gas cylinders and other policies directed at women, they are trying to change that.
Now, one of the features of the book is your highly positive assessment of the Indian voter: independent-minded, looks at how these programmes translate at his or her level. Their living conditions. That is the most important issue. Throwing people out if they do not perform. Then the Election Commission itself. It must be one of the wonders of the world...
It must be. With all the faults, it is an organisation that is brilliant.
Let us take the EVMs first. Now, you look into it pretty thoroughly.
We have followed and tested and analysed this since 1977. And then in 1983. Actually in Kerala in 1982, and then in 1983 in Karnataka and Andhra they used it. And we saw it. And we saw it in the Election Commission offices. We pressed it. We checked it. We looked at whether it works or not. The main thing about them is they cannot be tampered with because they have no connectivity with the outside world. There is no Bluetooth. There is no Wi-Fi. There is no Internet connection. It is just a machine that is recording rather than a piece of paper, ballot paper. And it is the suspicion of any technology that one feels, oh, something is wrong. And only those who lose say the EVM was tampered with. When they win, it is fine. There have been a lot of hackers who say globally it has been tampered with, it is a complete flop. They cannot prove that these can be tampered with. It is very unlikely [to prove it has been tampered with] because so many people have tried. At the moment, I think, we should have faith that they are recorded properly. And VVPAT, which is a piece of paper that follows it, is just to assuage the doubts. It is the same thing. It is an analog version of a digital recording. But we are all analog minds I think. We are suspicious of anything digital, although digital is probably more accurate.
The Election Commission, I think, has done pretty well in inviting everybody to come and test it in situ. Not to take it out of their homes and offices. And so far, it has stood up.
Absolutely. They have done a lot.
I remember Julian Assange telling me that the best way to get leaked documents is hand it over personally. Standalone. No Internet. No mobile phones. Don’t keep your mobile phones nearby. They were working on all kinds of things to see.
But this is like a giant calculator. It does not even have an operating system.
Absolutely right. It is just a calculator. And you say, a calculator saying 3+3 is 6, there is something wrong. Let us get a printout.
So, this VVPAT is really a bit of a luxury.
A very expensive luxury to assuage...
It costs much more...
Than the machine itself. And I hate to say, a lot of trees are being cut because of the papers being used.
So this five per Assembly constituency. Do you think it is reasonable?
I don’t think it is needed at all. But it is psychological.
Now, the role of money in elections. This is an area where it is unfair to blame the Election Commission. They are completely helpless. Because there is political finance that operates outside the system of electoral funding. You already have these resources. Political parties collect money in various ways. And now you have new, innovative ways that are less transparent. Electoral bonds, which make an ass of the law. The law is truly an ass. The election law, when it comes to spending limits. And everybody knows. In Tamil Nadu, anecdotal... the minimum expenditure for a Lok Sabha constituency is ₹30 crore. In some cases, they will say ₹50 crore.
Yes, Karnataka too. Again, anecdotal.
So, what do we do about it? Is that your next book?
Yes, we are actually going to make this into a more academic piece, which nobody will read, of course. We don’t have tables with lots of columns. We try to bring it down to one figure. A key figure. There are some areas like money which we touched once and we got a lot of feedback from the field. But it is still too early to actually come to a proper conclusion. In America, a lot of research has been done, which says money plays a huge role in re-electing Senators and Governors. There, the re-election rate is about 90%. Pro-incumbency. Money effects pro-incumbency there. Here, as well. As you said, the kind of numbers involved. In addition to that, there are advertisements on television, on the roads, quiet handouts. And I probably have anecdotal evidence for this as well. Indian voters are a little smarter than the American voters. They take money from both sides and vote the way they want, because they know that the final vote is secret. They have great faith in that that nobody will find out how they voted eventually.
You’ve mentioned incumbency. And in the book, you say, quoting Ruchir Sharma, that in the U.S., nobody understands what anti-incumbency means. But in your book, you have a historical discussion on three periods. The first period is pro-incumbency, from 1952-1977. Then you have a period of distinct anti-incumbency, 1977-2004.
…2002, yeah. 25 years.
And then 50-50.
50-50 is the current stage where...
It is just the data because in the first 25 years, over 80% of the governments were voted back. So it was pro-incumbency. The next 25 years, when they were, they found that the politicians had failed them, they just threw out everybody. Good or bad, over 70% of the governments were thrown out. While earlier, 80% used to be thrown back, there was a complete reversal.
And since 2002, 50-50. Half the governments are thrown out, half have come back. And the governments that are voted back tend to be governments that have worked on the ground. They are doers. It is clear now that the voters are not taken up by pure oratory and great speeches because, the most successful politicians...
Naveen Patnaik.
A doer. Not known for his oratory. Even Raman Singh is not known for his oratory. Shivraj Singh Chouhan. Manik Sarkar earlier. Even Sheila Dikshit is a doer, she did a lot for Delhi. Not an orator.
They want things done on the ground. Your oratory may help. But if you have not done things on the ground, you will not be voted back
Now, on corruption. Let us look at issues. Usually, livelihood issues come right at the top. They are the most important.
Now the thing is, for an issue to be an election issue, it has to fulfil two criteria: one, it has to be important in my life — corruption is important to most people. You go to a police station, there is corruption. Widows go to a bank to collect their ₹1,000 pension every month. The bank manager charges ₹500 to give her ₹1,000. He’ll say, ‘Oh, big problem. Computer mein kuch...’ So a bank manager sees a poor, widowed woman and he still takes ₹500. I mean, it is just awful. And it is all computerised, so nothing to do. Corruption beats computers. So, corruption is always very high.
But the second criterion: one party must be seen to be better at solving it than the other. Take corruption, there must be a clear distinction in the parties for it to become an election issue. I think the Aam Aadmi Party once swept in Delhi because that is the one distinction they did have at the time that they were not a corrupt party, while all the others were seen as corrupt, apart from other issues.
V.P. Singh in the Hindi areas...
Yes, he was considered to be non-corrupt. Made a big difference. Cho Ramaswamy once said, that how can corruption be an issue when on the one side you have a pickpocket and on the other side you have a thief?
One thing you mention is that opinion polls generally get the winner right if not the extent of the win. One exception was the 2004 election. What do you think went wrong then, and have pollsters adjusted for it?
You know, people are not yet 100% sure why it went wrong but every poll got it wrong in 2004. One of the reasons is — and many pollsters are worried this time as well — the ‘play safe’ factor among voters when they are responding to questions. They tend to say their vote is for the ruling party. If you had done it immediately after the Emergency, who would have said, I am not voting for Indira Gandhi? You would have said Indira Gandhi to be safe because there was a feeling of fear. Out of 100 voters, five may do that, but 5% is huge. So, if the fear factor is 2%, 5% or 7%, it changes the forecast completely. So that is a worry in every election and in some elections a little more. And we have found that in the villages and even the small towns in the north, there is a lot of tension today. It is not like it was five years ago, 20 years ago. There is a lot of tension between castes and religions. And whether in this atmosphere there is going to be an additional fear factor, pollsters are finding it very difficult to assess that and it is a big worry.
