This episode is the 11th in a miniseries of weekly short episodes featuring young scholars entering the academic job market who discuss their latest research. In this episode, Shruti talks with Apurav Yash Bhatiya about how voters behave in simultaneous versus sequential elections, possible reasons why they behave differently, politicians’ use of the bully pulpit and much more. They also discussed his paper titled “Behavioral Voters in a Decentralized Democracy” (co-authored with Vimal Balasubramaniam and Sabyasachi Das). Bhatiya is a Ph.D. student in economics at the University of Warwick. His research areas include political economy, development and labor economics.
SHRUTI RAJAGOPALAN: Welcome to Ideas of India, a podcast where we examine academic ideas that can propel India forward. My name is Shruti Rajagopalan, and this is the 2021 job market series where I speak with young scholars entering the academic job market about their latest research on India. I spoke with Apurav Yash Bhatiya, a Ph.D. candidate in economics at Warwick University. He also has a B.Sc. in economics from the Indian Institute of Technology, Kharagpur, and an M.Sc. in economics from the Toulouse School of Economics.
We talked about his paper titled “Behavioral Voters in a Decentralized Democracy” (which is co-authored with Vimal Balasubramaniam and Sabyasachi Das). We talked about voter behavior in simultaneous versus sequential national and state elections in India, information overload, rational ignorance models, the bully pulpit effect and much more.
For a full transcript of this conversation, including helpful links of all the references mentioned, click the link in the show notes or visit Discourse Magazine DOT COM.
Hi, Apurav, thanks so much for coming on the show.
APURAV BHATIYA: Thank you. It’s an honor to be here.
Simultaneous Elections Lead to Bigger Party Gains
RAJAGOPALAN: You have a super interesting paper with Vimal Balasubramaniam and Sabyasachi Das on studying the effect of simultaneous elections in India. The listeners know India has a federal system. We have parliamentary state and local elections, and sometimes the cycles coincide. What you study is that during those cycles when there’s a general election and a state assembly election simultaneously—in certain parliamentary constituencies and assembly constituencies.
What you find is remarkable, that when there is a simultaneous election versus a sequential election, the likelihood of the same party winning increases by 21%. That’s a big effect. Can you tell us more about why this happens?
BHATIYA: In this paper, we study how synchronization of elections affects the voters’ behavior, and is it big enough to significantly affect the electoral outcomes? Maybe let me make it clear for the listeners what is synchronization of election. That means holding the elections for the different tiers of the government on the same day. Currently, there is a policy motivation to study this question, and there is an economic motivation to study this question. Let me go into both of them in detail.
In the policy motivation, the government of India has a proposal to hold the synchronized elections, and the NITI Aayog and the Law Commission of India are all in favor of it. The proposal’s core rationale is that elections are expensive to the exchequer, and multiple elections create disruptions to governance as the bureaucracy is occupied with election management. This policy proposal is not just in India, but even the European Union and South Africa are thinking about it. Over here, we ask a more fundamental question: Does synchronization of elections affect how voters vote, and what are its implications?
Now, on the economic motivation side, there are two different strands. The first is that theories of decentralization point to the economic and governance benefits for having multiple tiers in the democracy. These theories presume that voters are sufficiently sophisticated to make decisions in this manner. If the voters cannot behave in this way, it affects the degree of effective or de facto decentralization in the economy and the benefits that can result from it.
The second economic motivation strand is that there is a growing literature that studies the cognitive limitations and behavioral biases in the elections, such as voters prefer to vote for the winning candidate, they are sometimes overconfident, they are inattentive to information. This evidence studies how individual voters vote in a single election. Almost all of the democracies today are decentralized.
We study this question of synchronized elections and how it affects the voter behavior. We set up a model to illustrate cognitive constraints voters face during simultaneous elections. And we use, as you were saying, the natural variation in the electoral cycles in India to empirically test models’ predictions.
We have three sets of results. The first is that synchronized elections increase party salience among voters. By party salience, I mean when the voters have to think about party affiliation or the personal characteristics of the candidates, they’re more likely to focus on the party affiliation of the candidates.
The second is that there is an increase in likelihood that the voters vote for the same party across both tiers. In the Indian context, that means when a voter goes for voting, they’ll vote for the same party for the Lok Sabha seat and the Vidhan Sabha seat. The third result that we find is that there is an increase in probability that the same party wins both constituencies and that this increase is about 21%.
Theories About the Indian Voter
RAJAGOPALAN: I want to understand a little bit more about what is your theory or model of the Indian voter. What is going on with them, either with the assembly election or the parliamentary election or when both of them happen in a synchronized way?
BHATIYA: We don’t have a fancy theory for how the Indian voter works. We’re taking a very simple theory of what the decision-making process is for a voter when they go for the elections. To understand how the simultaneity of election changes voter behavior, we basically use the timing of election across tiers to understand the vote behavior changes. We set up a model where once the voters are voting for both tiers on the same day, and once they are voting then there is a gap between the two elections. We set up a voting choice model with two parties, and the voters have to decide on who they will vote for.
They have the party affiliation of the candidates. There are personal characteristics of the candidates, such as what is the work that they have done, what is the caste, what is the religion? They give some weights to these different characteristics of the candidates when they’re deciding whom to vote for. We try to understand how this decision-making rule and the weights for different characteristics change when the elections are synchronized or when they are held sequentially.
Now, in the setting, we take the party affiliation as a characteristic which is more easily available. You know which parties are contesting, so you know the parties, but the information on the personal characteristics of the candidates is more difficult to acquire. When the elections are sequential, this decision is split up and the voters vote based on their characteristics. When the elections happen on the same day, a joint decision has to be made on how to vote.
When that happens, our model gives three predictions. The first is that parties become more salient when the elections are synchronized in the voters’ decision-making rules. Party gets more weight in their decision-making rule. The second is that the fraction of voters who vote for the same party also increases in simultaneous elections. The third is we find that the probability that the same party wins both elections is higher when they are held simultaneously than when they are held sequentially.
Rational Ignorance and Voter Confusion
RAJAGOPALAN: This is in one sense a very classic and canonical model of rational voter ignorance. This is your Anthony Downs origins of public choice. The voter is rationally ignorant because there is a very high cost to assimilating information about, as you mentioned, the details of the candidate, the party affiliation, so on and so forth. The benefit is fairly low because the probability of overturning an election because of a single vote, because of an informed voting choice, is incredibly low. In a sense, is this just the most obvious result—that none of this is surprising?
I mean, you’re going to have a marginal effect if they are held in a simultaneous way, but it’s not like the Indian voter was doing a lot of research or any voter is doing a lot of research when it comes to nonsynchronized elections, right?
BHATIYA: Yes. Essentially, I think what you’re thinking is that, is it an information overload story that the voters are voting on the same day, or is it something else? What we’re seeing here is that there are cognitive constraints that the voters face. These cognitive constraints versus information overload hypothesis, we try to understand in our framework through the empirical findings. We use the Lokniti CSDS surveys to understand the voters’ decision-making when they are held in a synchronized way or in a sequential way—so the simultaneous and sequential elections.
We take the questions from the survey, and we try to understand the voter behavior. We take the question, “What is the most important issue of the election?” In the survey, the responses could be what their issue was, and we classify that issue as a national issue if it falls under the central government according to the constitution. And if it becomes an issue of the state government if it falls under the state government’s area of working.
Let me take an example, and maybe that becomes clear. If the voter says the most important issue of the election is national security, that’s a central government issue, and the voters are thinking of national issues. If the voter is thinking of the hospitalization or medical care as the most important issue, then that’s the responsibility of the state government.
We try to understand first how the election issues change, how the most important issue of the election changes, when they’re held in a synchronized way. What we find is that synchronized election doubles the probability that the voter says, “I don’t know what is the main election issue.” What we see here is that synchronized election dramatically increases confusion about the election issues. Now, this confusion could be through the information overload hypothesis or the cognitive constraints.
We delve into it further, and we try to understand if the confusion is explained by the voter characteristics. Essentially, trying to understand if female voters are different, if older or younger voters are different, how the voters are behaving differently by their educational background or social category. And we don’t find any of these characteristics explains this confusion. The confusion is equal for all voters.
Effects of Party Behavior
RAJAGOPALAN: Let me go in a different direction. You’re right in that you don’t find much difference in voter characteristics based on gender and caste and so on and so forth. Are you maybe looking in the wrong place, in the sense that is there a different explanation coming from political party behavior that could explain it?
There is a cost to elections and campaigning and so on. When there is a party that could have the possibility of winning both the assembly election seat and the national election seat, then it might strategically allocate more resources. It might fly more VIPs because you’re hitting two birds with one stone, so to speak. Does the party salience come because of the supply-side effect from the effort that the party is putting in, and not just what the voters are doing?
BHATIYA: No, we don’t find any supply-side changes when the elections are synchronized. To understand this, we use the dataset available from the Trivedi Center of Political Data. They’ve assembled an amazing dataset for Indian elections. We try to understand the candidate selection strategy through the data. First, we try to understand if candidates running for the first time go up when the elections are synchronized. We don’t find that to be the case. We don’t find any turnout or turncoat effects—so essentially, candidates are changing their parties between two elections.
Turncoats don’t happen between sequential and simultaneous elections. Also, we don’t find that the candidates are re-contesting candidates. A candidate is re-contesting from the same constituency, or from a different constituency within the same state. And we don’t find that to be happening. The total number of candidates who have lost their deposit in the elections does not change between simultaneous and sequential elections. The total number of candidates who run for the election also is not different between the two.
We also look at the share of parties which are fielding candidates. As you were saying, you can hit two birds with one stone. It’s not the case that the share of party fielding candidates in both years changes when this happens. Importantly, this result is also true for both national parties and state parties, because the level of funding can be drastically different between the two groups of parties. And we don’t find the supply-side changes to explain the results.
RAJAGOPALAN: There’s one part of supply side which is talking precisely about the characteristics of the political candidate, but the other aspect of supply side in an election is information. This is basically campaigning. So it might make more sense to bring in the president or the leader or the face of your party to a constituency where you have both assembly candidates as well as a parliamentary candidate. and the election happens to be simultaneous.
I’m thinking more from a point of view of information, because if you say there is a capacity constraint to absorb information and it leads to party salience, then maybe the parties understand that, and they try to supply greater information to increase party salience.
BHATIYA: Indeed. To understand this effect, we try to do a couple of things; one, we look again from the survey evidence, and we find that the political campaigning is not different. The survey asked the question, “Has a political party visited your house to ask for votes?” We don’t find that to be different between two elections.
Next, most importantly, we find that when we change the time window of the gap between the two elections—we’re comparing an election which happens on the same day with elections which happened with a gap of 180 days. If we vary that gap to, let’s say, 210 days, 240 days, 270, 300, 360, 480—changing this time window of sequential election does not affect our results. If information was playing a role, maybe the time gap would have widened, changed the degree of information flow, and that could explain the results. But that doesn’t seem to be happening.
Information and Media Fatigue
RAJAGOPALAN: Another question in terms of the information. Do you think some of it could be just media fatigue? There are only so many column inches in a newspaper or only so many minutes in a primetime news show. And it’s harder to cover in-depth detail, both the national general election parliamentary seat candidate, as well as the assembly seat candidate. There is much less information, so there’s greater reliance on a heuristic like party affiliation.
BHATIYA: Essentially, what you’re suggesting is that it may be there’s a choice fatigue that’s leading to decision shortcuts. We also test that by understanding if there are coattail effects. Coattail effects have been prominently documented in the economics and political science literature. And by coattail effects, I mean that there is a prominent candidate in one election that attracts the votes for their party for the other seats when they are held sequentially. We’ve tested for the star candidates, as we call it, in the coattail effects.
We try to look at if prominent candidates such as the prime minister candidate or the chief minister candidate, or it’s Sunil Dutt, O.P. Jindal, Shatrugan Sinha, who are contesting for the elections. We try to decompose the effect that’s coming from these coattail effects or the star candidates and the non-coattail effects. We don’t find that the effect is explained by just the coattail effects.
RAJAGOPALAN: You’re not in a position to look at media transcripts and do textual analysis of newspapers and things like that. That would be a slightly different direction, but that could also provide supporting evidence for your theory.
BHATIYA: No, we couldn’t do that much. In terms of the time period, we have analyzed between 1977 and 2019. We analyzed all the elections and, in that period, we’ve tried to understand how the synchronized elections play an effect.
RAJAGOPALAN: I just want to go back to the canonical, rational ignorance voter models. One of the weaker versions of that hypothesis is that, because this information processing and acquiring is so costly to voters, that individual voters will generally choose to remain rationally ignorant. A strong version of that hypothesis is that because of this, in large elections, voter preferences are not really reflected in political choices and political outcomes.
Now, in India, there is another very interesting natural variation, which is that each constituency size is not the same. I don’t mean just the parliamentary versus the assembly constituency. Of course, those are different. Every parliamentary constituency is not the same number of voters as it was at the birth of the republic. This is, of course, thanks to delimitation, and they have to figure this problem out in the next five years, in 2026.
Do you think that the size of the constituency changes this in any meaningful way? That is, if the rational voter model is correct—if it’s an information question and not a cognitive constraints question—then you should see that in larger elections where a voter has a low probability of flipping the result, you should see more of a particular behavior versus—you should see more party salience, let me put it that way, versus in smaller constituencies. Have you studied that at all, or what do you think about my hunch?
BHATIYA: In that context, I think what you’re thinking is that if there is a large constituency, that my marginal vote would not make a difference, so maybe let me not go and vote. So essentially, the turnout changes. Over here, actually, what we find is that state elections do not experience an increase in turnout when the elections are synchronized. When the state elections are happening on their own, there is some turnout, but when they are clumped with the national election, there is no difference in turnout between these two cases. In the national elections, we do see an increase in turnout.
National election—general election—happening on its own has some turnout, but when it is clumped with the state election, the national election turnout actually is higher. We find that this increase in turnout is uniform. Again, from the survey evidence, we find it’s uniform across all individual characteristics. So it’s not the case that, again, women are more likely to vote, or younger or older people are voting differently by caste groups or by education. These turnout changes are uniform across individual characteristics, and we find that our main result is not driven by constituencies which have a higher turnout. This turnout cannot completely explain the effect.
RAJAGOPALAN: One part is turnout, but also maybe you choose to be rationally ignorant. And you just show up in a form of expressive voting, but you inform yourself less when you’re in a larger constituency because you’re not going to change the result, right? Only one part of the cost that the voter takes on is showing up. The remaining is everything you’re talking about, which is getting to learn the characteristics and so on so forth. That would be an interesting thing to study.
I have another question about what might be driving voter behavior that leads to this party salience, which is do the voters perceive that there is some coordination benefit if you have the same candidate at the state level and the union level from the same party, so they could coordinate better as opposed to slippage? Do you think that could be driving some of this result?
BHATIYA: No. So we analyzed it, and we don’t find that any changes in economic activity explain this result. We look at both synchronized elections and synchronized representation, and we look at different economic activity outcome variables. We look at what is the overall agricultural production in those areas. We look at the gross cropped area, we look at the credit dispersed by the RBI [Reserve Bank of India]. We look at the total investment both from the public and private sources.
Lastly, we look at the night lights, which are frequently used as an overall indicator of the economic activity. Across all of these outcome variables, we don’t find any significant differences, either from synchronized elections or synchronized representation. Essentially, any economic benefits that could result from synchronized election does not seem to be there. That really makes us ask the question, “Are there any benefits to synchronization in terms of decentralization, and what implications does it have? And do we really want to play with how the voters vote by changing drastically the way elections have been happening in the country?”
Normative Implications of Synchronized Elections
RAJAGOPALAN: I think you landed on one of the very important normative implications of your paper, which is that in the framework of India’s democratic structure in the republic, we have a set of horizontal checks and balances. Of course, you have the parliament, you have the executive and the cabinet, and you have the judiciary.
But there’s also very clear vertical set of checks and balances, which is the three tiers of government. Some of these vertical checks and balances have been quite important over the years. Especially since India walked away from a very strong form of central planning, you’ve seen greater amounts of federalism in India. Also, fiscal federalism has started playing a bigger and bigger role, of course, diluted a little bit by the GST [goods and services tax]. One of the important normative implications of your paper is that holding synchronized elections would actually weaken the federal structure, because now party salience is going to dominate other effects. Is that a correct reading of your work?
BHATIYA: Yes, it is in some sense that—because the debate currently is all about the costs of holding elections across tiers and the implications of it. Our paper is highlighting the question, “Do we want to play with how the voters are behaving, and what implications does it have on the decentralization as well, going forward?”
RAJAGOPALAN: I have a question on long-term voter behavior. Is it a reasonable assumption that as the voting system changes and tries to game voter behavior, eventually the voters will also get smarter and will know that they are being gamed, and these effects will go away? In other words, if India does switch to a system of synchronized elections, is there over a period of time, do you think the voters are also going to start changing their behavior to get what they want?
BHATIYA: I don’t know, sorry.
RAJAGOPALAN: No, that’s fine because this is why I was going back to the earlier question of what is a model for the voter. If the voter is rationally ignorant, then that’s not going to be the case. If it’s only a question of cognitive capacity and there are huge costs to voting in this particular way, then hopefully, over a period of time, voters, the media, the entire surround sound that that’s there in an election will start accommodating for this particular fact of synchronized election. Very similar to how it happens in developed countries.
RAJAGOPALAN: If it’s okay, I want to also pivot to another question that you’ve been working on, which is fascinating. This is the paper where you’re talking about a very salient political party or a political leader using what you term their “bully pulpit” to drive election results. Here you find that there’s a very interesting effect of the attacks that happened in India in Pulwama quite recently. There’s been a lot of narrative that that changed the outcome of the election; it led to this big BJP [Bharatiya Janata Party] wave. We know all that from media reports, and we know that through anecdotal evidence. Can you tell us a little bit more about exactly what is this bully pulpit and how politicians maneuver it or exploit it?
BHATIYA: It’s a joint project that my colleague from Warwick, Yatish Arya, and we try to understand how political messages from the head of the state—or “bully pulpit,” as the U.S. President [Teddy] Roosevelt called it—how voters are receptive to political messages, and how it affects the voting behavior. We take this focus on India, and we try to understand the incumbent prime minister’s focus on soldier deaths and secessionist conflicts during the run-up to the 2019 general elections. What we do is we try to exploit the exogeneity of the home constituency of the dead soldier.
Think of it this way, 40 soldiers died in the Pulwama attacks, but they all came from different constituencies all across India. What we do is we try to understand that if a constituency in a state in Haryana gets a soldier death, how do voters in that constituency vote versus some other constituency? We do this analysis, not just for the Pulwama attacks, but actually, we take all the soldier deaths which have happened in India since 2004 to study the voter behavior.
We have three sets of results in this study. First, using textual analysis of the prime minister speeches, we find that only the soldier deaths that happened in secessionist regions increase the nationalistic content in his speeches. If a soldier died from a Maoist conflict, the speeches of the prime minister do not mention friends and nationalistic issues.
Second, what we find is that the BJP and the NDA’s [National Democratic Alliance’s] vote share increased in the constituencies that received the soldier deaths from secessionist regions only, but the soldier deaths from the left-wing extremism or the Maoist conflict had no impact on the voter behavior. Essentially, think of it this way: The soldiers are dropped from all across India. The soldiers don’t know in which conflict region they might be posted, or which soldier might die at what time in which region. Essentially, the shock that any region of India receives from the soldier death is very random in nature, but the voters are only responding to the messages of the bully pulpit, the head of the state, and they vote accordingly.
The third result is that we find, also from the survey evidence, that the voters are more likely to mention secessionist issues as important in the 2019 elections in areas that had a soldier death. And then they vote for the incumbent BJP and the NDA party.
Research vs. Memory Recall
RAJAGOPALAN: This is really fascinating. I have a follow-up question to that. One explanation is, as you point out, this bully pulpit effect that is going on. Do the prime minister speeches reference simply the attack and the fact that soldiers died, or do they also mention the specific soldier’s name and characteristics from that particular constituency?
BHATIYA: No, the prime minister speeches do not mention specifically the name of the soldier or anything. It’s just general speech. What we’re thinking is that is it an information effect or a memory recall effect, and that’s what we study in this paper.
Let me take an example again. It’s possible that there is an event, a soldier death, and the politician is talking about that event. It could, A, make the voter more informed about the issue, and then they go and they learn more about the issue—what is this conflict about, what has my politician done—and they change their voting behavior.
The second is that the voters are not more informed, but essentially it acts as a cue to recall their past experiences, and they vote for him. That’s what we find, that it’s the memory recall of these events that makes the voter change their behavior. Essentially, when we drop all the soldier deaths which happened just in the Pulwama attack from the analysis, the results still exist.
Again, from the survey evidence, the Lokniti CSDS data, we check how informed are voters about these issues. We find that voters are equally informed about the issues, but they are more likely to credit the prime minister in the response of the attacks solely in areas which got soldier deaths, but not in other areas.
Also, local campaigning, local political participation does not explain the results. It’s not that the parties were strategically going to constituencies where soldier deaths happened. So that didn’t happen. Local incumbency does not play a role. And the result is, again, uniform across voter characteristics.
Immigration and Its Impact on Political Behavior
RAJAGOPALAN: What are the other questions you’re looking at in terms of political economy of elections, or what’s the larger research program that you have planned over the next few years?
BHATIYA: In another project, in my job market paper, I try to understand how enfranchised immigrants affect politicians’ behavior in the U.K. International migrants are currently the most extensive and growing disenfranchised group in Western democracies. The literature has these two links: On the one hand, immigration has been linked to the rise of populism and polarization in the host countries, but on the other hand, the evidence also suggests that immigrants’ naturalization leads to more integration.
The issue is that this process takes many years. In the U.S., actually, it can be decades for many immigrants. In response, in Europe, in Sweden and Switzerland, there have been referendums in recent years to enfranchise foreign-born noncitizens after a few years, much before nationalization is possible.
In my job market paper, I study the unique context of the U.K., where immigrants from Commonwealth countries and Ireland have the right to vote as soon as they come into the country, while the remaining immigrants are not enfranchised. I do textual analysis of the parliament speeches and I look at the voting behavior on the bills. What I find is that when there is more enfranchised immigration, the politicians are spending more time in the parliament talking about immigrants. They’re also talking positively about the immigrants, but at the same time, they’re voting to increase immigration restrictions.
Overall, I find that enfranchisement leads to a higher political engagement of the immigrants, and the politicians are now responding to it. But still, because the immigrant community is small and there is a small voting bloc, there is an anti-immigrant sentiment among the natives. The incumbents are compensating for their actions for enfranchised immigrants by restricting future immigration.
RAJAGOPALAN: Yes, so basically, it’s not a large enough group to swing the voting outcome, right?
RAJAGOPALAN: That’s really what’s driving this result. If it became a larger group, it would swing the vote, which is also the push-and-pull effect of what natives don’t want. No, that’s a fascinating research program. What have you been up to during the pandemic?
BHATIYA: The coping mechanism is exercises. I’ve been doing some outdoor running, and that’s when I’ve been listening to podcasts. Actually, I got hooked to podcast while running outdoors and trying to find something to listen to. Just got bored of the music. Over the pandemic, I also got married, so that’s also been something.
RAJAGOPALAN: Congratulations, that’s wonderful. Now the most important question. What are you binge-watching?
BHATIYA: I’ve been binge-watching “The Crown” recently, and I’ve been enjoying it a lot. I’ve been in the U.K. for five years now, and I can relate to it and connect with the story a bit more.
RAJAGOPALAN: Fantastic. I really enjoyed the series too. Thank you so much for doing this, Apurav. This was such a pleasure. Thank you for making the time, and I’m really look forward to all your research in the future.
BHATIYA: Thank you so much. It’s a pleasure to be here.