VB Singh on survey research in India since the 1960s


One of the privileges of conducting research at the Centre for the Study of Developing Societies is the presence of so many accomplished scholars of Indian politics. One such scholar is VB Singh, a Senior Fellow. Professor Singh joined CSDS in 1966, and worked on the institution’s first National Election Study in 1967. He has conducted research on electoral politics for many decades and his publications include a five volume series on elections in India giving detailed results, including by-elections from 1952 to 1985 for all the states in the Union of India. From 1987 to 1990, Professor Singh directed a program on survey research and training, which conducted research methodology courses in the social sciences. He served as the director of CSDS from 1997 to 2003.

On December 19, I met with VB Singh in his office at the CSDS campus to talk about his experience working on election studies since the 1960s. The transcript of the conversation below has been lightly edited for length and clarity. The transcript starts abruptly because I did not turn on the audio recorder at the start of our conversation.

On CSDS’ origins and the first National Election Study in 1967

VB Singh: …building an organization where the problems are not studied only in the classroom situation, but the area is much wider. Go to the people. Understand their views. Understand their problems and their reflections on various issues confronting this country.

So the first two studies undertaken by Rajni Kothari at that time where one was a values in active community, that was an international project headed by Philip Jacob and his wife. In that, they were trying to reach the district-level and block-level leaders. By that time, the three-tiered government was already in existence. So how the values of these active people at this lower level is being inculcated to understand the problems. So that was the project which was done here in India in three states: Uttar Pradesh, Maharashtra, and Gujarat.

Sam Solomon: And what year was this?

VBS: This was 1965.

SS: 1965. And when was CSDS started?

VBS: CSDS was founded in 1963.

SS: By Rajni Kothari?

VBS: By Rajni Kothari.

The first election study was undertaken by Rajni Kothari and his team. At that time, his team constituted Rajni Kothari himself, DL Sheith, Ramashray Roy, and Bashiruddin Ahmed.

SS: And what about you?

VBS: I joined a little later. I joined actually, some time, early ‘66.

SS: So that was before the ‘67 [national election and election study].

VBS: Before ‘67. I came to CSDS through one project which was [a] cross-national project on social and political change. Sidney Verba. Have you heard of him?

SS: I haven’t. Sidney Verba?

VBS: Sidney Verba. Civic Culture [is] his book. Civic Culture. He was from Stanford University. And he had a cross-national study of social and political change. That study was conducted here in India, in Yugoslavia, in Japan, in Canada, and I think America as well.

CSDS was the India partner. And that was the first major fieldwork in which we traveled [to] almost every major state of this country. That was a kind of cross-section [of] people: the panchayat leaders, the block-level leaders, and then district leaders. To cover the three layers of administration, and their views with regard to various issues, and their perceptions, and their evaluations. There is one book based on that data, published by Verba, Bashiruddin Ahmed, and Anil Bhatt. That book may be available in our library [gets up and looks through bookshelf in his office]. You will get it in the library.

SS: This study was in ‘66?

VBS: That was 1966. And then, [following] that study, the ‘67 general election came. So part of that Social and Political Change Studies resources were used in that study. And some new grant from other interested scholars came. We had partial funding from various sources. SRC [Survey Research Center at University of Michigan], America, and then this cross-national project money. This transnational studies fund was utilized to study voting behavior in 1967.

SS: So that wasn’t part of the social and political change study. It was a separate thing.

VBS: That was earlier than this. But that spilled over. The same team went to the field again.

SS: So the same team that did that study did the study for 1967?

VBS: Yeah. By that time the kind of foundation of the survey team was already made. Basiruddin Ahmed happened to be the project director on both the projects, that cross-national social and political change [study] as well as this election study.

SS: What was your role for the 1967 national election study?

VBS: I came from a very rural background. Trained in psychology in Lucknow University and just happened to be here. So I was at that time a research assistant in that project. But I was quite enthusiastic, so at that time Rajni Kothari and Basiruddin Ahmed, they gave me a brand [new] name, “Field Marshal.” It means carrying out, supervising the fieldwork, recruiting and training the researcher team, sending them to the field.

But not alone. At that time, whenever the training of the investigators used to take place, all these go, the senior people: Rajni Kothari, Ramashray Roy, Bashiruddin Ahmed, DL Sheith. [They] assisted them because everyone was interested to know the problem from the ground. They used to visit some of the field along with the field researchers, had their own experience. Their point was that the data were collected by the field investigators, but we have to give the meaning to the data. So we must also go to field, and interact with some of the local people then we will be able to understand the responses from the people and will be able to give meaning to that.

So I associated myself from that. Questionnaires used to be conceived in English. I used to translate that into Hindi.

SS: The original questionnaire was written in English?

VBS: Yeah. Because the basic thing is that with survey research, we had borrowed this technology from some of the established research centres. Mostly, actually, in America. So many of the questions on motivations, on other larger issues, theoretical issues, those questions were in English. So that we brought into [an] Indian English version. And then that had to be translated into the local language, that language which illiterate people can also comprehend. So I did a kind of translation of those and going to the field, testing whether the translation is done properly, whether we are getting the kind of responses we intended to get. Pre-testing of the questions. So that was done.

And then the survey used to be conducted and at that time we didn’t have this kind of mechanism of state-based transferring answers from questionnaires to the computer. At that time, questionnaires used to be quite lengthy. There used to be a large number of open-ended questions and that was the period actually… You c[an] have pre-coded questionnaire questions only when you have a kind of paramount idea of what type of answers you are going to get. Since that was the early stage of our research, we were not very sure. We were asking this question and what can be the answers? But now we can very easily, based on our earlier experiences, we can pre-code. We can pre-conceive the type of answers we are going to get and then you can provide these spaces in your questionnaire.

Earlier everything used to be recorded, the answers [for] open-ended questions. So the filled-in questionnaires used to come, then a team used to sit together. All those open-ended questions’ answers used to be jotted down with a kind of frequency marking. And then a team of seniors, with the project director being the head of that team, the coding category used to be constructed, be it caste, be it problems, be it aspirations, be it expectations, type of leadership. So that used to be done.

And then we used to have the large 80-column codesheets, and we had IBM punchcards. I’m sorry, actually, I have to describe all those things. So the data used to be punched by a punching machine on those IBM cards that used to be of this size. [holds up a large book] Of this size. And there used to be columns 1 to 80, and they used to be filled [with] 0 to 9. And then data used to be punched and then these cards used to be fed to the computer.

SS: So you had a computer here at CSDS.

VBS: We didn’t have a computer. We used to actually go to Delhi University and for some bigger allowances, at that time IIT Kanpur had a very big computer. So the team used to go with that data to IIT Kanpur computer.

SS: So was the data in the punch cards?

VBS: Yeah, punch cards. We prepared a kind of box, in which [there were] actually two thousand cases because of our first study, after election study, and the cross-national study. Tthe total number of completed interviews were less than 2300.

SS: So you had a card for each respondent?

VBS: One card for each respondent. That data used to be run into several cards for one person. So eight cards, nine cards. Card one, card two, card three, card four, card five, card six. So there was variable 1 to, say, 20 in one card. 21 to 40 in other card. 41 to 80 in other card[s]. So that is how this data used to be handled.

You used to get a table, you could ask [for] cross-tabs, bivariate tables, control tables. So the data handling was a big problem. Now whatever you conceive, you feed in the computer and get the out. At that time, getting one table worked out, it used to take days! Even with computers.

SS: Because you had to put all these cards into the machine.

VBS: And then the machine read [the data]. You had to write programs. SPSS was not available in those days. SPSS came into being some time in the ‘70s. So the programmer used to write: this variable, this field, compare this with this variable. So we had the programmers. So that was the case, actually.

At that time our fieldwork was a very vigorous thing, actually. We used to be in field for months. These days actually sending a person to the field for four, five days becomes very, very tiresome. We used to travel. I have myself, actually, in 1966 and 1967, I mean, two months in the field! And we had four or five teams going in very different directions, spending one, two days in one village. Complete the interview during the daytime, most of the time, actually, since it is a kind of working population, you had to be there during the night. For light, we had torches. We used to interview with that. That was a kind of very vigorous thing.

SS: How did you assemble your field team for ‘67?

VBS: Not through newspapers, but we had contacts in different universities. So suppose we have to do fieldwork in Uttar Pradesh. Lucknow University, Allahabad University, Aligarh University. We are looking for these kinds of persons.

Suppose we want to recruit thirty investigators. We used to give training to fifty during all-day class–coming, going, staying, and training them. And out of that fifty, we select thirty, [and] put some persons in the reserve field. Suppose somebody gets ill or somebody decides to leave in between. Always through personal contact, so that the investigator who is selected, he is directly responsible to me only, he is responsible to the person who has recommended his name, and through him, to the person who has recruited. So we used to have this kind of very, very committed team.

SS: For each state?

VBS: For each state. For each state.

SS: The 1967 national election study, was that fielded in every state?

VBS: Yes, and there were state in charge. In 1967, by that time, I had already acquired some experience. So I was made in charge of [the] Madhya Pradesh team. I went to Madhya Pradesh in 1967. In Uttar Pradesh there was another colleague who left us, Mr. Ayep, and one Dhirendra Vajpeyi, who is maybe still alive actually, at Ohio University. These two persons were in charge of Uttar Pradesh. And the training used to be done, generally, [with] two or three states together.

At that time, we used to get a letter from the Election Commission and from the central government. “Such-and-such research team is going to such-and-such area. Provide them basic help and assistance.” We used to approach the district headquarters. From there we used to book our stay in some circuit house, guest house, and the local officials of the state or the block development officer were asked to coordinate our fieldwork. So generally they used to provide local transport from wherever we stayed.

SS: The Electoral Commission would transport you?

VBS: Sometimes, actually, because the letter was from the Election Commissioner and the Home Secretary. They used to transport us from headquarters where we were staying to the village and thereafter they disappeared. We would be there in the field for one day or two days and then they would come and take us back. At some times, if there was no staying facility in that village, then we would do that fieldwork until late and then we would come in that Jeep to the headquarters.

So it was quite tiresome at that time and we traveled, actually, to remote areas where Jeeps can’t go. Even the carriages run by horses, they [would] not. Sometimes bullock carts, you had to hire. [laughs] One field, we had gone by elephant. Riding the elephant, we crossed.

People actually by and large were very, very committed. There used to be team of four or five persons. One person used to just roam around and see if things are doing done properly or not. If somebody is interviewing somebody and that person is surrounded by a large number of spectators, somebody used to go, “Please come and talk to me. Let him do his interview.” So this was our earlier practice.

SS: Was the fieldwork done before election day or after? How was the fieldwork timed to coincide with the election?

VBS: Good question. You see, if it is a kind of voting behavior study, at that time we used to do post-election. Post-election means the process used to start before the election, but the preparing of questionnaire used to take something like [a] fortnight, or a month. Because there used to be four or five scholars, pulling their minds, preparing questions, deliberating, fighting with each other. So finalizing a questionnaire itself was very, very time-consuming. And then the field investigators go to the field after the results are more or less declared. Then we used to ask all these questions: To whom did you vote for? Why did you vote for this person? All other questions.

Later on, we realized that there are some kinds of gaps. Gaps in the sense that once you are going to the field when the results are already announced, then at that time, if you ask the question, “To whom did you vote for?”, there is more likelihood of the person responding in favor of [the] winning candidate. Because it’s human psychology to side with the winning person. So we used to get much more responses in favor of the winning candidate and the winning party. Then you had to test the validity of that response along with the other questions. He said that this party was good, that party was bad on this count, how far the response of that person is in favor is correct. Number one.

Number two, another thing was when you ask a person, “Whom did you vote for?” That respondent used to answer for, “I voted for Mr. X.” Either that person’s name, or she will name the symbol, or name the party. Later on we realized that this is not a very correct and honest response we are getting. Why? Because there are other people who are also hearing that. So suppose I am supporter of Congress Party, and I am being interviewed by you. You ask me this question, “To whom did I vote for?” My natural response would be, “I voted for Congress Party.” But I am not saying that. Why? Because there is one boy standing there whose father is the landlord of that village, or is the pradhan of that village, or is the leader of one political party. And he has come to me and asked me to vote for some other party, not the Congress. Suppose he asked me to vote for BJP. At that time, it was Jan Sangh. Even if I did not vote for that party, to communicate to that person, through that boy, that I voted for BJP, so that that boy will go and he will tell his father, “I was listening and he was asked this question, ‘To whom did you vote for?’ Papa, he did vote for BJP.” So instead of answering my question, he was answering and satisfying that person. We used to get wrong information.

So there were two changes in our survey technique that were introduced. One, asking this question, not to respond openly, but we created a kind of dummy ballot box and dummy ballot paper. Asking them to mark your vote here — don’t tell me, don’t show me — and put it here. And the other was that since the results are announced, there is a kind of bandwagon effect on the respondent. To reduce that, we started doing pre-poll studies. But by learning this technique, the two general elections were already gone. In ‘71, we didn’t do it. In 1980 election, we didn’t do it. But later on, in the 90s, we started doing this pre-poll — pre-poll means pre-results — and through ballot paper and ballot box.

SS: That wasn’t introduced until the 90s.

VBS: And that’s why the answers used to be biased in favor of the winning party, at that time.


On sampling methodology, and CSDS studies during the 1970s and 1980s

SS: How did you do the sampling for the ‘67 survey, and how did it change from ‘71 to ‘80?

VBS: Sampling actually, we have not deviated a lot. Because that was quite perfect during that period.

In 1967, we selected constituencies based on their competition typology. The different types of competition between the parties. The competition between parties where Congress, which used to be considered a kind of centrist party, where there is a right-wing party as well as a left-wing party — triangular contests. The constituency where there is only Congress and right-wing parties. The constituencies where there was only left-wing parties. It was a direct contest, or triangular contest, or four-cornered contest. Since the survey was being done after the results were announced, we knew what the different political parties got and in what type of contest typology that constituency [fell]. That was applied only in 1967 and through that we are not getting the real representation of our whole universe. Because certain kinds of constituencies fell only in a few areas. So that area used to get less representation.

Soon after ‘67, in ‘71, we made a kind of very universal sampling design in which we decided how many interviews would be good enough for the country. So that was at that time 2,500 to 3,000 responses, would be good enough to do a kind of analysis at the national level. Then how that had to be distributed. At that time, there was 400 some odd constituencies. It was enhanced in 1974 to 543. So this many constituencies are there. So we had to select this many parliamentary constituencies, about — I don’t exactly remember the percentages of those constituencies — suppose it was decided, 15% of parliamentary constituencies will be studied. So depending on the size of the state and other things, 15% of constituencies were selected. So from some states, twelve constituencies. I think it was more than 20%, because a state like Uttar Pradesh always got more than fifteen constituencies. From smaller states, two parliamentary constituencies. From middle-level states like Kerala, four constituencies. Like that. You decided that this many constituencies would be studied from this state, this many from this state.

Then how [were] those fifteen constituencies from Uttar Pradesh selected? For that selection of the parliamentary constituencies was done on PPS [probability-proportionate-to-size] basis. You take the electorate size of all the constituencies, get the cumulative total, and then divide it, get the mean, mode, median like that. Then you select one — you always allow a larger chance to the larger constituencies. So once the parliamentary constituencies were selected, we decided to select two assembly constituencies from that parliamentary constituency.

SS: So just to be clear, the assembly constituencies are located within the parliamentary constituencies.

VBS: Each parliamentary constituency, two assembly constituencies.

SS: This was at that time, in 1971.

VBS: Yes. To select those two constituencies was again the same basis.

SS: PPS, random starting point with a step.

VBS: Yeah. Total of that parliamentary constituency, divide it by two, then draw a random number from the first half and add the half to that and get the second constituency.

SS: So they typically had more than two assembly constituencies in the parliamentary constituencies. And you would pick two by doing PPS.

VBS: By doing this PPS. Then from each assembly constituency, again two polling stations. And those two polling stations again on the same basis.

SS: On the same basis. Most of them are pretty much the same size, but you’d still do PPS.

VBS: The larger polling stations used to get a little larger chance of being found in the sample.

SS: How did you get all this data on the size of the parliamentary constituencies?

VBS: Parliamentary constituency and assembly constituency data were available at the national level. Below that, one had to go to the field. By that time, you know which are the assembly constituencies, which are the parliamentary constituencies you are going to cover.

SS: And you got this data from the Electoral Commission, or the Census?

VBS: The electoral data was Election Commission. We used to go to the Election Commission here for the parliamentary and the assembly constituency [data]. And once the assembly constituency was selected, then to select the polling station we had to go to the district headquarters of that assembly constituency. And there the polling stations were selected. At that time, once the polling station is selected, we used to take the whole list of voters at that polling station and draw a random number.

SS: And so the lists were also at the district headquarters. The list of every registered voter in the polling station.

VBS: Yeah. Team leaders’ work used to be going in advance. Somebody [went] to the district headquarters first before the team arrived. By that time, you are ready with the sampling of polling station and the list of respondents whom you are going to interview. So that list used to be prepared in advance before the survey team reached that area. So that would be handed over to that team, and the person would move to another district, make arrangements for the team’s stay — where they will come and stay and get the data, and how to reach the polling sample area.

The only difference which also was later introduced in [the] 90s, I think a little later — not the first survey but the second survey — earlier we used to say that the distribution of respondents [was] done based on the electorate size of the assembly constituency. Suppose a number has to be interviewed in Uttar Pradesh, and you have fifteen parliamentary constituencies. How [was] that number distributed? Based on their PPS size. So they will get their number, and that number has to be divided in assembly constituencies. That used to be done [with] PPS. Then how many respondents had to be interviewed from the polling station? That was also done on the basis on PPS.

Suppose one polling station is of an electoral size of 1,200 and the other is 700. So the 1,200 polling station will get a little more and this 700 will get a little less. Which was later made equal. Actually, thirty interviews needed to be done from each sampling point.

And then earlier, it was only a kind of random number. Suppose we have 1 to 700. So the three-digit column was taken from the random table. You select, you may get 413. That will be your first respondent. Then you get another one. That may be 102. Third may be 435. So you are from the same area getting this third respondent. There used to be a chance actually in that period that sometimes you got two respondents from the same household.

SS: Sorry, I didn’t follow. How did you draw the respondents from within the polling station? Did you do a random start and step?

VBS: Random sampling, but the random number, actually, have you seen the random tables?

SS: No. What are random tables?

VBS: [searches bookshelf] That book is not there. There used to be a random number table prepared. Where there is only figures. There would be actually two digit rows. So there would be three-one, two-one, one-three, zero-two, zero-five, seven-one, nine-one. All these random tables were prepared randomly by the machine. And that table used to be available. We only had to decide that, we had to draw from how many digits. If it is a kind of two digit number, then you have to have only first row. And then suppose the first number was 71. Second number was 03. Third number was 75. You go on doing that until you reach the end. The random number was prepared in such a manner that the same number cannot fall twice in the same poll.

But the danger was that to get two respondents from the same family at the same locality.

SS: What would you do then?

VBS: We used to interview them. But later on we realized that there is some flaw in this technique.

What to do then? Then we decided that we had to interview thirty persons from this polling station, and we had a sample size of 900. 900 has to be divided by thirty, so you get a quotient of 30. You draw a random number, which is less than 30. Suppose you get 14. So the fourteenth will be your first respondent. Then keep on adding 30 to that. And you will get thirty in your list. Doing this, you get the representation of all the area of that polling station.

SS: When was the first survey that you used that?

VBS: I think this was done some time in… ‘95? I think it was done some time in ‘96 or ‘98.

SS: So for the early projects, ‘67, ‘71, ‘80–

VBS: That was a kind of random number where there was a danger of getting the proximal respondent in our sample and some areas were getting not properly represented.

SS: Because of the way you drew the respondents within the polling station. It seems like you did that circular sampling for the parliamentary constituency, then the assembly constituency, then the polling station. But then you didn’t do it for the respondent.

VBS: Not in the first two or three studies. ‘67, ‘71, ‘69 — ‘69, there was a kind of mid-term poll in four states.

SS: Which states?

VBS: Uttar Pradesh, Bihar, West Bengal, and Punjab. We did that ‘69 study as well. That was [a] mid-term poll. And then ‘71 study. Then ‘80. ‘77 was done — only a few constituencies were studied.

SS: There were a few studies in constituencies in ‘77?

VBS: ‘77, we didn’t do that survey research. It was a few case studies.

SS: I’m wondering about the timing of it. Why you chose to do it for ‘67, and then the mid-term [in] ‘69, and then ‘71, and then you didn’t really do a full study again until 1980. Why not do 1977?

VBS: Two things. One was the funds. And the other was a kind of lack of faculty support. Because I was the youngest in the first generation of researchers. So by the time we reached [the] 70s, this first generation of scholars, they lost interest.

SS: They lost interest? Really?

VBS: Not that enthusiastic, yes. But not energetic enough to undertake this kind of rigor of survey research.

Because that was quite, as I said, that was quite time-consuming, unlike today actually. [CSDS Director] Sanjay, suppose today he conceives that [he] has to do a nationwide survey, he can plan and do it, implement in seven days’ time. Because the team is already there. There are trained people, actually. You don’t have rush experts to train Tamil Nadu team or Karnataka team. There are local teams, people available. Every person of our network has some team of committed field workers who take charge and recruit how many are required, and they send and collect the data. And they deliver it. And the questionnaires are now in such a manner that they will send the data directly to you.

Earlier it was not possible. As I said, you had to prepare the questionnaire. And then you had to quote the data, check the data, clean the data, get it punched, get it on the cards, take it to the computer. The computer will take so much time. There used to be a long queue in the computer. Your data will be processed tomorrow. So you have to wait. And then tables used to come. Sometimes there need[ed] to be a kind of merging of various categories. Somebody had to sit and manually add two columns together and then work out the percentage once again. It was quite time-consuming.

So they thought that they had got sufficient information to theorize, with the support of whatever they had got. Number one.

Number two, I was still young and I was not that competent to theorize, so I needed to divide this activity. So they used to say, “Okay. Go ahead. Get it done.” I did try but at that time, actually, in 1971, the Indian Council of Social Science Research (ICSSR) was already founded by 1971. It was founded some time in 1969. So they funded a large number of election studies in 1971. They gave grants, small, big, to various professors located in different universities. They got the project and they never submitted the report. So ICSSR decided to not fund any more.

SS: Because they just gave money away and didn’t get anything for it. Professors just took the money.

VBS: Yeah. So the funds were not available. At that time I was not that competent enough and known enough to get funding from some other sources. So the funds were not available. That’s why in 1977 I got a kind of very small grant from Indian Council of Social Science Research so we had four case studies. One in Kanpur, one I did in Azamgarh, one my colleague did in Muzaffarnagar. In 1986, we got a small grant to do a study of Haryana assembly elections. But not a large-scale survey project because the money was not available.

SS: It was in ‘80 though? In ‘80, you could get it?

VBS: ‘80, actually, we didn’t do directly. In 1980, we didn’t get a direct grant. De Costa, actually, [of] the Indian Institute of Public Opinion (IIPO), he managed the grant from, I think, ICSSR or some government [body]. But he didn’t have the expertise, so we helped them. IIPO got the grant [and] did the study, but we helped them to prepare their instrument, sampling, and team. The promise was that, “You’ll [IIPO] get the money, we’ll [CSDS] save the data. We will get access to the data.”

And then the first major study we did was in 1995, the Bihar assembly election study.


On the establishment of Lokniti, and survey research in the 1990s and 2000s

SS: How did that come about? In ‘80, IIPO did the study. The last one CSDS was 1971. So what happened after twenty-four years, that you decided to do a study of Bihar?

VBS: The interest was always there actually. Yogendra [Yadav], he was spotted by DL Sheith in one of the seminars in the Institute of Advanced Study, Shimla. He was making one presentation there. At that time he was teaching in Punjab, Chandigarh. DL Sheith was very much impressed by his presentation there. Then he invited him to make another presentation here in CSDS. He made [a] presentation here and then all of those people who were there at that time persuaded him to join us.

SS: You saw him and knew, “We want this guy.”

VBS: So he joined. He joined and by that time he had already seen my writings and my books were there on elections. So he said, “Singh saheb, let’s start once again.” I got a kind of good support in him, then we got a kind of small proposal for this Bihar study. And it was granted. How much money? The quantitity was 1 lakh, 11 thousand… or 91 thousand? Less than 1 lakh rupees. It was a post-poll survey. So we went there to Bihar, and the poll was to take place two or three days after that. We thought that we would be training our investigators two days in advance. They will take two days to prepare themselves to go to the field. By that time, the election would be over. At that time, the idea was to do a post-poll but before the results were declared.

We went there and we completed the training. The second day morning, when we were to take our investigators to have field experience, the news came that the elections had been postponed. Then we were really in a fix. Subrata Mitra was there, Yogendra was there, myself, and Sanjay [Kumar]. Then we sat together. What to do? We decided that suppose we ask our investigators to go home and then again come and do it after the poll, then we don’t have that kind of money. So what to do? I don’t remember who was the first to forward this idea: why not make this [a] pre-poll survey? Instead of asking, “whom did you vote for?” let us ask, “For whom you are likely to vote supposing the elections take place tomorrow?”

So we jumped on this idea, and a few questions needed to be altered. Other questions, any time they can be asked. Before the election, they can be asked after the election. But about the decision-making, and what happened during the voting day period, those kinds of questions were altered [to include] “suppose the elections are going to take place tomorrow.” And we did it. We sent it out. And we ourselves also got two days’ time. We roamed around. Somewhere we went to drop our team members to different places and came back. Talking to the people.

And that was going to be over Lalu Prasad Yadav. He was fighting for the second term. By that time actually, the Bihar intelligentsia, the opinionated people, the journalists, everybody had already written off Lalu.

SS: They thought he was going to lose.

VBS: In 1995. That he is going to be defeated. And we used to be ask people, “How will he be defeated? Because there is no single front against him.” People used to argue, those who are informed: “No need to align against him. The voters have made up their mind to defeat him and they will choose their own candidate to defeat Lalu. They have already made up their mind to vote Lalu Prasad Yadav out, so whoever candidate they find in the constituency competing, they will vote for that person.”

When the data came, actually, then we got a fair amount of idea that Lalu — although he had entered into some kind of understanding with Left parties, some seat-sharing — people were saying that at the most he will be able to get a kind of minimum majority and with the support of the supporting parties he will be able to form the government. Our data showed that he is going to get single-handed the majority. In this month — this was February — we were deliberating what to do. We had got the data, but none of us, we didn’t have that amount of courage and we had never done it advance. And we didn’t want at that time to publish that Lalu is going to win before the elections were completed.

SS: Because this was such against conventional wisdom. Everyone thought he was going to lose.

VBS: Yeah. But we trusted our data. We had very good faith in whatever we had been able to see. We were very confident that Lalu is returning to power. But we couldn’t get that courage to publish it before the elections. One, we didn’t want in the eyes of the Election Commission that we ha[d] done something wrong. And the other, we didn’t have that kind of courage also.

But we couldn’t hold it. The day the counting was scheduled, we gave handouts to three, four newspapers and only two of them published. One is Indian Express, the other was The Statesman. Times of India didn’t publish and Hindustan Times also didn’t publish. And we said, “We have done such survey, this and that, this is the sample size, this is the conduct, and this is the result.” And in the morning, the counting started and those people who read the newspaper started phoning us: “You people have also turned out to be admirers [of Lalu].” There is one saying at that time, Lalu chalisa. It means, those who are praising Lalu like Hanuman. “You people also started praising Lalu Prasad Yadav.” We said, “No, actually our data is showing that.” They said, “This time your data is going to fail you!”

By the afternoon, the trends were available and the same people started phoning back. “Sorry, your data seems to be right.”

And soon after that, a fortnight after that, there was a director’s meeting at the ICSSR. All directors of different institutes in a year, once or twice, they meet in ICSSR. Sometimes in Delhi, sometimes outside. At that time, Jiram Reddy was the chairperson of the ICSSR. By that time this report came in the paper and the results were there. Jiram Reddy asked me, “Why don’t you give the copy?” I made an almost nine-page article, “Why did Lalu win?” The elections were staggered in four, five phases. And every phase forward, Lalu’s chances improved. This has happened this time [in the 2015 elections] also. He asked [for] the copy. I sent that. He got something around one hundred copies of that article and he circulated [it] to relevant places. In that director’s meeting, while he was speaking to the different directors that use public money and decide what kind of research we used to do: “Here is the example. Dr. Singh has done this study.”

SS: So it was great publicity.

VBS: He was so… I knew him earlier also. He had visited us earlier, actually. He was also carrying out some study. For that, he came to Bashiruddin Ahmed and then took our help. So he was just kind of ga-ga about our findings and our quality of research. And I became a kind of darling of Jiram Reddy.

And at that time in that meeting during the lunchtime he used to talk to me. At dinner time, he spent time. And then he said, “Singh sahib, you do [a] similar study for the Lok Sabha elections.” I asked him, “Will it be possible for ICSSR to fund such a mega-budget project?” He said, “How much?” I said, “Something around 20, 25 lakh rupees.” He said, “That’s all?” I said, “Yes.” “Okay. Done! Start working on that.” And he said this thing in front of the members secretary.

SS: And this was the head of ICSSR.

VBS: After four, five months, then we started working on the proposal and other things. Yogendra Yadav really prepared a very good research proposal. The model research proposal, actually. He submitted that proposal to ICSSR.

It was going to be considered. Meanwhile, [Jairam Reddy] was in a meeting in Oxford University and there during the dinnertime, something happened, and he collapsed and he never returned. So we lost all support.

SS: [gasps]

VBS: But the member secretary, he was very loyal to [the] chairperson, because he brought him there. And our conversation took place in his presence. So he said, “Singh sahib, don’t worry. We’ll do it. But we are not able to fund the same amount.” But he funded something. 17, 18 lakh rupees.

Subrata Mitra also managed some funds and then we polled together and we did all sorts of studies in 1996. We did exit polls. We did pre-polls. We did post-polls. Because we had funds.

And thereafter, that exit poll was done for one Nalini Singh and associates who were working with Doordarshan, Delhi Doordarshan. [They] funded that. And we had that data. That time was the first time I saw [a] mobile telephone.That was double the size of this [holds up landline phone on his desk and drops it]. And very heavy weight. So we got most of the data, some of the data we used to get on telephone. Yogendra used to be sitting on the screen and we used to supply data to him for the deliberation. All of us, actually, one after the other, appeared. And that happened to be a 15,000 [respondent] sample. For all India!

And we did predictions and it was 99% correct.

SS: That was the election where–

VBS: 1996.

SS: Where the BJP got the largest vote and formed a government for thirteen days.

VBS: We went wrong only in Karnataka because of our sampling. In there, some tribal constituencies were overrepresented. The share of Karnataka should have only half the constituencies of [the] tribal belt, but we had two constituencies. I don’t know how that happened. So we went wrong in Karnataka, but nationally — since the sample was national[ly] drawn — we were very correct.

Thereafter we were so far in media and other thing[s]. And then we were very, very critical of ourselves [in] predictions of seats and doing research for media. It is not a kind of serious research. We collect data and we make presentations on electronic media, and we do some small writings for newspapers, and then we forget about our data. So it [was] a very great loss of our academic caliber.

We should disassociate ourselves from the prediction of seats. But we have to monitor and do the surveys of all elections so that our database for all these stages are available, so that we can understand the political processes operating at different points in time. Based on that, actually, we can very easily — if someone is serious to do — we can predict what is going to happen in the next ten years.

That should not be last. Therefore, we keep on doing surveys but not for seat predictions. But for [understanding the] parameters changing, considerations changing, support bases of different political parties changing, what that change indicates and why that change is taking place, whether it has moved or going back. All these kinds of things. That interest is there. Since we don’t do predictions — we are happy enough to do post-polls but before the election [results are announced] — we get the real picture.

Apart from this, still there has been surveys… there have been Ashis Nandy’s survey of displaced persons in the party, and block-level planning projects, and some other projects. But when Lokniti was constituted in CSDS, that was the period when CSDS was going through a very, very critical financial constraint. For doing these election-related surveys and other things, the funder used to approach us. We [didn’t] have to go the funder. So There was a large amount of money available. Ashis Nandy at the time was the director. He said, “Today we have a large amount of funds for this. Tomorrow we will not have. So let us have a kind of separate programme for these electoral studies.” They decided that we should have an Institute of Comparative Democracies. That was the English name, and the Hindi name was Lokniti. Later on we decided that it should not be Institute for Comparative Democracy but [a] Programme for Comparative Democracy so that we can block this money for the rainy days. Suppose we don’t get a grant. There should be something we can invest from our own. And therefore there was a need of a separate programme and [to] block this money for some institutional purpose.

Any more questions?

SS: Yes. How it evolved after that, after the 1996 National Election Study and the formation of Lokniti– It’s become a very regular thing now. For every state election, every national election, CSDS does an election study. It seems like it’s shifted between doing some pre-polls, and then they did exit polls, and they had a relationship with CNN-IBN and they were doing projections for CNN-IBN. How that whole process unfolded… It seems like there’s been a kind of constant evolution in CSDS’ mission and its approach to survey research. It seems like it’s something that CSDS is still even now trying to develop.

For this last Bihar election, they did a pre-poll study but they also did a post-poll study. I remember when I first arrived here talking to Sanjay Kumar about it. And he was saying, “We used to do seat projections, but we don’t like to do them anymore.” Because it’s exactly as you were saying. It’s not as scientific a process, and you’re kind of playing this media game where if you project it and you get it right, then people think you’re a genius. And then you get it wrong, even if the principles behind it were good, people are going to criticize you.

In terms of your own involvement, are you involved with the surveys these days? It sounds like you were very involved in the 1990s and the reconstitution of survey research. What was your involvement from the end of the 90s going into the 2000s with Lokniti and surveys here?

VBS: I am still with Lokniti. I am still with Lokniti. I have been a very active partner of Lokniti. Formally, I retired in 2006. Until that, I had a very, very formal role to play in Lokniti. We are partly to take a collective decision with regard to what to do and what not to do since, at one point in time, fund-raising was kind of our goal.

You see, actually, why we fell into that media trap. Because we wanted to get some money. Because at that time this IDRC funding was not available and ICSSR was very, very minimal, only supporting up to 60% or 70% of our salaries. So to pay the salaries and get some money for our in-house research, somebody who is not working on projects or not on funded projects. If that person wants to go abroad or go to some other library in some other state, from where can that money be got? That was the kind of consideration we wanted to [address], and we were able to sell ourselves better at that time. So the funds were available.

But at the same time actually, there was always a kind of unease among the collectivity. What is the difference between CSDS and the other media, these poll companies? They do it, and we do it, and then there is a kind of bickering in the institution. If we say something favorable for one party or unfavorable for another party, our people say that, “You are pursuing this kind of ideology.” Or: ”You have preference.” Or: “You are misrepresenting. This is not the collective thinking of the institution.”

This conversion of votes into seats in India is really very, very difficult because [of] the complexities of electoral politics. Competition typologies. Regional parties. There are some parties which cover only three districts in a state. And you are doing a pre-poll survey. For seat predictions, suppose somehow that that area doesn’t come into the [sample]. So predicting vote is scientific as far as the Indian situation is concerned. Converting that vote into seats has a lot of problems. So chances of your being right and chances of your being wrong is almost 50/50.

Suppose there is a kind of direct contest, as happened in Bihar, it was very easy. And this has happened in the history of Bihar since 1952, until date, no election was as directly contested as this time. 1952 onwards, we always have a three-cornered, four-cornered [contest]. Independents in different constituencies. And you see here, the chances of independents winning elections was quite high. [It] used to be that one-third, one-fourth of candidates selected in the house were independents. But later on they had been marginalized. They had been sidelined by the elected [officials] saying, “Independents can’t play any role, any constructive role in [the] house.”

So a large amount of discussion took place in our Lokniti group to restrict ourself to telling about the mood of the people, but not converting them into seats. Not spending too much time on this rigmarole of survey research. We should do what we require for our understanding of the subject. So that our understanding is not hampered by the lack of data to analyze and understand the democratic institutions, and the problems confronted by them in this country, and where we are heading, and what problems we confront. So for that whatever is required, we should do. But we should not enter into this kind of entertainment competition of different media. Number one.

Number two, [because of] the kind of method we apply in our research it is very difficult to get that kind of funding from this media. Our surveys, actually, we spend something around, not less than 500 rupees per interview. And these media people, they give them results with 120, 125 rupees per interview. Because they don’t have to spend so much, they will just increase [the] number only on computers. They will do interviews of 10,000 people and they will say they have done it for 60,000 people based on that. They have their network of commercial people in different areas. They will select people. They will send them. Say, “Go to this district, to this constituency. From the headquarters you go three kilometres west and interview one village. And you go five kilometres south to that headquarters, and visit another village there and do thirty interviews. And you are paid per interview completed.”

If we sent our people with a randomly selected list of names and ask[ed] them to meet people and interview only those persons, if out of thirty you are able to meet only three persons — you turn back with an explanation for why you have not been able to interview twenty-seven — we are more than happy. It won’t affect your payment. We don’t pay the type of investigator… we recruit through their teachers. So for them, it was not only a type of earning proposition, it was there for learning. They come here to learn. Only at the end of the day would we take care of their expenses and give some money at the end so that they can have some pocket money. Saying, “I spent one week outside my class or my home and I have three thousand, four thousand rupees in my pocket.”

The same survey, if they want to get it done from us, that will be five times, three times costlier than that. So they’re not ready to do [it]. But still there are some people, some media, they still want [us] to do [it]. But we say that we can’t predict the seats. And for them, actually, the greatest entertainment is only to cover the votes-into-seats. That made the change, when we stopped our pre-poll surveys. Because in pre-poll, whatever stage you do — suppose you do it three minutes in advance — even then you will be asked to predict. Predicting seats based on a survey conducted three months earlier, and then what happens during that three months — God knows — but some people will remember that, “You said that the BJP will get this much seats in Bihar. But they got only 53.” If they are not aware of what you have done. So we have slowly kept ourselves out of that and we are happier for that. We have no complaints. It was a collective, wise, well-deliberated decision.

SS: When was that decision?

VBS: This was… You see, actually, we decided not to do that prediction but we helped them to get one. Rajeeva Karandikar, located here in Bangalore, he used to get data and used to do something, and used to convert that. But people were not able to differentiate between Rajeeva Karandikar and CSDS because the survey was done by CSDS, Rajeeva Karandikar made the prediction, but Rajeeva Karandikar’s name is lost. So it happened from 1998, ‘99, 2004… I don’t remember whether [it was] 2009 we stopped it, or ‘11. It was decided, I think, three general elections before. It must be 2004, or ‘09. But this you can ask [someone else].


On reflections

SS: So you’ve worked with CSDS then during the 60s, 70s, 80s, 90, 2000s, 2010s. How has the public profile of surveys and quantitative methods as a means to study politics in India changed — or not changed — over the decades?

VBS: There is not much qualitative change, but in terms of approaches, time and energy spent, with the media exposure, technological development, means of transportation, these things have made surveys much easier, number one.

Number two, I would say that this younger generation who are engaged in survey research, there [are] some sort of shortcut methods. Very few survey researchers who are doing major writings based on survey data have much exposure to the pre-poll. Earlier, the people who used to analyze data — it was always analyzed by the technical hands — but use that analyzed data and give meaning to that, used to spend some time in field to understand the nerves of the people.

Number three, you save survey data for long-term theoretical considerations and reflecti[ons]. That kind of research. We had very, very great hopes from Yogendra Yadav. We were expecting something will come from Yogendra. What Rajni Kothari did in the 60s and 70s, Yogendra will do in the 2000s, the 21st century. That kind of dissatisfaction. The kind of writings we are seeing in the form of reports and other things, that depth is missing. We are sitting on a mine of data and I am pained to say that we are not able to use more than 75% of the data which we have collected.

SS: Why not?

VBS: Because this is the kind of thing, there is a lack of strength in our CSDS as well, apart from Sanjay. There are hardly any persons actually. Earlier there used to be at least more than three persons, senior-level, using this empirical data and doing some long-term work. But Sanjay has too [many] tasks. He is now holding the post of director of this place, so there is institutional responsibility. Elections are too frequently coming here and there. So that kind of research is missing.

SS: More theoretical, long-term, systematic looks at the data across elections, rather than one election study after the other.

VBS: Suhas Palshikar. We requested [to] him several times, “Why don’t you come here? Spend a few years just concentrating on using the data.” We need that kind of thing. Yogendra has already gone into active politics.

That is the kind of problem, or sadness, that we face. I am ready to say [it is] my responsibility, but not able to create that amount of time, but now actually we have money, we have time. But [we need] to find the right kind of people.

The change that has taken place in the current generation of academicians, no one has time to do this: long-term using cross-sectional data for two decades, three decades, digging out information. It takes time. If you are working on this, you are not going to produce a book after six months. You are going to produce a book after three years. That kind of patience in the younger generation is missing. There we have some expectations and hope from people just like you [laughs].

SS: Thank you very much.


2 thoughts on “VB Singh on survey research in India since the 1960s

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s