|HOME | NEWS | SPECIALS|
The Rediff Special/Dr N Bhaskara Rao
To survey or not to survey
Staple newspaper fare at election time, poll surveys have suddenly become suspect with the body politic. Last month, most parties demanded that they be outlawed by the Election Commission. In this fascinating feature, one of India's leading pollsters outlines the scenario behind election surveys.
Thirty years ago I wrote in my book, Politics of Leadership in an Indian State, about how public opinion sets the base for leadership and moulds that very process. My concern then -- as a sociologist -- was with perceptions and group dynamics. It was a couple of years later that I turned to survey research methodology to enlarge the scope of such inquiries into preferences and practices as a political analyst.
Until the mid-1960s, the concern of such surveys in the national context was limited mostly to four metropolitan areas by the doyen of public opinion surveys in India, Eric da Costa. It was on my return from the United States 25 years ago with a Ph D in communication research that I implemented national level public opinion surveys using country-wide field infrastructure developed by a premier market research agency. Soon after, opinion surveys on public affairs became a focus of market research firms.
Pre-poll surveys became a serious activity. The credit for elevating the status of pre-poll surveys belongs to the late Girilal Jain of The Times of India, Aroon Purie of India Today and Ramoji Rao of Eenadu. I had the privilege of discussing surveys and sampling with each of these eminent editors two decades ago and of convincing them of the relevance and objectivity of pre-poll surveys. It is these very publications that have sustained public opinion surveys in India over the years. More recently, since 1994, the state-owned Doordarshan has been a leader in featuring the findings of pre-poll surveys.
Opinion polls, in the context of national elections, are perceived as indicators of not just the dimension of victory or defeat, but also the attributes for differences and margins at micro and macro levels. Generally speaking, such an exercise reflects at best the trend, mood, and direction at any given point in time. Of course, by using past voting trends from previous elections, one can go much further.
Opinion polls are not always sponsored by the media with wide scope, although they would certainly be more relevant. Such rolling polls are not always feasible for newspapers to sponsor because of the resource constraints involved. India Today is perhaps an exception. Opinion polls sponsored by newspapers were traditionally meant to bring to the fore the mood of the people objectively. The statistical accuracy of surveys has become a concern only recently.
A poll survey should indicate at the outset whether it is meant to capture opinions and beliefs for preferences and practices or value and perceptions. A combination of these, if accommodated, will make the findings more relevant to the Indian scene.
In the Indian situation, an opinion poll conducted much earlier than the announcement of the actual poll schedule would neither elicit the same kind of response nor would it be able to attain the degree of reliability as a poll conducted after the kick-off of the campaign. Theoretically, an agency engaged in periodic opinion surveys on voting intentions at various critical points would be in a better position to identify the undercurrents and differences on the eve of an actual election and, thereby, make better projections on the outcome.
Public opinion surveys on the eve of polls, which started as a hobby and sideline activity of a couple of academicians 35 years ago, have come a long way. Starting from a sample size of 500 voters in the early 1960s and increasing to 10,000 by the 1970s and 1980s, samples have grown more recently to include 20,000 to 30,000 participants on the eve of national elections.
Today, India easily has 20 to 25 agencies across the country that are seriously involved in conducting pre-poll surveys nationally and regionally. From a budget of less than Rs 10,000 per survey, the budgets for some of today's surveys is as high as Rs 1 million. In 1996, more than Rs 10 million was spent in conducting pre-poll opinion surveys on the eve of national and state level elections, reflecting how pre-poll surveys have become business propositions for market research agencies.
Unlike earlier surveys which were sponsored primarily by newspapers, most of today's surveys are commissioned, directly or indirectly, by political parties, accounting for two-thirds of the spending. In the last few general elections, the leading national political parties sponsored more than one round of pre-poll surveys, as well as exit polls.
Impact of Opinion Surveys
Results of barely one-third of the 20 to 25 poll-eve surveys conducted across the country get into the media. One of the reasons for this trend is the reluctance of newspapers to sponsor, or even to publish, survey findings. The Election Commission's directive on the eve of the 1996 national poll -- advising newspapers against publishing exit poll findings where poll hours are staggered -- added to the apprehensions. Some surveys are not published because the results are not favourable to the sponsors or not consistent with the stand of the newspaper.
However, there is no evidence that pre-poll surveys have made a qualitative difference in the poll-eve scene in terms of the character of the campaign or the process of candidate selection. Pre-poll surveys also have not earned a reputation for influencing agenda-setting, nor is there any evidence that any political party has benefited from such surveys. The only exception so far is isolated: M Karunanidhi of the DMK, who had a landslide victory in the 1996 election for the Tamil Nadu assembly, publicly credited the CMS survey as one of two factors that contributed to his party's sweep.
With increased coverage of the findings of pre-poll surveys in the news media as part of their campaign coverage, these surveys have now become campaigns. It is no wonder, then, that three distinct influences of pre-poll surveys emerged from one of our 1996 post-poll validation studies.
Need for a code
Some motive-driven pre-poll surveys, like the one by an advertising agency in Andhra Pradesh in 1995, have led to controversy and undermined the very relevance of such surveys. The exit poll on the eve of the Maharashtra assembly poll in 1995 is yet another example of a similar kind of eroding of the very credibility of poll-eve surveys.
The onus of avoiding such instances of motive-driven surveys, however isolated they may be, lies with the media. Hence the need for a code for the media, particularly since all editors are not discriminating enough, nor do they have the background to judge research methodologies.
Writing in India Today and A&M more than 15 years ago, I called for a code of conduct for pollsters, as well as for newspapers publishing poll results, to minimise the possibility of polls misleading voters and contaminating the election scene. In fact, I have even suggested that the Election Commission should take the initiative to develop a common code for pre-poll surveys so that political parties do not resort to using them as part of their campaigns.
Overall unease with the quality of poll-eve surveys in India can be attributed to the use of inappropriate or inadequate methodologies in all phases of the surveys, including setting sample size, questionnaire design, interviewing practices, weighting, analysis, and, of course, reporting outcomes by both the research agencies and the media.
While the approach to sampling has, of late, become more systematic and more or less uniform among agencies, the actual selection practices vary considerably and include street corners, voter lists, different physical settings, and broad economic-criteria-based classifications. Street corner interviews, of course, cannot be expected to posses the same attributes and degree of reliability as interviews which are conducted at the household level.
Opinion polls often suffer on account of unexpected developments once the electoral process starts, such as the death of a political leader (as in the case of the late Rajiv Gandhi). Extended and rescheduled poll dates can also have a significant impact, as happened in the case of Orissa. I also remember a poll survey going haywire when essential supplies under PDS were blocked to many areas just prior to the poll dates for a state assembly.
Fragmentation (or otherwise) of political parties between any two elections, frequently changing alliances between parties (formal and informal), and the role of dissident candidates, and even delimitation of constituencies are other factors which contribute to the accuracy level of opinion polls conducted on the eve of state and national elections in India.
The recent round of poll surveys bring home the idea that small sample surveys can be good indicators of poll outcomes even in the Indian situation. For the second time in a row, independent CMS, with a sample of approximately 3,000 voters, was able to forecast accurate results better that the surveys with three-to five-times bigger sample sizes. Furthermore, an analysis of such polls conducted during recent years showed that market-research-based and commercially-oriented polls which tend to have larger samples, implying larger volumes of business, do not necessarily have more accurate results. Hence the relevance of an alliance with psephologists (those who study political elections) who have merged recently with more reliable and promptly-available data on past elections and their outcomes.
Also the recent performance of poll surveys by different agencies showed that a market research background alone is not enough to guarantee reliable and insightful poll surveys. The analytical skills needed for such surveys have to be interdisciplinary and more dynamic in light of the constantly changing political scene in the country and in poll practices.
Techniques of investigation
The gap between the field work day and the actual voting day also determines accuracy of the final results. The shorter the process of field work, analysis and publication, the more relevant the results will be. Also, when surveys are conducted in various parts of the country, the questions may carry different connotations in the various local languages and dialects. This is a problem that is, to a large extent, peculiar to India. The manner in which the questions are posed by investigators, therefore, is important.
The interaction of the person analysing the field data with field investigators can prove to be a major advantage. If the analyst, himself, is involved in the field work, it greatly enhances the credibility and accuracy of the analysis and avoids mistakes like ignoring the significance of no response or other such responses that occur quite often.
Poll results published with a clear byline for the research reflect a higher degree of confidence in the survey. This practice also ensures greater responsibility by both the publication and the researcher and helps achieve improvement in subsequent efforts.
In the Indian context, predicting the results of elections precisely can only be a coincidence. Like the contents of a boiling pot, election-time dynamics change continuously until the actual polls hours, unlike in more politically stable countries such as the United Kingdom or the United States. But opinion polls in India are not futile if they can identify the trend, the mood, and the direction of the outcome accurately -- as is true in all countries.
Dr Bhaskara Rao is founding chairman of two leading agencies, Centre for Media Studies and Marketing and Development Research Associates, and a pioneer in opinion research with three decades of distinguished background and experience.
INFOTECH | TRAVEL | LIFE/STYLE | FREEDOM | FEEDBACK