Analytical Frameworks for Studying Elections in India

Every five years (and often more frequently at the state level), millions of voters in the world’s largest democracy walk to polling booths and quietly reshape the political landscape. But how do scholars actually make sense of this enormous exercise? Counting votes tells us who won. Understanding why people voted the way they did, and what their choices reveal about society, requires something more systematic. This is where analytical frameworks come in. Political scientists have developed distinct methods to study elections, each offering a different lens on the same complex reality. Two of the most influential approaches are survey research and ecological analysis.

Why we need frameworks to study elections 🔗

Elections are far more than a contest for seats. They are an expressive medium of democratic participation, a moment when citizens communicate their preferences, anxieties, and aspirations through the act of voting. Studying them helps us track the health of democracy itself: how representation works, how social groups engage with politics, and how power shifts over time.

The challenge is scale. Indian national elections are the largest electoral exercise on the planet, and analysing them is genuinely difficult given the size of the country and its population. A single Lok Sabha constituency can contain millions of voters spread across diverse castes, religions, languages, and economic conditions. No single observation can capture this. Frameworks give researchers structured, repeatable ways to collect evidence and draw conclusions, rather than relying on guesswork or impressions.

Broadly, election studies divide into two families based on the kind of data they use. Some rely on aggregate data, such as official results compiled at the constituency or district level. Others rely on survey data gathered from individual voters. As the Institute of Developing Economies notes in its review of election studies in India, this division is useful precisely because each type of data serves different analytical purposes.

Survey research: listening to the individual voter 🔗

Survey research is built on a simple but powerful idea. If you want to know why people vote a certain way, ask a carefully selected, representative sample of them. By collecting data on voter profiles, attitudes, and reported choices, researchers can analyse broader democratic trends with statistical confidence.

This method is the speciality of psephology, the statistical study of elections. The technique travelled to India from American political science, where scholars at the University of Michigan had pioneered the National Election Study model. Indian researchers adapted it to local conditions, and over the decades it has become the dominant academic approach to understanding voting behaviour.

The role of the Centre for the Study of Developing Societies 🔗

No institution is more central to this story than the Centre for the Study of Developing Societies (CSDS) in Delhi. It was founded in 1963 by the political theorist Rajni Kothari, who had trained in survey methods in the United States. CSDS was one of the first Indian organisations to venture into election studies, conducting its first survey-based study of an election in 1965 during the Kerala state assembly polls.

The CSDS team went on to study the general elections of 1967, 1971, and 1980, before a gap and a revival in the 1990s. Today its data unit holds one of the largest archives of survey data on political behaviour and attitudes outside Western Europe and North America, spanning nearly five decades. Crucially, it also maintains election results for all national and state elections held since 1952, giving it a rare combination of both survey and aggregate data.

Lokniti and the National Election Study 🔗

In 1995, a team at CSDS created Lokniti, a network of scholars based across Indian states who study democracy and elections. Lokniti runs the National Election Study (NES), a large-scale nationwide survey of voters’ opinions and attitudes conducted during Lok Sabha elections. The series has been unbroken since 1996, with studies in 1996, 1998, 1999, 2004, 2009, 2014, and 2019, building on the earlier surveys of 1967 and 1971.

The NES is designed as a scientific study of the political behaviour, opinions, and attitudes of the electorate. One of its notable methodological innovations was the use of a dummy ballot paper and dummy ballot box to ask the sensitive vote-choice question, helping respondents answer more honestly. The survey also records a wide range of background variables to document the social profile of each respondent, and the questionnaire is carefully translated into all the major Indian languages. Parallel State Election Studies conducted during Vidhan Sabha elections extend the same rigour to state-level politics, capturing local issues that national surveys might miss.

The value of this work lies in what it reveals. As one analysis describes, voter surveys have illuminated how caste, class, gender, and region shape political preferences, and they tracked the decline of the Congress system and the rise of regional parties and a BJP-led alternative. The data has become so foundational that most serious studies of Indian politics now draw on it in some way.

Strengths and limits of survey research 🔗

Survey research has a clear advantage: it captures individual attitudes directly. The voting preference of the individual is the mainstay of electoral democracy, and opinion surveys are exceptionally good at measuring those individual attitudes. They let us ask not just who won but why, connecting votes to reasons.

The limits matter too. Surveys depend on a representative sample and a well-designed questionnaire. A poorly framed question or a skewed sample can distort findings. There is also the problem of over-reporting, where respondents claim to have voted for the eventual winner. And because pre-poll survey data is collected before counting, it has sometimes been pulled into the business of forecasting seats, which is a different and riskier exercise than academic analysis. The Lokniti pre-poll for the 2015 Bihar assembly election, for instance, estimated a lead for one alliance that the actual results reversed.

Ecological analysis: reading patterns in the aggregate 🔗

The second major framework takes a step back from the individual. Ecological analysis works by correlating electoral results with other kinds of aggregate data, such as census figures on the social and economic characteristics of an area. Instead of asking individuals about their choices, it studies areas rather than individuals as the unit of analysis, looking for relationships between the way a region votes and its underlying social composition.

The scholar most associated with advocating this approach in the Indian context is Paul R. Brass, an American political scientist who spent his career studying the politics of India. In works such as his 1985 study of caste, faction, and party, Brass combined electoral data with regional socio-economic profiles to understand the relationship between electoral performance and social groups. His analysis of Muslim electoral politics in Bihar from 1952 to 1972 is a classic example of using aggregate data to study how a particular community engaged with elections over time.

How ecological analysis complements survey data 🔗

Ecological analysis is at its most powerful when combined with survey research rather than used alone. Survey data tells you about individuals; aggregate data tells you about places. When a researcher overlays the two, layering survey insights onto the socio-economic profile of a region, the result is a more comprehensive view of the political process than either could offer separately.

This combined approach helps explain phenomena that pure survey work might overlook. Social divisions in India are extremely local, and parties build support from the village upwards through door-to-door campaigning. Aggregate data captured at a fine geographic level can reveal these local concentrations of support that broad national surveys smooth over.

The danger of the ecological fallacy 🔗

Ecological analysis carries one famous risk that every student should understand: the ecological fallacy. This is the error of assuming that what is true of a group is automatically true of the individuals within it. If a district with many members of a certain community votes heavily for a particular party, it is tempting to conclude that members of that community voted for that party. But this need not be true; the votes could have come from other residents of the same district.

Researchers have noted that the enormous size of Indian parliamentary constituencies makes this risk especially serious, because so much variation is hidden when data is observed at a high level of aggregation. This is one reason ecological correlation has not caught on in India as widely as survey research, despite the availability of good census data. Working at finer levels, such as the polling-station or village level, reduces the danger but demands far more detailed data.

Putting the frameworks together 🔗

It is a mistake to think of these frameworks as rivals where one must win. They answer different questions. Survey research excels at explaining individual motivation and mapping attitudes; ecological analysis excels at detecting structural patterns across regions and linking them to social context. The richest election studies use both, and institutions like CSDS are valuable precisely because they hold both survey and aggregate data under one roof.

There is also a deeper intellectual lineage worth remembering. Rajni Kothari, who built the survey tradition at CSDS, also gave Indian political science some of its most enduring concepts, including his theory of the Congress ‘system’, which reframed the Indian National Congress as a system of accommodation rather than an ordinary party. This reminds us that frameworks for studying elections are not just technical tools. They shape the very categories through which we understand democracy. Choosing how to study an election is, in part, choosing what questions about democracy we think are worth answering.

What do you think? If survey data and aggregate data sometimes point to different conclusions about why a region voted the way it did, which should a researcher trust more, and why? And as media-driven forecasting increasingly overshadows academic psephology, do you think election studies risk losing their original purpose of explaining democracy rather than predicting it?

References

1. https://www.ide.go.jp/English/Publish/Reports/Dp/098.html
2. https://www.lokniti.org/page/accessing-data
3. https://www.lokniti.org/national-election-studies
4. https://swarajyamag.com/politics/when-numbers-become-narratives-how-politicised-polling-is-manufacturing-a-credibility-crisis
5. https://www.cambridge.org/core/journals/american-political-science-review/article/abs/minority-electoral-politics-in-a-north-indian-state-aggregate-data-analysis-and-the-muslim-community-in-bihar-19521972/C0FF9722CF296BA581CF037E5BBBE9DC
6. https://www.nature.com/articles/s41597-025-05418-6
7. https://jan.ucc.nau.edu/~sj6/Kothari%20Congress%20System.pdf