The survey method gathers the same items of information, in the same form, from a large number of people, so that the answers can be tabulated and analysed statistically. Its defining feature is standardisation: every respondent faces an identical stimulus — the same question, in the same wording, in the same order — on the assumption that only then are differences in the answers attributable to differences among the respondents rather than to differences in how they were asked.

Two instruments dominate. The questionnaire is self-administered, whether on paper, by post or online. The structured interview has an interviewer read a fixed schedule and record the replies, which raises the response rate, permits probing within limits, and works where literacy cannot be assumed — a decisive consideration in much Indian fieldwork. Both differ sharply from the unstructured interview, where the wording and sequence are allowed to vary with the conversation.

Census and sample survey

A census enumerates every unit in the population; a sample survey studies a selected subset and infers to the whole. The census yields figures for the smallest administrative units and settles disputes about totals, but it is slow, hugely expensive, and for that reason can carry only a short and simple set of questions. A sample survey can ask far more, ask it better, employ trained investigators and return results within months — at the price of sampling error, which is, however, calculable when the sample is drawn by probability methods. The paradox noted by survey statisticians is that a well-designed sample is often more accurate than a complete count, because the errors of an enormous field operation (mis-enumeration, coverage gaps, poorly trained enumerators) are larger than the sampling error of a good sample.

Steps in a survey

The sequence is conventional and examinable. First, the research problem is formulated and its concepts operationalised into observable indicators — social class into occupation, education and income; religiosity into attendance, belief and practice. Second, the population is defined and a sample design chosen. Third, the schedule is drafted: question wording, response categories, sequence, filter and skip patterns.

Fourth comes the pre-test or pilot, administered to a small number of respondents resembling the target group, to expose ambiguous wording, missing categories, questions that offend and an interview that runs too long. Skipping the pilot is the commonest avoidable failure in student research. Fifth, fieldwork: recruiting and training investigators, assigning workloads, and supervising through back-checks and spot re-interviews. Sixth, editing, coding and data entry, then analysis — univariate distributions, cross-tabulation, tests of significance, multivariate models — and finally the report.

Open and closed questions

Closed questions offer fixed alternatives. They are quick, cheap to code, and comparable across respondents, but they force answers into the researcher's categories and can manufacture opinions that the respondent did not hold before being asked. Open questions let respondents answer in their own words, which preserves meaning, reveals unanticipated frames of reference and suits exploratory work, but they demand skilled recording, expensive post-coding and are vulnerable to differences in articulacy. Most schedules mix the two, using open questions early to explore and closed questions to measure. Question craft matters more than beginners expect: double-barrelled items, leading wording, loaded terms, double negatives and long recall periods each generate error of their own.

Strengths and limitations

The strengths are breadth, comparability and inference. A survey can cover a whole country, produce the same measurement for every respondent, and — if the sample is probabilistic — attach a confidence interval to an estimate about a population of hundreds of millions. It is the only practical means of establishing distributions and trends, and it makes secondary re-analysis by other researchers possible.

The limitations are equally structural. Surveys tend to superficiality: they record opinions and reported behaviour, not the processes and meanings behind them, and the respondent has no chance to explain what a question meant to them. Response bias takes several forms — social desirability, acquiescence, prestige-seeking, faulty recall, proxy answers given by a household head on behalf of women and juniors. Non-response is more damaging than a small sample, because non-respondents are rarely a random subset. And in a multilingual country translation is a substantive problem, not a clerical one: caste, community, work, household and even income have no clean equivalents across Indian languages, and back-translation is a minimum safeguard.

Surveys in India

India runs some of the world's largest social surveys. The Census of India, a decennial enumeration with an unbroken series from 1881, remains the backbone of demographic and social classification. The National Sample Survey, established in 1950 on the initiative of P. C. Mahalanobis and the Indian Statistical Institute, pioneered large-scale sample survey practice — including the interpenetrating subsample design for estimating non-sampling error — and supplies the consumption, employment and unemployment data on which poverty estimation rests. The National Family Health Survey, conducted by the International Institute for Population Sciences since 1992–93, generates district-level data on fertility, health, nutrition and, increasingly, women's autonomy and domestic violence. Sociologists use all three heavily, while remembering that the categories in the schedule are administrative constructions with histories of their own.

For the UPSC answer

Define the survey by standardisation rather than by size, and distinguish the questionnaire from the structured interview at the outset. Set the census against the sample survey explicitly, noting that a good sample can beat a bad complete count, then take the examiner through the steps with the pilot study named, since it is the step most answers omit. Weigh breadth, comparability and statistical inference against superficiality, response bias, non-response and translation, and use the NSS, Census and NFHS as your Indian illustrations. A closing line on Mahalanobis lets you claim, accurately, that India shaped survey methodology and did not merely import it.

References & further reading

  1. Moser, C. A., and Kalton, G. (1971). Survey Methods in Social Investigation. Heinemann.
  2. Payne, S. L. (1951). The Art of Asking Questions. Princeton University Press.
  3. Groves, R. M., Fowler, F. J., Couper, M. P., Lepkowski, J. M., Singer, E., and Tourangeau, R. (2009). Survey Methodology. Wiley.
  4. Babbie, E. (2010). The Practice of Social Research. Wadsworth.
  5. Mahalanobis, P. C. (1946). Recent Experiments in Statistical Sampling in the Indian Statistical Institute. Journal of the Royal Statistical Society, 109(4), 325–378.
  6. International Institute for Population Sciences (2021). National Family Health Survey (NFHS-5), 2019–21: India Report. IIPS, Mumbai.