Secondary analysis is the use of data that someone else collected for some other purpose, and documentary research is the analysis of texts and records produced independently of the research. The two overlap and are usually treated together, since both confront the same fundamental condition: the researcher did not control the production of the evidence and must therefore reconstruct the circumstances under which it came into being before it can be used.
The range is wide. Official statistics include the Census of India, the National Sample Survey, the National Family Health Survey, and vital registration of births and deaths. Administrative records cover land records, school and hospital registers, police First Information Reports, court judgments and programme databases. Archives hold state papers, colonial reports, gazetteers, missionary and company records. Personal documents include diaries, letters, autobiographies and, now, digital traces. Mass media supply newspapers, pamphlets, films and broadcast material. And existing survey datasets, increasingly deposited for public reuse, permit re-analysis with new questions. Documents may be primary — produced by participants at the time — or secondary accounts compiled later, and solicited or unsolicited.
Scott's four criteria
John Scott proposed the standard framework for appraising any document. Authenticity asks whether the document is genuine and of unquestioned origin — whether it is what it purports to be, by the author claimed, and free of later alteration. Credibility asks whether it is free from distortion and error, given what the author knew, believed and wanted the reader to conclude; a village official's report of programme coverage is written for an audience that rewards compliance. Representativeness asks whether the surviving documents are typical of the class they belong to, and whether the sample available is complete or the residue of selective preservation. Meaning asks whether the document is intelligible — whether the researcher can recover both the literal sense of its terms and the interpretive context in which its categories signified something to the people who wrote and read it.
Advantages
The gains are substantial and are the reason so much sociology rests on such data. Scale: no individual researcher can generate a dataset comparable to the Census of 1.4 billion people, or an NSS round covering more than a hundred thousand households. Time depth: comparison across decades or centuries is possible only through records, which makes documentary work indispensable for studying social change, and hence for the historical and comparative method. Low cost: the expensive stage has already been paid for, which opens serious research to scholars without funds and permits replication and re-analysis of others' claims. Unobtrusiveness: because the data were not produced in reaction to a researcher, they escape the interviewer effects, social desirability bias and observer effects that afflict fieldwork. And documents are stable, in the sense that they can be consulted repeatedly by different analysts, which supports reliability of coding in a way that a single field encounter cannot.
Hazards
Every advantage carries a matching danger, and the root of all of them is that the data were produced for another purpose. The researcher's concepts and the compiler's categories rarely coincide, so operationalisation must be forced into whatever the source measured; questions the source did not ask cannot be answered, and information about the fieldwork that generated the numbers is usually thin.
More seriously, official categories embed the assumptions of the state. The colonial census is the classic illustration: from 1871 the enumeration of caste required that fluid, local and contextual jati identities be fixed into standard, countable and rankable categories, and — as Bernard Cohn and Nicholas Dirks have argued — the act of enumeration itself contributed to hardening those identities and to the emergence of caste associations that mobilised around census classification. Statistics here do not merely record a social fact; they participate in making it.
Systematic undercounting follows the same logic. Women's work has long been underestimated in Indian labour statistics because unpaid family labour on the farm, in animal husbandry and in home-based production fits poorly with a definition of employment built around wage work, so that recorded female work participation fluctuates with definitions and reference periods as much as with reality. The informal economy, which employs the overwhelming majority of Indian workers, is by its nature poorly captured by establishment-based enumeration. Vital registration misses deaths and births in exactly the populations most at risk. Survival bias compounds all of this in archives: the literate, the propertied and the official leave records, while the illiterate, the poor and the itinerant do not, so an archive is a sample selected by power.
Durkheim's suicide statistics
Emile Durkheim's Suicide (1897) is the founding demonstration and the standing warning. Working entirely from official statistics of European states, he showed that suicide rates were stable over time, varied systematically between countries, religions, marital and occupational categories, and rose or fell with political and economic conditions — evidence that rates were social facts requiring social explanation rather than the sum of individual pathologies, and the ground of his theory of egoistic, altruistic, anomic and fatalistic suicide.
The critique, developed most fully by Jack Douglas, attacks the data rather than the theory. A suicide statistic is the end point of a social process of classification: a death is discovered, reported, investigated, and categorised by coroners, police and officials who apply local conventions and respond to the wishes of families. Where suicide is shameful, or criminal, or bars religious burial, deaths are recorded as accidents, and the very variables Durkheim used as explanations — religion, integration, community concern — plausibly affect the recording of suicide as much as its incidence. Catholic communities may conceal more, not kill themselves less. The lesson is general and applies to Indian farm suicide data as much as to nineteenth-century Europe: an official statistic must be treated as a bureaucratic product whose rules of construction are themselves an object of enquiry.
For the UPSC answer
Open by defining secondary and documentary sources and listing the range, from Census and NSS to archives, diaries and newspapers, so the breadth is on the page. Use Scott's four criteria as your evaluative framework — authenticity, credibility, representativeness, meaning — since a named framework structures the answer. Balance scale, time depth, low cost and unobtrusiveness against the central hazard that data made for other purposes carry the state's categories, illustrating with the colonial census and caste and with the undercounting of women's work and the informal economy. Close with Durkheim's use of suicide statistics and Douglas's reply, which converts the whole discussion into the sociological point that official data are socially constructed.
References & further reading
- Scott, J. (1990). A Matter of Record: Documentary Sources in Social Research. Polity Press.
- Durkheim, E. (1897). Suicide: A Study in Sociology. Free Press (1951 translation).
- Douglas, J. D. (1967). The Social Meanings of Suicide. Princeton University Press.
- Cohn, B. S. (1987). An Anthropologist among the Historians and Other Essays. Oxford University Press.
- Dirks, N. B. (2001). Castes of Mind: Colonialism and the Making of Modern India. Princeton University Press.
- Bryman, A. (2012). Social Research Methods. Oxford University Press.