Correlation is a statistical association: two variables move together, so that knowing one improves prediction of the other. Causation is the stronger claim that change in one produces change in the other. Every causal relationship shows up as a correlation, but the reverse does not hold — which is why the warning that correlation does not establish causation is the first rule of quantitative sociology.
How it works
Three conditions must be satisfied before a correlation supports a causal inference.
- Association. The variables must actually covary; a relationship that vanishes under measurement cannot be causal.
- Temporal order. The presumed cause must precede the presumed effect. Where the sequence is uncertain, causal direction is undecidable, and cross-sectional data often leaves it so.
- Non-spuriousness. The association must not disappear once other relevant variables are held constant.
A spurious correlation is one produced by a third variable acting on both. The classic form is the confounder: two indicators rise together only because urbanisation, income or education is rising behind both. Removing the third variable removes the relationship.
Why it matters
The practical tool is the control variable. By holding a suspected confounder constant — comparing rural with rural, or one income band with itself — the analyst tests whether the original association survives. If it holds within every category, the case for causation strengthens; if it collapses, the correlation was spurious. Emile Durkheim used a version of this logic in Suicide, eliminating climate and race before locating variation in the degree of social integration. Experimental designs settle temporal order by manipulation, but sociology usually cannot manipulate its variables and must rely on statistical control instead.
For the UPSC answer
Give the three conditions as a named list — association, temporal order, non-spuriousness — and illustrate spuriousness with a clear confounder rather than a joke example. Add that sociology's inability to run true experiments makes control variables and comparative method its substitutes, and note Durkheim's successive elimination of rival factors as the classical demonstration.
References & further reading
- Durkheim, E. (1895). The Rules of Sociological Method. New York: Free Press.
- Goode, W. J. and Hatt, P. K. (1952). Methods in Social Research. New York: McGraw-Hill.
- Babbie, E. (2010). The Practice of Social Research. Belmont: Wadsworth.