Yup. In social science (IIRC, I wish I could find the reference), three conditions much exist to infer causality: 1) Temporal precedence, that is A must precede B in time, 2) correlation, and 3) (IIRC) explanation that confounds have been considered and (statistically) rejected.
Of course, the best way to demonstrate causality is empirical test with a treatment and control group, but this is rarely possible in social sciences, and in studies like this would be downright unethical.
Thus, we're stuck trying to determine causality, and correlation is a strong tool to help us get there.
Of course, the best way to demonstrate causality is empirical test with a treatment and control group, but this is rarely possible in social sciences, and in studies like this would be downright unethical.
Thus, we're stuck trying to determine causality, and correlation is a strong tool to help us get there.