Causal and measurement hypotheses cannot be tested simultaneously.

Evidence

One rests on the general principle that there is no recognized epistemological rationale for privileging causal hypotheses over measurement hypotheses. Both begin as conjectures, both must be conceptually coherent, and in the end both must meet certain empirical standards for acceptance. But, again as a matter of principle, both cannot be evaluated simultaneously; one must always be assumed while the other is tested. — Gary Goertz, Social Science Concepts, 906

Categorization

• XREF: Pairs with the earlier chapter’s point that a measure’s validity is a theoretical claim — measurement is theory all the way down, which is exactly why the two can’t be co-tested. — connects to: Overlapping measures of NPI and affective polarization make testing their theorized causal link impossible. — Sabrina J. Mayer and Luana Russo, What one is not: a new scale to measure…, 11
• SEED: Angle: ‘AHEM validation’ — how social science routinely assumes the measure and tests the hypothesis, and what quietly breaks when the measure is wrong.