Shared acquiescence can create spurious correlations between any two measures

Evidence

Furthermore, researchers noted long ago that unbalanced scales usually correlate best with other unbalanced scales (Hyman and Sheatsley, 1954; Campbell, Siegman, and Rees, 1967)—or indeed with anything potentially affected by yea-saying, including single-item questions. Shared acquiescence could thus create a connection between any two things, no matter what they (supposedly) measure. — Bob Altemeyer, The Authoritarian Specter, 775 (#corroborates, via linked claim)

The Zealot Scale, like my other 0-6 measures such as the Religious Emphasis Scale, does not control for response sets. So the interitem correlations and alphas on these scales are doubtless inflated, perhaps considerably. And their correlations with other unbalanced measures will also be pumped up. But their correlations with balanced measures such as the RWA, Religious Fundamentalism, and DOG Scales would be lessened, not increased. — Bob Altemeyer, The Authoritarian Specter, 4274 (#complicates, via linked claim)

Categorization

• XREF: connects to: Big data correlations alone cannot deliver causal knowledge to prediction systems — Shoshana Zuboff, The Age of Surveillance Capitalism, 930; Low internal consistency lets tests correlate with other measures for many spurious reasons; Response acquiescence can forge fake statistical links between unrelated measures (+6 more)

Broader Connections

• Response acquiescence can fabricate correlations between unrelated survey scales #corroborates (0.98) — Bob Altemeyer, The Authoritarian Specter, 775
• Response acquiescence can forge fake statistical links between unrelated measures #corroborates (0.97)
• Unbalanced survey scales produce inflated internal correlations but deflated cross-measure correlations #complicates (0.85) — Bob Altemeyer, The Authoritarian Specter, 4274