Treating Likert responses as interval data distorts estimates when aggregating across disparate items
Evidence
The difficulties come from the ordinal nature of the Likert scale. For instance, the attitudinal distance between strongly disagree and disagree on one scale may be bigger or smaller than the one between agree and strongly agree. Not the mention differences across indicators. It actually causes statistical and interpretative problems to understand both estimand (Lundberg, Johnson, and Stewart 2021) and analysis unit in modeling, especially when one uses an additive scale using multiple Likert-scale items. — Ruiqian Li and Paul Froese, The Duality of American Christian…, 96
Categorization
- XREF: Speaks to a measurement concern underlying the reader’s question about whether ‘Christian nationalism’ is a genuine latent schema or a collinear proxy artifact of ordinal scaling. — connects to: Response acquiescence can forge fake statistical links between unrelated measures; One-directional survey items make scales falsely look valid and consistent — Bob Altemeyer, The Authoritarian Specter, 776; Identical survey responses can stem from opposite interpretations — Bob Altemeyer, The Authoritarian Specter, 3977 (+6 more)
- QUESTION: Worth checking whether this ordinal-distance critique indicates some of the collinearity with conservatism could be an artifact of crude Likert aggregation rather than substantive overlap.
Broader Connections
- Weighted sums in latent variable models silently convert ordinal indicators to interval.
#expands(0.92) — Gary Goertz, Social Science Concepts and Measurement, 832 - Unbalanced survey scales allow response bias to inflate apparent reliability
#expands(0.80) - Response acquiescence can forge fake statistical links between unrelated measures
#expands(0.78) - One-directional survey items make scales falsely look valid and consistent
#expands(0.78) — Bob Altemeyer, The Authoritarian Specter, 776 - Response bias makes balanced surveys appear spurious multifactor structures
#expands(0.78) — Bob Altemeyer, The Authoritarian Specter, 779 - Response acquiescence can fabricate correlations between unrelated survey scales
#expands(0.78) — Bob Altemeyer, The Authoritarian Specter, 775 - Scale items sharing a distinct thematic content inflate shared variance
#expands(0.75) — Sherman A. Lee, Chelsea A. Reid, Stephen D. Short, Jeffrey A. Gibbons, Ruth Yeh, and Matthew L. Campbell, Fear of Muslims, 65 - Identical survey responses can stem from opposite interpretations
#qualifies(0.72) — Bob Altemeyer, The Authoritarian Specter, 3977 - Strongly held beliefs reduce response-set bias in survey measures
#complicates(0.71) — Bob Altemeyer, The Authoritarian Specter, 816