Per-module ingestion notes

Prepared for the NORC/ANES requests. Item counts and exact item selections are not determinable without Online Appendix E, which is paywalled (HTTP 403). Everything below about module content comes from the paper’s Table 8; everything about format comes from general knowledge of GSS/ANES instruments and should be verified against the GSS variable index before ingestion.

Summary

Module	Wave	Response format	Ordinal as-is?	IRW handling
Science, religion, spiritualism	1988 GSS	Likert agreement	yes	standard
Occupational prestige	1989 GSS	1–100 rating	see below	needs decision
Music tastes	1993 GSS	like/dislike scale	yes	standard
Economic attitudes	1996 GSS	Likert agreement	yes	standard
Mental health causes	2006 GSS	likelihood rating	yes	standard
Job ideals	2016 GSS	importance rating	yes	standard
Government spending	2016 GSS	too little / about right / too much	yes (3-cat)	standard, but see note
Trust in institutions	2016 GSS	confidence rating	yes (3-cat)	standard, but see note
Political attitudes	2012 ANES	mixed	verify	verify

Occupational prestige (1989) — the one that needs a decision

Respondents rate occupations on a 1–100 prestige scale. This is a continuous or near-continuous rating, not a Likert item, and it is the module that made their Table 9 an outlier (OverallSD 2.288 against 0.691–1.584 for everything else; VarColPC1 47.99 against 1.34–11.31).

Three things follow:

IRW has precedent. The eligible-table scan found tables at n_categories = 101 already in the archive (test_taking_much_2025_cm, climatechange_geiger_2025). So a 1–100 response is representable under the existing standard — this is not a new structural problem.
It should not be collapsed to a coarser scale. Its structural distinctiveness is precisely what makes it valuable in a corpus dominated by 4–7 category Likert items, and binning it would destroy that.
It changes what “item” means. In the other modules an item is a survey question. In the prestige module, the rated occupation is the item and the prestige score is the response. Worth stating explicitly in the table description so users don’t treat it as an attitude battery.

Three-category modules (government spending, trust in institutions)

Both use 3-point response scales. These are ordinal and ingest cleanly, but note they fall below the n_categories >= 4 cutoff used in the eligibility scan for RCA/CCA work — the reasoning being that on very short ordinal scales the person-by-person similarity matrices become badly tied. They are perfectly good IRW tables; they would just not be eligible for that particular analysis.

ANES 2012 political attitudes

The paper describes this as a mixed module of political attitude items. Response formats across ANES batteries vary considerably (7-point ideology scales, feeling thermometers at 0–100, binary items). Verify per-item before ingestion — a feeling thermometer and a 7-point agreement item should not be pooled into one table without noting the mix, and under the IRW standard (contribute.qmd, point 4) responses from distinct scales should be split into separate files anyway.

Access terms are separately unresolved — see 02_anes_2012_note.md.

General notes for all GSS modules

Missing-value codes must be stripped. GSS uses negative and reserved codes for don’t-know / refused / not-applicable. The IRW standard is explicit that missing values cannot be coded as numbers (contribute.qmd, point 1). These need converting to NA, not carried through as responses.
One table per scale. Where a wave’s module spans more than one distinct instrument, split it (contribute.qmd, point 4). This is also what would make the modules merge-eligible under the DOI-keyed grouping used elsewhere in this scouting run.
Respondent ids must be stable within a wave if any cross-module merging is intended. GSS respondent ids are wave-specific; do not assume comparability across waves.
Item text is available for GSS and would be worth including via the itemtext pipeline, since the modules are well documented.