Skip to content
Benchmarks

Side by side with real FGIs and surveys

Each case here pairs one SSR study with a published FGI or panel survey on the same questions. Every setting on both sides — recruitment, weighting, fieldwork dates, original question text, Likert labels, the SSR seeds — is disclosed in the open. The point is not to claim a win. It is to make the comparison reproducible.

Coverage so far
Paired studies
18
Question pairs
98
Pairs to unlock headline r
202

Cross-study correlation, Top-2 box agreement, and Bland-Altman bias appear once at least 5 studies and 300 question pairs have been collected. Before that, every case is shown but no aggregate headline is claimed.

Catalogue shape mix: 34 multimodal-symmetric 31 bimodal-asymmetric 16 unimodal 14 left-skewed 3 right-skewed
Reuse the catalogue: CSV JSON all 98 paired questions across 18 fixtures in one file

Each shelf below collects studies that share an instrument and a topic. The Δ chip on each shelf reports the biggest mean-to-mean gradient inside that shelf — the SSR has to reproduce these gradients to be useful, so they are also the shelf's acceptance test.

Cross-shelf landscape
sorted by within-shelf Δ ↓

Which topics show the biggest internal disagreement? Below: every shelf with a computable Δ, plotted on the same 0 → max scale. A wide bar is a wide gradient the SSR has to reproduce; a short bar is a flat consensus the SSR must not over-polarise. The |s| column reports the shelf-mean |skewness|: ≥ 0.5 (amber) flags shelves where at least one item is bimodal or heavily skewed, so a re-run that matches the mean from a unimodal cluster is still wrong. σ (mean SD on the SSR axis) catches dispersion that |s| misses: a symmetric trimodal distribution — say, partyid spiking at strong-Dem / independent / strong-Rep — reads as skew ≈ 0 but its SD is large. ≥ 1.2 (sky) flags wider-than-unimodal shelves the |s| chip alone would let through.

Abortion attitudes →

Dichotomous yes/no on the seven AB conditions, GSS 2024. The within-battery gradient separates traumatic from elective scenarios.

Top Δ 1.59
4.65 → 3.06

Well-being (HAPPY + SAT*) →

Three- and four-point happiness and satisfaction items, GSS 2024. The within-battery gradient separates marital happiness and job satisfaction (high) from financial satisfaction (low) on the same wave and the same respondents.

Top Δ 1.31
4.15 → 2.83

Generalised social trust →

Three classic NORC social-trust items (TRUST + FAIR + HELPFUL) on the same wave and split-ballot, GSS 2024. The within-battery gradient is the gap between how trustworthy Americans think other people are (low) versus how helpful and fair they expect them to be (higher).

Top Δ 0.71
2.91 → 2.20

Political-policy trio →

Three dichotomous policy items spanning the U.S. partisan spectrum — death penalty, gun permit, marijuana legalisation — asked of the same 2024 wave. The shelf-Δ is intentionally narrow: all three sit at majority-favor, so an SSR re-run has to reproduce a flat consensus rather than a wide gradient.

Top Δ 0.25
3.74 → 3.48

Sexual-morality trio →

Three 4-pt sexual-morality items (HOMOSEX + XMARSEX + PREMARSX) on the same GSS 2024 wave. The within-battery gradient is the second largest in the catalogue (Δ ≈ 2.34) — Americans are overwhelmingly tolerant of premarital sex but overwhelmingly intolerant of extramarital sex, with same-sex relations reading as a sharply bimodal distribution that an SD/skew chip surfaces.

Top Δ 2.34
4.00 → 1.66

Political identity (polviews + partyid) →

Two 7-pt political-identity items on the same GSS 2024 wave — ideology (polviews) and partisanship (partyid). Both averages sit near the SSR centre, but the shapes differ sharply: polviews is unimodal at "moderate" while partyid is trimodal across strong-Dem / independent / strong-Rep. The shelf is the within-fixture case for the round-38 shape-mismatch finding.

Top Δ 0.11
3.12 → 3.01

Religion belief (postlife + god + pray) →

Three religion-belief items on the same GSS 2024 wave — afterlife (dichotomous), god (6-pt ascending), prayer frequency (6-pt descending). pray is textbook bimodal-asymmetric (31.5% several-times-a-day + 21% never, with only 6% at the once-a-week midpoint), god is left-skewed at "know God exists", postlife is left-skewed at "yes". The shelf is the round-38 finding rendered across three different scales.

Top Δ 1.02
4.29 → 3.27

Institutional confidence →

3-pt Likert across 13 U.S. institutions, GSS 2024. The same wave, the same respondents — the gradient is which institutions still earn confidence.

Top Δ 1.58
3.63 → 2.05

Federal spending priorities →

3-pt Likert across 23 federal-spending items, GSS 2024 — including the natfare/natfarey split-ballot that isolates the welfare vs assistance-to-the-poor wording effect.

Top Δ 2.53
4.38 → 1.85

Civil liberties (SPK + LIB) →

Dichotomous tolerance for speech and library books across four target groups, GSS 2024. Same wave isolates target × action interactions.

Top Δ 1.53
4.18 → 2.65

Hedonic taste tests (CPG) →

9-pt hedonic panel scored on four SKU per category in the Visalli et al. open dataset. Same panel, multiple products — the gradient is product-vs-product within one consumer.

Top Δ 1.65
3.63 → 1.97

Aquaculture sensory panels →

9-pt hedonic on aquaculture species across feed systems (Control / Insects / Algae / Bioactive), iFishIENCi D3.6. The gradient is which feed system the panel actually prefers.

Top Δ 0.60
4.10 → 3.50