N = 300 (Cohorts 1–3 combined). Hornik & Woolf
(1999) percentage-to-gain analysis. Outcome: govtfund_1 —
approval of federal funding for university research. Outcome dichotomy:
“Strongly approve” (top of 4-point scale). Predictor dichotomy: top
response category if positively correlated with the outcome; bottom if
negatively correlated. Ceiling flag: ≥ 60% already at desired position
(e.g., items like “approve cuts to health research” where most people
already hold the desired belief). Demographics excluded from modifiable
beliefs.
Note on missing ROR values: 55% of participants in this sample already said “Strongly approve” — a substantial ceiling on the outcome. When splitting by a predictor, subgroups often have everyone (rate = 1.0) or no one (rate = 0.0) at “Strongly approve,” which makes the odds ratio undefined. The % to gain values are not affected and remain the primary metric here. ROR is omitted wherever either cell rate is at a boundary.
Included for reference. These items reflect prior behavior rather than modifiable beliefs.