Step 1: PTG Interim Summary — Promising Belief Targets
Nicole Cooper
2026-05-22
Overview
N = 300 (Cohorts 1–3 combined). This document summarizes findings from a Hornik & Woolf (1999) percentage-to-gain (PTG) analysis examining which beliefs are most associated with four behavioral intentions related to government funding of university research, and integrates belief network centrality (EBIC-GLASSO, 58 nodes, N/p = 5.2, 11 communities) to prioritize among similarly-ranked targets. It is intended as an interim working summary, not a final analysis.
Methods note
PTG = (rate of “Definitely will” among those at the desired belief position) − (overall rate of “Definitely will”). Predictor dichotomized at the top response if positively correlated with intent composite, bottom if negatively correlated.
Reliability threshold: estimates based on fewer than
30 participants at the desired position are flagged ([!]).
None of the targets in this summary trigger that flag.
Ceiling flag: beliefs where ≥ 60% already holds the desired position are excluded.
Lower bound = PTG − 1 SE (floored at 0). A lower bound ≥ 5 on at least two outcomes is the working criterion for a “credible” target (shaded rows in the table below).
Centrality = eigenvector centrality from the EBIC-GLASSO partial correlation network (58 nodes). Higher centrality = more connected to other influential beliefs. Percentile rank shown in parentheses.
‡ From govtfund_6 battery (dropped from
network; α = 0.41). PTG valid but lower priority. † Not
included as a node in the belief network.
Cross-outcome summary table
Values are PTG (lower bound). Centrality column shows eigenvector centrality (percentile rank). Shaded rows meet the credibility criterion (lb ≥ 5 on ≥ 2 outcomes).
| Belief | Dir | 1 |
% to gain (lower bound)
|
Composite PTG2 | Centrality (pctile)3 | |||
|---|---|---|---|---|---|---|---|---|
| Talk to others | Share online | Oppose cuts | Support funding | |||||
| Govt funding benefits | ||||||||
| Results become publicly available | + | 13.4 (7.8) | 13.3 (8.0) | 8.9 (4.4) | 9.6 (4.9) | 11.3 | 0.37 (33%) | |
| Funded researchers held to higher standards | + | 12.5 (5.9) | 12.5 (6.3) | 12.3 (6.5) | 7.7 (2.4) | 11.2 | 0.08 (2%) | |
| Exciting research is sped up | + | 16.8 (10.7) | 11.7 (6.2) | 3.3 (0.0) | 5.6 (1.1) | 9.3 | 0.25 (14%) | |
| Institutional trust | ||||||||
| State public university system | + | † | 10.8 (5.0) | 10.2 (4.9) | 7.8 (3.1) | 10.2 (5.2) | 9.8 | — |
| Private colleges and universities | + | 12.7 (5.9) | 14.5 (7.9) | 3.3 (0.0) | 4.8 (0.0) | 8.8 | 0.51 (53%) | |
| Local / economic impact | ||||||||
| Local area would be negatively impacted by cuts | − | 12.9 (7.6) | 9.9 (5.2) | 10.1 (5.7) | 8.3 (4.0) | 10.3 | 0.47 (47%) | |
| boosts the economy | + | 11.5 (6.6) | 8.4 (4.0) | 4.9 (1.3) | 9.3 (5.2) | 8.5 | 0.81 (86%) | |
| Personal relevance | ||||||||
| People you care about | + | 22.9 (16.6) | 19.3 (13.4) | 7.8 (3.1) | 11.7 (6.6) | 15.4 | 0.67 (72%) | |
| Disadvantaged individuals | + | 14.0 (7.9) | 14.8 (9.0) | 7.2 (2.5) | 6.5 (1.8) | 10.6 | 0.30 (19%) | |
| Local community | + | 16.7 (10.4) | 12.3 (6.5) | 6.0 (1.4) | 7.0 (2.1) | 10.5 | 0.60 (58%) | |
| You personally | + | 14.4 (8.5) | 8.1 (3.0) | 8.8 (4.1) | 9.6 (4.7) | 10.2 | 0.64 (67%) | |
| Researcher integrity | ||||||||
| Researchers lie to maintain funding | − | 17.7 (10.7) | 17.6 (10.9) | 7.0 (1.8) | 6.3 (1.1) | 12.2 | 0.50 (51%) | |
| Money spent on less important things | − | 17.6 (10.2) | 12.3 (5.6) | 8.9 (3.1) | 8.3 (2.5) | 11.8 | 0.46 (46%) | |
| University researchers motivated primarily by profit | − | ‡ | 21.1 (14.6) | 16.9 (10.9) | 4.1 (0.0) | 5.1 (0.5) | 11.8 | — |
| Taxpayer money is wasted | − | 11.3 (5.1) | 15.3 (9.2) | 4.9 (0.3) | 4.3 (0.0) | 9.0 | 0.65 (68%) | |
| Science salience | ||||||||
| I think about the impact of science on my daily life | + | 19.5 (12.9) | 11.4 (5.6) | 16.8 (11.0) | 19.6 (13.5) | 16.8 | 0.37 (30%) | |
| Clear how to apply research findings in my life | + | 13.0 (6.3) | 11.0 (4.9) | 10.7 (5.1) | 16.0 (9.8) | 12.7 | 0.41 (39%) | |
| University contributions | ||||||||
| Increased compassion and tolerance | + | 16.6 (11.0) | 14.3 (9.1) | 7.9 (3.6) | 7.2 (2.9) | 11.5 | 0.63 (63%) | |
| Better international relations | + | 20.1 (14.2) | 14.7 (9.3) | 2.1 (0.0) | 4.2 (0.1) | 10.3 | 0.59 (56%) | |
| 1 ‡ govtfund_6 battery dropped from network (α = 0.41). † Not a network node. | ||||||||
| 2 Composite = mean PTG across all four outcomes. | ||||||||
| 3 Eigenvector centrality from EBIC-GLASSO network (58 nodes). Percentile rank in parentheses. | ||||||||
Outcome-by-outcome highlights
Talk to others about cuts
| Block | Belief | Dir | PTG (lb) | ROR | n |
|---|---|---|---|---|---|
| Research Beliefs | People you care about | positive | 22.9 (16.6) | 6.52 | 66 |
| Govt Funding | University researchers motivated primarily by profit | negative | 21.1 (14.6) | 5.30 | 61 |
| University Perceptions | Better international relations | positive | 20.1 (14.2) | 5.90 | 74 |
| Research Beliefs | I think about the impact of science on my daily life | positive | 19.5 (12.9) | 4.57 | 58 |
| Govt Funding | Researchers lie to maintain funding | negative | 17.7 (10.7) | 3.71 | 49 |
The personal relevance cluster dominates. People you care about (PTG = 22.9, ROR = 6.52) is the single strongest target: proximate personal concern mobilizes interpersonal communication. Multiple research_3 items appear (local community, you personally, disadvantaged individuals), reinforcing a stakes/proximity framing.
Contact reps to oppose cuts
| Block | Belief | Dir | PTG (lb) | ROR | n |
|---|---|---|---|---|---|
| Research Beliefs | I think about the impact of science on my daily life | positive | 16.8 (11.0) | 9.31 | 58 |
| Govt Funding | Funded researchers held to higher standards | positive | 12.3 (6.5) | 4.82 | 51 |
| Research Beliefs | Clear how to apply research findings in my life | positive | 10.7 (5.1) | 4.00 | 50 |
| Govt Funding | Local area would be negatively impacted by cuts | negative | 10.1 (5.7) | 6.25 | 86 |
PTG values are lower overall — contacting representatives is a higher-effort behavior. Science salience (thinking about the impact of science on daily life) becomes the top target (PTG = 16.8, ROR = 9.31). Local impact also has a reliable estimate (lower bound = 5.7): connecting research cuts to local constituents may be particularly effective for motivating representative contact.
Contact reps to support funding
| Block | Belief | Dir | PTG (lb) | ROR | n |
|---|---|---|---|---|---|
| Research Beliefs | I think about the impact of science on my daily life | positive | 19.6 (13.5) | 11.14 | 58 |
| Research Beliefs | Clear how to apply research findings in my life | positive | 16.0 (9.8) | 6.26 | 50 |
| Govt Funding | The general public | positive | 13.9 (6.4) | 4.13 | 32 |
| Research Beliefs | People you care about | positive | 11.7 (6.6) | 4.97 | 66 |
| Higher Education | State public university system | positive | 10.2 (5.2) | 4.11 | 66 |
Science salience is the strongest target here as well (PTG = 19.6, ROR = 11.14). Clear how to apply research findings ranks 2nd (PTG = 16) — applied relevance framing may be particularly effective for motivating proactive support.
Network centrality
Eigenvector centrality indicates how well-connected each belief is to other influential beliefs in the network. A high-centrality belief is embedded in the broader structure and may cascade: shifting it could move adjacent beliefs as well. A low-centrality belief predicts behavior but is structurally isolated — it may respond well to direct messaging but is unlikely to produce downstream ripple effects.
The priority zone is the upper-right quadrant below: above-average on both composite PTG and centrality.
Quadrant interpretation
High PTG + High centrality beliefs are the primary targets: shifting them is both associated with behavior change and likely to propagate through the belief network.
High PTG + Low centrality beliefs are strong direct targets but structurally isolated: they predict behavior without being well-connected to other beliefs. Useful for direct messaging, but less likely to produce broader attitude ripple effects.
Key takeaways
Single best cross-cutting target: “Think about the impact of science on daily life” (composite PTG = 16.8%, lb ≥ 5 on 4 of 4 outcomes). However, its centrality is at the 30th percentile — it is a strong behavior predictor but peripheral in the belief network. It is a high-value direct messaging target, not a gateway belief.
Best combined target: personal relevance cluster (research_3). “People you care about” (composite PTG = 15.4%, 72th centrality percentile) sits in the priority quadrant: high PTG and moderately central. The full personal relevance cluster (items 1, 2, 3, 13) all fall in community 3 together, confirming construct coherence. Shifting personal relevance beliefs may cascade through the network.
Social behaviors vs. political action require different frames. Personal relevance drives talking and sharing; science salience and applied relevance (“clear how to apply research findings,” composite PTG = 12.7%) drive contacting representatives. A single message strategy is unlikely to move all four outcomes equally.
Researcher skepticism cluster is specific to online sharing and is highly central. “Taxpayer money is wasted” (centrality 68th percentile), “researchers lie,” and “money spent on less important things” cluster together in community 11 and are each other’s neighbors in the network. These beliefs form a coherent structural unit — but their behavioral payoff is concentrated in sharing behavior, not political action.
“Funded researchers held to higher standards” is the most outcome-consistent govt funding item (lb ≥ 5 on three of four outcomes) but has near-zero centrality — structurally isolated. Strong for direct messaging; unlikely to cascade.
Local impact is the most reliable political action frame. “Local area negatively impacted by cuts” has lower bounds ≥ 4% on all four outcomes and is the most actionable frame for constituent-targeted messaging.
Caveats
- Correlational estimates only. PTG identifies co-variation with intent; experimental manipulation is required to confirm causal effects.
- N = 300 limits per-outcome power. Lower bounds below ~5% should be treated as noise.
- Network N/p = 5.2 (58 nodes). At the lower edge of acceptable range; centrality estimates should be treated as directional, not precise.
- govtfund_6 items (‡) excluded from network due to poor coherence. PTG estimates included here but lower priority.
- research_4_6 (“Science makes our way of life change too fast”) appears in unexpected direction for political outcomes; dropped from network due to reverse-coding instability. Do not treat as an intervention target.
Generated 2026-05-22. PTG source: 04_promising_beliefs.csv. Network source: 05_belief_network.csv (N = 300, 58 nodes, N/p = 5.2).