Record
How the score worksWhat controlled tests found

617 campaign ads, each tested against a control group

Swayable archive · 107 randomized experiments, 2018 and 2020
FavorabilityVote choiceVote choice (yes/no)
+1.38 ptsAverage ad
1.09 ptsSpread between ads
83%Ads above zero
0%Of a new study explained by style
Average ad
+1.4
One spread above
+2.5

Points of favorability.

0%of the next study explained, fitted on all the others

So every script is a test to run, and the chart below is a prior, not a rule.

helps unclear hurtsVolunteer
Helps · 2
  1. Explains factsVolunteer+0.60 ±0.5
  2. Offers a fact the viewer likely hasn't heardVolunteer+0.60 ±0.4
Unclear · 26
  1. High production valueVolunteer+0.57 ±0.6
  2. Anger+0.56 ±0.6
  3. Fear+0.53 ±0.7
  4. Plain-folks appealVolunteer+0.43 ±0.6
  5. Sounds scripted+0.38 ±0.5
  6. Cites a source for its facts+0.35 ±0.4
  7. Negative tone+0.35 ±0.4
  8. Policy facts+0.34 ±0.4
  9. Humor+0.26 ±0.9
  10. Negative testimonial+0.20 ±0.5
  11. Everyday people as messengersVolunteer+0.12 ±0.4
  12. Politician as messenger+0.02 ±0.5
  13. Somber+0.02 ±0.6
  14. Explicit ask to vote forVolunteer0.00 ±0.5
  15. Personal storyVolunteer-0.02 ±0.5
  16. Celebrity as messenger-0.06 ±1.4
  17. Animation-0.07 ±0.8
  18. Contrast-0.17 ±0.5
  19. Veteran as messenger-0.18 ±1.1
  20. Positive tone-0.28 ±0.4
  21. Amateur, phone-shot styleVolunteer-0.28 ±0.5
  22. Positive testimonialVolunteer-0.29 ±0.4
  23. Enthusiasm-0.34 ±0.4
  24. Pride-0.36 ±0.7
  25. Advocates an opinion-0.37 ±0.5
  26. Text on screenVolunteer-0.45 ±0.9

Points of favorability, ads with the feature against ads without, with the 95% interval. Random-effects meta-regression, one feature at a time. Hewitt, Broockman, Coppock, Tappin, Slezak, Coffman, Lubin, Hamidian. How Experiments Help Campaigns Persuade Voters. APSR 2024. Dataverse doi:10.7910/DVN/LBPSSV (CC0).