SLEZAK - InnovatingForPublicGood Nov 15 2023 1_09PM talk (7 min).pptx

1、1Messagingto drive supportJames Slezak,SwayableNovember 15,2023Responses10.8 millionAd treatments9,724Experiments2,461Agreement-Age Largest dataset in existence of high-quality RCT survey experiments on persuasionNOTE:Statistics include all experiments up to November,2023Each dot in the background i

2、s one complete response from 2018-2020 on election tests3American Political Science Review(Hewitt et al 2023,accepted)Important:In this presentation,Swayable is presenting its own analysis and conclusions and not speaking for the academic group.4Things our work has bearing on Is there a magic-bullet

3、 theory to win more votes with ads?-Is median voter theory(aka popularism)it?-Is advertising later better than early?What approaches have most promise?-What research do we need more of?-What is less useful?Some choices clearly worked better than others within a cycle,but no persistent patterns Presi

4、dential support metrics.ATEs reported at 68%confidence.*indicates statistically significant difference(95%confidence;p Liberal-Conservative-Very conservative-High correlation:r=0.67High correlation:r=0.86High correlation:r=0.87High correlation:r=0.68Results from presidential support metricsCorrelati

5、on in persuasiveness of adsCorrelation in persuasiveness of adsOlder voters were differentLow correlation:r=0.13Age18-24 25-34 35-5455+18-2410.940.720.1325-340.9410.850.3335-540.720.8510.7355+0.130.330.731Persuasive adBacklashing adNo measurable impact at 95%confidence levelSample:723,158Treatments:

6、696Results from presidential support metrics Some content persuaded both young and old However,a lot worked only with one or the other55+18249Sharp critiques of the right performed wellBetrayalDeceptionExtremismAngerSadnessDegradationHarmConcerningAverage adPersuasion,percentage points+2.02*0.28+2.0