Dissuasion and polarization vs. persuasion in nano-targeting: experimental evidence from climate policy messaging

Micro-targeting is entering a new era: interacting a large set of voter characteristics with multiple message sub-components -- "nano-targeting'' -- will become increasingly cheap and accessible. We combine evidence from two survey experiments (N=5,000) and predictive modeling to test the effectiveness of nano-targeted messages on climate policy. We find that, on average, nano-targeting does not perform better than an overall "best message'', but that nano-targeted messages were more persuasive than the best message for individuals who were already supportive of climate action. We also discover backlash effects from "mis-targeting'', and find an important asymmetry: being randomly assigned messages predicted to be far from one's best policy has stronger dissuasive effects than the persuasive effects of being shown one's best message. Together these results suggest that these technologies may not be practical for most actors, but may be effective for entrepreneurs who aim to dissuade, demobilize or polarize.