Nano-targeting or mass appeal, what makes persuasive climate communications?

@miriamsorace.bsky.social @lsedatascience.bsky.social, Thomas Robinson @lsemethodology.bsky.social, @simonhix.bsky.social @epssnet.bsky.social and @fresejoris.bsky.social @eui-eu.bsky.social for @lseimpactblog.bsky.social

@poalab.bsky.social
@uniofreading.bsky.social @lseimpactblog.bsky.social @eui-sps.bsky.social

Key Takeaways:
➡️ Nano-targeting isn't great for positive persuasion: broad appeals work best.
➡️ It's more useful for dissuading or attacking, by targeting people with what they dislike most.
➡️ It can help mobilize the ‘already converted’.
This explains why anti-system actors are most likely users.

Most-disliked messages had stronger dissuasive effects than most-liked messages had persuasive ones – suggesting nano-targeting may be more effective for dissuasion and de-mobilization than persuasion.

To quantify dissuasion effects, we turned to the control group respondents (N=1,000). By virtue of randomization, some individuals received – by chance – climate messaging that they personally disliked (mis-targeted), and some messaging that they liked (akin to the nano-targeted treatment).

Uniquely, our study does not simply test the persuasive effects of nano-targeting, but also its polarizing and dissuasive effects. We found that nano-targeting exacerbates polarization: nano-targeting was most persuasive among already pro-climate respondents and not among the undecideds/moderates.

This is great news for climate policy communication: we for example moved overall likelihood to sign an emission reduction petition to 52% with the best climate policy message, without the need for overly sophisticated message tailoring pipelines or to selectively hide policy pledges.

We find that – on all our measures of persuasion – the nano-targeted group **never** out-performs the best climate message outlined above (i.e. positive + energy security frames; tax incentives on sustainable buildings, energy and trade; investment in sustainable transport and farming)

For phase two (message testing), we partnered with @opiniumresearch.bsky.social to randomly assign respondents to: nano-targeted messages, the best-performing climate message, or a control condition where each individual received a bespoke but non-targeted climate policy message.

We then nano-target by training a predictive model (xgboost) on first-phase conjoint data, predicting petition support by interacting policy components with individual demographics, past vote and vote intention.

The calibration stage – run with @yougov.co.uk - shows that the best climate policy messaging has an optimistic frame and highlights energy security benefits, proposes tax incentives on sustainable buildings, trade and renewable energy and investment on the transport and natural investment sector.

We use climate policy to illustrate and test the technique in the strongest, most realistic, and most ethical way possible, given its multi-dimensionality, the strong heterogeneity in sub-dimension support, and a clear uni-directional goal (emissions reduction) seen as a public good.

We first illustrate nano-targeting as a technology, which calibrates messaging via conjoint analysis on natl representative samples and then applies eXtreme Gradient Boosting (xgboost) predictive modelling to assign the nano-targeted climate policy messages to individuals from a separate sample.

Micro-targeting is seen as persuasive weapon, but experiments shows it underperforms broad appeal messaging. We introduce and test ‘nano-targeting’ a deeper micro-targeting approach which is now increasingly scalable. Is this blend of micro-targeting and message sub-units tailoring more effective?