Moral Reframing and Political Persuasion: Talking Across the Divide: Moral reframing and its replication record

In brief

Making an argument in your audience's moral vocabulary can shift attitudes in some controlled experiments, usually by a modest amount. It has not held up in the largest tests of its central premise: messages matched to a group's values do not reliably beat messages that are not matched. Where researchers measured behaviour, such as donations or mobilisation, the technique mostly produced nothing. The best field result came from conversations that uncovered what individual voters actually valued, not from inferring values from a category. Treat it as a real but conditional technique, not a general key to changing minds.

How to use this

Before commissioning a values-matched campaign, ask what your frame is being tested against. Compare a matched message with a mismatched one and a no-message control, rather than assuming the match is the active ingredient. Run the test in your own setting, with your own audience and issue, because the largest tests have not found a reliable matching advantage. Measure behaviour, not just attitudes; donations and mobilisation are where the technique has mostly failed. If you want the version with the strongest field support, invest in conversations that discover voters' values rather than inferring them from a category. Ignore the temptation to buy automated personalisation; it has not been tested at scale.

What the story is about

Moral reframing means making your case in the moral language your audience speaks, rather than the one you speak yourself. Say you want a conservative to back a climate policy. You can argue that it protects the vulnerable, which is your value, or that it honours tradition and keeps the country strong, which may be theirs.

Advocates rarely do the second. Across six studies, Feinberg and Willer (2015) found that "political advocates spontaneously make arguments grounded in their own moral values, not the values of those targeted for persuasion". Isiminger and Giner-Sorolla (2024) asked 249 US liberals to pick a message for a hypothetical conservative audience, and 73% chose the reframed one. Whichever they chose, though, they felt "more positive and less negative integrity feelings for the message that was not morally reframed". In plain words, the message that was not reframed felt truer to their own beliefs, so the move costs the speaker something.

In controlled experiments, moral reframing does move attitudes. Thomas et al. (2023) ran three online survey experiments with 1,888 US adults. They presented a universal basic income as a route to the conservative value of financial freedom. Support for the policy rose among conservatives by about a third of a standard deviation (0.36), with a likely range of roughly 0.27 to 0.46, a modest but real shift. These are attitude scales in an online experiment, not votes.

How much of a win that is depends on what you compare it against. Brugman (2024), a systematic review of 114 moral framing experiments, names "the choice of non-moral control group(s)" as a priority for future work. Kim, Hammond and Milfont (2023) measured a moral environmental message against a plain environmental one and found nothing. The studies that chose the harder comparison are the ones the field treats as most informative.

Two large tests have now looked for the pattern the technique depends on. Voelkel et al. (2026) pooled five replication studies with 3,216 people, then ran a very large preregistered test with 13,544 US participants, testing the ten most-cited climate messages. Six messages shifted preregistered attitudes, by one to four percentage points. Persuasiveness varied little across party lines, which is not what a theory of matched messages predicts. No message increased donations.

Hernandez, Ottone and Kalla (2025) ran an online field experiment with a real political organisation. Frames tailored to supporters' identity mobilised them no better than ordinary broad frames. Misfired targeting did not demobilise anyone either, that study found. The one positive behaviour result belongs to Spaiser and Nisbett (2025), a study of 156 people who reported their own daily habits, which is a thin base for any campaign claim.

Knowing what your audience actually values is the hard part. The first study pointed at that question is a preprint, so treat it as provisional. Reinero et al. (2026) ran three experiments with 3,455 people and report that "most liberals and conservatives shared the same top moral value: care and compassion". Guessing wrong about someone's top value backfired among moderates and conservatives. Earlier work found that people's guesses about the other side got the direction of the difference right but exaggerated its size (Graham, Nosek and Haidt, 2012).

That leaves the decisive experiment unrun. Nobody has compared a mismatched frame against both a matched frame and no frame at all in one design. So nobody knows whether a wrong guess is worse than saying nothing.

Another question is whether a moral frame beats simply giving people the facts. The two largest tests disagree. Kaplan et al. (2023) gave loyalty-framed messages about mask-wearing to 597 self-identified US conservatives, and the loyalty frame beat a purely scientific-information message. The change in belief lasted at least a week. Hackenburg et al. (2025) compared an instruction to use facts and evidence against an instruction to use moral reframing among 76,977 UK adults, and facts won. The designs differ in who writes the message, in country, in issue, and in whether the frame was validated in advance. Nobody has run the design that would reconcile them. No field test has yet taken place inside a candidate campaign.

Almost all of this evidence comes from the United States. The single direct replication outside the country failed: a German arm of 269 people, one half of a two-country failure (Voit, Twardawski and Fischer, 2026). So the technique has mainly been tested inside one national conversation.

There is also no pooled number. Brugman (2024) reviewed 114 moral framing experiments and reports no average effect size, and the wider reviews of tailoring messages to audiences mix moral values in with other traits. Ten years in, nobody can say how big this effect is on average.

Put the parts together and the picture is narrower than the promise. Moral reframing shifts attitudes in careful experiments on some issues. It has not survived the largest tests of its own premise, which is that a message matched to an audience's values works better than one that is not. Where behaviour was measured, it mostly did not move. The technique is real, and it is smaller and more conditional than the case for it suggests.

So what

Moral reframing is a real effect of modest size, not a general key to changing minds. That changes how you spend money. Paying to match messages to values is a bet on an unproven premise, and the largest tests suggest the premise is the weak part. Testing whether a particular frame works in your own setting is cheap by comparison. Do that before anyone builds a strategy on the technique.

For political parties

A party tempted by the cheap version of this should know it does not exist yet. Personalising messages at scale is easy now. Measuring a voter's moral values and then having a model write to them has not been tested at scale, which is the version everyone imagines.

So there is nothing to buy. Matching messages to values is a live research question, not a service with a track record. A campaign that tests its own messages against a plain alternative learns something either way. A campaign that tunes the wording until a poll agrees with it has bought a mirror, and mirrors do not warn you before election night.

For government

Governments run the largest public persuasion campaigns there are, so the temptation to mandate a favoured style is real. The evidence does not support that step. In the largest test that ranked strategies against each other, a plain instruction to use facts and evidence beat an instruction to use moral reframing, and the biggest message tests moved no one's behaviour. Treat values-matched framing as unproven rather than settled practice. If a department commissions it, the fair question is what the message is compared against. A frame that only beats saying nothing has not told you much. The test is whether the framing serves the decision or the conclusion the sponsor already preferred.

Case studies

The strongest field result moral reframing has comes from Kalla, Levine and Broockman (2022), and it did not come from guessing. Fifty-two canvassers had preregistered conversations with 1,034 voters. They listened for each voter's own moral values, then made the case in those terms. The result splits in an instructive way: "large effects on interest in taking action and some evidence of changes in policy attitudes". Planned Parenthood of Northern New England, the organisation behind it, treated the audience's values as something to discover in conversation rather than to infer from a category. That is the version the evidence supports, and it depends on conversation rather than automation.

References

Brugman, B.C. (2024) 'How the effects of emphasizing ethics are examined: a systematic review of moral framing experiments', Annals of the International Communication Association, 48(4), pp. 436–455. Available at: https://doi.org/10.1080/23808985.2024.2393845 (Accessed: 11 September 2026).

Feinberg, M. and Willer, R. (2015) 'From gulf to bridge: when do moral arguments facilitate political influence?', Personality and Social Psychology Bulletin, 41(12), pp. 1665–1681. Available at: https://doi.org/10.1177/0146167215607842 (Accessed: 11 September 2026).

Graham, J., Nosek, B.A. and Haidt, J. (2012) 'The moral stereotypes of liberals and conservatives: exaggeration of differences across the political spectrum', PLoS ONE, 7(12), e50092. Available at: https://doi.org/10.1371/journal.pone.0050092 (Accessed: 11 September 2026).

Hackenburg, K. et al. (2025) 'The levers of political persuasion with conversational artificial intelligence', Science, 390(6777), eaea3884. Available at: https://doi.org/10.1126/science.aea3884 (Accessed: 11 September 2026).

Hernandez, N., Ottone, N. and Kalla, J. (2025) 'Targeted abortion frames do not mobilize political action-taking', American Politics Research, 53(4), pp. 353–368. Available at: https://doi.org/10.1177/1532673x251324137 (Accessed: 11 September 2026).

Isiminger, A. and Giner-Sorolla, R. (2024) 'Willingness to use moral reframing: support comes from perceived effectiveness, opposition comes from integrity concerns', Social Psychological Bulletin, 19, e13053. Available at: https://doi.org/10.32872/spb.13053 (Accessed: 11 September 2026).

Kalla, J.L., Levine, A.S. and Broockman, D.E. (2022) 'Personalizing moral reframing in interpersonal conversation: a field experiment', The Journal of Politics, 84(2), pp. 1239–1243. Available at: https://doi.org/10.1086/716944 (Accessed: 11 September 2026).

Kaplan, J.T., Vaccaro, A., Henning, M. and Christov-Moore, L. (2023) 'Moral reframing of messages about mask-wearing during the COVID-19 pandemic', Scientific Reports, 13(1), 10140. Available at: https://doi.org/10.1038/s41598-023-37075-3 (Accessed: 11 September 2026).

Kim, I., Hammond, M.D. and Milfont, T.L. (2023) 'Do environmental messages emphasising binding morals promote conservatives' pro-environmentalism? A pre-registered replication', Social Psychological Bulletin, 18, e8557. Available at: https://doi.org/10.32872/spb.8557 (Accessed: 11 September 2026).

Reinero, D.A., Sinclair, A.H., Cosme, D. and Falk, E.B. (2026) Red, blue, and morally alike: moral reframing of climate change works when we don't make incorrect assumptions. PsyArXiv preprint, posted 25 June 2026. Available at: https://doi.org/10.31234/osf.io/a9wpd_v1 (Accessed: 11 September 2026).

Spaiser, V. and Nisbett, N. (2025) 'Mobilising climate action with moral appeals in a smartphone-based 8-week field experiment', npj Climate Action, 4(1), 81. Available at: https://doi.org/10.1038/s44168-025-00282-x (Accessed: 11 September 2026).

Thomas, C.C., Walton, G.M., Reinhart, E.C. and Markus, H.R. (2023) 'Mitigating welfare-related prejudice and partisanship among U.S. conservatives with moral reframing of a universal basic income policy', Journal of Experimental Social Psychology, 105, 104424. Available at: https://doi.org/10.1016/j.jesp.2022.104424 (Accessed: 11 September 2026).

Voelkel, J.G. et al. (2026) 'A registered report megastudy on the persuasiveness of the most-cited climate messages', Nature Climate Change, 16(2), pp. 214–225. Available at: https://doi.org/10.1038/s41558-025-02536-2 (Accessed: 11 September 2026).

Voit, M., Twardawski, M. and Fischer, M. (2026) 'The limits of moral framing in promoting pro-environmentalism: a preregistered replication of Wolsko et al. (2016)', Journal of Experimental Social Psychology, 124, 104871. Available at: https://doi.org/10.1016/j.jesp.2026.104871 (Accessed: 11 September 2026).