>[S]ocial science research (including some of our [own](https://onlinelibrary.wiley.com/doi/abs/10.1111/ajps.12724)) suggests that Democrats should not be focused on attacking Trump. A huge new survey we fielded — testing dozens of messages among over 100,000 people — finds the same. Voters have been hearing about Donald Trump for almost ten years now. If they’re willing to vote for him based on that near-decade of experience, a few ads or a new quip are unlikely to change their minds about him. In our survey, we found that every attempt at attacking Trump — from overturning *Roe* to his threat to democracy and calling him “weird” — didn’t persuade voters to support Harris.
>
>Instead of attacking Trump, Democrats should talk about *Harris*. The bad news for Democrats is that, to the extent voters do know Harris, they think she is [very liberal](https://www.cbsnews.com/news/poll-harris-trump-cbs-news/) and that her policies [would not make them better off financially](https://www.cbsnews.com/news/poll-harris-trump-cbs-news/). The good news is that voters have heard much less about Harris than Trump: in fact, many don’t know basic facts about whether she supports protecting Social Security or taxing the rich. That means there should be much more room to change voters’ views about her. Our survey finds exactly this: Only messages praising Harris’s achievements and describing her vision for America win her votes. Messages attacking Trump don’t.
>
>[…]
>
>Some commentators treat this question as a proxy for the battle between the Democrats’ progressive and moderate factions, hoping Harris will voice support for their faction’s signature policies. But our mega-survey finds neither side of this debate is right. Rather than tacking to the left or to the center, it’s messages that present Harris like a *normal* Democrat that most persuade voters. **That means running on mainstream “kitchen table” Democratic ideas to reduce the cost of living, protect Medicare and Social Security by taxing the rich, keep abortion legal, and raise the minimum wage.** Other messages that don’t map onto ideological divides in the Democratic party, such as touting her achievements as a prosecutor and casting the tie-breaking vote for the American Rescue Plan, also perform well.
>
>In other words, a good rule of thumb for Harris is that if *both* AOC *and* Joe Manchin would say they’re for something, she should probably be saying that, too.
>
>[…]
>
>Either way, attacking Trump instead of praising Biden in 2020 was a mistake. As detailed in a [piece Matt wrote at the time](https://www.vox.com/2020/6/9/21284758/broockman-kalla-biden-ads-trump-ads), [we found](https://www.dataforprogress.org/blog/2020/6/9/dems-should-run-pro-biden-ads) that attacks on Trump didn’t do much (and we tested hundreds!), but that messages both for and against Biden were much more persuasive. We also found that the more specific — in other words, informative — the Biden messages were, the more persuasive they were. This is because voters already knew a lot about Trump, including that he made a lot of mistakes as President and had questionable moral character. Even the specific bad things about Trump people *didn’t* know yet still weren’t surprising to them based on what they already knew, and so didn’t move votes. Democrats’ anti-Trump ads in 2020 were likely a huge waste.
>
>Democrats should not make the same mistake this time. Instead of repeating a single negative adjective about Trump over and over, Democrats should be informing voters about what *Harris* would do.
>
>People who read political Substacks might take it for granted that all Democratic candidates will support taxing the rich to protect Social Security and Medicare. But voters (especially persuadable voters) actually don’t. In another recent survey of swing state voters, we found that *half* don’t know that Harris opposes cutting Social Security benefits, and a third don’t know that she wants to raise taxes on the rich and corporations. In other words, that Harris is a normal Democrat actually *is* new information to many voters. And it’s likely a mistake to focus on new pejorative adjectives for Trump rather than on letting voters know where Harris stands on these bread-and-butter Democratic issues.
>
>[…]
>
>That’s not to say that progressives and moderates don’t have some messages that can persuade voters too. For example, a moderate “fiscal responsibility” message (about taxing the rich to both reduce the debt and protect Social Security and Medicare) performs well, as do progressive messages about abortion. But messages from both wings of the party generally perform worse than messages that span the ideological factions. For example, messages which paint Harris as moderate *or* as liberal on immigration do not appear to be as effective as simply talking about Social Security and Medicare.
>
>(You can see all the individual messages and the raw results for each one [here](https://docs.google.com/spreadsheets/d/15ixzsZffRvpDa1LEXJlS9mv6m00zzbDdMfvoG6IIXiA/edit?usp=sharing).)
#### Surveys aren’t everything
>To state the obvious, surveys like ours which ask voters how they’d vote immediately after seeing an ad are not a be-all end-all. Campaigns also have to think about other considerations. For example, just because certain policy ideas tested well in this survey doesn’t mean Harris would gain votes from supporting them after the other side hits Harris for supporting them. Likewise, as Matt has [written](https://www.slowboring.com/p/kamala-harris-should-shake-the-etch), videos of Harris taking unpopular positions during the 2020 Democratic presidential primary are likely to continue to haunt Harris, and Harris might be able to blunt the effectiveness of these attacks if she disavows some of those positions. The media might also be more interested in covering some messages than others. Our survey didn’t examine these dynamics.
>
>With that said, a common complaint about message testing — that it doesn’t measure what will excite voters — didn’t seem to hold much water. The messages that were persuasive were also the messages that excited the Democratic base to say they would turn out to vote this fall or share the messages on social media. We also see that pro-Harris messages *both* reduce support for Trump *and* increase support for Harris (our survey had third party and “won’t vote” options, which don’t drive the results).
>
>When pundits argue about how Democrats should talk to voters, data like this rarely makes an appearance, but as we’ve [written elsewhere](https://www.cambridge.org/core/journals/american-political-science-review/article/how-experiments-help-campaigns-persuade-voters-evidence-from-a-large-archive-of-campaigns-own-experiments/FF5BE6ED1553475F8321F7C4209357F7), it’s a mainstay of how campaigns actually make decisions. In our view, the increasing role of data in guiding campaign messaging is a positive development; human beings are [bad](https://osf.io/preprints/osf/8un6a) at predicting what will persuade others. Even if it needs to be taken with a grain of salt, data like this is almost certainly better than simply relying on gut instinct.
>
>Right now, the data’s message for Democrats is clear: Just be normal.
5 Comments
Summary:
>[S]ocial science research (including some of our [own](https://onlinelibrary.wiley.com/doi/abs/10.1111/ajps.12724)) suggests that Democrats should not be focused on attacking Trump. A huge new survey we fielded — testing dozens of messages among over 100,000 people — finds the same. Voters have been hearing about Donald Trump for almost ten years now. If they’re willing to vote for him based on that near-decade of experience, a few ads or a new quip are unlikely to change their minds about him. In our survey, we found that every attempt at attacking Trump — from overturning *Roe* to his threat to democracy and calling him “weird” — didn’t persuade voters to support Harris.
>
>Instead of attacking Trump, Democrats should talk about *Harris*. The bad news for Democrats is that, to the extent voters do know Harris, they think she is [very liberal](https://www.cbsnews.com/news/poll-harris-trump-cbs-news/) and that her policies [would not make them better off financially](https://www.cbsnews.com/news/poll-harris-trump-cbs-news/). The good news is that voters have heard much less about Harris than Trump: in fact, many don’t know basic facts about whether she supports protecting Social Security or taxing the rich. That means there should be much more room to change voters’ views about her. Our survey finds exactly this: Only messages praising Harris’s achievements and describing her vision for America win her votes. Messages attacking Trump don’t.
>
>[…]
>
>Some commentators treat this question as a proxy for the battle between the Democrats’ progressive and moderate factions, hoping Harris will voice support for their faction’s signature policies. But our mega-survey finds neither side of this debate is right. Rather than tacking to the left or to the center, it’s messages that present Harris like a *normal* Democrat that most persuade voters. **That means running on mainstream “kitchen table” Democratic ideas to reduce the cost of living, protect Medicare and Social Security by taxing the rich, keep abortion legal, and raise the minimum wage.** Other messages that don’t map onto ideological divides in the Democratic party, such as touting her achievements as a prosecutor and casting the tie-breaking vote for the American Rescue Plan, also perform well.
>
>In other words, a good rule of thumb for Harris is that if *both* AOC *and* Joe Manchin would say they’re for something, she should probably be saying that, too.
>
>[…]
>
>Either way, attacking Trump instead of praising Biden in 2020 was a mistake. As detailed in a [piece Matt wrote at the time](https://www.vox.com/2020/6/9/21284758/broockman-kalla-biden-ads-trump-ads), [we found](https://www.dataforprogress.org/blog/2020/6/9/dems-should-run-pro-biden-ads) that attacks on Trump didn’t do much (and we tested hundreds!), but that messages both for and against Biden were much more persuasive. We also found that the more specific — in other words, informative — the Biden messages were, the more persuasive they were. This is because voters already knew a lot about Trump, including that he made a lot of mistakes as President and had questionable moral character. Even the specific bad things about Trump people *didn’t* know yet still weren’t surprising to them based on what they already knew, and so didn’t move votes. Democrats’ anti-Trump ads in 2020 were likely a huge waste.
>
>Democrats should not make the same mistake this time. Instead of repeating a single negative adjective about Trump over and over, Democrats should be informing voters about what *Harris* would do.
>
>People who read political Substacks might take it for granted that all Democratic candidates will support taxing the rich to protect Social Security and Medicare. But voters (especially persuadable voters) actually don’t. In another recent survey of swing state voters, we found that *half* don’t know that Harris opposes cutting Social Security benefits, and a third don’t know that she wants to raise taxes on the rich and corporations. In other words, that Harris is a normal Democrat actually *is* new information to many voters. And it’s likely a mistake to focus on new pejorative adjectives for Trump rather than on letting voters know where Harris stands on these bread-and-butter Democratic issues.
>
>[…]
>
>That’s not to say that progressives and moderates don’t have some messages that can persuade voters too. For example, a moderate “fiscal responsibility” message (about taxing the rich to both reduce the debt and protect Social Security and Medicare) performs well, as do progressive messages about abortion. But messages from both wings of the party generally perform worse than messages that span the ideological factions. For example, messages which paint Harris as moderate *or* as liberal on immigration do not appear to be as effective as simply talking about Social Security and Medicare.
>
>(You can see all the individual messages and the raw results for each one [here](https://docs.google.com/spreadsheets/d/15ixzsZffRvpDa1LEXJlS9mv6m00zzbDdMfvoG6IIXiA/edit?usp=sharing).)
#### Surveys aren’t everything
>To state the obvious, surveys like ours which ask voters how they’d vote immediately after seeing an ad are not a be-all end-all. Campaigns also have to think about other considerations. For example, just because certain policy ideas tested well in this survey doesn’t mean Harris would gain votes from supporting them after the other side hits Harris for supporting them. Likewise, as Matt has [written](https://www.slowboring.com/p/kamala-harris-should-shake-the-etch), videos of Harris taking unpopular positions during the 2020 Democratic presidential primary are likely to continue to haunt Harris, and Harris might be able to blunt the effectiveness of these attacks if she disavows some of those positions. The media might also be more interested in covering some messages than others. Our survey didn’t examine these dynamics.
>
>With that said, a common complaint about message testing — that it doesn’t measure what will excite voters — didn’t seem to hold much water. The messages that were persuasive were also the messages that excited the Democratic base to say they would turn out to vote this fall or share the messages on social media. We also see that pro-Harris messages *both* reduce support for Trump *and* increase support for Harris (our survey had third party and “won’t vote” options, which don’t drive the results).
>
>When pundits argue about how Democrats should talk to voters, data like this rarely makes an appearance, but as we’ve [written elsewhere](https://www.cambridge.org/core/journals/american-political-science-review/article/how-experiments-help-campaigns-persuade-voters-evidence-from-a-large-archive-of-campaigns-own-experiments/FF5BE6ED1553475F8321F7C4209357F7), it’s a mainstay of how campaigns actually make decisions. In our view, the increasing role of data in guiding campaign messaging is a positive development; human beings are [bad](https://osf.io/preprints/osf/8un6a) at predicting what will persuade others. Even if it needs to be taken with a grain of salt, data like this is almost certainly better than simply relying on gut instinct.
>
>Right now, the data’s message for Democrats is clear: Just be normal.
!ping Democracy
Inmate.
Saying Vance had sex with a couch.
Creepy can work too.
Having him eat a bunch of Big Macs. Like 30 a day