A voter is reflected in the glass frame of a poster while leaving a polling site. (David Goldman/AP)

Political professionals in both parties are busy strategizing about how to persuade voters. Democratic operatives are fiercely debating how to win back working-class voters in key swing states. Meanwhile, the Republican Party establishment is puzzling over how to beat back populist primary challengers.

Will either be successful? At first glance, our new research “The Minimal Persuasive Effects of Campaign Contact in General Elections” might appear discouraging for such efforts. In the paper, we analyzed 40 existing experiments and nine new experiments that rigorously measure the persuasive effect of political campaigns’ voter contact. Vox’s Dylan Matthews summarized one of our main findings thus: “Campaigns’ attempts to win swing voters appear to not work at all.”

Thankfully for campaigns, however, such headlines do not tell the whole story about our findings. There are also important exceptions that suggest ways forward for campaigns. Campaigns might wish to heed important caveats to our conclusions.

1. Campaigns can use experiments to identify persuadable voters.

We find that persuasion can work in partisan general elections when campaigns use experiments to identify persuadable voters and remove voters that react negatively to their messages.

For example, Working America, the community affiliate of the AFL-CIO, conducted experiments in 2016 that found surprising pockets of voters persuaded by their messages. It then targeted these voters more intensively going forward, and stopped talking to the voters who reacted negatively.

It’s a powerful approach. From validation studies, we estimate that Working America’s changed targeting generated many thousands of additional votes. These votes would have taken many millions of dollars to capture with other tactics.

These so-called experiment informed programs (EIPs) thus offer data-driven targeting recommendations that perform better than campaign gurus’ current recommendations based on polling, focus groups or gut instinct. Campaigns have been using these techniques for years, but new research methods we published last year make these experiments much cheaper.

2. Campaigns can persuade voters months in advance — but they don’t yet know how to make those changed minds stay changed.

We find that more than a couple of months before Election Day, campaigns can persuade voters to support their candidates. However, this persuasion does not last until Election Day; as campaigns get closer to an election, these tactics stop working. This suggests that an important task for campaigns is to learn how to make that early persuasion last. Voters do appear more open to changing their minds early on in an election cycle.

3. Campaigns are better at persuading voters when party labels aren’t available.

Campaign persuasion appears to work in primaries and in ballot measure elections, even competitive ones. When voters can’t rely on party labels on ballots, it is much easier for campaigns to influence their choices.

4. There’s much more to campaigning than the additional direct voter contacts we studied.

There are also many kinds of campaign activity our study does not consider, which means that campaigns should not overreact to our findings.

First, experiments like the ones we report measure the marginal impact of additional persuasion attempts, not the total impact of a campaign. We did also find minimal persuasion in quiet, uncompetitive races, suggesting “saturation” is not the entire reason we found persuasion was usually ineffective. However, we do not yet know how much — or whether — campaigns’ persuasive efforts with direct voter contact matter in aggregate.

Further, we do not study the effects of a campaign’s overall “message,” its media coverage from journalists, or other ways that voters might get information about candidates, such as from co-workers, friends, family and neighbors. We have the least evidence on the effectiveness of TV ads.

Finally, our findings in no way contradict the vast body of research demonstrating the efficacy of voter registration and mobilization tactics for getting supporters to the polls.

These caveats still might not provide much solace to campaign operatives with a big vision, like Democrats hoping to win back the Rust Belt or Republicans hoping to slow down the support swell for outsiders. But the history of research on voter turnout also suggests another possibility.

5. If campaigns try new and creative ideas for how to persuade voters, they may succeed — dramatically.

The history of voter turnout campaigning shows the dramatic gains campaigns can reap from trying new tactics and testing them rigorously. When Alan Gerber and Donald Green started conducting field experiments on the effects of voter turnout efforts in the late 1990s, they found that typical consultant-designed tactics — including campaign phone calls, mailers and robo-calls — did little, if anything, to mobilize voters. Soon, however, campaign professionals started imbuing those tactics with new messages informed by psychology research, using such angles as social pressure and gratitude.

Those new efforts were dramatically more effective at getting voters to the polls. Data from experiments that Gerber and Green collected for their 2015 book “Get Out The Vote!” makes clear just what a revolution in effectiveness these tactics represented. When we examine data from experimental measurements of GOTV efforts that don’t use one of these “psychologically informed” tactics, their effect on voter turnout appears positive but very close to zero — 0.2 percentage points on average. However, tactics that use at least one psychological enhancement have effects of about 2 percentage points on average.

In other words, the data suggests that rethinking many traditional GOTV tactics with help from psychology research may have made campaigns’ turnout efforts 10 times more effective.

Had campaign operatives and funders simply given up on stimulating voter turnout by mail in reaction to the early Gerber and Green findings, researchers could never have done the research necessary to learn how to increase turnout more effectively. The lesson: Don’t give up on persuasion; look for new ideas, whether new modes, messages or messengers.

But to succeed, campaigns will need not only to try new ideas but also rigorously test which ones work. Just as in medical trials where doctors randomly assign patients to different treatments to see which ones improve patient health, campaigns will be most successful if they randomize their persuasive contacts and measure the effects against a control group to know whether their efforts actually make a difference.

Aaron Blake at The Washington Post’s The Fix wondered whether our results should lead to “the end of political campaigns as we know them.” We don’t think so. But at the very least, our findings should spur more campaigns to carefully measure which of their efforts actually work.

Joshua Kalla is a PhD candidate in political science at the University of California, Berkeley. Follow him on Twitter @j_kalla.

David Broockman is an assistant professor of political economy at the Stanford Graduate School of Business. Follow him on Twitter @dbroockman.