Steven Cherry Hi, this is Steven Cherry for Radio Spectrum.
For some years and still today, there’s been a quiet but profound schism among political strategists. There are those who favor modern methods and modern media—mass mailings, robocalling, television advertising, and, increasingly, social-media advertising. On the other hand are those, including my guest today, who not only still see a value in traditional person-to-person messaging, but see it as, frequently, the better bang for the campaign buck.
Just last week [this was recorded Oct 5, 2020—Ed.] the attorney general of Michigan— a state that has been a battleground, not just for electoral delegates, but this methodological dispute—announced that two political operatives were charged with felonies in connection with robocalls that made a number of false claims about the risks of voting by mail, in an apparent attempt to discourage residents of Detroit from voting by mail. And last week as well, the Biden campaign announced a complete turnaround on the question of door-to-door canvassing, perhaps the gold standard of person-to-person political campaigning. Are they perhaps afraid of Democratic standard-bearers making the same mistake twice?
In the endless post-mortem of the 2016 Presidential election, an article in Politico argued that the Clinton campaign was too data-driven and model-driven, and refused local requests, especially in Michigan, for boots-on-the-ground support. It quoted a longtime political hand in Michigan as describing quote “months of failed attempts to get attention to the collapse she was watching unfold in slow-motion among women and African-American millennials.”
I confess I saw something of that phenomenon on a recent Saturday. I’m living in Pittsburgh these days, and in the morning, I worked a Pennsylvania-based phone bank for my preferred political party. One of my first calls was to someone in the Philadelphia area, who told me he had already made his absentee ballot request and asked, while he had me on the phone, when his ballot would come. “There used to be someone around here I forget what you call her but someone I could ask stuff of.” That was strike one.
In another call, to a man in the Erie area, the conversation turned to yard signs. He said he would like to put one out but he had no idea where to get it. Strike two. In the late afternoon, two of us went to a neighborhood near us to put out door-hangers, and if we saw someone face-to-face we would ask if they wanted a yard sign. One fellow said he would. “We were supposed to get one,” he told us. When he saw we had a stack of them in our car, he sheepishly added, “We were supposed to get two in fact, one for a friend.” That was my third indication in one day that there was a lack of political party involvement at the very local level—in three different parts of what could well be the most critical swing state of the 2020 Presidential election.
When I strung these three moments together over a beer, my partner immediately thought of a book she owned, Get Out the Vote, now in its fourth edition. Its authors, Donald Green and Alan Gerber, argue that political consultants and campaign managers have underappreciated boots-on-the-ground canvassing in person and on the phone, in favor of less personal, more easily-scaled methods—radio and TV advertising, robocalling, mass mailings, and the like.
Of particular interest, they base their case with real data, based on experimental research. The first edition of their book described a few dozen such experiments; their new edition, they say, summarizes hundreds.
One of those authors is Donald Green, a political scientist at Columbia University focusing on such issues as voting behavior and partisanship, and most importantly, methodologies for studying politics and elections. His teaching career started at Yale University, where he directed its Institution for Social and Policy Studies. He joins us via Skype.
Steven Cherry Don, welcome to the podcast.
Donald Green Thank you very much for having me.
Steven Cherry Modern campaigns can employ an army of advisers, consultants, direct mail specialists, phone bank vendors, and on and on. You say that much of the advice candidates get from these professionals comes from war stories and not evidence. Robocalls seem to be one example of that. The study of a 2006 Texas primary found that 65 000 calls for one candidate increased his vote share by about two votes.
Donald Green Yes, the robocalls have an almost perfect record of never working in randomized trials. These are trials in which we randomly assigned some voters to get a robocall and others not and allow the campaign to give it its best shot with the best possible robocall. And then at the end of the election, we look at voter turnout records to see who voted. And in that particular case, the results were rather dismal. But not just in that case. I think that there have been more than 10 such large-scale experiments, and it’s hard to think of an instance in which they’ve performed well.
Steven Cherry The two robocallers in Michigan allegedly made 12 000 calls into Detroit, which is majority black—85 000 calls in total to there and similar areas in other cities. According to a report in the Associated Press, calls falsely claimed that voting by mail would result in personal information going into databases that will be used by police to resolve old warrants, credit card companies to collect debts, and federal officials to track mandatory vaccines. It quoted the calls as saying, “Don’t be finessed into giving your private information to The Man. Beware of vote-by-mail.” You’ve studied plenty of affirmative campaigns, that is, attempts to increase voter participation. Do you have any thoughts about this negative robocalling?
Donald Green Well, that certainly seems like a clear case of attempted voter suppression—to try to scare people away from voting. I don’t think I’ve ever seen anything like this. I haven’t heard the call. I’d be curious to know something about the voiceover that was used. But let’s suppose that it seemed credible. You know, the question is whether people take it seriously enough or whether they questioned the content, maybe talking to others in ways that undercut its effectiveness. But if robocalls seldom work, it’s probably because people just don’t notice them. Not sure whether this one would potentially work because it would get somebody to notice at any rate. We don’t know how effective it would be. I suspect not terribly effective, but probably effective enough to be concerning.
Steven Cherry Yeah, it was noticed enough that complaints about it filtered up to the state attorney general, but that doesn’t give us any quantitative data.
For decades, campaigns have spent a lot of their money on television advertising. And it can influence strategy. To take just one example, there’s a debate among Democrats about whether their candidate should invest in Texas because there’s so many big media markets. It’s a very expensive state to contest. What does the experimental data tell us about television?
Donald Green Experience on television is relatively rare. One that I’m most familiar with is one that actually I helped conduct with my three coauthors back when we were studying the Texans for Rick Perry campaign in 2006. We randomly assigned 18 of the 20 media markets in Texas to receive varying amounts of TV advertising, and various timings at which point it would be rolled out. And we conducted daily tracking polls to see the extent to which public opinion moved as ads rolled out in various media markets. And what we found was there was some effect of Rick Perry’s advertising campaign, but it subsided very quickly. Only a few days passed before it was essentially gone without a trace, which means that one can burn quite a lot of money for a relatively evanescent effect in terms of the campaign. I really don’t think that there’s much evidence that the very, very large amounts of money that are spent on television in the context of a presidential campaign have any lasting effect. And so it’s really an open question as to whether, say, the $300 million dollars that the Clinton campaign spent in 2016 would have been better spent least as well spent on the ground.
Steven Cherry In contrast to war stories, you and your colleagues conduct true randomized experiments. Maybe you could say a little bit more about how hard that is to do in the middle of an election.
Yes, it’s a juggling act for sure. The idea is, if we wanted to study, for example, the effects of direct mail on voter turnout, one would randomly assign large lists of registered voters, some to get the mail, some to be left alone. And then we’d use the fact that voting is a public record in the United States—and a few other countries as well—to gauge voter turnout after the election is over. This is often unsatisfactory for campaigns. They want to know the answer ahead of time. But first, we know no good way of answering the question before people actually cast their ballots. And so this is something that’s been done in increasing numbers since 1998. And now hundreds of those trials have been done on everything ranging from radio, robocalls, TV, direct mail, phone calls, social media, etc, etc.
Steven Cherry One thing you would expect campaign professionals to have data on is cost-effectiveness, but apparently they don’t. But you do. You’ve found, for example, that you can generate the same 200 votes with a quarter of a million robocalls, 38 000 mailers, or 2500 door-to-door conversations.
Donald Green Yes, we try to not only gauge the effects of the intervention through randomized trials but also try to figure out what that amounts to in terms of dollars per vote. And these kinds of calculations are always going to be context-dependent because some campaigns are able to rely on inexpensive people power, to inspire volunteers in vast numbers. And so in some sense, the costs that we estimate could be greatly overstated for the kinds of boots-on-the-ground canvassing that are typical of presidential elections in battleground states. Nevertheless, I think that it is interesting to note that even with relatively cautious calculations, to the effect that people are getting $16 an hour for canvassing, canvassing still acquits itself rather well in terms of its comparisons to other campaign tactics.
Steven Cherry Now that’s just for turnout, not votes for one candidate instead of another; a nonpartisan good-government group might be interested in turnout for its own sake, but a campaign wants a higher turnout of its own voters. How does it make that leap?
Donald Green Well, typically what they do is rely on voter files—and augmented voter files, which is, say, voter files that had other information about people appended to them—in order to make an educated guess about which people on the voter file are likely to be supportive of their own campaign. So Biden supporters have been micro-targeted and so have Trump supporters and so on and so forth, based on their history of donating to campaigns or signing petitions or showing up in party primaries. And that makes the job of the campaign much easier because instead of trying to persuade people or win them over from the other side, they’re trying to bring a bigger army to the battlefield by building up enthusiasm and mobilizing their own core supporters. So the ideal for that kind of campaign is a person who is very strongly aligned with the candidate that is sponsoring the campaign but has a low propensity of voting. And so that that kind of person is really perfect for a mobilization campaign.
So that could also be done demographically. I mean, there are zip codes in Detroit that are 80 percent black.
Donald Green Yes, there are lots of ways of doing this based on aggregates. No, you often don’t have to rely on aggregates because you typically have information about each person. But if you were to basically do it, say, precinct by precinct, you could use as proxies for the left—percentage-African-American—or proxies for the right demographics that are associated with Trump voting. So it’s possible to do it, but it’s probably not state of the art.
Steven Cherry You mentioned door-to-door canvassing; it increases turnout but—perhaps counterintuitively—apparently, it doesn’t matter much whether it’s a close contest or a likely blowout, and if it doesn’t matter what the canvasser’s message is.
Donald Green This is one of the most interesting things, actually about studying canvassing and other kinds of tactics experimentally. It appears that some of the most important communication at the door is nonverbal. You know, you show up at my door, and I wonder what you’re up to—are you trying to sell me something, trying to, you know, make your way in here? I figure, oh, actually you’re just having a pleasant conversation. You’re a person like me. You’re taking your time out to encourage me to vote. Well, that sounds okay. And I think that that message is probably the thing that sticks with people, perhaps more than the details of what you’re trying to say to me about the campaign or the particularities about why I should vote—should I vote because it’s my civic duty or should I vote because I need to stand up in solidarity with my community? Those kinds of nuances don’t seem to matter as much as we might suppose.
Steven Cherry So it seems reminiscent of what the sociologists would call a Hawthorne effect.
Donald Green Some of it is reminiscent of the Hawthorne effect. The Hawthorne effect is basically, we increase our productivity when we’re being watched. And so there’s some sense in which being monitored, being encouraged by another person makes us feel as though we’ve got to give a bit more effort. So there’s a bit of that. But I think partly what’s going on is voting is a social activity. And just as you’re more likely to go to a party if you were invited by a person as opposed to by e-mail. So too, you’re more likely to show up to vote if somebody makes an authentic, heartfelt appeal to you and encourages you to vote in-person or through something that’s very much like in-person. So it’s some gathering or some friend to friend communication as opposed to something impersonal, like you get a postcard.
Steven Cherry So without looking into the details of the Biden campaign flip-flop on door-to-door canvassing, your hunch would be that they’re making the right move?
Donald Green Yes, I think so. I mean, putting aside the other kinds of normative concerns about whether people are at risk if they get up and go out to canvass or they’re putting others at risk ... In terms of the raw politics of winning votes, it’s a good idea in part because in 2018, they were able to field an enormous army of very committed activists in many of the closely contested congressional elections and showed apparently very good, good results. And the tactic itself is so well tested that if they can do it with appropriate PPE and precautions, they could be quite effective.
Steven Cherry In your research you found by contrast, door-hangers and yard signs—the way I spent that Saturday afternoon I described—have little or maybe even no utility.
Donald Green Well, yard signs might have some utility to candidates, especially down-ballot candidates who are trying to increase their vote share. It doesn’t seem to have much of an effect on voter turnout. Maybe that’s because the election is already in full swing and everybody knows that there’s an election coming up—the yard sign isn’t going to convey any new information. But I do think the door hangers have some residual effect. They’re probably about as effective as a leaflet or a mailer, which is not very effective, but maybe a smidge better than zero.
Steven Cherry You’re more positive on phone banks, albeit with some qualifiers.
Donald Green Yes, I think that phone banking, especially authentic volunteer-staffed phone banking, can be rather effective. You know, I think that if you have an unhurried conversation with someone who is basically like-minded. They’re presumably targeted because they’re someone who shares more or less your political outlook and you bring them around to explain to them why it’s an important and historic election, giving them any guidance you can about when and how to vote. You can have an effect. It’s not an enormous effect. It’s something in the order of, say, three percentage points or about one additional vote for every 30 calls you complete. But it’s a substantial effect.
And if you are able to extract a commitment to vote from that person and you were to be so bold as to call them back on the day before the election to make sure that they’re making good on their pledge, then you can have an even bigger effect, in fact, a very large effect. So I do think it can be effective. I also think that perfunctory, hurried calls by telemarketing operations are rather ineffective for a number of reasons, but especially the lack of authenticity.
Steven Cherry Let’s turn to social media, particularly Facebook. You described one rather pointless Facebook campaign that ended up costing $474 per vote. But your book also describes a very successful experiment in friend-to-friend communication on Facebook.
Donald Green That’s right. We have a number of randomized trials suggesting that encouragements to vote via Facebook ads or other kinds of Facebook media that are mass-produced seem to be relatively limited in their effects. Perhaps the biggest, most intensive Facebook advertising campaign was its full-day banner ads that ran all day long—I think it was the 2010 election—and had precisely no effect, even though it was tested among 61 million people.
More effective on Facebook were ads that showed you whether your Facebook friends had claimed to vote. Now, that didn’t produce a huge harvest of votes, but it increased turnout by about a third of a percentage point. So better than nothing. The big effects you see on Facebook and elsewhere are where people are, in a personalized way, announcing the importance of the upcoming election and urging their Facebook friends—their own social networks—to vote.
And that seems to be rather effective and indeed is part of a larger literature that’s now coming to light, suggesting that even text messaging, though not a particularly personal form of communication, is quite effective when friends are texting other friends about the importance of registering and voting. Surprisingly effective, and that, I think, opens up the door to a wide array of different theories about what can be done to increase voter turnout. It seems as though friend-to-friend communication or neighbor-to-neighbor communication or communication among people who are coworkers or co-congregants ... that could be the key to raising turnout—not by not just one or two percentage points, but more like eight to 10.
Steven Cherry On this continuum of personal versus impersonal, Facebook groups,—which are a new phenomenon—seem to lie somewhere in between. Some people are calling them “toxic echo chambers,” but they would seem to maybe be a godsend for political engagement.
Donald Green I would think so, as long as the communication within the groups is authentic. If it’s if it’s automated, then probably not so much. But to the extent that the people in these groups have gotten to know each other or knew each other before they came into the group, then I think communication among them or between them could be quite compelling.
Steven Cherry Yes. Although, of course, that person that you think you’re getting to know might be some employee in St. Petersburg, Russia, of the Internet Research Agency. Snapchat has been getting some attention these days in terms of political advertising. They’ve tried to be more transparent than Facebook, and they do some fact-checking on political advertising. Could it be a better platform for political ads or engagement?
Donald Green I realize I just don’t know very much about the nuances of what they’re doing. I’m not sure that I have enough information to say.
Steven Cherry Getting back to more analog activities, your book discusses events like rallies and processions, but I didn’t see anything about smaller coffee-klatch-style events where, say, you invite all your neighbors and friends to hear a local candidate speak. That would seem to combine the effectiveness of door-to-door canvassing with the Facebook friend-to-friend campaign. But maybe it’s hard to study experimentally.
Donald Green That’s right. I would be very, very optimistic about the effects of those kinds of small gatherings. And it’s not that we are skeptical about their effects. It’s just, as you say, difficult to orchestrate a lot of experiments where people are basically opening their homes to friends. We need to talk to rope in more volunteers to bring in their friends experimentally.
Steven Cherry The business model for some campaign professionals is to get paid relative to the amount of money that gets spent. Does that disincentivize the kind of person-to-person campaigning you generally favor?
Donald Green Yes, I would say that one of the biggest limiting factors on person-to-person campaigning is that it’s very difficult for campaign consultants to make serious money off of it. And that goes double for the kind of serious money that is poured into campaigns in the final weeks. Huge amounts of money tend to be donated within the last three weeks of an election. And by that point, it’s very difficult to build the infrastructure necessary for large-scale canvassing or really any kind of retail-type politics. For that reason, the last-minute money tends to be dumped into digital ads and in television advertising—and in lots and lots of robocalls.
Steven Cherry Don, as we record, this is less than a week after the first 2020 presidential debate and other events in the political news have maybe superseded the debate already. But I’m wondering if you have any thoughts about it in terms of getting out the vote. Many people, I have to say, myself included, found the debate disappointing. Do you think it’s possible for a debate to depress voter participation?
Donald Green I think it’s possible. I think it’s rather unlikely to the extent that political science researchers have argued that negative campaigning depresses turnout, tends to depress turnout among independent voters, not so much among committed partisans who watched the debate and realize more than ever that their opponent is aligned with the forces of evil. For independent voters, they might say, “a plague on both your houses, I’m going to participate.” But I think that this particular election is one that is so intrinsically interesting that the usual way that independents feel about partisan competition probably doesn’t apply here.
Steven Cherry On a lighter note, an upcoming podcast episode for me will be about video game culture. And it’ll be with a professor of communications who writes her own video games for her classes. Your hobby turns out to be designing board games. Are they oriented toward political science? Is there any overlap of these passions?
Donald Green You know, it’s strange that they really don’t overlap at all. My interest in board games goes back to when I was a child. I’ve always been passionate about abstract board games like chess or go. And there was an accident that I started to design them myself. I did it actually when my fully-adult children were kids and we were playing with construction toys. And I began to see possibilities for games in those construction toys. And one thing led to another. And they were actually deployed to the world and marketed. And now I think they’re kind of going the way of the dinosaur. But there’s still a few dinosaurs like me who enjoy playing on an actual physical board.
Steven Cherry My girlfriend and I still play Rack-O. So maybe this is not a completely lost cause.
Well Don, I think in the US, everyone’s thoughts will be far from the election until the counting stops. Opinions and loyalties differ. But the one thing I think we can all agree on is that participation is essential for the health of the body politic. On behalf of all voters, let me thank you for all that your book has done toward that end and for myself and my listeners, thank you for joining me today.
Donald Green I very much appreciate it. Thanks.
Radio Spectrum is brought to you by IEEE Spectrum, the member magazine of the Institute of Electrical and Electronic Engineers.
For Radio Spectrum, I’m Steven Cherry.
Note: Transcripts are created for the convenience of our readers and listeners. The authoritative record of IEEE Spectrum’s audio programming is the audio version.