Popular in Social Sciences

Political Science

V-Dem’s Conceptions of Democracy and Their Consequences

University of Notre Dame Michael Coppedge

Central Monetary Authorities

Macroeconomic modelling of CBDC: a critical review

European Central Bank Ulrich Bindseil

Social Identities and Divisions

A new explanation of antisemitism: Jew hatred as a civilisational phenomenon

Boston University Liah Greenfeld

Central Monetary Authorities

Financing Private Credit

Federal Reserve Bank of New York Nina Boyarchenko

Economics

Failing Banks

Massachusetts Institute of Technology Emil Verner

Culture and Society

The City as a Growth Machine: Toward a Political Economy of Place

New York University Harvey Moltoch

American Politics

Do Billboard Advertisements Increase Voter Turnout?

View more insights in Social Sciences - Political Science

View more figures from Columbia University

View the page of Donald Green

Read the Study

Next in Political Science

Stronger Issues, Weaker Predispositions

Paul Goren | University of Minnesota

Allied Air Attacks and Civilian Harm in Italy, 1940–1945 Bombing among Friends

Matthew Anthony Evangelista | Cornell University

Can billboard advertisements really sway voter participation? A comprehensive study during the 2020 U.S. general election explores their impact across diverse locales, revealing minimal overall effects on turnout.

Image courtesy of interviewee. December 17, 2024

Log-in or Sign-up to Faculti
Currently viewing this subject insight as a guest. You have insight(s) remaining for this month. Login to view 8000+ figures on the platform.

Generating Workspace...

Export as PDF

This video discusses the effectiveness of billboard advertising in influencing voter turnout, particularly in the context of the 2020 presidential election in the United States. Scholars have long questioned the impact of outdoor signage in both marketing and political science, but research on this topic is limited and often lacks rigorous methodology.

The authors conducted a nationwide randomized control trial to evaluate how billboard advertisements, promoting voting, affected turnout in over 150 metro areas. They focused on placing billboards in locations likely to reach local residents rather than along highways, allowing for a diverse geographical representation. The study aimed to detect subtle effects, especially in non-battleground states that typically receive less campaign attention.

Using cell phone data, the researchers tracked how many people passed by the billboards, ensuring no individual voter data was compromised. Overall, the findings revealed that while there may be slight increases in voter turnout very close to the billboards, the overall effects were generally close to zero. Interestingly, the billboards did not significantly raise registration rates, which suggests that passive advertising might not effectively motivate voters to take action.

The authors noted that the high turnout rates in the 2020 election might have made it harder to detect the billboard's influence. They also acknowledged that future research could explore whether more compelling messages or different election contexts could yield better results in advertising effectiveness.

Language:

Export as PDF

X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:90000

1
00:00:00.065 --> 00:00:03.445
For many, many years, scholars both in marketing

2
00:00:03.585 --> 00:00:06.085
and in political science have wondered about the

3
00:00:06.085 --> 00:00:09.085
effectiveness of billboards or signage more generally.

4
00:00:09.385 --> 00:00:11.365
Uh, they've wondered it with respect to,

5
00:00:11.365 --> 00:00:14.405
they've wondered about it with respect to, its ability

6
00:00:14.465 --> 00:00:16.085
to, sell products,

7
00:00:16.105 --> 00:00:18.885
but also its ability to mobilize voters

8
00:00:19.265 --> 00:00:21.365
or contribute to,

9
00:00:21.365 --> 00:00:23.005
political participation more generally.

10
00:00:23.505 --> 00:00:26.765
Um, surprisingly, the literature is rather thin.

11
00:00:27.315 --> 00:00:28.805
It's not as though that there are no studies,

12
00:00:28.865 --> 00:00:31.165
but there are no studies that are both,

13
00:00:31.225 --> 00:00:32.765
experimentally crafted

14
00:00:32.945 --> 00:00:36.445
and done on a sufficiently large scale to, produce,

15
00:00:36.545 --> 00:00:38.285
you know, fairly clear conclusions.

16
00:00:38.425 --> 00:00:41.565
And there are some rather extravagant claims out there.

17
00:00:41.785 --> 00:00:43.245
So we wanted to test those.

18
00:00:43.355 --> 00:00:46.405
Broader context is the 2020 presidential election.

19
00:00:46.405 --> 00:00:47.845
Of course, this was in the middle of Covid.

20
00:00:48.145 --> 00:00:51.445
Um, and so it was a special election in many ways,

21
00:00:51.905 --> 00:00:55.925
but the idea was to deploy something that is, kind

22
00:00:56.125 --> 00:00:59.405
of staple for advertising, outdoor advertising in general.

23
00:01:00.025 --> 00:01:02.245
And we wanted to do it in the broadest possible way.

24
00:01:02.465 --> 00:01:05.605
And because this was a nonpartisan, evaluation,

25
00:01:05.945 --> 00:01:08.805
we did not specifically seek out battleground states.

26
00:01:08.905 --> 00:01:12.365
As you can see from the, the map in the, in the article.

27
00:01:12.745 --> 00:01:17.325
Uh, it is a nationwide, romp through,

28
00:01:17.435 --> 00:01:19.765
more than 150, metro areas.

29
00:01:20.715 --> 00:01:22.205
What did we do specifically?

30
00:01:22.205 --> 00:01:23.965
Well, we, we wanted to make sure

31
00:01:23.965 --> 00:01:26.565
that we weren't just advertising to roadsides.

32
00:01:26.565 --> 00:01:29.605
We wanted to, put billboards where people

33
00:01:29.985 --> 00:01:32.285
who are living nearby could see them.

34
00:01:32.705 --> 00:01:35.405
And so that meant that we were, picky about

35
00:01:35.405 --> 00:01:37.045
where we selected our billboards.

36
00:01:37.045 --> 00:01:40.165
We wanted them to be in metro areas, not along,

37
00:01:40.235 --> 00:01:42.285
roadsides and long country roads.

38
00:01:42.585 --> 00:01:45.285
And we also wanted them to be geographically diverse.

39
00:01:45.565 --> 00:01:47.885
I think the idea is we wanted to have a conclusion

40
00:01:48.115 --> 00:01:51.245
that would, have legs about, about billboards.

41
00:01:51.265 --> 00:01:53.965
We did not want it to be specific to,

42
00:01:54.025 --> 00:01:56.525
one particular presidential campaign in one area.

43
00:01:56.945 --> 00:02:00.565
We also wanted to make sure that to the extent that,

44
00:02:00.635 --> 00:02:02.125
that the 2020 election

45
00:02:02.125 --> 00:02:04.605
or any presidential election oversaturate voters

46
00:02:04.955 --> 00:02:08.205
with advertising, we could, have an opportunity

47
00:02:08.225 --> 00:02:09.845
to detect the possible effects

48
00:02:09.845 --> 00:02:12.925
of billboard advertising in non battleground states, which,

49
00:02:13.105 --> 00:02:15.045
as you may know, are altogether ignored

50
00:02:15.065 --> 00:02:16.325
by the, the campaigns.

51
00:02:16.585 --> 00:02:18.805
So to the extent that a person is living

52
00:02:19.385 --> 00:02:22.005
off the beaten presidential campaign path,

53
00:02:22.105 --> 00:02:24.805
but nevertheless might be seeing a, a billboard

54
00:02:24.945 --> 00:02:26.885
and, you know, ready to act on it

55
00:02:26.885 --> 00:02:27.965
because of course, it's just

56
00:02:27.985 --> 00:02:30.645
as much a presidential election for them.

57
00:02:31.185 --> 00:02:32.685
Uh, we wanted to be able to,

58
00:02:32.685 --> 00:02:34.085
to detect, to detect that effect.

59
00:02:34.195 --> 00:02:36.965
Okay. This is a randomized control trial akin

60
00:02:36.985 --> 00:02:38.565
to say, a pharmaceutical trial.

61
00:02:38.665 --> 00:02:42.205
So we wanted to do it in the most sciencey, sort of way.

62
00:02:42.825 --> 00:02:45.525
And so the idea was to take areas

63
00:02:45.525 --> 00:02:48.245
where billboards were available for purchase

64
00:02:48.345 --> 00:02:51.005
and randomly assign them to either a treatment con

65
00:02:51.265 --> 00:02:52.285
or control condition.

66
00:02:52.425 --> 00:02:55.525
The treatment condition received often several weeks

67
00:02:55.585 --> 00:02:58.765
of billboard advertising, displaying a

68
00:02:59.085 --> 00:03:00.845
Prominent message, encouraging people to vote

69
00:03:01.265 --> 00:03:03.005
or the control condition receive nothing.

70
00:03:03.465 --> 00:03:05.845
And so the idea was to

71
00:03:06.435 --> 00:03:09.725
examine voter turnout patterns in the catchment areas

72
00:03:09.905 --> 00:03:11.005
of these billboards.

73
00:03:11.065 --> 00:03:15.005
Now, how large those catchment areas are, it's a kind

74
00:03:15.005 --> 00:03:16.405
of, open question.

75
00:03:16.465 --> 00:03:17.845
And that's, I think, part of

76
00:03:17.955 --> 00:03:19.965
what makes this research kind of fun and interesting.

77
00:03:20.195 --> 00:03:24.085
Cell phone data were never linked to any individual voters,

78
00:03:24.185 --> 00:03:26.925
so this was not a case of a privacy concern.

79
00:03:27.305 --> 00:03:28.365
Um, it's just

80
00:03:28.365 --> 00:03:32.725
that a group was collecting cell phone ping data,

81
00:03:32.725 --> 00:03:34.125
during the 2020 election

82
00:03:34.185 --> 00:03:38.205
and had them available to, nonpartisan groups like ours.

83
00:03:38.505 --> 00:03:43.325
Um, and the idea is many people leave their cell phone pings

84
00:03:43.505 --> 00:03:45.645
on and available to aggregators.

85
00:03:46.105 --> 00:03:50.765
Um, and essentially these aggregators simply form little

86
00:03:50.765 --> 00:03:53.685
miniature maps of every voter or every person.

87
00:03:53.715 --> 00:03:55.645
They don't necessarily have to be a voter,

88
00:03:55.745 --> 00:03:58.725
as they travel around, on their daily,

89
00:03:58.835 --> 00:04:00.565
path from home to wherever they're going.

90
00:04:00.905 --> 00:04:03.005
And so they can sort of figure out

91
00:04:03.005 --> 00:04:04.165
where people live based on

92
00:04:04.165 --> 00:04:06.165
where their cell phone is pinging all night,

93
00:04:06.505 --> 00:04:08.325
and they can sort of figure out where people are working,

94
00:04:08.545 --> 00:04:12.885
based on sort of an equivalent, assumption,

95
00:04:12.885 --> 00:04:14.645
about where they're standing, staying around

96
00:04:14.705 --> 00:04:16.685
for several hours during the day every day.

97
00:04:16.985 --> 00:04:20.285
And so they, map out their travel routes.

98
00:04:20.285 --> 00:04:22.365
And the reason we used that was not

99
00:04:22.365 --> 00:04:24.045
because we were linking any individual

100
00:04:24.265 --> 00:04:25.685
to any individual voter record.

101
00:04:26.145 --> 00:04:29.565
Um, it was just that we were interested in knowing the flows

102
00:04:29.745 --> 00:04:33.965
of, people from, area to area so that we'd know

103
00:04:33.965 --> 00:04:36.605
how many people would come within a few hundred meters

104
00:04:36.785 --> 00:04:37.805
of our billboards.

105
00:04:37.945 --> 00:04:40.325
And that would give us a sense of not only

106
00:04:40.425 --> 00:04:41.965
how many people were seeing the billboards,

107
00:04:42.265 --> 00:04:43.685
but at what frequency.

108
00:04:44.205 --> 00:04:46.285
I would say that the findings are somewhat disappointing,

109
00:04:46.285 --> 00:04:47.765
perhaps for those who are hoping

110
00:04:47.765 --> 00:04:49.565
that billboards would have a massive effect.

111
00:04:50.145 --> 00:04:53.405
Um, they have a bit of a murky effect, it seems as though

112
00:04:53.665 --> 00:04:55.885
for the most part, the effect is close to zero.

113
00:04:56.265 --> 00:04:59.765
But there are glimmers of a possible effect very, very close

114
00:04:59.865 --> 00:05:00.885
to the billboard.

115
00:05:00.985 --> 00:05:04.485
So half a mile, as we Americans would say,

116
00:05:04.745 --> 00:05:06.045
and I would, I would say, you know,

117
00:05:06.275 --> 00:05:08.445
roughly less than one kilometer.

118
00:05:08.665 --> 00:05:12.285
Um, I would say that the, the, the tricky part is

119
00:05:12.285 --> 00:05:15.125
that even at that very, very close distance,

120
00:05:15.145 --> 00:05:16.405
the results are tantalizing,

121
00:05:16.545 --> 00:05:18.445
but not decisive despite the fact

122
00:05:18.445 --> 00:05:21.725
that we have almost 300 billboards in our study in

123
00:05:22.005 --> 00:05:23.445
155 catchment areas.

124
00:05:23.905 --> 00:05:26.325
So, you know, I would say that we did our level best

125
00:05:26.325 --> 00:05:28.005
to find a subtle effect.

126
00:05:28.355 --> 00:05:29.565
Here are many possible reasons.

127
00:05:29.865 --> 00:05:32.925
One is that this is a, by us standards,

128
00:05:32.965 --> 00:05:36.085
a high turnout election, and many of the people who are,

129
00:05:36.635 --> 00:05:40.005
otherwise on the border of voting have already voted of

130
00:05:40.135 --> 00:05:43.685
among the registered voters, which are roughly 75% of the,

131
00:05:43.825 --> 00:05:46.765
the American, eligible electorate,

132
00:05:47.435 --> 00:05:49.245
turnout is roughly 90%.

133
00:05:49.305 --> 00:05:51.965
And that's how we end up with our roughly two thirds,

134
00:05:51.965 --> 00:05:53.485
rate of turnout in 2020.

135
00:05:54.805 --> 00:05:56.905
The billboards were not effective at,

136
00:05:57.385 --> 00:05:59.285
Um, raising rates of registration.

137
00:05:59.785 --> 00:06:02.445
So it may be that this kind of passive,

138
00:06:03.135 --> 00:06:06.845
impersonal message broadcasted to people along the roadside,

139
00:06:07.265 --> 00:06:10.205
was not terribly effective in getting them to act.

140
00:06:10.545 --> 00:06:13.925
Now, that may be different from roadside signs with respect

141
00:06:13.925 --> 00:06:16.765
to restaurants or some other kinds of thing,

142
00:06:16.765 --> 00:06:18.485
that directs people where to pull off

143
00:06:18.485 --> 00:06:19.645
and, and eat immediately.

144
00:06:19.995 --> 00:06:23.965
Here, you'd have to let a little time go by to, get out

145
00:06:23.965 --> 00:06:26.565
of your car and go vote either by mail or in person.

146
00:06:27.025 --> 00:06:29.805
That's right. I, I think that, it would be hard

147
00:06:29.805 --> 00:06:34.485
to imagine billboards, for restaurants

148
00:06:34.505 --> 00:06:37.325
or other kinds of roadside conveniences not working,

149
00:06:37.325 --> 00:06:39.765
because you would think that eventually those,

150
00:06:39.985 --> 00:06:42.005
establishments would stop paying for them.

151
00:06:42.425 --> 00:06:45.565
But in the context of voter turnout, it's really never,

152
00:06:45.795 --> 00:06:47.845
it's never been tested in any direct way.

153
00:06:47.845 --> 00:06:50.125
And so that's why it was interesting to, to do this.

154
00:06:50.645 --> 00:06:52.245
I actually, I I take that back.

155
00:06:52.335 --> 00:06:55.085
There were tests of the sort that we did,

156
00:06:55.425 --> 00:07:00.045
but unlike ours, they, they, they randomized,

157
00:07:01.085 --> 00:07:02.565
locations within a city.

158
00:07:02.785 --> 00:07:03.845
So these were billboards

159
00:07:03.845 --> 00:07:04.965
that were very close to one another.

160
00:07:04.965 --> 00:07:08.485
Whereas ours basically is a, a study that randomized,

161
00:07:08.815 --> 00:07:12.205
randomized as the assignment of different catchment areas,

162
00:07:12.585 --> 00:07:14.765
that are quite far apart from one another.

163
00:07:14.885 --> 00:07:17.005
I think the one question is whether there's something

164
00:07:17.005 --> 00:07:19.205
special about the presidential election context.

165
00:07:19.585 --> 00:07:21.125
Uh, it's a high turnout context

166
00:07:21.345 --> 00:07:24.125
and, the United States has the most frequent election

167
00:07:24.365 --> 00:07:25.405
calendar in the, in the world.

168
00:07:25.665 --> 00:07:28.845
Um, we have very low salience elections, quite often,

169
00:07:29.425 --> 00:07:31.765
and it would be interesting to know whether,

170
00:07:31.935 --> 00:07:35.365
billboards work more effectively in low salience contexts.

171
00:07:35.625 --> 00:07:38.605
Um, it would also be interesting to know whether

172
00:07:39.625 --> 00:07:40.885
our rather bland,

173
00:07:40.955 --> 00:07:43.645
nonpartisan message could be amped up a bit

174
00:07:43.755 --> 00:07:45.525
with a different kind of message,

175
00:07:45.525 --> 00:07:46.645
either a partisan message

176
00:07:46.745 --> 00:07:48.565
or a message that is a bit,

177
00:07:48.565 --> 00:07:51.605
more striking than our quite, quite, you know,

178
00:07:52.035 --> 00:07:56.525
milk toast, black and white, billboard.

179
00:07:56.525 --> 00:07:59.405
Because we, we wanted to stay away from partisan coloration

180
00:07:59.425 --> 00:08:00.045
of all kinds.

Faculti AI can make mistakes

Research

Policy

Analysis