by Hugo Mercier
from the regime who appeared to accept its values.19 Even now,
132 ch ap t er 9
Rus sian propaganda efforts— for example, in Ukraine— follow
a familiar pattern: succeeding modestly when preaching to the
choir, backfiring when targeting opponents.20
Propaganda by the other great communist power, China, was
barely more convincing, even under Mao. Po liti cal scientist Shao-
guang Wang studied in detail what motivated the vari ous actors
of the Cultural Revolution in Wuhan, a large city of central
China.21 Instead of reflecting a “blind faith in Mao,” citizens’ en-
gagement with the Cultural Revolution was “a product of
[their] perception that Mao’s initiative would provide solutions
to [their] personal prob lems.”22 Those who stood to benefit from
taking up the cause did so; many others resisted.
Work on more recent propaganda attempts by the Chinese
government confirms its broad in effec tive ness. A study of Chi-
nese citizens’ attitudes toward the government in the mid-1990s
revealed that consumption of the state- controlled news media
correlated with lack of trust in the government, making it very
unlikely that the media successfully instil s trust in the leader-
ship.23 Lack of trust in official media means that Chinese citizens
are “always eager to get other information from diff er ent chan-
nels,” as one of them put it.24 Shortly after Weibo, the Chinese
equivalent of Twitter, started up, 70 percent of the Chinese who
used social media admitted relying on them as their primary
source of information.25 Mistrust of official media and increased
reliance on other sources mean that rumors presenting a nega-
tive view of the government are quickly taken up and prove dif-
ficult to fight.26 That Chinese citizens do not passively accept
government propaganda is also shown by their many acts of pro-
test. Journalist Evan Osnos, whose work I have drawn on here,
reports that, on average, in 2010 there were nearly five hundred
“strikes, riots, and other ‘mass incidents’ ” taking place across
China every day— and that is according to official statistics.27
p r o p a g a nd is t s, c a mp a i g ne r s, a d v e r t ise r s 133
Having had time to learn the limits of propaganda, the Chi-
nese Communist Party has shifted strategies for control ing the
public, away from brute persuasion toward what po liti cal scien-
tist Margaret Roberts calls “friction and flooding.”28 Friction
consists in making sensitive information more difficult to
access— blocking keywords, forcing people to use VPNs, or sim-
ply not collecting such information in the first place (e.g., infor-
mation on how wel , or poorly, such and such government agency
is performing, information that only the state could reliably
gather). Flooding consists in distracting people from sensitive
issues by bombarding them with official propaganda. The gov-
ernment is suspected of having recruited as many as two mil ion
people to spread messages online— known as the 50 Cent Party,
after the sum these shil s receive per post. Yet, the government
seems to have essentially given up on using these legions of pro-
pagandists to change people’s minds: they “avoid arguing with
skeptics . . . and [do] not even discuss controversial issues.”29 In-
stead, they try to either bolster the views of citizens supportive
of the regime in the first place (a significant number, as we will
see presently) or talk about other topics, such as celebrity gos-
sip, distracting the attention of citizens who do not care all that
much about politics.
In his book on nondemo cratic regimes, po liti cal scientist
Xavier Márquez notes several other failures of propaganda:
“Nearly 40 years of Francoist propaganda did not turn Spaniards
against democracy . . . constant exposure to the cult of Ceaușescu
did not turn most Romanians into his partisans . . . unrelenting
propaganda turned many East Germans into habitual cynics who
did not believe anything the regime said.”30
By and large, government propaganda fails to convince the
public. It can even backfire, leading to widespread distrust of the
regime. At most, propaganda surfs on preexisting opinions and
134 ch ap t er 9
frees people to express what might other wise be seen as socially
objectionable views.31
Why, then, do some people in authoritarian regimes behave as
if they had been brainwashed, saluting the führer in unison, buy-
ing bil ions of Chairman Mao badges, wailing at Kim Jong- il’s fu-
neral? The answer is simple. Every authoritarian regime that relies
on propaganda also closely monitors and violently represses signs
of dissent. Failure to perform the Nazi salute was perceived as a
symbol of “po liti cal non- conformism,” a potential death sen-
tence.32 In North Korea, any sign of discontent can send one’s
entire family to prison camps.33 Under such threats, we cannot
expect people to express their true feelings. Describing his life dur-
ing the Cultural Revolution, a Chinese doctor remembers how “to
survive in China you must reveal nothing to others.”34 Similarly, a
North Korean coal miner acknowledged, “I know that our regime
is to blame for our situation. My neighbor knows our regime is to
blame. But we’re not stupid enough to talk about it.”35
When it comes to genuine support, rather than empty dis-
plays, it seems carrots work better than sticks. Chinese citizens
might not widely trust the state media, but, on the whole, they
re spect and support the central government and the Chinese
Communist Party— which typically garners more than
70 percent approval, higher than any Western government.36 It
could be propaganda. Or it could be that under the party’s di-
rection, China has had high growth rates for de cades, lifting eight
hundred mil ion people out of poverty.37
Campaigners
As we have seen repeatedly, forceful attempts at mass persuasion
by propagandists in authoritarian regimes fail to sway the popu-
lation. Rather than the public exercising due vigilance, however,
p r o p a g a nd is t s, c a mp a i g ne r s, a d v e r t ise r s 135
could it be that these failures reflect instead the propagandists’
lack of sophistication or skil ? Goebbels, for example, didn’t seem
to have been much of a master influencer: by 1940 already, citi-
zens had lost any interest in official propaganda, on account of
its “boring uniformity” (as reported by the Nazis’ intel igence
ser vice).38
The campaign man ag ers, spin doctors, marketers, pol sters,
crisis con sul tants, and other specialists who proliferate in con-
temporary democracies may be more astute. Authoritarian pro-
paganda relies on monopolistic control of the media: Maybe the
lack of competition has blunted the instincts and motivations of
the chief propagandists? By contrast, modern po liti cal campaigns
are fiercely fought, providing p
lenty of opportunities for profes-
sionals to refine their skil s and learn how to guide a candidate
to victory, as well as for candidates to figure out who can best help
them get elected.
I focus here on U.S. politics, for two reasons. First, U.S. politi-
cians vastly outspend other politicians: in 2016, $6.4 bil ion was
spent on po liti cal campaigns (a third on the presidential race).39
Second, that’s where the vast majority of studies are conducted.
If the amounts involved are extraordinary, U.S. electoral cam-
paigns—in par tic u lar the most high- profile races— are other-
wise similar to campaigns elsewhere in being presented by the
press as dramatic events, full of plot twists, with candidates going
up and down in the pol s as a function of devastating ads, mov-
ing speeches, and per for mance in public debates. Indeed, given
the means available— the army of volunteers canvasing door to
door, the hours of TV ads, the countless robocalls—we would
expect commensurately dramatic results.
Yet research on whether po liti cal campaigns and the media
can win elections, or sway public opinion more generally, has
given surprisingly ambiguous results. In the first de cades of the
136 ch ap t er 9
twentieth century, a popu lar model was that of the “hypodermic
needle” or the “magic bullet,” according to which people would
pretty much accept what ever the media were tel ing them.40 This
model was based on the innovative (but likely in effec tive) use
of propaganda in World War I, and on a view of the public as re-
acting reflexively to just about any stimulus they encountered.41
One thing this model was not based on, though, was data. As
opinion pol s, tracking of voting be hav ior, and proper studies of
media influence arose, in the 1940s and 1950s, the era of “mini-
mal effects” began.42 Summing up years of research, Joseph Klap-
per stated in 1960 that po liti cal communication “functions
more frequently as an agent of reinforcement than as an agent
of change” (a conclusion reminiscent of what the research on
propaganda shows).43
The 1970s and 1980s saw the rise of experimental studies in
po liti cal science. Instead of mea sur ing people’s opinions in the
field, researchers would bring participants into the lab, expose
them to vari ous stimuli— campaign materials, TV news, and so
forth— and mea sure the influence of these stimuli on the par-
ticipants’ opinions. These techniques revealed that the media had
the potential to influence public opinion: not by tel ing people
what to think but by tel ing people what to think about (agenda
setting), how to best understand issues (framing), and what crite-
ria to use when evaluating politicians (priming).44 Although these
effects are less direct than those suggested by the hypodermic-
needle model, they could still be power ful: people who evaluate
politicians according to their economic policies rather than
their views on abortion (say) are likely to vote differently.
The advantage of these lab- based techniques is the rigor
of their methods, as they allow researchers to perform well-
controlled experiments, with participants randomly exposed
to diff er ent stimuli, their reactions carefully monitored. The
p r o p a g a nd is t s, c a mp a i g ne r s, a d v e r t ise r s 137
drawback of these methods is their lack of so- called ecological
validity: it is hard to tell whether the phenomena observed in
the lab happen in the uncontrolled environment of real life. For
instance, some studies showed that exposure to diff er ent pieces
of news on TV could lead to changes in po liti cal opinions. In
real life, however, people aren’t passively exposed to TV news:
they choose which news to watch, or even whether to watch the
news at all. Po liti cal scientists Kevin Arceneaux and Martin
Johnson conducted a series of studies in which participants had
greater leeway in which channel to watch. They observed not
only that many people simply tuned out but that those people
who chose to watch the news were those with the most po liti cal
knowledge, people who were also less likely to change their
minds in reaction to what they saw on the news.45 Still, even
looking at more ecologically valid studies, it is clear that po liti cal
campaigns and the media can shape public opinion on some issues.
But the way they do so reveals that people do not unquestion-
ingly accept what ever message po liti cal campaigns put forward.
By far the most impor tant moderator of whether campaigns
or the media influence public opinion is the strength of people’s
prior opinions. On the vast majority of po liti cal issues, people
have no strong opinion, or even no opinion whatsoever— which
makes sense, given the time and effort required to garner infor-
mation on any topic. For example, in the run-up to the 2000 U.S.
presidential election, few voters were aware of what position
George W. Bush and Al Gore (the two main candidates) held on
Social Security.46 As a result, when people were told that the can-
didate from the party they favored had such and such opinion,
they tended to adopt this opinion, following “party cues.”47 Fol-
lowing party cues reflects the (largely) sound working of trust
mechanisms: if you have come to trust a party over many years,
it makes sense to follow its lead on issues about which you have
138 ch ap t er 9
little knowledge. Citizens are also quite skilled at recognizing
who among them is most knowledgeable on po liti cal issues, and
at taking their opinion into account.48 On the whole, people are
more influenced by reliable signals. For example, a newspaper
sends a more reliable signal when it endorses a surprising
candidate— one who doesn’t belong to the party usually sup-
ported by the newspaper— and people are only influenced (if
at all) by these surprising endorsements.49
In the first de cade of the twenty- first century, po liti cal scien-
tists began conducting large- scale experiments on the effective-
ness of po liti cal campaigns, sending flyers to a random subset of
counties, canvassing a random subset of houses, calling a random
subset of potential voters, and so forth. Opinion surveys or vot-
ing outcomes were then recorded, allowing the researchers to
precisely estimate the effects of their intervention— the letter,
the face- to- face discussion, the call—on participants who had
been exposed to it, compared with other wise similar participants
who hadn’t. This methodology offered the best of both worlds:
it was rigorous yet ecologically valid.
In 2018, po liti cal scientists Joshua Kalla and David Broockman
published a meta- analysis of all the studies that respected these
rigorous methods, to which they added some new data of their
own.50 Some of the campaign efforts carried out a long time be-
fore the election had a small but significant effect on voting
inte
ntions. Early on in the election cycle, people have had less
time to develop fixed ideas about who they are going to vote for,
making their opinions slightly more labile. However, these effects
were never long- lasting and had all but dis appeared by Election
Day, so that the campaign efforts had no net effect on voting be-
hav ior.51 Other studies have shown some effects of campaign
efforts on elections for which voters have few preconceived ideas,
p r o p a g a nd is t s, c a mp a i g ne r s, a d v e r t ise r s 139
as they can’t rely on candidates’ affiliations, such as primaries or
ballot mea sures.52
When it comes to the big prize— voting on congressional or
presidential elections— the overall effect of the campaign efforts
studied was nil.53 This is quite a remarkable result. In spite of the
huge sums sunk into mailing, canvassing, calling, and advertis-
ing, campaign interventions in the most salient elections (in the
United States at least) seem, as a rule, to have no effect.
The most recent, sophisticated- looking techniques have done
nothing to challenge this conclusion. Many of you will have
heard of Cambridge Analytica, the infamous firm that harvested
data from Facebook users (often without their consent), created
psychological profiles of these users, and offered po liti cal cam-
paigns ads targeted specifically to these psychological profiles.
According to the Guardian, Cambridge Analytica allowed “de-
mocracy [to be] highjacked.”54
In fact, it was a scam.
Targeted advertising can, it seems, have some limited effects,
but these have only been proven on product purchases, with rel-
evant data on the users’ profiles, and the effects were tiny, add-
ing a few dozen purchases after mil ions of people had seen the
ads.55 Cambridge Analytica was attempting to influence presi-
dential elections (something no ad has been shown to do) with
dubious data on users. Even if the influence of Cambridge Ana-
lytica’s campaign had been as large as that recorded in experi-
ments on beauty products, it would only have swayed a few thou-
sand voters. In real ity, its influence was likely nil. Republican
po liti cal analysts remember the Cambridge Analytica employ-
ees “throwing jargon around,” but they never saw “any evidence
it worked”— unsurprisingly as it was based, still in their words,