ArticlePDF Available

The Global Resonance of Human Rights: What Google Trends Can Tell Us



Where is the human rights discourse most resonant? We use aggregated cross-national Google search data to test two divergent accounts of why human rights appeal to some populations but not others. The top-down model predicts that nationwide interest in human rights is attributable mainly to external factors such as foreign direct investment, transnational NGO campaigns, or international legalization, whereas the bottom-up model highlights the importance of internal factors such as economic growth and persistent repression. We find more evidence for the latter model: not only is interest in human rights more concentrated in the Global South, but the discourse is also most resonant where people face regular state violence. In drawing these inferences, this article confronts high-level debates over whether human rights will remain relevant in the future, and whether the discourse still animates counter-hegemonic modes of resistance. The answer to both questions, our research suggests, is “yes.”
American Political Science Review (2023) 122
doi:10.1017/S0003055423000199 © The Author(s), 2023. Published by Cambridge University Press on behalf of the American Political
Science Association. This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://, which permits unrestricted re-use, distribution and reproduction, provided the original article is
properly cited.
The Global Resonance of Human Rights: What Google Trends Can
Tell Us
GEOFF DANCY University of Toronto, Canada
CHRISTOPHER J. FARISS University of Michigan, United States
Where is the human rights discourse most resonant? We use aggregated cross-national Google
search data to test two divergent accounts of why human rights appeal to some populations but
not others. The top-down model predicts that nationwide interest in human rights is attributable
mainly to external factors such as foreign direct investment, transnational NGO campaigns, or interna-
tional legalization, whereas the bottom-up model highlights the importance of internal factors such as
economic growth and persistent repression. We find more evidence for the latter model: not only is interest
in human rights more concentrated in the Global South, but the discourse is also most resonant where
people face regular state violence. In drawing these inferences, this article confronts high-level debates over
whether human rights will remain relevant in the future, and whether the discourse still animates counter-
hegemonic modes of resistance. The answer to both questions, our research suggests, is yes.
What does the future hold for human rights?
Scholars offer numerous and competing pre-
dictions (Baxi 2008; Hopgood, Snyder, and
Vinjamuri 2017; Moyn 2014; Rodríguez-Garavito
2014a;2014b). A revisionist camp with roots in struc-
tural realist and critical theory traditions expects that
the lingua francaof human rights will gradually lose
its relevance.
The ideological export of a Western
cosmopolitan elite, the human rights discourse has
since its postwar incarnation grown only more profes-
sionalized, legalistic, and detached from the plight of
the poor and downtrodden (Englund 2006; Hopgood
2013; Moyn 2018). As such, those seeking empower-
ment against the vagaries of global capitalism and other
forces of domination may soon decide that human
rights is not the right languagefor their struggle
(Mutua 2008, 1034).
Another camp with allies in liberal and constructivist
schools forecasts the opposite for the future of the
Human rights will continue to serve as a
resilient and flexible tool of social movements
worldwide (Chase 2012; Stammers 1999), commanding
an ever-expanding popular constituency that resorts to
rights claims for self-empowerment and resistance to
injustice in various forms (Benhabib 2009; Simmons
2009). Therefore, one should expect a transnational
human rights discourse to adapt, evolve, and thrive
well into the foreseeable future (Dancy and Sikkink
2017; Rodríguez-Garavito 2014a;2014b).
Though ostensibly about the future, this debate is
really about how to understand the current moment.
Are human rights claims still relevant, or is the world
on the cusp of a collective moving-on? There is much
at stake here: if the human rights discourse is no
longer salient, and thus fails to inspire mobilization
on behalf of the oppressed, then investments should
shift to developing more promising political pro-
grams. However, the truth is that it is difficult to
generalize with confidence about the popularity of
human rights. Worldwide, some populations are more
likely than others to embrace the language of human
rights. Therefore, to anticipate future trends, we must
first ask a more fundamental question: what might
explain the mass appeal of human rights in different
Some scholars have referred to wide uptake of the
human rights vocabulary as resonance (see Chase
2012;2016), which could be defined as the regular
appropriation, adoption, or amplification of a repro-
ducible language by a designated population. Any
statements predicting a bright or bleak outlook for
human rights, or discussing how the discourse gains or
loses traction, presume knowledge about resonance.
However, scholars are rarely forthcoming about how
they arrive at that knowledge. In the words of Chase
(2012,506),standard explanations in the literature
explaining human rightsresonance are insufficient.
Our goal is to confront this problem. We assume that
Geoff Dancy , Associate Professor, Department of Political Sci-
ence, University of Toronto, Canada,
Christopher J. Fariss , Assistant Professor, Department of Political
Science, University of Michigan, United States,
Received: February 17, 2021; revised: December 02, 2021; accepted:
February 24, 2023.
On lingua franca,see Law (2018), among others.
We do not advance a specific definition of discourse, or engage
extensively with discourse theory. Following Gordon and Berkovitch
(2007, 244) and others, we use the term discoursein a Foucauldian
sense,to mean the relationship between language and a set of
observable social, legal, and political practices.
1 Published online by Cambridge University Press
an observable indicator of human rights resonance,
which possesses an intersubjective dimension, is the
degree to which a population demonstrates a collec-
tive interest in the subject. And, one way a group of
people may convey collective interest in a subject is by
regularly searching for it on the internet. Based on this
assumption, we use Google aggregate search trends
data to test competing hypotheses about where the
human rights discourse will take hold and flourish.
This article proceeds in four parts. First, we identify
how current scholarship makes claims about human
rights without paying sufficient attention to the empir-
ical validity of key assumptions. Second, we explain
why Google aggregate search trend data can help us
evaluate these assumptions, and ultimately resolve
some problems of evidence that plague studies of res-
onance. In short, analyzing Google search data on
human rightscreates unique opportunities to capture
the latent, or hard to observe, interests of populations
over time, and to do so in a way that addresses short-
comings in archival and survey research.
Third, drawing on a wide array of literatures, we
construct two predictive models of domestic human
rights resonance. The top-down model treats human
rights talkas a Western cultural script that is
imposed on, transmitted to, or diffused across foreign
polities through various channels of influence.
bottom-up model presumes that human rights interest
primarily evolves out of processes endogenous to
those polities, including economic development or
local resistance to repression. The advantage of these
models is that they provide specific, testable hypoth-
eses. The top-down model holds that cross-national
variations are best explained by international factors,
while the bottom-up model holds that those variations
are rooted in domestic political and economic
Our fourth move is to evaluate these models empir-
ically. We start by examining broad temporal and
spatial trends in human rights interest, and uncover a
little-known fact: worldwide, the countries with the
highest volume of web-based searches for human rights
are not advanced democracies in the OECD, but chron-
ically under-privileged developing countries in Central
America and Africa. With regard to cross-national
variations, we find that the most powerful correlates
of human rights interest across countries are not factors
like foreign direct investment from the Global North or
transnational NGO campaigns. Rather, across the
board, the search for human rights is most pronounced
in countries where economic growth is afoot, and citi-
zens are routinely subjected to state violence. In other
words, the human rights discourse remains resonant,
especially in places where it is needed most.
Thinkers fundamentally disagree over the nature of
human rights. A great deal of that disagreement
revolves around the perceived contributions of the
human rights project”—past, present, and futureto
the aim of emancipation (Engle 2021). The following
two statements are illustrative:
Statement 1:Social and political systems become
hegemonic by turning their ideological priorities into
universal principles and values. In the new world order,
human rights are the perfect candidate for this role.
Their core principles, interpreted negatively and eco-
nomically, promote neoliberal capitalist domination
(Douzinas 2008).
Statement 2:Since the end of the Second World
War, a wide range of social movements have sought to
challenge existing forms of power. Indeed in the
previous 60 years, it was the oppressed of the world
mobilized in and through social movementswho were
the hidden authors of developments in human rights
(Stammers 2013).
Statement 1 interprets the discourse as hegemonic.
To this line of thinking, in a historical moment dominated
by Western economic, military, and cultural power,
human rights serve as a tool of subjugation to, or mar-
ginalization within, Empire (Hardt and Negri 2001).
In this vein, contemporary critics charge that the human
rights corpus: instantiates legacies of colonialism
(Inayatullah and Blaney 2012); reproduces neoliberal
subjecthood and exploitation (Odysseos 2010;Whyte
2019); justifies violent Western interventionism (Abu-
Lughod 2002;Zizek2005); supports liberal do-
goodingand paternalism (Budabin and Pruce 2018;
Gourevitch 2009); reinforces white saviorism (Mutua
2001; Tascón and Ife 2008); or too easily becomes
hijacked by right-wing forces (Perugini and Gordon
2015). Central in these accountsis the notion that human
rights are imbricated in structures of domination. At best,
the language fails to empower activists in their struggle
for more equitable social arrangements (Brown 2004).
At worst, it actively disempowers would-be agents of
change by blinding them to alternative avenues for
resistance (Engle 2021). In the words of Kapur (2014),
human rights are something we cannot not want, even
though they cannot give us what we want.
Statement 2 represents an opposite, counter-
hegemonic orientation toward human rights, the logic
of which is this: despite the fact that great powers like
the United States played an influential role in the
formation of the postwar human rights doctrine, and
at times instrumentalize human rights in the conduct of
We use the term latentto refer conceptually to unobservable
phenomena that have observable manifestations, but we are not
estimating a latent variable model, as in Fariss (2014;2019) and
Fariss and Dancy (2017). For instance, we do not have direct access
to individualssearch queries, nor can we make predictions about
individual motivations. However, based on Google data, we can
make certain assumptions about aggregate interests in a population
(see the section Google as Method).
For rights talk, see Glendon (1993).
For our purposes, it does not matter whether one defines
hegemonyin the Gramscian terms, or in International Relations
Geoff Dancy and Christopher J. Fariss
2 Published online by Cambridge University Press
foreign policy, the West does not own or control the
discourse (Sikkink 2017). Instead, human rights belong
to the multitude (Hardt and Negri 2004). Those who
mobilize this language do not constitute a vertical, elite-
driven social movement, but a decentralized, polycen-
tric, and multivocal amalgamation of subaltern move-
ments (Baxi 2008; Chase 2012;OConnell 2007;
Stammers 1999). Aligned with this position are works
that: recover the alignment between decolonization
and the global human rights regime (Jensen 2016);
document the use of human rights in community-based
organizing against neoliberal policies (Dunford 2019);
recount how rights activists struggle doggedly against
American militarism (Neier 2012); theorize how popu-
lations engage in human rights localization, vernacular-
ization, and reverse standard-setting (Acharya 2004;
Destrooper 2017; Merry 2006); show how human rights
movements have emerged even in localities insulated
from Western cultural influence (Mokhtari 2009; Nick-
erson 2020); and outline how existing movements
amplified their message by adopting a human rights
frame (Langlois 2020). Crucial in these accounts is the
role of agency. Neither the content nor the utility of
human rights claims is determined by power structures.
Instead, the language can serve as an enduring resource
for expressing political grievances and pursuing eman-
Whether one conceives of human rights discourse as
hegemonic or counter-hegemonic will shape ones pre-
dictions about discursive resonance. If human rights are
an extension of American imperium, then the fragmen-
tation and decline of American power will inevitably
weaken the pull of rights claim-making worldwide
(Hopgood 2013, chap. 5). This is why Hopgood states
that the thing most likely to stall human rights pro-
gress is people around the world simply not considering
them to be as important as their advocates would have
us believe(Hopgood 2018). If, however, the human
rights discourse is powered by a global constituency of
local activists, then the decline of American power is
unrelated to expectations about resonance. In response
to Hopgood, Chase (2013) makes this exact point: The
impetus in defining human rights has long since ceased
to be unipolar, if it ever was. Human rightsendtimes
will only come when states no longer bother rebutting
them and activists no longer seek to own them.
How do scholars arrive at such divergentoften
totalizinginterpretations? How do they know what
they know? A first place they turn is philosophy.
Assessments of the human rights discourse are driven
in part by views on the nature of language and agency.
Theorists differ over whether people can use a dis-
course to accomplish their good aims, or whether dis-
courses constitute people, and hence their aims too
deeply for individual intention to guide their use
(Wahl 2019, 15). One may hold that people instrumen-
tally select human rights from a menu of claim-making
grammars to pursue their goals, or one may hold that
actors are in effect rendered choiceless by the all-
consuming and productive force of human rights
tropes. Insofar as these characterizations derive from
a priori reasoning, they cannot be assessed with
empirical research. However, it is likely that ones
ontological commitments are influenced, at least in
part, by evidence drawn from the observed world
(Wendt 1999,356). A thinker cannot entirely decide
what human rights mean, or what the future holds for
human rights, if one has never seen human rights in
For evidence, some thinkers turn to a second place:
history. Scholars use archival data on the original draft-
ing and dissemination of human rights doctrine to draw
inferences about its meaning today. A growing body of
scholarship focuses on the historical precursors to
human rights (Hunt 2008; Weitz 2019); the intellectual
authors of human rights statements in the immediate
postwar period (Borgwardt 2005; Glendon 1999; Sik-
kink 2017); and the evolution of human rights norms,
law, and advocacy in the decades following the Univer-
sal Declaration (Buergenthal 1997; Clark 2001). The
implicit assumption of this work is that the history of
human rights provides insight into the way human
rights presently operate. For example, much has been
made of the simultaneous emergence of global rights
advocacy and the rise of neoliberal orthodoxy in the
1970s (Moyn 2010). Some treat this historical fact as
dispositive: it proves either that human rights help
justify neoliberalism (Whyte 2019), or at the very least
serve as its powerless companion(Moyn 2018). How-
ever, it is unclear whether the history of human rights is
determinative (Alston 2013). It is possible that human
rights were once aligned with neoliberal hegemony but
have since been appropriated and re-fashioned for
counter-hegemonic purposes (Rajagopal 2006). For
this reason, it might be a futile quest to seek out the
human rights foundation as a way to understand what
human rights are today(Chase 2012, 514).
A third resource for assessing human rights is social
science on contemporary practice. If one thinks of the
human rights discourse as a market, most scientific
studies focus on the supply side. Researchers shine light
on the dynamics of norm promotion, detailing where
transnational advocacy groups originate (Keck and
Sikkink 1998; Smith, Pagnucco, and Lopez 1998); how
transnational organizations are funded (Ron, Pandya,
and Crow 2016); and in what ways international human
rights NGOs behave strategically (Murdie and Bhasin
2011; Wong 2012). These questions have inspired vig-
orous data collection efforts on human rights organiza-
tions and the mounds of written material they produce
(Cordell et al. 2020; Lebovic and Voeten 2006; Ron,
Ramos, and Rodgers 2005). Based on what they know
about the supply side of human rights, social scientists
make guesses about who is consuming, or using, the
human rights message. But in general, they possess very
few details on this demand side of the equation. Some
systematic case study research delves into the emer-
gence and mobilization of human rights in widely dis-
parate contexts such as Brazil (Dunford 2019), China
(Merry 2006), Israel/Palestine (Perugini and Gordon
2015), or Iran (Nickerson 2020). However, much less
attention is devoted to collecting valid cross-national
data on human rights uptake. Instead, observers offer
reductionistclaims about the global (non)resonance
The Global Resonance of Human Rights
3 Published online by Cambridge University Press
of human rights based mostly on impressions or anec-
dotal data (Dancy and Fariss 2017; Rodríguez-Garavito
One reason variations in resonance are under-
studied is that overall interest in human rights is a latent
population characteristic. It exists, but it is very difficult
to observe directly. One way to capture latent popula-
tion variables is to survey samples of that population.
The most thorough cross-national survey study of pub-
lic attitudes toward human rights is Ron et al.s(2017)
Taking Root. The authors find that only a small portion
of individuals in developing countries are directly
familiar with the work of human rights organizations,
though they generally trust rights defenders. The
authors base this argument about exposure to human
rights in the Global South on a total of 13,277 sampled
individuals from six different countries: Colombia,
Ecuador, Mexico, Nigeria, Morocco, and India (Ron
et al. 2017, chap. 2). On the basis of these data, Ron
et al. note the widespread familiarity with the concept
of human rights, but worry that it is mostly elites who
are ever exposed to human rights NGOs and advocacy
networks. In a separate article, the authors claim that
human rights are more toprootsthan grassroots
(Ron, Crow, and Golden 2013).
One benefit of sophisticated social science like Ron
et al.sTaking Root is that it produces precise results.
Compared to philosophical interpretations, archival
sources, or interviews drawn from convenience sam-
ples, randomized surveys allow for methodologically
sound inferences about human rights knowledge and
reception. However, two problems remain. The first is
that surveys are limited instruments. They are frozen in
particular geographical locations in particular
moments. As such, it is difficult to use surveys to
observe trends. It is simply not feasible to draw samples
from all countries continuously over time, and thus we
cannot know how or why interest in human rights
changes (Mellon 2013). Even the World Values Survey,
with its extensive geographic and temporal coverage,
can only present us with a few data points related to
rights for any given country, every few years. If we are
to study the resonance of human rights, we need a
reliable tool that at once provides more geographical
and temporal coverage of discursive shifts. In the fol-
lowing section, we propose that Google aggregate
search data is fit for this purpose.
A second problem is that the literature by and large
still lacks a causal theory for why the language of
human rights resonates in some populations but not
in others. According to Gordon and Berkovitch (2007,
245), two of the only authors to confront this question
directly, very little research has focused on how
human rights discourse surfaces in the domestic
sphere.What the literature does contain are two basic
modelsthe top-down and bottom-up accounts of
rights resonanceeach of which might help us gener-
ate expectations about where, and when, interest in
human rights will be most pronounced.
On the one hand, the top-down model treats human
rights as a language whose worldwide popularity is
attributable mainly to the influence of Western powers.
Some authors characterize the spread of the discourse
as a type of foreign imposition(Dicklitch and
Lwanga 2003, 485), where others use more passive
terms such as transmissionor diffusionto
describe how non-Western polities come to embrace
human rights idioms (Goodman and Jinks 2004,673;
Greenhill 2010). Regardless of terminology, these
accounts emphasize exogenous, outside-in factors to
explain why the language becomes a more relied-upon
source of political and legal claims in some contexts.
On the other hand, the bottom-up model approaches
this phenomenon from the opposite direction, often
referring to the outcropping of human rights interest
as organic and the development of human rights
movements as grassrootsor from below.While
proponents of this model also recognize that the dis-
course of human rights is a widely available script
attached to the liberal international order, they would
still more likely explain variations in discursive reso-
nance across countries by pointing to endogenous,
inside-out, or local factors. Though similar to the
opposition between hegemonic and counter-
hegemonic conceptions, top-down and bottom-up
accounts more readily lend themselves to empirical
evaluation, as we will show. Before we develop test-
able hypotheses based on these models, however, we
first turn to our method for utilizing Google aggregate
search data.
To address lingering questions about human rights
resonance, we may need to explore new strategies of
measurement and assessment. Google Trends data,
which capture average internet search patterns among
defined populations in time and space, may help fill
gaps left by survey work. Google data are utilized
widely in public health, economics, and media studies
primarily to measure information-seeking behavior
and the trending concerns of large groups (Jun, Yoo,
and Choi 2018). For example, Google searches are used
to predict influenza epidemics (Ginsberg et al. 2009)
and to forecast stock market swings (Preis, Moat, and
Stanley 2013). In political science research, though,
Google Trends data are not as commonly used as other
sources of large-scale digital sources of information.
Notable exceptions are Pelc (2013), who employs Goo-
gle search data to measure information-seeking about
the WTO and global trade relations, and Kalmoe
(2017), who demonstrates that Google-driven news-
seeking on drone attacks increases in the United States
and Pakistan after major strikes on militant leaders.
More germane to our topic, researchers have selec-
tively used Google data to make arguments about
media framing of human rights crises (Dabbous
2018), or the hidden impactof International Criminal
Court interventions on human rights salience in states
under investigation (Dancy 2021).
Our aim is broader than most of these studies. We
examine aggregate worldwide searches for human
rightsin the Google search engine across five different
Geoff Dancy and Christopher J. Fariss
4 Published online by Cambridge University Press
language groups: English, Spanish, Portuguese,
French, and Arabic. In using aggregate Google
searches as evidence, we make the following simple
assumption: the more a defined population searches for
human rights,the more that population possesses a
latent interest in human rights as a field of knowledge
and practiceergo, the more the discourse resonates
(cf. Foucault 1990, chap. 1). If our assumption is con-
ceptually valid, then the aggregate search data are
generated through the following four-step process:
(1) individuals in a population jointly receive some
cue that sparks their interest in human rights; (2) they
turn to the internet to privately seek information;
(3) when they do, they decide to use Google as a search
tool; and (4) they key in human rightsto begin
learning or discovering resources. If any step in this
data-generating process is broken, it casts serious doubt
on our assumptions.
Throughout the article, we use Google search rates
as a dependent variable. This publicly available mea-
sure is a transformation of publicly unavailable search
ratios, which are the total number of Google searches
containing human rightsdivided by the total number
of all Google searches in a defined population for a
particular time period. Notably, Google the company
does not provide raw search totals to researchers;
instead, the Google Trends tool offers data that
are already transformed using minmax normalization
(see Section A of the Supplementary Material for
more details). The unit of analysis is the population-
period, which we can define as global-week, global-
year, country-week, or country-year. The observed
population-period with the maximum ratio of human
rights searches to all Google searches receives a score
of 100. The observation with the minimum ratio of
human rights searches to all Google searches would
receive a score of 0. All other population-period values
are defined in relation to these maximum and minimum
values. This minmax transformation allows one to
make relative comparisons across units but not with
respect to the absolute number of Google searches.
Using search rates, countries are scored not by how
wired they are, but by how much they look up a
particular search term, human rights,relative to
other search terms. In addition, because the denomi-
nator is total searches, the search rate indicators we
employ control for variation in population size and
internet penetration across countries.
A justifiable concern is that Google search rate is a
simplistic indicator that cannot be validated (Mellon
2013). An indicator of what proportion of global
country-wide searches contains the phrase human
rightsis seemingly a crude measure. Would it not be
more advantageous to gather data on specific searches
related to, say, torture or the right to food? To this point,
a couple of facts. First, human rightsis searched in
high volume across the globe.
Figure 1 plots pair-wise
comparisons, using minmax transformations, of vari-
ous search terms at the global-week level over a 5-year
period. Though not as frequently searched as terms
such as time,war,god,orrace,human rightsis still far
more commonly searched than quotidian political
terms such as terrorism,national security,orsocial
justice. To put it in perspective, human rightsis
Google searched in English about as often as malaria,
a very common and dangerous disease.
The high
volume of global searches for human rightsevery
year makes for richer and more meaningful variations
across cases in comparison to (1) other similar, high-
volume search terms like torture, which also yields a
flood of unrelated (often sadistic) content; or (2) spe-
cific, low-volume search phrases like right to food,
which is seldom keyed in across cases in various lan-
Second, how do we know that individuals are not
searching human rightsbecause they view this dis-
course in a negative light? Might they key in human
rights are neo-colonialor human rights are
pointless? Of course, not all searches for human rights
are positively motivated. However, Google Trends
does provide tools for observing what additional search
terms or phrases are most related to queries about
human rights. These are called co-occurrenceterms
because they follow similar search patterns. In nearly
every case we observe, aggregate interest in human
rightsis connected to information-collection on sub-
jects like the Universal Declaration of Human Rights,
the United Nations, and local Human Rights Commis-
Collectively, individuals turn to Google not to
dig up dirt but to learn and make connections about
human rights norms and institutions (co-occurrence
patterns are described in more detail in Section D of
the Supplementary Material).
As it pertains to questions of resonance, collecting
aggregate search data on human rightsis not all that
different from methods employed in other studies.
Scholars make empirical arguments based on the pres-
ence of human rightsin print mediaincluding fre-
quency of use in The New York Times and other Anglo-
American periodicals through the twentieth century
(Dancy and Fariss 2017; Moyn 2010); the number of
One could calculate raw search totals if one knew the maximum and
minimum value for the set of units we are comparing. But Google
does not share these data.
Written information on human rightshas increased markedly in
various languages over the last 50-year period (Dancy and Fariss
2017). See Section K of the Supplementary Material.
On this point, we note that Google searches for malariaover-
whelmingly occur in countries where malaria infection is high. To
validate the use of Google searches as a measure of latent interest
within a population, we show that relative rate of Google searches for
the term malariaby country are correlated with the actual preva-
lence of the malaria mortality rate out of one hundred thousand
people at 0.82 (n=75) for the English (or Spanish) language group,
0.99 (n=12) for the Portuguese language group, and 0.81 (n=44) for
the French language group. See Section I of the Supplementary
Material for additional details.
Not all human rights content co-occurs in the aggregate. For
example, searches for Amnesty Internationalare globally fewer
and concentrated in different countries than searches for human
rights.See Section H of the Supplementary Material.
The Global Resonance of Human Rights
5 Published online by Cambridge University Press
FIGURE 1. Pairwise Comparisons of Relative Search Term Rates
Global: time vs. war
2015 2016 2017 2018 2019
Global: war vs. god
2015 2016 2017 2018 2019
Global: god vs. race
2015 2016 2017 2018 2019
Global: race vs. injustice
2015 2016 2017 2018 2019
Global: injustice vs. human rights
2015 2016 2017 2018 2019
human rights
Global: human rights vs. malaria
2015 2016 2017 2018 2019
human rights
Global: human rights vs. torture
2015 2016 2017 2018 2019
human rights
Global: human rights vs. terrorism
2015 2016 2017 2018 2019
human rights
Global: human rights vs. national security
2015 2016 2017 2018 2019
human rights
national security
Global: human rights vs. social justice
2015 2016 2017 2018 2019
human rights
social justice
Global: human rights vs. right to food
2015 2016 2017 2018 2019
human rights
right to food
Global: human rights vs. worker's rights
2015 2016 2017 2018 2019
human rights
worker's rights
Note: In each plot, the purple line represents the higher searched term relative to the green line that represents the lower searched term.
Moving from left to right, in the top-left panel, the term timeis searched more often than war.In the top-middle panel, the term waris
searched more than god,and so on. The term human rightsis searched slightly more than the terms terrorismand malaria,and more
often than the term injustice.
Geoff Dancy and Christopher J. Fariss
6 Published online by Cambridge University Press
articles referring to human rightsover a 13-year
period in the Israeli newspaper Haaretz (Gordon and
Berkovitch 2007); or the number of references to
human rightsin four Filipino newspapers over a
recent 4-year period (Chaudoin 2020). The difference
between those studies and ours is that they are focused
on human rights in the public sphere, while we focus on
human rights interest at the private, individual level,
which in the aggregate produces a measure of
population-wide interest.
In this sense, Google search data are similar to survey
data, but they have three comparative advantages.
First, these data allow one to observe aggregate search
totals down to the weekly level over a series of 5 years,
yielding very detailed geographical and temporal pat-
terns. For any given country, we possess hundreds of
observations measuring the relative frequency with
which the population uses Google to access informa-
tion on human rights, as well as the city-wide and
regional breakdown of aggregate searches. A second
advantage is that, in most cases, people are typing
human rightsinto their digital devices on their own
volition, which means that aggregate data reflect indi-
vidual cognitive impulses that drive the search for
human rights. People will not admit certain things even
on anonymous questionnaires, and their attitudes are
often shaped by their interaction with survey scientists.
Internet browser data offer a way around this source of
error, to some extent (Stephens-Davidowitz 2017).
A third and final advantage of Google search data is
that they may be leveraged to evaluate theories that are
panoramic in scope. Claims about the waxing or waning
relevance of human rights can be sweeping, which
trades off with empirical specificity. For example, Hop-
good defines the endtimes of human rights by a general
decline in the appeal of the global human rights dis-
course to address practical problems. But what exactly
is the empirical expectationa uniform and simulta-
neous dip in human rights interest across the globe? Or
would the endtimes begin in certain countries or
regions, only to become contagious? In what follows,
we draw on the literature to develop precise hypotheses
about human rights resonance in space and time, and
we test these hypotheses using Google data.
Why human rights, a concept of relatively recent vin-
tage, catches on in some contextsbecoming politically
salient, localized,or vernacularized(Acharya
2004; Merry 2006)remains an open question. We
are aware of only a handful of studies that address
variations in human rights uptake in comparative per-
spective (Gordon and Berkovitch 2007; Ron et al.
2017). As a result, we have few empirical studies of
resonance from which to derive expectations. That said,
many accounts offer historically informed perspectives
on the appearance and evolution of rights claims in
certain contexts. We separate these into top-down and
bottom-up accounts.
The Top-Down Model
The top-down model of human rights resonance ech-
oes elements of the hegemonic interpretation of the
discourse, framing its spread as outside-in or vertical.
In this formulation, human rights ideals pulse from the
Western heart of the liberal world order to that orders
hard-to-reach extremities. This transnational flow of
ideas resembles a global hierarchy. Elites in the cul-
tural and economic core gin up concepts like human
rights, and then they transmit these concepts, via
policies and institutions, to populations in the periph-
eryeven when those populations are unprepared or
unwilling to receive the message. But how? Building
on the work of Goodman and Jinks (2013), we theo-
rize that top-down transmission of human rights
norms could occur through three main mechanisms
of influence: material inducement, persuasion, or
Material inducement entails states or institutions
increasing the benefits of conformity or the costs of
nonconformitywith human rights ideals (Goodman
and Jinks 2013,23).Traditionally,internationalfinan-
cial institutions like the International Monetary Fund
(IMF) and the World Bank have a poor track record
with human rights promotion. Structural adjustment
programs put a squeeze on public goods provision,
leading to protest and repressive violence (Abouharb
and Cingranelli 2007). However, foreign direct
investment by transnational firms is empirically asso-
ciated with greater protection of human rights
(Richards, Gelleny, and Sacko 2001). One reason is
that investment firms demand devotion to rule of law,
contract-enforcement, and political stability, of which
human rights protections are an indicator (Blanton
and Blanton 2007). Organizations like PRS Group
generate revenue by producing political risk rankings
for investors. These reports incorporate information
on rule of law and human rights practices.
this, government leaders in potential recipient coun-
tries might be inclined to adopt the human rights
script, at the very least to maintain a veneer of com-
pliance with international norms. For example, one
study finds an empirical relationship between FDI
inflows and pro-forma domestic criminal proceedings
against human rights abusers in post-conflict settings
(Appel and Loyle 2012). If it is true that government
leaders publicly support human rights norms in order
to attract investors, even if it is a cheap concession,
this might rhetorically entrapleaders and open up
new avenues for human rights mobilization (Risse
and Sikkink 1999,27).Therefore,wemightexpect
the following hypothesis:
FDI inflows hypothesis: Inhabitants of countries with
higher FDI inflows should show greater interest in
human rights.
See, for example, the International Country Risk Guide: https://
The Global Resonance of Human Rights
7 Published online by Cambridge University Press
The second hypothesis involves the role of human
rights non-governmental organizations, or HRNGOs,
that serve as instruments of persuasion by engaging in
promotional campaigning. These bodies, which often
operate transnationally, are usually seen as a primary
agent of norm transmission from the West to the rest of
the world (Mutua 2001). The number of operating
transnational NGOs devoted to human rights increased
dramatically in the late 1990s (Smith, Pagnucco, and
Lopez 1998). The modus operandi for these HRNGOs
is information politics: to produce reports, stoke media
attention, shine a spotlight onto abusive states, provide
consultation, and ultimately, convince target states of
the wrongfulness of their behaviors. Research shows
that HRNGO campaigns are in fact persuasive (Korey
1999). Murdie and co-authors discover that negative
attention from HRNGOs is correlated with poorer
opinions of government and higher numbers of protests
(Davis, Murdie, and Steinmetz 2012; Murdie and Bha-
sin 2011). However, HRNGOs do not target certain
countries on the basis of need alone; they act strategi-
cally, in a way that boosts market share(Cooley and
Ron 2002). For instance, HRNGOs are more likely to
direct their information campaigns at countries that
feature hotissues or high-profile conflicts, that gain
widespread attention in other print media, and that
have a larger number of local partner NGOs
(Meernik et al. 2012; Ron, Ramos, and Rodgers
2005). This may strike some as a positive development,
but others portray HRNGOs as playing an artificial and
unevenly distributed constituency-building role in for-
eign countries. As such, these organizations are some-
times characterized as agents of disseminationin a
kind of norm-promoting industry(Gready 2010, 5). If
it is true that HRNGOs effectively transplant ideas, but
more so in the countries they target directly, then we
might expect the following hypothesis to hold:
NGO hypothesis: Inhabitants of countries that
receive more attention from international HRNGOs
should show greater interest in human rights.
A third mechanism of outside-in transmission is accul-
turation, which can occur in and through international
human rights institutions. Though state participation in
the human rights regime is voluntary, some argue that
international bodies change behavior by serving as sites
for socializing state actors into the prevailing order
(Goodman and Jinks 2013; Greenhill 2010). This
extends insights from sociological models that document
processes of global cultural homogenization,of which
formal human rights organizations are a critical part
(Finnemore 1996, 328). World Society theory holds that
global culture gradually advances over time, leading to
institutional isomorphism and decoupling; for instance,
we should expect states to ratify more and more human
rights treaties, incorporate identically worded rights pro-
visions into constitutions, and establish similarly
designed national human rights institutionseven when
these actions are increasingly detached from local con-
text (Meyer et al. 1997). The reason is that state actors
seek to conform, behaving in line with dominant cultural
scripts that favor governance practices centered around
bureaucracy, markets, and individual rights. However,
whether offering public commitments to international
human rights norms is catalytic, filtering down into a
population-wide human rights resonance, is an open
question, and one often left unaddressed by state-
centric, macro-sociological approaches. Some scholar-
ship does find that human rights legalization exerts a
gradual impact by inspiring domestic constituencies to
mobilize and generate domestic legal reforms (Berlin
and Dancy 2017;Simmons2009); by helping to translate
a universal language of justice to the disempowered
(Benhabib 2009;Merry2006); or at a deeper level, by
producing the concept of the individual as an autono-
mous actor(Finnemore 1996, 332). Presumably, these
trickle-down effects of international acculturation, if
they exist, would be more pronounced in those states
that have ratified more human rights treaties, making
the international legal discourse more available to their
own citizens. Thus, the following expectation:
Human rights regime hypothesis: Inhabitants of coun-
tries with more commitments to international human
rights agreements should show greater interest in
human rights.
The Bottom-Up Model
A great deal of research also questions whether these
top-down mechanisms of human rights promotion
FDI, transnational NGOs, or human rights treaties
have any observable impact (e.g., Clark and Kwon
2018; Hafner-Burton 2008; Posner 2014). But if it is
not top-down dissemination that makes the discourse
resonate, then what does? An alternative bottom-up
theory holds that interest in human rights emerges
primarily when local conditions are ripe. The discourse
diffuses and roots down in various contexts because it
serves a useful function for social movements and
agents of change. Therefore, the adoption of human
rights is not caused by actors or institutions in the global
core, but an outgrowth of local political causes strug-
gling to reach a wider audience. For instance, even
though they emphasize the important role of transna-
tional advocacy organizations from the Global North,
Keck and Sikkink (1998) fit into the bottom-up
approach because they argue that local actors are the
prime movers in the boomerang model of human rights
pressure. According to this model, concern for human
rights starts with the efforts of local civil society, who in
the face of domestic obstacles, seek outside assistance.
This is more likely to happen in situations of dire need,
when physical integrity violations are rampant.
If human rights interest is diffuse, but does not travel
through lines of transmission from the top down, then
what would predict the appropriation of the human
rights idiom in any particular context? Two ancillary
theories may assist in generating cross-national expec-
tations. The most influential of these is modernization
theory, which holds that economic development
Geoff Dancy and Christopher J. Fariss
8 Published online by Cambridge University Press
produces systematic changes in society and politics
(Inglehart and Welzel 2010, 553). This account is prom-
inent in research on endogenous democratization. Neo-
institutionalists posit that the desire for seventeenth-
century English wealth-holders to secure property
rights against the sovereign generated constitutional
rule of law (North and Weingast 1989). The lesson is
that that rights claims thrive in moments when old
eliteseconomic power weakens, and with it their
political control. Extrapolating, one may use this theory
to explain fluctuations in rights-based claim-making.
As new pockets of wealth emerge in a countrys econ-
omyfor example, a growing middle classso too do
rights claims and demands for public goods. This is
rooted in the individual desire to protect property from
government expropriation. Previously disenfranchised
out-groups with new wealth will demand political
concessions in return for tax compliance(Ansell and
Samuels 2010, 1549).
North and Weingasts contractarian version of the
historical emergence of rights is focused on elites, and
may not necessarily explain why a discourse of human
rights might become widespread in a nation. A second
version of modernization theory does consider the
role of ordinary peoplein linking economic growth
to liberal democracy. Inglehart and Welzels(2010)
mass prioritiesaccount holds that with economic
growth in a society comes less everyday worry over
existential security,(553) and more emphasis on
self-expression valuesa syndrome of trust, tolerance,
political activism, support for gender equality, and
emphasis on freedom of expression…” (557). This
package of concernswhich correlate strongly with
economic development and technological advance-
mentechoes human rights ideals. It is not a stretch,
then, to think that economic growth may also inspire
ordinary citizens eager for self-expression to explore
the concept of human rights.
Economic growth hypothesis: Inhabitants of coun-
tries experiencing sustained economic growth should
show greater interest in human rights.
A second endogenous explanation is similar to the
first, with the exception that it is not exclusively mate-
rialist. In short, human rights claims emerge organically
in domestic political interactions, but these need not be
determined by deeper economic structures. This kind
of formulation might be found in literature on social
movements. Essentially, one may conceive of human
rights as one claim-making technology in a larger rep-
ertoire of contention, which comprises what people
know they can do when they want to oppose a public
decision they consider unjust or threatening(della
Porta 2013, 1). Human rights claims take their place
alongside other tools in the arsenal of collective actors
challenging the legitimacy of certain governing actions.
When applied to actual examples, those who sub-
scribe to this human rights-as-resistance model see
state repression and rights expression to be mutually
reinforcing. In the Swedens of the world, where indi-
vidual dignity is generally respected by government
and where any incidents are transparently reported
and addressed, there is not a pressing need to mobilize
around protection of human rights (Eck and Fariss
2018). However, in the Zimbabwes of the world, where
citizens live in fear of their governments, and do not
trust courts to act as neutral arbiters of disputes, people
look to human rights as an external resource for indi-
vidual and legal empowerment, precisely because
human rights are not dependent on a flawed domestic
constitution. Though not a grammar of violence,
human rights do embody a radical proposition: that
local government should be reformed or replaced in
line with internationally defined norms. In circum-
stances where government resorts to unauthorized vio-
lence, human rights become a logical language of
resistance. Theoretically, this aligns with Donnellys
(2003) argument that rights claims themselves are
self-liquidating: where rights are respected, human
rights claims cease; inversely, where rights are not
respected, human rights claims abound.
The resistance model explains a number of findings
within comparative scholarship. Some scholars discuss
how transnational human rights movements emerge in
response to lack of reform (Keck and Sikkink 1998),
while others presume regimes follow a law of coercive
responsiveness(Davenport 2007). Under this law,
governments are reactionary, responding to protests
and other challenges with violence when it strengthens
their hold on power. Because government violence and
human rights mobilization so often coincide, it is hard
to know which is causally prior. Yet regardless of
whether governing elites or resistors are the prime
movers in the repression-resistance cycle, the empirical
expectation should be the same:
Human rights violations hypothesis: Inhabitants of
countries with more human rights violations should
show a higher interest in human rights.
Global Search Rates
How is relative interest in the human rights discourse
distributed across time and space? We answer this
question, and thus test the hypotheses produced by
the top-down and bottom-up models, by studying
trends in aggregate Google searches. Put simply, in
which countries are people most prone to input Google
searches containing human rights,relative to people
in other countries? To investigate, we examine search
data from 109 countries across five language groups:
English, Spanish, Portuguese, French, and Arabic. We
use the data produced to empirically evaluate theories
about cross-national variations in human rights interest
over the 5-year period between 2015 and 2019.
the English language group, we study searches for the
We chose this time period because it is the most recent period in
which data for a number of model covariates are available. We
The Global Resonance of Human Rights
9 Published online by Cambridge University Press
term human rights,and for other language groups,
we analyze searches for the most common translation
of the phrase. For example, in Spanish, the search term
is derechos humanos,and in Arabic, it is huquq
Figure 2 plots weekly search rates for each language
group over a 260-week period.
The y-axis is the min
max value indicating how much searching there is at the
global level across all weeks. High values, and peaks in
the trend, mean that there was more searching in a
given week compared to all other weeks on the x-axis.
Two things to note are, first, that the smoothed global
trend in human rights searching is relatively flat. It is
not appreciably increasing or decreasing over this
period of time, in any language.
Second, human rights
searching appears to be seasonal. While it increases in
the spring and fall, it decreases during the mid-summer
and in December. It is yet unclear why this is the case.
We further analyze temporal trends in aggregate
searches across different countries, in order to locate
which populations are the top searchers in each lan-
guage. Our initial expectation was that the most
searches for human rights would be located in OECD
countries, which are wealthy, educated, and filled with
university students and professionals in the NGO sec-
tor. This expectation is wildly off the mark. Figures 36
present the geographic distribution and the list of top
searchers in English and Spanish, respectively.
Almost universally, human rights interest is most pro-
nounced in countries in the Global South. In English,
the three populations with the most relative searching
for human rightsare in the sub-Saharan countries of
Zambia, Zimbabwe, and Uganda. The United King-
dom and the United States do not make the top
12 searchers presented in these figures. The United
Kingdom is 17th on the list, and the United States is
28th, behind Bangladesh and Qatar. In the Spanish
language group, the top five searchers are Central
American states. Guatemala, El Salvador, Honduras,
and Mexico are ranked 14. Argentina, which has
formerly been referred to as a global human rights
protagonist,is all the way down in the 15th position
(Sikkink 2014).
One way of synthesizing these findings is that popu-
lations in the Global South are steady-state searchers,
plugging the phrase human rightsinto Googles
engine on a more regular basis than populations in
the Global North. This, we argue, represents a differ-
ence between sustained human rights interest and pass-
ing curiosity in headline events. Take Guatemala and
Uganda, which across all 5-year samples are consis-
tently in the top three most interested in human rights.
Though we know neither the raw total of net Google
searches nor the actual number of human rights
searches made by Guatemalans or Ugandans, we do
know the relative rate of these searches. Based on
calculations of rate, we draw comparative inferences.
For example, an average Ugandan searches approxi-
mately 7.3 times for human rights every one time an
average American does.
And individuals in Guate-
mala on average search 10 times more often for dere-
chos humanos than do individuals in Spain. This wide
gap between Global South countries and countries in
the Global North is striking, and it leads back to our
central research question: what determines variation in
human rights interest across countries?
Cross-National Variations
Why are Guatemalans and Ugandans top searchers for
human rights in Spanish and English? The theories we
outlined above offer very different answers. The top-
down model would hold that these countries likely
receive the most pressure from foreign investors,
attract more attention from NGOs, or interact more
with the international legal regime. The bottom-up
model would probably chalk up the prevalence of
human rights discourse to growing GDP, or to upticks
in levels of political repression and violence.
Here, we analyze which expectations bear out across
cases. We generated a dataset including five years of
observations for each language-country-year in our
sample. Each panel of this dataset includes yearly
aggregate Google searches in each language spoken
in a given country over the period 201519. For every
single country-year observation, we record three dif-
ferent Google measures. Search mean represents the
average weekly Google search rate for a country in an
entire year. Search median is the median weekly Goo-
gle search rate for each country in year. And search
max is the maximum yearly search rate. Higher values
on each of these variables indicate that inhabitants of a
particular state more regularly search for human
rightseach week over an entire calendar year com-
pared to populations in other states. The range of
values for search mean is 0 to 68, with a standard
deviation of 13.37; the range of values for search
median is 0 to 68.5, with a standard deviation of
present findings on other 5-year periods in Section L of the Supple-
mentary Material (201216, 201317, and 201418). Almost all of our
statistical findings are robust to different samples.
We use search termto refer to what Google Trends labels
queries.Google Trends also allows one to study topics,which
are based on bundles of related (co-occurring) search terms and are
language-agnostic. Google does not provide full information about
the process by which they create topics, so we have focused most of
our analysis on natural language search terms. But the statistical
results are similar when using the human rights topic to measure our
dependent variable in place of the language-specific search terms.
Searches appear to be decreasing in Spanish, but this is somewhat
sensitive to the time period chosen. When other time periods are
analyzed, the slope is flatter. See Section C of the Supplementary
We are exploring this in greater depth in a related project. The
answer may have to do with seasonal patterns in violence.
See Section E of the Supplementary Material for additional lan-
guage groups. Also see Sections F and G of the Supplementary
Material for a discussion and comparison of variation in search terms
in the French and English languages.
Though Uganda is the third-highest searcher in Figure 4,we
highlight this country because its inhabitants are routinely top
human rightssearchers across all samples of Google Trends data.
Geoff Dancy and Christopher J. Fariss
10 Published online by Cambridge University Press
FIGURE 2. Global Weekly Search Rates from Google Trends for Five Language Groups (201519)
Global: 'human rights' (search term)
2015 2016 2017 2018 2019
Global: 'derechos humanos' (search term)
2015 2016 2017 2018 2019
Global: 'direitos humanos' (search term)
2015 2016 2017 2018 2019
Global: 'droits' (search term)
2015 2016 2017 2018 2019
Global: 'human rights' (topic)
2015 2016 2017 2018 2019
Global: 'huquq al−insan' (search term)
2015 2016 2017 2018 2019
Note: Absolute comparisons of the global rate are valid within each figure but not between figures because of the minmax transformation
described above (95%CI). Relative comparisons of change in the trend over time are possible between panels. Note that the human rights
topic (lower right panel) pools searching across language group. In Section C of the Supplementary Material, we present global trends for
other 5-year periods: 201216, 201317, and 201428.
FIGURE 3. Map of Google Search Rates for Human Rightsin the English Language
100 0 100
English Language: 'human rights'
Note: Darker colors indicate a higher relative rate of searching compared to other countries conducting the same search. The rectangular
projection (i.e., Plate Carrée projection) is defined by equally spaced parallels, equally spaced straight meridians, and is true to scale at
0 latitude.
The Global Resonance of Human Rights
11 Published online by Cambridge University Press
13.68; and the range of values of search max is 0 to
100, with a standard deviation of 25.1. These measures
are meant to capture the latent interest in human rights
in different countries (see Section B of the Supplemen-
tary Material for descriptive statistics).
The reason that we use yearly aggregates for our
dependent variables is that doing so places greater
emphasis on sustained attention, rather than short-term
spikes in interest. The United States would score high for
1 week in the summer of 2018when family separations
on the Mexican border became headline newsbut the
United States regularly scores low on search mean,
search median, and even on search max, because its
average yearly values are much lower than the values
in this single week. This, we argue, is because the
U.S. population has lower latent human rights interest
compared to populations in other states.
To assess the hypotheses presented above, we esti-
mate a regression model, with a fixed effects parameter
for each different language group. We include six
independent variables. The first, FDI inflows as a
percentage of total GDP, we take from the World
Bank. The second, Amnesty report rate, is defined by
the number of Amnesty Internationals (AI) reports
FIGURE 4. Rate of Google Searches for Human Rightsin the English Language across Country-
Weeks, 201519
2015 2017 2019
2015 2017 2019
2015 2017 2019
South Africa
2015 2017 2019
2015 2017 2019
2015 2017 2019
2015 2017 2019
2015 2017 2019
2015 2017 2019
2015 2017 2019
2015 2017 2019
New Zealand
2015 2017 2019
Note: Top 112 countries displayed in descending order.
Geoff Dancy and Christopher J. Fariss
12 Published online by Cambridge University Press
published on a country in any given year, per one
hundred thousand residents of that country.
AI is one of the two foremost HRNGOs in the world
(with Human Rights Watch), this measure should cap-
ture the degree to which a country is subject to global
human rights campaigning. To derive this indicator, we
scraped Amnestys website and built a corpus of over
seventy-five thousand documents, including news
reports, background reports, and urgent actions,
among others. We used AIs own labels to determine
which country is covered in each publication. Amnesty
report rate is more extensive and more updated than
other similar variables used in the literature, which
either rely on specific sub-samples of AI written con-
tent or are available only through the early 2000s (Ron,
Ramos, and Rodgers 2005). Third, we include a mea-
sure of total human rights treaty ratifications
(HR treaties), which we derive by summing the number
of ratifications each country has made to the 27 agree-
ments listed by the UN Treaty Collection as promoting
human rights. These include signature agreements like
the International Covenant on Civil and Political
Rights, but also lesser-known agreements like the
International Convention for the Protection of All
Persons from Enforced Disappearance.
Fourth is a
measure of percentage GDP growth, also drawn from
the World Banks World Development Indicators.
Finally, we include a variable called HR violations,
which is the inverse of Farisss normalized Human
Rights Protection Scores (4) data; for our models,
the higher the score, the more regular government
violence is directed at a countrys inhabitants (Fariss
2014;2019; Fariss, Kenwick, and Reuning 2020).
A final measure that we include is a control variable
called Internet censorship. A valid concern with our
research design is that searching for human rights will
be less prevalent in countries like China, where the
central government makes a concerted effort to limit
internet access, or to block certain information from
users. Three points here. First, China is not a case in our
cross-national statistical models, but not because it fails
to respect internet privacy. It is excluded because we
only include searches in language groups that span a
variety of countries. Doing so provides us with varia-
tion to analyze with statistics. Chinese-language Goo-
gle searches overwhelmingly occur in only three
countries: China, the United States, and Malaysia. That
does not give us much leverage when considering
covariation with other state-level variables. Second,
we can examine the China problem”—that internet
control would affect Google searches for human rights
by examining its effects in other cases. This is why we
include the Internet censorship variable, derived from
the Varieties of Democracy Project.
It captures
how much, in practice, statesgovernments filter the
Figure 7 visualizes results from the three models, and
Table 1 presents coefficients and levels of significance.
The first thing to note is that some expectations derived
from the top-down model are supported, but not in
overwhelming fashion. FDI inflows are statistically
insignificant in all but the third model, suggesting that
countries with higher FDI inflows are no more or less
likely to show greater mean interest in human rights,
though they are associated with greater yearly maxi-
mum searches. Given the small magnitude of these
findings, however, one might say that material
FIGURE 5. Map of Google Search Rates for Derechos Humanosin the Spanish Language
100 0 100
Spanish Language: 'derechos humanos'
Note: Darker colors indicate a higher relative rate of searching compared to other countries conducting the same search. The rectangular
projection (i.e., Plate Carrée projection) is defined by equally spaced parallels, equally spaced straight meridians, and is true to scale at
0 latitude.
We use population-adjusted data in our main models because our
dependent variable is also adjusted by population. In Section L of the
Supplementary Material, we specify our models with alternative
measures, including raw number of Amnesty International Reports
and the number of HRNGOs operating in a given country.
We thank Timo Thoms for sharing these data, which we updated
beyond 2017.
See V-Dem version 9:
All continuous independent variables in the analysis are standard-
ized. A table of descriptive statistics are available in Section B of the
Supplementary Material.
The Global Resonance of Human Rights
13 Published online by Cambridge University Press
inducements do not seem to exert much influence on
human rights uptake in a country.
Amnesty report rate is positive and statistically sig-
nificant in all three models, meaning that the more a
country is targeted for exposure by groups like
Amnesty International, the more that countrys popu-
lation searches for human rights in Google, on average.
While this is grounds for rejecting the null NGO
hypothesis, we still urge some caution in interpreting
this result. For one, the substantive significance of the
Amnesty report rate coefficient is fairly small: a one-
standard-deviation change in Amnesty reporting leads
to just a 0.3-point increase in mean or median search
rates (the rate DVs range from 0 to 100 points). But
also, additional tests show this finding to be sensitive to
variable operationalization and model specification; in
robustness checks, some models produce findings in
the opposite direction, and others yield statistically
insignificant coefficients (see Section L of the Supple-
mentary Material). The relationship between NGOs
and measurable interest in human rights certainly war-
rants further investigation, but at the very least, it does
FIGURE 6. Rate of Google Searches for Derechos Humanosin the Spanish Language across
Country-Weeks, 201519
2015 2017 2019
El Salvador
2015 2017 2019
2015 2017 2019
2015 2017 2019
2015 2017 2019
2015 2017 2019
2015 2017 2019
2015 2017 2019
Dominican Republic
2015 2017 2019
2015 2017 2019
Costa Rica
2015 2017 2019
2015 2017 2019
Note: Top 112 countries displayed in descending order.
Geoff Dancy and Christopher J. Fariss
14 Published online by Cambridge University Press
not appear that the work of Western NGOs is a primary
driver of human rights resonance across the globe.
Finally, disconfirming the expectations of the top-down
model, the effect of HR treaties is statistically indistin-
guishable from zero. It does not seem to be the case that
embeddedness in the human rights regime, on its own,
drives collective interest in human rights.
The bottom-up modelsexpectations appear more on
the mark. The GDP growth and HR violations vari-
ables perform in the model as expected, and their
substantive effects are more robust than the other vari-
ables. In contexts where the national economy is grow-
ing, there are more people regularly searching human
rightson the internet. This is the second most power-
ful predictor in the model, judged by magnitude; a one-
standard-deviation change in growth corresponds with
around a 0.9-point increase in mean human rights
search rates.
By far the most powerful predictor in the model,
though, is HR violations. A one-standard-deviation
increase in this variable leads to a 1.9-unit increase in
Google search rates. By comparison, this effect is 2.2
times greater than GDP growth, and 6.4 times greater
than the effect of Amnesty reporting. This finding is
robust to every single model we have specified, and it is
not conditional on external inducements or persuasion.
For instance, to account for the possibility that human
rights interest spikes precisely in scenarios with a com-
bination of external pressures and internal repression,
we specified a set of models with an interaction
between HR violations and Amnesty report rate (see
Section L of the Supplementary Material), only to find
that the interaction is negative and statistically signifi-
cant in all models.
Another set of models incorporates an interaction
between HR violations and HR treaty ratifications
(see Section L of the Supplementary Material). This
variable is positive and statistically significant in the
three main models. So while the embeddedness of a
country into the human rights treaty regime has no
independent effect on the local appeal of the dis-
course, that is, it does not affect all populations
evenly, it does magnify the relationship between gov-
ernment violence and interest in human rights
throughout the population (e.g., Simmons 2009). This
finding validates world society theories, but only in
part (Meyer et al. 1997). Human rights as a legal and
cultural script have spread nearly everywhere, but
people are more likely to access that script in situa-
tions of acute need (cf. Law 2018; Merry 2006). While
not the final word on the matter, we take these results
as evidence that concern foror interest inhuman
rights is mostly generated within countries, when
government violence is an everyday reality.
FIGURE 7. Coefficient Plot of Results from Regression Models with Language Fixed Effects
Standarized Coefficients
Lower <−−−−−−−−−−− Google Search Rate −−−−−−−−−−−> Higher
Internet Censorship
HR Violations
GDP Growth
HR Treaty Ratifications
Amnesty Report (Rate)
FDI Inflows
−6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6
mean rate
median rate
max rate
Note: The search mean, search median, and search max dependent variables are measures of the yearly mean, median, or max of the
country-week search rate value. Independent variables are measured annually for each country-year unit (201519). Lines represent 90%
and 95%confidence intervals.
The Global Resonance of Human Rights
15 Published online by Cambridge University Press
Measurement Validity
Our findings show that human rights is a discourse
meant to directly challenge excesses of state coercion.
It is not a leap to think that people facing widespread
violence would seek out ways to empower themselves,
and that they would do so in private on the internet. In
short, the reason that we see more searches for human
rights in countries of the Global South is that these
countries are in the greatest need of rights. However,
these findings may be legitimately challenged from the
perspective of measurement theory. How could it be
that these data capture an actual social process? To
check the validity of the theory and the measures used
to test the hypotheses, we introduce additional case
study evidence to study the links of the four-step data-
generating process described in the Google as
Methodsection. We focus specifically on Guatemala,
one of the top human rights-searching countries in the
Step 1 is interest-formation. Are Guatemalans con-
cerned about human rights because of government
repression? Indeed, evidence from a Latin American
Public Opinion Project (LAPOP) survey in 2016/17
(Latin American Public Opinion Project (LAPOP)
2020) suggests that Guatemalans demonstrate a high
level of concern for lack of freedom of expression,
second only to Colombia in the entire region (see
Figure 8b). Moreover, there is strong evidence that
actual repressive events are correlated with higher
human rights search rates. Section M of the Supple-
mentary Material presents an analysis of weekly level
data on government violence against civilians and
Google searches. One can see a statistically significant
lagged relationship: number of government-initiated
violent events in a week is associated with more
searches for human rights in the following week. These
data, while preliminary, establish support for the mech-
anisms driving our causal theory; people respond in real
time to cues, in this case real instances of state violence,
by seeking information about human rights on the
The second step in the data-generating process is use
of the internet. While the same LAPOP survey project
from above finds that Guatemalans use the internet
much less than other countries in Latin American on
average (Figure 8c), World Values Survey wave 7 data
from 2017 to 2020 (Haerpfer et al. 2020) show that
Guatemalans are at the top of the list for use of the
internet as a source of information generally, and infor-
mation about political events specifically (Figure 8e,f).
Note that the World Values Survey was conducted in
TABLE 1. Country-Year Regression Analysis with Language Fixed Effects
Dependent variable: Google search rate
Search mean Search median Search max
FDI inflows 0.497 0.487 1.203**
(0.312) (0.307) (0.604)
Amnesty report rates 0.310** 0.310** 1.162***
(0.130) (0.139) (0.345)
HR treaty ratifications 0.703 0.720 1.347
(0.516) (0.509) (1.001)
GDP growth 0.897** 0.854** 1.521*
(0.396) (0.390) (0.816)
HR violations 1.987** 1.920*** 3.908***
(0.307) (0.308) (0.561)
Internet censorship 0.743 0.761 0.507
(0.496) (0.487) (0.996)
Spanish (derechos humanos) 18.602*** 18.637*** 38.850***
(1.462) (1.524) (2.837)
Portuguese (dereitos humanos) 10.304*** 9.697*** 27.233***
(1.328) (1.224) (3.922)
French (droit) 21.340*** 21,070*** 37.077***
(1.468) (1.465) (2.493)
English (human rights) 7.799*** 7.543*** 16.594***
(0.555) (0.537) (1.198)
Arabic (huquq al-insan) 5.858*** 5.225*** 16.280***
(0.528) (0.498) (1.356)
0.567 0.561 0.570
Sample size 753 737 741
Note: Models include a fixed effects parameter for each language group (Spanish, Portuguese, French, English, and Arabic). Search mean,
search median, and search max dependent variables are measures of the yearly mean, median, or max of the country-week search rate
value for each country-year unit. Independent variables are measured annually for each country-year unit. *p< 0.1, **p< 0.05, ***p< 0.01.
Geoff Dancy and Christopher J. Fariss
16 Published online by Cambridge University Press
Guatemala in 2019, 3 years after the LAPOP survey
was conducted in 2016. These data suggest that while
on average Guatemalans use the internet more sparingly
than citizens in other countries, when they do, they
search for political news and information at a higher
The third step in the process is use of Google.
Evidence suggests that Guatemalans overwhelmingly
use this search engine when scouring the internet.
Particularly over the period 201319, Googlesshare
of searches in Guatemala compared to other browsers
is over 94.5%98.5% across all monthly periods
(Figure 9).
The fourth and final step involves information-
seeking. What are Guatemalans looking for when they
type human rightsinto the Google search bar? The
Google Trends portal suggests that the queries most
related to human rightssearches in Guatemala are
what are human rights, declaration of human rights,
human rights inspector, and the interamerican court of
human rights. Related search topics compiled by Goo-
gle Trends include Law,Universal Declaration of
Human Rights,American Convention on Human
Rights, and Constitution (Figure 10).
All of this evidence suggests that the data-generation
process assumed by our theory is plausible, and that we
are capturing facets of that process with our indepen-
dent and dependent variables. However, one problem
remains. Even if this is true, Google aggregate data may
still lack representativeness. After all, internet access is
not equally distributed within states. Therefore, it could
be that searches for human rights in developing coun-
tries are driven primarily by urban city-dwellers, stu-
dents, and visitors from abroad working in the NGO
FIGURE 8. Human Rights Survey Validation
Latin American Public Opinion Projectabc
(LAPOP 2016/2017):
Internet Use
El Salvador
Dominican Republic
Costa Rica
United States
Less than
monthly Monthly Weekly Daily
Latin American Public Opinion Project
(LAPOP 2016/2017):
‘Very Little' Freedom of Expression
0.0 0.2 0.4 0.6 0.8 1.0
United States
Costa Rica
Dominican Republic
El Salvador
No Yes
Latin American Public Opinion Project
(LAPOP 2016/2017):
‘Very Little' Human Rights
0.0 0.2 0.4 0.6 0.8 1.0
United States
Costa Rica
Dominican Republic
El Salvador
No Yes
World Values Survey (WVS 2019):
How much respect is there for individual
human rights nowadays in this country?
No respect
at all
Not much
Fairly much
A great
deal of respect
World Values Survey (WVS 2019):
Internet as a Source for Information
Less than
monthly Monthly Weekly Daily
World Values Survey (WVS 2019):
Online Searching for Information
about Politics and Political Events
Would never do Might do Have done
Note: We compare country-level averages of subject response rates by question across two surveys for their most recent waves for LAPOP
in the top row (administered in 201617) and the World Values Survey wave 7 in the bottom row (administered in 201720). We highlight
and compare Guatemala (green) and Argentina (pink). Guatemalans are at the top of the list for use of the internet as a source of information
generally, and information about political events.
See Section J of the Supplementary Material for regional compar-
isons of Google search volumes relative to other search engines
search volumes.
Translated from Spanish. See Section D of the Supplementary
Material for additional country examples for co-occurring search
The Global Resonance of Human Rights
17 Published online by Cambridge University Press
FIGURE 9. Google Search Proportions
2013 01
2013 03
2013 05
2013 07
2013 09
2013 11
2014 01
2014 03
2014 05
2014 07
2014 09
2014 11
2015 01
2015 03
2015 05
2015 07
2015 09
2015 11
2016 01
2016 03
2016 05
2016 07
2016 09
2016 11
2017 01
2017 03
2017 05
2017 07
2017 09
2017 11
2018 01
2018 03
2018 05
2018 07
2018 09
2018 11
2019 01
2019 03
2019 05
2019 07
2019 09
2019 11
Guatemala: Monthly Search Engine Use (2013–2019)
Note: Google search (dark blue) dominates all other search engines (all other colors) for all regions of the world. In Guatemala, Google
accounts for approximately 94.5%98.5%of search requests. Data are taken from the statcounter Globalstats website: https:// (last accessed: February 12, 2021).
FIGURE 10. Related Co-occurring Search Queries and Search Topics for Guatemala (Top Row) and
Relative Search Rates by City and Region (Bottom Row)
Attorney −PDH− Guatemala Human Rights Headquarters
Arzobispado de Santiago de Guatemala
United Nations
Inter−American Court of Human Rights
American Convention on Human Rights
Universal Declaration of Human Rights
Human rights
Rate of Related
instituciones de derechos humanos en guatemala
constitucion politica de guatemala
convencion americana de derechos humanos
ejemplos de derechos humanos
derechos humanos definicion
10 derechos humanos
procuraduría de los derechos humanos
historia de los derechos humanos
procurador de los derechos humanos guatemala
corte interamericana de derechos humanos
procuraduria de los derechos humanos
cuales son los derechos humanos
procurador de los derechos humanos
declaracion universal de los derechos humanos
declaracion de los derechos humanos
que son los derechos humanos
que son derechos humanos
los derechos humanos en guatemala
derechos humanos en guatemala
derechos humanos de guatemala
derechos humanos guatemala
los derechos
los derechos humanos
Rate of Related
Searches (Queries)
Guatemala City
Antigua Guatemala
Villa Nueva
San Jorge Muxbal
Puerto Barrios
Rate of Searches
by City
Chiquimula Department
Chimaltenango Department
Santa Rosa Department
Jutiapa Department
Petén Department
Sololá Department
Totonicapán Department
Guatemala Department
El Progreso Department
Baja Verapaz Department
Izabal Department
San Marcos Department
Retalhuleu Department
Quiché Department
Jalapa Department
Quetzaltenango Department
Zacapa Department
Alta Verapaz
Huehuetenango Department
Rate of Searches
by Region
Note: Rates are relative to 100 for human rights as a topic (left panel) and the search term derechos humanos.A google queryis
equivalent to a search termas we have used the term and is language specific. Topics are based on bundles of related search terms and
are language agnostic. Google does not provide full information about the process by which they create topics, so we have focused most of
our analysis on natural language search terms (queries).
Geoff Dancy and Christopher J. Fariss
18 Published online by Cambridge University Press
sector. This certainly is a possibility. However, concen-
tration of interest in major cities does not accurately
describe Guatemala, where searches for human rights
are distributed widely across the country. The cities
that top the list of searchers are Huehuetenango and
Cobán. These metros are one-tenth the size of Guate-
mala City, and each was a site of extreme violence
during the civil war in the 1980s and remain a hotbed
for indigenous protest and repression. This lends fur-
ther credence to the theory that collective interest in
human rightsexpressed by population-wide searches
in Googleresponds to government abuse.
It is now in vogue to say that we are heading into a
post-human rights world(Strangio 2017). But this
revisionist claim is seldom grounded in good evidence.
In this study, we have used Google aggregate search
data to answer a little understood question: do human
rights still resonate, and if so, where? We find that, by
and large, the most human rights-interested popula-
tions, defined by the willingness to search for the phrase
human rights in Google, are located in the Global
South. People in countries like Uganda, Zambia,
Mozambique, Guatemala, and El Salvador seek infor-
mation about human rights at far higher rates than their
counterparts in the United States, the United Kingdom,
and even Argentina.
Why? The answer is not that rights ideals are being
pushed onto them by neo-imperial Western actors, nor
that rights uptake is a direct product of the interna-
tional human rights regime. Instead, the answer is more
straightforward: people show more interest in human
rights when they are subject to coercive regimes, and,
just like the malaria search validation demonstrates,
people look to the internet for details about how to
protect themselves. In the end, the language of human
rights continues to appeal where people need rights the
most. This fact has so far eluded direct observation,
even though it appears to be a major point of dispute.
Interdisciplinary scholarship on human rights is con-
sumed with high-level debates over the nature of the
discourse. Now ascendant is the posture of human
rights endism(Teitel 2020). Here, it is worth reiter-
ating how serious people around the worldare not
particularly impressed by the human rights criticism
that has been recently formulated in academic debates
(Mahlman 2020, 69). On the ground, the oppressed still
seem to want human rights.
Though our findings dispute currents of thought that
frame human rights resonance as a top-down phenom-
enon, they cannot be used to reach definitive conclusions
about the hegemonic or counter-hegemonic nature of
the discourse. For instance, the discovery that economic
development is associated with country-wide interest in
human rights could easily be construed as a function
either of neoliberal hegemony or of anti-neoliberal
counter-hegemony: one may surmise that domestic
elites who benefit from liberalization increasingly mobi-
lize rights to protect their assets, or alternatively, one
could posit that the economically dispossessed use
human rights to resist their marginalization by forces
of globalization. Of course, it is also possible that human
rights are simultaneously beneficial for both hegemonic
and counterhegemonic projects(Perugini and Gordon
2015, 17). Our data are simply too limited to weigh in on
this matter. However, what we can say is that the
concrete linkage between government violence and
internet searches for human rights, which is evident at
the yearly and weekly levels, strongly hints at a counter-
hegemonic role for human rights.
Still, more research is needed. For instance, how is
internet searching on government violence associ-
ated with social mobilization, collective action, and
protest? And is it possible that interest in human
rights is skewed by domestic class politics? In other