Presidential attention focusing in the global arena: the impact of international travel on foreign publics.
Cohen, Jeffrey E.
Presidents allocate a large proportion of their time and energy to
foreign affairs. In pursuing their foreign policy goals, presidents (and
their secretaries of state) have increased their travel to other nations
since the end of the Second World War. Figure 1 plots the annual number
of foreign trips by U.S. presidents from 1946 to 2011. From a handful of
country visits during the mid-twentieth century, by the 2000s presidents
routinely visit one and a half dozen nations per year, a sixfold
increase from the 1950s. Secretaries of state are globally even more
peripatetic than presidents, with an average of fifty-one nations
visited per year in the 2000s, compared to a dozen in the 1950s. (1)
On these visits, presidents meet with foreign leaders, performing
classic diplomatic activities such as attending formal negotiation
meetings that are held in secret. But now presidents also routinely
appear in public when visiting other nations. For instance, presidents
now commonly hold joint press conferences and public announcements with
the leader of the foreign nation, give interviews with foreign
journalists, visit locations of local symbolic importance, and directly
address the local citizenry. As one prominent example, on August 3,
2009, President Obama held a U.S. campaign-style town hall meeting in
Strasbourg, France, with French citizens. Goldsmith and Horiuchi (2009)
have dubbed these public activities "high-level public
diplomacy."
[FIGURE 1 OMITTED]
High-level public diplomacy activities resemble "going
public" techniques that presidents use in domestic politics and
policy making (Kernell 1993). Domestic going public is predicated upon a
two-step process. First, through public activities, like making
speeches, presidents try to rally or mobilize public opinion around an
issue. Then presidents use that activated public opinion to pressure
members of Congress to enact their policy initiatives (Canes-Wrone
2006). A large literature has investigated the effects of presidential
going public on domestic public opinion and Congress, finding mixed
effects on both. In the domestic arena, presidential going public is far
from consistently effective. (2)
In several respects, high-level public diplomacy efforts parallel
those for domestic going public. First, like domestic going public,
public diplomacy activities target foreign public opinion, with the aim
of improving the image of the United States, opinion concerning an
international issue, and/or the president. The administration's
hope is that an improved foreign public opinion climate can be used as a
resource in negotiating with the leaders of the other nation, to
increase the likelihood that the president will realize his foreign
policy goals vis-a-vis the visited nation. An underlying assumption of
high-level public diplomacy appears to be that foreign leaders are
responsive to public opinion pressures within their own country, like
members of Congress and the president are thought to be responsive to
U.S. voters.
Currently, there is only a limited literature on high-level public
diplomacy. Several early studies have looked at what was then termed
presidential "going international" (Rose 1988; Smith 1997).
These studies focused on the logic of why presidents increasingly go
international in the post--Cold War era (Rose 1988) and tracked trends
in presidential going international activities like foreign trips (Smith
1997). A more recent literature has turned its attention to the
effectiveness of high-level public diplomacy on foreign public opinion
and support for U.S. policies. Several studies find that such public
diplomatic efforts can affect foreign public opinion, at least under
some conditions (Dragojlovic 2011, 2013; Goldsmith and Horiuchi 2009).
Furthermore, the climate of public opinion in the foreign nation is
associated with greater support for U.S. positions on foreign policy
issues, for instance, on roll-call voting in the United Nations (Datta,
2009; Goldsmith and Horiuchi 2012).
One of the major barriers to assessing the effectiveness of
high-level public diplomacy, especially on public opinion, is the dearth
of cross-national data on public opinion about the United States, its
policies, and the president. (3) Moreover, the extant literature has not
investigated all of the causal steps from a presidential visit to
foreign public opinion. This article asks whether a presidential visit
to a nation can focus public attention on the president. (4) The
agenda-setting (McCombs and Shaw 1972) and political communication
(Edwards 2006; Zaller 1992) literatures argue that prior to the
president being able to influence public opinion targeted citizens must
receive the president's communication. Can the president focus
foreign public opinion on his communication to them? To test the
presidential international attention focusing hypothesis, I construct a
pooled cross-section time series of weekly Google searches for Barack
Obama across forty-two nations during the president's first term.
Analysis finds that, during the week of the president's visit, such
searches increased by about 25%, indicating relatively potent attention
focusing effects from a presidential visit.
This article proceeds as follows. First, I discuss agenda-setting
and political communication theories as they pertain to presidential
attention-focusing activities. Then I review the literature on
high-level public diplomacy by American presidents and discuss why
presidential trips to another nation should focus local public attention
on the president. The following section presents the Google Trends data,
which is followed by the analysis. The conclusion puts the findings into
perspective and raises suggestions for future research.
Agenda Setting, Presidential Communications, and Attention Focusing
Early research on agenda setting has looked at the impact of news
coverage on the public's issue priorities. That research found that
when news organizations increase their reporting on specific issues or
topics, those issues and topics would rise in the public's priority
rankings (McCombs 2013; McCombs and Shaw 1972, 1993). Later research
asked whether other communications, such as major presidential speeches
like the State of the Union Address, can also affect the public's
issue priorities (Cohen 1995; Eshbaugh-Soha and Peake 2011; Hill 1998).
That research found that presidential public communications can
influence public issue priorities but not issue preferences (Edwards
2006).
Priming appears to be one mechanism that accounts for these
communication effects on the public's agenda. For a presidential
message to affect public opinion, the voters must receive the message
(Edwards 2006; Zaller 1992). Voters may learn about the president's
speech directly, for instance, by watching it. They may also learn about
the president's speech indirectly, from conversations with family,
friends, coworkers, and so on. (Cohen 2010; Edwards 2006; Zaller 1992).
Even with these two transmission mechanisms from the presidential speech
to the vote, as Edwards details, in modern American politics, there are
many barriers to voters receiving presidential communications (Edwards
2006; see also Baum and Kernell 1999; Kernell and Rice 2011). For
instance, Young and Perkins (2005) show that as the audience for major
presidential addresses has decreased in size, the effect of the
president's public rhetoric on the public's issue agenda has
also weakened. Contemporary presidents seem less able to affect the
public agenda of American voters than was the case a generation ago.
In contrast to public communication from the president to domestic
audiences, which are frequent and routine, a presidential trip to a
foreign nation is rare and special. Presidents do not travel to each
nation every year, and most nations will receive at most one visit
during the entire time a president is in office. Because such trips by
the president to the overwhelming number of nations are rare, the news
media in the visited country will report on the trip with high volumes
of news coverage. Politicians and other elites in the visited nation may
also want to take credit for the president visiting their nation and
perhaps have some of the prestige of an American president passed onto
them. To do so, local political leaders, especially heads of state, may
orchestrate events that show them in the president's company. These
attributes of a presidential trip to a foreign nation increase the
likelihood that large numbers of citizens in the visited nation will
become aware of, and perhaps even interested in, the president's
visit. This broadscale awareness lays the foundation for potential
presidential influence over foreign public opinion.
High-Level Public Diplomacy and Attention Focusing of Foreign
Public Opinion
There is empirical evidence that, under certain conditions,
high-level public diplomacy by the president, and the secretary of
state, can influence foreign public opinion. Goldsmith and Horiuchi
(2009), for example, find that presidential, and secretary of state,
visits are associated with a more positive image of the United States,
as long as the public in the host country views the United States as
credible and trustworthy. (5) Once that credibility begins to flag,
visits may no longer influence foreign public opinion. And if the United
States or the president is viewed as "noncredible," high-level
diplomatic visits may actually heighten negative attitudes toward the
United States (Goldsmith and Horiuchi 2009). Dragojlovic (2011) employs
an experiment on Canadian college students, adding support to the
Goldsmith and Horiuchi (2009) perspective. When presented with an Obama
prime shortly after his inauguration, when Obama enjoyed high
credibility, treated subjects in the experiment displayed more positive
attitudes toward the United States than nontreated subjects. However, if
exposed to a George W. Bush prime, when Bush's credibility was
weak, attitudes toward the United States became more negative.
There is also some empirical support for the second claim of the
high-level public diplomacy literature, that foreign public opinion will
affect the nation's foreign policy. Datta (2009) and Goldsmith and
Horiuchi (2012) find that support for the U.S. position on issues of
interest to the United States, like support for the Iraq War and voting
patterns in the United Nations, is positively associated with public
opinion about the United States.
But as noted above, prior to a high-level public diplomatic
activity affecting public opinion, citizens in the visited country must
receive the president's communication. Can a presidential trip to
another nation effectively compete with all of the other stimuli that
vie for the attention of citizens in that host nation? Where the
president's ability to focus public attention appears to be waning
within the United States, we expect that trips to other nations are able
to focus the attention of foreign publics on the president and his
policy objectives in visiting the nation.
First, although presidents travel to foreign nations frequently,
they rarely visit any nation more than once a year. Still, presidents do
visit a handful of nations multiple times per year. Generally, these
nations are powerful in international politics and are often important
U.S. partners on critical issues such as the United Kingdom, Germany,
and/or France. The international power and importance of these nations
leads presidents to visit them with some frequency. Similarly, the
leaders of these nations are likely to visit the United States several
times per year for much the same reasons that the president visits them,
indicating a high degree of high-level interactions between the United
States and these nations. Presidents may also visit less powerful
nations more than once a year, primarily when these nations are involved
in timely and important issues that affect the national security and/or
other interests of the United States. South Korea and Indonesia are two
such examples. South Korea is strategically important to the United
States, while Indonesia is becoming an economic power in a region that
has been gaining increased attention from the United States.
Table 1 presents a breakdown of the number of presidential visits
to forty-two nations during President Obama's first term of office
(2009-2012), nations for which we have weekly Google search data. France
was the most frequently visited nation, with five visits during those
four years, while the president visited the United Kingdom and Japan
three times each and Germany twice. President Obama visited South Korea
three times, neighboring Mexico three times, and Indonesia three times,
and several other nations received two visits from the president during
his first term.
With only from one to two dozen presidential international trips
per year, most nations do not receive even a single presidential visit
in any given year. A number of reasons may lead a president to visit one
of these nations, at least occasionally. For instance, a specific issue
may have arisen that requires attention, a new regime has come to power,
the president wants to establish a better working relationship with the
leader of the host nation, and/or the president has not visited the
nation for an extended period of time. There are many other, and perhaps
idiosyncratic, reasons that motivate a president to visit any particular
nation from this set of infrequently visited nations. Of the forty-two
nations on Table 1, twenty did not receive a presidential visit during
Obama's first term in office, and another ten were visited only
once. Presidential visits to foreign nations, although growing in the
aggregate over time, are still rare events for most countries most of
the time.
The rarity of a presidential visit may lead politicians and
citizens of the visited nation to view the president's visit as
important and noteworthy. If citizens view a president's visit to
their nation as special and important, it is likely that the president
will be able to focus public attention on himself and the international
issues of interest to the president. In this sense, a presidential visit
to a foreign nation may be comparable to a prime-time presidential
televisions address to U.S. voters. Presidents rarely broadcast to the
United States on television during prime time, doing so only when the
issue is both timely and important (Foote 1990; Kernell 1993). A
presidential prime-time address to the nation signals to voters that the
president thinks something is important and wants the public to pay
attention to the president and the issue. Although presidential trips to
another nation do not necessarily involve critical issues, as is often
the case for a domestic prime-time address, the infrequency of
international visits may make them important in the eyes of foreign
citizens.
Hence, we can think of a presidential trip to a foreign nation, and
the public activities that he performs while on the trip, as a
presidential attention-focusing activity. The question here is: does the
president's trip focus the attention of the foreign public on the
president? To test this idea, we need data on foreign public opinion
and/or behavior. The next section discusses the data used in this
article, derived from Google search behavior.
Measuring Public Attention with Google Trends Data
One of the aims of a presidential public activity is to alter
public opinion. In the international context, presidents may want to
affect foreign public opinion about the United States, the president,
and/or a particular U.S. foreign policy issue. By producing a more
favorable foreign opinion climate, the president's negotiating
position with the leader of that nation may be improved, enabling the
president to obtain a policy response from the foreign leader closer to
the president's foreign policy goal than would result with a less
favorable opinion climate.
The attention-focusing hypothesis tests whether the target (foreign
public opinion) receives the message (the president's public
activity) from the messenger (the president). At the individual level,
for the president's message to affect opinion, the target must
receive or become aware of the message. The targeted individual(s) may
receive the message either directly or indirectly. An individual may
receive the president's message directly, for instance, by watching
a television broadcast of a presidential speech and/or by reading about
the speech in a newspaper. Individuals also may receive the message
indirectly, for instance, by being informed about the president's
speech from a family member, coworker, or friend. Through this indirect
route, a president's message potentially can affect opinion of a
large number of people even if not many people learned of the speech
directly. Message reception, however, does not guarantee that the
president's speech will influence opinion; message reception is
merely a necessary precondition to opinion change. If an individual does
not receive a message, either directly or indirectly, the message cannot
affect his or her opinion.
Generally, public opinion polls are used to measure message
reception and opinion change, but there are several complications in
using public opinion polls to test for presidential leadership effects
on opinion when visiting other nations. First, there is often a temporal
mismatch between visits and polls--for some nations there may be visits
but no polls and for others, polls but no visits. Rarely do we have
polls being administered in a foreign nation that coincide with the
president's visit. Second, as Goldsmith and Horiuchi (2009) point
out, polling attitudes about the United States, and polling in general,
is not randomly distributed around the world. It is more common in more
well-developed nations such as in Western Europe. Thus, there is a
selection bias to the nations for which we have polling data. This type
of selection bias may affect statistical estimates.
Third, when there are relevant polls in the visited nation, there
may be a large time gap between the visit and the poll. The larger the
time between the visit and the poll, the more likely that a factor
besides the presidential visit may be affecting public opinion such as
other local or international events. Moreover, the longer the time gap
between the visit and the poll, the more likely that the effects of a
visit on opinion may have dissipated. In their study, Goldsmith and
Horiuchi (2009) find the effects of a high-level visit on public opinion
appear to persist for up to five or six weeks. This may appear to be
quite a long time for visit effects to persist, but the effects of a
presidential visit may persist if the news media and local leaders
publicly discuss the issue raised by the visit in the following days and
weeks and if the presidential visit set the public and political agenda.
Finally, that presidential visits are rare makes it difficult to
locate matching polls. Trips by the secretary of state around the globe
are more common. For instance, in the 2000s, the secretary of state
consistently traveled to fifty or more nations per year. In Goldsmith
and Horiuchi (2009), of 353 country polls that they gathered, there are
only thirty-eight high-level visits within the five-week lag that they
use in estimating their statistical model. And of these thirty-eight
visits from 2001 (after 9/11) through 2006, there are only fourteen for
the president. (6) To conduct their analysis, Goldsmith and Horiuchi
combine presidential and secretary of state visits into one variable. By
combining presidential and secretary visits, which is reasonable given
the relative rarity of presidential visits, we cannot test whether a
presidential visit has stronger effects on foreign public opinion than a
visit by the secretary of state. Despite these limitations, matching
visits and poll data is important and useful, but there is still much
that we do not understand about the effects of a high-level visit on
foreign opinion. In addition to the limitations noted above, the polls
used in research on foreign attitudes toward the United States, such as
in Goldsmith and Horiuchi (2009), lack measures on respondent awareness
and/ or exposure to high-level U.S. visits.
For these reasons, to test the presidential foreign public's
attention focusing hypothesis, I use Google Trends data. Research has
increasingly employed Google Trends data, especially for making
forecasts in fields as diverse as disease outbreaks, economic behavior,
and politics. (7) Google Trends data have not always proved accurate for
forecasting. But rather than forecasting, this article uses Google
Trends data as a behavioral indicator for hypothesis testing: do we find
an increase in Google searches for the president in response to a
foreign visit?
Google Trends is a database of searches on Google, the most
frequently used Internet search engine. One can retrieve the Google
Trends data for specific geographic regions, such as a country, as well
as time periods. After experimenting with several search terms, I
selected "Barack Obama" for the period from February 2009
through October 2012. I limited the search to this time period because
search volume for "Barack Obama" spiked during November 2008
to January 2009, and again in November 2012, a function of the
presidential election contest and inauguration in January 2009. Google
Trends data are presented as the percentage of searches compared to the
peak search. The peak search is given the value of 100 in the Google
Trends report. For privacy, competitive, and other reasons, Google does
not release the actual number of searches. The volume of searches during
the election/inauguration periods was extremely high, even outside of
the United States, compared to other periods, making it difficult to
discern variance during these other periods because of Google's
reporting practices--Google only reports whole numbers, such a
"0," "1," "10." By restricting the search
period as described, we are better able to see the peaks and valleys in
searches for "Barack Obama."
To see if a presidential trip affected the relative number of
searches, I searched "Barack Obama" for each nation
separately, only using data that were disaggregated by week. Google
Trends provides weekly data when the volume of searches is large enough
to allow such a refined disaggregation; otherwise it presents the data
in monthly units. (8) The weekly data are temporally refined enough for
testing the attention-focusing hypothesis. This short, weekly time unit
will increase confidence that the visit, and not some other event,
stimulated the search volume. Wider time units, like months, may be too
broad for us to isolate presidential trip effects. Other
events/processes may occur within a month that might affect foreign
public attention to the U.S. president. Finally, the way Google makes
the data public requires using the same search terms and time spans to
retrieve comparable data. Excluding the United States, I was able to
collect this weekly data for forty-two nations for 196 weeks, for an n
of 8,232 in country-week units. The nations, along with descriptive
information on country-specific search patterns, are listed on Table 1.
Unlike the opinion data gleaned from polls, the Google Trends data
provide us with a behavioral measure, search behavior. Still, there are
limits to these data. The Google Trends data are not based on
representative samples, nor are the search data completely comparable
across countries. First, to search on Google requires Internet access,
which is not randomly or equitably distributed within nations, and
Internet penetration varies across nations, with higher penetration
levels in more advanced nations. But Internet searching is not an
especially difficult or costly activity and has become increasingly
commonplace for large segments of the population. Second, of those
individuals with Internet access, only a nonrandom subset will use a
search engine, and an even smaller, and perhaps more unrepresentative,
group will search for "Barack Obama." Notably, the search
traffic volume in these forty-two nations was high enough for Google to
quantify in weekly units. Third, without demographic and other data on
who searches and for what reasons, we have to be careful in making
comparison statements across nations. Lastly, the Google Trends data
only tell us something about interest in "Barack Obama" by
searchers, not by citizens at large. These data do not convey
information on attitudes toward Obama such as whether they like or
dislike him, agree or disagree with his policies, and so on. Still,
given these caveats, we can test the attention focusing hypothesis
across a large number of countries over a relatively sustained period of
time with comparable data, which enables us to address the following
question: are there more searches for Barack Obama during the week when
he visited a country than for weeks without visits?
Estimation Issues
With data across nations and time, the analysis employs pooled
cross-sectional time series techniques. With a relatively large number
of panels (forty-two nations) and 196 time units (weeks), both temporal
and cross-sectional disturbances may affect results (Beck and Katz 1999,
2011; Eberhardt 2012; Franzese and Hays 2007; Wilson and Butler 2007).
It is highly likely that these data will exhibit cross-sectional
correlations, that is, search interest in "Barack Obama" in
one nation will affect similar searches in other nations. First, there
is now a global news media, such that major events transpiring in one
locale are reported around the globe. Many presidential trips are of
interest not only in the nation visited, but in other nations, and
presidential international travel tends to be reported by the global
media (Farnsworth, Lichter, and Schatz 2013). Thus, a presidential trip
in one country may spark interest among citizens in another country,
leading to increased search volume beyond the visited country. This
spillover means that cross-sectional observations in these data are not
necessarily independent of each other, a requirement for statistical
analysis.
I employ several diagnostics to test for these issues in these
Google search data. First, the Pesaran CD statistic tests for
cross-section dependence in panel time-series data (Pesaran 2004).
Essentially, the Pesaran test estimates the cross-sectional correlations
of each panel with all the others and then presents a summary statistic
(CD), which can be thought of as akin to an average correlation. The CD
test statistic of 153.44 (p < 0.000), with an (average) correlation
of 0.37 strongly suggests cross-sectional correlation in these data. I
also use the Breitung test for unit roots in panel data (Breitung and
Pesaran 2008). The Breitung robust lamba statistic of -2.89 (p = 0.002)
does not indicate unit roots in the series. Further, the robust lambda
statistic, with four lags, a trend term, and the series demean, of -4.14
(p = 0.000), reinforces that interpretation. (9)
Analysis
To test the presidential attention-focusing hypothesis, I regress
the weekly searches for "Barack Obama" on a presidential visit
variable, scored one if the president visited the nation during that
week and zero otherwise. The analysis also controls for several other
variables. The first is a dummy variable for whether Secretary of State
Hillary Clinton visited the nation in a solo trip, that is, without the
president (1 = secretary of state visit, 0 = no visit). It is possible
that a visit from the secretary of state will affect search interest in
the president. Also, I added this variable to the estimated model
because Goldsmith and Horiuchi (2009) combined presidential and
secretary of state trips into one variable in their measure of
high-level public diplomacy. The analysis also controls for the
announcement of the killing of Osama Bin Laden, an event that received
worldwide attention as well as credit for President Obama. This variable
is scaled one for the week of that announcement, zero otherwise. (10) A
simple comparison indicates an increase in searches for Barack Obama
during this announcement week (44.3) compared to other weeks (18.4).
Finally, the estimation controls for general searches for the
"United States." This variable is intended to pick up other
factors that may stimulate someone to search for Barack Obama, perhaps
as a byproduct or spillover of searching for the United States.
Table 2 presents an initial estimation, which clusters on country,
the clustering meant to control for country-specific effects. All
variables are significant predictors of search levels for Barack Obama
except for solo secretary of state visits. The coefficients indicate
that during the week of the announcement of Bin Laden's killing,
searches for "Barack Obama" rose by 18.4%, in line with the
simple comparisons presented above. Second, general searches for the
"United States" also led to increased searches for the
president. The coefficient suggests that each 1% increase in searches
for the "United States" leads to 0.44% more searches for
"Barack Obama." Searches for the "United States" in
these data range from 12 to 100, with a mean of 26.4 and a standard
deviation of 15.5. A one-standard-deviation shift in general
"United States" searches will produce a shift of about 6.8% in
searches for "Barack Obama."
Most important for this research is the positive effect of a
presidential trip on searches for "Barack Obama." The
coefficient indicates that during the week of a presidential trip,
searches increased by a massive 43%. We should be suspicious of this
massive impact, however. As discussed above, factors such as temporal
dependence in the dependent (and other variables), as well as
cross-panel correlation, may affect estimation. Does the finding that a
presidential trip stimulates increased searches for the president hold
when using other estimation techniques that can account for these
statistical issues?
As a first test, Table 2 also presents results of an estimation
using panel corrected standard errors. Panel corrected standard errors
assume that the disturbances in the data are heteroskedastic and
correlated across panels. The specific correction that I used assumes
AR1 processes for each panel, but that the coefficient for each AR1
process is specific to each panel. Inspecting the results for this
estimation on Table 2, the first thing to notice is that the rho
statistic falls short of statistical significance, indicating that
correcting for autoregression is not necessary.
Yet, the results closely resemble those for the cluster effects
estimations (Model 1) on Table 2. Again, all variables are significant,
except for the secretary of state trip dummy, as before. Each of the
significant independent variables exhibits slighted smaller regression
coefficients but much larger significance and z-score values. The
inflation in significance tests to such high levels often results from
applying the panel corrections and thus may not be the proper
specification for these data. I have presented the panel corrected
standard error estimation because that approach is so common in
political science research (Beck and Katz 1995, 2011; Wilson and Butler
2007). Other estimations techniques may be more appropriate for the data
used here, especially for dealing with the cross-sectional correlation,
which appears to be a consequential concern based on the above
diagnostics.
Table 3 presents results of two estimations that better account for
cross-panel, or spatial, correlation, the Pesaran Common Correlated
Effects Mean Group (CCEMG) estimator, and the Eberhardt and Teal
Augmented Mean Group (AMG) estimator (Pesaran 2006). The CCEMG
estimation deals with cross-sectional correlation by including
cross-sectional averages of the dependent and independent variables as
regressors in the estimation. Although the regression coefficients for
these averaged variables are reported, they do not have a meaningful
interpretation. Where the CCEMG considers the averaged variables as only
having relevance for estimating the equation, the AMG estimator
represents a cross-dynamic process that is shared across the panels, in
this case, country. (11) For purposes here, we are not concerned with
these estimators other than their ability to correct for cross-panel
correlation, and that cross-panel correlation may be affecting the
effects of the substantive variables in the analysis.
Results on Table 3 generally parallel those in Table 2 in
substantive terms, except that the regression coefficients and z-scores
are smaller, albeit still significant. Again, results indicate that
visits by the secretary of sate do not affect searches for "Barack
Obama." Moreover, the announcement of the Bin Laden killing is no
longer significant. As interest in President Obama peaked worldwide due
to the assassination, it makes sense that the CCEMG and AMG estimators,
which are meant to pick up properties common across panels, would do so
for the Bin Laden variable, because searches for the president spiked
worldwide as a result of the announcement. The two estimations, however,
differ in the effects of general "United States" searches,
with the CCEMG relegating that control variable to statistical
insignificance, while the AMG estimation finds it to be significant,
healthily so with a z-score of over 12. This variable however is of
limited theoretical interest here and was used mainly as a control.
Both estimations however find that a presidential trip boosts
searches for the president at statistically significant levels,
repeating the above results. The z-score is the same for both
estimations (4.75), although it is considerably lower than found above,
but still clearly significant. Plus, both estimations report similarly
sized regression coefficients, about 24.6 to 24.7. These effects are
about 12% less than found in the panel corrected standard error
estimation on Table 2 (Model 2) but still substantively impressive.
Based on these results, when a president visits a foreign nation, the
level of Google searches for the president increases by approximately
25% compared to a week when the president did not visit the nation.
Finally, I estimate an Error Correction Model (ECM) for panel data
that employs the mean group averaging to deal with cross-panel
correlation, an estimation that allows for correction for temporal as
well as cross-sectional issues. In this estimation, the variables are
expressed as changes, with an ECM mechanism that picks up long-term
trends. Table 4 presents these results. The ECM mechanism is positive
and significant (b = 0.39, z-score = 11.34). Again, foreign visits by
the secretary of state have no impact on searches for the president, but
the Bin Laden dummy and general "United States" searches do.
The results report that searches for "Barack Obama" increased
by nearly 16% during the week announcing the assassination of Bin Laden
and that an increase of 1% in searches for the "United States"
corresponds with a 0.33% increase in searches for the president. Most
importantly, presidential trips affect searches for the president. The
coefficient indicates that changing from a week with no presidential
trip to a week with a trip results in a 23.5% increase in searches for
the president, an effect in line with that found for the cross-panel
correlation analyses presented on Table 3. No matter the specification
employed, all estimations indicate that when the president visits
another nation, the percentage of Google searches rises. And depending
on the specification employed, the effect ranges from 23 to 40%. These
are substantively meaningful increases, as well as being statistically
significant.
Conclusion
A growing literature has found that high-level public diplomacy by
presidents, under some conditions, appears to affect foreign public
opinion toward the United States and foreign policies toward the Unites
States. The causal mechanisms from these public diplomatic efforts to
public opinion have not been specified fully, although some linkages
have been identified. First, it appears that the United States must be
viewed as credible for a high-level diplomatic trip to enhance the image
of the United States among foreign publics (Goldsmith and Horiuchi
2009). This article investigates another linkage in the casual process
from high-level diplomatic trips to foreign public opinion: message
reception. As agenda-setting and communication theories argue, before a
communication can affect behavior or opinion, the target must receive
the message. Message reception is commonly tracked with public opinion
surveys that ask respondents about awareness of events and actions of
political leaders. Unfortunately, for the topic under investigation
here, there are not enough surveys that coincide closely enough in time
to a presidential trip for analysis. Without such data, it is hard to
assess whether a person received the communication, but we can look at
the behavioral and/or opinion implications of communication reception.
That is the tack taken in this article, which asked whether trips
resulted in increased foreign public attention to the president, as
tracked with Google searches. With data spanning forty-two nations
across 196 weeks during Barack Obama's first term, the analysis
found that searches for the president in the visited nation rise by
about 25% during the week of the visit. These results support the notion
that presidents can focus public attention through high-level public
diplomacy activities and converge with other studies that find that such
activities appear to affect actual opinions, at least under some
conditions (Dragojlovic 2011, 2013; Goldsmith and Horiuchi 2009).
We did not find, however, that trips by the secretary of state also
led to increased searching for the president. At first blush, this
stands in contrast to Goldsmith and Horiuchi (2009). Due to the paucity
of presidential trips, they combined presidential and secretary of state
trips into one variable and thus did not distinguish the opinion effects
of a presidential versus a secretary of state visit. Their modeling
decision is reasonable given their data and the limited number of
presidential trips to the nations during the time period that they
studied. It is possible, though, for a visit by the secretary of state
to change foreign opinion without altering search behavior for the
president. A visit by the secretary of state may lead local political
elites and the mass media to discuss the issue that promoted the visit.
This heightened elite and media discourse may in turn affect citizen
opinions. (12)
Still there are some notable limitations of the research reported
here. First, it is for only one president across approximately a
four-year period. Second, the Google search data do not tell us anything
about why an individual searched for the president online or about
whether the trip (and/or the search) affected the individual's
attitudes toward the United States and the president. Third, Internet
searches are not performed by a representative sample of citizens, so we
cannot generalize the findings reported here to citizens in general. In
fact, from the Google Trends data, we know virtually nothing about the
individuals who searched for the president such as their demographics or
motivations to conduct the search. Nor do we know if the search affected
their level of information or their opinions. Still, Internet access and
searching is becoming very widespread; hence, the results apply to a
significant element of the mass public in these nations. Fourth, it is
not clear exactly how or if Internet search behavior translates into
opinion change, for the searchers as well as throughout the country.
Information retrieved from a search may alter someone's opinion
about the president, the United States, and/or the policies of the
United States. Searchers may also communicate the results of their
searches to nonsearchers, diffusing information gained from their search
to others. The linkages among presidential trips, Internet searching,
other expressions of attention focusing (like reading newspapers), and
opinion change are as yet not understood fully. This article represents
one attempt to fill in some of that gap and I hope will stimulate
continued research on the effects of high-level public diplomacy.
References
Barrett, Andrew W. 2004. "Gone Public: The Impact of Going
Public on Presidential Legislative Success." American Politics
Research 32 (May): 338-70.
Baum, Matthew A., and Samuel Kernell. 1999. "Has Cable Ended
the Golden Age of Presidential Television?" American Political
Science Review 93 (March): 99-114.
Beck, Nathaniel L., and Jonathan N. Katz. 1995. "What to Do
(and Not to Do) with Time-Series Cross-Section Data." American
Political Science Review 89 (September): 634-47.
Beck, Nathaniel, and Jonathan N. Katz. 2011. "Modeling
Dynamics in Time-Series-Cross-Section Political Economy Data."
Annual Review of Political Science 14: 331-52.
Breitung, Jorg, and M. Hashem Pesaran. 2008. "Unit Roots and
Cointegration in Panels." Advanced Studies in Theoretical and
Applied Econometrics 46: 279-322.
Canes-Wrone, Brandice. 2006. Who Leads Whom?: Presidents, Policy,
and the Public. Chicago: University of Chicago Press.
Choi, Hyunyoung, and Hal Varian. 2012. "Predicting the Present
with Google Trends." Economic Record 88 (June): 2-9.
Cohen, Jeffrey E. 1995. "Presidential Rhetoric and the Public
Agenda." American Journal of Political Science 39 (February):
87-107.
--. 2010. Going Local: Presidential Leadership in the
Post-Broadcast Age. New York: Cambridge University Press.
Datta, Monti Narayan. 2009. "The Decline of America's
Soft Power in the United Nations." International Studies
Perspectives 10 (August): 265-84.
Dragojlovic, Nicolas Isak. 2011. "Priming and the Obama Effect
on Public Evaluations of the United States." Political Psychology
32 (December): 989-1006.
--. 2013. "Leaders without Borders: Familiarity as a Moderator
of Transnational Source Cue Effects." Political Communication 30
(May): 297-316.
Eberhardt, Markus. 2012. "Estimating Panel Time-Series Models
with Heterogeneous Slopes." Stata Journal 12 (1): 61-71.
Edwards, George C., III. 2006. On Deaf Ears: The Limits of the
Bully Pulpit. New Haven, CT: Yale University Press.
--. 2009. "Leading the Public." In The Oxford Handbook of
the American Presidency, eds. George C. Edwards III and William G.
Howell. New York: Oxford University Press, 182-207.
Eshbaugh-Soha, Matthew, and Jeffrey Peake. 2011. Breaking through
the Noise: Presidential Leadership, Public Opinion, and the News Media.
Stanford, CA: Stanford University Press.
Farnsworth, Stephen J., S. Robert Lichter, and Roland Schatz. 2013.
The Global President: International Media and the US Government. Lanham,
MD: Rowman & Littlefield.
Foote, Joe S. 1990. Television Access and Presidential Power: The
Networks, the Presidency, and the "Loyal Opposition." New
York: Praeger.
Franzese, Robert J., and Jude C. Hays. 2007. "Spatial
Econometric Models of Cross-Sectional Interdependence in Political
Science Panel and Time-Series-Cross-Section Data." Political
Analysis 15 (2): 140-64.
Goldsmith, Benjamin E., and Yusaku Horiuchi. 2009. "Spinning
the Globe? U.S. Public Diplomacy and Foreign Public Opinion." The
Journal of Politics 71 (July): 863-75.
--. 2012. "In Search of Soft Power: Does Foreign Public
Opinion Matter for US Foreign Policy?" World Politics 64 (July):
555-85.
Hill, Kim Quaile. 1998. "The Policy Agendas of the President
and the Mass Public: A Research Validation and Extension." American
Journal of Political Science 42 (October): 1328-34.
Kernell, Samuel. 1993. Going Public: New Strategies of Presidential
Leadership. Washington, DC: CQ Press.
Kernell, Samuel, and Laurie L. Rice. 2011. "Cable and the
Partisan Polarization of the President's Audience."
Presidential Studies Quarterly 41 (December): 693-711.
Lammers, William W. 1982. "Presidential Attention-focusing
Activities." In The President and the Public, ed. Doris Graber.
Philadelphia: Institute for the Study of Human Issues, 145-71.
Lui, Catherine, Panagiotis T. Metaxas, and Eni Mustafaraj. 2011.
"On The the Predictability of The the US Elections through Search
Volume Activity." Paper presented at the Proceedings of the IADIS
International Conference on e-Society, Avila, Spain, March.
McCombs, Maxwell. 2013. Setting the Agenda: The Mass Media And
Public Opinion. New York: Wiley.
McCombs, Maxwell E., and Donald L. Shaw. 1972. "The
Agenda-Setting Function of Mass Media." Public Opinion Quarterly 36
(Summer): 176-87.
--. 1993. "The Evolution of Agenda-Setting Research:
Twenty-Five Years in the Marketplace of Ideas." Journal of
Communication 43 (June): 58-67.
Pelat, Camille, Clement Turbelin, Avner Bar-Hen, Antoine Flahault,
and Alain-Jacques Valleron. 2009. "More Diseases Tracked by Using
Google Trends." Emerging Infectious Diseases 15 (August): 1327-28.
Pesaran, M. Hashem. 2004. "General Diagnostic Tests for Cross
Section Dependence in Panels." IZA Discussion Paper No. 1240.
http://repec.iza.org/dp1240.pdf (accessed October 23, 2014).
Pesaran, M. Hashem, and Ron Smith. 1995. "Estimating Long-Run
Relationships from Dynamic Heterogeneous Panels." Journal of
Econometrics 68 (July): 79-113.
Powell, Richard J., and Dean Schloyer. 2003. "Public
Presidential Appeals and Congressional Floor Votes: Reassessing the
Constitutional Threat." Congress & the Presidency 30 (Autumn):
123-38.
Reilly, Shauna, Sean Richey, and J. Benjamin Taylor. 2012.
"Using Google Search Data for State Politics Research an Empirical
Validity Test Using Roll-off Data." State Politics & Policy
Quarterly 12 (June): 146-59.
Rose, Richard. 1988. The Postmodern President: The White House
Meets The World. Chatham, NJ: Chatham House.
Smith, Michael J. 1997. "Going International: Presidential
Activity in the Post-modern Presidency." Journal of American
Studies 31 (July): 219-33.
Vosen, Simeon, and Torsten Schmidt. 2011. "Forecasting Private
Consumption: Survey-based Indicators vs. Google Trends." Journal of
Forecasting 30 (September): 565-78.
Weeks, Brian, and Brian Southwell. 2010. "The Symbiosis of
News Coverage and Aggregate Online Search Behavior: Obama, Rumors, and
Presidential Politics." Mass Communication and Society 13 (4):
341-60.
Wilson, Sven E., and Daniel M. Butler. 2007. "A Lot More to
Do: The Sensitivity of Time-series Cross-section Analyses to Simple
Alternative Specifications." Political Analysis 15 (2): 101-23.
Young, Garry, and William B. Perkins. 2005. "Presidential
Rhetoric, the Public Agenda, and the End of Presidential
Television's 'Golden Age.'" The Journal of Politics
67 (November): 1190-205.
JEFFREY E. COHEN
Fordham University
Jeffrey E. Cohen is a professor of political science at Fordham
University. His most recent book is Presidential Leadership in Public
Opinion: Causes and Consequences. He is also the author of numerous
articles in academic journals.
(1.) Each visit to a foreign nation is counted separately, even if
the president visited the nation two or more times. For the secretary of
state totals, I do not count joint presidential-secretary trips. The
secretary almost always accompanies the president on a foreign trip. The
source for these data is the Office of the Historian of the State
Department, http://history.state.gov/.
(2.) The literature on gong public and public opinion is extensive.
See Edwards (2006, 2009) and Cohen (2010, 14-17) for useful reviews. The
major studies of the impact of going public on Congress are Canes-Wrone
(2005), Barrett (2004), and Powell and Schloyer (2003), but they differ,
with Canes-Wrone suggesting conditional effectiveness, Barrett finding
more general impacts, and Powell and Schloyer not finding going public
to be effective in affecting voting in Congress.
(3.) See the efforts that Goldsmith and Horiuchi (2009) go through
to create such an indicator.
(4.) The term "attention focusing" was first coined by
Lammers (1982).
(5.) Goldsmith and Horiuchi (2009, 865) distinguish presidential
credibility from the president being controversial or noncredible. A
credible president is one who is trustworthy, a noncredible one is a
president who is not trustworthy, while a controversial president is one
who takes positions that rouse debate and opposition around the world.
The credible--controversial--not credible distinctions come from the
literature on messengers attributes in communications.
(6.) The secretary of state always accompanies the president when
the chief executive visits another nation but may visit a nation alone,
without the president.
(7.) For studies that use Google Trends data, see Choi and Varian
(2012), Pelat et al. (2009), and Vosen and Schmidt (2011). For
applications to political data, see Lui, Metaxas, and Mustafaraj (2010),
Reilly, Richey, and Taylor (2012), and Weeks and Southwell (2010).
(8.) If enough search volume exists, Google Trends presents the
data in daily units.
(9.) The robust lambda accounts for the cross-sectional
correlation. Demeaning subtracts the cross-sectional means from each
series.
(10.) The week of May 1-7, 2011.
(11.) See Pesaran and Smith (1995), Pesaran (2006), and Eberhardt
and Teal (2010) for technical discussions of these estimators.
(12.) Also, secretary of state visits may affect other kinds of
search behavior such as for the secretary.
TABLE 1
Descriptive Statistics on Google Searches for "Barack Obama" and
the Number of Presidential and Secretary of State Visits,
by Country, February 2009 through October 2012
Barack Obama Searches
Country Mean Std. Dev. Min Max
Argentina 15.79082 13.37898 0 100
Australia 23.79082 11.43912 9 100
Austria 26.75 13.98291 9 100
Belgium 27.64286 13.35953 9 100
Brazil 12.11224 10.89567 4 100
Canada 23.72449 13.57999 10 100
Chile 7.311224 7.664776 0 100
China 13.9898 12.05117 0 100
Colombia 13.7398 11.87795 0 100
Czech 4.632653 8.23985 0 100
Denmark 16.39286 11.1587 0 100
Finland 32.5102 17.15584 0 100
France 32.19898 15.54567 12 100
Germany 28.36224 13.91463 10 100
Hungary 22.67857 13.68262 0 100
India 8.311224 8.012175 3 100
Indonesia 10.12755 13.92688 2 100
Ireland 3.443878 7.284822 1 100
Israel 18.22959 18.6488 0 100
Italy 9.255102 8.504371 3 100
Japan 20.14796 17.01714 0 100
Malaysia 30.07653 17.97518 0 100
Mexico 13.94388 9.875217 5 100
Morocco 13.13265 16.03595 0 100
Netherlands 25.11224 12.87485 7 100
New Zealand 38.55102 18.81959 0 100
Nigeria 20.63776 18.9476 0 100
Norway 8.831633 8.656832 2 100
Peru 24.45408 16.26804 0 100
Philippines 10.46939 10.10479 3 100
Poland 5.806122 7.954561 2 100
Portugal 12.87755 9.042651 4 100
Romania 13.86224 15.2257 0 100
Saudi Arabia 21.26531 27.07314 0 100
South Africa 21.84694 13.33348 6 100
South Korea 14.20918 22.58391 0 100
Spain 27.15306 14.02734 4 100
Sweden 32.64286 17.06293 9 100
Switzerland 27.30102 14.21001 11 100
Turkey 4.928571 7.354399 0 100
United Kingdom 20.53061 10.19674 8 100
Venezuela 20.43367 20.34751 0 100
Visits
Secretary
Country President of State
Argentina 0 1
Australia 1 2
Austria 0 0
Belgium 0 6
Brazil 2 4
Canada 2 4
Chile 1 1
China 1 7
Colombia 2 1
Czech 3 0
Denmark 0 2
Finland 0 1
France 5 5
Germany 2 7
Hungary 0 1
India 2 3
Indonesia 3 4
Ireland 1 1
Israel 0 5
Italy 1 1
Japan 3 5
Malaysia 0 1
Mexico 3 4
Morocco 0 2
Netherlands 0 2
New Zealand 0 1
Nigeria 0 2
Norway 1 2
Peru 0 2
Philippines 0 2
Poland 1 1
Portugal 1 0
Romania 0 0
Saudi Arabia 1 2
South Africa 0 3
South Korea 3 6
Spain 0 1
Sweden 0 1
Switzerland 0 5
Turkey 1 8
United Kingdom 2 5
Venezuela 0 0
Source: Google Trends for Barack Obama searches and the Office
of the Historian of the State Department for President and
Secretary of State Visits, http://history.state.gov/.
TABLE 2
Impact of Presidential Trips on Weekly Google Searches
for "Barack Obama," February 2009-October 2012
Model (1)
Regression Model (2) Panel
Clustered on Corrected
VARIABLES Country Standard Errors
Presidential Trip 42.99 *** 37.21 ***
(5.493) (1.228)
Secretary of 1.160 0.895
State Trip (1.125) (0.761)
Bin Laden 18.37 *** 17 90 ***
Announcement (3.419) (2.235)
Searches for 0.436 *** 0.342 ***
"United States" (0.0359) (0.0231)
Constant 6.696 *** 9.552 ***
(0.765) (0.917)
Observations 8,232 8,232
[R.sup.2] 0.273
[R.sup.2] Overall 0.202
[R.sup.2] Within 0.285
[R.sup.2] Between 0.051
sigma_u 8.737
sigma_e 12.04
rho 0.345
Wald chi2/p 375.8 1256
0.000 0.000
N = 8,232 Groups = 42, Time Units (weeks) = 196.
Robust standard errors in parentheses.
*** p < .01, ** p < .05, * p < .1.
Source: Google Trends for Barack Obama searches and the Office
of the Historian of the State Department for President and
Secretary of State Visits, http://history.state.gov/.
TABLE 3
Impact of Presidential Trips on Weekly Google Searches
for "Barack Obama," February 2009-October 2012,
Corrected for Cross-Panel Correlation
Model (3) Model (4)
CCEMG AUG
VARIABLES Estimator Estimator
Presidential 24.64 *** 24.71 ***
Visit (5.186) (5.205)
Secretary of 0.319 0.699
State Visit (1.198) (1.178)
Bin Laden 0.274 14.45 ***
Announcement (3.944) (3.452)
U.S. Search 0.00228 0.660 ***
(0.0351) (0.0528)
"Barack Obama" 0.995 ***
Searches Average (0.0989)
Presidential -39.03 ***
Visit Average (12.13)
Secretary of State -0.605
Visit Average (5.287)
Bin Laden Average 0
(0)
U.S. Searches Average 0
(0)
cdp+ 0.991 ***
(0.0964)
Constant 0.0257 0.683
(1.023) (0.949)
Wald chi2 22.88 462.6
N = 8,232, Groups = 42, Time Units = 196.
Standard errors in parentheses *** p < .01, ** p < .05,
* p < .1.
(+) cdp refers to the "common dynamic process"
Source: Google Trends for Barack Obama searches and the Office
of the Historian of the State Department for President and
Secretary of State Visits, http://history.state.gov/.
TABLE 4
Impact of Presidential Visits on Weekly Google Searches
"for "Barack Obama", February 2009-October 2012
(5) Error
Correction,
Mean Group
Averaging
VARIABLES Estimation
Error Correction 0.389 ***
Mechanism (0.0343)
Presidential Trip 23.51 ***
(changes) (4.862)
Secretary of State 0.587
Trip (changes) (0.592)
Bin Laden Announcement 15.84 ***
(changes) (2.170)
United States Searches 0.330 ***
(changes) (0.0309)
Constant -2.426 ***
(0.380)
Observations 8,190
Groups 42
Time Units 195
Sigma 62.17
Standard errors in parentheses *** p < .01,
** p < .05, * p <.1.
Source: Google Trends for Barack Obama searches and the Office
of the Historian of the State Department for President and
Secretary of State Visits, http://history.state.gov/.
