round the Flag Effect: International Conflicts

advertisement
An Analysis of the Rally ‘round the Flag Effect: International Conflicts, Presidential
Approval Ratings, and Multilateralism
Fowler Paper Competition
Roanoke College
April 22, 2014
1
Table of Contents
Abstract
p. 2
Introduction
p. 3
Literature Review
p. 7
Argument and Hypothesis
p. 13
Research Design
p. 19
Findings
p. 25
Conclusion
p. 28
2
Abstract
The rally ‘round the flag effect is an expression that has been used to describe the
noticeable spikes in U.S. presidential approval ratings during instances of international conflict.
Although the president can generally expect an increase in public support during crises, the size
of the effect varies widely across different conflicts. Several studies have examined this
phenomenon and yet the literature lacks consensus on why in certain instances of conflict rallies
barely register, while in others they can reach upward of 30 percentage points. This paper
contributes to this discussion by proposing a conditional explanation of the effect of conflict
initiation on presidential approval ratings. Specifically, I argue that instances of militarized
interstate disputes that gain multilateral support from the United Nations Security Council or
North Atlantic Treaty Organization tend to lead to larger increases in approval ratings, which can
explain the variations in spikes. To evaluate my argument empirically, I compiled a monthly data
set of presidential approval ratings, conflict, and displays of multilateral support from 1949 to
2010. This study furthers our understanding of the complex connections between the domestic
and international policy spheres.
3
Introduction:
Following the September 11, 2001 terrorist attacks on the United States, George W.
Bush’s job approval ratings surged to a record 90 percent on September 21st, 2001. Such a drastic
increase in presidential approval ratings coincides with the idea that during instances of
international crisis, the President will experience a period of increased public support (Mueller
1970). For example, following the September 11th terrorist attacks on the United States, not only
did President George W. Bush’s approval ratings swell to 90 percent but they also lingered in the
high 80’s in the subsequent months. This demonstrates a quintessential example of the rally
‘round the flag effect, yet this phenomenon is not consistently seen in all instances of
international conflict involving the United States. In June of 1950, 75,000 North Korean soldiers
advanced over the border into Western backed South Korea and by July, President Truman had
sent American troops to fight on South Korea’s behalf (Stueck 2010, 168). In contrast to the
2001 example, President Truman’s job approval ratings did not show a massive inflation
following the involvement in an international crisis. In the months following the military
intervention, President Truman’s approval ratings hovered between the same 35 to 45 percent
range, the same as prior to the initiation of the conflict.
The tendency of the public to bolster support for the current President of the United
States during times of international crises, the rally ‘round the flag effect, has produced an
unexplainable range of variations of spikes in approval ratings. In some instances, this effect is
barely seen, as within the Korean War under President Truman, while yet in others the increases
in the ratings are massive, such as the 33% point increase as demonstrated after the 9/11 attacks
in 2001 (Gallup). This discrepancy in the public’s reaction to conflicts raises the question of
4
what other factors might be responsible for this variation in presidential approval ratings during
different instances of militarized interstate disputes.
Given the unexplained variations in spikes of approval ratings during different instances
of conflict, it becomes necessary to determine the underlying factors. This paper will analyze
support from the UNSC and NATO as the principal factor intervening in the relationship
between conflict engagement and presidential approval ratings. By examining the UNSC and
NATO’s influences, this project can further the understanding of international institution’s
impact on foreign policy in the domestic sphere. The more specific research question becomes:
what is the effect of support from the UNSC and NATO during instances of foreign conflict on
presidential approval ratings in the United States?
International politics can be seen as having a significant impact on domestic policies, as
demonstrated through the relationship between foreign militarized disputes and presidential
approval ratings. Domestically, Abramson, Aldrich and Rohde (1987) and Fiorina (1981)
illustrated that voting behavior in both presidential and congressional elections are effected by
the current Presidents approval ratings. Presidents who have a higher job approval rating are
symbolically given a general vote of confidence by the American public. This is then translated
into voting patterns in congressional elections and how the presidential incumbent will most
likely fair in the coming elections. Literature on the subject has established that voters with low
opinions of the current president are less likely to vote for representatives of the same political
party (Cover 1986; Kernell 1977; Marra and Ostrom 1989; Gronke, Koch and Wilson 1998).
“The policies, programs, and outcomes associated with the current administration serve as
important voting cues in these races, and citizens cast the congressional ballot at least in part as
an expression of their attitudes towards the president’s agenda” (Gronke, Koch and Wilson 2013,
5
785). This allows for the more popular presidents to gain leverage in influencing other political
actors to support administrative policies, and provides more options without the fear of
congressional resistance. On the other hand, a president with a low job approval rating can also
influence the next generation of elected congressional and presidential leaders. The effect that
approval ratings has over elections at both the congressional and the presidential level can serve
as a determinant for a president’s use of force on an international level.
John Mueller, the first to operationalize the rally ‘round the flag effect, theorized that
during a militarized interstate dispute, public opinion will generally favor the president due to the
fear that opposing him would hinder a successful outcome to the crisis. During crises, job
approval ratings can then become a major factor for the president when deciding how to respond.
Wittkopf and DeHaven (1987) found that direct and forceful action by the current president in
response to an international crisis have a positive relationship with public approval. Therefore,
the presidents who hold lower approval ratings may be more likely to “rattle the saber” in an
attempt to create a favorable surge in public opinion. Ostrom and Simon (1987) indicate that as a
president can influence the environmental stimuli and therefore public opinion during instances
of international conflict that they may choose a course of action specifically to influence popular
support. Further, Morgan and Bickers (1992) suggest that “aggressive foreign behavior is a
useful tool for dealing with domestic political problems” (p. 26). Through this relationship, job
approval ratings can become significant in determining not only the future in American domestic
politics, but it can be used to influence international affairs as well.
The UNSC and NATO then allow for the public to more efficiently judge the presidents
the actions as they are outside, legitimate, and independent organizations with varying members
that hold a myriad of interests, goals and values. By providing support for the U.S. president,
6
resolutions of support from these multilateral organizations may signal to the American public
that the president is acting appropriately, and therefore may explain for the variations in the
rallying effect. For the purposes of this paper, the definition of the UNSC support is an issued
resolution from the Security Council while NATO support includes special declarations of
approval and missions and interventions directed by the organization. I collected conflict
information from the Militarized Interstate Dispute (MID) data set and used primary source
documents from the UN and NATO to determine if support was given or not. Within the paper, a
means comparison and linear regression analysis were used. This paper continues with a
literature review and goes on to explain the methodology, research design, and findings.
This paper contributes to the literature that attempts to explain the underlying factor
affecting the variations in increases in presidential approval ratings during instances of
militarized interstate disputes. Up until this point, little has been written connecting the ways in
which international institutions, presidential approval ratings, and the rally ‘round the flag effect
are correlated, making this paper a significant addition to international relations scholarship. In
understanding the ways in which an international event affects an American phenomenon, it
becomes necessary to look outside of purely domestic influences. Although much of the
literature is concerned with domestic aspects will affect this phenomenon, it is important to
instead examine how international influences impact the rally ‘round the flag effect. For this
reason, this paper ties together an interesting effect of and relationship between international and
domestic politics. Further, as Chapman and Reiter (2004) produced a different set of findings
from their examination of the relationship between the United Nations and presidential approval
ratings, this study becomes important in the furthering of the ways in which international affairs
influence domestic policies.
7
Literature Review:
During instances of foreign conflict in the United States, the logical expectation would be
that approval ratings for the president would increase due to a surge in public patriotism (Mueller
1973, Bradley and Oneal 1993, Oneal and Baker 2001). Mueller labeled this concept the rally
‘round the flag effect and it has since been analyzed by a number of scholars through different
conditional factors that serve to expose the variations in presidential approval ratings stemming
from this phenomenon (Mueller 1970). Different studies have proposed that the inconsistencies
in the rallying effect are based on the impact of a specific factor such as the media, the
president’s political party, or the temporal span of the conflict. Despite the fact that these studies
have proposed a number of explanations for the variations in spikes, there is still little consensus
on the underlying causes for this phenomenon.
Lian and Oneal (1998) argued that the rallying effect of an instance of conflict is greater
when it is heavily reported by the media, as the American public cannot support something of
which they are unaware. Although they found a slight increase in mean change in the president’s
popularity when prominently reported by the New York Times, the study still does not explain
why in some instances the rallying effect generates a spike of 30 percent or more while in others
the event barely registers. Baum (2002) believes that the phenomenon is more complex than just
recording if there is a rallying effect or not. Instead, he argues that different types of Americans
respond differently to instances of conflict based on various environmental factors. In his study
he finds that different constituent groups do rally according to varying circumstances yet it is due
to a number of possible economic and political environmental factors. Further, he found that
individual political party identification, if there is a divided government, the economic status,
and the timing of the use of force all affect the likelihood and strength of a rally. While this study
8
is useful in explaining the patterns and trends of the rallying effect, it does not provide a specific
and definitive answer to why the phenomenon is inconsistent. Baum’s study instead indicates the
possibility that these factors may play a part in determining the level of the spike, but based on
the broad nature of his findings, it cannot point to the specific underlying cause for the variation
in the rallying effect.
Lee (1977) argued that the type of foreign event plays a more significant impact on the
size and duration of the rally. He specifically examines important events that include “1)
outbreak of wars or military crises that involved actual or potential use of American troops; 2)
end of war or reconciliation of major international conflict that involved American resources; 3)
Summit conferences between the U.S. and other major powers; 4) New initiatives in foreign
policy which received a wide public attention; 5) international setbacks (or achievements),
including break off of negotiations and other embarrassing events of failures; 6) events involving
the personality of the president such as an assassination attempt” (p. 252). He argues that this is
due to the tendency of the public to support the president directly after an international event due
to the lack of information and therefore the public is more likely to grant the president the benefit
of the doubt, but that the approval ratings will digress after a brief period of time. Further, his
study serves to explain that during crises, the president becomes the focus of national attention
and instigates a feeling of patriotism that does not exist without the crisis. Lee argues that the
type of major event serves as the reason for the variations. Further, international setbacks were
found to reduce approval ratings and he concluded that wars and military crises have the largest
and most durable rallying effect while policy initiatives tend to be associated with the largest
average change in job approval ratings. Although this study is significant in analyzing the ways
in which the public perceives certain international events, Lee’s study is more focused on the
9
duration aspect of the rally ‘round the flag effect and therefore further examination is necessary
in creating a complete analysis that describes the causes of the phenomenon.
Although there is a general consensus among the different studies that war does tend to
increase presidential approval ratings, they fail to explain for the variations in the level of the
rallying effect. The empirical findings from these studies suggest that factors other than solely
conflict initiation account for the differences in public support as demonstrated in the
discrepancies in spikes in approval ratings. Although other scholars have attempted to explain
the differences through an analysis of several specific factors, no study has been able to
definitively answer why the phenomenon occurs. Each study has only been able to conclude that
a certain variable may be able to influence the differences in the magnitude of the rallying effect,
but it does not explain if one specific factor conditions the larger increase in some instances over
others.
This may be due to the fact that there is no generally agreed upon definition of foreign
conflict and therefore each study may be analyzing different examples, which would explain the
differences in findings. The most widely used definition comes from Mueller (1970): “In
general, a rally point must be associated with an event which 1) is international and 2) involves
the United States and particularly the President directly; and it must be 3) specific, dramatic, and
sharply focused” (21). Although this is the most popular definition found in the rally ‘round the
flag literature, I believe that it is too broad to allow for a definitive solution to the affect as it
does not explain the type of foreign event. The definition could be referring to military
intervention, a declared war, a direct attack, or a negotiation or summit meeting.
Due to the generic nature of this definition of a rallying point, it may explain for the
literatures inability to produce a conclusive answer. It is unclear how other scholars have
10
interpreted this definition and therefore analyzed the rallying points. While some studies may
believe that certain instances fit each of the criteria for Mueller’s definition of a rallying point,
this may not be true across all of the studies. Mueller’s generic definition does not further
explain what specific, dramatic, or sharply focused includes and therefore scholars may be using
different instances of militarized interstate disputes based on what they believe to be specific,
dramatic, and sharply focused. Each aspect of this portion of the definition is generalizable in a
number of different ways and based on the varying nature of militarized interstate disputes in
that no one is an exact replica of another this may differ from study to study. Therefore, since it
is unclear how other scholars have interpreted Mueller’s general definition or chose to identify
foreign conflict through a different set of terms, this lack of a uniform conceptualization can be
used to explain the inconsistencies in the literature. Without this commonality, it cannot be
possible for the current literature to create a substantial and definitive understanding of the rally
‘round the flag effect if the authors are not defining the same category of events.
The differences in findings could also be due to the fact that they are not all studying the
same time periods. In some studies, the authors only examine specific instances of foreign
conflict. Callaghan and Virtanen (1993) analyze only the Iran hostage Crisis under President
Carter, arguing that the rallying effect is not unconditional and must be explained along party
lines, and find that partisan groups must be understood as individual and not uniform in a
comprehensive analysis of the rallying effect. Eichenberg, Stoll and Lebo (2006) examined
President George W. Bush’s approval ratings in terms of the 2001 terrorist attack and the
subsequent Iraq War. They include five variables: “a ‘honeymoon effect, an autoregressive
function that tracks a decline in approval, measures of economic performance, measures of
important ‘rally events,’ and a measure of the costs of war – in this case, the U.S. death toll in the
11
Iraq War” (p 783). They find that the terrorist attacks instigated a steep rallying effect but it was
soon deflated by the Iraq War, which cost Bush a large percentage of his support, and that only
truly dramatic events can determine a significant rallying effect. Schubert, Stewart, and Curran
(2002) analyzed public approval ratings of George W. Bush immediately following the 2001
terrorist attacks through five factors: 1) the terrorist attack; 2) Bush’s speech following the
attacks; 3) media exposure; 4) partisan support; and 5) gender effects. This study found that
Bush’s September 11 speech was the most critical factor influencing this rally effect while none
of the other factors were significant.
Another approach has been to examine periods of 10 to 59 years. Oneal and Bryan 1995.
Baker and Oneal (2001) examined the influence of how the crises are presented to the public
from 1933 to 1992 in media coverage, bipartisan support, and from White House statements, and
concluded that rallies are associated due to a number of conditional factors: higher levels of
economic confidence, escalation of the hostility level of a dispute, and media coverage.
Chapman and Reiter (2004) analyzed the effect of the UNSC on the rallying effect from 1945 to
2001 and their findings suggest that the UN Security Council can have a positive effect on
approval ratings. Groeling and Baum (2012) attempted to explain the phenomenon using the idea
that the persuasiveness, credibility, and influence of the messages from the political elite will
determine the rallying effect from 1979 to 2003. They found that the relationship between the
content of the message, public opinion, and rallies is not solely based on elite debate but instead
the party affiliation of the elite. Without a consistent period of study, the scholars are examining
different environmental factors that may only be influential at that time. This does not account
for the phenomenon as a whole and therefore cannot result in a sole explanation that defines the
underlying cause for the rallying effect
12
The lack of a consistent period of study can also be the cause of the inconsistent findings
in the literature as without a consistent scope of research, the findings will not be able to
represent and explain the phenomenon as a whole. As each investigates a different time period,
the studies are analyzing different types of events and different environmental factors. The shortperiod studies draw conclusions on a very specific set of circumstances while they do not
consider the long term effects of other stimuli such as if there is a divided government, the
current political party of the White House, or how the economy may affect the ways in which the
rallying effect is interpreted. Therefore, it becomes necessary to use a consistent temporal scope
in order to account for those influences on job approval ratings.
While this phenomenon has been studied with Chapman and Reiter in 2004, it did not
include NATO and interpreted the variables of UNSC support and presidential approval ratings
differently in contrast with this paper. Therefore this examination into the relationship between
presidential approval ratings and multilateral support may be able to more fully explain the
phenomenon. Our studies differ in our definition of the dependent and explanatory variables and
the sample used to describe the phenomenon. Chapman and Reiter use a temporal scope of 1945
to 2001 while I instead analyze conflicts from 1949 to 2010. The dependent variable used in the
Chapman and Reiter study is based on the change in public approval of the president
immediately prior to and following a militarized interstate dispute and I instead use a monthly
data set of presidential approval ratings which allows me to observe more adequately the true
nature of the relationship between presidential approval ratings, conflict initiation, and
multilateral support. Further, they measure support from the UN on a scale of -1 to 3 as -1
indicates a condemnation of force, 0 states no action, 1 is a proposed resolution that did not pass
or a noncommittal resolution, 2 is a supportive resolution and 3 indicates only a resolution
13
calling for a cease-fire or establishment of peacekeeping force. In contrast, I only examine
official and public declarations of support from the UNSC and NATO. I chose to use only a
supportive resolution as my indicator as it is unlikely that the U.S. public would be notified of
proposed but not passed resolutions and it is doubtful that if it was publicized that the public
would significantly respond within the approval ratings.
This paper will attempt to fill the gap in the literature by analyzing how support from the
UNSC and NATO influence presidential approval ratings when the United States is involved in
conflict. It contends that approval from the institutions serves as a public way to legitimize the
acts of aggression and therefore may demonstrate why in some instances the spikes are larger
than in others. Prior studies have determined that the rallying effect may not be presented in
every instance, but are unable to propose a definitive solution to the puzzle of occasional and
inconsistent spikes. In this paper, I propose that the spikes in approval ratings are due to the
declarations of support from a legitimate international institution, such as the UNSC or NATO.
Argument and Hypothesis:
Conflict Initiation and the Rally ‘round the Flag Effect
Beginning with Mueller (1970, 1973), there has been substantial literature that
demonstrates that presidential approval ratings will increase following an event or crises that is
international, involves the United States and the president directly, and is specific, dramatic, and
sharply focused. Mueller theorized that this surge in public support stems from the ability of the
president to evoke and manipulate more patriotic responses due to a number of reasons. First,
only international events were capable of uniting the U.S., as domestic crises would instead
exacerbate previously existing political divisions. In instances of foreign policy crises, the
14
political and social integrity of the U.S. is at stake, which then influences the public to refrain
from criticism of the president, and results in a larger public backing of the president as seen in
the job approval ratings. The effects of the international crises are furthered through the lack of
media criticism as those who are typically more inclined to criticize the policies of the president
either say nothing or support the president for the fear of appearing unpatriotic (Baker and Oneal
2001, p. 668). Subsequently, without media criticism that would otherwise influence the public
to refrain from such an immediate patriotic response, the effects illustrated within presidential
approval ratings are larger and more significant. This results in the rally ‘round the flag effect in
which following a conflict, the president can therefore expect that there will be an increase in
their job approval ratings. Much of the relevant literature supports Mueller’s findings (Lee
1977;Wittkopf and DeHaven 1987; Ostrom and Simon 1985; etc) although there is much debate
on what influences the magnitude of the rallying effect. Therefore, I argue that the
differentiations seen in the level of the spikes in presidential approval ratings following a conflict
is due to support from such multilateral organizations the UNSC and NATO.
The Effect of Multilateralism on the Rally Effect
Resolutions of support from these multilateral institutions during instances of foreign
conflict signal to the public that the way in which the United States government is acting is the
correct form of response. Approval from these international institutions for military actions
undertaken by the U.S. legitimizes and validates acts of aggression. Specifically, it can be argued
that receiving approval from NATO and the UNSC is the underlying factor that accounts for
variations in the magnitude of the spikes in presidential approval ratings as seen in some
instances than in others. During times of interstate militarized disputes, the Americans tend to
15
monitor the president as to ensure that their actions are representative of the preferences of the
public. When citizens are able to feel confident in that the president is demonstrating the best
method of response to the international conflict, the approval ratings will then tend to be higher
than in instances without that same conviction. The UNSC and NATO allow for the public to
monitor the president’s actions, and therefore by legitimizing that the president’s actions are
necessary and correct can lead to a higher spike in the approval ratings, as it would not be seen
otherwise. Therefore, as a result, the rallying effect will be larger when the UNSC or NATO
issues a statement of support for the president’s method of response during an instance of
international dispute. Declarations from NATO or the UNSC can further account for the varying
spikes in approval ratings as the president cannot always depend on support from these
organizations as they are given inconsistently and can explain for the variations in the spikes
seen following a conflict.
Despite the fact that the government is supposed to operate in a way which coincides
with the public opinion of its citizens, the support for the governments decisions on how to react
during times of international disturbances are not always in the public’s best wishes. This can be
explicitly seen in the current example of Syria. According to a Gallup survey, the public support
for the United States taking military action against the Syrian government following its
suspected use of chemical weapon on its constituents is the lowest it has been in the last 20
years. In May 2013, 68% opposed using military action in Syria while 24% supported the
intervention. Following this trend, in September 2013, only 36% of Americans favored military
intervention, while 51% opposed and 13% had no opinion. In contrast, in February of 2003
immediately prior to the invasion of Iraq, 59% supported that intervention, 37% opposed and 4%
had no opinion (Gallup Politics). Despite the trends in the polling data which indicate that the
16
majority of the public does not support military action in Syria, President Barack Obama has
publicly announced his support for the intervention and has called for a congressional vote. It is
due to this possible difference in opinion that the rallying effect may not work in every instance
but to explain the occasional larger spikes in approval ratings, it can be proposed that it is due to
the support from an international institution that claims that the way in which the United States
government is acting is the best course of action. Multilateral organizations signal that the
militarized interaction is the best situation from a global standpoint. Since war is typically a
societies last resort, multilateral support indicates to the U.S. public that the president’s mode of
response is truly the best option, allowing for an increase in job approval ratings.
Thomspon (2006) theorizes that such institutions are actors within the international
community that “serve to constrain and assess the policies of potential coercing states, thereby
generating political important information that can be used to screen desirable from undesirable
actions” (3). This becomes especially important when the organizations are politically neutral as
the institutions are less likely to share the beliefs and values of the involved states. The UNSC
and NATO are not required to issue a resolution of support and there is no incentive to do so.
Due to this, such a resolution of support then may allow for an increase in public support
illustrated within presidential approval ratings. Domestically, this assures the American public
that the President’s actions are truly legitimate, as it had been voted on by an unprejudiced group
of member states that hold individual and varying interests and goals for either offering or
denying support in the interstate dispute.
Although the U.S. serves as a major influence in both institutions, the variety included in
each of the voting procedures ensures that one state cannot overrule the opinions of other
members. The UNSC has five permanent member states with veto power: China, France, Russia,
17
Great Britain and the U.S. Further, ten non-permanent member states, give of which are elected
each year by the General Assembly to serve a two-year term comprise a majority of the Security
Council. When voting, nine members of the council, including supporting votes from all of the
five permanent member states, must be present in order to adopt a resolution. Based on the
myriad of goals, interests, and values demonstrated within the UNSC that must be aligned in
order to pass a resolution of support for a militarized action issued by the U.S., it demonstrates
that such a resolution will be significant in terms of both global and domestic politics (United
Nations Security Council 2013).
The inability of the U.S. to dominate its agenda within international institutions is further
demonstrated within NATO’s voting procedures: “Decisions are agreed upon on the basis of
unanimity and common accord…This means that policies decided upon by the North Atlantic
Council (NAC) are supported by and are the expression of the collective will of all the sovereign
states that are members of the Alliance are accepted by all of them” (North Atlantic Council
2013). As both the UNSC and NATO require a general consensus from a number of different
states, each holding to its own interests in the global arena, it further legitimizes the idea the
resolutions of support from these institutions is significant. Such a resolution therefore indicates
to the American public that the threat is real, the militarized response to the threat is the best
response and that the militarized response to the threat is globally legitimate. In terms of job
approval ratings for the president, support from international institutions allows for the American
public to assess the danger differently. Based on the resolutions of support from these
multilateral institutions, the threat is real, impending and the action taken by the president is
necessary and befitting to the situation. This then translates into the public’s perception of the
18
president by increasing patriotism and decreasing criticism, thereby illustrating the rallying
effect.
Further on a domestic level, multilateralism and support allow for the United States to
initiate the idea of burden-sharing. As Tago (2005) explains, “although acting with allies may
not, in reality, substantially reduce costs for the United States, it may provide domestic audiences
with the image that burdens of the operations are being shared with other nations” (588). By
implementing this idea that the entire human and economic cost will not be placed solely on the
American public, it defeats some of the common criticisms that wars are expensive and therefore
not always in the best interest of the country. As burden-sharing inhibits economic criticisms that
would elicit further cause for uncertainty for Americans when interpreting the conflict,
multilateral support further allows for the public to interpret the president in a more positive light
as demonstrated through an increase in presidential approval ratings.
Despite the fact that some believe that the United Nations and NATO serve as an
example of a hegemonic institution in the global arena, U.S. public opinion values the
organizations. A recent study shows that 83% of voters state that it is important for the United
States to maintain their influential role within these institutions (Mcinturff, Harrington and Garin
2012). In addition, 70% of Americans believe that the United Nations is still a necessary
institution within the current global arena (Mcinturff, Harrington and Garin 2012). These
statistics reject the idea that the majority of the public views the United Nations and NATO in a
negative light, and justify the consideration of UNSC and NATO resolutions of support as an
explanation for the variations in presidential approval rating spikes during militarized interstate
disputes. Although some are critical of the UN and NATO, public opinion overall supports such
19
institutions, therefore making the organizations a viable factor to help explain the rally ‘round
the flag effect.
Therefore, this paper will test two hypotheses:
H1: Presidential approval ratings tend to increase in years when the U.S. is involved in an
interstate militarized dispute.
H2: The positive effect of conflict initiation on US presidential approval ratings depends on
support from the United Nations Security Council and NATO on the conflict. Holding all else
constant, the presidential approval ratings will then be higher when the conflict initiated by the
U.S. receives support in resolutions of approval by the UN and/or NATO.
Research Design:
Sample:
Within this study, 12 presidents have been examined over a period of 62 years with
monthly data that includes most major modern conflicts following the creation of the UNSC and
NATO and was selected based on the availability of the data. This then allows for the rally
‘round the flag effect to be analyzed across multiple presidencies and conflicts, which will
ensure that the results are representative of the actual phenomenon. This sample includes a total
of 744 observations.
Dependent Variable
20
The dependent variable, presidential approval ratings, is measured at the interval level,
using data from the Gallup survey. The survey is records the percentage of respondents who
approve of the current president. To obtain the approval ratings, Gallup asks the question “Do
you approve or disapprove of the way that the current president is handling his job?” (Gallup).
Based on the collected responses, Gallup reports the percentage of people in the sample who
responded with “approve.” The usual sample size is around 1,000 adults, with a margin of error
of plus or minus 4% (Gallup). The household is used as the polling unit, and is contacted through
its landline phone. The caller asks to speak to the adult (18 or over) who has had the most recent
birthday. The procedure used by Gallup ensures that the survey data are fully representative of
the population of U.S. voters.
Within the 744 observations collected, the average presidential approval rating was
54.17, with the lowest rating at 22% under Truman in February 1952 and the highest at 90%
under George W. Bush in September 2001. The mean approval rating of 54% in this sample (e.g.
George H.W. Bush in October of 1990 or Johnson in April of 1966). The distribution of the
variable is not significantly skewed (skewness/std.error is less than 2 (-1.39). Out of 744 total
observations, 40 are missing for this variable between 1949 and 1989. The majority of the data
points that are not available for the purposes of this study were under the Truman administration,
but some missing observations are also scattered throughout Eisenhower’s, Johnson’s, Nixon’s,
Ford’s and Regan’s terms.
Main Explanatory Variables
21
My argument focuses on two concepts that explore the variations in presidential approval
ratings during instances of militarized interstate disputes: conflict initiation by a U.S. president,
and multilateral support from the UNSC or NATO for that action.
Conflict Initiation
Conflict initiation was coded with a 1 indicating an initiation of conflict and a 0
otherwise. The data showed that conflict initiation was not an entirely common event, yet it is
not extremely rare either. Using data that from the Correlates of War Index, which outlines each
military conflict that the United States was involved in, as well as information collected from the
Correlates of War Index, the mode and variation ratio were then used to interpret whether or not
conflict initiation had impacted presidential approval ratings. This variable does not include or
explain ongoing conflicts as that is analyzed further by the variable of conflict involvement.
Therefore, among the 62 presidential years that were observed, the modal value for conflict
initiation was 0 with a variation ratio of 13.7. In this time period, there were 122 conflicts
initiated yet in 83.6% of this sample, no conflicts were initiated. Given the relatively small
variation ratio, it can be concluded that the mode is representative of the sample. Further, it can
be said that conflict initiation is not an explicitly common event in the recent history of the
United States, yet on the other hand, it is obvious that it is not entirely rare either.
Any Support
Any support, as coded with a 0 or 1, indicated that only 2.4% of the time did the
multilateral organizations offer their support in an instance of a militarized interstate dispute.
With the next explanatory variable labeled any support, which observes if there was any sort of
multilateral support garnered by the UNSC or NATO, it was measured nominally based off
22
information collected through the Correlates of War table and data recorded by analyzing the UN
resolutions during instances when conflict was initiated by the United States. From this, the
variable is defined only by if there was any support from the UNSC or NATO regarding the
conflict, as it was coded with a 1 indicating endorsement or a 0 demonstrating no support. Within
the 62 presidential years that were observed in this study, the mode was 0 and the variation ratio
was 2.4, making the mode representative of the sample.
Control Variables
UNSC Support
Within support from the UNSC, the measurement is based on a nominal level scale that
indicates if the UNSC offered its support or not during times of conflict in which the United
States was involved between the years of 1949 to 2010. The data was collected from the
Correlates of War Index, which listed summaries of conflicts. From there, I examined summaries
of documents taken from the UN website of each of the resolutions passed in that year in order to
determine whether support was given or not.
When there was a conflict, it was coded with a 1 and otherwise with a 0 to recognize no
conflict, which was then measured and indicated that it is incredibly rare for the UNSC to offer
its support. Since it is a nominal variable, the central tendency was measured through the mode
and the variation ratio was used as to calculate dispersion. When evaluated, the data showed that
the mode is 0, with a variation ratio of 2%. This indicates that as seen within the 744
observations within the sample, only 2% of the time did the U.N. offer its support for a conflict.
Both the mode of 0 and the variation ratio of 2 are very representative of the sample and
23
demonstrate that within the data, it is incredibly rare for the UNSC to offer its support during an
instance of foreign conflict which involves the United States.
NATO Support
For NATO support, I used information from the Correlates of War Index that described
instances when the U.S. was involved in NATO missions or when NATO endorsed U.S. actions
in a foreign conflict. NATO support was coded in the same manner as the UNSC and it also
showed that NATO support is incredibly rare. It was measured nominally as 0 indicated that
there was no support and 1 reflected an instance of endorsement. Within 744 observations, there
were only 6 instances when NATO issued their support. The mode in this study is 0 and the
variation ratio is .8%. The data then indicates that only .8% of the time did NATO offer its
support for the U.S. From this very small variation ratio, it is clear that the mode is therefore
representative of the sample.
Conflict Involvement
Conflict involvement was coded with either a 0 or a 1 and demonstrated that 82.5% of the
time, the U.S. was involved in a conflict. Within the variable of conflict involvement, I
nominally measured the data based off of the information retrieved from the PBS database and
the Correlates of War Index that detailed the military conflicts in which the United States was
involved. This variable is defined only by if the United States is involved in a conflict in that
month and year. This includes ongoing conflicts, as well as militarized interstate disputes that are
initiated in that month and year as well. Among the 62 presidential years that were observed, the
modal value for conflict in that particular year was 1 with a variation ratio of 14.4. From this, it
indicates that 82.5% of the time, the U.S. was involved in a conflict through the context of this
24
study. This variable was included for the fact that if the United States is already involved in
another conflict, public fatigue with the war may reduce the rallying effect.
Dispute Severity
Dispute severity was measured as it was taken from the Correlates of War Index with a 3,
4, or 5 as to indicate a range from display of force to declared war. The results show that the
quartile range and median of 0 were representative of the sample. Under the variable dispute
data, which analyzes the severity of a conflict, the median and the quartile range were 0. When
coded as a 5, this indicates that the dispute was a declared war. A 4 illustrates a use of force and
a 3 describes a display of force. Out of the 744 observations, there were 69 cases of a dispute
categorized as a 3, or labeled as a display of force, .47 cases demonstrated a use of force under
the classification of a 4 and only 4 were marked as a declared war, or classified as a 5. Both the
quartile range and median of 0 are representative of the sample.
Unemployment
Unemployment was used as a control variable as it allows for the effects of the economy,
which can influence job approval ratings, to be separated from the effects of multilateralism.
There was a high level of skewness in the data, making the median more representative of the
sample. Unemployment has been collected and inputted using the data from the Bureau of Labor
Statistics, is measured at the interval level for every month from January 1949 to January 2010.
This control variable will allow for me to separate the effects of conflict and multilateralism
from the effects of the economy, which has been identified in previous studies as an important
determinant of presidential approval ratings in the U.S. The rate of unemployment is a good
indicator of the state of the economy and other studies have proven that there is a negative
25
relationship between unemployment and presidential approval ratings (Norpoth and Yantek,
1983). Over the 62 presidential years and 744 observations examined in this study, the average
unemployment rate is 5.74%, with a standard deviation of 1.62. The lowest rating was 2.5% in
June 1953 under Eisenhower and the highest was 10.8% under Reagan in December 1983. When
the skewness (.658) is divided by the standard error (.090), the results equaled 7.31, indicating
that there is a high level of skewness within the data. Since there is a positive and significant
skew in this variables distribution the median of 5.6% is a better representation of the central
tendency.
Method
To analyze the data, a simple and controlled comparison of means and linear regression
analysis was used. First, the comparison of means will be used to summarize the relationship
between the explanatory and the dependent variables. A linear regression analysis will be used to
estimate size and direction of the relationships and to make inferences from the sample about the
population.
Findings:
Bivariate Exploratory Statistics: Comparison of Means
The means comparison showed that presidential approval ratings tend to be higher when
the U.S. is involved in a conflict. In Table 1, the mean of approval for years with conflict
initiation is 58.75, as compared to presidential approval in years without a conflict initiation,
which is 53.24. These findings provide support for my first hypothesis, as it illustrates that when
conflict is initiated, the mean for the presidential approval ratings are higher.
26
When support from the UNSC and NATO were incorporated, the means of presidential
approval ratings combined with multilateral support were lower than if multilateral support was
given. When conflict is initiated and there is no multilateral support from the UNSC or NATO,
the mean is 58.75, yet when these institutions do offer support during instances of militarized
interstate disputes, the mean is only 46.00. As this indicates that the mean approval ratings for
the president are lower during foreign conflict when there is support from the UNSC or NATO,
this does not support my hypothesis. This finding contradicts my second hypothesis as the effect
of UNSC and NATO support on presidential approval ratings when involved in a foreign conflict
is in fact negative.
When analyzing the severity of the dispute as compared against multilateral support,
these means also do not support my second hypothesis. Dispute severity also shows that the
means are higher than if there was no support from the international institutions. When the
United States is involved in a level 3 conflict with no multilateral support, the mean is 57.77
while when there is support, the mean is 51. With no multilateral support and the United States is
involved a conflict categorized as a 4 in terms of its hostility level, the mean is 59.17 while with
support, the mean is 45.75. Based on disputes labeled as a 5, the mean is 73.00 with no support,
and 44 with support. The second means comparison which includes multilateral support as it
correlates to dispute severity and presidential approval ratings, seen in Table 2, demonstrates that
during instances of militarized interstate disputes, presidential approval ratings will actually be
lower if the UNSC or NATO displays support and therefore the means comparison also does not
support my second hypothesis.
Unemployment and conflict involvement have a positive relationship with presidential
approval ratings. When the control variable of unemployment is factored in, it then becomes
27
clear that if there is high unemployment, only 52.95% of the American public would show
support through presidential approval ratings yet if there is a time of low unemployment, 60.30%
would approve, as seen in Table 3. For unemployment, as seen in Table 4, when the United
States is already involved in conflict and multilateral support is given, the mean approval rating
is 56.11 while when support is not granted, the mean is 54.73. Due to the fact that the mean of
conflict initiation is higher than the mean of no conflict, one can see that the positive effect of
foreign conflict on presidential approval ratings is consistently a variable that will continue to
bolster the same results and therefore supports my first hypothesis. Even separated through
periods of high and low unemployment, the means of conflict initiation are consistently higher
than without conflict initiation as seen at 54.04% for low unemployment and 50.02% for high
unemployment. Although this supports my first hypothesis, these results still do not support my
second hypothesis. With high unemployment and conflict initiation, presidential approval ratings
when combined with multilateral support is 45 while with low unemployment is 46.16, as
compared with 60.30 and 52.95 for conflict initiation without multilateral support. Therefore,
although support from the UNSC and NATO causes presidential approval ratings to decrease,
conflict initiation regardless of high or low unemployment has a positive relationship with
approval ratings.
Inferential Statistics: Linear Regression
The linear regression analysis demonstrates the same results as from the means
comparison, therefore contradicting my second hypothesis. When conflict is initiated in that
year, presidential approval ratings will increase 4.577 points and proves to be statistically
significant with a p value of .001, as seen in Table 4. In contrast, the other variables analyzed,
28
unemployment, dispute severity, and conflict involvement, are not statistically significant
according to the linear regression analysis.
Based on the observations from the means comparison and the linear regression analysis,
I cannot reject the null hypothesis and therefore cannot prove that support from international
institutions serves as the underlying factor in the variations in approval rating spikes. Instead of
indicating that UNSC or NATO support during instances of foreign conflict does increase
presidential approval ratings, the results indicate the opposite, that there in fact is a negative
correlation. The means comparison showed that unemployment and conflict initiation were
related in terms of presidential approval ratings and therefore should be analyzed at a different
time to create a definitive explanation for the rally ‘round the flag effect.
Conclusion:
In an attempt to explain for the varying levels of spikes in presidential approval ratings
following a militarized interstate dispute, this paper analyzed the effect from the UNSC and
NATO on such ratings. Such international institutions serve as a legitimate and outside source
that could validate to the American public that the president is acting appropriately. Further,
resolutions from multilateral organizations signal that the threat is real and impending, the
militarized response is legitimate on a domestic and international level, and that the United
States may not be forced to pay the human and economic cost of foreign conflicts alone. I
therefore argued that the differences seen in the magnitude of spikes in presidential approval
ratings following an international conflict were due to resolutions of support from the UNSC and
NATO.
29
I proposed and tested two hypotheses. The first one simply establishes the rally effect,
the second explores the effect of multilateral support on rally spikes
H1: Presidential approval ratings tend to increase in years when the U.S. is involved in an
interstate militarized dispute.
H2: The positive effect of conflict initiation on US presidential approval ratings depends on
support from the United Nations Security Council and NATO on the conflict. Holding all else
constant, the presidential approval ratings will then be higher when the conflict initiated by the
U.S. receives support in resolutions of approval by the UN and/or NATO.
Using a sample of 744 observations, which examined 12 presidents over a period of 62
years from 1949 to 2010, I tested these hypotheses through a means comparison and linear
regression analysis. Both the means comparison and the linear regression analysis demonstrated
that conflict initiation has both a positive and statistically significant relationship on presidential
approval ratings, thus supporting the rally effect hypothesis. In contrast, the means comparison
and linear regression analysis did not allow me to reject the null for my second hypothesis. The
means comparison instead illustrated that support from the UNSC and NATO has a small
negative effect on presidential approval ratings following conflict initiation; the linear regression
analysis also estimated a negative coefficient for multilateral support, but showed no statistical
significance for this effect.
The inability to reject the null concerning my second hypothesis may be due to the fact
that creating consensus among sovereign nations, each of whom is focused on differing interests,
goals and values, is a slower process. Therefore, if the UNSC or NATO does not respond
30
immediately by issuing a resolution of support, then it cannot be included in analyzing changes
in presidential approval ratings directly following the initiation of a militarized interstate dispute.
As most of my sample includes resolutions that were issued while the United States was already
involved in conflict and not at the direct initiation, the lack of data corresponding to only conflict
initiation makes it difficult to efficiently assess the relationship between presidential approval
ratings and multilateral support. Therefore, as the UNSC and NATO do not always issue
supportive resolutions directly following conflict initiation, further research may prove that
instead support from multilateral organizations may have a larger impact when the United States
is already involved in a conflict. By analyzing the relationship between multilateral organizations
and presidential approval ratings when the conflict is already ongoing, this may be a more
accurate analysis due to the often lagging nature of these resolutions and could still explain the
relationships between presidential approval ratings and multilateral support. Another interesting
avenue for future research is to explore if and how presidents use multilateral support in their
public discourse on the conflict.
31
Appendix
Table 1 – Comparison of approval ratings across instances of no conflict and conflict initiation
Report
presidential approval ratings
conflict initiation
Mean
N
Std. Deviation
no conflict initiation
53.4201
588
13.23753
conflict initiation
57.9828
116
12.70740
Total
54.1719
704
13.25149
Table 2 - Comparison of approval ratings across instances of no conflict and conflict initiation
controlling for multilateral support
Report
presidential approval ratings
conflict initiation
no conflict initiation
conflict initiation
Total
any support
Mean
N
Std. Deviation
no multilateral support
53.2461
577
13.07438
multilateral support form UN
62.5455
11
18.61915
Total
53.4201
588
13.23753
no multilateral support
58.7523
109
12.58391
multilateral support form UN
46.0000
7
8.14453
Total
57.9828
116
12.70740
no multilateral support
54.1210
686
13.14395
multilateral support form UN
56.1111
18
17.21111
54.1719
704
13.25149
or NATO
or NATO
or NATO
Total
32
Table 3 – Comparison of approval ratings across instances of no conflict and conflict initiation
controlling for unemployment
Report
presidential approval ratings
conflict initiation
unemployment rate by year
Mean
N
Std. Deviation
(Binned)
no conflict initiation
conflict initiation
Total
Low
54.3213
470
13.88411
High
49.8305
118
9.49611
Total
53.4201
588
13.23753
Low
59.3804
92
12.59761
High
52.6250
24
11.90474
Total
57.9828
116
12.70740
Low
55.1495
562
13.79888
High
50.3028
142
9.95236
Total
54.1719
704
13.25149
Table 4 – Comparison of approval ratings across instances of no conflict or conflict involvement
controlling for multilateral support
Report
presidential approval ratings
conflict involvement
any support
Mean
N
Std. Deviation
no multilateral support
51.1207
116
12.29574
Total
51.1207
116
12.29574
no multilateral support
54.7316
570
13.23705
multilateral support form UN
56.1111
18
17.21111
Total
54.7738
588
13.35971
no multilateral support
54.1210
686
13.14395
multilateral support form UN
56.1111
18
17.21111
54.1719
704
13.25149
no conflict involvement
conflict involvement
Total
or NATO
or NATO
Total
33
Table 6 – Effect of conflict initiation and multilateral support on presidential approval ratings
Explanatory Variable
Model 1
Model 2
Conflict Initiation
4.805***
5.651
Sig .000
.358
3.236
2.882
Sig .332
.398
-.591*
-.502
.054
.105
-17.719*
-16.671
.074
.104
Multilateral Support
Unemployment
Conflictin*anysupp
Conflict Involvement
2.447*
.076
Dispute Severity
-.396
.824
Constant
56.748
54.283
N
744
744
R2
0.026
0.03
Note: *p,<0.1, **p<0.05, ***p<0.01
34
References
Abramson, P.R., John H. Aldrich, and David. W. Rohde. 1987. Change and Continuity in the
1984 elections. Washington, DC: Congressional Quarterly.
Baker, William and John R. Oneal. 2001. “The Nature and Origins of the Rally Round the Flag
Effect.” The Journal of Conflict Resolution (October) : 661-687.
Baum, Matthew A. 2002. “The Constituent Foundations of the Rally-Round-the-Flag
Phenomenon.” International Studies Quarterly 46 (2): 263-98.
"Bureau of Labor Statistics Data." United States Department of Labor. N.p.. Web. 23 Apr 2013.
<http://data.bls.gov/pdq/SurveyOutputServlet>.
Callaghan, Karen and Simo Virtanen. 1993. “Revised Models of the Rally Phenomenon: The
Case of the Carter Presidency.” The Journal of Politics (August) : 756-764
Chapman, Terrence, and Dan Reiter. 2004. “The United Nations Security Council and the Rally
‘Round the Flag Effect.” The Journal of Conflict Resolution (December) : 886-909
Cover, Albert D. 1986. “Presidential Evaluations and Voting for Congress.” American Journal of
Political Science 30 (4): 786-801.
Eichenberg, Richard C., Richard J. Stoll and Matthew Lebo. 2006. “War President: The
Approval Ratings of George W. Bush.” The Journal of Conflict Resolution (December) :
783-808.
Fiorinia, Morris. 1981. Retrospective voting in American national elections. New Haven, CT:
Yale University Press.
35
Gallup Politics. 2013. “U.S. Support for Action in Syria is Low vs. Past Conflicts.” September 6.
http://www.gallup.com/poll/164282/support-syria-action-lower-past-conflicts.aspx.
(September 22, 2013).
Gartner, Scott and Gary M. Segura. 1998. “Opening Up the Black Box of War: Politics and the
Conduct of War.” The Journal of Conflict Resolution (June) : 278-300.
Groeling, Tim and Matthew A. Baum. 2012. “Crossing the Water’s Edge: Elite Rhetoric, Media
Coverage, and the Rally-Round-The-Flag Phenomenon.” Journal of Politics (September)
: 1065 – 1085.
Gronke, Paul, Jeffery Koch and J. Matthew Wilson. 1998. “Follow the Leader? Presidential
Approval, Perceived Presidential Support, and the Representatives’ Electoral Fortunes.”
Journal of Politics. (October) 104-37.
Kernell, Samuel. 1978. Explaining presidential popularity. American Political Science Review
72:506-22.
Lee, John R. 1977. Rallying ‘round the flag. Presidential Studies Quarterly 7:252-56.
Lian, Bradley and John Oneal. 1993. “Presidents, the Use of Military Force and Public Opinion.”
The Journal of Conflict Resolution. (June) : 277-300.
Marra, Robin F., Charles W. Ostrom Jr. 1989. Explaining seat change in the U.S. House of
Representatives, 1950-86. American Journal of Political Science 27: 165-92.
36
Mcinturff, Bill, Elizabeth Harrington and Geoff Garin. 2012. “Index of Public Opinion on
International Issues and the United Nations.” Public Opinion Strategies. (April)
"Militarized Interstate Disputes (v3.10)." Correlates of War. N.p.. Web. 23 Apr 2013.
<http://www.correlatesofwar.org>.
Morgan T. Clifton, and Kenneth N. Bickers. 1992. Domestic discontent and the external use of
force. Journal of Conflict Resoltuon. 36:25-52.
Norpoth, Helmut and Thom Yantek. 1983. “Macroeconomic Conditions and Fluctuations of
Presidential Popularity: The Question of Lagged Effects.” American Journal of Political
Science (November) : 785-807.
North Atlantic Treaty Organization. “Consensus decision-making at NATO”.
http://www.nato.int/cps/ar/natolive/topics_49178.htm. (September 30, 2013).
Oneal, John and Anna Lillian Bryan. 1995. “The Rally Round the Flag Effect in U.S. Foreign
Policy Crisis, 1950-1985.” Political Behavior (December) : 379-401.
Ostrom, Charles W. Jr., and Dennis M. Simon. 1987. P”romise and performance: A dynamic
model of presidential popularity”. American Political Science Review 79: 334-58.
Schubert, James N., Patrick A. Stewart and Margaret Ann Curran. 2002. “A Defining
Presidential Moment: 9/11 and the Rally Effect.” Political Psychology. 23: 559-83
Stueck, William. 2010. The Korean War in World History. Lexington.: The University Press of
Kentucky. http://ehis.ebscohost.com/eds/detail?vid=2&sid=a0bb7ea9-103f-4457-bd90-
37
abe45ee1a640%40sessionmgr112&hid=105&bdata=JnNpdGU9ZWRzLWxpdmU%3d#d
b=nlebk&AN=413895 (July 19, 2013)
Tago, Atsushi. 2005. “Determinants of Multilateralism in US Use of Force: State of Economy,
Election Cycle, and Divided Government.” Journal of Peace Research. 42 (5): 585-604.
Thomspon, Alexander. 2006. “Coercion through IOs: The Security Council and the logic of
information transmission.” International Organization 60 (1): 1-34.
United Nations Security Council. “Security Council.” http://www.un.org/en/sc/ (September 30,
2013).
Witkopf, Eugene R., and Mark J. DeHaven. 1987. “Soviet behavior, presidential popularity, and
the penetration of open political systems”. In New directions in the study of foreign
policy, edited by Charles F. Hermann, Charles W. Kegley Jr., and James N Rosenau.
Boston: Unwin Hyman.
2012. “How Does Gallup Polling Work?” Gallup Public Opinion Polling.
2010. “How Are Polls Conducted?” Gallup Polling Data.
Download