Public Perception of the Presidential Toolkit
Kenneth S. Lowande∗and Thomas R. Gray†
January 15th, 2015‡
Scholars typically study a single tool of presidential action—executive orders, memoranda, proclamations,
and signing statements—in relative isolation. More accurately, presidents boast a diverse toolkit with particular actions possessing variable suitability for a given political circumstance. Our study investigates one
mechanism by which presidents may choose one tool over another: political cost. Specifically, we ask: does
public perception of policy movement vary with the means used to alter the status quo? Leveraging a survey
experiment conducted after the 2014 midterm elections, we find support for the idea that presidents have
strong incentives to take action—any action—but that more salient means like executive orders have the
potential damage respondents’ evaluations of policy change. By implication, the means of “unilateral” action
are endogenous to political circumstances, rendering any study that considers them in isolation vulnerable to
bias.
∗ Ph.D.
Candidate, University of Virginia, [email protected]
Candidate, University of Virginia, [email protected]
‡ Prepared for presentation at the 2015 meeting of the Southern Political Science Association, New Orleans, LA. We thank
Adam Hughes, Paul Freedman, Nick Winter, and Lynn Sanders for helpful comments and suggestions. We thank the Political
Psychology Working Group at the University of Virginia for supporting our survey experiment.
† Ph.D.
Scholars typically study a single “tool” of presidential action in relative isolation. More
accurately, presidents boast a diverse toolkit with particular actions possessing variable
suitability for a given political circumstance.1 The list of means by which presidents may
initiate policy change is extensive and malleable: executive orders, presidential memoranda,
signing statements, proclamations, agency directives, etc. Though studies have made progress
aggregating data and analyzing the use of these tools from year to year, a theoretical treatment
of why presidents use one tool over another is noticeably absent. To move forward this
research agenda, we investigate public response to unilateral actions. Though past studies
have regressed directive issuance on public opinion ratings, mixed findings and under theorized
expectations render the precise relationship between public opinion and unilateral action
largely unknown. Accordingly, our study investigates this mechanism by which presidents
may select one tool over another. We argue that presidential tools vary in their level of
perceived assertiveness, which results in variation in how the public responds to policy change.
As a result, we argue presidents may avoid more overt means of action when alienating the
president’s political base is particularly damaging.
To begin evaluating this theoretical perspective, we test its foundational assumptions.
Specifically, we ask: does public perception of policy movement vary with the means used to
alter the status quo? We leverage a survey experiment conducted after the 2014 midterm
elections. This approach allows us to evaluate what may be the micro-foundations of a
larger theory before investigating it using observational data. We find support for the idea
that presidents have strong incentives to take action—any action: when respondents favor
the policy in question, respondents consistently rate the president higher when the policy
change comes directly from the president. However, at the same time, more salient means
like executive orders may damage respondent’s views of policy change itself. That is, we find
that the more salient tool decreases public approval of policy change.
1 This
much has been recognized by foundational studies in president’s “acting alone” (Mayer 2001, Howell 2003).
1
Unilateral Action and the Presidential Toolkit
As Mayer (2009) notes in a recent review of the unilateral action research agenda, this
important paradigm contains two important assumptions. The first is that presidential
character, personality traits, and background do not predictably or systematically alter the
data generating process for policy outcomes from president to president. The second is that
the president’s means of action originates in the office’s authority—whether that authority
originates in constitutional grant, statutory delegation, or some other (likely informal) source.
Mayer summarizes the perspective in the following way:
Broad grants, ambiguity, and the potential for swift action are a compelling
combination. Presidential authority is amplified by a unique institutional setting
and powerful claims to legitimacy. In recognizing the potency of formal authority,
unilateral theory presents a direct challenge to the Neustadian conception of
presidential weakness. (2009, 433)
While the degree to which this challenge is direct is likely a matter of one’s own interpretation
of Neustadt (1960), it cannot be denied that the unilateral action paradigm has directly led to
a dramatic increase in systematic considerations of presidential actions. This enterprise began
in earnest with executive orders (Glieber & Shull 1992; Mayer 1999, 2001; Mayer & Price
2002; Howell 2003; Warber 2006; Fine & Warber 2012; Rudalevige 2012; Bolton & Thrower
2014), but more recently, scholars have analyzed signing statements (Kelly & Marshall 2008,
2009; Korzi 2011; Ostrander & Sievert 2013a, 2013b; Crouch et al. 2013; Thrower 2014),
proclamations (Rottinghaus & Maier 2007; Bailey & Rottinghaus 2013), and memoranda
(Cooper 2002; Dodds 2013; Lowande 2014). Though not all of the research referenced above
subscribes to some variant of the unilateral action model, they all reflect a contemporary
focus on presidential directives and their potential to impact public policy. But a critical,
unaddressed question arises from this proliferation and the paradigm that preceded it. If,
as the unilateral action model suggests, all presidential directives accomplish a similar goal
(movement of the status quo), what motivates presidents to choose one tool over the other?
2
Though some reasons may be obvious2 or uninteresting, we argue presidents have enough
discretion in the means by which they act alone that this decision will be governed by an
important strategic process.
It is also important to note that the unilateral politics paradigm does not distinguish
between means by which the president might change the status quo. Howell (2003), for
instance, develops a general theory of unilateral action and uses executive orders as a plausible
dependent variable. There is nothing unique about Howell’s depiction of spatial movement
in policy that confines it to executive order use. Research on the issuance of presidential
documents has taken this generalization seriously. Ostander and Sievert (2013a) write, for
example, that signing statements often “take on a character similar to executive orders or
line-item veto [...] and can be characterized in the tradition of Howell (2003) as examples
of power without persuasion.”(143) Bailey and Rottinghaus (2013), as well, see presidential
proclamations as “[involving] substantive matters of policy and [carrying] significant legal
weight.”(188) In other words, scholars have worked to integrate presidential documents beyond
executive orders into the unilateral action paradigm—acknowledging that, at a minimum,
these actions have been used to make substantive changes to public policy.
This is not to say that existing work has confined the proposed functions of these orders
to policy movement. When the issuance of these directives does not follow the proposed
empirical patterns of the unilateral action model, their purpose is appropriately called into
question. Ostrander and Sievert acknowledge that signing statements can be used to alter
the status quo, but also argue that they play an important role in defining the institutional
boundaries between the President and Congress. More recently, Jackman (2014) has developed
a model of presidential reputation building that frames signing statements as a response to
Congressional encroachment on institutional territory. Thus, though presidential documents
are most often treated as attempts to directly change policy, studies have also looked to
define their role in terms of the larger institutional struggle among separated powers. To
answer our main research question, it is necessary to restrict our analysis to efforts within
2 For example, the use of a signing statement can only occur in response to a law—so presidents are necessarily limited to
responding to status quos that have passed during their tenure when using this tool.
3
the unilateral action paradigm. That is, we will confine our attention to directives which
attempt to accomplish the same goal: policy change. If we assume that all presidential
directives can satisfy this function, then two critical points become clear. First, considering
them in isolation may be inappropriate—subjecting empirical analyses to the basic problems
of missing data. Second, studying when presidents will use one tool over another becomes
critical, since the implications of that question explain how previous studies will be biased.
However, it is not enough to assert that these tools often accomplish the same goal. For
presidents to issue different types based on some strategic process, they have to possess the
discretion to choose between directive types in a given situation. That is, if the president wants
to move the status quo to some policy closer to his or her ideal policy, then each of the directive
types have to be somewhat comparable in their capacity for policy change. The motivation to
select one tool over another should not merely be the function of constitutional construction,
legislative statute, or institutional circumstance. Demonstrating this is difficult in the context
of a short paper, but a few examples typify the general trend of interchangeability. Consider,
for instance, an area of public policy that president’s traditional have considerable discretion
over: immigration. Presidents have routinely “acted alone” in the area of immigration,
such that there are a variety of data points and types of actions to compare. To those with
knowledge of this area, the three excerpts below (which were taken from three executive actions
made by presidents) will be familiar. However, when the content of these directives is removed
from the larger documents from which they originate, the appropriate category of directive
is not immediately apparent. The first excerpt is part of Executive Order 12324, issued by
Ronald Reagan in 1981. The second is taken from the well known “prosecutorial discretion”
memorandum issued by then Secretary of Homeland Security, Janet Napolitano. The third
comes from a recent presidential memorandum issued by Barack Obama. The relevant
question is: could the policy movement initiated by these actions have been accomplished by
other means? Or, more broadly, is it plausible that presidential administrations engaged in
strategic decision-making when choosing which tool to use?
The Secretary [...] shall issue appropriate instructions to the Coast Guard in order
4
to enforce the suspension of the entry of undocumented aliens and the interdiction
of any defined vessel carrying such aliens.3
[U.S. Immigration and Customs Enforcement] should exercise prosecutorial discretion, on an individual basis, for individuals who meet the above criteria by deferring
action for a period of two years, subject to renewal, in order to prevent low priority
individuals from being removed from the United States.4
[The Secretaries] shall develop: (i) in consultation with private and nonfederal
public actors, including business people, labor leaders, universities, and other
stakeholders, recommendations to streamline and improve the legal immigration
system – including immigrant and non-immigrant visa processing – with a focus
on reforms that reduce Government costs, improve services for applicants, reduce
burdens on employers, and combat waste, fraud, and abuse in the system.5
The first two brought about changes in enforcement patterns for existing agencies. The
clear difference, however, is that the executive order prescribes that changes be made, whereas
the second actually describes the changes themselves. Moreover, the perceived source is
different, in that the first came directly from the president, whereas the second was issued by
a cabinet official. However, in both cases, presidents claimed credit for the policy changes
they brought about. Moreover, though the use of “prosecutorial discretion” in the Napolitano
case was recommended by John Morton, the Director of ICE a year earlier, it was largely
embraced by the White House as a means of partially implementing the DREAM Act.6 The
first and third excerpts direct officials to develop policies and recommendations. Moreover,
though the third was issued via memorandum, presidents have issued dozens of executive
orders accomplishing similar tasks. That is, the creation of inter-agency task forces to
produce policy recommendations has been accomplished both via memorandum and executive
order. The codification of specific policies exemplified by the second excerpt, has also been
accomplished by memorandum and executive order.7 Moreover, in none of the cases was
the president required (either by constitutional mandate or legislative statute) to issue a
3 Ronald Reagan. “Executive Order 12324 - Interdiction of Illegal Aliens,” September 29, 1981. Online by Gerhard Peters and
John T. Woolley, The American Presidency Project. http://www.presidency.ucsb.edu/ws/?pid=44317.
4 Janet Napolitano. “Exercising Prosecutorial Discretion with Respect to Individuals Who Came to the United States as
Children,” Memorandum of June 15, 2012.
5 Barack Obama. “Presidential Memorandum ŋŋ Modernizing and Streamlining the U.S. Immigrant Visa System for the 21st
Century,” The White House, November 21, 2014.
6 First introduced in 2001 as S. 1291.
7 See, for example, E.O. 10940, 13639, and 12071.
5
particular kind of directive. In addition, it is clear that the second and third excerpt contain
policy prescriptions that could have been accomplished by alternative means—but that those
means may have been sub-optimal given the political climate. In particular, both could have
been accomplished by executive order. In the months leading up to the memorandum, the
President and his advisors repeatedly answered charges from Republicans about “imperial
overreach” by noting that the president had issued the fewest executive orders since any
president in 100 years.8 Our central contention is that asking why an executive order was
not used is both an appropriate and important research question.
In addition to latitude related to the means of action, presidents have discretion over the
definition and title of the actions they take. That is, what presidents choose to call their
directives is rarely mandated by an external authority. As Mayer (2001) notes, importantly,
executive orders are not defined by the Constitution or congressional statute. The same can be
said of proclamations, and most of the other 27 distinct types of presidential directives (Relyea
2007). Categories of presidential directives are created intermittently. Determinations, for
example, where once issued by executive orders and memoranda, but have now emerged as a
separate class of presidential documents in the Federal Register. Categories of directives have
changed names. For example, presidential memoranda were once referred to as “letters”(Cooper
2001). Moreover, the distinctions between these documents—however presidents define them—
are not uniformly adhered to.
Thus, given that their are few formal standards associated with presidential directives, it
is not surprising that we would observe presidents using a variety of tools to accomplish the
same ends. Few statutory rules constraint presidents—so few, in fact, that presidents can
invent, rename, or do away with entirely, categories of directives at their discretion. We argue
that if presidents possess this kind of latitude, then we should observe maximizing behavior
when they choose the means by which they change the status quo. That is, presidents will
strategically choose directives which produce the best overall outcome. A key question, then,
8 “The truth is, even with all the actions I?ve taken this year, I?m issuing executive orders at the lowest rate in more than 100
years. So it?s not clear how it is that Republicans didn?t seem to mind when President Bush took more executive actions than I
did.” Statements made by President Obama on July 10, 2014.
6
is what establishes a particular means as the appropriate tool. There are several reasons a
president might choose one tool over another—some of which, we outline below. However,
for the purposes of this paper, we focus on the political costs in terms of public opinion (for
reasons which we go on to specify).
First, there may be non-strategic reasons associated with statutory requirements or
institutional circumstance. Legislation may require the president issue a particular kind of
directive when implementing a statute. Note, however, that the cases in which this kind of
constraint would apply likely fall outside of the scope conditions of the unilateral politics
model. How often does the president change the status quo policy when explicitly complying
with the statutory requirements within a particular law? Additionally, as noted earlier, a
signing statement is only issued (by definition) following the enactment of a piece of legislation.
So the circumstances in which the president may issue a signing statement will be limited. It
should be noted, however, that the policy prescriptions contained in signing statements could
also be accomplished via other directives. So the institutional circumstance only limits when
a signing statement may be issued—it does not determine when it will be issued.
Second, there may be strategic reasons based upon the likelihood of litigation and judicial
challenge. As Mayer (2001) and Howell (2003) both show, the Courts have a well-established
precedent of overturning executive orders which are deemed to go beyond the scope of
the president’s authority. Other directives, like presidential memoranda, have not been
successfully challenged—or simply have never been the subject of a lawsuit. This is not to say
that the president’s directives are more legally resilient when they are not executive orders.
The judiciary could, in theory, expand its review of presidential directives, or begin overriding
new categories of actions. It may be that interested parties opposed to the directive’s policies
may be less likely to litigate, given the lack of precedent.
Finally, there may be strategic reasons based upon public perception of the presidential
toolkit. That is, the means of action will reflection considerations of how the policy change
will be reported by mass media and perceived by the general public. Put simply, how the
president changes policy is partly a matter of political messaging. We focus our attention
7
on these considerations for two reasons. First and foremost, the contemporary salience of
unilateral action renders investigation into public reaction especially critical. “Acting alone”
(or acting dependent of Congress) is increasingly a public endeavor—a vital component of
president’s public strategy to push its legislative agenda and secure electoral gains. Though
all modern presidents have used directives to implement their preferred policies, President
Obama’s “We Can’t Wait” campaign, as well as other actions, are unique, in that it they
represent packaged efforts designed to secure electoral gain (Lowande & Milkis 2014). Figure
1 reproduces the web-presence of the White House at the time of President Obama’s recent
immigration actions. The rhetorical devices employed by the administration in this context
highlights the fact that presidential actions may be designed to claim credit for policy change
and bolster public support.
Figure 1: Public Rollout of November 2014 Immigration Actions
8
Moreover, the increased salience of the particular tools of unilateral action is unmatched
in American history. Figure 2 depicts the frequency of New York Times and Wall Street
Journal articles about the president containing the word “executive order” or “presidential
memorandum.” Though the post-millennium period is itself unique in coverage of these
directives, the dramatic increase during the Obama administration is particularly noteworthy.
Thus, unilateral action is fully incorporated into the president’s public strategy, and receiving
more informed and specific coverage from major media outlets.
Figure 2: Solid lines indicate coverage in the New York Times, dashed indicate coverage in the Wall Street Journal
80
Coverage of Unilateral Action
Number of Articles
40
60
Executive Order
0
20
Memorandum
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
Second, the precise relationship between public opinion and the frequency of unilateral
action is unclear in past research. Though Deering & Maltzman (1999), Mayer (1999), and
Fine & Warber (2012) find a negative relationship between approval ratings and the frequency
of executive orders, the theoretical explanations for these findings are largely unsatisfactory.
More often, approval ratings and public opinion are relegated to control variables in analyses
which focus on inter-branch conflict. Fine & Warber, for instance, argue that the negative
relationship indicates that presidents issue executive orders in an attempt to raise their
approval ratings through policies which benefit particular constituencies. If we grant this
9
explanation (though whether executive orders are targeted at particular constituencies is an
empirical question in itself) questions remain: Why should we expect presidents to increase
executive order issuance when they are unpopular? If executive orders have a linear impact
on public opinion, why not simply issue them consistently? Are presidents mistaken in
their belief that executive actions shore up their approval ratings? Moreover, the proposed
explanation specifies than endogenous relationship that is not modeled in previous work.
That is, past research argues that public opinion causes presidents to issue directives in order
to influence public opinion. Moreover, the efficacy of the latter relationship is suspect, given
the limitations of presidential influence on the mass public (Edwards 2006).
In summary, we proceed with a few basic points about presidential action and public
opinion: Presidents have a variety of tools which they can use to alter the status quo policy.
Those tools are not well defined, and presidents possess substantial discretion in selecting a
particular tool for a given task. Though there are a variety of reasons presidents may select
one tool over another, an important and relevant set of considerations arises out of the public
nature of the unilateral presidency—the focus of the proceeding theoretical and empirical
sections.
Presidential Action and Public Response
Given the preceding the section, assume there are a variety of tools presidents can use to the
accomplish the same goal. For the purpose of our argument, these “tools” vary on a single
dimension—how overt or assertive they are perceived to by by the public. This variation
can be thought of in two ways. First, there might be accumulated historical connotations
associated with a particular tool. For instance, as Copeland (1983) notes, several antebellum
presidents did not use the power of the veto—in part, because of the personal and public
conception of what was an “appropriate” use of the tool itself. Presidents began using the
veto more frequently after use of the veto based on pure policy-related rationale became
publicly appropriate. Contemporary use of executive orders bares a similar kind of historical
attachment. Media efforts to explain these directives to viewers inevitably involve discussions
10
of the most infamous executive orders: 9066, which ordered Japanese internment during
World War II; and 10340, which ordered the seizure of steel mills during the Korean War. Any
in-depth investigation of the tool itself invokes an immediate negative connotation. Second,
(and more plausibly, for the average voter) mere semantic differences between the directives
may frame issues and prime voters. As considerable research has shown, semantic (as opposed
to substantive) differences in the content of news stories influences how respondents evaluate
candidates and issues (e.g. Nelson et al 1997). The language of “proclamations”, “executive
orders”, “decrees” will undoubtedly appear more assertive than “memoranda”, “statements of
administrative policy”, and “signing statements.”
If presidential tools vary in terms of their level of assertiveness, then how will this variation
influence how the public perceives the actions themselves? Moreover, how will this impact
their evaluation of the policy, and the president himself? We begin with the initial hypothesis
that actions which explicitly name the president as the source of policy movement will
inevitably tie one’s evaluation of the policy change to one’s approval of the president. For
example, when the public disapproves of the policy, and the president claims credit for it, we
should expect that relationship to depress approval. While this hypothesis is fairly intuitive,
it makes an important point. Presidents often claim credit for policy movement, but they are
equally capable of obfuscating the source of policy movement. Policy changes often occur at
a lower administrative level, even if they are authorized and approved by the White House.
H1 : Explicitly presidential actions will tie presidential approval to policy approval.
Next, we argue that overt means of presidential action will negatively impact respondent’s
evaluation of the policy change itself. As Reeves & Rogowski (2014) show, attitudes about
the president and the appropriate scope of executive power are not merely identical. Thus,
we expect that policy evaluation will operate somewhat independently of attitudes about the
president. Actions perceived to be more assertive may garner an implicit stigma that has a
negative impact on a respondent’s evaluation of the policy.
H2 : More overt means of unilateral action will depress approval of the policy
movement.
11
We also expect the means of policy change to impact the public’s view of the president.
Though the level of assertiveness in the action itself may have a negative impact on the policy
evaluation, we expect it to polarize attitudes about the president himself. That is, overt
actions prime respondents to take more extreme positions on the president’s job approval.
For example, those who would have (counterfactually) weakly approved of the president will
observe the overt action and approve with greater intensity. We argue these actions prime
existing attitudes about the appropriate scope of executive power, reducing the probability
that respondents will lean or select “no opinion”.
H3 : More overt means of unilateral action will polarize views of the president.
Finally, we argue that the source of the observed negative response expected in H2 will be
largely driven by Democrats. We do not expect that something as comparatively subtle as
the means of policy will override presidential approval or partisan identification. That is, if
Republicans already rate Obama low, we do not expect that the means of action would have
a major influence. The difference, then, should be among moderates and Democrats, whose
opinions are not defined in relation to their opposition.
H4 : Members of the president’s party will be less likely to remain neutral, the more
overt the means of unilateral action.
As stated earlier, we consider these hypotheses to be potential bases for a larger theory of
presidential tool selection based upon public response. So, in the interest of advancing toward
that end, we conducted a survey experiment designed to directly evaluate these proposed
relationships.
Public Response Study
To test the impact of the method of executive-branch policy-making on public opinion about
the executive and the policy, we conducted a randomized survey experiment about United
States policy on commercial drone use. Drone policy is effective as an issue because it
meets a “Goldilocks” standard of being neither so obscure that it is unreasonable to expect
12
an average voter to be able to form an opinion of it nor so salient that existing opinions
would be overwhelming and largely determined by factors unrelated to our survey. For
example, a policy change related to healthcare or the Affordable Care Act might invariably
tie respondent’s opinions to partisan identification or ideology.
Each participant saw an article that appeared to be from CNN.com which offered a very
brief summary of a new policy in which the government would be limiting licenses for private
and commercial drone use.9 The policy—and thus the article summarizing it—is fictional,
but the presentation was crafted to seem as realistic as possible.10 Because the focus of
the experiment is on how the policy was enacted, we sought to create a “control” and set
of treatment CNN.com articles that would be identical (verbatim, where possible) except
in identifying how the policy came to be. The control article does not identify any source
or method for the policy’s genesis, simply that the “government” issued the policy. This
approach has limitations. Identifying the government generically as the source may imply
different things to different people. Some may read into this a particular or most common
method (for example, an Act of Congress). Though it is explicitly imprecise, that does not
mean it is content-less. Thus, the effect of any intervention must be properly understood not
as an absolute effect but rather the effect relative to how voters assess policy making in the
absence of details about methods of policy formation.
The first treatment is the “Agency” treatment, in which the source of the policy is an FAA
regulation. The FAA stands in for agencies more generally because of its topical relevance,
but this also means that this treatment brings in specific feelings and assessments respondents
may have had about the FAA. We do not believe that the FAA comes pre-loaded with
sufficient information to impact our result, but it is worth mentioning. More salient agencies,
such as the Department of Health and Human Services, the Environmental Protection Agency,
or the Food and Drug Administration may prime existing attitudes. Our issue selection
side-steps this potential confounder.
9A
reproduction of the survey treatments and questions appears in the Appendix.
however, that discussions of commercial drone regulation have been covered by major news outlets. For a recent
example, see Lowy (2014).
10 Note,
13
The final two treatments analyze two different “tools” in the Presidents executive arsenal. In
the first, the new drone policy is credited to a Presidential Memorandum signed by President
Obama. In the second, the policy is attributed to the President signing an Executive Order.
Otherwise, the two treatments are identical. In these and all treatments, the method is
mentioned in the headline and multiple times throughout the article.
One potential problem that these treatments pose may be thought of as the “Obama
Problem.” President Obama is a highly polarizing figure in American politics and previous
scholarship shows that his attachment to a policy may substantially alter the public’s response
to it (Tesler 2012). Therefore, if two treatments say “President Obama signed this” and two
others make no mention of President Obama, then the two presidential action treatments
become a test not just of Presidential Memoranda and Executive Orders, but also of the
respondents’ opinions of the current president himself. To control for this, we inserted a
simple line stating that President Obama approves of the new policy into both the Control
and Agency treatments.
The respondents were obtained from Survey Sampling International, INC. (SSI), which
paid respondents in terms set by SSI. The respondents were directed by SSI to a University
of Virginia website with a survey set up using LimeSurvey software. SSI used demographic
quotas to create a nationally representative sample of 640 respondents.11 Generally the quotas
were successful in creating a sample balanced across age, gender, race, and partisanship,
though the sample does over represent Democrats, as is common in online surveys. In addition
to our survey, respondents completed additional survey questions within the ten minutes
allotted. Because of the randomization of treatments and heterogeneity of subjects in the
treatments, we do not believe that there are any systematic spillover effects between surveys.
Table 1 provides an overview of the content of each treatment, with the differences relative
to the “Control” italicized.
11 The
actual numbers of respondents to any particular question varies based on incompletion rates and non-responsiveness.
14
Table 1: Summary of Treatments
Treatment
Content
Control
Drone policy outline
Obama approval
Generic “government” source
Agency
Drone policy outline
Obama approval
FAA source
Memo
Drone policy outline
Presidential Memorandum source
Obama approval (implicit in above)
Order
Drone policy outline
Executive Order source
Obama approval (implicit in above)
After receiving their stimulus, respondents were then asked a short series of questions that
form the dependent variables of subsequent empirical testing. First, all respondents were
asked whether they approved of the new drone policy. The responses allowed were “Approve
Strongly,” “Approve Somewhat,” “No Opinion,” “Disapprove Somewhat,” and “Disapprove
Strongly.” With the same possible responses, they were also asked whether they approved of
the President and whether they approved of the Executive Branch’s overall performance under
the current President. Finally, they were all asked how often President Obama stays within
the limits of the Constitution, with options “Rarely,” “About Half of the Time,” “Often,” and
“No Opinion.” Respondents were also asked a battery of demographic questions (age, race,
gender, education level, as well as ideological and partisan self-placement) and an attention
check to test whether they could correctly identify the stimulus.
15
Results
An important strategic decision for executive policy making is how “high” up the executive
pyramid the policy will originate from. Some policies emerge obscurely from a long, quiet
process in an agency sub-department. Other policies are issued by agency heads. The
President may also choose to claim credit for the policy himself through one of his own
unilateral tools. We initially test whether the President’s choice to be personally responsible
for the policy has ramifications for voter evaluations of the policy and the President. Recall
that (H1 ) predicts that overt personal involvement by the president will tie the president’s
approval rating to the policy’s approval rating.
To test this, we collapse the four treatment groups into two: those actions taken by the
President (Memo and Order) and those taken by anyone else (Control and Agency). We
will now use T-tests to compare the differences across question responses between these two
groups. To allow for simpler testing, we dichotomize the results for the policy, president, and
executive-branch performance questions. In each, both “Approve Somewhat” and “Approve
Strongly” become simply “Approve” and the same is performed on the “Disapprove” responses.
All of those who selected “No Opinion” are dropped. For the question regarding how often the
president remains within his constitutional limitations, we present the average response, with
higher numbers corresponding to strong belief in the president’s adherence to constitutional
limitations. Additionally, respondents who failed the attention check were dropped. In our
view, the attention check question is the best means to isolate those who actually received
the treatment. That is, those who cannot recall an obvious feature of the story they read less
than a minute ago may have not read the article at all. While it may be the case that our
treatments also operate on an implicit basis, we have opted to maintain the most attentive
sample of survey respondents.
16
Table 2: Impact of Personal Presidential Involvement in Executive Policy-Making on Voter Evaluations
Group
Policy
Presidential
Executive Well
President Within
Approval
Approval
Functioning
Constitution
Non-Presidential
74.4
45.2
38.0
2.09
Presidential
70.4
56.7
46.8
2.31
Difference (S.E.)
-4.0 (5.8)
11.5 (5.7)*
8.8 (5.9)
0.22 (0.09)*
N
242
303
276
314
Standard errors in parentheses. Differences may appear incorrect due to rounding in cells.
*=p<0.05, two-sided tests.
In Table 2, we present the group-level averages for all four questions and T-tests on the
between-group differences. When the president is the direct source of the policy change,
respondents were generally more positive in their evaluations of the president. They were
about 11.5 percent more approving of the president and were almost more likely to say that
the president stays within his constitutional limitations. This matches the prediction in
H1 . The policy received high approval ratings (above seventy percent) in each treatment
group. When the policy was attributed to the President, the president’s approval rating was
significantly higher. Thus, when the President “chose" not to put his name directly on the
policy, he did not gain the approval benefits. Recall that each treatment indicated that the
president’s support for the policy, thus these gains are tied directly to him taking the action,
not simply because he supports a policy the voter agreed with.
When respondents were told that the president approved of the new policy but not that
the president himself had created it, they were more likely to think the president sometimes
oversteps his constitutional limitations. This presents an interesting and counterintuitive
finding: when the president was seen as acting unilaterally, the respondents thought the
president remained within his constitutional requirements. There is a similar, but statistically
insignificant result (p<0.2) for the question of whether the executive is functioning well under
President Obama. Finally, the President’s personal involvement had no meaningful effect on
policy approval.
17
Comparing Presidential Memoranda to Executive Orders
Next, we test H2 , which predicted that more overt or salient actions would depress policy
approval. If the president chooses to claim explicit credit for the action, a second-level
strategic choice is which of his unilateral tools to use. Does the choice of method impact
voter assessments? In order to determine what the most “overt” or salient means of unilateral
presidential action were, we examined whether respondents reported having some prior
knowledge of the FAA, executive orders, and presidential memoranda. According the selfreporting among those who had received the corresponding treatments, 83 percent had at least
“heard of” the FAA, whereas 75 percent had heard of executive orders, and only 44 percent
claimed to have heard of presidential memoranda.12 Thus, the executive order treatment, as
expected, was the most salient means of action which could be taken directly by the president.
To test whether being more overt decreases policy support, we compare the evaluations
of the policy by those who received the Memo treatment to those who received the Order
treatment. These treatments provide a useful comparison, because there is little difference
(in terms of legal effect) between presidential memoranda and executive orders. We have
maintained the simplified, dichotomous coding using in Table 2.
Table 3: Impact of Choice of Presidential “Tool” on Voter Assessments of Policy-Making and Presidential Performance
Group
Policy
Presidential
Executive Well
President Within
Approval
Approval
Functioning
Constitution
Memorandum
80.0
62.3
56.1
2.33
Order
61.5
51.3
38.4
2.29
Difference (S.E.)
-18.5 (8.1)*
-11.1 (7.9)
-17.7 (8.4)*
-0.04 (0.12)
N
125
157
139
163
Standard errors in parentheses. Differences may appear incorrect due to rounding in cells.
*=p<0.05, two-sided tests.
The results match H2 . Participants in this survey responded very differently to the
new policy when it originated from an Executive Order compared to when it came from a
12 Note
that the difference between these executive order and memo self-reports is statistically significant (p<0.05).
18
Presidential Memorandum, offering significantly less approval when the policy originated in
an Executive Order. And their evaluations of how the President was running the executive
branch were also much more negative. Each of these results was statistically significant, even
with the small sample size. Participants in the Order treatment were also more negative in
their evaluations of the president himself than those in the Memo treatment, though this
finding is not statistically significant (p<0.2). Surprisingly, constitutional norms do not
appear to explain these differences, as both groups responded equally to that question.
Does Presidential Unilateral Action Drive Polarization?
To test whether the president’s personal involvement in executive policy-making drives
polarization, we reform the questions to measure intensity of response rather than simple
approval. Recall that the three approval questions were asked with a 5-point scale. We
collapse this to three points: “0” for those who responded “No Opinion,” “1” for those who
responded either “Somewhat Approve” or “Somewhat Disapprove,” and “2” for those who
responded either “Strongly Approve” or “Strongly Disapprove.” This creates a measure of
response intensity. If the president’s involvement drives polarization, we expect it would create
more rather than less extreme reactions in our respondents. For the constitutionality question,
we collapse it into a dichotomous variable which takes the value “0” when respondents claimed
that the president remained within his constitutional bounds “About half of the time” and “1”
when they said “Rarely” or “Often.” Note that this measure only runs from zero to one, while
the first three measures run from zero to two.
19
Table 4: Impact of Personal Presidential Involvement in Executive Policy-Making on the Intensity of Responses
Group
Policy
Presidential
Executive Well
President Within
Approval
Approval
Functioning
Constitution
Non-Presidential
0.98
1.36
1.29
0.74
Presidential
0.93
1.32
1.11
0.72
Difference (S.E.)
-0.05 (0.08)
-0.05 (0.07)
-0.18 (0.08)*
-0.02 (0.05)
N
343
345
276
314
Standard errors in parentheses. Differences may appear incorrect due to rounding in cells.
*=p<0.05, two-sided tests.
Table 4 does not show any polarization of responses for those who believed the policy
was created through a presidential action. In fact, responses to all four questions were less
extreme for those in the presidential group. For responses to the president’s handling of the
executive branch, this was statistically significant.
H3 predicts that more overt types of presidential action have a polarizing effect. To test
that, we re-run the analysis of Table 4 between those in the Memo group and those in the
Order group. The results are presented in Table 5.
Table 5: Impact of Method of Presidential Action on the Intensity of Responses
Group
Policy
Presidential
Executive Well
President Within
Approval
Approval
Functioning
Constitution
Memo
0.92
1.26
1.03
0.70
Order
0.93
1.37
1.19
0.73
Difference (S.E.)
0.01 (0.11)
0.12 (0.10)
0.15 (0.11)
0.03 (0.07)
N
179
180
180
163
Standard errors in parentheses. Differences may appear incorrect due to rounding in cells.
*=p<0.05, two-sided tests.
Table 5 indicates that there are no statistically significant differences between the two
groups in the extremity of their responses, failing to support H3 . While responses were always
more polarized for those in the Order treatment, most differences were small and all were
20
insignificant. We are therefore unable to draw any conclusions in support of the hypothesis
that the president’s involvement in executive branch policy-making or the use of executive
orders increases polarization.
Co-Partisans and Executive Orders
Finally, turning to H4 , we ask, do Democrats respond differently than Republicans, given
the means of policy change? If we follow the procedure in previous analyses, then there is no
statistically distinguishable difference between the response of Democrats and Republicans
to the Executive Order treatment, when compared to the memoranda treatment. So, the
negative effect uncovered earlier may not be driven by differential effects on the opposition
and co-partisans. However, subsetting the data to Democrats in these treatment groups who
recalled the treatment reduces the sample size to 98, further limiting our ability to draw
reasonable inferences.13
An Accidental Test: Measuring the Effects of President Obama’s Immigration
Action Speech
Conducting an experimental survey takes months of planning, drafting and re-drafting of
stimuli, IRB approval, pilot testing, and so on. And after all of these preliminary phases, the
researchers are still subject to the randomness of the universe to throw up the unexpected
and compromise the initial intentions of the survey. And so it was that our survey went live
in the week of November 17, 2014. On November 20, 2014, President Obama unveiled his
long-awaited executive actions on immigration. In conjunction with the actions, he gave a
speech on national television that evening. This fell right in the middle of our experiment.
This posed a problem that roughly the second half of our respondents took the survey after
executive action had become suddenly hyper salient with extensive news coverage and a
national speech by President Obama.
While this may raise issues for our above tests, it also presents an interesting opportunity
13 Note,
that we do observe that the subset of all Democrats (including those that failed the attention check) did yield an
association. Democrats had more intense opinions about Obama in the executive order treatment. They were 14 percentage
points more likely to voice an opinion in the executive order treatment (p<0.05).
21
to measure potential effects of the president’s speech. Did those who came after the speech
react to the executive order treatment differently? We test this by separating out those who
took the survey before 8 pm on November 20, 2014 from those who took it after. First, we
measure whether there is evidence that those who took it after were actually aware of the
coverage or speech. Because we did not have specific questions about the speech in the survey,
we resort to an indirect test: had the person heard of executive orders before? If they had
been exposed to news coverage or the speech, they likely would have. And that bears out.
Only 63 percent of respondents before the speech had heard of executive orders while a full 91
percent in the “after speech” group reported hearing of them. This difference is statistically
significant at the p<0.001 level.14
With some evidence that watching the speech had some effect, we now analyze whether
the speech impacted the way the respondents reacted to the executive order treatment. To
do this, we break down evaluations of the policy for those who received the Order treatment.
The first group is those from before the speech and the second group is those from after the
speech. Again, we use the dichotomized responses with those who failed the attention check
dropped. Table 6 shows the results.
Table 6: Impact of November 20 Presidential Speech on Effect of Order Treatment
Group
Policy Approval
Before
0.50
After
0.74
Difference (S.E.)
0.24 (0.12)*
N
65
Standard errors in parentheses.
*=p<0.05, two-sided tests.
Even with the small sample size, the speech shows a significant effect on responses to the
policy approval for those in the Order group. After hearing the speech, respondents were far
14 The results in this section are robust to multiple specifications of “before and after” the speech. There were reports about
the speech in the preceding day. Moreover, some respondents may have received this “treatment” the following morning. The
results do not substantively change if the break is specified in either fashion.
22
more positive. This implies that the speech may have given overt actions a bit of “good press”
or extra legitimacy. Having seen the president defend unilateral action in a different context
may have been persuasive for evaluations in the drone context.15
Discussion
When presidents choose to “go it alone”, they face important strategic decisions about the
means by which they change the status quo. Our study has investigated one dimension of that
strategic decision: anticipated public response. Importantly, by leveraging an experimental
research design, we have shown that the public may respond to policy change differently,
depending upon the means by which it was accomplished. Specifically, we find that though
explicitly presidential actions tie approval of the policy to the approval of the president (in
this case, raising approval), more salient means of action like executive orders actually have a
negative impact on respondents’ view of the policy. How the president initiates policy change
sends an important political message to the media and to the public.
However, there are important limitations to the findings we present here. First, roughly
half of our respondents were receptive enough to the treatment to recall it minutes later. It is
possible that this is simply a limitation of our instrument, in that paid survey respondents may
simply be less attentive than the typical member of the public. Our experiment attempted
to reproduce the circumstance in which a typical person would consume news, but there are
limits to our approach. Moreover, it may be the case that the group who paid attention are
qualitatively different, in which case, our results only travel as far as similarly attentive viewers
in the public at large. Additionally, the reduction in sample size implies that replication and
expansion of these tests is necessary. Second, the results of the survey may have been impacted
by circumstances beyond our control. SSI recruited the respondents as President Obama
focused momentary, intense media attention on political issues associated with unilateral
action. We have attempted to look into possible contagion effects, but there is ultimately no
15 One additional limitation of this analysis is that assignment to “before” and “after” was not random. Because of the use of
quotas, some groups were more likely to get into the sample before rather than after. Multivariate regression may weaken these
results, but there is insufficient data to give a thorough testing.
23
way to determine who received this rogue treatment.
Despite these limitations, we have taken an important step toward demonstrating that the
“tool” selected by the president has implications for how the public perceives policy change
and the president. Importantly, this implies that the means of unilateral action—in addition
to the more general decision to “act alone”—may be endogenous to political circumstances. It
is important to note that the degree to which presidential administrations have made choices
based on upon this kind of political messaging is likely limited by historical scope. After all, it
may be that executive orders’ apparent stigma only operates within the context of heightened
polarization and contemporary news reporting. Experiments, after all, are immediate historical
artifacts once they are fielded. However, in an environment with considerable reporting of
Washington-related news (especially stories which concern the president), we would expect the
relationships uncovered here to continue for future administrations. Ultimately, substantively
equivalent means of action carry variable suitability based upon political circumstance and
anticipated public response.
References
Bailey, Jeremy D. & Brandon Rottinghaus. 2013. “The Development of Unilateral Power
and the Problem of the Power to Warn: Washington through McKinley," Presidential
Studies Quarterly 43(1): 186-204.
Bolton, Alexander, and Sharece Thrower. 2014. “Legislative Capacity and Executive Unilateralism.” Working Paper.
Cooper, Phillip J. 2001. “Presidential Memoranda and Executive Orders: Of Patchwork
Quilts, Trump Cards, and Shell Games," Presidential Studies Quarterly 31(1): 126-141
———–. 2002. By Order of the President: The Use and Abuse of Direct Executive Action.
Lawrence: University Press of Kansas.
Crouch, Jeffrey, Mark J. Rozell, & Mitchell A. Sollenberger. 2013. “President Obama’s Signing
Statements and the Expansion of Executive Power," Presidential Studies Quarterly 43
(4): 883-898.
Deering, Christopher J. & Forrest Maltzman. 1999. “The Politics of Executive Orders:
Legislative Constraints on Executive Power," Political Research Quarterly 52, 4: 767783.
Fine, Jeffrey A. & Adam L. Warber. 2012. “Circumventing Adversity: Executive Orders and
Divided Government," Presidential Studies Quarterly 42 (2): 256-274.
Howell, William H. 2003. Power without Persuasion: The Politics of Direct Presidential
Action. Princeton: Princeton University Press.
24
Jackman, Saul P. 2014. “Presidential Pens and Power: Signing Statements, Reputation, and
the Battle for Inter-Branch Authority.” Working Paper.
Korzi, Michael J. 2011. “ ‘A Legitimate Function’: Reconsidering Presidential Signing
Statements," Congress & the Presidency 38 (2): 195-216.
Lowande, Kenneth S. 2014. “After the Orders: Presidential Memoranda and Unilateral
Action,” Presidential Studies Quarterly 44 (4): 724–741.
Lowande, Kenneth S. and Sidney M. Milkis. 2014. “ ‘We Can’t Wait’: Barack Obama,
Partisan Polarization, and the Administrative Presidency,” The Forum 12 (1): 3–27.
Lowy, Joan. 2014. “Key Decisions on drones likely from Congress,” December 28, Associated
Press (http://bigstory.ap.org/article/02208c41dbcd464dba64cbe219c07f88/congress-likelymake-key-decisions-drones)
Mayer, Kenneth R. 1999. “Executive Orders and Presidential Power," The Journal of Politics
61 (2): 445-466.
———–. 2001. With the Stroke of a Pen: Executive Orders and Presidential Power, Princeton:
Princeton University Press.
———–. 2010. “Going Alone: The Presidential Power of Unilateral Action," in The Oxford
Handbook of the American Presidency, ed. George C. Edwards III and William G.
Howell, Oxford University Press, pp. 427-454.
Nelson, Thomas E., Rosalee A. Clawson, and Zoe M. Ozley. 1997. “Media Framing of a Civil
Liberties Conflict and Its Effect on Tolerance,"’ American Political Science Review.
91(3): 567–583.
Ostrander, Ian & Joel Sievert. 2013a. “The Logic of Presidential Signing Statements,"
Political Research Quarterly 66 (1): 141-153.
———–. 2013b. “What’s So Sinister About Presidential Signing Statements?” Presidential
Studies Quarterly. 43(1): 58–80.
Reeves, Andrew and Jon Rogowski. 2014. “Public Opinion Toward Presidential Powers,”
Working Paper.
Relyea, Harold C. 2007. “Presidential Directives: Background and Overview,” Congressional
Research Service, April 23.
Tesler, Michael. 2012. “The Spillover of Radicalization into Health Care: How President
Obama Polarized Public Opinion by Racial Attitudes and Race,"’ American Journal of
Political Science. 56(3): 690–704.
Thrower, Sharece. 2014. “Presidential Action and the Supreme Court: The Case of Signing
Statements.” Working Paper.
25
Appendix: Survey Treatments and Questionnaire
Figure 3: Treatment Conditions
(a) Condition: Control
(b) Condition: Agency
Figure 4: Treatment Conditions
(a) Condition: Executive Order
(b) Condition: Memoranda
26
27