Presidential Unilateral Power

Contrary to stylized accounts of policy making in democracies, it is routine for presidents, governors, and other chief executives to issue directives such as decrees and executive orders to make law on their own. This article evaluates what political scientists have learned about presidential unilateral power. In our view, while a quarter century of scholarship on the topic has yielded a variety of theoretical predictions, the empirical record offers conflicting and perhaps unreliable evidence to substantiate and adjudicate between them. We review the dominant theoretical perspectives, which focus largely on constraints related to the separation of powers and political accountability. We then evaluate the evidence supporting these arguments and conclude with recommendations for conceptual, theoretical, and empirical advancement.


INTRODUCTION
Claims of presidential overreach are persistent features of modern political life in the United States. Political scientists, historians, and legal scholars frequently express unease with the increased salience and use of unilateral powers. Presidential unilateral powers in (advanced) democracies extend to determining the immigration status of thousands, public funding of abortion, regulation of fossil fuels, the conservation of public lands, and extrajudicial detention and killings. No area of policy seems to be out of reach. Observers argue that this subversion of the separation of powers has a direct, negative impact on the democratic process itself. While President Eisenhower once saw fit, for instance, to remind the press that "the presidency is. . .part of the legislative process," contemporary presidents appear to routinely issue unilateral directives to change policy without engaging the legislative process at all (Eisenhower 1959).
The salience of presidential unilateral power coincides with developments within Congress. First, polarized political parties and increasingly narrow partisan majorities govern the chambers of Congress and contribute to congressional deadlock and dysfunction, rendering the legislative process less available as a means for presidents to pursue their agendas. Second, Congress has tended to delegate greater statutory authority to the executive branch to exercise discretion in how legislative provisions are implemented. Both sets of developments may create incentives for presidents to use administrative directives, including executive orders, memoranda, and the like, to make law on their own.
By acting on their own, presidents can seemingly recast the national political agenda and reshape the nation's public policies. The capacity for presidents to pursue their objectives through unilateral action, rather than by consulting with Congress, "virtually defines what is distinctively modern about the modern American presidency" (Moe & Howell 1999, p. 133). Beyond its impact on policy outcomes, the politics of unilateral action reveals core insights about the contemporary American presidency and larger questions that are central to both normative and positive political science, including the separation of powers, democratic performance, accountability, and representation.
We evaluate what political science has learned from a quarter century of scholarship on presidential unilateral power. In short, our view is that while the field has produced several distinct theoretical approaches to the study of unilateral power, the empirical record offers conflicting and perhaps unreliable evidence to substantiate and adjudicate between them. Though our discussion focuses mostly on the American presidency, we also highlight how the concept of presidential unilateralism has influenced understandings of politics in US states and in Latin America. We begin by reviewing how unilateral power has been defined and discussing its relationship with various forms of presidential actions. We then present the two main theoretical perspectives used to explain unilateral action. While most theoretical perspectives emphasize the importance of institutional constraints related to the separation of powers, we argue that these perspectives have largely neglected important agency problems in the exercise of unilateral power. A second, more recent literature emphasizes political accountability and the role of public opinion. This emerging work, however, features two largely siloed approaches. Some have developed formal theories of presidential unilateralism that center on the agency relationship between the president and the public, while others mostly gauge the public's reaction to unilateral action with surveys-but both tend to neglect potential complementarities.
We evaluate the evidence for these perspectives using King's (1993) classic analysis of the thenstandard methodology in presidency research as a model. After surveying nearly 100 articles and books on the topic, we examine what progress has been made in terms of reliability of research design, measurement, and generalizability. We argue that while the technical sophistication of empirical research in this area has increased, many research designs are underpowered, and most studies develop their own individual measures of key concepts, which renders findings difficult to compare or take as stylized fact. While research practices that limit this evidence are widespread, the opportunities for improving them are straightforward. As a starting point, we advocate for a centralized online repository to catalog unilateral action and systematize its application in empirical research. We conclude by noting other promising avenues for conceptual, theoretical, and methodological innovation.

CONCEPTUALIZING UNILATERAL ACTION
consent of Congress" (Moe & Howell 1999, p. 133). 1 This definition emphasizes two key characteristics. First, unilateral directives produce a new policy outcome. Because presidential documents such as statements of administration policy and signing statements do not themselves create new policy outcomes or assume the force of law, but rather are more rhetorical in nature (Kelley & Marshall 2010, Rice 2010, a unilateral politics framework is likely inappropriate for explaining their use.
A second key feature of this definition of unilateral power is the substitution of a presidential directive for a legislative enactment that could accomplish a similar outcome. Since most directives are considered management documents internal to the executive branch, they are generally less sweeping in scope and less permanent than legislation. Nonetheless, presidents' ability to create new policy outcomes on their own distinguishes the use of executive orders and similar directives from other actions that do not directly create new policies, such as the strategic use of appointments and personnel.
Finally, unilateral power is distinguished by its agenda-setting advantages (see Moe & Howell 1999, p. 138). Unlike negative powers such as the veto, for instance, which can only be wielded in response to legislation passed by Congress (also true of signing statements), presidents deploy unilateral powers as first movers. The other branches are then confronted with whether and how to respond. These conceptual criteria are linked to the strategic context in which presidents consider the use of unilateral powers and the parameters that are posited to affect their behavior.

THEORY
Under what conditions do presidents make law on their own? Theoretical accounts to answer this question generally focus on the strategic logic that shapes a president's decision to exercise unilateral power in a specific instance. These accounts typically fall into two classes, each of which emphasizes a different explanatory factor or set of mechanisms impacting unilateral power. One class of explanations studies how checks and balances between political institutions affect presidents' use of unilateral power, while a second focuses on accountability relationships between presidents and voters.

Unilateral Action and the Separation of Powers
The first and most common approach to modeling presidents' use of unilateral power emphasizes separation-of-powers issues. In these models, presidents typically seek to achieve political outcomes that best reflect their policy preferences-subject to the potential response from other political institutions. Madisonian checks and balances provide Congress and the courts with the opportunity to respond to the president's exercise of unilateral power. Congress may pass legislation that supersedes presidential directives, and the courts may use judicial review to overturn or strike down unilateral actions that receive a legal challenge. This theoretical approach argues that strategic presidents anticipate these institutional responses when contemplating unilateral action and refrain from issuing unilateral directives when subsequent action from Congress or the courts would undermine the president's policy goals (Chiou & Rothenberg 2017, Deering & Maltzman 1999, Howell 2003, Krause & Cohen 2000. The president's decision to issue a unilateral directive thus is embedded in the larger policy-making process that reflects the separation of powers. Howell (2003) presents the benchmark model in formal terms. Initially, there exists a randomly chosen status quo policy and some amount of presidential discretion; the former parameter characterizes the location of the status quo in policy space, and the latter parameter describes the extent of the president's authority to change existing law through unilateral action. Presidents decide whether to modify the status quo via unilateral action; if they do, Congress and the courts each have the opportunity to respond. Subject to supermajoritarian institutions characterized by Krehbiel (1998), Congress can veto or modify the new policy. If the president's directive is not overturned by Congress, the judiciary decides whether to uphold or overturn the directive based on whether the president exceeded the initial level of discretion provided by the original status quo. The policy outcome reverts to the original status quo if the courts strike down the president's action; otherwise, the new status quo established by presidential directive is allowed to stand.
This class of models posits two circumstances in which presidents use unilateral power. First, they do so when Congress is gridlocked over a given status quo policy. With Congress collectively unable to agree on whether and how to modify an existing policy, presidents can create new policies that could not have been produced through legislation. 2 Second, presidents can preempt legislative action to which they are opposed. Here, the president fends off more sweeping changes to status quo policies. In both instances, new policy outcomes better reflect the president's preferences, relative to the alternative that would be enacted by Congress alone.
This theoretical approach corresponds with a reconsideration of the nature of presidential power. The traditional view, owing to Neustadt (1960, p. xix), characterized the presidency as a relatively weak institution lacking formal authority. On this view, presidents' political influence is as strong as their ability to haggle with other politicians. Yet, unilateral power offers the potential for presidents to eschew legislative negotiations and instead create new policy outcomes through direct action. Not only may unilateral power enable presidents to wield greater influence on policy outcomes than they might without it, but also the increased reliance on unilateral power may have implications for the distribution of political power across the branches of government.
A key interpretive question in this literature concerns whether unilateral action indicates that a president has circumvented Congress, or instead represents the president's exercise of administration powers over the executive branch that have been delegated by Congress. The former interpretation suggests that unilateral action represents an assertion of presidential power insofar as presidents achieve policy outcomes that otherwise would elude them and is variously termed the "strategic model" (Deering & Maltzman 1999) and the "strong form" of unilateral action (Mayer & Price 2002). The latter perspective, in contrast, suggests that unilateral actions are exercised by presidents with tacit or explicit congressional approval. Adjudicating between these competing characterizations has implications for interpreting unilateral actions as assertions of presidential power. To address this issue, Chiou & Rothenberg (2017) extend the framework offered by Howell (2003) and build multiple models that generate competing predictions about the use of unilateral power, each with distinct implications for how unilateral action relates to presidential power. Distinguishing between these models empirically is important for evaluating how political institutions constrain presidents' use of unilateral authority as well as for characterizing the formal bases of presidential power. We summarize this evidence in Section 4 and evaluate it in Section 5.
3.1.1. Comparative perspectives on unilateral power. Though most scholarship has focused on the American chief executive, presidential unilateral power is a global phenomenon. Research outside the US context typically focuses on Latin America, as the rise of decree authority in Brazil and elsewhere prompted scholars to take notice (e.g., Neto 2006, Neto et al. 2003, Palanza 2019, Pereira et al. 2008, Reich 2002, Shair-Rosenfield & Stoyan 2017. Treatments of Western Europe (Huber 1998, Sala & Kreppel 1998), Russia (Parrish 1998, and Africa (Opalo 2019) also apply versions of the unilateral action framework. Relatedly, a small but growing literature examines unilateral politics in American state governments (Barber et al. 2019, Cockerham & Crew 2017, Sellers 2017. Each of these approaches largely adopts the emphasis on chief executives' strategic behavior in response to institutional constraints related to the separation of powers.
The promise of these comparative perspectives is clear. Partisan support, ideological polarization, divided government, and other measures of political context vary over time but may be endogenous to unilateral policy making. Moreover, considered in isolation, they say little about how institutional arrangements influence presidential power. A possible implication is that if the unilateral politics literature on the separation of powers is to have something useful to say about institutional reforms, system performance, or normative democratic theory, it will require institutional variation simply absent at the US federal level in the post-World War II period.
We review three important points this approach has yielded. First, this scholarship turns to the relative strength or capacity of legislatures as a determinant of unilateralism. Studies suggest that institutional weakness in legislatures promotes unilateralism, an insight that was later applied to the US context (e.g., Barber et al. 2019, Bolton & Thrower 2016. For instance, Carey & Shugart's (1998) theory of decree issuance suggests that informational deficiencies in the legislature should promote unilateral action. Reich (2002) demonstrates that formal legislative oversight results in more frequent amendments to presidential actions. Shair-Rosenfield & Stoyan (2017) and Cockerham & Crew (2017) argue that professionalized legislatures moderate the effect of political support, with more professionalized legislatures providing an effective check on executives.
A second insight is that the formalization of presidents' first-mover authority has contestable and even counterintuitive effects. Relative to other presidents in the Americas, US presidents have few formal constitutional powers. Yet, even early work noted that this lack of formal agendasetting power did not limit US presidents' policy impact (Sala 1998). Subsequent work by Pereira et al. (2008) suggests that reforms intended to limit decree issuance in Brazil increased its use and strengthened the president's relative bargaining position. By reissuing decrees prior to expiration, past presidents could extend temporary policies indefinitely. But along with constitutional reforms that put in place a firm expiration date, the national legislature mandated that legislators vote on decrees-effectively moving presidential initiatives to the top of the legislative agenda and incentivizing future presidents to flood the calendar. Moreover, the consolidation of executive power in Russia and post-Soviet states has relied mostly on informal networks of elites rather than the formal authority of chief executives (Chaisty et al. 2014).
Third, comparative research has more often focused on the broader toolkit of presidential governance. In contrast to most research on the American presidency, most studies of presidential policy making in Latin America consider the strategic choice between instruments-either decrees or the introduction of legislation (e.g., Neto et al. 2003, Pereira et al. 2005. This is partly a product of formal arrangements. In Brazil, for example, the constitution grants the president both powers, and so they are both plausibly interchangeable and relatively easy to track. Unilateralism, then, can be measured as the relative reliance on one instrument or the other. This point, as well, was later adopted by studies of the American presidency (e.g., Dickinson & Gubb 2016).
Nonetheless, these perspectives have important limits. Some are not unique to comparative politics. The basic problems of theory, data, and research design that we raise in subsequent sections of this review do not vanish with the addition of cases. For example, in countries with more formalized unilateral powers, actions are typically easier to count. But, as we later highlight, reliance on counts and ratios introduces important assumptions in the transition from theory to testing. In addition, though unilateral power manifests in some form in most democratic governments, comparative perspectives still focus almost entirely on the Americas. Finally, in multistate treatments, few studies demonstrate that the decree instrument is equivalent across policy-making contexts. The American politics literature has demonstrated there is meaningful variation in the significance of proposals within US states. Ignoring systematic variation of this kind between states is potentially more problematic.

Agency problems in unilateral action.
Agency problems in the executive branch present one obvious limitation of the emphasis on separation of powers. Presidents never act alone. Taken literally, so-called unilateral power is unilateral only with respect to Congress, the judiciary, or other nonexecutive actors. Every presidential directive is an order to an administrator. We suspect few researchers would disagree with these points, but how they ought to inform theories of unilateral power is contestable. To date, they mostly have not.
Formal theories of unilateral power set aside the potential for administrative noncompliance by focusing on the process of policy selection by presidents (e.g., Chiou & Rothenberg 2017, Howell 2003. Others define unilateral action itself as a compliance-inducing initiative on the part of presidents (e.g., Mayer 2001, Sala 1998. Rich, descriptive research often contains cases illustrating bureaucratic noncompliance and cooperation issues (e.g., Cooper 2014, Dodds 2013. But even these emphasize interbranch conflict in their summary treatments of the topic. Of course, bureaucratic obstacles in the way of presidential initiatives are not new to presidency scholarship (e.g., Burke 1992, Nathan 1983. They seem to be at the core of the argument presented by Neustadt (1960), as bureaucrats are potential targets of presidential persuasion. The idea of persuasion itself implies that presidents have power only if they can get bureaucrats to go along. Moe & Howell (1999), among many others, minimize or ignore these obstacles because of the apparent formal authority presidents enjoy through the use of directives-along with the myriad tools (e.g., appointments, regulatory review, budgetary control, etc.) that presidents wield to influence bureaucratic behavior. But there are sufficient anecdotes of policies that do not change and bureaucrats who quit rather than obey to suggest that a Goldilocks position is in order.
Put simply, the president faces agency problems. The key questions are "what kind?" and "how do they limit unilateral power?" The answer to the first likely influences the second, as the category of agency problem will influence its severity. We think it is useful to classify these agency problems as one of two types, either top-down or bottom-up. Top-down problems involve post-proposal compliance among (mostly) bureaucratic actors. Presented with some presidential directive, will these agents carry it out? Bottom-up problems involve the formulation of presidential proposals prior to enactment. Presented with the opportunity to formulate a presidential policy, what will agents send to the president's desk? Though these two phenomena are distinct, in general, agency problems should be a moderating force-they diminish presidents' ability to enact policies that make them uniquely better off. This is because the agents in question rely on multiple principals to set the scope of their duties, working conditions, and even tenure of service.
Contrary to what the unilateral politics literature typically asserts or assumes, in some respects, compliance problems are worse for the president than for Congress. Prevailing case law considers executive orders nonjusticiable-they cannot be enforced by private lawsuit (Newland 2015). In contrast, many congressional statutes are enforced by private litigants (Farhang 2010). In other words, for unilateral presidential initiatives, the main avenue used to coerce bureaucratic compliance in the United States is closed. Directives have the force of law-as many point out-but they do not bind the government to act. A president enacts policy but must rely on a better-informed agent to implement it. These compliance issues suggest the importance of accounting for the politics of delegation.
Happily, presidency scholars do not have to reinvent the wheel. A vast theoretical literature addresses delegation in political science and economics (Gailmard & Patty 2012). Lowande (2018) adopts this approach by modeling the president's selection of agents in light of Congress's power to sanction administrators directly. However, this does not address a broader question of how presidents' knowledge of this class of agency problems influences unilateral power. By contrast, Turner (2020) models the trade-off between policies pursued with Congress and those implemented unilaterally-on the basis of how effectively policy is implemented by bureaucrats. If unilateral policy making is less durable, then administrators have reduced incentives to exert costly effort. Thus, Turner shows how intrabranch politics affects interbranch bargaining by changing the appeal of presidential unilateral action relative to legislating.
But the primary challenge to understanding these trade-offs and the compliance problems they engender is measurement. To date, Kennedy (2015) presents the most systematic effort to measure compliance by hand-coding executive order citations in rule promulgation. Unfortunately, executive orders are one of many means of unilateral action, many actions do not require regulations, and rule-making dockets driven by presidential orders do not always include them. Beyond its generic definition in spatial models, what compliance means is context-specific because unilateral action itself is so diverse. Tracking rules written, projects funded, laws enforced, contracts delineated, or any other conceivable outcome is as challenging as connecting any of it to a presidential directive.
Presidents, moreover, do not write the orders they sign. Their proposals are formulated by agents. Presidents may give the final "green light" on an order, but it is conceivable they are not perfectly informed about its technical details or about the universe of alternatives rejected before it reached their desk. Most policy loss might be eliminated by the selection of faithful agents who serve at the president's pleasure-but most administrators do not fit that description. This suggests that agency problems might extend to a proposal stage. Empirical work by Rudalevige (2012Rudalevige ( , 2015 suggests that this is the case. Between 1947 and 1987, he finds that 65% of a set of randomly sampled orders were written outside of the Executive Office of the President-the institutional context in which agency problems are most plausibly ignorable. This is a fundamental challenge to theories of unilateral action, since their aim is to understand policy formulation. Put differently, presidential initiatives may confront veto players within the executive branch-long before the threat of legislative or judicial checks. This matters for understanding presidential power, more broadly, if some alternatives to the status quo that would otherwise overcome the separation of powers are foreclosed because of bureaucratic actors. This possibility, too, presents measurement challenges for empirical work. Though Rudalevige (2012Rudalevige ( , 2015 finds evidence of directives never signed, record-keeping practices in the executive branch, along with the nature of this strategic interaction, imply that we cannot recover the full set of failed proposals.

Unilateral Action and Political Accountability
Agency problems extend beyond the president and bureaucracy. They also characterize the relationship between presidents and their voters. Accountability-and its absence-has been frontand-center in normative debates over presidential power. In the most prominent indictment of presidential power in the late twentieth century, Schlesinger (2004, p. ix) lamented what he saw as the erosion of presidential accountability: [T]he American Constitution. . .envisages a strong Presidency within an equally strong system of accountability. When the constitutional balance is upset in favor of Presidential power and at the expense of Presidential accountability, the office can be said to become imperial.
Political scientists and other observers have leveled similar criticisms of presidents' use of unilateral authority. For instance, Mayer & Price (2002, p. 9) summarize the belief held by some that "the executive order is an example of unaccountable power and a way of evading both public opinion and constitutional constraints." But for theories of democratic representation, which posit that voters penalize and reward officeholders based on the policies they advance, unilateral action implies different normative conclusions. Unilateral action may be particularly well-suited for studying presidential accountability because of the clear attribution of unilateral directives to presidents. While presidents routinely express their support for or opposition to legislation pending in Congress, the president plays no formal role in shaping the content of legislative initiatives. Moreover, though the media, political observers, and presidents themselves frequently link the state of the American economy to presidential action, the effect of a particular presidential administration on economic outcomes is murky at best. In contrast, presidents cannot sidestep responsibility for unilateral directives they issued (and cannot claim credit for unilateral directives they did not). To the extent that the public has preferences over the policies presidents advance via unilateral directives and is aware of the president's unilateral directives, the conditions exist for political accountability in the context of unilateral action.
Not surprisingly, then, other theoretical approaches link unilateral action to a president's electoral considerations. This more recent class of theories portrays presidents as motivated to exercise unilateral power based on how it translates into public support. For instance, presidents may have incentives to exercise unilateral power if voters prefer skilled leaders and unilateral action provides information about a president's skill. Judd (2017) analyzes such a model, showing that presidents are more likely to act unilaterally when existing policy is low quality and as the value of holding office increases. These electoral incentives can be perverse, however, as they can sometimes lead presidents to enact lower-quality policy via unilateral action than they otherwise would in the absence of an election.
The public's reaction to the substance of a potential directive may also factor into presidential decisions. This accountability relationship can impact the authority of successors. Howell & Wolton (2018) show that when aligned with the public, presidents have incentives to act aloneeven when that action may empower opponents in the future. In another model, Kang (2020) analyzes the differential effects that unilateral action may have on a president's electoral supporters and opponents. This framework posits that voters have constitutional concerns about presidential unilateralism, which will demobilize the president's supporters unless unilateral action addresses policies in which they are invested. This theoretical perspective implies that unilateral directives can overcome potential constitutional objections when their policy content is sufficiently popular.
Several related perspectives, though less specifically concerned with unilateral action, highlight other aspects of agency relationships that may be relevant for presidential unilateralism. One such perspective suggests that, at least in some circumstances, presidents' incentives for unilateral action are asymmetric, with the political costs of inaction exceeding any potential benefits from adopting a more cautious approach. According to this view, "Presidents who fail to act, even when the statutory or constitutional basis for action is dubious, face the prospect of a substantial political backlash against them and their party" (Howell 2013, p. 105). A second perspective links separation-of-powers issues to leaders' incentives to enact bold but potentially unwise policies (Fox & Stephenson 2011). In this model, the potential for another political institution, such as the judiciary, to reverse an unwise directive could lead presidents to engage in higher rates of unilateral activity than they otherwise might. Finally, related models on pandering consider the conditions under which presidents will advance policies that are in the public interest (e.g., Canes-Wrone 2006). This research clarifies the conditions under which elections produce incentives for presidents to use unilateral action to advance popular policies in light of those policies' potential impacts.

Separation of Powers
In the past two decades, a large volume of scholarship in political science has investigated the exercise of unilateral powers by the president (e.g., Belco & Rottinghaus 2017;Bolton & Thrower 2016;Chiou & Rothenberg 2017;Dickinson & Gubb 2016;Fine & Warber 2012;Howell 2003;Krause & Cohen 1997, 2000Mayer 2001;Moe & Howell 1999;Warber 2006). In contrast to the theoretical perspectives outlined in the previous section, the available empirical evidence focuses mostly on the macropolitics of unilateral power. Much of this research studies the relationship between structural features of the political environment and patterns of unilateral activity, particularly in examining how Congress and the courts constrain presidents' exercise of unilateral powers. Rather than studying presidents' decisions to advance specific policies via unilateral directives, this research area mostly examines the production of unilateral directives at the annual or biennial level, and occasionally at the monthly level (see Table 1). We return to the impacts of these modeling decisions in Section 5.
Taken on their own terms, empirical studies offer a relatively mixed view of how the president's political relationship with Congress interacts with unilateral activity. Accounting for factors such as divided government and the share of congressional seats held by the president's party, some studies find that presidents issue unilateral directives at greater rates during periods of interbranch disagreement (e.g., Chiou & Rothenberg 2017, Howell 2003, Lowande 2014, while others show that interbranch conflict either has no relationship with unilateral activity or is associated with increased use of unilateral powers (Deering & Maltzman 1999, Fine & Warber 2012, Mayer 2001, Thrower 2017b, Williams 2019. Some research argues that the relationship between interbranch conflict and presidential unilateralism is more conditional in nature. For example, Bolton & Thrower (2016) report that divided government is associated with increased unilateral activity for the first half of the twentieth century, but less in the latter half, which is attributed to the increased capacity of Congress to constrain the president through statutory means after 1947. Many studies argue that there is heterogeneity in the effects of particular variables based on subclassifications of unilateral action. For example, Fine & Warber (2012) argue that the relationship between interbranch conflict and unilateral action may vary across measures of conflict (i.e., divided government versus preference-based disagreement) and based on the substantive content of the directive. The findings are similarly mixed about the association between unilateral activity and the distribution of preferences within Congress. Assuming a uniform distribution of status quo policies at the start of each term, pivot-based theories posit that presidents make greater use of unilateral power as the gridlock interval increases in width. Scholars have studied this prediction using a variety of measures and produced inconsistent results. While Howell (2003) finds that smaller and more internally divided majorities are associated with more frequent use of unilateral powers, other research finds that the relationship is inconsistent across measures and chambers of Congress (Belco & Rottinghaus 2017). Chiou & Rothenberg (2017) further report that presidents issue more executive orders as the gridlock zone expands on the president's side of the median, but increased gridlock in the region of the policy space opposite the president's side is not associated with greater unilateral activity. As the authors explain, this finding weighs against the claim that presidents exercise unilateral power to assert their policy preferences despite a hostile Congress; it implies instead that they do so with the tacit support of the majority party in Congress.
By comparison, little empirical research examines how judicial review affects unilateral activity. In a notable exception, Thrower (2017a) suggests that presidents issue executive orders at greater rates as their ideological distance from the courts increases. To the extent that presidents would anticipate judicial objections to their directives, this finding runs counter to the expectation that presidents temper their unilateral ambitions as ideological conflict with the courts increases. On the unilateral directives that are issued, moreover, the courts overwhelmingly side with the president. According to Howell's (2003, ch. 6) survey of judicial responses to executive orders, 83% of the executive orders challenged in federal court between 1942 and 1998 were ultimately upheld. Measurement is a key empirical challenge for evaluating how the courts constrain presidential action; specifically, courts are posited to overturn a unilateral directive if the president exceeded the discretion permitted in the existing status quo. Absent reliable measures of discretion, however, these predictions are difficult to assess.
The micropolitics of unilateral action based on separation-of-powers theories are most clearly examined by Lowande (2021). Rather than examine aggregate patterns of unilateral activity, Lowande generates issue-specific estimates of status quo policies and tests whether presidents use unilateral action to modify status quo policies in ways consistent with theoretical predictions (Chiou & Rothenberg 2017). Though this approach more closely conforms to the parameters of the theoretical models summarized above, Lowande (2021) finds no support for separation-ofpowers theories of unilateral action; instead, the data indicate that presidents routinely change status quo policies that theories predict they should not. These results suggest that separation-ofpowers constraints are weaker than posited by the models, that the available data and measures are simply inadequate to test the theories, or both.
Overall, then, existing scholarship provides somewhat limited support for theories of unilateral action that emphasize institutional conflict and the separation of powers. It is not immediately clear, however, whether these inconsistencies reflect the inadequacies of this theoretical perspective or limitations of the chosen empirical strategies. We return to this latter possibility in Section 5.

Public Opinion and Accountability
Empirical research has generally taken two approaches to the study of public opinion and presidential unilateralism. One line of research on the macropolitics of unilateral action includes presidential approval ratings as predictors of aggregate numbers of unilateral directives. The primary theoretical expectation from much of this research is that less popular presidents should exercise unilateral powers at greater rates. Unpopular presidents are thought to secure less legislative success in Congress and thus must turn to unilateral directives to realize their policy goals. The evidence regarding this expectation is quite mixed. Several studies have found that, indeed, decreases in presidential popularity are associated with increased numbers of unilateral directives (Deering & Maltzman 1999, Mayer 2001. Others, however, find no evidence of a relationship between approval and unilateral action (Krause & Cohen 1997, Lowande 2014, while Fine & Warber (2012) report inconsistent relationships across model specifications and types of presidential directive.
A second, and more recent, line of scholarship evaluates the micropolitics of unilateral power and reverses the hypothesized relationship by studying public reactions to specific instances of its use (e.g., Christenson & Kriner 2020; Lowande & Gray 2017; Reeves & Rogowski 2015. This scholarship is motivated by the claim that "public opinion. . .serves as the primary check on the unilateral executive" (Christenson & Kriner 2020, p. 8) and draws from public opinion surveys and survey experiments. Some of the findings from this literature exhibit consensus while others are inconsistent. Several studies show that individuals are less likely to approve of the president following the use of unilateral power (e.g., Christenson & Kriner 2020; Reeves & Rogowski 2016), yet these same studies disagree about whether the public exhibits a preference for legislation compared with executive action. According to Christenson & Kriner (2020), the public does not respond to the means by which presidents achieve their policy goals, while Reeves & Rogowski (2018) show that the public imposes a penalty on presidents for the use of unilateral power rather than legislation. Overall, these results are generally consistent with research showing that vetoes reduce presidential approval (Groseclose & McCarty 2001), yet the findings on unilateral power implicate different mechanisms. While Groseclose & McCarty (2001) argue that vetoes signal that the president is out of step with the public's policy preferences, scholarship on unilateral power presents somewhat competing arguments that the public opposes the use of unilateral action due to their core democratic values (Reeves & Rogowski 2016) or in response to mobilization by political elites (Christenson & Kriner 2020).
Christenson & Kriner (2020) seek to unite these two strands of research on the macro-and micropolitics of unilateral action by clarifying the temporal relationship between presidential unilateralism and public approval. Using monthly data and time-series techniques, the authors show that increases in approval ratings predict increases in unilateral activity, but unilateral activity does not predict presidential approval ratings. The results are surprising in that they are inconsistent with the theoretical arguments and empirical findings from both lines of research summarized above. However, the authors explain the findings by arguing that greater popularity reduces the incentives for the president's political opponents to challenge unilateral directives and prevents popular presidents from suffering political or electoral penalties from the use of unilateral power.
Finally, a nascent literature explores issue politics in the context of unilateral action. An analysis of voters' responses to presidential unilateral action shows that presidential approval ratings are responsive to the public's level of agreement with the policies presidents have created (Ansolabehere & Rogowski 2020). This research shows that individuals' agreement with the ideological content of unilateral directives is associated with their evaluations of the president who issued them. Other research shows that presidents issue a greater number of executive orders in a given issue area as public opinion on that issue moves in a more liberal direction, when public opinion is aligned with the president's ideological perspective, and if the issue is publicly salient (Rogowski 2019). Together, this research suggests that presidents perceive incentives to issue unilateral directives in ways that correspond to the public's issue preferences. The standards of issue accountability that apply in other contexts may also discipline presidents' exercise of unilateral authority.

ASSESSING THE EVIDENCE
How strong is the evidence for theories of presidential unilateral power? 3 That depends on your reference point. King (1993) provides a useful one, as his critique of the methodological state of presidency research predates most studies we review. King in 1993 made several points worth reviewing: He pointed out that most models included far too many moving parts; that analyses were underpowered; and that, due to presidential studies' historical engagement with practitioners, claims intended to influence presidential practice exceeded the credibility of the evidence in support of them.
The strength of the evidence, of course, also depends on the typical approach adopted by these studies. We examined this by coding features of research design and data for the 90 books and articles applicable to this review. As Table 1 reports, the modal research design for empirical, noncase-study research is a time-series regression of a count of presidential documents on a vector of political covariates. Though there is variation in the theories tested and data employed by these studies, we think much can be learned by examining their common features.

Research Design
One of King's (1993, p. 403) central points was that "the common practice of using the president as the unit of analysis for causal inferences is extremely unlikely to yield reliable empirical conclusions." Since then, the "n = 1" problem has been widely recognized and, as Table 1 indicates, few studies have fallen into this basic methodological trap. But suppose the typical analysis from the following quarter century included only variables measured without error and plausible identification assumptions. 4 By King's standard, are contemporary research designs any more reliable? We do not think so.
A simple way to assess this is to use simulations to replicate King's power analysis and compare it to the modal research study in our review. Most strikingly, King's analysis (replicated in Figure 1a) showed that any research using the president as the unit of analysis was unlikely to 3 Theories can provide a narrative or heuristic that (sporadically) describes important political events, or they can explicate causal mechanisms that are fundamentally unobservable (Paine & Tyson (2020). Instead, we take as given that the purpose of this family of theories is to organize observed political behavior. 4 Questions of causal identification have largely escaped attention in research on unilateral presidential power. For that reason, we simply note this as a general limitation. However, it is clear that studies in this area share the same basic issues and potential strategies related to causal inference as studies of legislative productivity (Clarke et al. 2018).

Figure 1
Contemporary research designs have elevated false-negative rates. Panel a plots the proportion of simulations that returned statistically significant estimates of a simulated effect, based on assumptions outlined by King (1993)-most importantly, bivariate models with a dichotomous independent and dependent variable. Panel b plots the proportion of simulations that returned statistically significant estimates of a reported effect in three studies with differing units of analysis, based on simulations generated with replication data and reduced-form models reported by study authors. Abbreviations: Pr, proportion; SD, standard deviation. recover any but very large effects-or that otherwise, scholars would have to wait hundreds of years and presidential terms to come away with reliable estimates. This analysis was conservativeit included only dichotomous predictor and outcome variables, and set aside the conventional practice of including all potential confounders in a single model. To assess progress, we obtained replication data for three published studies leveraging three different units of analysis: directive, year, and Congress. 5 The dependent variables are a dichotomous attribute of a directive, a count of orders in a year, and a count of orders within a Congress, respectively. We then repeat the same exercise as in Figure 1a, simulating each dependent variable using the observed covariates and the reduced-form models of each study, which we assume to be correct. As Figure 1b suggests, within the range of "massive effects" [defined by King (1993) as 0.1-0.2 standard deviation], these studies are about as well powered as the psychobiographical research King was addressing. Put simply, the standard research design in this area of study is (still) very unlikely to produce reliable estimates of the association between unilateral action and variables of theoretical interest. Part of the problem is that scholars typically strain these models with additional estimated parameters-intercept shifts like presidential fixed effects and what can be long lists of control covariates. But the issue persists after excluding these controls. Comparative approaches can improve this outlook, but the small number of additional cases typically considered (e.g., all states in Latin America), along with necessary modeling adjustments for panel data, means they do not overcome this limitation. Unilateral action is simply too noisy, and there are too few cases. 6 This issue precedes questions of causal inference. And, to be sure, this issue is not specific to studies of unilateral action. Presidency research presents some inherent challenges that are difficult to overcome with the standard methodological toolkit associated with causal empiricism in the social sciences, so the strength of the evidence for unilateral action theories depends on how well they organize patterns in data. This presumes the patterns are reliably estimated. If this is not possible given a reasonable set of assumptions, then systematic empirical evidence for or against the theory cannot accumulate.

Measurement
What counts as unilateral action? The analysis above presumes the key outcome-unilateral action-is measured without error. But unlike the decisions of Congress or the Supreme Court, a president's policy decisions are not neatly cataloged. The most obvious way to measure unilateral action is to collect presidential directives. Though unilateral powers are expressed through a variety of directives-for instance, executive orders, memoranda, national security directives, proclamations, and international agreements-existing literature focuses overwhelmingly on executive orders alone (e.g., Bolton & Thrower 2016;Chiou & Rothenberg 2017;Howell 2003;Krause & Cohen 1997, 2000Mayer 2001;Warber 2006). Because presidents can exercise unilateral powers through several different tools, any single directive type will not provide a comprehensive summary of the degree of presidential unilateralism. Moreover, some of them are classified, and many more are never published. Worse still for measurement, some unilateral policy initiatives have no presidential directive attached to them-they are informal orders carried out by administrators.
To address these issues, studies have simultaneously expanded the population of directives under consideration and innovated means of pruning to the appropriate sample. Research began with the population of executive orders but eventually included proclamations (e.g., Rottinghaus & Maier 2007) and memoranda (e.g., Lowande 2014)-first separately but later simultaneously (e.g., Kaufman & Rogowski 2017, Williams 2019. As scholars commonly recognize (e.g., Dodds 2013, Howell 2003, Mayer 2001, Warber 2006, not all unilateral actions are equally important or consequential. Like studies of legislative productivity (e.g., Mayhew 1991), studies of presidential actions confront the challenge of distinguishing significant unilateral actions from more routine and administrative directives. For instance, Howell (2003) distinguishes significant executive orders as those that received coverage in national media outlets such as the New York Times, while Chiou & Rothenberg (2017) use an item-response model to estimate executive order significance using a wider range of media outlets and contextual covariates. In contrast, Warber (2006) characterizes executive orders' significance based on hand-coding whether their content is "policy," "routine," or "symbolic," and Bolton & Thrower (2016) focus on "nonceremonial" executive orders. These efforts represent important advances in reducing measurement error.
Nonetheless, these approaches have limits. Most obviously, what constitutes unilateral action varies from study to study, which renders research difficult to compare and slows the accumulation of knowledge. In addition, the reliance on media coverage likely introduces time-dependent error. As media coverage of the presidency has changed over time, and as the nature of media itself has evolved, front-page newspaper coverage of an executive order becomes a noisier measure of its significance. In addition, because some unilateral tools receive greater press coverage than others, it is not clear that this measurement strategy is equally effective for classifying significance across ate to simply dismiss the prospect for improvement, or even claim that presidency research is much worse than other areas of the social sciences.
www.annualreviews.org • Presidential Unilateral Power unilateral tools. Moreover, expanding the study of unilateral action across tools and time introduces practical concerns about the feasibility of manually coding the substantive content of each action.
Relatedly, while smaller bodies of research have studied proclamations (e.g., Cooper 1986, Rottinghaus & Maier 2007, international agreements (e.g., Krutz & Peake 2006, Martin 2005, and national security directives (e.g., Gordon 2007), existing scholarship largely overlooks potential interdependencies between unilateral tools. This omission is particularly important if presidents strategically use unilateral tools based on their expectations about the scrutiny they may attract from Congress and other actors. Furthermore, the president's unilateral toolkit has expanded over time, and the interpretation of particular unilateral tools has evolved along with it. The lack of attention to the range of unilateral tools that presidents utilize and the potential complementarities between them raises the possibility that the findings reviewed in the previous section are either incorrect or misleading.
In addition, for the purposes of most analyses (again, see Table 1), unilateral action is typically measured in the aggregate-as a generic indicator of executive productivity over some arbitrary unit of time. This not only contributes to the power and variability issues noted in the previous section but also creates distance between the theory and data, which opens up potential ecological inference problems. Put differently, existing measures of unilateral action do not characterize the quantities of interest generated from theory. Aggregate productivity is conveniently measured, but benchmark models from Howell (2003) and Chiou & Rothenberg (2017) make predictions about which status quo policies presidents choose to amend through unilateral action and the magnitude of the policy shifts that presidents can induce.
However, virtually every existing empirical study of unilateral power focuses on the frequency of action rather than the policy shifts induced by unilateral action. [Indeed, this limitation is not specific to research on unilateral power; pivot-based theories of the legislative process (e.g., Krehbiel 1998) present similar empirical challenges.] While this body of research has examined variation in executive productivity, itself an important outcome of interest, much less is known about the extent to which unilateral action enables presidents to achieve specific policy outcomes that would otherwise elude them. Existing work is essentially silent on seemingly straightforward empirical questions like "how often do presidential directives lead to policy change?" Similar issues pervade the measurement of proposed determinants of unilateral action. Typical predictors include divided government, strength of the majority party, policy disagreement between the president and other actors, economic indicators such as GDP growth or unemployment, some within-administration periodization (e.g., first half of term), and war. These aggregate measures are straightforward to incorporate in the typical research design, but they have notable limitations. First and foremost, they are not well connected to the underlying theories or spatial models. For example, the justification for expecting more or less unilateral action during the first half of a presidential term involves an argument about some immeasurable latent variable like political capital, which is connected to the theory by further assuming that it exogenously shifts the distribution of preferences.
Second, the basic limitations of examining an exhaustive set of variables have been mostly ignored. Periodizations like presidential fixed effects reduce all other effect estimates to withinadministration variation-further reducing the number of empirically relevant cases. In addition, many of these predictors themselves covary, raising concerns about unbiased estimation-in addition to the basic inferential problem that nature did not produce a sufficient set of cases for the kind of multidimensional comparisons that researchers' hypotheses demand. If the political context implied by these analyses rarely (or never) occurs, it is worth considering whether such exhaustive theorizing is fruitful.

Generalizability
Empirical research on unilateral action exhibits two kinds of generalizability problems. In studies of both institutional constraints and (informal) public constraints on unilateral action, the vast majority of scholarship focuses on the post-World War II era (for important exceptions, see Bolton & Thrower 2016, Dodds 2013, Williams 2019. While the emphasis on the modern era may be reasonable given Moe & Howell's (1999) argument regarding the centrality of unilateral action to the modern presidency, the use of unilateral directives to craft important new policies is not limited to modern presidents. Indeed, presidents since George Washington have used unilateral tools to enact politically controversial policies, including Washington's Neutrality Proclamation, Jackson's Nullification Proclamation, and Lincoln's Emancipation Proclamation. In the nation's first hundred years, presidents frequently used unilateral directives to acquire public lands, manage Indian affairs, and address personnel issues within the bureaucracy. Dodds (2013, p. 107) further outlines a "surprisingly large" set of nineteenth-century unilateral actions that were "surprisingly important" and addressed issues including domestic unrest, Mormon settlement, Reconstruction, private participation in foreign conflicts, and labor issues.
Studies focused specifically on the modern era miss the opportunity to examine the politics of unilateral action during an era in which public expectations of presidents differed markedly from the present and institutional arrangements were considerably different. While Bolton & Thrower (2016) leverage institutional changes with respect to Congress in studying unilateral action over the twentieth century, history offers even more variation in theoretically relevant institutional and electoral arrangements the further back one looks. Moreover, incorporating the nineteenth century into scholarship on unilateral action can shed light on how the use of unilateral powers coincided with development of the American presidency (see Williams 2019); therefore, understanding change and continuity in the predictors of unilateral action could be especially revealing.
Many of the inferential problems we discuss do not apply to recent studies on unilateral action, public opinion, and accountability because they typically rely on survey experiments. However, this approach presents its own generalizability concerns. Experiments are historical artifacts once completed. Most scholarship acknowledges the possibility that any causal effect estimate may be bound by time, the current president in office, or the particular issue covered by the survey.
Surveys provide a stylized simulation of how the public engages with presidential unilateralism, but it is unknown whether this setup is applicable to how the typical person learns about or responds to policy making. This problem is not unique to studies of unilateral action, but it is uniquely important in this context. Suppose, for example, there was a consensus supported by surveys that issuing an executive order was associated with a significant decrease (relative to legislation) in public support for the president. To understand whether presidents actually paid this penalty, we would have to understand how media sources typically report instances of unilateralism and how this information reaches voters. Media reporting that is routinely unclear about the source of policy change could reduce or eliminate the penalty observed in the context of surveys. Thus, if the ultimate goal of this line of research is to understand how public reactions shape presidents' strategic incentives, this is a missing link in the causal chain.

CONCLUSIONS AND RECOMMENDATIONS
More than 30 years ago, Cooper (1986, p. 255) noted that "[e]xecutive orders and proclamations are very important but little understood mechanisms of governance." Few would disagree today, and scholarly attentiveness to presidential unilateralism has increased markedly in the last several decades. Important limitations and opportunities for development remain. We conclude, therefore, with four recommendations.

Collectively Define and Measure Presidential Unilateralism
The basic measurement problems identified in Section 5.2 have a relatively straightforward solution: make defining and measuring presidential unilateralism a collective enterprise with an open repository. By this, we do not mean merely collecting presidential documents for searching and batch downloading. 7 We mean that contributors define and code instances of unilateral action, along with a set of attributes of theoretical interest, with updates as the actions occur. This addresses several related issues. It renders future empirical studies more comparable. It reduces the cost of innovation for each successive study. It creates a clear benchmark for what kind of political behavior scholars are attempting to explain.
The Correlates of War Project (https://correlatesofwar.org/) provides a useful analogue. Since the early 1960s, it has unified empirical research on the causes of interstate conflict. It has been ruthlessly critiqued, updated, and revised. Nothing comparable exists for presidential unilateralism or the presidency; each successive study collects its own dependent variable, with readers left to judge its quality based on limited information and ad hoc standards.
To be clear, we do not think a unified source of this kind would settle, once and for all, the important conceptual questions raised earlier in the review. The point of a centralized measure is to provide a standard-any standard-against which to clarify the contributions and judge the quality of future work. For example, dozens of quantitative studies we reviewed report an estimate of the effect of divided government on unilateral action. They vary dramatically. It is not clear how much of this variation is due to differences in inclusion criteria, outright errors in data sources, or something with substantive implications for theories of presidential unilateralism.

Model Unilateral Action at the Directive or Policy Level
We recommend that empirical research focus on organizing variation in unilateralism at the directive or policy level. We contrast this with aggregate analyses of productivity, which we examined in Section 5.1 and which make up the plurality of research (41%). This recommendation addresses two related issues. First, it more closely aligns empirical models with theories-jettisoning the additional assumptions otherwise necessary to get from theory to testing, and forcing the theories themselves to be more relevant to observed phenomena.
Second, all else equal, it will increase the reliability of the stylized facts reported by any statistical analysis. As we show, the modal research design is (still) vulnerable to high false-negative rates for reasonable effect sizes. Across scientific disciplines, this pattern-together with researcher career incentives-has been shown to contribute to adverse selection of false positives in published research. Adjustments of research practices like these are simple, but they can often separate reliable knowledge claims from stories told around statistical noise.
We are also cognizant that this recommendation may change the types of questions that researchers ask of their data. It invites researchers to pursue different implications of the underlying theory-explaining attributes of presidential actions (e.g., the ideological content or effectiveness of a presidential initiative), instances of actions in a set of policy areas, or cases of action in subgovernments (e.g., states, agencies). This does not mean that hypotheses of the variety "divided (unified) government leads to more (less) action" should not be considered. Multilevel models, after all, are designed to simultaneously estimate relationships like these, which operate with different units of analysis. More generally, King's (1993) conclusion bears repeating: Because this enterprise is important, we should insist that research on presidents fairly assess its uncertainty and adopt standard practices of scientific inquiry.

Investigate Other Quantities Implied by Theoretical Models
Theories of unilateral action do not simply characterize the relationship between the composition of Congress and unilateral directives. They include a number of other parameters that are related to questions that are (or should be) central in the study of unilateral power. What, for example, is the policy impact of unilateral action? Pivot-based models provide a theoretical framework for answering this question. If we can measure the inherited status quo, then, in principle, we can locate the ideological position of the new status quo achieved through unilateral action. The difference between these locations provides an assessment of the degree of policy change achieved through presidential unilateralism.
In many ways, this quantity would be a more theoretically informed means of evaluating the theory itself. Efforts to distinguish significant executive directives are fundamentally concerned with identifying directives that created new and substantively significant policies relative to the status quo. Yet, as we discussed above, the theoretical origins of existing approaches for measuring directive significance are relatively unclear. And from a research design perspective, with estimates of policy change in hand, these data would easily facilitate directive-and policy-level analysis.
Likewise, as we also noted above, empirical understanding of how the courts affect unilateral policy making has been hindered by the absence of measures of discretion. Yet reliable estimates of this quantity stand to facilitate important theoretical and empirical advances in the study of unilateral power. Recent research by , for instance, generates agency-level estimates of discretion based on appropriations reports. Generalizing this approach to measure discretion at the legislation or issue level may be a promising avenue for making this kind of progress.

Contextualize Policy Implications
We recommend that scholars pay greater attention to the policy implications of presidential directives. To what extent, for example, do presidential directives meaningfully change the status quo, and when are they window dressing? Presidents, particularly modern presidents, may have incentives to appear to take action on or prioritize a particular policy goal. Yet the mere fact that a president issues a directive does not imply it produces substantive consequences. Identifying these policy consequences, and comparing them to the consequences of legislative enactment, would be helpful for understanding the president's direct policy impact.
It would also be helpful to catalog the conditions under which presidential directives are implemented. Irrespective of their policy or ideological content, for instance, when are they implemented by executive branch officials and for how long? Addressing these questions is an important opportunity to link the unilateral presidency with the administrative presidency.
Finally, how durable are the policy impacts of unilateral action, and how does their durability compare with that of legislation? A key stylized fact about unilateral directives is that they can be revoked by subsequent presidents with relative ease, while revoking or rescinding legislation is believed to be substantially more difficult. Is this true? Addressing this question can shed light on the strategic calculus employed by presidents. Thrower (2017b) makes important progress on www.annualreviews.org • Presidential Unilateral Power this question by studying the revocation of executive orders. Subsequent work can build upon it by expanding the universe of relevant cases and examining whether a given policy is likely to persist for a longer period of time depending on the means by which it is implemented. This research could provide new insight about the capacity of presidents today to tie the hands of their successors.

DISCLOSURE STATEMENT
The authors are not aware of any affiliations, memberships, funding, or financial holdings that might be perceived as affecting the objectivity of this review.