Delegation or Unilateral Action?

Delegation or Unilateral Action? Abstract Unilateral presidential actions often face implementation problems in the executive branch. I argue these actions are better studied as delegation. I model the conditions under which a president is likely to delegate and provide discretion to subordinates either insulated or uninsulated from their control. I find legislators benefit from agency discretion when presidents pursue policymaking in the executive branch. The threat of legislative sanction induces agents to deviate from presidential priorities, and inter-branch disagreement increases bureaucratic non-compliance in insulated agencies. Nonetheless, in equilibrium, the president is more likely to delegate to insulated agents. Ultimately, the model demonstrates how the politics of direct action are influenced by the need for bureaucratic cooperation. Case studies on US presidential directives mandating public funding of gun violence research and security reforms at government facilities illustrate key features of the model. 1. Introduction In separation of powers systems, executives are often said to act “unilaterally” to enact policies without the explicit consent of legislatures. Most research argues the primary constraints on this power are legislative and judicial veto points (Moe and Howell 1999; Howell 2003; Chiou and Rothenberg 2014; Bolton and Thrower 2016). However, executives—like legislatures—rely on bureaucrats with multiple principals to implement policy. For example, when Harry Truman issued an executive order that seized the property of American steel mills, it delegated the authority to the Secretary of Commerce and gave him substantial discretion to both determine the terms of employment and even return the mills to their owners.1 The Secretary used that discretion to delay action until the order was invalidated (Neustadt 1960). The central premise of this article is that directives like these are more effectively studied instances of delegation, as opposed to policy change “with the stroke of a pen” (Mayer 2001). I study executive policymaking by developing a model of delegation that approximates the implementation of directives. In the model, a political principal (e.g., the president) makes policy “in-house” or delegates to an outside agent. These agents can pay an exogenous penalty for not complying, or choose to implement policy. After selecting policy, outside agents are subject to punishment from another principal (e.g., a congressional median or committee). A key component of the model is the analysis of the implications of agency insulation. The non-compliance penalty relative to the legislative penalty determines the relative insulation of the outside agent. Insulated agents pay more severe legislative penalties (and are thus insulated with respect to the president), whereas uninsulated pay more severe non-compliance penalties. Thus, the model organizes observed features of executive policymaking like bureaucratic non-compliance, agency insulation, and legislative sanctioning in a familiar and simple framework. As a result, it provides a way forward for empirical work by redirecting attention to the content of directives—rather than their issuance. Placed in the context of research on executive unilateralism, I argue that bureaucratic agency is a mechanism by which legislatures secure better policy outcomes when executives “act alone.” The general point that directives are not self-executing has been made by presidency research—much of which suggests that non-compliance and negotiation on the part of bureaucrats is common (e.g., Dickinson 2009; Krause 2009; Krause and Dupay 2009; Rudalevige 2012; Kennedy 2015).2 While studies argue implementation matters, it is less clear how these features should systematically influence outcomes. Accordingly, the theory highlights several broad implications of bureaucratic implementation. First, the model suggests that inter-branch disagreement induces bureaucratic non-compliance. Thus, this disagreement not only limits opportunities to shift the status quo—as Howell (2003) shows—it reduces the probability such orders will be carried out. Second, the theory implies that executives are more likely to delegate when the agent is insulated from potential punishment. Insulated agencies are less compliant by design. This generates uncertainty about when they will use their discretion to implement a policy unfavorable to the principal. Thus, as long as policy development is costly for executives to undertake in-house, delegating to insulated agencies makes the president better off. Counterintuitively, these are precisely the agents most likely to bend to congressional pressure. More generally, the dynamics of agent non-compliance and policy selection demonstrate that “unilateral” policy outcomes can reflect legislator preferences, even in the absence of laws that explicitly overturn directives. 2. Unilateral Action, Agency Problems, and the Executive Branch Howell (2003) writes that “modern presidents often exert power by setting public policy on their own and preventing Congress and the courts—and anyone else for that matter—from doing much about it” (14). This perspective influenced a generation of quantitative research on the presidency that attempts to identify political circumstances that enable the president to act alone. Much of this research focuses on the empirical study of presidential directives.3 More recently, the conceptual focus on what presidents can accomplish alone has informed investigations of the president’s role in the distribution of federal spending (e.g., Kriner and Reeves 2015; Rogowski 2016). But there are several reasons to believe an alternative theoretical base might be appropriate. First, this stylized depiction of policy change does not adequately capture the presidents’ orders. Second, scholars have long recognized that managerial challenges within the executive branch often thwart policy change (e.g., Nathan 1983; Burke 1992). Recent work argues this past emphasis is largely at odds with the implicit assumption of perfect implementation (Dickinson 2009; Krause 2009; Rudalevige 2009). It is important to build on a central point: when Congress passes a law or the president issues an executive order, both rely on bureaucrats for policy change. Absent that recognition, theories risk aggrandizing the capabilities of presidents and downplaying the agency of the bureaucrat. In contrast, my model builds on well-known agency problems in the executive branch. That is, when presidents want policy changed, they must rely on subordinates with agency. This feature of the presidency has been the basis for the study of public management and political influence over policy outcomes. For example, the notion that presidential and bureaucratic preferences often diverge is at the core of most explanations for the politics of appointments (Lewis 2008). Moreover, bureau responsiveness, even after politicization, is not guaranteed (Dickinson and Rudalevige 2004). However, nearly every action the president takes can be considered an act of delegation. A “delegation-all-the-way down” perspective risks returning to an understanding of the president as overwhelmed and ultimately incapable of seriously influencing policy. Therefore, I conceive of the president’s decision to delegate as a dichotomous choice: delegate to “external” agents—those in government corporations, independent agencies, and cabinet departments—or, delegate to actors within what has been called the “presidential” branch—the White House (WH) and Executive Office (EOP).4 Presidents have the capacity to develop policy in a wide variety of issue areas with the resources at their disposal within the WH and EOP. This preserves an essential point, made first by Moe (1985): that within the presidential branch, presidents face fewer collective action problems and lower risks of policy drift. 3. The Model I present a spatial model of delegation within the executive branch with three actors: the president (P), a congressional committee (C), and an external agent (A). The congressional committee may be thought of as the committee with oversight jurisdiction over the corresponding agency.5 In the model, the president attempts to secure a policy outcome (realized by the equation x^=x+ω, where x is the policy selected and ω is a uniformly distributed shock) nearest to their preferred (xP), while avoiding the resource cost (τ) associated with developing policy in the WH and EOP. To avoid that cost, the president may delegate via directive to an external agent ( e∈{0,1}) and provide them with a level of discretion ( d∈[0,R+]).6 The congressional committee looks to obtain a policy closest to its bliss point (xC) through sanctioning ( s∈{0,1}), which imposes an exogenous cost on the agent, θs, if the agency opts in to policymaking. This sanction can be thought of as a law that directly punishes the agent, a rider on an appropriations bill, or the inconvenience of having to give formal testimony on Capitol Hill.7 This leads to two important assumptions. First, Congress only punishes agents for deliberate policymaking action on behalf of the President. That is, agents who do not cooperate with the President are not subject to congressional sanction. This is often the case in practice, as agents who have taken deliberate action under direction of presidents are then subject to hearings and investigation. I discuss several examples later. This also highlights scope conditions of the model, which is confined to understanding bureaucratic response after a president has become the first-mover. Second, I assume that s = 1 when −|x^−xP|>−|x^−xC|. This implies the committee does not know exactly where the policy is, but does know when the president is made better off by agency actions. In this way, the committee is modeled as largely reactionary. This allows me to focus on the relationship between the president and agencies without overcomplicating the model. It is problematic, for example, to assume that congressional committees have perfect knowledge of the policy implementation process. But assuming they have no knowledge seems equally divorced of the basic political phenomenon. Alternatively, granting a unitary C the power of ex post veto is likely inappropriate because collective action problems limit their capacity to respond (Moe and Howell 1999). Here, I assume they know, broadly, whether the policies make the president better off. It is also important to note that the model does not investigate congressional incentives to provide resources to the executive branch; τ is exogenous. I defer discussion of potential implications to the conclusion. The agent, if chosen to make policy, can opt out by refusing to comply with the president’s directive ( v∈{0,1}). If it opts in, it selects a policy conditional on the level of discretion supplied by the president. If the agency opts out of policymaking, x^=ω, and the agency pays a non-compliance cost that can be thought of as an ex post presidential sanction (θv).8 Presidents have a variety of tools to impose costs on non-compliant agencies—including the removal of appointed program managers and, in some cases, the reassignment of agency functions. Moreover, variation in this parameter allows the model to analyze institutional variation among line agencies.9 More specifically, I define the following to aid in describing the results: Definition An “insulated” or “independent” Agency (A) has non-compliance costs less than the cost of congressional sanction ( θv<θs). Conversely, an “uninsulated” Agency (A) has non-compliance costs greater than the cost of congressional sanction ( θv>θs). Scholars have long recognized that some agencies are designed to render them less subject to presidential manipulation (e.g., Moe 1985; Lewis 2003; Selin 2015). I operationalize this in terms of an agency’s vulnerability to presidential punishment—relative to congressional punishment. The effectiveness and availability of punishments are often due to features of agencies that are difficult to change. For example, fixed service terms effectively inhibit the president’s ability to punish agencies by replacing (or leaving vacant) their appointed heads.10 Nonetheless, there are other aspects of agency structure this definition cannot capture (e.g., ex post review of decision-making or reporting requirements). Since I do not model the president’s choice among different external agents, a more complete consideration of these institutional forms is beyond the scope of the present paper—but could be taken up in future research. The utility of the president is given by   UP=−|x^−xP|−(1−e)τ, so that they attempt to minimize the disutility associated with distant policy and the cost of developing policy in-house.11 The committee’s utility is governed solely by the policy outcome, such that   UC=−|x^−xC|. Finally, the Agent’s utility is governed by policy outcomes and the costs associated with congressional sanction ( θs∈[0,R+]) and non-compliance ( θv∈[0,R+]).   UA=−|x^−xA|−sθs−(1−v)θv. Sequence of play: Nature selects a random shock, ω, which is a common prior, uniformly distributed between [−R,R]. A observes ω. The president chooses whether to delegate or develop policy within the WH and EOP ( e∈{0,1}). If e = 0, then the president pays a resource cost, τ, and the EOP selects a policy, x. If e = 1, then the president selects a level of discretion, d. The agent chooses whether or not to comply, v∈{0,1}, after observing ω. If v = 0, then s = 0 and x = 0. If v = 1, the agent selects a policy, x∈[−d,d]. If −|x^−xP|>−|x^−xC|, then s = 1, otherwise, s = 0. Play ends and payoffs are distributed. Player preferences (and all parameters except ω) are common knowledge, and all arbitrary spatial orientations reduce to two cases. Let xm denote the midpoint between xP and xC, a policy outcome that would result in s = 0. Without loss of generality, I assume that xP = 0, xC>0 so that the relevant cases are xm > xA in which, the agent is closer to the president; and xm < xA, when the agent is closer to the committee. I define the following to simplify discussion. Definition A “proximate” agency (A) has an ideal point closer to the President (xA < xm). Conversely, a “distal” agency has an ideal point closer to the committee (xm < xA). so that in the description of the results, I refer to agencies as one of four types:   A∈{insulated,uninsulated}×{proximate, distal}. The model differs from related work in several respects. First, in contrast to Howell (2003), critical pivots in Congress and the Judiciary are absent. Whereas the unilateral politics model outlines conditions that lead to presidential action, this model attempts to explain how the agency of bureaucrats influences final outcomes—after the decision to act alone has been made. Second, in contrast to many models of delegation (Bawn 1995; Epstein and O’Halloran 1999; Volden 2002; Gailmard and Patty 2007), the principal’s key trade-off is not primarily motivated by gains in expertise. In some cases, delegation is attractive because it reduces the cost of policymaking for the president—presumably reserving limited resources for other policy initiatives. A few simple examples underscore the applicability of this setup. Presidents, in pursuit of their goals, are faced with an initial choice of whether to use resources within their immediate domain, or delegate to external agents. In 1976, for instance, Gerald Ford chose to delegate policymaking functions to the Federal Energy Office within the EOP, rather than vest those functions in the newly created Federal Energy Administration.12 If a president delegates, they are faced with the task of formulating the agent’s mandate to make policy, which can be limited or provide latitude. The president could, require that the agent consult with other agencies, as Harry Truman did when he redelegated wartime employment functions to the Department of Labor in 1945.13 Relatedly, the president could simply specify the new policy in great detail. Members of Congress can sanction agents directly. As Barack Obama’s issued a series of immigration-related directives in November 2014, then-Senator Jeff Sessions (R-AL) reminded the applicable departments that “Congress has the power and every right to deny funding for unworthy activities.”14 In another instance, a proposed ban on armor-piercing ammunition resulted in the introduction of legislation to abolish the agency responsible.15 Although the president pays political costs for overturned or impeded policies, bureaucrats endure punishments which vary in severity. 4. Results The model produces several key results relevant to executive-driven policy initiatives. The first is that for insulated agencies, inter-branch conflict influences the likelihood of non-compliance. Specifically, as disagreement between the committee and the president increases, these agents are more likely to opt-out. However, despite that dynamic, when the agent is insulated from presidential punishment, the president is always better off delegating. This is because the guaranteed compliance of uninsulated agents lets presidents know when they will opt-in and pick a policy that makes the president worse off. Thus, the model demonstrates how bureaucratic structure can influence the president’s dependence on agents when engaging in unilateralism. 4.1 When Do Agencies Serve Congress? The solution concept is subgame perfection and the model is solved via backward induction. All proofs appear in the Appendix. Let θs¯>|xA−xm|, the value of the sanction punishment which exceeds the agency’s policy payoff from moving the outcome from its ideal point to the sanction-free point. Working backwards, it is apparent that the optimal policy choice for the bureaucrat (denoted by x*) depends on θs and xA. When the cost the committee imposes is sufficiently low ( θs<θs¯), or the bureaucrat is distal, then the agency’s choice follows the intuitive, well-known result regarding policy selection. Lemma 1 When xA> xm, or θs<θs¯  x*={d if −R<ω≤xA−dxA−ω if xA−d<ω≤xA+d−d if xA+d≤ω<R. For any given random shock, the agent pulls policy as close as possible to its ideal point, given the bounds of its discretion. However, the results diverge from Lemma 1 when sanctioning imposes a sufficient cost. Lemma 2 When θs≥θs¯ and xA < xm  x*={d if −R<ω≤xA−dxA−ω if xA−d<ω≤xm−dxm−ω if xm−d<ω≤xm+d−d if xm+d≤ω<R. When the random shock is between—R and xm−d, the agent shifts policy as close as possible to its ideal point because it cannot select xm. On the other hand, when the shock exceeds xm−d, the Agent shifts policy as close to xA as possible, while avoiding sanctioning by selecting xm−ω at minimum. Figure 1 illustrates this change in bureaucratic behavior. To illustrate the additional policy loss (highlighted in red), I set d=|xA−xm|. 16 Delegation always implies some policy loss for the president. However, the presence of a powerful congressional committee implies an additional cost. In Figure 1, the agent selects the policy d+ω=xm when ω=xA, despite the fact this will move policy away from its ideal point. So under the right conditions, the bureaucrat’s optimal strategy is to appease the congressional committee. Figure 1. View largeDownload slide Committee Influence and Presidential Policy Loss. Figure 1. View largeDownload slide Committee Influence and Presidential Policy Loss. 4.2 Bureaucratic Non-compliance Given x*, I characterize whether the agent will comply with the president’s order, or opt out and allow the state of the world (ω) to obtain. In general, the agent complies if the following inequality is satisfied:   −|x*+ω−xA|−s*θs>−|ω−xA|−θv. The right-hand side of the equation shows the policy that would result if the agency refused to implement the order. This comparison has direct implications on compliance, discretion, and ultimately, the utility of delegation. Proposition 1 (Uninsulated). In any equilibrium, both proximate or distal agencies are perfectly compliant. If θv>θs,v*=1. Insulated, Distal. Agencies comply when the state of the world permits avoidance of or proper policy inducement outweighing congressional sanction. If θv<θs, when θs−θv>|xA−ω|   v*={0 if −R<ω≤xm−R+xm−2xa2R1 if xm−R+xm−2xa2R≤ω<R. when |xA−ω|>θs−θv>0   v*={0 if −R<ω≤θs−θv−R+xm−2xa2R1 if θs−θv−R+xm−2xa2R≤ω<R.Insulated, Proximate. Agency are complies when the state of the world permits avoidance of or proper policy inducement outweighing congressional sanction. If θv<θs, when θs−θv>|xA−ω|   v*={0 if −R<ω≤xm−R−xm21 if xm−R−xm2≤ω<R, when |xA−ω|>θs−θv>0   v*={0 if −R<ω≤θs−θv−R−2xA−θs−θv21 if θs−θv−R−2xA−θs−θv2≤ω<R. Insulated agencies comply with the directive only when the policy shock puts a beneficial policy within reach.17 A beneficial policy is one that either allows the agent to avoid sanctioning, or allows the agent to select a policy that effectively outweighs the cost of congressional sanction. This means—somewhat counterintuitively—that insulated, proximate agencies are less compliant (and receive less discretion) the more powerful the presidential punishment. This also reveals some important points about compliance with presidential directives. First, it’s worth noting that as discretion increases, so does the region of random shocks for which the insulated bureaucrat will opt-in to policymaking.18 This means that discretion increases the probability of compliance with the president’s directive. Figure 2 illustrates these basic implications. An increase in discretion contracts the non-compliance region (in red). Because the agent is insulated from presidential control, they comply when ω allows them to implement xm. For proximate agencies, increases in policy disagreement between the president and the committee increase the probability of non-compliance. In Figure 2, the shift from xm1 to xm2 results in a reduction of the compliance region. Note also, that because this region is partly a function of the “state of the world” as determined by ω, in any equilibrium in which x∈[−d,d]≠[−R,R] and A is insulated, there is some probability of non-compliance. Most importantly, however, since the optimal level of discretion is linked to committee preferences, when the committee is sufficiently powerful, the non-compliance region for insulated agencies is increasing in the distance between the president and the committee. Figure 2. View largeDownload slide Discretion and Non-compliance. Figure 2. View largeDownload slide Discretion and Non-compliance. 4.3 Bureaucratic Discretion Let d* denote the president’s optimal choice of d. Given the results above, the president’s maximization problem takes one of several forms, which suggest some differing predictions based on agency insulation and ideological divergence between president, agency, and committee. Proposition 2 (Uninsulated). For distal agencies, the optimal level of discretion is increasing in policy uncertainty and decreasing in ideological distance. When the agency is proximate and the committee is sufficiently powerful, discretion is increasing in policy uncertainty and decreasing in divergence between the president and the committee,   d*={R−xA if xm<xA or θs<θs¯R−xm if xA<xm or θs≥θs¯.Insulated, Distal. The optimal level of discretion is increasing in policy uncertainty and committee ideological distance, but decreasing in agency distance,   d*=max[0,R+xm−2xA2R] if xm<xA.Insulated, Proximate. If there is a proper policy inducement outweighing congressional sanction, then discretion is decreasing in agent distance and non-compliance cost, and increasing in policy uncertainty and congressional sanction. If there is not a sufficient inducement, discretion is increasing in policy uncertainty and decreasing in committee distance,   d*={R−xm2 if xA<xm and θs−θv>xAmax[0,R−2xA+θs−θv2] if xA<xm and θs−θv<xA. Equilibria with insulated agencies are a function of ideological proximity, policy uncertainty, and the fixed costs imposed by Congress and the president.19 For distal agencies, there is a standard result associated with delegation—that divergence between the principal and agent results in a reduction in discretion. Importantly, however, divergence between president and committee marginally increases discretion because non-compliance motivated by Congress makes the president slightly better off. Distal, insulated agencies sometimes use their discretion to pull policies away from the president. Non-compliance restricts the region for which that occurs. However, importantly, this discretion is essentially flat (zero), alongside the other cases in Figure 3. A similar dynamic is at play in when the committee is sufficiently weak and the agent is proximate—where, the president’s punishment seems to “backfire.” Even proximate agents sometimes pull policy away from the president—a powerful Congress prevents them from doing that, whereas a high-compliance cost encourages it. So in this case, discretion is increasing in θs and decreasing in θv. Figure 3. View largeDownload slide Inter-branch Disagreement, Discretion, and Insulation. Note: All comparative statics assume R>xi; insulated, proximate assumes θs−θv>xA; d*(insulated, distal)>0 when R>2xA−xm. Figure 3. View largeDownload slide Inter-branch Disagreement, Discretion, and Insulation. Note: All comparative statics assume R>xi; insulated, proximate assumes θs−θv>xA; d*(insulated, distal)>0 when R>2xA−xm. As Figure 3 shows, for insulated, proximate agencies, discretion is again a function of ideological divergence between the president and either the agency or the committee. This compounds the dynamic shown in Figure 2. A more distant committee means a higher probability of non-compliance and less discretion—which, in turn, leads to an even higher probability of non-compliance. Overall, it is also apparent that so long as R>xi, uninsulated agencies receive more discretion than insulated ones. I return to these implications in Section 5.20 4.4 Delegation and the Presidential Branch Finally, the model reveals the conditions that must be satisfied for the president to delegate to an external agency. Let e*be the optimal choice of delegation. Under most conditions, there are values of τ that render delegation beneficial. However, the circumstances under which the president does not delegate are confined to uninsulated agencies. I explain the rationale below. Proposition 3 (Uninsulated, Distal). In any equilibrium, the president delegates when the cost of developing policy “in-house” exceeds a critical value decreasing in the distance between the agent and policy uncertainty. If θs<θv and xm < xA,   e*={0if RxA−xA22R>τ1otherwise.Uninsulated, Proximate. The president delegates when the cost of developing policy “in-house” exceeds a critical value decreasing in the distance between the agent, committee, and policy uncertainty. If θs<θv,θs>θs¯, and xA < xm  e*={0 if 4Rxm+2xAxm−3xA2−3xm24R>τ1otherwise.Insulated. In equilibrium, the president delegates. If θs>θv, then e*=1. For distal agencies, this holds if τ>0. For proximate, this holds even if τ = 0. When the agent is insulated from presidential control, the president delegates. Irrespective of policy disagreement, delegation results in gains in expertise and efficiency that cannot be matched within the presidential branch. For all agencies, as the distance between the president and the insulated agent increase, the utility of delegation declines. Thus, the cost of in-house policy development must be sufficiently high to offset the policy loss incurred through delegation and committee influence. For uninsulated agencies, the thresholds for delegation are straightforward.21 For insulated agencies, there are no thresholds. The logic behind this finding is directly tied to the role non-compliance plays in unilateral action. Non-compliance generates the possibility that the state of the world will prevail. This uncertainty prevents the president from isolating the circumstances in which in-house development is more optimal. Uninsulated agencies are perfectly compliant. That allows the president to know when they will use their discretion to select a policy worse than the one generated by a presidential branch that lacks expertise. Absent that guarantee the president delegates to avoid wasting resources needlessly. Thus, under these conditions, agency insulation can increase delegation. The importance of this result should not be understated. It implies that if the president wants to act in an area of policy governed by an insulated delegate, then unilateral action is dependent on bureaucratic agency. For proximate agents, this holds even if the policy is “free” (τ = 0) to produce. 5. Applications Given the substantive motivation of the model, this section details several empirical applications. First, I provide two case studies to demonstrate its usefulness for interpreting instances of unilateral presidential action. In each case, absent the model, unilateral action theory would provide an incomplete explanation for the observed policy outcomes. Second, I describe expectations for more systematic analyses of president-driven policymaking. To recap, I argue the model applies to scenarios in which policy could be developed by either the EOP or line agencies. In other words, I expect that many policies developed in the President (e = 0) will be carried out by bureaucrats outside the EOP. This is analogous to the situation Congress finds itself in (a la Epstein and O’Halloran 1999)—since legislators themselves implement no policy. Thus, according to the model, not delegating means dedicating in-house resources to making a policy—which might then be imposed on an agency. 5.1 Policy Drift, Congressional Influence, and Gun Violence Research In a 2015 interview with the BBC, President Obama referred to the lack of policy movement on firearm regulation as his “biggest frustration” in office.22 Editorials and commentaries cite congressional gridlock and the influence of the National Rifle Association (NRA) for this outcome. But this leaves unanswered a key question: what prevented the President from “going it alone”—as the administration had done with other important issues like carbon capture technology and deferred action for undocumented immigrants (DACA)? The model sheds light on this question by highlighting the incentives of bureaucrats faced with implementing presidential orders. As the political struggle to fund gun violence research shows, bureaucrats with discretion often select policies to appease multiple principles—especially when those principals can impose substantial costs on implementors. Despite a presidential directive ordering the Centers for Disease Control (CDC) and National Institutes of Health (NIH) to conduct gun violence research, both agencies have yet to take major steps—due in large part to the political threat of Congress. In the United States, violence associated with firearms has been referred to as a public health epidemic that eclipses yearly causes of death like influenza.23 However, for nearly 20 years, the federal government has funded no research into the causes of that violence—in part, due to an annual limitation rider in the budgets of the CDC and NIH. The so-called “Dickey Amendment” prohibits agency funding from being used to “advocate or promote gun control.”24 Although the NRA itself does not interpret the rider as a ban on gun violence research, the rider (which first appeared in 1996) has had a clear chilling effect on discretionary funding priorities in both agencies.25 After a series of legislative defeats and the Sandy Hook Elementary School shooting, President Obama ordered the CDC and NIH to “conduct or sponsor research into the causes of gun violence and the ways to prevent it.”26 Although the order left the level of funding up to the discretion of the agency, the President’s preferences were clear: he followed up the order with a request from Congress for $10 million in funding for gun violence research. That year, the CDC had more than $30 million allotted for intentional injury prevention.27 Both agencies had a clear presidential directive, as well as the resources and discretion to act.28 Neither agency did precisely what the president asked for, but differences in the agencies’ responses are instructive. The CDC funded no research, but instead conducted a short internal study that was released 2 years later.29 The NIH, on the other hand, released a formal call for research proposals that referenced the president’s directive. However, it funded only two projects—totaling just $1.62 million. Although both agencies complied with the vague language of the memorandum, the NIH moved considerably closer to the president’s preferred policy by funding independent research. As the theory implies, this variation in policy drift is attributable to the relative cost Congress can impose on bureaucrats implementing orders. Although both are agencies of the Department of Health and Human Services, the CDC’s budget is a series of line-item appropriations for specific programs. The NIH, on the other hand, receives block funding by institute—insulating its research priorities.30 Moreover, congressional opposition to the President’s order was clear: support for the limitation rider and opposition to new firearm regulation remained strong. Thus, the CDC was comparatively vulnerable to congressional sanction, and both agencies had incentives to avoid significant steps to publicly fund gun violence research. The case highlights important features of the model. The president can direct agencies to use their discretion to pursue a policy agenda. The existing perspective on unilateral action would label the president’s directive as an instance of policy change. But that obscures important variation in the final outcome. In this instance, the federal government continues to fund very little research on gun violence because agencies with discretion operate among forces other than presidential decree. 5.2 Discretion, Non-compliance, and the Implementation of HSPD-12 The September 11, 2001, attacks ushered in dramatic changes in national security policy and the structure of the administrative state. Among these were efforts to enhance security protocols for government facilities deemed potential targets. Homeland Security Presidential Directive-12 (HSPD-12) initiated one such program in August 2004.31 The directive mandated the use of “SmartCard” identification technology for all federal employees and contractors in order to standardize existing, disparate programs, and limit facilities’ vulnerability to intrusion.32 Despite this agreeable aim, HSPD-12 was subject to widespread, systemic non-compliance by agencies: as of the most recent quarterly report, it has still yet to be fully implemented, over 10 years later.33 Its implementation illustrates an important dynamic in the model: the impact of discretion on compliance. The model implies that agencies “opting out” of policymaking results from disagreement, insufficient discretion to deal with unforeseen exigencies in the implementation process, and the threat of sanction from Congress. HSPD-12 demonstrates the first two, as the text of the directive provides specific, technical requirements for government-wide identification cards.34 By design, the identification standard dramatically constrained the discretion of agencies. Prior to HSPD-12, agencies operated their own security programs. This variation provided agencies with the flexibility to determine the appropriate level of intrusion protection based on the facilities they operated. The directive meant that all agencies (regardless of functions or security vulnerabilities) would have to implement a high standard—so high, in fact, the technology needed to do so had never been produced on the scale required.35 Thus, agencies were provided little-to-no discretion to implement a policy for which the technical uncertainty was substantial. Moreover, the President’s policy was substantially more stringent than the one preferred by many agencies. The standard for acquiring the necessary access card requires an intensive screening process that includes criminal and defense background checks conducted by multiple agencies. According to the model, HPSD-12 presents precisely the scenario in which non-compliance via “opting out” of the policymaking process should occur. Compliance problems immediately followed the issuance of the general standard.36 The directive required agencies put a program in place to achieve compliance within 4 months and required full compliance within 8 months. Few agencies successfully met the deadline. This was not merely a function of the technical challenges of implementation. Non-compliance was concentrated in agencies whose functions posed few security risks or did comparatively little work through contractors and would likely have preferred to meet less stringent fraud protection standards.37 Moreover, the full extent of that non-compliance is unclear, as agency self-reporting is infrequent and sometimes includes errors. Implementation at the National Aeronautics and Space Administration (NASA) highlights critical features of this non-compliance. NASA began developing its own SmartCard system prior to HSPD-12, so that compliance involved sunk research and development costs.38 In 2004, implementation was assigned as an additional staff function of the Facilities and Administrative Services Division, contrary to the instructions of the directive—which required the development of a dedicated program. For 3 years, the agency made no significant progress.39 After years of inaction at NASA, the president dedicated personnel in the WH and Office of Management and Budget to help implement HSPD-12. At NASA, WH officials met with the Deputy Administrator and a newly appointed Program Director in 2007 to ensure the order was implemented. Full compliance was finally achieved 13 months later. In sum, HSPD-12 demonstrates of the basic dynamics of non-compliance in the unilateral presidency. Non-compliance occurred in the area national security policy—often considered the president’s domain—even in relative absence of congressional opposition. In this case, the security of government facilities was dramatically delayed because agencies—driven by their own preferences—chose to not implement the president’s order. 5.3 Implications for Empirical Research Although the model is consistent with the cases above, it points to the need for analyses of the content of presidential directives like executive orders, memoranda, and proclamations. For this reason, it is important to highlight several empirical implications of the model, which all involve observable behavior that can be measured. Although I have mentioned several potentially interesting results, five merit emphasis: Prediction 1. All else equal, presidential directives should delegate more authority to agencies insulated from presidential control (Proposition 3). Institutional variation among bureaucratic agencies—and its impact on presidential control—has been the subject of a large body of research (e.g., Moe 1989; Lewis 2003; Selin 2015). Much of this research has sought to determine how politically responsive agencies are to elected institutions. Here again, presidential directives offer an opportunity to validate the intuition of the model. Prediction 2. All else equal, presidential directives should provide less discretion to agencies insulated from presidential control (Proposition 2). Nonetheless, uninsulated bureaus are generally more reliable agent of the president. Directives—like legislation—often included procedural limitations designed to limit the discretion of agents. Prediction 3. All else equal, uninsulated agencies should be more compliant with presidential directives (Proposition 1). Although it remains difficult to track, recent studies have attempted to measure non-compliance through rule promulgation (e.g., Kennedy 2015; Bertelli and Doherty 2017). Prediction 4. All else equal, for insulated agencies, as the policy disagreement between the President and the relevant oversight committee increases, bureaucratic non-compliance with presidential directives should increase (Proposition 1). The prediction highlights an important, secondary consequence of disagreement between Congress and the executive. Whereas Howell (2003) shows disagreement tends to decrease the frequency of unilateral action, this suggests the same decreases its effectiveness—since insulated bureaucrats become less likely to implement such action. Prediction 5. All else equal, presidential directives should delegate more often as the cost of developing policy in-house increases (Proposition 3). Although operationalizing the cost of policy development in the presidential branch presents a clear measurement challenge, several contributing variables seem apparent. For example, the resource cost might be thought of as the relative proportion of the president’s institutional apparatus that the policy requires to be produced. For that reason, one might look to WHO/EOP budgetary authorizations and staff levels. Additionally, since administrations tend to accumulate policy objectives over time, the resource cost might be extrapolated from the president’s time in office. 6. Conclusion I sought to address two general limitations of theories of executive unilateralism. First, these theories may say more about the frequency of attempts at unilateral action, rather than their efficacy. We observe the issuance of a directive, but theories often cannot explain whether (and why) that order will translate into a preferred outcome. Second, taken to its logical conclusion—the theory aggrandizes the power of chief executives. Presidential policymaking is remarkably contingent, both as a result of constitutional checks and the inherent challenge of wielding the administrative state. The model suggests that unilateral action is dependent on bureaucratic agency in several key ways. Put simply, bureaucrats enable legislators to influence final outcomes. This result provides the basis for future empirical work looking to study a variety of other political phenomena: bureaucratic non-compliance, congressional oversight, and resources in the presidential branch. Moreover, the model reveals an important secondary consequence of agency insulation—that the threat of non-compliance effectively prevents the president from determining when in-house policymaking would be optimal. Counterintuitively, this means directives targeted to insulated agencies should delegate more authority. This leads to a number of directions for future work. As the previous section laid out, the model motivates a variety of possible empirical studies. While the model itself incorporates some important components of presidential efforts to implement policy “unilaterally”—these are worth further study. For instance, the model demonstrates that even a relatively weak legislative actor may meaningfully influence agents implementing presidential policy. But the “in-house” resources of the president are partly a function of congressional preferences. This is because presidential institutionalization (e.g., WHO and EOP resources) is known to be dependent on congressional efforts to build their own expertise and capacity (e.g., Krause 2002; Dickinson and Lebo 2007). In other words, the model may understate legislative influence, since legislators are capable of influencing the cost of policy development for the executive. In addition, future work might analyze the implications of a more complex definition of insulation—since agencies meaningfully vary in ways not captured by the relative capacity of elected institutions to sanction them. Another direction for future study is to incorporate features of delegation by chief executives within the broader framework of separation of powers in order to address the original question empirically evaluated by Howell (2003): when will presidents act alone? Ultimately, the key dependencies revealed by the model likely impact the president’s initial decision to act—and indeed, legislatures’ decisions to veto those actions. Indeed, this notion is more broad than presidential attempts to change policy. Scholarship on policy development in legislatures also tends to ignore the “shadow of implementation”—which surely influences their ability to enact policy. Thus, more generally, integrating this siloed work is both appropriate and promising. For helpful comments and suggestions, I thank Randy Calvert, Justin Fox, Sean Gailmard, Thomas Gray, Will Howell, Jeff Jenkins, George Krause, Dave Lewis, Eric Patashnik, Rachel Potter, and Craig Volden. Previous versions were presented at the 2015 Southern California Law and Social Science Forum, Los Angeles, CA, the 2015 annual meeting of the Midwest Political Science Association, Chicago, IL, USA, and the 2016 annual meeting of the Southern Political Science Association, San Juan, PR, USA. Footnotes 1. The Secretary was permitted to “act through or with the aid of such public or private instrumentalities or persons as he may designate,” to “determine and prescribe terms and conditions of employment under which the plants […] shall be operated,” and to issue regulations. 2. Moreover, this point is acknowledged by Howell (2005): “All presidents, and all politicians, struggle to ensure that those who work below them will faithfully follow orders” (433). 3. For the most recent review of this work, see Mayer (2009). 4. Work by Epstein and O’Halloran (1999) applied the conceptual distinction between policy development “in-house” and external delegation to congressional delegation. Thus, the model’s scope conditions are similar. As I argue in Section 5, the model applies to policies that can be conceivably developed “in-house” and imposed on agencies. That is, I do not assume that the WH/EOP has the staff, resources, and legal authority to carry out policy. By “WH” and “EOP,” I mean the collection of administrative agencies for whom the President submits a yearly budget under the “EOP of the President.” I consider cabinet departments, independent agencies, and government corporations external delegates. 5. This is particularly important, given the broad understanding of sanctioning behavior I put forth. Drafting punitive legislation and holding hearings occur at the committee level, such that assuming C to be the congressional floor median may inappropriately limit the threshold of political support that determines whether Congress engages in sanctioning activity. 6. I define “discretion” (like Epstein and O’Halloran 1999) as delegated authority, together with the severity of procedural and oversight constraints placed on that authority. 7. This is akin to the “subversion cost” highlighted by Gailmard (2002), wherein Congress imposes a cost via investigation. 8. Following Huber and McCarty (2004), agency punishments (θ) are fixed and exogenous. 9. Here, “line agency” refers to any department or agency outside the EOP that implements public policy. 10. Note, however, this category need not be defined on institutional features alone. As Lewis (2016) demonstrates, a majority of agencies—even those that do not bear the institutional hallmarks of insulation—report that Congress has more influence on agency spending post-appropriation. That is, agencies tend to perceive Congress as the principal to whom they must be more responsive. Other recent work suggests that the president’s influence is limited to agencies explicitly prioritized by their administrations (Potter and Shipan 2017). 11. The main results of model are not sensitive to the choice between linear and quadratic utility. Although I am largely agnostic as to whether the actors are risk averse (e.g., Bendor and Meirowitz 2004), linear utility produces results and comparative statics which are more readily interpretable. 12. E.O. 11930—“Performance by the Federal Energy Office of Energy Functions of the Federal Energy Administration” (July 30, 1976). Note that this order is included in most studies which regress macro-political characteristics on counts of executive orders. 13. “E.O. 9617—Transfer of Certain Agencies and Functions to the Department of Labor,” (September 19, 1945). 14. November 20, 2014. Quoted in Shabad (2014). 15. Marcos, Cristina. 2015. “Republican proposes abolishing the ATF amid bullet ban controversy,” The Hill, March 5. The ATF has since withdrawn the proposal. 16. Naturally, without sufficiently high d, delegation loss is entirely a function of ω, and there is no sanction loss. 17. Not surprisingly, uninsulated agencies comply because congressional threats pale in comparison to the cost of presidential sanction. 18. Note that in Proposition 1, optimal discretion is the second term of each compliance threshold. These values are discussed in Proposition 2 and proved in the Appendix. 19. The results for uninsulated agencies replicate a standard finding in the delegation literature, with one intuitive (but notable) exception. Uninsulated, proximate agencies receive less discretion in equilibrium as the president and legislative committee diverge. This implies an exception to the “ally principle,” which suggests that principals ought to rely more on ideologically proximate agents. The solid line in Figure 3 plots this relationship. Here, the agent and the president have similar policy preferences—yet, ideological divergence with a sufficiently powerful committee results in less discretion. In this case, the president’s ally cannot be expected to act in the president’s best interest because it serves two masters. This highlights a broader point in the context of “unilateral action.” Even when Congress is powerless to overturn a presidential directive outright, it can change outcomes by influencing those who implement the president’s policies. 20. Note, following Epstein and O’Halloran (1999), all results hold subject to the R>xi assumption. 21. In general, the threshold increases the higher the ideological distance, and decreases the greater the policy uncertainty. Moreover, the threshold is strictly lower for proximate agencies compared with distal ones. In other words, delegation is more likely to make the president better off over developing policy in the WH apparatus. 22. “Obama admits US gun laws are his ‘biggest frustration’,” BBC, July 24, 2015. http://www.bbc.com/news/world-us-canada-33646704. 23. McNeill (2016) referring to a 1995 statement by the American College of Physicians. 24. See, for example, “Notice of Legisaltive Mandate for Fiscal Year 2012,” NIH. http://grants.nih.gov/grants/guide/notice-files/NOT-OD-12-034.html. 25. Chris Cox, Executive Director of the NRA’s Institute for Legislative Action (NRA-ILA), wrote in a 2015 Op-Ed for Politico that “the National Rifle Association is not opposed to research that would encourage the safe and responsible use of firearms and reduce the numbers of firearm-related deaths. Safety has been at the core of the NRA’s mission since its inception. But that is not the goal of the gun control advocates who are behind the calls for CDC funding.” goo.gl/1PfiKD. 26. Memorandum of January 16, 2013. “Engaging in Public Health Research on the Causes and Prevention of Gun Violence,” 78 Federal Register 4295–4296. 27. “FY 2013 Full Year CR Operating Plan,” CDC and Prevention. https://www.cdc.gov/budget/documents/fy2013/fy-2013-operating-plan.pdf. 28. As Stephen Teret, director of the Center for Law and the Public’s Health at the Johns Hopkins Bloomberg School of Public Health noted, “even in the absence of an additional $10 million, CDC could have said, ‘Let’s reallocate some of our spending’.” Quoted in Rubin (2016). 29. http://dhss.delaware.gov/dhss/dms/files/cdcgunviolencereport10315.pdf. 30. See, for example, the Consolidated Appropriations Act of 2010 (P.L. 111–117; December 16, 2009). 31. HSPD were first issued in October 2001. The first such directive defines an HSPD as “record[ing] and communicat[ing] presidential decisions about the homeland security policies of the United States.” Thus, their basic function is indistinguishable executive orders and presidential memoranda (HSPD-1, October 29, 2001). 32. HSPD-12, August 27, 2004. url: goo.gl/G6xzMl. 33. Q1 FY14—HSPD-12 Public Report Summary, url: https://goo.gl/5YCjha. 34. The directive mandated the new forms of identification include several features: (1) verification of individual identity prior to issuance, (2) fraud resistance, (3) “rapid authentication,” and (4) verification of vendor through an accreditation process. 35. Author interview with former HSPD-12 Program Director, NASA, May 11, 2015. 36. The order’s provisions contain several delegations of authority. First, it tasks the Secretary of Commerce (in consultation with other agency heads) with setting the specific government-wide standard to fulfill the technical requirements. Second, it tasks all agencies with developing their own program to meet that standard. 37. A 2014 report indicates several of these agencies remain at or below 50% coverage for their federal contractors: the Federal Labor Relations Authority, Federal Maritime Commission, Federal Communications Commission, Commodity Futures Trading Commission, Merit Systems Protection Board, US Commission on Civil Rights, Small Business Administration, National Credit Union Association, and the Department of Health and Human Services (Q1 FY14—HSPD-12 Public Report Summary). 38. Sturgill, Emily. 2007. “New Badge Enrollments on the Rise; Benefits Too,” Glenn News, url: http://www.nasa.gov/centers/glenn/news/AF/2007/Nov07_HSPD-12HQ.html. 39. Author interview with former HSPD-12 Program Director, NASA, May 11, 2015. Appendix Lemma 1 For discussions of this result see Epstein and O’Halloran (1999: 248), Volden (2002a: 113), or Gailmard and Patty (2007: 879). Lemma 2 Given the discussion, its proof is omitted. Proposition 1 Given the discussion, its proof is omitted. Proposition 2 Given Lemmas 1 and 2, the President’s maximization problem reduces to one of five scenarios. For uninsulated agencies, when xA > xM and θs<|xA−xM|,   maxdEUP=−∫−RxA−d(ω+d)12Rdω−∫xA−dxA+dxA12Rdω−∫xA+dR(ω−d)12Rdω, which reduces to   maxdEUP=xA2−2xAd−(d−R)22R. Since ∂EUP∂d=R−xA−dR,d*=R−xA. When xA < xM and θs>|xA−xM|,   maxdEUP=−∫−RxA−d(ω+d)12Rdω−∫xA−dxA+mxA12Rdω−∫xm−dxm+dxM12Rdω−∫xm+dR(ω−d)12Rdω, which reduces to   maxdEUP=3xA2−2d2−2xAxm−4dxm+xm2+4dR−2R24R. Since ∂EUP∂d=−R−d−mR,d*=R−xm. For insulated agencies, when θs−θv>xA and xA > xm,   maxdEUP=−∫−Rxm−dω12Rdω−∫xm−dxA−d(w+d)12Rdω−∫xA−dxA+dxA12Rdω−∫xA+dR(w−d)12Rdω, which reduces to   maxdEUP=d(R−2xA−d+xm)2R. Given that ∂EUP∂d=R+xm−2xA2R,d*=R+xm−2xA2. Conversely, when θs−θv>xA and xA < xm,   maxdEUP=−∫−RxM−dω12Rdω−∫xM−dxM+dxM12Rdω−∫xM+dR(ω−d)12Rdω, which reduces to   maxdEUP=−d(xm+d−R)2R, and ∂EUP∂d=−2d+xm−R2R,d*=R−xm2. Finally, when xA>θs−θv and xA < xm,   maxdEUP=−∫−Rθs−θv−dω12Rdω−∫θs−θv−dxA−d(ω+d)12Rdω−∫xA−dxA+dxA12Rdω−∫xA+dR(ω−d)12Rdω, which reduces to   maxdEUP=−d(2xA+d−R−θs+θv)2R and   ∂EUP∂d=−2xA+2d+θs−θv+R2R. Given ∂EUP∂d=0, this equation has one solution: d*=R−2xA+θs−θv2. ▪ Proposition 3 In general, the President delegates when the following condition is satisfied   EUP(e=1)>EUP(e=0). (1) This means, in general, substitution of d*in EUP for each of the five cases above. First, I analyze the two cases of uninsulated agencies. When xA > xM and θs<|xA−xM|,   EUP(e=1)=xA2−2xAd*−(d*−R)22R, which, given d*=R−xA, reduces to the following comparison. If the following inequality holds, the President does not delegate.   RxA−xA2R>τ. When xA < xM and θs>|xA−xM|,   EUP(e=1)=3xA2−2d*2−2xAxm−4d*xm+xm2−4Rd*−2R4R, which, given d*=R−xA, reduces to the following comparison. If the following inequality holds, the President does not delegate,   4Rxm+2xAxm−3xA2−3xm24R>τ. The three cases of complete delegation for insulated agencies appear below. When θs−θv>xA and xA > xm,   EUP(e=1)=d*(−2xA−d*+xm+R)R. Given d*=R+xm−2xA2R, this equation reduces to zero. Thus, so long as τ is positive, the non-delegation inequality (Equation (1)) cannot be satisfied. When θs−θv>xA and xA < xm,   EUP(e=1)=−d*(d*+xm−R)2R, which, given d*=R−xm2, reduces to (xm−R)28R and is strictly positive so that Equation (1) cannot be satisfied. When xA>θs−θv and xA < xm,   EUP(e=1)=−d*(2xA+d*−R−θs+θv)2R and given d*=R−2xA+θs−θv2, this reduces to the following comparison   (R−2xA+θs−θv)28R<−τ, which, since the left side is strictly positive, cannot be satisfied. So the president always delegates to insulated agencies. ▪ References Bawn Kathleen. 1995. “Political Control versus Expertise: Congressional Choices about Administrative Procedures,” 89 American Political Science Review  62– 73. Google Scholar CrossRef Search ADS   Bendor Jonathan, Meirowitz Adam. 2004. “Spatial Models of Delegation,” 98 American Political Science Review  293– 310. Google Scholar CrossRef Search ADS   Bertelli Anthony M., Doherty Kathleen. 2017. “Setting the Regulatory Agenda: Statutory Deadlines and the Conditions for Compliance.” Working Paper , 1– 25. Bolton Alexander, Thrower Sharece. 2016. “Legislative Capacity and Executive Unilateralism,” 60 American Journal of Political Science  649– 63. Google Scholar CrossRef Search ADS   Burke John P. 1992. The Institutional Presidency . Baltimore, MD: Johns Hopkins University Press. Chiou Fang-Yi, Rothenberg Lawrence S.. 2014. “The Elusive Search for Presidential Power,” 58 American Journal of Political Science  653– 68. Google Scholar CrossRef Search ADS   Dickinson Matthew J. 2009. “We All Want a Revolution: Neustadt, New Institutionalism, and the Future of Presidency Research,” 39 Presidential Studies Quarterly  736– 70. Google Scholar CrossRef Search ADS   Dickinson Matthew J., Rudalevige Andrew. 2004. “Presidents, Responsiveness, and Competence: Revisiting the ‘Golden Age’ at the Bureau of the Budget,” 119 Political Science Quarterly  633– 54. Google Scholar CrossRef Search ADS   Dickinson Matthew J., Lebo Matthew. 2007. “Reexaming the Growth of the Institutional Presidency, 1940–2000,” 69 Journal of Politics  206– 19. Google Scholar CrossRef Search ADS   Epstein David, O’Halloran Sharyn. 1999. Delegating Powers: A Transaction Cost Politics Approach to Policy Making under Separate Powers . Cambridge, NY: Cambridge University Press. Google Scholar CrossRef Search ADS   Gailmard Sean. 2002. “Expertise, Subversion, and Bureaucratic Discretion,” 18 Journal of Law, Economics, and Organization  536– 55. Google Scholar CrossRef Search ADS   Gailmard Sean, Patty John W.. 2007. “Slackers and Zealots: Civil Service, Policy Discretion, and Bureaucratic Expertise,” 51 American Journal of Political Science  873– 89. Google Scholar CrossRef Search ADS   Howell William G. 2003. Power without Persausion: The Politics of Direct Presidential Action . Princeton, NJ: Princeton University Press. Howell William G. 2005. “Unilateral Powers: A Brief Overview,” 35 Presidential Studies Quarterly  417– 39. Google Scholar CrossRef Search ADS   Huber John D., McCarty Nolan M.. 2004. “Bureaucratic Capacity, Delegation, and Political Reform,” 98 American Political Science Review  481– 94. Google Scholar CrossRef Search ADS   Kennedy Joshua B. 2015. “‘Do This! Do That!’ and Nothing Will Happen: Executive Orders and Bureaucratic Responsiveness,” 43 American Politics Research  59– 82. Google Scholar CrossRef Search ADS   Krause George A. 2002. “Separated Powers and Institutional Growth in the Presidential and Congressional Branches: Distinguishing between Short-Run versus Long-Run Dynamics,” 55 Political Research Quarterly  27– 57. Google Scholar CrossRef Search ADS   Krause George A. 2009. “Organizational Complexity and Coordination Dilemmas in U.S. Executive Politics,” 39 Presidential Studies Quarterly  74– 88. Google Scholar CrossRef Search ADS   Krause George A., Dupay Brent M.. 2009. “Coordinated Action and the Limits of Presidential Control over the Bureaucracy: Lessons from the George W. Bush Presidency,” in Provost Colin, Teske Paul, eds., President George W. Bush’s Influence over Bureaucracy and Policy: Extraordinary Times, Extraordinary Powers . 81– 104. New York, NY: Palgrave. Google Scholar CrossRef Search ADS   Kriner Douglas L., Reeves Andrew. 2015. The Particularistic President: Executive Politics and Political Inequality . Cambridge University Press. Lewis David E. 2003. Presidents and the Politics of Agency Design . Stanford University Press. Lewis David E. 2008. The Politics of Presidential Appointments: Political Control and Bureaucratic Performance . Princeton University Press. Google Scholar CrossRef Search ADS   Lewis David E. 2016. “Political Control and the Presidential Spending Power.” Working Paper. Mayer Kenneth R. 2001. With the Stroke of a Pen: Executive Orders and Presidential Power . Princeton University Press. Mayer Kenneth R. 2009. “Going Alone: The Presidential Power of Unilateral Action,” in Howell William G., Edwards, eds George C., The Oxford Handbook of the American Presidency , 427– 98. Oxford University Press. McNeill Matthew B. 2016. “Gun Violence: A Public Health Concern?” Clinical Correlations  June 9. Moe Terry M. 1985. “The Politicized Presidency,” in Chubb J. E., Peterson P. E., eds., The New Direction in American Politics . 235–272. Washington, DC: Brookings Institution Press. Moe Terry M. 1989. “The Politics of Bureaucratic Structure,” in J. E. Chubb and P. E. Peterson, eds., Can the Government Govern ? 267–330. Washington, DC: Brookings Institution Press. Moe Terry M., Howell William G.. 1999. “The Presidential Power of Unilateral Action,” 15 Journal of Law, Economics, and Organization  132– 79. Google Scholar CrossRef Search ADS   Nathan Richard P. 1983. The Administrative Presidency . Prentice Hall. Neustadt Richard E. 1960. Presidential Power and the Modern Presidents: The Politics of Leadership from Roosevelt to Reagan . Wiley & Sons. Potter Rachel Augustine, Shipan Charles R.. 2017. “Agency Rulemaking in a Separation of Powers System,” Journal of Public Policy  1– 25. https://doi.org/10.1017/S0143814X17000216. Rogowski Jon C. 2016. “Presidential Influence in an Era of Congressional Dominance,” 110 American Political Science Review  325– 41. Google Scholar CrossRef Search ADS   Rubin Rita. 2016. “Tale of 2 Agencies: CDC Avoids Gun Violence Research but NIH Funds It,” 315 Journal of the American Medical Association  1689– 92. Google Scholar CrossRef Search ADS PubMed  Rudalevige Andrew. 2009. “The Administrative Presidency and Bureaucratic Control: Implementing a Research Agenda,” 39 Presidential Studies Quarterly  10– 24. Google Scholar CrossRef Search ADS   Rudalevige Andrew. 2012. “Executive Orders and Presidential Unilateralism,” 1 Presidential Studies Quarterly  138– 60. Google Scholar CrossRef Search ADS   Selin Jennifer L. 2015. “What Makes an Agency Independent?,” 59 American Journal of Political Science  971– 87. Google Scholar CrossRef Search ADS   Volden Craig. 2002. “A Formal Model of the Politics of Delegation in a Separation of Powers System,” 46 American Journal of Political Science  111– 33 Google Scholar CrossRef Search ADS   © The Author(s) 2018. Published by Oxford University Press on behalf of Yale University. All rights reserved. For Permissions, please email: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Journal of Law, Economics, and Organization Oxford University Press

Delegation or Unilateral Action?

Loading next page...
 
/lp/ou_press/delegation-or-unilateral-action-BBgiMtqXD6
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of Yale University. All rights reserved. For Permissions, please email: journals.permissions@oup.com
ISSN
8756-6222
eISSN
1465-7341
D.O.I.
10.1093/jleo/ewy001
Publisher site
See Article on Publisher Site

Abstract

Abstract Unilateral presidential actions often face implementation problems in the executive branch. I argue these actions are better studied as delegation. I model the conditions under which a president is likely to delegate and provide discretion to subordinates either insulated or uninsulated from their control. I find legislators benefit from agency discretion when presidents pursue policymaking in the executive branch. The threat of legislative sanction induces agents to deviate from presidential priorities, and inter-branch disagreement increases bureaucratic non-compliance in insulated agencies. Nonetheless, in equilibrium, the president is more likely to delegate to insulated agents. Ultimately, the model demonstrates how the politics of direct action are influenced by the need for bureaucratic cooperation. Case studies on US presidential directives mandating public funding of gun violence research and security reforms at government facilities illustrate key features of the model. 1. Introduction In separation of powers systems, executives are often said to act “unilaterally” to enact policies without the explicit consent of legislatures. Most research argues the primary constraints on this power are legislative and judicial veto points (Moe and Howell 1999; Howell 2003; Chiou and Rothenberg 2014; Bolton and Thrower 2016). However, executives—like legislatures—rely on bureaucrats with multiple principals to implement policy. For example, when Harry Truman issued an executive order that seized the property of American steel mills, it delegated the authority to the Secretary of Commerce and gave him substantial discretion to both determine the terms of employment and even return the mills to their owners.1 The Secretary used that discretion to delay action until the order was invalidated (Neustadt 1960). The central premise of this article is that directives like these are more effectively studied instances of delegation, as opposed to policy change “with the stroke of a pen” (Mayer 2001). I study executive policymaking by developing a model of delegation that approximates the implementation of directives. In the model, a political principal (e.g., the president) makes policy “in-house” or delegates to an outside agent. These agents can pay an exogenous penalty for not complying, or choose to implement policy. After selecting policy, outside agents are subject to punishment from another principal (e.g., a congressional median or committee). A key component of the model is the analysis of the implications of agency insulation. The non-compliance penalty relative to the legislative penalty determines the relative insulation of the outside agent. Insulated agents pay more severe legislative penalties (and are thus insulated with respect to the president), whereas uninsulated pay more severe non-compliance penalties. Thus, the model organizes observed features of executive policymaking like bureaucratic non-compliance, agency insulation, and legislative sanctioning in a familiar and simple framework. As a result, it provides a way forward for empirical work by redirecting attention to the content of directives—rather than their issuance. Placed in the context of research on executive unilateralism, I argue that bureaucratic agency is a mechanism by which legislatures secure better policy outcomes when executives “act alone.” The general point that directives are not self-executing has been made by presidency research—much of which suggests that non-compliance and negotiation on the part of bureaucrats is common (e.g., Dickinson 2009; Krause 2009; Krause and Dupay 2009; Rudalevige 2012; Kennedy 2015).2 While studies argue implementation matters, it is less clear how these features should systematically influence outcomes. Accordingly, the theory highlights several broad implications of bureaucratic implementation. First, the model suggests that inter-branch disagreement induces bureaucratic non-compliance. Thus, this disagreement not only limits opportunities to shift the status quo—as Howell (2003) shows—it reduces the probability such orders will be carried out. Second, the theory implies that executives are more likely to delegate when the agent is insulated from potential punishment. Insulated agencies are less compliant by design. This generates uncertainty about when they will use their discretion to implement a policy unfavorable to the principal. Thus, as long as policy development is costly for executives to undertake in-house, delegating to insulated agencies makes the president better off. Counterintuitively, these are precisely the agents most likely to bend to congressional pressure. More generally, the dynamics of agent non-compliance and policy selection demonstrate that “unilateral” policy outcomes can reflect legislator preferences, even in the absence of laws that explicitly overturn directives. 2. Unilateral Action, Agency Problems, and the Executive Branch Howell (2003) writes that “modern presidents often exert power by setting public policy on their own and preventing Congress and the courts—and anyone else for that matter—from doing much about it” (14). This perspective influenced a generation of quantitative research on the presidency that attempts to identify political circumstances that enable the president to act alone. Much of this research focuses on the empirical study of presidential directives.3 More recently, the conceptual focus on what presidents can accomplish alone has informed investigations of the president’s role in the distribution of federal spending (e.g., Kriner and Reeves 2015; Rogowski 2016). But there are several reasons to believe an alternative theoretical base might be appropriate. First, this stylized depiction of policy change does not adequately capture the presidents’ orders. Second, scholars have long recognized that managerial challenges within the executive branch often thwart policy change (e.g., Nathan 1983; Burke 1992). Recent work argues this past emphasis is largely at odds with the implicit assumption of perfect implementation (Dickinson 2009; Krause 2009; Rudalevige 2009). It is important to build on a central point: when Congress passes a law or the president issues an executive order, both rely on bureaucrats for policy change. Absent that recognition, theories risk aggrandizing the capabilities of presidents and downplaying the agency of the bureaucrat. In contrast, my model builds on well-known agency problems in the executive branch. That is, when presidents want policy changed, they must rely on subordinates with agency. This feature of the presidency has been the basis for the study of public management and political influence over policy outcomes. For example, the notion that presidential and bureaucratic preferences often diverge is at the core of most explanations for the politics of appointments (Lewis 2008). Moreover, bureau responsiveness, even after politicization, is not guaranteed (Dickinson and Rudalevige 2004). However, nearly every action the president takes can be considered an act of delegation. A “delegation-all-the-way down” perspective risks returning to an understanding of the president as overwhelmed and ultimately incapable of seriously influencing policy. Therefore, I conceive of the president’s decision to delegate as a dichotomous choice: delegate to “external” agents—those in government corporations, independent agencies, and cabinet departments—or, delegate to actors within what has been called the “presidential” branch—the White House (WH) and Executive Office (EOP).4 Presidents have the capacity to develop policy in a wide variety of issue areas with the resources at their disposal within the WH and EOP. This preserves an essential point, made first by Moe (1985): that within the presidential branch, presidents face fewer collective action problems and lower risks of policy drift. 3. The Model I present a spatial model of delegation within the executive branch with three actors: the president (P), a congressional committee (C), and an external agent (A). The congressional committee may be thought of as the committee with oversight jurisdiction over the corresponding agency.5 In the model, the president attempts to secure a policy outcome (realized by the equation x^=x+ω, where x is the policy selected and ω is a uniformly distributed shock) nearest to their preferred (xP), while avoiding the resource cost (τ) associated with developing policy in the WH and EOP. To avoid that cost, the president may delegate via directive to an external agent ( e∈{0,1}) and provide them with a level of discretion ( d∈[0,R+]).6 The congressional committee looks to obtain a policy closest to its bliss point (xC) through sanctioning ( s∈{0,1}), which imposes an exogenous cost on the agent, θs, if the agency opts in to policymaking. This sanction can be thought of as a law that directly punishes the agent, a rider on an appropriations bill, or the inconvenience of having to give formal testimony on Capitol Hill.7 This leads to two important assumptions. First, Congress only punishes agents for deliberate policymaking action on behalf of the President. That is, agents who do not cooperate with the President are not subject to congressional sanction. This is often the case in practice, as agents who have taken deliberate action under direction of presidents are then subject to hearings and investigation. I discuss several examples later. This also highlights scope conditions of the model, which is confined to understanding bureaucratic response after a president has become the first-mover. Second, I assume that s = 1 when −|x^−xP|>−|x^−xC|. This implies the committee does not know exactly where the policy is, but does know when the president is made better off by agency actions. In this way, the committee is modeled as largely reactionary. This allows me to focus on the relationship between the president and agencies without overcomplicating the model. It is problematic, for example, to assume that congressional committees have perfect knowledge of the policy implementation process. But assuming they have no knowledge seems equally divorced of the basic political phenomenon. Alternatively, granting a unitary C the power of ex post veto is likely inappropriate because collective action problems limit their capacity to respond (Moe and Howell 1999). Here, I assume they know, broadly, whether the policies make the president better off. It is also important to note that the model does not investigate congressional incentives to provide resources to the executive branch; τ is exogenous. I defer discussion of potential implications to the conclusion. The agent, if chosen to make policy, can opt out by refusing to comply with the president’s directive ( v∈{0,1}). If it opts in, it selects a policy conditional on the level of discretion supplied by the president. If the agency opts out of policymaking, x^=ω, and the agency pays a non-compliance cost that can be thought of as an ex post presidential sanction (θv).8 Presidents have a variety of tools to impose costs on non-compliant agencies—including the removal of appointed program managers and, in some cases, the reassignment of agency functions. Moreover, variation in this parameter allows the model to analyze institutional variation among line agencies.9 More specifically, I define the following to aid in describing the results: Definition An “insulated” or “independent” Agency (A) has non-compliance costs less than the cost of congressional sanction ( θv<θs). Conversely, an “uninsulated” Agency (A) has non-compliance costs greater than the cost of congressional sanction ( θv>θs). Scholars have long recognized that some agencies are designed to render them less subject to presidential manipulation (e.g., Moe 1985; Lewis 2003; Selin 2015). I operationalize this in terms of an agency’s vulnerability to presidential punishment—relative to congressional punishment. The effectiveness and availability of punishments are often due to features of agencies that are difficult to change. For example, fixed service terms effectively inhibit the president’s ability to punish agencies by replacing (or leaving vacant) their appointed heads.10 Nonetheless, there are other aspects of agency structure this definition cannot capture (e.g., ex post review of decision-making or reporting requirements). Since I do not model the president’s choice among different external agents, a more complete consideration of these institutional forms is beyond the scope of the present paper—but could be taken up in future research. The utility of the president is given by   UP=−|x^−xP|−(1−e)τ, so that they attempt to minimize the disutility associated with distant policy and the cost of developing policy in-house.11 The committee’s utility is governed solely by the policy outcome, such that   UC=−|x^−xC|. Finally, the Agent’s utility is governed by policy outcomes and the costs associated with congressional sanction ( θs∈[0,R+]) and non-compliance ( θv∈[0,R+]).   UA=−|x^−xA|−sθs−(1−v)θv. Sequence of play: Nature selects a random shock, ω, which is a common prior, uniformly distributed between [−R,R]. A observes ω. The president chooses whether to delegate or develop policy within the WH and EOP ( e∈{0,1}). If e = 0, then the president pays a resource cost, τ, and the EOP selects a policy, x. If e = 1, then the president selects a level of discretion, d. The agent chooses whether or not to comply, v∈{0,1}, after observing ω. If v = 0, then s = 0 and x = 0. If v = 1, the agent selects a policy, x∈[−d,d]. If −|x^−xP|>−|x^−xC|, then s = 1, otherwise, s = 0. Play ends and payoffs are distributed. Player preferences (and all parameters except ω) are common knowledge, and all arbitrary spatial orientations reduce to two cases. Let xm denote the midpoint between xP and xC, a policy outcome that would result in s = 0. Without loss of generality, I assume that xP = 0, xC>0 so that the relevant cases are xm > xA in which, the agent is closer to the president; and xm < xA, when the agent is closer to the committee. I define the following to simplify discussion. Definition A “proximate” agency (A) has an ideal point closer to the President (xA < xm). Conversely, a “distal” agency has an ideal point closer to the committee (xm < xA). so that in the description of the results, I refer to agencies as one of four types:   A∈{insulated,uninsulated}×{proximate, distal}. The model differs from related work in several respects. First, in contrast to Howell (2003), critical pivots in Congress and the Judiciary are absent. Whereas the unilateral politics model outlines conditions that lead to presidential action, this model attempts to explain how the agency of bureaucrats influences final outcomes—after the decision to act alone has been made. Second, in contrast to many models of delegation (Bawn 1995; Epstein and O’Halloran 1999; Volden 2002; Gailmard and Patty 2007), the principal’s key trade-off is not primarily motivated by gains in expertise. In some cases, delegation is attractive because it reduces the cost of policymaking for the president—presumably reserving limited resources for other policy initiatives. A few simple examples underscore the applicability of this setup. Presidents, in pursuit of their goals, are faced with an initial choice of whether to use resources within their immediate domain, or delegate to external agents. In 1976, for instance, Gerald Ford chose to delegate policymaking functions to the Federal Energy Office within the EOP, rather than vest those functions in the newly created Federal Energy Administration.12 If a president delegates, they are faced with the task of formulating the agent’s mandate to make policy, which can be limited or provide latitude. The president could, require that the agent consult with other agencies, as Harry Truman did when he redelegated wartime employment functions to the Department of Labor in 1945.13 Relatedly, the president could simply specify the new policy in great detail. Members of Congress can sanction agents directly. As Barack Obama’s issued a series of immigration-related directives in November 2014, then-Senator Jeff Sessions (R-AL) reminded the applicable departments that “Congress has the power and every right to deny funding for unworthy activities.”14 In another instance, a proposed ban on armor-piercing ammunition resulted in the introduction of legislation to abolish the agency responsible.15 Although the president pays political costs for overturned or impeded policies, bureaucrats endure punishments which vary in severity. 4. Results The model produces several key results relevant to executive-driven policy initiatives. The first is that for insulated agencies, inter-branch conflict influences the likelihood of non-compliance. Specifically, as disagreement between the committee and the president increases, these agents are more likely to opt-out. However, despite that dynamic, when the agent is insulated from presidential punishment, the president is always better off delegating. This is because the guaranteed compliance of uninsulated agents lets presidents know when they will opt-in and pick a policy that makes the president worse off. Thus, the model demonstrates how bureaucratic structure can influence the president’s dependence on agents when engaging in unilateralism. 4.1 When Do Agencies Serve Congress? The solution concept is subgame perfection and the model is solved via backward induction. All proofs appear in the Appendix. Let θs¯>|xA−xm|, the value of the sanction punishment which exceeds the agency’s policy payoff from moving the outcome from its ideal point to the sanction-free point. Working backwards, it is apparent that the optimal policy choice for the bureaucrat (denoted by x*) depends on θs and xA. When the cost the committee imposes is sufficiently low ( θs<θs¯), or the bureaucrat is distal, then the agency’s choice follows the intuitive, well-known result regarding policy selection. Lemma 1 When xA> xm, or θs<θs¯  x*={d if −R<ω≤xA−dxA−ω if xA−d<ω≤xA+d−d if xA+d≤ω<R. For any given random shock, the agent pulls policy as close as possible to its ideal point, given the bounds of its discretion. However, the results diverge from Lemma 1 when sanctioning imposes a sufficient cost. Lemma 2 When θs≥θs¯ and xA < xm  x*={d if −R<ω≤xA−dxA−ω if xA−d<ω≤xm−dxm−ω if xm−d<ω≤xm+d−d if xm+d≤ω<R. When the random shock is between—R and xm−d, the agent shifts policy as close as possible to its ideal point because it cannot select xm. On the other hand, when the shock exceeds xm−d, the Agent shifts policy as close to xA as possible, while avoiding sanctioning by selecting xm−ω at minimum. Figure 1 illustrates this change in bureaucratic behavior. To illustrate the additional policy loss (highlighted in red), I set d=|xA−xm|. 16 Delegation always implies some policy loss for the president. However, the presence of a powerful congressional committee implies an additional cost. In Figure 1, the agent selects the policy d+ω=xm when ω=xA, despite the fact this will move policy away from its ideal point. So under the right conditions, the bureaucrat’s optimal strategy is to appease the congressional committee. Figure 1. View largeDownload slide Committee Influence and Presidential Policy Loss. Figure 1. View largeDownload slide Committee Influence and Presidential Policy Loss. 4.2 Bureaucratic Non-compliance Given x*, I characterize whether the agent will comply with the president’s order, or opt out and allow the state of the world (ω) to obtain. In general, the agent complies if the following inequality is satisfied:   −|x*+ω−xA|−s*θs>−|ω−xA|−θv. The right-hand side of the equation shows the policy that would result if the agency refused to implement the order. This comparison has direct implications on compliance, discretion, and ultimately, the utility of delegation. Proposition 1 (Uninsulated). In any equilibrium, both proximate or distal agencies are perfectly compliant. If θv>θs,v*=1. Insulated, Distal. Agencies comply when the state of the world permits avoidance of or proper policy inducement outweighing congressional sanction. If θv<θs, when θs−θv>|xA−ω|   v*={0 if −R<ω≤xm−R+xm−2xa2R1 if xm−R+xm−2xa2R≤ω<R. when |xA−ω|>θs−θv>0   v*={0 if −R<ω≤θs−θv−R+xm−2xa2R1 if θs−θv−R+xm−2xa2R≤ω<R.Insulated, Proximate. Agency are complies when the state of the world permits avoidance of or proper policy inducement outweighing congressional sanction. If θv<θs, when θs−θv>|xA−ω|   v*={0 if −R<ω≤xm−R−xm21 if xm−R−xm2≤ω<R, when |xA−ω|>θs−θv>0   v*={0 if −R<ω≤θs−θv−R−2xA−θs−θv21 if θs−θv−R−2xA−θs−θv2≤ω<R. Insulated agencies comply with the directive only when the policy shock puts a beneficial policy within reach.17 A beneficial policy is one that either allows the agent to avoid sanctioning, or allows the agent to select a policy that effectively outweighs the cost of congressional sanction. This means—somewhat counterintuitively—that insulated, proximate agencies are less compliant (and receive less discretion) the more powerful the presidential punishment. This also reveals some important points about compliance with presidential directives. First, it’s worth noting that as discretion increases, so does the region of random shocks for which the insulated bureaucrat will opt-in to policymaking.18 This means that discretion increases the probability of compliance with the president’s directive. Figure 2 illustrates these basic implications. An increase in discretion contracts the non-compliance region (in red). Because the agent is insulated from presidential control, they comply when ω allows them to implement xm. For proximate agencies, increases in policy disagreement between the president and the committee increase the probability of non-compliance. In Figure 2, the shift from xm1 to xm2 results in a reduction of the compliance region. Note also, that because this region is partly a function of the “state of the world” as determined by ω, in any equilibrium in which x∈[−d,d]≠[−R,R] and A is insulated, there is some probability of non-compliance. Most importantly, however, since the optimal level of discretion is linked to committee preferences, when the committee is sufficiently powerful, the non-compliance region for insulated agencies is increasing in the distance between the president and the committee. Figure 2. View largeDownload slide Discretion and Non-compliance. Figure 2. View largeDownload slide Discretion and Non-compliance. 4.3 Bureaucratic Discretion Let d* denote the president’s optimal choice of d. Given the results above, the president’s maximization problem takes one of several forms, which suggest some differing predictions based on agency insulation and ideological divergence between president, agency, and committee. Proposition 2 (Uninsulated). For distal agencies, the optimal level of discretion is increasing in policy uncertainty and decreasing in ideological distance. When the agency is proximate and the committee is sufficiently powerful, discretion is increasing in policy uncertainty and decreasing in divergence between the president and the committee,   d*={R−xA if xm<xA or θs<θs¯R−xm if xA<xm or θs≥θs¯.Insulated, Distal. The optimal level of discretion is increasing in policy uncertainty and committee ideological distance, but decreasing in agency distance,   d*=max[0,R+xm−2xA2R] if xm<xA.Insulated, Proximate. If there is a proper policy inducement outweighing congressional sanction, then discretion is decreasing in agent distance and non-compliance cost, and increasing in policy uncertainty and congressional sanction. If there is not a sufficient inducement, discretion is increasing in policy uncertainty and decreasing in committee distance,   d*={R−xm2 if xA<xm and θs−θv>xAmax[0,R−2xA+θs−θv2] if xA<xm and θs−θv<xA. Equilibria with insulated agencies are a function of ideological proximity, policy uncertainty, and the fixed costs imposed by Congress and the president.19 For distal agencies, there is a standard result associated with delegation—that divergence between the principal and agent results in a reduction in discretion. Importantly, however, divergence between president and committee marginally increases discretion because non-compliance motivated by Congress makes the president slightly better off. Distal, insulated agencies sometimes use their discretion to pull policies away from the president. Non-compliance restricts the region for which that occurs. However, importantly, this discretion is essentially flat (zero), alongside the other cases in Figure 3. A similar dynamic is at play in when the committee is sufficiently weak and the agent is proximate—where, the president’s punishment seems to “backfire.” Even proximate agents sometimes pull policy away from the president—a powerful Congress prevents them from doing that, whereas a high-compliance cost encourages it. So in this case, discretion is increasing in θs and decreasing in θv. Figure 3. View largeDownload slide Inter-branch Disagreement, Discretion, and Insulation. Note: All comparative statics assume R>xi; insulated, proximate assumes θs−θv>xA; d*(insulated, distal)>0 when R>2xA−xm. Figure 3. View largeDownload slide Inter-branch Disagreement, Discretion, and Insulation. Note: All comparative statics assume R>xi; insulated, proximate assumes θs−θv>xA; d*(insulated, distal)>0 when R>2xA−xm. As Figure 3 shows, for insulated, proximate agencies, discretion is again a function of ideological divergence between the president and either the agency or the committee. This compounds the dynamic shown in Figure 2. A more distant committee means a higher probability of non-compliance and less discretion—which, in turn, leads to an even higher probability of non-compliance. Overall, it is also apparent that so long as R>xi, uninsulated agencies receive more discretion than insulated ones. I return to these implications in Section 5.20 4.4 Delegation and the Presidential Branch Finally, the model reveals the conditions that must be satisfied for the president to delegate to an external agency. Let e*be the optimal choice of delegation. Under most conditions, there are values of τ that render delegation beneficial. However, the circumstances under which the president does not delegate are confined to uninsulated agencies. I explain the rationale below. Proposition 3 (Uninsulated, Distal). In any equilibrium, the president delegates when the cost of developing policy “in-house” exceeds a critical value decreasing in the distance between the agent and policy uncertainty. If θs<θv and xm < xA,   e*={0if RxA−xA22R>τ1otherwise.Uninsulated, Proximate. The president delegates when the cost of developing policy “in-house” exceeds a critical value decreasing in the distance between the agent, committee, and policy uncertainty. If θs<θv,θs>θs¯, and xA < xm  e*={0 if 4Rxm+2xAxm−3xA2−3xm24R>τ1otherwise.Insulated. In equilibrium, the president delegates. If θs>θv, then e*=1. For distal agencies, this holds if τ>0. For proximate, this holds even if τ = 0. When the agent is insulated from presidential control, the president delegates. Irrespective of policy disagreement, delegation results in gains in expertise and efficiency that cannot be matched within the presidential branch. For all agencies, as the distance between the president and the insulated agent increase, the utility of delegation declines. Thus, the cost of in-house policy development must be sufficiently high to offset the policy loss incurred through delegation and committee influence. For uninsulated agencies, the thresholds for delegation are straightforward.21 For insulated agencies, there are no thresholds. The logic behind this finding is directly tied to the role non-compliance plays in unilateral action. Non-compliance generates the possibility that the state of the world will prevail. This uncertainty prevents the president from isolating the circumstances in which in-house development is more optimal. Uninsulated agencies are perfectly compliant. That allows the president to know when they will use their discretion to select a policy worse than the one generated by a presidential branch that lacks expertise. Absent that guarantee the president delegates to avoid wasting resources needlessly. Thus, under these conditions, agency insulation can increase delegation. The importance of this result should not be understated. It implies that if the president wants to act in an area of policy governed by an insulated delegate, then unilateral action is dependent on bureaucratic agency. For proximate agents, this holds even if the policy is “free” (τ = 0) to produce. 5. Applications Given the substantive motivation of the model, this section details several empirical applications. First, I provide two case studies to demonstrate its usefulness for interpreting instances of unilateral presidential action. In each case, absent the model, unilateral action theory would provide an incomplete explanation for the observed policy outcomes. Second, I describe expectations for more systematic analyses of president-driven policymaking. To recap, I argue the model applies to scenarios in which policy could be developed by either the EOP or line agencies. In other words, I expect that many policies developed in the President (e = 0) will be carried out by bureaucrats outside the EOP. This is analogous to the situation Congress finds itself in (a la Epstein and O’Halloran 1999)—since legislators themselves implement no policy. Thus, according to the model, not delegating means dedicating in-house resources to making a policy—which might then be imposed on an agency. 5.1 Policy Drift, Congressional Influence, and Gun Violence Research In a 2015 interview with the BBC, President Obama referred to the lack of policy movement on firearm regulation as his “biggest frustration” in office.22 Editorials and commentaries cite congressional gridlock and the influence of the National Rifle Association (NRA) for this outcome. But this leaves unanswered a key question: what prevented the President from “going it alone”—as the administration had done with other important issues like carbon capture technology and deferred action for undocumented immigrants (DACA)? The model sheds light on this question by highlighting the incentives of bureaucrats faced with implementing presidential orders. As the political struggle to fund gun violence research shows, bureaucrats with discretion often select policies to appease multiple principles—especially when those principals can impose substantial costs on implementors. Despite a presidential directive ordering the Centers for Disease Control (CDC) and National Institutes of Health (NIH) to conduct gun violence research, both agencies have yet to take major steps—due in large part to the political threat of Congress. In the United States, violence associated with firearms has been referred to as a public health epidemic that eclipses yearly causes of death like influenza.23 However, for nearly 20 years, the federal government has funded no research into the causes of that violence—in part, due to an annual limitation rider in the budgets of the CDC and NIH. The so-called “Dickey Amendment” prohibits agency funding from being used to “advocate or promote gun control.”24 Although the NRA itself does not interpret the rider as a ban on gun violence research, the rider (which first appeared in 1996) has had a clear chilling effect on discretionary funding priorities in both agencies.25 After a series of legislative defeats and the Sandy Hook Elementary School shooting, President Obama ordered the CDC and NIH to “conduct or sponsor research into the causes of gun violence and the ways to prevent it.”26 Although the order left the level of funding up to the discretion of the agency, the President’s preferences were clear: he followed up the order with a request from Congress for $10 million in funding for gun violence research. That year, the CDC had more than $30 million allotted for intentional injury prevention.27 Both agencies had a clear presidential directive, as well as the resources and discretion to act.28 Neither agency did precisely what the president asked for, but differences in the agencies’ responses are instructive. The CDC funded no research, but instead conducted a short internal study that was released 2 years later.29 The NIH, on the other hand, released a formal call for research proposals that referenced the president’s directive. However, it funded only two projects—totaling just $1.62 million. Although both agencies complied with the vague language of the memorandum, the NIH moved considerably closer to the president’s preferred policy by funding independent research. As the theory implies, this variation in policy drift is attributable to the relative cost Congress can impose on bureaucrats implementing orders. Although both are agencies of the Department of Health and Human Services, the CDC’s budget is a series of line-item appropriations for specific programs. The NIH, on the other hand, receives block funding by institute—insulating its research priorities.30 Moreover, congressional opposition to the President’s order was clear: support for the limitation rider and opposition to new firearm regulation remained strong. Thus, the CDC was comparatively vulnerable to congressional sanction, and both agencies had incentives to avoid significant steps to publicly fund gun violence research. The case highlights important features of the model. The president can direct agencies to use their discretion to pursue a policy agenda. The existing perspective on unilateral action would label the president’s directive as an instance of policy change. But that obscures important variation in the final outcome. In this instance, the federal government continues to fund very little research on gun violence because agencies with discretion operate among forces other than presidential decree. 5.2 Discretion, Non-compliance, and the Implementation of HSPD-12 The September 11, 2001, attacks ushered in dramatic changes in national security policy and the structure of the administrative state. Among these were efforts to enhance security protocols for government facilities deemed potential targets. Homeland Security Presidential Directive-12 (HSPD-12) initiated one such program in August 2004.31 The directive mandated the use of “SmartCard” identification technology for all federal employees and contractors in order to standardize existing, disparate programs, and limit facilities’ vulnerability to intrusion.32 Despite this agreeable aim, HSPD-12 was subject to widespread, systemic non-compliance by agencies: as of the most recent quarterly report, it has still yet to be fully implemented, over 10 years later.33 Its implementation illustrates an important dynamic in the model: the impact of discretion on compliance. The model implies that agencies “opting out” of policymaking results from disagreement, insufficient discretion to deal with unforeseen exigencies in the implementation process, and the threat of sanction from Congress. HSPD-12 demonstrates the first two, as the text of the directive provides specific, technical requirements for government-wide identification cards.34 By design, the identification standard dramatically constrained the discretion of agencies. Prior to HSPD-12, agencies operated their own security programs. This variation provided agencies with the flexibility to determine the appropriate level of intrusion protection based on the facilities they operated. The directive meant that all agencies (regardless of functions or security vulnerabilities) would have to implement a high standard—so high, in fact, the technology needed to do so had never been produced on the scale required.35 Thus, agencies were provided little-to-no discretion to implement a policy for which the technical uncertainty was substantial. Moreover, the President’s policy was substantially more stringent than the one preferred by many agencies. The standard for acquiring the necessary access card requires an intensive screening process that includes criminal and defense background checks conducted by multiple agencies. According to the model, HPSD-12 presents precisely the scenario in which non-compliance via “opting out” of the policymaking process should occur. Compliance problems immediately followed the issuance of the general standard.36 The directive required agencies put a program in place to achieve compliance within 4 months and required full compliance within 8 months. Few agencies successfully met the deadline. This was not merely a function of the technical challenges of implementation. Non-compliance was concentrated in agencies whose functions posed few security risks or did comparatively little work through contractors and would likely have preferred to meet less stringent fraud protection standards.37 Moreover, the full extent of that non-compliance is unclear, as agency self-reporting is infrequent and sometimes includes errors. Implementation at the National Aeronautics and Space Administration (NASA) highlights critical features of this non-compliance. NASA began developing its own SmartCard system prior to HSPD-12, so that compliance involved sunk research and development costs.38 In 2004, implementation was assigned as an additional staff function of the Facilities and Administrative Services Division, contrary to the instructions of the directive—which required the development of a dedicated program. For 3 years, the agency made no significant progress.39 After years of inaction at NASA, the president dedicated personnel in the WH and Office of Management and Budget to help implement HSPD-12. At NASA, WH officials met with the Deputy Administrator and a newly appointed Program Director in 2007 to ensure the order was implemented. Full compliance was finally achieved 13 months later. In sum, HSPD-12 demonstrates of the basic dynamics of non-compliance in the unilateral presidency. Non-compliance occurred in the area national security policy—often considered the president’s domain—even in relative absence of congressional opposition. In this case, the security of government facilities was dramatically delayed because agencies—driven by their own preferences—chose to not implement the president’s order. 5.3 Implications for Empirical Research Although the model is consistent with the cases above, it points to the need for analyses of the content of presidential directives like executive orders, memoranda, and proclamations. For this reason, it is important to highlight several empirical implications of the model, which all involve observable behavior that can be measured. Although I have mentioned several potentially interesting results, five merit emphasis: Prediction 1. All else equal, presidential directives should delegate more authority to agencies insulated from presidential control (Proposition 3). Institutional variation among bureaucratic agencies—and its impact on presidential control—has been the subject of a large body of research (e.g., Moe 1989; Lewis 2003; Selin 2015). Much of this research has sought to determine how politically responsive agencies are to elected institutions. Here again, presidential directives offer an opportunity to validate the intuition of the model. Prediction 2. All else equal, presidential directives should provide less discretion to agencies insulated from presidential control (Proposition 2). Nonetheless, uninsulated bureaus are generally more reliable agent of the president. Directives—like legislation—often included procedural limitations designed to limit the discretion of agents. Prediction 3. All else equal, uninsulated agencies should be more compliant with presidential directives (Proposition 1). Although it remains difficult to track, recent studies have attempted to measure non-compliance through rule promulgation (e.g., Kennedy 2015; Bertelli and Doherty 2017). Prediction 4. All else equal, for insulated agencies, as the policy disagreement between the President and the relevant oversight committee increases, bureaucratic non-compliance with presidential directives should increase (Proposition 1). The prediction highlights an important, secondary consequence of disagreement between Congress and the executive. Whereas Howell (2003) shows disagreement tends to decrease the frequency of unilateral action, this suggests the same decreases its effectiveness—since insulated bureaucrats become less likely to implement such action. Prediction 5. All else equal, presidential directives should delegate more often as the cost of developing policy in-house increases (Proposition 3). Although operationalizing the cost of policy development in the presidential branch presents a clear measurement challenge, several contributing variables seem apparent. For example, the resource cost might be thought of as the relative proportion of the president’s institutional apparatus that the policy requires to be produced. For that reason, one might look to WHO/EOP budgetary authorizations and staff levels. Additionally, since administrations tend to accumulate policy objectives over time, the resource cost might be extrapolated from the president’s time in office. 6. Conclusion I sought to address two general limitations of theories of executive unilateralism. First, these theories may say more about the frequency of attempts at unilateral action, rather than their efficacy. We observe the issuance of a directive, but theories often cannot explain whether (and why) that order will translate into a preferred outcome. Second, taken to its logical conclusion—the theory aggrandizes the power of chief executives. Presidential policymaking is remarkably contingent, both as a result of constitutional checks and the inherent challenge of wielding the administrative state. The model suggests that unilateral action is dependent on bureaucratic agency in several key ways. Put simply, bureaucrats enable legislators to influence final outcomes. This result provides the basis for future empirical work looking to study a variety of other political phenomena: bureaucratic non-compliance, congressional oversight, and resources in the presidential branch. Moreover, the model reveals an important secondary consequence of agency insulation—that the threat of non-compliance effectively prevents the president from determining when in-house policymaking would be optimal. Counterintuitively, this means directives targeted to insulated agencies should delegate more authority. This leads to a number of directions for future work. As the previous section laid out, the model motivates a variety of possible empirical studies. While the model itself incorporates some important components of presidential efforts to implement policy “unilaterally”—these are worth further study. For instance, the model demonstrates that even a relatively weak legislative actor may meaningfully influence agents implementing presidential policy. But the “in-house” resources of the president are partly a function of congressional preferences. This is because presidential institutionalization (e.g., WHO and EOP resources) is known to be dependent on congressional efforts to build their own expertise and capacity (e.g., Krause 2002; Dickinson and Lebo 2007). In other words, the model may understate legislative influence, since legislators are capable of influencing the cost of policy development for the executive. In addition, future work might analyze the implications of a more complex definition of insulation—since agencies meaningfully vary in ways not captured by the relative capacity of elected institutions to sanction them. Another direction for future study is to incorporate features of delegation by chief executives within the broader framework of separation of powers in order to address the original question empirically evaluated by Howell (2003): when will presidents act alone? Ultimately, the key dependencies revealed by the model likely impact the president’s initial decision to act—and indeed, legislatures’ decisions to veto those actions. Indeed, this notion is more broad than presidential attempts to change policy. Scholarship on policy development in legislatures also tends to ignore the “shadow of implementation”—which surely influences their ability to enact policy. Thus, more generally, integrating this siloed work is both appropriate and promising. For helpful comments and suggestions, I thank Randy Calvert, Justin Fox, Sean Gailmard, Thomas Gray, Will Howell, Jeff Jenkins, George Krause, Dave Lewis, Eric Patashnik, Rachel Potter, and Craig Volden. Previous versions were presented at the 2015 Southern California Law and Social Science Forum, Los Angeles, CA, the 2015 annual meeting of the Midwest Political Science Association, Chicago, IL, USA, and the 2016 annual meeting of the Southern Political Science Association, San Juan, PR, USA. Footnotes 1. The Secretary was permitted to “act through or with the aid of such public or private instrumentalities or persons as he may designate,” to “determine and prescribe terms and conditions of employment under which the plants […] shall be operated,” and to issue regulations. 2. Moreover, this point is acknowledged by Howell (2005): “All presidents, and all politicians, struggle to ensure that those who work below them will faithfully follow orders” (433). 3. For the most recent review of this work, see Mayer (2009). 4. Work by Epstein and O’Halloran (1999) applied the conceptual distinction between policy development “in-house” and external delegation to congressional delegation. Thus, the model’s scope conditions are similar. As I argue in Section 5, the model applies to policies that can be conceivably developed “in-house” and imposed on agencies. That is, I do not assume that the WH/EOP has the staff, resources, and legal authority to carry out policy. By “WH” and “EOP,” I mean the collection of administrative agencies for whom the President submits a yearly budget under the “EOP of the President.” I consider cabinet departments, independent agencies, and government corporations external delegates. 5. This is particularly important, given the broad understanding of sanctioning behavior I put forth. Drafting punitive legislation and holding hearings occur at the committee level, such that assuming C to be the congressional floor median may inappropriately limit the threshold of political support that determines whether Congress engages in sanctioning activity. 6. I define “discretion” (like Epstein and O’Halloran 1999) as delegated authority, together with the severity of procedural and oversight constraints placed on that authority. 7. This is akin to the “subversion cost” highlighted by Gailmard (2002), wherein Congress imposes a cost via investigation. 8. Following Huber and McCarty (2004), agency punishments (θ) are fixed and exogenous. 9. Here, “line agency” refers to any department or agency outside the EOP that implements public policy. 10. Note, however, this category need not be defined on institutional features alone. As Lewis (2016) demonstrates, a majority of agencies—even those that do not bear the institutional hallmarks of insulation—report that Congress has more influence on agency spending post-appropriation. That is, agencies tend to perceive Congress as the principal to whom they must be more responsive. Other recent work suggests that the president’s influence is limited to agencies explicitly prioritized by their administrations (Potter and Shipan 2017). 11. The main results of model are not sensitive to the choice between linear and quadratic utility. Although I am largely agnostic as to whether the actors are risk averse (e.g., Bendor and Meirowitz 2004), linear utility produces results and comparative statics which are more readily interpretable. 12. E.O. 11930—“Performance by the Federal Energy Office of Energy Functions of the Federal Energy Administration” (July 30, 1976). Note that this order is included in most studies which regress macro-political characteristics on counts of executive orders. 13. “E.O. 9617—Transfer of Certain Agencies and Functions to the Department of Labor,” (September 19, 1945). 14. November 20, 2014. Quoted in Shabad (2014). 15. Marcos, Cristina. 2015. “Republican proposes abolishing the ATF amid bullet ban controversy,” The Hill, March 5. The ATF has since withdrawn the proposal. 16. Naturally, without sufficiently high d, delegation loss is entirely a function of ω, and there is no sanction loss. 17. Not surprisingly, uninsulated agencies comply because congressional threats pale in comparison to the cost of presidential sanction. 18. Note that in Proposition 1, optimal discretion is the second term of each compliance threshold. These values are discussed in Proposition 2 and proved in the Appendix. 19. The results for uninsulated agencies replicate a standard finding in the delegation literature, with one intuitive (but notable) exception. Uninsulated, proximate agencies receive less discretion in equilibrium as the president and legislative committee diverge. This implies an exception to the “ally principle,” which suggests that principals ought to rely more on ideologically proximate agents. The solid line in Figure 3 plots this relationship. Here, the agent and the president have similar policy preferences—yet, ideological divergence with a sufficiently powerful committee results in less discretion. In this case, the president’s ally cannot be expected to act in the president’s best interest because it serves two masters. This highlights a broader point in the context of “unilateral action.” Even when Congress is powerless to overturn a presidential directive outright, it can change outcomes by influencing those who implement the president’s policies. 20. Note, following Epstein and O’Halloran (1999), all results hold subject to the R>xi assumption. 21. In general, the threshold increases the higher the ideological distance, and decreases the greater the policy uncertainty. Moreover, the threshold is strictly lower for proximate agencies compared with distal ones. In other words, delegation is more likely to make the president better off over developing policy in the WH apparatus. 22. “Obama admits US gun laws are his ‘biggest frustration’,” BBC, July 24, 2015. http://www.bbc.com/news/world-us-canada-33646704. 23. McNeill (2016) referring to a 1995 statement by the American College of Physicians. 24. See, for example, “Notice of Legisaltive Mandate for Fiscal Year 2012,” NIH. http://grants.nih.gov/grants/guide/notice-files/NOT-OD-12-034.html. 25. Chris Cox, Executive Director of the NRA’s Institute for Legislative Action (NRA-ILA), wrote in a 2015 Op-Ed for Politico that “the National Rifle Association is not opposed to research that would encourage the safe and responsible use of firearms and reduce the numbers of firearm-related deaths. Safety has been at the core of the NRA’s mission since its inception. But that is not the goal of the gun control advocates who are behind the calls for CDC funding.” goo.gl/1PfiKD. 26. Memorandum of January 16, 2013. “Engaging in Public Health Research on the Causes and Prevention of Gun Violence,” 78 Federal Register 4295–4296. 27. “FY 2013 Full Year CR Operating Plan,” CDC and Prevention. https://www.cdc.gov/budget/documents/fy2013/fy-2013-operating-plan.pdf. 28. As Stephen Teret, director of the Center for Law and the Public’s Health at the Johns Hopkins Bloomberg School of Public Health noted, “even in the absence of an additional $10 million, CDC could have said, ‘Let’s reallocate some of our spending’.” Quoted in Rubin (2016). 29. http://dhss.delaware.gov/dhss/dms/files/cdcgunviolencereport10315.pdf. 30. See, for example, the Consolidated Appropriations Act of 2010 (P.L. 111–117; December 16, 2009). 31. HSPD were first issued in October 2001. The first such directive defines an HSPD as “record[ing] and communicat[ing] presidential decisions about the homeland security policies of the United States.” Thus, their basic function is indistinguishable executive orders and presidential memoranda (HSPD-1, October 29, 2001). 32. HSPD-12, August 27, 2004. url: goo.gl/G6xzMl. 33. Q1 FY14—HSPD-12 Public Report Summary, url: https://goo.gl/5YCjha. 34. The directive mandated the new forms of identification include several features: (1) verification of individual identity prior to issuance, (2) fraud resistance, (3) “rapid authentication,” and (4) verification of vendor through an accreditation process. 35. Author interview with former HSPD-12 Program Director, NASA, May 11, 2015. 36. The order’s provisions contain several delegations of authority. First, it tasks the Secretary of Commerce (in consultation with other agency heads) with setting the specific government-wide standard to fulfill the technical requirements. Second, it tasks all agencies with developing their own program to meet that standard. 37. A 2014 report indicates several of these agencies remain at or below 50% coverage for their federal contractors: the Federal Labor Relations Authority, Federal Maritime Commission, Federal Communications Commission, Commodity Futures Trading Commission, Merit Systems Protection Board, US Commission on Civil Rights, Small Business Administration, National Credit Union Association, and the Department of Health and Human Services (Q1 FY14—HSPD-12 Public Report Summary). 38. Sturgill, Emily. 2007. “New Badge Enrollments on the Rise; Benefits Too,” Glenn News, url: http://www.nasa.gov/centers/glenn/news/AF/2007/Nov07_HSPD-12HQ.html. 39. Author interview with former HSPD-12 Program Director, NASA, May 11, 2015. Appendix Lemma 1 For discussions of this result see Epstein and O’Halloran (1999: 248), Volden (2002a: 113), or Gailmard and Patty (2007: 879). Lemma 2 Given the discussion, its proof is omitted. Proposition 1 Given the discussion, its proof is omitted. Proposition 2 Given Lemmas 1 and 2, the President’s maximization problem reduces to one of five scenarios. For uninsulated agencies, when xA > xM and θs<|xA−xM|,   maxdEUP=−∫−RxA−d(ω+d)12Rdω−∫xA−dxA+dxA12Rdω−∫xA+dR(ω−d)12Rdω, which reduces to   maxdEUP=xA2−2xAd−(d−R)22R. Since ∂EUP∂d=R−xA−dR,d*=R−xA. When xA < xM and θs>|xA−xM|,   maxdEUP=−∫−RxA−d(ω+d)12Rdω−∫xA−dxA+mxA12Rdω−∫xm−dxm+dxM12Rdω−∫xm+dR(ω−d)12Rdω, which reduces to   maxdEUP=3xA2−2d2−2xAxm−4dxm+xm2+4dR−2R24R. Since ∂EUP∂d=−R−d−mR,d*=R−xm. For insulated agencies, when θs−θv>xA and xA > xm,   maxdEUP=−∫−Rxm−dω12Rdω−∫xm−dxA−d(w+d)12Rdω−∫xA−dxA+dxA12Rdω−∫xA+dR(w−d)12Rdω, which reduces to   maxdEUP=d(R−2xA−d+xm)2R. Given that ∂EUP∂d=R+xm−2xA2R,d*=R+xm−2xA2. Conversely, when θs−θv>xA and xA < xm,   maxdEUP=−∫−RxM−dω12Rdω−∫xM−dxM+dxM12Rdω−∫xM+dR(ω−d)12Rdω, which reduces to   maxdEUP=−d(xm+d−R)2R, and ∂EUP∂d=−2d+xm−R2R,d*=R−xm2. Finally, when xA>θs−θv and xA < xm,   maxdEUP=−∫−Rθs−θv−dω12Rdω−∫θs−θv−dxA−d(ω+d)12Rdω−∫xA−dxA+dxA12Rdω−∫xA+dR(ω−d)12Rdω, which reduces to   maxdEUP=−d(2xA+d−R−θs+θv)2R and   ∂EUP∂d=−2xA+2d+θs−θv+R2R. Given ∂EUP∂d=0, this equation has one solution: d*=R−2xA+θs−θv2. ▪ Proposition 3 In general, the President delegates when the following condition is satisfied   EUP(e=1)>EUP(e=0). (1) This means, in general, substitution of d*in EUP for each of the five cases above. First, I analyze the two cases of uninsulated agencies. When xA > xM and θs<|xA−xM|,   EUP(e=1)=xA2−2xAd*−(d*−R)22R, which, given d*=R−xA, reduces to the following comparison. If the following inequality holds, the President does not delegate.   RxA−xA2R>τ. When xA < xM and θs>|xA−xM|,   EUP(e=1)=3xA2−2d*2−2xAxm−4d*xm+xm2−4Rd*−2R4R, which, given d*=R−xA, reduces to the following comparison. If the following inequality holds, the President does not delegate,   4Rxm+2xAxm−3xA2−3xm24R>τ. The three cases of complete delegation for insulated agencies appear below. When θs−θv>xA and xA > xm,   EUP(e=1)=d*(−2xA−d*+xm+R)R. Given d*=R+xm−2xA2R, this equation reduces to zero. Thus, so long as τ is positive, the non-delegation inequality (Equation (1)) cannot be satisfied. When θs−θv>xA and xA < xm,   EUP(e=1)=−d*(d*+xm−R)2R, which, given d*=R−xm2, reduces to (xm−R)28R and is strictly positive so that Equation (1) cannot be satisfied. When xA>θs−θv and xA < xm,   EUP(e=1)=−d*(2xA+d*−R−θs+θv)2R and given d*=R−2xA+θs−θv2, this reduces to the following comparison   (R−2xA+θs−θv)28R<−τ, which, since the left side is strictly positive, cannot be satisfied. So the president always delegates to insulated agencies. ▪ References Bawn Kathleen. 1995. “Political Control versus Expertise: Congressional Choices about Administrative Procedures,” 89 American Political Science Review  62– 73. Google Scholar CrossRef Search ADS   Bendor Jonathan, Meirowitz Adam. 2004. “Spatial Models of Delegation,” 98 American Political Science Review  293– 310. Google Scholar CrossRef Search ADS   Bertelli Anthony M., Doherty Kathleen. 2017. “Setting the Regulatory Agenda: Statutory Deadlines and the Conditions for Compliance.” Working Paper , 1– 25. Bolton Alexander, Thrower Sharece. 2016. “Legislative Capacity and Executive Unilateralism,” 60 American Journal of Political Science  649– 63. Google Scholar CrossRef Search ADS   Burke John P. 1992. The Institutional Presidency . Baltimore, MD: Johns Hopkins University Press. Chiou Fang-Yi, Rothenberg Lawrence S.. 2014. “The Elusive Search for Presidential Power,” 58 American Journal of Political Science  653– 68. Google Scholar CrossRef Search ADS   Dickinson Matthew J. 2009. “We All Want a Revolution: Neustadt, New Institutionalism, and the Future of Presidency Research,” 39 Presidential Studies Quarterly  736– 70. Google Scholar CrossRef Search ADS   Dickinson Matthew J., Rudalevige Andrew. 2004. “Presidents, Responsiveness, and Competence: Revisiting the ‘Golden Age’ at the Bureau of the Budget,” 119 Political Science Quarterly  633– 54. Google Scholar CrossRef Search ADS   Dickinson Matthew J., Lebo Matthew. 2007. “Reexaming the Growth of the Institutional Presidency, 1940–2000,” 69 Journal of Politics  206– 19. Google Scholar CrossRef Search ADS   Epstein David, O’Halloran Sharyn. 1999. Delegating Powers: A Transaction Cost Politics Approach to Policy Making under Separate Powers . Cambridge, NY: Cambridge University Press. Google Scholar CrossRef Search ADS   Gailmard Sean. 2002. “Expertise, Subversion, and Bureaucratic Discretion,” 18 Journal of Law, Economics, and Organization  536– 55. Google Scholar CrossRef Search ADS   Gailmard Sean, Patty John W.. 2007. “Slackers and Zealots: Civil Service, Policy Discretion, and Bureaucratic Expertise,” 51 American Journal of Political Science  873– 89. Google Scholar CrossRef Search ADS   Howell William G. 2003. Power without Persausion: The Politics of Direct Presidential Action . Princeton, NJ: Princeton University Press. Howell William G. 2005. “Unilateral Powers: A Brief Overview,” 35 Presidential Studies Quarterly  417– 39. Google Scholar CrossRef Search ADS   Huber John D., McCarty Nolan M.. 2004. “Bureaucratic Capacity, Delegation, and Political Reform,” 98 American Political Science Review  481– 94. Google Scholar CrossRef Search ADS   Kennedy Joshua B. 2015. “‘Do This! Do That!’ and Nothing Will Happen: Executive Orders and Bureaucratic Responsiveness,” 43 American Politics Research  59– 82. Google Scholar CrossRef Search ADS   Krause George A. 2002. “Separated Powers and Institutional Growth in the Presidential and Congressional Branches: Distinguishing between Short-Run versus Long-Run Dynamics,” 55 Political Research Quarterly  27– 57. Google Scholar CrossRef Search ADS   Krause George A. 2009. “Organizational Complexity and Coordination Dilemmas in U.S. Executive Politics,” 39 Presidential Studies Quarterly  74– 88. Google Scholar CrossRef Search ADS   Krause George A., Dupay Brent M.. 2009. “Coordinated Action and the Limits of Presidential Control over the Bureaucracy: Lessons from the George W. Bush Presidency,” in Provost Colin, Teske Paul, eds., President George W. Bush’s Influence over Bureaucracy and Policy: Extraordinary Times, Extraordinary Powers . 81– 104. New York, NY: Palgrave. Google Scholar CrossRef Search ADS   Kriner Douglas L., Reeves Andrew. 2015. The Particularistic President: Executive Politics and Political Inequality . Cambridge University Press. Lewis David E. 2003. Presidents and the Politics of Agency Design . Stanford University Press. Lewis David E. 2008. The Politics of Presidential Appointments: Political Control and Bureaucratic Performance . Princeton University Press. Google Scholar CrossRef Search ADS   Lewis David E. 2016. “Political Control and the Presidential Spending Power.” Working Paper. Mayer Kenneth R. 2001. With the Stroke of a Pen: Executive Orders and Presidential Power . Princeton University Press. Mayer Kenneth R. 2009. “Going Alone: The Presidential Power of Unilateral Action,” in Howell William G., Edwards, eds George C., The Oxford Handbook of the American Presidency , 427– 98. Oxford University Press. McNeill Matthew B. 2016. “Gun Violence: A Public Health Concern?” Clinical Correlations  June 9. Moe Terry M. 1985. “The Politicized Presidency,” in Chubb J. E., Peterson P. E., eds., The New Direction in American Politics . 235–272. Washington, DC: Brookings Institution Press. Moe Terry M. 1989. “The Politics of Bureaucratic Structure,” in J. E. Chubb and P. E. Peterson, eds., Can the Government Govern ? 267–330. Washington, DC: Brookings Institution Press. Moe Terry M., Howell William G.. 1999. “The Presidential Power of Unilateral Action,” 15 Journal of Law, Economics, and Organization  132– 79. Google Scholar CrossRef Search ADS   Nathan Richard P. 1983. The Administrative Presidency . Prentice Hall. Neustadt Richard E. 1960. Presidential Power and the Modern Presidents: The Politics of Leadership from Roosevelt to Reagan . Wiley & Sons. Potter Rachel Augustine, Shipan Charles R.. 2017. “Agency Rulemaking in a Separation of Powers System,” Journal of Public Policy  1– 25. https://doi.org/10.1017/S0143814X17000216. Rogowski Jon C. 2016. “Presidential Influence in an Era of Congressional Dominance,” 110 American Political Science Review  325– 41. Google Scholar CrossRef Search ADS   Rubin Rita. 2016. “Tale of 2 Agencies: CDC Avoids Gun Violence Research but NIH Funds It,” 315 Journal of the American Medical Association  1689– 92. Google Scholar CrossRef Search ADS PubMed  Rudalevige Andrew. 2009. “The Administrative Presidency and Bureaucratic Control: Implementing a Research Agenda,” 39 Presidential Studies Quarterly  10– 24. Google Scholar CrossRef Search ADS   Rudalevige Andrew. 2012. “Executive Orders and Presidential Unilateralism,” 1 Presidential Studies Quarterly  138– 60. Google Scholar CrossRef Search ADS   Selin Jennifer L. 2015. “What Makes an Agency Independent?,” 59 American Journal of Political Science  971– 87. Google Scholar CrossRef Search ADS   Volden Craig. 2002. “A Formal Model of the Politics of Delegation in a Separation of Powers System,” 46 American Journal of Political Science  111– 33 Google Scholar CrossRef Search ADS   © The Author(s) 2018. Published by Oxford University Press on behalf of Yale University. All rights reserved. For Permissions, please email: journals.permissions@oup.com

Journal

The Journal of Law, Economics, and OrganizationOxford University Press

Published: Mar 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off