Abstract
How do presidents decide to delegate? Research on unilateral powers has (often informally) assumed that executive orders start and end with the president, but as in all principal-agent relationships, executives rely on subordinates to carry out their directives. Political scientists have explored when and why Congress delegates to the president, but such action often necessitates a response from the White House; namely, the president delegates to the bureaucracy the authority delegated to him by Congress. Bureaucrats may in turn have a wide degree of latitude when implementing the president’s orders. I argue that presidents must make strategic decisions in choosing to whom to delegate, much in the same way that Congress makes certain calculations in deciding whether to delegate to the executive branch.
Introduction
The title of William Howell’s 2003 book on executive orders, Power Without Persuasion, is illustrative when one considers the politics of presidential unilateral powers. The book draws its title from Richard Neustadt’s (1990) seminal work on the presidency, in which Neustadt argues that a president’s chief asset in securing a desired policy outcome is through persuasion. As the author writes, “despite his status, [the president] does not get action without argument” (Neustadt, 1990: 11). Howell’s book is intended to establish that this is not the case; rather, presidents have methods of directly controlling various governmental actors through unilateral powers such as the executive order.
However, executive orders do not have to contain only directives intended to influence the behavior of agencies. Frequently, these orders involve explicit decisions to delegate presidential authority (usually granted by statute) to the head of an agency. Previous scholarship has investigated the conditions under which Congress will cede authority to the president (e.g. Epstein and O’Halloran 1994, 1999), but such action does not always end at the White House doorstep. Occasionally, there is a second step in which the president decides whether and where to delegate authority that was itself granted to him by the legislature. This paper asks this central research question: why do presidents delegate? I argue that numerous agency characteristics as well as legislative factors drive the president to make strategic decisions concerning delegation of authority that has been granted by Congress. The paper proceeds as follows: I briefly review previous scholarship on executive orders and the politics of delegation, then advance a number of hypotheses to explain what factors influence the presidential decision to delegate. The hypotheses are tested using data from 1989 through to the end of Barack Obama’s first term. The results indicate support for applying a transaction costs theory, similar to that advanced by Epstein and O’Halloran (1999) with respect to legislative delegation, to intra-executive delegation.
Executive orders in brief
Presidential scholarship in the past decade or so has increasingly focused on the president’s ability to make policy unilaterally (e.g. Cooper, 2002, 2005; Howell, 2003; Kelley and Marshall, 2008, 2009, 2010; Mayer, 1999, 2001). One particularly fruitful research track looks to executive orders, a tool of unilateral policy-making, to explore the development of presidential power (e.g. Deering and Maltzman, 1999; Howell, 2003; Krause and Cohen, 1997, 2000; Mayer, 1999, 2001). To summarize, executive orders are directives most commonly addressed to the bureaucracy that set policy or require the implementation of new rules and regulations. The orders can be exceedingly broad (applying to every agency in the executive branch of government) or quite narrow (focused only on a sub-agency of a specific department). Such orders can attract significant attention depending on their subject matter (e.g. Howell, 2003; Mayer, 2001) or can go virtually unnoticed, in part because of the minute details of policy that the orders sometimes deal with. These orders have their limits: they can be overturned by succeeding presidents, Congress, or the courts (e.g. Howell, 2003), and executive agencies play a vital role in their formulation and implementation, even having the means to resist presidential directives if they so choose (e.g. Kennedy, 2015; Rudalevige, 2012).
Executive orders are highly variable in their content. Here, my interest is in those orders that result from Congress giving the president explicit authority to implement specific provisions of a statute. A good example of this is Richard Nixon’s (1971) EO 11615, which was designed to put wage and price controls in place based on the power granted to the president by the Economic Stabilization Act of 1970. Directives such as this one typically involve very explicit delegations of authority to agencies or agency heads. This project focuses particularly on the president’s decision about when and why to delegate statutorily granted authority. To date, little research has been conducted on the utility of executive orders as a tool of delegation. Therefore, in the next section I will discuss the politics of delegation and how they apply to the presidency.
The politics of delegation
Scholars have approached the issue of political delegation from a number of different perspectives; some of the most prominent research on delegation principally involves relations between Congress and the executive branch (e.g. Epstein and O’Halloran, 1999; McCubbins et al., 1987, 1989), and much of what we know about delegation is explicitly rooted in principal-agent theory (as discussed, for example, by Miller (2005)). Epstein and O’Halloran’s (1999: 43) study on Congress’s decision to delegate authority to the executive branch is a strategic one; the authors argue that “Congress will delegate to the executive when the external transaction costs of doing so are less than the internal transaction costs of making policy through the normal legislative process.” My purpose in this paper is to adapt this transaction costs theory of Epstein and O’Halloran to the context of the president’s decision to use executive orders as a tool of intra-executive delegation. The central theoretical assumption is that presidents are strategic and will consider numerous relevant factors before delegating.
Ideological distance between the president and the agency is one such factor. In general, a conservative president should be less likely to delegate to a liberal agency than to a conservative agency, as ideological differences have the potential to enhance the principal-agent problem and increase the likelihood of agent defection (see, for example, Miller (2005) for a review). An ideologically incongruent agency should give strategic presidents more pause than an ideologically congruent agency when it comes to delegation.
But I argue further that the extent to which ideological distance matters for the president is moderated by another variable: political insulation. This concept refers to the agency’s independence from political processes and its general autonomy, or, as Epstein and O’Halloran (1999: 155) put it, its independence “from both legislative scrutiny and executive interference.” The principal-agent problem should be magnified when the president cannot exert control; a conservative president may be particularly reticent to delegate to a liberal agency that is well outside the bounds of their control, but may still delegate to a liberal agency that can be overseen. That is, when an agency is not insulated, presidents may worry less about ideological considerations because they know they have some control over the organization. Similarly, the president may delegate to an agency outside of White House control so long as that agency shares the president’s preferences. This suggests an interactive hypothesis:
Hypothesis 1: As ideological distance between the president and an agency increases, the probability of presidential delegation decreases; this effect is particularly pronounced in agencies that are insulated rather than in agencies that are not insulated.
Further, I expect Congress to play a role in the president’s delegatory decisions. Agencies are not subjected to a single principal; they must answer to a number of political actors, including the legislature, which has its own vested interests in bureaucratic outcomes (e.g. McCarty, 2004; Moe, 1985, 1989; Whitford, 2005). Here I expect that a strategic president may hesitate to delegate authority under conditions of divided government. Knowing that an ideologically incongruent Congress has greater incentives to rein in agency behavior, presidents ought to tamp down the use of delegatory orders that give agencies latitude and instead adopt a more command-oriented approach through their directives. This leads to the second and final hypothesis.
Hypothesis 2: Presidents will be less likely to delegate in times of divided government to any agency than they will be in times of unified government.
Data and methods
To test these hypotheses, I collected data on all executive orders from 1989 through to the end of Barack Obama’s first term (American Presidency Project, 1999–2016). These data allow for variation across numerous presidencies and Congresses of varying partisan makeups and ideologies. The first step in the process is to identify to which agencies a given executive order applies. Generally, this is a straightforward proposition because the relevant agencies are at least named at some point in the text of the executive order. Occasionally, to whom the order applies is quite vague, as the president may decide to issue a blanket order that applies to the entire federal bureaucracy or may choose to except a small subset of agencies from the order while simultaneously applying it to all others. These orders are subsequently identified and controlled for.
The second step is paring the data down for a workable subset of cases. I compile a master list of agencies by referring to work done by Clinton et al. (2012), who researched mean ideal point estimates for 24 different organizations within the executive branch. 1 With this information in hand, I determined which executive orders applied to which of the 24 agencies and coded each as a separate observation. Thus, for example, a single directive that applies to the Department of Justice, the Department of State, and the Department of the Treasury constitutes three observations in the dataset.
I then determined whether the president explicitly delegated statutorily granted authority to an agency through the executive order. For example, consider the following portion of EO 12517 from Ronald Reagan (1985): “All functions vested in the President by the United States-India Fund for Cultural, Educational, and Scientific Cooperation Act…are delegated to the Secretary of State.” This is easily classified as a delegatory order given that the president expressly provides for the transfer of some of his authority, granted by congressional act, to another officer of government. Other executive orders are slightly more vague, but still easily classified, like the following order from George W Bush (2001; emphasis not in original): “The Secretary of the Treasury…is hereby authorized to take such actions, including the promulgation of rules and regulations, and to employ all powers granted to the President by IEEPA and UNPA, as may be necessary to carry out the purposes of this order.”
Although this order does not contain any form of the word “delegate,” it is still clearly transferring authority from the president to a federal agency.
Central independent variables include, first, the level of bureaucratic insulation. To measure this, I turn to work by Selin (2015), who devises estimates of this concept based on two separate categories: essentially, the independence of the agency heads from presidential influence as well as the agency’s policy-making autonomy. 2 I use only the variable representing the independence of agency heads to classify insulation on the understanding that those individuals subject to direct sanction from the president should feel more compelled to obey White House directives than those whose employment is more secure. Selin (2015: 977) constructs this measure by examining “each legal mandate that places limitations on who may serve in an agency’s or a bureau’s key leadership positions,” estimating via a “Bayesian latent factor model.” The variable ranges from a minimum in my model of −0.781 (indicating low insulation) to a maximum of 1.815 (indicating high insulation).
I construct the ideological distance measure based on work done by Clinton and Lewis (2008), who construct ideal points for a number of agencies based on expert surveys used to obtain statistical estimates. In this case, when an estimate is negative, the agency is considered liberal, and when it is positive, the agency is conservative. I classify an agency on a three-point scale from −1 (Liberal) to +1 (Conservative) if the Clinton–Lewis estimates are significantly different than 0. Estimates whose confidence intervals cross 0 result in the agency being classified as moderate. Ideological distance is measured as the absolute value of the agency’s assigned score from the absolute value of the president’s assigned score, and this variable ranges from a value of 0 (ideological congruence) to 2 (ideological incongruence). 3 Table 1 presents a list of the 24 agencies studied herein and their ideological classifications.
List of agencies and their ideological classifications.
I also account for the president’s popularity as measured by Gallup to control for the possibility that presidents vary their executive order usage based on public standing (e.g. Mayer 1999, 2001). Additional controls include dummy variables for whether the executive order was principally advisory in nature (i.e. the order creates an advisory council or task force and other agencies are involved only insofar as their relationship with the council requires), whether the executive order is a blanket directive (i.e. it is directed to all agencies of the federal government), and a dummy variable for the Department of the Treasury, which is a frequent target for delegation because of a number of orders dealing with economic sanctions directed to certain countries. Finally, president-specific fixed effects are included. The model is estimated using logistic regression with standard errors clustered by agency.
Results and discussion
I estimate two models below to account for some critical variation in executive orders. Model 1 includes all executive orders in the sample, but this runs the risk of missing key information because so many directives are of the blanket variety. This inflates the number of orders applicable to different organizations. To account for this, I estimate a second model that includes only those agencies explicitly named in the order. Table 2 presents the results.
Logistic regression of delegation in executive orders, 1989–2012.
p<0.05; bp<0.01; cp<0.001. Clustered standard errors in parentheses. Ldr: Agency Leadership; EO: Executive Order.
The results in Model 1 indicate some support for Hypothesis 1 (interaction between insulation and ideological distance) and no support for Hypothesis 2 (divided government). Though the interaction term is itself insignificant, the significance of one of the component terms merits a closer look at the effects. 4 Results do lend some support to Hypothesis 1, though only in certain circumstances. At either very low or very high levels of insulation, the marginal effect of ideological distance is not statistically significant, and changes we do observe otherwise are relatively small (always less than 0.01). This is demonstrated in Figure 1.
Marginal effects of ideological distance on the probability of delegation in executive orders, by insulation.
Some of the additional controls do exert independent effects. The marginal effect of orders dealing principally with the creation of advisory committee on the probability of delegation is around −0.03 and the marginal effect of blanket directives is approximately −0.04. Finally, the marginal effect of an order to the Treasury, as expected, is positive and significant (0.06).
The effect sizes with respect to Hypothesis 1 in Model 1 are comparatively small given the universe of executive orders in which I am dealing. Model 2, which focuses only on orders where the agency was explicitly named presents a much clearer picture. 5 In this case the interactive and both constituent terms are statistically significant, and when calculating marginal effects, the changes are considerably starker. Figure 2 depicts the marginal effects of ideological distance on the probability of delegation across various levels of insulation for orders in which an agency has been explicitly named.
Marginal effects of ideological distance on the probability of delegation in executive orders (named only), by insulation.
Ideological distance does appear to be a negative predictor of delegation, though this is not the case in agencies that are comparatively non-insulated. The marginal effects indicate that at the lowest levels of agency insulation, ideological distance has no effect on the probability that the president will delegate. However, as agencies become more autonomous the probability of delegation drops precipitously as ideological distance increases. At one standard deviation below the mean of insulation, ideological distance has no effect; at the mean level of insulation, the marginal effect of ideological distance is approximately −0.03, and at one standard deviation above the mean level of insulation, this marginal effect is approximately −0.06. There is thus some support for Hypothesis 1: the effects of ideological distance are moderated by insulation, though in contrast to the hypothesis, ideological distance does not always predict less delegation in agencies that are not insulated from presidential control.
It is also worth noting that insulation is not a universally negative predictor of the probability of delegation. As a matter of fact, examining the marginal effects of insulation show that it is a significantly positive predictor (0.05) of delegation in agencies that share the president’s ideology. This finding may be explained by the fact that those agencies that are not insulated are agencies less able to institute significant policy change, and thus presidents are slightly less likely to delegate to them because of their structure. Presidents also do not seem reticent to delegate to agencies far outside their control, so long as those organizations share the president’s ideological proclivities. Remaining control variables behave largely as they did in Model 1 with the exception of blanket orders, which are no longer predictive in Model 2.
Conclusion
Much of what we know about the politics of delegation up to this point principally concerns legislative delegations of authority to the executive branch. In this paper I have sought to apply existing theories of strategic delegation to a different context: intra-executive delegation (i.e. delegation from the president to other actors within the executive branch). The results indicate that, much like Congress does when delegating to the White House, presidents are strategic in deciding when to authorize the use of executive authority by other actors. Why, then, do presidents delegate? The answers are numerous and varied. Presidents seem more reluctant to delegate to agencies that are further ideologically from their ideal, and this tendency is exacerbated as agency heads become more insulated from presidential control. It also appears that presidents are willing to lose some control by delegating to insulated agencies so long as those organizations are ideologically compatible with the White House’s own broad policy goals. Divided government does not predict the use of delegation in executive orders, however, suggesting that the strategic decision on the part of the president depends more on agency characteristics than what is occurring in Congress. Executive orders are not the only tool presidents have with which to delegate however, and while this research provides a glimpse into the strategic presidency, more work is needed to determine the full causes and consequences of delegation.
Further, more research on the utility of executive orders and other unilateral powers (i.e. how they are applied by the agencies to which they are directed) is needed as the field of presidential scholarship continues to make strides toward more comprehensive theories of presidential power. Presidents may be delegating their authority under some circumstances precisely because they have concerns about how a given order will be carried out; perhaps anticipation of non-compliance motivates the executive to make the issue “someone else’s problem.” In any case, it is not sufficient to argue that unilateral powers are important simply because the president can take those actions completely independently of Congress. Establishing the importance of such powers necessarily depends upon evaluating their effectiveness, which remains an ongoing challenge in the realm of executive branch research.
Footnotes
Acknowledgements
A previous version of this paper was presented at the 2013 Annual Meeting of the Midwest Political Science Association. The author wishes to thank Kenneth Lowande, Jose D. Villalobos, and the anonymous reviewers for their helpful comments. All errors are mine.
Declaration of conflicting interest
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Notes
Carnegie Corporation of New York Grant
The open access article processing charge (APC) for this article was waived due to a grant awarded to Research & Politics from Carnegie Corporation of New York under its ‘Bridging the Gap’ initiative.