CARDOZO LAW REVIEW [Vol. 20:947 1999]
THE NONDELEGATION DOCTRINE
SEPARATION OF POWERS:
A POLITICAL SCIENCE APPROACH
David Epstein* &
The passage of the Line Item Veto Act (“LIVA”)
and its subsequent demise at the hands of the United States Supreme
Court have led to a resurgence of interest in the nondelegation
doctrine. LIVA granted the President unprecedented authority to cancel not only
specific spending items in appropriations bills, but certain targeted tax
benefits as well. Opponents of LIVA consistently argued that it infringed on
the separation of powers by delegating legislative authority to the Executive,
and they urged the courts to strike it down as a violation of the nondelegation
doctrine. LIVA’s supporters portrayed it as good public policy and a
permissible delegation of authority. Clinton v. City of New
York thus serves as a convenient lens through which to view
competing theories of the nondelegation doctrine in particular, and the role of
delegation in a system of separate powers more generally.
Even though LIVA was eventually overturned on grounds other than the
nondelegation doctrine, the majority opinion distinguished
Clinton from Field v. Clark by embracing what
Richard Stewart terms the “transmission belt” theory of
delegation. According to this view, Congress provides the
Executive with a policymaking algorithm, specifying which actions are to be
taken in which circumstances, leaving to the Executive only the duty of
determining the facts in a particular instance. “When enacting the
statutes discussed in Field, Congress itself made the decision to
suspend or repeal the particular provisions at issue upon the occurrence of
particular events subsequent to enactment, and it left only the determination
of whether such events occurred up to the President.” Thus,
the President did not legislate; Congress had made the fundamental
policy decisions itself, and subsequent to any given finding of facts, the
President was obligated to take a certain set of actions.
By taking this view of policymaking, the Court assumes away the problem
of uncontrolled delegation by denying the possibility that Congress ever
delegates legislative authority in the first place. All delegations according
to this theory are well-regulated, with Congress laying out the policymaking
program and the Executive merely filling in the details. The problem with LIVA,
by this reckoning, was that it gave the President discretionary authority to
change a previously enacted statute, one which made no mention of allowing the
Executive to fill in any details.
The Court thus skirted the issue of delegation in Clinton by
denying its existence outside of the case. Were this an accurate portrayal of
reality, the remainder of this Article would be superfluous. This convenient
stance, however, ignores the numerous cases in which the President and
executive agencies are given real policymaking discretion in their own right,
with either no statutory guidance or guidance that is so broad that it imposes
almost no constraints on executive actions. It also fails to consider other
instances of delegation, such as the Reciprocal Tariff Act of
1934, which allowed the President to amend explicit policy
prescriptions in previous acts of Congress.
Thus, proponents of a revitalized doctrine protest that congressional
delegation of authority to the Executive is real, and that it undermines the
very foundations of our representative government. Professors David Schoenbrod
and Marci Hamilton, for instance, argue in their amicus curiae brief for
Raines v. Byrd (an earlier version of Clinton v. City
of New York) that, in delegating authority to the Executive, “Congress
generally avoids accountability by leaving the hard policy choices to unelected
and unaccountable agencies.” Delegation thus leads to a
dangerous concentration of power in one branch of government, hidden from
public scrutiny and operating at the expense of the public good for the benefit
of a few well-placed individuals.
Peter Aranson, Ernest Gellhorn, and Glen Robinson similarly construe
delegation as a means of delivering private benefits to favored
constituents. They pose the problem as one of having legislation
spelled out in statutes, in which case it will be enforced by the courts or by
delegating legislative power to agencies. By having the executive branch fill
in the regulatory details, Congress can shift some degree of both the credit
and blame to the agency. If delegation helps Congress shift to the agency a
preponderantly large part of the blame, then it will prefer agency regulations
to judicially enforced statutes, and vice-versa if delegation would shift more
credit to executive agents.
Elsewhere, David Schoenbrod argues forcefully that Congress oversteps
its constitutional bounds when delegating broad discretion to executive
agencies; that Congress could find the time and resources necessary to write
detailed laws if it really wanted to; and that, in practice, delegation creates
perverse incentives that lead to such nonsensical policies as burning oranges
during the Great Depression. “Delegation allows Congress to
stay silent about [regulatory costs and benefits], so it severs the link
between the legislator’s vote and the law, upon which depend both
democratic accountability and the safeguards of liberty provided by Article
According to these arguments, LIVA was just the tip of the iceberg of
unconstrained, unconstitutional delegation to the Executive. Advocates of an
enhanced nondelegation doctrine fear that the concentration of power in one
branch of government inevitably leads to tyranny and a loss of individual
liberty, citing The Federalist No. 47 as affirmation: “The
accumulation of all powers, legislative, executive and judiciary, in the same
hands ... may justly be pronounced the very definition of
tyranny.” One way that power can come to reside within a
single branch is by the practice of Congress delegating the details of
policymaking to executive agencies. As Justice Kennedy stated in his concurring
opinion in Clinton, “Liberty demands limits on the ability of any
one branch to influence basic political decisions.... Abdication of
responsibility is not part of the constitutional design.”
Therefore, the argument goes, to ensure liberty it is necessary to limit such
delegations of authority.
We agree that unconstrained delegation does pose a threat to individual
liberties. We disagree, however, with the assumption that Congress does in fact
delegate, either de jure or de facto, unrestrainedly. Legislators
delegate authority in those areas — such as pork barreling in
appropriations bills, military base closings, and trade policy — where the
legislative process produces inefficient outcomes. Congress is also wary,
though, of ceding too much authority to executive branch actors who may pursue
their own policy goals rather than those of the enacting legislative coalition.
Legislators therefore set the limits of executive branch discretion so that
these costs and benefits of delegation balance at the margin. Thus, legislators
may well delegate authority to executive actors, but they will rarely, if ever,
do so without constraints. Moreover, legislators will delegate those issue
areas where the normal legislative process is least efficient relative to
regulatory policymaking by executive agencies.
In particular, we would argue that, in the case of LIVA, delegation
served as a vehicle for good public policy. The delegation of authority
involved was clearly defined, and, more importantly, was targeted specifically
at a problem that the normal legislative process handles poorly — the
curtailing of pork barrel benefits. Perhaps, in a perfect world, legislators
could have come to the same policy outcomes themselves. In reality, though,
they could only choose between two flawed alternatives — the committee
system or delegation to the Executive. Moreover, it is doubtful that
restricting Congress’ power to delegate would have engendered more fair or
more effective final policy outcomes, particularly given Congress’
predilections for logrolling and omnibus legislation.
If our theory of delegation as a balancing of competing inefficiencies
is correct, the balance of power between the branches will be continuously
recalibrated to reflect changing contingencies of the day. As political factors
— such as constituent demands, legislators’ policy goals, and
partisan control of the branches of government — change, so too will the
terms of delegation. In our view, this state of affairs is a testament to the
health of our political system, allowing neither committees nor agencies to
dominate policymaking. Congressional delegation is, therefore, a
self-regulating system, and any attempts to revive the nondelegation doctrine
would merely strengthen the hands of congressional committees, sub-committees,
and interest groups at the expense of agencies, thereby reducing accountability
and forcing Congress to make policy in exactly those areas that it handles
least effectively relative to executive agencies. Delegation should, thus, be
seen as a complement to, rather than a substitute for, the separation of
We develop our argument by first reviewing the political science
literatures on legislative organization and on congressional oversight of the
bureaucracy, two areas that serve as the building blocks for our theory of
delegation. Next, we explicate our argument relating to delegation and the
separation of powers and derive a set of testable propositions that, if our
approach is correct, should consistently predict variations in the amount of
authority granted to the executive branch over time and across issue areas.
This Article describes the data used to test these propositions, provides some
useful summaries of changes in executive statutory discretion in the postwar
era, and statistically tests our predictions. We conclude with a few remarks
about the nondelegation doctrine and the role of the courts in a system of
I. DELEGATION AND THE POLICYMAKING PROCESS
Attacks on delegation rest on three assumptions: first, legislators
delegate many important policy decisions to executive agencies; second, once
authority has been delegated, legislators show little interest in overseeing
its exercise; and third, restricting Congress’ ability to delegate will
improve the quality of public policy. Hence, delegation is equivalent to the
abdication of Congress’ legislative duties under Article I of the
Constitution, resulting in policy detrimental to the public good.
If this situation corresponded with reality, it would be clear that
delegation represents a severe departure from the policy-making process
originally envisioned by the Founders, and hence a threat to our system of
separate powers and the individual liberties that it was meant to protect.
However, we will present a theory of delegation that, if correct, contradicts
the predictions that follow from these three assumptions. First, though, we
briefly review the two areas of literature within political science upon which
our approach is built — those on the committee system and legislative
organization, and on the oversight of delegated authority. The former revolves
around the committee system and its impact on policy outputs, while the latter
focuses on the means by which 535 relatively uninformed legislators can monitor
the actions of a vast federal bureaucracy.
A. Legislative Organization
To understand policymaking through the committee system, it is
convenient to begin with David Mayhew’s book, The Electoral
Connection, which lays out a vision of congressional
organization as a rationally constructed, conscious choice made by
reelection-seeking individuals. Mayhew emphasized the importance of the
committee system as a basis for members to pursue their electoral goals through
the activities of advertising, position taking, and credit claiming.
Legislators, thus, use committees as platforms to translate their
constituents’ preferences into policy outcomes. Mayhew’s study
remains influential mainly due to its methodological focus: it took a ground-up
view of legislative organization and built upon this edifice a theory of public
The second strand of modern theories of Congress has its roots in what
has come to be known within political science as the new
institutionalism. Its key premise is that “institutions
matter,” meaning that in the details of legislative procedures —
including floor voting and amendment rules, committee referral precedents, and
conference committee proceedings — lie the secrets of congressional power.
What Congress did was seen as inseparable from how it did it. Thus, a number of
studies concerning the theoretical properties of amendment agendas, gatekeeping
powers, and proposal rights, were launched. All of these studies showed the
links between institutional arrangements and policy outcomes.
As with Mayhew, the committee system took center stage in this line of
reasoning, solving a distributional puzzle: how to pass programs that benefit a
few members intensely when all floor votes must be made by majority rule? The
answer was that, via the committee system, legislators requested (and were
granted) membership on the committees that most directly affected their
constituents. Committees could legislate on their subject areas in peace, free
from outside interference. These committee jurisdictions were upheld through
the use of various procedural devices: gatekeeping powers, closed rules, and
domination of conference committees. Thus, legislative organization is designed
to maximize district-specific benefits and help ensure members’
reelection. Weingast and Marshall capped off a decade of research with a
summary of this distributive view of congressional organization (a
rather cynical view, which makes very little mention of which policies are good
for the nation as a whole) in their article, The Industrial Organization of
An alternative view of the committee system posits that the central
problem facing legislators is producing reasonable policy in the face of an
uncertain, complex environment. Committees, once again, solve this problem:
they are populated by experts who put their expertise to work when fashioning
legislation. Therefore, any procedural advantages that committees receive, such
as closed rules, can be read as inducements to gather information, and any
deference they receive on the floor is due to their superior expertise. This
informational theory of committees (much less cynical than the
distributive view, but still not necessarily wrong) is laid out at length by
Krehbiel and complements the distributive theory’s emphasis
on the provision of particularistic benefits.
Despite their different emphases, both approaches share much in common
when describing legislative action. First, both analyses put committees at the
forefront by emphasizing the role committee power plays, even if the two sides
cannot agree on its basis. This committee-based structure is no accident.
Legislators designed it this way, along with its attendant procedural
accouterments, in order to pass legislation that maximizes their reelection
chances. Under both views, the committee system has built-in inefficiencies.
For the distributive view, it is unrestrained logrolling, the most spectacular
instance of which is probably the Smoot-Hawley Tariff Act of
1930, but less disastrous examples of wasteful pork barrel
legislation occur every session. For the informational view, the problem is
that committees and floor members may differ in their preferences, so
policymaking will display some residual uncertainty over outcomes, unable to
fully incorporate committees’ expertise.
Moreover, neither theory provides much support for the proposition that
decreased delegation improves the quality of public policy. Commissioning
legislators with making even more of the fine policy details would only
exacerbate their informational dilemmas, considering the fact that even giving
bureaucrats more explicit goals and guidelines for the exercise of delegated
authority requires significant time and expertise. In addition, it undermines
one of the primary reasons for delegating in the first place — the ability
of agencies to respond flexibly to changing conditions. Furthermore, committees
have shown themselves to be quite adept at providing pork barrel benefits
through omnibus legislation, so it is not clear how shifting a greater
legislative burden to them will decrease the provision of particularistic
policy. As Farber and Frickey note, “A classic example of rent seeking was
the Smoot-Hawley tariff, in which the statute provided enough numerical
certainty to satisfy the most dedicated opponent of delegation. The arduous
task of developing the numbers was largely left to the industries themselves,
which had a field day writing the statute.”
B. Delegation and Oversight
Our other building block comes from the recent literature discussing
administrative procedures, in which a long normative debate has taken a
positive turn. As noted above, those in favor of a strengthened nondelegation
doctrine base their arguments on the hypothesis that delegation is equivalent
to Congress’ abdicating its legislative prerogatives. Some social
scientists have echoed these concerns as well. For instance, the “original
sin” hypothesis advanced by Stigler asserts that bureaucracies are created
by politicians specifically to serve constituents’ needs, with the amount
of protection being determined by the point where supply meets demand; that is,
where the costs to the politician of adding more protection equal the benefits
to the affected constituents.
This point was also argued vigorously by Lowi, who accused legislators
of abandoning their duties by delegating power to unelected bureaucrats and
omnipotent congressional committees. Lowi believed that the
original delegations of power, as in the Interstate Commerce Commission, were
well-conceived and structured so as to make agencies adhere to congressional
intent. But, over time, delegation became less and less tied to specific
mandates and more open-ended, allowing agencies an illegitimate amount of
discretion. Legislators had abdicated responsibility for the execution of
public policy, to be replaced by “interest group liberalism,” meaning
that agencies reacted to the wishes of those organized groups that pressured
them for favorable policy decisions. This, in turn, was wont to
devolve into agency capture — public power exercised for the benefit of a
few private interests against the public good, and unsupervised by
democratically elected legislators. Similarly, iron triangles or subgovernments
may form, in which congressional committees, interest groups and bureaucrats
combine in an unholy trinity to deliver benefits to an interest group’s
members at the public’s expense.
Until the late 1970s, most scholars agreed with this portrayal of
Congress as less important in the ongoing control of agencies, relative to the
courts and interest groups. Much of the support for this viewpoint came from
the observation that Congress rarely exercised its most obvious levers of
control, such as cutting agency budgets, revising original mandates, and
rejecting presidential nominees.
In the early 1980s, however, political scientists began to reassess the
assumption that Congress has relatively little interest in overseeing delegated
authority. The battle lines were drawn clearly by Weingast and Moran, who
distinguished the “bureaucratic” approach (agencies are not
influenced by Congress) from the “congressional dominance” approach
(Congress does exert significant influence). Weingast and Moran
made the important argument that the behavioral patterns emphasized by
advocates of the bureaucratic approach — the scarcity of conspicuous
oversight activities — were indeed consistent with a world not only in
which Congress has little influence over bureaucrats, but also where Congress
perfectly controls the bureaucracy. If the mere threat of congressional
retaliation is enough to cower executive branch agents into submission, then
these agents will never step out of line and legislators need never impose any
overt sanctions. Thus, it is possible that the traditional tools of
congressional control are so effective that they are never actually used. This
is the problem of observational equivalence.
This theme of congressional oversight-at-a-distance was also the subject
of an article by McCubbins and Schwartz, which examined the question of how a
relatively uninformed Congress could possibly control bureaucrats much more
knowledgeable about their particular policy area. True, they
could go out and gather their own information or force the agent to disclose
information at oversight hearings (“police patrol” oversight), but
this would quickly become prohibitively costly, consuming legislators’
scarce time and energy. On the other hand, legislators have access to a cheap
source of information, namely, those interest groups affected by the
agency’s decisions. These groups are generally well-informed about the
relevant issue area and are more than willing to let their representatives know
when an agency is acting contrary to their interests. Thus, legislators would
be able to control agencies simply by sitting back and waiting to see if any
groups come to their doors with complaints (“fire alarm” oversight).
As in Weingast and Moran, if the fire alarm system works perfectly, then
bureaucrats will never step out of line and no fire alarms will actually be
In a similar vein, McCubbins, Noll, and Weingast and
Kiewiet and McCubbins point to administrative procedures as a
key mechanism of congressional control. Along with establishing an agency and
giving it an initial mandate, Congress also specifies the procedures that
agencies must follow in reaching decisions. These procedures may influence
bureaucrats to favor a certain constituency, avoid making rulings in a certain
area, or otherwise bend them to legislators’ will. Thus, congressional
control is woven into the very fabric of the agency, exerting influence in a
powerful, yet subtle, manner. Notice the similarities between this argument and
those of the new institutionalist congressional scholars — one controls
outcomes by controlling the procedures through which they are reached.
According to the administrative procedures literature, the basic problem
that Congress faces when delegating authority is one of bureaucratic
drift, or the ability of an agency to enact policies different from those
preferred by the enacting coalition. This phenomenon, illustrated in Figure 1
below, occurs because agencies can make regulations that can only be overturned
with the combined assent of the House, Senate, and the President. Let the ideal
points of these three actors be H, S, and P, respectively, and
assume that Congress passes and the President signs legislation designed to
implement policy X. Then the Pareto set, or the set of outcomes for
which no improvement for one actor is possible without disadvantaging any other
actor, is the triangle joining the three ideal points. Now assume that the
agency has policy preference A. The agency maximizes its utility by
setting policy equal to X’, the point in the Pareto set closest to
its ideal point. Even though this policy is not what Congress and the President
originally intended for it to be, the necessary coalition to overturn agency
decisions cannot be formed.
Figu r e 1 : B ur e au c r a t ic D r i ft a nd A ge ncy Discretion
Total [X' A]
As detailed by Epstein and O’Halloran, two general categories of
administrative devices have been identified as means of controlling
bureaucratic drift. The first category, ex ante controls,
concerns issues of agency design. What are the procedures, including reporting
and consultation requirements, that an agency must follow in making policy? Who
are the agency’s key constituents, and how will they influence
decisionmaking? What standards or criteria must an agency consider when
promulgating regulations? In which executive department will the new agency be
located, and how far down the organizational ladder will political appointments
reach? These are all threshold questions that legislators must answer when
drafting the authorizing legislation.
The second category consists of ongoing controls by the
institutions and procedures that regularly check agency actions. These
institutions and procedures include instruments of congressional oversight,
such as the direct and indirect monitoring discussed above, and renewing or
withholding. They also include judicial oversight implemented
through existing administrative law and presidential appointment
powers. Both ex ante and ongoing controls, then, serve to limit
the degree to which agencies can act contrary to congressional intent.
In addition, a more direct method of circumscribing agency influence is
available. This method avoids the problems of costly monitoring and complicated
administrative procedures, explicitly limiting the discretion of an
agency to move outcomes from the status quo. Assume, as shown in Figure 1, that
the relevant statute allows the agency to move policy only a limited distance
away from X, e.g., by limiting agricultural price support levels to
somewhere between sixty and ninety percent of parity. Outcomes are now equal to
X’’ which is much closer to Congress’ original intent
than X’. Whereas McCubbins, Noll, and Weingast define discretion as
the actions that no political coalition can overturn and equate discretion with
the limits of bureaucratic drift, we argue that Congress may
prefer to set more stringent limits on agency discretion than those
implied by its ex post power to overturn agency decisions alone.
In fact, Congress limits executive branch discretion in this manner all
the time, simply by passing specific, detailed legislation. If Congress had
specified in the Occupational Safety and Health Act of 1970 that
airborne concentrations of benzene emissions were to be limited to one part per
million, then the Secretary of Labor would have little choice but to implement
these standards. Any attempt to do otherwise would be overturned not by a
coalition of the House, Senate, and President, but by the courts who, in turn,
could not disregard Congress’ explicit mandates without losing
credibility. Thus, the discretion continuum connects congressional and
executive policymaking; stating that there is no discretion is equivalent to
saying Congress creates policy single-handedly.
Congressional control over executive agencies, however, will always be
imperfect. Although legislators may try to influence agency actions through
administrative procedures, these controls can only ameliorate, but never
completely resolve, the basic problem of trying to oversee better-informed
executive actors. As Moe states:
Experts have their own interests — in career, in autonomy —
that may conflict with those of [legislators]. And, due largely to
experts’ specialized knowledge and the often intangible nature of their
outputs, [Congress] cannot know exactly what its expert agents are doing or
why. These are problems of conflict of interest and asymmetric information, and
they are unavoidable. Because of them, control will be
The conclusion is that Congress does oversee agency actions, but it must
still resolve the principal-agent problem of oversight and control by
delegating the appropriate amount of authority to agencies in the appropriate
way. Delegation that is too limited or tightly constrained will deny Congress
the benefits of agency expertise and reduced workload that motivated the
initial delegation of authority. On the other hand, by delegating too much
power to an agency, Congress runs the risk of allowing policies to be enacted
that are contrary to the wishes of legislators and their constituents. It is
this trade-off between expertise and control informational gains and
distributive losses that lies at the heart of this view of administrative
II. THEORY: THE POLITICAL LOGIC OF DELEGATION
We now juxtapose the legislative organization literature with the
delegation literature to address our central questions of how much authority
Congress delegates to the executive branch, and why Congress delegates more
authority in some policy areas than in others. The starting
point for our analysis, as with all political analyses, is the equation that
preferences are filtered through institutions to produce policy. That is,
individuals and their preferences are the fundamental building blocks for
political outcomes, but they do not, in and of themselves, predict a particular
policy; one must also know the institutions that aggregate these preferences,
as well as how they operate in a given circumstance. When predicting the
outcome of a legislative election, for instance, one must not only know the
partisan preferences of the electorate, but also the electoral system being
used, e.g., proportional representation or plurality-winner elections,
district-based or at-large, primaries or no primaries, and so on.
Here, we wish to explain delegation from the legislative to the
executive branch and the impact of this delegation on public policy. The key
actors in this situation are legislators, the President, and executive
agencies. We assume the preferences of legislators and the President to be,
first and foremost, reelection. They may have other concerns as well, such as
the desire for power, rewarding friends, and good government, but to satisfy
any of the above they must first retain public office. The preferences of
bureaucrats are more difficult to specify, as they lack any direct electoral
motivation. The bureaucracy literature suggests that they may be controlled by
their political superiors, driven by the desire to expand their budgets, seek
to protect their professional reputation, or angle for lucrative post-agency
positions. We will concentrate here on the former motivation
— control by other political actors — as it is the most sensitive to
variation in external political conditions.
We further assume that political actors who seek reelection will, on any
given policy, attempt to bring final outcomes as close as possible to the
median voter in their politically relevant constituency. Note that legislators
will not necessarily take into account the preferences of all voters in their
district if only a subset are mobilized on a particular issue, hence the
possibility for special interest politics. On some issues, though, such as
Social Security, minimum wage, and health care, a large proportion of the
electorate will be mobilized, in which case the legislator will try to satisfy
this broader constituency. Furthermore, we assume that actors’ preferences
over policy outcomes differ because they respond to different constituents. For
instance, the fact that presidents have a national constituency usually means
that they will be less susceptible to the demands of any one special interest,
as opposed to House members who represent more narrowly defined geographical
A. Choosing How to Decide
Our institutional analysis begins with the observation that there are
two alternative modes for specifying the details of public policy. Policy can
be made through the typical legislative process, in which a committee considers
a bill and reports it to the floor of the chamber, and then a majority of the
floor members must agree on a policy to enact. Alternatively, Congress can pass
a law that delegates authority to regulatory agencies, allowing them to fill in
some or all of the details of policy. The key is that, given a fixed amount of
policy details to be specified, these two modes of policymaking are substitutes
for each other. To the degree that one is used more, the other will perforce be
Note also that it is Congress who chooses where policy is made.
Legislators can either write detailed, exacting laws, in which case the
executive branch will have little or no substantive input into policy, they can
delegate the details to agencies, thereby giving the executive branch a
substantial role in the policymaking process, or they can pick any point in
between. Since legislators’ primary goal is reelection, it follows that
policy will be made so as to maximize legislators’ reelection chances.
Thus, delegation will follow the natural fault lines of legislators’
In making this institutional choice, legislators face costs either way.
Making explicit laws requires legislative time and energy that might be
profitably spent on more electorally productive activities. After all, one of
the reasons bureaucracies are created is for agencies to implement policies in
areas where Congress has neither the time nor expertise to micro-manage policy
decisions, and by restricting flexibility, Congress would be limiting
agencies’ ability to adjust to changing circumstances. This tradeoff is
captured well by Terry Moe in his discussion of regulatory structure:
The most direct way [to control agencies] is for today’s
authorities to specify, in excruciating detail, precisely what the agency is to
do and how it is to do it, leaving as little as possible to the discretionary
judgment of bureaucrats — and thus as little as possible for future
authorities to exercise control over, short of passing new legislation. ...
Obviously, this is not a formula for creating effective organizations. In the
interests of public protection, agencies are knowingly burdened with
cumbersome, complicated, technically inappropriate structures that undermine
their capacity to perform their jobs well.
Where oversight and monitoring problems do not exist, legislators would
readily delegate authority to the executive branch, taking advantage of agency
expertise, conserving scarce resources of time, staff, and energy, and avoiding
the logrolls, delays, and informational inefficiencies associated with the
Consider, for example, the issue of airline safety, which is
characterized on the one hand by the need for technical expertise, and on the
other hand by an almost complete absence of potential political benefits. That
is, policymakers will receive little credit if airlines run well and no
disasters occur, but they will have to withstand intense scrutiny if something
goes wrong. Furthermore, legislative and executive preferences
on this issue would tend to be almost perfectly aligned — have fewer
accidents as long as the costs to airlines are not prohibitive. The set of
individuals receiving benefits, the public who use the airlines, is diffused
and ill organized, while those paying the costs of regulation, the airline
companies, are well-organized and politically active. Furthermore, keeping in
mind that deficiencies in the system are easily detectable, delegated power is
relatively simple to monitor. For all these reasons, even if legislators had
unlimited time and resources of their own (which they do not), delegation to
the executive branch would be the preferred mode of policymaking.
However, delegation implies surrendering at least some control over
policy, and legislators will be loathe to relinquish authority in politically
sensitive policy areas where they cannot be assured that the executive branch
will carry out their intent. To the extent that legislators delegate to the
executive branch, they face principal-agent problems of oversight and control
since agencies will be influenced by the President, by interest groups, by the
courts, and by the bureaucrats themselves. If agencies are so influenced, they
may abuse their discretionary authority and enact policies with which Congress
is likely to disagree.
Take, by way of illustration, the issue of tax policy, where Congress
uses considerable resources to write detailed legislation that leaves the
executive branch with little or no leeway in interpretation. The
political advantages of controlling tax policy do not come from the duty of
setting overall rates, which taxpayers tend to resent, but from the possibility
of granting corporations and other well-organized lobby groups special tax
breaks, so-called corporate welfare. If designed correctly,
these benefits can target a specific industry or group and are paid for by the
general public, either through taxes paid into general revenues or by the
decrease in revenue stemming from the tax break. Such political benefits are
not lightly foregone, and they would be difficult to replicate through a
delegation scheme with open-ended mandates. Thus, Congress continues to make
tax policy itself, despite the demands of time and expertise that this entails.
Figure 2: Boundaries of the Administrative State
Policy Details Made by Congress Policy Decision
Policy Details Made by Executive Agencies
So, when deciding where policy will be made, legislators will trade off
the internal costs of policymaking in committees against the external costs of
delegating authority to regulatory agencies. We can think of Congress’
decision of where to make policy as equivalent to a firm’s make-or-buy
decision — legislators can either produce policy internally, or they can
subcontract it out (delegate) to the Executive. In making this decision,
legislators face a continuum of possibilities: Congress can do everything
itself by writing specific legislation and leave nothing to the Executive; it
can give everything to the Executive by writing very general laws and do
nothing itself; or it can choose any alternative in between. So, Congress will
choose the point along this continuum — how much discretionary authority
to delegate — that balances these two types of costs at the margin.
As a result, Congress delegates to the Executive in those areas which it
handles least efficiently, where the committee system is most prone to
over-logrolling and/or the under-provision of expertise. Conversely, it retains
control over those areas where the political disadvantages of delegation —
loss of control due to the principal-agent problem — outweigh the
advantages. Just as companies subcontract out the jobs that they perform less
efficiently than the market, legislators subcontract out the details of policy
that they produce at a greater political cost than executive agencies.
B. The Pyramid of Power
To illustrate our argument, Figure 2 summarizes legislators’
alternatives over the structure of policymaking. Each node of the tree
represents a policy decision to be made in a given law, ranging from the most
elemental decisions at the top (e.g., who will receive Social Security, and on
what basis benefits will be paid) to the fine-grained details of implementation
on the bottom (e.g., when and how to mail the benefit checks). The shaded area
represents those details spelled out in the enacting legislation, so the
boundaries between the shaded and unshaded regions are the boundaries of the
administrative state in that policy area.
Note that the uppermost nodes fall within the sphere of legislative
policymaking. This is where the nondelegation doctrine comes into play.
Congress must make the basic policy decisions in any law; it cannot delegate
without also specifying an “intelligible principle” for agencies to
follow. But, this standard imposes only a minimal constraint on
Congress, which in practice is easily satisfied by reference to some
reasonableness standard for bureaucrats. At the bottom of the
diagram, Congress will usually not find it useful or even possible to specify
the most intricate details of policy implementation. Consequently, the scope of
detail in a given piece of legislation will be limited, making explicit some
policy choices and leaving others to executive actors. And the boundaries
between legislative and executive action will be adjusted over time as one mode
of policy production or the other better fulfills legislators’ reelection
C. Empirical Predictions
If this theory aptly describes legislators’ preferences over
delegation, then what patterns should we expect to see in executive discretion
from law to law? To begin with, legislators should be more willing to delegate
authority to executive branch actors who share their preferences than to those
who do not, as such actors are less likely to use their discretion in the
pursuit of policy goals contrary to legislators’ desires. To the extent
that partisan affiliation can serve as a proxy for preferences, and to the
extent that the President can control agency actions through appointment
powers, we should expect to see Congress delegate more authority to Presidents
of their own party than to Presidents of the opposite party.
In a similar vein, legislators should be more apt to rely on committees
whose membership mirrors that of the floor, and to distrust outlying
committees. The legislative organization literature reviewed above emphasizes
that committees whose preferences differ from those of the floor will not
receive procedural benefits such as closed rules and deference in the
policymaking process. If our theory is correct, then these should also be the
committees that lose authority to their executive branch counterparts.
A third prediction arising from our approach concerns the informational
as opposed to distributive nature of policymaking. As policy becomes more
complex, Congress will rationally rely more on the executive branch to fill in
policy details. This occurs for two reasons. The first and most obvious reason
is that the executive branch is filled (or can be filled) with policy experts
who can run tests and experiments, gather data, and otherwise determine the
wisest course of policy, much more so than can 535 members of Congress and
their staff. The second, less obvious reason has to do with the
fact that expertise garnered in legislative committees cannot be transformed
directly into policy outcomes. Rather, it must first pass through the floor,
which may decide to make some alterations to the committee’s proposals.
The existence of the floor as a policy middle-man gives committees less
incentive to gather information in the first place. Executive agencies, on the
other hand, are not hampered by the need to obtain congressional approval;
their rulings become law directly. Therefore, even purely policy-motivated
executive agencies will be more informationally efficient than will be
Bringing together these statements, we predict that Congress would
delegate more authority to the Executive:
1. The closer are the preferences of the Executive to the median floor
voter, so that divided government leads to less discretion;
2. The higher the level of conflict between the committee and the
median floor voter, so that committee outliers lead Congress to delegate less
3. The more complex is a policy area.
Note that our theory, if correct, contradicts the key predictions of the
nondelegation forces. Rather than portraying Congress as delegating
ever-increasing authority to executive actors, we assert that levels of
delegation will rise and fall over time in response to changing external
factors. Instead of assuming that legislators have no interest in monitoring
delegated authority, we assert that they will empower interest groups, the
courts, and other actors to challenge agency actions through administrative
procedures as well as direct oversight. Finally, a revitalized nondelegation
doctrine would have the effect of shifting back to Congress precisely those
policy areas, such as the reduction of pork barrel benefits, that it handles
poorly relative to the Executive, so limits on delegation would only tend to
diminish the efficacy of the political process.
III. TEST: PATTERNS OF DELEGATION IN THE POSTWAR ERA
We now turn to the question of whether or not our predictions are borne
out in practice as well as theory. We use a sample of laws based on David
Mayhew’s list of important legislation in the postwar era contained in his
book, Divided We Govern. Mayhew’s purpose in
compiling this list was different than our current enterprise — he sought
to show that the number of important pieces of legislation enacted during times
of divided partisan control of government does not differ markedly from those
enacted under unified government. For our present purposes, what counts is that
Mayhew devised a methodology for identifying important pieces of legislation,
which will serve as our sample of laws to be analyzed.
Mayhew’s data set includes 267 enactments over the 1947 to 1990
period. Starting from this number, we added the important laws passed in the
102d Congress, then eliminated the entries that were either treaties or
constitutional amendments and those laws that had insufficient legislative
summaries in the Congressional Quarterly, bringing the final number to
257 public laws in our data set.
A. Measuring Executive Discretion
We measure the total amount of discretion ceded to the Executive in law
i as the amount of authority delegated (r[i]) less the
total constraints (c[i]) placed on the executive’s use
of that discretionary authority: di = ri -
ci . Here, the delegation ratio
(r[i]) is defined as the number of major provisions listed in
the Congressional Quarterly summary of each law that delegated
discretionary authority to the executive branch, over the total number of
Provisions with Executive Delegation
Thus, bills in which every provision delegated authority to the
Executive would receive a delegation ratio of one, and laws in which no
provision delegated authority would receive a ratio of zero. Summary statistics
for the delegation ratio and the other measures in our analysis are provided in
Table 1: Summary Statistics
|| Std. Dev.
|Delegation Ratio (ri)
|Constraint Ratio (ci)
|Total Discretion (di)
The second component of the discretion index is the administrative
procedures, the hoops and hurdles, that Congress attaches to the
Executive’s exercise of delegated authority. After canvassing the
congressional oversight literature and reading through the legislation in our
data set, we settled on a list of fourteen procedural constraints that might
appear in any law. These categories, along with their
descriptions, are listed in Table 2.
Table 2: Categories of Constraints
|APPOINTMENT POWER LIMITS
Are there any constraints on executive appointment powers that go
beyond the advice and consent of the Senate?
Are there sunset limits on delegation? That is, does any delegated
authority expire after a certain fixed time period?
Does the act define a maximum amount that the agency can allocate
to any activity or set of activities, either stated explicitly or in a
|LEGISLATIVE ACTION REQUIRED
Do agency determinations require the action (approval) of
Congress to take effect?
|EXECUTIVE ACTION REQUIRED
Do agency determinations require the action (approval) of another
agency or the President to take effect?
Does Congress retain a unilateral ex post veto over the enactment
of an agency regulation?
Are any specific reporting requirements imposed on agency
Are consultations with any other actor, either public or private,
required prior to final agency actions?
Are public hearings explicitly required?
Is there a procedure stated in the act for a party adversely
affected by agency actions to appeal the agency’s decision?
Do explicit mandates require rulemaking or adjudicatory processes
to be carried out in a certain manner (beyond the requirements of the
Administrative Procedure Act)?
Is there a procedure defined in the implementing legislation by
which a non-agency actor reviews the agency’s activities — i.e., a
General Accounting Office audit of the agency?
Is any particular group, product, or affected interest exempt
from any aspect of regulation for a given period of time?
Are any groups, industries, or states given a specific
compensation? In particular, does the act mention any group as receiving either
additional time to adjust to the new regulations, or some concession because of
the costs that may be imposed?
Using the Congressional Quarterly summaries, we coded each law
for the incidence of these constraints. Descriptive statistics for all
categories are shown in Table 1, with the means of each category indicating the
percent of bills that contain at least one instance of that type of
administrative procedure. The constraint index was then
calculated as the number of constraints observed in a law over the total
possible number of constraints, scaled by the delegation ratio:
Categories of Constraints in a Law
ci = Total Categories * ri .
Total discretion is then simply calculated as the difference between the
delegation and constraint ratios. Defining discretion in this way accomplishes
three goals: 1) it avoids negative values of discretion; 2) it translates
delegation and constraints into common units that can then be compared; and 3)
it provides a continuous measure of discretion that falls between zero and
A histogram of the discretion index for the laws in our sample is
provided in Figure 3, with normal distributions overlaid. Nondelegation
advocates would predict high values of r[i] and low
values of c[i]. That is, high levels of delegation
with few constraints, so that overall executive discretion is high. Note,
however, that the distribution in Figure 3 is skewed towards lower values of
discretion. In fact, the average value is only 18% on a zero to one scale. This
skew is corrected somewhat when the zero-delegation laws are removed from the
figure, making the distribution track a normal curve more closely, but there
are still only nine laws with discretion indexes above 0.5, as opposed to
twenty-five that delegated no authority. In short, Figure 3 does
not portray a situation in which Congress delegates large swaths of
discretionary authority to executive actors; those delegations that do exist
appear to be moderate and circumscribed by the imposition of restrictive
Figure 3: Histogram of Discretion Index, with and without
0 .2 .4 .6 .8 1
0 .2 .4 .6 .8 1
B. Patterns and Postwar Trends
As a first check on the plausibility of our measure of discretion, we
identified those bills that contain no executive delegations, the bills for
which d[i] = 0. Table 3 provides the year,
Congress, public law number, and title for the twenty-five bills that met this
criterion. The table reveals some noteworthy results. Some entries, such as the
Consumer Goods Pricing Act of 1975 and the Birthday of Martin
Luther King, Jr., had only one major provision with no
delegation: the former prohibited state laws allowing manufacture-dealer price
fixing, while the latter established a national holiday. Some laws were
instances of rare legislation that do not involve executive authority, such as
the granting of Alaskan statehood, the Public Health Cigarette Smoking Act of
1969 (banning cigarette advertising on television), the Civil
Rights Restoration Act of 1987 (which overruled the 1984 Supreme
Court ruling in Grove City College v. Bell), and the
granting of reparations to Japanese-Americans who were relocated during World
Table 3: Acts with No Executive Discretion
Year Congress PL Num. Title
1948 80 80-471 Revenue Act of 1948
1950 81 81-734 Social Security Act Amendments of 1950 1950 81 81-814
Revenue Act of 1950
1950 81 81-909 Excess Profits Tax Act of 1950
1952 82 82-590 Social Security Increase of 1952 1953 83 83-31 Submerged
1954 83 83-591 Internal Revenue Code of 1954
1954 83 83-761 Social Security Amendments of 1954 1955 84 84-381 Minimum
Wage Increase of 1955
1958 85 85-508 Alaska Admission into Union
1961 87 87-30 Fair Labor Standards Amendment of 1961 1963 88 88-38 Equal
Pay Act of 1963
1965 89 89-44 Excise Tax Reduction Act of 1965
1968 90 90-364 Revenue and Expenditure Control Act of
1970 91 91-222 Public Health Cigarette Smoking Act of
1972 92 92-336 Public Debt Limitation — Extension 1972 92 92-512
State and Local Fiscal Assistance Act of
1973 93 93-233 Social Security Benefits — Increase 1975 94 94-12
Tax Reduction Act of 1975
1975 94 94-145 Consumer Goods Pricing Act of 1975 1981 97 97-34 Economic
Recovery Tax Act of 1981
1983 98 98-144 Public Holiday — Birthday of Martin
Luther King, Jr.
1988 100 100-259 The Civil Rights Restoration Act of 1987 1988 100
100-383 Wartime Relocation of Civilians 1989 101 101-157 Fair Labor Standards
Other bills, though, were more substantive and contained a fair number
of provisions, including the Internal Revenue Code of 1954, the
Tax Reduction Act of 1975, and the Economic Recovery Tax Act of
1981, with seventy-one, forty, and eighty major provisions,
respectively. These bills pertain to tax and fiscal policy, areas in which
Congress has traditionally been reluctant to cede power to the Executive. A
similar story holds for the social security increases, changes in minimum wage,
and other money bills on the list. These are all areas where the usual mode is
for Congress to write specific, detailed legislation that leaves the Executive
with little room for interpretation when enacting the law.
Table 4 shows the acts with the highest levels of executive discretion.
These laws are obviously quite different in nature from those above.
High-delegation laws tend to concern issues of defense and foreign policy (the
National Aeronautics and Space Act of 1958 and the Selective
Service Amendments Act of 1969), environmental policy (the Air
Quality Act of 1967, the Water Quality Improvement Act of
1970, and the Clear Air Act Amendments of 1966),
and health and general social policy (the Maternal and Child Health and Mental
Retardation Act of 1963, the Health Research Facilities
Amendments of 1965, and the Stewart B. McKinney Homeless
Assistance Act of 1987). Traditionally, Congress has willingly
ceded the Executive great leeway in these areas, as the results of ill-formed
policy are often drastic and the political advantages of well-formulated laws
are not nearly as evident — they have only a political downside.
Table 4: Acts with Most Executive Discretion
Year Congress PL Num. Title Discretion
1978 95 95-619 National Energy Conservation 0.48
1967 90 90-148 Air Quality Act of 1967 0.49
1970 91 91-224 Water and Environmental Quality 0.50
Improvement Act of 1970
1958 85 85-568 National Aeronautics and Space 0.51
Act of 1958
1987 100 100-77 McKinney Homeless Assistance 0.53
Act of 1987
1966 89 89-675 Clean Air Act Amendments of 0.57
1958 85 85-864 National Defense Education Act 0.64
1963 88 88-156 Maternal and Child Health and 0.64
Mental Retardation Planning
Amendments of 1963
1963 88 88-164 Mental Retardation Facilities and 0.79
Community Health Centers
Construction Act of 1963
1965 89 89-115 Health Research Facilities 0.93
Amendments of 1965
1961 87 87-41 Inter-American Program — 1.00
1969 91 91-124 Selective Service Amendments 1.00
Act of 1969
These results give some support to our contention that legislators
systematically vary the degree of executive branch discretion from one issue
area to the next, in line with their desires for reelection. In those policy
areas with the greatest potential political benefits, Congress makes specific,
detailed policy, and it delegates in those areas where executive policymaking
is more advantageous. On the other hand, if Congress delegated solely as a
means to avoid difficult choices, as the advocates of nondelegation argue, then
we would not expect legislators to enact direct laws in such contentious policy
areas as taxation, social security, and minimum wage.
It is also enlightening to examine trends in executive discretion over
the postwar period. Figure 4 displays a graph of average discretion, year by
year, for our data, with a trend line overlaid. The graph shows considerable
amounts of variation over time — higher rates of delegation during the
Great Society Era of the 1960s and lower rates during the gridlocked 1980s
— with an overall downward trend. Given these patterns, it is not hard to
understand intellectual trends in the politics of delegation. Viewing the
situation in the late 1960s, it is natural that some observers, such as
Lowi, would fear an ever-expanding rate of delegation with
relatively little congressional control. Those writing in the late 1970s, such
as Fiorina, would be less worried about this trend, as the
Nixon, Ford, and Carter administrations saw a lessening of delegated authority,
and by the late 1980s, scholars like McCubbins, Noll, and Weingast would be
more apt to speak of carefully delimited delegation. Fears that
Congress will abdicate its constitutional role by delegating increasing amounts
of authority to the Executive would therefore seem to be misplaced, at least
from our current vantage point. To put it another way, the only way for these
findings to be consistent with the arguments of nondelegation proponents would
be in the unlikely circumstance that policy choices have become significantly
less difficult over time, rather than more difficult, so that Congress prefers
to delegate less now than before.
Figure 4: Average Discretion, by Year
Era I EraIII EraIV
1947 1949 1951 1953 1955 1957 1959 1961 1963 1965 1967 1969 1971 1973
1975 1977 1979 1981 1983 1985 1987 1989 1991
Furthermore, decomposing this trend reveals that average rates of
delegation to the Executive have remained fairly constant on average over the
postwar era, while the constraints on delegated authority have risen. From the
legislators’ perspective, these findings make a good deal of sense since,
as government activity widens in scope and becomes more complex, a fixed number
of legislators will have a harder time overseeing executive branch activities.
They will, therefore, enlist outside interest groups, the courts, and other
executive actors as allies in the task of congressional oversight. Once again,
delegation to executive officials is not monolithic. Rather, it varies over
time as legislators readjust the division of policymaking authority between the
branches in response to changing conditions.
C. Predicting Discretion
As outlined above, our theory makes explicit predictions concerning the
effect of divided government, committee outliers, and issue area
characteristics on delegation. We now proceed to test this theory for the laws
included in our data set — our 257 pieces of major postwar legislation.
Our approach will rely on multiple regression analysis, using the discretion
measure derived above as the dependent variable and proxies for interbranch
conflict, committee-floor tension, and the informational content of issue areas
as independent variables.
To measure the former, we start with divided government, which is
coded zero if both chambers of Congress and the presidency are controlled by
the same party, and one otherwise. The seat share measure is a more
sensitive indicator of divided government, defined as the average percent of
seats held in the House and the Senate by the party opposite the President. For
example, when the President’s party holds 45% of the seats in both
chambers and the opposition 55%, seat share equals 10%. Descriptive statistics
of these and other variables in the regressions are also provided in Table 1
above. The prediction is that, as the percent of seats held by the opposite
party increases, Congress grants less discretionary authority to the
Figure 5 provides a quick summary of these partisan conflict data,
showing the bivariate relation between seat share and discretion, with one data
point for each Congress, unified government in bold, and divided government in
italics. Notable congressional overachievers in terms of discretion include
President Nixon’s first term, the Kennedy/Johnson
Congress, and the start of President Eisenhower’s second
term. Underachievers include one Congress each for Reagan and
Bush and Carter’s entire tenure. As shown
clearly in the graph, higher levels of partisan conflict are associated on
average with less delegation to executive actors, thus confirming our first
We next measured the degree to which each committee was a preference
outlier with respect to the floor. As of the 102d
Figure 5: Discretion vs. Seat Share, by Congress
Kennedy I Eisenhower III
Discretion Nixon I
Truman IV Bush II Ford
Carter I Eisenhower II
Carter II Truman II
Truman III Reagan I Reagan IV
.1 Reagan III Eisenhower IV
-.2 -.1 0 .1 .2
Congress, twenty-two standing committees composed the
legislative machinery of the House of Representatives. Relying
on the Garrison Nelson data set, we constructed a list of all committee
assignments from the 80th to 102d Congresses. We then combined
these rosters with Poole and Rosenthal Nominate scores, which order members
along a general liberal-conservative continuum, to calculate the median
committee preferences, as well as median party contingent preferences by
committee and by Congress. From this data we constructed the
committee outlier variable, defined as the median majority party
Nominate score on the committee that reported a given bill, less the median
Nominate score for the party overall.
A graph of committee outliers and discretion shows, like the previous
figure, a strong pattern supporting our predictions. As the committees
considering a bill became more of an outlier, Congress delegated more authority
to the executive branch. One interesting data point in Figure 6 is the
Agriculture Committee, long noted as representing farm interests much more
heavily than the average member of Congress. The data suggest that the
committee appears as a preference outlier, but it also shows that considerable
amounts of authority have been delegated to executive
Figure 6: Committee Outliers and Executive Discretion
Science, Space & Tech.
Gov Ops Education
Interior Internal Security
0 .05 .1 .15
branch actors in this policy area, thereby curbing to some extent the
committee’s logrolling tendencies.
Finally, we measured the informational content of an issue area,
traditionally a difficult task. We have therefore developed our own measures of
issue area complexity based on committee hearings data available in the
Congressional Information Service’s Congressional
Masterfile. This source lists all congressional hearings by
committee, with a subject description of the hearing, witnesses, and dates
held. Our measure is based on the assumption that committees dealing with more
informationally intense issue areas will be the ones that hold the greatest
number of hearings. In fact, many authors have described the key role of
hearings in gathering both technical and political information relating to the
measure at hand. For example, Polsby states:
All [these hearings] are necessary¾to make a record, to
demonstrate good faith to leaders and members of the House and Senate, to
provide a background of demonstrated need for the bill, to show how experts
anticipate that the bill’s provisions will operate, to allay fears, and to
gather support from the wavering. Not only does it tell congressmen what the
technical arguments for and against a bill are, but, even more important, it
tells them who, which interests and which groups, are for and against
bills and how strongly they feel about them.
The same can be said of oversight hearings in particular, which Aberbach
identified as a key element of committees’ “intelligence
systems.” Committees that hold more oversight hearings, in
general, invest greater resources to monitor details of policy implementation.
This yields two ways to measure the informational component or complexity of
legislation: by the average number of hearings, and by the percent of
oversight hearings of the committee or committees that reported the
A first look at these data is provided in Figure 7, which graphs the
average number of hearings per Congress and the average level of executive
discretion, by committee. As shown, the trend is indeed generally positive.
Armed Services has relatively high discretion given the number of hearings
held, and Appropriations relatively low discretion. Otherwise, though, this
simple predictor does a fairly good job of estimating the amount of discretion
in laws reported by the committees.
Figure 7: Discretion and Number of Hearings, by Committee
creis omm .3 Armed Services
e y Agriculture
Banking Gov Ops
Ex epo Ed & Lab Energy
e For Aff
R PubWorks Judiciary Appropriations
er ws .1 Post Office
v a W& M
L House Admin
0 50 100 150
Average Number of Hearings Per Congress
As a check on these measures, we also employ another proxy for
informational intensity: committee scope, as defined by Smith and
Deering’s fragmentation of a committee’s jurisdiction.
The broader the scope of issues that come under a committee’s aegis, the
more expertise will be required to manage them. Our fragmentation
variable is constructed as a combination of the number of departments and
agencies under each committee’s jurisdiction, and the number of areas of
legislative jurisdiction in the chamber rules. Since those
committees with more fragmented jurisdictions would be expected to hold more
hearings, and especially more oversight hearings, this serves as a useful check
on our other measures. Indeed, the scope measure correlates with the number of
hearings at 0.63 and with oversight hearings at 0.59. Also, this measure ranks
the committees in roughly the same order as the hearings-based measures.
We are now in a position to present our results. Regressing the amount
of discretion given to the executive branch, law by law, on
congressional-executive conflict, committee outlier status, and informational
intensity, Table 5 suggests that in all cases our major variables were
significant in predicting the level of discretionary authority delegated to the
Executive. More congressional-Executive conflict leads to less discretionary
authority; a 4% decrease in discretion on average. The committee measure
indicates that outlying committees induce the floor to delegate more authority
to the Executive. Finally, the table shows that as issue areas become more
informationally intense, no matter how information is measured, more authority
is delegated as well.
Table 5: Regression Analysis
Dependent Variable: Executive Discretion
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
Divided -0.04 -0.04 -0.04
(-2.90)** (-3.12)** (-3.10)**
-0.17 -0.18 -0.18
(-2.50)** (-2.59)** (-2.66)**
Committee 0.38 0.27 0.34 0.39 0.29 0.36
Outlier (3.68)** (2.60)** (3.37)** (3.77)** (2.77)** (3.50)**
Issue Area Characteristics
Avg. Hearings 0.001 0.001
Avg. Oversight 0.01 0.01
Hearings (4.07)** (3.99)**
Scope of 0.01 0.01
Jurisdiction (5.02)** (5.00)**
Constant 0.05 0.10 0.06 0.02 0.07 0.04
(2.16)** (5.39)** (2.97)** (1.12) (4.37)** (1.82)*
k 16.42** 13.93** 17.41** 15.65** 13.13** 16.68** F
Note: t-statistics in parentheses, calculated from robust standard
errors; one-tailed test *< 0.10; **< 0.05. N=292. The data includes
multiple referrals. High leverage outliers with Cook’s distances greater
than 0.04 omitted from the sample.
Furthermore, analysis of standardized coefficients shows that each of
these three factors had roughly similar effects on delegation, so no one
dimension dominates Congress’ decision to delegate. These results confirm
our hypothesis that a political logic underlies the process of delegation.
Executive discretion increases when it better suits legislators’ need for
reelection, and it decreases when legislative policymaking becomes politically
more efficient. Combined with the finding that in major legislation the norm is
for Congress not to delegate large amounts of its authority (recall that the
mean of the discretion variable was only 0.18), our findings imply a measured
view of delegation. It certainly exists, but it does not overwhelm
congressional policymaking, and, if anything, the trend over time shows it to
be decreasing rather than increasing.
Arguments in favor of strengthening the nondelegation doctrine rest on
the assumption that congressional policymaking is centered around the provision
of particularized benefits to favored constituents, leaving legislators too
little time to consider important pieces of public policy. A stronger
interpretation of the nondelegation doctrine would, therefore, force Congress
to take up major issues that it had been leaving to the executive branch, thus
restoring accountability to public policy and curbing a runaway
This Article provides three responses to this requirement. First, our
data show that Congress does not delegate wholesale to the Executive. Even on
important policy issues, some of which, like the budget and tax policy, require
considerable time and expertise, Congress takes a major role in specifying the
details of policy. Second, when Congress does delegate, it also constrains
executive discretion with restrictive administrative procedures. In fact,
legislators carefully adjust and readjust discretion over time and across issue
areas so as to balance the marginal costs and benefits of legislative action
against those of delegation. Congress is not free from particularistic
legislation, but neither does it devote its energies solely to narrow,
individually tailored policy at the expense of larger issues.
Third, delegation is not only a convenient means to allocate work across
the branches, but it is also a necessary counterbalance to the concentration of
power in the hands of committees. In an era where public policy becomes ever
more complex, the only way for Congress to make all important policy decisions
internally would be to concentrate significant amounts of authority in the
hands of powerful committee and subcommittee leaders, once again surrendering
policy to a narrow subset of its members. From the standpoint of floor voters
this is little better than complete abdication to executive branch agencies. As
it now works, the system of delegation allows legislators to play off
committees against agencies, dividing the labor across the branches so that no
one set of actors dominates. Given this perspective, a resuscitated
nondelegation doctrine would not only be unnecessary, but also would threaten
the very individual liberties that it purports to protect.
What, then, should the attitude of the courts be towards delegation? In
our view, delegation is a self-regulating system, not in need of closer
attention from the judiciary. Legislators will, over time, adjust the
boundaries of the administrative state so that the executive branch considers
those issue areas that Congress handles less effectively itself, keeping the
system in a rough equilibrium. Forcing Congress to do more legislating would
only push back into the halls of the legislature those issues on which the
committee system, with its lack of expertise and tendency towards
uncontrollable logrolls, produces policy most inefficiently. This would be
hardly a step in the right direction.
Consider the position, implied in the majority opinion from Clinton
v. City of New York, that, when delegating, Congress should provide
agencies with a road map or algorithm for translating technical findings into
policy, rather than rely more substantially on agencies’ policy
judgment. The assumption is that, since legislators are elected
and bureaucrats are not, policy decisions made in Congress gain legitimacy that
agency rulings lack.
But consider a fairly complex policy area — say, airline safety
— in which legislators are confident that the regulator’s policy
goals are nearly identical to their own. Delegating under such circumstances
with broad discretion would result in outcomes close to those preferred by
legislators and, by transitivity, to those preferred by their constituents as
well. Where is the illegitimacy here? And why would Congress delegate broadly
to those who do not share their policy goals? In contrast,
forcing Congress to provide a detailed policy algorithm may result in the
regulator’s having to implement policy that neither she nor the
legislators in the enacting coalition would have preferred. True, a good
algorithm will deliver good policy, but relatively uninformed legislators may
not even possess sufficient expertise to be sure that their road map will lead
to the desired destination, rather than a dead end. Must we sacrifice
unanimously agreed upon policy goals in the name of strict procedural
Rather than meddle in the details of Congress’ subcontracting
arrangement with the Executive, the courts should content themselves with their
three traditional roles in the area of delegation: 1) enforcing a minimally
specified nondelegation doctrine, so that some policy goals always accompany
delegated authority; 2) vigorously policing agencies’ rulemaking
procedures, to ensure that those constraints Congress incorporates into the law
are in fact carried out in practice as well; and 3) continuing to enfranchise
any and all affected interests into the policymaking process, so that
bureaucratic malfeasance will be less likely to go unnoticed. In this way, the
courts can provide the infrastructure for an effective balance of powers across
the branches without denying legislators and their constituents the real
benefits of delegation.
* Department of Political Science and Stanford Graduate School of
Business, Columbia and Stanford University.
** Department of Political Science and the School of International and
Public Affairs and Hoover Institution, Columbia and Stanford University.
The authors gratefully acknowledge funding by the National Science
Foundation under grant 95-11628 and by the John M. Olin Foundation.
 2 U.S.C. §§ 691-692 (Supp. II 1996). LIVA
amends the Congressional Budget and Impoundment Control Act of 1974, Pub. L.
No. 93-344, 88 Stat. 297 (codified as amended in scattered sections of 31
U.S.C.). The Supreme Court held LIVA unconstitutional in Clinton v. City of
New York, 118 S. Ct. 2091 (1998).
See Clinton v. City of New York, 118 S. Ct. 2091
 LIVA was ruled unconstitutional for violating the
presentment clause. In the majority opinion, Justice Stevens stated: “If
the Line Item Veto Act were valid, it would authorize the President to create a
different law — one whose text was not voted on by either House of
Congress or presented to the President for signature.” Id. at
 143 U.S. 649 (1892) (upholding Congress’ delegation
of the authority to reduce tariff rates under certain circumstances).
See Richard B. Stewart, The Reformation of
American Administrative Law, 88 HARV. L. REV. 1667, 1675-77 (1975).
Clinton, 118 S. Ct. at 2105.
 Pub. L. No. 73-316, 48 Stat. 943 (1934) (giving the
President the authority to unilaterally reduce tariffs specified in the
Smoot-Hawley Tariff Act of 1930, Pub. L. No. 71-361, 46 Stat. 590, when
negotiating reciprocal agreements with foreign countries).
 117 S. Ct. 2312 (1997).
 Brief Amicus Curiae of David Schoenbrod & Marci
Hamilton in Support of Appellees, Raines v. Byrd, 521 U.S. 811 (1997) (No.
See Peter H. Aranson et al., A Theory of
Legislative Delegation, 68 CORNELL L. REV. 1, 55-63 (1982).
See DAVID SCHOENBROD, POWER WITHOUT
RESPONSIBILITY: HOW CONGRESS ABUSES THE PEOPLE THROUGH DELEGATION (1993);
see also David Schoenbrod, Delegation and Democracy: A Reply to My
Critics, 20 CARDOZO L. REV. 731 (1999).
 SCHOENBROD, supra note 12, at 17.
 Clinton v. City of New York, 118 S. Ct. 2091, 2109
(1998) (Kennedy, J., concurring) (citing THE FEDERALIST NO. 47, at 301 (James
Madison) (Clinton Rossiter ed., 1961)).
Id. (Kennedy, J., concurring).
 DAVID MAYHEW, THE ELECTORAL CONNECTION (1974).
 Key articles in this tradition include: Arthur T. Denzau
& Robert J. Mackay, Gate-keeping and Monopoly Power of Committees: An
Analysis of Sincere and Sophisticated Behavior, 27 AM. J. POL. SCI. 740
(1983); Thomas Romer & Howard Rosenthal, Political Resource Allocation,
Controlled Agencies, and the Status Quo, 33 PUB. CHOICE (Issue No. 4), at
27-43 (1978); Kenneth A. Shepsle, Institutional Arrangements and Equilibrium
in Multidimensional Voting Models, 23 AM. J. POL. SCI. 27 (1979).
Book-length expositions include excellent works by: STANLEY BACH & STEVEN
S. SMITH, MANAGING UNCERTAINTY IN THE HOUSE OF REPRESENTATIVES: ADAPTATION AND
INNOVATION IN SPECIAL RULES (1988); WALTER OLESZECK, CONGRESSIONAL PROCEDURES
AND THE POLICY PROCESS (3d ed. 1989); STEPHEN SMITH, CALL TO ORDER: FLOOR
POLITICS IN THE HOUSE AND SENATE (1989).
 Barry R. Weingast & William J. Marshall, The
Industrial Organization of Congress; or, Why Legislatures, Like Firms, Are Not
Organized as Markets, 96 J. POL. ECON. 132 (1988).
See KEITH KREHBIEL, INFORMATION AND LEGISLATIVE
 Pub. L. No. 71-361, 46 Stat. 590 (1930).
 DANIEL A. FARBER & PHILIP P. FRICKEY, LAW AND PUBLIC
CHOICE: A CRITICAL INTRODUCTION 86 (1991).
See George J. Stigler, The Theory of Economic
Regulation, 2 BELL J. ECON. & MGMT. SCI. 3 (1971), reprinted in
GEORGE J. STIGLER, THE CITIZEN AND THE STATE: ESSAYS ON REGULATION 114
See THEODORE J. LOWI, THE END OF LIBERALISM: THE
SECOND REPUBLIC OF THE UNITED STATES (2d ed. 1979).
See HAROLD SEIDMAN & ROBERT GILMOUR,
POLITICS, POSITION AND POWER (4th ed. 1975); James B. Pearson, Oversight: A
Vital Yet Neglected Congressional Function, 23 KAN. L. REV. 277 (1975). A
commonly cited figure was the fact that the average Senate confirmation hearing
lasted 17 minutes.
See Barry R. Weingast & Mark J. Moran,
Bureaucratic Discretion or Congressional Control? Regulatory Policymaking by
the Federal Trade Commission, 91 J. POL. ECON. 765, 775-800 (1983).
See Mathew McCubbins & Thomas Schwartz,
Congressional Oversight Overlooked: Police Patrols Versus Fire Alarms,
28 AM. J. POL. SCI. 165 (1984).
See Weingast & Moran, supra note
See Mathew D. McCubbins et al., Administrative
Procedures as Instruments of Political Control, 3 J.L. ECON. & ORG. 243
(1987) [hereinafter McCubbins et al., Administrative Procedures]; Mathew
D. McCubbins et al., Structure and Process, Politics and Policy:
Administrative Arrangements and the Political Control of Agencies, 75 VA.
L. REV. 431 (1989) [hereinafter McCubbins et al., Structure and
See RODERICK D. KIEWIET & MATHEW D.
MCCUBBINS, THE LOGIC OF DELEGATION: CONGRESSIONAL PARTIES AND THE APPROPRIATION
See David Epstein & Sharyn O’Halloran,
Administrative Procedures, Information, and Agency Discretion, 38 AM. J.
POL. SCI. 697 (1994).
See generally Randall L. Calvert et al.,
Congressional Influence over Policy Making: The Case of the FTC, in
CONGRESS: STRUCTURE AND POLICY (Mathew McCubbins & Terry Sullivan eds.,
See generally Jerry L. Mashaw, Explaining
Administrative Process: Normative, Positive, and Critical Stories of Legal
Development, 6 J.L. ECON. & ORG. 267 (1990) (special issue).
See Randall L. Calvert et al., A Theory of
Political Control and Agency Discretion, 33 AM. J. POL. SCI. 588
See McCubbins et al., supra note 28.
 Pub. L. No. 91-596, 84 Stat. 1590.
 Terry M. Moe, The Politics of Bureaucratic Structure,
in CAN THE GOVERNMENT GOVERN? 271 (John E. Chubb & Paul E. Peterson
 Despite their obvious importance to the operation of our
system of separate powers, these questions have received relatively little
attention in the previous literature. Important exceptions include: MORRIS P.
FIORINA, CONGRESS: KEYSTONE OF THE WASHINGTON ESTABLISHMENT (2d ed. 1977);
Morris P. Fiorina, Legislative Choice of Regulatory Forms: Legal Process or
Administrative Process?, 39 PUB. CHOICE 33 (1982) (emphasizing the benefits
of delegation when legislators wish to shift the blame for unpopular
decisions); Mathew D. McCubbins, The Legislative Design of Regulatory
Structure, 29 AM. J. POL. SCI. 721 (1985) (positing that legislators
delegate when they cannot reach an internal consensus on policy).
See JAMES Q. WILSON, BUREAUCRACY: WHAT GOVERNMENT
AGENCIES DO AND WHY THEY DO IT (1989).
 Terry M. Moe, Political Institutions: The Neglected
Side of the Story, 6 J.L. ECON. & ORG. 213, 228 (1990) (special
 Airline regulation is an issue with only a political
downside, and failures tend to be spectacular and well publicized.
 The Congressional Quarterly summary of the Tax
Reform Act of 1969, Pub. L. No. 91-172, 83 Stat. 487, for instance, listed 136
major provisions, of which only three delegated substantive authority to the
 For instance, one of the first items on which President
Clinton used his line item veto, coming from the Taxpayer Relief Act of 1997,
Pub. L. No. 105-34, 111 Stat. 788, would have provided a single individual,
Texan multi-millionaire and major Republican party contributor, Mr. Harold
Simmons, with a tax savings of $84 million over five years.
 These are Justice Taft’s words from J.W. Hampton
Jr. & Co. v. United States, 276 U.S. 394 (1928).
 In fact, only two New Deal-era cases have nullified
legislation for lack of compliance with the nondelegation doctrine. See
A.L.A. Schechter Poultry Corp. v. United States, 295 U.S. 495 (1935); Panama
Ref. Co. v. Ryan, 293 U.S. 388 (1935). Both cases invalidated key provisions of
the National Industrial Recovery Act (“NIRA”) as over-broad
delegations of authority, not only to executive branch officials but also to
private industry boards as well, which were to be allowed to set their own
rules of conduct. NIRA was actually an extraordinarily ambitious act, and it
would probably be struck down today as well.
 True, more staff can always be hired, but this creates
ever more intense problems of supervising and monitoring the staff’s
 DAVID R. MAYHEW, DIVIDED WE GOVERN: PARTY CONTROL,
LAWMAKING, AND INVESTIGATIONS, 1946-1990 (1991) (defining important legislation
in a two-step — or as he calls it, a “two-sweep” process). The
first sweep of legislation includes those bills reported in the year-end
roundups of both The New York Times and Washington Post. This
sweep captures contemporary judgments about the productivity of national
government and identifies those pieces of legislation that were thought to be
of historic significance. The second sweep captures those laws which historians
and political observers, in hindsight, identify as being of lasting importance.
This second sweep covered laws only through the early 1980s, as laws passed
more recently would not yet be covered in the secondary sources that Mayhew
relied on for posterior judgment).
 For a more complete description of this process, see
DAVID EPSTEIN & SHARYN O’HALLORAN, DELEGATING POWERS (1999).
 In our survey we relied in particular on the works by:
JOEL D. ABERBACH, KEEPING A WATCHFUL EYE: THE POLITICS OF CONGRESSIONAL
OVERSIGHT (1990); RONALD CASS & COLIN DIVER, ADMINISTRATIVE LAW: CASES AND
MATERIALS (1987); SHARYN O’HALLORAN, POLITICS, PROCESS, AND AMERICAN TRADE
POLICY (1994); WALTER J. OLESZEK, CONGRESSIONAL PROCEDURES AND THE POLICY
PROCESS (3d ed. 1989); PETER L. STRAUSS, AN INTRODUCTION TO ADMINISTRATIVE
JUSTICE IN THE UNITED STATES (1989); McCubbins et al., Administrative
Procedures, supra note 29; McCubbins et al., Structure and Process,
supra note 29; Stewart, supra note 6.
 For instance, rulemaking requirements appeared in nearly
80% of the laws in our sample, reporting requirements and spending limits
appeared in about half, and legislative action, executive action, and a
legislative veto were included in just above 10%.
 Nor is this result simply an artifact of Congress’
delegating large amounts of authority and then adding administrative procedures
to constrain executive actors. In our data set of 257 laws, only 31 had
delegation ratios of 0.5 or greater, or less than one-eighth of the sample.
 Pub. L. No. 94-145, 89 Stat. 801.
 Pub. L. No. 98-144, 97 Stat. 917 (1983).
 Pub. L. No. 91-222, 84 Stat. 87 (1970).
 Pub. L. No. 100-259, 102 Stat. 28.
 465 U.S. 555 (1984).
See Civil Liberties Act of 1988, Pub. L. No.
100-383, 102 Stat. 903.
 Pub. L. No. 83-591, 68A Stat. 3.
 Pub. L. No. 94-12, 89 Stat. 26.
 Pub. L. No. 97-34, 95 Stat. 172.
 Pub. L. No. 85-568, 72 Stat. 426.
 Pub. L. No. 91-124, 83 Stat. 220.
 Pub. L. No. 90-148, 81 Stat. 485.
 Pub. L. No. 91-224, 84 Stat. 91.
 Pub. L. No. 89-675, 80 Stat. 954.
 Pub. L. No. 88-156, 77 Stat. 273.
 Pub. L. No. 89-115, 79 Stat. 448.
 Pub. L. No. 100-77, 101 Stat. 482.
See LOWI, supra note 23.
See FIORINA, supra note 38.
See generally McCubbins et al., Administrative
Procedures, supra note 29; McCubbins et al., Structure and Process,
supra note 29.
 91st Congress (1969-1971).
 88th Congress (1963-1964).
 85th Congress (1957-1958).
 99th Congress (1985-1986) and 101st Congress
 95th Congress (1977-1978) and 96th Congress (1979-1980),
 Committees that changed name during this period were
identified by their name as of the 102d Congress. Only one committee was
abolished completely, the Internal Security Committee, which from the 80th to
90th Congresses was entitled the Committee on Un-American Activities.
 These data were compiled while the authors were
participants of the Harvard-MIT Research Training Group. We thank Charles
Stewart for his assistance. All data were checked against the relevant volumes
of the Congressional Directory; committee rosters used were those as of
the beginning of each Congress.
 For a complete discussion of Nominate scores, which
range from: -1 (liberal), to 1 (conservative), see KEITH T. POOLE & HOWARD
ROSENTHAL, CONGRESS: A POLITICAL-ECONOMIC HISTORY OF ROLL-CALL VOTING
 © 1997 Congressional Information Service, Inc. Used
 NELSON W. POLSBY, CONGRESS AND THE PRESIDENCY (4th ed.
See JOEL D. ABERBACH, KEEPING A WATCHFUL EYE: THE
POLITICS OF CONGRESSIONAL OVERSIGHT (1990).
See STEVEN S. SMITH & CHRISTOPHER J. DEERING,
COMMITTEES IN CONGRESS (2d ed. 1990).
See id. at 80.
See Clinton v. City of New York, 118 S. Ct. 2091
 Some may protest here that the shared policy goals
between legislators and bureaucrats might be the protection of special
interests, rather than the public good. But then this is a dissatisfaction with
politics in general, not delegation in particular; legislators have provided
benefits to favored constituents from Day One of the Republic, and there is no
evidence that the problem has intensified in the present Era of