Professional Documents
Culture Documents
Structural Reform Litigation: Remedial Bargaining and Bureaucratic Drift
Structural Reform Litigation: Remedial Bargaining and Bureaucratic Drift
Anthony M. Bertelli
University of Georgia
Department of Public Administration
and Policy
Sven E. Feldmann
and
Northwestern University
Kellogg School of Management
MEDS Department
The authors wish to thank John Brehm, Richard Epstein, Morris Fiorina, Tom Hammond, George Krause,
Jonathan Bendor, David Baron, Barry Weingast, and an anonymous reviewer for helpful discussions and
comments. Ellen Rubin provided excellent research assistance.
Introduction
Structural reform litigation has been the source of lively debates in legal commentary (e.g.,
Chayes 1976; Fiss 1982; Horowitz 1982; Epstein 1996). Initiated by interest groups representing the unified interest of a class of agency clients, this form of public litigation shapes the
administration of important policy domains, particularly in the social services. The prison
systems of most American states, many state-level child welfare agencies, welfare programs,
public housing authorities, school districts, police departments, and other agencies have been
defendants in structural reform cases, and their administration continues to be significantly
affected by the outcomes. Court sanctioned reforms mandate and interdict specific sets of
agency actions and procedures. Moreover, since structural reform cases are typically brought
against state agencies in federal court, issues of federalism arise.1
Though it is widely accepted that structural reform litigation is a strategy for interest groups to change policy outcomes,2 we illustrate how such cases provide administrative
agencies, in conjunction with interest groups, an institutional tool for systematically creating
bureaucratic drift. The reason for this drift lies in remedial bargaining. Due to the complexity and political difficulty of structural reform cases, remedies are generally negotiated
by the defendant agency and interest group plaintiff.3 This arrangement allows the litigants
to obtain a mutually beneficial consent decree at the expense of the legislature, which is absent from the bargaining table. Surprisingly, this occurs even where the judiciary is entirely
policy neutral and non-activist.
Given the practice and reality of structural reform litigation, the legislature cannot prevent policy drift and cannot realistically anticipate all drift that would result from future
litigation. As a result, the legislature may create an agency with the potential to create unintended policy (see Banks and Weingast 1992). Moreover, once an administrative agency is
1
established, the legislature may not have an incentive (Fiorina 1977) or the ability (Weingast
and Moran 1983) to revisit the delegation. In fact, the drift historically observed in structural reform cases often occurs through changes in the constellation and litigation capacity of
interest groups in a particular policy area.4 Seeking ex post redress from litigation-induced
drift is costly and consequently rare. As a consequence, the drift we analyze presently occurs de facto. If such drift is anticipated to be sufficiently large, the legislature may take
precautions, including the narrow definition of administrative discretion in enabling statutes
(e.g., Epstein and OHalloran 1999; Gailmard 2002) or a decision not to create an agency
at all (Banks and Weingast 1992).5 In a separate paper, we explore the tradeoffs for the
legislature involved in such enforcement issues (Bertelli and Feldmann 2005).
In 1987, the United States District Court for the Western District of Missouri
ordered salary increases for teachers and staff in the Kansas City Metropolitan School District (KCMSD) as one portion of a vast desegregation remedy (Joondeph 1996: 622). KCMSD had difficulty in producing the funds to
comply with this judicial prescription, and it so advised the court. Under
Missouri law, the capacity for local governments to increase property taxes
was restricted. Nevertheless, the District Court ordered property taxes in the
KCMSD to be raised from $2.05 to $4.00 per $100 in assessed value. The
United States Supreme Court held that though the District Court could not
order the tax increase itself, it could have authorized or required KCMSD to
levy property taxes at a rate adequate to fund the desegregation remedy and
could have enjoined the operation of state laws that would have prevented
KCMSD from exercising this power (Jenkins v. Missouri 495 U.S. 33, 51,
1990).
Jess McDonald, director of the Illinois Department of Children and Family
Services (DCFS, the states child welfare agency), and the American Civil
Liberties Union, plaintiffs attorney in a structural reform suit known as B.H.
were accused of collusion by a member of the Illinois Senate. After a consent
decree was in place that held the agency to a significant expansion in service
provision, then-Illinois State Senator Peter Fitzgerald publicly stated his suspicion that DCFS had solicited pressure tactics and threats to seek contempt
citations from the plaintiffs attorneys. As a countermeasure, Fitzgerald proposed a Bill, which was approved and signed by Governor Edgar to become
Public Act 89-645 on 9 August 1996, requiring the Attorney General to notify
3
both the House Speaker and the Senate President before entering into any
remedial agreement whose initial or cumulative effect will require or involve
the appropriation or expenditure of $10 million or more (Chicago Daily Law
Bulletin, 12 August 1996).
At the behest of Governor Orville Faubus, mediaembraced scandals regarding the treatment of inmates in Arkansas state prisons were investigated by
the state police, who found evidence of significant maltreatment. In response,
the Arkansas legislature increased the systems budget and commissioned further study. Another scandal emerged just as the governorship passed to a
Republican, the first since reconstruction, who considered prisons to be a
particularly important issue. This confluence of events led to the creation
of a consolidated Arkansas Department of Corrections and its first commissioner. Commissioner Sarver helped the plaintiffs draw up complaints against
his agency in a structural reform suit (Holt v. Sarver 309 F. Supp. 362, E.D.
Ark. 1970), and the court found prevailing practices unconstitutional (Feeley
and Rubin 1998: 61).
Judicial power over agency resources is crucial to the incentives discussed in this paper.
Courts do not, in principle, order the spending of public funds. Rather, they suggest that
such funds must be appropriated, or that the government, through legislative enactment,
must cease providing the particular service or program in question. Exemplifying this Hobsons choice is the courts language in Holt v. Sarver: If Arkansas is going to operate a
penitentiary system, it is going to have to be a system that is countenanced by the Constitution of the United States.(385) As noted above, the legislature, with its mechanisms for
aggregating preferences of varying intensities over all governmental operations, is given no ef4
fective place at the bargaining table in the ex post enforcement negotiations we subsequently
analyze.
The influence of courts on policy outcomes has been widely recognized in the political
science literature, but is generally attributed to judges policy preferences (cf. Segal and
Spaeth 1993; Rogers 2001; Shipan 2000). Brandice Canes-Wrone (2003) lists the following
cumulative requirement as a necessary condition for courts to have policy influence: (1)
an agency has discretion, (2) courts have discretion in the interpretation of the statute, (3)
judges ideology must affect their ruling, (4) judges preferences are known to the agency, and
(5) agency officials are adverse to being overturned. These are very reasonable conditions
under which ideological judges may be influential. Moreover, in the qualitative literature
on structural reform, the preferences of judges are central to any policy drift achieved in
structural reform litigation (cf. Cooper 1988; Mezey 2000). By contrast, our model demonstrates that none of these conditions is necessary in order for legislative policy choices to be
systematically diverted from intended outcomes. To this end, our model assumes judicial
preferences to be policy neutral.6 We argue that the prevalence of bargaining and interest
representation in public law litigation, rather than judicial ideology, offers interest groups a
technology for influencing policy making that can be more fruitful than lobbying.7
The following section presents a spatial Rubinstein bargaining model, and is followed by
a discussion of its results. The paper concludes with some empirical implications.
The Model
istrative inputs, p and q, which together constitute its policy choice.8 One dimension (q)
may be quantifiable and is easily defined ex ante monitored, and enforced, while the second dimension (p) is more difficult to specify for all contingencies and harder to verify by
objective standards ex post. We will refer to the p dimension as the process by which
the agency implements the statute, while the q dimension may be thought of as personnel
and budget available to the task, i.e., as the resources appropriated by the legislature for
the agency. Both process and quantifiable inputs, (p, q), are required for an agency
action, and the result of their choice produces policy outcomes according to some (possibly
stochastic) production function, such as the placement of a child within foster care no later
than 24 hours after referral to a social services office, the provision of physical and mental
health to inmates, or the maintenance of safety and order in a community.
The agencys enabling statute defines a range of discretion, i.e., a range of permissible
(p, q) from which the agency is to choose its policy. In principle, the range of discretion
could be an arbitrary set in R2 , though we assume that the legislature constrains p and q
separately and independently of each other and that the set is connected. Thus, the statute
specifies a window of p [p, p] and q [q, q] from the agency may choose its policy.9
For simplicity, we consider only one quantifiable and process dimension, respectively,
though the implementation of policy may involve much more than the choice of, say, budget
outlays and effort of public service provision (process). The argument extends straightforwardly to a multi-dimensional choice environment. The gravamen of the process dimension
is that the agency has some de facto degree of discretionregardless of the legislatures intent. The agency may, of course, explicitly be given some discretion to determine policy, but
questions of whether the agencys exercise of discretion falls within the scope of its mandate
and complies with constitutional provisions can often only be addressed ex post as in the
it is natural to assume that the plaintiffs ideal process input exceeds the level the agency
ideally prefers to provide: pg > pa . (ii) The agency would like to perform less than the
statutorily mandated level of p: pa < p. (iii) The statutory level of quantifiable inputs q is
insufficient to satisfy either the agency or the plaintiff: min{qa , qg } > q.
We distinguish two revealing preference configurations. In the first case, the agency
considers the quantifiable input q to be mainly a private good providing benefits to its staff
and management by alleviating odious tasks, increasing leisure or prestige, such as stateof-the-art administrative technology, employment of more bureaucrats, or funds for subcontracting tasks to consultants, and so forth.11 Consequently, the agency generally prefers
a larger q to a smaller one up to some level, and only when q is very large do responsibilities
such as the supervision of consultants and administration of equipment offset the agencys
private benefit. Given Euclidean preferences, larger q can compensate the agency for higher
levels of p in the range where q < qa and p > pa . The plaintiff, on the other hand, only
marginally benefits from these additional q, and its ideal q consequently is smaller than that
desired by the agency: qa > qg .
In the second case, the quantifiable input q is by and large a public good that primarily
benefits the agencys clients, such as a larger budget for direct payments or services. Larger
q directly improve the quality of client services, while creating only marginal benefit for the
agency. Thus, we have qa < qg .
2.1
Since the the quantifiable input is easily monitored and enforced, its appropriation creates
a legislatively generated hard constraint on the agencys policy choice: only combinations
in B = {(p, q) R2 | q q} are feasible. Assumption 2(iii) assures that the constraint is
8
binding.12
In the absence of any further constraints the agency maximizes its utility subject to the
budget constraint and chooses
= arg max ua (x).
x
xB
Since preferences (under A1) are separable and (under A2(iii)) input-constrained, the agency
chooses its preferred process and exhausts the entire quantifiable input permitted under the
= (pa , q).
statute: x
If the legislature wants to induce a different (i.e., higher) process choice, it may specify
the mandate p > pa (as assumed in A2(ii)) and rely on the courts to enforce compliance. Of
course, the legislature may also control the agency, either directly by policing its policies
or indirectly, using a fire alarm mechanism in which interest groups report agency underperformance to an oversight committee (McCubbins and Schwartz 1984; McCubbins, Noll,
and Weingast 1987). If the legislature still has little incentive ex post to sanction the agency
by imposing costs or making future appropriations contingent on compliance, then even the
fire alarm will be ineffective.
In contrast, we are presently interested in judicially enforced constraints. The legislature
specifies a minimum process, or a minimum level of services provided to the client population.
Underperformance relative to this mandated level, however, does not automatically trigger
a sanction; for the constraint to have any effect, an interest group or client group of the
agency needs to file suit for violation of the mandate.13
The set of legal choices is L = {(p, q) | p p p}, and choices outside L generate
a potential case for a law suit.14 By allegedly violating its statutes the agency creates a
justiciable claim for an injured party. Given standing, the prospective plaintiff will sue if
litigation introduces a sufficient probability that, as a result of the suit, the agency will be
9
10
If the legislature responds to the crisis, litigation may be averted.19 If not, the trial and
negotiation stages ensue. Bargaining over remedy is prompted by a finding (or expected
finding) of agency liability. We assume that this condition is satisfied and turn to the
negotiation of the remedy to be embodied in a consent decree.
2.2
R2 {accept, reject} when i is the receiver of an offer. The judge applies a decision rule
d : H R2 {approve, reject}, such that d = approve whenever x L, and d = reject
otherwise. The order of moves is represented in Figure 1.
The decision rule of the policy neutral judge is equivalent to a constraint on feasible offers.
We will momentarily demonstrate the substantive effects of the requirement of obtaining the
judges approval on the bargaining outcome. But first, we characterize the subgame perfect
equilibrium outcome of the game in the following proposition, and present its formal proof
in the appendix.
Proposition 1. For any > 0, the above bargaining game has a subgame-perfect equilibrium
with the following properties: There is a unique pair of policy alternatives, x , y L s.t.
(1) x and y are constrained Pareto efficient, and (2) the following equations are satisfied:
ua (y ) = (1 )ua (x ) + ua (c)
and
(1)
ug (x ) = (1 )ug (y ) + ug (c).
The equilibrium strategies are such that the agency always proposes x and rejects any offer
that yields utility less than ua (y ), and the plaintiff always proposes y and rejects any offer
that yields less than ug (x ). The judge applies the decision rule given above.
In light of the equilibrium strategies in Proposition 1 bargaining ends in the first period
with the agencys legal offer, x L, being accepted by the plaintiff and approved by the
judge. The agency enjoys a slight advantage in making the first offer and having x accepted
instead of y . This advantage becomes negligible as 0.
12
the set of illegal outcomes, an analysis similar the one presented in the cases below might be
done, with the agency and plaintiff splitting the surplus beyond the disagreement outcome,
c. Additionally, if both the agency and plaintiff have ideal points which lie in the legal set
of outcomes, then the agencys constrained maximization problem yields an outcome that
cannot be challenged in court.
14
15
the judges approval must lie on the constrained portion of the Pareto set, i.e., where the
approval constraint is binding.
Figure 3 about here
We have seen that the legal constraint on substantive policy procedure is more likely to
affect the equilibrium bargaining outcome for the remedy if the quantifiable input is a public
good, i.e., when the input does not directly, or only to a limited extent, benefit the agency.
This may seem surprising. However, when the input is primarily a private good, the agency
is more willing to trade off procedure for additional resources.
Both cases of remedial bargaining, however, entail an improvement for both parties relative to the legislatively mandated policy c by violating the legislated constraint on q. As
we shall see, the courts have been willing to approve structural reform of this sort by prioritizing the constitutional and statutory rights of agency clients over the legislative budget
allocations (see Frug 1978). In this sense, structural reform litigation creates an opportunity
for interest or client groups to collude with the agency against the legislatures manifest
interest.25
16
unconstrained utility possibility set is linear.26 The Nash Bargaining Solution corresponds
with the utility possibility set. In the
to the tangency point of the highest hyperbola h
N lies at the midpoint of the Pareto
unconstrained problem with linear utility, this point v
frontier, whereupon the parties split the surplus evenly.
Figure 4 about here
The constrained utility possibility set is a subset of V , as some of the agencys highest
utility levels can no longer be achieved due to the requirement that p > p. The hatched
, which maximizes
region in Figure 4 depicts V . The figure illustrates Proposition 2: If v
the Nash product for the unconstrained utility possibility set, is available in V , then it
lies outside the
clearly maximizes the Nash product for the constrained set. However, if v
constrained set (as depicted in the figure), then the Nash product is maximized at a point
v that must lie on the portion of V where the legal constraint is binding.
Figure 4 also shows that with the requirement of judicial approval, the surplus is no longer
split evenly. The constraint makes the agency necessarily worse off, whereas the plaintiff
may be better off than if the parties bargained without the constraint.27 The surplus for the
two parties is not maximized whenever the legal requirement is binding at the bargaining
outcome. In Figure 2 inefficiency comes entirely at the agencys expense; the plaintiff is a
net beneficiary.28 However, when the constraint is tighter and further reduces the bargaining
set, sufficient inefficiencies may emerge such that both agency and plaintiff are worse off than
at the unconstrained solution. In such a case, both parties still benefit from litigation when
compared with the minimal legislative mandate, c. During settlement bargaining, however,
both would rather do away with the requirement of judicial approval, since a large portion
of the joint surplus is unavailable to them under that condition.
17
The legal constraint, of course, also determines the default outcome when bargaining
breaks down, c. If the constraint p moves toward pg , both constrained and unconstrained
bargaining outcomes move toward G. Thus, ex ante the plaintiff benefits directly from
existing lawspecific statutes and precedential (or at least persuasive) case lawrelating to
process.
Once again, though we have relied on the Nash Bargaining Solution in deriving these
results, the equilibrium of the extensive-form alternating offer game in Section 2.2 converges
to the NBS as the probability of bargaining breakdown vanishes, so the results remain
qualitatively identical.
Discussion
We have shown that an agency can improve its position relative to minimal, but faithful,
compliance with its legislated mandate by (1) violating the mandate, (2) being sued, and
(3) entering into a consent decree. This argument holds when the courts value norms in
different terms than the legislature. Since courts, generally unwilling to substitute a rulebound remedy for agency expertise in reforming bureaucratic structures (see Diver 1979),
encourage remedial bargaining between the litigants, the version of policy drift that results
from negotiated consent decrees is very difficult for legislatures to mitigate ex post. The
legislature is not party to the court proceedingsincluding the negotiation of remedyit
often lacks the interest (due, for example, to low issue salience) or the capacity to monitor
or influence litigation outcomes. Legislators, of course, could revisit the policy issue after
it observes the court-sanctioned drift, and change relevant statutory law, but the payoff for
doing so is often small relative to other legislative tasks. If the court renders its decision on
18
constitutional grounds, a change in statutory law may be insufficient to rein in the drifting
agency, exaggerating the imperfect judicial enforcement effect. In Bertelli and Feldmann
(2005) we consider how the legislature ex ante chooses an agency mandate given that the
enforcement of the mandate by the courts is biased and extremely difficult to surmount.
The direction and magnitude of drift induced by negotiated consent decrees depends
significantly on the agencys preferences over quantifiable input and process choices. In
the private good input case, i.e., where the agency benefits more directly from additional
resources than does the client, the agency is, somewhat ironically, more willing to accede to
the plaintiffs process claims. A consent decree provides the basis for additional funding for
its present and future operations, and these resources compensate the agency for substantive
(process) concessions made to the plaintiff. This case clearly illustrates the incentives for
collusion between agency and plaintiff through structural reform litigation. By signing a
consent decree, the agency admits no wrongdoing, while the remedy provides both litigants
with a policy outcome they desire. This was the situation in our earlier example of the Illinois
DCFS, which led Senator Fitzgerald to introduce a bill to protect the state legislature from
such a hold-up strategy in the courts.
Collusion in the public good input case may be more subtle. Resources in this scenario primarily benefit the agencys clients, and increased resources do less to compensate
the agency for process concessions. Furthermore, many mutually beneficial bargains are
preempted by the legal constraint. Since the bargaining outcome lies directly on the legal
constraint, we anticipate doctrinal debate about what the law actually requires of the agency,
or, in other words, a hard fought case. Recall, though, that our model assumed perfect
common knowledge of the position of the legal constraint. With this degree of information,
the process component of the remedy is entirely determined by the law, and the plaintiff
19
gets the larger share of the bargaining surplus (as Fig. 4 illustrates). To an outside observer,
it may therefore look as if the agency is recalcitrant, with the consent decree minimally satisfying statutory requirements.29 This occurs even though (1) our model assumes common
knowledge of the bargaining environment, (2) the surplus is split using the same bargaining
structure, and (3) as in the private-good input case, neither party has a bargaining power
advantage. While the litigants may seem more adversarial in this case, the nature of the collusion between them is identical. The legislatures ex post enforcement loss is, furthermore,
smaller since the interests of the litigants are less complementary.
The contrast between these scenarios suggests several empirical implications:
a) Structural reform cases in which there is little scope to compensate the agency in some
other dimension for process concessions would likely be more adversarial. We expect
discord to ensue in these cases with regard to the minimal process requirements of the
law (the location of L).
b) An optimal strategy for the plaintiff (or for the judge) to achieve cooperation among
the litigants is to identify a mechanism (such as the private good input in our generic
setup) to compensate the agency for process concessions.
c) Independent of the ostensible degree of collusion between the litigants, the nature of
remedial bargaining implies agency drift from minimal statutory compliance that is
mutually beneficial to the litigants.
Our third prediction clearly distinguishes enforcement through remedial bargaining from
the resolution of a thirdparty enforcement problem in a principalagent relationship. Once
an agency is sued, bargaining leads to policy drift, and the more agency and plaintiff
20
preferences complement each other in their divergence from legislative intent, the more costly
the settlement outcome becomes for the legislature.
21
incur the costs of litigation while costlessly benefiting from the outcome.
Second, groups may observe each others behavior over time, but they presumably lack
the punishment mechanisms to enforce cooperation. Thus, litigation may be delayed. If
one group eventually does file the suit, others may then have an incentive to influence the
bargaining outcome by filing amicus briefs that suggest to the judge particular interpretations
of the agencys statutory and constitutional mandate. Assuming that amicus briefs are costly,
groups with preferences identical to the plaintiff should not file a brief, and allow the plaintiff
to do their bidding in the bargaining process.
Third, one possible consequence of this free-riding is specialization among interest groups.
Groups like the American Civil Liberties Union or Childrens Rights, Inc. are lawyers groups
focusing primarily on litigation strategies, while other groups pursue their common policy
interests in other arenas (such as legislative lobbying). Related to this division of labor
is two-tiered lobbying which would occur when interest groups first successfully lobby the
legislature for legislation that supports their position. Then, to the extent that the implementation of this legislation does not meet their interests, these groups recruit a class
of plaintiffs to sue the agency charged with implementing the legislation. In this way, interest groups can circumvent portions of legislation with which they do not agree, change
agency mandates by bargaining directly with the administration, and build constituencies
for programs and initiatives that, absent court involvement, could not be built. We have not
explored the full consequences of this behavior for policy design and outcomes, but consider
it an interesting avenue for further research.
22
Conclusions
We have shown that structural reform litigation can be used by interest groups and administrative agencies to achieve mutually favorable outcomes that contradict legislative intent.
The divergence of legislative preferences and court enforcement relies on the notion that
courts, when approving remedial consent decrees, do not consider the same trade-offs, budgetary and otherwise, as the legislature in its ex ante attempts to achieve political control
over the agency. In particular, when deciding a case or approving a consent decree, courts do
not consider the full range of policy issues that generate debate in legislatures and the polity
at large. Courts defer a consensus on these trade-offs to the parties of the dispute, citing
the limitations of substituting judicial judgment for administrative expertise. However, the
agency and plaintiffs approach policy analysis through a narrower window of implementation
options than does the legislature.
Our analysis has broader implications beyond public litigation. Since mass torts, shareholder derivative actions, enforcement of environmental statutes, civil rights cases, and other
important classes of litigation are frequently settled by negotiating consent decrees, incentives exist for similar strategic behavior to that exhibited in public law litigation. These
issues are left for future investigation.
23
Appendix
Proof of Proposition 1
We show that the (constrained) bargaining set X is non-empty and convex. We then note
that X maps into a utility possibility set V , to which we apply the standard bargaining
results. Since the correspondence between X and V along the Pareto frontier is one-to-one,
the policy proposals are unique, which completes the proof.
Let c c + (, 0), and note that c = (b, p) L. Since preferences are input-constrained,
> 0 s.t. c L and c i c, i = a, g, thus X is non-empty. Since preferences and the
set L are convex, it follows that X, the intersection of the preferred-sets and L, is convex.
Every proposal in L is approved by the judge.
Each policy in X induces a utility tupel (va , vg ) (ua (c), ug (c)). Let V = {(va , vg ) | vi =
ui (x), j = a, g, for some x X} be the utility possibility set. Generically, each point in
V corresponds to two distinct policies in X (since the players indifference curves intersect
twice, except where they are tangent). However, we note that each point on the Pareto
frontier of V corresponds to a unique policy in X, since either (a) the indifference curves
on the Pareto set of X are tangent, or (b) they cross for the second time outside of L when
x lies on the boundary of L. Furthermore, linearity (or concavity) of the utility functions
assures that V is (weakly) convex.
What follows is the standard bargaining result (cf. Myerson 1991: 3957). Given the
non-empty, convex utility possibility set V , the SPNE is unique; there exists a unique pair of
utility allocations, (va0 , vg0 ) and (va00 vg00 ) that is Pareto efficient and satisfies (2) for vi0 = ui (x )
and vi00 = ui (y ). The equilibrium strategies are such that the agency always demands va0
and rejects any offer strictly below va00 , and the plaintiff always demands vg00 and rejects any
24
Proof of Proposition 2
N maximizes the Nash
For brevity, denote ui (x) by vi and ui (c) by zi , i = a, g. Then x
25
Endnotes
1. State structural reform actions do occur and our analysis applies similarly to them. We do
not take up concerns of federalism here, except to say that, among other sources of authority,
Article III of the U.S. Constitution, the principle of constitutional judicial review enunciated
by the U.S. Supreme Court in Marbury v. Madison 5 U.S. (1 Cranch) 137 (1803), and
the notion of judicial protection of the disenfranchised through the Fourteenth Amendment
enunciated in United States v. Carolene Products Co. 304 U.S. 144, 152-53, n. 4 (1938) have
provided federal courts and commentators with ample bases for justifying a public law
form of adjudication that is distinct from a notion of judges as passive arbiters of disputes
(cf. Fuller 1978; Fiss 1982; Chayes 1976).
2. This kind of litigation reflects an understanding by organized interests that a class action
suit could call governmental attention to problems that were not being addressed by Congress
or were being improperly handled by administrative agencies. (Schlozman and Tierney
1986: 369)
3. Courts encourage remedial settlement because it not only conserves scarce judicial resources
but also promotes compliance, while enabling the judge to avoid the difficult task of distilling and articulating detailed institutional standards from highly abstract constitutional
and statutory principles and minimizes his involvement in the institutions internal administrative detail. (Diver 1979: 78)
4. In Southern school desegregation litigation, for example, once the Supreme Court had de26
clared in Brown v. Board of Education 347 U.S. 483, 495 (1954) that [s]eparate educational
facilities are inherently unequal, plaintiffs could invoke the equitable powers of the courts to
enjoin school districts not only from future violations of rights, but to remedy past violations
as well. Thus, the institutional reform remedy sought by the National Association for the
Advancement of Colored People would facilitate the integration of schools.
5. Litigation addressing inequitable conditions in Southern prisons during the 1970s is illustrative. State legislatures were reluctant to address the issue, as it was inextricably linked to
race discrimination (Feeley and Rubin 1998). While legislators might have sought to claim
credit for improving conditions for minority prisoners, they shunned the public debate of
race relations. Moreover, the legislatures probably never seriously considered closing all of
the prisonsthe equivalent of rescinding the delegation. Thus, the legislatures did act to
supersede the outcomes of structural reform litigation. Recent moves to privatize prisons
are a potential legislative response, but were not widely employed by Southern legislatures
at the time.
6. An important effect of this neutrality is that the judge, when examining proposed consent
decrees, will act as though she does not know the ideal points of the agency and plaintiff. Thus, the parties statements about policy options are evaluated with regard to their
substance, but not their potential strategic value.
7. As Horowitz (1982: 137) notes, If political forces block action in a legislature, . . . [t]he
judicial process reduces the number of participants, so that the blocking action . . . may not
27
13. Whether the chosen process is in compliance with the agencys enabling statute and the
relevant constitutional provisions is a issue of fact. As a simplification, we assume that
agency and plaintiff-interest group have no uncertainty about the interpretation of the statute
and the process-requirement a judge would impose. In other words, we assume away any
issues of law.
14. It is clear that by A2(ii) the agency will never choose p > p and that only the lower bound is
binding; we will therefore ignore the upper bound constraint in the remainder of our analysis.
In some cases this condition might place an additional constraint on the bargaining outcome.
15. A controversy is justiciable if it is appropriate for judicial determination, as opposed to
contingent, hypothetical, or abstract. Standing requires that a plaintiff have in fact suffered
a harm such that it would be proper to grant it the relief sought.
16. Recall that the judge is considered policy neutral in our model. Courts interpret the whether
the agencys process choice violates a plaintiffs legal rights, and do not consider a lack of
resources to be a valid defense.
17. If, as in the judicial politics literature, judges are considered ideological, any known bias
would simply affect the location of the commonly perceived legal constraint.
18. In practice, the option of a structural injunction without the agencys cooperation constitutes
a last resort after lesser forms of reform and supervision have failed. Accordingly, we consider
the threat of receivership as one example of an outcome that occurs when bargaining breaks
down. Where exactly the outcome under receivership lies is irrelevant for the model, as long
29
as this outcome is not on the Pareto set between agency and plaintiff.
19. Due to established legal doctrine noted above, the legislature has the same Hobsons choice
at the threat of suit as it does upon the realization of suit. It may refuse to respond to a
prelitigation threat, since nothing with the force of a court ordera consent decreeis in
place; however, if a suit is imminent, the legislature cannot avoid the foreseeable remedial
action.
20. The seminal works on bilateral alternating-offer bargaining are the finite horizon game of
St
ahl (1972) that converges to the infinite horizon model of Rubinstein (1982).
21. In many structural reform cases, a state attorney general acts as counsel for the agency
(though some private counsel is retained), and in virtually every case must approve the
consent decree on behalf of the state. Our agency presents, accepts and rejects offers directly.
The attorney general is simply acting as a lawyer in this model. In practice, this relationship
has been problematic. In the B.H. case discussed in the introduction, Fitzgeralds bill was an
attempt to use this approval requirement as a fire alarm for the legislature. In a similar child
welfare case in Missouri, the agency director met privately with plaintiffs counsel, and only
informed the attorney general of the negotiation when announcing the agreement afterward
(Bertelli 2004). Such strategic manipulation of the relationship suggests that it is not a
sound measure of political control over the remedial negotiation process even in instances of
preference conflict between the agency director and attorney general. Such conflict is only
likely to arise in the public good input case discussed below.
30
22. Why would the court allow the parties to engage in this specific type of bargaining (alternatingoffer)? In noncooperative game theory, the alternating offer procedure is one of several methods of eliminating the dividers advantage in fair division problems. By giving the plaintiff
the opportunity not only to reject, but also to propose a counter-offer, the court eliminates
the agencys dividers advantage.
Alternating-offer bargaining is not the only means of doing this, but it is similar to what
courts have done in these cases. In comprehensive structural reform litigation, negotiation
almost universally begins with the judge asking the agency to propose a plan for its own
reform (Diver 1979). This plan is often prepared in consultation with a third party. A
hearing, at which plaintiffs and the agency are present, is held to adopt the plan after
negotiations occur.
23. The Nash Bargaining Solution (Nash 1953), xN X, is a policy alternative that satisfies
a number of normative properties, viz., symmetry, Pareto optimality, scale invariance of
the utility representation, and independence of irrelevant alternatives. xN is unique and
maximizes the Nash product, xN = arg maxxX (ua (x) ua (c)) (ug (x) ug (c)) (see Osborne
and Rubinstein 1990).
24. In the interest of simplicity, the figures below are drawn under Assumption A1, such that
preferences are Euclidean and utility is linear in the distance from the ideal policy. Linearity
then causes the (unconstrained) NBS to lie at the midpoint of the contract curve between
the the two players reservation utilities.
31
25. Normative views on such collusion may differ, and we do not presently engage them. For
example, Horowitz (1982: 138) casts such behavior in a distinctively negative light, stating
that a reforming judiciary can be a weapon in the hands of bureaucrats seeking to escape
from political accountability. Alternatively, a view of courts as protecting disenfranchised
minorities as in Carolene Products fn. 4, might see such collusion as a way to circumvent a
legislative debate that would give short shrift to the plaintiffs concerns.
26. Quadratic disutility functions, as often used in spatial model, would yield a strictly concave
(unconstrained) utility possibility frontier.
27. The first part of the statement is easily shown: V shares its Pareto frontier with V on the
unconstrained portion. Since V is convex it follows that the slope is steeper on the constrained than on the unconstrained portion of the boundary. Since the family of hyperbolas
must lie to the left of va whenever
h is homothetic, the tangency with the hyperbola h < h
is not available. The second part of this statement is depicted in the figure; however, one
v
can also construct an example in which vg < vg .
28. There is a literature on plaintiffs motivations in public law and other interest group litigation
(e.g., Olson 1990).
29. The effect in each case is more pronounced the steeper the contract curve between the
litigants: in the public good case, steepness induces a decrease in the scope for bargaining
on process and reduces the equilibrium payoff to the agency, while in the private good case,
the agency would make larger process concessions.
32
30. Litigation costs for interest groups in structural reform cases are large, including not only pecuniary, but also politicaleven physicalcosts. Southern social and political resistance to
integrated schools, for example, was overwhelming, and the strategy of the NAACPs Legal
Defense Fund was designed to keep the political costs as low as possible. By initially pursuing litigation in arenas less directly tied to widespread invidious discriminationgraduate
schoolsthe interest group was able to contain litigation costs while establishing legal precedents for the fight against discrimination in public primary education (Kluger 1977).
33
References
[1] Bendor, Jonathan (1988) Formal Models of Bureaucracy, British Journal of Political
Science 18(2): 35395.
[2] Bertelli, Anthony M. and Sven E. Feldmann (2005) Delegation in the Shadow of Public
Law Litigation, Mimeo, Kellogg School of Management, Northwestern University.
[3] Bertelli, Anthony M. (2004) Strategy and Accountability: Structural Reform Litigation
and Public Management, Public Administration Review 64(1): 2842.
[4] Binmore, Kenneth G., Ariel Rubinstein, and Asher Wolinsky (1986) The Nash Bargaining Solution in Economic Modelling, Rand Journal of Economics 17(2): 17688.
[5] Canes-Wrone, Brandice (2003) Bureaucratic Decisions and the Composition of the
Lower Courts: An Analysis of Wetlands Permitting, American Journal of Political
Science 47(2): 20514.
[6] Cooper, Phillip J. (1988) Hard Judicial Choices. New York: Oxford University Press.
[7] Chayes, Abram (1976) The Role of the Judge in Public Law Litigation, Harvard Law
Review 89(7): 1281316.
[8] Conybeare, John A.C. (1984) Bureaucracy, Monopoly, and Competition: A Critical Analysis of the Budget-Maximizing Model of Bureaucracy, American Journal of
Political Science 28(3): 479502.
[9] Diver, Colin S. (1979) The Judge as Political Powerbroker: Superintending Structural
Change in Public Institutions, Virginia Law Review 65: 43106.
[10] Epstein, Richard (1996) Recognizing the Limits of Judicial Remedies: The Remote
Causes of Affirmative Action, or School Desegregation in Kansas City, Missouri,
California Law Review 84(4): 110120.
[11] Feeley, Malcolm and Edward L. Rubin (1998) Judicial Policy Making in the Modern
State: How the Courts Reformed Americas Prisons. New York: Cambridge University
Press.
[12] Fiorina, Morris (1977) Congress: The Keystone of the Washington Establishment. New
Haven, CT: Yale University Press.
[13] Fiss, Owen (1982) The Social and Political Foundations of Adjudication, Law and
Human Behavior 6(2): 1218.
34
[14] Frug, Gerald E. (1978) The Judicial Power of the Purse, University of Pennsylvania
Law Review 126(4): 715794.
[15] Gailmard, Sean (2002) Expertise, Subversion, and Bureaucratic Discretion, Journal of
Law, Economics, and Organization 18(2): 53655.
[16] Heckelman, D. (1996) Attorney General Must Tell Legislature of Settlements, Chicago
Daily Law Bulletin August 12: 1.
[17] Horowitz, Donald L. (1982) The Judiciary: Umpire or Empire?, Law and Human
Behavior 6(2): 129143.
[18] Kluger, Richard (1977) Simple Justice: The History of Brown v. Board of Education
and Black Americas Struggle for Equality. New York: Random House.
[19] McCubbins, Matthew D., Roger G. Noll, and Barry R. Weingast (1987) Administrative Procedures as Instruments of Political Control, Journal of Law, Economics, and
Organization 3(2): 24377.
[20] McCubbins, Matthew D. and Thomas Schwartz (1984) Congressional Oversight Overlooked: Police Patrols versus Fire Alarms, American Journal of Political Science
28(1): 16579.
[21] Mezey, Susan G. (2000) Pitiful Plaintiffs. Pittsburgh: University of Pittsburgh Press.
[22] Miller, Gary J. and Terry M. Moe (1983) Bureaucrats, Legislators, and the Size of
Government, American Political Science Review 77(2): 297322.
[23] Myerson, Roger B. (1991) Game Theory: Analysis of Conflict. Cambridge, MA: Harvard
University Press.
[24] Nash, John F. (1953) Two-Person Cooperative Games, Econometrica 21(1): 128140.
[25] Niskanen, William (1971) Bureaucracy and Representative Government. Chicago: Aldine.
[26] Niskanen, William (1975) Bureaucrats and Politicians, Journal of Law and Economics
18(3): 617643.
[27] Olson, Susan M. (1990) Interest-Group Litigation in Federal Court: Beyond the Political Disadvantage Theory, American Journal of Political Science 52(3): 854882.
[28] Osborne, Martin J. and Ariel Rubinstein (1990) Bargaining and Markets. San Diego,
CA: Academic Press.
35
[29] Rogers, James R. (2001) Information and Judicial Review: A Signaling Game of
Legislative-Judicial Interaction, American Journal of Political Science 45(1): 8499.
[30] Rothman, David J. (1982) The Courts and Social Reform: A Postprogressive Outlook,
Law and Human Behavior 6(2): 113119.
[31] Rubinstein, Ariel, Zvi Safra, and William Thomson (1992) On the Interpretation of
the Nash Bargaining Solution and Its Extension to Non-expected Utility Preferences,
Econometrica 60(5): 117186.
[32] Rubinstein, Ariel (1982) Perfect Equilibrium in a Bargaining Model, Econometrica
50(1): 97109.
[33] Segal, Jeffrey A. and Harold J. Spaeth (1993) The Supreme Court and the Attitudinal
Model. New York: Cambridge University Press.
[34] Schlozman, Kay Lehman and John T. Tierney (1986) Organized Interests and American
Democracy. New York: Harper and Row.
[35] Shipan, Charles R. (2000) Design of Judicial Review: A Formal Analysis, Journal of
Theoretical Politics 12(3): 269304.
[36] St
ahl, Ingolf (1972) Bargaining Theory. Stockholm: Stockholm School of Economics.
[37] Weingast, Barry R. and and Mark J. Moran (1983) Bureaucratic Discretion or Congressional Control? Regulatory Policy Making by the Federal Trade Commission, Journal
of Political Economy 91(5): 765-800.
[38] Wildavsky, Aaron (1984) The Politics of the Budgetary Process. Boston: Little, Brown.
36
remedial
plan
accept
reject
approve
reject
remedial
plan
remedial
plan
accept
Agency
Group
reject
Decision rule
Receiver
approve
J
reject
Proposer
Judge
q
L
X
A
xN
^
x
c
ug
ua
38
q
L
G
~
xN
xN
A
X
q
^
x
ua c u g
39
vg
v
~
~
h
h
~
V
va
40