A Theory of Jurisdictional Assignments

A Theory of Jurisdictional Assignments in Bureaucracies
Author(s): Michael M. Ting

Source: American Journal of Political Science, Vol. 46, No. 2 (Apr., 2002), pp. 364-378
Published by: Midwest Political Science Association
Stable URL: http://www.jstor.org/stable/3088382
Accessed: 22-11-2015 07:39 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
Midwest Political Science Association and Wiley are collaborating with JSTOR to digitize, preserve and extend access to
American Journal of Political Science.
http://www.jstor.org
This content downloaded from 132.174.254.159 on Sun, 22 Nov 2015 07:39:55 UTC
All use subject to JSTOR Terms and Conditions
A of Jurisdictional
Theory Assignments
in Bureaucracies
Michael M. Ting University of North Carolina at Chapel Hill
What determines the allocation of fundamental issue in the design of bureaucratic systems is the as?
tasks among government agencies? signment of tasks to agencies. Between 1924 and 1973, Congress
This article develops a formal model A established 246 agencies, many of which assumed responsibility for
of task allocation that argues that new tasks (Kaufman 1976). But during that period Congress also assigned
jurisdictions are strategically de? countless new tasks to existing agencies, some of which were accepted hap?
signed to achieve legislative goals. In pily while others generated resistance. What accounts for these variations?
the model, agencies choose unob- Federal and state agencies today collectively handle a multitude of tasks,
servable policies, and political out? but to a large degree the rationales for particular configurations of agency
comes are a noisy indicator of these jurisdictions remain unclear.
choices. The legislature therefore Perhaps the most obvious explanation for how tasks are allocated is
faces a compliance issue when the that they are arranged according to technical criteria. For any new task, an
agencies' policy preferences are existing organization familiar with related tasks would seem likely to per?
different from its own. The legislature form it most effectively. Thus, tasks involving naval warfare should be as?
exerts control by defining agency signed to the Navy, tasks related to retirement insurance should be assigned
jurisdictions, setting ex ante budgets to the Social Security Administration, and completely new and unrelated
and choosing ex post contractual tasks should be assigned to new agencies. While the argument no doubt
inducements. The principal result is characterizes an important incentive for the designers of bureaucratic ju?
that tasks will be consolidated under risdictions, it (perhaps surprisingly) fails to capture a great many interest?
a single roof when that agency pre? ing cases. Technical similarities are often difficult to discern ex ante. It was
fers lower levels of policy than the not obvious to policy-makers in the 1930s, for example, whether food and
legislature. In other cases, separating drug advertising should be regulated by the Food and Drug Administration
tasks prevents resources from being or the Federal Trade Commission (Quirk 1980). Moreover, the argument
allocated in a manner undesirable for seems hard-pressed to explain many cases of agency specialization ex post.
the legislature. Why, for instance, are alcohol, tobacco, and firearms enforcement separated
from the Drug Enforcement Agency? Finally, even in some cases where the
technical explanation works, more general conclusions about the extent to
which technology matters remain elusive. As the Social Security Adminis?
tration's ineffective assumption of the Disability Insurance program illus?
trates (Derthick 1990), it is often clear retrospectively that agencies have
difficultiesperforming new and different tasks. Yet other agencies, like the
Forestry Service, have historically performed seemingly vast portfolios of
dissimilar tasks with great success (Kaufman 1960).
The prevalence of such technically counterintuitive task assignments
naturally leads to a consideration of alternative rationales for task alloca-
Michael M. Ting is Assistant Professor of Political Science, University of North Carolina

at Chapel Hill, Chapel Hill, NC 27599-3265 (mmting@unc.edu).
The author wishes to thank David Baron and Jonathan Bendor for useful discussions.
Justin Adams, Sean Gailmard, Tim McKeown, Ken Shotts, James Snyder, panel partici?
pants at the 1997 Midwest Political Science Association Meetings, and two anonymous
referees also provided helpful comments. All mistakes are of course my own.
American Journal of Political Science^Vol. 46, No. 2, April 2002, Pp. 364-378
?2002 by the Midwest Political Science Association ISSN 0092-5853
364
A THEORY OF JURISDICTIONAL ASSIGNMENTS IN BUREAUCRACIES 365
tion. One important argument is that an optimal juris- task allocation. Because task sets affect agencies' perfor?
dictional allocation might entail separating tasks which mance incentives, and designers anticipate these incen?
share technical similarities across agencies. By making tives, jurisdictional allocations are usefully analyzed as
bureaucratic systems redundant, political principals can equilibrium phenomena. The game theoretic model de?
insure themselves against random failures in individual veloped here will therefore address a host of important
agencies (Landau 1969). While this perspective does not but as yet unanswered questions.3 How exactly do mul?
embody a comprehensive theory of task allocation, its tiple tasks affect the incentives of agencies? Is compliance
amenability to formalization has resulted in the valuable better assured by handing a task to a new agency (or
theoretical contribution of making explicit the incentives commission) or assigning it to a "friendly" agency?
of political principals (Bendor 1985; Heimann 1993). Would legislative principals attempt to induce better per?
formance out of "unfriendly" agencies by merging them
with "friendly" ones? By answering these questions, the
The Strategic Allocation of Agency Tasks
model will also aid the construction of empirical tests.
This article adopts a third perspective, which differs from
redundancy theory in two ways. First, instead of asking Task Allocations
Explaining
when a task is assigned to multiple agencies, it considers
when an agency will be assigned multiple tasks. Second, The model is centered around a hierarchical relationship
it additionally considers the strategic incentives of agen? between agencies and their most important principal, the
cies and other relevant interests in the political process. legislature.4 As Wilson and Moe suggest, this environ?
These topics are explored to some degree by two ment is characterized by numerous tensions. The actors'
prominent works. Moe (1989) argues that a task's bu? policy preferences may diverge because legislatures do
reaucratic home reflects the preferences of the interest not fully control the appointment of bureau heads, or
groups and legislators concerned with its performance. because of pressure from an assigned task's constituen?
For example, an enacting coalition might shield a task cies (e.g., Hammond 1979). Combined with imperfect
from future political attacks by placing it within a monitoring, the bureaucratic allocation of resources can
"friendly" department, or by assigning it to an indepen? then pose serious compliance issues.5 A natural frame?
dent commission. Wilson (1989) attempts to explain why work, then, for analyzing problems with this structure is
even expansionist bureaus frequently shun new responsi? the principal-agent contracting model (Dixit 1995; de
bilities.1The Federal Bureau of Investigation, for in? Figueiredo 1998).
stance, resisted participation in narcotics investigations,
and the Navy and Air Force have historically neglected 3 To see
why formalization of these insights is necessary, consider
sealift and airlift, respectively. Far from being aberra? Wilson's explanations for why agencies resist new tasks. First, the
tions, these tasks have a clear technical relation with the "sense of mission' argument is susceptible to the same criticisms
as the assessment of technical similarities between tasks. Second,
activities of their host agencies, as well as an ability to in?
the cost argument cannot explain why a legislature would not fund
flate budgets. Nevertheless, such tasks were resisted be? additional tasks adequately. It is possible that budgeters operate
cause they might cause agencies to lose their "sense of under a fiscal illusion when facing multi-task agencies, perhaps in
the form of a lag between the time at which additional costs are in?
mission," impose additional costs, and introduce addi?
curred and when a sufficient budget is provided. But if a legislature
tional opportunities for failure.2 were well apprised of policy costs, the real problem for the agency
These arguments are informal and leave some room might well be a loss of "slack" rather than an increase in costs. Fi?
for interpretation, but they establish that the placement nally, policy failures are certainly bad for bureaus, but the argu?
ment lacks a discussion of the implications of failure. Without
of tasks is of critical importance to its supporters and op? some notion of the punishment scheme that an agency faces, it is
ponents, as well as to the agencies themselves. Further, difficult to draw conclusions about whether it would resist an ad?
of a ditional task.
they suggest some of the central features theory of
4See Kaufman (1981). Legislatures are the most natural choice for
Perhaps echoing some popular attitudes in America, a staple of agencies* political principals for this model, since they exert direct
economic analyses of bureaucratic politics is the assumption that control over agencies' budgets and task sets. An interesting exten?
agencies maximize budgets or size (Tullock 1965; Niskanen 1971). sion would be to explore the addition of other principals such as
In fact, this assumption is often accepted as the cardinal distinc? interest groups and the president.
tion between bureaus and firms (see Miller and Moe 1983). 5 The model therefore does not feature
incomplete information.
2In a somewhat different vein from the argument developed here, However, it is worth considering the possibility that a consolidated
Wilson also argues that adding tasks can introduce competing agency would be better informed than the principal about the
agencies or additional principals (e. g., more Congressional over? proper allocation of resources. Such a model might resemble mod?
sight committees). The latter scenario is addressed to some degree els of delegation under incomplete information (e. g., Epstein and
by common agency models (e. g.>Dixit 1995). O'Halloran 1999).
366 MICHAEL M. TING
Unfortunately, this approach presents a major diffi? proval of friendly appointees, or favorable future bud?
culty in the context of agency budgeting. A contractual getary treatment.8
payment does not constrain the ex ante feasibility of an Several nonobvious and testable predictions con?
agent's policy choices, but rather serves as ex post com? cerning the allocation of tasks result. The equilibrium al?
pensation. However, a crucial part of a legislature's location depends on the relative effectiveness of the ex
"Power of the Purse" is the ability to use the budget not ante and ex post controls, which in turn depend on the
only to reward agencies but also to delimit their feasible policy preferences of the players. Consolidation allows
actions ex ante. the principal to write a contract that disproportionately
The game developed here is based on a contracting awards good performance on all tasks. This gives a multi?
model, but it addresses this concern by considering both task agency a greater incentive to perform on all tasks
ex ante and ex post controls. It makes two assumptions than a contract that rewards each task independently.
about the policy technology available to agencies. First, Thus if the legislature desires higher levels of output than
budgets may be freely allocated across assigned jurisdic? the agency on all tasks, it may be in her best interest to
tions or consumed as slack. This comports with the ca? consolidate tasks. Otherwise, if the legislature wants less
sual observation that agencies in practice frequently have effort on a task, consolidation may distort the allocation
considerable leeway in exercising policy priorities.6 Sec? of effort toward the less desired task. In this case, separa?
ond, utilities across tasks are additively separable. Thus, tion ensures that the ex ante budgets are not misallo-
there are no exogenous efficiency gains from consolidat? cated, and the legislature can typically achieve her opti?
ing tasks. This assumption allows the strategic aspect of mal policy on at least one task at minimum cost and with
task allocation to be studied in isolation of technological no contractual payment.
factors. The model complements a variety of formal work on
In the game, the legislature first chooses whether to task and resource allocation in hierarchical relationships.
consolidate tasks under one agency, or separate them Its "hidden action" setup distinguishes it from previous
into a set of single-task agencies. It then sets a budget game theoretic models of agency budgeting, which as?
and a contract for each agency. The budget delimits sume informational asymmetries in a single-period set?
ex ante an agency's feasible policy space, while the con? ting (Banks 1989; Banks and Weingast 1992) or observ?
tract specifies the ex post payment to be delivered for able policies in a repeated setting (Silver 1996; Calvert,
each outcome in an agency's jurisdiction. Each agency McCubbins, and Weingast 1989). The latter category of
then chooses a costly policy that is feasible under its models appeal to Folk Theorems to derive their results,
assigned budget, unobserved by the legislature.7 Upon showing that many policies are supportable in equilib?
observing the outcomes of these policies (which are rium when actions are observable. By comparison, the
correlated with the policy choices), the legislature deliv? equilibrium budgets and policies for each jurisdictional
ers the promised contractual payments. These payments subgame in this model are unique. The model also con?
may be construed in a number of ways, such as less trasts usefully with contracting models. Perhaps the most
stringent oversight, greater program autonomy, ap- relevant work in this area is Holmstrom and Milgrom's
(1991) study of multi-task agents, which finds that a
6While principal will want to ban tasks that agents value much
legislatures often specify how funds are distributed across
tasks, agencies that undertake complex tasks have various methods more highly than the principal, since these tasks will di?
of working around such requirements to satisfy urgent needs or vert the agents' efforts excessively.9 This result is similar
pursue other policy goals. In the 1990s, for example, the California
State Assembly and Senate attempted with limited success to im?
pose a new accounting system on the Department of Transporta? 8A further difficulty with contracting models is the assumption of
tion (Caltrans) to ensure that it complied with legislative spending commitment to the compensation (or punishment) schedule. This
wishes. A report produced by the consultancy SRI International assumption is a nontrivial matter in political settings, where there
(1996, 15) concluded that "Unfortunately, yet not unusually for are no courts to make promises of payment credible. Since agen?
government agencies, no consistent measure of output could be cies must request new budgets annually, it is sometimes justified as
identified. Furthermore, Caltrans had no set of overall department a convenient way to incorporate the effects of repeated play in a
measures that desegregated into division, functional unit, project, single-period game (e.g., Bendor, Taylor, and Van Gaalen 1987).
and individual staff targets to be used as the basis for regularly Nevertheless, it is sometimes unclear whether the behavior pre?
tracking achieved performance versus target and for annual per? dicted by such a single-period game is supportable in a noncoop-
formance reviews." erative context (see, e.g., Banks 1989). It is possible to show, how?
ever, that the equilibrium strategies of the game developed here are
7This scenario accords particularly well with Wilson's (1989) no? sustainable as the Nash equilibrium strategies of a repeated game
tion of craft agencies, or organizations whose activities are difficult
(proof available upon request from author).
to observe, such as the Army Corps of Engineers or Food and Drug
9 The model also finds that the cost of
Administration. Under these conditions, budgets assume a par? providing optimal task in?
ticularly important role because administrative rules (e.g., centives leads a principal to assign agents tasks that are similarly
McCubbins, Noll, and Weingast 1987) are less likely to be effective. difficult to monitor.
in spirit to that of this article, but the model developed consume part of its budget as slack, or allocate resources
here differs in two ways. First, all tasks are mandated by away from one task and toward another.11
law and so must be allocated to agents somehow. Second, The game sequence, depicted in stylized form in Fig?
the principal enjoys ex ante budget control, which power? ure 1, clarifies the incentives of these players. Before the
fully affects her ability to obtain optimal outcomes under first period, P chooses an allocation of tasks across agen?
some conditions. cies. In particular, she may either consolidate tasks under
The remainder of the article proceeds as follows. I an existing agency that already performs another task, or
next introduce the model and characterize task alloca? separate tasks by creating a new agency. In the former
tions in the simple case where no ex post contracts are case, or the consolidation subgame, one agency (A) per?
possible. The full game is examined in the following sec? forms both tasks. In the latter case, or the separation
tion, which also establishes some simple requirements subgame, each task is performed by separate agencies
for effective ex post contracts. Subsequently I derive the (Al, A2) who care only about their own budgets and
main results, comparing the principal's anticipated pay? policy domains and cannot transfer utility amongst
off from consolidating policies to her payoff from assign? themselves. In this regime P effectively collects payoffs
ing tasks to different agencies. The final sections discuss from two independent games.
a few empirical applications and conclude. Following the jurisdictional allocation is a single
cycle of budgeting and policymaking. P chooses an ex
ante budget Bc > Oin the consolidation subgame or a
budget vector B = in the separation
(Bl,B2) subgame,
where B{ > 0 is the budget given to Ai. P also chooses an
The Model ex post compensation for each agency in the form of a
Environment and Sequence contract, which links all possible observable outcomes
for that agency with a payment vector. It is assumed that
The model is single-period game of imperfect informa? P can commit credibly to delivering its announced pay?
tion. There are two tasks or policy dimensions, the out? ments. For the consolidated agency, the contract vector is
comes of which are represented by the random vector X Pc > 0. It will be convenient to define the elements of Pc
= The outcome for task i is either a success as follows. Let P00 represent the payment for failures on
(Xl,X2).
= 1) or failure {X{ - 0). Outcomes are the result of a both tasks, P10 the payment for success on task 1 only,
(X{
two-element policy vector <J>or <|>c, depending on the al? P01 the payment for success on task 2 only, and P11 the
location of tasks. It is assumed that (|>-e[0,l] and payment for success on both tasks. In the separation sub-
= l} =
Pr{Xi tyj. Policies are costly, as defined by the game, P chooses a contract 0, = (P^P1) > 0 for each Ai,
= =
function c(<|>) X^c))-), where c,-((|>t-) mfy}. The where P^ and P1 represents the payment for failure and
policy vector <|)c has identical properties. success, respectively. The requirement that contractual
There are two kinds of players; a political principal payments be nonnegative captures two features of bu?
(P) and one or two agencies. Each player has linear pref? reaucratic politics. First, agencies do not have any re?
erences over policy outcomes and quasilinear utility over sources other than those given by their principals (this is
money. P receives one unit of utility for each success and in the spirit of the assumption of ex ante budgets), and so
zero otherwise. She chooses payment schemes for the P's ability to extract payments out of the agency is highly
agencies, but cannot directly affect outcomes. Each constrained. Second, individual bureaucrats have reser?
agency handles at least one task, though each task is vation values for continuing their employment.12
handled by only one agency. The agency in charge of task The agencies observe the budget and contract
/ receives w{ for a success on that task and z{ for a fail? choices and then adopt policies within their jurisdic?
ure.10 This potential divergence in preferences conve? tions, unobserved by P. In the consolidation subgame, A
niently incorporates the influence of external
parties chooses policy vector <|>c = ((|)f, ty%), while in the separa?
(most notably the executive branch) on agencies. As a re? tion subgame each agency Ai chooses component (|>? of
sult, an agency may exploit imperfect information to
11While slack
pervades organizations (Cyert and March 1963), the
preference for slack also complements the policy preference as?
sumption. A bureaucrat with an ideal policy would ideally wish to
10Since w{and z- can be varied arbitrarily, normalizing P's outcome
spend just enough to attain her most desired policy and keep the
payoffs at one and zero does not sacrifice any generality. Further, remainder as slack.
P's strict preference for successes is innocuous, for if P's payoff
from failure were greater than her payoff from success, then she 12The
assumption also prevents a trivial solution to the game:
need not deal with an agency to attain her most preferred out? without it, P could induce an agency to choose its ideal policy with
come. an appropriately chosen schedule of negative payments.
368 MICHAEL M. TING
Figure 1 Game Sequence
Consolidation Subgame:
Principal Agency Nature

Chooses budget Chooses feasible Determines policy
and contract for policies for both successes and
agency. tasks. failures.
Principal
Chooses to
Consolidate Separation Subgame:
or Separate
tasks.
Principal Agencies Nature
Chooses separate Simultaneously Determines policy
budgets and con? choose feasible successes and
tracts for each policies for own failures.
agency. tasks.
the policy vector <J>.Agencies are subject to the con? policy responses must maximize uA(<J>c,?c,pc) and
straints of their particular budgets: c(<J>c) < Bc for A, uAl{?{;B{,pi), subject to feasibility constraints.
<
and c{ (((>,?) B{ for Ai. Agencies must consume any unex? For the subsequent analysis, it will be useful to define
pended funds as slack. Finally, Nature determines policy "unconstrained" ideal policies for each player. Let
outcomes X based on the selected policies, and P delivers <|>p= arg max ^^-Cj-(<!>,?) be the policy P would
its stated contractual payments. choose if she paid for the policy herself. By the concavity
In the consolidation subgame, the expected utilities of P's objective, <|>p is characterized by dcjdtyi = 1 for
for A and P are therefore: each i at an interior solution, and tyf = 1 otherwise. Like?
wise, let <|>A = arg max ^T ^{w{ + (1 - (^)z- - c,-(<(>,?)]be
= the optimal policy vector for A if the budget constraint is
+ + BC
?A(*c;JBC^c) X,-[ivl^ z,.(l-^)-c|.(<|.?)] slack. Since As objective is concave, <|>A is characterized
by the first-order condition =
+i pu(()P(t)C+plo0P(l-0c) dci/dî w{ -z{ for each i
at an interior solution, so that marginal policy utility
rfi-^^+rfi-^X1-^).
equals marginal policy cost in every dimension. At a cor?
= 0 if
ner solution, tyf w{ < z{ and (|)f = 1 if dci{\)ldî
= < - Because As
^(Bc,Pc;<|)c) X?[^]-BC w{ Zj. utility is separable across dimen?
sions, agency A/s ideal policy is identical to As ideal in
-P%c^-P10*f(l-^) dimension i. Agency policy preferences therefore do not
depend on the jurisdictional allocation. This effectively
assumes that the same forces generate agency preferences
Similarly, the expected utilities for Ai and P in the separa? (e.g., interest group pressure) regardless of whether tasks
tion subgame are: are consolidated. It also implies that there are no econo?
mies of scope to consolidation. As a result, the motives
= + Bj for particular jurisdictional allocations in the model are
?A?(<|?j;BI.,pI.) Wl.(|?j+z,-(l-?|?|.)-c|.(<|?|.)
- strategic, rather than economic.
+ Pft,- +
p?(i <|>.)for each At,
=
?>^1>P2;<|,) XI[0i-.Bi-P!0I-P?(l-0I)]- The No-Contracting Case
Note that there are no income effects, so budgets affect As a baseline case, it will be helpful to derive jurisdic?
policy choices only if they constrain feasibility. tional assignments in the reduced game in which P can?
Subgame perfect Nash equilibria for this game are not write contracts. In other words, suppose that con?
therefore characterized by an optimal jurisdictional tractual payments are constrained to P, = 0 and pc = 0
choice by P, along with budgets and contracts that max? in their respective subgames. For generality, I consider an
imize wp(?c,pc;<|>c) and wp(B,p1,p2;<|)). Agency environment with n > 1 tasks.
This game has a unique subgame perfect equilib? Figure 2 Two-Task Equilibrium of
Case
rium. Consider first the consolidation subgame, where A _the No-Contracting
chooses a policy <|>c = (<|>f,...,<|>?). Using backward in?
Budget Constraint (Bc)
duction, A chooses the following policy vector for any
budget Bc:
A's policy indifference curve
4>c*(Bc)e argmaxX^?^
subject to: c(<|>c)<?c,<|>c >0,<|>c <1.
P's policy indifference

Call <J)C*(Bc) the policy curve. This curve defines the set curve
of policies such that, for any two tasks, the ratios of mar?
ginal utility and marginal cost are equal.13 Thus, policy
vectors along it maximize As return to effort. As Figure 2 *i
illustrates case, this curve consists of the
for the two-task
tangent points between As policy indifference curves and
the budget constraint when Bc < c(<|>A). Since As ob?
jective function is quasilinear and strictly concave,

is continuous and weakly in? < <J)p.
<|>c*(Bc) single-valued, policies along the policy curve as long as <J)C*(Bc)
creasing in Bc. If <|>p > <|>A, then the closest point to <|>p on the policy
Anticipating the policy response, the optimal budget curve is <|>A:BC* = c($A) and<|>c* (c(<|>A)) = <|>A. For
Bc* satisfies: other values of <J>p,P may choose an "interior" point on
the policy curve (as in Figure 2). Note that <|>p cannot be
- Bc, > 0.
Ba e arg max ]T {<$ff
(?c) subject to: Bc the equilibrium policy unless it lies along the policy
curve.
Thus, P chooses a budget that induces the policy on the In the separation subgame, equilibrium budgets and
policy curve that is "closest" to her ideal, <J)P. In general, policies are easier to characterize because each Ai maxi?
this budget equates P's marginal benefit from increasing mizes over one dimension and P's payoffs are additively
the budget (possibly resulting in higher policies) with her separable. Thus, each task may be addressed separately. If
marginal cost. It is easily shown that Bc* is unique. The > (j)f(i.e.>w{ -z{ > 1), then tyf lies along the policy
?f
equilibrium of the consolidation subgame is thus charac? curve. P can get exactly her ideal policy by setting a bud-
terized by the pair {jBc*,<|>c*(?c* )}. get B*
=
cM)> as tf (cf-(4>f)) = tf. If <tf < 4>f > the re?
Because of the simplicity of the functional forms, it sult is similar to the special case above of a principal that
is straightforward to derive closed forms for equilibrium is more ambitious than a consolidated agency:
allo? =
strategies. However, this is not essential for the task B* = c-((j)f) and $ (c^f)) <$>f.Thus for each task i,
of the
cation decision. What matter are the boundaries (|>*= min {tyf ,(|)f}. Aggregating across tasks, the equi?
for <J)C*(Bc). At one <
policy curve and its implications librium policy vector satisfies <J>* <J>P.14
extreme, it is clear that (0)
<J)C* = 0. At the other extreme, Now consider P's jurisdiction allocation problem.
(Bc) = for jBc > c(<J)A), *by the quasilinearity of From the above derivation, it is clear that if <J)A<<J)P,
<J>C* <J)A
As preferences. In general, <j)f (jBc) ? [0,(|)f] for all Bc, then c|)c* = c|)*=<J)A and the jurisdictional allocation is
since As utility is additively separable and (|)f is his ideal inconsequential. The following result uses the above-
choice (if feasible) for task i. Thus, the equilibrium bud? mentioned properties of both subgames to show that in
< c(<|>A), and there is no slack in equilib?
get satisfies Bc* other cases, P would prefer to separate tasks.
rium (i.e., Bc* = c(<J)c* (Bc*))) because A is either bud?
get-constrained or not given any funds above c(<|>A). Proposition 1. In the subgame perfect equilibrium of the
The absence of slack implies that P strictly prefers higher n-task, no-contracting jurisdictional allocation game, P
?
separates tasks.
13This is subject to the policies being "interior," i.e., for is* j, 14This intuition is quite robust across models that consider ex ante
(|>f e (0,1) and tyf e (0,1), which is assured if <|>A<1. Then budget controls and agency policy preferences. Silver (1996) exam?
solving the constrained maximization, it is easily shown that along ines a continuous, unidimenstional outcome space. Bendor and
the policy curve, tyf /tyf = m j(w {- z,) j'm^w;- - z j). As a result, Moe (1985) study a simulation model with boundedly rational
the policy curve in the no-contracting case is linear if <|>A< 1. agencies.
370 MICHAEL M. TING
Proof. Proofs of all Propositions and Comments may be payments by min{p?,p}}. But any contract such that
found in the Appendix. = 0 and
P) p? > 0 cannot be optimal, since Ai would
choose a weakly higher policy for lower payments if
Thus without contracts, and without assuming increas? =
P? 0. Thus, as intuition would suggest, P will never re?
ing returns to scope from policy production or other cost ward the agency for a failure.
benefits, P cannot gain from consolidating any number Second, case (i) (the budget-constrained case) ob?
of tasks. Unless P's policy goals are all more ambitious tains if and only if tyf > tyf. To see this, suppose that
than the agencies', consolidation reduces her control of <[>f> <[>f.Then as in the no-contracting case, P could ob?
budgetary resources across policies. Consolidation re? tain her ideal policy simply by choosing a zero contract
stricts P to maximize over only one budget, and hence and a budget equal to the cost of her ideal policy
over only the set of points along the policy curve. This re? = in case (i). If ty(-<?fy then
(B{ Cj(<|>f)), resulting
striction passes much of the surplus of the interaction to by the concavity of wAl(#)> (i) may only obtain if
A. By contrast, separating the tasks allows P to fine-tune duAi{^\)ldî > 0. But if this were the case, then P could
budgets, thereby protecting her from over-ambitious reduce p] so that duAi(^*)/dî = 0. With this revised
agencies. The result perhaps accords with the casual view contract, P would receive the same policy choice ?* for a
that legislative principals have little leverage over multi? strictly lower payment. Thus, when <j)f < c|)f, Ai's ideal
task agencies, but the subsequent section shows how ex policy is always fully funded, though P may use a con?
post inducements can alter this conclusion. tract to induce a new (higher) ideal policy.
To find this contract, P maximizes:
The Separation and

v ''
2m,- \ 2m-, ) *'{ 2m, )
Consolidation Subgames
I now develop some preliminary results for each sub- Solving, we find that |$j" = (l-2(w; -z,))/3 at an inte?
rior solution. Substituting back into the previous expres?
game that will both establish some intuition and serve as
sions yields the following characterization.
the basis for P's optimal jurisdictional allocation.
Comment 1. The unique subgame perfect Nash equilib?

The Separation Subgame rium strategies for each agency Ai in the separation
Because P's payoffs are additively separable across agen? subgame are:
cies and agencies cannot transfer utility amongst them?

selves, I focus on the equilibrium strategies between P
0 W,-Z;<-1
and a single, separated agency, Ai Given a budget B ? and
contract ($,(*}), Ai maximizes w^jB^p^p})) sub? (^-Z,-+1)2

-? ( , 1
w,-z,e
' ' -1,-
ject to Bj > 36m I 2
?,(<));). This problem is strictly concave, so
the first-order condition is sufficient for characterizing BT =
the maximum. for the constrained maximum: (w,.-z,)2
Solving ,1
4ra;
(w,-zi+P!-P?)2
i-/_ Wj'-Z,- >1,
(l ) 4ra,
=
?(BI.,(p?,p}))
(ii)
2m, 0 w{ -z{ <-l
(1) Wi-Zt + 1 f 1 1
Now consider P's problem. Two observations simplify ?? = W: - Z:

the problem considerably. First, P? = 0. It is clear from ,1
2m{
[ 1 ] that only the difference in payments, p ?- p^, matters
1
in Ai's decision; thus, for any contract such that Pj > 0
Wi-z^h
and p? > 0, P could do strictly better by reducing both 2m;
A THEORY OF JURISDICTIONAL ASSIGNMENTS IN BUREAUCRACIES 3J1
0 Expression [2] also makes clear why at least one pay?

wx-zx<-\
ment must be zero in equilibrium. An equal shift in all of
( i the payments does not affect As policy choice, and hence
= 1-2^^)
-*-L -1,-
PT Wi-Z^
' ' a contract with strictly positive payments cannot be opti?
3 ^ '2
mal. Beyond this intuition, closed-form solutions for the
0 equilibrium contracts and policies are unfortunately dif?
";-Z;^>
ficult to derive, especially when tyf > <|)f for some task
i.15 However, this difficulty will not hinder the derivation
=0. ?
andpf P's choice of jurisdictional allocation. This is because a
simple but important class of contracts, which I call inde?
In equilibrium, contracts are used to induce Ai to between P's
pendent contracts, facilitates comparisons
choose a higher policy only for moderate values of
payoffs in the two subgames.
wi\-z-x. In the other cases, outcomes are identical to the Independent contracts are simply those which specify
?-
no-contracting game. For w zt- < -1, any contract that
payments for each task that do not depend on the out?
can induce Ai to choose a nonzero policy is too expensive come of the other task. Because there are only two out?
for P. For wi -zt> l/2, the policy Ai would have chosen comes for each separated agency, their contracts are nec?
in the absence of a contract is good enough for P not to
essarily independent. For a consolidated agency, a
use a contract. In this case, ex ante budget control is all contract is independent if p11 = p10 + p01 - p00. Inspection
that is necessary to secure an adequate policy. of [1] and [2] reveals that such a contract induces identi?
cal unconstrained ideal policies for A and each Ai if
_ = - _ = _
The Consolidation pio poo pi po and poi poo p^ po But a consoli?
Subgame
dated agency may also receive a nonindependent contract.
subgame, given a budget B P could thus offer A an additional inducement for pro?
Turning to the consolidation
ob? two > + -
and contract |$c, A maximizes uA{$f ,?2>BC>$C) ducing successes, by choosing P11 P10 P01 P00.
> c(<|>c). This problem is also strictly concave, The next comment establishes that nonindependence is a
ject to Bc
so the first-order condition is sufficient for characterizing basic requirement for inducing a consolidated agency to
the maximum. Solving for the constrained maximum: perform better than its separated counterparts.
+ Comment 2. If contracts are independent, then P weakly

W?r ac\ w1-Zl+P10-P00+(PII-PI0-P?1 r)<D2 ?
* prefers to separate tasks.
(BC*C)=-i^nxj--'
.C?^CoC\_W2-22+P01-P00+(Pn-P10-P?1+P00)<t>1 The intuition behind this result is straightforward.

= *
?* \B 'p >
i^Xj With an independent contract, As marginal utility for
each task is unaffected by her effort on the other task.
(2)
The consolidated agency's incentives can thus be dupli?
cated by presenting separate agencies with essentially the
where X is the Lagrange multiplier on the budget con?
same contract. So, critical to consolidation is P's ability to
straint, which equals zero when Bc > c(<J>c*).
choose nonindependent contracts for that agency. Such
This expression makes clear the relation between the
contracts link As payments across tasks, in effect creating
policy selection problem in the two subgames: compared
externalities from the performance of each.
to [1], A's choice in each dimension depends not only on
the benefit of performing that task, but potentially also
on her performance of the overtask. From [2], it is eas?
ily seen that the "first-order" effect of increasing A's pay?

Optimal Jurisdiction Allocation
ment from a success in one task (e.g., P11 or P10 for task
1) is to increase his effort along that dimension. How?
The preceding development suggests that, in contrast to
ever, increasing p10 also decreases A's incentive to pro?
the conclusion of Proposition 1, jurisdictional arrange-
duce along task 2, and increasing P01 similarly hinders
his performance on task 1. Intuitively, then, each con?
15When <J>^<<J>P,P would never try to induce a budget-
tractual payment has the effect of shifting the policy
constrained policy (i.e., she does not use ex ante control), and
curve toward directions that are rewarded, and away hence X = 0. The closed form for equilibrium policies can then be
from directions that are not. derived by solving [2].
372 MICHAEL M. TING
ments can matter. If the principal can credibly commit to As policy curve, then P can of course achieve her ideal
a nonindependent contract, it is no longer obvious that outcome with a consolidated agency. In addition, if tyf is
separation should always be chosen. I now complete the only slightly larger than (|)f and ?f (j * i) is sufficiently
backward induction analysis of the game by characteriz? low, then P may do as well with a consolidated agency. As
ing P's task allocation decision. This is done simply by the proof demonstrates, however, P's preference for sepa?
deriving conditions under which the payoff from one ration is strict over a wide range of parameters.
subgame will be superior to that from the other.
Consolidated Agencies
Separated Agencies
When will P consolidate? Under the conditions of Propo?
I first examine conditions under which P chooses
sepa? sition 2, a consolidated agency would allocate its resources
rate agencies. The first result addresses the case where the in a way not desired by P. By separating the agencies, P
agency or agencies have higher policy ambitions than P could receive her ideal in at least one dimension without
for at least one task. There are two cases to consider. a contract, and the displacement of effort is reduced.
When <|>A><|>p, P does at least as well by separating When these conditions do not obtain; that is, when
agencies because this allows her to allocate resources op? <|>A< <|>p, P cannot receive her ideal policy on any di?
timally. As with the no-contracting case, P can obtain her mension without a contract. In this case, budgetary con?
optimal policy at a minimum cost with no contractual trols are insufficient for attaining a policy closer to <J>p.
payment. Budgetary controls alone are sufficient for Here a nonindependent contract can be useful, since it
making agencies fully pliant. can give A an added incentive to increase his effort on
When tyf > (|)f for only task i, matters are more both tasks simultaneously. The following result shows
complicated. In this situation, a consolidated agency dis? that this possibility causes P to prefer consolidation.
agrees with P not so much over the "quantity" of policy,
but over the distribution of resources across tasks. P can Proposition3.If <J>A<<J>p (alternately if wt- -z-x<\ for
induce A to choose all tasks i) then P consolidates tasks. ?
policies closer to her ideal by provid?
ing contractual inducements, and consolidation may
Intuitively, when <J>A<<|>p, As preferences differ
help a bit by providing a wider range of such induce?
with P's over the extent, rather than the distribution, of
ments. However, P may more easily prevent an excessive
policy. As a result, an independent contract can be de?
allocation of resources toward task iby simply separating
signed for a consolidated agency that duplicates the in?
the agencies. Proposition 2 combines these two cases.
centives of separated agencies. With a nonindependent
contract, P can do even better, inducing A to choose a su?
Proposition 2. if tyf > (|)f (alternately, if wt- -zt- >l)for
perior policy vector more efficiently than separated agen?
some task i, then P separates tasks. ?
cies would. Again, there is a set of preference configura?
tions under which P will be indifferent
between the two
The result is proved by showing that when tyf > <|)f
forms. These occur when -
for some i, P can find a contract for separate agencies
organizational wt- z,are suffi?
ciently low for all i.16 In such cases, P would offer no con?
that does at least as well as that for a consolidated
tractual payments under either structure, and thus the
agency. The intuition is straightforward: for any such
trivial outcome of zero budgets and policy levels result.
task i, P can receive her ideal policy simply by offering a
Figure 3 graphically depicts Propositions 2 and 3.
budget Bt = ct($f), without any contractual payment. The graph plots task assignments as a function of <J>A.
This implies that if <|>A><j>p, P can achieve her ideal
Since <j)A is defined by wi - z-, the graph is also in effect
outcome across both tasks. But if <|> < ty? for j]& i, then
f a plot of P's choice as a function of wt- - z{ e [0,2m{] for
agency A; can be induced to choose the same policy as A both tasks i.
would by offering them both the same expected pay?
ment for a success on task j. Thus, the contracts for
separated agencies will be more efficient than that for a A Numerical Example
consolidated agency, as the inducements for good per?
Figure 4 presents a numerical example of the jurisdiction
formance on task j do not impose any externalities on
game, where the cost function is c(<j>)= 2<|)j*+?\ and
the performance of task i.
zx =z2 =0. The graph plots principal utility in each
There exists
a "small" set of preference configura?
tions under which P will be indifferent between the two
16Asthe proof of Proposition 3 shows, if w j - z -x> -1 for all i, then
structures. For example, if <|>A><J>P and <J>Plies along the preference is strict.
Figure 3 Versus Consolidation - 1.5. Here, the con?

Separation only in the knife-edge case where wx
solidated agency's preferences differ with P's only in the
(j>.A extent, and not the distribution, of policy. P's ideal lies
V
along the policy curve, and thus by limiting the budget
appropriately, P does equally well under either task allo?
P separates tasks cation. Finally, in all cases, as wx increases beyond 1, As
preferences over the allocation of resources across tasks
diverge increasingly from P's. Consolidation then be?
comes progressively worse, while separation promises a
constant payoff.
The examples also give a sense of the optimal con?
tracts in the consolidation subgame. For example, in the
= 1.5 case, all contractual payments are zero except
w2
for p10, which is positive and decreasing for low values of
wx. When wx is low, raising p10 is the most efficient
<?/ means for offsetting As strong preference for devoting
resources toward task 2. Similarly, in the w2 = 0.25 case,
p01 is increasing in wx so that P can maintain As interest
in performing task 2.
subgame for three values of w2 (0.25, 0.75, and 1.5) as a

function of wx, which ranges from 0 to 4.
The calculations are fully consistent with the results Empirical Implications
derived above. As Proposition 3 predicts, for both w2 =
0.25 and w2 = 0.75, consolidation is preferred for wx < 1. The model may be applied to a range of task allocation is?
As Proposition 2 predicts, for w2 = 1.5 or wx > 1, P sepa? sues. For example, in the American military, debate has
rates the tasks. In the case where w2 - 1.5, P is indifferent persisted over the Army and Marine Corps' role in the
Figure 4 Comparing Jurisdictional Assignments
0.40
Wo = 1.5
0.35 Separated
Consolidated
0.30
Wo = 0.75
Separated
Consolidated
?_ 0.25
"o
c
W2 = 0.25
0.20
-Separated
- Consolidated
0.15
0.10 ?i?
0.0 1.0 2.0 3.0 4.0
374 MICHAEL M. TING
provision of air support for ground troops (also known as (H2) Agencies will merge when they are lower demand-
CAS, for Close Air Support). Reformers have often sug? ers than Congress; agencies will be split when they are
gested that the current system, under which each of the higher demanders.17
four military services runs its own "air force," is redundant
and wasteful (Builder 1989). However, this model pro? Examples of HI include the formation of the Consumer
vides a rationale for such a setup. Both the Army and Ma? Product Safety Commission and the Federal Bureau of
rines have a strong preference for CAS by virtue of their Investigation's investigation of organized crime. Ex?
primary missions: the deployment of ground forces. amples of H2 include the 1947 split of the US Army in to
These missions typically require fliers to operate in close the modern Air Force and Army, and the 1958-1960 con?
concert with ground forces in order to deliver munitions solidation of the National Advisory Committee for Aero?
sometimes no more than a few hundred yards from nautics, the Naval Research Laboratory, and the Army
friendly troops. In fact, Marine pilots receive training both Ballistic Missile Agency's
Jet Propulsion Laboratory and
as infantry and often also as "Forward Air Controllers," or Developments Operation Division into the National
ground personnel who direct air strikes. The Air Force, by Aeronautics and Space Administration.
contrast, has a less direct commitment to support ground Simple tests of both hypotheses can be constructed
forces, and has therefore traditionally stressed the destruc? using data on presidential preferences (as a crude mea?
tion of targets farther from the line of battle. This task, sure of agency demand) and median congressional roll-
known as "battlefield air interdiction" (BAI), has numer? call scores. The latter can be augmented with the compo?
ous attractions, as it emphasizes strategic (as opposed to sition of relevant legislative committees and, in some
tactical) gains, requires larger and more expensive aircraft, specifications, party control. Additionally, measures of
requires less inter-service coordination, and carries less each agency's demand can be refined even further by
risk of friendly casualties. its institutional
considering setting; for example, highly
How might the Air Force react if given the CAS insulated agencies (e.g., independent commissions) cre?
role? A consolidated agency with a strong predilection ated by liberal coalitions more resist
might effectively
for one task (i.e., BAI) over another (i.e., CAS) can be conservative presidents. For the dependent variables,
expected to divert resources away from the latter toward data on bureaucratic reorganization (for H2) and the as?
the former. As Proposition 2 suggests, the principal signment of new tasks (HI) can be compiled from sur?
would be better off separating the tasks across agencies. veys of significant legislation 1991).
(e.g., Mayhew
By contrast, both the Army and Marine Corps have a
relatively high demand for CAS, along with their tradi?
tional roles of deploying ground forces. These services
would likely suffer fewer allocative distortions between Conclusions
CAS and their other tasks, and to the extent that Con?
gress agrees with these goals, the consolidation of these It has long been recognized that bureaucratic
responsi?
tasks is sensible. Thus, despite the apparent overlap of bilitiesare not haphazardly distributed, but rather serve
hardware, the existing integrated air-ground force struc? specific political goals. Beyond this insight, however, little
tures in the Army and Marines can be expected to per? is known about a wide variety of circumstances under
form better than one in which the Air Force handled all which legislatures face the challenge of assigning tasks to
military aircraft. organizations. The theory developed here addresses the
The model is also amenable to a more rigorous issue by examining an environment in which the bureau?
analysis. It predicts variations in the assignment of new cracy may disagree with a legislature over the expendi?
tasks, as well as the rearrangement of existing jurisdic? ture of resources across tasks. To improve bureaucratic
tions, depending on Congress' and the agencies' relative compliance under uncertainty, the legislature may use
demands for bureaucratically provided services. In par? budgets as ex ante constraints on feasible policies and
ticular, two of the most straightforward hypotheses are as contracts as ex post inducements for better performance.
follows: This setup corresponds to the actual budgeting process
more closely than those of standard contracting models.
(HI) Agencies will be created for new tasks when they
are higher demanders than Congress; new tasks will be
17Amore refined hypothesis consistent with Proposition 2 is that a
folded into existing agencies if they are lower multi-task agency will be split when it is high demander for at least
demanders. one task.
A THEORY OF JURISDICTIONAL ASSIGNMENTS IN BUREAUCRACIES 3J5
The model derives intuitive

predictions about task
_Appendix_
assignments. In equilibrium, consolidation allows princi?
pals to link inducements across tasks, thus giving an Proposition 1. In the subgame perfect equilibrium of the
agency greater incentive to choose higher policies for all n-task, no-contracting jurisdictional allocation game, P
tasks. Separation prevents the allocative distortions that separates tasks.
are possible when agencies control multiple tasks. As a
result, when a principal prefers higher policy levels than Proof. Consider three cases. For convenience, I abuse
the agencies on all tasks, tasks will be consolidated. Oth? notation for wp(?) to reflect the zero contracts in the obvi?
erwise, tasks will be separated. In this situation, the prin? ous manner. First, if <|>A<<|>p, then <|>A is the equilib?
cipal can achieve her ideal policy on at least one dimen? =
rium policy under both regimes, so mp(B*;<J>*(B*))
sion at minimum cost, and this performance is insulated up(Bc*;4>c*(Bc*)).
from that of the other task. Second, if <|>p<<|>A, then <|>p may only be chosen if
Future research should expand on the model in sev? (J>p=cj)c* (Bc) for some Bc. If there exists such a Bcy
eral directions. An immediate extension would be to then wp(B*;<|>*(B*)) = up(Bc*;$c* (Bc*)); otherwise,
make agencies more active participants in the task alloca? up(B*;V(B*))>up(Bc*;4>c*(Bc*)).
tion process. Among other things, this would allow a for- Third, without loss of generality, suppose that for
malization of Wilson's (1989) logic for why agencies pro? ie{l,...,;}, <|)A<(|>f and for ie{/ + l,...,?}, <|>A^<|>f>so
tect their "turf." In its current form, the model suggests a that <|>*= (tyf,...ytyf,?j+v...,typ). By the convexity of <;(?)
rationale for why even budget-maximizing agencies and additive separability of utility functions,
would resist new tasks: in equilibrium a legislature will
add tasks to realize efficiency gains, and these gains come for all<|><<|>*. [3]
wp(B*;<|>*(B*))>wp(c(<|>);<|>)
at the expense of agency contractual payments. If, in ad?
dition to choosing policies, agencies could offer the legis? Thus if <|)c*<<|)*,wp(B*;<|)*(B*))>wp(5c*;<|)c*(5c*)). If
lature inducements for favorable treatment or exploit not, then <|>?*>?*k for some fce{/ + l,...,n}. For any such
=
asymmetric information about their production tech? <|)c* define a new policy vector <|>c',such that <|)f' <j)pif
*
nology, then they may resist unfavorable tasks or com? <|>f >(()? and <|>f'=(()? otherwise (so that <|>c'< 4>*). Then
pete over favorable ones. The extended model could then for each such kyP receives <j>?'- ck ((()?') under policy vector
-
serve as a basis for an equilibrium theory of when and <|>c',and (|>?* ck (()>?*)under policy vector <|>c*.But since
= - -
how agencies expand, and of how an endogenous prefer? C ?*>?*' ck (?*') > <Mf ^fc(<MT)?Bythe additive sep?
ence for turf might arise. arability of utility functions, we obtain: wp(c(<|>c');<|>c')
A second set of extensions should address the policy- >up(Bc*;<bc*(Bc*)). However, <|>c'<<|>*, so by [3],
making context. For example, the policy environment wp(B*;<|>*(B*))> wp(c(<|>c');<|>c'). Combining expressions,
may be enriched to suit specific policy areas better. Such wp(B*;<J>*(B*))> up(Bc*;4>c* (Bc*)). Therefore, P chooses
might include more dimensions, fixed or to separate tasks. ?
generalizations
nonseparable costs, and correlations in success probabili?
ties across policy dimensions. But most importantly, the Comment 1 and Proof. This result is by derivation in the
institutional context deserves closer scru? text. ?
bureaucracy's
tiny. The legislature may be given a broader set of choices
over a new agency's type (e.g., independent commissions Comment 2. If contracts are independent, then P weakly
versus departmental agencies) or departmental home prefers to separate tasks.
(which may affect the agency preferences). Finally, since
new jurisdictions are often accompanied by new princi? Proof. It is sufficient to show that if pc satisfies P11 =
pals, the model can be usefully adapted to flesh out the p10+P01-p00, all outcomes under consolidation can be
implications of common agency for the assignment of replicated with separate agencies. Substituting into
tasks. wA(<|>c;?c,Pc) and simplifying, A's objective under an in?
dependent contract is:
Manuscript submitted luly 7, 2000.
Final manuscript received August 24y 2001.
- - + p00.
+bc +
(p10 p00 )(|>f + (p01 p00 )$
376 MICHAEL M. TING
Let <|>c* be As solution to this problem. Now suppose Al and A2 choose (|)P and ty%*,respectively. P thus receives
that P offers each separated agency Ai a budget cx(<|>f*) and a strictlysuperior policy and budget combination with
the following contracts: P1=P10-P??, P2 =P01-p00, separate agencies. P also makes lower expected contractual
P?=0 and P5=0.Then 'duA)^l=duA^d^l for each payments with separated agencies if: pj>'=0 or
task i. By the concavity of all agency objective functions, for 4>f [(P01* -P00* )(i-4>f )+(Pn* -P10* )4>f ] ^ P00* (i-tf*)
- _
(1 <|>f) pio* $? (1 q>c*} +por (1 _ tyc* ^c* +pir ^c* ^c* ?
any task such that tyf* >0, duA(^)/d^t>0. Hence, +
duîtyf^/dtyi >0 and Ai would choose <|>f* if offered or p00* Cl? <t>F*
) + P10>f >0, which holds trivially. Con?
*
budget c{ ((|)f *). For any task such that <|>f > 0, Ai would tradiction.Thus P prefers separation if c^f* ^ <|>p.
clearly choose <|>f* if offered budget ct-(<|>f*). Since <|>c* C/aira 2: Pprefers consolidation only if ?2* = (|)J. Sup?
must be feasible under Bc, Bc >c(<J>c*). Moreover, since pose not. Given Claim 1, <|>c*= (^Py^2*)> while <|>*= <|>p.
P^^PI + pi, P10>P1, P01>pi, and P00>P?+P^, the to? By choosing P2 as defined above, P can induce A2 to
tal payments to Al and A2 for any outcome are weakly less choose ty%*with the same expected contractual payments
than the payment to A. Thus <|>c* may be implemented by as in the consolidation subgame. But since P ideally prefers
separated agencies at weakly less cost and weakly lower con? a contract inducing A2 to choose <|>J,she prefers separa?
tractual payment than with a consolidated agency. ? tion: contradiction.
Thus consolidation can only be preferred by P if
Proposition 2. If tyA > tyf (alternately if w{ -zt>\) for <|)c* =<j)* (note that <|>p= $). Since <|>*is achieved with
some task i, then P separates tasks. the minimal budget and expected contractual payment un?
der separation, separation is weakly preferred for <|>A> <|>p
Proof. There are two cases. First suppose that <|>A><|>p. and (|)A <(|)P.
Then by Comment 1, if P chooses separate agencies, she at? Note finally that if w2-z2>l/2, the preference for
tains her ideal outcome; B* = c{($f), pf =pf =0, and separation is strict. In this case, P makes no contractual pay?
= for each Ai. P therefore
<|>* (|)f, weakly prefers to separate ments under separation, but since <|>*^<j)c* (?C;PC =0)
tasks. This preference is strict unless A would choose for any budget Bc> a positive contractual payment is neces?
<|>c*(?c,0c) = <|>pwhen Bc=c($p) and 0C = 0. To deter? sary to induce A to choose <J>*. ?
mine conditions under which this is possible, consider As
maximization problem. There are five constraints; (i)
Bc>c($\ (ii) (h<l, (iii) (|)2<1, (iv) (h>0, and (v) Proposition 3. // <|>A< <|>p (alternately, ifwt- -z{<\ for all
?2 >0. Because it is assumed that <|>A><|>p and <|>p< 1, tasks i) then P consolidates tasks.
(ii)-(v) are slack, and I disregard them. Given budget Bc
and 0C = 0, As Lagrangian is:
Proof. It is sufficient to show that P does at least as well un?
der the consolidation subgame as under the separation
subgame. Let ?*, P*, and(|)* represent the equilibrium
Differentiation yields: strategies in the separation subgame. Now let P offer budget
Bc' =B? +B2 and the following contract 0C' to the con?
= solidated
w.-z.-2ml^l-2miî<0y agency, A: p00'=0, p10'=pf, p01'=Pi\ and
pn'=Pf + p2*. Then A's objective is: uA(<MC',0C') =
so that tyx= ?2m2(wx-zx)/(mx(w2 Since <|>P=<|>A
-z2)). I>^-+z-(l-^)-m^]5C,+Pi>i+p2>2- Since uAC)
when wx-zx =w2-z2=l, =
<|>c*(c(<J>p),pc 0) = <|>p only if is concave, first-order conditions are sufficient to character?
wx-zx=w2-z2>\. This is clearly a measure zero case. ize a maximum. A would therefore choose, if feasible, (|)f' =
Second, suppose without loss of generality that tyxA> ?xp max{0,(wi-z-+pn/2mf}. By Comment 1, <|>f'=<|>* for
and (|)A < ?2.1 first establish two necessary conditions for all tasks /. Thus, <(>f' is feasible under budget Bc'. P there?
consolidation to be preferred to separation. fore receives the same utility by offering budget Bc' and
*
Claim 1: P prefers consolidation only if <|>f =(|)1p. Sup? contract Pc in the consolidation game as she receives in
pose not. I construct contracts for separated agencies such equilibrium of the separation subgame. Therefore, P weakly
*
that <|>f = tyf that result in strictly superior outcomes for P. prefers to consolidate tasks.
Define contracts 0/ ,02') so that pf =p}' =p2r =0, and I show finally that if additionally w{ ?z{ >-l for all
P? =max{0,(p01* -p00*)(l-(|)f) + (pir -p10*)<|>f}. Then tasks f, this preference is strict. Suppose otherwise; i.e., that
and =
[2], for budgets Bx cx(tyxp) and B2 = c2 ($*),
by [1] Pc is optimal. Calculating 3?p/9p11, we obtain:
A THEORY OF JURISDICTIONAL ASSIGNMENTS IN BUREAUCRACIES 3JJ
Builder, Carl H. 1989. The Masks of War: American Military

Ml ,?(?)a<t,f -Pu :.m: Ml Styles in Strategy and Analysis. Baltimore: Johns Hopkins
2m,<ft *! ?+<t>r
ap11 fj ap11 ap11 ap11 ap11 University Press.
Calvert, Randall L., Mathew D. McCubbins, and Barry R.
'Ml fcc-Ml_*c?Ml "A Theory of Political Control and Agency
-?t>r Weingast. 1989.
"
apu apn ap?j Discretion American Journal of Political Science 33:588-
? 611.
_R01 Ml Ml
1 Cyert, Richard M., and James G. March. 1963. A Behavioral
ap11 apn ap11
Theory of the Firm. Englewood Cliffs: Prentice-Hall.
Ml ?Wi de Figueiredo, Rui. 1998. "The Strategic Use of Agency Jurisdic?
_poi 1 Y1 tions: Applications of Multi-Task Agency Theory to the Un?
apn apn ap1 ap11
derstanding of Bureaucracies." Presented at the Annual
Meetings of the American Political Science Association.
Since P00' =0 andp11' =p10' +p01', dup/d$n evaluated at
Derthick, Martha. 1990. Agency Under Stress: The Social Secu?
Pc simplifies to: rity Administration in American Government. Washington
D.C.: The Brookings Institution.
dup($c'A , .^tiP Dixit,AvinashK. 1995. The Making of Economic Policy: A Trans?

-X?-i= action Cost Politics Perspective. Cambridge: MIT Press.
l 1-p10
H -2m^'^_
m
apn -'ap11
Epstein, David, and Sharyn O'Halloran. 1999. Delegating Pow?
ers: A Transaction Cost Politics Approach to Policy Making
+1 2V2 / Under Separate Powers. New York: Cambridge University
(i-por
ap11 Press.
> 0 if "Evaluation of the Organizational Structure and Management

Substituting from Comment 1, 1 + P10 -2m${
Practices of the California Department of Transportation -
W[-z, <1, and likewise l + P01'-2m2<]>2 >0 if w2 -z2
Progress Report on Caltrans' Implementation Efforts." SRI
<1. Thus, dup/d$n>0 if a<|)f /apu>0 for all j. Now International, March 1996.
a<t>f*/apH evaluated at Pc simplifies to: Hammond, Thomas H. 1979. Jurisdictional Preferences and the
Choice of Tasks: Political Adaptation by Two State Wildlife
,c* or Departments. Doctoral Dissertation, University of Califor?
dtf _ w2 +P1
nia at Berkeley.
ap11 4mxm2
Heimann, C. F. L. 1993. "Understanding the Challenger Disas?
ter: Organizational Structure and the Design of Reliable
Then by Comment 1, 3(t)f*/^PU >0 for ^2 ~zi >-l. A
Systems." American Political Science Review 87:421-435.
symmetric statement obtains for d?2*/d$n. Thus if
Holmstrom, Bengt, and Paul Milgrom. 1991. "Multi-task Prin?
for all tasks i, 3wp(pc>)/3pn >0.P would
w,--*:,?? (-1,1) cipal-Agent Analyses: Incentive Contracts, Asset Ownership
achieve a higher utility by choosing P11 > Pf + P2*: contra? and Job Design." Journal of Law, Economics, and Organiza?
diction. Therefore Pc cannot be the optimal contract, and tion 7:24-52.
P can achieve a strictly higher utility under consolidation Kaufman, Herbert. 1960. The Forest Ranger: A Study in Admin?
istrative Behavior. Baltimore: Johns Hopkins University
than with the optimal contract under separation. ?
Press.
Kaufman, Herbert. 1976. Are Government Organizations Im?
mortal*'Washington, D.C.: The Brookings Institution.
Kaufman, Herbert. 1981. The Administrative Behavior of Fed?
References eral Bureau Chiefs. Washington, D.C.: The Brookings Insti?
tution.
Banks, Jeffrey S. 1989. "Agency Budgets, Cost Information, and Landau, Martin. 1969. "Redundancy, Rationality, and the Prob?
Auditing." American Journal of Political Science 33:670-699. lem of Duplication and Overlap." Public Administration Re?
Banks, Jeffrey S. and Barry R. Weingast. 1992. "The Political view 29:346-358.
Control of Bureaucracies Under Asymmetric Information."
Mayhew, David. 1991. Divided We Govern: Party Control, Law-
American Journal of Political Science 36:509-524.
making, and Investigations, 1946-1990. New Haven: Yale
Bendor, Jonathan B. 1985. Parallel Systems. Berkeley: University University Press.
of California Press. McCubbins, Mathew, Roger Noll, and Barry R. Weingast. 1987.
Bendor, Jonathan B. and Terry M. Moe. 1985. "An Adaptive "Administrative Procedures as Instruments of Political Con?
Model of Bureaucratic Politics." American Political Science trol." Journal of Law, Economics, and Organization 3:243-
Review 79:755-774. 277.
Bendor, Jonathan B., Serge Taylor, and R. Van Gaalen. 1987. Miller, Gary J., and Terry M. Moe. 1983. "Bureaucrats, Legisla?
"Politicians, Bureaucrats, and Asymmetric Information." tors, and the Size of Government." American Political Science
American Journal of Political Science 31:796-828. Review 77:297-322.
378 MICHAEL M. TING
Moe, Terry M. 1989. "The Politics of Bureaucratic Structure." Silver, Jay. 1996. "Influencing Agency Policies Through Con?
In Can the Government Govern? ed. J. E. Chubb and P. E. gressional Appropriations." Mimeo. Stanford GSB.
Peterson, Washington, DC: The Brookings Institution. Tullock, Gordon. 1965. The Politics of Bureaucracy. Washington,
Niskanen, William A. 1971. Bureaucracy and Representative D.C.: Public Affairs Press.
Government. Chicago: Aldine-Atherton. Wilson, James Q. 1989. Bureaucracy. New York: Basic Books.
Quirk, Paul J. 1980. "Food and Drug Administration." In The
Politics of Regulation ed. J. Q. Wilson, New York: Basic
Books.

A Theory of Jurisdictional Assignments

Uploaded by

Copyright:

Available Formats

You might also like

A Theory of Jurisdictional Assignments

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Theory of Jurisdictional Assignments

Uploaded by

Copyright:

Available Formats

A Theory of Jurisdictional Assignments in Bureaucracies

Author(s): Michael M. Ting

Michael M. Ting University of North Carolina at Chapel Hill

Michael M. Ting is Assistant Professor of Political Science, University of North Carolina

Figure 1 Game Sequence

Principal Agency Nature

A's policy indifference curve

subject to: c(<|>c)<?c,<|>c >0,<|>c <1.

P's policy indifference

jective function is quasilinear and strictly concave,

The Separation and

Comment 1. The unique subgame perfect Nash equilib?

cies and agencies cannot transfer utility amongst them?

contract ($,(*}), Ai maximizes w^jB^p^p})) sub? (^-Z,-+1)2

Now consider P's problem. Two observations simplify ?? = W: - Z:

0 Expression [2] also makes clear why at least one pay?

+ Comment 2. If contracts are independent, then P weakly

.C?^CoC\_W2-22+P01-P00+(Pn-P10-P?1+P00)<t>1 The intuition behind this result is straightforward.

ily seen that the "first-order" effect of increasing A's pay?

Figure 3 Versus Consolidation - 1.5. Here, the con?

subgame for three values of w2 (0.25, 0.75, and 1.5) as a

Figure 4 Comparing Jurisdictional Assignments

The model derives intuitive

Builder, Carl H. 1989. The Masks of War: American Military

dup($c'A , .^tiP Dixit,AvinashK. 1995. The Making of Economic Policy: A Trans?

> 0 if "Evaluation of the Organizational Structure and Management

You might also like