You are on page 1of 19

Bureaucrats and Public Procurement

Author(s): Dieter Bös


Source: FinanzArchiv / Public Finance Analysis, Vol. 58, No. 2 (2001), pp. 103-120
Published by: Mohr Siebeck GmbH & Co. KG
Stable URL: http://www.jstor.org/stable/40912958
Accessed: 14-12-2016 06:06 UTC

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.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

Mohr Siebeck GmbH & Co. KG is collaborating with JSTOR to digitize, preserve and extend access to
FinanzArchiv / Public Finance Analysis

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
FinanzArchiv vol.58 no. 2 103

Bureaucrats and Public Procurement


Dieter Bos*

This paper deals with a Niskanen type of public-procurement agency.


We show that the procurement game should be separated into an
investment game and a project game, the first game to be played
before nature determines the actual realizations of benefit and costs
of the project, the second game to be played afterward. In the first
game the relationship-specific investments of agency and seller are
determined, in the second game the decision on the production of the
project is taken. In contrast to many other incomplete-contract papers,
in our Niskanen setting it is meaningless to write only one contract
which refers to both investment and production. We obtain the
following results with respect to second-best welfare optimality:
optimal procurement of the project can be attained under relatively
weak assumptions; optimal investments of the seller (and only of the
seller) may result under special circumstances; optimal investments of
both agency and seller cannot be reached - both agents will overinvest.
(JEL: D23, D73, H57)

1. Introduction

This paper deals with the purchase of a public project. The parliament spon-
sors a procurement agency which, in turn, pays a private seller who provides
his services. As in Niskanen's (1971, 1975) model1, the parliament is willing
to appropriate money for the completion of the project up to its evaluation
of the social benefit of the project. The agency is a budget-maximizing bu-
reaucrat2. It wants to maximize its power, which can be achieved best if it
spends as many dollars as possible. The private seller is a profit maximizer.

* This paper was finished while I was visiting EPRU, University of Copenhagen. The hos-
pitality of EPRU is gratefully acknowledged. The paper benefitted from comments by
an anonymous referee of this journal and by T. Ihori, Tokyo, C. Keuschnigg, St. Gallen,
K. Leipold, Washington, D.C., C. Liilfesmann, Bonn, S0ren B. Nielsen, Copenhagen, and
Peter Birch S0rensen, Copenhagen.
1 For a simple explanation of Niskanen's model see Mueller (1989, pp. 250-254).
2 The assumption of a budget-maximizing procurement agency is in contrast to several
other incomplete-contract models on public procurement, for instance Bos and Liilfes-
mann (1996) and Bos (2000), which assume a welfare-maximizing procurement agency.

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
104 Dieter Bos

Given these objectives of parliament, agency, and seller, we expect the


sult to be far from welfare optimality. However, in this paper we show th
under particular circumstances second-best welfare optimality may result.
At the beginning of their relationship the procurement agency and
private seller face a situation of uncertainty about the actual realizations o
benefit and costs of the project. However, the probabilities of higher bene
fit and of lower costs can be influenced by relationship-specific investmen
of the seller and of the agency. The seller's investments consist of spec
innovative activities, for instance the development of a particular tech
ogy for military interceptor planes, or a special design for a particular bu
ing for elderly handicapped people. However, the agency is also intere
in investing because the parliament's benefit of the project determines th
agency's budget and, therefore, the agency directly benefits from an
crease in the probability of higher benefits. The specific investments of t
agency typically refer to complementary goods which are essential in ensu
ing the success of the project. By way of example, one can think of invest
ment in the development of a new approach control radar which ma
landing of a military interceptor plane more secure, or the development o
special infrastructure if a new hospital is to be built.
The seller provides two services: his specific investments and the produc
tion of the project. We shall assume that it is the same seller who prov
both services3. This is the most interesting problem. Otherwise, in a f
stage seller A would be paid for the investment costs and in a second s
seller B would be paid for the project, and these two stages would not
connected. It is just the connection between the investment and the projec
stage which constitutes the really interesting problem. The relationship-sp
cific investments have to be made before the veil of uncertainty about th
actual realizations of benefit and costs of the project is lifted by a draw o
nature. But the final distribution of rents will only be determined af
nature's draw, when it is finally decided whether the project is carried ou
or not. However, at this moment the costs of the specific investments
sunk. In standard incomplete-contract models this implies the well-kn
hold-up problem4: when the rents are finally distributed, no party can ar
that it deserves a higher share of the ex-post surplus because of its high s
cific investments. This is anticipated by the buyer and by the seller and c

3 In various public-procurement guidelines it is forbidden that the same firm provi


both the R&D investments and the completion of the project. These provisions aim
a promotion of competition because many smaller firms may be specialized in R&D
in project implementation.
4 The hold-up problem was first articulated by Klein et al. (1978) and by Williamson (1
1979).

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
Bureaucrats and Public Procurement 105

stitutes a tendency toward underinvestment. In contrast, there is a tenden-


cy toward overinvestment in this paper. The budget-maximizing agency i
interested in high expenditures for investment which, in turn, will also in-
crease its expenditures for the project because high investments lead t
higher benefits, which increase the parliament's willingness to pay for th
project. The agency's overinvestment increases the expected production
rent of the seller and, therefore, the seller also has an incentive to overin-
vest.

If only one seller provides his services to the agency, it seems natural that
only one contract is written which refers to both relationship-specific invest-
ments and the production of the project. However, in our model, agency
and seller should write two contracts. Their contractual relationship begins
with a contract referring to the specific investments of the seller. Then the
parties wait until the draw of nature determines the parliament's willing-
ness to pay for the project (project budget). On the basis of this budget the
parties write a second contract which refers to the production of the pro-
ject. The separation into an investment contract and a project contract is
rooted in the budget-maximizing attitude of the bureaucratic procurement
agency: it would be meaningless to stipulate a fixed project price before the
draw of nature determines the project budget, which is always fully exhaust-
ed by the agency5.
The setting of this paper is in contrast to the usual theory of incomplete
contracts, such as Hart and Moore (1988), Aghion et al. (1994), and Noldeke
and Schmidt (1995). In these papers the buyer and the seller write one and
only one contract at the beginning of their relationship, that is, before the
agents' uncertainty about benefit and costs is dissolved by a draw of nature.
Typically, in this ex-ante contract two prices are stipulated: a price p1 to be
paid if the project is completed, and a default price p0 if this is not the case.
After nature has revealed the actual realizations of benefit and costs, the
prices px or pQ may be renegotiated. In contrast, in our paper the two
prices are contracted upon at different stages of the game. p0 is stipulated
in a contract which is signed when the procurement agency starts with the
planning of the particular project; it is not a default price, but is paid or en-
cashed by the agency when the investment stage of the game ends./?! is stip-
ulated in a contract written after nature revealed the actual realizations of
benefit and costs. Since both contracts are written under certainty, neither

5 In the practical application the procurement agency should enter into a public-private
partnership, writing a long-term skeleton contract which states that the parties' total re-
lationship consists of two short-term contracts: if the investment contract is satisfactor-
ily executed, agency and contractor continue in their cooperation and write the project
contract.

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
106 Dieter Bos

agency nor seller face any incentive to change any price at a later st
there is no renegotiation in our model.
The paper is organized as follows. Section 2 presents the sequencin
the game, the observability and verifiability assumptions, a detailed desc
tion of the interplay between the parliament and the procurement agenc
and, finally, the second-best benchmark which serves as the basis of
parison for the actually played game. Then, in section 3 we deal with
equilibrium analysis. We begin with the project game, where parliament
agency determine the price to be paid for the project, and the seller deci
whether to sign the project contract or not. Subsequently, we analyz
investment game, where agency and seller choose their relationship-speci
ic investments on the basis of an investment contract which, in turn, is b
on the parliament's investment budget. A brief summary concludes.

2. The model

2.1. Benefits, Costs, and Investments

If the public project is carried out, the parliament's benefit is B, and the pro-
duction costs are C. At the beginning of the game, the actual realizations of
benefit and costs are not yet known; they are determined by a draw of na-
ture at a later stage. This random draw selects the actual realizations of ben-
efit and costs from the following lists of deterministic variables:

B = Bi<...<Bi<...<BI = B' />2, (1)

C=C1>...>Cj>...>Cj = C' J>2. (2)


The probability that a particular value of benefit or
on relationship-specific investments a of the agency
convenience, these investments are normalized
Following Hart and Moore (1988) we specify the

jii(a) = ajzt + (l-a)jtj (benefit), (3)

Oj(e) = eof + (l-e)oj (costs). (4)


The probability distributions Jt+ and n~ are define
jzpjtj is increasing in / (monotone likelihood ratio pr
o+ and o~ are probability distributions over (Clv.., Q
ing in j. A particular choice of investment determines a
of two probability distributions and our assumptions im
vestments increase expected benefit and reduce expec

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
Bureaucrats and Public Procurement 107

ly. From the definitions of the probability distributions jv(a) and o(e), it fol-
lows directly that the first derivatives are constant and that they sum up to
zero:

7t' = Jit - n~i ; o) = o) - aj ; £'=1 x' = £y=1


The relationship-specific investments are costly. We define in
functions ji{a) and ip(e) and assume that both functions are
arguments. We also assume that the Inada conditions are fu
Until now we have only treated the direct costs which re
completion of the project and from the specific investments.
social evaluation of costs matters, there are indirect costs whi
taken into account. This paper starts from the realistic as
lump-sum taxation cannot be applied in practice. Hence, wh
ernment spends money, this money comes from distortion
any partial-equilibrium model, therefore, the shadow costs
A > 0 must be explicitly considered: if the government app
ject budget of B, the actual social costs are B (1 + A) 6. The exc
sum taxation puts our paper into the realm of second best. Th
ever we speak of welfare optimality, we mean second-best wel
ly.

2.2. The Stages of the Game

The sequence of events is illustrated in table 1.


The relationship between parliament and agency is characterized by two
budgets. The investment budget refers to the investment costs of the agency
and of the seller. The project budget finances the costs of the completion of
the project. Similarly, the relationship between agency and seller is gov-
erned by two contracts. The investment contract refers to a lump-sum pay-
ment stipulated at the beginning of the relationship. The project contract
refers to the completion of the project.
Why is this the only meaningful setting in our model? Assume, for the
sake of the argument, that at the beginning of their relationship ("ex ante")
agency and seller write a single contract which refers to both investment
and production of the seller. Which price for the completion of the project
should be set in such a contract? The procurement agency must never con-

6 In a theoretical setting Laffont-Tirole's (1993) use of A in their welfare considerations


has become particularly well-known. Typical empirical estimations lead to 0<A<l,
where values between 0 and 0.5 can be found more often, at least in developed coun-
tries; see, in particular, Ballard et al. (1985, p. 136). For a brief overview of various em-
pirical estimates see Jones et al. (1990, pp. 28-30).

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
108 Dieter Bos

Table 1

Sequencing of Events

date 0 investment budget


date 1 investment contract, agency's investments investment game
date 2 seller's investments

date 3 draw of nature

date 4 project budget


date 5 project contract project game
date 6 completion of project

tractually promise to pay a price which may exceed its budgetary possibil-
ities at the moment when the project is finally carried out. Therefore, the
agency could only offer a price which equals the lowest budget which could
arise after the draw of nature. However, since the agency is a budget max-
imizer, no party would believe in this low price. Both agency and seller
would correctly anticipate that later on the agency will enforce a higher
price, thus exhausting the total budget appropriated by the parliament af-
ter the draw of nature7. Therefore, any ex-ante project price is meaningless.
Meaningful, however, is the ex-ante stipulation of a lump-sum payment p0
which is a compensation for the specific investments of the seller, or a pay-
ment for his participation in the procurement game. This explains the sep-
aration of investment contract and project contract. As already mentioned
in the introduction, there is no renegotiation in this model since both con-
tracts are written under certainty.
After motivating the separation of contracts, let us now consider the sev-
eral stages of the game (see table 1). At date 0 the investment game is
opened: the parliament appropriates the investment budget b > 0. A nega-
tive budget cannot be appropriated because of the procurement agency's
limited liability. The investment budget covers the costs of the agency's in-
vestments ji{a) and a lump-sum payment p0 to be encashed or paid by the

7 Any project budget which is set before the draw of nature would be adjusted by down-
ward or upward renegotiations after the draw of nature, because the parliament would
not be willing to pay more than its actual benefit from the project (downward renego-
tiation) and the agency, knowing the actual realization of the parliament's benefit will
always go for a budget which fully exploits the parliament's willingness to pay (upward
renegotiation). Anticipating this, the only meaningful procedure for the parliament is to
set the project budget after the draw of nature.

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
Bureaucrats and Public Procurement 109

seller; it is split into its two components by a decision of the procuremen


agency at date 1. Since choosing jii(a) implies the choice of a, the agency acts
as a Stackelberg leader in the investment game; it chooses its investment
prior to the seller. The agency's decision on how to split the investment bud-
get into jii(a) and p0 is the basis of the investment contract which is signed
by the seller unless the stipulated lump sum p0 violates his participation
constraint. This constraint is based on the agency's commitment not t
choose another seller for the project game which definitely must be give
at date 1 8. Then, at date 2 the seller chooses his relationship-specific invest-
ments. The agency pays or encashes the lump sum p0. The investment game
ends. Only then, at date 3 the actual realizations of benefit and costs are
determined by a draw of nature. The uncertainty is thus dissolved and a
date 4 the project game is opened: the parliament sets the project budge
according to nature's draw. The agency fully extracts this budget which
therefore, is available for the seller as the price to be paid for the complet-
ed project. On the basis of this price and his knowledge of the actual cost
at date 5 the seller decides whether to sign a contract about the completion
of the project. If he makes a negative decision, the game ends. Otherwise
at date 6 the project is produced, and the seller is paid for the complete
project9.

2.3. Observability and Verifiability

The supports of B and C, and the probabilities it (a) and o(e) are common
knowledge, that is, they are known to parliament, procurement agency, and
seller. The same holds for the investment-cost functions.
The actual realizations of benefit and costs, as determined by nature, are
observable as follows: The benefit B is known to the parliament and to the
procurement agency. It does not matter whether it is also known to the
seller. The costs C, however, in this model are assumed to be private knowl-
edge of the seller, that is, the producer is the only one who knows the pro-
duction costs. In Niskanen's original two-person model of a sponsor and a
bureaucrat, the bureaucrat produces the good and only he knows the costs.
We have extended Niskanen's model to a three-person setting, and it seems
natural to assume that, once again, only the producer knows the costs. Note,
however, that the results of the paper remain unchanged if we assume that
both seller and procurement agency know the costs (as long as the parlia-

8 Recall footnote 5 above.

9 Disputes on delivery and on payments would be decided upon after the end of a game,
either the investment game or the project game. However, in the subgame-perfect equi
librium no such disputes occur.

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
110 Dieter Bos

ment does not know them)10. The actual amount of relationship-specific in


vestments which the agents choose may be private or public informat
this does not influence the results of the paper.
The above informational assumptions show that the parliament does
employ the agency to gain information about the production costs. H
ever, the agency is the only one which is able to prepare the project by ap
propriate relationship-specific investments. (Moreover, there is always
usual justification that in practice a parliament cannot do everything itsel
it always needs a special agency which looks into the particulars of a p
lic project.)
Let us next turn to the verifiability assumptions which are important be-
cause any contract can only be conditioned on variables that are verifiable
before a court. In this paper we assume that the relationship-specific invest-
ments a and e are non-verifiable since they are effort levels which contain
many subjective elements. Further, since the costs C are not observable,
they are non-verifiable, and, as usual in most models on incomplete con-
tracts, the benefit B is non-verifiable. After all, the evaluation of the bene-
fit of the project is a subjective attitude of the parliament. Note, however,
that a budget is always an officially published figure, which therefore can
be verified before a court. This holds even if the budget functionally de-
pends on the non-verifiable benefit, for instance, if the parliament appro-
priates a budget of Bl(' + A). Furthermore, the completion of the project
and all payments are verifiable.

2.4. The Interplay between Parliament and Procurement Agency

Parliament and procurement agency are modelled according to Niskanen


(1971, 1975). A modern formulation of the Niskanen model has to impute
to the parliament a lexicographic preference ordering with respect to allo-
cative efficiency and payments11. The parliament first wants to attain allo-
cative efficiency, therefore at date 4, it favors the completion of the project
iff

Bi>{' + k)Cj. (6)

10 This is proved at the end o


11 For the use ot such a preterence tunction see also tfos and Luitesmann (^yyoj. /'
precise formulation of the parliament's utility function <J> would be as follows: one
must define 0 = <£(*,;y), where x denotes the level of allocative efficiency and y the pay-
ments to the procurement agency. In this formulation, lexicographic preferences over
these two arguments can be expressed as follows: &1><P2^=> (a) xi > xi or (P) x' - X2
<mdyx<y2.

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
Bureaucrats and Public Procurement 111

If the project is completed, its costs C;- have been realized. Unfortunatel
when it has to appropriate the project budget at date 4, the parliament can-
not observe the actual cost realization which is private information of th
seller. Since payments are only the second-ranked part of the parliament
lexicographic preference ordering, however, it does not mind if it has to pay
more than the actual costs as long as the first-ranked objective of allocative
efficiency is not violated12. Therefore, the parliament is willing to spend up
to Btl{' + A) dollars if the project is carried out. This can be captured by the
following constraint:

/?•
project budget < -- l- . (7)
1+A

The parliament's lexicographic preference ordering is exploited by the bud-


get-maximizing procurement agency. Since the agency wants to spend as
many dollars as possible, it fully extracts the parliament's willingness to pay:
the agency makes a take-it-or-leave-it offer to the parliament which equates
the budget to the parliament's social evaluation of the project. (This take-it-
or-leave-it offer is the core of Niskanen's theory of bureaucratic behavior.)
The parliament accepts this offer and appropriates a budget of Bt/(1 + A) dol-
lars conditional upon the completion of the project.
Let us now step back and consider the interplay between parliament and
procurement agency at date 0, when the investment budget is to be appro-
priated. The first-ranked part of the parliament's lexicographic preference
ordering aims at efficient investments of both procurement agency and
seller. The parliament anticipates the subgame-perfect continuation of the
game, in particular the splitting of the investment budget by the agency
(b = Li(a) +/?o)> and the final decision on the completion of the project. Since
payments rank only second, the parliament would be willing to choose any
budget b which maximizes welfare <W,

W= lilj ^(«)^(^)[A-C7(1 + A)]-(1 + A)[^) + ^)], »>


5,->Cy(l+A)

with respect to a and e13. Note, however, that the parliament has to con-
sider the agency's limited liability which requires b > 0.

12 Therefore, the parliament does not implement any direct mechanism in order to induce
the seller to reveal the actual cost realization.
13 The respective first-order conditions are necessary and sufficient for a unique and inter-
ior solution {a*,e*}>0. Formally, the existence of an interior solution is ensured since
expected welfare as defined in (8) is concave in the investments and the Inada condi-
tions with respect to fi(a) and ip(e) are assumed to be fulfilled. The maximum is unique
if one assumes | Wkk | > | (Wkm ',k,m e {a, e).

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
112 Dieter Bos

2.5. A Benchmark

It is a well-known paradigm of second-best theory that in the presence of


distortions it is meaningless to strive for a first best. Therefore, in an econ-
omy without lump-sum taxation it would be meaningless to refer to a first-
best benchmark. Such a first-best welfare optimum would not be optimal in
a second-best world. Therefore, in this paper we refer to a second-best
benchmark where the shadow costs of public funds are taken into account.
We consider a fully informed parliament which wants to maximize welfare.
It does not engage in public procurement, but 'does it alone'. Both invest-
ment costs and project costs are born by the parliament, paid from distor-
tionary taxation, and raise shadow costs.
Applying backward induction, let us first define project efficiency which
refers to the decision made with respect to the completion of the project.
Project efficiency requires that the project is carried out if and only if this
increases welfare. Since we deal with procurement of an indivisible good,
the project, let q = 1 and q = 0 be the quantities to be procured. Therefore,
project efficiency requires:

q*=l <=> Bi>Cj(l + X), (9)


q*=0 t=> Bi<Cj(l + X), (10)
where Bt and C; are the realizations of ben
the draw of nature. The investment costs are
fore, do not influence the parliament's decisi
Second, we define investment efficiency wh
mal choice of the investments a and e. Anticip

(fl*e*)eargmaxflte'W>= X; Xy Jti
£;>C,(1+A)
-(1 + A)b(a) + V(e)].

We obtain the following first-order conditions:

K = 0-B&Cj(l+k)
X;Xy ^g;(g)gg'"fi(a'+A)=^(fl)»
^i + A'
+ 02)

<We = 0: X;Xy jti(a)(/j l~ ' * =V'(<0- 03)


These conditions are necessary and sufficient for a unique and interior so-
lution {#*,£*} >014. Note that the benchmark implies that at date 0 a gen-

14 Compare footnote 13.

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
Bureaucrats and Public Procurement 113

eral efficiency constraint ^>0 holds. Therefore, at date 0 the expected net
value of the project should be weakly positive, otherwise the project should
not be started at all.

3. Equilibrium Analysis

3.1. The Project Game

At date 5, the agency knows the project budget which allows for an expen-
diture of Pi dollars for the completed project. The agency is always interest-
ed in having the project carried out because otherwise its budget is reduced
to zero. Therefore, it wants to spend exactly pt dollars: the agency is not
interested in spending less, and it is unable to spend more. Therefore, at
date 5 the procurement agency has no freedom of decision and offers the
seller a procurement contract with price pt to be paid if the project is car-
ried out. Note that the agency cannot take the money ph pay part of it to
the seller and retain the rest for personal emoluments. The agency rather
has to forward all of the money to the seller, since payments are verifiable
and, accordingly, the agency could be sued if it used parts of the budget for
other purposes than for the particular public project.
Since the agency has no freedom of decision, at date 5 only the seller's
decision counts. He compares two situations: if the project is carried out, he
gets Pi and faces the production costs Cy. Otherwise, he has no production
costs, and does not get anything. Therefore, the seller will sign the project
contract if and only if15

Pi-Cj>0. (14)
Let us now step bac
This is the typical s
his model of bureau
learned the actual re
that the parliament
completion of the p
theagency makes a t
for the completed pr
the agency and the p

(£;/(l + A) if q =

15 This is also the partic


of the agency is /?/>0, w

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
114 Dieter Bos

Note that Bt is the project evaluation of the parliament, whereas pt


are the social project costs which the parliament faces. The actual p
costs Cj do not matter for the parliament. It does not know them, and t
are not explicitly relevant for the project budget, because the Niskanen
liament cares only about the evaluation of the project and not abou
seller's costs. When setting the project budget at date 4, the parliament
ticipates that the agency will offer the seller a project contract with pr
and that, at date 6, #,7(1 + ^) dollars will be paid to the seller if the pro
is carried out, and 0 dollars if it is not carried out.
Let us finally investigate whether the decisions of parliament, agency
seller attain project efficiency. The seller signs the contract iff pt>C
project budget has fixed the expenditures for the completed project
Pi = 5,7(1 + A). Hence, although the seller does not know it, he signs the
tract iff

which is equivalent to

5/>Cy(l + A). (17)


Therefore, the seller signs th
only if this is welfare-optim
efficiency is guaranteed.
There is an important differ
and Niskanen's original mod
reaucrat is expanded until t
to the costs of the project16.
the seller, and he retains th
and the production costs as
agency's budget-maximizing
attain is given by the parliam
will always make sure that t
Since payments are verifiab
budget, and the seller will get
Cj. It is remarkable that thi
agency knows the production
will not plead for a lower budg
And it will have to pay all o
verifiable. Therefore, a procu

16 Ignoring Niskanen's demand-co

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
Bureaucrats and Public Procurement 115

duction costs will 'throw away' this information because it could only lea
to a reduction of the project budget which is unwanted by the agency. (This
shows that all of our results hold as well if the agency knows the produc
tion costs, which proves our statement in subsection 2.3 above.)

3.2. The Seller's Investment Decision (Date 2)

When choosing his relationship-specific investments the seller anticipates


the continuation of the game and maximizes the expected profit given the
agency's investments. His objective function at date 2 is as follows:

Us= 1,-Iy Jit (a) oj (e) Bi " ^(1 + + A) + po - V(e) , (18)


Bj>Cj(i+2l) ii + /v

where the lump sum p0 is included in the objective function becaus


encashed or paid by the seller at date 2. The seller's choice of specif
vestments is determined by the following marginal condition:

1/1/ Jti(a)o'jB^^^ + = ip'(e). (19


Bi>Cj(i+x) l1 + A;

The solution of this condition is the seller's investment function e(a) which
describes the adjustment to the agency's investment choice. We have de(a)l
da > 0: if the agency increases its investments, this implies an increase of the
left-hand side of (19), which induces a higher value of the right-hand side,
ip'(e), and, hence, a higher investment e. In other words, an increase in the
agency's investment increases the expected production rent of the seller
and, therefore, the seller has an incentive to invest more.
Note that the seller's marginal condition (19) is equivalent to the bench-
mark marginal condition (13). Therefore, the seller's investment choice will
lead to investment efficiency iff the procurement agency has chosen effi-
cient investments at date 1. Otherwise, if the agency has chosen positive but
inefficient investments, the seller will overinvest if the agency overinvested,
and underinvest if the agency underinvested (since de(a)/da>0). Unfortu-
nately, only by chance will the agency choose efficient investments, as we
shall see in the analysis of the agency's decision at date 1.
However, let us mention a particular exception. Assume that n(a)-ji for
all a. Then, the agency has no incentive to invest and a - 0. In this particu-
lar case of one-sided investments we attain investment efficiency, and since
project efficiency is always guaranteed, the (second-best) welfare optimum
is achieved.

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
116 Dieter Bos

3.3. Agency Investments and the Investment Contract (Date 1)

At date 1, the agency splits the parliament's investment budget into a bud
get for its own investments fi (a) and a lump sum p0 to be paid to, or
cashed from, the seller. We assume that the agency is in a stronger positio
than the seller. Therefore, the investment contract at date 1 is based
take-it-or-leave-it offer the agency makes to the seller. Since the agency i
a budget maximizer, it wants to maximize its utility consisting of the sum
its two budgets,

UA= I,I;- xMojW-^j + nia). (20)


Bi>Cj{'+k)

When maximizing this objective function with respect to a andpo> the agen-
cy has to consider the following two constraints. First, at date 1 the agency
knows that the parliament's investment budget amounts to zero (as we shall
prove when stepping back to the discussion of the parliament's decision at
date 0):

P0 + /x(«) = 0. (2D

Second, the agency has to c

Us>0. (22)
What are the res
ed in attaining t
er investments
project game an
ponents of the
tracts all of the
that is, by requi
ment game. Us=

A) = I/I; *i
Bt>c7(i+A) v1+A;
where the value of e(a) is given by equation (19). The payment of
is then used to finance the investment costs of the procurement ag

/*(*) = | A) | . <24>
Substituting (23) into (24), we obtain an e
fines the agency's investment choice.

17 The agency's participation constraint does not


equation (20), the agency is always willing to sign

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
Bureaucrats and Public Procurement 117

Figure 1
Overinvestment by Procurement Agency

0 a* a°' a

The background behin


procurement agency e
Why is this the best s
easily be seen if we re

^(fl,e(fl)) = t/5(l + A)-(A) + Mfl))(l + A), (25)


where W(a, e(a)) refers to the fact that the agency controls welfare in a dou-
ble way: directly by its own investment choice at date 1 and indirectly by in-
fluencing the seller's investment choice at date 2. Now recall that the
agency's strategy at date 1 extracts the seller's surplus, Us=0, accepting the
parliament's investment budget p0 + ti(a) = 0. Therefore, welfare is equated
to zero, (W = ^. This implies overinvestment by the agency (and by the seller
because de(a)lda > 0). A simple illustration is presented in figure 1. Since the
investments are extended until welfare is equated to zero, the agency will
choose a = a0, which is an overinvestment.
Formally, this can be proved as follows. Let us first show that the welfare
function W is concave in a. We differentiate (W(a,e(a)) with respect to a

18 Substitute Us, equation (18), into the welfare function <W, equation (8).

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
118 Dieter Bos

and apply the envelope theorem19 to obtain the following result:

, ,/ / xx f>0 for a<a*,


dV(ae(a)) , ,/ / xx = dy_ = K_{1 + k
da da [<0 for a>a*,
where K := I, Xy ;r • a; (e) [£, - Cj (1 + A)]. This term does not d
5(>C,(1+A)

a, hence it is a constant for this proof (this is particularly due to the speci-
fication of the probabilities JT/(«) whose derivative is a constant, see (5)).
Therefore, a constant is reduced by a function (1 + A) in'(a) which is mono-
tonically increasing in a. Since welfare maximization implies a unique inter-
ior solution <2* >0 with d(W/da = 0, the first derivative dW/da must be posi-
tive for a<a* and negative for a>a*. The proof of the concavity of <W is
completed by considering the second derivative

& = -(! + A) Ax»<0. (27)


Finally, note that the welfare function may intersect the ho
where the second point of intersection a would imply W =
ever, a will never be chosen by the procurement agency
ways increase its utility UA by choosing an investment leve

3.4. The Investment Budget (Date 0)

At date 0, the parliament wants to choose the budget b whi


efficient investments of agency and seller. Anticipating th
vestment problem at date 1, therefore, the parliament wou
the agency's investment incentives. This could be achieved b
budget b < 0. Recall from the previous subsection that
written as <W = Us{' + A) - (p0 + ii{a)) (1 + A) [equation (
ment agency will always fully extract the seller's ex-ant
let the parliament choose an investment budget b = (p0 + j*
vents the agency from equating welfare to zero; welfare
be -b. Therefore, the agency will reduce its investment lev
vest efficiently if

b = - <W(a*) (28)

is chosen as investment b
at figure 1. There is a uni

19 In other words, we substit


20 Although the specific invest
information to calculate W(a

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
Bureaucrats and Public Procurement 119

welfare. An investment budget of b < 0 requires the agency to accomplish a


positive welfare level which implies a shift of investments from a0 to the
left. If the required welfare is ^(0*), the agency would invest 0*.
Unfortunately, however, the limited liability of the procurement agency
only allows an investment budget of b > 0. Therefore, the parliament cannot
choose an investment budget which induces the first best. Instead, it chooses
the lowest possible budget, b = 0, and has to accept the overinvestment by the
agency (and by the seller). As already mentioned, this policy implies ^=0,
so it is always guaranteed that the general efficiency constraint W>0 holds
at date 0.

4. Summary

At a first glance, bureaucratic behavior of public-procurement agencies


seems to exclude the possibility of welfare-optimal procurement. However,
as this paper shows, this is not necessarily the case:
First, if we assume one-sided investments by the seller, the seller will
choose welfare-optimal investments and the project will only be carried out
if this is welfare-improving. Hence, we have complete (second-best) welfare
optimality in the case of one-sided investments.
Second, both-sided investments by agency and seller can be attained if
the seller pays a lump sum for the right to participate in the procurement
game and this lump sum is used to finance the investments by the agency.
Both agency and seller will overinvest in this case. This result is driven by
the fact that the bureaucratic procurement agency maximizes its budget giv-
en the parliament's willingness to pay up to its evaluation of the benefits of
the project. Project efficiency will be achieved in this case: after the invest-
ment costs have been sunk, only projects will be carried out whose social
benefit (weakly) exceeds the production costs.
The positive results of the paper are driven by the parliament's choice to
appropriate a project budget equal to the social benefit of the public pro-
ject. The parliament knows that the bureaucratic procurement agency will
never choose a smaller budget than these social benefits (regardless of
whether it knows the production costs or not); it will pass on this budget to
the seller and, therefore, the seller is offered a price equal to the social ben-
efit of the project. Since the seller will only sign the project contract if this
price is at least cost-covering, he decides in favor of the project if this is wel-
fare-improving and otherwise does not sign the project contract. Therefore,
the seller earns a weakly positive ex-post rent which gives the correct in-
centives to induce his project-efficient decision. This ex-post rent, therefore,
must not be taken away from the seller. On the other hand, the procure-

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms
120 Dieter Bos

ment agency can fully extract the ex-ante rent of the seller by stipulatin
negative lump sum to be paid by the seller for the right to participate in
procurement game.

References

Aghion, P., M. Dewatripont and P. Rey (1994), Renegotiation Design with Unverifiable
Information, Econometrica 62, 257-282.
Ballard, C, J. Shoven and J. Whalley (1985), General Equilibrium Computations of the
Marginal Welfare Costs of Taxes in the United States, American Economic Review 75,
128-138.

Bos, D. (2000), Incomplete Contracts in Public Procurement, Working Paper No. 265,
Stern School of Business, New York University.
Bos, D., and C. Liilfesmann (1996), The Hold-up Problem in Government Contracting,
Scandinavian Journal of Economics 98, 53-74.
Hart, O., and J. Moore (1988), Incomplete Contracts and Renegotiation, Econometrica 56,
755-785.
Jones, L. P., P. Tandon and I. Vogelsang (1990), Selling Public Enterprises, Cambridge/Mass.
Klein, C, R. G. Crawford and A. Alchian (1978), Vertical Integration, Appropriable
Rents, and the Competitive Contracting Process, Journal of Law and Economics 21,
297-326.

Laffont, J.-J., and J. Tirole (1993), A Theory of Incentives in Procurement and Regulation,
Cambridge/Mass.
Mueller, D. C, 1989, Public Choice II, Cambridge.
Niskanen, W. (1971), Bureaucracy and Representative Government, Chicago.
Niskanen, W. (1975), Bureaucrats and Politicians, Journal of Law and Economics 18,
617-643.
Noldeke, G, and K. M. Schmidt (1995), Option Contracts and Renegotiation: A Solution
to the Hold-up Problem, Rand Journal of Economics 26, 163-179.
Williamson, O. E. (1975), Markets and Hierarchies: Analysis and Antitrust Implications,
New York.
Williamson, O. E. (1979), Transaction Cost Economics: The Governance of Contractual
Relations, Journal of Law and Economics 22, 233-261.

Dieter Bos

University of Bonn
Department of Economics
Adenauerallee 24-42
D-53113 Bonn

Germany
dieter.boes@uni-bonn.de

This content downloaded from 213.55.83.49 on Wed, 14 Dec 2016 06:06:43 UTC
All use subject to http://about.jstor.org/terms

You might also like