001 - THE BASIC PUBLIC FINANCE OF - Engel Fisher

THE BASIC PUBLIC FINANCE OF
PUBLIC-PRIVATE PARTNERSHIPS
Eduardo Engel Ronald Fischer

Yale University Universidad de Chile
and Universidad de Chile
Alexander Galetovic
Universidad de Los Andes
Abstract
Public—private partnerships (PPPs) have been justified because they release public funds or save on
distortionary taxes. However, the resources saved by a government that does not finance the upfront
investment are offset by giving up future revenue flows to the concessionaire. If a PPP can be justified
on efficiency grounds, the PPP contract that optimally balances demand risk, user-fee distortions, and
the opportunity cost of public funds has a minimum revenue guarantee and a revenue cap. The optimal
contract can be implemented via a competitive auction with reasonable informational requirements.
The optimal revenue guarantees, revenue sharing agreements, and auction mechanisms are different
from those observed in the real world. In particular, the optimal contract duration is shorter in
demand states where the revenue cap binds. These results also have implications for budgetary
accounting of PPPs, as they show that their fiscal impact resembles that of public provision, rather
than privatization. (JEL: H21, H54, L51, R42)
1. Introduction and Motivation
The use of public private partnerships (PPPs) to replace the public provision of
—
infrastructure services has become increasingly common. Projects that require large
upfront investments, such as highways, water and sewage, bridges, seaports and
airports, hospitals, jails, and schools are being provided via PPPs. The average
annual value of €22,900 million for PPP projects signed in Europe between 2002
and 2006 was three times the annual average over the preceding decade. Similarly,
The editor in charge of this paper was George Marios Angeletos.

Acknowledgments: We thank Maham Mela for research assistance, and Raj Chetty, Peter Diamond, Hadi
Esfahani, Antonio Estache, John Friedman, Aldo Gonzalez, James Hines, Mervyn Lewis, Guillermo Perry,
Eduardo Saavedra, John Shoven, and participants at various conferences and workshops for insightful
comments on an earlier version of this paper. We also thank the Corporacion Andina de Fomento (CAF),
the World Bank and the Instituto Milenio de Sistemas Complejos en Ingenieria (Chile) for financial
support, and the Stanford Center for International Development for its hospitality. The views are our
own and do not represent those of any of the above mentioned institutions. Engel is a member of the
NBER.
E-mail: emraengel@gmail.com (Engel); rfischer@dii.uchile.cl (Fischer); alexander@galetovic.cl
(Galetovic)
Journal of the European Economic Association February 2013 11(1):83-111

© 2012 by the European Economic Association DOI: 10.1111/j.1542-4774.2012.01105.x
This content downloaded from 34.192.2.131 on Fri, 21 Jun 2019 21:25:57 UTC
All use subject to https://about.jstor.org/terms
84 Journal of the European Economic Association
investment in PPPs in developing countries grew at an average annual rate of 28.3%

between 1990 and 1997, followed by a slowdown after the East Asian crisis, and
a new growth spurt beginning in 2003, reaching 154,400 million dollars during
2008.1
The main characteristic of a PPP, compared with conventional provision, is that
it bundles investment and service provision in a single long-term contract. For the
duration of the contract, which typically lasts several decades, the concessionaire
manages and controls the assets, usually in exchange for user fees and government
transfers, which compensate for investment and other costs. At the end of the
concession, the assets revert to government ownership.2
As the economics of PPPs is still imperfectly understood, practice has run ahead of
theory. Many practitioners and governments claim that PPPs relieve strained budgets
and release public funds,3 while others suggest that PPPs are appealing because finance,
investment, and management is delegated to private firms, which are more efficient.
Despite these seemingly reasonable arguments, the experience with PPPs has been
mixed. In some cases expectations have been met, but in many more cases contracts
have been renegotiated in favor of the concessionaire, and sometimes firms have
been affected by regulatory takings (Guasch 2004).4 The reason seems to be that the
profitability of PPP projects is subject to large exogenous demand uncertainty, which
is often not considered properly when designing the contracts. This explains why
renegotiations take place when demand is lower than expected, as well as the array of
risk sharing agreements that are observed.
The purpose of this paper is to contribute to the normative analysis of PPPs by
answering two public finance questions. First, what is the structure of the optimal risk-
sharing contract between a government and a private firm when there is substantial
exogenous demand risk? Second, what is the impact of PPPs on the government
budget? In order to answer these questions we use a stylized model in which the
government contracts a risk-averse firm to build, operate and maintain an infrastructure
1. See Tables B.1 and B.2 in Appendix B for details on PPP investments in Europe and developing
countries.
2. There are several definitions for "public—private partnership". In this paper we mean an infrastructure
project such that (i) assets are controlled by a private firm for a (possibly infinite) term; (ii) during the
duration of the contract, the firm is the residual claimant, while the government is the residual claimant
at the end of the concession, and (iii) there is considerable amount of public planning in the design of
the project. Note that the claims in (ii) are ambiguous. We use the term "concession" as synonymous
to PPP.
3. "The boom is good news for governments with overstretched public finances: many local and national
authorities have found themselves sitting on toll roads, ports and airports that they can sell for billions of
dollars to fund other public services." Financial Times, 5 July 2007.
4. This does not mean that the traditional approach to infrastructure provision, with the government
contracting a private firm to build the project, would have done better. For an early evaluation of
infrastructure PPPs, see Economic Planning Advisory Commission (EPAC) (1995), Final Report of the
Private Infrastructure Task Force, Australian Government Publishing Service, Canberra. For more recent
evaluations, see Engel, Fischer, and Galetovic 2003; Grimsey and Lewis 2007.
Engel, Fischer, and Galetovic Public Finance of PPPs 85
project.5' 6 The investment is upfront, there are no further costs and the demand for the
project is perfectly inelastic, exogenous and stochastic. The concessionaire receives
a combination of state-dependent user fees and subsidies (i.e., direct transfers) as
compensation for its efforts. Thus our model encompasses the range from conventional
provision of infrastructure, in which the government pays for the project and the firm
collects no user fees, to the case of a traditional concession, in which the firm's only
source of income is user fees.
In standard fashion, we assume that there is a cost of raising public funds, so that a
dollar in government revenues costs more than one dollar to society. This leads to our
first result, namely that contrary to common wisdom, PPPs do not release public funds
even when totally financed by user fees. The reason is that while private financing
reduces the need for current taxes, this is followed by a period where the government
foregoes user fee revenues that are cashed in by the concessionaire. These revenues
could have been received by the government and used to reduce distortionary taxes, so
there is no net gain to the government in discounted value.
Having shown that private financing cannot justify a PPP raises the question
of whether there exist other reasons to prefer them. Hart (2003) argued that the
main characteristic of PPPs is that they bundle investment expenditure with life-cycle
operation costs.? A PPP achieves the most efficient mix of these costs and is therefore
superior to conventional provision when the benefits of cost-cutting investments during
the building phase are not undone by the cost to users of lower service quality.8
In this paper we focus on an alternative efficiency justification for PPPs, by
considering the financial aspect of bundling in PPPs. We note that in a PPP the firm
can be compensated with a combination of user fees and subsidies, and assume that
user fees are a more efficient way of putting money in the hands of the concessionaire,
because the private sector pays lower overhead and has better incentives to control
corruption. Agency problems faced by the budgetary authority when monitoring the
government agency in charge of the resource transfer justify this assumption.
In the presence of this second wedge, the optimal contract is characterized by two
thresholds: a revenue cap and a minimum income guarantee. If discounted income is
above the revenue cap, the contract ends when this cap is attained. On the other hand,
if revenue never reaches the minimum income guarantee, the firm is compensated with
5. Martimort and Pouyet (2008) also assume a risk-averse concessionaire, see also Dewatripont and
Legros (2005) and Hart (2003). Others are skeptical, and point out that private firms can use the capital
market to diversify risks at least as well as the government: Hemming (2006); Klein (1997). For a discussion
of the controversy in economics see Brealey, Cooper, and Habib (1997). Finally, note that similar results
to the ones we derive are obtained by assuming a risk neutral firm subject to expropriation risk (Engel and
Fischer 2010).
6. Note also that firms routinely ask for minimum revenues guarantees as part of PPP contracts because
they consider demand risks to be excessive. Apparently, the same reason underlies the move in some
European countries away from shadow toll contracts towards availability payments.
7. Bennet and Iossa (2006); Bentz, Grout, and Halonen (2005); Martimort and Pouyet (2008); Iossa and
Martimort (2008).
8. See also Grout (1997).
the difference. Last, in states with revenues between both thresholds, the concession
has an infinite duration but the firm does not receive a subsidy.
The optimal contract does not provide full insurance to the concessionaire, even
though the government could eliminate all risk for the firm by remunerating it with any
combination of user fees and transfers that add up to investment costs. The planner can
improve on this contract, however, by trading off increased risk for the concessionaire
against paying lower subsidies. It can lengthen the term of the concession in high
demand states, so that user fee revenues exceed investment costs, thus lowering the
need for subsidies in low demand states. This is an improvement on full insurance
because user fees foregone by the planner in high-demand states are less valuable
at the margin than subsidies required in low-demand states; this wedge drives our
results.
Even though the optimal contract would appear to be difficult to implement, we
show that it can be attained in a competitive auction. The government announces
the probability density of the different states, and the wedge between the shadow
cost of public funds and subsidies. Firms bid both a minimum income guarantee
and a revenue cap. The government scores the bids according to a weighted average
of expected payments to the firm via user fees and subsidies, and chooses the one
with the lowest score. Assuming firms with identical costs and competition, the
auction reproduces the optimal revenues caps and guarantees. The auction neither
requires the government to know investment costs nor the degree of risk aversion
of firms.
The simplicity of our results—both for the optimal contract and the auction that
implements it—relies on the assumption that quality of service is contractible, which
allows us to ignore moral hazard. That quality of service can be defined and enforced is
arguably the case for the main types of PPP infrastructure in the transportation sector,
which account for 84% of PPP investments in continental Europe, 67% of which are
roads.9 For example, in the case of roads and highways the quality of service provided
can be ascertained by independent third parties using equipments such as laser/inertial
profilometers, to measure roughness, unevenness, texture, surface skid resistance, and
rutting problems. Measures of the time needed to remove a broken-down car are also
easy to implement (for details see Engel, Fischer, and Galetovic 2009a). And when
seaports are contracted as a PPP, service standards, such as the time ships need to wait
before obtaining a berth and the speed with which cargo is unloaded, can be specified
and enforced.
The optimal contract we derive in this paper has implications for the ongoing
debate on whether PPPs should add to public debt or not.10 We show that if demand
9. See Blanc-Brude, Goldsmith, and Valila 2007. Also, in the case of the United States, where until
recently PPPs played a smaller role than in many European countries, financing of transportation
infrastructure via PPPs increased almost tenfold, on an annual basis, between 2006-2008 and the preceding
decade (1996-2005). 24,698 million dollars of investment were financed in this sector via PPPs during the
1996-2008 period.
10. See, for example, Daniels and Trebilcock 1996, Daniels and Trebilcock 2000; Gerrard 2001; Savas
2000; Starr 1988.
risk is allocated optimally, the impact of a PPP on the intertemporal government budget
is usually the same as under conventional provision of infrastructure. Most, or even all,
risk is borne by the government and the concessionaire recovers the upfront investment
in most states. By contrast, under privatization, assets and cash flows are transferred
forever to a private firm in exchange for a one-time payment. This means that the link
between the project and the public budget is permanently severed. Under a PPP this
link continues to exist, even when the compensation to the concessionaire is derived
solely from user fees. It follows that from a public finance perspective PPPs are much
closer to conventional provision, and therefore should be accounted for in the same
way.
There is a growing literature on PPPs related to this paper.11 Risk sharing between
the government and the concessionaire has always been a concern among practitioners
and policy makers. The standard prescription is that each risk should be allocated to
the party best able to manage it.12 Martimort and Pouyet (2008) study this problem in
a moral hazard model where effort during investment affects both the quality of the
infrastructure and its operating cost, and their analysis is extended in various
directions by Iossa and Martimort (2008). Bentz, Grout, and Halonen (2005), on the
other hand, study a model with moral hazard in building and adverse selection in
operation.
Our paper, by contrast, studies the implications of the optimal allocation of demand
risk, when subsidy finance is less efficient than user-fee finance. We show that variable,
state-contingent concession lengths are a key component of the optimal risk-sharing
contract. In addition, we provide a rigorous foundation for minimum income guarantees
and revenue caps and show that the optimal guarantees and caps bear little relation to
observed guarantees and revenue-sharing agreements.
This paper is also related to the literature on franchise bidding pioneered by
Chadwick (1859) and Demsetz (1968), according to which competition for a monopoly
infrastructure project replicates the competitive outcome (see Stigler (1968), Posner
(1972), Riordan and Sappington (1987), Spulber (1989, chapter 9), Laffont and Tirole
(1993, chapters 7 and 8), Harstad and Crew (1999) and Engel, Fischer, and Galetovic
(2001) for papers within this tradition, and Williamson (1976, 1985) for a criticism).
We contribute to this literature by considering projects that require subsidies to make
them feasible.
Finally, in Engel, Fischer, and Galetovic (2001), we studied the optimal private
provision of infrastructure projects by solving a Ramsey problem with variable
concession lengths. In that paper we assumed a 'self-financing constraint', which
11. See Grimsey and Lewis (2004) for a survey, and Grimsey and Lewis (2005) for a collection of
papers on PPPs. Vaillancourt-Rosenau (2000) and Akintoye, Beck, and Hardcastle (2003) include useful
collections of essays.
12. Irwin (2007, p. 14) is more precise: each risk should be allocated to maximize project value, taking
account of moral hazard, adverse selection, and risk-bearing preferences. See also the discussion in
Dewatripont and Legros 2005.
ruled out government transfers to the concessionaire. In the present paper, demand-
contingent government subsidies play a central role, thus providing a framework to
study the public finance of PPPs.
The remainder of the paper is organized as follows. Section 2 describes the model
and presents the basic irrelevance result as well as the optimal contract. This section
also contains extensions of the model to the case of price responsive demand, when
there are operational and maintenance costs, as well as to the moral hazard case
in which effort in the investment stage increases the likelihood of higher demand.
Section 3 shows that the optimal contract can be implemented with an auction with
reasonable information requirements. Section 4 discusses the practical relevance of
the results. We document the increasing popularity of flexible term contracts similar to
those suggested by the optimal contract we derive. We also examine the implications
of our results for the fiscal accounting of PPPs. Section 5 concludes and is followed
by two appendices.
2. Model and Optimal Contract
A risk-neutral benevolent social planner must hire a concessionaire to finance, build

and operate an infrastructure project with exogenous technical characteristics. There
are no maintenance nor operation costs, the upfront investment does not depreciate,
and the concessionaire is selected among many firms that can build the project at
cost I > 0.13 All firms are identical, risk-averse expected utility maximizers, with
preferences represented by the strictly concave utility function u.
Demand uncertainty is summarized by a probability density over the present value
of user fee revenue that the infrastructure can generate over its entire lifetime, f(v),
with c.d.f. F(v).14 This density is common knowledge to firms and the planner, and is
bounded from below by vinin and from above by vinax. Also, for simplicity we assume
that v equals the discounted private willingness to pay for the project's services.15
2.1. Planner's Problem
Let PS(v) denote producer surplus in state v, CS(v) denote consumer surplus in state
v, and a E [0, 1] be the weight that the planner gives to producer surplus in the social
13. That is, we ignore construction cost uncertainty and focus instead on demand uncertainty, which is
considerably larger for many PPP projects.
14. We assume that the instantaneous user fee revenue process is "well behaved", in the sense that the
expected time for a v 1 -income trajectory—that is, a revenue trajectory with discounted value equal to
v1—to reach a given threshold M is larger than for a v2-trajectory when M < vi < v2. This will be the case,
for example, if all demand trajectories grow at the same rate and differ only in their initial values. It also
holds for more general instantaneous revenue processes such as a Brownian motion. This assumption is
used only when associating higher values of v with shorter contract terms when the optimal contract term
is finite.
15. In Appendix A of Engel, Fischer, and Galetovic (2009b) we show that this simplification does not
affect the structure of the optimal PPP contract.
welfare function.16 The planner's objective is to maximizel7
f[CS(v) + aPS(v)] f (v) dv, (1)
subject to the concessionaire's participation constraint
fu(PS(v)) f (v) dv > u(0),

where u(0) is the concessionaire's outside option. Assuming that the firm's utility is
concave in its discounted profit simplifies the dynamic problem facing firms and the
government, but is all that is needed for the purpose of determining the long run fiscal
impact of PPPs, and the optimal contract.
To maximize (1), the planner chooses how much user fee revenue and subsidy the
concessionaire should receive in each state v. Denote by R(v) the present value of user
fee revenue collected by the concessionaire in state v, and by S(v) the present value of
the subsidy it receives. Hence
PS(v) = R(v) + S(v) — I. (2)
Note that by "subsidy" we mean any cash transfer from the government to the
private concessionaire. It may be the upfront payment made by the government with
conventional unbundled provision (in which case S(v) is the same for all v), but it could
also be a cash transfer made over time, contingent on v, to supplement revenue from
the project under a Build-Operate-and-Transfer (BOT) contract (a so-called "minimum
revenue" or "minimum income" guarantee).
Since the concessionaire receives R(v) in state v the government receives v — R(v),
and we have 0 < R(v) < v. If the term of the concession is finite and v — R(v) > 0,
these funds can be used to reduce distortionary taxation elsewhere in the economy.
Moreover, assuming that the willingness to pay is positive at all points in time, we
have that R(v) = v only if the concession lasts forever. Letting 1 + A > 1 denote the
marginal costs of public funds (see, e.g. Dahlby (2008)) we then have
CS(v) = [v — R(v) — (1 + A)S(v)] ATI) — R(v)] = (1 + A)[v — R(v) — S(v)].

(3)
The first term in the expression between equal signs, v — R(v) — (1 + A.)S(v), is the
difference between users' willingness to pay in state v and the amount transferred to
the concessionaire, where the cost of the subsidy is augmented by the cost of the tax
distortion required to finance it. The second term, [v — R(v)], is the value of the
16. In many countries foreign firms are important investors in PPPs, which implies a < 1.
17. This objective function assumes that the income of users is uncorrelated with the benefit of using the
project, so that if users spend a small fraction of their incomes on the services of the project they will value
the benefits produced by the project as if they were risk neutral. See Arrow and Lind (1970).
reduction in distortionary taxes allowed by the fact that the government collects v —
R(v) in toll revenue (net of toll collection costs) after the concession ends.18
As is traditional in economics, the expression for consumer surplus in (3) includes
the social costs caused by tax collection. Nevertheless, it ignores the inefficiencies
associated with disbursing those revenues, which are relevant when comparing PPPs
with conventional provision. We present a simple model which goes beyond the
standard costs of collecting taxes and adds administrative and agency costs incurred
when disbursing money.19 These include standard overhead costs which are present
in any organization as well as the additional inefficiencies of government agencies:
overstaffing, the lack of a board to pressure management to control overhead, the need
to follow and comply with rigid administrative procedures and controls imposed by the
budget and the comptroller's offices, or even the diversion of public funds and outright
corruption.2°' 21 Our point is that the government has to disburse more than $1 when it
provides $1 to the recipient.
We assume that the social planner can allocate spending across n government
agencies indexed by i = 1, 2, . . . , n. The surplus created when agency i spends Gi on
its best projects is Si(Gi), with Si (0) = 0, > 0 and Si' < 0. Nevertheless, to achieve Gi
dollars spent on projects the government agency i disburses Gi Zi(Gi) in total, with
Zi (0) = 0, Z; > 0 and Z;' < 0. 22 Let T be the total amount of taxes raised
18. Note that this assumes that the government is as efficient in collecting tolls as the private
concessionaire. This is optimistic, as anecdotal evidence suggests: for example when the city of Chicago
transferred the Chicago Skyway to private management under a PPP in January 2005
[...] not a single City toll collector took a job with the toll collection contractor hired by
Skyway Concession Company because they were offering positions at $12/hour, about 60%
of the rates they were being paid by the City and with less health and other benefits. The
hundred or so City collectors were found jobs elsewhere in City employment under labor
agreements signed by the City.
("ITR pays near market wages—unfilled vacancies", http://www.tollroadsnews.com./node/1466.)

19. See Appendix C in Engel et al. (2009b) for additional details.
20. Consider, for example, the National Flood Insurance Program of the United States, which hands
"[...] one-third of its collected premiums—equivalent to a 50 percent loading cost—to

financial intermediaries who do not bear any risk."
In comparison, private insurance, with typical loading costs of 20%-30%, must pay for rate-making,
product development and must hold capital against risk (Michel-Kerjan (2010)). Even though a diversion
of public funds might be a transfer, it usually involves wasteful activities and expenditures to conceal
it. Even in the case of pure diversion, a diversion of $1 costs more than $1, because it is financed with
distortionary taxation at the margin.
21. The distinction between the collection and disbursement costs becomes clear when considering
charities. The Forbes index, for instance, ranks charities according to the percentage of private support
remaining after fundraising expenses (the equivalent of collection costs in tolls, 10% for the Forbes 200
Charities in 2008) and on the charitable services provided as a percentage of total expenses (15% for the
Forbes list). The difference between the first and second term is a rough measure of the cost of disbursing
resources. We have assumed collection costs of zero to simplify the notation and because it makes no
difference to our analysis, unless the cost is very high.
22. This implies that costs increase in the size of the disbursement by the agency. While some costs are
probably fixed, it is clear that others increase with the size of the project.
and A(T) the costs of the distortions caused by taxes, with A A' > 0 and A" > 0.
The planner chooses (G i)n 1 and T to maximize
Si(Gi) — T — A(T)
i =1
subject to the budget constraint T = =i[G Zi(Gi)]. Solving the planner's
allocation problem leads to
57(G i)
______ = 1 + i = 1, 2, ... , n.
1
This is a standard result, which can be traced back to Atkinson and Stern (1974)23
government should expand until the marginal benefit of spending an additional dollar
equals the marginal cost of public funds, 1 + A, and the marginal benefit of spending
should be the same across agencies. But in addition, the marginal benefit of projects
undertaken by agency i must be adjusted by the agency's relative inefficiency, (1 + ..)
to account for the fact that resources are spent and wasted when disbursing funds to
projects. Agencies that use up more resources to disburse $1 should spend less,
ceteris paribus.
We can now insert a PPP into this framework. In what follows, a PPP is assumed
to be undertaken by one agency within the government. The project is small relative
to the overall budget, so that and are evaluated at the optimal resource allocation
and can be assumed constant for the PPP agency.
Substituting equations (2) and (3) into (1) shows that maximizing the planner's
objective function (1) is equivalent to maximizing
—[(1 A) — a] f R(v) f (v) dv — [(1 A)(1 + — a] f S(v) f (v) dv,
and therefore to minimizing
f{[(1 + A) — a]R(v) [(1A)(1 + — a]S(v)} f (v) dv, (4)
where we have dropped al and (1 + )1/4.)f vf(v)dv from the objective function because
they do not depend on the planner's choice variables, R and S.
In equation (4) the marginal cost of funds is scaled up by 1 + when the
concessionaire receives subsidies, but not when it receives toll revenue. This captures
the fact that toll revenues go directly into the concessionaire's pocket. Thus, by
remunerating the concessionaire with toll revenue, society avoids the costs incurred
when disbursing subsidies.24' 25
23. See Dahlby (2008, Chapter 2.2) for a recent exposition.
24. One might argue that the government has to incur in costs to monitor truthful reporting of toll
revenues by the concessionaire. Nevertheless, these costs are likely to be negligible, for the concessionaire
is monitored by investors and tax authorities. Moreover, it is probably far cheaper to monitor the
concessionaire's report than to directly monitor toll collection.
25. For simplicity we have assumed that each dollar collected by the concessionaire (net of toll collection
costs) ends in its pocket. Of course, private firm pays some overhead. But parameter is to be understood
as capturing a relative difference between a private and a public bureaucracy.
Defining > 0 via26

1+ =
_ (1 + A.)(1 + 0 — a
1 + ),, — a
we have that the planner's program can be written as
{R(v),S(v)}
min f {R(v) + (1 + -)S(v)} f (v) dv, (4a)
subj. to: f u(R(v) + S(v) — I) f (v) dv > u(0), (4b)
0 < R(v) < v, (4c)
S(v) > 0. (4d)
2.2. Irrelevance Result
It is often claimed that PPPs relieve the public budget by substituting private finance
for distortionary tax finance. Does this argument make the case for PPPs?
If we only consider the distortions associated with taxation, so that 4- = 0, the
planner's objective described in (4a) is equivalent to minimizing
fR(v) f (v) dv + f S(v) f (v) dv.

The per-dollar cost of paying the concessionaire with either user fee revenues or
subsidies is the same, so social welfare only depends on total transfers T(v) =
R(v) + S(v) to the concessionaire, not on the division of payment between subsidies
and user fee revenue. This is the insight behind the following result.
PROPOSITION 1. (Irrelevance of the cost-of-funds argument) Assume 4' = 0. Then any

combination of user fee and subsidy schedules that satisfies constraints (4c) and (4d)
and such that T(v) = I for all v solves the planner's program (4a)—(4d).
Proof . See Appendix A.1. ❑
What is the economics of this result? The standard reasoning in favor of PPPs
points out that subsidies are an expensive source of finance, because they are financed
with distortionary taxes. Yet the multiplicity of optimal subsidy-revenue combinations
indicates that distortionary taxation (A > 0) is not a sufficient reason to prefer private
provision. One solution is R(v) m- 0 and S(v) = I, which in our framework corresponds
to public provision. Another solution is that the concessionaire invests I, collects user
26. From the definition of > 0 < ___ > > 0 an d < 0 .; = > - < O. Fu rt herm o re, = wh en a = O.
fee revenues equal to I in present value, and no subsidies are paid.27 In addition, there
is a continuum of combinations where the government provides a partial subsidy.
The intuition for this result is that if the user fee revenue collected by the
concessionaire increases by $1 (and thus government revenue decreases by $1), the
government has to levy $1 in additional taxes, which costs society 1 + X. This is
the same cost that society bears when paying $1 in additional subsidies. Hence, at
the margin the opportunity cost of paying with user fee revenue or subsidizing the
concessionaire is exactly the same. The rich set of optimal combinations of state-
contingent subsidies and concession terms reflects that user fees and subsidies are
perfect substitutes in the planner's objective function.
Un argumento similar muestra que el planificador satisfará la restricción de
participación del concesionario con igualdad. Un dólar adicional para las
concesionarias aumenta el bienestar social en a, pero cuesta 1 + X a los usuarios.
Como 1 + A> a, el planificador extrae todas las rentas del concesionario.
Finalmente, tenga en cuenta que el contrato óptimo proporciona un seguro completo
al concesionario.
El resultado irrelevante implica que el caso de las PPP no puede basarse en la
afirmación de que alivian los presupuestos restringidos. ¿Cuándo se justifican las
PPP? Como se mencionó en la introducción, una justificación de las PPP es que la
agrupación puede mejorar la eficiencia productiva. Una ventaja adicional de las
PPP, destacada en este documento, es que reducen las sumas que fluyen a través del
presupuesto público, reduciendo las ineficiencias asociadas con las transferencias
de subsidios. Esto corresponde al caso donde> 0 en nuestro modelo. El resto de esta
sección (y el documento) está dedicado a obtener (e implementar) el contrato
óptimo para este caso.
2.3. Optimal Risk-Sharing Contract: High- And Low-Demand Projects
To derive the optimal contract when > 0, note that equation (4a) implies that the
planner will pay subsidies in state v only after exhausting user fees—otherwise she
could slightly reduce subsidy payments, which would save (1 + A)(1 + — a; and
increase R(v), which would cost only 1 + A — a. This rules out the possibility that
R(v) < v and S(v) > 0 simultaneously, and it follows that if the firm receives subsidies
in state v, it must be the case that user fees revenues are exhausted in this state. This
insight motivates classifying demand states into high-, intermediate- and low-demand
states as follows. In a high-demand state, the contract length is finite and no subsidies
are paid out—that is, R(v) < v and S(v) = 0. By contrast, in low-demand states
the contract lasts indefinitely and the firm is remunerated with subsidies: R(v) = v
and S(v) > 0. There remain the intermediate-demand states, where the contract lasts
indefinitely, as in low-demand states, but no subsidies are paid out, as in high-demand
states.
The planner faces the following tradeoff: On the one hand, she would like to
utilize user fee revenues as far as possible to compensate the concessionaire, in order
27. This is only possible if vim!, > I, for otherwise the project cannot be financed with user fees in all
states.
TABLE 1. Taxonomy of demand states and taxonomy of projects.

Demand state Project
High: R(v) < v, S(v) = 0 v min > I

Intermediate: R(v) = v, S(v) = 0 v min < I < v m a x
Low: R(v) = v, S(v) > 0 vm <
to avoid paying subsidies. On the other hand, using only user fees may expose the
concessionaire to excessive risk, and an efficient contract would insure against low
demand states through subsidies.
The planner can avoid the subsidy—risk tradeoff when user fees alone always pay
for the project, that is, when vmin > I. The optimal contract then fully insures the firm,
financing it solely out of user fee revenues, while providing no rents. This contract
sets R(v) = I < v and S(v) = 0 for all v—all states are high demand states when
vmin > I.
Consider next the case where user fees are always insufficient to pay for the project,
that is, vmax < I. Then R(v) = v combined with S(v) = I — v > 0 for all v also defines
the optimal contract, since this contract exhausts user fees in all states of demand
before resorting to subsidies, while avoiding a risk premium, and without giving rents
to the firm. As in the previous case, there is no tradeoff between minimizing subsidy
payments and reducing risk exposure for the firm.
We refer to a project with vmin > I as a high-demand project, while one with
vmax < / is a low-demand project; intermediate-demand projects satisfy vmin < I < vmax.
Table 1 summarizes our taxonomies for states and projects. We summarize the optimal
contract for high- and low-demand projects in the following proposition.
PROPOSITION 2. (Optimal contract for high- and low-demand projects). The optimal
contract for high- and low-demand projects requires that R(v) S(v) = I for all v.
Given demand realization v, the government collects v — I in each state if the project
is high demand, while it pays a subsidy of I — v in each state if the project is low
demand.
2.4. Optimal Risk-Sharing Contract: Intermediate Demand Project
The subsidy—risk tradeoff becomes relevant when designing the optimal contract for
an intermediate-demand project, that is, a project with vmin < I < vmax. In this case full
insurance does not implement the optimal program.
To see that even at the lowest cost, full insurance (R(v) + S(v) = I) in all states
is suboptimal, we argue in what follows that a minor modification of this contract—
referred to as the "full insurance contract"—improves welfare. Consider lowering
insurance from I to I — A/ for states with v below I, and use the resources that are
freed up to raise the revenue cap for states with v above I by an amount proportional to
A/.28 Since user fee revenue is substituting for subsidy financing, every dollar saved
in low-demand states has a social value that is (1 + times that of every additional
dollar transferred to the firm in high demand states. It follows that for small A/ this
leads to a welfare gain proportional to A/. It is true that the firm bears risk under the
modified contract, yet the compensation needed to satisfy its participation constraint
is only of order (A/)2, since we are starting from a situation where it bears no risk.
Thus there exists a contract with minimum revenue guarantee I — AI and revenue cap
I c AI that is feasible and improves welfare. 29
For intermediate-demand projects, the optimal contract is characterized by two
thresholds, a minimum revenue guarantee m and a revenue cap M. In low demand
states v < m, R(v) = v and S(v) = m — v. By contrast, in high demand states v > M
and R(v) = M. The remaining states, with m < v < M, are intermediate-demand states,
with R(v) = v and S(v) = 0. This contract cuts off the tails of the distribution and is
the most efficient means of reducing the variance of revenues using subsidies in bad
states and caps on revenue in good states in order to satisfy the participation constraint
(see Appendix A). The optimal contracts of Proposition 2 may be viewed as particular
cases of two-threshold contracts, where either the minimum income guarantee or the
revenue cap are not binding.
Figure 1 shows how the subsidy—risk tradeoff is resolved optimally when

vimn < I < vmax . The horizontal axis plots the support of v while the vertical axis
shows the total revenue received by the concessionaire in each state, R(v) S(v).
Implicit is the assumption that subsidies are used in a particular state only once user
fees are exhausted.
In any state with a finite concession term, the social opportunity cost of the
last dollar received by the concessionaire is 1 + A — a; this justifies equalizing the
concessionaire's revenue across high-demand states by fixing a revenue cap M. On the
other hand, in any low-demand state the last dollar paid to the concessionaire comes
from a subsidy and costs society (1 + + — a. Again, this justifies equalizing
revenue across low demand states at the minimum revenue guarantee m < M.
-
As can be seen from Figure 1, the wedge between 1 + A — a and (1 + +

— a creates an interval M — m of intermediate-demand states. To see the intuition,
consider one such state, 15. It is straightforward to obtain the following inequalities:
1 u'(13 — I) u'(13 — I)
_______ < 1 < _________ < 1+ -e .
1+ u'(m — I) u'(M — I)
These inequalities imply that the concessionaire's marginal utility evaluated at v — I
is lower than the marginal utility at m, but higher than the marginal utility at M. In
other words, the shadow value of the last dollar received by the concessionaire in state
28. More precisely, this modification of the full insurance contract frees up resources F(I)A1 in expected
value, that can be used to finance a new revenue cap / c01 with c = F(I)I(1 — F(I)), where 1 — F(I) > 0
since this is an intermediate-demand project.
29. The constant c will be slightly lower than the original F(I)I(1 — F(I)), by an amount on the order of
A/ to compensate for the added risk.
R(v) + S(v),I
A
Revenue for the

government (A)
v + S(v) = m
Subsidy
0,4)
z
Mvm„ x
FIGURE 1. Optimal contract, intermediate demand project.
is too low to warrant a subsidy and too high to warrant a revenue cap. Consequently,
the concession lasts forever, but no subsidies are paid. The following proposition
characterizes the optimal values of both thresholds:
PROPOSITION 3. (Optimal contract for intermediate-demand projects). Consider a

project with v min < 1 < vmax(intermediate-demand project). Assume u' (v min — I) > (1 +
(vma„ — 1).3° Then the optimal contract is characterized by thresholds m and
M, with vmin < m < I < M < v max, such that states with v > M are high demand, states
with m < v < M are intermediate demand and states with v < m are low demand. 31
M and mare determined from the concessionaire's participation constraint
M
F(m)u (m — I ) f u (v — I ) f (v) d v + (1 — F(M))u (M — I) = u (0 ) (5 )
30. This condition ensures that m > v„,,„ and M < v„,„„, so that condition (6) holds with equality.
Two possibilities arise if u'(v„„„ — /) < (1 + )u'(v„,„„ — /). First, if f u(v — I)f(v)dv > u(0), the optimal
contract involves no subsidies (m < v,„,„) and M is determined from
L n U(V I)f (v) dv + (1 — F(M))u(M — I) = u(0).
When the inequality is reversed, f u(v — I)f(v)dv < u(0), the optimal contract involves no revenue cap and
the minimum income guarantee is determined from
F(m)u(m — I) + f u(v — 1)f (v) dv = u(0).
31. See Table 1 for the definition of high-, intermediate- and low-demand states.
102
Engel,
100
98
Engel, Fischer, and
Fischer, and Galetovic
Galetovic Public
Public Finance
Finance of
ofJournal
Journal
PPPs of
PPPs of the
theEuropean
EuropeanEconomic 99
EconomicAssociation
Association
101
97
and the condition

tti (m — I) = (1 + (M — I). (6)
2.5. Extensions
These results can be extended in several directions. Here we briefly discuss the intuition
underlying two extensions of the model, and refer the reader to Engel, Fischer, and
Galetovic (2009b) for a formal treatment. First, we consider the case where demand
responds to price changes and the concessionaire faces a standard convex short-run
cost curve. Second, we incorporate moral hazard, by assuming that demand responds
to the concessionaire's unobservable effort.
Price-Responsive Demand. Assuming a totally inelastic demand simplifies the

derivations, but is not realistic. Nevertheless, our insights carry through to the case
with a price-responsive demand. Once tolls are set appropriately, the optimal contract
continues to be characterized by a minimum guarantee and a cap on revenues.
In Engel, Fischer, and Galetovic (2009b) we consider a continuum of verifiable
demand states where, for tractability, we assume that the demand curve becomes known
immediately after the project is built and remains constant over time.32 This means
that for every demand state, the planner chooses two prices, the user fee paid during
the concession, and the user fee collected by the government after the concession ends.
The planner also sets a demand-contingent concession length.
While the determination of optimal user fees is no longer trivial, the structure of
the optimal contract remains identical to the case of perfectly inelastic demand. Thus,
the present value of the cash flow received by the concessionaire is equal to M across
all high-demand states, and m across low-demand states, with m < M. As before, the
cash flow received by the concessionaire in intermediate-demand states lies between m
and M. Moreover, high-, intermediate- and low-demand projects are defined as before.
User fees are set taking into account the relevant margin of fiscal revenue which
they substitute. For example, user fees after the concession ends are set so as to create
a distortion commensurate with the cost of public funds, since these fees substitute for
this source of government revenue. It follows that there will be marginal cost pricing
after the concession ends only when = 0.
The same principle applies, mutatis mutandis, to the different types of demand
states during the life of the concession. Somewhat surprisingly, in a high demand
32. The results that follow extend easily to the case where the demand schedule grows at an exogenous
rate that may vary over time and with the demand state, since the price-elasticities of demand do not vary
over time in this case. We can also introduce production costs that are increasing and convex in quantity
demanded. The problem becomes considerably harder when demand is allowed to evolve arbitrarily.
R(v)+ S(v)
Revenue for the

4... 5 government (A)
H1 M Vmax
FIGURE 2. Optimal contract with moral hazard and > 0.
state the relevant margin also is the shadow cost of public funds, despite the fact that
the planner values a dollar in the concessionaire's pocket at a < 1 +)c. The reason
for this is that the planner can recoup the extra cash flow that the concessionaire
receives per period in a high-demand state as a result of a higher user fee during the
concession because the concession duration is shorter. This implies that at the margin
the higher revenue generated by raising the user fee during the concession substitutes
for distortionary taxation after the concession ends. Similarly, the user fee during
the concession in a low-demand state is set commensurate with the cost of subsidy
financing, (1 + ?.)(1 + O. In low-demand states the planner can recover any extra
dollar of user fee revenue received by the concessionaire by lowering the subsidy. 33
Moral Hazard. Another extension is to allow for demand that depends on

unobservable and costly effort by the concessionaire. An additional motive to have
the firm bear risk emerges in this case, as risk induces optimal levels of effort by
the concessionaire. As before, two thresholds, m and M, suffice to partition states
into high-, intermediate- and low-demand states. Even though in the optimal solution
the total revenue collected by the concessionaire always increases with the demand
realization v, subsidies are still paid out only in low-demand states (v < m), while the
government still receives user fees only in high-demand states (v > M).
Figure 2 shows that for an intermediate-demand project with > 0 we have a
range of values of v where the contract lasts indefinitely and there are no subsidies.
33. For intermediate-demand states the user fee is set so that the resulting distortion lies between that
associated with financing from general taxes and that associated with subsidies. See Engel, Fischer, and
Galetovic 2009b.
This range of intermediate-demand states (and intermediate-demand projects) emerges
only when > 0, leading to an increase in risk borne by the concessionaire beyond
the level predicted by the standard principal—agent model for the case = 0. Contrary
to the optimal contract for the case with no moral hazard depicted in Figure 1, the
concessionaire's revenue is not equal across all high-demand states or all low-demand
states. But the gap between m and M emerges, as before, because subsidy finance is
more expensive than user fee revenue, at the margin. Again as in the case with no
moral hazard, the optimal contract also involves a state-contingent concession term,
with shorter terms in high-demand states.
3. Implementation
The informational requirements needed to implement the optimal contract might seem
formidable, but somewhat surprisingly, this is not the case. We show next how to
implement the optimal contract with a competitive auction when the planner knows
neither I nor the risk aversion of firms.
3.1. High- and Low-Demand Projects
Consider first a high-demand project. Then an auction where the bidding variable 8 is
the total present value of user fee revenues (PVR) collected by the concessionaire
implements the optimal contract. This follows from observing that rents will be
dissipated in a competitive auction, so that p will satisfy
f 14(,3 — I) f (v) dv = u(0). (7)
Hence the winning bid will be /3 = I, which corresponds to the optimal contract derived
in Proposition 2. Denote by T(v) the time it takes for user fee revenue accumulated
in state v to attain I. The concession term is shorter when demand is high, that is,
when T(v) is sma11.34 The concessionaire bears no risk because users pay him the
same amount in all states of nature.35 Furthermore, the planner can implement the
optimal contract using a PVR auction even if she does not know I, the density f(v),
or the concessionaire's degree of risk aversion. All that the planner needs to know
is that the project can finance itself in all states of demand, that is, that vnnn > I.
Furthermore, moving from a fixed-term contract to the optimal contract can lead to
substantial welfare gains.36
Consider next a low-demand project. A PVR auction will implement the optimal
contract in this case as well, as long as the government subsidizes the difference
34. This requires that the instantaneous user fee revenue process be well behaved, see footnote 14.
35. Uncertainty in I, which may be important in some projects, cannot be eliminated with a variable (or
fixed) term contract.
36. Depending on the degree of risk aversion and revenue uncertainty, Engel, Fischer, and Galetovic
(2001) find welfare gains between 16% and 64% of the upfront investment.
between the winning bid and the present value of user fees collected. In this case
firms end up bidding on a minimum income guarantee and the winning bid ensures a
total revenue of I. Informational requirements are modest again, since the planner only
needs to know that vmax < I, and be able to verify revenue in each state. Note that the
concession lasts forever in this case. We summarize both cases reviewed so far with
the following proposition.
PROPOSITION 4. (High- and low-demand projects). The optimal contract can be

implemented with a PVR auction, or by a simple extension, for both high- and low-
demand projects. Furthermore, bidders reveal I in the auction and there is no need to
know f or u.
3.2. The General Case
Next we consider the case where the planner does not know if the project is high-,
intermediate- or low-demand. We also assume that the planner does not know the
firms' degree of risk aversion, but does know the probability density f (v) .37 We show
next how to implement the optimal contract with an auction where firms bid on the
minimum revenue guarantee and on the revenue cap they desire.38
PROPOSITION 5. (Optimality of the two-threshold auction). The following two-

threshold, scoring auction implements the optimal contract:
 The government announces the probability density of expected discounted user
fee revenue flow from the project, f(v), and the parameter e that summarizes the
wedge between the shadow cost of public funds and subsidies.
 Firms bid on the minimum revenue guarantee, m, and on the cap on user fee
revenue, M.
 The firm that bids the lowest value of the scoring function
W (M , m) = M(1 — F (M)) + v f (v) dv + (1 + e)f (m — v) f (v) dv

(8)
wins the concession.
What is the intuition underlying this result? Note first that the planner's objective
function does not require knowledge of I. The objective function only depends
on the probability distribution of the present value of revenue that the project can
generate and the distortions associated with government expenditures, as summarized
37. The government should have as much information on demand as third parties, because it either
provides the service or it must compare the PPP with conventional provision. Furthermore, substantial
public planning is needed to design most PPP projects, and this requires an assessment of demand.
38. See Engel, Fischer, and Gasletovic (2009b) for an extension of the results to the case with price-
responsive demand.
by . By awarding the PPP to the bidder that maximizes this objective function,
and assuming competitive bidding, the planner induces the concessionaire to solve
society's problem without knowing the cost of the project or the firms' degree of risk
aversion.
In the case of a high-demand project, the two-threshold auction is equivalent to a
PVR auction. If all states have high demand, any bid with M = I and m < I will win
the auction. No subsidies are paid out and the concession term is shorter if demand
is higher. 39 Similarly, in the case of a low-demand project, a bid with m = I and
M > I wins the concession, since this time the upper threshold is irrelevant. In this
case the two-threshold auction reduces to the extension of the PVR auction described
in Proposition 4. However, the two-threshold auction is more general than a PVR
auction, as it can be used for intermediate-demand projects or, more importantly, for
projects where the planner does not know whether the project is of low, intermediate
or high demand.
4. Practical Relevance
For the duration of a typical PPP contract, which can be thirty years or more, the
concessionaire will build, manage, maintain, and control the assets, in exchange for
some combination of user fees and government transfers, which are its compensation
for its investment and other costs. For example, the main source of compensation for
European PPP contracts for roads, bridges, and tunnels are user tolls (61.1% of all
projects), followed by shadow tolls (32.6%).4°
Usually PPP contracts are conferred via competitive (Demsetz) auctions, the
most common bidding variables being user fees, length of the contract, government
subsidies, and payments to the government. Minimum income guarantees, which
ensure that revenue reaches a predetermined sum, are common, and so are revenue-
sharing arrangements when revenues exceed a predetermined threshold. At the end of
the contract, the asset reverts to the government.
In this section we discuss the practical relevance of the optimal contract and the
implementation derived in Sections 2 and 3. Even though most observed contracts
have a fixed duration, which is suboptimal, flexible-term contracts have become
increasingly common. These nonstandard contracts share characteristics with those of the
optimal contract derived in this paper.
We begin by describing the experience with flexible-term contracts. Next we
compare minimum income guarantees and revenue-sharing agreements observed
in practice with those suggested by our results. Finally, we use our model to
contribute to the debate on whether investment in PPPs should add to the public
debt.
39. This assumes the instantaneous revenue process is well behaved, see footnote 14.
40. S ee Tab le B .3 for d et ai ls.
4.1. Experience with Flexible Term Contracts
The United Kingdom was probably the first country to use a contract similar to the
optimal contract derived in Section 2.3. In 1987 the British government concessioned
the construction and operation of the Queen Elizabeth II Bridge in Essex county. While
demand for bridge crossings was uncertain, there was little doubt that the project was
financially sound provided that the concession term was long enough. In our terms, it
was a high-demand project. The contract specified that the concession would end after
20 years or when toll revenue was sufficient to repay principal and interest, whichever
happened first. The £190 million project relied 100% on debt financing, with no equity.
The bridge was inaugurated in October 1991 and the concession ended in March of
2002. This contract is identical to the optimal contract we derived for high-demand
projects, except for the upper bound on the concession duration.
The Route 68 concession in Chile, linking Santiago with Valparaiso and Villa
del Mar, was the first flexible-term contract assigned via an open auction along the
lines suggested by our implementation results in Section 3.1. This concession involved
major improvements and extensions of the 130 kilometer highway, and was auctioned
in February of 1999. Bidders were given a choice of fixed or flexible inflation adjusted
interest rates (plus a 4% risk premium) to use in discounting their annual toll revenue.
The winner, which chose the fixed rate, asked for a present value of revenue of US$373
million, which was lower than the US$379 million construction and maintenance costs
estimated by the Ministry of Public Works. A possible explanation is that the risk
premium was set too high.
After a first generation of highway concessions using shadow tolls and fixed-
duration contracts, Portugal began using flexible-term concessions.41 The first flexible-
term auction was the 98.4km Litoral Centro highway along the Atlantic Ocean, with
an estimated cost of €795 million euros. The concession duration depends on the net
present value (NPV) of toll revenue reaching a limit of €784 million. If the limit is
reached before year 22, the concession lasts 22 years; if is reached between years
22 and 30, the concession ends once the limit is attained; and if it is not reached by
year 30, the concession ends. Toll revenue is discounted at a market rate and tolls
are indexed to the consumer price index. This is a flexible-term contract that limits
the up and downside risk for the concessionaire, as does the two-threshold optimal
contract discussed in Section 3.2. The project won the Eurofinance prize in 2004 and
the Portuguese government has announced that it will use flexible-term concessions
for all future highway concessions.
Other countries also have used flexible-term auctions, yet it is worth speculating
on why flexible-term contracts have not been adopted more broadly throughout the
world. Opposition has come mainly from the concession lobby, which fears that a
flexible contract will limit their ability to renegotiate contracts. 42 In most countries,
41. What follows is based on Project Finance, February 2005.

42. Guasch (2004) presents evidence on frequent renegotiations of concession projects and argues that
they often reflect opportunistic behavior by the concessionaire.
the Public Works Authority tends to support the position of the concession lobby,
since its governance structures usually provide incentives for road building rather than
supervision and regulation that eventually benefits users. This requires good relations
with future bidders for concessions.
By contrast, the Ministry of Finance usually favors flexible-term contracts, since
they reduce the demand for guarantees given the lower demand side risk. It i s
not surprising, therefore, that flexible-term concessions have been adopted when
the budgetary authority has the upper hand over the public works authority. For
example, in the case of Portugal, the shadow tolls used to remunerate highway
concessions auctioned beginning in 1999 became an increasing burden on the budget.
By 2004 it was estimated that the shadow toll obligations would increase to €660
million per year by 2008, which was almost equal to the total annual road budget.
Thus, it became necessary to shift towards financing schemes that relied on user
charges.
In Chile, the Ministry of Finance was able to push for flexible-term contracts after
the Ministry of Public Works overspent between 2000 and 2003.43
4.2. Minimum Income Guarantees and Revenue Sharing
Minimum income guarantees and revenue sharing are common in PPP contracts.
However, they differ from the optimal contract because contract lengths are fixed and
do not last longer in low-demand states, implying that guarantees in those states are
excessive. The usual profit- and revenue-sharing agreements in fixed-term contracts
split revenues above a threshold in fixed proportions. By contrast, Propositions 2 and
3 suggest assigning all the revenue in excess of a given threshold to the government,
so the windfall profits tax rate should be 100%.44
More generally, the rationale behind real-world guarantees and revenue-sharing
schemes is to reduce the risk borne by the concessionaire. By contrast, the rationale
behind the optimal contract in Propositions 2 and 3 is to optimally trade off insurance
on one hand, and user fees and subsidies on the other. This is why the concession lasts
indefinitely when subsidies (i.e., guarantees) are granted and why in high-demand states
the term is variable and the concessionaire's revenue is higher than in low-demand
states.
43. PVR auctions were used recently in Chile to auction Route 160 (February 2007), the road accessing
Santiago's main airport (December 2007), the Melipilla-Camino de la Fruta highway (January 2008), and
the Vallenar-Caldera highway (January 2008).
44. The optimal contract under moral hazard discussed in Section 2.5 is closer to observed contracts in
high-demand scenarios, since the firm receives a fraction of marginal revenues in these states as is the
case under most observed sharing agreements. Yet the moral hazard contract implies a variable term in
high-demand states, and that the firm will share in marginal user fee revenue in low-demand states, which
is not the case under the often-used minimum income guarantees. The contract derived in Proposition
3 does better than the moral hazard contract in low-demand states. More importantly, as argued in the
Introduction, in the case of transport infrastructure, the moral hazard problem is not relevant, as service
quality is contractible and thus there is no way to increase demand significantly.
TABLE 2. Average discounted budget: public provision vs. PPPs (high and low demand).
Public provision PPP Privatization
Upfront surplus: —I 0 f of (v)dv — I — Risk Premium

Discounted user fees: f vf(v)dv f vf (v)dv — I 0
Total: f vf (v)dv — I f vf(v)dv — I f vf(v)dv — I — Risk Premium
4.3. Are PPPs Private or Public?
It has often been argued that the main reason why governments find PPPs an attractive
option is that investments financed this way do not contribute to the public debt.45
We use this paper's results to argue that, as far as the risk profile of the government's
budget is concerned, PPPs are much closer to public provision than to privatization.
Our starting point is that when thinking about the risk allocation implied by
PPPs, what matters is the risk profile of discounted cash flows, not the year-to-year
risk profile. This has interesting implications: for low- and high-demand projects, an
optimal PPP contract replicates the net cash flow streams of conventional provision,
state by state (see Table 2, which assumes an additive risk premium). Essentially, all
residual risk is transferred to the government, and the concessionaire recovers 1 in all
states, as with conventional provision.
This result implies that some of the most influential guidelines on when to classify
a PPP asset as government owned may be misguided, since they are based on whether
the concessionaire bears risk. For example, Eurostat, the Statistical Office of the
European Community, recommends that the assets involved in a PPP be classified as
nongovernment assets, and therefore recorded off balance sheet for government, if
both of the following conditions are met: (1) the private partner bears the construction
risk, and (2) the private partner bears at least one of either availability or demand
risk (Eurostat Press Office, February 11, 2004). When quality is contractible, as we
argued is the case for most of PPP investments in the transportation sector, demand risk
will be mainly exogenous and therefore this risk does not generate useful incentives.
Furthermore, in the case of a high- or low-demand project, the optimal contract
eliminates risk for the firm, with implications for the public budget that are identical
to those of public provision.
For intermediate-demand projects, Proposition 3 shows that some risk-sharing is

optimal. The extent to which the firm bears risk now depends on the extent to which
subsidies are a more costly source of financing than user fees, as captured by the
parameter When subsidy financing is very inefficient, it is too expensive to reduce
the firm's risk via subsidies, and it is best to have the firm bear most (sometimes all) of
the risk. The efficient PPP contract resembles privatization in this case. On the other
hand, if subsidy financing is only slightly less efficient than user-fee financing, the
45. "Cynics suspect that the government remains keen on PFI not because of the efficiencies it allegedly
offers but because it allows ministers to perform a useful accounting trick." The Economist, 2 July 2009.
minimum income guarantee and the cap on user fee revenues that characterize the
optimal contract are very similar, and the government bears most of the risk. As with
high- and low-demand projects, risk-sharing arrangements resemble public provision
in this case.
Under privatization, the project is sold for a one-time payment and all risk is
transferred to the firm. Moreover, the link between the project and the public budget
is permanently severed. This is not the case with a PPP, where at the margin cash
flows from the project always substitute for either taxes or subsidies. The conclusion,
then, is that from a public finance perspective there is a strong presumption that PPPs
are analogous to conventional provision—in essence, they remain public projects, and
should be treated as such.
4.4. Least Subsidy Auctions
Low-demand projects are sometimes awarded to the firm that bids the lowest subsidy
required to build, operate and maintain the project. However, these are fixed-duration
concessions. For that reason, and even under a competitive auction, the winning bid
includes a risk premium, because the concessionaire is forced to bear the risk of
varying revenue in different states of demand. This implies that the winning bid will be
higher than the difference between the cost of the project and expected revenues. On
the other hand, a PVR auction for a low-demand project provides complete insurance
to the concessionaire, so the winning bid in a competitive auction does not require
a risk premium. This leads to the somewhat counterintuitive result that the average
subsidy paid out with a PVR auction is lower than the winning bid in a lowest-subsidy
auction.46
5. Conclusion
The insights of this paper have implications for the question of whether PPPs are closer
to private or public provision of infrastructure. The existing literature focuses on the
incentives wrought by PPPs and concludes that they have several attributes typically
associated to privatization. For example, in a PPP the concessionaire owns assets
(Hart 2003);47 retains control over how to produce the service and may unilaterally
implement any cost-saving innovation (Bennet and Iossa 2006); and directly collects
user fees (Grout and Stevens 2003).48
We use a different approach, which derives the optimal PPP contract from a
public finance perspective. In this case the optimal contract provides full insurance for
46. For more details and a formal statement, see Engel, Fischer, and Galetovic (2009b).
47. Though usually it needs authorization to sell assets that are comprised in the concession.
48. In a different vein, Besley and Ghatak (2001) show that when contracts are incomplete, transferring
the ownership of the assets needed to produce a public good to a private party is efficient, when that party
values the benefits created by the public good relatively more.
high- and low-demand projects, and is characterized by a minimum income guarantee

and revenue cap for intermediate-demand projects. The impact on the intertemporal
budget is similar to that of public provision, suggesting that PPPs should be counted
as public projects in the budget.
Our framework stresses the importance of risk allocation in the face of large
demand uncertainty present in many PPP projects. We emphasize that the temporary
nature of PPP contracts can be used to improve welfare by allowing state-contingent
contract terms, making feasible risk allocations that are not available under privatization
or public provision.
The crisis that started in 2008 has left many countries with large budget deficits.
Hence, these countries may find PPPs an attractive solution to finance infrastructure
projects without increasing their apparent debt burden. We have shown that this should
not be a basis for choosing PPPs over conventional provision. PPPs should be favored
only when they lead to efficiency gains. To ensure this happens, PPPs should be given
the same treatment in budgetary accounting than publicly provided infrastructure.
Appendix A: Proofs of Propositions
A.1. Proposition 1
Proof . We solve the program (4a)—(4d) for the case where e = 0 (which is equivalent
to = 0, see footnote 26).
Since u is concave, applying Jensen's inequality to the concessionaire's
participation constraint leads to
u (f [R(v) + S(v)] f (v) dv — I) > f u(R(v) + S(v) — I) f (v) dv = u(0),
where we used that the planner leaves no rents to the firm (see the main text). And
since u is strictly increasing, the above inequality implies that
E[R] + E[S] > I,
where E[R] = f R(v)f(v)dv denotes the expected revenue before demand is realized
and E[S] denotes expected government expenditure on subsidies.
It follows that if the solution to
min E[R] + (1 + e)E[S]
R>O,S>0 (A.1a)
subj. to: E[R] + E[S] > I
satisfies equations (4c)—(4d), then it solves program (4a)—(4d) as well.

Hence, if = 0, any combination of revenue and subsidy schedules that satisfies
(4c), (4d), and R(v) + S(v) = I for all v, solves the planner's problem. 111
110
Engel,
108 Fischer, and Galetovic Public
Public Finance
Finance of Journal
PPPs
ofJournal
PPPs ofof the European Economic Association
111
109
107
Association
A.2. Proposition 3
Proof . Let A > 0 denote the multiplier of the concessionaire's participation constraint
(4b).49 The FOC with respect to R(v) for a state v such that the term of the concession is
finite leads to
1+A—a
u'(R(v) — I) = ______ (A.2)
it
while the FOC with respect to S(v) for a state where subsidies are paid leads to
(1 + AM + 0 — a
ut(v + S(v) — I) = (A.3)
It
where in both cases we have used that revenue financing dominates subsidy financing.
Define m and M via
(1 + AM -1- 0 — a
ut(m — I) = (A.4)
1+A—a
ut(M — I) = _______________________________ (A.5)
il
Since > 0 we have m < M and
ut(m — I) = (1 + Out(M — I),

It follows from (A.2) and (A.5) that in states with v > M no subsidies are paid out
and the concession lasts until the concessionaire collects M in present value. The
government, on the other hand, collects v — M after the concession ends. Thus, in
high-demand states the concessionaire's revenue is capped by M and the term of the
concession is variable.5°
Similarly, from (A.3) and (A.4) we have that a subsidy equal to m — v is paid in
states with v < m. Therefore, in low-demand states the concession lasts indefinitely
and the concessionaire receives a minimum revenue guarantee.
Finally, there is a third class of states of demand such that m < v < M. In these states
the concession lasts indefinitely, for otherwise they would be high-demand states. But
no subsidies are paid out by the government, for otherwise they would be low-demand
states. It follows that R(v) = v and S(v) = 0 in this class.
Having established that the optimal contract is a two-threshold contract, the
planner's problem is equivalent to finding m and M that minimize
M m
M(1 — F (M)) ±f v f (v) dv + (1 + OF (m) i (m — v) f (v) dv, (A.6)
o o
49. Note that the participation constraint will hold with equality because 1 + ? > a, hence a > 0.
50. Under the assumption that the user fee revenue process is well behaved, this implies that higher
values of v correspond, on average, to shorter concession terms.
subject to the concessionaire's participation constraint (5). Noting that (5) implicitly
defines M as a function of m, we have that
(1 — F(M))u'(M — I)
A similar calculation shows that the rate at which M and m have to change to keep the
objective function (A.6) unchanged is given by
(1 + OF(m)
•
M'(m) = (A.8)
1 — F(M)
Equating (A.7) and (A.8) for M'(m) leads to (6) and completes the proof.51 ❑
A.3. Proposition 5
Proof. Since all firms are identical, the winning bid of the competitive auction
minimizes the scoring function subject to firms' participation constraints. And since
the scoring function is equal to the planner's objective function, where we use the
fact that the optimal contract is characterized by thresholds m and M, it follows
that the winning bid maximizes the planner's objective function subject to the firm's
p ar t i c ip a t io n c o n s tr a in t , th e r e b y so l v i n g th e p lan n e r 's p ro b le m. ❑
Appendix B: PPP Statistics
TABLE B.1. PPP investment in Europe.

Total investment Fraction of public investment
Country (1990-2006, MM EUR) (2001-2006, %)
Belgium 2,112 3.5

France 7,670 1.3
Germany 5,658 1.5
Greece 7,600 5.9
Hungary 5,294 7.3
Italy 7,269 2.5
Netherlands 3,339 2.2
Portugal 11,254 22.8
Spain 24,886 6.9
UK 112,429 32.5a
aIf the London Underground is excluded, this becomes 20%.

Source: Blanc-Brude, Goldsmith, and Valila 2007. Ten countries with most investment.
51. This proof assumes that F(m) > 0 and F(M) < 1. Footnote 30 outlines the proof when this is not the
case.
TABLE B.2. PPP investment in developing countries (1990-2008, USD).

Country Energya Telecoma Transport Water-sewage Total
Argentina 29,540 29,328 14,094 8,176 81,137
Brazil 75,993 107,554 32,142 4,576 220,265
China 37,339 14,518 47,449 8,427 107,732
India 45,868 52,898 24,766 331 123,864
Indonesia 15,492 24,972 3,743 1,020 45,228
Malaysia 14,313 9,596 16,552 10,144 50,605
Mexico 10,753 54,068 25,374 1,675 91,869
Philippines 19,268 14,280 3,478 8,071 45,096
Russia 30,484 48,813 706 2,225 82,228
Turkey 12,678 24,293 8,170 942 46,082
aThe extent to which the projects in this sector fit with what we consider as a PPP in this paper, since they
correspond to infrastructure that is privatized (thereby severing any link with the public budget) and are
regulated as natural monopolies.
Source: World Bank-PPIAF PPI database. Ten countries with most investment.
TABLE B.3. Toll types for PPP roads, bridges and tunnels in Europe.
Country Availability Real toll Shadow toll Total
Austria 0 2 0 2
Finland 2 0 0 2
France 0 8 0 8
Germany 0 8 0 8
Greece 0 6 0 6
Hungary 0 5 0 5
Ireland 0 8 0 8
Italy 0 7 0 7
Latvia 1 0 0 1
Netherlands 2 0 1 3
Norway 0 3 0 3
Poland 0 1 1 2
Portugal 0 6 11 17
Spain 0 31 14 45
UK 4 3 20 27
Total 9 88 47 144
Source: Timo Valila at the European Investment Bank.
References
Akintola, Akintoye, Matthias Beck, and Cliff Hardcastle (2003). Public-Private Partnerships:
Managing Risks and Opportunities. Blackwell.
Arrow, Kenneth J. and Robert C. Lind (1970). "Uncertainty and Public Investment Decisions."
American Economic Review, 60, 364-378.
Atkinson, Anthony B. and Nicholas H. Stern (1974). "Pigou, Taxation and Public Goods." The
Review of Economic Studies, 47, 119-128.
Bennett, John and Elisabetta Iossa (2006). "Building and Managing Facilities for Public Services."
Journal of Public Economics, 90, 2143-2160.
Bentz, Andrew, Paul Grout, and Maija Halonen (2005). "What Should Governments Buy from the
Private Sector-Assets or Services?" Working paper, University of Bristol.
Besley, Timothy and Maitreesh Ghatak (2001). "Government Versus Private Ownership of Public
Goods." Quarterly Journal of Economics, 116, 4, 1343-1372.
Blanc-Brude, Frederic, Hugh Goldsmith, and Timo Valila (2007). "Public-Private Partnerships in
Europe: An Update." Economic and Financial Report No. 2007/03, European Investment Bank.
Brealey, Richard, Ian Cooper, and Michel Antoine Habib (1997). "Investment Appraisal in the Public
Sector." Oxford Review of Economic Policy, 13, 12-27.
Chadwick, Edwin (1859). "Results of Different Principles of Legislation in Europe." Journal of the
Royal Statistical Society, A22, 381-420.
Dahlby, Bev (2008). The Marginal Cost of Public Funds. MIT Press.
Daniels, Ronald J. and Michael J. Trebilcock (1996). "Private Provision of Public Infrastructure: An
Organizational Analysis of the Next Privatization Frontier." University of Toronto Law Journal,
46, 375-426.
Daniels, Ronald J. and Michael J. Trebilcock (2000). "An Organizational Analysis of the
Public-Private Partnership in the Provision of Public Infrastructure." In Public-Private Policy
Partnerships, edited by P. Vaillancourt-Roseneau. MIT Press.
Demsetz, Harold (1968). "Why Regulate Utilities." Journal of Law and Economics, 11, 55-66.
Dewatripont, Mathias and Patrick Legros (2005). "Public-Private Partnerships: Contract Design and
Risk Transfer." EIB Papers, 10, 120-145.
Economic Planning Advisory Commission (EPAC) (1995). Final Report of the Private Infrastructure
Task Force. Australian Government Publishing Service, Canberra.
Engel, Eduardo and Ronald Fischer (2010). "Optimal Resource Extraction Contracts Under Threat
of Expropriation." In Populism and Natural Resources, edited by W. Hogan and E Sturzenegger.
MIT Press.
Engel, Eduardo, Ronald Fischer, and Alexander Galetovic (2001). "Least-Present-Value-of-Revenue
Auctions and Highway Franchising." Journal of Political Economy, 109, 993-1020.
Engel, Eduardo, Ronald Fischer, and Alexander Galetovic (2003). "Privatizing Highways in Latin
America: Fixing What Went Wrong." Economia, 4, 129-158.
Engel, Eduardo, Ronald Fischer, and Alexander Galetovic (2009a). "On the Efficient Provision of
Roads." Working paper, Yale University.
Engel, Eduardo, Ronald Fischer, and Alexander Galetovic (2009b). "The Basic Public Finance
of Public-Private Partnerships." Working paper, http://cowles.econ.yale.edu/engel/pubs/basic-
pub.pdf.
Gerrard, Michael B (2001). "Public-Private Partnerships." Finance and Development, 38, 48-51.
Grimsey, Darrin and Mervyn Lewis (2004). The Economics of Public-Private Partnerships. Edward
Elgar.
Grimsey, Darrin and Mervyn Lewis (2005). Public-Private Partnerships. Edward Elgar.
Grimsey, Darrin and Mervyn Lewis (2007). "Public Private Partnerships and Public Procurement."
Agenda, 14, 171-188.
Grout, Paul (1997). "The Economics of the Private Finance Initiative." Oxford Review of Economic
Policy, 13, 53-66.
Grout, Paul and Margaret Stevens (2003). "The Asessment: Financing and Managing Public
Services." Oxford Review of Economic Policy, 19, 215-234.
Guasch, J. Luis (2004). Granting and Renegotiating Infrastructure Concessions: Doing it Right. The
World Bank.
Harstad, Ronald and Michael Crew (1999). "Franchise Bidding Without Holdups: Utility Regulation
with Efficient Pricing and Choice of Provider." Journal of Regulatory Economics, 15, 141-
163.
Hart, Oliver (2003). "Incomplete Contracts and Public Ownership: Remarks and an Application to
Public-Private Partnerships," Economic Journal, 113, C69-C76.
Hemming, Richard (2006). "Public-Private Partnerships, Government Guarantees and Fiscal Risk."
IMF.
lossa, Elisabetta and David Martimort (2008). "The Simple Microeconomics of Public–Private
Partnerships." Working paper, University of Toulouse.
Irwin, Timothy (2007). Government Guarantees: Allocating and Valuing Risk in Privately Financed
Infrastructure Projects. The World Bank.
Klein, Michael (1997). "The Risk Premium For Evaluating Public Projects." Oxford Review of
Economic Policy, 13,29-42.
Laffont, Jean-Jackques and Jean Tirole (1993). A Theory of Incentives in Procurement and Regulation.
MIT Press.
Martimort, David and Jerome Pouyet (2008). "To Build or Not to Build: Normative and Positive
Theories of Private–Public Partnerships.", International Journal of Industrial Organization, 26,
393-411.
Michel-Kerjan, Erwann 0. (2010). "Catastrophe Economics: The National Flood Insurance
Program.", Journal of Economic Perspectives, 24,165-186.
Posner, Richard A (1972). "The Appropriate Scope of Regulation in Cable Television." Bell Journal
of Economics, 3,335-358.
Riordan, Michael H. and David E. M. Sappington (1987). "Awarding Monopoly Franchises."
American Economic Review, 77,375-387.
Savas, Emanuel S (2000). Privatization and Public-Private Partnerships. Chatham House.
Spulber, Daniel (1989). Regulation and Markets. MIT Press.
Starr, Paul (1988). "The Meaning of Privatization." Yale Law and Policy Review, 6,6-41.
Stigler, George (1968). The Organization of Industry. Richard D. Irwin, Homewood.
Vaillancourt-Roseneau, Pauline (2000). Public–Private Policy Partnerships. MIT Press.
Williamson, Oliver E (1976). "Franchise Bidding for Natural Monopolies—In General and with
Respect to CATV." Bell Journal of Economics, 7,73-104.
Williamson, Oliver E (1985). The Economic Institutions of Capitalism. The Free Press.

001 - THE BASIC PUBLIC FINANCE OF - Engel Fisher

Uploaded by

Copyright:

Available Formats

001 - THE BASIC PUBLIC FINANCE OF - Engel Fisher

Uploaded by

Document Information

Original Title

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

001 - THE BASIC PUBLIC FINANCE OF - Engel Fisher

Uploaded by

Copyright:

Available Formats

THE BASIC PUBLIC FINANCE OF

Eduardo Engel Ronald Fischer

1. Introduction and Motivation

The editor in charge of this paper was George Marios Angeletos.

Journal of the European Economic Association February 2013 11(1):83-111

investment in PPPs in developing countries grew at an average annual rate of 28.3%

2. Model and Optimal Contract

A risk-neutral benevolent social planner must hire a concessionaire to finance, build

2.1. Planner's Problem

welfare function.16 The planner's objective is to maximizel7

f[CS(v) + aPS(v)] f (v) dv, (1)

subject to the concessionaire's participation constraint

fu(PS(v)) f (v) dv > u(0),

PS(v) = R(v) + S(v) — I. (2)

CS(v) = [v — R(v) — (1 + A)S(v)] ATI) — R(v)] = (1 + A)[v — R(v) — S(v)].

("ITR pays near market wages—unfilled vacancies", http://www.tollroadsnews.com./node/1466.)

"[...] one-third of its collected premiums—equivalent to a 50 percent loading cost—to

and therefore to minimizing

f{[(1 + A) — a]R(v) [(1A)(1 + — a]S(v)} f (v) dv, (4)

Defining > 0 via26

min f {R(v) + (1 + -)S(v)} f (v) dv, (4a)

subj. to: f u(R(v) + S(v) — I) f (v) dv > u(0), (4b)

0 < R(v) < v, (4c)

S(v) > 0. (4d)

2.2. Irrelevance Result

fR(v) f (v) dv + f S(v) f (v) dv.

PROPOSITION 1. (Irrelevance of the cost-of-funds argument) Assume 4' = 0. Then any

2.3. Optimal Risk-Sharing Contract: High- And Low-Demand Projects

TABLE 1. Taxonomy of demand states and taxonomy of projects.

High: R(v) < v, S(v) = 0 v min > I

2.4. Optimal Risk-Sharing Contract: Intermediate Demand Project

Figure 1 shows how the subsidy—risk tradeoff is resolved optimally when

As can be seen from Figure 1, the wedge between 1 + A — a and (1 + +

Revenue for the

FIGURE 1. Optimal contract, intermediate demand project.

PROPOSITION 3. (Optimal contract for intermediate-demand projects). Consider a

L n U(V I)f (v) dv + (1 — F(M))u(M — I) = u(0).

F(m)u(m — I) + f u(v — 1)f (v) dv = u(0).

and the condition

Proof . See Appendix A.2. ❑

Price-Responsive Demand. Assuming a totally inelastic demand simplifies the

Revenue for the

FIGURE 2. Optimal contract with moral hazard and > 0.

Moral Hazard. Another extension is to allow for demand that depends on

3.1. High- and Low-Demand Projects

f 14(,3 — I) f (v) dv = u(0). (7)

PROPOSITION 4. (High- and low-demand projects). The optimal contract can be

3.2. The General Case

PROPOSITION 5. (Optimality of the two-threshold auction). The following two-

W (M , m) = M(1 — F (M)) + v f (v) dv + (1 + e)f (m — v) f (v) dv

41. What follows is based on Project Finance, February 2005.

4.2. Minimum Income Guarantees and Revenue Sharing

Upfront surplus: —I 0 f of (v)dv — I — Risk Premium

4.3. Are PPPs Private or Public?

For intermediate-demand projects, Proposition 3 shows that some risk-sharing is

4.4. Least Subsidy Auctions

high- and low-demand projects, and is characterized by a minimum income guarantee

Appendix A: Proofs of Propositions

u (f [R(v) + S(v)] f (v) dv — I) > f u(R(v) + S(v) — I) f (v) dv = u(0),