Contractual design and PPPs for hospitals: lessons for the Portuguese model

Recently the Portuguese Government announced the launching of public–private partnerships (PPPs) to build hospitals with the distinctive feature that infrastructure construction and clinical activities management will be awarded to separate private parties. Also, one of the parties will be in charge of providing soft facilities. We explore alternative configurations of contracts and assess whether the equilibrium allocations attain the first-best solution.


Introduction
Public-private partnerships (PPPs) have become a popular strategy in the production and delivery of public services. The European Commission (2005) in its Lisbon strategy to strengthen Europe's position as a technologically innovative country, recognizes the role of PPPs involving the industry, the research community, and the public authorities in meeting those challenges. In particular, the 7th Framework Program introduced the concept of joint technology initiatives as a new way of realizing public-private partnerships at European level.
Given the financial constraints faced by governments, PPPs allow the possibility of providing high quality public services, thus extending the scope of analysis to what is known as the "new public management" in the provision of public services.  and Webb and Pulle (2002) provide complete introductions to this topic; Grimsey and Lewis (2004) present an in-depth analysis of the PPPs, and Grimsey and Lewis (2005) review a selection of the main papers in this relatively new literature. Also, Zarco-Jasso (2005) and Patel et al. (2007) provide a general approach to PPPs as defined by their ownership, funding and control.
PPPs have been most developed in the UK's so-called Private Finance Initiatives (PFIs), but PPPs are spreading across the world (EU countries, Canada, USA, Japan, Australia, New Zealand, and South Africa are representative examples) 1 in multiple forms. 2 Key sectors where PPPs are often applied are health care provision, education, transport, 3 and defense.
The introduction of the PPPs has not avoided controversy. Gaffney et al. (1999) show concerns on the fact that (health care) facilities funded through PFI in most cases provided less capacity than those they were intended to replace; Ahadzi and Bowles (2004) and Pollock et al. (1997) focus on the high costs and pre-contract time overruns that are often reported ;Shortell et al. (2002), Spitz and Ritter (2002), Emanuel and Titlow (2002) or Deakin (2002) alert on the negative or mixed results of the evaluations of the public private partnerships as a device to improve community health. Bazzoli et al. (1997) however, sustain that public-private partnerships have great potential 1 Information on PPP projects by sectors and countries is available at www.pppbulletin.co.uk. For the UK, the HM Treasury Department (www.hm-treasury.gov.uk) updates biannually the information on the PFI signed projects.
2 See, for instance, the European Commission (2003) guidelines for PPPs as an attempt to harmonize criteria and practices across the EU countries.
to improve coordination and effectiveness in community health delivery. Carvel (2002) supports the use of the PFI, and Shaw (2003) explores the pros and cons of the use of PFIs within the New Labour's policy paradigm of the Blair government.
From a different perspective, Atun and McKee (2005) relate the failure of the PFIs in the UK to the accountancy criteria used in the public budget rather than to the mechanism itself. Nevertheless, they note that a fundamental flaw at the heart of the PFI scheme is its lack of flexibility due to the combination of long term contracts with the faster pace of change in the delivery of health care. De Bettignies and Ross (2004) review the fundamental underlying economics of the PPPs to clarify the controversies driven more by ideology than careful analysis.
In general, there are two main reasons for public private partnerships: high risk of what?
should we say something about it? and the appropriability issue (Bachula, 1998). again, should we clarify what this means? Technology-based competition erodes appropriability and increases risk. Thus, these are factors working against the social objectives of an effective competition policy. In turn, competition policy is designed on the premise that innovation leads to a faster technological advance which increases economic growth and competitiveness of firms.
Accordingly, there is a trade off between multiple independent firms competing and the benefits of the coordination to reduce risk and appropriability problems. One approach to this trade off is the design of guidelines for a public private partnership competition policy. Link and Scott (2001) contribute to this literature. Grimshaw et al. (2002) argue that there is little evidence of mutual gains from PPPs and, when they arise, their distribution is not equitable. In a somewhat similar line, Wettenhall (2003) and Estache (2006) study conditions for successful public private partnerships aimed at protecting the public interest against the power exerted by market forces. Pongsiri (2002) argues that the merit of public private partnerships is that they are oriented towards the mutual benefit. But as the role of the government is also to monitor the market place, it is necessary a well-defined regulation framework. Such regulatory scheme should ensure efficiency and optimality in the resources available to the partnership in line with the general policy objectives. Also, it should provide protection to the private sector against expropriation, respect for contractual agreements, and legitimate recovery of costs and profit proportional to the risk undertaken. On a different viewpoint, Timmons and Marx (2004) address consumer protection and safety under the presence of public private partnerships in the health care sector.
In this paper we will concentrate in the design of PPPs in the health care sector. Drevdahl (2002) looks at the consequences of the partnership between the public health agencies and managed care organizations in terms of the sustainability of the provision of population-oriented care, given the conflict between concern for populations and communities and the interest of the stockholders. Nishtar (2004)  From a different perspective,  note that there are many criteria to guide whether PFIs in the UK National Health Service should be pursued at the pre-decision stage. 5 Concerned by the lack of post-project evaluation, they suggest a system for post-project evaluation.
We can summarize the across country differences in the design of public private partnership contract in two general structures. One consists in bundling investment and service provision into a single contract. Engel et al. (2007) claim that this is the usual contracting format in road building (which is the activity attracting the highest share of PPP arrangements) and study the characteristics of the optimal contract. The other, sometimes used in the health care sector, contemplates two different contracts for the investment and for the provision of the service. These differences among contractual structures and among countries lies behind the poor understand-  Figure 1: The roadmap of models studied.
Inspired by the Portuguese approach to the launching of public private partnerships to build new hospitals (see section 2 for details), we assume that the Government offers two contracts.
One with the institution in charge of the construction and maintenance of the hospital; the other with the entity in charge of clinical activities. Also, one of these parties will be responsible for the subcontracting of the soft facilities, thus yielding two variations of the contracting set-up.
Alternatively, we could envisage a scenario where the hospital is already built. The Government decides whether to offer a contract to the clinical facilities entity who in turn, is responsible for subcontracting the soft facilities, or propose two separate contracts for the soft and clinical services. On the figure -since we are in the introduction, we should avoid parameters and formulas. I could not edit the figrues. I suggest we just write "investment in infra-structures" for k, "soft facilities" for s, clinical services for c, and "no externality from infra-structures investment to cost of clinical services", on the top part; on the bottom part, use "single contract" and "separate contracts" In all cases benefit and cost functions are left unspecified and we assume linear payments on costs, which conforms well with observed practice; also, contracts are assumed to specify in full all economic transactions between the entities and the Government. Finally, we abstract from the usual asymmetric information problems. These are dealt with in Pérez-Castrillo and Macho-Stadler (1988), Jelovac and Macho-Stadler (2002), Tommassi and Weinschelbaum (2007), and Boadway et al. (2004). The model of Jelovac and Macho-Stadler (2002) looks precisely at the issue of centralization and decentralization in the context of health care provision. Their primary example uses hospitals and physicians as the relevant agents. Decentralization in their setting means that an entity contracted by the payer (a hospital) will set the contract to a second entity (the physicians). Centralization means the payer offers directly contracts to both physicians and hospitals. Tommassi and Weinschelbaum (2007) study this problem inspired in the policy decisions taken at a higher level (such as monetary policy designed by the European Central Bank) and other policy decisions taken "closer to the people" (such as education policy that is designed at the national level). Also, Engel et al. (2007) look at the characteristics of a single optimal contract where construction and service provision are bundled together.
We depart from the analysis of Jelovac and Macho-Stadler (2002) in that (i) we consider the application of linear payment schedules and (ii) we allow patients' welfare to be a concern of providers as well, leading us to a version of Ellis and McGuire (1986) model.
We assess whether the equilibrium allocations of the seven models proposed attain the firstbest (socially optimal) solution (see figure 1). When the hospital is already in place, we find that this is the case. When building the hospital (investment in infra-structures) is part of the contract, the first-best allocation is always reached except when hard and soft facilities are bundled together in the contract. However, in this case if restrictions are given by budget constraints we attain the first-best allocation when the externality arising from the investment in infrastructure only affects the level of quality of the software facilities, but not the the cost of providing clinical services.
The paper is organized in the following way. Section 2 describes the Portuguese case of public private partnerships for hospitals. Section 3 introduces the model and characterizes the first-best allocation. Section 4 presents the alternative contract scenarios when the construction of the hospital is part of the contract and discusses the conditions allowing to achieve the firstbest allocation. Section 5 reviews the previous models when objective functions are subject to budget constraints instead of participation constraints. Section 6 studies the situation where the hospital is already in place and only clinical and soft facilities are to be contracted. Section 7 contains some final remarks. are awarded in the PPP contract (although in some cases, Italy adopted the UK model). Another country that uses the PPP instrument is Spain, in which we find both the UK model and a more integrated view. In this integrated view, management of clinical activities are included in the PPP contract and, in a broader approach, we find also a case where primary care is added to the contract with the private partner. In any case, a single contract is used, covering both primary care and hospital care, and both building (and maintenance) and operations management.
The use of two contracts with separate, although closely related, entities, at least during the first time period of 10 years, is an important departure from existing PPP experience and deserves attention.
A particular point of discussion, when two contracts are employed, is where to include management of activities that could be, in principle, awarded to any of the two contracts. management is probably in better position to monitor quality of soft facilities provision. On the other hand, functional integration with hard facilities and heavy equipment may allow for less costly operation and better coordination.
Since typical soft facilities activities (like laundry, food catering, cleaning, security and waste disposal, for example) are contracted out, the question of to which contract they should be allocated can be seen as which party (clinical activities management or hard facilities management) entails smaller distortions in subcontracting such soft facilities provision.
The answer to this question needs to consider the nature of the contract that is established between the Government and each entity, as well as the type of contract that it is used to ensure provision of soft facilities. The remaining of the paper is devoted to obtain a policy-relevant answer to this issue, which given the diversity of PPP models is also of wider interest.

The model
Hospital production (to be thought of as patients' health) is described by a benefit function , where x is the input level of clinical activities, q is the quality of soft facilities, and k denotes the investment in hard facilities. This health production function B has the standard regularity properties: positive marginal effects of each input (∂B/∂x > 0; ∂B/∂q > 0; ∂B/∂k > 0) at a decreasing rate (∂ 2 B/∂x 2 < 0; ∂ 2 B/∂q 2 < 0; ∂ 2 B/∂k 2 < 0).
Costs of providing clinical services are given by C(x, k), costs of providing quality are given by S(q, k), and cost of hard facilities are H(k). We consider positive and increasing marginal costs in all activities (∂C/∂x > 0; ∂ 2 C/∂x 2 > 0; ∂S/∂q > 0; ∂ 2 S/∂q 2 > 0; ∂H/∂k > 0; ∂ 2 H/∂k 2 > 0). We assume away the existence of economies of scope. 7 The output measure B, is known to the Government, although its individual components are not, and thus they cannot be used to define payment rules. Cost functions are not known to the Government either. It only observes the realized cost in each activity.
The Government's social welfare function is defined as total benefits to the patients net of the total cost. We assume benefit B already defined in monetary units. Therefore, the social 7 Its existence would naturally favor joint production.
welfare function is: and we are assuming that participation constraints for providers of clinical services, soft facilities and infra-structures hold at the first best. The first-order conditions characterizing the optimal solution are given by: 8 These first-order conditions will serve as a benchmark to the analysis in the next sections. Note that at this level of generality different combinations of (x, q, k)-values may satisfy the system of first order conditions.

Contract choice
The Government may decide upon two contract (payment rule) structures to allocate hard, soft and clinical facilities. In one scenario, the Government signs a contract with the hard facilities provider who in turn will subcontract the soft facilities. Also, it signs a separate contract for clinical activities. In this situation, there appears a moral hazard problem because the level of quality provided by the soft facilities provider is not observable by the hard facilities provider.
In the second scenario, the clinical activities provider subcontracts the soft facilities. A separate contract determines the provision of hard facilities. Here we assume that clinical facilities provider can monitor the quality of the soft activities.
We denote by T 1 the contract between the Government and the clinical facilities provider, and by T 2 the contract between the Government and the hard facilities provider.
Finally, Ω denotes the payment rule in the subcontracting. Naturally, the structure of these payments is contingent to the scenario under analysis. Figure 2 illustrates. Accordingly, the contract between the Government and the clinical facilities provider contemplates the following (linear) payment schedule: The objective function of the clinical facilities provider (W c ) includes the benefits of the patients and its own profits (Π c ), that is: 9 The problem of the clinical services provider is to choose the level of service that maximizes its objective function. The associated first order condition is: The objective function of the hard facilities provider (W h ) also takes into account the benefit of patients and its own profits (Π h ). Given that the provider of hard facilities subcontracts the provision of soft facilities, its optimization problem must include the participation constraint and the incentive compatibility constraint induced by the soft facilities provider.
Finally, the objective function of the provider of soft facilities (W s ) depends of the benefit of patients, and its own profits (Π s ).
The contract between the Government and the hard facilities provider contemplates a (linear) payment including the finance for the provision of infrastructure and soft facilities, that is, where Ω denotes the payment rule associated to the subcontracting of the soft facilities: The problem of the hard facilities provider is thus: The system of first order conditions is (where λ 1 and λ 2 are the Lagrange multipliers associated with the two restrictions): From (10) and (11) we obtain, Also, we compute w 0 as the value satisfying (12), and the value w 1 as the value satisfying (13): We can characterize the equilibrium values for (x, q, k) from (5), (8), (9) after substituting the values of λ 1 , λ 2 and (13).
To assess whether the equilibrium solution allows to reach the first-best allocation, we compare the systems of first-order conditions. Note however, that as different values of (x, q, k) may satisfy the first-best allocation, we can only obtain a definite answer if both systems of first-order conditions coincide.
The equilibrium value of x is given by (5), while its first-best is given by (1). Substituting (1) into (5) and using (4), we obtain that the equilibrium value of x hits the first-best for The equilibrium values of q and k are given by (8) and (9), while their first-best levels are given by (2) and (3). Substituting (2) into (9) we obtain, where we have made use of (7).
As we have assumed that participation constraints are satisfied in the first-best, let us ignore the last term in (8) and also assume away the externalities, i.e. ∂C/∂k = ∂S/∂k = 0. Then, we Given that the values of β 1 just obtained do not coincide, we obtain our first result: A PPP procedure where provision of soft facilities is subcontracted by the hard facilities provider is not able to reach the first-best solution.

Clinical facilities provider subcontracts soft facilities
Let us now consider the scenario where the government signs a contract with the clinical facilities provider who bears the responsibility to (sub)contract the soft facilities. In this case, the contract between the Government and the hard facilities provider is defined by a payment scheme as follows: The objective function of the hard facilities provider includes the benefits of the patients and its own profits, that is The problem of the hard services provider is to choose the level of investment that maximizes its objective function. The associated first order condition is: In a similar vein as in the previous section, the contract between the Government and the clinical facilities provider contemplates a (linear) payment including the finance for the provision of clinical and soft facilities, that is, where Ω denotes the payment rule associated to the subcontracting of the soft facilities: The problem of the clinical facilities provider includes the participation constraint induced by the provider of soft facilities: The system of first order conditions is parallel to the system (8)-(11) substituting the superindex h by c, and k by x. Condition (12) remains unaltered and so it does the value of w 0 .
We characterize the equlibrium values for (x, q, k) from (15), and the parallel expressions corresponding to (8) and (9) after substituting the values of λ 1 and (13).
To assess whether the equilibrium solution allows to reach the first-best allocation, we compare the systems of first order conditions. Assuming away externalities and taking into account that we assume that participation constraints are satisfied in the first-best solution, we obtain that (x, q, k)-equilibrium values implement the first-best allocation for This leads to our second result. Assume a Government is designing a PPP procedure with award of three activities (clinical services, soft facilities and infra-structure investment).
Assume away cost externalities. Assume that only linear payment schedules on observed costs are feasible. Assume clinical services management has better information on soft facilities quality. Then, a PPP procedure where soft facilities are subcontracted by the clinical services management is able to implement the first-best solution.

Participation vs. budget constraints
The analysis presented thus far assumes that the subcontracting of soft facilities induces a restriction in the optimization problem of the subcontractor in the form of a participation constraint.
Alternatively, we can condition the optimization problem to satisfying a budget constraint.

Social welfare
This means that in characterizing the first-best allocation, the government maximizes a social welfare function under the condition that contracts T 1 and T 2 allow to cover the costs of provision of hard, soft, and clinical facilities. Formally, The system of first order conditions with µ 1 and µ 2 standing for the Lagrange multipliers is, Now (21) yields µ 1 = 0, and (23) yields µ 2 = 0. Substituting these values in the system of first order conditions, it reduces to Therefore, β 1 and α 1 can be anything as long as α 0 and β 0 are used to balance the budget (introducing cost of public funding will not change this implication). The same will hold true for the other problem. The budget constraints will have a zero shadow cost at the optimum.

Bundling Hard and soft facilities
In a parallel analysis to the one in section 4.1, the problem of provision of clinical services yields: The contract of the provider of hard facilities is given by: The system of first-order conditions is, where From (32) and (33) it follows that λ 2 = 0 and From (30) we obtain w 0 , and from (31) we obtain the value of w 1 : Substituting the values of λ 1 and λ 2 into (28) and (29) we obtain, The first-order condition of the clinical services provider is, Rearranging (34), (35) and (36) we obtain, The equilibrium characterized by this system of first-order conditions is also a first-best allocation if Note, that the last condition eliminates the externality over clinical management cost. Therefore, if such an externality exists the first best allocation cannot be achieved. Therefore, we have our third result:In the absence of cost externalities from investment in infra-structure relative to provision of clinical services, the PPP procedure in which soft facilities are subcontracted by the hard facilities management can achieve the first-best solution. Otherwise, in the presence of such cost externalities, the first-best solution is not attainable.

Joining clinical and soft facilities with moral hazard
The first-order condition of the problem of the provider of hard facilities is The problem to be solved by the clinical activities management is The first-order conditions are given by Rearranging the first three conditions we obtain, In this case the cost externalities are internalized through β 1 and contract (w 0 , w 1 ) solves the moral hazard problem.

Provider of clinical activities sets directly q
Let us assume away the problem of moral hazard by letting the clinical activities provider set directly thr level of quality (q) of the soft facilities. the the problem to be solved is, The system of first-order conditions is, Equation (39) yields the value of w 0 , and from (40) and (41) we obtain, Then, we can rewrite equations (37) and (38) as Note that the same value of α 1 aligns variables x and q to achieve the first best solution. Also, we obtain the value of β 1 that corrects the externality. Finally, we have a degree of freedom to determine the value of w 1 from equation (39).

Contracting clinical and soft facilities only
Consider a scenario where a hospital is already in place and the Government needs to contract soft and clinical facilities only. This amounts to have a fixed value of k in the previous section. We maintain all the remaining relevant assumptions on the properties of functions B(x, q), C(x), and S(q), the linearity of the payment rules and the informational structure.
We assume that either soft facilities are included in the contract of clinical activities management, or alternatively, they are included in the infrastructure institution, being directly paid by the Government. In the first option, the clinical activities management has to contract with the soft facilities provider its input. Another option would be the clinical activities management to provide directly the soft facilities services, a situation we term vertical integration.
When the Government sets two contracts, T 1 , T 2 , these are defined by If a single contract exists, the payment rule is described by where Ω(q) is the payment made by the entity in charge of clinical activities management for the the soft facilities input.
The objective function of each entity takes into account benefits for patients and its own profits: for the entity that explores clinical activities, and for the entity that is in charge of soft facilities management. We assume for simplicity, that That is, both entities value in the same way patients' welfare and own profits.
Under integration of both activities in the same entity, we have which may differ from the single contract case as long as Ω(q) differs from S(q).
Our interest lies in the comparison of two alternative contracting strategies by the Government with the first-best allocation.
The social optimum solution is given by (1) and (2).

Two separate contracts
Solving by backward induction, we first present the optimal choice of each entity, given the linear payment rule. Then we discuss the optimal choice by the Government, of the parameters defining the payment rules.
The problem faced by the soft facilities management is max q W s = V (B(x, q; k), T 2 (q) − S(q)).
The associated first-order condition is given by: 10 where Π 2 = T 2 (q) − S(q) is the financial surplus accruing to the management of soft facilities.
Similarly, the problem faced by clinical activities management is: The first-order condition is: Finally, the problem of the Government is where x * and q * are the optimal solutions resulting from the previous problems.
Given the simple structure of the problem, the implementation of the social optimum requires and (α 0 , β 0 ) are chosen such that V * 1 = 0 and V * 2 = 0. A first question is whether these conditions give origin to a well-defined set of parameters.
Since ∂T 1 /∂x = α 1 and ∂T 2 /∂q = β 1 , notice that in equation (42) we are interested in the relative weights for the clinical activities management between patients' welfare and profits 10 The standard regularity assumptions assumed earlier ensure second-order conditions for a maximum hold. and in equation (43) the relevant agent is the soft facilities management. Thus, by appropriate definition of the parameters, the first-best allocation is achieved with the Government contracting with both entities, even if the Government does not control directly the input of soft facilities management and of clinical activities. This is, essentially, an extension of the Ellis and McGuire (1986) result, to the case of two decision variables.
Note that we need both conditions to hold simultaneously, as marginal valuations for one entity are conditional on the decision taken by other entity. There is no efficiency cost, under our assumptions, from the existence of two different contracts. Therefore, the possibility of monitoring soft facilities management by clinical activities management can, at best, lead to the same optimal allocation of resources (although surplus allocation across economic agents may well differ). Therefore, the following results: Assume infra-structure investment has already been done (alternatively, is specified in high detail in the contract), then it is neutral to the allocation of resources which entity has the responsibility to sub-contract soft facilities quality.

Sequential contracts
We take now the case where the Government sets a single contract with the clinical activities provider who in turn, has to set the contract to the soft facilities management.
A consequence of this cascade, or sequence, of contracts is that payments by the clinical activities entity to the soft facilities management is a cost to clinical activities management that is eligible for reimbursement. As such, it is included in the allowable cost basis for payment purposes.
We take first the case where clinical activities management can perfectly monitor soft facilities quality. Thus, clinical activities management can effectively choose q and set payment Ω(q) so that the participation constraint of soft facilities management is satisfied.
The second-stage choice problem is given by where the constraint requires that a minimum level of utility is achieved by soft facilities managers.
The corresponding first-order conditions are (with λ being the Lagrange multiplier): Inspection of first-order conditions reveal that equations (46) and (47) define the same constraint for S(q) > 0, which introduces one degree of freedom in the payment rule. Moreover, from these conditions, λ = 1 − α 1 , which allows to simplify equation (44) to Comparison with the first-best allocation reveals that the socially optimal choice of x will result if α 1 is such that the following condition holds: Finally, substituting the value of λ into (45) we obtain, Accordingly, the definition of α 1 above is also sufficient to induce the first-best choice of quality q if ∂V /∂Π 1 = ∂V /∂Π 2 . Therefore, the first-best allocation of resources can also be achieved in a sequence of contracts, though with a different parameter for the cost sharing between the Government and clinical activities management: This parameter can be higher or lower than 1, and it can be higher or lower than the one in the case of separate contracts.
Assuming ∂V /∂Π > ∂V /∂B, it follows α 1 < 1. Therefore, the optimal contract set by the Government will have now less cost sharing than in the case of separate contracts.
The reason for this is relatively simple, as the clinical activities management has a smaller interest in introducing incentives to efficient provision of soft facilities quality the more it benefits from cost sharing.
Two remarks are worth noting. First, if the contract between the Government and the clinical activities management may specify constraints over w 0 and w 1 the first-best may be easier to attain. Suppose that full cost reimbursement is imposed in the payment for soft facilities quality, though clinical activities management can perfectly monitor this quality, and the same financial rent as in the separate contracts situation is given to the parties. Then, the optimal contract between the Government and clinical activities management is equal to the one of the case of two separate contracts.
On the other hand, if there is no clear information advantage in monitoring soft facilities quality by the clinical activities management relative to the Government, then the sequence of contracts cannot implement the first-best allocation.

Final Remarks
The use of public-private partnerships has become popular with budget-constrained Governments to produce and deliver services of public interest, covering interventions in different sectors. Recent assessments and discussions are calling into question the advantages highlighted by PPP promoters.
We concentrate here on a specific type of PPPs: the ones for hospitals construction and operation. Motivated by a distinctive feature, the possibility of having different types of activities contracted out in a PPP procedure, we address under which conditions the design of the PPP procedure can achieve the allocation of resources that would be selected by the Government in a first-best situation. The basic trade-off we unveil is between a better monitoring role performed by clinical activities management over the quality of soft facilities provision and the internalization of infra-structure investments impact on the costs of providing clinical services and on costs of providing quality in soft facilities output. This latter aspect reflects the belief that a wellthought hospital design facilitates clinical activities and lowers the cost of providing quality in non-medical services (the so-called soft-facilities). Since most discussions on PPPs focus on the financial incentives, this trade-off, admittedly specific to health care, is often neglected.
The definition of the PPP procedure, which entails two contracts plus a sub-contract (for soft facilities quality) is not neutral from the social welfare viewpoint. Under the more general setting, there is no easy trade-off between taking advantage of better monitoring (includes soft facilities subcontract in the responsibilities of clinical activities management) and internalizing cost externalities (includes soft facilities contract in the responsibilities of the infra-structures company). The option taken in the Portuguese PPP procedure was the former. We show here this is the right approach to implement the first best whenever there are no cost externalities from infra-structure investment on the cost of clinical services. Otherwise, none of the two options achieves the first-best allocation of resources. Note that externalities from investment on the cost of quality from soft facilities are always internalized through the contract.
The existence of a sequence of contracts does not add additional complexity, as we have assumed away the informational issues that have been already treated in the economics literature on decentralization within organizations.
Interestingly, the results mentioned above depend crucially on the role of investment in infrastructures generating cost externalities. To show it more clearly, we have also considered a situation where this investment was already committed. Then, the option of including soft-facilities services under the contract of the Government with the clinical activities management, or not, is irrelevant to achieve the first-best solution. Naturally, the optimal contract that allows the Government to achieve the first-best solution is different in each case, involving less cost sharing to clinical services management when it sub-contracts soft facilities. Although the careful reader may find this discussion a side step to the main issue, its relevance is uncovered by the notice that one possible solution to difficulties with PPPs which were initially just for the heavy infra-structure design, construction and maintenance, is to give also operation of clinical activities under a PPP system. The stylized description also matches the award, in 1996, of a private management contract to run a publicly built hospital in Portugal, where soft facilities were included in the contract with the private party.
Taking all together, there is no simple answer to the PPP procedure design issue, though whenever the impact of infra-structures investment on costs of clinical services is small, then the management of the latter should also be in charge of sub-contracting soft facilities. Since in the Portuguese first wave of PPPs the terms of reference include detailed dispositions about the construction project, freedom to innovate and generate the relevant cost externalities at investment in infra-structures was small. Under these conditions, the option selected by the Portuguese Government of leaving soft facilities to be managed by the clinical services provider seems to be the better option.
Several caveats need to be pointed out. An issue not addressed is the timing of the contracts.
For clinical activities management, contracts are predicted to have a duration of 10 years, while for building and maintenance the contract will be awarded for 30 years. How the different time horizons will change, or not, incentives for investment is an open issue. A second issue, valid for both options, is the scope for future renegotiation of contracts. Experience has shown hat health care delivery is sensitive to cost shocks due to technological progress (which usually entails more costs), and, in Portugal at least, to "soft budget constraint" problems, making future renegotiation of contract terms a concern that may have a bearing on contract design. A third issue, intentionally left aside, is asymmetric information, which would result under decentralization in a sequence of incentive compatible contracts. This problem has been addressed, in a slighlty different context, by Boadway, Marchand and Sato (2004). A final issue is the possibility of collusion between the two entities, something that may be fostered by the existence of common shareholders in both entities. If we take collusion as meaning that both entities behave as a single containing both activities within it, then the first-best is again attainable, and the existence of two contracts or a sequence of contracts is irrelevant, as the effects of the contract between clinical activities management and the soft facilities entity will be fully internalized.
Our analysis complements, therefore, the existing literature and helps to explain observed choices. Future research to address extension of the main results to more general settings is welcome.