OPTIMAL COORDINATION,

TAXATION AND GUARANTEES

FOR PRIVATE PARTICIPATION
IN PUBLIC INVESTMENT

Renato Reside Jr., Ph. D.

June 2007
 DISCLAIMER

“The views expressed in this report are strictly those of the authors and do not necessarily reflect those of
the United States Agency for International Development (USAID) and the Ateneo de Manila University”.
 Optimal Coordination, Taxation and Guarantees for
Private Participation in Public Investment

Renato E. Reside, Jr., PhD1

Abstract

A simple model is constructed which aids in determining optimal quantities, taxes and
guarantees for contracts for public investment projects undertaken with private sector
participation. Under symmetric information and coordination among responsible
agencies, a full government guarantee and an income tax holiday are optimal solutions;
they should be provided to investors. But under asymmetric information conditions, this
leads to adverse selection. The optimal coordinated solution in this case is a set of
contracts that grants partial guarantees and non-zero income taxes for all private
investors. The answer challenges conventional wisdom in investment promotion that
guarantees and tax holidays are necessary to attract investment.

1
University of the Philippines School of Economics, Diliman, Quezon City, Philippines. Correspondence
should be sent to email: renato_ reside@hotmail.com
“Should we give a government guarantee?”

“Should we give an income tax holiday?”

- Frequent questions raised by the Philippines’ Department of Finance

I. Introduction

The introduction of private sector participation (PSP) in infrastructure

development in the Philippines in the 1990’s has led to palpable improvements in the
quality and quantity of services in transportation and utilities. Initially developed to
address crippling electricity shortages in the 1980’s, templates for PSP in infrastructure
involve a variety of modalities for private investor-contractors to build and/or own
infrastructure facilities. In the Philippines, these templates are better known by the
acronym, “BOT” (Build-Operate-Transfer, referring to the most common template:
modalities for eventual transfer of ownership to government of an infrastructure facility
built by private investor-contractors, after a period in which they have earned a sufficient
return from operating it). Since their inception, BOT projects have posed great challenges
to the Philippine government because they are complex contractual relationships with
investors, contractors and creditors with often conflicting interests and greater familiarity
with project finance.2

Dealing with firms better-versed with structured finance for infrastructure projects
has made it imperative for the Philippine government to enhance its capacity: 1) to screen
applicant-investors; and 2) to structure BOT contracts that offer internationally
competitive terms, investor protection and rewards, and with efficient incentives for
optimal effort to be exerted. Unfortunately, this capacity has not developed at a pace that
adequately addresses increasingly complex requirements of private project stakeholders.

Shortcomings in screening and contracting private firms are not benign. Standard
government guarantee packages in BOT contracts are subject to the same moral hazard
and adverse selection problems inherent in any insurance transaction. Improper
structuring of guarantees and incentives within contracts may lead to adverse selection,
attracting a risky and/or inefficient mix of investors. Insufficient monitoring capacity and
lax reporting requirements give rise to more unwanted incentive and moral hazard
problems. This study points out that additional problems arise from the lack of
coordination between agency that implements the project, the agency that provides
fiscal incentives and the agency that provides government guarantees. These problems
manifest in foregone opportunities to write contracts for PSP in infrastructure that
provide stronger incentives for transparency, accountability and efficiency. This is
highly costly. This also underscores the dangers of reducing the level of checks and
balances in the approval and implementation of these projects.

2
The Philippines was one of the original proponents of the BOT format because it greatly facilitates
investment. Such advantages have since convinced many other developing countries to adopt it.

1
 Unfortunately, the government has not found adequate mechanisms for properly
“incentivizing” contracts. The absence of adequate incentive mechanisms in contracts
aggravates the country’s fiscal problems. Specifically, contingent liabilities, arising from
generous provision of government guarantees in BOT contracts have been blamed for
severely disrupting government cash flows. The unintended selection of risky investor-
contractors by the government in the 1990’s has resulted in a rash of large and sudden
claims on the treasury by guaranteed private investors. Moreover, where fiscal incentives
(income tax holidays and other tax and duty exemptions) given to projects which would
have been undertaken anyway even without them, these incentives have added to the
fiscal costs of PSP projects. Thus, another important policy question arises: should a
potentially fiscally costly income tax holiday be granted to investors?

The fiscal squeeze from BOT projects has forced government to suspend private
investment guarantees for private investment and/or enforce tighter screening of foreign
and domestic infrastructure investors, but this has sharply curtailed investment in
essential infrastructure (see Figure ). However, since the Philippines is way behind its
neighbors in terms of developing its infrastructure, solutions for the risk-sharing and
contracting problem need to be found. Thus, one of the main objectives of this paper is
to build a simple benchmark model, broadly applicable to the country, usable for policy
and capable of being extended.

Given the issues discussed above, one of the central issues regarding government
guarantee and incentives policy is how to address adverse selection and moral hazard.
Adverse selection (the tendency of the government to select risky BOT projects) arises
because government may not be sufficiently informed about the capabilities of the private
investors it contracts, resulting in sub-optimal risk-sharing. There are several
explanations for this:

a) many of the investors are foreign; so government has encountered them for the first
time;

b) government lacks familiarity with the market and industry altogether, so it may have
insufficient basis for distinguishing efficient from inefficient contractors; and

c) government does not charge risk-adjusted guarantee premiums.

For most infrastructure projects with private sector participation, the Philippines’ BOT
Law mandates the government to provide guarantees and fiscal incentives under the
country’s Omnibus Investments Code (this includes income tax holidays for up to 8
years). This study argues that the rigidity of the legal framework for applying
government guarantees and fiscal incentives to infrastructure project with private
sector participation (PSP) makes it difficult for the government to gain any leverage
(whatsoever) in the contracting process and opens the window for adverse selection
problems. The rigidity results in standard BOT contracts (government guarantee
plus four-year income tax holiday) applying to most (if not all) projects, regardless
of project circumstances. The standardized application of guarantees and fiscal

2
incentives in BOT projects is not capable of distinguishing between risky and less
risky contracts and contractors. This leads to the selection of (fiscally) risky project
proponents and risky project contracts, raising the fiscal cost of BOT projects,
contributing to the pressure to reduce supply. This study points out that optimizing
contracts for PSP, including guarantees and fiscal incentives, requires the government
agencies providing them (the Department of Finance, DoF, and the Board of Investments,
BOI) to coordinate with one another in considering and offering alternative contract
configurations (more on this later) – of course, written by the government (!), so as to
improve the state’s bargaining position and power. Recent moves by top officials of the
country to reduce the role played by NEDA’s (National Economic Development
Authority) Investment Coordinating Council (ICC) in the review and approval of
PSP infrastructure projects further erodes the government’s ability to write
welfare-optimizing contracts and potentially raises fiscal risks.

In this study, I seek to find optimizing roles for production quantities, guarantees
and taxes (and perhaps, even prices) to play in “incentivizing” project contracts. The
previous paragraphs suggest that policies which balance interests among government and
investors, should help create a better environment for infrastructure investment overall.
For this to succeed, the approach should address the need for government to distinguish
low from high risk investors, to allow them to exploit this information in order to
minimize their risk and contingent liabilities, while at the same time giving proper
incentives to low risk investors. Due to the demands for sustainable credit and sources of
finance, BOT contract negotiations usually center around or emphasize the level of
guarantee coverage by government. This invariably leads to government’s assumption of
an inordinate amount of risk. This study suggests that a better strategy for government
during the contracting process is to coordinate its actions, utilize all variables within its
control (quantities, guarantees, fiscal incentives), in order to create more incentive-
compatible environments in projects.

The study may help agencies seek answers to the following questions:

a) When are full guarantees optimal? When are partial guarantees necessary?
b) When is an income tax holiday optimal (if at all)?
c) Can the Philippines do better than its current legal and regulatory framework for
infrastructure project approvals? What are the features of an optimizing legal and
regulatory regime? Are there benefits to greater coordination among agencies? Are
there costs to less coordination?

Throughout the study, the following timing of events in the contracting process
will be assumed to hold

3
Figure 1

t=0 t=1 t=2 t=4

Agent/ Principal/ Agent/ The contract is (is

investor Government Investor accepts not) executed; the
discovers designs then or refuses the agent/investor
his type T offers a contract contract supplies (does
not supply) effort

Contracts are offered in the interim stage (time t = 1), so if there is asymmetric
information between the contracting parties, it is already present when the government
makes the offer. To further simplify this benchmark model, we have also assumed that
the process above is a static, one-shot game. Extension into a dynamic setting is left for
further study. Note that the events in Figure 1 preclude cases where applicant investors
bid for project contracts, such as in the case of concessions. In the Philippines, therefore,
the timing of events more closely resembles the process of awarding contracts in projects
solicited by government.3

II. Adverse Selection

a) A Baseline Symmetric Information Model: Infrastructure projects that the

private contractors can do better than the government

Consider the case of a risk-neutral government that offers contracts to risk-averse

investors for the production and delivery of infrastructure services to a risk-averse public
constituency. The contract specifies a project to be undertaken by the investor with
payment based on the delivery of an observable output. The project to be undertaken,
such as a toll road, a power generating plant, or an urban rail transit facility, creates
important infrastructure services for the public, which pays for these services at tariffs
regulated by the government. The project is inherently risky, since the value of the
investor’s revenue and/or expenditure flows are uncertain. However, the investor has a
minimum threshold for income flows received, so that he will not participate in the
project unless the flows he receives are higher than this threshold.

Assume that the project to be undertaken by the investor is so vital that the
government offers to insure, or guarantee all or part of the flow of income received by the
investor. If the government chooses to guarantee the entire flow of income, it offers the
investor a full guarantee. If the guarantee is only on a portion of the flow, it is a partial
guarantee.

3
The majority of BOT projects in the Philippines are solicited by government. But this has not prevented
unsolicited projects from getting similar forms of support.

4
 There are two states of nature, good and bad. Assume that these states are
verifiable and project outcomes under each state are observable to the government. In the
bad state, all or portion L of the value of the investor’s flow of income is lost. But if a
government guarantee is built into the contract, the investor can call on this guarantee and
claim the amount q from the government, the value of the guarantee coverage. In the
good state, the flow is not lost and the guarantee is never called. π is the probability that
the bad state will occur. Of course, 0 < π < 1. For a full guarantee, q = L : the entire loss
is covered. For a partial guarantee, q < L : only a portion is covered. Note that q, the
guarantee coverage of the investor, is a contingent liability of the government since it is
to be paid by the Treasury to the investor contingent upon the realization of the state
which the contract states can trigger a claim (and such a remedial action is so stipulated).
It is assumed that q ≤ L : the government can never compensate an investor for
more than its loss.

For simplicity, I initially assume that the government levies a lump sum tax τ on
the contractor in the good state and no tax in the bad state (the assumption of a lump sum
tax can be relaxed later). It is assumed that τ ≥ 0. If the optimal tax equals zero as a
result of the optimization that follows, the implication is that an income tax holiday
should be granted to the project proponent.

It is assumed that the government does not charge guarantee premiums (as is done
in insurance). This is a crucial and unfortunate omission, because risk-adjusted premiums
are a mechanism often used to discriminate between risky individuals or firms. With no
premiums charged for guarantees, all the government can do to discriminate between
risky individuals or contractor-firms is to offer them various contracts that impose
different combinations of guarantee coverage, tax rates and outputs in order to maximize
public welfare.

The outcome of the project for the investor can be described by a binomial
distribution (only two outcomes are possible):

Table 1: Probability distribution of investor outcomes

State of Nature Probability Outcome
Good (No Loss) 1-π B [pQ – cQ – F – τ]
Bad (Loss) π B [pQ – cQ – F – L + q]

The investor’s utility is described by the function B [ . ]. Because the investor is assumed
to be risk averse, his utility function is concave: B ’ [ . ] > 0 and B ’’ [ . ] < 0. The
argument within the utility function is the flow of income (net revenues, wealth or
endowment) received by the investor across states. The quantity of goods produced by
the investor is Q, sold to the public at the per-unit tariff p. 4 Revenues equal pQ. In
producing each unit of Q, the investor incurs a constant variable cost c. A fixed cost of
production, F, is also incurred. Think of the variables c and F in terms of variable and
4
In practice, most infrastructure projects with PSP do involve the production of goods and services at
regulated tariffs. It is very common for road tolls, electricity and water tariffs to be sold at regulated tariff
rates.

5
fixed costs commonly incurred in the process of constructing and operating infrastructure
facilities (see Table 2).

Table 2: Fixed and variable costs of infrastructure construction and operation

Infrastructure Facility Fixed costs, F Variable costs, c
Toll roads Cost of toll booths Maintenance costs, cost of
asphalt or concrete, labor
Power plants Cost of constructing the Maintenance costs, cost of
power plant fuel inputs and wires, labor
Water supply Cost of constructing Maintenance costs, cost of
pumping stations water treatment and/or
sewerage, pumping, pipes,
electricity consumed in
pumping water, labor
Urban rail (MRT-III) Cost of rolling stock, cost of Maintenance costs, cost of
constructing train stations fuel and/or electricity, labor
and laying train tracks

It is assumed that both parties make reasonable estimates of the project’s return
before the contract is signed, so that both c and F are planned costs. In the good state, the
project proceeds with all costs being realized according to plan. In the bad state, the
project loses the flow of income L due to unexpected increases in costs, due to delays or
cost overruns. In the bad state, it is assumed that the investor is not taxed. Instead, the
investor gets guaranteed compensation for losing the value L. Calls or claims on
contractual guarantees are assumed to be triggered by shortfalls of cash flow or
income to the investor. This in turn triggers outflow q from the national treasury to
the investor. By restricting the loss L to a flow, the scope of work is restricted to the case
where a flow of revenues or costs (or net revenues) accruing to the investor is at risk of
being lost.

The Philippine experience with BOT projects suggests in fact that the loss of
flows is most closely associated with delays, cost overruns, and other situations where
unexpected reductions in revenues occur, leading to insufficient cash flow for amortizing
a given stock of debt falling due. The loss of flows could also be broadly consistent with
any situation where the government assumes the investor’s (contractor’s) payment for
liabilities to sub-contractors (investor income or cash flow shortfall leads to default -- the
trigger event -- which leads to outflows from the treasury). 5 The information in Table 2
suggests that cost overruns are most often associated with risky fixed costs (F).

The outcome for the public can also be described by a binomial distribution:

5
This is in fact, similar to what occurred in 2000 when the Philippines’ MRT-3 project called on a
government guarantee. Due to various cost overruns, the MRT-3 consortium of private investors required
additional cash infusions to finance payments to their contractors (Reside, 2000 and Reside, 2001).

6
Table 3: Probability distribution of public outcomes
State of Nature Probability Outcome
Good (No Loss) 1-π U(Q) – pQ + τ
Bad (Loss) π U(Q) – pQ – q

Since the public is risk averse, its (aggregate) utility function is concave: U ’ [ . ] > 0 and
U ’’ [ . ] < 0. It is assumed that in the bad state, the government ultimately passes onto the
public the cost of claims made by investors, q through taxation. Thus, the public’s utility
is reduced by q in the bad state. We also assume for simplicity that all arguments outside
the function U’( ) are expressed in (or converted into) utils of social welfare. That is, pQ
in the good state and pQ - q in the bad state are multiplied by the variable ν, where ν is
the social cost of public funds and ν = 1. In the good state, however, the public gains τ in
the value of tax revenues paid by the contractor.

The contract between government and investor includes the terms of the
guarantee, tax rate and tasks to be carried out by the investor. It contains three
parameters:

a) Q, the quantity of the infrastructure good or service produced and purchased by the
government (let this proxy for the extent of the size of the infrastructure facility being
contracted;
b) τ, the lump-sum tax on corporate income levied in the good state; and
c) q, the value or extent of the guarantee coverage.

It is assumed that government, when offering the contract to the investor, chooses the
values of these parameters (this effectively imposes the assumption that the agencies
responsible for setting Q (the implementing agency), defining τ (the BOI in the present
case), and defining q (the DoF), act in coordination with one another – in drafting a
PSP contract that maximizes public welfare. In other words, the task at hand is to find
incentive-compatible contracts that maximize social welfare and mitigate agency
problems, fiscal risk, fiscal costs and contingent liabilities of government – but this
requires coordination between responsible agencies – and for agencies to yield to one
another as needed (for example, for the BOI to resist in granting an income tax holiday if
doing so maximizes public welfare).

The government must choose private infrastructure contractors from among a

pool of applicants. Suppose that in this pool, there are two types of investors of varying
competence, low risk (W) and higher risk (H). They have the same planned variable and
fixed costs for given project. The low risk investor is distinguished from the higher risk
investor by having a lower probability of failure in a project (or, a lower probability of
calling on a guarantee, perhaps because its management is more competent and
experience fewer delays and cost overruns). That is, πH > πW. We also assume, as a
baseline, that either the government has no funds to undertake the project on its
own OR πW < πH < πGOVERNMENT: government has a higher probability of failure
than the riskiest private investor and government observes this, so it solicits private
sector investors to undertake the project (we could also make the stronger

7
assumption that government does not have the capability of implementing the
project itself). For the remainder of the paper, we shall then refer to low risk investors as
“safer” investors and those with higher risks as “riskier” investors.

Under conditions of symmetric information, the government has complete

information about the characteristics of the investor relevant to the task at hand. The
government is able to observe investor type. If T is an index for investor type, the
government’s objective function is assumed to be the expected value of social welfare:

[ ( ) ] [ ( )
(1 − π T ) U QT − pQT + τ T + π T U QT − pQT − qT ] for T = H, W (1)

For simplicity, social welfare is separable in all of its arguments. Because the
government has the ability to observe the type of each investor that applies for a
contract and guarantee, it performs the maximization problem separately for each
investor. Also note that social welfare in the bad state is reduced by two factors: total
payments to the investor, pQ, as well as payment of claims on government guarantees, q.6

If the government were contracting with a type T investor (T = H, W) under

symmetric information, the government would maximize social welfare subject to two
inequality constraints:
[ ( ) ] [ ( )
(1 − π T ) U QT − pQT + τ T + π T U QT − pQT − qT ≥ R ] G
(2)

[ ] [
(1 − π T ) B pQT − cQT − F − τ T + π T B pQT − cQT − F − L + qT ≥ R ] (3)

The first constraint is government’s participation constraint. The expected utility

derived from purchasing the good or service must exceed the government’s
reservation level expected utility, RG. Think of RG as the level of expected social
welfare the public could achieve if government had undertaken the project itself (instead
of contracting the services of a private investor). Thus, the government will only entertain
offers from competent enough private contractors (albeit differing in degree of risk) who
will provide at least as high a level of expected social welfare as the government could
provide by itself. This is the rationale for privatizing the provision of infrastructure goods
and services in the first place. The second constraint is the investor’s participation
constraint. For the investor to agree to undertake the project, his expected utility
must exceed his reservation expected utility level ( R ). Think about the reservation
expected utility R, to be the utility associated with a level of income sufficient to satisfy
the investor’s required rate of return on the project.

6
The author also earlier analyzed cases with no corporate income taxes and with the per-unit price or tariff
as a choice variable of government. The price or tariff is normally regulated and it is usually determined
outside of BOT contracts, so the model presented here is probably a better approximation of reality.

8
 Given these assumptions, the government faces the following problem of
maximizing social welfare subject to the two constraints.

Max
(Q
T T T
,τ , q )
[ ( ) ]
(1 − π T ) U QT − pQT + τ T + π T U QT − pQT − qT [ ( ) ]
subject to

[ ( ) ] [ ( )
(1 − π T ) U QT − pQT + τ T + π T U QT − pQT − qT ≥ R ] G

[ ] [
(1 − π T ) B pQT − cQT − F − τ T + π T B pQT − cQT − F − L + qT ≥ R ] (4)

The problem can be solved using conventional Kuhn-Tucker conditions. The first
derivatives of the lagrangean with respect to QT, τT, and qT give rise to the following
Kuhn-Tucker conditions:

∂L
∂Q T
[ ( ) ]
= (1 + λ ) U ′ Q T − p

[
⎧⎪(1 − π T ) B ′ pQ T − cQ T − F − τ T
+ ( p − c )γ ⎨ T
] ⎫⎪ = 0 (5)
[
⎪⎩+ π B ′ pQ T − cQ T − F − L + q T ]⎬⎪⎭
∂L
∂τ T
{ [
= (1 + λ ) − γ B ′ pQ T − cQ T − F − τ T = 0 ]} (6)

∂L
∂q T
[
= −(1 + λ ) + γB ′ pQ T − cQ T − F − L + q T = 0 ] (7)

also, γ ≥ 0 and λ ≥ 0.

Equations (5), (6) and (7) give rise to the condition for efficient production:

( )
U ′ QT = c (8)

Marginal utility must equal the marginal cost of production.

Proposition 1a: In the case of symmetric information with lump-sum taxes, the optimal
guarantee q is a full guarantee on the loss, L. The optimal tax equals τ T = q T – L ≥ 0
and there is an immediate implication in the symmetric information case: Since it is
impossible to overcompensate the private firm for its loss (q T > L) and also impossible
to levy a negative tax (τT < 0), government provides an income tax holiday, it exempts

9
the firm from paying taxes (τ T = 0) and fully guarantees or compensate the contractor
for its loss (q T = L).

Proposition 1b: In the case of symmetric information, optimal output is set so that
marginal utility for the contractor equals unit price less the tax: U’(QT)= p = c. The
outcome is a competitive solution.

Proof of Proposition 1a: Equations (6) and (7) yield the condition that marginal utility in
the bad state for each type of investor equals the expected value of marginal utility across
states:

[
B′ pQ T − cQ T − F − L + q T = ]
{(1 − π T
[ ] [
) B′ pQ T − cQ T − F + τ T + π T B′ pQ T − cQ T − F − L + q T ]} (9)

This condition will only hold if τ T = q T – L (the optimal amount of the lump-sum tax
equals the guarantee compensation due to the contractor less the loss it incurs). If qT < L,
then τ T = qT - L < 0 (the optimal policy is to compensate the firm and not tax it, but this is
impossible since we assumed taxes cannot be negative). On the other hand, the
assumption that qT > L (over-guarantee) is impossible rules out taxes greater than zero.
On impulse, one could question the result – why give a tax holiday or a full guarantee to
both types? Recall first that this is a symmetric information result. Second, recall that the
government’s probability of incurring a loss is higher than any private contractor’s. So if
the project is so important, then the government must agree to give generous contractual
terms. Q.E.D.

Proof of Proposition 1b: Because B’( ) > 0 by assumption and the fact that 0 < πT < 1,
then from equation (6), we know that the term [(1- πT) B’(pQ T – cQ T – F) + πT B’(pQ T –
cQ T – F – L + qT)] > 0. Since λ ≥ 0 and – (1 + λ) < 0, it follows that γ > 0. Now, from (5),
(1 + λ) > 0, γ > 0 and [(1- πT) B’(pQ T – cQ T – F) + πT B’(pQ T – cQ T – F – L - qT)] > 0.
Thus, the only way for the left hand side of equation (5) to be equal to zero is for τT = L -
qT = 0 (so that L = qT and p = marginal cost (= c) and U’(QT) = p - τT = c - τT = p = c.
The optimal contract under symmetric information fully compensates the investor
for its loss and levies on the investor a positive tax or an income tax holiday. Q.E.D.

Note that in the case of symmetric information, the optimal action for government
is to set output at a level that achieves efficiency in production. This it does by setting
output at a level where marginal utility equals unit price (equals marginal cost).

What are the characteristics of the set of optimal contracts {(QW, τ W, qW), (QH, τ H
H
, q )} under symmetric information? Since marginal cost c is assumed to be the same for
all investors, it turns out that Q = QW = QH: optimal quantity is the same across investor
types. Because we assume that there is only one possible value of the loss L, we also
have qW = qH = q = L. Each investor type gets the same type of guarantee (full guarantee
to the extent of L) and each investor type gets an income tax holiday, τ T = 0.

10
 Because QT, τ T and qT are the same for all T, the same contract is offered to (and
chosen by) both low and riskier investors. The solution is a pooling equilibrium. But this
creates incentive problems when asymmetric information prevails. Although the safer
investors have a lower probability of failure, or calling and claiming on the government
guarantee, the government offers identical contracts to both investor types. Because
guarantees and lump-sum taxes for both low and riskier investors are the same, riskier
investors have a greater incentive to participate in the project than the safer ones do. Safer
investors, aware of their own characteristics, know they are better than riskier ones, but
they cannot obtain a better contract from the government. Because the symmetric
information contract treats high and safer investors exactly the same way, it follows that
when the government is unable to observe investor types, the pooling equilibrium
contract is sub-optimal due to adverse selection. This study argues that a major reason
for fiscal losses in infrastructure contracts with PSP is that the government provides
a standard contract (as mandated by the existing BOT legal framework), is not able
to distinguish safer (of filing a claim on government) from riskier contractors and is
unable to offer contract variations that will properly incentivize projects in light of
asymmetric information. The present framework discourages safer contractors, and
encourages (and rewards) riskier ones.

One important difference between the model just discussed and the real
world is the assumption that the government acts in a coordinated fashion when
choosing the level of output, QW, the tax τ W, and the level of government exposure, qW.
In this model, the coordination of government agencies is manifested in the
assumption that the government chooses QW, τ W and qW simultaneously: there are
three partial derivatives with respect to the choice variables, equations 5 – 7, and the
results of setting these partial derivatives equal to zero (for optimization) are used to
arrive at the equilibrium conditions (8) and (9).

In reality, however, the decisions on QW, τ W and qW are made most of the time
independently by separate institutions (in fact, most projects qualify for income tax
holidays because the BOT law specifies that projects can qualify for EO 226 incentives,
which are approved by the Board of Investments (BoI)).On the other hand, government
guarantees (q) are provided by the Department of Finance (DoF). Meanwhile, the
National Economic Development Authority (NEDA) oversees project review and
approval (this includes the setting of Q). Government agencies working on their own
(without coordination), take partial derivatives using ONLY with respect to the
variable under its control. Since it will not be possible to solve for optimal values of
QW, τ W and qW under these conditions, it follows that it will be impossible to
optimize social welfare when government agencies are acting on their own.

The optimization process described above and in the rest of this study is
therefore meant to demonstrate how greater coordination among agencies can
strengthen the government’s hand and make actual solutions and contracts closer to
optimal. It is also meant to demonstrate why moves to decentralize approvals of
infrastructure projects even further are potentially costly from a public welfare
(and fiscal risk) standpoint.

11
 b) A Baseline Asymmetric Information Model

With asymmetric information, again assume there are two types of investors: safer
and riskier. The latter have a higher probability πH of incurring a given loss L than the
former (πW < πH). Suppose the proportion of safer investors among all investors is x,
where 0 < x < 1. The government can observe the proportion x and the loss probabilities
(πH and πW), but cannot completely observe the exact type of each investor/applicant that
approaches it and exhibits interest in undertaking a given project. Low and riskier
investors differ only in their loss probabilities (both investor types have the same c and
F). Assume that the functional form of the utility function is the same across investor
types. The government’s problem is to maximize social welfare:

Max (1 − x ){(1 − π W )[U (QW ) − pQW + τ W ] + π W [U (QW ) − pQW − qW ]}+

{(Q W
,τ W W
,q H
), (Q ,τ H
,q H
)}
{ [ ( ) ] [ ( )
x (1 − π H ) U Q H − pQ H + τ H + π H U Q H − pQ H − q H ]} (10)

subject to

[ ( )
(1 − π W ) U QW − pQW + τ W + π W U QW − pQW − qW ≥ R ] [ ( ) ] G
(11)

[ ( )
(1 − π H ) U Q H − pQ H + τ H + π H U Q H − pQ H − q H ≥ R ] [ ( ) ] G
(12)

[
(1 − π W ) B pQW − cQW − F − τ W + π W B pQW − cQW − F − L + qW ≥ R ] [ ] (13)

[ ]
(1 − π H ) B pQ H − cQ H − F − τ H + π H B pQ H − cQ H − F − L + q H ≥ R [ ] (14)

[
(1 − π W ) B pQW − cQW − F + τ W + π W B pQW − cQW − F − L + qW ≥ ] [ ]
[ ]
(1 − π W ) B pQ H − cQ H − F + τ H + π W B pQ H − cQ H − F − L + q H [ ] (15)

[ ]
(1 − π H ) B p H Q H − cQ H − F − τ H + π H B p H Q H − cQ H − F − L + q H ≥ [ ]
[
(1 − π ) B p Q − cQ − F − τ
H W W W W
] + π B[ p
H W
Q − cQ − F − L + qW
W W
] (16)

The first two constraints are the participation constraints of government. The level of
expected social welfare under each investor type must exceed RG (otherwise, the
government will do the project, but it has the highest probability of failure – and this
yields worse results for the public). Constraints (13) and (14) are the participation
constraints of each investor. Each investor should receive an expected utility greater than
or equal to the reservation utility (assumed to be the same). The (15) and (16) are the self-
selection constraints. The safer investor will never select the contract intended for the
riskier investor because the expected utility from the contract intended for him will be
greater than or equal to the expected utility from the contract intended for the riskier

12
 investor. Similarly, the riskier investor will never select the contract intended for the safer
investor because the expected utility from the contract intended for him will be greater
than or equal to the expected utility from the contract intended for the safer investor.

Note that the asymmetric information approach to the government’s optimization

problem is very different from the symmetric information approach. Instead of running
separate maximization problems for each investor type, the government performs the
maximization problem only once for both investor types. The objective function includes
the expected utilities (for consumer surplus) for both investor types.

It turns out that only a subset of combinations of Q, τ, and q will ensure that each
investor type has the incentive to participate in the project. That is, the incentive-
compatibility, or self-selection constraints for both types (15) and (16) will only hold for
a subset of combinations of Q, τ, and q. Also, participation constraints for both types will
not be binding. For some range of values of Q, τ, and q, one or both types of investors
will not participate in the project. The complete Lagrangean for the asymmetric
information case is in an Appendix. The following is a summary of the first-order
conditions with respect to the contract variables QH, τ H, qH, QW, τ W, and qW

∂L
∂Q W
{[ ( ) ]}
= (1 − x + λ ) U ′ QW − p + ( p − c ){(α + θ )A − ηD} = 0 (17)

∂L
= - (1 − x + λ ) + {(α + θ )A − ηD} = 0 (18)
∂τ W

∂L
∂q W
{ (
= - (1 − x + λ ) + (α + θ )C − η π H π W C = 0)} (19)

∂L
∂Q H
{[ ( ) ]}
= (x + δ ) U ′ Q H − p + ( p − c ){(γ + η )J − αK } = 0 (20)

∂L
= - ( x + δ ) + {(γ + η )J − αK } = 0 (21)
∂τ H

∂L
∂q H
{ (
= - ( x + δ ) + (γ + η )N − α π W π H N = 0 ) } (22)

{ [ ] [
A = (1 − π W ) B′ pQW − cQW − F − τ W + π W B′ pQW − cQW − F − L + qW ]} (23)

{ [ ] [
D = (1 − π H ) B′ pQW − cQW − F − τ W + π H B′ pQW − cQW − F − L + qW ]} (24)

13
 [
C = B′ pQ W − cQW − F − L + qW ] (25)

{ [ ] [
J = (1 − π H ) B′ pQ H − cQ H − F − τ H + π H B′ pQ H − cQ H − F − L + q H ]} (26)

{ [ ] [
K = (1 − π W ) B′ pQ H − cQ H − F − τ H + π W B′ pQ H − cQ H − F − L + q H ]} (27)

[
N = B′ pQ H − cQ H − F − L + q H ] (28)

We can now proceed to determine the optimal values for QW, τ W, qW, QH, τ H, and
qH. The general plan for the succeeding parts of the paper are as follows:

1) Show that optimal Q is identical for both agent types;

2) Show that since QW = QH , the self-selection constraints imply that qH < qW < L and 0
< τH < τ W in equilibrium. The optimal solutions with asymmetric information are to
partially guarantee firms and to levy a nonzero lump-sum tax (no income tax
holiday).

a. Optimal Production

Proposition 2: The contracts offered by government to (and accepted by) both low and
riskier investors specify the same quantity of the good or service to be produced.

Proof: Combine (17) and (18) and we derive the condition for efficiency in production for
the safer agent:

( )
U ′ QW = c (29)

Combine (20) and (21) and we derive the condition for efficiency in production for the
riskier agent:

( )
U ′ QH = c (30)

Thus, QW = QH = Q. Q.E.D.

b. Optimal Guarantee Coverage and Taxation of Infrastructure Services

Proposition 3a: The self-selection constraints will force the government to

discriminate in equilibrium between low and riskier investors by offering differing

14
guarantee coverage. It less than fully guarantees the investors: qH < qW < L. The
riskier investor gets lower guarantee coverage.

Proposition 3b: The safer investor will be levied a higher and positive tax relative to the
riskier investor and the government will levy a positive (non-zero) lump-sum income
tax on the riskier investor: 0 < τH < τW.

Proof: Since the self-selection constraints must be satisfied in equilibrium, it follows that
equilibrium will also be characterized by the outcomes in Proposition 3a and 3b. We can
show that Proposition 3a and 3b hold by taking (15) and (16):

[ ] [
(1 − π W ) B′ pQW − cQW − F − τ W + π W B′ pQW − cQW − F − L + qW ≥ ]
[ ] [
(1 − π W ) B′ pQ H − cQ H − F − τ H + π W B′ pQ H − cQ H − F − L + q H ] (15)

[ ] [ ]
(1 − π H ) B′ pQ H − cQ H − F − τ H + π H B′ pQ H − cQ H − F − L + q H ≥
[ ] [
(1 − π H ) B′ pQW − cQW − F − τ W + π H B′ pQW − cQW − F − L + qW ] (16)

We can factor and rearrange the terms in (15) to get:

⎧1 − π W ⎫
{B′[pQ W
] [ ]}
− cQW − F − τ W − B′ pQ H − cQ H − F − τ H ⎨ W ⎬ ≥
⎩ π ⎭
{ [ H H H
] W
[
B′ pQ − cQ − F − L + q − B′ pQ − cQ − F − L + qW
W
]} (31)

We can factor and rearrange the terms in (16) to get:

⎧1 − π H ⎫
{B′[pQ ] [ ]}
− cQ H − F − τ H − B′ pQ H − cQ H − F − τ H ⎨ H ⎬ ≥
H

⎩ π ⎭
{ [ W W W
] H
[
B′ pQ − cQ − F − L + q − B′ pQ − cQ − F − L + q H
H
]} (32)

Noting that the RHS of (31) is merely the negative of the RHS of (32), we can combine
(31) and (32) and impose QW = QH = Q at the optimum to get:

15
{B′[pQ − cQ − F − τ ]− B′[pQ − cQ − F − τ ]}⎧⎨1 −ππ ⎫
H

⎬≥
H W
H
⎩ ⎭

{B′[pQ − cQ − F − L + q ]− B′[pQ − cQ − F − L + q ]}≥

W H

⎧1 − π W ⎫
{B′[pQ − cQ − F − τ ]− B′[pQ − cQ − F − τ ]}
H W
⎨ W ⎬ (33)
⎩ π ⎭

⎧1 − π H ⎫ ⎧1 − π W ⎫
Now, since ⎨ H ⎬ < ⎨ W ⎬ , the above condition can only hold if τ W > τ H and
⎩ π ⎭ ⎩ π ⎭
q >q .
W H

Next, it can be shown that that neither contract will offer an income tax
holiday NOR a full guarantee. Equations (18) and (19) in the first order conditions
imply that

⎧ ⎛π H ⎞ ⎫
{(α + θ )A − ηD} = 0 = ⎨(α + θ )C − η ⎜⎜ W ⎟⎟C ⎬ (18)
⎩ ⎝π ⎠ ⎭

⎧ ⎞ ⎫
(α + θ )( A − C ) = 0 = η ⎨D − ⎜⎜ π W ⎟⎟C ⎬
⎛ H
(19)
⎩ ⎝π ⎠ ⎭

⎧ ⎛π H ⎞ ⎫
⎨D − ⎜ ⎟C ⎬
(α + θ ) = ⎩ ⎜⎝ π W ⎟⎠ ⎭ ≥ 0 (34)
η {A − C}

⎧ ⎛π H ⎞ ⎫
For the above to be true, it must be true that ⎨ D − ⎜⎜ W ⎟⎟C ⎬ ≥ 0 and A ≥ C . This can be
⎩ ⎝π ⎠ ⎭
⎧ ⎛π H ⎞ ⎫ ⎛π H ⎞
demonstrated as follows. ⎨ D − ⎜⎜ W ⎟⎟C ⎬ ≥ 0 implies that D ≥ ⎜⎜ W ⎟⎟C . This can be
⎩ ⎝π ⎠ ⎭ ⎝π ⎠
simplified into:

16
 ⎛ ⎞
⎜ ⎟
⎜ 1− π H
> ⎜ H
[
⎟ B ′ pQ − cQ − F − L + q W ]
1 ⎟ ≥ B ′ pQ − cQ − F − τ W
[ ] > 0 (35)
⎜ ⎛⎜ π ⎞⎟ − π H ⎟
⎜⎜πW ⎟ ⎟
⎝⎝ ⎠ ⎠

which is greater than zero because by assumption, B’( ) > 0. The fact that the fraction is
less than one implies that:

−τ W > − L + qW or L − qW > τW (36)

that is, if the lump-sum tax is greater than zero (τW > 0), this means that qW < L : the
guarantee is a partial one (the case in which A ≥ C can be shown to yield exactly the
same result). That τW > 0 must be true follows from the fact that τW > τH . Thus, the
contract aimed at attracting the safer investor levies a positive tax and a partial guarantee,
but the positive tax levied on the safer investor is higher than the positive tax levied on
the riskier investor. On the other hand, since qH < qW < L, it follows by transitivity that
qH < L (the riskier investor has a lower partial guarantee – it gets compensated less when
it loses).

Since τ W > τ H, can it be that τ H = 0 (income tax holiday)? To show that this is
not possible, take equations (21) and (22) from the first order conditions:

⎧ ⎞ ⎫
{(γ + η )J − αK } = 0 = ⎨(γ + η )N − α ⎜⎜ π H ⎟⎟ N ⎬
⎛ W
(21)
⎩ ⎝π ⎠ ⎭

⎧ ⎞ ⎫
(γ + η ){J − N } = α ⎨ K − ⎜⎜ π H ⎟⎟ N ⎬
⎛ W
(22)
⎩ ⎝π ⎠ ⎭

⎧ ⎛πW ⎞ ⎫
⎨ K − ⎜ ⎟N ⎬
(γ + η ) = ⎩ ⎜⎝ π H ⎟⎠ ⎭ ≥ 0 (37)
α {J − N }

since
(γ + η ) ≥ 0 because γ , η ≥ 0 and α > 0 . This will only be the case if (ruling out J
α
= N):

⎛πW ⎞
1) Case 1: K ≥ ⎜⎜ H ⎟⎟ N and J > N; or
⎝π ⎠

17
 ⎛πW ⎞
2) Case 2: K ≤ ⎜⎜ H ⎟⎟ N and J < N.
⎝π ⎠

It can be shown that only case 1 will work and this implies a positive non-zero tax.

6444444444444 47J 4444444444444 8 6444447 N 44444 8

[ H H H H
] H H
[ H H H
]
(1 − π ) B′ pQ − cQ − F − τ + π B′ pQ − cQ − F − L + q > B′ pQ − cQ − F − L + q H
H
[ ]
can be simplified into

[ ] [
(1 − π H ) B′ pQ H − cQ H − F − τ H > (1 − π H ) B′ pQ H − cQ H − F − L + q H ]
For the last inequality to hold, it must be the case that:

− τ H > −L + q H

L − qH > τ H ≥ 0

Now, since τ W > τ H , it follows that τ W > 0: the safer investor is not offered an income
tax holiday. It remains to be shown that the riskier investor will not be offered an income
tax holiday at the optimum either. Note that since we have shown that L > q H and we
have assumed that τ H ≥ 0 , it is not possible that L − q H = τ H = 0 and nor will it be
possible that L − q H < τ H . This leaves as the only possibility τ H > 0 . Therefore, it
follows that τ W > τ H > 0. Neither investor is offered an income tax holiday.

A summary of the key results that hold in equilibrium are:

1) QW = QH = Q;
2) qH < qW < L (partial guarantees); and
3) 0 < τ H < τ W (no income tax holidays).

Under conditions of asymmetric information (applicant investors hold more

information about themselves than does the government), It is possible for the agencies of
government, acting in coordination with one another, to offer a contract that will
maximize public welfare without offering an income tax holiday and a full government
guarantee on a given investment project. This it can do by giving better guarantee terms
to safe investors while taxing them more than the riskier ones. In other words, one can do
this while still maximizing welfare by partially compensating the investors for their losses
while nevertheless levying on them a relatively generous non-zero tax. The intuition

18
behind the results is the fact that government needs to keep both types interested in
participating in the project. Otherwise, this important project will have no takers.

Note that the optimum under asymmetric information is strikingly different from
the optimum under symmetric information. In the latter, the optimum is characterized by
a full guarantee and an income tax holiday for all investors. In the asymmetric
information case, the optimal solution is a partial guarantee and a positive, non-zero
income tax for all investors.

It is further possible to prove that there are a range of possible values for these
results to hold.

(THIS BEING REFINED FURTHER)

Proof:

Take the self-selection constraint ()

{B′[pQ − cQ − F − τ ] − B′[pQ − cQ − F − τ ]}⎧⎨1 −π π ⎫ H

⎬≥
H W
H
⎩ ⎭
{B′[pQ − cQ − F − L + q ] W
[
− B′ pQ − cQ − F − L + q H ]} (31)

This can be rewritten and interpreted as:

1 − π }{B′[ pQ − cQ − F − τ ] − B′[ pQ − cQ − F − τ ]}
{1 H H W
≥
4444444444244444444443
expected differential between marginal benefit of riskier and safer investors when outcome is favorable

{B4′[pQ
π1H4 − cQ − F − L + qW ] − B′[ pQ − cQ − F − L + q H ]}
444444444244444444444 3
expected differential between marginal benefit of riskier and safer investors when outcome is unfavorable

Note that the left hand side is comprised of the bracketed term denoting the differential
between the marginal private benefit of the riskier and the safer investor when the project
outcome is favorable (i.e., no adverse event occurs). This should be at least as large as the
right hand side: the differential between the marginal private benefit of the riskier and the
safer investor when the project outcome is not favorable (i.e., the adverse event occurs
and the investor loses L). Notice that for the inequality to hold (or for all investor types to
have an incentive to participate in the project):

1) the lump sum tax on the riskier investor should be sufficiently low;
2) the lump sum tax on the safer investor should be sufficiently high;
3) the value of the guarantee (insurance) policy for the safer investor should be
sufficiently low; and

19
 4) the value of the guarantee (insurance) policy for the riskier investor should be
sufficiently high.

The inequality suggests that the expected differential between the marginal private
benefits for high and safer investors in good times should be at least as large as the
expected differential between the marginal private benefits for low and the riskier ones
investors in bad times. This is intuitive: in order for all firms to be interested in
participating in the project, the expected marginal benefit of after-tax profits for riskier
investors should be sufficiently high (or its lump-sum tax τH sufficiently low) to
compensate it for the disadvantage of having a lower than expected marginal benefit of
profit in adverse times when it incurs a loss L (relative to the low risk firm). This also
ensures that investors self-select: that the riskier (safer) investor will not have an
incentive to accept the contract intended for the safer (riskier) investor.

IV. Conclusion

This paper constructs a simple model which aids in determining optimal

quantities, taxes and guarantees for contracts for public investment projects undertaken
with private sector participation. Under symmetric information and policy coordination
by government agencies, it can be shown that a full government guarantee and an income
tax holiday are optimal solutions and therefore, these should be provided to investors.
However, it can be shown that this leads to adverse selection under asymmetric
information. The optimal coordinated solution under asymmetric information is a set of
contracts that provides partial guarantees and non-zero income taxes for all private
investors.

A key feature of the model is that optimal solutions can be found only if
government agencies responsible for reviewing projects, providing fiscal incentives
and other means of support such as guarantees, coordinate their actions.
Unfortunately, the current practice of granting guarantees and the automatic granting
income tax holidays (as prescribed by the legal and regulatory framework for PSP in
infrastructure) does not require agencies to coordinate and obscures the potential gains
from greater policy coordination. The lack of coordination also makes it difficult for the
government to gain any leverage (whatsoever) in the contracting process and opens the
window for adverse selection problems. The standardized application of guarantees and
fiscal incentives in BOT projects is not capable of distinguishing between risky and less
risky contracts and contractors. This leads to the selection of (fiscally) risky project
proponents and risky project contracts, raising the fiscal cost of BOT projects,
contributing to the pressure to reduce supply. This study points out that optimizing
contracts for PSP, including guarantees and fiscal incentives, requires the government
agencies providing them (the Department of Finance, DoF, and the Board of Investments,
BOI) to coordinate with one another in considering and offering alternative contract
configurations (more on this later) – of course, written by the government (!), so as to
improve the state’s bargaining position and power. Recent moves to reduce the role
played by NEDA’s (National Economic Development Authority) Investment
Coordinating Council (ICC) in PSP infrastructure projects in project approvals further

20
erodes the government’s ability to write welfare-optimizing contracts and raises fiscal
risks.

21
Appendix

The complete Lagrangean is:

{ [ ( ) ] [ ( )
L = (1 − x ) (1 − π W ) U QW − pQW + τ W + π W U QW − pQW − qW ]}
{ [ ( ) ] [ ( )
+ x (1 − π H ) U Q H − pQ H + τ H + π H U Q H − pQ H − q H + ]}
λ {(1 − π W )[U (QW ) − pQW + τ W ] + π W [U (QW ) − pQW − qW ]}+

δ {(1 − π H )[U (Q H ) − pQ H + τ H ] + π H [U (Q H ) − pQ H − q H ]}+

θ {(1 − π W ) B[ pQW − cQW − F − τ W ] + π W B[ pQW − cQW − F − L + qW ]}+

γ {(1 − π H ) B[ pQ H − cQ H − F − τ H ] + π H B[ pQ H − cQ H − F − L + q H ]}+

[ ] [
⎧⎪(1 − π W ) B pQW − cQW − F − τ W + π W B pQW − cQW − F − L + qW −
α⎨
] ⎫⎪
⎬+
⎪⎩ [ ] [
(1 − π W ) B pQ H − cQ H − F − τ H + π W B pQ H − cQ H − F − L + q H ]⎪⎭

η⎨
[ ] [
⎧⎪(1 − π H ) B pQ H − cQ H − F − τ H + π H B pQ H − cQ H − F − L + q H − ] ⎫⎪
(A.1)
⎬
⎪⎩ [ ] [
(1 − π H ) B pQW − cQW − F − τ W + π H B pQW − cQW − F − L + qW ]⎪⎭

22