The World Bank
November
PREMnotes
1998
number 10
Economic Policy
Private infrastructure, public risk
Whether private infrastructure will realize its expected benefits depends on how
governments allocate risks. Many governments have assumed risks that private
investors should bear, reducing incentives for efficiency and exposing govern-
ments to contingent liabilities. Governments can, however, reduce project risks
and improve how guarantees are measured and budgeted.
Facing budget constraints and recognizing aire if traffic was less than 90 percent of the
their failure to provide infrastructure services specified level. The Colombian government
efficiently, governments in many develop- also provided a minimum revenue guaran-
ing countries are increasingly relying on the tee when it awarded a build-operate-trans-
private sector to provide infrastructure ser- fer concession for a new runway at Bogotá’s
vices. Private cross-border finance for infra- El Dorado airport.
structure projects in developing countries Such guarantees can undermine the ben-
grew from $0.1 billion in 1988 to $20.3 bil- efits of privatization. First, they can blunt pri-
lion in 1996. More than one hundred gov- vate investors’ incentives to choose only good
ernments have involved the private sector in projects and to run them efficiently. If the
areas such as power, gas, telecommunications, government bears the risk of project fail-
water, roads, railroads, ports, and airports. ure—for example, by guaranteeing
But infrastructure projects are often risky: demand—the private investor has little incen-
they are typically long term and subject to tive to choose financially sound projects.
political pressures to keep service prices low. Second, guarantees may impose excessive
To cover these risks, investors often ask for— costs on the host country’s taxpayers or con-
and sometimes receive—government guar- sumers. Because government guarantees—
antees against risks such as changes in the a contingent public liability—rarely show up
political or regulatory climate, reneging on in government accounts or budgets, gov-
contracts by state enterprises, cost overruns, ernments are more willing to assume risks
low demand, or changes in exchange and that are better borne by investors. (For more
interest rates. (Here the term government details on contingent liabilities, see
guarantee refers to any undertaking by gov- PREMnote 9.) Governments may not even
ernment to assume risk.) For example, to know the extent of their risk exposure. At
attract private investment in power genera- worst, the issue of guarantees could lead to
tion, the Pakistan and Philippine govern- a fiscal crisis by encouraging excessive risk-
ments have committed to honor the payment taking ("heads the investor wins, tails the
obligations of public utilities to purchase government loses").
power at predetermined prices, regardless
of demand. In the recent El Cortijo-El Vino Reducing project risks
toll road project in Colombia, the govern- Governments are often forced to bear risk
ment agreed to reimburse the concession- because of shortcomings in current and
f ro m the developm ent economics vice p res iden cy a nd pover t y re d u ct i o n a nd e co no mi c ma na ge me nt n e t w o r k
expected policies. By improving policies, on? Is a government guarantee preferable
government can avoid granting guarantees to a budget subsidy?
while still being able to attract private infra-
structure investment. For example, stable Guidelines for allocating risk
macroeconomic policies reduce the likeli- Two factors determine whether an agent
hood of large changes in exchange and inter- should bear risk: the degree to which the
est rates, reducing pressures on government agent can influence or control the outcome
to provide exchange rate guarantees or at risk and the agent’s ability to bear risk.
discontinue currency convertibility or trans- Other things being equal, risk should be
ferability. Similarly, the regular disclosure allocated to agents who can best control the
of timely, reliable information on the econ- risky outcome and to agents who can bear
omy and government finances makes it eas- the risk at the lowest cost. But these two
ier for investors to forecast revenues. factors often push in different directions—
Firms are less likely to insist on guaran- the entity that has the most control over the
tees if a country has a good regulatory frame- risky outcome may not be in the best posi-
work, nonpolitical regulatory agencies, and tion to bear the risk. Other factors to con-
a strong and independent judiciary. Firms sider are whether the entity assuming a risk
investing in the United States, for exam- has an incentive to reduce it and what alter-
ple, do not need government guarantees natives there are to a government guaran-
against opportunistic government behavior tee. How can these principles guide the
because they are confident that the courts allocation of common types of risk?
will protect them in the event of expropri-
ation or regulatory takings. A firm operat- Political and regulatory risks
ing in a competitive environment or under Expropriation and currency inconvertibil-
tariff regulations enforced by an indepen- ity or nontransferability are under direct
dent regulator is less likely to insist on guar- government control, and there is good rea-
antees on tariffs. By the same token, son to encourage governments not to cre-
permitting international arbitration eases ate losses associated with these risks. Thus
investors’ fears that they will be mistreated it makes sense for government to assume
by local courts that are not independent. these risks. But regulatory risks—whereby
Few industrial countries find it necessary to government commits not to change the laws
provide government guarantees for infra- and regulations affecting an investment pro-
structure projects. In addition, some devel- ject, or to compensate if it does—are trick-
oping countries have been able to attract ier. While they are under government
private infrastructure investment without control, it is sometimes desirable for gov-
assuming risks. In Argentina, for example, ernment to change laws in ways that
restructuring and privatization of the power adversely affect investment projects. It may
industry have enabled the government to be beneficial to increase taxes to fund
attract private investment without assuming needed public investment, or to impose reg-
major risks or issuing guarantees. In Chile ulations that mitigate newly recognized envi-
private firms recently invested in telecom- ronmental problems. Regulatory risks are
munications, power, and gas without gov- best handled on a case-by-case basis following
ernment guarantees. the principles above.
In many cases, however, good policies
cannot be brought about overnight, and Quasi-commercial risks
private investment is unlikely to be forth- When an investor contracts with public sup-
coming unless government assumes certain pliers or purchasers that may renege on con-
risks or provides subsidies. How should gov- tractual commitments (as in the Pakistan
ernments decide whether to bear risks in a and Philippine power projects), the deci-
private infrastructure project? If they decide sion on whether the government should
to bear risks, which risks should they take assume that risk depends on the degree to
which it can influence the behavior of the follow stable macroeconomic policies. But
public entity. If the utility is fully several other factors suggest that investors
autonomous, there is little to be gained from should bear these risks.
a government guarantee. But if the utility First, a government guarantee may
is completely beholden to government deci- encourage an investor to expose itself to
sions, guarantees may be desirable. Even so, excessive exchange and interest rate risk.
it is preferable to increase the agency’s auton- Losses from currency depreciation, for
omy by privatizing it. example, could be blamed on the govern-
ment, which allowed the currency to fall, or
Demand and construction cost risks the firm, which left itself exposed by bor-
In road, bridge, or tunnel projects, gov- rowing in foreign currencies. Second,
ernments are often asked to bear demand exchange rate guarantees may change gov-
and construction cost risks. The rationale ernment behavior for the worse—for
for such support is weak. The concession- instance, discouraging governments from
aire usually has much more control than allowing their currencies to depreciate in
the government over construction costs. the wake of a terms of trade shock. Third,
And even though government policies can many governments and the taxpayers who
influence demand, assigning demand risk back them may already be exposed to the
to government reduces investors’ incentives risks associated with exchange and interest
to screen projects carefully. rate shocks. An adverse terms of trade shock,
Government can, however, reduce for example, might lead to both a depreci-
demand risk for some infrastructure pro- ation and a drop in local incomes, forcing
jects. Instead of auctioning the right to oper- the government to compensate investors
ate a service for a fixed period, as is typical, just when its tax base has shrunk. Finally,
the term of an operating concession can be the private sector may have more incentive
allowed to vary with demand. If demand is to manage the risk. For example, if foreign
higher than expected, the concession will be banks had borne the exchange rate risk in
shorter; if demand is lower, the concession Spain’s highway projects in the 1960s and
will be longer. The method has been used 1970s, they would have hedged the risk at
in the United Kingdom for bridges. An inge- a much lower cost than the $2.7 billion
nious variant of this method is to award the that it eventually cost Spanish taxpayers.
concession to the bidder seeking the lowest
present value of revenue, calculated with a Measuring and budgeting
discount rate specified by government. The guarantees
concession ends when the concessionaire’s Whichever risks government takes on, it needs
revenue reaches the present value it sought. to measure them accurately and incorporate
The concessionaire still bears some demand them in its accounts and budgets. Otherwise,
risk—if demand is too low, revenue may never it is difficult to make good decisions about
reach the target value—but it is much less. whether and which risks to assume, and could
lead to financial disaster.
Exchange and interest rate risks
Because many infrastructure investments Identifying and listing guarantees
are funded by floating-rate loans in for- The first and simplest step that governments
eign currency, profits are highly sensitive to can take to improve the monitoring and
changes in exchange and interest rates. management of risks is to compile and pub-
Should government accede to investor lish a consolidated list of their contingent
demands to assume exchange and interest liabilities and the maximum amounts they
rate risks? At first glance it appears that gov- stand to lose. The New Zealand government,
ernment should do so because it can bet- for example, presents this information in
ter control exchange and interest rates and its statement of contingent liabilities (see
doing so will give it a stronger incentive to http://www.treasury.govt.nz).
Calculating expected losses pare guarantees with cash subsidies. When
While helpful, the listing of guarantees and guarantees are not valued, a government
associated maximum possible losses does not may prefer to provide a guarantee instead
indicate what losses government should of a subsidy, even when the guarantee is
expect. For example, if a government guar- more costly than the subsidy, because the
antees a $10 million payment by a state enter- costs of the guarantee are hidden and may
prise and there is a 10 percent chance of the be borne by a future administration. When
enterprise defaulting (and a 90 percent guarantees are valued, decisions are more
chance of full payment), the expected cost likely to based on real rather than appar-
to the government of the guarantee is $1 mil- ent costs and benefits.
lion. In more realistic cases, however, it is
harder to calculate the expected cost. There Incorporating expected losses in accounts
may be more than two relevant possibilities, and budgets
and estimating the probabilities may be Once expected losses can be reliably cal-
extremely difficult. Still, it is sometimes pos- culated, they should be incorporated in gov-
sible to calculate expected losses using straight- ernment accounts and budgets. Most
forward techniques. Where a government government budgets and accounts are cash-
has issued a large number of similar guar- based. While it is possible and desirable to
antees for many years and has recorded infor- note guarantees and other noncash items
mation on defaults, the expected cost of the in cash-based budgets and accounts, fully
guarantees can be estimated actuarially in the incorporating them requires moving away
same way as, say, car insurance premiums. In from cash-based systems. With standard
other cases econometric modeling or out- accrual-based accounts and budgets, many
come simulations based on multiple scenar- noncash expenditures show up in the gov-
ios with different probabilities (Monte Carlo ernment’s budget and operating statement,
studies) may be feasible. and the government has no fiscal incen-
The techniques developed in the past tive to prefer noncash to cash expenditures.
25 years to value financial derivatives But although standard accrual accounting
(options, futures, swaps) can also be used discloses guarantees, it records them as
to value guarantees and contingent liabili- expenses only if the loss is considered prob-
ties. The value of a guarantee can be used able and can be quantified. From an eco-
to calculate the government’s expected loss. nomic point of view this distinction between
Extending a credit guarantee, for example, probable and improbable losses is not always
is equivalent to the government selling— useful; a 10 percent chance of losing $10
at zero cost—a put option to the lender. million is worse than a 90 percent chance
This option can be valued using option pric- of losing $1 million. More useful is an esti-
ing techniques. The valuation of some guar- mation of the present value of the expected
antees is difficult, however, requiring the loss arising from the contingent liability.
skills of financial specialists. Moreover, the An ideal system of accounting and bud-
feasibility of timely, reliable, and cost-effec- geting would record the expected present
tive valuation has not been widely tested. value of all contracts into which the gov-
But the possibilities are not merely theo- ernment has entered. Under such a system
retical—guarantees have already been val- government would have no fiscal incentive
ued using option pricing techniques in to issue guarantees instead of giving subsi-
Colombia and the United States. dies of equivalent value, because both would
Valuing the government’s guarantees and show up as expenditures affecting the deficit
other contingent liabilities—and not sim- and both would require appropriation by
ply noting maximum exposure—has impor- the legislature. While full present value
tant advantages. By calculating the expected accounting and budgeting is not feasible,
cost of government guarantees, the gov- the adoption of accrual accounting—and
ernment and observers can more easily com- the systematic recording of present values
where they are significant and quantifiable, This note was written by Mateen Thobani
even when losses are not probable—appears (Principal Economist, PREM Unit, Latin America
to be a crucial step toward the better man- and the Caribbean). Please contact Hana
agement of guarantees. Polackova (x30182) if you want to participate
in the Quality of Fiscal Adjustment Thematic
Further reading Group, which focuses on the analysis, manage-
Irwin, Timothy, Michael Klein, Guillermo E. ment, and fiscal implications of contingent gov-
Perry, and Mateen Thobani, eds. 1997. ernment liabilities.
Dealing with Public Risk in Private Infrastructure.
Washington, D.C.: World Bank.
This note series is intended to summarize good practice and key policy findings on
PREM-related topics. PREMnotes are distributed widely to Bank staff and are also
available on the PREM website (http:/ /prem). If you are interested in writing a
PREMnote, email your idea to Asieh Kehyari. For additional copies of this PREMnote
please contact the PREM Advisory Service at 87736.
Prepared for World Bank staff