. 13
( 18)


potential than an average revenue constraint to distort output incentives, producing
prices above the monopoly level in some markets. Cowan (1997b) compares the
dynamic case for three different types of regulatory constraints: average revenue,
Laspeyres base-weighted tariff basket constraint, and the average revenue lagged
regulation first studied by Sappington and Sibley (1992). He confirms that the average
revenue lagged constraint and average revenue may not only be inefficient but is likely to
reduce overall levels of welfare, while a Laspeyres index-based constraint can induce
efficient prices even when the firm is not myopic.

This section has made it clear that there are substantial difficulties in attempting to
design an optimal incentive contract for regulation while taking account of factors such
as consistency, uncertainty, and welfare. More generally, where a principal seeks to
create incentives for multiple tasks, there is a risk that the one may create adverse
incentives for the other. In particular, multiple objectives need to be carefully balanced.
Baron (1985) shows that if an economic and environmental regulator develop their
constraints independently, the firm will produce a higher environmental standard, less
output, charge higher prices, and make more profit that in the maximum efficiency case.
This is exactly the conflict which arose in the UK water industry as new price levels were
set from 2000. Baron shows that the optimum can be restored if the regulators can
internalize the trade-off between their different objectives before setting the constraints, a
result with wide-reaching implications for regulatory structure.
Uncertainty for the firm raises questions of regulatory commitment. If a firm has an
investment choice that will result in a return only at some future date, lack of commitment
by the regulator to enable recovery of the investment may result in the abandonment of a
cost reducing investment. The regulator's time inconsistency problem arises from the
initial desire for reduced costs, followed by the wish to minimize prices by not allowing
the costs of the investment itself, once it has been undertaken. The optimal reward ex
ante for the investment (to ensure it is undertaken), is no longer optimal ex post,
tempting the regulator to change the rules and strand the invested assets. Where it is
possible to write some form of binding contract upheld by both parties or some third party
(such as the introduction of judicial reviews in the United Kingdom) the time
inconsistency problem may be mitigated. One infamous example of stranded assets has
arisen as a result of regulatory reform in the UK gas industry. The incumbent monopolist
signed several take-or-pay contracts which committed it to paying for gas supplies even
if they were not used. As competition was introduced the incumbent lost large portions of
its market, especially for industrial consumers, and found itself committed to pay for gas
which it could no longer sell. In the United States the Federal Communications
Commission adjusted prices downwards for local telecom exchange carriers in 1995,
arguing that productivity had been substantially higher than forecast. Such a
manipulation of the contract could result in reduced incentives and increased uncertainty,
especially where firms' previous investment decisions had been based upon the original
productivity assumptions.
Clawing back profits ex post, whether explicitly or implicitly, has the result of dulling the
firm's incentives, to the extent that the regulated firm may prefer an earnings-sharing
plan within the regulatory contract. Such arrangements reduce incentives to lower costs
and innovate because the regulated firm will receive only a proportion of its earnings. In
the United States this has been recognized since 1993, when no incentive contracts
have incorporated an explicit earnings-sharing portion. In general some level of flexibility
in incentive regulation is optimal to allow exogenous shocks to be factored in and the
reasonable recovery of stranded assets, but frequent changes in regulatory rules create
uncertainty and a corresponding reduction in investment.[9]
One major difference between UK and US regulatory bodies is the degree of
transparency in decision-making. The UK regulator has more discretion and less need to
reveal the basis for decisions than its US counterpart. It is an interesting quirk of the UK
regulatory system that because of the lack of transparency, the price review is often
something of a bargaining process between the regulator and the firm. If agreement
cannot be reached, the Competition Commission acts as arbiter, and typically supports
the arguments of the regulator.

This issue is explored in Green and Waddams Price (1995).

The estimated budget for Oftel in 1999 was £12.6 million, which may still be thought of as a
regulatory bargain when considered against the turnover of a UK telecommunications market
estimated at £22 billion. Similarly Ofgas' budget of £12.9 million in 1998 may be compared to
Centrica and Transco's combined revenue of £11 billion. What is less clear is the cost of the
regulatory burden placed on the companies to meet the regulator's information demands, and
in pursuing rent-seeking behavior.

See Tardiff and Taylor (1996) for a good summary of some of the detailed issues.

See Shleifer (1985) for the formal theory behind yardstick competition.

See Crew and Kleindorfer (1999) for a discussion of stranded assets and a means of
allowing fair recovery.
4 Incentive regulation and competition
The Conservative Government of the 1980s and 1990s believed that regulation should
be only an interim stage in the drive towards competition. This desire was at first
tempered only by the practicalities of ensuring that competition did not open up a host of
other problems such as the wasteful duplication of sunk costs, and incorrect billing of
consumers. Criticism of the government's treatment of BT and British Gas, which were
privatized as vertically integrated monopolies, induced the government to separate
transmission from generation, and introduce some horizontal separation of generation
when electricity was privatized in 1990“1. The possibilities of competition in water were
much more limited owing to the absence of a national network and geographical
problems, and the companies were left as vertically integrated regional monopolies.
Although competition has proved successful in many of the industries such as telecoms,
gas, and electricity, there now exists a mix of old regulatory problems and new issues
created through the introduction of competition.

4.1 Industry structure

Development of competition has raised new questions about the boundaries of
regulation. Parts of the industries which are intrinsically naturally monopolistic will require
some form of long-term control, while technological advances continually expand the
areas where competition is viable. This raises questions about the optimal structure of
the industry, the development of competition, and terms of access to natural
monopolistic elements which are "essential facilities" for competition. There may be a
danger that policy-makers lose sight of the final goal during the quest for competition,
which is fundamentally a means to engender higher welfare. Especially in the rapidly
changing telecoms market, the assumption that competition brings higher welfare is
being questioned, and if competition should be allowed, how should it be structured?
Gilbert and Riordan (1995) compare a vertically integrated monopolist producing the
end-good with that of a vertically separated monopolist who is prevented from operating
in the end-market, and fringe competition bidding in the form of an auction with the right
to enter the end-market. An integrated monopolist avoids the cost information asymmetry
that results in an effect analogous to double marginalization, but this benefit may or may
not be greater than the benefit derived through the competitive auction to supply the
downstream market; the balance depends on the degree of complementarity between
the upstream and downstream products.

Industries which have remained vertically integrated (such as telecoms) have proved
more difficult both to regulate and for the introduction of competition, than those where
the natural monopoly element has been separated. However there may be a loss in
economies of scope through separation, and an overall reduction in welfare. Further loss
results from a potential problem of double marginalization, if both the upstream and the
downstream firms are imperfectly competitive or inadequately regulated. In this case the
regulator must ensure that the upstream firm (i.e. the natural monopoly that supplies an
input for the downstream firms) prices at or close to marginal cost, raising practical
problems of providing subsidy. This is equivalent to internalizing the transaction,
resulting in greater final output and welfare for all. Vertical restraints such as franchise
contracts or price ceilings by the monopoly may generate the same problems as vertical
integration and are best controlled by an external regulator.[10]

Comparing models of integration and separation, Vickers (1995) looks at the trade-off
between allowing a vertically integrated monopolist to operate downstream under
Cournot competition with entrants, and forbidding the monopolist to enter the market
downstream. He shows that because of the link between the number of firms
downstream and the access price, it is optimal to have the access price higher than
marginal costs to prevent duplication of fixed costs. The monopolist's incentive to raise
rivals' costs may be outweighed by the advantage of producing at a lower average cost if
fixed costs are not duplicated. However where there is no link between the number of
firms and the access price (Bertrand competition), the access price should be set below
marginal cost to compensate for the incentive to raise rivals' costs, and it is less likely
that the monopolist should be allowed access. Thus the excess entry result common in
many product differentiation models drives the incentive to integrate.

This basic framework has been extended by Lee and Hamilton (1999) by relaxing the
assumption of identical constant marginal costs between the monopolist and
downstream firms. However they still maintain the somewhat unrealistic assumption of
the regulator knowing the costs of the downstream firms (owing to difficulty in assuming
asymmetric information both upstream and downstream). The advantage in cost for the
incumbent increases the likelihood that the integrated structure will be optimal. Iossa
(1999) presents a similar model looking at asymmetrical information in demand rather
than costs, finding that the optimal market structure depends upon the level of correlation
between the upstream and downstream goods.

4.2 Access pricing, bypass, and cross-subsidies

Where the incumbent is vertically integrated and owns an essential facility, the access
price itself becomes a crucial issue. In the United Kingdom, even as the telecoms
industry becomes increasingly competitive, the access price is now controlled by a
network price cap. With a vertically integrated upstream network and downstream
consumer supplier, the incentive for the monopoly to favor its own business with lower
access prices than the entrant's is strong. De Fraja and Waddams Price (1999) show
that welfare can be higher when the incumbent's access price is not directly regulated
but he is rewarded according to the level of entry downstream. Laffont and Tirole (1990)
model a multiproduct monopolist where the regulator cannot observe effort. The model
illustrates a number of interesting points when considering the trade-off between
consumer gain through allocative efficiency and possibly harmful distribution effects.
Asymmetric cost information between the regulator and firm increases the likelihood of
bypass, as the access cost is higher. The greater the asymmetry in costs the higher the
access cost is likely to be and hence the more chance that an entrant will build its own
network. If the competitors are unregulated and their profits take away incumbent profits
used to subsidize social obligations, then competition may be undesirable.

Curien, Jullien and Rey (1998) use a model similar to Laffont and Tirole (1990) to show
that bypass can increase welfare for all consumers connected to the network, while the
incumbent's profits are reduced and the regulator's subsidies increase at the expense of
the tax payer. They demonstrate that where subsidies to the incumbent are not allowed,
it is the largest consumers that benefit from bypass, while small consumers who have
been supported by the social obligations policy face falling welfare through exclusion.
Such exclusion may be mitigated through the diversification of the incumbent into other
markets in which it is free to adjust prices.

The most useful models look at the cross-subsidization problem and the incentive to
increase access price trade-off as a whole rather than separately. Laffont and Tirole
(1993) provide some insight into this problem where the regulator is able to control both
the access price and the final product price of the monopolist. The optimal access price
will exceed the marginal cost of access in proportion to elasticity (for Ramsey reasons).
This access price will be greater when the regulator attaches higher value to the
monopolist's profit than to that of the competitive fringe (perhaps for financing social
obligations). To facilitate social obligations and bypass problems of asymmetric
knowledge, one suggestion has been the use of "global" price caps on the entire
incumbent's output. Laffont and Tirole (1996) argue that when access is simply added
into a weighted basket of products, the incumbent is induced to choose the optimal
Ramsey prices, subject to the consumer surplus weights in the basket. This regime
allows the incumbent to change prices as long as the average value remains within the
global price cap, but has been criticized as potentially leading back to subsidizing
predation by the incumbent.[12]

4.3 Competition in telecoms networks
These issues have been reflected in practice through the gradual creation of new
telecoms networks bypassing the existing incumbents, and creating problems as well as
benefits. Laffont and Tirole (1998a, 1998b) model an unregulated telecoms industry
characterized by several interconnected networks. Their findings cast doubt on the
general conclusion that increased competition results in lower prices. Assuming that
there is balance between outflows and inflows of calls, a non-price-discriminating firm
that reduces the usage price increases the number of consumers joining the network, but
does not increase profitability (owing to balancing). The greater number of consumers on
the network increases the number of calls made, raising the number of off-network calls
and triggering an access price deficit to the other network. Where price discrimination is
allowed, firms are able to charge relatively high prices to calls off the network compared
to on network calls, reducing the access deficit. Consequently high access prices may
trigger intense competition for market share and consequently low "on-network" prices.
Where firms are forced to adopt uniform prices, reducing final price results in a higher
market share, but this increases the call volume, resulting in a higher access deficit and
a loss in profitability. Hence they conclude that under uniform pricing there is little
incentive to reduce prices and competition is weakened. This result is strengthened by
disincentives for a new limited coverage entrant (under uniform prices) to undercut the
full network incumbent, as it will again incur an access deficit. The entrant may prefer to
under-invest in coverage (taking the role of "puppy dog") to transform the incumbent into
a pacified "fat cat".

The introduction of non-linear price competition increases competition among the
networks as the firms can reduce the fixed fee and build market share without incurring
an access deficit (keeping usage price high). Laffont and Tirole find that price
discrimination may increase competition but creates inefficiencies compared with uniform
prices. For small-scale entry, an incumbent may use price discrimination to squeeze the
small firm out by raising terminating access charges. The high access charge results in a
large access deficit for the entrant and eventual eviction from the market. Unless
entrants can quickly build a large network they are unlikely to remain in the market. This
implies that in reality where price discrimination is practiced it may well be necessary to
regulate access prices until the networks are of a similar size.

Whether the firm is vertically integrated or not, competition has usually been introduced
in the presence of the former monopolist as incumbent. How, then, might a competitive
fringe behave? Caillaud (1991) analyzes the optimal regulation of a dominant firm facing
Bertrand competition from a fringe of competitive firms which are not regulated. He finds
that the fringe will potentially produce more efficiently than the regulated firm, and its
presence reduces the asymmetry in cost information when the fringe's costs are
correlated with the incumbent.

Taylor and Weisman (1996) discuss an incentive contract proposed for regulating
Canadian Radio-Television and Telecommunications. The contract is designed to avoid
some of the problems of price cap regulation discussed in this chapter. They consider
the introduction of a yardstick means of sharing productivity gains. Where the firm's
productivity is higher than industry productivity, the firm is allowed to keep some
specified proportion of this increase in the form of a proportionately higher revenue
constraint. The new contract also incentivizes increasing the quality of service through
comparison with the last period's quality, and through managing the weights may provide
an effective disincentive to manipulate quality. Lastly they explicitly adjust for the loss of
the incumbent's market share to entrants. Each service has a weight attached to it
dependent upon the incumbent's market power; as market power decreases the weight
of that service within the tariff basket falls, until eventually at some set level it is removed
altogether. The authors propose that the strategic manipulation of two-part tariffs
along the line of Sappington and Sibley (1992) will be welfare enhancing as long as
consumers may choose between the non-linear and original tariff. This type of incentive
contract illustrates the development of practice through a detailed contract specifying as
closely as possible the optimal price for the firm, while preserving the incentive structure
for cost reductions and efficiency savings.
4.4 Competition and distributive concerns

The issue of access deficit charges for social obligations raises more general questions
about the nature of these "utility" industries, which are likely to form part of a larger social
distribution policy. The "Universal Service Obligation" (USO) policy found in many
developed economies is based on the argument that each citizen has a right to clean
water, electricity, and other basic utilities, regardless of income or location. Consequently
many of the prices for goods in such monopoly industries do not fully reflect costs. One
example of this is in the gas industry's charges. In the United Kingdom, British Gas
traditionally charged the same amount for gas regardless of the level of costs their
service entailed. Those consumers whose costs of service are low (i.e. customers that
pay by direct debit) indirectly subsidized those with high costs of service (customers
paying by coin meters). When competition in domestic gas was first introduced in 1996,
the entrants logically targeted the most profitable direct debit market. This resulted in a
reduction of the incumbent's profits that were further reduced by the regulator initially
preventing British Gas from raising prices to customers not paying by direct debit. Had
this continued, over time British Gas Trading would no longer have been able to fulfill its
social obligations to higher-cost consumers and still make a profit. To compensate for
this, the regulator has allowed some rebalancing of prices between the high- and low-
cost consumers, though it has limited the extent to which the incumbent is able to do this,
ostensibly on grounds of discrimination. Direct regulation was removed from the retail
gas market in April 2001, with the only remaining constraint (initially for one year) being
an undertaking that the difference between charges to the less and more competitive
markets should not increase.

Because of these social dimensions and the desire for universal service, there may be
greater weight attached to profits made by the incumbent monopolist than by the
entrants when social obligations and cross-subsidization are issues. All the privatized
industries have inherited such cross-subsidies, hence the challenge is to incorporate
such objectives within the incentive contract in the presence of competition.

The solution to the conundrum of promotion of competition and social obligations such
as a USO may be finally solved only through the unravelling of cross-subsidized
products, leaving price regulation only on services deemed essential. The problem
however is how to treat this in the transitional period. Helm and Jenkinson (1997)
propose four possible policy responses: allowing a long transition time; increasing prices
to deal with costs; relying on social security policy; and relying on the efficiency gains to
offset the cross-subsidies. Another approach is through a competitively neutral tax on the
revenues of all providers. The provider of a universal service would then receive a lump-
sum subsidy to compensate for the difference in price and cost of the service. The
solution requires the controversial deregulation of many of the incumbents' markets once
they are deemed competitive. The energy regulator appointed in 1999 has addressed
this issue directly, recognizing the potential of these concerns to prevent or distort the
development of competition (Ofgem 1999). The best solution is often to reinstate the
efficiency/equity division and provide direct government support for vulnerable groups to
replace the cross-subsidies. Unfortunately governments are not always amenable to
such solution, despite their benefits. Particular problems may arise if there are
simultaneous changes across several industries which adversely affect some vulnerable
or politically sensitive group (Waddams Price and Hancock 1998).

A sign of the increasing importance that the government has placed upon the consumer
welfare aspects of competition has been the passing of the Utilities Act 2000. This
changed the primary task of the energy regulator to one of protecting consumers and
introduced a new duty to take account of the interests of individuals with low income,
providing an explicit (if unspecific) distributional remit. The Act also provided for the
government to issue guidance on environmental or social concerns which the regulator
must take into account, but need not act upon. As important as these remit change has
been the new institutional framework which the Act introduced. The powers of the
individual regulator have been transferred to an Authority, with a majority of non-
executive directors, which "the regulator" now chairs. A new consumer body has been
created with both a representative and an advocacy remit, although there remain
questions about what differentiates this role from that of the Regulatory Authority with its
own new consumer-oriented duties. Similar legislation was expected to be introduced for
the water industry (except that the regulator will remain an individual) and the
communications sector in 2002.

A good introduction to the problem of vertical integration and restraints can be found in
Waterson (1996).

See Baron and Myerson (1982) for an influential model of asymmetric information in which
they assume that the firm's effort to reduce costs is exogenous and concentrate on the
information asymmetry between the regulator and the firm.

See Armstrong, Doyle and Vickers (1996) for an analysis of this criticism.

Taylor and Weisman are clear that market share is not a sufficient indication of market
power, as first it may not be a good measure of power in nationalized industries, and secondly
it may be strategically manipulated by both the incumbent and the entrants.

See Waddams Price and Bennett (1991) for an account of domestic gas competition and
the types of consumers who have switched their gas supplies from British Gas Trading to the
new entrants.
5 Conclusions
The very success of incentive regulation in generating cost reductions and profit
increases led to political calls to reform the UK system, because of perceptions of
excessive profits from lax regulation. The "windfall tax" introduced by the Labour
government in 1997 runs contrary to the original spirit of incentive regulation. Because
gains to consumers through the price cap are independent of the firm's earnings,
announcement of high profits generates political pressure to claim that the price cap is
flawed and consumers should gain a larger share of the realized gains. The clawback
may take place explicitly (i.e. through a "windfall tax") or implicitly through higher
demands for quality or increasing competition at an earlier date, but we have already
seen that this weakens incentives.

This chapter shows that there is no simple answer to creating the optimal regulation
regime. Over time regulatory contracts have become more incentive-based under the
different types of price cap and tariff basket schemes. However there still remain many
problems such as quality control, strategic manipulation, and cross-subsidization, among
others. Consequently the design of an optimal incentive contract depends largely on the
regulator's main goal. The goal itself depends on the regulatory structure, and the
institutional relationship between regulator and government. How much discretion should
the regulator have in determining objectives and the means to achieve them? Who sets
the goals, how well are they defined, and how closely do they reflect society's
preferences? These are the very issues which prompted the second round of regulatory
reform leading to the Utilities Act in the United Kingdom.

These reforms underline the trade-offs inherent in the regulation process itself. The
literature demonstrates that identifying the correct level of incentive is like walking a
tightrope: too much discretion to the firm may result in a reduction of total welfare; too
little dulls the incentives for cost reduction. In designing an incentive contract it is almost
impossible to shift all of the producers' gains into consumer gains without destroying the
incentives which the original contract so carefully sought to create. At the same time the
overall system needs to be politically acceptable in terms of the distributional

What is clear from application in the United Kingdom and United States is that regulatory
contracts are no longer simply matters of price caps concentrating on a single
performance measure. Incentive contracts must be designed to take account of several
different dimensions, and have become increasingly complex. In the United Kingdom this
has generally evolved behind closed doors at the discretion of the regulator, whilst the
United States has adopted a more transparent methodology using their system of public
hearings. A parallel may be drawn with Williamson's (1976) criticism of franchising in that
for an efficient regime to exist, the contract must by specified in its entirety. Even where
the aims are explicit, it is doubtful whether such a detailed contract is compatible with the
original aim of light handed regulation.

The question is how regulation should proceed where such complete contracts are not
feasible. Some regulators (notably in the United Kingdom) have moved to an informal
contract framework in which the regulator has considerable unofficial influence over the
actions of the monopolist. Several economists have argued that one of the main
advantages of the UK system over the US is the United Kingdom's ability to use
judgment in setting prices. A range of factors may be considered when setting price
constraints which do not necessarily explicitly enter the contract. Littlechild (1983) saw
the process of resetting X as a bargaining process between the regulator and firm taking
account of a whole range of factors, rather than relying heavily on total factor productivity
measurements, as had been common in the United States. However we return to the
danger of weakening regulatory commitment where the ability of the regulator to
manipulate the rules may result in stranded assets, uncertainty, a reduction below
optimal investment levels, and higher costs.
In reality, all regulation constitutes a series of choices with various associated trade-offs.
Incentive mechanisms within utilities are limited both by the nature of the industries and
the need to achieve political consensus in the design and outcome of their regulation.

See Sappington and Weisman (1996) for a description of the limits and myths of incentive
The authors acknowledge funding from the ESRC under grants R00429834287 and
R022250147, respectively, and are very grateful for many helpful comments from Morten
Hviid, though he bears no responsibility for the outcome. The first version of this chapter
was written when both authors were members of the Centre for Management under
Regulation at the University of Warwick.
1. See Sherman (1989, chapter 8), for a detailed discussion of these
2. For empirical evidence of the AJ model hypothesis, Courville (1974) finds
that for all 110 rate of return regulated plants analyzed, the ratio of input
prices exceeds the ratio of marginal products as the AJ model suggests.
He finds that costs are up to 40 percent higher than the minimum
efficient level, with the average being 11.6 percent higher. See also
Petersen (1975) and Jones (1983) for other studies confirming the
general bias result.
3. The possibility that the government's attitude to industry profits changes
post-privatization is considered in section 3.
4. The problem of completely specifying a contract is discussed in more
detail at the end of sub-section 3.2.
5. This issue is explored in Green and Waddams Price (1995).
6. The estimated budget for Oftel in 1999 was £12.6 million, which may still
be thought of as a regulatory bargain when considered against the
turnover of a UK telecommunications market estimated at £22 billion.
Similarly Ofgas' budget of £12.9 million in 1998 may be compared to
Centrica and Transco's combined revenue of £11 billion. What is less
clear is the cost of the regulatory burden placed on the companies to
meet the regulator's information demands, and in pursuing rent-seeking
7. See Tardiff and Taylor (1996) for a good summary of some of the
detailed issues.
8. See Shleifer (1985) for the formal theory behind yardstick competition.
9. See Crew and Kleindorfer (1999) for a discussion of stranded assets and
a means of allowing fair recovery.
10. A good introduction to the problem of vertical integration and restraints
can be found in Waterson (1996).
11. See Baron and Myerson (1982) for an influential model of asymmetric
information in which they assume that the firm's effort to reduce costs is
exogenous and concentrate on the information asymmetry between the
regulator and the firm.
12. See Armstrong, Doyle and Vickers (1996) for an analysis of this criticism.
13. Taylor and Weisman are clear that market share is not a sufficient
indication of market power, as first it may not be a good measure of
power in nationalized industries, and secondly it may be strategically
manipulated by both the incumbent and the entrants.
14. See Waddams Price and Bennett (1991) for an account of domestic gas
competition and the types of consumers who have switched their gas
supplies from British Gas Trading to the new entrants.
15. See Sappington and Weisman (1996) for a description of the limits and
myths of incentive regulation.
Contractual Choice and Performance-
Chapter 24:

The Case of Water Supply in France
Claude M©nard, St©phane Saussier
1 Introduction
A great variety of contractual arrangements coexist today in the provision of public
utilities such as water supply, urban transportation, and electricity. In the extensive set of
modes of governance to which these arrangements correspond, the "purely" integrated
form of a service provider owned and managed as a public "bureau" appears as a very
specific case, and maybe one in extinction. The general reexamination of public
provision for these services that developed in the 1980s raises the issue of the extension
of government activities. This question by far exceeds the problem of privatization, with
which it is too often identified. Beyond the transfer of property rights, important decisions
must be made about the choice of the most satisfactory mode of governance for
providing these services. Research by Hart, Shleifer and Vishny (1997) and Williamson
(1999) looks for more rigorous analytical foundations to the resulting trade-off.

With regard to these issues, the case of water supply is a particularly rich domain. There
is no doubt about the importance of guaranteeing safe and regular provision of water to
the population. However, the choice of the most relevant mode of governance for doing
so efficiently, i.e. at a low price and with high quality, remains an open question. Studies
such as M©nard and Shirley (1999) show a significant dispersion of results for similar
contracts, suggesting a major impact of institutional factors. Depending on the context,
public providers sometime perform quite well while, symmetrically, private operators also
fail. Other studies claim that disengagement of local authorities in favor of private sector
participation systematically improves performance, at least under certain conditions
(World Bank 1995; Gatty 1998). Last, empirical surveys show innumerable malfunctions,
whatever the mode of governance is (Cour des Comptes 1997).

The French situation presents an exceptional terrain for studying these questions. Water
supply has been under local responsibility for centuries, generating a wide variety of
solutions. At the same time the rules of the game constraining choices (e.g.
environmental laws) are the same for all, making the institutional environment continuous,
stable, and homogeneous. Thus, it becomes feasible to compare alternative modes of
governance that monitor similar activities. In this chapter, we take advantage of this
situation to shed light on two questions. How much does the choice of a governance
structure for providing public utilities depend on economic choices related to
characteristics of the good to be distributed and the transactions that are involved in
doing so? And do some modes outperform others systematically?
More precisely, this chapter presents results based on a detailed comparative analysis of
performance for different contractual arrangements in the water sector. The study put
aside factors that may depend on institutional elements (e.g. political influence) in order
to focus on variables related to the governance per se. We used a database that
provides information on all units supplying water (WSU) to towns of more than 5,000
inhabitants. This panel includes 2,109 WSU, for a period of three years (1993-5); it
represents 73 percent of the French population.
After a short overview of the organization of the water sector in France (section 2), we
introduce our analytical framework, based on recent developments in transaction cost
economics (section 3). The propositions derived from that framework are then tested on
our data set, in order to shed light on the economic rationale behind the choice of a
mode of governance (section 4) and on the links between the arrangements chosen and
their performance (section 5). We show that these choices, although they are made in a
sector that is particularly sensitive to political decisions, obey significant economic
determinants. Neglecting the latter in making the choice of a contractual arrangement
translates immediately into decrements in performance.
A forthcoming study will complete these data by a set of contracts that covers all the main
cities, with information about a wide variety of variables (such as size, demography, and
geological factors).
2 Contractual arrangements: characteristics of our sample
Before proceeding to the analysis itself, we need to briefly introduce some major
characteristics of the organization of the water sector in France. Considering the goal of
this chapter, we will not report strictly institutional characteristics (e.g. laws regulating the
entire sector).

Water supply is different from other French network industries providing services to the
public, such as mail, rail transportation, and electricity, in that it has traditionally been
decentralized. The choice of the mode of governance and its monitoring depend
primarily on local authorities. Successive laws have defined the general rules within
which these choices operate. There are three main types of law that govern the sector:
(1) Laws defining quality standards, because of the externalities on public health; (2)
Laws compelling decision-makers to obey rules intended to make these choices
transparent, in order to reduce risks of "capture" by operators and risks of corruption; (3)
Laws oriented toward the protection of the environment and of a scarce resource.

Within these general rules, which allow flexibility unknown in most other public utilities in
France, there is a wide variety of contractual arrangements and of their accompanying
modes of governance. It is standard to differentiate three families of arrangements.
The first one is that of public bureaus ("R©gies") involving direct ownership and control
by local authorities. This mode is called "gestion directe" (direct management). Three
sub-varieties can be identified. The "r©gie directe" is actually a public department
through which local authorities directly manage the provision of water. The "r©gie
autonome" characterizes a situation in which the agency providing water acquires
financial autonomy but remains without legal independence: legally, it is not distinct from
the local government. Last, the "r©gie personnalis©e" identifies a public agency with
financial autonomy and some autonomy in its corporate governance (with a Board of
Administration, usually appointed by local authorities, and a director elected by the
A second mode of governance is characterized by the involvement of an external partner,
a private operator acting as a manager, while the water system remains publicly owned.
This is called "gestion interm©diaire" (intermediary management), with an associated
governance structure identified as "R©gie assist©e". In one sub-variety, the "r©gie
int©ress©e," the operation and maintenance of the service are outsourced to a contractor,
while local authorities remain responsible for investments and financial risks. The
operator is involved in determining the price of the service and is paid a fixed amount for
the service provided, usually complemented by revenue based on performance. The
other sub-variety, the "g©rance," differs essentially with regard to the incentive
mechanism, since the operator is not involved in price-setting and receives a fixed
amount for his services.
The third family covers different forms of "franchising" and is called "gestion d©l©gu©e"
(delegated management). Typically, this is a contractual arrangement in which the
franchiser, i.e. the local government, delegates to a franchisee, i.e. a private operator,
the responsibility of providing water. In the case of "affermage," which corresponds to a
lease, the franchiser delegates the operation and maintenance of the system as well as
some investments to the franchisee, with the contract specifying goals and constraints
(e.g. delays for connections), while the local government remains in charge of all major
investments and bears financial risks. The franchisee assumes the risks related to the
daily maintenance and operation, and is paid by collecting bills from users according to
rules (e.g. prices) negotiated in the contract. The other case is that of a "concession," in
which local authorities delegate investments, maintenance, and daily operation
(connecting, billing, collecting) to a private operator through a long-term contract. The
operator bears the financial risks and gets its revenues by collecting bills from users,
under constraints (e.g. prices) negotiated in the contract. At the end of the contract, all
assets remain the property of local authorities.

One last arrangement to be mentioned, although it is extremely marginal in France, as in
almost all countries,[2] is privatization, in which case a private operator fully owns and
operates all assets related to the provision of water.
To summarize, there is a wide spectrum of arrangements, and all of them are present in
France (see table 24.2). However, most of our study will focus on the three dominant
forms, i.e. public bureaus, lease and concessions, notwithstanding the diversity
introduced by the sub-varieties. Together, these three forms represent over 95 percent of
the arrangements. The number of fully private operators in our sample is too small to be
significant in our tests. The distribution of contractual arrangements among the three
forms is provided in table 24.1. We have indicated the size of populations concerned,
since this variable is important in measuring the full significance of the distribution
system adopted; moreover, this variable will play an important role in our analysis.

Table 24.1: Permanent average population, by type of arrangement
Contractual Observations Average Std Min Max
arrangeme populati error
nts on

Direct 534 18,704 41,745 528 606,147
Lease 1,416 16,619 32,709 200 586,501
Concession 102 58,112 116,550 3,065 698,127

Source: Direction G©n©rale de la Sant©.

Table 24.2: Distribution of contractual arrangements, by regional agencies

Regional agencies

Contractual Seine- Loire- Rhône- Adour- D©partments Rhin- Artois
arrangements Normandi Bretagn M©diterran©e Garonne d'Outre-mer Meuse Picardie
e e

Direct 16.7 24.6 23.1 22.7 0 43.2 30.9

Assisted direct 1.1 1.5 1.2 1.8 15.1 3.4 2.3

Lease 71 61.5 74.3 65.5 84.9 53.4 57.6

Concession 7.8 69. 1.2 5.2 0 0 8.6

Privatization 2.5 0 0.2 0.3 0 0 0.6

Others 0.9 5.5 0 4.5 0 0 0

Observations 438 468 520 287 73 148 185

Source: Direction G©n©rale de la Sant©.

One last thing needs to be mentioned. All the operators, whatever their status, are
coordinated and partially supervised by regional agencies ("Agences de l'Eau"). These
agencies correspond to the main rivers defining the major basins that provide water.
These agencies are designed to coordinate the usage of a collective resource by the
different users and to prevent and control pollution. Their main interest for our study is
that they provide us with a geographical dimension, thus allowing a more precise
distribution of contractual arrangements that includes geological and climatic factors.
These factors have an important impact on costs and on consumption. In 1995, for the
WSU serving more than 5,000 inhabitants, table 24.2 shows the distribution of
contractual arrangements.
These data demonstrate the interest of a study of the water sector for the economy of
organizations and contracts. They show that, for the same sector, producing goods and
services that are relatively homogeneous, using well-known technologies, and sharing
characteristics with most network industries, we have a large variety of contractual
arrangements. This raises questions that are at the core of our study: How do we explain
such a diversity of arrangements for organizing similar transactions? Does this diversity
translate in significant differences in performance? And is there a logical and coherent
distribution of these performance differences (if they exist) among the modes of

The United Kingdom is the only significant exception so far, with the privatization of water in
England and Wales in 1989. The sector remains highly regulated by OFWAT (the Office of
Water Services).

In an on-going project we are planning case studies to examine their performance.

Corsica and Oversea Territories (DOM) are exceptions: they correspond to an area, not a
3 Our analytical framework
Three main approaches to the problem of the choice of contractual arrangements have
been developed in recent economic literature. A first approach put the emphasis on
asymmetry of information between the government and the operator as the key factor in
the provision of public utilities (Laffont and Tirole 1993). Choosing the best information
revealing scheme ex ante is therefore at the core of the trade-off among alternative
modes of governance. For example, if asymmetries are such that the franchiser (the
government) can not obtain the relevant information, it may be better for him to provide
the service directly, which is a form of integration. As a result, this type of analysis
focuses essentially on the incentive mechanisms and neglects ex post adaptation that
requires devices built into the mode of governance. A second approach emphasizes the
allocation of residual property rights in the decision to outsource a service versus
providing it "in-house" (Hart, Schleifer and Vishny 1997). There is a trade-off between
quality and cost in providing a collective service with the assumption that there exists an
adverse effect between quality and cost (i.e. it is not possible to increase quality and
decrease cost at the same time). The choice of the mode of governance must be made
according to the priority, with public bureaus emphasizing quality factors, since their lack
of control over residual rights provides them little incentive to reduce costs, while private
operators react the other way around. This analysis raises important issues, since the
trade-off between quality and cost is so central in the provision of water; but it ignores
the variety of potential contracts between the polar cases of private versus public
operators. A third approach analyzes the choice of a mode of governance as the search
for a form that proposes relevant incentives ex ante without neglecting the role of
contractual hazards that will require adaptation ex post. The degree of adaptability
required, and therefore the form of the contract, will depend on the characteristics of
transactions at stake. Initially developed for explaining the trade-off between making or
buying, and progressively extended to take into account intermediate modes of
governance ("hybrid arrangements"), the framework of the economics of transaction has
recently been applied to the decision that a government must make between providing a
service itself, or outsourcing it through contractual arrangements (Williamson 1999).
In order to answer the questions raised at the end of section 2, we will use this last
approach that has been so successful empirically.[6] The analytical framework, largely
developed in Williamson (1985; see also 1996), is now well known. Let us assume that
agents are looking for efficient modes of organization, i.e. arrangements that will
minimize both their costs of production and their costs of transaction, under the
constraint that represents the risk of opportunistic behavior of their partners. The theory
then predicts that the trade-off among different possible arrangements and the adequacy
of the resulting choice depend on the characteristics of the transaction that the mode of
governance has to organize. Identifying these characteristics makes the central
proposition testable: efficient modes of governance are those in correspondence with the
degree of specificity of the assets required by the transaction and the degree of
uncertainty surrounding this transaction. As a consequence, misalignment of an
arrangement increases transaction costs, providing incentives to shift to another
arrangement. A very large number of econometric tests confirms the robustness of this
prediction, particularly for cases in which the tradeoff for a firm is between buying on the
market or making in-house.

More recent studies have extended the initial model, showing a wide array of
arrangements between markets and integrated firms. Moreover, some of these studies
have shown circumstances in which several substantially different arrangements coexist,
without significant differences in performance (M©nard 1996). At first sight, the data
above suggest that this is the case for water supply in France, since several modes of
governance have persisted over time within the same institutional environment. A main
goal of this chapter is to determine whether there is a relationship between modes of
governance and performance. If performance were similar across very different
arrangements operating on the same transactions within the same environment, then
transaction-cost theory would be weakened. On the other hand, if performance differs,
then the persistence of different forms would have to be explained by other factors, e.g.
the political dimension involved in choosing the mode of governance for providing water,
path dependency, and so forth.
In order to explore the determinants of the mode of governance and the resulting
performance, we will define propositions based on the hard core of transaction cost
economics, i.e. the hypothesis that a mode of governance performs much better if it fits
the characteristics of the transaction it supports, namely, specificity of assets and
uncertainty. Space constraints prevent us from looking at these determinants and their
rationale, we will restrict ourselves to applying the basic propositions to the case under
review, in order to focus on our data and our test.
Proposition 1

The more a geographic area requires specific investments to provide water, the lower is the
probability of outsourcing these investments (i.e. delegating), everything else remaining

This proposition results directly from Williamson's hypothesis, one of the most often
tested, according to which a higher degree of specificity in investments pushes towards
more integration. In our version, this means that when highly specific investments are
required, it is likely that integrated forms (i.e. "r©gies") will prevail over arrangements that
are closer to market forms (e.g. concessions).
Proposition 2

With specific investments required for distributing water in a certain area, the higher the
uncertainty in that distribution, the lower the probability of outsourcing these investments
(i.e. delegating), everything else remaining constant.

Again, this proposition simply expresses Williamson's hypothesis that there is a close
relationship between the degree of uncertainty surrounding a transaction and the degree
of integration. Indeed, increasing uncertainty pushes us towards the adoption of a mode
of governance that allows tight control, the polar case being full integration. In our
typology of arrangements, direct management by a public bureau ("r©gie directe") is the
extreme expression of such integration.

These two propositions, now quite standard in transaction-cost economics (TCE), do not
shed light on the institutional dimension involved in the decision to choose a specific
mode of governance. Indeed, the logic underlying these propositions focuses on
economic determinants. So far, we have assumed that agents have a strong incentive to
choose the most efficient mode of governance. This assumption is quite reasonable
when we study actors operating in highly competitive markets. It can be seriously
challenged, however, in an analysis of the decisions made by local authorities for utilities
that are largely protected from competition. In these circumstances, it is likely that
important factors other than economic efficiency, e.g. support of key political
constituencies, will play an important role. For example, local authorities may choose a
form that will allow them to influence local employment, a much easier task with a public
bureau ("r©gie") than with a private operator whose autonomy of decision is protected by
a long-term concession. Political orientation may also be a factor.[8] We plan to come
back to these issues in another paper.

One last thing that we want to consider, because of its importance to local authorities, is
the role of financial constraints. Specific investments are usually costly and can hardly
(or not at all) be redeployed. Water is a sector with very important sunk costs, and these
costs represent a very high proportion of total costs (up to 80 percent: Shirley and
M©nard 2002). Many local governments will therefore be subject to financial constraints
that do not allow them to chose the mode of governance they would otherwise prefer for
that type of investment. This can actually be considered as another side of specific
investments. We translate this into the following proposition:
Proposition 3

Local authorities with limited budgets are more likely to choose to outsource than to provide
the service themselves, when significant specific investments are involved, everything else
remaining constant.

What follows is a highly simplified summary of the different approaches. Space constraints
notwithstanding, it is important to make explicit and in comparative terms some reasons for
our choice of the approach developed in this chapter.

For surveys of this empirical literature, see Joskow (1988a); Klein and Shelanski (1995);
Crocker and Masten (1996); Coeurderoy and Quelin (1997), and Masten and Saussier see
chapter 16 in this volume, pp. 273-291.

The heuristic model is in Williamson (1985, chapter 4). More is developed in Williamson
(1996) and, with more technical details, in Saussier (1997, 1999).

A previous study, based on a limited number of cities, concluded that the political orientation
of local authorities did not play any significant role in the choice of the mode of governance
(Derycke 1990). But political factors may still be involved that transcend delineation of political
parties (e.g. influence, corruption).
4 The choice of the mode of governance: our variables
Our analysis is based on a sample of 2,109 Water Supply Units (WSU), serving all towns
of over 5,000 inhabitants, for the period 1993“5. These units represent only 7.3 percent
of the total units providing water to the French population, but they cover the needs of
72.6 percent of the total population. In order to test our propositions, we have identified
for each unit, during the period under review, information relevant to the characteristics
of transactions identified in our theoretical framework, namely: investments, uncertainty,
and the financial constraint.

4.1 Investments
According to our proposition 1, geographical areas that require large investments to
guarantee a reliable supply of water should push toward integration by local authorities,
i.e. WSU should be under their direct control ("r©gie"). So far, we do not have coherent
data on investments required for each WSU. However, we were able to identify proxies
that are closely correlated with the level of investments.
4.1.1 Properties of raw water
One indicator of the volume of investments needed is the quality of raw water available
and the related treatment it requires from the WSU. The worse the quality of raw water,
the greater the investments required for its treatment. Quality of surface water is
indicated by a standardized typology: A1 is for raw water that requires only simple
mechanical filtering with light disinfection; at A2, raw water requires a combination of
physical and chemical treatment, plus disinfection; for A3, raw water needs all of the
previous treatments, plus a refining process; last, the level OS ("out of standard")
designates quality that poses exceptional problems. To represent this quality factor, for
which we have the relevant information, we use the variable A3OS which takes value 1 if
the WSU operates in departments (the French administrative unit) where there exist raw
water of quality A3 or OS, 0 otherwise.
4.1.2 Origin of water
As for underground water, we do not have information on its initial quality before
treatment. However, it is well known that underground water is of much better quality
than surface water. Hence, units for purifying underground water are less complex and
less expensive. On the other hand, underground water is more costly to exploit. Pumping
requires investments significantly larger than does routing surface water into canalization.
For similar quality, different sources of water therefore require significantly different
amounts of investment. To capture this characteristic, we have isolated the WSU that
operate in departments where all water comes from underground. This variable is
labeled WATUND.
4.1.3 Population affected
Last, the size of the population for which a WSU provides water also plays an important
role in the size of investments as well as in the dependency of local authorities on a
potential private operator. First, the larger the size of a population, the more rapid
amortization can be. This will reduce the incentive to have long-term contracts in which
control is more diffuse, thus favoring the risk of opportunistic behavior by the operator.
Second, the size of the population also influences the economic and technical capacities
that local authorities can mobilize. Small towns have fewer internal resources either to
produce water themselves or to monitor and control private operators, while using
external expertise is costly, since private operators have little interest in managing
smaller systems. This may explain the tendency of small towns to create pools, either to
provide water directly through a joint bureau or to outsource. When the population is
large, local authorities can much more easily hire technical expertise and, simultaneously,
their market is more attractive to the private operators. With a large population, the
choice of a contractual arrangement is much more open. We capture this effect with the
variable PERMPOP.
To summarize, we have three proxies that can indicate the degree of specificity of
investments required: A3OS,WATUND, and PERMPOP.

4.2 Uncertainty
Our proposition 2 suggests that areas in which transactions are plagued with a high level
of uncertainty should be "integrated," i.e. water should be provided through direct
management ("r©gie"). Sources of uncertainty may include climate (rainfall, drought) and
other unknown factors that influence the volume of water to be distributed (economic
development of the area, variation of future population) or its quality. The available data
do not provide us with fully satisfying proxies for these factors. However, taking into
account the basins through dummies allows us to approximate part of the problem, since
they correspond to natural geographic area (climate) and to areas with specific urban
and economic development.

4.3 Financial constraints
Last, our proposition 3 emphasizes that the size of investments also translates into
financial constraints. In addition to the size of the population, which obviously affects the
potential budget of local authorities (see our variable PERMPOP), another factor plays
an important role: the gap between average and permanent population, a factor largely
owing to seasonal variation. Indeed, such variations, when they are substantial (e.g.
winter resorts, or the Riviera in the summer) require substantial investments to meet the
seasonal demand, and these investments are often very significant relatively to the
financial resources available to local authorities. We capture this with our variable

4.4 Performance

In our introduction, we stressed that one important goal of this chapter was to evaluate
performance of each mode of governance. Indeed, a key point of our analysis is to
identify whether or not we can observe significant differences according to the mode of
governance chosen, and to determine if there is a mode better adapted to the
characteristics of the distribution of water. As is well known, choosing the relevant
variables for measuring performance is not trivial. Several dimensions can be taken into
account, and several indicators can be chosen: financial, economic, or even physical. In
this chapter, we adopt a simple criterion with a clear rationale for water service, the
capacity of WSU to provide water that meets legal standards.[9]

In France, standards of quality are defined by a legal decree (no. 89.3, from January 3,
1989).[10] Their implementation and control are under the responsibility of powerful
regional administrations ("Directions Departementales des Affaires Sanitaires et Sociales,
DDASS). Any anomaly detected by controllers of the DDASS or of specialized
organizations must be reported to DDASS. It is followed, according to the severity of the
anomaly, by additional controls, by imposition of measures to correct the situation or,
when threat to health is serious, by prohibition of the incriminated water for consumption.
Standards of quality changed significantly over the twentieth century, with increasingly
tighter requirements. At the beginning of the century, drinkable water was defined
through six chemical parameters and the identification of two microorganisms. Before the
decree of 1989, twenty-one parameters were taken into consideration. Now there are
sixty-two parameters used for determining quality of drinkable water. Obviously, these
parameters cover a very diversified set of factors. Some serve essentially as indicators
of the good condition of facilities (e.g. indicators of turbidity), so that they do not
necessarily signal a risk for consumers. But most have a direct relation to health.
Another important point to mention relates to the potentially large variation in the quality
of water. The quality of raw water depends on where it is captured. It is subject to
hazards related to natural conditions (hydrogeology, meteorology) as well as to
temporary pollution. It also varies according to the type of treatment. Last, it changes in
the distribution process, by getting mixed with other sources of water, by contact with
materials used, and by exogenous sources of pollution. Since our goal here is to
measure as directly as possible performance of contractual arrangements, we focus on
the quality of water after treatment but before transportation and distribution to final
consumers. We use the variable DETECT, which takes value 1 for a WSU that has
been identified as producing water not meeting the standards, zero otherwise.

4.5 Checklist of our variables
Table 24.3 summarizes all variables used in our econometric tests.

Table 24.3: Variables and their meaning
Variables Definition

Dependent variables Variable taking value 1 when the
REGIE mode of organization is direct
DELEG Variable taking value 1 when the
mode of organization is direct
management; value 2 for leasing;
value 3 for concession
DETECT Variable taking value1 for a WSU
that has been identified
distributing bad-quality water, at
least once within a year, 0
Investments Variable equals the gap between
DELTAPOP average and permanent
PERMPOP Variable equals the permanent
population concerned by the
WSU A3OS Variable taking value
1 when the WSU operates in a
department where there exists
raw water of bad quality (A3 or
OS quality levels)
WATUND Variable taking value 1 when the
WSU operates in a department
where all water comes from
Control variables Variable taking value 1 when the
SN WSU operates in an area
supervised by the Seine-
Normandie regional agency
LB Variable taking value 1 when the
WSU operates in an area
supervised by the Loire-Bretagne
regional agency
RMC Variable taking value 1 when the
WSU operates in an area
supervised by the Rhône-
M©dit©rann©e-Corse regional
AG Variable taking value 1 when the
WSU operates in an area
supervised by the Adour-Garonne
regional agency
DOM Variable taking value 1 when the
WSU operates in an area
supervised by the DOM regional
RM Variable taking value 1 when the
Table 24.3: Variables and their meaning
Variables Definition

WSU operates in an area
supervised by the Rhin-Meuse
regional agency

France being a highly developed country, we assume that all population is connected. Rate
of connection is a major issue in developing countries (see Shirley and M©nard 2002).

General quality standards are based on those established by the World Health
Organization (WHO) in 1986. Sanitary standards for water for human consumption are
defined more precisely in another decree (no. 98-3, from January 3, 1989). Also relevant are
the decrees adopted by the EU (no. 75-440, no. 79-869, and no. 80-778).

Indeed, in transportation and distribution, several factors can interfere to change the quality
of water without the responsibility of the WSU being involved (e.g. negative effects of
roadwork, or of pollution originating outside of the water system).
5 Results
As already mentioned, our econometric regressions intended to clarify two main issues:
what are the determinants of contractual choice? And what is the relationship between
the arrangement chosen and its performance? Our results confirm the robustness of the
predictions we made using transaction cost economics.

5.1 Determinants of contractual choice
In order to analyze the determinants of the choice of the arrangement which
characterizes a WSU, we have defined a variable DELEG. This variable reflects the
degree of delegation chosen by local authorities (see table 24.3). It takes value 1 when
the mode of organization is direct management by local authorities ("r©gie"), i.e. there is
no delegation to a private operator; value 2 for leasing, which corresponds to a partial
delegation of authority to a private operator; and value 3 if the contract is a concession,
which is the maximum involvement of a private operator short of full privatization.[12] The
results of our tests are in table 24.4.

Table 24.4: Determinants of contractual choice
#] #] Logit REGIE (5)
Multinomial logit DELEG (3)[ Multinomial logit DELEG (4)[
Ordered logit

DELEG (1) DELEG (2) R©gie Concession R©gie Concession

0.67 0.72 0.81 1.03 0.87 0.07 0.84
***] ***] ***] ***] (0.19)
( 4.48) ( 3.94)
[ [ [ [
(3.41) (3.47) (3.034) (3.99)
0.21 0.20 0.20 1.23 0.25 0.05 0.23
(1.09) (1.04) (1.29) (0.156)
( 1.24) ( 1.21)
(3.63[ )
0.15 0.06 0.50 1.44 0.56 2.65 0.37
***] ***] ***] ***]
(0.803) [ [ [ [
( 3.33) ( 2.90) ( 2.79) ( 5.09) ( 1.90)
( 0.31)
0.23 0.13 0.30 0.59 0.36 0.54 0.28
[*] [*]
(1.135) (0.63) (1.53)
( 1.69) ( 1.67) ( 1.28) ( 1.33)
0.50 0.28 - - - - 1.03
*] (0.90) ( 2.79)
0.75 - - - - 0.55
***] ***] ***
3.30)[ 4.15)[
( ( (2.34)[
- 0.043 0.018 0.077 0.01 0.32 0.001
***] ***] ***]
(1.21) (0.26) ( 0.14)
(3.70)[ (4.47)[ (4.83)[
2 12
0.042 0.97
( 2.83)
( 0.13)
- 0.30 0.63 0.28 0.65 0.25 0.61
**] **]
*] (0.63)
2.37)[ 2.42)[
( ( ( 2.31)
( 0.63)
- 0.55 0.018 2.96 0.50 3.20 0.13
***] ***] ***] (0.75)
[ [ [
3.33 ( 4.04) ( 4.35)
( 0.10) ( 0.29)
Table 24.4: Determinants of contractual choice
#] #] Logit REGIE (5)
Multinomial logit DELEG (3)[ Multinomial logit DELEG (4)[
Ordered logit

DELEG (1) DELEG (2) R©gie Concession R©gie Concession

- 0.34 0.19 2.23 0.21 2.41 0.042
[***] [***] [***] [***] (0.31)
2.63 ( 5.43) ( 5.43) ( 5.85)
( 1.44) ( 1.52)
0.85 1.02 0.56 2.72 0.47 1.94 0.69
[***] [***] [***] [***]
***] ***] ( 4.91) ( 9.88) ( 2.50) ( 5.63) ( 3.94)
(5.22)[ (5.68)[

Log likelihood 1522 1505 1458 1285 1145

Observations 2052 2052 1831 1831 2052

Notes: For all our estimations, we took into account the possiblity that PERMPOP would have non-linear effects on the decision to choose a mode of governance. There were no significant
effects except in regession DELEG (4).
WSU operating in overseas territories (DOM) and in the basin monitored by the Rhin-Meuse agancy have been removed from the regression, because in these two cases, there is no concession
contract. Hence, the total number of observations is down to 1,831. Results are identical in the constraint model in which the variable DELTAPOP intervenes only in the decision whether or not to
outsource water provision (i.e. it is not involved in the decision to choose the specific form of outsourcing, lease versus concession.)

Significant at the 1 percent level;

significant at the 10 percent level.

significant at the 5 percent level;

A preliminary comment is necessary with regard to column DELEG (1) in table 24.4, in
which there are significant differences according to the basins. This was already
noticeable in table 24.2. Local authorities in the Seine-Normandie basin delegate much
more water provision than in other regions. Conversely, local authorities in Rhin-Meuse
delegate much less. Other basins are in between. A similar result has been observed
previously on a much more limited sample of WSU (Derycke 1990).
Let us now introduce the variables that measure the key characteristics of transactions
involved in the choice of the mode of governance. For all of them, results are significant
(see column DELEG (2) in table 24.4). Indeed, these choices are unambiguously related
to the explanatory variables that we have identified.
First, our results show a clear impact of PERMPOP. The larger the population concerned,
the more we observe delegation by local authorities. This supports proposition 1: the
larger the population, the smaller the investment per capita, and the better the
profitability for an operator. Indeed, anticipation of good profitability gives local authorities
the choice between providing "in-house" or delegating to an operator; it also provides an
incentive for operators to bid, since they can reasonably expect normal amortization of
their investments within the limit of the duration of the contract. In these circumstances,
there is an incentive to delegate.
Second, for the WSU operating in areas in which water comes exclusively from
underground, or in areas in which there exist surface water of bad quality, our test shows
a clear predominance of direct management through public bureaus ("r©gies") and, to a
lesser degree, of lease contracts. These modes allow local authorities to exert tighter
control over the operator, public ("r©gie") or private (lease), than they could over a
concession. This result substantiates proposition 2. Raw water of bad quality or of
underground origin requires much larger investments; shaving costs or being vulnerable
to opportunistic behavior by a private operator would have a negative effect on quality of
water and on the health of the population, with political consequences as a direct
Third, our test shows that the more variable the population served by a WSU, the more
likely it is that the arrangement adopted will be delegation to a private operator. Indeed,
these modes relax the financial constraint for the local authorities. The result confirms
proposition 3.
Considering the quality of the data available, we decided to go a step further and to
check the robustness of our results. One possibility is to proceed to an estimation in
assuming that the variable DELEG is a qualitative variable, but not an ordered one.
The results, based on a regression in a multinomial model, confirm our propositions (see
DELEG (3) in table 24.4). They also provide more precision on the effect of each variable
on the choice of arrangements open to local authorities. The most noticeable effect is
that strong seasonal variation in the population (DELTAPOP) has a significant impact on
the decision to not provide water through a public bureau ("r©gie"). The other variables
do not play a determinant role in that choice with this model. This is confirmed by
another estimation, in which the dependent variable is binary (see column DELEG (5) in
table 24.4). One interesting result is that larger populations (PERMPOP) increase
significantly the probability that water will be provided through a concession contract
rather than a lease, with the possibility of a non-linear effect (see column DELEG (4) in
table 24.4). On the other hand, bad water quality, or an underground source of water,
increases the probability that distribution will be through a public bureau ("r©gie") or a
lease, rather than through a concession that would escape the control of local authorities.
To summarize, our results seem robust. They also suggest that the choice of a mode of
governance proceeds in two steps. The decision to outsource or not depends centrally
on the financial constraint, particularly when investments are major ones. If the decision
is to outsource, then the choice between a lease and a concession depends largely on
the density of the population and the concomitant investments. This last point reinforces
the idea that control over potential opportunistic behavior plays an important role in the
decision process. Indeed, local authorities have much more control over the private
operator under a lease than under a concession. In the former arrangement, investments
that the operator will engage directly are almost always much less than in the latter, and
major investments remain under the control of local authorities. Moreover, the duration of
a lease being significantly shorter, control over the private operator and the capacity to
put him under competitive pressure are easier.

Therefore, it seems that the choice of a mode of governance is not random, nor is it
based purely on political determinants. There are factors involved that suggest economic
rationale in these choices. This being said, we must also acknowledge that, with the data
available for this chapter, a significant part of the variation in choices remains
unexplained, which suggests that important explanatory factors have been neglected.

5.2 Mode of governance and performance

Another goal of our chapter is to test if there is a close relationship between contractual
arrangements and performance. One puzzling aspect that confronts transaction-cost
economics (TCE) is the coexistence in some sectors, for long periods of time, of different
modes governing the same transactions (M©nard 1996). Again, our data set is
particularly useful for examining aspects of this issue since, within the same rules of the
game, we have an array of arrangements that have been operating for years, some for
decades. If the theory is right, different modes of governance monitoring transactions
with similar characteristics should have different performance. Indeed, local authorities
having chosen the "wrong fit," i.e. a contractual arrangement that is not well aligned with
the transactions having the characteristics that we have identified above, should be
much more exposed to opportunistic behavior from the operator, e.g. under-investments,
repeated renegotiations. These malfunctions should reflect in the quality of the product
delivered, which is precisely what our data measure.
As mentioned very briefly in sub-section 4.4, in order to measure the impact of
contractual choice on performance of our WSU, we selected a simple, observable, and
unchallenged criterion when it comes to provision of water, i.e. quality (which involves
safety in this sector). More precisely, we considered the probability for a WSU to be
identified as failure to meet at least one parameter of quality as defined by the law, at
least once a year, whatever this parameter is. Hence, our variable DETECT takes
value 1 for a WSU that has been identified as failing at least one quality parameter, at
least once within a year, 0 otherwise.
Our sample covered three years. Data were available for 1,942 of the 2,109 WSU of our
initial sample. Results of our econometric tests are summarized in table 24.5.

Table 24.5: Modes of organization and performance
Logit Logit Logit
variables [#] [@]
(1) (2) (3)

SN 0.50 0.94
[***] [***]
( 4.11) ( 3.70)
LB 0.86
0.47 0.45
(4.22)[***] (2.53)[***] ( 2.44)
RMC 3.80
0.69 0.41
(6.03)[***] ( 5.19) (1.27)
AG 1.19 1.01 -
[***] [***]
(9.79) (4.88)
DOM 4.43 - -
RM 0.53 - 0.32
(3.73) (0.25)
PERMPOP 2.64 6.65 3.41
(2.99)[***] (2.68) (1.21)
DELTAPOP 2.90 - -
WATUND 0.78 - -
A3OS 0.68 - -
CONTROL NUMBERS 0.10 0.27 0.14
[***] [***]
( 2.95) ( 2.59) ( 1.19)
0.55 -
(3.57)[***] ( 0.82)
REGIE 0.12
0.89 0.10
(5.64)[***] ( 0.33)
Constant 1.76 0.24 0.42
( 10.16) ( 0.08) (1.05)
3650 673 504
Log likelihood
Observations 5826 1101 795

2 3
Notes: In all our estimations, variables PERMPOP and PERMPOP are not
Table 24.5: Modes of organization and performance
Logit Logit Logit
variables [#] [@]
(1) (2) (3)

This estimation concerns only small units (less than 50,000 inhabitants) in which
there is no significant variation of population during the year (DELTAPOP = 0) and
operating in areas with water surface of bad quality (A3O3 = 1).

This estimation concerns only small units (less than 50,000 inhabitants) in which
there is no significant variation of population during the year (DELTAPOP = 0) and
operating in areas with underground raw water only (WATUND = 1).

Significant at the 1 percent level; ** significant at the 5 percent level; * significant at
the 10 percent level.

Results of our tests show that concession is the mode of governance that performs the
best (see column DETECT (1) in table 24.5), even when the specific characteristics of
the different basins are taken into account. In contrast, public bureaus ("r©gies") have the
worst performance, in that their probability of distributing water that is below some legal
standards is significantly higher.
More precisely, this is the result we obtain if we assume that the contractual
arrangement is given, i.e. we consider the arrangement as exogenous. But one
important contribution of TCE is to make the choice of the mode of governance
endogenous: each mode has its advantages and its disadvantages, with the "right"
choice depending on the characteristics of the transactions that the arrangement will
have to organize. In that respect, the decision for a government to make "in-house," i.e.
through its own "bureau," rather than outsourcing, should correspond to the same logic
(Williamson 1999). If it is so, there should be situations in which the "integrated" form
that is a public bureau ("r©gie") should perform at least as well as other forms. According
to the theoretical explanation of integration (and a public bureau is a form of integration
into the government), this should occur in areas that require heavy investments per
capita to produce and distribute water that meets quality standards and in areas in which
costly water treatment installations are required.
In order to test this proposition, we first focused on WSU serving less than 50,000
inhabitants and operating in areas with bad quality surface water (A3 and OS). WSU
operating in areas in which raw water comes exclusively from underground sources (i.e.
is of much better quality) are excluded. Thus, we are concentrating our analysis on areas
in which important investments are required and in which quality is a real problem. Our
sample then shrinks to 1,101 WSU, among which only nine operate under a concession;
we eliminate these nine units in order to focus on the measure of the respective
performance of public bureaus ("r©gies") and lease contracts. In the situation thus
described, public bureaus perform at least as well as lease units (see column DETECT
(2) in table 24.5); this is consistent with what the theory suggests.
In other terms, we need to reexamine our initial result that showed a comparative
advantage of concessions over all other forms. More precisely, a more refined test
shows that WSU under lease or concession perform better than public bureaus only
when the latter do not correspond to what the theory suggests to be the most adapted
form with regard to the characteristics of the transactions. But when these characteristics
correspond to those for which one would expect integration according to predictions
made by TCE, then the comparative advantage of lease and concessions disappears. In
a second step, we extended our analysis to WSU operating in areas with raw water of
underground origin and with populations of less than 50,000 inhabitants. The result is
identical to the previous one (see DETECT (3) in table 24.5). Hence, the two approaches
converge: when public bureaus ("r©gies") have been chosen in situations with
characteristics that correspond to what TCE predicts, these integrated forms perform at
least as well as lease or concession.
Therefore we obtain quite consistent results. First, the choice of the mode of governance
seems to follow an implicit economic logic that conforms to what TCE predicts,
notwithstanding the influence of other factors, e.g. politics. Moreover, this choice of a
mode of governance does have a direct impact on the performance of the WSU, as
measured by the criterion of quality relative to legal standards. There are significant
differences in performance among WSU. But these differences do not express the
absolute advantage of one mode of governance over the others. Rather, they follow logic
predicted by TCE. Indeed, integrated arrangements ("r©gies") are used in situations in
which problems of raw water quality are the most acute, and in which investments
required are significantly greater. To put it the other way around, when the integrated
form ("r©gies") is adopted in such circumstances, its performance is comparable to and
sometimes better than the performance of private operators working in similar conditions.

We have already mentioned that for towns of more than 5,000 inhabitants in France, which
is the base of our data set, there are not enough cases of fully privatized modes of
governance to be significant in our tests.


. 13
( 18)