Research Institute of Applied Economics 2009 Working Papers 2009/23, 33 pages
Is Private Production of Public Services Cheaper than Public
Production? A meta-regression analysis of solid
waste and water services
Germà Bel, Xavier Fageda and Mildred E. Warnerd
Germà Bel is Professor at Universitat de Barcelona and Guest Professor at Barcelona Graduate School of
Economics. Departament de Política Econòmica, Avd. Diagonal 690, 08034 Barcelona, Spain, email@example.com.
Xavier Fageda is Assistant Professor at Universitat de Barcelona. Departament de Política Econòmica, Avd.
Diagonal 690, 08034 Barcelona, Spain, firstname.lastname@example.org.
Mildred E. Warner is Professor at Cornell University. Department of City and Regional Planning, 215 W.
Sibley. Cornell University. Ithaca, NY 14853, USA; e-mail email@example.com .
Abstract: Privatization of local government services is assumed
to deliver cost savings but empirical evidence for this from
around the world is mixed. We conduct a meta-regression
analysis of all econometric studies examining privatization for
water distribution and solid waste collection services and find no
systematic support for lower costs with private production.
Differences in study results are explained by differences in time-
period of the analyses, service characteristics, and policy
environment. We do not find a genuine empirical effect of cost
savings resulting from private production. The results suggest
that to ensure cost savings, more attention be given to the cost
characteristics of the service, the transaction costs involved, and
the policy environment stimulating competition, rather than to
the debate over public versus private delivery of these services.
Keywords: Privatization, contracting-out,
governments, meta-regression analysis.
JEL classification: L33, R51, H72, C25.
Acknowledgments: Germà Bel and Xavier Fageda are thankful for support provided for this research from the
Spanish Commission of Science and Technology (SEJ 2006-04985), and from the Fundación Rafael del Pino. Mildred
Warner’s privatization research was supported by funding from the US Dept of Agriculture National Research
Initiative (NYC-121524). An earlier version of this paper was presented at the American Public Policy and
Management Conference Los Angeles, CA (November 6, 2008). We are thankful for comments from Daniel Albalate,
Trevor Brown, Raymond Gradus, and Eoin Reeves.
One of the promises of privatization is that it offers efficiency gains and reduces the costs of public
service delivery. However, after more than three decades of experimenting with contracting out local
government services, the evidence for cost savings is mixed. Some studies find that private
ownership does indeed result in a reduction in costs, but others fail to find statistically significant
differences between public and private ownership. Descriptive meta-analyses conducted by Boyne
(1998), Hirsch (1995), and Hodge (2000) report savings in some cases but not in others, with the
weight of evidence failing to support systematic cost savings for private service delivery.
Interestingly, the literature has increasingly turned its attention to factors that might undermine
savings from privatization such as hefty transaction costs (Brown & Potoski, 2003, 2005), market
concentration, and the lack of competition (Bel & Costas, 2006; Dijkgraaf & Gradus, 2007). Most of
the empirical economic literature has focused on the delivery of solid waste services and water
distribution, the two local government services with the greatest contracting experience. Recent
descriptive meta-analyses of these two services report limited evidence of cost savings—especially
among the most recent studies (Bel & Warner, 2008).
The analysis presented here constitutes a systematic test of cost differences across a range of
studies of public and private production controlling for several variables. In our study, we conduct a
meta-regression analysis of 27 empirical studies that compare the costs of private and public
production for large samples of municipalities with different attributes. We focus our attention on
cross-sectional empirical analyses that use multivariate methods to study the two local services with
the most contracting experience: solid waste collection and water provision. To date we have found
38 published papers and working paper on such studies conducted in the US, the UK, and around
the world, forming the population from which we select our sample. By statistically analyzing the
characteristics of each regression model, we are able to assess the weight of evidence concerning the
costs of public or private production.
Our empirical analysis is based on studies that compare the costs of public and private
production rather than studies that examine the impact of privatization within services as they
change ownership or between services that have changed ownership and those that have not. To
gauge the impact of privatization on costs as accurately as possible, a time-series approach to assess
the before and after effects of the change in ownership would be preferred. However, only one
paper in the extant body of literature (Lynk, 1993) used such an approach1 to determine the impact
of the change in ownership on efficiency. Inevitably, a meta-regression analysis is limited to the
available empirical evidence and, thus, our goal is to assess whether private production is less costly
than public production when controlling for attributes that can be compared across studies.
Costs are a driving factor in the decision concerning the form of service delivery (Bel & Fageda,
2007). Our objective is to analyze whether privatization (private delivery)2 constitutes an effective
alternative for reducing costs in the provision of solid waste and water distribution services at the
metropolitan scale. Expectations for cost savings stem primarily from the notion that competition
increases the pressure to achieve efficiency at lower costs. However, water distribution is a service
characterized by high asset specificity and tends toward a natural monopoly with few expectations of
competition. This might account for the fact that there have been relatively few instances of
privatization in water services. Solid waste collection tends to have more competition and more
common privatization. However, considerable industry concentration has been seen in the waste
sector over the last 20 years (Bel & Costas, 2006; Davies, 2007; Dijkgraaf & Gradus, 2008; Warner &
Bel, 2008). Therefore, in neither of the two service areas is competition expected to persist over time.
One primary benefit from opening up public services to competition from the private sector may
be efficiency improvements generated among public producers as a response to competitive pressure
and benchmarking from private providers (Hatry, 1988; Bel, 2006). Hence, one likely effect of
privatization is that it can spur improvements in public sector efficiency. This could lead to an
underestimation of the dynamic benefits of the broader privatization process. However, we are
unable to test for competition, market dynamics, or public sector improvement in our meta-
regression analysis because almost none of the empirical studies considered here controlled for these
Our meta-regression analysis does not reveal a systematic relationship between cost savings and
private production. Indeed, we find that the most recent studies, those from the US, and those on
water services, are less likely to show any savings. The rest of this paper is organized as follows. First,
1 The specific characteristics of water privatization in England and Wales allowed Lynk (1993) to adopt this
approach. Water is a provincial service in England and Wales, and privatization was undertaken
simultaneously in 1989, when 10 public agencies were privatized. Lynk compared the efficiency of these 10
agencies before privatization and of the corresponding private firms following privatization.
2 Privatization or contracting out separates the provision decision (which remains public) from the production
or actual delivery of the service (which can be contracted out or privatized). Pure privatization of municipal
services, or service shedding, is rare. All the studies in our sample examine cost differences between public
and private production (via a contract or concession agreement). In both cases, the government retains
responsibility for service delivery.
we review the literature related to the relationship between privatization and costs. Then, we explain
the methodology and describe the meta-regression analysis of the results reported in previous
studies. Finally, we present a discussion of our results in light of our theoretical concerns.
RELATIONSHIP TO THE THEORETICAL AND EMPIRICAL LITERATURE ON
PRIVATIZATION AND COSTS
In this section, we address several reasons why privatization may or may not produce cost savings.
Issues related to the dilemma between competition and monopoly and between public and private
ownership have been emphasized by public choice and property rights literatures. These approaches
consider the incentives available to managers and the role that competition can play in reducing
excessive public supply of public services (Niskanen, 1971), or in providing greater incentives for
cost reduction under private property (Shleifer, 1998), and thus reducing costs. Other approaches
have put more emphasis on the effects of the principal-agent problems associated with privatization.
In this way, transaction costs and industrial organization literatures are more concerned with the
nature of the service, and put more emphasis on the importance of the costs of contracting and
monitoring (Williamson, 1999), the importance of economies of scale (Donahue, 1989), and market
structures (Vickers & Yarrow, 1988). We assess the empirical evidence of privatization’s cost savings
through these theoretical lenses. This reflects our awareness that a comprehensive theoretical
approach that can focus both on actors and on incentives as well as on market and regulatory
structure is required to understand why privatization has not delivered cost savings. However, the
more recent theoretical propositions (outlined below) have not been incorporated in the extant
empirical evidence (especially the earlier studies). Hence, our meta- regression is unable to take full
account of all the theoretical insights discussed below.
Competition Versus Monopoly in Service Delivery
One of the primary ways in which privatization may produce cost savings is by replacing monopoly
with competition in the public services market, which is in line with the belief that competition will
restrict excessive supply of public services and lower costs. Replacing monopoly with competition
can be achieved by assigning contracts to external producers through competitive procedures or by
promoting competition between governmental units (Tiebout, 1956; Osborne & Gaebler, 1992).
Particular consideration to promoting competition is found in the public choice literature,
which gives primary emphasis to incentives and is centered on the view that politicians and
bureaucrats behave like the typical neoclassical individual (Niskanen, 1971). Hence, the central actors
in the government service delivery process seek to maximize their personal utility and interests. If
public service delivery is a monopoly in the hands of politicians and bureaucrats, the result will be an
excess supply of public services and consequently, inefficiency, because the services will be managed
with the objective of extracting material rents and political power (Savas, 1987). These insights
provide a strong rationale for the expectation of cost savings from privatization—provided
competition is present. However, public services are at best quasi-markets with a limited number of
alternative private suppliers (Lowery, 1998; Sclar, 2000; Warner & Hefetz, in press). Competitive
markets rarely exist for public services and this undermines the basis for cost savings. Thus,
governments need to play a role in creating competition in public service markets and in careful
monitoring to ensure cost savings (Warner & Hebdon, 2001; Warner & Hefetz, 2008).
Private Property, Property Rights, and Cost Reducing Innovations
Grossman and Hart (1986) and Hart and Moore (1990) argue that asset ownership gives the owner
control and bargaining power in situations in which the contractor cannot perfectly foresee the
evolution of the activity (Shleifer, 1998). Ownership is an important factor because it confers the
right to obtain the benefits from actions related to the assets such as profit, as well as the benefits
from innovation and efficiency gains. Bureaucrats have control rights under public ownership, but
they do not enjoy property rights, and thus cannot directly benefit from the profits generated by cost
reduction. By contrast, private owners have control rights and can appropriate benefits from cost
reduction (Shleifer & Vishny, 1994).
Hart, Shleifer, and Vishny (1997) apply the theory of incomplete contracts and property rights to
the choice between public or private production of public services. Their study suggests that under
private production, incentives exist to reduce costs at the expense of quality. Under this framework,
incentives work as follows:
With private ownership, the manager has incentives to reduce costs through quality
deterioration. The manager does not need authorization from the government, which will
bear the political costs of quality reduction. To give the manager incentives to innovate to
increase quality, the manager would need to negotiate price increases with the government
to ensure compensation for his investment. Most likely, this negotiation will not result in a
full appropriation of benefits from the innovation, which reduces the manager’s incentives
Under government ownership, incentives work in the opposite direction. Because the
manager is government-employed, he will take into account potential quality erosion when
considering the implementation of cost reducing innovations. In addition, the public
manager will need government permission for any innovation he wants to undertake
(either quality improvement or cost reduction). In the absence of a pay-for-performance
scheme, the public manager will not fully benefit from the results of innovation.
Overall, private ownership offers more incentives for cost reduction, but these incentives can induce
quality erosion. Ensuring quality under privatization requires increased oversight, which can blur the
line between public and private ownership (Guttman, 2000; Bozeman, 1987). As the difference
between public and private ownership disappears, the potential for cost savings from private
ownership may disappear as well.
Principal-agent Problems, Transaction Costs, and Market Structure Conditions
According to the seminal work by Ronald Coase (1937), transactions occur inside the firm when
market transactions incur higher costs than internal ones. The transaction costs approach takes the
choice to “make” or “buy” within a private firm framework and applies it to government decisions
concerning public services delivery. Following Williamson (1999), transactions have three basic
dimensions: (1) uncertainty regarding how the transaction develops, and its results; (2) the frequency
with which transactions are repeated; and (3) the relative requirement of long-term investments
specifically related to the transaction, or sunk costs. Because of these factors, the institutional
organization required to establish and apply contracts can be very complex.
The theoretical analysis of privatization and contracting out uses the concept of transaction
costs in a broad sense, which includes administrative costs as well as costs from incomplete
contracts. In their theoretical analysis of the choice between public and private production,
Sappington and Stiglitz (1987) argue the main factor explaining the choice of production form is a
function of the transaction costs derived from the delegation of authority. Monitoring and control play
a central role, and cost minimization refers to both the production and transaction costs implied by
contracting out. Cost savings are likely to emerge when transactions costs are not great. Thus,
depending on the characteristics of the specific service (with respect to the three dimensions above),
the likelihood of savings will vary.
Stein (1990) used this approach to classify local government services and assess the form of
delivery. Transactions costs have been used to explain government choice in the decision to contract
out (Nelson, 1997; Sclar, 2000; Hefetz & Warner, 2004, 2007). While some authors downplay
contracting costs and argue that the costs of bureaucracy are higher (Savas, 1987; Eggers & O'Leary,
1995; Osborne & Plastrick, 1997), others find transactions costs to be significant factors in
explaining decisions to privatize or reinternalize production (Ferris & Graddy, 1994; Lowery, 1998;
Kavanagh & Parker, 1999; Brown & Potoski, 2003; Hefetz & Warner, 2004, 2007; Bel & Fageda,
2008; Levin & Tadelis, in press). Cost savings expectations, when considered from this perspective,
are dependent on the nature of the service and local market conditions.
The industrial organization literature shares with the property rights approach the core relevance
given to the relationship between incentives and ownership. However, it also emphasizes the duality
between principals and agents, as in the transaction cost approach. The central problem here is how
incentives might encourage the manager to behave in accordance with the owner’s objectives. When
comparing public and private ownership, the ability to align managerial actions with ownership
objectives is the rationale for differences in efficiency between private and public ownership.
When there is a strong separation between ownership and management, a few key factors can work
as control mechanisms to improve the alignment between the two (Vickers & Yarrow, 1988). According
to the industrial organization literature, private ownership is preferred to public ownership when (1)
owners benefit from devoting time and money to obtaining the information needed for supervision, (2)
firms can be taken over, and (3) firms are at risk of bankruptcy. The way in which the market structure
influences how these three factors work is extremely important. These factors are more common in
competitive markets that are not subject to strong government regulation (Kay & Thompson, 1986;
Vickers & Yarrow, 1988). In markets prone to concentration, the impact of factors that align principal
and agent objectives is weak. In those services characterized by scale economies, market concentration
trends will emerge. The same is true of markets in which strong government regulation persists after
privatization. Dnes (1995) emphasized the potential advantage incumbents enjoy in markets
characterized by long-term specific assets. The industrial organization approach emphasizes the design
of contracts and bids to specify properly the conditions that stimulate dynamic competition and thus
reduce the likelihood of future monopolization (Laffont & Tirole, 1993; Bolton & Dewatripont, 2005).
The industrial organization literature also recognizes that privatization can introduce pressure and
incentives for internal reform in those municipalities that retain public service delivery (Hatry, 1988),
because including privatization within the menu of available alternatives creates pressure on the public
manager similar to that placed on private managers by the risk of bankruptcy (Bel, 2006). In this way,
privatization could stimulate efficiency improvements in public production in those services kept under
The empirical studies we analyze in our meta-regression draw on these different approaches (public
choice, property rights, transaction costs, and industrial organization) to privatization and costs. By
different approaches, we do not imply that they are incompatible or radically distinct. Instead, they differ
in their emphases on ownership, competition, principal agent alignment, and market structure when
analyzing the relationship between privatization and costs. Our data set contains studies of both solid
waste collection and water distribution—two services with different cost structures. Scale economies
(related to output) affect solid waste collection, while density economies (related to population
density) are critical for water distribution. Fixed assets are required to produce both services, but
water distribution has network features associated with a high level of sunk (and specific)
investments. Hence, the transaction costs involved in privatizing should be lower for solid waste
Theoretically, we might expect cost savings in both water and waste if competition were present.
Competition is a factor common to all four theoretical approaches (albeit of lesser relevance in the
case of property rights). However, the emphasis on conditions governing competition and on market
dynamics differs between the approaches.
In the case of waste collection, property rights theory suggests private production can be cheaper
due to incentives to invest in new technologies. Public choice theory emphasizes the benefits of
competition in reducing costs, while transaction cost theory recognizes complete contracts are more
likely in waste collection than in water (because of lower asset specificity and less measurement
difficulty). Industrial organization theory tends to emphasize the benefits of economies of scale but
recognizes the challenges faced by market structures that are prone to concentration. While each of
these theoretical perspectives offers possibilities for cost savings, the mixed empirical evidence may
be explained by limited competition and technological improvement, and the importance of market
3 In a 2002 survey of public managers in local governments in the US, Brown and Potoski (2005) measure
perception of transaction costs from asset specificity and ease of measurement on scales from 1 (low) to 5
(high). The asset specificity of residential and commercial solid waste is at 3 on the scale while water
distribution is at 4. The ease of measurement is 2 for solid waste and 2.4 for water. Thus, both asset specificity
and difficulty of measurement are lower for solid waste services than they are for water distribution. Levin and
Tadelis (in press) build indicators of contract difficulty, as perceived by U.S. city managers, and find that
contract difficulty is above average for water services and below average for waste. Warner and Hefetz (in
press) in a national U.S. 2007 survey of transactions costs and competition also find city managers’
assessments of asset (AS) specificity and difficulty of contract management (Mgmt) are higher for water (AS =
4.5, Mgmt = 3.5) than for waste (AS = 2.9, Mgmt = 2.2). In addition, they measure levels of competition in
local government service markets and find that competition for waste is above average at 2.6 providers, while
competition for water service is less than one (0.8).
In the case of water distribution, the theoretical predictions are less optimistic. From property
rights theory we would expect cost savings, but at the expense of service quality, though if
governments maintain careful quality regulation, cost savings through quality reduction in water are
unlikely. Competition is harder to achieve due to the fixed network infrastructure on which water
delivery depends. As a result, private contracts are usually of longer duration than in other services,
and this places incumbents in a strong position in the event of a new tendering process. Transaction
cost theory predicts problems due to the long duration of contracts and the large (and sunk)
investments that create a high degree of incompleteness in water contracts. Industrial organization
theory emphasizes that sunk costs prevent competitive discipline among private providers. Thus, the
theoretical basis for cost savings under private production is weaker than it is for waste distribution.
Summing up, two main propositions derived from our theoretical expectations are (1) systematic
cost savings with private production are not expected due to problems related to competition and
market dynamics, and (2) cost savings with private production will be less frequent for water service
than waste collection, due to water service’s higher transaction costs (from incomplete contracts) and
higher quality protections (preventing quality erosion).
Empirical Background on Lack of Costs Savings
We provide a brief review of the empirical literature to explore potential reasons for the lack of cost
savings from privatization. To do this, we focus on the empirical works included in the meta-
regression that have found either higher costs with private production or no significant difference of
costs between public and private delivery.4
Papers published in the 1970s and early 1980s in the US offer ad-hoc explanations for the lack
of differences in costs between public and private delivery. Mann and Mikesell’s (1976, p. 1003)
analysis on water suggests that scale of operations is much more important than ownership regarding
efficiency. Bruggink (1982, p. 121) offers two basic explanations for the lack of differences in cost:
(1) The operation of government firms within hostile “free enterprise” environments exerts
continuous pressure on public firms’ performance; and (2) public firms attract higher quality
management because of factors like longer job tenure. Regarding solid waste collection, Stevens
(1978, p. 445) finds that private competitive arrangements are more costly than monopoly
4 Some works that do not find lower costs with private delivery do not provide any discussion of potential
reasons for this result. This is the case of Hirsch (1965), Dubin and Navarro (1988), Callan and Thomas
(2001) for solid waste, and of Teeples and Glyer (1987) and Byrnes (1991) for water. Two papers, Kemper
and Quigley (1976) and Collins and Downes (1977)—both on solid waste—provide some discussion on why
private provision (competition in the market) is more costly than municipal provision, but they do not discuss
the differences in costs between public and private delivery within municipal (monopoly) provision.
arrangements (either public or private) and attributes this to higher billing costs and extra expenses
derived from non-exclusivity within the market area born by the firm under private market
arrangements. When comparing public and private monopolistic delivery, Stevens finds that private
delivery is less costly in cities over 50,000 inhabitants, but there are no significant differences in cities
below 50,000. Stevens’ explanation (1978, p. 447) is based on the main differences in productivity
between private and public providers that come from smaller crews with larger capacity vehicles, as
well as lower absentee rates under private delivery. While these differences have a relevant dimension
in larger cities, they become non-relevant in smaller cities below 50,000 inhabitants.
The importance of competition to explain costs differences is stressed in the analyses on solid
waste collection in the United Kingdom published in the mid 1980s and early 1990s. Domberger,
Meadowcroft, and Thompson (1986, pp. 79–80) find that tendering is cheaper than in-house
production, but when contracts are awarded by tender, public and private units do not show
significant differences in costs. They conclude that competition matters much more than ownership
regarding cost savings. Similarly, Szymanski and Wilkins (1993, p. 127) find differences exist between
tendering and in-house production with no tendering, but not between different property regimes
under tendering. The message is the same as in Domberger, Meadowcroft, and Thompson:
Competition is more important than ownership.
Through the current decade many papers have found no differences in costs between public
and private delivery. Discussion of the potential reasons for lack of cost differences has become
richer. Problems related to transaction costs involved in contracting out are emphasized in the
empirical analysis by Kirkpatrick, Parker, and Zhang (2006, pp. 155–158), Bel and Fageda (in press),
and Bae (in press). Competition failures, particularly those derived from concentration in the market
for private providers, are discussed in Dijkgraaf and Gradus (2003, p. 155; 2007, p. 582), Bel and
Costas (2006, p. 20), Bel and Fageda (in press), and Bae (in press). Another factor that may explain
the lack of difference in costs between private and public production is the reform of services
retained under public delivery spurred by the threat of privatization, as in Estache and Rossi (2002, p.
146-147), Bel and Costas (2006, p. 17), Dijkgraaf and Gradus (2007, p. 583), Bel and Mur (2009, p.
2777), and Bae (in press). Finally, Ohlsson (2003, p. 467) points out that municipalities that choose
public delivery have conducted more careful yardstick competition, in the sense that they have done
cost analysis. In all, recent empirical analysis that finds lack of differences stresses the importance of
competition failures due to concentration, high transaction costs associated with contracting out, and
improvement of efficiency in publicly delivered services due to internal reform.
We develop a meta-analysis to provide a systematic test for cost differences across a range of studies
of public and private production. A major objective of a meta-analysis is to provide a statistical
explanation for the differences in results reported in the empirical literature on a given topic (Stanley
& Jarrell, 1989). Meta-analysis provides tests concerning the true effect of the relationship analyzed,
along with tests for possible publication bias (Stanley, 2005b).
Meta-analysis is a statistical technique that has been widely used in the social and medical
sciences. Since the late 1980s, about 150 meta-analyses of empirical studies have been published on a
great variety of issues in economics and economic policy (Connor & Bolotova, 2006). Some relevant
examples include minimum wage effects (Card & Kruger, 1995), the value of air quality (Smith &
Huang, 1995), productivity spillovers of multinational companies (Görg & Strobl, 2001), the value of
life (Mrozek & Taylor, 2002), the effect of immigration on wages (Longhi, Nijkamp, & Poot, 2005),
environmental inequities (Ringquist, 2005), the natural rate hypothesis (Stanley, 2005a), cartel
overcharges (Connor & Bolotova, 2006), voter turnout (Geys, 2006), productivity gains of urban
agglomeration economies (Melo, Graham, & Noland, 2009), and factors explaining local
privatization (Bel & Fageda, 2009).
Economics and public policy studies do not use data collected from controlled experiments so
meta-analysis is implemented differently in those disciplines than in the medical sciences. Stanley and
Jarrell (1989) called the use of meta-analysis in economics “meta-regression analysis” because this
methodology is generally applied to data collected from studies employing regression based
The starting point of our meta-regression analyses is the following equation (Stanley & Jarrell,
bj = β + ΣαkZjk + ej j =1,2,….L (1)
where bj is the reported estimate of β of the jth study, β is the true value of the parameter of
interest5, Zjk are the meta-independent variables that measure relevant characteristics of an empirical
study, and αk are the coefficients associated with those independent variables. In this equation the
5 We mean here by the parameter of interest the coefficient that measures the sign and magnitude of the
relationship analyzed; in our case, this parameter is the t-statistic for the coefficient associated with the
variable of private production in a regression where the dependent variable is the production costs of the
magnitude or significance of the intercept term, β, will determine whether there is a true effect
regarding the relationship analyzed.
The meta-independent variables, Zk, will allow us to examine the influence of different study
characteristics on the results. For example, if year of data collection is significant then we will have
evidence that results of the relationship analyzed are conditioned by the period of analysis.
To tackle any potential problem of heteroscedasticity, Stanley and Jarrell (1989) suggest a
transformation of Equation (1). This transformation implies dividing all the terms in Equation (1) by
the standard error of the coefficient bj. Hence, we obtain the following equation:
Tj = bj/Sb = β/Sb + ΣαkZjk/Sb + ej/Sb j =1,2,….L (2a)
where Tj is the t-statistic of the coefficient bj and Sb is the standard error of the coefficient bj.
Note that the intercept term in Equation (1) and the coefficient associated with the inverse of the
standard error (1/Sb) in Equation (2a) should be the same parameter β. As noted by Stanley (2005b),
the inverse of the standard error may be substituted by sample size. Thus, the parameter that
measures the true effect is now the coefficient associated with the variable of sample size. In this
regard, from an econometric point of view Equation (2a) would lead to the following equation:
Tj = α0 + α1Sample_size + ΣαkZjk/Sb + ε (2b)
This latter Equation (2b) is the basis of our empirical analysis in which the magnitude and
statistical significance of α1 will provide empirical evidence of the relationship examined. Note here
that a standard statistical property is that the magnitude of the t-statistics will vary systematically with
sample size (and degrees of freedom) only if there is in fact a systematic empirical effect.
A major concern in meta-regression analysis is to identify the possible existence of publication
bias. Indeed, papers are more likely to be published when significant relationships between the
variables of interest are found. Studies with statistically significant findings may be more likely to be
published, leading to an incorrect conclusion that a policy is effective, when in fact it is not. To
detect and correct for possible publication bias in a meta-regression, Stanley (2005b, 2007) suggests
the funnel asymmetry test (FAT). The FAT test is based on the statistical property that standard
errors of estimates become smaller as the number of observations in the study increases. Hence,
studies with larger samples are expected to contain smaller publication biases.
The FAT test estimates the relationship between a study’s reported effect and the standard errors
of its coefficients. We estimate the following equation:
Ti = β0 + β1(1/SEi) +εi , (3a)
where T is a study’s reported t-statistic and 1/SE is the inverse of the standard error. Stanley (2005b,
2007) suggests that the statistical significance of the intercept in Equation (3a), β0, is a test for
publication bias and that its sign indicates the direction of this bias. Evidence of publication bias will
be found when β0 ≠ 0. Additionally, the coefficient β1 in Equation (3a) provides an estimate of the
true empirical effect of the parameter of interest. The independent variable, 1/SE, may have some
measurement errors that condition the econometric estimates. Hence, the square root of sample size
is used as an instrument for the inverse of the standard error. Sample size is not subject to estimation
error, and standard errors and sample size are highly correlated. Observed effects should vary
randomly around the true value if publication bias is absent. This will be the case if the intercept
term of Equation (3a) is not statistically significant. Beyond publication bias, some study’s
characteristics may influence the t-statistics obtained. Hence, the FAT test may be embedded into
multivariate FAT-tests where some other explanatory variables related to study’s characteristics are
Ti = β0 + β1(1/SEi) +ΣβkXik+ εi , (3b)
The relationship between a study’s t-statistic and its degrees of freedom using the logarithmic
form can also serve as a meta-significance test (MST) to identify a genuine empirical effect (Stanley,
2005b, 2007). The meta-significance test is based on the statistical property that the magnitude of the
t-statistics will vary systematically with the degrees of freedom if there is in fact an overall genuine
empirical effect. We estimate the following equation:
log(|T|i) = α0 + α1log(df) +εi , (4)
where |T| is a study’s reported t-statistic (in absolute value) and df is the corresponding degrees of
freedom. If we find that α1 = 0, then estimates of α1 will vary randomly around zero and the t-
statistics will not show any clear relationship with the degrees of freedom. By contrast, if we find that
α1 ≠ 0, the observed magnitude of the t-statistics will vary with its degrees of freedom. This would
provide evidence of a systematic effect, as we mention above. In our context, we are interested in the
t-statistics for the coefficient associated with the variable of private production in a regression where
the dependent variable is costs of producing the service. Where α1 = 0 in Equation (4), we have
evidence that the genuine empirical effect is zero.
THE EMPIRICAL STRATEGY
Our empirical analysis has two goals. First, we want to examine the influence of such characteristics
as sample size, time-period, geographic area, and service characteristics on the findings from
empirical studies of the relationship between private production, public production, and costs.
Although a more complete model might also include measures of competition, market dynamics,
transaction costs, and service quality, we are unable to measure these aspects with any degree of
accuracy in our meta-regression because virtually none of the previous studies directly tested for
these effects. Second, we want to analyze the cost effects of privatization and examine whether these
effect are contaminated by possible publication bias.
The Sample of Studies
To the best of our knowledge, the sample used here includes all studies, both published and
unpublished, that use multivariate regression techniques to examine the cost effects of privatizing the
delivery of local solid waste collection or water distribution services. Our meta-regression includes
articles published in academic journals in the fields of economics, political science, and public policy
and public administration. Additionally, we have conducted an extensive search in relevant working
paper series (such as the Social Science Research Network—SSRN).
All the studies in our sample are concerned with publicly provided services supplied either by
municipalities (public production) or via government contracts with private firms (private
production). All use the total or average costs of producing the service as their dependent variable,
and production at the local level as their unit of analysis. Our explanatory variable of interest
regarding costs is the form of service delivery (that is, public or private production).
Individual studies can provide more than one observation in our data if they contain several
estimations containing different data sets or different explanatory variables. Table 1 provides a list of
the 27 studies used in our analysis and their key differentiating characteristics including the t-statistics
for the privatization variable, sample size, period of analysis, and the number of observations each
study contributes to the sample (total observations = 46). Note that the studies included in our
analysis use local municipality data, have a cost variable for which we can use the t-statistic in our
meta-regression, and involve cross-sectional data in a linear regression, log-linear regression, two-
stage estimation, or maximum likelihood analysis.
Insert Table 1 about here
The first econometric study of waste collection (Hirsch, 1965) used a production cost model that
controlled for amount, quality, and service conditions that affect input requirements, factor prices,
technology, density, and form of finance (user fee or general budget). Later studies used similar
models (Collins & Downes, 1977; Kempler & Quigley, 1976), with variables that addressed features
of property rights, transaction costs, and industrial organization theories. Stevens (1978) improved
the model and paid greater attention to scale economies. Dubin and Navarro’s (1988) model was the
first to consider the choice of production and the comparison of costs jointly. Recent studies have
used more comprehensive databases and more sophisticated econometric techniques. Cost savings
were more frequent in the earlier studies (during the 1970s and 1980s), and less so in later studies.
The empirical literature on urban water distribution and costs dates back to the mid-1970s.
Between the mid-1970s and the mid-1990s, econometric studies of this question were limited to the
US. Mann and Mikesell (1976) and Morgan (1977) set up the basic model, focusing on operational
costs, which was adopted in most subsequent studies. Feigenbaum and Teeples (1983) used a
hedonic costs model, and Fox and Hofler (1986) introduced the multi-product characteristic
(production and distribution) of water firms. Subsequent studies in the US followed this latter
approach. Since 2000, the first econometric studies conducted outside the US and the UK have
appeared, covering a wide and mixed set of geographical areas including Estonia and Latvia (Jones &
Mygind, 2000), Asia and the Pacific (Estache & Rossi, 2002) and Africa (Kirkpatrick et al., 2006).
Additional details on all these papers are found in Bel and Warner (2008).
Of the original 35 studies identified by Bel and Warner (2008), we excluded those that did not
use an explicit cost variable and included three additional recent papers on solid waste collection (Bel
& Mur, 2009; Bel & Fageda, in press; Bae, in press). For water, we excluded the works on the UK by
Ashton (2000a, 2000b) and Saal and Parker (2000, 2001) because they examine efficiency using
productivity data at the national level. Interestingly, they do not find significant efficiency
improvements after privatization. We also excluded several studies that estimate a production
function to obtain productivity or efficiency indicators, because their dependent variable is not
consistent with our cost variable target. These works are Feigenbaum and Teeples (1983), Fox and
Hofler (1986), Teeples and Glyer (1987), Link (1993), Bhattacharyya, Parker, and Raffie (1994),
Bhattacharyya, Harris, Narayanan, and Raffie (1995), and Jones and Mygind (2000). None of these
studies finds private production to be significantly less efficient than public production.
For solid waste, we excluded from our meta-regression analysis Szymanski’s (1996) study in the
UK because he jointly considers the effects on costs of public and private ownership and the use of a
competitive tendering process. Szymanski (1996) finds both private and public ownership with
competition are related to cost savings when compared to public production without competition.
However, cost savings obtained from competition erode overtime, and they do so more quickly
under public production than under private production.
Our analysis focuses on cross-sectional studies in which public and private production are
compared. A better measure of the effects of privatization on costs would require the use of time-
series data to account for the change from public to private production with regard to the same unit
of analysis. However, we were only able to identify seven studies (Domberger, Meadowcroft, &
Thompson, 1986; Lynk, 1993; Szymanski, 1996; Szymanski & Wilkins, 1993; Reeves & Barrow 2000;
Dijkgraaf & Gradus, 2007; Bae, in press) that used time-series data, and five of these (all but Lynk,
1993 and Szymanski, 1996) are included in our analysis. However, even these five studies did not
measure the before and after effects of privatization within the same locality, but rather changes over
time across localities.
From this descriptive review, we can see that some of the studies report significant cost savings
from private production (see Table 1). However, we are not able to identify a systematic pattern in
the relationship between private production and costs. If we look in detail at the t-statistics reported,
about half of the studies find statistically significant costs savings with private production, while the
remaining studies find an ambiguous or negative relationship between private production and costs.
Note that most of the evidence for cost savings comes from the early studies conducted back in the
1970s. Since then, few studies report any cost savings from private production for solid waste, and
only Raffie, Narayanan, Harris, Lambert, and Collin (1993) find cost savings for water distribution
services. While private production may lead to cost reductions in some instances, these studies fail to
provide a strong argument in favor of private production.
The Meta-regression Equation
The linear equation that we estimate to examine the influence of different study characteristics on
reported results is as follows:
T = α + β1Sample_size + β2Year + β3Functional_form + β4Service + β5US + β6UK + β7Panel + ε (5)
where the dependent variable measures cost differences under public and private ownership. The
dependent variable is the t-statistic (T) of the dummy variable for private ownership in the
corresponding estimated cost equation. Studies reporting lower privatization t-statistics suggest a less
statistically significant or negative relationship between service costs and private production as
compared to public production. This equation comes from Equation (2b), which was derived in the
Our parameter of interest, T, the dependent variable in Equation (5), is the t-statistic for the
coefficient associated with the variable of private production in a regression where the dependent
variable is total costs of producing the service. If the true effect is a systematic relationship between
private production and lower costs, then that parameter should be negative and different from zero.
To assess if it is different from zero, we must examine the statistical significance of the coefficient
associated with sample size, β1. Recall that the magnitude of the t-statistics will vary systematically
with sample size only if there is in fact a systematic empirical effect.
Furthermore, Equation (5) will allow us to analyze the influence of study characteristics like year,
method, type of service, or country on the t-statistics obtained (that is, on the estimated effect of
private production on costs).
The explanatory variables of the meta-regression equation (Equation 5), sometimes also called
moderator variables, concern particular characteristics of the empirical studies. A negative sign in the
coefficient of an Equation 5 explanatory variable means that studies with a higher value of this
variable are more likely to find cost savings from private production (that is, it increases the
likelihood of finding negative t-statistics or lower positive t-statistics). A positive sign means that a
higher value of this variable makes cost savings less likely.
A common meta-regression moderator variable is the number of observations in the analyses
included in each study, Sample_size, as studies with larger numbers of observations are considered
more robust. Indeed, the estimated effects of the parameter of interest (for example, the effect of
privatization on cost savings) are expected to be closer to their true effects for larger sample sizes. If
the sample size variable is negative and significant then we have some evidence of lower costs under
private production.6 Given that one of the contributions of the meta-regression is to provide
evidence about the genuine empirical effect of the relationship between private production and costs,
we conduct several tests to obtain further insights into this question.
Another moderator variable commonly used in meta-regression analysis is the year in which the
data were collected, Year. We are interested in examining whether recent studies find smaller
differences between public and private ownership than was the case in earlier studies. Recent studies
may, for example, capture rising costs among private producers (Ohlsson, 2003), the potential for
collusion (Dijkgraaf & Gradus, 2007), or efficiency gains under public production (Bel & Costas,
6 Indeed, this would imply that studies with larger sample sizes (whose results should be closer to the true
effect) are obtaining larger negative t-statistics for the coefficient associated with the variable of private
production in the cost equation, and larger negative t-statistics for that coefficient would lead to the
conclusion of systematic lower costs with private production.
2006). However, we cannot make any strong claims regarding dynamics over time because almost all
of the studies in our analysis are cross sectional.
We also account for the role of functional form in explaining variation in the reported results
across studies by using a dummy variable, Functional_form, that takes the value one when the linear
form is used, and zero in other cases (usually the double-log form). This is another common
moderator variable used in meta-regression analyses. However, we do not expect substantial
differences in results from our sample of studies due to the choice of functional form. This choice
might influence the estimated magnitude of the relationship between private production and costs
but not necessarily its statistical significance.
Aggregating studies for solid waste collection and water distribution could result in some
estimation bias as the estimates refer to two different services. We include a dummy variable that
takes the value of one for solid waste collection and zero for water distribution. We also estimate
Equation (5) for the sub-sample of studies that analyze solid waste collection. Eighty percent of the
studies cover solid waste. By looking at the sub-sample of studies on solid waste collection, we can
examine the influence of different characteristics avoiding any distorting effect derived from jointly
considering services with very different cost structures. A distinctive characteristic of water is the
high transaction costs associated with its delivery. We cannot estimate Equation (5) for the sub-
sample of water distribution studies because the subsample is not large enough, and thus we are not
able to isolate the services with high transaction costs.
We also include moderator variables for the geographic area analyzed. We include two dummy
variables that take the value of one when the study refers to United States (US) or the United
Kingdom (UK), respectively. These two countries account for the highest number of observations in
our sample of studies, and are the countries in which these reforms in local public services have been
most widely tested. Note that both countries are characterized by a competitive market framework in
the delivery of local services relative to other countries. Public units in these countries often actively
participate in competitive bidding in their own and neighboring municipalities. In addition, many
local governments in the US mix both public and private production within their jurisdiction as a
way to encourage competition (Miranda & Lerner, 1995; Warner & Hefetz, 2008).
The quality of data used in a cross-sectional analysis may not be as strong as in a longitudinal
analysis, so we include a moderator variable that captures this aspect of data quality. The dummy
variable, panel, takes the value of one for longitudinal studies.
Variables for the intensity of competition or other institutional features aside from type of
ownership would provide additional insights, but measures of these are found only in the studies by
Bel and Costas (2006) and Dijkgraaf and Gradus (2007). It would be interesting to examine the
influence of the multi-product nature of the services analyzed, but most of the studies do not
account for multi-product effects. Note that while we present the current state of knowledge in the
theoretical background (which is cumulative), only the most recent studies (after 2005) reflect the
current propositions regarding transaction costs, competition, or market characteristics. Thus, we
lack enough data to consider these as moderators in our meta-regression.
Results of the Meta-regression Equation
Table 2 presents descriptive statistics for the variables included in the meta-regression equation.
The mean t-statistic is negative indicating that private production typically results in lower costs than
public production, but there is considerable dispersion in the value of the t-statistic. Hence, these
data suggest there may not be significant cost differences between public and private production.
Our sample of studies covers estimations made for the period 1960 to 2005.7
Insert Table 2 about here
Table 3 shows the results from the estimation of the meta-regression equation (Equation (5)) that
comes from Equation (2b) in the methodology section. The estimation is made both for all studies
and for the studies that focus on solid waste. Recall that the statistical significance of the coefficient
of the variable for sample size will allow us to examine the underlying empirical effect in the
relationship between privatization and costs. Additionally, the statistical significance of the
coefficients associated with the variables of year, methods, service, or country will allow us to
examine the influence of these study characteristics on the reported results of the relationship
between private production and costs.
Note that some of the studies contribute more than one observation to the sample because they
include different estimations with different data sets or different explanatory variables. To correct for
possible violation of the uncorrelated errors assumptions between observations from the same study,
we adjust our estimates by clustering observations from the same study.
Insert Table 3 about here
Importantly, the coefficient of the variable for Sample Size is negative but not statistically
significant. This means that larger sample studies are not more likely to report cost savings from
7 When a study uses a data set with a cross-section and time-series form, we compute the first year of the
period for the moderator variable, year.
private production. That is, they are not more likely to find larger negative t-statistics for the
coefficient associated with the variable of private production in their cost equation. The magnitude
of the t-statistics will vary systematically with sample size only if there is in fact a systematic empirical
effect. In other words, results of studies with larger sample sizes should be closer to true effect.
Thus, we do not find clear evidence that the true effect is a systematic relationship between private
production and costs.
From the result for the variable of sample size, we conclude there is no statistical support for an
empirical effect of private production on costs. This is true for studies that analyze solid waste
collection and water distribution and studies that only analyze solid waste collection. In the following
sub-section, we implement several tests that confirm this result.
The coefficient for the variable Year has a positive sign and is statistically significant. This means
more recent studies are less likely to find cost differences between public and private production.
This is true for studies that analyze solid waste collection and water distribution and studies that only
analyze solid waste collection.
The coefficient functional form is not statistically significant; functional form does not explain
why some studies find cost savings and others do not. The coefficient of the dummy variable for
services is negative although its statistical significance is modest. This provides some evidence that
lower costs under private production are more likely in solid waste collection than they are in water
distribution. This may be explained by differences in transaction costs. Privatizing water distribution
involves higher transaction costs for local governments; whereas, privatizing the delivery of solid
waste collection may allow for exploitation of scale economies.8
Our geographic variable shows the US studies are less likely to find differences between public
and private production in both the full sample meta-regression estimation and the solid waste only
estimations. Finally, the coefficient of the dummy variable for longitudinal studies is negative and
significant. Thus, studies that use a data set with a cross-section and time-series form seem to find
more differences between public and private production. However, these studies did not measure the
before and after effects of privatization but rather changes over time across localities.
8 Privatization can facilitate scale economies because private companies can distribute fixed costs efficiently
over several geographic units, as they are not constrained to a single municipality (Bel & Miralles, 2003).
Several studies have found evidence of scale economies in solid waste collection (Stevens, 1978; Callan &
Thomas, 2001; Bel & Costas, 2006; Bel & Fageda, in press), but scale economies are exhausted as population
increases. Inter-municipal cooperation is another alternative for promoting scale economies.
Results of the Meta-regression Tests
As we mention above, publication bias is a major concern of meta-regression analysis because papers
are more likely to be published when significant relationships between the variables of interest are
found (Stanley, 2005b; 2007). In our context, the direction of the bias is likely to be as follows:
Devotees of contracting uncover statistically significant savings (significant negative t-statistics),
while detractors might be satisfied with no savings, rather than finding that diseconomies are
statistically significant.9 As a result, the empirical literature that compares the performance of public
and private production may find larger cost differences than those related to the true effect of
Table 4 shows results of funnel asymmetry tests (FAT) using both the inverse of standard errors
(column 1) and the square roots of the sample size (column 2) as independent variables, and results
of the meta-significance test (MST) are indicated in column 3.
Recall that the FAT test estimates the relationship between a study’s reported effect and the
standard errors of its coefficients (or the square root of sample size). The FAT test comes from
Equation (3a) in the methodology section. We will find evidence of publication bias if the intercept
term is different from zero. Additionally, we will find evidence of a true empirical effect (that is, a
systematic relationship between private production and costs) if the coefficient associated with the
inverse of the standard error (or the square root of sample size) is different from zero. Additionally,
the MST estimates the relationship between a study’s t-statistic and its degrees of freedom using the
logarithmic form. The MST test comes from Equation (4) in the methodology section. We will find
evidence of a true empirical effect in the relationship between private production and costs if the
coefficient associated with degrees of freedom is different from zero.
We do not find any evidence of a genuine empirical effect in the relationship between private
production and costs because our estimates show that the coefficients associated with the inverse of
the standard errors (column 1), the square root of sample size (column 2) or the logarithm of degrees
of freedom (column 3) are not statistically different from zero. Hence, our analysis provides
empirical evidence that private production of local services is not systematically less costly than that
of public production.
Insert Table 4 about here
9 We thank one of the anonymous referees for this insight.
Table 4 presents mixed evidence of publication bias. The FAT test shows that the intercept term
is statistically different from zero when using the inverse of the standard error as an independent
variable (column 1). However, this intercept term is not significant when using the square root of the
sample size (column 2). Recall that the inverse of the standard errors may have measurement errors
that the square root of the sample size does not have. Hence, we should be more confident with the
results of the estimation using the square root of the sample size as the independent variable. If there
is publication bias, the direction is negative as expected; the studies are obtaining overly large cost
savings—negative t-statistics—of private production from a statistical point of view. This evidence
of publication bias, somewhat equivocal as it is, strengthens our finding that a strong association
between cost savings and private production is not justified.
Concerning the explanatory variables of the meta-regression, publication bias may be filtered by
estimating a multivariate FAT-meta-regression model (Stanley, 2005b). The multivariate FAT-meta
regression model comes from Equation (3b) in the methodology section. In practice, this will imply
re-estimating Equation (5) to include the inverse of standard errors or the square root of sample size
instead of sample size as the independent variable. This allows us to test whether publication bias
distorts our results in relation to the other study characteristics we use as explanatory variables. In
addition, this provides additional tests of publication bias. The estimated effects should vary
randomly around the true value if publication bias is absent, and this will be the case if the intercept
term of the FAT-meta-regression model is not statistically significant. Additionally, the FAT-meta-
regression model includes other explanatory variables (for example, study characteristics) that may
influence the reported effects. This enables us to obtain results for these explanatory variables given
the value of the intercept term that measures any possible publication bias.
Table 5 shows the results of these estimates. Our previous results in relation to the moderator
variables in Equation (1) are confirmed since the sign and statistical significance of each explanatory
variable does not change with the new estimation. In fact, the variable, service, is now statistically
significant at the 10 percent level (not just at the 11 percent level) when using the inverse of standard
errors as an explanatory variable. Furthermore, we find stronger evidence of publication bias—that
the empirical literature comparing the performance of public and private production may find larger
negative t-statistics than those related to the true effect of privatization.
Insert Table 5 about here
The advantage of a meta-regression analysis is that it allows us to determine the overall impact of
phenomena across a wide range of studies. Previous narrative meta-analyses of cost differences
between public and private production have reported mixed findings (Boyne, 1998; Hirsch, 1995;
Hodge, 2000; Bel & Warner, 2008). Our meta-regression analysis allows us to determine the impact
of certain variables that account for these differences across studies.
In the studies we examine, the average t-statistic for privatization is negative, implying lower
costs on average for private production. However, this result does not hold up in the more rigorous
meta-regression analysis once we have controlled for sample size and other factors. Indeed, our
results can be seen as another warning that studies conducted with small sample sizes be interpreted
with caution. Moreover, we find some evidence of publication bias, which means that papers
obtaining significant cost savings are more likely to be published. We also find cost differences to be
less likely in more recent studies. As expected, cost savings are more likely in solid waste collection
than in water distribution services given the lack of competition and higher asset specificity of the
latter. Our results also show that lower costs are dependent on service characteristics, geographic
area, and time-period of the study. Transaction costs are of significance in that they are responsible
for the greater likelihood of savings for solid waste collection than for water distribution. Many
public services are natural monopolies with high asset specificity, as in the case of water distribution,
and private production in these cases is unlikely to yield cost savings.
The major theoretical approaches to privatization suggest that competition can have a positive
effect on cost savings. Unfortunately, our meta-regression analysis is unable to capture properly the
impact of competition on cost differences between private and public production. Our results
indicate that the more competitive policy environment in the US in comparison to that in other
countries reduces the likelihood of finding cost savings with private production. It could well be that
the primary benefit from opening up competition for public services is the efficiency improvements
this generates among public producers as a result of benchmarking pressure from potential private
Public choice, property rights, transaction costs, and industrial organization approaches all
contribute to our understanding of privatization, but individually each one only captures part of the
costs and service delivery equation. Competition, market dynamics, transactions costs, and service
quality are also important, but the existing empirical evidence does not allow us to understand their
impact fully. Policymakers should pay careful attention to these factors when considering the choice
between public or private production. Future empirical analyses should give increased emphasis to
the characteristics of the service (potential for technological improvement, sunk costs), the
characteristics of the contract (specification and monitoring, transaction costs), the characteristics of
the market (competition), and the managerial and policy environment. These are the priorities for
Ashton, J. K. (2000a). Cost efficiency in the UK water and sewerage industry. Applied Economics Letters, 7,
Ashton, J. K. (2000b). Total factor productivity growth and technical change in the water and sewerage
industry. Service Industries Journal, 20, 121–130.
Bae, S. (in press). Public versus private delivery of municipal solid waste services: The case of North Carolina.
Contemporary Economic Policy.
Bel, G. (2006). Economía y política de la privatización local. Madrid: Marcial Pons.
Bel, G., & Costas, A. (2006). Do public sector reforms get rusty? Local privatization in Spain. Journal of
Policy Reform, 9, 1–24.
Bel, G., & Fageda, X. (2007). Why do local governments privatize public services? A survey of empirical
studies. Local Government Studies, 33, 517–534.
Bel, G., & Fageda, X. (2008). Reforming the local public sector: Economics and politics in privatization of
water and solid waste. Journal of Economic Policy Reform, 11, 45–65.
Bel, G., & Fageda, X. (2009). Factors explaining local privatization: A meta-regression analysis. Public Choice,
Bel, G., & Fageda, X. (in press). Empirical analysis of solid management waste costs: Some evidence from
Galicia, Spain. Resources, Conservation and Recycling.
Bel, G., & Miralles, A. (2003). Factors influencing privatisation of urban solid waste collection in Spain,
Urban Studies, 40, 1323–1334.
Bel, G., & Mur, M. (2009). Intermunicipal cooperation, privatization and waste management costs: Evidence
from rural municipalities, Waste Management, 29, 2772–2778.
Bel, G., & Warner, M. (2008). Does privatization of solid waste and water services reduce costs? A review of
empirical studies. Resources, Conservation and Recycling, 52, 1337–1348.
Bhattacharyya, A., Parker, E., & Raffie, K. (1994). An examination of the effect of ownership on the relative
efficiency of public and private water utilities. Land Economics, 70, 197–209.
Bhattacharyya, A., Harris, T. R, Narayanan, R., & Raffie, K. (1995). Specification and estimation of the effect
of ownership on the economic efficiency of water utilities. Regional Science and Urban Economics, 25,
Bolton, P., & Dewatripont, M. (2005). Contract theory. Cambridge, MA: MIT Press.
Boyne, G. A. (1998). Bureaucratic theory meets reality: Public choice and service contracting in US local
government. Public Administration Review, 58, 474–484.
Bozeman, B. (1987). All organizations are public: Bridging public and private organizational theories. San
Francisco, CA: Jossey Bass.
Brown, T. L., & Potoski, M. (2003). Managing contract performance: A transaction cost approach. Journal of
Policy Analysis and Management, 22, 275–297.
Brown, T. L., & Potoski, M. (2005). Transaction costs and contracting: The practitioner perspective. Public
Performance & Management Review, 28, 326–351.
Bruggink, T. H. (1982). Public versus regulated private enterprise in the municipal water industry: A
comparison of operating costs. Quarterly Review of Economics and Business, 22, 111–125.
Byrnes, P. (1991). Estimation of cost frontiers in the presence of selectivity bias: Ownership and efficiency of
water utilities. Advances in Econometrics, 9, 121–137.
Callan, S. J., & Thomas, J. M. (2001). Economies of scale and scope: A cost analysis of municipal solid waste
services. Land Economics, 77, 548–560.
Card, D., &Krueger A. B. (1995). Time-series minimum wage studies: A meta-analysis. American Economic
Review, 85, 238–243.
Coase, R. H. (1937). The nature of the firm. Economica, 4, 386–405.
Collins, J. N., & Downes, B. T. (1977). The effect of size on provision of public services: The case of solid
waste collection in smaller cities. Urban Affairs Quarterly, 12, 333–347.
Connor, J. M., & Bolotova, Y. (2006). Cartel overcharges: Survey and meta-analysis. International Journal of
Industrial Organization, 24, 1109–1137.
Crain, W. M., & Zardkoohi, A. (1978). A test of the property-rights theory of the firm: Water utilities in the
United States. Journal of Law and Economics, 21, 395–408.
Davies, S. (2007). Politics and markets: The case of UK municipal waste management. (Working Paper #95)
Cardiff, Wales, UK: School of Social Sciences, Cardiff University.
Dijkgraaf, E., & Gradus, R. H. J. M. (2003). Cost savings of contracting out refuse collection. Empirica, 30,
Dijkgraaf, E., & Gradus, R. H. J. M. (2007). Collusion in the Dutch waste collection market. Local
Government Studies, 33, 573–588.
Dijkgraaf, E., & Gradus, R. H. J. M. (2008). Institutional developments in the Dutch waste collection market.
Government & Policy, 26, 110–126.
Dnes, A.W. (1995). Franchising and privatization. Washington DC: The World Bank.
Domberger, S., Meadowcroft, S. A., & Thompson, D. J. (1986). Competitive tendering and efficiency: The
case of refuse collection. Fiscal Studies, 7 (4), 69–87.
Donahue, J. D. (1989). The privatization decision. Public ends, private means. New York, NY: Basic Books.
Dubin, J. A, & Navarro, P. (1988). How markets for impure public goods organize: The case of household
refuse collection. Journal of Law, Economics & Organization, 4, 217–241.
Eggers, W. D., & O'Leary, J. (1995). Revolution at the roots: Making our government smaller, better, and
closer to home. New York, NY: The Free Press.
Estache, A., & Rossi, M. A. (2002). How different is the efficiency of public and private water companies in
Asia? World Bank Economic Review, 16, 139–148.
Feigenbaum, S., & Teeples, R. (1983). Public versus private water delivery: A hedonic cost approach. Review
of Economics and Statistics, 65, 672–678.
Ferris, J., & Graddy, E. (1994). Organizational choices for public service supply. Journal of Law, Economics
& Organization, 10, 126–141.
Fox, W. F., & Hofler, R. A. (1986). Using homothetic composed error frontiers to measure water utility
efficiency. Southern Economic Journal, 53, 461–477.
Geys, B. (2006). Explaining voter turnout: A review of aggregate-level research. Electoral Studies, 25, 637–
Görg, H, & Strobl, E. (2001). Multinational companies and productivity spillovers: A meta-analysis. Economic
Journal, 111, 723-740.
Grossman, S.J., & Hart, O. D. (1986). The costs and benefits of ownership: A theory of vertical and lateral
integration. Journal of Political Economy, 94, 691–719.
Guttman, D. (2000). Public purpose and private service: The twentieth century culture of contacting out and
the evolving law of diffused sovereignty. Administrative Law Review, 52, 859–926.
Hart, O. D., & Moore, J. (1990). Property rights and the nature of the firm. Journal of Political Economy, 98,
Hart, O. D., Shleifer, A., & Vishny, R. W. (1997). The proper scope of government: Theory and an
application to prisons. Quarterly Journal of Economics, 112, 1127–1161.
Hatry, H. P. (1988). Privatization presents problems. National Civic Review, 77, 112–117.
Hefetz, A., & Warner, M. E (2004). Privatization and its reverse: Explaining the dynamics of the government
contracting process. Journal of Public Administration, Research and Theory, 14, 171–190.
Hefetz, A., & Warner, M. E (2007). Beyond the market vs. planning dichotomy: Understanding privatisation
and its reverse in US cities. Local Government Studies, 33, 555–572.
Hirsch, W.Z. (1965). Cost functions of an urban government service: refuse collection. Review of Economics
and Statistics, 47, 87-92.
Hirsch, W. Z. (1995). Contracting out by urban governments: A review. Urban Affairs Review, 30, 458–472.
Hodge, G. (2000). Privatization. An international review of performance. Boulder, CO: Westview Press.
Jones, D. C., & Mygind, N. (2000). The effects of privatization on productive efficiency: Evidence from the
Baltic republics. Annals of Public and Cooperative Economics, 71, 415–439.
Kavanagh, I., & Parker, D. (1999). Managing the contract: A transaction cost analysis of externalisation. Local
Government Studies, 26, 1–22.
Kay, J. A., & Thompson, D. J. (1986). Privatisation: A policy in search of a rationale. Economic Journal, 96,
Kemper, P., & Quigley, J. (1976). The economics of refuse collection. Cambridge, MA: Ballinger.
Kirkpatrick, C., Parker, D., & Zhang, Y. Z. (2006). State versus private sector provision of water services in
Africa. World Bank Economic Review, 20, 143–163.
Kitchen, H. M. (1976). A statistical estimation of an operating cost function for municipal refuse collection.
Public Finance Quarterly, 4, 56–76.
Laffont, J. J., & Tirole, J. (1993). A theory of incentives in procurement and regulation. Cambridge (MA): MIT
Levin, J., & Tadelis, S. (in press). Contracting for government services: theory and evidence from U.S. cities,
Journal of Industrial Economics.
Longhi, S., Nijkamp, P., & Poot, J. (2005). A meta analytic assessment of the effect of immigration on wages.
Journal of Economic Surveys, 19, 451–477.
Lowery, D. (1998). Consumer sovereignty and quasi-market failure. Journal of Public Administration Research
and Theory, 8, 137–172.
Lynk, E. L. (1993). Privatisation, joint production and the comparative efficiency of private and public
ownership: The UK water industry case. Fiscal Studies, 14 (2), 98–116.
Mann, P. C., & Mikesell, J. L. (1976). Ownership and water system operation. Water Resources Bulletin, 12,
Melo, P. C, Graham, D. J., & Noland, R. B. (2009). A meta-analysis of estimates of urban agglomeration
economies. Regional Science and Urban Economics, 39, 332–342.
Miranda, R., & Lerner, A. (1995). Bureaucracy, organizational redundancy, and the privatization of public
services. Public Administration Review, 55 , 193–200.
Morgan, W. D. (1977). Investor owned vs. publicly owned water agencies: An evaluation of the property
rights theory of the firm. Water Resources Bulletin, 13, 775–781.
Mrozek, J. R., & Taylor, L. O. (2002). What determines the value of life? A meta-analysis. Journal of Policy
Analysis and Management, 21, 253–270.
Niskanen, W. A. (1971). Bureaucracy and representative government. Chicago, Il: Aldine.
Nelson, M. A. (1997). Municipal government approaches to service delivery: An analysis from a transactions
cost perspective. Economic Inquiry, 35, 82–96.
Ohlsson, H. (2003). Ownership and production costs. Choosing between public production and contracting-
out in the case of Swedish refuse collection. Fiscal Studies, 24 (4), 451–476.
Osborne, D., & Gaebler, T. (1992). Reinventing government: How the entrepreneurial spirit is transforming
government. Reading, MA: Addison-Wesley.
Osborne, D., & Plastrick, P. (1997). Banishing bureaucracy. Reading, MA: Addison-Wesley.
Pommerehne, W. W., & Frey, B. (1977). Public versus private production efficiency in Switzerland: A
theoretical and empirical comparison. Urban Affairs Annual Review, 12, 221–241.
Raffie, K., Narayanan, R., Harris, T. R., Lambert, D., & Collins, J. M. (1993). Cost analysis of water utilities: A
goodness-of-fit approach. Atlantic Economic Journal, 21 (3), 18–29.
Ringquist, E. J. (2005). Assessing evidence of environmental inequities: A meta-analysis. Journal of Policy
Analysis and Management, 24, 223–247.
Reeves, E., & Barrow, M. (2000). The impact of contracting-out on the costs of refuse collection services. The
case of Ireland. Economic and Social Review, 31, 129–150.
Saal, D. S., & Parker, D. (2000). The impact of privatization and regulation on the water and sewerage
industry in England and Wales: A translog cost function model. Managerial and Decision Economics, 21,
Saal, D. S., & Parker, D. (2001). Productivity and price performance in the privatized water and sewerage
companies of England and Wales. Journal of Regulatory Economics, 20, 61–90.
Sappington, D. E. M., & Stiglitz, J. E. (1987). Privatization, information and incentives. Journal of Policy
Analysis and Management, 6, 567–582.
Savas, E. S. (1987). Privatization. The key to better government. Chatham, NJ: Chatham House.
Sclar, E. (2000). You don’t always get what you pay for. The economics of privatization. Ithaca, NY: Cornell
Shleifer, A. (1998). State versus private ownership. Journal of Economic Perspectives. 12 (4), 133–150.
Shleifer, A., & Vishny, R. W. (1994). Politicians and firms. Quarterly Journal of Economics, 109, 995–1025.
Smith, V. K, & Huang, J. C. (1995). Can markets value air quality? A meta-analyses of hedonic property value
models. Journal of Political Economy, 103, 209–227.
Stanley, T. D. (2005a). Integrating the empirical tests of the natural rate hypothesis: A meta-regression
analysis. Kyklos, 58, 611–634
Stanley, T. D. (2005b). Beyond publication bias. Journal of Economic Surveys, 19, 309–345.
Stanley, T. D. (2007). Meta-regression methods for detecting and estimating empirical effects in the presence
of publication selection. Oxford Bulletin of Economics and statistics, 70, 103–127.
Stanley, T. D., & Jarrell, S. B. (1989). Meta-regression analysis: A quantitative method of literature surveys.
Journal of Economic Surveys, 3, 54–67.
Stein, R. M. (1990). Urban alternatives: Public and private markets in the provision of local services.
Pittsburgh, PA: University of Pittsburgh Press.
Stevens, B. J. (1978). Scale, market structure, and the cost of refuse collection. Review of Economics and
Statistics, 60, 438–448.
Szymanski, S. (1996). The Impact of compulsory competitive tendering on refuse collection services. Fiscal
Studies, 17 (3), 1–19.
Szymanski, S., & Wilkins, S. (1993). Cheap rubbish? Competitive tendering and contracting out in refuse
collection. Fiscal Studies, 14 (3), 109–130.
Teeples, R., & Glyer, D. (1987). Cost of water delivery systems: Specifications and ownership effects. Review
of Economics and Statistics, 69, 399–408.
Tickner G., & McDavid, J. C. (1986). Effects of scale and market structure on the costs of residential solid
waste collection in Canadian cities. Public Finance Quarterly, 14, 371–393.
Tiebout, C. (1956). A pure theory of local public expenditures. Journal of Political Economy, 64, 416–424.
Vickers, J., & Yarrow, G. (1988). Privatization: An economic analysis. Cambridge, MA: MIT Press.
Warner, M. E., & Bel, G. (2008). Competition or monopoly? Comparing US and Spanish privatization. Public
Administration: An International Quarterly, 86, 723–735.
Warner, M. E., & Hebdon, R. (2001). Local government restructuring: Privatization and its alternatives.
Journal of Policy Analysis and Management, 20, 315–336.
Warner, M. E., & Hefetz, A. (2008). Managing markets for public service: The role of mixed public/private
delivery of city services. Public Administration Review, 68, 150–161.
Warner, M. E. & Hefetz, A. (in press). Service characteristics and contracting: The importance of citizen
interest and competition. In The municipal year book 2010. Washington, DC: International City County
Williamson, O. E. (1999). Public and private bureaucracies: A transaction cost economics perspective. Journal
of Law, Economics, & Organization, 15, 306–342.
Table 1. Characteristics of Studies Included in the Meta-Regression Analysis
-0.8 24 1960
-7.4+ 48 1970
Kemper & Quigley (1976) -1.7/-10.7+ 15/57
Man & Mikesell (1976) 1.9 214
Collins & Downes (1977) 1.5 53
Country Regression method Number
Linear regression &
Two stage estimation
Panel & OLS
OLS, OLS corrected
Pommerehne & Frey (1977) -2.8/-4.2+ 103 1970 Switzer. 2 Solid waste
Crain & Zardkoohi (1978)
Tickner & McDavid (1986)
Domberger et al. (1986)
Dubin & Navarro (1988) -0.5 261 1974 USA
1 Solid waste
Byrnes (1991) 0.6 154 1976 USA 1 Water
Szymanski & Wilkins (1993) -1.7/-2.7+ 165/3351984-88 UK 5 Solid waste
Raffie et al. (1993)
Reeves & Barrow (2000)
Callan & Thomas (2001)
Estache & Rossi (2002)
Dijkgraaf & Gradus (2003)
Bel & Costas (2006)
Kirkpatrick et al (2006)
Dijkgraaf & Gradus (2007) -2.3+ 491 1998-2005 Holland1 Solid waste
Bel & Mur (2009)
Bel & Fageda (in press)
Bae (in press)
Note: In the column of t-statistics, the superindex + marks those studies in which at least one of the estimations finds lower costs with
private production with a statistical significance below 10 percent.
Table 2. Summary statistics
Continuous Variables Mean Standard
Number of observations (total = 46)
Variable = 1
Functional_form (1 = linear,
0 = other)
Service (1 = Solid waste,
0 = water)
Variable = 0
Table 3. Meta-regression estimates (OLS)
Dependent variable: Cost differences between public and private
Note 1: Standard errors in parentheses (robust to heteroskedasticity and adjusted by correlation
between observations in the same study).
Note 2: Significance at the 1 percent (***), 5 percent (**), 10 percent (*) level.
Note 3: The variable of service is significant at the 11 percent level in the estimation that considers the
Studies for solid waste
F (joint sig.)
33 Download full-text
Table 4. Meta-regression tests (OLS)
Dep variable: t-
Dep variable: t-
Dep variable: log (t-statistic
in absolute values)
sqrt (sample size)
Note 1: Collins & Downes’ (1977) study cannot be included in some tests since neither coefficients nor
standard errors are reported.
Note 2: Standard errors in parentheses (robust to heteroskedasticity and adjusted by correlation between
observations in the same study).
Note 3: Significance at the 1 percent (***), 5 percent (**), 10 percent (*) level.
Table 5. FAT-Metaregression estimates (OLS)
Mod. Variables Full Sample
Model 1: 1/SE Model 2: sqrt
1/SE -0.006 (0.0049) -
sqrt (sample size) - -0.003 (0.12)
Year 0.11 (0.03)*** 0.11 (0.033)***
Functional_form -0.92 (1.03) -0.64 (1.06)
Service -1.73 (0.99)* -1.50 (0.90)
USA 1.65 (0.87)* 1.86 (0.90)**
UK 0.59 (0.72) 0.58 (1.18)
Panel -2.80 (0.61)*** -2.87 (0.82)***
Intercept -228.14 (67.03)*** -221.11 (65.66)***
F (joint sig.)
Note 1: Standard errors in parentheses (robust to heteroskedasticity and adjusted by correlation between
observations in the same study).
Note 2: Significance at the 1 percent (***), 5 percent (**), 10 percent (*) level.
Note 3: The variable of service is significant at the 11 percent level in the estimation that uses the square
root of the sample size as explanatory variable.
Studies for solid waste
Model 1: 1/SE Model 2: sqrt