Post on 27-Jul-2020
DIPARTIMENTO DI SCIENZE ECONOMICHE AZIENDALI E STATISTICHE
Via Conservatorio 7 20122 Milano
tel. ++39 02 503 21501 (21522) - fax ++39 02 503 21450 (21505) http://www.economia.unimi.it
E Mail: dipeco@unimi.it
THE REFORM OF NETWORK INDUSTRIES,
PRIVATISATION AND CONSUMERS’ WELFARE:
EVIDENCE FROM THE EU15
CARLO V. FIORIO MASSIMO FLORIO
Working Paper n. 2009-41 OTTOBRE 2009
1
The reform of network industries, privatisation and consumers’ welfare: evidence from the EU15
Carlo V. Fiorio and Massimo Florio
Department of Economics, Business and Statistics, Università di Milano and Econpubblica; Department of Economics, Business and Statistics, Università di Milano
1. 0B0BINTRODUCTION
In the past two decades privatisation and liberalisation of network industries providing
services of general economic interest (SGEI), such as electricity, gas, fixed and mobile
telephony, water, and railways, have been particularly significant in the European Union.
The key feature of these industries is that they include both a major fixed-cost
component, the network, under increasing returns to scale, and several potentially
competitive upstream or downstream operations. This feature leads to natural monopoly
for the network services, and potential market dominance of the vertically integrated
network owner. Before the 1970s wide disappointment with earlier private monopolies
or oligopolies, sheltered by various forms of long-term concessions, led most European
governments to take control of industries often plagued by collusion, underinvestment
and price discrimination. Millward (2005) offers a masterly account of the often more
pragmatic than political reasons behind nationalisation and consolidation of energy,
telecoms and transport in Europe in the last century. Historically, the industry structure
preferred by most governments was a vertically integrated public monopoly. Networks
and operations were nationalised at the same time. Tariffs were strictly regulated and
2
included cross-subsidies for some segments of the residential users (low-income, rural or
other disadvantaged areas, etc.). After World War II service provision and investment in
the networks under public ownership increased significantly in many countries,
particularly in energy and telecoms. In the 1970s, however, there was a dramatic policy
reversal. In the mid-1980s the United Kingdom was the front-runner of reform, while,
among the EU Member States, France has often been regarded as a country averse to
moving away from public monopoly. In fact, in the last 15-20 years virtually all
European countries have undertaken dramatic regulatory reforms in the network
industries. Wide variations around a common policy trend can, however, be observed
across countries and sectors. This leads us to treat this story as an experiment in policy
reform, and to study its welfare effects on users. We focus on a simple and
straightforward research question: “Do consumers benefit from the reform of network
industries as it has been designed and implemented in the EU?”
A typical ‘European-style’ reform package has four main dimensions: divestiture of
public ownership; unbundling of the network from service operations; price regulation
by an independent office (usually in the form of price capping); lifting of restrictions to
market entry and finally full liberalisation. According to some early views (Beesley and
Littlechild, 1983, 1989), price controls had to be considered as a transitory mechanism to
protect the consumer before full liberalisation, so that only generic anti-trust vigilance
was needed at the end of the process.
In general, the EU institutions have been strongly supportive of the reforms. While
neutral on public ownership divestiture, over the years the European Commission has
proposed a number of important directives on transport and energy sectors that push the
Member States towards a homogenous pattern of regulatory legislation, see e.g. CEC
(2007). As for telecommunications, the European Parliament recently approved
(September 26, 2008) a compulsory unbundling of local telephone loops, a significant
change for most incumbent operators. Other directives have involved postal services,
water, airports and local transport. A new paradigm has emerged, and legislation is
shaped by it across Europe. We want to test the paradigm on empirical grounds,
disentangling the effect of privatisation from that of other reforms.
We consider EU15 only, because data for the New Member States are less reliable, the
time series are shorter, and privatisation and regulatory change from former planned
economies is less comparable with change in industrial organisation elsewhere. We
consider three industries (electricity, gas, and telecom), each representing different
stages and features of the reform story. This is a subset of all the network industries, but
the core ones for most consumers.FF
1FF Moreover, government-owned providers in energy
and telecoms were not loss makers, their prices covered costs, and comparison with
pricing of private firms is easier.
The welfare calculation involved in the empirical analysis of any regulatory reform is
rather complex. In principle, it would involve a social cost-benefit analysis of outcomes
for different agents, such as consumers, tax-payers, workers, suppliers, competitors,
1 Water services would have been another good candidate, but in general they are unreformed in the EU in the years we consider.
3
investors, etc, see e.g. Galal et al. (1994), Newbery and Pollitt (1997), Florio (2004).
This approach needs to build a counter-factual scenario where states-of-the-world ‘with’
and ‘without’ the reform are compared. This research line is very demanding in terms of
data and assumptions. Hence, it is probably more appropriate for specific industry or
country case studies. An alternative empirical approach is to take advantage of cross-
country variability of the reform pattern and trends. To do this, after having identified
the hypotheses to be tested, we need (country by country) a standardised, comparable set
of reform indicators, such as those provided in the OECD Indicators of Regulation in
Energy, Transport and Communications (henceforth ECTR, and formerly known as
REGREF), a database recently (April 2009) released by the OECD, and covering years
1975-2007.
Under this approach, one outcome variable is selected, e.g. productivity, investment or
wages and, after considering country- or industry-specific control variables, a set of
predictions about the impact of the reforms is tested. For example, Azmat et al. (2007)
use REGREF variables to test the impact of privatisation and market liberalisation on
employment and wages in several network industries in the OECD. Alesina et al. (2005)
also use REGREF data for OECD countries to study product market regulation and
investment. Nicoletti and Scarpetta (2003) use REGREF to study privatisation,
liberalisation and productivity growth patterns across countries, particularly between
continental Europe, the US and the UK. We review other studies in later sections of this
paper.
We contribute to this strand of the literature by shifting the research focus to objective
and subjective evidence on consumer prices. After all, we think that the main
justification of a regulatory reform in SGEI should lie in the potential benefits for users.
Previous empirical research shows that it is unlikely that there is a positive net benefit of
a policy reform if consumers do not get a fair dividend from it, see e.g. Ugaz and
Waddams Price (2003). Moreover, if consumers do not perceive any net benefit, at a
certain stage they can shift from indifference to opposition, as happened with similar
reforms in Latin America, see e.g. Checchi et al. (2009).
Under very general assumptions consumer surplus changes are inversely related to
price changes. Hence, focussing on consumer prices is a sensible shortcut for welfare
analysis. While there is no comparable information on the prices paid by individual
households in Europe, we consider both actual average price changes, as recorded by
Eurostat, and perceived price fairness by individual users, as recorded by Eurobarometer
surveys. Hence, we can split our research question into two: Are average consumer
prices lower (after controlling for non-reform related factors) in countries that
implemented privatization and liberalization reforms? Are consumers happier with the
price they pay in the latter countries (after controlling for individual characteristics and
country features)?
Clearly, if the answer is ‘yes’ to both questions, the reform is on a promising track,
because it is supported by objective evidence and it enjoys public support at the same
time. If prices are low, but perceptions are less positive, reformers should carefully
consider why people are unhappy. If there is no evidence that prices are lower and that
4
people are happier in countries that adopted the reforms, then a reconsideration of the
policy is in order.
In a nutshell, our empirical findings for EU15 in the three industries we consider, are
mixed. Privatisation per se consistently leads to higher consumer prices. There are,
however, beneficial effects of market opening reforms, but often smaller and more
uncertain than we would have expected. In terms of perceptions, consumers are happier
with the price they pay under public ownership while the evidence is mixed for the other
reforms. Thus, objective and subjective evidence are broadly mutually consistent, as far
as privatisation is concerned. These findings are entirely new and tend to reject the
earlier, more optimistic views by the European Commission on the outcomes of the
reforms CEC (2007).
The structure of the paper is the following: first, we briefly review some key facts
about the reform trends in the European industries under consideration (Section 2). Then
we look at price trends (Section 3). Earlier literature is briefly discussed and a very
simple conceptual model of a reform sequence in a generic network industry is presented
in Section 4. Then we present our empirical modelling strategy (Section 5), the data we
use (Section 6), and our results (Section 7). In Section 8, we discuss our findings and
conclude offering some policy implications.
2. 1B1BREFORM TRENDS
The typical network industries reform package has three pillars: privatisation,
unbundling, and market opening. In the last two decades the EU has witnessed a
considerable variability of policy combinations thereof. Thus we want to test separately
the effects of privatisation and of the other two reforms.
Our research strategy is to focus on the three industries selected because they cover
different technological and institutional features:
a) fixed telephony, where the reform story started earlier in the UK with the
divestiture of British Telecom (1984): a sector with common technology
shocks across countries and limited differences in technology adoption;
b) electricity: a sector where the reforms started in the 1990s, with persistent
differences in technologies and in the endowment of fuel resources across
countries;
c) gas: a sector that is a late-comer in the reform, where the technology is
similar across countries, but where there are variations in dependency on a
homogenous natural resource.
Public enterprises in these three sectors were (or are) profit-makers in Western Europe,
and usually had no difficulty in financing their investments, hence their pricing is usually
not affected by government transfers, as in other network industries (e.g. railways).
For each of these sectors the regulatory history was summarised in the ECTR database
by scores that go from 0 to 6, under different headings and sub-headings. For example
for telecoms there are three headings: public ownership that is scored 6 for the vertically
integrated public monopoly and 0 for full privatisation; entry regulation, where 6 means
5
no entry is allowed and 0 means full market opening; market share, where 6 means that
the new entrants have zero market share and 0 that they have a large market share. These
scores result in fact from the aggregation of sub-heading scores.FF
2FF
[Tables 1, 2 and 3 about here]
To present the history of regulatory reform in concise way, one can look at Figures 1-
2-3-4, where trends are reported for the ECTR reform indicator, and for the reform
components. In each figure we report the median value of the indicator, the 25 and 75
percentiles, and individual country observations in each year since 1975. Additionally,
one may also look at the year when the independent regulatory authorities in each sector
and each country was officially established, which accompanied the reform (Table 5).
[Figures 1, 2, 3, 4 and Table 5 about here]
As mentioned, telecom is the sector where the reform story started in the UK more
than 20 years ago. Around 1995 there was a steep and convergent reform process in the
EU15, with nearly all countries converging to the same value in 2007 (Figure 1, first
panel). However, looking at public ownership, entry regulation and market structure, it is
clear that there is substantial country variability around the common policy pattern
(Figure 2). While formal market entry in 2007 shows full opening everywhere, market
structure is much more dispersed across countries. Moreover, since 2000 the process
seems to have converged around a value indicating a stable oligopoly. As for public
ownership, there is also some evidence of cross country variability and of a slow-down
of the privatisation process in recent years (some governments retain their stake in the
incumbent operator).
The reform of the electricity industry also shows a clear trend, but the starting date was
around ten years later than for telecom (Figure 1, second panel). While unbundling and
entry regulation were forced by EU legislation, it is interesting to notice that public
ownership still plays a significant role in the industry (Figure 3). Market structure is not
reported in ECTR for this sector, but there is evidence elsewhere (see below) of a
tendency towards oligopoly with a competitive fringe.
Finally, the gas sector is a relative latecomer in the process. Unbundling and entry
liberalisation started around 2000, only in recent years has the market structure evolved
from a near monopoly to a rather concentrated oligopoly, while public ownership is still
important in several countries (Figure 4). In fact, after a start-up in the mid-1980s,
privatisation proceeded very slowly and has stopped in several countries in recent years.
The ECTR country scores provide strong evidence of our two tenets: there is in the EU
a common reform trend, but there is also enough variability across countries and
industries.
3. 2B2BPRICE TRENDS
How did consumer prices evolve since the end of 1970s? Figure 4 depicts the median
price of each SGEI across the EU15, using the longest times series available of average
2 Tables 1, 2 and 3 show the definition of the ECTR scores joint with our own coding, which we are going to use in this paper. Details about ECTR data coding can be found in Conway and Nicoletti (2006).
6
national prices, deflated using the yearly national consumer price index and setting at
100 the median price in 1978 (left vertical axis). The median price was contrasted with
the ECTR synthetic indicator providing an overall picture of the reform trends in each
sector (right vertical axis). Let us start with the electricity sector. The first panel shows
the clear downward trend of electricity prices during the 1980s and the 1990s although
the trend was the opposite before and after this period. It is however worth noting that
the downward price trend started well before the reform wave showed its first signs:
about one third of the median price reduction happened before 1990, when the reforms
started in the UK (the front runner of reforms in the EU15). Another third happened
before 1996, when the first EU Directive was approved. After 2003, when the second
EU directive was approved, the price started increasing, most likely due to the increasing
cost of oil, a relevant input in energy production. The correlation between prices and
reforms is even less clear in the gas industry (second panel): prices increased up until
1985, while no reform had started. Major price reductions took place between 1985 and
2000, while the ECTR index had barely reduced and in the following years the two
series present a negative correlation. In telecoms, prices and reform trends are both
roughly decreasing, although the time series starts in 1990 and correlation is low since
mid 1990s.
It is also informative to look at price figures deflated using the consumer price index
for some relevant years in two key countries, the UK and France often regarded as at the
two extremes of SGEI reform policy implementation (Table 4). In 1978 electricity prices
in France and in the UK are equal to the EU15 median, but by the end of 2006 the
French prices are lower and UK ones are higher than the median, besides the
convergence of prices across the EU15 as showed by the decreasing standard deviation.
In 1978 French gas prices where about 50% higher than the UK prices, which were
about at the median level. After thirty years gas prices for these two countries converged
close to the median. Finally, as for telecoms, UK prices slightly reduced since 1995
while French ones increased and they are both higher than the median level.
The message from this example is twofold: first, relative prices show strong and
different dynamics across countries; second, to study the impact of privatisation and
liberalisation we need to control for possible cost and demand shifters across sectors.
4. 3B3BEARLIER LITERATURE AND A THOUGHT EXPERIMENT
Can we predict how prices may respond to privatisation and regulatory reforms of
utilities? Two strands of theoretical literature are relevant to discuss the issue: the theory
of privatisation and the literature on regulation of networks industries. The case for
privatisation is reviewed, inter alia, by Bös (1993),Vickers and Yarrow (1993,Ch 2),
Parker (1998), Newbery (2000), Florio (2004, Ch.2), Megginson (2005), Roland (2008,
Ch 1). On the issue of the impact of privatisation on welfare a seminal paper by
Sappington and Stiglitz (1987) established the conditions for indifference between
public and private ownership. They show that under different information structures a
benevolent policy-maker who can write complete contracts achieves the same welfare
7
outcomes of private owners. This result prompted two research lines: privatisation under
incomplete contracts, see Schmidt (1996), where the government is unable to get all the
information needed to achieve the Sappington-Stiglitz result; or privatisation where the
policy makers have a private agenda, see Shapiro and Willig (1990). On the incomplete
contract side the divestiture of public ownership is seen as a mechanism that prevents
governments to achieve full information and to wipe away agent’s rents. The regulator
gives information rents to private agents in exchange of efficiency-enhancing
investment. Hence social welfare increases. The model rests, inter alia, on the
assumption that the private agent only has the capital, or knowledge, or the incentives to
sink new investment. On a different vein Laffont and Tirole (1993, Ch.16) stress the
trade off that the managers face when they need to respond to private owners and to
regulators. This is a multi-principal context in which privatisation can be socially
beneficial if the objectives of managers and of shareholders are more aligned than the
objectives of the manager and the regulator. In this context, similar outcomes can be
obtained by different layers of government or different offices in charge of ownership,
management and regulation of the utility. Thus, the hypothesis that privatisation is
rooted in the limited commitment credibility of government does not provide a clear-cut
answer to our question. If privatisation enhances efficiency through information rents
given to the agent, unit cost may decrease more than the mark-up on costs, but the same
result may be achieved by a combination of public ownership, managerial discretion, and
independent regulation. When quality is also considered, as by Hart, Shleifer and Vishny
(1997), several outcomes are possible, and in general privatisation is less desirable
where competition is limited. The distinction between privatisation and liberalisation is a
core one. In Vickers and Yarrow (1993, p. 44) words: “ Where product markets are
competitive, it is more likely that the benefits of private monitoring (e.g. improved
internal efficiency) will exceed any accompanying detriments (e.g. worsened allocative
efficiency)[…]. In the absence of vigorous product market competition, however, the
balance of advantage is less clear cut, and much will depend upon the effectiveness of
regulatory policy”.
Turning to the second strand of privatisation theory, it rests on the realistic hypothesis
that policy makers have private agendas and they distort the management of utilities to
favour rent extraction. Earlier models, such as Boycko, Shleifer and Vishny (1996) tend
to say that the divestiture of public ownership is welfare enhancing because transfers
between the Treasury and the utility are more transparent, and this limits rent extractions
by politicians. It is, however unclear, in this context, why politicians would sell public
enterprises, as they did on a huge scale in the last twenty years. Florio (2004) observes
the asymmetry of assuming benevolent policy makers under privatisation and non-
benevolent ones under nationalisation. Laffont (2005) presents a model where
privatization occurs with a corrupted government, and shows under some assumption
that the welfare outcome is non linear in corruption, and that in some case privatization
occurs for the “wrong” reasons, i.e. as an alternative way to extract rents by the
politician. Thus, also the second strand of privatization theory, while in general more
suspicious about the role of government ownership, cannot provide a clear message
8
about the direction of change of prices after privatization, because much will depend
again upon a public office, the regulator. If the regulator is non-benevolent, collusion
between the privatised incumbent and the regulated utility can reverse the expected
benefits of the divestiture of public ownership.
Turning now to regulatory issues, important new insights are given (selectively citing
from a very wide literature) by Newbery (2000), Laffont (2005), Rey and Vergè (2008),
Rey and Tirole (2008). According to Newbery (2000) the restructuring of network
industries should take advantage from differences in economies of scale of different
segments, with the physical network usually showing sub-additive costs, while several
upstream and downstream activities operate under a regime of constant or decreasing
returns to scale. This leads to a paradigm of reform where vertical disintegration of the
network is a crucial step, and access regulation is the institutional mechanism that would
allow for competition in other activities. One way to evaluate this reform paradigm is to
see unbundling as a structural remedy to market foreclosure. Rey and Tirole (2008)
mention divestiture as the extrema ratio measure, with milder forms of vertical
separation as an alternative to be assessed case by case. These include common
ownership of the network by competitors, access price control, access quantity control,
price benchmarking, ‘common carrier’ policies, transparency in contracts. Rey and
Tirole (2008) distinguish between vertical (limiting access to an essential input) and
horizontal foreclosures (across goods). Drawing from previous literature and new
models, they formalise an often very complex discussion and show a number of counter-
intuitive results. Particularly, they tend to see vertical integration as a mechanism that
helps the upstream monopolist to defend monopoly pricing against non-integrated
competitors. They, however, show several mechanisms alternative to structural
separation to prevent vertical foreclosure. Moreover, they also review potential cases for
socially beneficial market foreclosures (e.g. when there is excessive entry, or free-riding
by down-stream competitors). In fact, foreclosures can be seen as part of a wider
discussion on the economics of vertical restraints ( Rey and Vergé, 2008). These are
contracts between producers and distributors that are aimed at softening competition
(exclusive dealings are a well known example). In network industries, even after
unbundling of the network, the question arises about the contract arrangements between
the upstream agents (e.g the electricity generators, or the gas importers), the network
owners (e.g in telecoms the owners of the physical lines), and the downstream suppliers.
Rey and Vergé (2008) show that the welfare effects of vertical restraints are crucially
different according to a number of features. They conclude that while the impact on
aggregate profits of a ‘vertical structure’ (and vertical integration can be seen as an
extreme case of such arrangement) is positive, the impact on consumer surplus is
ambiguous. Provisions that wipe away double marginalisation, occurring where each of
the different players has some market power, may be welfare enhancing. Thus, their
policy conclusion is that “the optimal policy towards vertical restraints cannot be one
such that some particular provisions are deemed illegal per se while some others are
always acceptable”. In fact, the balance of the welfare impact of unbundling depends
upon the extent of the double marginalisation effect versus the entry effect.
9
When we combine this relativistic policy conclusion, with the also relativistic
conclusion of the discussion on privatisation in a realistic setting with incomplete
contracts and various degrees of government benevolence, it seems difficult to predict
price direction changes based on a robust general reform paradigm of network industries.
To add some intuition to the previous discussion, even if we need here to simplify a
lot, let us see price changes in a simple thought experiment (for detailed welfare
calculations see Ceriani and Florio, 2008).See Fig. 6.
[Figure 6 about here]
We consider, first, a vertically integrated public monopoly; second, a privatised, but
still vertically integrated monopoly, without any (binding) price regulation (except for
the constraint of no price discrimination across users); third, a price-cap regulation of the
privatised incumbent monopoly; fourth, unbundling as a preliminary step to
liberalisation; fifth, a duopoly market, without price regulation; sixth, full market entry,
as an extension of the previous case. One may interpret the different steps either as
variations across countries, or across industries, or over time.
Let us assume that network is under decreasing average costs and the marginal cost is
zero. In contrast, the operation component has constant average costs equal to marginal
costs. In a given initial year of our story, the network is a sunk cost: capital invested in
past times and financed by the taxpayer, who will continue to finance some maintenance
costs. Thus, the public monopolist needs only to cover its variable costs. The
government wants to maximise the consumer surplus, but also has a private agenda. The
public enterprise, in this simple example, implements tariffs equal to the marginal cost of
the operations (other assumptions are possible, but we are not exploring them here).
While, by construction, there is no allocative inefficiency, the public provider is,
however, possibly affected by x-inefficiencies because of rents distributed to managers,
workers and politicians. Thus, costs are not minimised, and unit cost c is greater than c*,
a minimum long-term marginal cost (we are not discussing here the complications that
arise in the long run because of technology shifts). The cost under public monopoly is
given by c1=c*(1+α), where α stands for inefficiency (this can be zero with a
benevolent government and symmetric information). Hence, the price for the electricity
service is simply p1=c*(1+α), see Figure 6a.
Consumers’ unhappiness (perhaps recorded by surveys such as Eurobarometer) leads
the (elected) government to privatise the industry. The private owner has an incentive to
minimise costs. At this early stage of the reform the government is unable or unwilling
to effectively regulate prices (e.g. to maximise privatisation revenues to the Treasury, or
because of lack of experience of the regulator). Thus the industry is now under vertically
integrated private monopoly. The new equilibrium price (see Figure 6b) is now p2. The
consumer surplus (and perhaps the perceived fairness of the price) will change
accordingly.
Popular dissatisfaction with the first step of the reform pushes the government to
appoint a (tougher) regulator with the power to impose a price-cap to the private,
vertically integrated incumbent. The regulator imperfectly knows the firm-specific
technological constraints and the managerial effort of the regulated firm, and he picks up
10
a price constraint sufficient to offer the private investor an ‘adequate’ profitability
(participation constraint) with some slack. This can be represented by a simple mark-up
(1+β) over the first best marginal cost, which is compatible with participation .The new
price is p3=c*(1+β). Obviously, p1>p3 only if α>β. Figure 6c, where p2>p1>p3, shows
one possible outcome.
The regulator imposes then on the incumbent the separation of the network (examples
are electricity in the UK since 1990, Sweden since 1996, or Finland and Portugal since
2000, and fixed telephony in the UK since 2006, etc). This new reform is best seen as
preparation for granting access to entrants, see Pollitt (2008). The simplest case is the
establishment of a duopoly under private ownership and the separation of the network.
The regulator lifts the price-cap for the operators. There are now some co-ordination and
transaction costs (allocation of time and capacity slots). Let us say that, under
unbundling, c* is no longer achievable, but only c*(1+γ). Moreover, there is now an
additional cost to the entrant because, for instance, he has to advertise his entry. Let us
say that δ is this extra cost. We assume that there is some market inertia. Under the
Chamberlin (1933) ‘small group’ competition model, the entrant targets just the demand
not satisfied by the incumbent at the original price. Thus the entrant fixes its optimal
quantity and price supply of the service based on the residual segment of the demand
curve. The optimal price for the entrant is clearly less than for the incumbent’s segment
(technology is common), and the incumbent must follow the entrant and fix the price at
the new lower level. Figure 6d shows the new situation.
If γ (loss of economies of integration, related to incomplete contracts and asymmetric
information) and δ (duplication of some sales and administration costs)are sufficiently
high and α or β were sufficiently small, then the duopoly equilibrium price, p4, can be
higher than each of the other three market structures.FF
3
The government in our virtual story learns this lesson, and takes further steps to lift
any market entry regulation. For example, since July 2007 all EU households can choose
a different electricity supplier and switch their contract. To study the impact of full
market liberalisation on prices, again, we want to be as simple as possible. After having
unbundled the network, the regulator is now ready to offer a licence to any new entrant.
Let n be the number of electricity or gas providers. There still is no price regulation, and
the network service costs are covered by the taxpayer. For the generic n-th entrant, profit
maximisation under Cournot (1897) competition implies p5=cn/[1- (σn/|ε|)], where cn is
the marginal cost for the n-th firm, p5 is the price when market entry is free, ε is the price
demand elasticity of the industry itself, and σn = qn/q, i.e. the share of the demand
covered by the entrant (e.g. Varian,1987, pp. 452-453). The smaller the entrant’s market
share, the more elastic the firm’s demand, whose elasticity is |(ε/σn)|. As the market
3 The intuition is clear. EDF, the public monopolist in France, offered electricity prices ‘near’ to the first best long run marginal costs (nuclear), cN*. To allow entry for a competitor from Germany (possibly gas powered, cN* < cG* ), both the incumbent and the entrant have to incur interconnection network-related transaction costs (currently there is an auction system), plus some output related costs such as advertising. In fact, Percebois (2008) shows that the consumer surplus change can be negative in France if the equilibrium tariff in certain hours is set at the higher marginal cost in Germany, while the producer surplus of EDF will increase. The opposite happens in Germany. If we disregard producer surpluses (which here are duopoly profits), the two combined changes in consumer surpluses in France and Germany do not need to cancel out. Thus, partial market integration may create some losers among the consumers.
11
share tends to zero, the condition for profit maximisation is the one associated with pure
competition. The firm’s cost, cn, however, is again not equal to first-best, because of the
inefficiency of unbundling (φ now replaces δ + γ of the duopoly example).FF
4FF Having five
gas suppliers competing in the same town for capturing a share of the demand, while
using the same network, accessing the same technology, and in fact probably importing
the same natural gas, has two partially offsetting effects. While competition gradually
reduces market power, hence prices, each firm has to face somewhat increased sales and
administration costs compared with our benchmark c*. Assume that cn=c*[1+φ(n)],
where φ(n) is a function increasing with the number of competitors. For example, a
regional gas supply oligopoly with five competitors, industry demand elasticity in the
equilibrium point |0.8|, and φ(5)= 0.05 (meaning that each entrant has a 5% extra-cost to
secure itself a 20% market share), implies p5=c*(1.05/0.75)=1.40c*. This outcome
implies that the vertically integrated public monopoly must have been affected by
α >40%, i.e. significant x-inefficiency, to deliver higher prices than the unregulated
oligopoly. Alternatively, under a price regulated monopoly the extra profits permitted
should be based on β>40%, i.e. a very lenient regulator, to get higher consumer prices.
In fact, the outcome of market liberalisation in the network industries in Europe seems
to be oligopoly, with very few competitors in (until now) poorly interconnected regional
markets. FF
5
Thus, even in our very simple story of the reform sequence, the prediction of prices
under different regulatory regimes depends on a set of (possibly country/industry
specific) parameters. Welfare dominance of one regime over the other is not warranted.
Even ‘full’ market opening can deliver higher prices than a regulated monopoly, if
entrants are few, elasticity of demand low, and unbundling and related costs high. Our
simple thought experiment shows in a precise sense why the evaluation of the reform
outcomes of network industries is essentially an empirical matter: it depends upon a set
of parameters that easily lead to non-linear outcomes along the reform ‘line’.
Table 4 presents a selection of empirical papers in this area. Unlike the papers
mentioned in the Introduction, these papers focus on consumer prices (and in some cases
on other variables, as well, not reported in the table). Our selection criterion is to show
only papers that use regulatory indicators as explanatory variables, with standard
demand and cost shifters as controls. Moreover, we consider only papers that use cross-
country evidence, as we do, and that use standard econometric testing. A ‘positive effect’
means that the reform increases prices. Hence, if public ownership has a positive effect,
or privatisation a negative one, the interpretation is that privatisation benefits the
consumer. For example, Li and Xu (2004), in a study of the price of local telephone calls
in 177 countries (1990-2001), are puzzled to find that full privatisation has a significant
positive (i.e. increasing) effect on residential prices, while it increases volumes.
Moreover they do not find a significant effect of competition on prices, or a modest one
4 It is worth noting that if the network owner is private or if it is a budget constrained public firm that applies the rule ‘price = average cost’ the inefficiency of network unbundling would be even worse, because zero access price is optimal. 5 Taking again the example of the electricity market, the combined market share of the two major competitors in 2006 was 68% in Germany, 75% in Spain, and over 80% in Belgium and France, with only the UK having established a competitive market, which was, however, not interconnected with continental Europe.
12
in containing the adverse effect of privatisation (interaction term). Viani (2007) in a
study of 67 countries (1984-2003) finds that while monopoly (private or public)
increases prices, vertical separation of the network has no effect. Grzybowsky (2008) for
telephone in EU15 (1998-2002) finds that liberalisation decreases prices. Edwards and
Waverman (2006) also consider EU15, but again for few years (1997-2005). They do not
study residential prices but interconnection rates and show that public ownership of the
incumbent increases them, while regulator’s independence decreases them. Boylaud and
Nicoletti (2000) study a basket of telecom prices in 23 OECD countries (1997-2003),
and find that liberalisation decreases prices, while state ownership has no effect. Similar
results are reported by Martin and Vansteenkiste (2001). More recently Gashi et al
(2006) study fixed charges for telephone and cellular prices in 29 developing and 23
developed countries (1985-1999). Their main focus is on regulatory governance, but
they find that privatisation increases the prices in less developed countries, while has no
effect on prices in developed economies. Estache et al. (2006a) for 204 countries (1990-
2003) find that privatisation increases prices, while regulation decreases them see also
Estache et al. (2006b) for a smaller sample (91 countries, 1990-2002). Wallesten (2001)
focuses on Africa and Latin America: they find a negative effect on prices of
privatisation, but the result is not statistically significant.
As for electricity, Hattori and Tsutsui look at 19 OECD countries (1987-1999) and
consider both industrial prices and the ratio between industrial/household prices. They
find some support for privatisation and liberalisation as determining price decrease.
Steiner (2001), however, in a study of 19 OECD countries (1986-1996), finds that
privatisation (and time to it) increases prices, while unbundling and liberalisation have
the opposite effect. Zhang et al. (2002) study electricity residential prices in 51
developing countries (1985-2000) and find no effects of the reforms. Martin and
Vansteenkiste (2001) do not find an impact of liberalisation on prices, and find that
public ownership increases prices in the EU15, in the very short period they consider
(1995-2000). More recently, Gassner et al. (2009), in a detailed empirical study for the
World Bank on 1,200 utilities in 71 developing and transition countries over ten years,
including publicly owned and private sector participated ones, and with a different
regulatory index, use differences-in-differences econometric techniques. They find that
privatisation does not have an impact on prices and investment.
Finally, there is very little, as far as we know, about rigorous testing regulatory reform
in the gas industry. Copenhagen Economics (2005) is supportive of the reforms while
Growitsch et al. (2008) do not find a significant impact of ownership change, while some
negative effect of market liberalisation.
To sum-up, earlier literature on network industries has only limited evidence that
privatisation per se is beneficial to consumers, and in fact several earlier papers conclude
that it increases prices or has no effect. In contrast, earlier research suggests that market
opening may decrease prices.
[Table 6 about here]
13
5. 4B4BOUR APPROACH TO THE EMPIRICAL ANALYSIS
Our empirical strategy is simple and straightforward. We want to answer the following
research question: are privatisation and liberalisation associated with lower consumer
prices? Moreover, we ask an additional question: Are consumers happier with the prices
they pay for the services in countries that have most reformed the network industries?
The perspectives of the two research questions are different but complementary. First,
we want to test whether consumers actually pay less when we observe a reform, looking
at average consumer prices and taking into consideration demand and supply conditions
and other potential explanatory variables. Second, we want to test whether consumers
are happier with the price they pay under stylised reform regimes, using individual
perceptions. The former research question tests hard ‘objective’ evidence on prices, but
it contains a potential aggregation error, as only prices for average consumers are
available in cross-country data sets. A realistic example is when there is a change in the
degree or the orientation of price discrimination. The latter research question tests softer
‘subjective’ evidence of individual users’ satisfaction with prices. This is perhaps a
proxy of consumer surplus, but it may contain a different type of error, if individuals are
biased in their perceptions. Looking, however, into both perspectives seems interesting.
If the empirical findings are mutually consistent, despite the different nature of the
potential errors, the evidence on the impact of the reforms would be reinforced.
Moreover, the second research question is also interesting in a political economy
perspective.
Our empirical analysis is based on two different class of models: dynamic panels for
prices and probit on repeated cross-sections for individual satisfaction. As for prices, a
reduced form model for prices is specified and estimated, including, as explanatory
variables, aggregate or detailed measures of the level of privatisation and liberalisation
of the sector, controlling also for year fixed effects and other macroeconomic variables.
As for individual satisfaction with the prices paid for each service, we assess whether
reforms are positively correlated with the likelihood of consumer satisfaction,
controlling for individual characteristics as well as for country and time fixed effects and
a set of individual characteristics of respondents.
The structural and evolutionary differences across the three network industries justify
why, while retaining a common empirical approach, our models have industry-specific
ingredients in terms of explanatory variables and controls, in addition to a core set of
indicators.
Following Blundell and Bond (1998, 2000), Bond (2002) and Bond et al. (2003) we
first study the autoregressive (AR) properties of the average consumer price for each
industry. As we find evidence of prices following AR(1) processes (cf. Section 7), the
econometric analysis of consumer prices is performed by using dynamic panel data
models, i.e. including among regressors also the lagged dependent variable for
explaining the strong persistence of prices measures.
For each network service sector, let it
P be a measure of current net-of-tax (log) prices
for country i at time t, itR the set of regulatory variables, which might include a score of
14
the level of regulatory regime in each industry or measures of entry regulation, public
ownership, market structure, vertical integration,FF
6FF and
itX a set of control variables,
including demand variables such as an indicator of per capita income levels, sector
specific variables such as type of energy source, input costs and efficiency indicators,
when available. The model for price levels is:
, 1 ' '
( ) with 1,..., ; 1,..., .
it i t it it t it
it i it
P P R X
i I t T
α β γ δ ε
ε η ν
−= + + + +
= + = = (1)
The year-specific intercept (tδ ) is included to account for common cyclical or trend
components in prices, preventing a likely form of cross-country correlation. A key
assumption of this kind of models is that of independence of the idiosyncratic
disturbances (it
v ) across countries. We treat the individual effects (i
η ) as stochastic,
which implies that they are correlated with the lagged dependent variable ( , 1i tP − ), unless
the distribution of the i
η is degenerate. We also allow the control variables (it
X ) and
the regulatory variables (it
R ) to be correlated with the individual effect i
η .
Maintaining that the it
v component of the error term is serially uncorrelated, we then
assume that it
X is strictly exogenous (i.e. uncorrelated with all past, present and future
realisation of it
v ). As for the set of reform variables it
R , we assume that they might be
correlated with the unobserved error term but that, due to the political and decisional
process involved, they react with some lag to changes in it
v . In other words, we assume
that it
R is predeterminate (i.e. it
R and it
v are uncorrelated, but it
R may be correlated
with , 1i tv − and earlier shocks).
The assumption of stochastic individual effects implies that they are correlated by
definition with it
P , and possibly also with it
X and it
R . If so, Ordinary Least Square
(OLS) estimator of { , ,α β γ } in the level equations of model (1) is inconsistent, since
the RHS variables are positively correlated to the error term due to the presence of the
individual effects, and this correlation remains even for ,T N →∞ . A Within
estimator would eliminate the source of OLS inconsistency, i
η , by transforming the
original observations as deviations from their own individual time mean, however it does
not completely solve the problem. In fact, for a small T , the Within transformation
induces a correlation between the transformed lagged dependent variable and the
transformed error term, producing a biased estimator (Nickel, 1981). Notwithstanding
their biased nature, we follow the established practice in the related literature the OLS
and Within estimates are presented for comparison along with other estimates based on
Generalised Method of Moments (GMM). GMM, by using extra moment conditions,
potentially produce consistent and efficient estimates, coping with endogeneity of the
lagged dependent variable.
In fact, as no external instrument is usually available outside the immediate data set, an
alternative to OLS and Within estimates is to either transform the data to eliminate the
individual effects or find some instruments which are orthogonal with the error term but
6 We also considered introducing as regulatory variable the years spanned since the establishment of the sectoral independent authority (cf. Table 5) but in no regression this variable was statistically significant, hence it was omitted from estimations.
15
not with endogenous regressors. Arellano and Bond (1992) suggested7 transforming the
data in first-difference obtaining
, 1 ' 'it i t it it t it
P P R X vα β γ δ−∆ = ∆ + ∆ + ∆ + ∆ + ∆ . (2)
This transformation eliminates the fixed effect, although the lagged dependent variable
remains potentially endogenous, as the , 1i ty − term in , 1 , 1 , 2i t i t i t
y y y− − −∆ = − is
correlated with , 1i tv − which is in , 1it it i t
v v v −∆ = − and predetermined variables such as
itR∆ become potentially endogenous because they may also be related to , 1i t
v − .
However, longer lags of the endogenous regressors ( , ,,i t s i t r
P R− −∆ ∆ with
2,3,..., 2; 1, 2,..., 2s t r t= + = + ) are orthogonal to the error and can provide
additional moment conditions working in the GMM framework.FF
8FF The Arellano and
Bond estimator in this context is called the differenced-GMM method (henceforth,
GMM-Dif).
As GMM methods, as well as other methods based on instrumental variables, crucially
rely on the existence of strong moment conditions we also investigated this issue. Hence
we provide estimates of our model using the Blundell and Bond (1998) system estimator
(henceforth, GMM-Sys). This estimator augments the GMM-Dif by estimating the
model (1) in level and instruments the lagged dependent variable by first differencing
each instrumental variable (including earlier lags of endogenous and predetermined
variables) to make it exogenous to the fixed effect of the equation in level. In all our
GMM estimates, we use a one-step GMM estimator, similarly to most applied work in
this area as simulation studies have suggested very modest efficiency gains from using
the two-step version, even in presence of considerable heteroskedasticity (Arellano and
Bond, 1991; Blundell and Bond, 1998; Blundell, Bond and Windmeijer, 2000).
Finally, we follow the practice of showing results for different number of instruments
and to use the Arellano and Bond (1991) autocorrelation test for testing whether
autocorrelation in the idiosyncratic disturbance term it
v would render some lags invalid
as instruments. In fact, a key identifying restriction in dynamic panel data models
estimated using GMM is that there is no serial correlation in the it
v disturbances and
this is tested using the first-differenced residuals. Since it
v∆ is mathematically related
to , 1i tv −∆ via the term , 1i t
v − , negative first order serial correlation is expected in
differences and evidence of it is uninformative. The test of r-order serial correlation is
asymptotically normally distributed under the null of zero serial correlation.9
To sum up, we shall propose estimations based on four estimation methodologies:
OLS and Within group which are flawed by the endogeneity of lagged dependent
variable but use standard moment conditions and GMM-Dif and GMM-Sys, which
exploits additional moment conditions for tackling correlation of some regressors with
the country fixed effect. As our analysis is unavoidably constrained by the number of
7 See also Holtz-Eakin et al. (1988). 8 Of course, also the orthogonality of the exogenous variables with the transformed error term is exploited and corresponding moment conditions included in all GMM estimations. 9 Arellano and Bond (1991) find that their test has greater power than the Sargan/Hansen test to detect lagged instruments being invalid due to autocorrelation.
16
countries included in the EU15, any estimation method in this context may lead to a
biased estimate. However, if different estimation methods deliver similar results we
interpret them – with some caution – as pointing to the unknown true parameter.
As, due to the limited cross-section dimension of our datasets ( 15N = ) none of them
is certainly consistent. As no analysis is feasible about which of them delivers the
smallest bias, we interpret similar results using different estimation methods as evidence
that true coefficients should not be significantly different from those estimated.
As a supplementary analysis of the effects of reforms on consumers we develop a
model of self-assessed satisfaction with prices paid for the considered services. As
mentioned in Section 4, price perceptions by consumers may capture individual
information misrepresented by average prices. Moreover, policy adoption or reversal can
be influenced by perceptions. It is, however, necessary to consider possible subjective
bias and information noise in perceptions. Thus, empirical modelling of subjective data
needs to control as far as possible for individual features of respondents. The advantage
is that, in contrast to standard analysis of price series, the number of observations is here
much higher.
As no other significantly more rich data is available for our analysis of effects on
prices of reforms across the EU15,FF
10FF we looked for possible confirmation of our main
results using a completely different type of data to be analised with a different approach:
we use subjective satisfaction data to be analysed with dichotomous dependent variable
models, such as probit models. In this case we can exploit a much larger sample size (cf.
Section 6), although available only as repeated cross-sections. These data can also be
criticised on several grounds as subjective evidence is notoriously fickle. For instance,
there might be a “hedonic treadmill” effect whereby satisfaction changes with outcomes
so that, for example, if a country has enjoyed a big price cut in one period, consumers
expect this again and are dissatisfied if this does not happen again in the next period.FF
11FF
Notwithstanding this limitations, in case results on individual perceptions were found
correlated to utilities’ reforms in a similar way to that of average prices, suggesting
similar welfare effects, we would interpret them as additional evidence complementing
that coming from the analysis of prices.
The model for consumer price satisfaction rests on the definition of a variable
disentangling satisfied from relatively dissatisfied consumers regarding the price paid for
a particular network service. Let this variable bejit
S , which is recorded equal to 1 if
individual j in country i at time t states that the price he pays for fixed telephone services
is fair, and is recorded equal to 0 otherwise. We estimate a standard probit model such as
*
Pr( 1| ' , ' , ' , , )
= Pr( 0 | ' , ' , ' , , )
jit jit it it t i
jit jit it it t i
S Z M R
S Z M R
τ λ ζ φ ϕ
τ λ ζ φ ϕ
= =
> (3)
10 For instance, there is no EU15 comparable data set on individual prices paid by households with different characteristics, nor data on average consumers by European regions at NUTS2 level. 11 We thank an anonymous referee for suggestion this effective way of presenting the endogeneity of subjective data. For a seminar paper on this issue, see Kahneman et al. (1986).
17
where *
jitS is the latent exact level of satisfaction, ,t iφ ϕ are year and country fixed
effects, respectively ( 1, 2,..., ; 1,2,...,t T i I= = ), jit
Z is a vector of individual
characteristics (i.e. sex, occupation, etc.) accounting for individual observed
heterogeneity, and it
M is a vector of macroeconomic control variables including
population density, GDP levels, consumer price index and the total price (including all
taxes) of a unit of each industry’s provided service, and it
R is the scalar or vector of
regulatory variables as presented in Section 2. Finally, , ,τ λ ζ are vectors of parameters
to be estimated.
The estimation methods used here is straightforward. Unlike previous models on
average net-of-tax prices, no panel data is available for studying consumer satisfaction,
hence we estimate model (3) with maximum likelihood using repeated cross-sections of
individual data. First we estimate model (3) under the constraint that 0ζ = , for a
simple check on whether satisfaction is correlated with average prices paid for each
service, with an expectation of a significant and negative coefficient, according to the
view that consumers are rationally less satisfied when the average price paid is higher.
Then, we estimate the full model (3) and interpret any significant coefficient in the
vector ζ as a sign of the fact that regulatory reforms have some effect on consumer
satisfaction regarding prices. A priori, we have no expectation as for the statistical
significance nor for the sign of the coefficients of the regulatory variables.
6. 5B5BTHE DATA
For this project we extensively used different data sources, focussing on EU15
countries. While presenting large variability in the timing and the extent of
implementation of regulatory reforms, EU15 countries also share similar institutional
characteristics and a common legislative direction. The primary sources for the average
(log) price variables are the International Energy Agency (IEA) for electricity and
natural gas and the World Telecommunication Indicators database, maintained by the
International Telecommunication Union (ITU), for the telecom sector.FF
12FF The only
alternative data source available for a EU15-wide analysis would be the Eurostat, which
however is available for a much shorter time series as it starts at best in 1991 and at
worst (for telecommunications) they start in 1998, thus comprising at most only eight
consecutive years. Instead, the time-series of IEA and ITU data start much earlier (1978
and 1983, respectively), although, we restrict the analysis for telecoms to data after 1989
as many missing values are present during the 1980s (some summary statistics are
provided in the Appendix). The IEA net-of-tax electricity and gas prices for households
are expressed in €/unit and present a correlation with household net-of-tax electricity
prices (yearly consumption of 3 500 kWh of which night 1 300) and gas prices (yearly
consumption: 83.70 GJ) from Eurostat data equal to 0.814 and 0.847, respectively. The
ITU measure of telephone price is the price of a 3-minute fixed local telephone call at
peak rate.
12 IEA and ITU monetary variables, which are expressed in US$, were converted into euro using the World Development Indicators (WDI) Eurostat euro/US$ exchange rate.
18
For consumer satisfaction we used self-rated satisfaction with the prices paid for each
service considered (electricity, gas and telecommunications) by a representative sample
of EU15 citizens, as measured in the Eurobarometer (EB) surveys, which were run
biannually between 2000 and 2006. Although formulated in a very generic way,FF
13FF the
question of price satisfaction was checked for consistency with the total prices paid by
an average consumer as in model (3) above. The EB data also included a large list of
individual characteristics, allowing us to control for some individual heterogeneity.
Using the EB data we had four separate cross-sections comprising over 50,000
individual opinions about electricity and telecommunication prices and around 30,000
about natural gas prices.FF
14
As for the regulatory reform variables, including measures of entry regulation, public
ownership, market structure, vertical integration, in varying levels of detail, we used data
provided by the ECTR data set (see also Section 2). These data have a yearly frequency,
are available since 1975 and up to 2007 and present detailed information allowing one to
capture the industry-specific trends of reforms in all the three sectors considered. As
each ECTR score, going from 0 to 6, is computed as a weighted average of public
ownership, vertical integration, market structure and entry regulation scores, by
assigning a cardinal measure to variables that are only ordinal, we also introduced our
own coding. In particular, we defined dummy variables where only ordinal variables are
available in first place. For instance, for each industry, we defined a dummy variable for
public ownership equal to 1 if the industry is “public” and zero otherwise. A dummy
variable for entry regulation equal to 1 if there is “no third party access, full entry
regulation” and zero otherwise. A dummy variable for vertical integration equal to 1 if
the industry is “vertically integrated” and zero otherwise. We prefer this approach in
order to avoid any possible measurement bias with cardinalisation as in the original
ECTR scores. No market structure measure is available for electricity, while the market
structure dummy for the gas industry is equal to 1 if the market share in all segments of
the industry is larger than 90%. As ECTR data for telecommunications record the share
of new entrants as a percentage between 0 and 100, we defined the market share in the
sector as its complement to 100, meaning that the larger this number is, the farther the
industry is from full implementation of the reform package, consistently with the
interpretation of the other ECTR variables. We left the market share variable in the
telecom industry in a non dichotomous form as it is the only one that allows a cardinal
definition without imposing any arbitrary value (for more details on the question
addressed in the ECTR database, the original and our coding, refer to Tables 1, 2 and 3).
7. 6B6BRESULTS
The estimation of the price model (1) is structured in various steps. First of all, we
provide some evidence supporting the assumption that 0α ≠ , using notation
13 The question for the electricity service is “In general, would you say that the price you pay for the electricity supply service is fair or unfair?”, and similarly for gas and fixed telecom services. 14 Respondents about gas prices were fewer as in some countries (e.g. Greece, Portugal, Sweden) there is often no natural gas service provision.
19
introduced in Section 5. In case (log) prices did not follow an autoregressive process,
one would not need to use the dynamic panel toolkit for dealing with the endogeneity of
the lagged dependent variable and, standard Within group transformation would be
enough to sweep out the correlation of some regressors with the error term because of
the country fixed effect. Second, we estimate a set of models using different estimators,
including the OLS, Within group, the difference GMM (GMM-Dif) and the system
GMM (GMM-Sys) using different set of instruments. Third, a diagnosis of different
models is provided supporting the use of different models for different sectors. Finally,
we will discuss what are results estimating the model (3) of subjective satisfaction with
prices paid for each service.
Let us now turn to the decision of using dynamic panel toolkits instead of simpler
static panel methods. In other words, what would one lose if model (1) was estimated
under the restriction that 0α = ? Our approach to answer this question is
straightforward and follows Blundell and Bond (2000). First of all we estimate simple
AR(1) equations of (log) prices for each sector separately, including year dummies in all
models using OLS and Within estimation methods. Table 7 presents results. It shows
that in all models the lagged dependent variable is highly statistically significant, with a
p-value smaller than 0.1% and with a large but statistically smaller than one
coefficient.FF
15FF This supports our expectations that the lagged dependent variable is a
relevant variable to include in the analysis. However, as the main focus of our paper are
the coefficients in vector β , we are mostly interested to assess whether the omission of
1tP− would bias its estimation. This would happen if , 1i t
P − was correlated with the
regulatory variables. Hence, we regress the one-period-lagged log prices first over the
sector score (0-6) and a full set of time-dummies and test whether the coefficient of the
sector score is significantly different from zero. Then we do the same replacing the
sector score with the set of regulatory dummies (0-1) described in Section 6 and perform
an F-test of the hypothesis that all coefficients of the regulatory dummies are jointly
zero. Results are shown in Table 8, where the relevant test statistics on the regulatory
variables’ coefficients is named F1. The p-values of these tests let us conclude that the
correlation is highly significant in all specifications, except for the ECTR score variable
in the gas sector. In other words, except for this case, we conclude that dynamic panel
models are necessary for avoiding omitted variable bias in the β coefficients.FF
16
We now turn to the estimation of various specifications of model (1). Tables 9, 10 and
11 present results for the electricity, gas and telecom prices, respectively. These tables
share the same structure, although they are estimated separately, sector by sector. The
first six columns report the estimation of model (1) where the reform variable R is the
ECTR sector score ranging from 0 to 6, the second six use as reform variable the set of
dichotomous dummy variables, as described in Section 6. In each block of six columns,
the OLS estimation is reported in the first column, the Within estimation in the second,
two selected GMM-Dif models in the following two and two GMM-Sys in the last two.
15 We also estimated AR(2) processes but found no clear evidence on the significance of the coefficient of the two-period lagged dependent variable, so we ruled out the possibility of using longer lags of (log) prices among regressors. 16 To simplify comparisons, we estimate dynamic panels also for the model of gas prices using the ECTR score variable.
20
As opposed to OLS and Within estimation, both difference- and system-GMM results
might depend on the choice of instruments. One might use all possible instruments,
including the dependent variable lagged two periods or more, the predetermined
variables lagged one period or more and all exogenous variables. However, as in GMM
methods the instrument proliferation is easy and poor instruments might cause biased
and inconsistent results, we also test what happens if only the most recent of available
lags of the dependent variable is used as internal instrument. Then, following Blundell
and Bond (2000) who discuss the possibility that the error term might have a small
degree of autocorrelation due to measurement errors, we also estimate these models
using all or the most recent lags of the dependent variable starting from period t-3. All
estimated models pass the key identification test allowing for zero second-order
autocorrelation of residuals. For matter of space limitation, in the main tables of results
we present only our preferred models, however the remaining models are also reported
in the Appendix. All models are estimated using robust estimation as tests of
homoskedasticity of residuals always the null.
Table 9 presents the results of the estimation of model (1) for the electricity industry.
The highly significant lagged dependent variable sheds doubt over similar panel data
models in earlier literature without a dynamic specification, whose estimates are very
likely to be affected by omitted variable bias, as discussed above. It is interesting to
notice that if the ECTR score, ranging from 0 (full privatisation, unbundling and
liberalisation) to 6 (no privatisation, vertical integration and no free entry in the
industry), is included in the regression testing whether the reform package as a whole
has any effect on average prices, one could conclude that no significant effect or a very
limited positive one is found, as an increase in the score (less reforms) increases average
prices, as found using GMM-Sys. If, instead of the ECTR score, a dummy variable for
each of the three dimensions of the electricity industry is included, it emerges that while
more restrictive entry regulation has a statistically significant and positive coefficient
with some estimation methods (namely GMM-Dif), vertical integration is consistently
significant and positive in all cases except for the OLS. Hence, entry liberalisation and
unbundling seem to contribute to the reduction of prices, although it should be noticed
that no information on market structure is available and some policy reforms, in
particular vertical integration, are likely to be positively correlated with the incumbents’
market share, hence to be upward biased due to omitted market structure variable.
Interestingly, except for the OLS estimation, public ownership has a consistently
negative sign meaning that average (log) electricity prices are between roughly 0.08
(using Within or GMM-Dif) and 0.04 (using GMM-Sys) log-points lower in countries
with public ownership. As average price in the IEA sample is equal to Euro 0.09 (per
kW/h), this amounts to a price lower by 0.4-0.8 Euro cents, which corresponds to about
4-8% of the unit price. Turning to control variables with a significant coefficient, they all
present the expected sign: elasticities of per capita GDP and residential consumption are,
respectively, positive and negative. In fact, earlier empirical literature shows that GDP
acts consistently as a demand shifter, while consumption for the service acts as a scale
factor. Hydroelectric source tends to reduce log prices.
21
[Table 9 about here]
Remarkably, similar conclusions are also reached with respect to regulatory variables
in the natural gas and telecommunication industries although not all coefficients are
statistically significant in all estimates. In the gas sector more restrictive entry regulation
has a consistently increasing effects on prices as well as more vertical integration,
although with very mildly. Public ownership and market structure, when statistically
significant, present both a negative sign, meaning that a supply of gas that is not too
fragmented and a large role of the public producer tend to reduce prices. According to
some GMM-Dif results, in the gas sector public ownership reduces average prices by
0.08 log points, corresponding to 8% of the average price. The elasticity of prices to per
capita GDP, when statistically significant, is positive. The log change of price of Brent
oil is instead consistently an important determinant of price dynamics, which is not
surprising as gas prices have been indexed using oil market prices since the beginning of
the 1990s.
[Table 10 about here]
Finally, as for telecommunications, notwithstanding the large reduction in the total
number of observations, public ownership remains statistically significant and negative
using the GMM-Sys, showing that publicly owned utilities reduces average log prices by
up to 0.10 log points, corresponding to roughly 1% lower average prices. All the other
dimensions of the reform package have insignificant coefficients, whatever the
estimation method used. As control variables only statistically significant variables are
introduced, i.e. the log per capita GDP and the log of mobile subscribers, which are
found to have a positive and significantly negative coefficient, using the GMM-Sys.
Mobile telephony diffusion can be interpreted as a substitute good for fixed telephony,
and a proxy of technical change as well.
[Table 11 about here]
At the end of the empirical analysis of prices we discuss which model of those
presented we trust more and why. The purpose of presenting different estimation
methods is to show how much our results depend on the estimation method used. While
the OLS and Within methods are flawed with endogeneity of the lagged dependent
variable, the GMM tries to overcome this problem. Among GMM methods we tested
whether using the dependent variable as internal instruments at periods t-2 and earlier or
at periods t-3 and earlier makes any difference. We found they produce very similar
results, ruling out the possibility that some moderate degree of moving average pattern in
the error terms due to measurement problem might be at work as in Blundell and Bond
(2000).
The estimation of AR(1) models as in Table 7 ruled out the existence of unit roots
although the α coefficient is estimated over 0.6 in all models suggesting the use of
GMM-Sys. On the other hand, the ratio of the variance of η over that of v is always
22
smaller than 1.17 Some insights about choosing among difference- or system-GMM can
be gained looking at the first-step regression results as, since Bound et al. (1995) it is
well known that weak instruments can provide inconsistent estimates. Hence, similarly
to Blundell and Bond (2000), we estimate a reduced form regression of the first
difference 1log( )t
price −∆ on 2 3 1log( ), log( ),..., log( )t t
price price price− − and a
reduced form regressions of the level 1log( )t
price − on
2 2log( ),..., log( ).t
price price−∆ ∆ A small coefficient of determination (2R ) in any
of these regressions and a Wald test of slope coefficients jointly equal zero would signal
a weak instrument set. Table 12 presents the results of these estimation for each sector.
For the electricity and gas results the story is similar, as the set of instruments in the first
difference equation are jointly significant whereas for the set of instruments in the level
equation it is not possible to reject the assumption of all instruments being insignificant
at any reasonable confidence level. For the telecom sector, instead, both set of
instruments are found to be highly significant. This results suggest that in the case of
electricity and gas, GMM-Dif are less affected by weak instrument problems, hence to
be considered as preferred estimation models, while in the case of telecoms the level
equation provides an additional and valuable set of moment condition for the estimation
of model (1) using GMM-Sys. Moreover, following Blundell and Bond (1998)’s results,
they may contribute reducing the small sample bias due to short time dimension of the
panel.
Turning now to consumer satisfaction with the price people pay for each of the three
services considered, Table 13 presents the estimation of model (3) under the constraint
that 0ζ = , for electricity, gas and telecom sector respectively. For each sector we first
estimate model (3) under the constraint that macro variables are not significant and then
we remove it. The probit estimation of satisfaction with prices of each service
considered, controlling for year and country fixed effects, shows that females between
31 and 45 years, unemployed and showing average or bad co-operation during the
interview are the least likely to be satisfied across all services considered. People aged
75 or over are more likely to be satisfied with electricity and telecom prices (although
not with natural gas) compared to people aged 15-30; satisfaction increases with the
number of years of education; other white collars, house persons and managers tend to
be more satisfied than the self-employed with the price paid for electricity and gas, while
no significant difference emerges for local phone calls, and students are consistently the
happiest; respondents with more polarised political views are less likely to be satisfied
with the price paid for electricity and telecom (although no significant difference
emerges for gas); finally, respondents showing average or bad cooperation during the
interview are less likely to be satisfied. Year dummies show that although over 60% of
respondents are satisfied with the price paid for each service, the trend is still positive.
More interestingly, introducing macro variables and in particular the average gross-of-
17 We studied the AR(1) process of prices and the ratio of the variance of η over that of v , as Blundell and Bond (1998)
showed that, if (log) prices P are close to a random walk or the variance of η over that of v is large, then GMM-Dif
performs poorly, especially in short panels, because past levels are little informative about future changes and untransformed lags are weak instruments for transformed variables.
23
tax price, as expected, we find that the average unit price paid by consumers is strongly
and negatively correlated with satisfaction, meaning that higher average prices reduce
the probability of consumer satisfaction. This finding justifies our cross-checking of
objective and subjective evidence.
[Table 13 about here]
Finally, the assumption that 0ζ = is removed. Tables 14 and 15 report only variables
measuring the regulatory reform package (using the ECTR score 0-6 and the
dichotomous dummies 0-1, respectively), omitting coefficients for individual
characteristics and most of the macro variables reported in previous table. Table 14
shows that the ECTR score (0-6) tends to be highly significant, especially when prices
have been controlled for. However, the sign of the ECTR score is negative for electricity
and positive for gas and telecoms, meaning that the whole reform package is positively
correlated with more consumer satisfaction as for gas and telecom services, but
negatively correlated as for electricity supply.
[Table 14 about here]
Once again, as the ECTR score might hide important information, we estimated model
(3) replacing the ECTR score with the variables described in tables 1 to 3. Table 15
shows that some dimensions of the recent regulatory reforms increase the likelihood of
satisfaction. In particular, where the electricity and gas industry is public satisfaction is
around 5% more likely for electricity prices and around 10% more likely for gas prices,
while the coefficient is not statistically significant for telecom prices. Evidence for the
other dimension of the reform is mixed: increasing free entry in the electricity and gas
industries, unbundling a vertically integrated electricity industry, and increasing the
market share of new entrants in the gas sector are correlated to higher satisfaction, while
other dimensions of the regulatory reforms had the effect of reducing satisfaction.
[Table 15 about here]
With all the caution one might reasonably have about using self-assessed satisfaction
with prices paid by consumers, these results show that there is a consistent difference in
satisfaction among EU citizens depending on how far their country is gone in
privatisation and other reforms. In particular, after controlling for a wide set of
individual characteristics and year and country fixed effects, satisfaction is more likely
where public ownership is still relevant. This finding is in striking agreement with the
previous analysis of price dynamics.
8. 7B7BCONCLUDING REMARKS AND POLICY IMPLICATIONS
In this concluding section we sum up our results, give a closer look to the policy
design of the reforms in the EU context, and we discuss how our empirical findings
contribute to the debate on next reform steps.
Our paper offers a double check of the impact of privatisation and of liberalisation of
network industries on consumer prices in the EU, looking -at the same time and with the
same framework of analysis- to both objective and subjective evidence. Moreover, while
the reform is often proposed as a policy package that includes privatisation, vertical
24
disintegration, and liberalisation, we disentangle the ownership effect after controlling
for other reforms. We explicitly consider dynamics, use time series longer than in earlier
literature, and data sources from international organisations, such as Eurostat, IEA, ITU,
OECD, the World Bank, Eurobarometer. These sources are widely available to any
researchers for further empirical analysis. While in Section 7 results were presented for
the evidence on prices and for consumer satisfaction, here we combine our reading along
the different components of the reforms, starting with the overall package, and then we
move to its components. Table 16 summarises the findings of the preferred models.
[Table 16 about here]
a) What is the overall effect of the reform package on prices and consumer
satisfaction? We find that the overall impact of the reform package on
consumer prices, as summarised in the ECTR industry score, is never
statistically significant. Hence, we cannot reject the hypothesis that the
combined effect of the reforms on average consumer prices of electricity, gas,
and fixed telephony is zero. This is what one would expect in our thought
experiment if the adverse allocative effect of privatization is just
counterbalanced by the beneficial effect of liberalisation, see Vickers and
Yarrow (1993) and our discussion in Section 4. An alternative interpretation
is that the linear weights that the OECD uses for the indicator misrepresent
the reforms. The results by Azmat et al. (2007), and by the earlier papers cited
in the Introduction, however, suggest that the OECD scores work well for
several other performance variables, including productivity and investment.
Our results, compared with Azmat et al. (2007), who find declining labour
shares, imply an increase of the profitability of reformed industries. This fact
may sustain investment, as in Alesina et al. (2005). The combined evidence of
lack of the overall reform impact on consumer prices, and of changes of the
factor shares, implies a redistribution effect, that may explain the overall
mixed or adverse perceptions of consumers.
b) Does privatisation per se decrease consumer prices? Our answer to the
question is certainly negative. Our discussion in Section 4 shows that if a
public enterprise is very inefficient, a privatised monopoly or oligopoly, even
without price regulation, can offer lower prices than the vertically integrated
public monopoly, because on balance its allocative inefficiency may be less
than its cost savings. We find, however, that this is certainly not true in the
European Union. As mentioned, private ownership is associated with around
6% higher prices in both energy sector, and 8% higher prices in telephony.
Consistently with this result, we find strong evidence of higher consumers’
satisfaction with the price paid under public ownership of the energy
incumbent, while for telephony the perceptions are not statistically
significant. We cannot rule out that part of the price increase is related to
quality change or additional services for telecoms, but this effect is unlikely
for the energy sectors.
25
c) Is vertical disintegration associated to lower prices? The ECTR database
does not consider unbundling of the telecom network, something that is now
discussed by EU policy-makers, but we can test the question for the two
energy sectors. In electricity vertical integration is associated to higher prices,
and we find that consumers are less satisfied. In gas, prices and vertical
integration are uncorrelated, while consumers are more satisfied in countries
and years with higher integration. We conclude that the evidence on
unbundling is mixed and different in the two sectors .
d) Does market opening deliver lower prices? This is clearly the crucial
question, because the bottom line of the reform is to achieve competition, and
competition is indeed expected to decrease prices. Our answer is here again
mixed. Entry legislation is not statistically significant in the electricity price
models, and is associated with lower consumer satisfaction. We find limited
evidence in favour of market opening in gas. Telecom would have been
certainly our key test, because market opening was here earlier and more
radical. We have found, however, no evidence that prices or perception of
prices respond in a discernible way to market opening legislation. We need to
be prudent in the interpretation of these rather disappointing results, because
we cannot test specific pro-competitive actions by each country/sector
regulator, but only the legal provisions per se. This, however, suggests again
that generic market entry legislation, as the one often embodied in EC
directives, is probably inadequate to promote effective competition, and
further research is needed, with more detailed information or the regulatory
environment.
Our bottom line is that we have been able to identify a clear ownership effect, while
the overall evidence on the impact of liberalisation reforms on prices and consumers’
satisfaction is mixed. We turn now to some policy issues raised by our results.
We focus on four issues: a) public versus private ownership; b) ownership versus legal
or accounting unbundling of networks; c) entry regulation, liberalisation, market
structure; d) EU wide market integration, price regulation and consumers’ protection. As
we need to be selective in our discussion, we focus mainly on policy issues related to
consumers (for earlier literature on investment, productivity, and employment, see
Section 1).
a) Consumer prices under privatisation may increase, and this creates discontent.
We have found that privatisation of the network industries, after controlling for other
factors, and despite price regulation, is correlated to higher average consumer prices in
the EU15. Consistently with this finding, we have discovered that consumers tend to be
less satisfied of the price they pay in the countries where the incumbent is under private
ownership. We are not entirely surprised by this finding, because one traditional
objective of public ownership was to offer low prices to consumers (sometimes by cross-
subsidies) even if these sectors were profitable. The result is, however, entirely new and
26
suggests two policy implications. First, while earlier reformers hoped that price caps
were to be removed with full liberalisation (see e.g. Littlechild, 2009), we suggest that
under privatisation, continued monitoring and regulation of prices should be a permanent
feature, because productivity gains are not necessarily passed to consumers (particularly
to most vulnerable ones, as found by Florio and Poggi (forthcoming, 2010) with ECHP
panel data. Second, we suggest that privatization is not a panacea, and that the European
Union must remain neutral about public ownership.
In fact, Article 295 of the EC Treaty states: “ This Treaty shall in no way prejudice the
rules in Member States governing the system of property ownership”. This article was
included in the 1957 Treaty to allow nationalisation of certain industries, and has not
been changed over half century. The European Court of Justice in several decisions
(Hunt, 2009) has clarified the interpretation: Member States are free in establishing
ownership rights, as far as they do not conflict with other aspects of the Treaty, namely
freedom of establishment, free movement of capital or non-discrimination.
The belief that public ownership is necessarily associated with inefficiency, corruption,
capture from vested interests (see e.g. Boycko, Shleifer and Vishny, 2003) must be
assessed against the reality that private ownership under oligopoly in essential services
may also be socially inefficient. Private interests have an incentive to buy regulators and
law-makers in order to be allowed to enjoy market dominance or, under oligopoly, they
may tend to collude. The balance of the public-private inefficiencies varies country by
country, and industry by industry, or even looking at specific segments of the industry.
Moreover, public ownership is not necessarily the enemy of liberalisation. Examples
are the national electricity transmission system operators (TSO) in the Nordic power
system, perhaps one of the most advanced in the world in terms of regional integration
and market opening. Interestingly, the ownership structure of the TSOs across the four
participating countries ranges from Svenka Krafnat (Sweden), a state agency, to the
Danish and Norwegian TSOs (Energinet.dk and Statnett), that are corporations with the
state as sole owner, to the Finnish TSO jointly owner by a private generator, institutional
investors, and the Finnish state. In this context, public ownership of the networks has
probably been an advantage for the promotion of cross-border market integration. Even
in telecoms, the industry where privatisation started in 1984 (British Telecom), public
enterprises did not disappear. For example, one of the most important mobile operator in
the world, Telenor, with a strong position in Norway, Denmark and Sweden is 54%
owned by the State of Norway. In fact some publicly-owned players in the EU has
contributed to competition through foreign direct investment (e.g. the French
government now, through France Telecom, has a stake in players in the UK, Greece,
Poland, Romania, Slovakia, Spain). We conclude that when there is a tradition of
reasonably effective management in the public sector, for example in the Scandinavian
countries, or in France, public ownership of part of the network industry, particularly the
network itself, can still play a role in protecting consumers from oligopolistic
exploitation. This role must be, however, assessed case by case, looking at the specific
institutional environment. Thus, in our view article 295 of the EC Treaty is a wise
provision, in that it delegates to member states to decide whether in their national
27
conditions public versus private provision is still an option to achieve certain objectives
in the public interest.
We suggest that if privatisation is considered on efficiency grounds, its implications
for consumer prices and overall satisfaction must be addressed (including compensation
mechanisms for the poor, who may suffer from tariff re-balancing and other adjustments
of tariff structures following the divestiture of public provision).
b) Ownership unbundling of network is a costly move and benefits are uncertain
The motivation for the EU legislation in electricity and gas that provided for
mandatory unbundling of the networks was to avoid restrictions to third party access to
entrants, in a context where duplication of physical facility is uneconomical. According
to Cameron (2007) vertically integrated companies which own both networks and
generation facilities have an incentive to discriminate against entrants. The First, Second
and the newest Third Energy Package proposed by the EC and adopted by the EU
institutions in the form of several directives (see Hunt, 2009, for a brief review of legal
issues) all included one form of unbundling. The first electricity directive (Directive
96/92/EC) included a provision for either negotiated or regulated third party access, and
a mild form of unbundling. The second Directive (2003) made regulated access
mandatory and provided for accounting unbundling (meaning that separate accounts
should be available for inspection by the regulators). The Third Package, very recently
agreed between the EC and the European Council (May 2009) has proposed to make
ownership unbundling mandatory in all the EU member states. This would have implied
that there should be one network company that owns the network assets, and the same
entity is legally excluded by any direct or indirect control over supply of the service.
This proposal was rejected by Germany, France and several member states and a
compromise solution had to be found. Thus, twelve years after the first directive,
unbundling is still a controversial policy issue. Are Germany and France just protecting
they national champions? Or are they voicing a legitimate concern that ownership
unbundling is inefficient?
Expert opinions are divided. See for example Thomas (2006) for a critical view on
unbundling in Britain, Glachant and Léveque (2005) for a midway view, and Pollitt
(2008) for a defense of ownership unbundling in the EU. Ownership unbundling is a
costly and complex move. The potential reform dimensions are in fact several ones: (a)
private versus public ownership of the network, (b) ownership versus other forms of
unbundling, (c) public versus private ownership of the incumbent. There are much more
than six possible combinations, if we consider different types of public/private
ownership and of unbundling forms. It is difficult to say in abstract terms which one
among all these combination is better in terms of consumers’ welfare.
Given the state of debate among experts, and the uncertainty surrounding the empirical
estimates on this issue, including our own, we suggest that the position of the EC in
favour of mandatory ownership unbundling in 27 Member States and in industries as
different as gas, electricity, telephony, (and railways, etc) is questionable and until now
is certainly not evidence-based.
28
c) Formal entry regulation does not imply competitive markets.
Earlier empirical findings, as our own to a certain extent, suggest that, in general, a
dose of competition pushes down prices, ceteris paribus. It is usually not entry
regulation per se, however, that has this effect. The European Commission itself has
pointed out a number of problems in the implementation of market opening. While
national legislation has been passed over a decade or more (since the Single European
Act, 1987) allowing entry in formerly monopolistic markets, the competitive outcomes
are less than expected. The Energy Sector Inquiry launched by the European
Commission (2007) concluded that
- market concentration is still very high
- markets are still mainly national and only loosely interconnected
- in the 2000s a new waves of mergers has lead to increased oligopolistic
dominance.
- in electricity concentration in generation was mirrored by concentration in spot
and forward markets
- in gas, the market is largely determined by long-term import contracts.
In fixed voice sector, the activity we have considered in our empirical analysis, the EC
Progress report admits that “ fixed incumbents have still not lost significant market
shares, and in some cases, have even strengthened their position”. According to
European Commission (2009), in the EU 27, the incumbents enjoy in 2007 on average
around 65% on retail revenues, and in 2008 more than 80% of fixed subscribers had to
rely on the incumbent’s infrastructure to access alternative networks (cable, unbundled
lines, fibre). In no country the market share of the incumbent is less than 50%.
Interestingly, the market share of the incumbent is now higher in the UK, that started
liberalisation much earlier, than in Germany. The average price for an international
call is now higher in the UK (for business users) than in any other EU country, and is
substantially over the average for residential users as well. Local call charge (10
minutes) in the UK is now higher than in other 21 countries, but it is lower than in
other 16 countries for national 3-minute calls (see European Commission, 2009 for the
details).
Thus, we do not dismiss the importance of liberalisation policy in the network
industries, but it seems that until now consumers have had less than a fair dividend from
the reforms (while the profitability of these industries is still usually quite high and risk
relatively low). We conclude, that having in place an entry legislation that offers some
choice to the consumer, but within domestic oligopolistic markets in fact closed to
international competition, and with a lot of information noise and high search costs
(Wilson and Waddams, 2007), is far from a competitive outcome. We turn then to the
role of price regulation and of EU institutions in this context.
d) Consumers’ price regulation and the internal market: a role for the EU institutions
Our empirical analysis suggests that industry-specific factors are so important in the
determination of prices and consumer welfare, that it is unlikely that there is a unique
29
recipe, in the form of required legislation, for everybody in the EU. This does not imply
that the EU should not play a pro-active role. Going back to the 1987 Single Act, the
overarching goal of the European Commission was to create an internal EU wide market.
As restated, for example, in the Energy Sector Inquiry (EC, 2007, p 4): “ Well-
functioning energy markets that ensure secure energy supplies at competitive prices are
key for achieving growth and consumer welfare in the European Union. To achieve this
objective the EU decided to open up Europe’s gas and electricity markets to competition
and to create a single European energy market”. Similar wording can be read in several
other directives for other SGEI including telecoms and transport. An optimistic view of
the last decade could conclude that progress has been made, albeit slowly, and that
reforms are on the right track. Our interpretation of the empirical evidence is that the EC
has focused excessively on legislation prescribing changes, country by country, in the
regulation of the industries. The result is a very intrusive legislation, now embodied in
literally hundreds of national laws in compliance with the EU directives. There are tens
of sector national regulatory authorities, each authority struggling to implement own
views on how to find a balance between national interests and the European perspective.
This legal outcome is only loosely related to creating a unique internal market, or
protecting the consumer. The frequency of new Directives shows that something is not
working as expected. One possible reason can be the de facto collusion of national
regulators and major players in each country, in order to find local equilibria. However,
nationally regulated oligopolies are different from a European internal market, whatever
the domestic regulations.
Are there alternatives? Our tentative answer is that EU institutions should focus on
consumer protection in a European perspective and on cross-border market integration.
The EU should play its own regulatory role and give more freedom to member states in
the way they solve the issues of ownership, vertical integration, and market design.
For example, the EC has recently acknowledged that roaming tariffs were still
‘ridiculously’ (sic) high. FF
18FF On June, 27, 2008, the EC proposed, and Council agreed, to
impose a EU wide price cap of 11 cents for SMS against the average price of 29 cents.
Here the EC acted not through directives to be embodied in national legislation, but as a
quasi-federal regulator. Interestingly, this happened in a sector that formally is one of the
most open to competition.
This example shows two things: first, formal liberalisation at country level does not
necessarily avoid oligopoly rents at European level; second, the EC has direct powers to
address the issue. Hence, we suggest that, given the mixed empirical support to its
strategy until now, after more than ten years of reforms, the EC should refocus its role.
The effort of creating an internal market for services of general economic interest should
be the core concern of the EU. This implies a policy aimed at abolishing legal, technical
and physical cross-border barriers to trade when this is beneficial to the consumers. We
suggest that this is more promising than trying to achieve exactly the same reform
architecture in each of the 27 Member States. After all, the successful integration of
18 See the website: http://ec.europa.eu/information_society/activities/roaming/tariffs/index_en.htm.
30
European markets was achieved, since its beginning in the 1960s (with six members), by
a totally different strategy. In general, in the first half century of its existence, the EC did
not try to tell any Member State how to reform, e.g., its manufacturing industry, but it
focussed more on lifting cross-country barriers to competition or on preventing
collusion. The EU has been often extremely successful in market integration, and very
questionable in market design (as with the Common Agriculture Policy). The market
integration strategy success has protected European consumers much more than any
other reform. This success story can be replicated in the SGEI by a bolder strategy of EU
institutions. They can use their unique cross-border regulatory powers, and perhaps
enhance articles 81 (anti-competitive agreements) and 82 (abuse of dominant position)
of the EC Treaty. The EU can give more voice to municipalities, consumers associations
or cooperatives, that, better than individual consumers, can discover fair deals abroad
when they are de facto or de jure prevented in one country to enjoy better terms.
At the same time, the EC can direct part of the EU Structural Funds on capital
incentives for wider interconnection infrastructure. At the EU level, a combination of
direct price regulation and of anti-trust policy, seems to us more promising than the
current approach. We suggest that in some countries and for some industries having
public provision or a publicly owned network or a range of different arrangements (part-
privatisation, mixed oligopoly, mutual ownership, etc.) should be allowed without any
interference by the EU, as for article 295 of the Treaty , provided that (a) borders are
open for capital investment and for trade, with the only limitation of national security
(e.g. selling to Gazprom the gas network of a country, that perhaps depends from
Gazprom itself for imports, does not seem a wise form of unbundling); and that (b) any
user is given the concrete right and opportunity to pick up the best possible deal in the
European economic space.
The network industries are still far from the competitive paradigm. One of the leading
British experts in the energy sector, after having reviewed two decades of reform in the
UK, perhaps the inspiring model for the current EC approach, concluded : ‘ in 1980s ad
1990s the pendulum swung too far the other way (from public monopoly). The market
enthusiasts failed to the recognize how far the electricity market deviated from the
normal commodity model. To recap, supply must instantaneously match demand as there
is limited scope for storage: the assets are sunk or long lived, the networks are natural
monopolies. There are very great environmental externalities; and critically, electricity
and gas are complementary to the rest of the economy, in that failure to supply has
(extremely) large costs to all economic activity. It is hard to think of any other activity in
modern developed economies with quite such coincidence of major market failures. If
the issue of fuel poverty and the distributional implications of electricity and gas pricing
and supply are also included, it is extraordinary that anyone could have regarded these
as anything other than political industries”. (Helm, 2003, p.407).
In telecoms (or in other SGEI such as transport or water) there are variations on the
features mentioned by Helm, but market failures and policy concerns are still pervasive
in all network industries. Introducing competitive markets by law is difficult in any
country. Introducing them in 27 countries, in several industries, by adopting one specific
31
model is a cumbersome strategy. Price regulation and consumer’s protection, incentives
to infrastructure investment, and a clear set of common policy objectives, such as
security of supply, environmental protection and adequate long-term investment (that
were often the rationale of vertically integrated public monopolies) are still needed. The
EU is potentially in a better position than the member states to play a direct role in this
policy arena. It just needs to be more directly the protector of the European consumer,
and less the designer of national legislations.
8B8BAcknowledgments
We are very grateful to the three anonymous referees and to the editor for helpful and
constructive comments, and to several colleagues, particularly to Massimiliano Bratti,
Paolo Garella, Marzio Galeotti, Matteo Manera and Franco Peracchi for fruitful
discussions on earlier versions of this paper. We also thank Elisa Borghi for competent
research assistance. This paper is part of a research project “Understanding privatisation:
political economy and welfare effects”, supported by the European Union, VI
Framework Programme, with FEEM as co-ordinating unit, and teams from Milan
University, Pompeu Fabra, Amsterdam University, CERGE-EI (Prague), IFO (Munich),
and OECD. The Milan team was co-ordinated by Massimo Florio (“Welfare Effects on
Consumers”) and included Emanuele Bacchiocchi, Rinaldo Brau, Lidia Ceriani, Raffaele
Doronzo, Pieralda Ferrari, Carlo Fiorio, Marco Gambaro, Simona Grassi, Stefania
Lionetti, Giancarlo Manzi, Ambra Poggi, Riccardo Puglisi, Silvia Salini. Several
working papers related to earlier steps of the project are available in REPEC at
http://edirc.repec.org/data/damilit.html. The usual disclaimer applies.
9B9BReferences
Alesina, A, Ardagna, S., Nicoletti, G. and F. Schiantarelli (2005). Regulation and investment, Journal of the European Economic Association, June, 3(4):791-825
Arano K.G. and Blair B.F. (2008). An ex-post welfare analysis of natural gas regulation in the industrial sector, Energy Economics 30, 789-806.
Arellano, M. and Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Review of Economic Studies, 58:277-97.
Azmat,G., Manning, A. and Van Reenen, J. (2007). Privatization, entry regulation and the decline of labour’s share of GDP: a cross-country analysis of the network industries, CEPR Discussion Paper No. 6348, June.
Bauer, J. (2003). Impact of license fees on the prices of mobile voice service, Telecommunications Policy, 27(5-6), 417-434
Blundell, R.W. and Bond, S.R. (1998). "Initial conditions and moment restrictions in dynamic panel data models," Journal of Econometrics, Elsevier, vol. 87(1), pages 115-143, August
Blundell, R.W. and Bond, S.R. (1998). GMM estimation with persistent panel data: an application to production functions, Econometrics Reviews, 19, 321-340.
Blundell, R.W., Bond, S.R. and Windmeijer, F. (2000), Estimation in dynamic panel data models: improving on the performance of standard GMM estimator in B. Baltagi (ed.) Advances in Econometrics, Volume 15: Nonstationary panels, panel cointegration and dynamic panels, JAI Elsevier Science.
Bond, S. (2002), “Dynamic panel data models: a guide to micro data methods and practice”, Cemmap Working Paper CWP09/02, April.
32
Bond, S., Elston, J., Mairesse, J., Mulkay, B. (2003). Financial factors and investment in Belgium, France, Germany and the UK: a comparison using company panel data. Review of Economics and Statistics 85, 153–177.
Bös, D. (1991), Privatization, A Theoretical Treatment, Clarendon Press, Oxford Bound, J., Jaeger, D.A. and Baker, R.M. (1995), Problems with instrumental variables estimation
when the correlation between the instruments and the endogenous explanatory variable is weak, Journal of the American Statistical Association, vol. 90, No. 430, 443-450.
Boylaud, O. and Nicoletti, G. (2000). Regulation, Market Structure and Performance in Telecommunications, OECD, Economics Department Working Paper, No. 237, OECD Publishing.
Boycko M., Shleifer A., Vishny R, (1996), A theory of Privatization, Economic Journal, 106, 309-319 Ceriani L., Florio, M. (2008). Consumer's Surplus and the Reform of Network industries: A Primer,
DEAS Working Papers n. 23. Chamberlin, E.H. (1933). The Theory of Monopolistic Competition. Harvard University Press. Checchi, D., Florio, M. and Carrera, J. (2009). Privatization discontent and utility reform in Latin
America, Journal of Development Studies, vol.45, march n.3, pp 333-351 CEC, (2007). Evaluation of the Performance of Network Industries Providing Services of General
Economic Interest. 2006 Report, Commission of the European Community. Conway, P. and Nicoletti, G. (2006). Product Market Regulation in the Non-Manufacturing Sectors of
OECD Countries: Measurement and Highlights, OECD Economics Department Working Papers 530 Copenhagen Economics (2005): Market Opening in Network Industries, Final Report prepared for DG
Internal Market, European Commission, Brussels Cournot, A. (1897). Researches into the Mathematical Principles of the Theory of Wealth. The
Macmillan Company. Edwards G. and Waverman L. (2006). The Effects of Public Ownership and Regulatory Independence
on Regulatory Outcomes, Journal of Regulatory Economics 29(1), 23-67 Espey, J.A, Espey, M. (2004). Turning on the lights: a meta-analysis of residential electricity demand
elasticities, Journal of Agricultural and Applied Economics, April, 36:1,pp 65-81. Estache A., Goicoechea A. and Manacorda M. (2006). Telecommunications Performance, Reforms,
and Governance, World Bank Policy Research Working Paper 3822. Estache A., Goicoechea A., Trujillo L. (2006). Utilities reforms and corruption in developing
countries, World bank Policy Research Working paper 4081. European Economy, (2007). Office for Official Publications of the EC. Luxembourg, July. Florio, M. (2004). The Great Divestiture. Evaluating the welfare impact of British privatizations 1979-
1997. The MIT Press. Galal, A. J., Leroy, T. and Vogelsang, I. (1994). Welfare consequences of selling public enterprises.
Oxford University Press. Gasmi F., Noumba Um P. and Virto, L. R. (2006). Political accountability and regulatory performance
in infrastructure industries: an empirical analysis, World Bank Policy Research Working Paper Series No. 4101.
Gassner K, Popov A., Pushak N. (2009). Does private sector participation improve performance in electricity and water distribution?, The World Bank
Grzybowski, L. (2005). Regulation of Mobile Telephony across the European Union: An Empirical Analysis, Journal of Regulatory Economics, 28 (1) , 47-67
Grzybowski, L. (2008). The impact of regulation on the retail prices in fixed-line telephony across the European Union, Telecommunications Policy Vol. 32(2), 131-144.
Hart O, Shleifer A., Vishny R (1997), The proper scope of government with an application to prisons, Quartely Journal of Economics
Hattori, T. and Tsutsui, M. (2004). Economic impact of regulatory reforms in the electricity supply industry: a panel data analysis for OECD countries, Energy Policy Vol. 32 (6), 823-832.
Holtz-Eakin, D., W. Newey, and H.S. Rosen. (1988) Estimating vector autoregressions with panel data. Econometrica 56: 1371–95.
Kahneman D, Knetsch, R.h, Thaler H, (1986) Fairness as a constraint to profit-seeking: entitlements in the market, American Economic Review, 76 (4), pp 728-741.
Laffont, J.J. (2005). Regulation and Development, Cambridge University Press Laffont J.J and Tirole J.(1993), A Theory of Incentives in Procurement and regulation, MIT Press,
Cambridge (Mass) Li, W. and Xu, L. C. (2004). The Impact of Privatization and Competition in the Telecommunications
Sector around the World, Journal of Law & Economics, October, 47(2), 395-430
33
Martin R. and Vansteenkiste I. (2001). EU telecommunications and electricity markets-Heading towards price convergence?, Intereconomics: Review of European Economic Policy Vol. 36(3), 131-140
Megginson W.L. (2005), The Financial Economics of Privatization, Oxford University Press Millward R. (2005). Private and Public Enterprise in Europe:Energy, Telecommunications and
Transport c.1830-1990, Cambridge University Press Newbery D. (2000), Privatization, Restructuring and Regulation of Network Industries, MIT Press,
Cambridge (Mass.) Nicoletti, G. and Scarpetta, S. (2003). Regulation, productivity and growth: OECD evidence,
Economic Policy, vol. 18(36), pp 9-72, 04. Nickel, S.J. (1981), Biases in dynamic models with fixed-effects, Econometrica, 49, 1417-1426. Parker D. (1998), Privatization in the European Union. Theory and Policy Perspectives, Routledge,
London Percebois J., (2008).Electricity Liberalization in the European Union: Balancing Benefits and Risks,
The Energy Journal, vol 20, pp 1-19 Pollitt M., (2007). The arguments for and against ownership unbundling of energy transmission
networks, Cambridge Working Papers in Economics, n 0737. Pollitt, M.G. and Newbery, D.M.G. (1997). Restructuring and Privatisation of the CEGB - was it
worth it?, Journal of Industrial Economics, Vol.45, No.3, pp.269-304 Rey P and Vergé T. (2008), Economics of Vertical restraints, in Buccirossi (ed), Handbook of Anti-
Trust Economics, MIT Press, Cambridge (Mass), pp 353-390 Rey P and Tirole J, (2007) A primer on foreclosure, in Armstrong M. and Porter Robs (eds),
Handbook of Industrial Organization, Vol. III, North Holland, pp 2145-2220 Roland G. (2008), Privatization: Success and failures, Columbia University Press Roodman, D. (2009). How to do xtabond2: An introduction to difference and system GMM in Stata,
Stata Journal 9(1): 86-136. Sappington D. and Stiglitz J., (1987), Privatization, Information, and Incentives, Journal of Policy
Analysis and Management, 6, 567-582 Schmidt K, (1996), The costs and benefits of privatization, An incomplete contracts approach, Journal
of law, Economics, and Organization, 12. 1-24 Shapiro C and Willig R. (1990), Economic rationales for the scope of privatization, in E.N. Suleiman
E.N. and Waterbury J (eds), The political economy of public sector reform and privatization, Westview Press, Boulder (Colorado), 55-87
Steiner F. (2000). Regulation, Industry Structure and Performance in the Electricity Supply Industry, OECD, Economics Department Working Papers No. 238.
Ugaz, C. and Waddams Price, C. (2003). Utility Privatization and Regulation: a Fair Deal for Consumers? Edward Elgar. Cheltenham.
Vagliasindi M., (2008). Governance Arrangements for State Owned Enterprises, The World Bank, Policy Research Working Paper 4542, Washington DC.
Varian, H.R. (1987). Intermediate Microeconomics. A Modern Approach. W.W. Norton and Company.
Viani, B. E. (2007). Consequences of vertical separation and monopoly: Evidence from the Telecom Privatizations, AEI-Brookings Joint Center for Regulatory Studies Working Papers 06-20.
Vickers J, Yarrow G. (1993), Privatization. An Economic Analysis, MIT Press, Cambridge (Mass) Wallsten S. (2001). An Empirical Analysis of Competition, Privatization, and Regulation in Africa and
Latin America, Journal of Industrial Economics, Vol. 49(1), 1-19. Zhang, Y., Kirkpatrick, C. and Parker, D. (2002). Electricity Sector Reform in Developing Countries:
An Econometric Assessment of the Effects of Privatisation, Competition and Regulation, University of Manchester, Institute for Development Policy and Management (IDPM) Working papers No. 30593.
34
02
46
1975 1980 1985 1990 1995 2000 2005year
Countries Median
25th perc. 75th perc.
ECTR Score (0-6)
Telecoms
02
46
1975 1980 1985 1990 1995 2000 2005year
Countries Median
25th perc. 75th perc.
ECTR Score (0-6)
Electricity
12
34
56
1975 1980 1985 1990 1995 2000 2005year
Countries Median
25th perc. 75th perc.
ECTR Score (0-6)
Gas
35
Figure 1: The yearly trend of the ECTR country score (0-6) for the telecom, electricityand natural gas industries.
0
24
6
1975 1980 1985 1990 1995 2000 2005year
Countries Median
25th perc. 75th perc.
ECTR Score (0-6)
Telecoms: Public Ownership
12
34
56
1975 1980 1985 1990 1995 2000 2005year
Countries Median
25th perc. 75th perc.
ECTR Score (0-6)
Telecoms: Market Structure
02
46
1975 1980 1985 1990 1995 2000 2005year
Countries Median
25th perc. 75th perc.
ECTR Score (0-6)
Telecoms: Entry Regulation
Figure 2. The three dimensions of the ECTR country score for the telecoms industry.
36
02
46
1975 1980 1985 1990 1995 2000 2005year
Countries Median
25th perc. 75th perc.
ECTR Score (0-6)
Electricity: Public Ownership
02
46
1975 1980 1985 1990 1995 2000 2005year
Countries Median
25th perc. 75th perc.
ECTR Score (0-6)
Electricity: Entry Regulation
02
46
1975 1980 1985 1990 1995 2000 2005year
Countries Median
25th perc. 75th perc.
ECTR Score (0-6)
Electricity: Vertical Integration
Figure 3. The three dimensions of the ECTR country score for the electricity industry.
37
02
46
1975 1980 1985 1990 1995 2000 2005year
Countries Median
25th perc. 75th perc.
ECTR Score (0-6)
Gas: Public Ownership
12
34
56
1975 1980 1985 1990 1995 2000 2005year
Countries Median
25th perc. 75th perc.
ECTR Score (0-6)
Gas: Market Structure
12
34
56
1975 1980 1985 1990 1995 2000 2005year
Countries Median
25th perc. 75th perc.
ECTR Score (0-6)
Gas: Vertical Integration
02
46
1975 1980 1985 1990 1995 2000 2005year
Countries Median
25th perc. 75th perc.
ECTR Score (0-6)
Gas: Entry Regulation
Figure 4. The four dimensions of the ECTR country score for the natural gas industry.
38
12
34
56
ER
60
70
80
90
100
110
Price
1975 1980 1985 1990 1995 2000 2005year
Price ER
Source: IEA
Electricity
23
45
GR
80
100
120
140
160
Price
1975 1980 1985 1990 1995 2000 2005year
Price GR
Source: IEA
Gas
12
34
56
TR
70
80
90
100
Price
1975 1980 1985 1990 1995 2000 2005year
Price TR
Source: ITU
Telecoms
Note: Price deflated using the national CPI (Source: WDI).
12
34
56
ER
60
70
80
90
100
110
Price
1975 1980 1985 1990 1995 2000 2005year
Price ER
Source: IEA
Electricity
23
45
GR
80
100
120
140
160
Price
1975 1980 1985 1990 1995 2000 2005year
Price GR
Source: IEA
Gas
12
34
56
TR
70
80
90
100
Price
1975 1980 1985 1990 1995 2000 2005year
Price TR
Source: ITU
Telecoms
Note: Prices deflated using the national consumer price index.
Figure 5: Trends of median household prices (left vertical scale) and ECTR scores (right scale) in each sector. ER, GR and TR are EU15 average of ECTR country
scores (see tables 1-3 for exact definitions).
39
Figure 6. Prices and consumer's surpluses under different regimes.
40
ELECTRICITY Original ECTR coding Our coding
Country scores 0-6 Sector
indic. Binary variable 0-1
Question weights
Weights by theme
Our label Coding Our label
PUBLIC OWNERSHIP:
What is the ownership structure of the largest companies in the generation, transmission, distribution, and supply segments of the electricity industry? (ERpo1)
1/3 1/3 1 if ownership is
public ERpo_d
ENTRY REGULATION:
How are the terms and conditions of third party access (TPA) to the electricity transmission grid determined? (ERen1)
1/3
Is there a liberalised wholesale market for electricity (a wholesale pool)? (ERen2)
1/3
What is the minimum consumption threshold that consumers must exceed in order to be able to choose their electricity supplier ? (ERen3)
1/3
1/3
1 if wholesale market for elect. is not liberalised & consumption
threshold is larger than 1MWatts
ERen_d
VERTICAL INTEGRATION:
What is the degree of vertical separation between the transmission and generation segments of the electricity industry? (ERvi1)
1/2
What is the overall degree of vertical integration in the electricity industry? (ERvi2)
1/2
1/3
ER
1 if overall degree of vertical
integration in the industry is mixed
or integrated.
ERvi_d
Source: ECTR coding taken from Conway and Nicoletti (2006).
Table 1: The ECTR indicator for the electricity industry with our coding.
41
GAS Original ECTR coding Our coding
Country scores 0-6
Sector
indic. Binary variable 0-1
Question weights Weights by theme
Our label
Coding (zero otherwise)
Our label
PUBLIC OWNERSHIP:
What percentage of shares in the largest firm in the gas production/import sector are owned by government?
What percentage of shares in the largest firm in the gas transmission sector are owned by government?
1/3 1/4
What percentage of shares in the largest firm in the gas distribution sector are owned by government?
1/3
1 if 100% of ownership of shares in all
segments of the industry is
public
GRpo_d
ENTRY REGULATION:
How are the terms and conditions of third party access (TPA) to the gas transmission grid determined?
1/3
What percentage of the retail market is open to consumer choice?
1/3
Do national, state or provincial laws or other regulations restrict the number of competitors allowed to operate a business in at least some markets in the sector: gas production/import
1/3
1/4
1 if no third party access, no
consumers' choice in the retail market,
and restrictions operate in all
markets.
GRen_d
MARKET STRUCTURE:
What is the market share of the largest company in the gas production/import industry?
1/3
What is the market share of the largest company in the gas transmission industry?
1/3
What is the market share of the largest company in the gas supply industry?
1/3
1/4
1 if market share in all
segments of the industry is
larger than 90%
GRms_d
VERTICAL INTEGRATION:
What is the degree of vertical separation between gas production/import and the other segments of the industry?
1/3
What is the degree of vertical separation between gas supply and the other segments of the industry? 1/3
Is gas distribution vertically separate from gas supply? 1/3
1/4
GR
1 if the industry is integrated in
all segments GRvi_d
Source: ECTR coding taken from Conway and Nicoletti (2006).
Table 2: The ECTR indicator for the natural gas industry with our coding.
42
TELECOM Original ECTR coding Our coding
Country scores 0-6
Sector
indic. Binary variable 0-1
Question weights Weights by theme
Our label
Coding (zero otherwise)
Our label
PUBLIC OWNERSHIP:
What percentage of shares in the PTO are owned by government?
1-wm
What percentage of shares in the largest firm in the mobile telecommunications sector are owned by government?
wm
1/3
1 if 100% of shares in the
PTO are owned by government
TRpo_d
ENTRY REGULATION:
What are the legal conditions of entry into the trunk telephony market?
1/4*wt*(1-wm)
What are the legal conditions of entry into the international market?
1/4*(1-wt)(1-wm)
What are the legal conditions of entry into the mobile market?
1/2*wm
1/3
1 if legal condition of entry into the
trunk tel. market is franchised to
one firm
TRen_d
MARKET STRUCTURE:
What is the market share of new entrants in the trunk telephony market?
1/4*wt*(1-wm)
What is the market share of new entrants in the international telephony market?
1/4*(1-wt)(1-wm)
What is the market share of new entrants in the mobile market?
1/2*wm
1/3
TR
100-market share of new
entrants in the trunk telephony
market (a)
TRms
Source: ECTR coding taken from Conway and Nicoletti (2006).
Notes: The weight wm is the OECD-wide revenue share from mobile telephony in total revenue from trunk, international, and mobile. The weight wt is the OECD-wide revenue share of trunk in total revenue from trunk and international telephony. "PTO" stands for "Public telecommunications operator". (a) the variable TRms is the only non-binary variable as it is defined as 100 minus the share of new entrants in the trunk telephony market variable, going from zero (new entrants replace the incumbent) to 100 (all market share to the incumbent).
Table 3: The ECTR indicator for the telecommunication industry with our coding.
43
1978 1990 1995 2000 2003 2004 2005 2006
ELECTRICITY
price 0.14 0.11 0.11 0.09 0.08 0.08 0.08 0.08 France
ECTR score 6.00 6.00 6.00 4.28 3.61 2.61 2.61 2.11
price 0.14 0.13 0.10 0.11 0.09 0.10 0.10 0.12 UK
ECTR score 6.00 0.83 0.11 0.00 0.00 0.00 0.00 0.00
price 0.14 0.12 0.11 0.10 0.09 0.09 0.10 0.11 EU15-median
ECTR score 6.00 5.50 5.50 2.78 2.00 1.61 1.50 1.50
price 0.24 0.04 0.03 0.02 0.02 0.02 0.02 0.02 EU15-st. dev.
ECTR score 0.61 1.36 1.57 1.54 1.05 0.96 0.82 0.81
GAS
price 12.63 8.89 8.11 7.65 8.87 8.26 8.78 10.22 France
ECTR score 6.00 6.00 6.00 6.00 3.67 3.49 2.24 2.24
price 8.75 8.33 6.35 7.21 6.64 7.06 7.81 10.02 UK
ECTR score 5.75 3.50 3.03 1.90 1.65 1.10 0.73 0.73
price 8.85 8.33 7.56 7.29 7.41 8.26 8.98 10.03 EU15-median
ECTR score 4.95 4.70 4.50 4.35 2.69 2.69 2.24 2.24
price 6.62 3.47 2.29 1.96 2.64 2.68 2.64 2.85 EU15-st. dev.
ECTR score 0.83 0.87 0.87 1.23 1.12 1.24 1.17 1.12
TELECOMS
price - 0.12 0.12 0.11 0.12 0.12 0.15 0.14 France
ECTR score 6.00 5.96 5.59 2.52 2.06 2.03 1.69 1.67
price - - 0.17 0.19 0.16 0.16 0.15 0.15 UK
ECTR score 6.00 3.74 1.33 0.60 0.46 0.49 0.52 0.55
price - 0.15 0.13 0.12 0.12 0.12 0.11 0.12 EU15-median
ECTR score 6.00 6.00 4.83 1.83 1.49 1.37 1.27 1.17
price - 0.05 0.04 0.03 0.20 0.23 0.04 0.04 EU15-st. dev.
ECTR score 0.34 0.65 1.39 0.94 0.72 0.69 0.66 0.64
Source: Authors' calculations using IEA, ITU, WDI and ECTR data.
Note: Prices are deflated using the national consumer price index.
Table 4: Some statistics on price trends for EU15, France and the UK.
44
Electricity Gas Telecom
Austria 2001 2001 1997
Belgium 2000 2000 1993
Denmark 2000 2000 1991
Finland 1995 2000 1988
France 2000 2003 1997
Germany 2005 2005 1998
Greece 2000 2000 1995
Ireland 1999 1999 1997
Italy 1997 1997 1997
Luxembourg 2001 2002 1997
Netherlands 1999 1999 1997
Portugal 1997 2000 1989
Spain 2000 2000 1996
Sweden 1998 2000 1992
United Kingdom 1986 1989 1983
Source: various sources, mainly individual authorities websites. Notes: We record the establishment year of the current regulatory office, but in some countries and sectors there were restructuring of the offices (including merges) which are not recorded here
Table 5: Year of establishment of national independent regulatory authorities in each sector considered
across the EU15.
45
Paper Sector and Dependent variable(s)
Regulatory indicators Controls Sample Main results
Boylaud and Nicoletti (2000)
TELECOM International calls: OECD tariff basket Revenue/outgoing minutes; Trunk service: OECD tariff basket
Degree of liberalization (# of competitors) Index of state ownership; Degree of internationalization (# of non-resident operators); Actual market structure (share of new entrants); Time to liberalization/Time to privatization
Income level Population density Input costs Percentage of digital lines Capital intensity
23 OECD countries 1991-1997 Panel FE
The degree of mkt.competition and the time to liberalization index are the two main explanations for reduction in prices. State ownership has a slightly significant positive effect on prices of international services and a negative significant effect on trunk services’ prices. Time to privatization and internat. of domestic markets. do not affect prices.
Edwards and Waverman (2006)
TELECOM Interconnection rate: per minute for call termination on incumb. fixed line networks
Government’s ownership share of the incumbent Public dummy EURI-I: index of regulatory independence. Number of years since liberalization
Population density Urbanization Lines (total #)
EU 15 1997-2003 Pooled OLS (time dum.) IV
Public ownership has a positive and significant effect on the interconnection rate. Regulatory independence has a negative and significant effect.
Estache, Goicoechea and Manacorda (2006a)
TELECOM Price of a 3-minute local phone call
Privatization (dummy) IRA (= 1 if an independent regulatory agency has been established). IRA*Privatization;Privatization*Developed countries dummy;Corruption of the political system (index);IRA * Corruption;Invest. risk index. Privat.*inv. risk; IRA*inv risk
GDP per capita Population Time trend
204 countries 1990– 2003 Panel FE with linear time trend
Looking at marginal effects, to capture the effect of explanatory variables and interaction terms, privatization has no effect, while IRA has a negative and significant effect both for developing and developed countries.
Gasmi, Noumba and Recuero Virto (2006)
TELECOM Residential monthly subscription to fixed
Regulatory governance index. Institutionalization (Corruption, bureaucracy, law and order, expropriation risk, currency risk). Checks and balances: # of veto players in a political system; Privatization (% sold to private in developing c.; dummy 0=state-owned). Comp.: 0 = monop.1 = o.w.
29 devel.c. 23devel. c. 1985–1999 DIF-GMM SYS-GMM
A lower expropriation risk or stronger checks and balance have a positive impact on prices in developing countries. Regulation and competition have a negative and slightly significant effect on prices in developing countries. For developed cou., regulation has no effect; currency risk is the only significant variable (negative impact).
Grzybowski (2008)
TELECOM Price of fixed-line services;Price for local and national, peak and off-peak 5-minute call.
Number portability in fixed-lines networks Carrier pre-selection of national calls Enforcing of unbundling of local loops Liberalization of the telecommunications market.Interconnection regulation (Access)
Average hourly labour cost in manufacturing Cost of capital: 10-year government bonds GDP per capita;Time trend
EU 15 1998 –2002 Panel FE (time trend)
Liberalization has a negative and significant effect on the basket prices and on national calls, both at peak and off-peak times. Portability has a negative and significant effect on local calls’ prices only.
Li and Xu (2004)
TELECOM Local price index (3-minute local call)
Full privatization (1 = control to private) Partial privatization (1 = control to public) Competition index (0 = national monopoly; 1 = more than one operator in either fixed-line or mobile market segment; 2 = more than one operator in both).
Population Share of urban population GDP per capita Consumer price index
177 countries 1990-2001 Panel FE (time trend)
Full privatization has a positive and significant effect. Partial privatization has a negative and slightly significant effect. Competition has a negative and significant effect only for countries with privatization.
Viani (2007)
TELECOM: Resid. price: price of one year of resid. local fixed service. Resid. price = Connect. charge + 12*month. rate.
Local monopoly: # of years until end of monopoly on local fixed telephony Vertical separation (# of years since separation)
Real GDP p.c. Pop. (active and > 65) Left-wing (party of chief executive); Residential lines; Faults/lines; Digital (% digitalization)
67 countries 1984-2003 Panel FE (time dummies)
Monopoly has a positive and significant effect on prices. No effect of vertical separation
46
(continued from previous page)
Wallsten (2001)
TELECOM Price of a 3-minute local call
Privatization = 1 if the incumbent has been privatized # of mobile operators (competition) Regulation = 1 if the country has a separate independent regulatory agency
GDP per capita; Urban pop. (%); Pop. Size; Risk of expropriation WB telecom project WB aid/GDP; Exports/GDP
30 count. 1984-1997 Panel FE
Competition has a negative and highly significant effect. Privatization and regulation have no effect on prices. The presence of a WB program has a positive effect on prices.
Estache, Goicoechea and Trujillo (2006)
TELECOM Price of e 3-minute local phone call ELECTRICITY Electricity end-user prices for households
IRA = establishment of an independent regulatory agency. Privatization (dummy for private participation) IRA*Privatization Corruption in the political system IRA*Corruption
GDP per capita Urbanization Time trend
47 count. (electr.) 91 countries (telecom) 1990-2002 Panel GLS (time trend)
Telecom: IRA and corruption have a positive and significant effect. Privatization has a negative and significant effect. Electricity: IRA has a negative and significant effect, privatization a positive and significant effect, corruption has no effect.
Martin and Vansteen-kiste (2001)
TELECOM Annual average prices for a 3-minute peak call (local, long distance, international). ELECTRICITY Semi-annual prices for small and large households.
Number of years prior liberalization Opening of the market to competition (dummy) % of publicly owned incumbent’s shares. Price of leased lines (Telecom) Carrier-select facilities (Telecom) Price of gas (Electricity) Percentage of electricity generated by gas (Electricity)
EU 15 1995 -2000 1990-2000 Panel FE
Telecom: Negative effect of liberalization for international and long distance calls. No effect on local calls. Negative effect of privatization for international and long-distance fixed-line calls, no effect prices of local calls. Carrier selection and number portability have a negative effect on local calls’ prices. Electricity: No effect of liberalization. Positive effect of public ownership only for small households.
Hattori and Tsutsui (2004)
ELECTRICITY Industrial price before taxes Industrial price/household price
Unbundling of generation Third party access Wholesale power market Private ownership Time to liberalization Time to privatization
Share of hydro capacity Share of nuclear capacity GDP
19 OECD 1987-1999 Panel FE and RE
Third party access has a negative and significant effect on both price variables. Unbundling has a positive and slightly significant effect on prices and a negative effect on the ratio. The existence of a wholesale power market has a positive and slightly significant effect on prices and the ratio. Private ownership has a negative effect on prices and no effect on the ratio.
Steiner (2001)
ELECTRICITY Industrial price before taxes Industrial price/household price
Time to liberalization Time to privatization Ownership Unbundling of generation and transmission Third Party Access Wholesale pool
GDP Hydro share in generation Nuclear share in generation Urbanization
19 OECD 1991-1996 1986-1996 Panel RE
Unbundling has a negative effect on the ratio and no effect on the price. TPA and the introduction of a wholesale market have a negative effect on both variables, while ownership has a positive effect. Imminence of liberalization and privatization has a negative effect on prices.
Table 6: Selected empirical papers on the effect of regulation on telecom and electricity prices.
47
ELECTRICITY GAS TELECOMS
OLS Within OLS Within OLS Within
log price (t-1) 0.972*** 0.838*** log price (t-1) 0.957*** 0.812*** log price (t-1) 0.789*** 0.667***
(lEAprinet_kw) (0.000) (0.000) (lGAprinet_gj) (0.000) (0.000) (lTIpri2) (0.000) (0.000)
N. obs. 414 414 N. obs. 328 328 N. obs. 199 199
ar1p 0.266 0.203 ar1p 0.412 0.076 ar1p 0.383 0.131
ar2p 0.975 0.820 ar2p 0.175 0.181 ar2p 0.661 0.889
Source: Authors' calculations using IEA data.
Notes: The price models estimated are: 1t t t itP Pα δ ε−= + + and are estimated separately for each sector. Variables are defined as in Section 5.
The price variables are lEAprinet_kw, lGAprinet_gj, lTIpri2, respectively for the electricity, gas and telecom sectors. For variable definitions and sources, refer to Table A1. Robust p-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Year dummies included in all models. Asymptotic standard erros in parentheses. ar1p and ar2p report the p-values of tests for first-order and second-order serial correlation, asymptotically N(0,1). These test the levels residual for OLS and Within groups estimation, and the first-differenced residuals in all other columns.
Table 7. AR(1) specification for the series of prices by sector.
ELECTRICITY GAS TELECOM
Dep. variable is 1tP−
Regulatory variables
ECTR score 0-6 yes no yes no yes no
Binary variable 0-1 no yes no yes no yes
year-fixed effect yes yes yes yes yes yes
Observations 417 417 334 334 202 202
R-squared 0.451 0.493 0.293 0.515 0.165 0.185
Prob>F 0.000 0.000 0.000 0.000 0.003 0.002
Prob>F1 0.005 0.000 0.636 0.000 0.002 0.003
Source: Authors' calculations using IEA and ECTR source data.
Note: The models estimated are: , 1 ,'i t i t t itP R β δ ε− = + + and are estimated separately for each sector. Variables are defined as in Section 5.
The price variables are are lEAprinet_kw, lGAprinet_gj, lTIpri2, respectively for the electricity, gas and telecom sectors. For variable definitions and sources, refer to Table A1. The null hypothesis of the F test is that all coefficients are jointly zero. The null hypothesis of the F1 test is that regulatory variables coefficients are jointly zero.
Table 8: Regressions of one-period lagged price variable over regulatory variables and year fixed effects.
48
Regulatory variables: Score 0-6 Regulatory variables: Binary variables 0-1
OLS Within GMM-Dif GMM-Dif GMM-Sys GMM-Sys OLS Within GMM-Dif GMM-Dif GMM-Sys GMM-Sys
log price (t-1) 0.968*** 0.732*** 0.747*** 0.772*** 0.931*** 0.914*** 0.957*** 0.670*** 0.618*** 0.634*** 0.917*** 0.901***
(lEAprinet_kw) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Score 0-6 0.004 -0.003 0.012 0.014 0.013* 0.013**
(ER) (0.390) (0.612) (0.184) (0.179) (0.050) (0.042)
Public. Ownership -0.006 -0.078*** -0.078** -0.072*** -0.041*** -0.047***
(ERpo_d) (0.554) (0.003) (0.021) (0.004) (0.000) (0.000)
Entry regulation 0.013 -0.027 -0.021 -0.019 0.027* 0.029**
(ERen_d) (0.510) (0.162) (0.191) (0.248) (0.055) (0.029)
Vertical integration 0.012 0.099*** 0.098*** 0.094*** 0.039** 0.041**
(ERvi_d) (0.548) (0.000) (0.000) (0.000) (0.035) (0.019)
Source: Hydroelectric -0.001 -0.007* -0.008*** -0.008*** -0.001 -0.002 -0.002 -0.008* -0.009*** -0.009*** -0.003** -0.003*
(lEAshydr) (0.402) (0.099) (0.000) (0.000) (0.256) (0.144) (0.295) (0.061) (0.000) (0.000) (0.038) (0.056)
Imports 0.001 -0.002 -0.001 -0.000 0.000 0.000 0.000 -0.002 -0.002** -0.002** -0.001*** -0.002***
(lEAimports) (0.571) (0.234) (0.733) (0.954) (0.492) (0.642) (0.866) (0.297) (0.045) (0.042) (0.005) (0.001)
Distribution losses 0.029* -0.021 -0.026 -0.026 0.025 0.025 0.032** -0.030 -0.032 -0.031 0.045** 0.046*
(lEAendiloss) (0.061) (0.535) (0.450) (0.427) (0.261) (0.322) (0.046) (0.367) (0.312) (0.325) (0.044) (0.064)
Log residential consumption -0.019 -0.096** -0.094*** -0.090*** -0.005 -0.004 -0.022 -0.133*** -0.142*** -0.138*** -0.029 -0.029
(lEArescons) (0.244) (0.012) (0.008) (0.004) (0.837) (0.886) (0.196) (0.001) (0.000) (0.000) (0.181) (0.216)
Log per capita GDP -0.011 0.179*** 0.152*** 0.138*** -0.029 -0.033 -0.013 0.181*** 0.201*** 0.195*** -0.021 -0.025
(lMWgdppc) (0.436) (0.000) (0.000) (0.000) (0.285) (0.228) (0.392) (0.000) (0.000) (0.000) (0.313) (0.242)
Log cost of oil 0.005 0.041* 0.031 0.029 -0.004 -0.006 0.004 0.053** 0.050* 0.050* 0.008 0.007
(lEAcoil) (0.793) (0.083) (0.390) (0.400) (0.892) (0.849) (0.844) (0.023) (0.094) (0.092) (0.747) (0.786)
N. obs. 282 282 267 267 282 282 282 282 267 267 282 282
ar1p 0.385 0.145 0.004 0.004 0.002 0.003 0.361 0.280 0.006 0.008 0.002 0.002
ar2p 0.358 0.906 0.288 0.278 0.144 0.139 0.370 0.496 0.287 0.284 0.145 0.143
N. instr. 91 90 93 92 130 129 130 129
Instr. (periods of dep.var.) t-2,t-3,…1 t-3,t-4,...,1 t-2,t-3,…1 t-3,t-4,...,1 t-2,t-3,…1 t-3,t-4,...,1 t-2,t-3,…1 t-3,t-4,...,1
Source: Authors' calculations using IEA, WDI source data. IEA data used for price series. For exact source and variable definition refer to the label (in italics in the first column) and Table A1. Notes: Robust p-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Year dummies included in all models. Asymptotic standard erros in parentheses. ar1p and ar2p report the p-values of tests for first-order and second-order serial correlation, asymptically N(0,1). These test the levels residual for OLS and Within groups estimation, and the first-differenced residuals in all other columns. GMM results are one-step estimates with heteroskedasticity-consistent standard errors and test statistics. Instruments used in all GMM equations include the predetermined regulatory variable (at time t-1 and possibly earlier) and all exogenous variables. All calculations done using outreg2 for Stata (Roodman, 2009).
Table 9: The price models for the electricity industry.
49
Regulatory variables: Score 0-6 Regulatory variables: Binary variables 0-1
OLS Within GMM-Dif GMM-Dif GMM-Sys GMM-Sys OLS Within GMM-Dif GMM-Dif GMM-Sys GMM-Sys
log price (t-1) 0.954*** 0.813*** 0.785*** 0.768*** 0.969*** 0.962*** 0.916*** 0.808*** 0.830*** 0.843*** 0.934*** 0.912***
(lGAprinet_gj) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Score 0-6 0.001 0.009 0.018 0.017 -0.001 -0.002
(GR) (0.803) (0.381) (0.307) (0.308) (0.869) (0.760)
Public. Ownership 0.025 -0.047 -0.079** -0.078** 0.002 0.014
(GRpo_d) (0.158) (0.138) (0.042) (0.034) (0.917) (0.458)
Entry regulation 0.035** 0.046** 0.041* 0.044* 0.054*** 0.039**
(GRen_d) (0.015) (0.029) (0.086) (0.082) (0.009) (0.039)
Market structure -0.050*** 0.008 0.034 0.019 -0.037** -0.048***
(GRms_d) (0.002) (0.840) (0.342) (0.568) (0.029) (0.005)
Vertical integration 0.005 0.008 0.009 0.005 0.004 0.015
(GRvi_d) (0.640) (0.646) (0.375) (0.621) (0.551) (0.119)
Log change of price Brent oil 0.729** 0.609** 0.567*** 0.553*** 0.754*** 0.751*** 0.678** 0.638** 0.671*** 0.679*** 0.772*** 0.754***
(lDGAbrent) (0.017) (0.041) (0.001) (0.001) (0.000) (0.000) (0.023) (0.032) (0.000) (0.000) (0.000) (0.000)
Log per capita GDP -0.009 0.030 0.030 0.030 -0.002 -0.005 -0.020 0.066 0.081** 0.075** 0.006 -0.013
(lMWgdppc) (0.621) (0.490) (0.488) (0.520) (0.859) (0.763) (0.373) (0.160) (0.015) (0.035) (0.710) (0.619)
N. obs. 328 328 314 314 328 328 328 328 314 314 328 328
ar1p 0.389 0.077 0.016 0.016 0.022 0.023 0.510 0.144 0.019 0.020 0.015 0.020
ar2p 0.175 0.174 0.475 0.471 0.514 0.515 0.0600 0.179 0.598 0.607 0.521 0.491
N. instr. 283 273 105 103 167 167 213 212
Instr. (periods of dep.var.) t-2,t-3,…,1 t-3,t-4,…,1 t-2,t-3,…,1 t-3,t-4,…,1 t-2 t-3 t-2 t-3
Source: Authors' calculations using IEA, WDI source data. IEA data used for price series. For exact source and variable definition refer to the label (in italics in the first column) and Table A1. Notes: Robust p-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Year dummies included in all models. Asymptotic standard erros in parentheses. ar1p and ar2p report the p-values of tests for first-order and second-order serial correlation, asymptically N(0,1). These test the levels residual for OLS and Within groups estimation, and the first-differenced residuals in all other columns. GMM results are one-step estimates with heteroskedasticity-consistent standard errors and test statistics. Instruments used in all GMM equations include the predetermined regulatory variable (at time t-1 and possibly earlier) and all exogenous variables. All calculations done using outreg2 for Stata (Roodman, 2009).
Table 10: The price models for the natural gas industry.
50
Regulatory variables: Score 0-6 Regulatory variables: Binary variables 0-1
OLS Within GMM-Dif GMM-Dif GMM-Sys GMM-Sys OLS Within GMM-Dif GMM-Dif GMM-Sys GMM-Sys
log price (t-1) 0.739*** 0.652*** 0.639*** 0.631*** 0.822*** 0.852*** 0.739*** 0.657*** 0.634*** 0.613*** 0.827*** 0.807***
(lTIpri2) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Score 0-6 0.008 0.041 0.041* 0.041** 0.007 0.013
(TR) (0.666) (0.188) (0.054) (0.047) (0.564) (0.322)
Public. ownership 0.001 -0.001 -0.021 -0.024 -0.054* -0.099***
(TRpo_d) (0.987) (0.983) (0.655) (0.694) (0.073) (0.005)
Market structure -0.000 0.002 0.001 0.002 -0.000 -0.000
(TRms) (0.979) (0.528) (0.596) (0.512) (0.926) (0.728)
Entry regulation 0.010 0.048 0.048* 0.040 0.030 0.019
(TRen_d) (0.864) (0.538) (0.075) (0.226) (0.355) (0.570)
Log per capita GDP 0.084 0.019 0.012 0.012 0.035 0.009 0.086 -0.037 -0.038 -0.042 0.066** 0.093**
(lMWgdppc) (0.172) (0.921) (0.925) (0.921) (0.131) (0.792) (0.185) (0.857) (0.841) (0.828) (0.037) (0.012)
Log mobile subscr. -0.010 0.024 0.024 0.024 -0.012* -0.010 -0.013 0.017 0.019 0.015 -0.013 -0.018**
(lTImobsubsc) (0.485) (0.561) (0.504) (0.502) (0.096) (0.106) (0.393) (0.730) (0.514) (0.671) (0.197) (0.038)
N. obs. 199 199 182 182 199 199 199 199 182 182 199 199
ar1p 0.481 0.201 0.061 0.060 0.095 0.088 0.455 0.172 0.054 0.049 0.108 0.119
ar2p 0.948 0.966 0.481 0.479 0.294 0.089 0.927 0.987 0.439 0.471 0.196 0.276
N. instr. 162 156 186 180 149 139 185 176
Instr. (periods of dep.var.) t-2,t-3,...1 t-3,t-4,...1 t-2,t-3,...1 t-3,t-4,...1 t-2,t-3,…,1 t-3,t-4,...1 t-2,t-3,…,1 t-3,t-4,...1
Source: Authors' calculations using ITU, WDI and ECTR data. ITU data used for price series. For exact source and variable definition refer to the label (in italics in the first column) and Table A1.
Notes: Robust p-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Year dummies included in all models. Asymptotic standard erros in parentheses. ar1p and ar2p report the p-values of tests for first-order and second-order serial correlation, asymptically N(0,1). These test the levels residual for OLS and Within groups estimation, and the first-differenced residuals in all other columns. GMM results are one-step estimates with heteroskedasticity-consistent standard errors and test statistics. Instruments used in all GMM equations include the predetermined regulatory variable (at time t-1 and possibly earlier) and all exogenous variables. All calculations done using outreg2 for Stata (Roodman, 2009).
Table 11: The price models for the telecommunication industry.
51
ELECTRICITY GAS TELECOMS
First differences Levels
First differences Levels
First differences Levels
R-squared 0.086 0.024 0.109 0.047 0.899 0.969
Wald (Prob>F) 0.037 0.877 0.051 0.675 0.016 0.000
Notes: First differences: reduced form regression of 1log( )t
price −∆ on 2 3 1log( ), log( ),..., log( ).t t
price price price− −
Levels: reduced form regressions of 1log( )t
price − on 2 3 2log( ), log( ),..., log( ).t t
price price price− −∆ ∆ ∆
Wald p-value testing 0H : slope coefficients are jointly zero.
Table 12: A diagnosis of main results using Dif- and Sys-GMM: reduced form equations testing significance
of lagged endogenous variable as instruments.
52
Electricity Gas Telecom
(E1) (E2) (G1) (G2) (T1) (T2)
Individual characteristics
Female -0.024*** -0.025*** -0.025*** -0.025*** -0.017*** -0.017***
(0.000) (0.000) (0.001) (0.002) (0.004) (0.004)
31 - 45 years -0.020** -0.020** -0.027** -0.028** -0.032*** -0.032***
(0.033) (0.030) (0.018) (0.018) (0.001) (0.001)
46 - 60 years -0.008 -0.009 -0.026** -0.026** 0.001 0.000
(0.398) (0.386) (0.037) (0.043) (0.916) (0.967)
61 - 75 years 0.011 0.012 -0.007 -0.008 0.036*** 0.035***
(0.406) (0.357) (0.665) (0.635) (0.006) (0.007)
75 + years 0.046*** 0.046*** 0.008 0.009 0.077*** 0.077***
(0.007) (0.008) (0.736) (0.681) (0.000) (0.000)
End ed. age: 16 - 19 years 0.015* 0.014* 0.003 0.003 0.038*** 0.038***
(0.057) (0.063) (0.723) (0.722) (0.000) (0.000)
End ed. age: 20 + years 0.038*** 0.038*** 0.028** 0.030*** 0.041*** 0.041***
(0.000) (0.000) (0.012) (0.009) (0.000) (0.000)
Single 0.011 0.010 0.005 0.002 0.009 0.008
(0.171) (0.246) (0.641) (0.812) (0.272) (0.327)
managers 0.056*** 0.055*** 0.055*** 0.055*** -0.004 -0.004
(0.000) (0.000) (0.000) (0.001) (0.778) (0.756)
other white collars 0.033*** 0.032*** 0.041*** 0.038** -0.007 -0.007
(0.007) (0.009) (0.005) (0.013) (0.555) (0.558)
manual workers 0.011 0.010 0.003 -0.000 -0.019* -0.019*
(0.332) (0.398) (0.855) (0.982) (0.096) (0.098)
house person 0.043*** 0.043*** 0.042*** 0.043*** 0.005 0.007
(0.001) (0.001) (0.006) (0.007) (0.684) (0.610)
unemployed -0.032** -0.030** -0.042** -0.038* -0.059*** -0.059***
(0.040) (0.049) (0.029) (0.055) (0.000) (0.000)
retired 0.023* 0.021 0.015 0.016 -0.017 -0.017
(0.096) (0.114) (0.366) (0.348) (0.220) (0.226)
students 0.087*** 0.089*** 0.070*** 0.073*** 0.057*** 0.057***
(0.000) (0.000) (0.000) (0.001) (0.000) (0.000)
pol. views: center 0.015** 0.015** 0.008 0.010 0.020*** 0.020***
(0.041) (0.041) (0.343) (0.285) (0.005) (0.005)
pol. views: right 0.007 0.005 0.010 0.008 0.001 0.000
(0.433) (0.560) (0.346) (0.432) (0.904) (0.969)
pol. views: dk/na -0.011 -0.014* -0.005 -0.010 -0.016* -0.016*
(0.181) (0.091) (0.588) (0.347) (0.063) (0.062)
resp. coop.: avg./bad -0.065*** -0.063*** -0.057*** -0.052*** -0.030*** -0.029***
(0.000) (0.000) (0.000) (0.000) (0.002) (0.003)
(continues in the following page)
53
(continues from previous page)
Electricity Gas Telecom
(E1) (E2) (G1) (G2) (T1) (T2)
Macro economic variables (a)
Population density 0.013*** 0.003 0.005***
(0.000) (0.130) (0.002)
GDP (billions of euro) -0.000 -0.001*** -0.000***
(0.216) (0.000) (0.000)
CPI inflation All-items -0.014** -0.025*** 0.016***
(0.041) (0.007) (0.008)
Price (b) -1.829*** -0.036*** -0.162*
(0.000) (0.000) (0.068)
Control dummies
Year: 2002 -0.004 -0.012 0.010 0.132*** 0.009 0.020**
(0.610) (0.165) (0.275) (0.000) (0.183) (0.021)
Year: 2004 0.082*** 0.057*** 0.082*** 0.218*** 0.168*** 0.188***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Year: 2006 0.093*** 0.077*** 0.048*** 0.309*** 0.244*** 0.273***
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Country fixed-effects yes yes yes yes yes yes
Observations 57153 57153 30757 29585 53090 53090
Wald chi2 p-value 0.000 0.000 0.000 0.000 0.000 0.000
log-likelihood -35342 -35264 -18842 -17998 -30495 -30475
Pseudo R-squared 0.0575 0.0596 0.0578 0.0628 0.128 0.128
Coefficients show marginal effects. All estimates are weighted using sampling weights provided.
Robust p-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Source: Authors' calculations using Eurobarometer data except for (a), which come from Eurostat.
Notes: Omitted variables are: Male, 15-30 years, End education age: up to 15 years, In a couple, Self-employed, Political views: left, Respondent's cooperation: excellent/fair. (b) for electricity price is households price, with Dc tariff (Annual consumption: 3 500 kWh of which night 1 300), including all taxes. For natural gas price is households price, with D3 tariff (year consumption: 83.70 GJ), including all taxes. For telecomms price is for local calls (10 minutes).
Table 13: Consumers’ price satisfaction for each of the three services considered.
54
Electricity Gas Telecom
(E1) (E2) (G1) (G2) (T1) (T2)
ECTR data
Score (0-6) -0.025*** -0.023*** 0.010 0.029*** 0.142*** 0.139***
(ER;GR;TR) (0.000) (0.000) (0.161) (0.001) (0.000) (0.000)
Price (a) (b) no -1.692*** No -0.038*** no -0.133
(0.000) (0.000) (0.136)
Individual characteristics yes yes Yes yes yes yes
Macro economic variables (a) yes yes Yes yes yes yes
Time and country fixed effects yes yes Yes yes yes yes
Observations 57153 57153 30757 29585 53090 53090
Wald chi2 p-value 0.000 0.000 0.000 0.000 0.000 0.000
log-likelihood -35269 -35253 -18791 -17985 -30409 -30405
Pseudo R-squared 0.0595 0.0599 0.0604 0.0635 0.130 0.130
Coefficients show marginal effects. All estimates are weighted using sampling weights provided. Robust p-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1. For names of variables referring to ECTR data, see tables 1-3.
Source: Authors' calculations using Eurobarometer data except for (a), which come from Eurostat.
Table 14: Consumers’ price satisfaction and the ECTR (0-6) score for each of the three services considered.
55
Electricity Gas Telecom
(E1) (E2) (G1) (G2) (T1) (T2)
ECTR data
Public ownership 0.034** 0.048*** 0.092*** 0.105*** 0.012 0.024
(ERpo_d;GRpo_d;TRpo_d) (0.034) (0.002) (0.001) (0.000) (0.444) (0.118)
Vertical integration -0.022* -0.033** 0.137*** 0.097***
(ERvi_d;GRvi_d) (0.092) (0.016) (0.000) (0.001)
Entry regulation -0.057*** -0.050*** -0.047** 0.041*
(ERen_d;GRen_d) (0.000) (0.000) (0.018) (0.050)
Market structure -0.051 -0.104*** 0.003*** 0.002***
(GRms_d;TRms) (0.163) (0.007) (0.000) (0.000)
Price (a) (b) no -2.052*** no -0.034*** no -0.063
(0.000) (0.000) (0.490)
Individual characteristics yes yes yes yes yes yes
Macro economic variables (a) yes yes yes yes yes yes
Time and country fixed effects yes yes yes yes yes yes
Observations 57153 57153 30757 29585 53090 53090
Wald chi2 p-value 0.000 0.000 0.000 0.000 0.000 0.000
log-likelihood -35247 -35224 -18753 -17970 -30467 -30459
Pseudo R-squared 0.0601 0.0607 0.0623 0.0643 0.129 0.129
Coefficients show marginal effects. Robust p-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Source: Authors' calculations using Eurobarometer, ECTR and Eurostat data.
Notes: (a) Eurostat data. (b) For electricity price is households price, with Dc tariff (Annual consumption: 3 500 kWh of which night 1 300), including all taxes. For natural gas price is households price, with D3 tariff (year consumption: 83.70 GJ), including all taxes. For telecomms price is for local calls (10 minutes). For exact source and variable definition refer to the label (in italics in the first column) and Table A1.
Table 15: Consumers’ price satisfaction and the main dimensions of the regulatory reforms for each of the
three services considered.
56
ELECTRICITY GAS TELECOM
price
analysis
satisfaction
analysis
price
analysis
satisfaction
analysis
price
analysis
satisfaction
analysis
PRIVATISATION Public ownership of the incumbent
increases welfare
increases welfare
increases welfare
increases welfare
increases welfare
not significant
LIBERALISATION High vertical integration
mixed results
decreases welfare
increases welfare
increases welfare
not significant
not available
Legal restrictions to entry not
significant decreases welfare
not significant
mixed results
not significant
not available
Large incumbent's share not available not available not
significant decreases welfare
decreases welfare
increases welfare
Notes: "increasing welfare" in the price analysis means that (e.g.) public ownership of the incumbent is positively correlated with lower prices, and vice versa. "increasing welfare" in the satisfaction analysis means that (e.g.) public ownership of the incumbent is positively correlated with increasing satisfaction. "not significant" means that the coefficient is statistically not different from zero.
Table 16: A summary of main results about the effects of regulatory reforms on consumers’ welfare.
57
APPENDIX
Label Variable definition
ELECTRICITY
lEAprinet Electricity net-of-tax price for households, submitted to the IEA Secretariat by Administrations (in €/unit) (SourceIEA; (a))
ER See Table 1
ERpo_d See Table 1
ERen_d See Table 1
ERvi_d See Table 1
lEAshydr Electricity source: Hydroelectric (GWh/Tj) (Source: IEA; (a))
lEAimports Electricity import (GWh) (Source: IEA; (a))
lEAendiloss Electricity distribution losses (GWh) (Source: IEA; (a))
lEArescons Electricity residential consumtion (GWh) (Source: IEA; (a))
lMWgdppc Nominal GDP (billion of euro) (Source: WDI; (a))
lEAcoil Cost of high sulphur fuel oil (in €/tonne of oil equivalent; (a))
GAS
lGAprinet Natural gas net-of-tax price for households, submitted to the IEA Secretariat by Administrations (in €/ unit) (SourceIEA; (a))
GR See Table 2
GRpo_d See Table 2
GRen_d See Table 2
GRvi_d See Table 2
lDGAbrent Change of price Brent oil (Source: IEA; (a))
lMWgdppc Nominal GDP (billion of euro) (Source: WDI; (a))
TELECOMS
lTEprice Telecom price of a 3-minute fixed telephone local call (peak rate - €) (Source: ITU; (a))
TR See Table 3
TRpo_d See Table 3
TRms See Table 3
TRen_d See Table 3
lMWgdppc Per capita GDP in € (Source: WDI; (a))
lTImobsubscr Mobile cellular telephone subscribers - (Post-paid + Pre-paid) (Source: ITU; (a))
Notes: Variables starting with an "l", means that they were transformed in logarithms. (a) Original data are in US$ and were converted to € using Eurostat euro/USD average yearly exchange rate.
Table A1: Definitions of main variables used in the price equation models.
58
ELECTRICITY
Variable Obs Mean Std. Dev. Min Max
year 495 1991 9.53 1975 2007
lEAprinet_kw 429 -2.46 0.32 -3.46 -1.82
ER 495 4.39 1.81 0.00 6.00
ERpo_d 495 0.48 0.50 0.00 1.00
ERen_d 495 0.77 0.42 0.00 1.00
ERvi_d 495 0.91 0.29 0.00 1.00
lEAshydr 494 8.18 3.43 -9.21 11.30
lEAimports 495 7.71 4.14 -9.21 10.95
lEAendiloss 480 8.39 1.47 3.18 10.44
lEArescons 480 9.75 1.39 5.65 11.91
lMWgdppc 495 -4.18 0.72 -6.20 -2.27
lEAcoil 305 4.99 0.35 4.17 6.13
GAS
Variable Obs Mean Std. Dev. Min Max
year 343 1992 8.40 1978 2007
lGAprinet_gj 343 1.92 0.43 0.59 2.91
GR 343 4.37 1.14 0.73 6.00
GRpo_d 343 0.25 0.43 0.00 1.00
GRen_d 343 0.39 0.49 0.00 1.00
GRms_d 343 0.27 0.45 0.00 1.00
GRvi_d 343 0.48 0.50 0.00 1.00
lDGAbrent 343 0.05 0.29 -0.90 0.74
lMWgdppc 343 -4.01 0.59 -5.46 -2.41
TELECOMS
Variable Obs Mean Std. Dev. Min Max
year 216 1998 4.59 1990 2005
lTIpri2 216 -2.10 0.43 -3.49 -0.08
TR 216 3.42 1.93 0.46 6.00
TRpo1_d 216 0.36 0.48 0.00 1.00
TRms1_d 216 85.42 18.90 37.00 100.00
TRen1_d 216 0.42 0.49 0.00 1.00
lMWgdppc 216 -3.70 0.38 -4.88 -2.50
lTImobsubsc 216 14.38 2.31 6.71 18.19
Notes: authors' calculations. For variable definition see Table A1. All monetary variables are in euro or converted from US$ using Eurostat euro/USD average yearly exchange rate.
Table A2: Descriptive statistics for main variables used in the price equations.
59
Variable Obs Mean Std. Dev. Min Max
Price paid for electricity is fair 57828 0.63 0.48 0.00 1.00
Price paid for natural gas is fair 30811 0.66 0.47 0.00 1.00
Price paid for fixed telephone calls is fair 51402 0.64 0.48 0.00 1.00
Female 57828 0.53 0.50 0.00 1.00
31 - 45 years 57828 0.28 0.45 0.00 1.00
46 - 60 years 57828 0.25 0.43 0.00 1.00
61 - 75 years 57828 0.19 0.39 0.00 1.00
75 + years 57828 0.05 0.23 0.00 1.00
End ed. age: 16 - 19 years 57828 0.38 0.48 0.00 1.00
End ed. age: 20 + years 57828 0.28 0.45 0.00 1.00
Single 57153 0.21 0.41 0.00 1.00
managers 57828 0.10 0.30 0.00 1.00
other white collars 57828 0.11 0.32 0.00 1.00
manual workers 57828 0.21 0.41 0.00 1.00
house person 57828 0.12 0.32 0.00 1.00
unemployed 57828 0.06 0.23 0.00 1.00
retired 57828 0.24 0.42 0.00 1.00
students 57828 0.08 0.27 0.00 1.00
pol. views: center 57828 0.35 0.48 0.00 1.00
pol. views: right 57828 0.20 0.40 0.00 1.00
pol. views: dk/na 57828 0.19 0.39 0.00 1.00
resp. coop.: avg./bad 57828 0.11 0.31 0.00 1.00
Population density (a) 57828 162.55 120.45 17.00 483.80
GDP at market prices (billions of euro) (a) 57828 761.36 755.06 22.00 2321.50
CPI all-items annual rate of change (a) 57828 2.27 0.96 0.10 5.30
Electricity yearly average price (a),(b) 57828 0.14 0.04 0.06 0.24
Gas yearly average price (a),(b) 49175 12.70 4.60 6.02 29.82
Telecomms yearly average price (a),(b) 57828 0.38 0.11 0.19 0.69
Year 2002 57828 0.26 0.44 0.00 1.00
Year 2004 57828 0.23 0.42 0.00 1.00
Year 2006 57828 0.25 0.43 0.00 1.00
Source: Eurobarometer various surveys, except for (a), which come from Eurostat.
Notes: Omitted variables are: Male, 15-30 years, End education age: up to 15 years, In a couple, Self-employed, Political views: left, Respondent's cooperation: excellent/fair. (b) for electricity price is households price, with Dc tariff (Annual consumption: 3 500 kWh of which night 1 300), including all taxes. For natural gas price is households price, with D3 tariff (year consumption: 83.70 GJ), including all taxes. For telecomms price is for local calls (10 minutes).
Table A3: Descriptive statistics of variable used in the consumers’ price satisfaction analysis.
60
Regulatory variables: Score 0-6 Regulatory variables: Binary variables 0-1
GMM-Dif GMM-Dif GMM-Sys GMM-Sys GMM-Dif GMM-Dif GMM-Sys GMM-Sys
log price (t-1) 0.784*** 0.783*** 0.991*** 0.943*** 0.683*** 0.702*** 0.931*** 0.896***
(lEAprinet_kw) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Score 0-6 0.009 0.006 0.011* 0.012**
(ER) (0.262) (0.483) (0.067) (0.046)
Public. ownership -0.054*** -0.050*** -0.044*** -0.054***
(ERpo_d) (0.006) (0.004) (0.000) (0.000)
Entry regulation -0.020 -0.021 0.035** 0.040***
(ERen_d) (0.223) (0.199) (0.025) (0.003)
Vertical integration 0.079*** 0.075*** 0.040** 0.043**
(ERvi_d) (0.000) (0.000) (0.028) (0.031)
Source: Hydroelectric -0.008*** -0.007*** -0.001 -0.002 -0.008*** -0.008*** -0.003** -0.004*
(lEAshydr) (0.000) (0.000) (0.648) (0.142) (0.000) (0.000) (0.041) (0.050)
Imports -0.000 -0.001 0.001 0.000 -0.002 -0.001 -0.001*** -0.002***
(lEAimports) (0.822) (0.692) (0.104) (0.379) (0.145) (0.167) (0.000) (0.000)
Distribution losses -0.024 -0.023 0.025 0.023 -0.027 -0.026 0.044* 0.044*
(lEAendiloss) (0.444) (0.461) (0.113) (0.270) (0.341) (0.337) (0.062) (0.095)
Log residential consumption -0.088** -0.088*** -0.010 -0.005 -0.125*** -0.121*** -0.028 -0.027
(lEArescons) (0.010) (0.010) (0.568) (0.828) (0.000) (0.000) (0.208) (0.287)
Log per capita GDP 0.137*** 0.141*** -0.014 -0.027 0.180*** 0.174*** -0.019 -0.028
(lMWgdppc) (0.000) (0.000) (0.524) (0.278) (0.000) (0.000) (0.392) (0.223)
Log cost of oil 0.033 0.034 0.004 -0.002 0.047* 0.047* 0.004 0.002
(lEAcoil) (0.330) (0.299) (0.886) (0.944) (0.093) (0.085) (0.865) (0.944)
N. obs. 267 267 282 282 267 267 282 282
ar1p 0.006 0.006 0.002 0.003 0.008 0.007 0.001 0.002
ar2p 0.269 0.269 0.146 0.134 0.271 0.265 0.152 0.147
N. instr. 65 65 67 67 104 104 104 104
Instr. (periods of dep.var.) t-2 t-3 t-2 t-3 t-2 t-3 t-2 t-3
Source: Authors' calculations using IEA, Eurostat, WDI source data. IEA data used for price series.
Notes: Robust p-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Year dummies and constant included in all models. For other notes see Table 5.
A4: Robustness checks of results produced in Table 5, by using different instruments.
61
Regulatory variables: Score 0-6 Regulatory variables: Binary variables 0-1
GMM-Dif GMM-Dif GMM-Sys GMM-Sys GMM-Dif GMM-Dif GMM-Sys GMM-Sys
log price (t-1) 0.841*** 0.831*** 0.955*** 0.948*** 0.809*** 0.809*** 0.923*** 0.900***
(lGAprinet_gj) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Score 0-6 0.013 0.008 -0.003 -0.002
(GR) (0.381) (0.597) (0.649) (0.797)
Public. ownership -0.047 -0.051 0.018 0.015
(GRpo_d) (0.193) (0.166) (0.265) (0.404)
Entry regulation 0.046** 0.045* 0.036** 0.052***
(GRen_d) (0.048) (0.052) (0.010) (0.000)
Market structure 0.007 0.010 -0.050** -0.070***
(GRms_d) (0.814) (0.749) (0.033) (0.001)
Vertical integration 0.008 0.008 0.003 0.018*
(GRvi_d) (0.390) (0.412) (0.716) (0.051)
Log change of price Brent oil 0.627*** 0.627*** 0.748*** 0.729*** 0.638*** 0.640*** 0.672*** 0.594***
(lDGAbrent) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.001)
Log per capita GDP 0.030 0.030 -0.005 -0.010 0.066 0.067 -0.015 -0.022
(lMWgdppc) (0.380) (0.412) (0.607) (0.511) (0.152) (0.143) (0.443) (0.306)
N. obs. 314 314 328 328 314 314 328 328
ar1p 0.014 0.017 0.019 0.022 0.017 0.017 0.015 0.013
ar2p 0.480 0.483 0.515 0.511 0.585 0.586 0.463 0.250
N. instr. 87 86 132 129 306 300 292 282
Instr. (periods of dep.var.) t-2 t-3 t-2 t-3 t-2,t-3,…,1 t-3,t-4,…,1 t-2,t-3,…,1 t-3,t-4,…,1
Source: Authors' calculations using IEA, Eurostat, WDI source data. IEA data used for price series.
Notes: Robust p-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Year dummies and constant included in all models. For other notes see Table 6.
A5: Robustness checks of results produced in Table 6, by using different instruments.
62
Regulatory variables: Score 0-6 Regulatory variables: Binary variables 0-1
GMM-Dif GMM-Dif GMM-Sys GMM-Sys GMM-Dif GMM-Dif GMM-Sys GMM-Sys
log price (t-1) 0.625*** 0.613*** 0.863*** 0.862*** 0.613*** 0.566*** 0.718*** 0.719***
(lTIpri2) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Score 0-6 0.035** 0.036** 0.013 0.004
(TR) (0.032) (0.029) (0.194) (0.680)
Public. ownership -0.056* -0.056 -0.029 -0.044
(TRpo_d) (0.100) (0.231) (0.293) (0.208)
Market structure 0.004 0.003 -0.001 -0.000
(TRms) (0.307) (0.298) (0.628) (0.651)
Entry regulation 0.042 0.062 0.020 0.016
(TRen_d) (0.227) (0.212) (0.492) (0.509)
Log per capita GDP 0.015 0.016 0.001 -0.004 -0.106 -0.078 0.113** 0.114**
(lMWgdppc) (0.907) (0.903) (0.971) (0.905) (0.628) (0.695) (0.024) (0.026)
Log mobile subscr. 0.022 0.023 -0.012 -0.017** 0.005 0.018 -0.016** -0.017*
(lTImobsubsc) (0.539) (0.532) (0.132) (0.013) (0.862) (0.655) (0.035) (0.058)
N. obs. 182 182 199 199 182 182 199 199
ar1p 0.057 0.056 0.094 0.088 0.037 0.034 0.054 0.056
ar2p 0.477 0.475 0.097 0.403 0.590 0.454 0.341 0.428
N. instr. 140 139 171 169 69 68 110 109
Instr. (periods of dep.var.) t-2 t-3 t-2 t-3 t-2 t-3 t-2 t-3
Source: Authors' calculations using ITU, Eurostat, WDI, ECTR data. ITU data used for price series.
Notes: Robust p-values in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Year dummies and constant included in all models. For other notes see Table 7.
A6: Robustness checks of results produced in Table 7, by using different instruments.