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.