Docstoc

The ladder of investment approach and the development of new

Document Sample
The ladder of investment approach and the development of new Powered By Docstoc
					The ladder of investment approach and the development of new

         access infrastructures: Which empirical evidence?

                (Preliminary version: Please do not cite)

                  Maya Bacache, Marc Bourreau and Germain Gaudin

                                      Telecom ParisTech

                                          March 2010



                                            Abstract

     In the telecommunications sector, the "ladder of investment" refers to a regulatory approach

  which aims at reconciling service-based competition (where entrants lease access to incumbents’

  facilities) and facility-based competition (where they build their own infrastructure). In this

  paper, we construct a precise model that provides a test of the ladder of investment and, using

  data from EC Implementation Reports and the COCOM Broadband access in the EU Reports

  covering 15 European member states, we provide an empirical test of the ladder of investment

  approach. We find no ladder effect between LLU and new infrastructure lines but a ladder effect

  between bistream access lines and LLU lines. Entrants climb up the ladder to LLU but they

  cannot reach the next floor and build their own infrastructure.

     Keywords: Telecommunications, Ladder of investment.

     JEL Codes: L51, L96.




                                                1
1    Introduction

Two forms of competition can be developed in the telecommunication industry: Service-Based
Competition (SBC) and Facility-Based Competition (FBC). FBC is considered as the only means
to achieve sustainable competition for it creates a superior potential for service and product innova-
tions than SBC and it can lead to a (partial) deregulation of the sector. For instance, Distaso et al.
(2006) established that competition in infrastructure was the main driver of broadband adoption
in 14 European countries from 2000 to 20041 . However FBC is difficult to assess directly because
of high sunk costs that new entrants have to bear.
    The ladder of investment approach articulated by Martin Cave in 2006 suggests that access
regulation is not only pro-competitive per se as it reduces barriers to entry but also is an indirect
device to promote facility-based competition. What is the intuition behind this approach? By
setting low access prices, the regulator encourages service-based competition. Then, once entrants
have gained a consumer base and information on the market they can move up "the ladder of
investment" and invest in their own facilities (Bourreau and Dogan, 2004). The regulator should
then increase the access price in order to encourage the entrants to climb up the next rungs, i.e. burn
up the rung on which the entrant was standing. Hence this approach considers that service-based
competition could be a stepping stone for facility-based competition
    However, other effects could offset the ladder effect and access regulation could discourage in-
vestment mainly for two reasons: because entrants can rent access to the incumbents’ infrastructure
at a regulated low price, they have low incentives to build new infrastructures (an effect which is
referred to as the “replacement effect”, see Bourreau and Dogan, 2006); besides, in a dynamic
setting, access regulation could lower the net value of infrastructure investment, hence, reduce the
incumbent’s incentives to invest. Hence, service-based competition and facility-based competition
could be either substitutes in developing new infrastructure or complements (See Bourreau et al..
(2009) for a critical review of the ladder of investment approach).
    The ladder of investment has been widely implemented by European member state national
regulator authorities (NRAs). Most of European countries do regulate bitstream access and local
loop unbundling (LLU). However the reality of the ladder of investment, though this theory is very
appealing, has never been demonstrated with scrutiny.

   1
     However, Hoffler (2007) shows that for the period 2000-2004 in western Europe, inefficient duplication of an
existing infrastructure outweighted the benefits of broadband penetration.




                                                     2
   In this paper, we propose to test the validity of the ladder approach: Are SBC and FBC
substitutes or complements in developing new infrastructures? Does multiplying the levels of access
(rungs of the ladder) stimulate facility-based competition?
   However, testing the ladder of investment approach is very challenging not only because of
the lack of data or because of the imperfections of its implementation by the regulators, but also
because this approach lacks precise formalization that allows for a direct test. In this paper, we
propose a precise test of the ladder of investment (LOI) approach.
   We try to formalize the ladder approach trough three different specifications. First, in a basic
model, we argue that a naïve view of the ladder approach stipulates that there should be a positive
relation between SBC and FBC. We estimate different regressions using the number of New Gen-
eration Access Networks (NGAN) lines built by new entrants as a dependent variable. We use the
number of rungs and the number of unbundled lines as the main explanatory variables. We don’t
find a significant relation between SBC and FBC. In a second specification we allow for lagged
effects in the exogenous variables. In a third specification, we build a more complex model to take
into account both the substitution effect and the complementarity effect and we consider the rate
of growth of new lines rather then the stock of new lines to test the LOI. We also propose to test
the LOI not only on its final goal, which is new infrastructure, but also on the last rung (LLU)
to assert wether the incumbent move up the ladder. We study the implementation of the ladder
of investment approach by using data covering incumbent and entrant fixed-broadband operators
in 15 European member states over 7 years. Data are extracted from EC Implementation Reports
and COCOM Broadband access in the EU Reports.
   On the empirical side, reports and papers found weak support for the ladder approach and robust
econometric studies that quantify the effect of access regulation on investment in infrastructure are
few (see Cambini and Jiang (2009) for an extensive survey). Hausman and Sidak (2005) made
an empirical review of unbundling experiences in five countries and found no stylized facts in
line with the ladder theory, namely no migration from service-based competition to facility-based
competition. Crandall et al. (2007) noticed on the contrary that new entrants seem to be stuck
at the lower rung of the ladder and hence advocate that governments should subsidize the price of
computers rather than local loop access. Hazlett and Bazelon (2005) study two natural experiences
in the US and find no econometric proof for the ladder of investment hypothesis (from1999 to 2004).
Our paper differs from those previous qualitative studies by doing an econometric analysis to test
the LOI approach.


                                                 3
    Waverman et al. (2007) model the impact of access prices (the price of unbundled local loops)
on the market share of alternative structure and estimate that a 10% reduction in Local loop
unbundling price causes a 18 percent fall in subscriber share of alternative infrastructure. Crandall
et al. (2004) found that the idea that low access prices would encourage new entrants to rent
at first, and then build facilities after they gained market experience, is not supported by the
data on 15 European countries from 2002 to 2006. Similarly, Distaso et al. (2009) look at the
link between access prices and alternative inputs to provide broadband services and find that the
national regulatory authorities adopted policies consistent with the ladder of investment approach.
Those papers do not precisely test the LOI approach. Our paper differs by formulating a precise
econometric test of the LOI. Grajek and Roller (2009) study a wide data set that covers 20 countries
and 10 years and differentiate incumbents’ investment and entrants’ investment. The specificity of
the paper is to use a regulatory index and to account for the possible endogeneity of regulation.
They find a negative relation between access regulation and individual investment of entrants
despite the fact that total investment by entrants increases. However they do not test directly
the ladder approach (for instance they do not account for the multiplicity of ladder rungs or the
dynamics of the entrants climbing the ladder) mainly because they cannot distinguish investment in
new telecommunication infrastructure. Our papers studies the number of new telecommunication
lines in new infrastructures and hence can provide a direct test of the LOI.
    We find no evidence of the ladder of investment. Even when controlling fort the robustness
of the results, we find no evidence of a positive effect of the numbered of opened lines on new
infrastructures. However we find evidence that support the view that new entrants climb up the
ladder until the last rung but they don’t manage to climb to the next floor and build their own
infrastructure.
    In section 2 we present some descriptive statistics of the development of broadband infrastruc-
tures. In section 3 we propose a model and specifications to test the LOI. Section 4 presents our
results and section 5 various robustness tests. Section 6 concludes.


2     The development and regulation of broadband in Europe

2.1   The development of broadband in Europe

The number of European broadband subscribers among the EU15 countries (Austria, Belgium,
Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal,


                                                 4
Spain, Sweden and United Kingdom) increased from 8,706,101 in June 2002 to 99,108,430 in January
20092 . However, if we look at each Member State individually, we will see a more mixed picture:
in June 2009, the Netherlands had 37.9 broadband subscribers per 100 inhabitants whereas Greece
only had 15.5.
       Broadband connection can be provided using different technologies as DSL, Cable internet,
optical Fibre (Fibre to the Home, Fibre to the Building), Power Line Communication (PLC),
Wireless Local Loop (WLL) or Satellite3 . If we notice some persistent dissimilarities between
countries for the use of these technologies (for example, Cable internet has never been implemented
in Italy, nor in Greece), we also observe table 1, Cable internet and New technologies (i.e. Fibre,
WLL and PLC) that DSL has known a significant growth, at the expense of Cable internet.


                                       DSL (%)       Cable (%)     New facilities (%)
                         June 2002       68.8          29.9              1.4
                         Jan. 2009       84.5          14.1              1.5

                         Table 1: EU15 average use of broadband technologies




       New technology infrastructures include Fibre, WLL and PLC. The share of these technologies
remains low compared to DSL or Cable in EU15, with a maximum of 19% in Sweden in January
2009, whereas in non European countries the share of New technologies, especially Fibre, is quite
higher. For instance, in December 2008, the percentage of fibre connections in total broadband
subscriptions in Japan was 47.9% (respectively 43.1% in Korea)4 .


2.2      European regulation of broadband access

Regulation 2887/2000 on unbundled access to the local loop imposed mandatory LLU because "It
would not be economically viable for new entrants to duplicate the incumbent’s metallic local access
infrastructure in its entirety within a reasonable time"5 . Thus, the copper local loop can be defined
as an essential facility that cannot be replicated.

   2
     As we focus on terrestrian technologies matching with the LOI approach, we do not take into account satellite
communication broadband.
   3
     resp.
   4
     OECD Broadband subscribers per 100 inhabitants, by technology, June 2009, OECD Broadband statistics,
available at oecd.org/sti/ict/broadband
   5
     REGULATION (EC) No 2887/2000 OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 18
December 2000 on unbundled access to the local loop, Official Journal of the European Communities, December 2000,
the 30th, available at http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2000:336:0004:0008:EN:PDF


                                                        5
       National Regulatory Agencies can enforce availability of the access to the incumbent’s network
at multiple levels. For instance, technically available type of accesses can be: Resale Access6 ,
IP-Bitstream Access7 , ATM-Bitsream Access8 , DSLAM-Bitstream Access9 , Shared Access, Full
Local Loop Unbundling or Sub-Loop Unbundling. We have seen that Shared- and Full-Local Loop
Unbundling are mandatory by regulation 2887/2000. Practically, DSLAM-Bitstream Access and
Sub-loop Unbundling are of little relevance and are not widely used. Moreover, in available datasets,
IP-Bitstream and ATM-Bitstream Accesses are usually counted together as "bitstream product".
Finally, regarding the lack of major investment incentive differences between Shared- and Full-LLU
for new entrants, we consider that these two types of access can be associated in our analysis in a
so-called "LLU" access. This is why in our analysis the LOI is composed of four rungs only (resale,
bitstream access, LLU and own facility), as in ERG (2005).
       Even if the LOI is not mandatory at a European level, representatives from the European
Commission often refer to it as a good approach for new market entrants to start generating revenue
and then to roll out their own infrastructures1011 and many European regulators, as ARCEP12 , rely
on the LOI approach. From June 2002 to June 2009, EU15 average use of accesses to incumbents’
networks moved from 40.7% in resale, 31.0% in bitstream access and 28.3% in LLU to 11.0% in
resale, 15.8% in bitstream access and 73.2% in LLU, providing a good intuition for the efficiency of
the LOI approach. However, one should keep in mind that the goal of this regulation is for entrants
to fully build their own infrastructures as quickly as possible. To test this relation between New
infrastructures investment by entrants and the LOI approach is the main goal of our analysis.


2.3      Descriptive statistics

We know from Table 1 that average use of New technologies did not increase as a proportion of
broadband subscribers from June 2002 to July 2009, but we can see from Figure 1 that the new
infrastructure evolution mainly concern optical fibre and WLL while PLC did not take up.


   6
       ERG’s Revised Common Position Option 4
   7
       ERG’s Revised Common Position Option 3
     8
       ERG’s Revised Common Position Option 2
     9
       ERG’s Revised Common Position Option 1
   10
       Viviane Reding, SPEECH/06/697, ECTA Conference 2006.
   11
       Telecoms     reform,      Press   Factsheet     #5:           Investment       :          breathing      new
life      into    Europe’s    telecoms   markets,      November       2007,      the     13th,       available    at
http://ec.europa.eu/information_society/newsroom/cf/itemlongdetail.cfm?item_id=3723.
   12
       According to ARCEP, "the development of competition in France since 1998 is a good illustration of the theory
of the ladder of investment". See ARCEP (2007), p.36.


                                                         6
                                  1600
                      Thousands



                                  1400




                                  1200




                                  1000



                                                                                                                                                                            WLL
                                  800
                                                                                                                                                                            PLC
                                                                                                                                                                            Fiber

                                  600




                                  400




                                  200




                                    0
                                     Jul.02   Jan.03   Jul.03   Jan.04   Jul.04   Jan.05   Jul.05   Jan.06   Jul.06   Jan.07   Jul.07   Jan.08   Jul.08   Jan.09   Jul.09




                            Figure 1: EU15 ’New’ broadband technologies use

3    A simple model

In this Section, we build up a simple model to study the relation between service-based competition
and facility-based competition.
    We consider a market with an incumbent operator and a potential entrant. There are two
periods. At the beginning of each period, the entrant can decide to lease access to the incumbent’s
infrastructure at a given level of access (which leads to service-based competition), to build its own
infrastructure (which leads to facility-based competition), or to remain outside of the market (it
then obtains a fixed payoff, which we normalize to 0).
    When it enters service-based, the entrant has to incur a fixed (and sunk) cost σ, which is drawn
from a distribution with density fσ and cumulative distribution Fσ . For each period under service-
based competition, the entrant has also to pay an access charge r to the incumbent, which is set
by the regulator.
    Since our focus is how an initial phase of service-based competition affects the probability of
facility-based entry, we assume that the cost of facility-based entry is so high in the first period,
that the entrant never builds its infrastructure in this period. However, due to technical progress,
facility-based entry is a possibility in the second period. In the second period, if it enters facility-
based, the entrant has to incur a fixed (sunk) cost ϕ, which is drawn from a distribution with
density fϕ and cumulative distribution Fϕ .


                                                                                                    7
      The costs of service-based and facility-based entry, σ and ϕ, are revealed to the entrant at the
beginning of the game, prior to the first period.13
      We also assume that the entrant’s profit depends on a variable κ, which corresponds to the
experience of the entrant in the operation of a telecommunications network and in the commercial-
ization of communications services. For instance, κ can correspond to the amount of "learning by
doing" by the entrant (hence, a higher κ would imply a lower marginal cost), or to the perceived
quality of service of the entrant. We denote by κ0 the initial experience of the entrant, before it
enters the market, and κ1 its experience once it has spent one period in the market, with κ1 ≥ κ0 .
      We denote by ΠS (κ, r) the entrant’s profit under service-based competition, and by ΠF (κ) its
profit under facility-based competition. We assume that ΠS (κ, r) ≥ 0 for all κ and r, and that
ΠS (κ, r) is decreasing with r, for all κ. Consistent with the approach of the ladder of investment,
we assume that ΠF κ1 ≥ ΠF κ0 , and that ΠS κ1 , r ≥ ΠS κ0 , r . This assumption implies in
particular that a phase of service-based competition can serve as a "stepping stone" for the entrant
and increase its market experience, which translates into higher profits. Finally, we denote by δ
the discount factor.
      We start by analyzing the decision of the entrant in the second period, depending on its entry
decision in the first period.


3.1      Second period entry decision

We have to consider two different cases, depending on whether the entrant entered service-based
in the first period, or not.


No phase of service-based competition If the entrant remained outside the market in the first
period, it can either enter service-based or facility-based in the second period, or remain outside
the market. The entrant enters the market if max ΠF κ0 − ϕ, ΠS κ0 , r − σ ≥ 0. It prefers to
enter facility-based than service-based if ΠF κ0 − ϕ ≥ ΠS κ0 , r − σ.


With a phase of service-based competition Suppose now the entrant has entered service-
based in the first period. Since ΠS κ1 , r ≥ 0 by assumption, staying under service-based com-
petition always dominates exiting the market. Therefore, if the entrant entered on the basis of
services in the first period, in the second period, it decides either to remain service-based or to

 13
      The cost of service-based entry is the same in the two periods.


                                                           8
enter facility-based. It decides to build its own infrastructure in the second stage if and only if
ΠF κ1 − ϕ ≥ ΠS κ1 , r .


3.2   First period entry decision

Since facility-based entry is prohibitively costly in the first period, the entrant decides either to
remain outside the market or to enter service-based. If the entrant remains outside the market in
the first period, its profit is equal to


                             δ max ΠS κ0 , r − σ; ΠF κ0 , r − ϕ; 0 .


However, we find that the entrant never enters service-based in period 2 if it stayed outside the
market in period 1, and hence, max ΠS κ0 , r − σ; ΠF κ0 , r − ϕ; 0 = max ΠF κ0 , r − ϕ; 0 .
Indeed, if the entrant decided to enter service-based in the second period, it would imply that
ΠS κ0 , r − σ > 0. But then, the entrant would have made a higher profit by entering service-
based in period 1 and remaining service-based in period 2 than by staying outside the market in
period 1.
   From this analysis, it follows that if the entrant remains outside in period 1, it either enters
facility-based in period 2 or remains outside in period 2. Without a initial phase of service-based
competition, the ex-ante probability that the entrant enters facility-based in the second period is
therefore
                         pr [FB entry] = pr ϕ ≤ ΠF κ0        = Fϕ ΠF κ0     .

   If the entrant enters service-based in the first period and remains under service-based competi-
tion in the second period, it obtains an expected profit of


                                    ΠS κ0 , r − σ + δΠS κ1 , r .


If the entrant enters service-based in the first period and builds its own infrastructure in the second
period, it obtains an expected profit of


                                ΠS κ0 , r − σ + δ ΠF κ1 , r − ϕ .




                                                  9
Therefore, the entrant enters service-based in the first period if


                             σ ≤ ΠS κ0 , r + δ max ΠS κ1 , r , ΠF κ1 , r − ϕ .


We have ΠF κ1 − ϕ ≥ ΠS κ1 , r with probability β (r) = Fϕ ΠF κ1 − ΠS κ1 , r . With
probability β (r), the phase of service-based competition is followed by a phase of facility-based
competition; and with probability 1 − β (r), it is followed by service-based competition. Notice that
β (r) is increasing with r; a higher access charge increases the probability of facility-based entry,
because it softens the replacement effect.
      Let α the ex-ante probability of service-based entry,14 we have


               α (r) = Fσ ΠS κ0 , r + δ β (r) ΠF κ1 − ϕ (r) + (1 − β (r)) ΠS κ1 , r             ,


where ϕ (r) is the expected average ϕ for ϕ ≤ ΠF κ1 −ΠS κ1 , r . Note that a decrease of the access
charge r has a priori an ambiguous effect on the probability of service-based entry, α (r). Indeed, if r
is decreased, ΠS κ0 , r increases. However, the term β (r) ΠF κ1 − ϕ (r) +(1 − β (r)) ΠS κ1 , r
can either increase or decrease with r. The term (1 − β (r)) ΠS κ1 , r is decreasing with r, but
the term β (r) ΠF κ1 − ϕ (r) can be either increasing or decreasing, as β (r) and ϕ (r) are both
increasing with r. For the rest of the analysis, we will nonetheless assume that α (r) is decreasing
with r (more attractive conditions for service-based competition increase the probability of service-
based entry).
      The ex-ante probability that the entrant enters facility-based in the second period is then given
by
                    Ψ (r) = (1 − α (r)) Fϕ ΠF κ0           + α (r) Fϕ ΠF κ1 − ΠS κ1 , r   .

We have

dΨ                                               ∂ΠS κ1 , r ∂α
   = −α (r) fϕ ΠF κ1 − ΠS κ1 , r                           +    Fϕ ΠF κ1 − ΠS κ1 , r          − Fϕ ΠF κ0   .
dr                                                  ∂r       ∂r

The firm term is positive as ∂ΠS κ1 , r /∂r ≤ 0. The second term is positive if and only if ΠF κ1 −
ΠS κ1 , r ≤ ΠF κ0 . If κ1 = κ0 , this inequality does not hold (due to the replacement effect) and
hence, dΨ/dr ≥ 0, that is, a higher access charge increases the probability of facility-based entry.

 14
      As shown earlier, service-based entry can take place only in the first period.



                                                           10
We can have dΨ/dr ≤ 0 only if κ1 is sufficiently high relative to κ0 , that is, if there is a strong
"stepping stone effect."
    To summarize this analysis, we have found that if there is a strong "stepping stone" or "ladder
of investment" effect, more favorable conditions for service-based entry (a lower r) lead to more
service-based entry (i.e., α (r) increases) and more facility-based entry (Ψ (r) increases). In this case,
we expect a positive relation between service-based entry at a period t and facility-based entry at
a period t + 1. However, if the "ladder of investment" is not sufficient and the replacement effect
dominates, more favorable conditions for service-based entry (a lower r) lead to more service-based
entry but also to less facility-based entry. In this latter case, we expect a negative relation between
service-based entry at a period t and facility-based entry at a period t + 1.
    In the rest of the paper, we now turn to an empirical estimation, to determine whether we
observe a positive or a negative relation between service-based entry at date t and facility-based
entry at date t + 1.


4    Empirical evidence

The approach of the ladder of investment assumes that mandatory unbundling would stimulate
investment in broadband infrastructure, and more specifically in new technologies, which would
stimulate broadband penetration. The ladder of investment is based on two hypothesis: first
that service-based competition promotes facility-based competition and second that the higher the
number of levels of access, the stronger the development of new access lines by alternative operators
    We seek to capture the correlation between service-based competition and facility-based com-
petition. Hence we want to test the following relation:


                number of new lines = f(number of opened lines, number of access)

    We explain the number of new lines built by new entrants with the number of DSL lines provided
by the incumbent to the new entrants with unbundling (shared & full LLU) and access indices.
More specifically we test:
    -Hypothesis n◦ 1: SBC promotes FBC. In the standard view we expect β 1 < 0 and in the Ladder
of Investment view we expect β 1 > 0
    -Hypothesis n◦ 2: The higher the number of levels of access, the stronger the development of
new access lines by alternative operators, so in the Ladder of Investment approach: β 2 > 0:


                                                   11
               log(Newlines) = β 0 + β 1 log(LLUlines) + β 2 access + β x controls + u              (1)

where access is the sum of the use of full LLU, shared LLU, bitstream and resale.
   However this first specification, equation 1, can be rather naive because it does not allow for
time spread and time lags. In other words one could expect that it takes some time for the new
entrants to develop their own infrastructure once they developed telecommunication services thanks
to the incumbent infrastructure. So we test an second specification where t and i are time indexes:




 log(Newlines)t = β 0 + β 1 log(LLUlines)t + β 1’ log(LLUlines)t−1 + ...β 1′ log(LLUlines)t−i +

                       +β 2 accesst + β 2’ accesst−1 ... + β 2” accesst−i + β x controls + u        (2)


   Moreover, the direct test of the LOI would be to test if the rate of growth of the infrastructures
lines is, ceteris paribus, an increasing function of the rate of growth of unbundled lines. However this
third specification is difficult to test because of the lack of historical date on new infrastructures.


4.1   The ladder approach revisited

Cave (2009) proposed a light version of the LOI approach, stating that the ladder would not help
building new infrastructures per se but only increase intra-modal facility-based competition and
investment in LLU. It can also be argued that should we not observe impact on new infrastructure,
our results could be due to the fact that new lines are still very recent.
   Hence, we test the LOI only on the last rung of the ladder. We explain LLU lines, i.e. the
number of unbundled lines, with bitstream and resale Access lines variable and access indices


         log(LLUlines) = β ′ + β ′ log(Bitstreamlines) + β ′ accessLLU + β ′ controls + u′
                           0     1                         2               x


where accessLLU is the sum of the use of bitstream and resale .
   We also test the second specification allowing for lags in the exogenous variables. We can also
test a third specification of the light LOI, that is in equation 3 we test if the rate of growth of the




                                                   12
unbundled line is an increasing function of the rate of growth of resale and bitstream access lines.


∆(LLUlines) = β 0 + β 1t−1 ∆(Bitstreamlines)t−1 + β 1t−2 ∆(Bitstreamlines)t−2 + β ′ (Bitstreamlines)t−
                                                                                  1t−3

                    +β 2t−1 accesst−1 + β 2t−2 accesst−2 + β 2t−3 accesst−3 + β x controls + u         (


5    The Data

Main data are extracted from the "European Commission Broadband access in the EU" Reports.
The data range is 7 years from July 2002 to July 2009, and our data set covers 15 countries with
no missing values, we have 225 observation, hence the data set is a balanced panel.
    The main variables used are in terms of numbers of broadband lines:

    • Newlines are the number of broadband lines that belong to the new entrants and are not
      DSL or cable technology but new technologies (FTTH, WLL or CPL).We use the number of
      broadband lines as a proxy for investment: new lines, i.e. mainly fiber, built by new entrants
      and LLU lines (Full LLU + Shared LLU).

    • LLUlines are lines that belong to the incumbent operators but are open to new entrants
      through unbundling.

    • We construct the variable access as the sum of four dummies. Each dummy represents a
      type of access: full LLU, shared LLU, bitstream and resale, and equals 1 if the number of
      lines through this type of access represents more that 7% of DSL lines in a country. We also
      construct access2poss as the sum of bitstream and resale dummies.

    • We construct the variable incmktshr as the ratio of the number of DSL or ’new’ lines that
      belong to the incumbent on total broadband lines, that is the incumbent market share.

    As the data is a balanced panel, we can control for a country fixed effect: broadband penetration
may depend heavily on specific demand factors that could be considered to be constant through
the period hence not affect our results.
    Other data are used to test the robustness of our model: We add a variable P laut which
is a regulatory density index in telecommunications constructed by Plaut Economics but only
available for the years 2002-2006 hence it narrows our data set when used. We also control for
the development of cable infrastructure and add data on Cable lines found in EC Implementation


                                                  13
Reports, and data on LLU prices (EC reports). We construct a dummy that equals 1 if monthly
tariffs increase LoI = 0 if ∆(LLUmonthlyf ee) ≤ 0, LoI = 1 if ∆(LLUmonthlyfee) > 0
    Finally, we completed the data set with data from OECD source concerning GDP per capita
(log(GDP percapita) is the constant GDP per capita in logarithm) and with data from Eurostat
and ITU concerning population density density and mobile penetration rate mobpenrate.


6    Estimation results

In the naive specification of the model, with the OLS method, we find, in table 2, no direct impact of
unbundled lines on new investment: the coefficient of the variable log(LLUlines) is not significantly
different from zero. Too many types of access is damaging for new investment: the coefficient
of the variable access (−0.813) is significantly negative. Hence, this naive test rejects the LOI.
However there seems to be a LOI effect on the last rung: bitstream access stimulates LLU (0.156)
but the number of accesses still negatively affect the number of unbundled lines. Concerning our
control variables, our results are in line with other studies: the population density and the mobile
penetration rate positively and strongly impact investment in new infrastructures. Regarding the
scope of our database (221 observations), our R2 values can be considered as satisfactory.
    Nonetheless, this naive specification cannot be considered as a decisive test of the LOI approach.
Our model in section 3 emphasized the fact that a decision process takes time. Hence we need
to test if the development of service-based competition induces the development of facility-based
competition in the following periods.
    We introduce two lags in order to match the investment decision process (results are reproduced
in table 3). We find no direct impact of past LLU on Newlines investment: all coefficients are not
significantly different from zero. Hence, in this lagged specification we find no evidence supporting
the LOI approach. However the development of Bitstreamlines favors two periods afterwards the
development of LLUlines. There seems to be a LOI effect on the last rung of the ladder.
    Eventually we proceed to an econometric analysis which differentiates flows from stock. We
cannot log-characterize our model anymore because of the deltas. This analysis is not available for
new investment, but only for unbundled investment, because of new infrastructures market youth.
Hence we can only test the light LOI approach. We find, in table 4, that even if the stock of lagged
opened lines has a positive impact (0.210) on unbundled lines, the rate of growth per se has no
effect (the coefficient does not significantly differ from zero). Hence the specification 3 gives no


                                                 14
support to the revisited LOI approach.
   To conclude, we find no ladder effect between LLUlines and Newlines but a ladder effect
between Bitstreamlines and LLUlines. Hence, new entrants climb up the ladder until the last
rung but they don’t manage to climb to the next floor and build their own infrastructure.

                                      Table 2: Naive specification
                                            (IV)           (OLS)            (OLS)
                                       log(Newlines) log(Newlines)      log(LLUlines)
              log(LLUlines)               -1.015*          0.0541
                                           (-1.84)         (0.54)

              log(Bitstreamlines)                                         0.156***
                                                                           (4.48)

              access                          0.397         -0.813***
                                              (0.58)          (-3.29)

              access2poss                                                  -0.491*
                                                                           (-1.66)

              incmktshr                     5.084***        3.077***       1.047
                                             (2.94)          (2.74)        (1.24)

              log(GDPpercapita)               0.428          -0.237        0.159
                                              (0.12)         (-0.09)       (0.07)

              density                        0.0484          0.161         -0.0144
                                             (0.35)          (1.60)        (-0.19)

              mobpenrate                    0.0920***       0.0502***    0.0470***
                                              (3.63)          (4.41)       (5.81)

              log(pop)                       109.4**         34.55**      71.84***
                                              (2.56)          (2.15)       (6.58)

              cons                          -1801.1***      -588.5**     -1172.9***
                                              (-2.63)        (-2.36)       (-7.01)
              N                                 221            221           221
              R2                                              0.425         0.702
              adj. R2                                         0.365         0.671
              t statistics in parentheses
              * p < 0.1, ** p < 0.05, *** p < 0.01




                                                       15
                         Table 3: Lagged specification
                                  (IV)          (OLS)         (OLS)
                            log(Newlines) log(Newlines)   log(LLUlines)
L.log(LLUlines)                  -3.112         0.0934
                                (-0.45)         (0.52)

L2.log(LLUlines)                  3.438      -0.0552
                                  (0.51)     (-0.36)

L.log(Bitstreamlines)                                        0.0186
                                                             (0.47)

L2.log(Bitstreamlines)                                      0.138***
                                                             (3.81)

L.access                         -0.292       -0.347
                                 (-0.60)      (-1.21)

L2.access                        -1.010       -0.132
                                 (-0.66)      (-0.55)

L.access2poss                                                -0.0259
                                                             (-0.12)

L2.access2poss                                               -0.0953
                                                             (-0.46)

incmktshr                        -6.850       2.311*         0.385
                                 (-0.52)      (1.67)         (0.52)

log(GDPpercapita)                -1.028       -0.884         -0.964
                                 (-0.36)      (-0.34)        (-0.71)

density                           0.677      0.324***        0.0103
                                  (1.07)      (3.09)         (0.19)

mobpenrate                       -0.0288    0.0337***      0.0450***
                                 (-0.45)      (2.96)         (8.26)

log(pop)                         -20.86       25.17         34.36***
                                 (-0.28)      (1.44)         (4.18)

cons                              258.2       -452.5        -550.4***
                                  (0.23)      (-1.65)         (-4.31)
N                                  162          192             192
R2                                             0.408           0.758
adj. R2                                        0.327           0.725
t statistics in parentheses
* p < 0.1, ** p < 0.05, *** p < 0.01



                                       16
Table 4: The ladder approach revisited: Flows and Stocks specification
                                        D.LLUlines
                 LD.Bitstreamlines         0.120
                                           (0.90)

                 L2D.Bitstreamlines               0.102
                                                  (0.82)

                 L3.Bitstreamlines               0.210***
                                                  (4.23)

                 incmktshr                     -1109977.2***
                                                   (-5.59)

                 log(GDPpercapita)               -219606.1
                                                   (-0.87)

                 density                        25705.1**
                                                  (2.32)

                 mobpenrate                       -178.7
                                                  (-0.17)

                 log(pop)                       -2871269.5*
                                                   (-1.68)

                 cons                          45718838.1*
                                                  (1.72)
                 N                                 177
                 R2                               0.335
                 adj. R2                          0.239
                 t statistics in parentheses
                 * p < 0.1, ** p < 0.05, *** p < 0.01




                                      17
7         Robustness

Since we do not find any significant effect of the ladder of investment approach, we can make two
different hypothesis. First, the ladder of investment simply does not work. Second, regulators have
wrongly implemented this approach or not yet implemented it to the very last rung (i.e., in order
for the entrants to build their own infrastructures). Hence in order to assess the robustness of
our conclusions we need to control for the regulator’s behavior. All our robustness regression are
reproduced in appendix.


7.1         Controlling for the regulator’s behavior

To control for the impact of the regulator, in particular the strength and legal framework concerning
unbundling, we add a control variable P laut.
         Plaut15 overall index evaluates the regulatory density in each country, i.e. this country “reg-
ulates the telecommunications sector more intensively than countries with a lower index value”.
However Plaut index is available for the length 2002-2006 only. This robustness test strengthens
our results on the “flows and stock” model, that is on specification 3, with a very significant and
highly positive effect of Plaut index.


7.2         Controlling for cable

DSL and cable are likely to be part of the same market as shown by Cardona et al (2009) who
derived this result from estimating DSL demand elasticities in regions of Austria. There are huge
disparities between European countries in broadband cable modem. We create a dummy variable
“cable” to classify countries into two groups according to their use of broadband cable modem. We
use this dummy as a cross-variable with the number of bitstream access lines, on LLU “flows and
stock” regression. This control does not affect our results on line stock.


7.3         Controlling for tariffs

We proceed to a test of the second hypothesis, by taking into account the LLU pricing evolution.
Data on LLU prices are from European Commission Implementation Reports and cover 6 years (on
an annual basis).
         No data is available for bitstream access prices so we cannot test the light LOI approach.

    15
         Plaut index is used in Grajek & Röller (2009)


                                                         18
         We use the tariffs of the LLU product defined as the monthly fee16 as given in the European
reports. We expect that there is a lag between the time tariffs are low and the time new firms start
new investment. Hence we regress the importance of new infrastructure in date t on tariffs in date
t − 2. We test specification 2 on new infrastructures. There is no direct impact of unbundled lines
on new investment even if LLU price increase (i.e., the ladder of investment is well implemented
for the last rung).


7.4         Controlling for idiosyncratic shock

We proceed to a test taking into account the possibility of idiosyncratic shocks. We introduce time
dummies in our regression, in table 9. This test strengthens our results on the “flows and stock”
model, that is on specification 3.


8         Conclusion

Those results undermine the ladder of investment approach, because mandatory unbundling doesn’t
encourage new entrants to invest in their own infrastructure.
         Neither the number of LLU lines, neither the number of levels of the ladder seem to have any
impact on investment in new infrastructures by new entrants.
         However, the number of bitstream access lines used, has a significant and positive impact on
LLU investment.
         Our results are consistent with the “standard view” of the relation between service-based com-
petition and facility-based competition for New infrastructures (i.e. for the LLUlines-Newlines
gap).
         Our results are consistent with the “Ladder of Investment” only for Local Loop Unbundling.
The ladder of investment seems to work well for the Bitstreamlines-LLUlines gap but does not
work for the LLUlines-Newlines gap, i.e., its final goal. Entrants can actually climb up the ladder
to LLU but they cannot reach the next floor.

    16
         We also controlled for more complicated LLU tariffs, including monthly fees and/or connection costs.




                                                            19
References

ARCEP, 2007, "Rapport d’activité 2006", available at http://www.arcep.fr/uploads/tx_gspublication/rap2006
light.zip.

Bourreau, M. and Dogan, P., 2004, "Service-based vs. Facility-based Competition in Local Access,
Networks", Information Economics and Policy, 16, 287-306

Bourreau, M. and Dogan, P., 2006, "Build or Buy Strategies in Local Markets", American Economic
Review, Papers and Proceedings, Vol. 96, n◦ 2, 2006, pp. 72-76

Bourreau, M., Dogan, P. and Manant, M., 2009, "A critical review of the ladder of investment
approach", Working paper, Telecom parisTech.

Cardona, M., Schwarz, A., Yurtoglu, B. and Zulehner, Ch., 2009, "Demand estimation and market
definition for broadband Internet services", Journal of Regulatory Economics, 35, 70—95

Cambini, C. and Jiang, Y., 2009, "Broadband Investment and Regulation: A Literature Review",
Telecommunications Policy, Vol. 33(10-11), 559-574

Cave, M., 2006, "Encouraging infrastructure competition via the ladder of investment", Telecom-
munications Policy, Vol. 30(3-4), 223-237.

Cave, M., 2009, "Snakes and ladders: Unbundling in a next generation world", Telecommunications
Policy, Online, Dec.

Crandall, R.W., Ingraham, A.T. and Singer, H.J., 2004. Do Unbundling Policies Discourage CLEC
Facilities-Based Investment? Topics in Economic Analysis and Policy, 4(1), article 14.

Crandall, R.W. and Sidak, J.G., 2007, "Is Mandatory Unbundling the Key to Increasing
Broadband Penetration in Mexico?      A Survey of International Evidence", Draft available at:
http://ssrn.com/abstract=996065

Distaso, W., Lupi P. and Manenti, G.M., 2006, “Platform Competition and Broadband Uptake:
Theory and Empirical Evidence”, Information Economics and Policy, Vol. 18(1), 87-106.

Distaso, W., Lupi P. and Manenti, G.M., 2009, "Static and Dynamic Efficency in the European
Telecommunications Market: The Incentives to Invest and the Ladder of Investment", Handbook of
Research on Telecommunications Planning and Management - (AEBR) Book Series, forthcoming.

                                                20
ERG,       2005,   "Broadband      market    competition    report",     ERG     05   (23),      available   at
http://erg.ec.europa.eu/doc/publications/erg_05_23_broadbd_mrkt_comp_report_p.pdf

Grajek, M. and Röller, L-H., 2009, "Regulation and Investment in Network Industries: Evidence
from European Telecoms", ESMT working paper june, n◦ 09-004.

Hausman, J.A. and Sidak, J.G., 2005, “Did Mandatory Unbundling Achieve its Purpose? Empirical
Evidence from Five countries”, Journal of Competition Law and Economics, 1(1), 173—245.

Hazlett,      T.   and      Bazelon,   C.,     2005,      "Regulated      Unbundling       of     Telecommu-
nications      Networks:          A    Stepping     Stone       to      Facilities-Based      Competition?",
George       Mason       University,   Analysis        Group,        September    2005,         available    at
http://web.si.umich.edu/tprc/papers/2005/503/Stepping%20Stone%20TPRC_9_20.pdf

Höffler, F., 2007, "Costs and Benefits from Infrastructural Competition: Estimating Welfare Effects
from Broadband Access Competition", Telecommunications Policy, Vol. 31, 401—418.

Waverman L., Meschi M., Reillier B. and Dasgupta, K., 2007, "Access Regulation and Infrastruc-
ture Investment in the Telecommunications Sector: An Empirical Investigation", LECG report,
available at http://www.etno.be/Portals/34/ETNO%20Documents/LECG_Final%20Report.pdf.




                                                    21
9      Appendix




                                   Table 5: Description of variables
 Variable               Description                                        Source
 Main variables
 Newlines               Number of subscribers using entrants’ New in-      European Commission*
                        frastructures (Fibre + WLL + PLC)
 LLUlines               Number of subscribers using Local Loop Un-         Ibid.
                        bundling (Shared + Full)
 Bitstreamlines         Number of subscribers using Bitstream Access       Ibid.
 Resalelines            Number of subscribers using Resale Access          Ibid.
 access                 Sum of access dummies                              Ibid.
 access2poss            Sum of Resale and Bitstream dummies                Ibid.

 Controls
 incmktshr              Incumbent (DSL + New facilties) market share on    European Commission*
                        broadband market (DSL + Cable + New facilities)
 log(GDPpercapita)      logarithm of Gross Domestic Product per capita     OECD (extracted Oct.2009)
                        in € (constant prices)
 mobpenrate             Mobile Penetration Rate                            ICT Eye - ITU
 density                Density                                            OECD and Eurostat
 pop                    Population                                         OECD**

 Robustness variables
 Plaut                  Plaut Economics Regulation Index                   Plaut Economics 2007
 Cable                  Dummy variable; equals 1 if the country is above   European Commission*
                        the mean in EU15 Cable internet use.
 LoI                    Dummy that equals 1 if monthly tariffs has in-      European Commission***
                        creased


    *: COMMUNICATIONS COMMITTEE, Working Document, Broadband access in the EU: situation at

1 July 2009 & Broadband access in the EU: situation at 1 July 2009 (for data from January 2006 to July

2009) and Broadband access in the EU: situation at 1 July 2008 (for data from July 2002 to July 2005).

    **: Except France: INSEE, because OECD Factbook 2009 doesn’t take into account overseas territories

whereas the French National Regulatory Agency does when counting broadband lines.

    ***: From Implementation reports n◦ 9, 10, 12 and 14




                                                   22
                        Table 6: Summary statistics
     Variable          Mean         Std. Dev.       Min.       Max.      N
Newlines             46232.049      96312.475          0      593000    223
LLUlines             743841.222    1542258.416         0     8301000    225
Bistreamlines        252909.067     473475.558         0     2214000    225
Resalelines          302996.533     869152.276         0     4550197    225
access                  2.267          0.813           1         4      225
access2poss             0.956          0.489           0         2      225
incmktshr               0.509          0.169           0       0.939    221
log(GDPpercapita)     9.491929      0.3747044     8.693965   10.36277   225
mobpenrate             110.609        21.109        66.607    173.823   225
density                159.053       122.401        17.079    484.693   225
pop                 25941358.893 26515002.274 446000         82807000   225
Plaut                   0.559          0.096         0.326     0.753    60
cable_dummy             0.467           0.5            0         1      225




                                    23
                Table 7: Robustness test: Plaut regulatory index and Cable
                                     (IV)           (OLS)            (IV)        (OLS)
                                log(Newlines) log(Newlines) log(Newlines)    log(Newlines)
L.log(LLUlines)                     -3.019          0.0906          -2.593       0.0402
                                    (-0.70)         (0.27)         (-0.47)       (0.13)

L2.log(LLUlines)                       2.584          -0.338     2.408          -0.227
                                       (0.69)         (-0.97)    (0.54)         (-1.03)

L.cable_dummy*log(LLUlines)                                      2.076          -0.0535
                                                                 (0.41)         (-0.14)

L2.cable_dummy*log(LLUlines)                                    -1.232          0.319
                                                                (-0.31)         (1.11)

Plaut                                  6.540          7.044
                                       (0.80)         (0.90)

L.access                               -0.595         -0.680    -0.385          -0.312
                                       (-0.83)        (-1.07)   (-0.89)         (-1.10)

L2.access                              -0.279         -0.139    -0.765          -0.162
                                       (-0.24)        (-0.28)   (-0.89)         (-0.67)

incmktshr                              -6.953         3.602     -8.239          1.498
                                       (-0.64)        (1.63)    (-0.71)         (1.07)

log(GDPpercapita)                      -8.263         9.564     -0.00133        0.424
                                       (-0.30)        (0.61)     (-0.00)        (0.16)

density                                0.992          0.340     0.374**        0.260**
                                       (1.13)         (0.97)     (2.46)         (2.36)

mobpenrate                             0.00801        0.0564    -0.00820      0.0382***
                                        (0.07)        (1.03)     (-0.38)        (3.29)

log(pop)                               -18.24         38.13      17.07         35.95**
                                       (-0.17)        (0.72)     (0.74)         (2.02)

cons                                   234.3          -770.5    -325.8         -630.4**
                                       (0.14)         (-0.98)   (-0.89)         (-2.26)
N                                        45             60        162             192
R2                                                     0.372                     0.433
adj. R2                                               -0.059                     0.347
t statistics in parentheses
* p < 0.1, ** p < 0.05, *** p < 0.01




                                                 24
  Table 8: Robustness test: Plaut regulatory index and Cable
                                        (1)            (2)
                                    D.LLUlines     D.LLUlines
LD.Bitstreamlines                      0.291        0.368***
                                      (0.54)         (3.52)

L2D.Bitstreamlines                            0.460          0.139
                                              (0.93)         (1.42)

L3.Bitstreamlines                             0.220**       0.127***
                                               (2.22)        (3.55)

LD.cable_dummy*Bitstreamlines                                -0.399*
                                                             (-1.84)

L2D.cable_dummy*Bitstreamlines                               0.0418
                                                             (0.21)

L3.cable_dummy*Bitstreamlines                               0.987***
                                                             (11.14)

Plaut                                       1466897.6**
                                               (2.18)

incmktshr                                   -817310.9**   -1101474.2***
                                               (-2.64)        (-8.20)

log(GDPpercapita)                           -1213585.8     462572.9**
                                              (-0.93)        (2.53)

density                                       7527.4       -17789.5**
                                              (0.31)         (-2.17)

mobpenrate                                    3664.1         533.9
                                              (0.78)         (0.73)

log(pop)                                    -1375021.1    -2636851.8**
                                              (-0.39)        (-2.22)

cons                                        32165891.0    42211441.9**
                                              (0.64)         (2.29)
N                                               45             177
R2                                            0.705           0.702
adj. R2                                       0.383           0.652
t statistics in parentheses
* p < 0.1, ** p < 0.05, *** p < 0.01




                                       25
                          Table 9: Robustness test: monthly LLU fee
                                (IV)            (IV)           (OLS)         (OLS)
                           (without tariff (with tariff) (without tariff)    (with tariff)
                           log(Newlines) log(Newlines) log(Newlines)     log(Newlines)
L2.log(LLUlines)               -0.144          -0.147         -0.0302       -0.0208
                               (-0.40)         (-0.41)        (-0.22)        (-0.15)

L2.log(LLUlines)*LoI                         0.0217                         0.0314
                                             (0.43)                         (0.52)

incmktshr                        -2.399      -2.192         -2.210          -1.975
                                 (-1.01)     (-0.88)        (-1.12)         (-0.97)

log(GDPpercapita)                -4.538      -4.372         -0.993          -0.671
                                 (-1.10)     (-1.03)        (-0.22)         (-0.15)

density                         0.417***    0.427***       0.412***        0.424***
                                 (3.14)      (3.15)         (3.38)          (3.40)

mobpenrate                       0.0392      0.0398         0.0193          0.0192
                                 (1.34)      (1.36)         (1.01)          (1.00)

log(pop)                         -0.932      -3.864         9.896           5.356
                                 (-0.04)     (-0.15)        (0.51)          (0.25)

cons                             -1.826      42.96          -211.5          -142.3
                                 (-0.00)     (0.11)         (-0.72)         (-0.44)
N                                  60          60             75              75
R2                                                           0.578           0.580
adj. R2                                                      0.421           0.414
t statistics in parentheses
* p < 0.1, ** p < 0.05, *** p < 0.01




                                            26
Table 10: Flows and Stocks specification with semester dummies
                                   (OLS)          (OLS)
                               log(Newlines)    D.LLUlines
       L.log(LLUlines)             0.0127
                                   (0.07)
       L2.log(LLUlines)            -0.164
                                   (-1.02)
       LD.Bitstreamlines                            0.0654
                                                    (0.47)
       L2D.Bitstreamlines                           0.0978
                                                    (0.75)
       L3.Bitstreamlines                          0.157***
                                                    (2.91)
       semester1-3                    .                .
                                      .                .
       semester4                   -0.245         -54608.4
                                  (-0.42)          (-0.47)
       semester5                   0.377          -12442.3
                                   (0.58)          (-0.12)
       semester6                   0.604           33682.9
                                   (0.86)           (0.34)
       semester7                   0.877           38717.5
                                   (1.14)           (0.43)
       semester8                   1.249           46996.4
                                   (1.47)           (0.58)
       semester9                   1.351           74931.2
                                   (1.44)           (1.00)
       semester10                  1.217           94739.0
                                   (1.17)           (1.35)
       semester11                  1.283         131735.9**
                                   (1.14)           (1.98)
       semester12                  1.635         152326.5**
                                   (1.38)           (2.38)
       semester13                  1.946         131797.4**
                                   (1.57)           (2.16)
       semester14                 2.550**          22319.5
                                   (2.01)           (0.41)
       incmktshr                   2.816*      -1072025.9***
                                   (1.95)          (-5.38)
       log(GDPpercapita)           0.589        -930930.0***
                                   (0.18)          (-2.87)
       density                    0.258**         21766.2*
                                   (2.45)           (1.97)
       mobpenrate                 0.00400           -226.8
                                   (0.21)          (-0.12)
       log(pop)                    21.09        -3901689.7**
                                   (1.14)          (-2.07)
       cons                        -386.2       69940491.2**
                                  (-1.31)           (2.33)
       N                             192              177
       R2                          0.877            0.790
       adj. R2                     0.852            0.742
       t statistics in parentheses
       * p < 0.1, ** p < 0.05, *** p < 0.01

                                  27

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:2
posted:3/10/2012
language:
pages:27