[PPT] - Unbundling the incumbent: Evidence from UK broadband M. Nardotto PowerPoint Presentation

SLIDE 1

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Unbundling the incumbent: Evidence from UK broadband

M. Nardotto (ParisTech), T. Valletti (Imperial and

ParisTech) and F. Verboven (KU Leuven) April 2012

SLIDE 2

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Introduction

Policies aimed at “opening” the incumbent have been

ften applied:
Electricity market
Transportation
OUR CASE: Telecommunications (all over EU, and

elsewhere – not US though)

Motivation: higher competition downstream should lead to lower prices, higher quality and more demand Few empirical studies answering the question: DOES IT WORK?

SLIDE 3

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

This paper

We study the broadband market in UK, where: BT owns the network (Backbone and Local Exchanges) Two possible entry strategies (downstream):

Bitstream: low investment, BT still provides most of the

services (reduced area of responsibility of entrant)

Local-Loop Unbundling (LLU): bigger investment by

entrants

Another competitor is cable (NTL:Telewest, now Virgin) with a competing “closed” network

SLIDE 4

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

This paper – findings

MAIN QUESTION: Did entry through LLU induce market expansion? RESULT: penetration is not improved by LLU Then, we show:

1

Incentives for LLU change the composition of entrants’ technological portfolio

2

Inter-Platform competition leads to more penetration (BT lines v.s. cable)

3

Quality improves with LLU

SLIDE 5

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Related literature

Theory papers (surveys of Armstrong, Laffont and Tirole) Very rich policy (and advocacy) debate. Very few empirical works, almost no micro-data Grajek and Roller (2012, JLawEc): Regulation negatively affects total investments in facilities. Panel of EU countries, aggregate data Sraer (2008): WP on LLU and penetration (French data), Fevrier, Gaubert and Sraer (2011): structural model (cross-section) Minamihashi (2011): effect of regulation of both LLU and fiber in

Japan. Regulation made more difficult to invest in fiber

Studies on entry in US using earlier data: Greenstein and Mazzeo (2006, JIndEc); Economides, Seim and Viard (2008, RAND), Xiao and Orazem (2011, IJIO), Goldfard and Xiao (forthcoming, AER)

SLIDE 6

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Broadband market

Broadband in UK is provided by:

Fixed line telephone operators
Cable operator

British Telecom group (BT retail/BT wholesale/ Openreach) is the incumbent of the fixed line telephone network Virgin Media is the only cable operator (network built in the 80s) BT’s network has a (fiber) backbone and serves consumers locally through delivery points (local exchanges, LEs) which cover the “last mile” Each household is connected to a specific LE in the grid

SLIDE 7

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

The network

BB Backbone

SLIDE 8

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Whole BT grid of LEs + cable

White dots: LEs areas
Red dots: LEs areas

where cable covers at least 65% of lines

SLIDE 9

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Broadband market

Late 90s: BT opened the last mile to entrants Regulation by Ofcom:

1

BT had to upgrade the network to allow entry in the LE (completed in 2002-03)

2

Regulates wholesale prices that entrants pay to BT

Entry options:

(Simple re-selling) Bitstream: low investment by the entrant in the LE, the connection is still managed by BT LLU: higher investment by the entrant. Entrants install DSLAM equipment, routers, cabinets, provide maintainance, etc.

SLIDE 10

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Regulation and goals

Regulator’s perspective (Ofcom WBA rev. 2006):

The importance of broadband continues to grow. It plays a central role in communications used by many consumers and businesses to keep in touch, access information and conduct business [. . . ] Competition has an important role to play in delivering what consumers and businesses need [. . . ] Competition at the retail level depends on ISPs having access to wholesale broadband services or local loop unbundling (LLU) to build their services. Ofcom has identified that competition between networks based on LLU, rather than just at the retail level based

n wholesale broadband products, is crucial to maintaining the

UK’s broadband progress.

SLIDE 11

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Broadband market

Called “Central Offices” in the US 5,587 LEs in the UK grid Becoming LLU supplier costs approx £100k per LE (one-off), plus

ngoing costs

SLIDE 12

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Main dataset

Quarterly panel data on investments and subscriptions in the period 2005 (Dec) – 2009 (Dec) at the LE level (5,587). For each LE we observe:

Total number of potential lines (households)
Number of lines potentially served by cable
Lines actually served by cable
Lines served by entrants through LLU (disaggregated by
perator)
Total lines served by bit-stream
Total lines with broadband (LLU + cable + bit-stream)

SLIDE 13

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Data on Demographics

We complement the dataset with data on demographics at the LE level:

Age structure (5 classes: 0-14, 15-19, 20-29, 30-44,

44-59, more than 60)

Number of cars or vans per household
Income (weekly gross)
Sector of occupation (e.g. agricultural, manufacturing,

financial intermediation etc.)

Education
Urban / Rural area

Data from the UK Office of National Statistics (ONS) and UK census We cover England and Wales

SLIDE 14

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Broadband market

Penetration increases significantly over

ur time span

Over time the relative share of BT retail is stable Entrants substitute bit-stream with LLU

SLIDE 15

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Descriptive statistics

Subscribers and coverage 2005Q4 2007Q4 2009Q4

Num. lines

27,576,261 27,658,092 28,219,684

Num. subscribers

10,052,446 15,624,059 17,664,344

BT

26% 26.3% 24.7%

Bit-Stream

41% 24.2% 15.3%

LLU

2.2% 25.4% 37.7%

Cable

30.8% 24% 22.4% Broadband penetration 36.5% 56.5% 62.6%

Num. of LEs

5,587 5,587 5,587 LLU entry 695 (12.4%) 1,733 (31%) 2,011 (36%)

Avg. num. of

1.79 3.44 3.31 LLU competitors Cable coverage –65% 953 (17%) 844 (15.1%) 829 (14.8%) Demographics LEs without LLU Unbundled LEs Test– Mean

Std. dev

Mean

Std. dev

Stat P-value Urban (%) 13 33.6 77.4 41.8

47.85

<0.001 Lines 1,243 1,463 12,135 8,444

57.56

<0.001 Income 568.8 110.5 514.6 126.4 15.63 <0.001 HS occupations 53.5 10.4 53.5 14.3 0.99 HS occupations sectors 26.5 8 29.9 11.6

2.39

0.017

Pop. 0-14 y.o.

17.4 2.7 16.8 4.5 .47 0.64

Pop. 15-60 y.o.

57.6 4.3 60 7.2

1.51

0.13

Pop. more 60 y.o.

25 5.7 23.2 7.6 1.32 0.19

SLIDE 16

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

LLU expansion

In few years the adoption of LLU becomes more widespread (left is Dec. 2005, right is Dec. 2009):

SLIDE 17

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Does LLU increase penetration?

In the graph the penetration rate over time (from Dec. 2005 to Dec. 2009)

Dashed line: LEs

receiving at least 1 LLU investment

Solid line: LEs with no

LLU Non-unbundled LEs seem to catch up. . .

SLIDE 18

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Does LLU increase penetration?

Main equation: P enit = ”i + fit + ˛LLUit + ‚xit + "it where, ”i; fit : LE area fixed effect, time effects LLUit: dummy variable for having received LLU xit: control variables (income, cable)

SLIDE 19

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Does LLU increase penetration?

Main equation: P enit = ”i + fit + ˛LLUit + ‚xit + "it where, ”i; fit : LE area fixed effect, time effects LLUit: dummy variable for having received LLU xit: control variables (income, cable) THREE identification strategies: Panel FE estimation (no time-varying instrument) IV Panel estimation (time and spatial lags as instruments) System GMM Blundell-Bond estimator

SLIDE 20

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Does LLU increase penetration?

Dependent Variable: Total Broadband Penetration Panel FE Panel IV Panel IV GMM GMM (Spatial lag) (Time lag) LLU entry LLU entry

0.009***
0.057***
0.009***

0.002 0.003* (0.001) (0.003) (0.003) (0.003) (0.002) Cable Coverage 0.01 0.012*** 0.022*** 0.045*** 0.016*** (0.008) (0.004) (0.001) (0.008) (0.005) Log(income) 0.053*** 0.064*** 0.041*** 0.057*** 0.034*** (0.007) (0.004) (0.004) (0.000) (0.002) Tot Penetrationt-1 0.470*** (0.039) Constant 0.009

0.054**
0.114***

(0.042) (0.027) (0.026) Period FE YES YES YES YES YES Observations 72,505 72,505 68,240 72,505 68,240 Number of LEs 4,265 4,265 4,265 4,265 4,265 R-squared 0.57 0.51 0.53 Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1

SLIDE 21

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Entry

We estimate a model of entry à la Bresnahan and Reiss extended to account for entry / exit sunk costs as in Xiao and Orazem (2011) In this model entry today is affected by past choices due to entry costs Main assumptions:

Markets are local Products and firms are homogeneous

SLIDE 22

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Model of entry

nt is the number of forms in the local market i in period t V (nt; Xt) is the value function of an active firm with exogenous characteristics (Xt; "t) EC and SV are respectively the entry cost and the scrap value In each period three regimes:

Exit: 8 > > > > > > < > > > > > > : nt = 0 if V (1; Xt) ` " < SV

r

nt = n > 0 if V (n; Xt) ` " – SV and V (n + 1; Xt) ` " < SV (1)

SLIDE 23

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Model of entry

Entry:

8 > > > > > > < > > > > > > : nt = N if V (N; Xt) ` " – EC

r

nt = n < N if V (n; Xt) ` " – EC and V (n + 1; Xt) ` " < EC (2)

Inaction:

nt = nt`1 iff 8 > > > > > > > > > > > > > < > > > > > > > > > > > > > : nt`1 = 0 and V (1; Xt) ` " < EC

r

nt`1 = N and V (n; Xt) ` " – SV

r

0 < nt`1 < N and V (n; Xt) ` " < EC and V (n + 1; Xt) ` " – SV (3)

SLIDE 24

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Model of entry

We pool all quarters (Dec. 2005 – Dec. 2009) and include a time trend (also interacted with some regressors) Estimated by conditional maximum likelihood Entry and exit thresholds can be derived from the model

SLIDE 25

Introduction Broadband market LLU and penetration LLU and entry LLU and quality Dependent Variable: LLU entry

Num. of competitors

BR BR spatial Sunk cost BR Sunk cost – spatial Log(lines) 0.519*** 0.463*** 0.578*** 0.567*** 0.742*** (0.016) (0.016) (0.030) (0.014) (0.022) Log(lines) ˆ Trend 0.067*** 0.065*** 0.011*** 0.082*** 0.032*** (0.002) (0.002) (0.003) (0.001) (0.002) Log(income) 0.106 0.172** 0.095 0.393*** 0.694*** (0.056) (0.057) (0.109) (0.048) (0.070) Working age 0.029*** 0.032*** 0.017** 0.028*** 0.030*** (0.003) (0.003) (0.005) (0.002) (0.003) White

0.018***
0.013***
0.004
0.019***
0.013***

(0.002) (0.002) (0.003) (0.001) (0.001) HS sector 0.019*** 0.012*** 0.008** 0.026*** 0.014*** (0.001) (0.001) (0.003) (0.001) (0.002) Urban 0.080*** 0.103*** 0.077* 0.062*** 0.065** (0.018) (0.019) (0.033) (0.016) (0.023) Trend

0.449***
0.440***
0.068**
0.569***
0.217***

(0.014) (0.014) (0.025) (0.012) (0.019) LLU within 5 miles 0.555*** 0.440*** (0.016) (0.020) Sunk cost 3.679*** 3.045*** (0.035) (0.026) Fixed effect firm 1 6.380*** 7.206*** 4.252*** 8.494*** 9.900*** (0.494) (0.499) (0.936) (0.410) (0.596) Fixed effect firm 2 9.349*** 10.414*** (0.410) (0.596) Fixed effect firm 3 9.853*** 10.758*** (0.411) (0.597) Model chi-square 54,481.065 55,678.629 76,880.746 71,327.260 10,952.368 Observations 72,505 72,505 68,240 72,505 68,240

SLIDE 26

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Entry thresholds

The number of lines to enter declines rapidly over time Sunk costs are relevant and can be exploited as an instrument Also spatial proximity is an important factor determining entry

SLIDE 27

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

LLU vs. Bit-Stream

Do firm change their technology? DSLit = ”i + fit + ˛LLUit + ‚xit + "it

Dependent Variable: Total DSL Penetration Panel FE Panel IV Panel IV GMM GMM (Spatial lag) (Time lag) LLU entry LLU entry

0.110***
0.157***
0.374***
0.108***
0.020***

(0.003) (0.002) (0.007) (0.007) (0.002) Cable Coverage

0.093***
0.085***
0.086***
0.394***
0.076***

(0.025) (0.007) (0.009) (0.017) (0.007) Log(income) 0.013 0.027*** 0.076*** 0.065*** 0.015*** (0.019) (0.008) (0.010) (0.000) (0.001) DSL Penetrationt-1 0.852*** (0.012) Constant 0.243** 0.203***

0.108*

(0.115) (0.053) (0.064) Period FE YES YES YES YES YES Observations 72,505 72,505 68,240 72,505 68,240 R-squared 0.55 0.54 0.57 Number of LEs 4,265 4,265 4,265 4,265 4,265 Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1

SLIDE 28

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Possible explanations

Lack of competition? Product differentiation? Effective DSL wholesale regulation? Locus of competition is not price, rather quality

SLIDE 29

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Quality of Internet connections

We bought more than 1M observations from speedchecker.com for 2009 on:

Speed of the connection Post code of the test (to compute the distance) Date and time of the test Operator’s name and tariff option

We computed quality indexes by operator/LE/tariff

ption considering BT, the 4 main entrants and the

cable (94% of the market)

Speed tests Download Speed (Mbit/s.) Sample Distance (miles) by operator Mean

Std. dev.

Frequency (%) Mean

Std. dev.

BT 2,887 2,002 29.9 2.057 9.135 LLU entrants 3,221 2,339 51.5 1.823 6.973 Virgin Cable 5,351 3,301 18.6 1.574 5.066

SLIDE 30

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Evidence on quality

ln speedj = ‚0+‚1LLU_OPj+‚2Bitstream_OPj+‚3Cable_OPj+˛Xj+"j

Dependent Log of download speed Log of download speed Variables: (all ISPs) (BT only) Coeff.

Std. err.

Coeff.

Std. err.

Coeff.

Std. err.

LLU 0.178*** (0.009)

0.01

(0.013) TalkTalk 0.207*** (0.011) O2 0.506*** (0.013) Orange 3 0.058*** (0.014) Sky 4 0.086*** (0.011) Bit-stream

0.185***

(0.012)

0.184***

(0.012) Cable 0.569*** (0.012) 0.568*** (0.012) 0.023 (0.015) Urban 0.015* (0.009) 0.016* (0.009) 0.005 (0.013) Log(income) 0.021 (0.015) 0.008 (0.015) 0.011 (0.023) Constant 8.016*** (0.098) 8.087*** (0.097) 8.214*** (0.152)

Dist. and dist.2

YES YES YES Hours & Day YES YES YES Observations 985,590 985,590 358,849 R-squared 0.195 0.204 0.199 Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1

SLIDE 31

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Evidence on quality

LLU operators tend to

propose many contracts with higher speed than BT

We analyze the fastest

(O2) and the most important in terms of mkt share (Sky). Do they sell different products in LLU areas? (w.r. to Bit-stream areas)

SLIDE 32

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Evidence on quality

Figure: O2 Figure: Sky

Non parametric tests strongly rejects the null of equality of the distributions LLU operators not only propose different quality in LLU areas (where they control the connections and they can invest), they also sell different products!

SLIDE 33

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Conclusions

Did regulation work as in the mind of the regulator? Well, partially..

SLIDE 34

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Conclusions

Did regulation work as in the mind of the regulator? Well, partially.. LLU regulation did not increase penetration. The estimated effect is very close to 0%

SLIDE 35

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Conclusions

Did regulation work as in the mind of the regulator? Well, partially.. LLU regulation did not increase penetration. The estimated effect is very close to 0% However, it affected investments, i.e., ISPs did change technology

SLIDE 36

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Conclusions

Did regulation work as in the mind of the regulator? Well, partially.. LLU regulation did not increase penetration. The estimated effect is very close to 0% However, it affected investments, i.e., ISPs did change technology Once obtained the control of the connections, investments made by Entrants led to +18% in avg. speed w.r. to BT

SLIDE 37

Introduction Broadband market LLU and penetration LLU and entry LLU and quality

Conclusions

Did regulation work as in the mind of the regulator? Well, partially.. LLU regulation did not increase penetration. The estimated effect is very close to 0% However, it affected investments, i.e., ISPs did change technology Once obtained the control of the connections, investments made by Entrants led to +18% in avg. speed w.r. to BT Consumers looking for high speed did choose LLU