Efficiency frontiers, stranded assets and the Xfactor for telecoms

1
Efficiency frontiers, stranded assets and the
X-factor for telecoms

• Gordon Hughes
• (University of Edinburgh)
• Mathieu Pearson
• Nigel Attenborough
• (NERA Economic Consulting)

2
The issue

• Consider a company with regulated revenues of 5
  bln per year
• PV of difference between X2 and X4 exceeds 1
  bln over a 5 year price review
• Estimates of inefficiency i.e. potential scope
  for catch-up often exceed 20
• Typical estimates of TFP growth in telecoms vary
  by 1.5-2 across companies countries
• Unknown impact of structural switched from
  analogue to digital technologies on the value of
  assets constructed in the past

3
How is the X-factor determined?

• In practice, as an educated guess based on
• Estimates of TFP growth for the sector or for the
  economy as a whole
• Scope for efficiency improvements
• Special factors public policy, etc
• Large scope for regulatory discretion
• Lack of certainty and transparency
• Analytical confusion about sources of
  productivity improvement e.g. outsourcing

4
Stochastic frontier analysis

• Standard cost function plus two error components
• u(i,t) gt 0 exponential or truncated normal N(µ,
  su2)
• v(i,t) N(0, sv2)
• The u(i,t) error is often interpreted as
  inefficiency but may also reflect heterogeneity

5
Illustration of SFA estimation
6
Sample and cost data

• Data on 68 US local exchange carriers from
  1996-2007 from FCC database
• Current costs from historic cost accounts
• Standard return on capital average 10
• Accounting cost depreciation return on
  depreciated replacement cost of capital
• Economic cost rental on replacement cost
• Stranded assets due to fall in demand for
  analogue switched lines since 2000

7
Ratios of peak to current demand for switched
lines
8
Panel data estimation 1

• Structural break after 1999
• Due to technical change as well as shift in
  composition of demand
• Note switch in sign for coefficient on leased
  (digital) after 2000
• Primary cost drivers
• Total sheath, proportion of fibre in the network,
  average pay, cost of laying underground cable
• Time trend -3.3 to 1999, -2.1 after 2000

9
Estimates for stochastic costs with without
time-varying efficiency
10
Panel data estimation 2

• Lagged adjustment of stranded assets
• Distributed lag with ? 0.888
• Coefficient on stranded assets gt 0.6
• Cumulative impact of 5 annual decline in
  switched lines is equivalent to a cost increase
  of 2.6 per year
• Slow convergence towards frontier costs
• Catch up at 1.3 per year or lt 10 over a price
  review period

11
Inefficiency or heterogeneity?

• Very large dispersion of efficiency error
• Conventional to use lowest decile for reference
• 7 LECs have costs gt 30 above decile
• Worst LECs consistent over time
• Heterogeneity - state fixed effects
• µ f(competition, cost environment)
• Significant factors (2005/06) size as
  quadratic, average household income, average
  retail electricity price, of population born
  outside US

12
Competition in phone markets

• Difficult to measure competition over time
• Focus on entry by Competing LECs (CLECs)
• Fixed effects for heterogeneity (µ)
• Positive coefficients on of switched lines
  provided by CLECs average mobile subscriptions
  per switched line (using 2006 figures)
• Cost frontier
• Entry by CLECs ( of zip codes with no CLECs)
  associated with higher costs

13
Accounting or economic cost of capital?

• Time trends are quite different
• Accounting cost -1.7 per year (after 2000)
• Economic cost 1.7 per year (after 2000)
• Results suggest that much of the apparent fall in
  costs is a consequence of the ageing of the
  capital stock
• The method of measuring capital costs has a
  critical impact on any assessment of the X-factor
  debate about LRIC price caps in US

14
Conclusions 1

• The annual shift in the cost frontier for the US
  LECs changed in 2000 from -3.3 p.a. to -2.1
  p.a. in nominal terms
• The impact of asset stranding due to the fall in
  demand for switched lines is equivalent to a
  further 2.6 p.a.
• Most efficiency differences seem to reflect
  heterogeneity hence the very slow rate of
  apparent catch-up in efficiency levels

15
Conclusions 2

• The effect of competition for local network
  services has a similar effect to the stranding of
  assets - it appears to increase costs over time
  and heterogeneity across companies
• Use of economic rather than accounting costs
  implies that costs have been increasing at 1.7
  p.a. in nominal terms a little less than
  inflation at an average of 2.2 p.a.

16
RPI X today what value for X?

• For US LECs
• Using accounting costs X 1.5 2.5 p.a.
  allowing for asset stranding competition
• Using economic costs X -0.5 -2.0 p.a.
  (prices increase faster than inflation)
• In Europe
• US figures 2.5 but this margin will fall as
  asset stranding starts to affect European network
  operators