OFCOM WBA Charge Control and Charge Control for LLU/WLR Services Cost of

TalkTalk Group response

May 2011

A. INTRODUCTION

1. This is TalkTalk Group’s (TTG) response ’s consultation on the BT Group and cost of capital / WACC that is part of the consultations for both the WBA Charge Control and Charge Control for LLU/WLR Services.

2. TalkTalk Group provides to over 4 million residential and business customers under the TalkTalk, AOL, TalkTalk Business and brands. We are the UK’s biggest local loop unbundler, operate the UK’s largest next generation network (NGN) and are BT’s largest wholesale customer.

3. The conclusion that Ofcom reaches on WACC will have a profound effect on UK consumers, on our business and more broadly on the effectiveness of competition and consumer benefits in the UK. For instance, setting the cost of capital too high will result in consumers paying excessive prices, less effective competition, reduced downstream innovation and investment and inefficient upstream investment. A 1% increase in WACC, for example, will result in a £3 per year increase in the MPF and WLR prices and the amount paid by UK consumers by £70 million every year. Setting the appropriate WACC is not a perfect science, nonetheless it is important that the assumptions are sound.

4. Our response draws extensively on the report commissioned by TalkTalk and Sky from Europe Economics (‘EE report’). This response also comments on other aspects of the WACC calculation that are not covered in the Europe Economics report such as how to take into account the impact of the BT pension scheme and the relative riskiness of Openreach versus network utilities. In the event of any inconsistency between this response and the Europe Economics report, this response represents TalkTalk’s view.

5. The key points in our response are as follows: • The so-called ‘Openreach’ WACC must be conceived of, calculated as and applied as the WACC for copper access services activities (.e. LLU and WLR) and not include the cost of capital of Openreach’s other businesses (such as NGA) • We consider the risk free rate is better set at 1.4% rather than 1.5% given the benchmark data • We consider that Ofcom’s debt premia range for BTG and Openreach of 2.0% to 2.5% is reasonable though there are several strong reasons for selecting a number towards the lower end of the range for Openreach most particularly since the cost of debt is inflated by the pension scheme risk (which should be properly excluded) and also because Openreach should have a lower cost of debt than BTG • We consider that an ERP range of 4.5% to 5.0% is more reasonable than the Ofcom (point) assumption of 5.0% • The assumption of an invariant debt beta that Ofcom used in deriving historical asset betas is, we consider, inappropriate since the variance in debt premia

page 2 over the last three years is indicative of variance in debt beta. Assuming a variable debt beta is more realistic and produces a less volatile and narrower asset beta range. Using this method we estimate the BTG asset beta at 0.55 to 0.64 (versus Ofcom’s estimate of 0.45 to 0.60) and the BTG equity beta at 0.90 to 1.03 (Ofcom 0.78 to 1.08) • There appears to be an error in the calculation and presentation of the network utility benchmarks – accordingly the asset beta range of the benchmarks is 0.25 to 0.35 (not 0.30 to 0.40). • On the basis of the corrected network utility benchmarks (0.25 to 0.35) and the risk characteristics of Openreach (which show much more similarity with network utilities than the rest of BT), we consider than the appropriate Openreach asset beta range is 0.40 to 0.50 (Ofcom assumed 0.40 to 0.55). This places the Openreach asset beta (mid-point) a little closer to the network utility asset beta benchmarks than to the RoBT asset beta which is consistent with its relative risk characteristics. Under our assumptions the ‘wedge’ between the asset / equity beta of Openreach and of BTG is higher than under Ofcom’s assumptions. However, this should not be a cause for any alarm since the ‘wedge’ is little more than an imputed output that, as Ofcom notes, has little ‘science’ about it • We remain of the view (and have provided cogent reasoning to support this) that the BT defined benefits pension scheme risk increases the cost of capital and that the additional cost that this causes should be excluded from the cost of wholesale products. Ofcom have provided no cogent reasoning to contradict this position. In the case of BT, which has a relatively large and risky defined benefits scheme the impact will be large. We accept that it is not simple to calculate the precise impact but we consider that making no adjustment and/or ignoring the impact would be wrong. Accordingly we consider that a possible way to take account of this is to select assumptions near the bottom of the range for debt premia and asset beta • We consider that there are very strong reasons for not ‘aiming up’ or selecting assumptions at the top end of the ranges. First, the asymmetry of consumer benefit argument that is often used to justify aiming up is not present in this case – in fact in this case it is more economically efficient to ‘aim down’. Second, the ranges are already skewed or biased in the sense that the most likely value lies towards the bottom end of the range particularly since the ranges do not account for Openreach having a lower debt premia than BTG or for removing the impact of the BT pension scheme risk • Ofcom will need to update its analysis prior to the final decision. In doing this Ofcom needs to be careful in the way different assumptions move. For instance, if a higher risk-free rate is assumed this would imply (under most assumptions) a fall in the ERP

6. This response laid out as follows: • Role / scope of Openreach WACC

page 3 • Risk free rate • Debt premia, debt beta • Gearing • Equity risk premia (ERP) • BT Group equity / asset beta • Openreach equity / asset beta • Impact of BT pension scheme • Overall results and conclusion

7. If there are any questions regarding this submission please contact Andrew Heaney ([email protected] or 07979 657965).

ROLE / SCOPE OF ‘OPENREACH’ WACC

8. The WLA consultation addresses the question of what is the appropriate WACC to use in the calculation of LLU and WLR costs (referred to as ‘copper access services’) As a short hand Ofcom refer to the WACC for the copper access services as the ‘Openreach’ WACC – for instance at footnote 113 Ofcom state that “Note that the cost of capital for Openreach is more specifically a rate for BT’s copper access services business.” This is also consistent with the approach that Ofcom used in the 2009 LLU Charge Control1.

9. We fully agree that the Openreach WACC should be the WACC for only the copper access services and not for other businesses that form part of Openreach (such as the NGA investment). Consequently the ‘Rest of BT’ WACC will represent the WACC for the other BT divisions (e.g. BT Global Services, BT Wholesale, BT Retail) as well as the non-copper access services parts of Openreach (such as NGA). The RoBT WACC should apply for IPStream (which is part of BT Wholesale).

10. As Ofcom has, in this response we use the terminology of ‘Openreach’ WACC to mean WACC for copper access services. To avoid any misunderstanding, it may be better in its Statement for Ofcom to describe this WACC as the ‘copper access services WACC’ (or ‘LLU/WLR WACC’).

RISK FREE RATE

11. Ofcom assumes a risk free rate (RFR) of 1.5%. This is based on the 5 and 10 year averages for 5 year gilts (which are 1.4% and 1.7% respectively, see §6.68). Ofcom says that “… we continue to favour the use of 5 year gilt yields when estimating the

1 See LLU Appeal Determination §2.253

page 4 risk-free rate …” (Consultation2 §6.58) though it ‘considers 10 year gilts’ (which are coincidently 1.4% and 1.7% as well).

12. Ofcom rejects giving too much weight to more recent figures for the RFR (which are much lower and even negative for some periods) since they ‘based on a period of unusual market activity’ such as the financial crisis and quantitative easing (Consultation §6.60). We agree that recent rates alone are not a robust indicator of the future RFR.

13. However, even ignoring the most recent market data there appears to be a downward trend in RFRs over the last 10 years. Using a 10 year average would give no weight to this downward trend in RFR. For this reason we consider that the 5 year average (of 1.4%) is a more relevant and robust assumption to use than the 10 year average (which is 1.7%). Thus 1.4% should be preferred over 1.5% as the point assumption for RFR.

14. This (point) assumption for RFR is also consistent (or not inconsistent) with the Bristol Water case where the CC used a range of 1% to 2%. As pointed out in the EE Report (§2.7(b)§§), the CC did not recommend 2% as the correct risk-free rate – rather for various reasons (that do not read across to this situation) they selected a WACC at the top of the range which was consistent with a RFR of 2%.

DEBT PREMIUM / DEBT BETA

15. Ofcom assumes a debt premium of 2% to 2.5% for both BT Group (BTG) and Openreach (i.e. copper access services). We concur with this range but feel that the appropriate figure to apply particularly for Openreach is near the bottom of this range. We explain our reasoning below.

16. First, this range implies that (combined with the RFR of 1.5% and inflation of 2.5%) that the nominal cost of debt is 6% to 6.5%. This compares to a recent observed yield for BT bonds of 4% to 6% (see consultation Fig 6.4). Thus in comparison to recent data the assumed range is high.

17. Second, this same debt premia is used for Openreach as for BT Group. There is broad acceptance that the cost of debt (and so debt premium) for Openreach is likely to be lower than for BTG. For example: • in the 2009 LLU charge statement Ofcom said “we consider Openreach to have many characteristics of a network utility, and therefore to carry less specific risk than the rest of BT Group”3 which implies a lower cost of debt (for a given level of gearing)

2 All references to the ‘Consultation’ are to the WBA Charge Control Consultation (unless otherwise stated) 3 LLU Charges Statement (May 2009) §A8.72

page 5 • In the LLU appeal Ofcom’s cost of capital advisor, Professor Franks, went as far as to say that Ofcom might wish to consider in future how to derive a Openreach specific debt premia4 • The CC also accepted that Openreach would have a lower cost of debt (and/or a higher gearing): “We accept that a business exposed to lower overall risk may be able to target a higher credit rating, and hence a lower cost of raising finance, even at a higher level of indebtedness” (LLU Determination §2.366) • That Openreach might have a lower debt premia is also consistent with the fact that utility companies have better credit ratings that other companies. • That Openreach will have a lower debt premia since its risk characteristics are more similar to those of network utilities than rest of BT is clear from empirical data. For example, in late 2009 the difference between the debt premia for network utilities and BT Group was about 1.5%. This implies that the Openreach debt premia was about 0.5% less than the BT Group debt premia5. This point alone would infer that the appropriate debt premia range for Openreach is 1.5% to 2.0%

18. Third, as described below §53, we consider that the WACC used for Openreach should exclude the impact of the BT pension scheme. One impact of the BT pension scheme (which is generally accepted) is to raise the BTG cost of debt. Therefore the appropriate cost of debt (and so debt premia) for Openreach will be lower than for BTG.

19. As we describe below (see §32) we consider that the appropriate debt beta range is 0.20 to 0.25 which is the current forward-looking debt beta level assuming that half of the rise in debt premia is due to an increase in debt beta (equal apportionment).

GEARING

20. Ofcom have assumed a gearing of 50% for BT Group and Openreach which is similar to the average level of gearing for BTG over the last few years (Consultation §6.84). We agree that the gearing has a limited effect on the overall WACC since Ofcom implicitly assumed that the cost of capital (excluding tax effects) is invariant to the

4 LLU Appeal Determination §2.310 “Professor Franks considered the proposal to separate the financial structure of Openreach from that of the BT Group, i.e. to differentiate between their debt ratios, credit ratings and debt premiums. He noted: ‘to try to distinguish in a rigorous way between the two entities will prove a difficult task although one which Ofcom may wish to cover in the future’” 5 see Europe Economics report for Ofwat which shows water debt premia averaging around 1.5% (http://www.ofwat.gov.uk/pricereview/pr09phase3/rpt_com_20091126fdcoc.pdf). At the same time BT Group debt premia was around 3% (see consultation Fig 6.3). If conservatively Openreach debt premia was half way between that of these water companies and rest of BTG then this would imply that the Openreach debt premia was 2.5% and rest of BT 3.5% (BTG is roughly 50% Openreach and 50% RoBT). Thus the Openreach debt premia (2.5%) is about 0.5% below that of BT Group (3.0%)

page 6 gearing since rises in gearing are offset by higher costs of debt and equity (this is consistent with the Miller-Modigliani theory).

21. However, notwithstanding we consider that a higher gearing would be appropriate for Openreach since it has risk characteristics much closer to that of network utilities (rather than RoBT) and thus can sustain a higher level of gearing as can be seem from historic data – for instance, see Brattle Report (Oct 2010) page 10 which shows that network utilities tend to have gearing levels which are typically 10% to 20% higher than BT.

EQUITY RISK PREMIUM (ERP)

22. Ofcom assume an ERP of 5%. This reflects Ofcom’s preference for an ERP based on using an arithmetic mean. We consider that a range of 4.5% to 5.0% is more appropriate.

23. First, this range is more consistent with the CC’s recommendation for total market return (TMR = RFR + ERP) of 5% to 7% which implies a range for ERP of 4% to 5% (see EE Report §3.5(a)). As pointed out in the EE Report (§3.5(a)), the CC did not recommend 5% as the correct ERP – rather for various reasons (that do not read across to this situation) they selected a WACC at the top of the range which was consistent with a ERP of 5%.

24. Second, the 5% figure is consistent with calculating the ERP using an arithmetic mean premium which implies that there is no mean reversion. However, in deriving the RFR Ofcom assume that there is mean reversion in the RFR. These two positions are not consistent (see EE Report §3.5(b)). To be consistent with the approach to RFR (where mean reversion is assumed) would require the use of geometric means to derive the ERP – this would give an ERP of 4.5% or even lower6.

25. Third, Ofcom has argued consistently that the ERP increases in periods of market volatility – for instance, Consultation §6.92, LLU Statement 2009 §§A8.41, A8.42. Since the CC determinations that it has relied on Bristol Water (4% to 5%, data from spring 20107) and LLU appeal (5%, data from spring 20108) were both during periods of market volatility and that market volatility has (or will) subsided it seems reasonable to set an ERP lower than 5%.

26. We also note in respect of ERP that if Ofcom increases the RFR it should consider that as a corollary the ERP should reduce (EE Report §3.6). This is for several reasons:

6 DMS 2009 showed the UK ERP geometric mean at 3.6% 7 This date represents the period when the research / analysis was done rather than the date that the decision was published which was several months later 8 This date represents the period when the research / analysis was done rather than the date that the decision was published which was several months later

page 7 • The total market return (TMR = RFR + ERP) tends to be more stable than the individual components suggesting that rises / falls in RFR would be correlated with opposing movements in ERP • If the rationale for increasing RFR was mean reversion then it would be appropriate to also assume mean reversion in the calculation of ERP which would imply the use of geometric means to derive the ERP which would give an ERP of 4.5% or even lower – see §24 above. • A rise in the RFR would imply a view on the part of Ofcom that the market trends of the past five and more years — which have included equity market turmoil (on which basis Ofcom has, since 2009, included a 0.5 per cent uplift to the ERP) as well as a deep and sustained fall in yields in gilt markets — have decisively reversed. Many of the key arguments that would provide a rationale for a higher RFR (i.e. end of current market volatility) would also provide a rationale for a lower ERP (see §24 above)

BT GROUP EQUITY BETA / BT GROUP ASSET BETA

27. Ofcom’s method to estimate the future BT Group asset beta has been to first derive the historic 1 year / 2 year BTG asset beta based on the historic 1 year / 2 year BTG equity beta, average gearing and the debt beta. The debt beta used in this calculation was assumed to remain constant at 0.125. This provides a data series for the 1 year / 2 year BTG asset beta that varies (quite widely) from 0.43 to 0.65 over the last 3 years (see EE report Figure 6.3). From this data series, Ofcom selects an estimated BTG asset beta range of 0.45 to 0.60. It then uses this figure to estimate the BTG equity beta at 0.68 to 0.98 based on gearing of 50% and debt beta of 0.125.

28. Though we broadly agree with the overall method used we believe that the use of a constant historic debt beta and the forecast debt beta of 0.125 are inappropriate.

29. During the financial crisis the debt premia rose significantly. Theoretically, increases in debt premia result from increases in the default risk and debt beta. By assuming that the debt beta remains constant (as Ofcom has) it implies that all of the rise in debt premia is due solely to increased default risk. We consider this unrealistic (see EE Report §5.220).

30. We think it more plausible to assume that half of the rise in debt premia was due to increased debt beta and half due to default. We refer to this as equal apportionment. There are a number of reasons to support that this is a plausible assumption: • equal apportionment should be the default position since it is rational that debt beta contributes to variation in the debt premia (as well as default risk). In other words, absent a good reason for doing otherwise, equal apportionment is the natural approach; • academic studies suggest ranges for the proportion of debt premia variation attributable to default is 34% to 73% (mid-point 53%);

page 8 • the Competition Commission assumed that the proportion attributable to debt beta was 28% to 60% • using historic data it can be derived that in the case of BT 38% to 60% (mid- point 49%) of the variation in debt premia was attributable to debt beta • direct estimates of debt beta, though unreliable and non-robust and not supportive of equal apportionment are, nonetheless, broadly compatible with it

31. More detail on this is provided the EE Report AnnexAppendix 1:Apportioning debt premium .changes between debt beta and risk of default.

32. Using this assumption the BTG debt beta rises from 0.125 in 2007 to 0.3 (two year debt beta) to 0.4 (one year debt beta) in late 2009 and falls back to 0.20 (two year) to 0.25 (one year) in Feb 2011 (see EE Report Figs 5.3, 5.4). Using this varying debt beta in the calculation of the asset beta produces a narrower and more stable range for the BTG asset beta. The two year BTG asset beta range is from 0.55 to 0.64 (from Apr 09 to Mar 11). This compares to the much wider 0.43 to 0.65 data range that Ofcom derived (from which Ofcom assumed a range of 0.45 to 0.60).

33. We consider that this approach for deriving asset beta is more robust not only because it uses the more plausible assumption that the debt beta varies with changing debt premia but also because the asset beta range is narrower (‘width’ of 0.09 versus 0.22). One would expect that even through the financial crisis the asset beta would not be highly volatile.

34. Re-levering this asset beta provides a BTG equity beta range of 0.90 to 1.03 using a 50% gearing and debt beta of 0.20 to 0.25. This is higher (at the mid-point) and narrower than Ofcom’s estimated range of 0.78 to 1.08.

OPENREACH ASSET BETA / OPENREACH EQUITY BETA

35. Ofcom estimates the Openreach asset beta at 0.40 to 0.55. Ofcom considers that the asset beta of Openreach is between that of BT Group and network utilities.

36. The BT Group asset beta is derived using empirical data for BT Group as described above (Ofcom assumes 0.45 to 0.60). The network utility data is based on analysis by Brattle of the asset beta of several network utilities (Ofcom assumes 0.30 to 0.40). The exact method of how these two data sets are combined to derive the Openreach asset beta is not fully specified. Ofcom seem to base the bottom end of the Openreach asset beta range on the top end of network utility benchmarks (rather than the Openreach asset beta being derived by deducting a ‘wedge’ from the BT Group asset beta – see §6.140). Indeed Ofcom support the view that the calculation of the differential or ‘wedge’ as between the BTG asset beta and the Openreach asset beta is ‘not an exact science’ (see Consultation §6.136, LLU 2009 Statement §A8.70). The top end of the range is effectively derived by assuming that the width of the Openreach asset beta range is the same as the BTG asset beta range.

page 9 37. We consider the approach of basing the bottom end of the Openreach asset beta range on network utilities as being sound since it reflects the utility nature of Openreach and because there is no robust basis for directly estimating the size of the wedge to be able to derive the Openreach asset beta from the BTG asset beta.

38. Given the approach of basing the bottom end of the range on network utilities this raises two subsequent questions. First what are the asset betas of benchmark network utilities and second where should the (bottom end of the) Openreach asset beta range be set relative to the network utilities range.

39. Brattle have estimated a range of network utilities from 0.29 to 0.36 (using a debt beta of 0.15). On the basis of this and rounding up Ofcom assumed the network utility asset beta range as 0.30 to 0.40. We think this is incorrect for three reasons: • First it seems rather ‘coarse’ to round up to the nearest 0.1 rather than 0.05 (or 0.01) • Second, Brattle used a debt beta of 0.15 in calculating the asset betas rather than Ofcom’s preferred number of 0.125. Using a debt beta of 0.125 gives an asset beta range of 0.25 to 0.35 (see EE Report §7.13). • Third, since the average of the five benchmarks is 0.31, the Ofcom assumed range of 0.30 to 0.40 is skewed

40. On the basis of these corrections we think that the appropriate range for network utility benchmarks is 0.25 to 0.35.

41. The second question is how close to this data should the bottom of the range for Openreach asset beta be. Ofcom assumed that the bottom of the Openreach asset beta range (0.40) should be equal to its estimate of the top of the network utility benchmark range (0.40) (Consultation §6.140). We view equating the bottom end of the Openreach asset beta range with the top end of the network utilities asset beta range as possibly a little aggressive though we do not believe there should be a large difference. We explain our reasons below.

42. A simple assessment of the copper access services business of Openreach shows that against the key areas of operating / asset risk Openreach (i.e. LLU / WLR) is much more akin to that of network utilities (e.g. water and electricity) than it is to the rest of BT (which includes BT Retail, BT Global Services, BT Wholesale and the NGA network investment). • whilst LLU/WLR have a higher risk level than water/electricity it is much lower than for RoBT which includes a large amount (45%) of non- traditional services9 (such as NGA, TV, ICT, managed networks) for which demand is highly uncertain and competition unpredictable as well as traditional services which are also more risky volatile than LLU/WLR. This much lower

9 About 55% of rest of BT revenue is ‘traditional’ i.e. lines, calls, leased lines and the remaining 45% is non-traditional (e.g. ICT, managed networks, broadband, convergence, TV, NGA). Derived from BT Annual Report 2011 p108 and others. The revenue is net of the cost of LLU/WLR inputs into these products

page 10 volatility is demonstrated clearly by Consultation Table 6.10 which shows the average forecast error for copper lines at 1.5% versus 4% to 20% for leased lines and IPStream (which are some of the least volatile elements of the RoBT business) • on prices, like water/electricity, LLU/WLR prices are regulated and are reset every few years to align prices with forecast costs. This substantially reduces price and margin volatility. Conversely for RoBT (even for most traditional services) prices are not regulated and prices are set by the market which means they are volatile and unpredictable (and so returns are volatile) • on costs the LLU/WLR business is reasonably mature and operating costs are predictable. Through the use of CCA FAC FCM methodology historic investments are fully recovered (whatever current price levels) and new investments are based on current prices. This model is similar to that used in water / electricity. Conversely, for RoBT costs are more unpredictable (even unknown in some cases) and there is no mechanism to ensure recovery of historic investments • on asset stranding the LLU/WLR business has a very low risk of asset stranding given very slow pace of technology change, innovation and volume change. Further, even to the degree that it happens the cost of stranded assets are recovered given the asset valuation approach. This is very similar to water / electricity. For RoBT, particularly for new non-traditional services the level of innovation and so asset stranding risk is higher and there is no mechanism to ensure recovery

43. A more detailed assessment in provided in the table below. It is important to recognise that risk depends not on whether revenue / costs change but rather whether change in them is predictable or not.

page 11 Network utilities (water / Openreach (LLU and WLR) RoBT (BTR, BTGS, BTW, NGA) electricity)

• Near universal adoption of • Demand for lines has low cyclicality though • Traditional services (line rental, calls) Overall demand connections some substitution by mobile (but reasonably reasonably predictable overall volume (market size / • Demand has limited cyclicality predictable) (though more volatile than LLU/WLR) but market share) competition creates significant volatility • Very little / no competition • Competition for lines from cable but Openreach market share stable and • Increasing proportion of non-traditional predictable services (e.g. broadband, corporate networks, NGA broadband, TV / content services, non-UK) which exhibit both

unpredictable market size and a high level of CONCLUSION: OPENREACH MUCH MORE substitution / competition which results in SIMILAR TO UTILITIES THAN RoBT high volatility

• Prices regulated and rebased to costs • Prices regulated and rebased to costs (over • Very few services where BT has SMP and are Price every 4-5 years so that outturn charge control period) every 3-4 years so that price regulated. Thus most prices and departures from assumed cost of outturn departures from assumed cost of margins subject to competition and so are capital will be small capital will be small unpredictable and volatile (particularly for CONCLUSION: OPENREACH SIMILAR RISK TO non-traditional services) UTILITIES

• Mature business with stable • Mature business with stable operational • Though for traditional services costs Costs (operating / operational practices and cost practices reasonably predictable, competition means capex) structures result in little volatility • CAPEX mainly replacement. CCA FAC FCM that cost recovery not ensured • CAPEX mainly replacement though in regulation ensures full recovery of previous • For non-traditional and new services costs water, major investment programme investments. New investment asset value not reliably predictable and significant new in improved quality and services adjusted to reflect current prices (e.g. copper) investment required e.g. NGA, NGN • new investments included in RAB CONCLUSION: OPENREACH ARGUABLY LESS RISKY THAN WATER

• All assets reflected in the RAB and • All assets reflected in the RAB and prices and • No regulatory price protection from Asset stranding prices and therefore operator therefore BT protected from stranding risk stranding and stranding risk high given protected from stranding risk • CONCLUSION: OPENREACH SIMILAR RISK TO innovative nature of investments UTILITIES

45. Given risk characteristics of Openreach, we consider that setting the bottom end of the Openreach asset beta range to equal 0.40 (i.e. 0.05 more than top end of utility benchmarks) is a conservative and defensible assumption – plausibly the bottom end could be lower than 0.40.

46. Regarding the top of the range, Ofcom have effectively assumed the same width for the Openreach asset beta range as for the BTG asset beta range (since they applied the used the wedge implied from the bottom end to set the top end, see consultation §6.140). We consider that using the same spread for the Openreach asset beta range as for the BTG asset beta range (0.10 under our estimates) is a sensible approach. This provides our estimated range for the Openreach asset beta as 0.40 to 0.50.

47. Another reason as to why this range is reasonable is where it places Openreach asset beta in the range between the asset beta of network utilities and the asset beta of rest of BT Group. As we showed above, Openreach’s risk characteristics are much closer to those of a network utility than to those of the rest of BT and so one would expect the asset beta of Openreach to lie closer to that of network utilities than the rest of BT Group. • Under Ofcom’s assumptions (mid range) the Openreach asset beta (0.48) is in fact closer to that of rest of BT Group10 (0.58) than network utilities (0.35) – this is implausible and not credible • In fact if the error in the network utility asset beta range is corrected (to 0.25 to 0.35) then Ofcom’s positioning of Openreach (0.48) is even more implausible since it is even closer to rest of BT Group (0.58) than network utilities (0.30)

48. Under our proposed assumptions Openreach’s mid range asset beta (0.45) is a little closer11 to that of network utilities than that of BT Group which is a more plausible and reasonable assumption (though still conservative).

49. Re-levering this asset beta range (using 50% gearing and a debt beta of 0.20 to 0.25) provides an Openreach equity beta range of 0.60 to 0.75.

50. These assumptions imply a ‘wedge’ between the BTG equity beta and the Openreach equity beta of about 0.30. This is above Ofcom’s implied wedge of 0.10. We do not consider the higher wedge as in any way implausible or any cause for concern. • First, the wedge is similar to what it was in 2005. If a 30-35% gearing is used (as in 2005) then the equity beta wedge under our assumptions would be 0.23. This is similar to the equity beta wedge of 0.20 that Ofcom implicitly assumed in 2005

10 The rest BT asset beta is derived from the BTG asset beta and Openreach asset beta assuming (as Ofcom has) that Openreach accounts for 50% of BT Group. This means that the BTG asset beta equals the average of the Openreach asset beta and the rest of BT asset beta. 11 at the mid-point the Openreach asset beta (0.45) is 34% of the way between the network utility asset beta (0.30) and the rest of BTG asset beta (0.74) • Second, as Ofcom accepts itself there is ‘no science’ about the wedge – that the implied wedge halved between 2005 and 2009 (from 0.2 to 0.1) with no underlying reason to expect such a movement reinforces the concept that there is no science at play • Third, there is good reason to consider that the ‘wedge’ might have increased since rest of BT now includes the high risk NGA activity

51. The wedge should be envisioned as merely an output of the calculation and not an input. The important consideration is where the Openreach asset beta is positioned relative to the asset beta of network utilities and the asset beta of rest of BT – the evidence indicates that it should be closer to that of network utilities than that of rest of BT.

52. A further consideration that Ofcom should bear in mind is that in the case that the BT Group asset beta increases (for instance due to changes in the underlying dataset used to derive asset betas) there is no case for increasing the Openreach asset beta range since it is (and should be) in effect anchored to the network utility benchmark range (and not be derived from the BT Group asset beta less the wedge).

IMPACT OF BT PENSION SCHEME

53. The existence of the BT pension scheme increases the observed cost of capital of BT Group due to the higher risk of the BT defined benefits pension scheme relative to the risk of the other parts of BT Group. This concept has a wide range of support: • There is a wide body of academic support both for the concept and also that the impact is material – most notably research by Jin, Merton and Bodie • Ofcom accepts that there is a possible impact (though claims that the evidence that underpins it not robust) (Pension Statement §7.11) • Ofcom’s advisor on its pension statement (Professor Cooper) agrees that there is an effect (though argues that its size is uncertain)12 • The CC accepted there is an effect: for example, “… whilst it is fairly straightforward to see that the BT Group’s credit quality might have been weakened by its pension scheme …”13

54. The effect of the pension scheme risk can be characterised as raising the cost of capital and so increasing the required return on capital employed. This raises the question of whether this additional cost is a relevant cost to include in the ‘cost stack’ used to derive wholesale charges (such as LLU and IPStream). We refer to this cost as the ‘pension risk cost’.

12 His reports implicitly accept that there is an effect though the reports focus on how measurable the impact is 13 LLU Determination §2.374

page 14 55. Whether the cost of the pension scheme risk impact should be included in wholesale costs can be assessed by reference to the six principles of cost recovery. Broadly most of the key principles are consistent with economic efficiency considerations (productive, allocative and dynamic). These principles (originally conceived by the Mergers and Monopolies Commission) are widely used by Ofcom to assess whether it is appropriate to include certain costs in the costs estimates used to derive wholesale charges. Indeed, Ofcom used this approach to assess whether pension deficit contributions should be included in wholesale costs. However, (and notably) Ofcom did not opt to use this framework to assess whether to include the cost of the pension scheme risk in wholesale costs – Ofcom did not explain why it departed from its usual practice.

56. Below we summarise our assessment of whether to include the additional cost caused by the pension scheme risk against the six cost recovery principles. This assessment shows clearly that it is appropriate to exclude the pension risk cost.

57. Against all the main principles it is clear that the pension risk cost should be excluded since: • The pension risk cost is not caused by provision of wholesale services • If the pension risk cost is included in wholesale charges then it will reduce cost minimisation incentives • The beneficiaries of the causes of the pension risk cost are shareholders so it is appropriate of shareholders to bear the pension risk cost • Inclusion of the pension risk cost will raise prices above efficient level and so distort competition

page 15 58. On only one principle (practicality) is it preferable to include the pension risk cost and, as Ofcom agrees14, this principle is of relevantly minor importance.

ASSESSMENT OF SIX COST RECOVERY PRINCIPLES FOR PENSION RISK COST

Principle Exclude / Reasons include ?

Cost Clear The pension risk cost is/has been caused by the existing causality exclude accumulated scheme asset / liabilities15. These are a result of past activities and not the on-going provision of wholesale services so there is no cost causality. An additional increment of production of a wholesale service does not cause any additional pension liabilities or pension risk and so cause an increase in wholesale costs. Furthermore, the pension risk cost is not an efficiently incurred cost – Ofcom have presented no evidence to suggest that an efficient operator would have a £40bn defined benefits scheme16

Cost Clear If this cost is passed through to consumer it unquestionably reduces minimisation exclude to some degree the incentive on BT to minimise the cost (for instance by reducing the pension asset risk)

Distribution Clear Shareholders are clear beneficiaries of this cost / the causes of this of benefits exclude cost – for example, increasing asset risk (which will increase scheme risk and cost of capital) will provide pension scheme higher returns (which only shareholders will benefit from given Ofcom’s approach on pension risk/surplus). Thus effectively the beneficiaries of a higher cost of capital are the shareholders

Effective Clear If pension risk cost is included it will result in prices being above competition exclude efficient forward-looking costs and will therefore distort competition in the wholesale and downstream markets

Reciprocity Neutral No clear conclusion from this

Practicality Minor Calculating the pension risk cost and so the adjustment that needs include to be made is complex (but not insurmountable)

59. Another approach to assessing whether to include the cost resulting from the pension scheme risk is to assess whether Ofcom’s approach is (internally) consistent

14 See Second Consultation §3.91 “We accept the CWU’s comment on the application of the practicability principle, and agree that it does not mean that the simplest approach must be taken. If there were a strong case, on the basis of the other principles for including pension deficit payments in wholesale charges then some practical difficulty would have to be accepted” 15 The main determinants of the size of the pension risk cost are scheme size (e.g. liabilities), beta of the liabilities and beta of the assets 16 including the pension risk cost is effectively concluding that the prices today should reflect promises made over last 50 years and that it is efficient practice to have a DBS that is (proportionally) one of the largest in the UK. This is absurd and clearly incompatible with a efficient forward looking approach for deriving cost given most companies in competitive markets do not operate defined benefits scheme and/or have closed them to new members and/or are closing them to new promises.

page 16 with the approach it took to the inclusion or not of pension deficit contribution in wholesale costs. Ofcom concluded (correctly) that the costs used to calculate wholesale charges should not include pension deficit contributions.

60. We think that a priori that it is incorrect and inconsistent to exclude one effect of the pension scheme (deficit contribution) from wholesale costs but include another (pension scheme risk cost). One can consider the pension scheme as a business unit within BT – under this model Ofcom’s approach is tantamount to including some of the costs of this business unit in wholesale costs (the pension scheme risk cost) but excluding others (pension deficit contribution). Ofcom’s approach is incoherent and inconsistent.

61. Ofcom’s reasoning for its position is vague and lacks cogency. There is one short comment in the Pension Statement. This appears to be the totality of Ofcom’s reasoning regarding the economic principle of whether to include the pension risk cost – this is aside of reliance on CC comments (see below) and issues of materiality and robust methodology which are not relevant considerations to the question of principle. The comment is: “We believe that it would be inconsistent for us to determine that the risks and rewards of the pension scheme sit with shareholders but not to allow the regulatory cost of capital to reflect these risks and rewards also.”17

62. This statement in and of itself is incoherent and inconsistent. It argues that because the risks/rewards (deficit / surplus) of the pension scheme sit with shareholders so the cost caused by the risk of the pension scheme should sit with wholesale customers / consumers. This makes no sense – if shareholders bear the risks/rewards then shareholders should also bear the costs of those risks (not wholesale customers and consumers). Ofcom’s position is effectively the same as saying that shareholders can borrow money and invest in risky assets but not pay the high cost of interest that is commensurate with high risk (and instead consumers pay the higher interest rate).

63. Ofcom also seek to rely on the CC Determination in the LLU Appeal. However, unhelpfully the CC’s position and reasoning for its position is also rather vague. Ofcom in particular draws on one passage from the Determination which says: “Shareholders are expected to pay for the cost of funding the deficit should a deficit arise and receive a cost of capital based on an unadjusted beta to reflect this risk: this is consistent with Ofcom’s past treatment. The cost of capital through the equity beta reflects investors’ perceptions of the risk to them of either deficit repair payments or pension scheme surpluses (and possibly associated payment holidays) at a given point in time, and customer charges reflect this”18

17 Pension Statement (Dec 2010) §7.29 18 LLU Determination §2.350

page 17 64. A close reading of this statement seems in fact to do little more than state the implication of Ofcom’s position. The first sentence merely says that the cost of capital reflects the risk of deficit and the second sentence says that consumer charges reflect risk.

65. Unsurprisingly, the strongest support that Ofcom draw from the CC’s comment is that “[t]his statement suggests to us that the CC supported Ofcom’s view”. Ofcom seem to accept that the CC comments do not strongly endorse Ofcom’s position.

66. In fact if one reads elsewhere in the CC’s Determination it becomes clear that the CC’s position is far from equivocal on this point. For example, CC say that ‘it is not clear’ that the cost impact should be excluded “… it is not clear to us that Ofcom should have adjusted the Openreach beta and gearing ratio as a matter of consistency [with deficit approach]”19. Indeed much of the CC’s reasons not to accept CPW’s argument that the cost of capital impact should be excluded was on the basis that CPW had not proven that the cost was an inefficient cost (see LLU Determination §2.351).

67. Therefore we do not consider that the CC Determination provides cogent or clear support for Ofcom’s position.

68. In the WBA and LLU/WLR Charge Control consultations Ofcom provided no additional reasoning for its incoherent position regarding whether the impact of the pension scheme risk should be excluded or not. This was a little disappointing given the lack of cogency in its Pension Statement. We would encourage Ofcom to provide cogent reasoning for it position in the upcoming charge control statements (or change its position to be consistent with the evidence).

69. Since Ofcom has presented no cogent reasoning to support its position (or reject TTG’s view) we remain firmly of the view that the cost of capital increase caused by the pension scheme risk should be excluded when assessing the cost of capital to apply in setting Openreach wholesale charges. We accept that there are difficulties in making precise calculations of the impact.

70. The pension scheme risk will have a number of effects on the BT Group WACC calculation. Most particularly the pension scheme will: • Increase the equity / asset beta. Ofcom broadly accepts this in principle (see Pension Statement §§7.11.1, 7.11.2) • Increase the cost of debt. The CC have accepted this: for example “This additional risk [from pension scheme] is likely to cause an increase in the debt premium …” (LLU Determination §2.347) • Reduce the possible gearing

71. Given, in BT’s case, the relatively large size of its defined benefits scheme and the relatively risky assets it holds the impact on the cost of capital of the scheme risk is

19 LLU Determination §2.348

page 18 likely to be large. PWC estimated that the pension scheme risk increases BT Group’s asset beta by 0.10 i.e. the asset beta of the non-pension activities / assets would be 0.10 lower.

72. In principle, all these impacts of the pension scheme need to be excluded from the Openreach WACC. As we mentioned above doing this is difficult due to lack of strong and reliable evidence to quantify the impact. However, this difficulty must not be used as a reason to make no adjustment whatsoever or ignore the impact particularly since the impact is so large.

73. We suggest the following approach as a pragmatic way to reflect the need to exclude the pension scheme risk in the Openreach WACC.

74. In respect of the cost of equity there are two adjustments that should be made: • the network utility benchmarks should be adjusted to exclude the effects of their pension schemes on their asset / equity beta to provide a ‘clean’ asset/equity beta. Given the lower relative size of pension schemes in network utilities the impact on asset beta will be less than that for BT Group (which PWC estimated at 0.10) • the BT Group asset beta should be adjusted downwards. In the response to the second pension consultation PWC (commissioned by TTG / Sky) provided an estimate of the impact of the pension scheme risk. This suggested that conservatively the BT Group asset beta would be 0.10 less if the pension scheme risk was excluded.

75. These adjustments could be reflected by using a lower range for Openreach asset / equity betas and/or by choosing a point lower in the range.

76. Further, we note that the high level of ‘contamination’ in the BTG asset / equity beta figures increases the weight that should be given to the network utility benchmarks (which are less contaminated).

77. In respect of adjusting the cost of debt Ofcom should look to use the lower end of debt premia range rather than the middle or top.

78. In respect of the gearing we accept that it is difficult to identify the adjustment to reflect the exclusion of the impact of the pension scheme. Rather, the need to exclude the scheme should provide further comfort that the 50% gearing is a reasonable assumption.

page 19 OVERALL RESULT AND CONCLUSION

79. The overall WACC resulting from the assumptions is shown in the table below (which includes both Ofcom’s proposed assumptions and TTG’s). This provides a pre-tax nominal WACC for Openreach of 7.4% to 8.3% (compared to Ofcom’s estimate of 8.0% to 9.2%).

TalkTalk Ofcom Openreach BT Group Rest of BT Openreach BT Group Rest of BT

Risk-free rate 1.4% 1.4% 1.4% 1.5% 1.5% 1.5%

Asset beta 0.40 - 0.50 0.55 - 0.64 0.70 - 0.78 0.40 - 0.55 0.45 - 0.60 0.50 - 0.65

Debt premia 2.0% - 2.5% 2.0% - 2.5% 2.0% - 2.5% 2.0% - 2.5% 2.0% - 2.5% 2.0% - 2.5% Debt Beta 0.20 - 0.25 0.20 - 0.25 0.20 - 0.25 0.125 0.125 0.125 Gearing 50% 50% 50% 50% 50% 50%

Equity Beta 0.60 - 0.75 0.90 - 1.03 1.20 - 1.31 0.68 - 0.98 0.78 - 1.08 0.88 – 1.18

ERP 4.5% - 5.0% 4.5% - 5.0% 4.5% - 5.0% 5.0% 5.0% 5.0%

Inflation 2.5% 2.5% 2.5% 2.5% 2.5% 2.5% Tax rate 25% 25% 25% 25% 25% 25%

WACC low 7.4% 8.3% 9.2% 8.0% 8.3% 8.5% mid 7.9% 8.8% 9.7% 8.6% 8.9% 9.3% high 8.3% 9.3% 10.2% 9.2% 9.5% 10.0%

80. Two general observations on the results: • One of the elements that we consider to be particularly implausible within Ofcom’s assumptions is the small difference in WACC as between Openreach and rest of BT. These divisions have extremely different risk characteristics yet under Ofcom’s assumptions the difference in WACC in only between 0.5% and 0.8% (at low end and high end of range respectively). • Though the WACCs are lower than in 2009 this should not be seen as a cause for concern – most of the change in the WACC is a natural consequence of improvements in macro-economic conditions

81. There is a separate question of where within the range Ofcom should select the WACC to apply to copper access services.

82. Unlike some other regulators and the Competition Commission Ofcom does not explicitly opt to choose a WACC high in its range — an approach sometimes referred to as ‘aiming up’ – though Ofcom does tend to err on the high side on ERP (see Consultation §6.106).

page 20 83. The standard argument for ‘aiming up’ is that there is an asymmetry of consequences: if the WACC is a little too high, then consumers lose out a little, by paying higher prices; but if the WACC is a little too low, then consumers lose out more since the regulated company does not invest in new technologies and equipment (and this reduced investment is considered to be larger than the losses from excessive prices). However, we do not consider that this argument is sound or relevant in the case of setting Openreach’s WACC for a number of reasons.

84. First, the impact of raised wholesale prices is not just raised retail prices but also reduced downstream investment and less effective competition. This occurs because though retail prices will in response to a rise in wholesale prices they will not rise by as much (thus resulting in a reduced margin). This is due to price elasticity and since they are set by BT who do not face the wholesale charges20.

85. We consider than a £1 of lost downstream investment by ISPs has a more detrimental impact that a £1 of lost upstream investment by Openreach since more innovation, competition and dynamic economic benefits result from investment at the ISP layer than at the Openreach layer • There is a large amount of innovation that occurs at the downstream layers such as unbundling more exchanges, new xDSL technologies, dynamic line management, traffic management, network-based parental control software, new bundling and pricing packages, novel customer service approaches • In contrast there is very little innovation in the LLU layer since the technology (copper lines) has essentially been the same for decades • In terms of investment there is greater investment occurring in the downstream layers. Openreach invest about £20 per line per year – in contrast LLU operators (such as TalkTalk) invest around £30 per line per year21

86. Raised wholesale prices will also weaken downstream competition (since non-BT players will experience a margin squeeze). Competition at the LLUO/ISP layer is the major driver of consumer benefits due to the innovation and choice that it delivers.

87. Second, there are other reasons as to why BT will invest such as its universal service obligation (USO). This will mean that reductions in allowed returns will in effect have minimal effect on investment levels.

88. Third, (and unlike many other network utilities) the level of investment in new network for copper access services is small – about £0.5bn versus a (net) asset base of £8bn. This reflects the long asset lives and highly depreciated nature of the asset.

20 BT’s retail prices are set with reference to BT’s actual costs (including RoCE). If for any given level of BT’s costs the wholesale prices rise then BT’s retail prices will not rise (since their costs will not have risen). Though equivalence of input notionally requires BT’s downstream operations to ‘pay’ the wholesale price this is not a real charge. 21 Openreach invest about £450m in 23m lines (Consultation Fig 7.11). In contrast TTG invest about £110m of CAPEX of which probably around £80m relates to its 2.5m MPF customers which equates to £30 per line

page 21 89. Forth, because the capital employed is based on CCA FAC costs rather than LRIC costs. Since LRIC costs are generally lower than CCA FAC the actual return on incremental CAPEX will be higher that the calculated WACC. In other words even if the allowed return (on CCA FAC assets) is less than the actual cost of capital the actual return on incremental investment will be greater than the actual cost of capital.

90. Fifth, aiming up will create incentives for excessive and inefficient investment.

91. We observe, also, that discussions of the argument for ‘aiming up’ have moved on from simply arbitrarily choosing a value at the top of the range to discussions of how much to aim up. For example, in Europe Economics' 2009 report for Ofwat22 they argued that one standard deviation was sufficient aiming up, even in the context of the serious financial crisis and great uncertainty at that time, and that the two- standard-deviation aiming up proposed by the Competition Commission in the London Airports judgement was excessively cautious (since it effectively set the probability of under-forecasting the cost of capital at 2.5%). Arguably now that the financial crisis is ending even less aiming up than one standard deviation is appropriate (for those regulators that choose to aim up).

92. Rather than an explicit ‘aiming up’ approach, Ofcom’s preferred use of ranges seems to be to select values for individual components so as to come to a judgement about the appropriate overall WACC – we concur with this approach. In considering where to select values within the range we consider that there are strong reasons to select values towards the bottom of the range since certain of the Ofcom and TTG ranges are ‘skewed’ or biased in the sense that the most likely appropriate value lies towards the bottom of the range. We see a number of areas where there is such bias: • The debt premia range is based on that for BT Group (since it uses BT Group debt spreads) and does not reflect that Openreach is likely to have a lower debt premia due to its lower specific risk. The CC / Ofcom accept that Openreach which has lower risk would attract a lower debt premia (see §17 above). • The nominal cost of debt (6.0% to 6.5%) is already well above the observed yield for BT bonds (4% to 6%) • The lower end of the Openreach asset beta (0.40) is already well above the high end of the corrected range for network utility benchmarks (0.35). Ofcom itself felt that the lower end of the Openreach asset beta range should equal the top end of the network utility asset beta range • Using Ofcom’s assumptions, selecting a figure at the top end of the ranges would mean that the Openreach asset beta is actually closer to that the rest of BT asset beta than that of network utilities. This is not credible, given that Openreach is more similar to the risk profile of network utilities the positioning

22 §§8.26-8.28 http://www.ofwat.gov.uk/pricereview/pr09phase3/rpt_com_20091126fdcoc.pdf

page 22 of the Openreach asset beta is implausible and the asset beta should be set at the lower end of the range • The ranges do not properly include the impact of stripping out the impact of the BT pension scheme risk which will reduce the cost of debt and cost of equity. In particular:

o the debt premia used in the Openreach WACC should be lower than for the BT Group

o the asset beta should be adjusted: at the low end by using network utility asset beta benchmarks which excludes pension risk impacts and at the high end being based on a BT Group asset beta that excludes pension risk impacts (which would reduce the asset beta by about 0.10)

93. The final decision that Ofcom makes on WACC later this year will also need to reflect developments between now and the time of the final statement such as revisions in the underlying datasets. We offer the following observations for how conditions in the coming months might affect suitable assumptions: • If gilts market yields stay low, 1.4% would remain a reasonable value for the risk-free rate • If gilt yields and the risk free rate rise then it may be appropriate to reduce the debt premia range (and/or select a figure towards the bottom of the range) since debt premia and risk free rates tend to contemporaneously move in opposite directions23 • If equity markets normalise and gilt market yields rise in reflection of more optimistic expectations for the sustainable growth rate of the economy, values less than 5.0% should be considered for the ERP (since 5% is considered appropriate for times of market volatility). Only if a value of 1.4% is chosen for the risk-free rate would values towards the upper end of the 4.5% to 5.0% ERP range be plausible. • If market debt premia continue at around 2.0%, a figure right at the bottom of the 2% to 2.5% range should be considered plausible. • Using the equal apportionment approach, the 1 year/2 year BTG asset beta has been pretty stable at around 0.60 for some time, meaning that, unless something significant and explicable happens to equity betas between now and the final determination, a figure towards the middle of our proposed BTG asset beta range (0.55 to 0.64) would be most defensible. This would imply a BTG equity beta of about 0.95-1.0 at 50% gearing and debt beta of 0.20 to 0.25. • If the BTG asset / equity beta rises this should not result in a raised lower end for the Openreach asset / equity beta since the lower end of the range should be anchored against network utility benchmarks and not be derived from the BTG asset / equity beta

23 See LLU Appeal Determination §§2.392 et seq

page 23 94. The recommendations and suggestions we provide are linked. For example, the higher BTG asset beta is based on a presumption on the link between of debt beta and debt premia. If the higher BTG asset beta is to be used then this same link should be used in calculating the equity beta. In other words, the proposals cannot be ‘picked and mixed’.

page 24