AUSTRALIAN COMPETITION TRIBUNAL
Application by ActewAGL Distribution [2010] ACompT 4
IN THE AUSTRALIAN COMPETITION TRIBUNAL | |
ACT 1 of 2010 |
RE: | APPLICATION UNDER SECTION 245 OF THE NATIONAL GAS LAW FOR A REVIEW OF AN ACCESS ARRANGEMENT DECISION MADE BY THE AUSTRALIAN ENERGY REGULATOR (AER), PURSUANT TO RULE 64 OF THE NATIONAL GAS RULES |
BY: | ACTEWAGL DISTRIBUTION (ABN 76 670 568 688), A PARTNERSHIP BETWEEN ACTEW DISTRIBUTION LIMITED (ABN 83 073 025 224) AND JEMENA NETWORKS (ACT) PTY LTD (ABN 24 008 552 663) (ACTEWAGL) Applicant |
TRIBUNAL: | JUSTICE FINKELSTEIN (PRESIDENT) MR R F SHOGREN DR J S MARSDEN |
DATE: | 17 SEPTEMBER 2010 |
PLACE: | MELBOURNE |
REASONS FOR DETERMINATION
Background
1 The applicant, ActewAGL, is a partnership which owns and operates a gas distribution network through which natural gas is delivered to consumers in the Australian Capital Territory and the Queanbeyan and Palerang districts. The distribution network is covered by the National Gas Law (Natioinal Gas (ACT) Act 2008 (ACT), s 8(1)) and the National Gas Rules made under the Law. The object of the Law is “to promote efficient investment in, and efficient operation and use of, natural gas services for the long term interests of consumers of natural gas”: s 23.
2 The Law required ActewAGL to submit to the Australian Energy Regulator (AER) for its approval under the Rules an access arrangement which set out the terms and conditions upon which third parties could have access to ActewAGL’s gas distribution network: Law, s 132 and Rules, r 52. The proposed access arrangement is required, among other things, to specify the proposed service to which the access arrangement relates and the “reference tariff” for that service: r 48. The reference tariff is the tariff or charge to the third party for access to, or for the provision of, the relevant service: s 2, r 48.
3 One of the components of the reference tariff is the rate of return on capital. ActewAGL was required to submit with its proposed access arrangement certain information, including information specifying the proposed rate of return on its capital, the assumptions upon which the rate is calculated and a demonstration of how it is calculated: r 72(1)(g).
4 Rule 87 deals with the rate of return and specifies the factors to be taken into account in its determination. The rule provides:
(1) The rate of return on capital is to be commensurate with prevailing conditions in the market for funds and the risks involved in providing reference services.
(2) In determining a rate of return on capital:
(a) it will be assumed that the service provider:
(i) meets benchmark levels of efficiency; and
(ii) uses a financing structure that meets benchmark standards as to gearing and other financial parameters for a going concern and reflects in other respects best practice; and
(b) a well accepted approach that incorporates the cost of equity and debt, such as the Weighted Average Cost of Capital, is to be used; and a well accepted financial model, such as the Capital Asset Pricing Model, is to be used.
Access arrangement
5 On 30 June 2009 ActewAGL submitted an access arrangement proposal and access arrangement information to the AER, as required by rr 43(1) and 52(1). The access arrangement information set out how ActewAGL proposed to calculate its reference tariff. That required ActewAGL to calculate its weighted average cost of capital (WACC). One element of the WACC formula is the nominal return on debt. Two elements of the nominal return on debt are the nominal risk free rate of return and the debt risk premium. It is necessary to explain each concept.
The nominal risk free rate
6 The nominal risk free rate of return is the rate of return an investor expects to receive from an asset with guaranteed (ie risk free) payments.
7 The AER uses the annualised yield on Commonwealth Government securities (ie bonds) with a maturity of 10 years as a proxy for the risk free rate. The regulator considers 10 year bonds to most closely reflect the financing strategy of private energy networks.
8 The 10 year Commonwealth bond rate had consistently been applied by regulators in the Australian energy sector prior to 2003. The appropriateness of the 10 year bond rate was affirmed by the Tribunal in Re Gasnet Australia (Operations) Pty Ltd [2003] ACompT 6. The dispute in that case was whether the risk free rate should be set by reference to a 10 year bond or a bond that corresponded to the access period (which in that case was five years). The Tribunal (at [48]) stated that the use of a 10 year bond to determine a rate of return on assets with a life of approximately 30 years was “conventional” and “a correct use of the [Capital Asset Pricing Model]”.
The debt risk premium
9 The debt risk premium (which is also referred to as the debt margin) is the margin over the risk free rate that debt investors in a benchmark efficient service provider are likely to demand in order to be prepared to fund the business’ operations.
10 There are various ways to estimate the debt risk premium. Estimates based on historical averages are one of the most common proxies for the debt risk premium. Surveying market participants is another method and has the advantage of better reflecting prevailing market conditions. The debt risk premium can also be estimated based on the yield (ie return) on corporate bonds, which is the method commonly adopted by Australian regulators.
11 Returning to ActewAGL’s access arrangement proposal, so far as the debt risk premium is concerned, ActewAGL proposed it to be estimated by the average of the Bloomberg and CBASpectrum estimates of the yield on corporate bonds. ActewAGL noted that “[w]hile preferring CBASpectrum ... ActewAGL Distribution believes the most reasonable method of estimating the debt premium is to use the average.”
12 On 11 November 2009 the AER released a draft decision concerning ActewAGL’s access arrangement proposal, as required by r 59(1). In the draft decision the AER stated that it should make a choice between Bloomberg’s or CBASpectrum’s estimate or an average of the two. It proposed to use CBASpectrum’s yield estimate. By r 41 the AER was required to accept ActewAGL’s proposal in its entirety or not at all. Because the AER’s proposed debt risk premium differed from ActewAGL’s, the AER did not accept ActewAGL’s access proposal.
13 On 6 January 2010 ActewAGL submitted a revised access arrangement, as it was entitled to do under r 60(1). In that revision ActewAGL reiterated its proposal that the debt risk premium should be estimated by the average of the two estimates.
14 On 30 March 2010 the AER issued its final decision on ActewAGL’s proposal as required by r 62. The AER considered refinements and augmentations to its methodology for choosing between Bloomberg and CBASpectrum or an average of the two before applying the methodology and selecting the CBASpectrum estimate. Consequently it refused to approve ActewAGL’s revised access arrangement proposal and instead published its own access arrangement proposal: r 64(1).
15 On 27 April 2010 the AER confirmed its final access arrangement, subject to the correction of certain identified errors and omissions: r 64(4).
The dispute
16 Put very simply, the controversy between the parties is this. Each of Bloomberg and CBASpectrum provide information from which the debt risk premium might be estimated. The AER’s chosen measure, based on the CBASpectrum measure, produced a debt risk premium of 3.35%. ActewAGL’s measure, an average of the two, produces a debt risk premium 53 base points higher than the AER’s estimate.
Estimating the yield on corporate bonds
17 Corporate bonds are generally traded over the counter in private party-to-party transactions. It is not mandatory to report publicly the price paid for a bond. Information regarding the estimated price of corporate bonds is obtained through various commercial information services such as Bloomberg and CBASpectrum and corporate bond traders such as UBS. The prices published by each firm often differ, in part because each firm may know about different transactions in a particular bond.
18 CBASpectrum and Bloomberg produce an estimate of the “fair value curve” for bonds of a specific credit rating. A fair value curve plots estimates of bond yields against terms to maturity. Fair value curves provide a summary of how bond yields vary with the bond’s term to maturity. Fair value curves are typically upward sloping, reflecting investors’ requirement for a higher return for tying up their money for longer and the increased risk of a bond provider defaulting at some point over the life of a longer term bond. The fair value curve for higher rated (ie less risky) bonds is expected to lie below the curve for lower rated bonds, indicating that a higher yield will be paid on a more risky bond at a given time to maturity.
19 To estimate the debt risk premium the AER uses the yield on a BBB+ rated corporate bond. Bonds are rated by various rating agencies such as Standard and Poor’s, Moody’s and Fitch. Ratings reflect each agency’s opinion of the relative probability of the corporation defaulting on bond repayments or defaulting generally. The AER only uses Standard and Poor’s rating system, which rates bonds (in order of less risk to most risk) as AAA, AA, A, BBB, BB, B, CC, CC, C or D. Plus or minus ratings are also made for particular credit ratings. Even within the same credit rating, bond yields differ because bonds do not have identical characteristics. Yields may also vary because of the delay in rating agencies re-rating bonds following new information becoming available. The BBB+ rating has been routinely used by Australian regulators and was affirmed following a five-yearly review of inputs to the WACC model in 2009. ActewAGL does not dispute the use of a BBB+ bond.
20 It is generally accepted practice for the debt risk premium to be measured across a time period consistent with the risk free rate. Hence, regulators adopt the 10 year term to maturity for both the risk free rate and the debt risk premium.
21 Bloomberg does not publish a fair value curve for BBB+ bonds with a term to maturity of up to 10 years. It does, however, publish a “BBB” fair value curve based on the yields of BBB+, BBB and BBB- rated bonds with terms to maturity up to seven years. Generally speaking, the yield on a BBB bond would be expected to be higher than on a BBB+ bond with the same term to maturity. Bloomberg’s fair value curve would thus be expected to contain an upward bias, all other things being equal. Both the parties agreed that it was possible to extrapolate Bloomberg’s curve to 10 years by adding to it the spread between Bloomberg’s AAA seven year and 10 year fair value curves.
22 CBASpectrum publishes a fair value curve for BBB+ 10 year bonds.
23 The importance of choosing the right estimate is driven by the divergence between the two curves. The divergence may be observed by examining Figure 5.4 of the AER’s final decision with the dates normalised to July of each year.
No doubt the divergence is a reflection of the different methodologies and data used to produce the respective estimates. Only limited information is known about the methodologies. Each involves exercises of judgment and discretion which are non-transparent. The differences in methodology can be observed by examining the fair value curves of both companies as at 12 March 2010:
The AER’s methodology
24 The AER decided to determine which of the CBASpectrum and Bloomberg yields, or an average of the two, gave what was the most accurate fair value estimate for a corporate bond with a BBB+ credit rating and a term to maturity of ten years. It stated in its draft decision that “evaluation of the output from each method against real world observations of yields (over a period) for a sample of actual bonds that reflect an efficient benchmark is the only impartial means of determining which method produces the best estimates.”
25 The method by which the AER carried out the task involved three steps:
Step 1 – The selection of a “population” of corporate bonds that reflected as closely as possible the characteristics of the bonds that would be issued by a benchmark service provider.
In performing Step 1, the AER applied the following criteria:
- it excluded all bonds that were not rated BBB+ by Standard and Poor’s;
- it excluded all bonds that did not have yield estimates available from all of UBS, CBASpectrum and Bloomberg;
- it excluded all floating rate bonds;
- it excluded all bonds that were not issued in Australia (even if the issuing company was Australian); and
- it excluded all bonds that were issued in Australia but were not issued by an Australian company.
Step 2 – The selection of a “sample” of bonds by considering whether any bonds should be excluded from the population on the basis that the yields for those bonds were not representative of their credit rating. The AER undertook both qualitative and quantitative analysis to determine this.
Step 3 – Comparing the observed yields of the sample bonds to the fair value curves of CBASpectrum, Bloomberg and an average of the two curves to determine which curve aligned most closely to the observed yields. This involved using the weighted sum of squared errors (WSSE). The fair value curve that produced the smallest WSSE was determined to have the best fit to the data.
26 The result from Step 1 produced a “population” of six BBB+ bonds maturing on dates in 2012, 2013 (two), 2014, 2015 and 2016. Their terms to maturity at the averaging period were between 2.5 years and 6.5 years. In Step 2, the AER excluded one bond (the Babcock and Brown Infrastructure (BBI) bond maturing in 2016), leaving a sample of five bonds with an average term to maturity of less than four years and a maximum term of less than six years. Finally, in Step 3, the AER concluded that the CBASpectrum curve was the best fit with the observed yields of the sample bonds.
The right to review
27 The Tribunal has power, by leave, to review decisions of the AER: s 245. The grounds of review are confined by s 246(1) to:
(a) the original decision maker making an error of fact in the decision maker’s findings of facts, and that error of fact being material to the making of the decision;
(b) the original decision maker making more than one error of fact in the decision maker’s findings of facts, and those errors of fact, in combination, being material to the making of the decision;
(c) the exercise of the original decision maker’s discretion being incorrect, having regard to all the circumstances; and
(d) the original decision maker’s decision being unreasonable, having regard to all the circumstances.
28 In reviewing a decision of the AER the Tribunal is confined to the material that was before the decision maker: s 261(1). In limited circumstances (ie if a ground of review is established) the Tribunal can have regard to new information if it would be of assistance to it: s 261(3).
29 If the Tribunal is of opinion that leave to review should be granted, it must make a determination in respect of the application for review: s 259(1). If a ground of review is made out the Tribunal may set aside or vary the decision under review: s 259(1)(a). It also has power to remit the matter to the AER to reconsider the matter in accordance with any direction or recommendation the Tribunal considers appropriate: s 259(2)(b).
30 During the hearing, there was some discussion about the nature of the review. The grounds seem simple enough: error of fact, incorrect exercise of discretion and unreasonableness. In reality, however, these concepts are not straightforward.
31 Take the meaning of “fact”. A glance at a dictionary shows its meaning to be something which is capable of being experienced or perceived and hence known to be true. On this basis a “fact” is something that actually exists independently of its acknowledgement in the mind of the perceiver.
32 In Australian Competition & Consumer Commission v Australian Competition Tribunal (2006) 152 FCR 33, 73-74 the Full Court gave the word “fact” a much wider meaning. It decided that, in a provision such as s 39 of the Gas Pipelines Access (South Australia) Act 1997 (SA), which is in the same terms as s 246 of the Law, “facts” include: (1) historical facts; (2) present facts; and (3) an opinion about the existence of a future fact or circumstance (if necessary, we would add a fourth category, namely negative facts). As regards meaning (3), the Tribunal relied upon the decision of McInerney J in Morley v National Insurance Co [1967] VR 566, 567 where the question was what constitutes a “fact” for the purpose of s 55 of the Evidence Act 1958 (Vic), which made admissible certain documents containing statements by deceased persons intending to establish facts about which they had personal knowledge. McInerney J said that “fact” should be given an expanded meaning. He said the word “fact” should include a statement of opinion by an expert.
33 This is a radical meaning to be given to the word “fact”. The generally accepted view is that an opinion is an inference which is drawn from facts. Yet, as Wigmore famously pointed out, there are many instances where it is difficult (if not impossible) to distinguish between “fact” and “opinion”. Take the statement: “He was driving on the left hand side of the road”. Ordinarily this would be regarded as a statement of “fact”. On the other hand a statement that: “He was driving carelessly” would usually be regarded as an expression of an opinion. The difference between the two statements, however, is between a more concrete and specific form of descriptive statement and a less specific and concrete form.
34 Describing the meaning of a discretionary decision is also a difficult matter. The description “discretionary” is often applied to several types of decision making processes. It is most commonly applied to decision making which involves essentially a weighing up of relevant facts. First the decision maker finds the facts. Then the decision maker undertakes a weighing up process which involves taking into account considerations that are found to be relevant, assessing the weight to be given to those considerations so assessed and determining what, as a result of that process, is the right result. Another approach is found in Norbis v Norbis (1985-1986) 161 CLR 513, 518. There Mason and Deane JJ described a discretionary decision as one which involves an assessment that calls for “value judgments in respect of which there is room for reasonable differences of opinion, no particular opinion being uniquely right.”
35 A test for what is an unreasonable decision in the context of limited merits review has been considered, albeit briefly, by the High Court. In East Australian Pipeline Pty Ltd v Australian Competition and Consumer Commission (2007) 233 CLR 229, [80], Gummow and Hayne JJ said that unreasonableness in legislation such as s 246 of the Law is not intended to include the concept of unreasonableness as applied in judicial review proceedings: ie what is often referred to as “Wednesbury unreasonableness”. It is, we think, neither possible nor necessary to give an exhaustive definition of what is an unreasonable decision. At one extreme a decision that is arbitrary or capricious will plainly be unreasonable. At the other extreme, it will not be sufficient merely to reach a different decision to the first instance decision maker; in many areas reasonable persons can perfectly reasonably come to opposite conclusions. But, as the High Court indicated in East Australian Pipeline (at [80]) the term unreasonable “provides the basis for inferring the presence of one or more of the well established grounds which render a decision ‘incorrect’”. In other words, if the decision maker fails to call to attention matters he/she is bound to consider or considers matters which are irrelevant, he/she will be acting unreasonably. Reference might also be made in this connection to the Tribunal’s comments on unreasonableness in Application by EnergyAustralia [2009] ACompT 8 at [63]-[64].
The grounds of review
36 ActewAGL contends that:
• the AER should have identified the fair value curve which best estimates the BBB+ 10 year corporate bond rate instead of asking whether it had selected a fair value curve which most closely aligned to the observed yields of the BBB+ corporate bond yields in its sample;
• by limiting the observed bonds to a sample only containing bonds with a term to maturity of between two and less than six years, the AER’s selected fair value curve could not reflect the yields of bonds with a maturity of 10 years if there were systematic differences present in either bond yields or fair value curves for bond terms greater than six years;
• the AER should have increased the number of bonds in the observed bond population to include some or all of the following:
• bonds which have observations available from at least one of Bloomberg, UBS and CBASpectrum;
• floating rate bonds which have had their yields converted to fixed rates using prevailing swap rates; and
• longer-dated bonds with different credit ratings;
• the AER misapplied statistical tests in excluding the BBI bond and had the BBI bond not been excluded, the AER's methodology would not have selected the CBASpectrum fair value curve to derive the debt risk premium; and
• the only reasonable conclusions open to the AER were that:
• it was not reasonable to identify the CBASpectrum data source as a better predictor of BBB+ corporate bond yields in its sample, or as a better source of data to estimate the benchmark debt risk premium;
• given the uncertainty associated with observing the current market yield for a bond and the lack of transparency in the derivation of the CBASpectrum and Bloomberg fair value curves, no reliable conclusion could be drawn from use of only one of the CBASpectrum and Bloomberg fair value estimates over a relatively short period; and
• since both CBASpectrum and Bloomberg are experienced and well-respected market operators, it was appropriate to consider both as separate sources of relevant evidence and to estimate the benchmark debt risk premium by using both sources of evidence, which is best done by taking an average from both sources.
37 The attack on the AER’s decision, while relying upon each of the available grounds of review, is best considered under the “unreasonableness” head. Perhaps some of the alleged errors (eg the application of the outlier tests) may be complaints about errors of fact, but most are not. It may be possible to characterise the final decision (or relevant aspects of the decision making process leading to the final decision) as being an exercise of a discretion, but even that is an awkward characterisation of those parts of the decision which are under challenge. Relevantly, what is put against the final decision (and the interim decisions that led to the final decision), would, if accepted, lead to the conclusion that the final decision is unreasonable in all the circumstances. For that reason, we think it unnecessary to be concerned with an analysis of whether the complaints assert errors of fact or incorrect exercises of discretion.
Selection of a fair value curve
38 ActewAGL says that the bonds selected by the AER do not provide a basis for comparison with the fair value curves because the number of bonds is too small and their maturities are too short to be sufficiently representative of the yield on 10-year bonds.
39 The Tribunal accepts this submission. In the Tribunal’s view, it is not reasonable to decide which of three non-linear curves best fits a set of data that consists of only five points, especially when those points cover little more than half of the range of the independent variable, namely the term to maturity. The AER is seeking to select a curve on the basis of how close the observed yields lie to the curves, closeness being measured by the weighted sum of squared differences. There is not sufficient information to conclude that because the shape and position of a curve up to six years provides a better fit, the same curve will provide a better estimate for greater terms to maturity. Moreover, from inspection of the curves (see above, Figure 5.4 of the AER’s final decision) it is evident that data for bonds with terms to maturity of less than 3.5 years could not possibly be used to differentiate between fair value curves estimating the yield on bonds with a term to maturity of even five years or more, as the curves only diverge after 3.5 years. Data for bonds with terms to maturity of 3.5 years or less are effectively irrelevant.
40 Of the bonds in the selected sample, only three have a term to maturity of more than 3.5 years, one with a term to maturity at the averaging period of less than four years, one with a term to maturity of 4.5 years and one with a term to maturity of just under 5.5 years.
41 Even if the BBI bond had not been excluded (ie the AER had applied its methodology to a sample of six bonds), the bond population would still be insufficient in terms of size and maturity to support the AER’s conclusion.
42 It is clear that the AER recognised the problem of basing its analysis on such a small sample of bonds. In its draft decision (at 245), the AER wrote that “[i]deally, the sample would also include BBB+ bonds with longer maturity dates but there are no such bonds currently available in the market.”
43 In a recent decision, Final Decision: New South Wales distribution determination 2009-10 to 2013-14 published 28 April 2009, the AER discussed the relative merits of Bloomberg and CBASpectrum, including their methodologies and data sources, and considered which curve had performed better in the past. The AER also used a methodology for selecting a fair value curve which is similar to the methodology used in this case. The AER observed (at 230) that “[b]ecause the longest term to maturity of the bonds considered ... was eight years the analysis does not indicate whether Bloomberg or CBASpectrum performed better at predicting the fair yield of BBB+ bonds with a 10-year maturity.” The Tribunal upheld the AER’s determination: Application by EnergyAustralia [2009] ACompT 8. There appears to have been no material change in the BBB+ bond market or in the AER’s methodology since 2009 that could have led the AER to change its view that short maturity bonds should not be used to choose a fair value curve to predict the yield on longer maturity bonds.
Increasing the bond population
44 ActewAGL suggested – and the AER considered – three ways to increase the bond population: to use bonds with observations available from one or two data sources only, to use floating rate bonds and to use bonds of a different credit rating.
Including bonds with observations from one or two sources
45 ActewAGL said that the AER should have included bonds which have observations available from at least one of Bloomberg, UBS and CBASpectrum. The AER rejected this proposal “to ensure consistency and completeness of the sample of bonds data” and remarked that “[t]he same degree of confidence cannot be given to bonds with less data available. It is also preferable to maintain a stable and consistent sample when testing the fair value curves as it allows for comparability between tests.”
46 Figure 5.5 in the final decision shows the effect of including bonds with an observation available from at least one data source. A revised version of that figure is reproduced below with the Bloomberg curve superimposed and the dates normalised to January of each year.
47 The Tribunal is of the view, in conformity with the view expressed in the January 2010 Competition Economists Group (CEG) report prepared for Country Energy, that “whether a bond has a yield estimate from all UBS, Bloomberg and CBASpectrum (as opposed to from two or one of these sources) does not make it unreliable or biased as a relevant source of information.” Further, the inclusion of data from less than three sources can easily be accommodated within the weighted sum of squared errors formula adopted by the AER. Prima facie the AER’s decision to exclude bonds with data from less than all three sources seems to be unreasonable.
48 Nonetheless, it is not at all obvious that including the additional bonds (there are nine in all, although bonds with observations available from two sources appear on the graph twice) would improve the selection of a fair value curve because eight of the nine bonds had a term to maturity of less than two years and one had a term to maturity of 6.5 years. The AER suggested that some bonds may, however, be outliers. The Tribunal finds no fault with the AER’s conclusion that, after having examined the additional data, “the extra data from yields on BBB+ bonds available from one or two data sources do not provide any additional information that can be used to draw a meaningful conclusion.”
Including floating rate bonds
49 ActewAGL also proposed expanding the population by including floating rate bonds which have had their yields converted to fixed rates using prevailing swap rates. Figure 5.6 in the final decision shows the effect of including BBB+ floating rate bonds with observations available from at least one data source provider. That figure, superimposed with the Bloomberg curve and the dates normalised to January of each year, is shown below.
50 The issuer of a bond (the borrower) makes payments, referred to as “coupons”, to the bond holder (the lender) at periodic intervals. If the bond is “fixed”, so is the coupon. If the bond is “floating”, its coupon varies with interest rates or some external measure eg the risk of default of the issuer, the coupon varies. This affects the “face value” (ie price) of the bond. At any point in time the yield to maturity on a bond is the discount rate which, when used to discount future expected cash flows (both coupons and face values) equates the value of these cash flows with the bond’s market price.
51 As the AER acknowledged in its final decision, there is an accepted mathematical basis for converting between fixed and floating rates, although it is not necessary to outline that basis here.
52 The AER considered and then rejected the proposal to include floating rate bonds in its population of bonds. It gave several reasons.
53 First, the AER said that floating rate bonds were not a perfect substitute for fixed rate bonds. The AER may be correct to say that the two are not perfect substitutes, but this is not determinative. The issue here is whether taking floating rate bonds into account could aid in determining the yield on 10-year BBB+ bonds. As a matter of theory, the yield on a floating rate bond will be an unbiased proxy for the yield on a fixed rate bond because of the law of arbitrage. Briefly summarised, the law of arbitrage says that if an investor has a choice between a fixed rate bond and a floating rate bond that are identical other than their yield, he/she could buy the floating rate bond and enter into a swap arrangement, which would give him/her a fixed income stream. Consequently, investors would choose to buy the bond with the higher yield until the yields equilibrate. This theory is supported by empirical evidence. For example, Figure 2 of CEG’s report shows the yield on 10 companies’ simultaneously issued floating rate and fixed rate bonds with the same maturity and same rating.
54 The AER’s second basis for excluding floating rate bonds was that “in producing their fair value curves Bloomberg and CBASpectrum aim to reflect the rates on fixed rate bonds, not floating or converted floating rate bonds. This means that neither of the fair value curves are necessarily going to closely align to observed yields on floating rate bonds” (at 46). ActewAGL disputes the factual premises. But, whether true or not, the Tribunal does not accept the AER’s second basis given that the yields do correspond closely, as is demonstrated in Figure 2 of CEG’s report.
55 Third, the AER said (at 47) that “including converted floating rate bonds in the sample for analysis will lead to each converted bond being given the same weight as each fixed rate bond in the analysis. The AER does not consider this to be appropriate given that converted floating rate bonds are not perfect substitutes for fixed rate bonds which comprise the sample of bonds that are used to estimate the benchmark debt risk premium.” In the Tribunal’s view this adds nothing to the AER’s first two points and, in any case, weights can be varied. Consequently, it was prima facie unreasonable for the AER not to include floating rate bonds in its population.
56 One considerable advantage of including floating rate bonds is that it would add to the bond population six bonds with terms to maturity between 6 and 11.5 years. Moreover, all six bonds lie above the CBASpectrum fair value curve and closer to the Bloomberg curve. That said, we assume that not all the floating bonds in the population would be included in the sample. The application of Step 2 would, we assume, result in at least some bonds being excluded.
57 As it turns out the AER did consider what the result might be if Step 2 was applied (perhaps necessarily somewhat hurriedly in completing its decision making process) and found qualitative reasons to exclude all six longer term floating rate bonds in the sample. ActewAGL did not make detailed submissions on the accuracy of the AER’s exclusion of the bonds. The Tribunal does not, therefore, have a sufficient basis upon which to assess whether or not these longer term to maturity floating rate bonds should have been excluded.
58 The Tribunal considers that, as a matter of principle, floating rate bonds ought to be taken into account and treated equivalently to fixed rate bonds. However, if it is the case that the longer term floating rate bonds would have been appropriately excluded, the inclusion of floating rate bonds would still not have provided the AER with a sufficiently representative sample (with respect to the number of observations and terms to maturity) to choose a fair value curve.
Including bonds with different credit ratings
59 ActewAGL’s third proposal was to increase the number of bonds in the population by using bonds with different credit ratings; bonds rated BBB and A- were put forward.
60 Figure 5.7 in the final decision shows the effect of including BBB and A- bonds in the population. Figure 5.7 is reproduced below with the Bloomberg curve superimposed and the dates normalised to January of each year.
61 The AER rejected this proposal on the basis that it would potentially give equal weight to bonds with higher and lower credit ratings than the benchmark of BBB+. We think this is too cursory a rejection of the relevance of differently rated bonds. It is one thing to hold that a differently rated bond should not be given equal weight. It is quite another to refuse to take it into account in any way.
62 Notwithstanding the AER’s rejection of the proposal to include BBB and A- bonds, the AER did consider what effect their inclusion in the population may have. The AER said that the observations showed no clear pattern. The Tribunal considers the AER’s analysis to be too superficial. In fact, the longer term A- bond yields were above the CBASpectrum curve, contrary to what would usually be expected. We also consider that the AER was wrong to conclude as it did (at 56) that “[g]iven that the observed yields do not reflect reasonable expectations it is difficult to compare the selected fair value curve to the observed yields.” The very fact that observed higher rated (A-) bond yields were higher than the CBASpectrum curve for lower rated (BBB+) bonds should have sent alarm signals calling for further analysis.
63 In the Tribunal’s view, if it were reasonable not to include A- and BBB bonds in the population (because they are not representative of BBB+ bonds), it was unreasonable for the AER not to consider whether useful information could be obtained from taking these bonds into account without including them in the population. That A- yields sat above BBB+ yields should have indicated to the AER that by use of its methodology it may not have selected the fair value curve most likely to provide the best estimate of the benchmark bond yield.
The application of step 2
64 Here the issue is whether the AER should have excluded the BBI bond from its sample.
65 The AER’s process for excluding bonds involves “inspect[ion] of graphs of yields of the sample of bonds over time to identify any obvious anomalies” – this should probably be “potential anomalies” – followed by testing of (potentially) anomalous bonds using statistical tests. It also involves considering information about the bond and its issuer to determine whether the bond is unrepresentative, eg perhaps it is wrongly rated by the rating agency.
66 ActewAGL disputes the validity of the use of certain statistical tests and whether they had been correctly applied by the AER.
67 Part of the difficulty with the AER’s application of statistical tests stems from its mischaracterisation of Step 2 as a procedure for selecting a sample from a population of BBB+ bonds. In fact, the AER was attempting to determine whether the BBI bond ought to be classified as a BBB+ bond at all (and hence whether it should be included in the population). Standard and Poor’s had already determined that the BBI bond was a BBB+ bond on the basis of its relative probability of default. Statistical outlier tests do not consider the risk of default of a bond, thus cannot be determinative of whether a bond is correctly classified. What is required to conclude that a bond is incorrectly classified is qualitative factors additional to those that would likely have been taken into account by Standard and Poor’s at the time of last reviewing a bond’s rating. It appears that this is what the AER did with respect to the BBI bond and floating rate bonds.
68 In addition, the Tribunal wishes to make two further comments about the statistical tests. First, the Tribunal is sceptical about any statistical testing for an outlier amongst a mere six candidates. With such a small number of observations, a finding that one or more bonds were outliers would be unsurprising, but ought to draw attention back to what, if anything, can be ascertained from statistical testing in such a small pool of data.
69 Second, if the AER is to undertake statistical testing in the future, it should reconsider its approach to data interpolation. For instance, the AER assumed that where no observations on yields are recorded, the missing yields are the same as the last recorded observations. This rule has the virtue of simplicity but results in a downward bias in the recorded yields when actual yields are rising and an upward bias when actual yields are falling. The Tribunal notes that there are several better options ranging from a linear interpolation between the bookend observations to a more sophisticated, albeit resource-intensive, investigation into whether the market as a whole moved up or down on a particular day/period or whether there were explanations specific to the individual bond.
70 Turning to the AER’s decision to exclude the BBI bond, the Tribunal notes that the AER had some basis to consider on qualitative grounds that the BBI bond was anomalous and that, although rated BBB+ at time of issue, its yield ought not to be taken into account in estimating the benchmark yield on BBB+ bonds. The reasons included the suspension from trading of BBI shares. If nothing else, the suspension from trading would have had a significant effect on the yield, even though the credit rating assigned by Standard and Poor’s remained unchanged.
71 In view of the foregoing, the Tribunal proposes not to comment further on the results of the application of the statistical tests, noting in passing that both sides agreed that there had been some errors in the application of the tests. However, the Tribunal reiterates its caution about the use of the statistical outlier tests in the future.
The appropriateness of a 10 year bond
72 There is another point worth noting about the AER’s methodology. It arises out of the difficulty in identifying a sufficient number of long term bonds to determine yield. The reason a 10 year bond was originally chosen was because, in the past, many firms favoured long term debt, albeit that it came at a higher cost, because it reduced refinancing or roll-over risks. The high rate was then hedged via interest rate swaps. That may no longer be the position. If not, the AER may need to be reconsider its approach in light of more current strategies of firms in the relevant regulated industry. Further, there seems to be little point in attempting to estimate the yield on a bond which is not commonly issued.
A summary
73 It might be useful to draw the threads together.
74 In a robust bond market, it would likely be possible for the AER to calculate the yield based on particular representative bonds issued in Australia in reasonably close proximity to the time of the AER’s determination.
75 In the absence of a deep market for corporate bonds, the AER will likely have to rely on published fair value curves to estimate benchmark debt financing costs.
76 If the fair value curves differ substantially, the AER will need to choose between them.
77 We have identified three ways the AER is able to distinguish between the competing curves, although this is not intended to be an exhaustive list:
(1) If there is sufficient available information, the AER could examine and compare the merits of the publishers’ methodologies and data sources, as it has in the past.
(2) The AER could determine which curve has performed better in the past. This approach may not, however, be appropriate if there has been a material change in the bond market or in the methodologies or data sources used by the publishers.
(3) The AER could, as it has done here, compare relevant observed yields against the published fair value curves and an average of these curves. This will require the AER to undertake the following process:
(a) assemble a representative population of observed yields of sufficient number and term to maturity. It is difficult for the Tribunal to provide any hard and fast rule for determining whether a population is “representative”. A representative population would contain many bonds after the point at which the curves diverge. It should contain bonds with a term to maturity close to 10 years. The AER should include floating rate bonds and/or bonds with observations available from one or two sources in the population unless there is good reason to exclude them. The inclusion of these bonds may raise questions which the AER will need to address in the future, such as the weighting that should be given to them;
(b) only exclude bonds where there are sufficient qualitative reasons to consider that they are not correctly classed as being part of the relevant population;
(c) once a representative set of bonds has been chosen and refined in this way, select the fair value curve that most closely corresponds to the relevant set;
(d) use any other available information, such as observed yields on other rated bonds, to check that the selected fair value curve remains likely to provide the best estimate.
If a representative set of bonds sufficient to determine a fair value curve cannot be ascertained, or if later checks throw doubt on the chosen fair value curve, then this method of distinguishing between the curves cannot be used.
78 If the AER cannot find a basis upon which to distinguish between the published curves, it is appropriate to average the yields provided by each curve, so long as the published curves are widely used and market respected.
79 Of course, we do not intend to discourage the AER from investigating other ways to estimate the debt risk premium.
Conclusion
80 Having found that the AER fell into reviewable error, it is necessary to decide whether to remit the matter or vary the AER’s decision. There is no utility in remitting the matter because, for reasons we have explained, the AER would be required to determine the relevant fair value estimate by taking the average of Bloomberg and CBASpectrum curves. This can be done by the Tribunal on the information which is before it.
81 It would, in the circumstances, be convenient for the parties to agree on the appropriate orders that should be made varying the AER’s decision.
I certify that the preceding eighty-one (81) numbered paragraphs are a true copy of the Reasons for Determination herein of the Tribunal. |
Associate: