The Final Decision (ROE)
142 The final decision comprises 483 pages together with a number of Appendices.
143 At pp 69 to 247 (pars 206 to 1,182), the ERA addressed the concept of Total Revenue for the purposes of rr 72 to 87A.
144 At pp 69 to 75 (pars 206 to 222), the ERA set out the detail provided in the owners' various proposals and the ERA's various responses concerning the Revenue Building Blocks.
145 At pp 73 to 75 (pars 219 to 222), the ERA said:
Considerations of the Authority
The Authority's Final Decision on DBP's Total Revenue requirement is documented in the following sections.
• Operating Expenditure;
• Opening Capital Base;
• Projected Capital Base;
• Rate of Return;
• Gamma;
• Depreciation;
• Taxation;
• Incentive Mechanism; and
• Allocation of Total Revenue between Reference Services and Other Services
As a result of the Authority's assessment of DBP's proposed total revenue building blocks as per rule 76 of the NGR, set out in detail below, the Authority has not approved DBP's proposed total revenue for the fourth access arrangement (AA4) period. The Authority's Final Decision for approved total revenue by building block in real and nominal dollars is set out in Table 7 and Table 8 respectively.
Figure 3 compares DBP's proposed revenue building blocks with the building blocks approved in the Authority's Final Decision.
Figure 4 compares DBP's proposed revenue building blocks with the building blocks approved in the Authority's Final Decision. The key changes relate to the inflationary gain and return on capital base (reflecting a reduction in forecast inflation and increase in the real rate of return) and taxation (reflecting an adjustment to the opening tax asset base).
146 Table 8 reproduced at par 220 of the final decision is the Total Revenue Building Blocks in nominal dollars, not real dollars as described in that paragraph.
147 It is apparent from the information contained in those paragraphs of the final decision that the quantum of the ERA's approved Revenue Building Blocks ($1,844.53 million in nominal dollars) was an increase over the quantum assessed in the ERA's draft decision ($1,762.18 million in nominal dollars) but a decrease over the amount claimed by the owners ($2,210.04 million).
148 The Rate of Return determination made by the ERA is referred to in summary form at pp 218 to 222 (pars 1058 to 1070) of the final decision and also in Appendix 4 Rate of Return to that decision.
149 At pp 218 to 220 (pars 1058 to 1066), the ERA said:
In its Draft Decision the Authority did not accept DBP's approach for estimating the rate of return and determined its own numbers.
The Draft Decision noted that, as provided in the Final Decision on Proposed Revisions to the Access Arrangement for the Mid-West and South-West Gas Distribution Systems (hereafter, the ATCO GDS Final Decision) published as amended on 10 September 2015, [Economic Regulatory Authority, Final Decision on Final Decision on Proposed Revisions to the Access Arrangement for the Mid-West and South-West Gas Distribution Systems, as amended 10 September 2015] the Authority had recently modified its approach to estimating the return on debt and the return on equity as outlined in the Authority's Rate of Return Guidelines [Economic Regulation Authority, Rate of Return Guidelines, 16 December 2013].
The Authority considered that the modified approach aligned with the regulatory requirements for the rate of return as specified in the National Gas Law (NGL) and National Gas Rules (NGR). The Authority considered DBP's proposal for estimating the rate of return, but was not convinced that it met the requirements of either the NGL or the NGR.
The detailed reasoning for the Authority's Draft Decision is set out in Appendix 4 and is summarised below. The Authority:
• continued to estimate the rate of return based on the debt proportion of total capital - the gearing - for the benchmark efficient entity of 60 per cent;
• with regard to the estimate of the return on equity:
- retained the Sharpe Lintner Capital Asset Pricing Model (SL-CAPM) as the primary relevant model for estimating the return on equity;
- utilised information from other relevant models - including the Black Capital Asset Pricing Model (Black-CAPM) and the Dividend Growth Model (DGM) - to establish the value of parameters in the Sharpe Lintner CAPM;
- estimated the risk free rate parameter for input to the Sharpe Lintner CAPM from Commonwealth Government Securities, based on a 5 year term to maturity;
- estimated a range for the 5 year forward looking Market Risk Premium (MRP) based on historic excess return data and the DGM, in recognition that it fluctuates in response to prevailing conditions;
- drew on a range of forward looking information to establish the point value of the MRP; and
- estimated the beta parameter based a benchmark sample of Australian firms with similar characteristics to the benchmark efficient entity.
• with regard to the estimate of the return on debt:
- continued to estimate the cost of debt as the sum of the risk free rate, relevant Debt Risk Premium (DRP), and relevant debt raising and hedging transactions costs;
- estimated the risk free rate from the bank bill swap rate with the same term as the regulatory period, that is, 5 years;
- adopted a hybrid trailing average approach to estimating the return on debt, with the risk free rate estimated once, just prior to the regulatory period, and the DRP estimated using an equally weighted 10 year trailing average;
- estimated the DRP based on a BBB band credit rating, for a term of 10 years, using the Authority's enhanced bond yield approach that included international bonds issued by domestic entities (and for estimates of the DRP prior to the proposed averaging period, utilise the Reserve Bank of Australia's credit spread data for the BBB band); and
- annually updated the estimate of the DRP using a set of specified automatic formulas.
The Authority's resulting indicative estimate for the overall post tax nominal rate of return for its Final Decision, for the 2016 calendar year, was 6.02 per cent:
• the indicative expected 5 year return on equity was 7.28 per cent, estimated as at 2 April 2015;
• the indicative estimate for the return on debt for the 2016 calendar year was 5.172 per cent, estimated as at 2 April 2015.
This rate of return was applied from 1 January 2016 to 31 December 2020 in the tariff modelling for the Draft Decision, in order to estimate indicative tariffs for the Draft Decision.
The Draft Decision noted the indicative estimate of the rate return on debt, would be updated in the Final Decision to account for DBP's nominated averaging period for the 2016 estimate. The overall method for determining that revised calendar year 2016 estimate would follow that for the indicative estimate set out in the Draft Decision. The resulting estimated rate of return for 2016 would be applied in the tariff modelling for the Final Decision for 2016 to 2020.
The Draft Decision noted the 2017 to 2020 rates of return would then be progressively annually updated through the remaining years of AA4. The resulting revised rate of return would be included in the relevant tariff variations which occur in each calendar year.
The Draft Decision noted the process for implementing the annual update would be as follows:
• For each annual update for 2017, 2018, 2019 and 2020, the Authority would estimate the updated DRP following the relevant annual averaging period, recalculate the rate of return, and then notify DBP of the outcomes as soon as practicable. This would allow DBP to check the rate of return estimate, prior to its incorporation in the proposed annual tariff variation to occur on 1 January in each year.
• Following that notification, DBP would be required to respond on any issues as soon as practicable, in order to allow the updated DRP and rate of return estimates to be finalised prior to submission by DBP of its proposed annual tariff variation.
• In the event that there was a disagreement on the DRP annual update estimate, the Authority would work with DBP to ensure that any misapplication of the automatic formulas in Appendix 4G of the Draft Decision were corrected in a timely manner.
150 As a result of the reasoning and determinations summarised in those paragraphs, the ERA required an amendment to the owners' amended proposal in respect of the Rate of Return.
151 The owners did not accept the ERA's proposed amendment. The ERA then determined its own numbers which it then recorded in Table 71 on p 221 of the final decision. That table is in the following form:
152 The line items under challenge by Ground 1 of the owners' grounds of review are the 0.7 figure for equity beta and the second last and third last line items in Table 71.
153 In addition, it is worth noting at this point that the second ground of review (the gamma ground) relates to the fourth last line item in Table 71, with particular reference to the figure in the last column (0.40).
154 The detailed reasoning of the ERA in relation to its ROE decision is found in Appendix 4. Appendix 4 is itself a lengthy document (233 pages).
155 At pp 2 to 10 (pars 7 to 42) of Appendix 4, the ERA summarised the terms of the owners' original proposal in respect of the rate of return. At pars 41 and 42 of Appendix 4, the ERA said:
In revisions to the Access Arrangement, DBP proposed an allowed post tax nominal rate of return for the benchmark efficient entity of 8.36 per cent (as at 30 September 2014).
With debt gearing of 60 per cent, DBP's proposed nominal rate of return was a weighted average of:
• a return on equity of 11.71 per cent; and
• a return on debt of 6.13 per cent.
156 At pp 10 to 13 (pars 43 to 60), the ERA described its response to the owners' original proposal as reflected in its draft decision.
157 At pp 11 to 12 (pars 51 to 53), the ERA said:
The following conclusions were reached in relation to the approach for estimating the return on equity in the Draft Decision for DBP:
• The SL-CAPM should be utilised to estimate the return on equity.
• The Fama French three factor model is not relevant and as such, this model should not be used for the purpose of estimating a return on equity.
• The Black CAPM is relevant for the purpose of estimating a return on equity. However, given it is not reliable and practical to estimate a robust return on equity using this model, the model will not be used directly, but only to inform the point estimate of the equity beta from within its range for input to the SL-CAPM.
• The DGM is a relevant model for informing the market return on equity and also the forward looking MRP.
• Other information such as historical data on equity risk premium; surveys of market risk and other equity analysts' estimates are also relevant for the purpose of estimating the MRP and the market return on equity. This other material should be used as a cross check for the return on equity.
Given that the only robust model for estimating the return on equity in the Australian context is the SL-CAPM, the Authority did not see any current need for data sourced from the SIRCA SPPR database, as suggested by DBP [DBNGP Transmission Pty Ltd, Proposed Revisions DBNGP Access Arrangement, 2016 - 2020, Rate of Return, Supporting Submission: 12, 31 December 2014, p. 55]. The SPPR database was required by DBP to form long time series of predictions for the model adequacy test [This need for a long time series was considered one of the weaknesses of the model adequacy test (Appendix 4B), one which can be circumnavigated by various approaches to cross-validation (Appendix 4Bi)]. As need for the model adequacy test was rejected, then so too was need for the SPPR database.
The Authority remained of the view that its reasons for adopting the SL-CAPM are sound. The Authority considered that its application of the SL-CAPM meets the requirements of the NGR, and the allowed rate of return objective.
• The Authority did not agree with DBP's submission that it had not taken all of the relevant information into consideration with respect to its estimate of the return on equity. The Authority was of the view that all of the issues raised by DBP and its consultants were considered in the Draft Decision.
• The Authority also disagreed with DBP's estimates of the rate of return on equity. The Authority conducted significant research into the rate of return and cross checked its estimate across various sources. The Authority's estimate for the rate of return was in line with other industry estimates.
• The Authority considered that the estimated return on equity adopted in the Draft Decision was commensurate with the equity costs incurred by a benchmark efficient entity with a similar degree of risk as DBP with respect to the provision of reference services. The Authority therefore considered that the estimated rate of return meets the allowed rate of return objective and the requirements of the NGR and NGL.
158 The reference to "DGM" in par 51 is a reference to the Dividend Growth Model.
159 It is clear from the ERA's references to its draft decision in Appendix 4 that the ERA had considered other models, information and material and was of the opinion that the only robust model for estimating the return on equity in the Australian context is the SL-CAPM.
160 At pp 14 to 32 (pars 61 to 156) of Appendix 4, the ERA recorded the substance of the owners' response to its draft decision. Its references to the owners' response in these passages included references to the owners' response by way of further submissions (see, in particular, pp 30 to 32 (pars 146 to 156)).
161 At pp 15 to 16 (pars 69 to 78) of Appendix 4, the ERA summarised the owners' responses in the following terms (references to 'betastar' being references to an algebraic formula adjustment propounded by the owners):
Return on equity
In response to the Draft Decision, DBP submits only a slightly amended approach to estimating the return on equity, as compared to that put forward in its initial proposal.
First, DBP updates its range of outcomes for the return on equity from its model adequacy test. This delivers a so-called 'unbiased' betastar range of 1.00 (25th percentile) to 1.70 (99th percentile), around a mean of 1.15 [A footnote to this sentence stated that DBP have revised slightly the betastar estimates to reflect a change in the benchmark efficient entity sample set. The revised sample set reflects the omission of Envestra and HDF, which are now both delisted from the ASX DBP states (DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Access Arrangement Period Supporting Submission: 56, 24 February 2016, p. 97): We note in Appendix 4A(ii) that the ERA has dropped Envestra and HDF from its original sample set as they are now dead stocks. We are unclear as to why it did not do this at its last estimation; Envestra was trading until September 2014, but HDF ceased trading in November 2012, a year before the ERA undertook the beta calculations in its Guidelines. The ERA has not explained this change in stance.].
Second, the resulting range for the return on equity, when the betastar range is applied within the SL-CAPM, is between 9.9 per cent and 14.82 per cent. This result builds on DBP's estimate of the 10 year risk free rate, of 2.87 per cent, and its estimate of the Market Risk Premium (MRP) of 7.03 per cent.
Third, DBP also utilises information from the return on debt to derive the final range of the return on equity, drawing on the insights of Merton. DBP considers that this ensures consistency between the return on debt and the return on equity. Based on its analysis, DBP argues that the range of the return on equity should be between 10.61 per cent and 11.06 per cent. This range is narrower, lying entirely within the range given by the betastar estimates. DBP then takes the midpoint estimate of the narrower range, which is 10.84 per cent, to be the best estimate of the return on equity.
In arriving at this position, DBP considers that two issues are central to the differences between DBP and the Authority, which are reflected in their respective approaches to the return on equity.
First, DBP is of the view that the Authority has not made a proper assessment of its betastar approach. DBP contends that the Authority had based its conclusions on superficial reasoning and irrelevant evidence, while ignoring relevant evidence. DBP submits that the Authority fails to make a proper application of the evidence which itself had produced in relation to the identification or quantification of bias within the SL-CAPM.
Second, DBP argues that the Authority did not test whether the outcome of its SL-CAPM approach to estimating the return on equity meets Rule 87(5). DBP, on the other hand, considers that it does this through the use of its model adequacy test. DBP also contends that the need to test outcomes as well as inputs is a fundamental aspect of the regulatory framework in the NGL and NGR.
DBP maintains substantially the same approach to determining the return on equity in its initial Access Arrangement Proposal; that is, the application of its 'model adequacy test'. DBP considers that this tests the outputs of models, and whether they give rise to a range of unbiased outcomes; such that the model results then neither systematically overstate nor understate actual returns.
DBP notes that one of the amendments from the Authority requires the DBP to implement the SL-CAPM using the five-year risk-free rate and a beta of 0.7, along with the estimate of the MRP. DBP argues that the first two could be done, but not the third. This problem arises because the Authority's estimate of the MRP changes at each regulatory decision, based upon how it interprets a number of "forward looking" indicator variables. DBP considers that the Draft Decision fails to outline the ERA's methodology for quantifying the correlation between changes in these variables and the change in the MRP.
DBP's reasons for rejecting the Authority's views on relevant asset valuation models
DBP submits that the Authority accepts that the Black CAPM, dividend growth model (DGM) and SL-CAPM are relevant in principle, as it does. However, a key difference arises with respect to the Fama-French model (FFM); DBP considers it to be relevant in-principle (based on the advice of CEG) but the Authority does not.
162 At p 17 (par 83), the ERA recorded the owners' submission that the notion that the SL-CAPM is biased downwards is hardly a new idea, noting that the ERA accepted that this downward bias exists when it chose 0.7 for beta, while specifically acknowledging that it was doing so in order to address the issue of bias. In the same paragraph, the ERA noted the owners' submission that the model adequacy test utilised by the ERA produced results that no-one else had found suggesting that the ERA's view was contrary to more than 40 years of empirical finance.
163 The ERA acknowledged that the owners had submitted that there was a need for an empirical test of outputs. The ERA also understood that the owners had submitted that there was no proper basis for the ERA's rejection of the model adequacy test.
164 At p 26 (pars 124 to 126), the ERA recorded the owners' submissions in the following terms:
Reasons for maintaining betastar
DBP considers that the Authority's approach, given a finding of downward bias for low beta stocks in the SL-CAPM, is completely irrational. DBP argues that [DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Access Arrangement Period Supporting Submission: 56, 24 February 2016, p. 55]:
… the ERA is acknowledging that bias exists, acknowledging that different models can supply information which might help overcome the bias, but then explicitly rejecting any information from those models in order to solve the bias problem in order to satisfy itself that it is not deviating from the SL-CAPM in any material way.
DBP is of the view that its betastar adjustment is transparent and can be easily followed by any observer [DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Access Arrangement Period Supporting Submission: 56, 24 February 2016, p. 55].
DBP agrees with the Authority's view that there are no literature or empirical studies which use a betastar approach. DBP accepts that it fails to provide a single reference to support its view that betastar is well established, or at least follows any standard economic or statistical theories. However, DBP argues that [DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Access Arrangement Period Supporting Submission: 56, 24 February 2016, p. 55]:
This, however, is not surprising; in the ordinary course of events, if the Black CAPM passed a test like the model adequacy test but the SL-CAPM did not, one would simply have used the Black CAPM. However, betastar was adopted so as to minimise departure from the Guidelines.
165 At p 28 (pars 134 to 137), the ERA recorded the following:
Beta
In its response to the Authority's Draft Decision in relation to the estimates of equity beta, DBP submitted that it has no issue with the estimation of beta as undertaken by the Authority [DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Access Arrangement Period Supporting Submission: 56, 24 February 2016, p. 63].
However, DBP considered that there are two issues in respect of the Authority's beta, including: (i) The estimate of beta the Authority has used of 0.7 produces a result which is not consistent with the approach it has used in the past, because it has failed to take into consideration the changes in its beta estimation; and (ii) a potential issue concerning the efficiency of the market portfolio.
First, DBP argued that as the confidence interval around beta has shifted upwards, the Authority's choice of beta has not changed [DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Access Arrangement Period Supporting Submission: 56, 24 February 2016, p. 64]. DBP considered that systematic risk is measured relative to the market and one would expect change as either the actual risks facing the firm changed or risks in the market changed. DBP was of the view that a consistent regulator would also choose a point two basis points below the upper end of the same confidence interval to address the same bias issue. As such, DBP argued that doing so would require the Authority to adopt the estimate of equity beta of 0.79 [DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Access Arrangement Period Supporting Submission: 56, 24 February 2016, p. 64].
Second, DBP submitted that if the market portfolio is inefficient, then the SL-CAPM fails to hold, and the conclusions the Authority has drawn in respect of beta are wrong. DBP argued that, more importantly, DBP concluded that [DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Access Arrangement Period Supporting Submission: 56, 24 February 2016, p. 66]:
DBP is to show that the predictions made by an SL-CAPM predicated on an inefficient market portfolio are downward-biased estimators of the actual returns made firms with (imperfectly measured) systematic risk similar to (likewise imperfectly measured) systematic risk exposure to the benchmark efficient firm, whilst the Black CAPM does not produce downward-biased estimators.
166 At pp 32 to 189 (pars 157 to 189), the ERA explained its reasoning in respect of the Rate of Return which it proposed to adopt for the purposes of its final decision.
167 At pp 32 to 33 (par 157), the ERA said that it did not accept the owners' proposed approach for estimating the rate of return, as that approach does not comply with the regulatory requirements for the rate of return as specified in the NGL and NGR. The ERA said that, in evaluating the owners' amended proposal, it had drawn on previous positions set out in the Rate of Return Guidelines, prior decisions, the draft decision and the owners' response to the draft decision.
168 The ERA commenced its consideration of the issue of Return on Equity at p 50 (par 240). At pp 50 to 52 (pars 240 to 244), the ERA said:
Return on equity
In line with the requirements of NGR 87(5), the Authority evaluated the relevance of a broad range of material for estimating the return on equity in the Rate of Return Guidelines, covering relevant estimation methods, financial models, market data and other evidence [A footnote to this sentence referred to the Australian Energy Market Commission, Rule Determination: National Gas Amendment (Price and Revenue Regulation of Gas Services) Rule 2012, 29 November 2013, p. 36].
The Rate of Return Guidelines set out that the Authority will utilise a five step approach for estimating the return on equity. The five steps are summarised in Figure 2 below.
In applying this approach, the Authority has assessed a wide range of material, and identified relevant models for the return on equity, as well as a range of other relevant information. For this Final Decision, the Authority has had regard to and given weight to relevant material, according to its merits, seeking to fully achieve the requirements of the allowed rate of return objective.
204 The Authority considers that the term:
- 'approach' refers to the overall framework or method for estimating the return on equity, which combines the relevant estimation methods, financial models, market data and other evidence;
- 'estimation material' refers to any of the relevant estimation methods, financial models, market data and other evidence that contribute the 'approach'; and
- 'estimation method' relates primarily to the estimation of the parameters of financial models, or to the technique employed within that model to deliver an output.
The Authority in the Rate of Return Guidelines determined that only a subset of the material evaluated at that time could be considered relevant in the Australian context, given the allowed rate of return objective. The Authority remains of the view that:
Rate of return estimate materials - the estimation methods, financial models, market data and other evidence - would need to be broadly consistent with the requirements of the NGL, the NGO, the NGR and the allowed rate of return objective to be considered relevant. Some estimation materials may perform better on some requirements and less well on others, and yet may still be considered relevant. Accordingly, the assessment is whether, on balance, estimation materials are consistent with the requirements of the NGL, the NGO, the NGR and the allowed rate of return objective.
Nevertheless, estimation materials would need to pass a threshold of adequacy to be considered relevant. To the extent that estimation materials failed the adequacy threshold, then they would be rejected. This rejection would be consistent with the AEMC's purpose for the guidelines [Australian Energy Market Commission, Rule Determination, National Gas Amendment (Price and Revenue Regulation of Gas Services) Rule 2012, 29 November, p. 58]:
In order for the guidelines to have some purpose and value at the time of the regulatory determination or access arrangement process, they must have some weight to narrow the debate.
Once over the threshold for adequacy, then, as noted, any particular estimation material may meet the requirements of the NGL, the NGO, the NGR and the allowed rate of return objective to a greater or lesser degree. With this in mind, the criteria would then be used as a means to articulate the Authority's evaluation of the estimation materials, in terms of how they performed in meeting the requirements of the NGL, the NGO, the NGR and the allowed rate of return objective. In this way, the criteria are intended to assist transparency around its exercise of judgement [Economic Regulation Authority, Explanatory Statement for the Rate of Return Guidelines, 16 December 2013, p. 12].
In that context, the following analysis provides the Authority's determination for this Final Decision of the return on equity for the DBNGP benchmark efficient entity. The Authority considers that the estimate is consistent with delivering an outcome that meets the allowed rate of return objective, as well as the NGL and NGR more broadly [The allowed rate of return objective is set out at NGR 87(3):
The allowed rate of return objective is that the rate of return for a service provider is to be commensurate with the efficient financing costs of a benchmark efficient entity with a similar degree of risk as that which applies to the service provider in respect of the provision of reference services].
169 The ERA then moved through the steps depicted in Figure 2 which is set out at p 51 (par 242) and explained its approach at each of those steps.
170 After referring to the Rate of Return Guidelines, the ERA then moved to consider the approach which it had adopted in its draft decision.
171 At pp 55 to 56 (pars 252 to 255) of Appendix 4, the ERA recorded its opinion, reflected in its draft decision, that the SL-CAPM and the Black CAPM are both relevant models for estimating the return on equity. However, the ERA took the view that it was not appropriate for it to regard the Black CAPM as suitable for use directly for estimating the return on equity. The ERA concluded that the owners' proposed betastar method had significant empirical flaws. These flaws were summarised at pp 55 to 56 (par 253) as follows:
• the zero beta portfolio is sensitive to the data set used, highly variable through time with a wide standard error, and is therefore not robust [Economic Regulation Authority, Draft Decision on Proposed Revisions to the Access Arrangement for the Dampier to Bunbury Natural Gas Pipeline, 22 December 2016, Appendix 4, p. 45];
- DBP's estimates - which use a single average estimate of zero beta premium - disguise the significant instability of the Black CAPM model;
- DBP's model adequacy test is selective in its interpretation of the Black CAPM model;
• the betastar approach does not produce sensible results [Economic Regulation Authority, Draft Decision on Proposed Revisions to the Access Arrangement for the Dampier to Bunbury Natural Gas Pipeline, 22 December 2016, Appendix 4, p. 45.];
- the indicative overall market return on equity for a long period, estimated at the time of the Draft Decision, was approximately 10.83 per cent, [Economic Regulatory Authority, Final Decision on Proposed Revisions to the Access Arrangement for the Mid-West and South-West Gas Distribution Systems, as amended 10 September 2015, p. 255] which is lower than DBP's estimated return for low asset beta entities such as the DBP;
- DBP is therefore suggesting that its return on equity is more risky than the market as a whole;
- but this is not sensible on conceptual grounds.
172 In addition, the ERA did not accept the owners' model adequacy test for the reasons set out at p 56 (par 254) of Appendix 4. Those reasons were:
• relying on the historical data alone - as DBP does - for testing the relative adequacy of the Authority's approach, is erroneous;
• other forward looking information needs to be taken into account, as the Authority does in its approach to estimating the return on equity;
• it follows that DBP's model adequacy approach does not actually test the Authority's approach in using the SL-CAPM for estimating the return on equity;
• DBP in essence compares two models that are not robust in the Australian context (the Black CAPM and FFM) with another method that is not relied on either - an ex post SL-CAPM with an MRP that is based on historic data only.
173 The ERA then moved on to address the owners' response to the ERA's draft decision. It noted that the owners continued to seek to apply its model adequacy test to determine the relevance of models for the return on equity. On this occasion, the owners tested the outputs of three models - the SL-CAPM, the FFM and the Black CAPM - in terms of their ability to predict actual (ex post) market outcomes. The ERA noted that the owners had submitted that only the outcomes of the Black CAPM provide for unbiased estimates of the return on equity whereas the results from the other two models are considered to be biased and hence poor forecast predictors.
174 As noted by the ERA, the owners then transformed the two percentile Black CAPM outcomes into "betastar" estimates, for use in the SL-CAPM, as a means to estimate an unbiased range for the return on equity. The ERA noted that the owners narrowed the return on equity range using the so-called "Merton framework" cross check method, before choosing the mid-point as the resulting return on equity.
175 At p 57 (pars 260 to 262) of Appendix 4, the ERA said:
Consequently, the core of DBP's proposal relates to the model adequacy test, the associated inference that only the Black CAPM leads to unbiased results for the return on equity, and the use of the betastar transformation so as to implement the results of the Black CAPM within the framework of the SL-CAPM. DBP's claims with regard to the Merton framework cross check method are also key to its estimate.
The Authority considers these four elements of DBP's response in turn regarding the return on equity in what follows.
Further evaluation of DBP's model adequacy test and application of betastar
The Authority has significant concerns - both conceptual and empirical - with DBP's model adequacy test and betastar transformation. The following two sections set out the Authority's reasoning regarding DBP's proposed approach from these conceptual and empirical perspectives.
176 Over the ensuing 30 pages or so, the ERA considered, in great detail, the conceptual and empirical elements of the owners' suggested approach.
177 At p 61 (par 278), the ERA concluded that there were significant issues with the construct of the owners' model adequacy test. The ERA observed that no method tests the ERA's actual implementation of the SL-CAPM and that all methods seek, erroneously, to compare expected returns with ex post actual returns.
178 The ERA then moved on to consider the question of whether there was bias in the performance of the SL-CAPM or merely an anomaly.
179 At pp 61 to 66 (pars 279 to 299), the ERA said:
Bias or anomaly?
DBP's model adequacy test is intended to uncover 'bias' in the performance of the SL-CAPM. DBP then makes an adjustment to the beta in the SL-CAPM, as a means to counter the perceived bias. That 'betastar' adjustment is based on the Black CAPM, and is of the form [DBP, Proposed Revisions DBNGP Access Arrangement, 2016 - 2020 Regulatory Period, Rate of Return, Supporting Submission: 12, 31 December 2014, p. 68]:
However, it is not the beta in the SL-CAPM that is biased. As noted by Pink Lake in its evaluation of the statistical properties of the SL-CAPM and the Black CAPM [Pink Lake Analytics, Statistical Advice to ERA on DBP Submission 56, May 2016, p. 4]:
Upon review it is clear that the positions of the ERA and DBP are divergent. The Authority derives an RoE calculation from the Henry [Henry, O.T., Estimating 𝛽: An update, April 2014] statistical version of the Sharpe-Lintner Capital Asset Pricing Model (SL CAPM). The statistical model itself is valid - in predicting the data it minimises the squared error difference between observations and model predictions. Furthermore, the model includes a free intercept term in excess of the risk-free rate (𝛼), so for its class of models (i.e., linear models with a single predictor) it provides an unbiased estimate of 𝛽, the measure of an asset's exposure to systematic risk in the market. The ERA then omits the 𝛼 estimate of abnormal returns from the Henry model in its implementation of the Sharpe-Lintner CAPM, deeming these abnormal returns as not reflective of the systematic risk in market prices that is faced by benchmark efficient firms…
In contrast, DBP implements the Black CAPM model by first estimating a zero-beta premium (ZBP). Effectively, this ZBP estimate is a measure of the abnormal returns in excess of the risk-free rate. As such, although the Black CAPM is marginally biased in terms of its predictions (as it does not include a free intercept term) this bias is statistically insignificant. Where DBP and the Authority differ in their positions is that DBP include the full weight of the ZBP, as a de facto measure of abnormal returns in their RoE calculation…
Pink Lake has set out very clearly the mathematical underpinnings of the two modelling approaches, so these are not reproduced here [Pink Lake Analytics, Statistical Advice to ERA on DBP Submission 56, May 2016, p. 4]. Pink Lake's evaluation confirms that the Authority's estimate of the SL-CAPM beta is not biased. However, by loading the betastar adjustment into the SL-CAPM beta, DBP biases the estimate of beta in its estimate. At the same time, in so doing, DBP imports all of the deficiencies of its Black CAPM into the resulting SL-CAPM estimate.
It is clear that DBP is mistakenly comparing one form of model (a model of ex-post actual returns - the Black CAPM with a full intercept term, where the zero beta return captures all ex-post anomalies) with another form of model (a model of ex-ante expected returns - the SL-CAPM with no α intercept included). As Partington and Satchell observe [Partington, G. and Satchell, S., Report to the ERA: The Cost of Equity and Asset Pricing Models, May 2016, p. 34.]:
We need to be clear what unbiased means. If it means that the DBP Black CAPM estimates, when subject to a model adequacy test as proposed by DBP, are such that the model adequacy test is not rejected, then they are generally unbiased, at least with respect to the beta sorted portfolios.
However, this view of unbiasedness then gets translated into a view that the regulator who uses the SL CAPM is providing investors with approximately 4% per annum less compensation. This treats low beta ex-post returns as equilibrium returns. Here and elsewhere in the document we take the view that the [SL-CAPM] low beta anomaly is indeed an anomaly. The correct regulatory return would be more sensibly based on subtracting the intercept term from [ex post SL-CAPM] returns, not adjusting the slope and certainly not treating the Black CAPM (unbiased) returns as fair compensation. The more so since the SL CAPM industry portfolios also pass the unbiasedness test.
It follows that if there is any 'bias' arising in the Authority's estimate, that bias occurs with the Authority's omission of the α intercept term from its statistical estimation process. DBP in essence agrees with this point [DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Access Arrangement Period Supporting Submission: 56, 24 February 2016, p. 51].
…the only theoretical difference between the SL-CAPM and the Black CAPM lies not in beta, but on the intercept... The practical effect of this theoretical change is to shift the intercept of the security market line upwards, and thus lessen its slope. This, in turn, makes the expected returns of low beta stocks higher and of high beta stocks lower than predicted by the SL-CAPM.
This is important when considering the bias adjustments made by the ERA and DBP (through its betastar model). The ERA, motivating the "theory" of the Black CAPM, changes beta, using a higher level of beta than the mean value it obtains from its own regressions. However, the theoretical change from the SL-CAPM to the Black CAPM has nothing whatsoever with beta, it is a shift of the intercept.
Pink Lake also recognises that the 'low beta bias' issue relates to the interpretation of the intercept in the SL-CAPM estimation process. Pink Lake points out that this makes the whole model adequacy exercise redundant, in so far as it is testing for beta bias [Pink Lake Analytics, Statistical Advice to ERA on DBP Submission 56, May 2016, p. 4]:
For the Authority, the statistical model employed is already an optimally fitting model under reasonable model assumptions. Hence, there is no reason to undertake further the model validation proposed by DBP when adopting the Authority's position. In contrast, DBP propose to apply the model validation to the RoE calculation itself. As the DBP RoE calculation is essentially the same as their statistical model, then it is self-evident that their RoE calculation does not exhibit significant model bias. Similarly, it is self-evident that the Authority's RoE calculation does exhibit model bias, as it deliberately excludes the abnormal return component estimated in the Henry model in excess of the risk-free rate. Both the Henry model and the Black CAPM are valid, depending on the position being adopted. The question of which position to accept - either the Authority's or DBP's - is therefore not a statistical question, but a question of economics, and one that falls outside the scope and expertise of this consultancy.
Consequently, the Authority now recognises that there is no justification for changing the value of beta in the SL-CAPM. The further implication is clear: DBP, by adjusting beta, is introducing a highly significant bias into the beta estimate in its implementation of the SL-CAPM. The Authority considers DBP's approach to be in error on this ground.
The case for an alpha adjustment
Having examined the implications of DBP's arguments with regard to the bias in the SL-CAPM, and rejected the case for any adjustment to beta in the SL-CAPM, the Authority now turns to consider whether there is any case to adjust for α in the estimates derived from the model tests.
The Authority considers that there is little compelling evidence about the degree to which the α intercept term, or even part of it, should be included.
A positive intercept in tests of the SL-CAPM does not automatically imply that the Black CAPM applies. Positive intercepts (α) in ex-post outcomes are not automatically estimates of a zero beta premium [Partington, G. and Satchell, S., Report to the ERA: The Cost of Equity and Asset Pricing Models, May 2016, p. 17].
The theory of the SL-CAPM does not include the α term. Rather, the presence of positive (or indeed, a negative value of) α relates to differences (so-called 'anomalies') between the required (or expected or equilibrium) returns and realised returns [Refer paragraph 265 for the links between required, expected and equilibrium returns]. The Authority seeks to ensure that investors in the benchmark efficient entity obtain the required return, consistent with NGR 87 (see paragraph 264). As Partington and Satchell observe [Partington, G. and Satchell, S., Report to the ERA: The Cost of Equity and Asset Pricing Models, May 2016, p. 7]:
When prices are in equilibrium this required return is equal to the expected return, but there is no guarantee that expectations will be realised, or that prices are always in equilibrium. If there were a guarantee that expectations would be realised then the asset would have no risk.
Consistent with that view, the α intercept in observed returns should be subtracted in its entirety, in order to establish the required forward looking equilibrium returns [Partington, G. and Satchell, S., Report to the ERA: The Cost of Equity and Asset Pricing Models, May 2016, p. 15:
This usual argument for the Black CAPM is based on the premise that actual returns are equal to equilibrium returns on average and thus a positive intercept in tests of the SL CAPM are assumed to be driven by the SL CAPM underestimating (overestimating) realised returns for low (high) beta stocks. An alternative premise is that the results are a consequence of actual returns outperforming (underperforming) equilibrium returns for low (high) beta stocks. In the parlance of funds management such outperformance is expressed as alpha. Thus low beta stocks have positive alphas. In this case an estimate of the equilibrium return is obtained by subtracting alpha from the actual return. Whether the resulting return is then higher or lower than the regulated return is an open question and will depend upon the magnitude of alpha.
For similar reasons, this subtraction of the intercept term was employed by Henry, in estimating beta, and indeed the same subtraction is adopted by the Authority in its updated estimates of beta for input to the SL-CAPM [It is noted that the alphas in tests of the SL-CAPM, based on beta portfolio sorts, are not identical to the intercept term identified in the Henry-style beta estimation process. Nevertheless, they originate from the same source - that is, from anomalous returns observed in the ex post outcomes]. That is, the intercept in excess of the risk free rate is ignored, forcing the SL-CAPM security market line through the origin, consistent with the theory of the SL-CAPM.
The Authority considers that there is no justification to 'add back in' any alpha from the observed returns to the SL-CAPM, where those are simply differences, ex-post, as compared to the ex-ante required returns.
At the same time, the Authority is not convinced there is any empirical evidence at the current time to justify an adjustment to the SL-CAPM for expected alpha for the benchmark efficient entity. As noted above at paragraph 265, theory suggests that if such an expectation was widespread among investors, it would be bid away as part of a movement toward equilibrium asset pricing.
To examine this, the Authority turns to DBP's own model adequacy test results (even though, for the reasons stated above, the Authority does not consider the model adequacy approach a valid rationale for rejecting the SL-CAPM). DBP's own estimates indicate that, based on industry sorts, the model adequacy tests conducted by DBP tends to support the SL-CAPM. DBP tests two versions of the SL-CAPM - a vanilla version and an 'ERA' version, where it takes the 95th percentile beta of beta for each industry - in two tables in Appendix D of its initial proposal [DBP explain this as follows (DBP, Proposed Revisions DBNGP Access Arrangement, 2016 - 2020 Regulatory Period, Rate of Return, Supporting Submission: 12, 31 December 2014, Appendix D, p. 14):
The results of tests of the SL-CAPM that use industry returns appear in Table 7 below while the results of tests of the ERA's version of the SL-CAPM, which uses the 95th percentile of an estimate of the distribution of an OLS estimator for beta rather than an estimate of the mean of the distribution (the OLS point estimate), appear in Table 8]. As noted by Partington and Satchell [Partington, G. and Satchell, S., Report to the ERA: The Cost of Equity and Asset Pricing Models, May 2016, p. 21]:
Tables 7 and 8 from DBP [2015]… provide statistics for the mean forecast error for the SL CAPM by industry. The description in DBP's text says that the results of the ERA's version of the SL CAPM are in Table 8, whereas according to the title on Table 7 it gives the ERA's version of the SL CAPM. We think the latter is correct, but fortunately, the labelling is of no real consequence as there is relatively little difference in the nature of the results between the two tables.
The results in Tables 7 and 8 generally are supportive of the SL CAPM. Across the 104 tests in the two tables significant bias is only observed with respect to 3 industries. These are retailing, pharmaceuticals and utilities, which provide six results significant at the 5% level. With the exception of retailing, these results are only significant for Method B. In short there is very little evidence of significant bias and the number of significant results is approximately the number expected by chance. With a type 1 error of 5% we would expect 5.2 of the 104 hypotheses to be rejected even if the null is true. Thus finding only 6 rejections suggests to us that the SL CAPM is supported by these testing procedures.
The Authority agrees with Partington and Satchell that the evidence in these tables is supportive of the SL-CAPM. This is particularly the case for method A, which is the more relevant test (method B does not test any form of expected return, as noted above). DBP dismisses this, on the basis that the industry tests are of low power. However, it is notable that DBP's argument relates to the - in the Authority's view - discredited method B [DBP, Proposed Revisions DBNGP Access Arrangement, 2016 - 2020 Regulatory Period, Rate of Return, Supporting Submission: 12, 31 December 2014, Appendix D, p. 14]:
The low power of the tests is illustrated by the fact that a Method B test of the null hypothesis that the ERA's version of the SL-CAPM provides an unbiased estimator of the return required on a portfolio of utilities is unable to reject at the five percent level the null despite the mean forecast error associated with the estimator being 0.557 percent per month.
The Authority therefore is not convinced that there is strong evidence from DBP's analysis to reject the standard theoretical form of the SL-CAPM.
The Authority now considers, given these insights, that there is inadequate evidence, at this time, to justify departure from an ex-ante alpha estimate of zero in its implementation of the SL-CAPM:
• a positive intercept in tests of the SL-CAPM does not automatically imply that the Black CAPM applies;
• the theory of the SL-CAPM does not include the α term; rather, the presence of positive (or indeed, a negative value of) α relates to anomalies; and
• DBP's own estimates indicate that, based on industry sorts, the model adequacy tests conducted by DBP tend to support the SL-CAPM.
On this basis, the resulting implementation of the SL-CAPM becomes consistent with the theoretical form of the SL-CAPM: ex-ante, the SL-CAPM security market line is expected to pass through the zero intercept on the y axis. If positive alpha was expected ex-ante, prices would be expected to adjust to restore equilibrium and an expectation of zero alpha (refer to paragraph 265 above for this rationale).
The corollary is that while the theoretical insights of the Black CAPM are relevant - for example, for informing the theoretical position of the efficient market portfolio on the frontier in mean variance space in the absence of a riskless asset - the thorough exploration of this issue by the Authority identifies that the empirical estimate of the zero beta return, adopted by DBP, contains a large measure of anomalous alpha, and hence overestimates the required return. It is therefore not fit for purpose for estimating the return on equity for the benchmark efficient entity
180 For the reasons discussed by the ERA in the passages which we have extracted at [179] above, the ERA expressed its views at p 65 (par 297) of Appendix 4 in the following terms:
The Authority now considers, given these insights, that there is inadequate evidence, at this time, to justify departure from an ex-ante alpha estimate of zero in its implementation of the SL-CAPM:
• a positive intercept in tests of the SL-CAPM does not automatically imply that the Black CAPM applies;
• the theory of the SL-CAPM does not include the α term; rather, the presence of positive (or indeed, a negative value of) α relates to anomalies; and
• DBP's own estimates indicate that, based on industry sorts, the model adequacy tests conducted by DBP tend to support the SL-CAPM.
181 The ERA then moved on to discuss the relative acceptability of the SL-CAPM and the Black CAPM.
182 This discussion led to the conclusions expressed at p 69 (par 308) that it was reasonable for the ERA to use the SL-CAPM and that the Black CAPM could not be relied upon.
183 The ERA then moved on to discuss empirical elements of the owners' return on equity estimates.
184 After referring to a number of reports and items in the financial literature (at pp 69 to 84 of Appendix 4 (pars 309 to 369), at pp 85 to 86 (pars 370 to 382) the ERA said:
On that basis, DBP acknowledges that there is a reversal of ranking of the Fama French model and the CAPM when the method of portfolio formation changes.
Partington and Satchell note that the SL CAPM does not fare particularly well in the Kan, Robotti and Shanken tests, although the Inter temporal CAPM is a clear winner. Their view is that the results of Kan, Robotti and Shanken show the difficulty of all attempts to fit asset pricing models to realised returns, including the work of NERA/HoustonKemp [Partington, G. and Satchell, S., Report to the ERA: The Cost of Equity and Asset Pricing Models, May 2016, p. 28].
The Authority notes the sensitivity of model performance ex post to model specification and portfolio formation. This flags further caution with regard to the findings of DBP, as their work is based on a beta sort, rather than an industry sort.
On balance, the Authority considers that there are still many unsolved issues in relation to the estimates of the zero beta premium. As such, the Authority considers that DBP's estimates - which use a single estimate of zero beta premium - disguises the significant instability in the model. Therefore, the Authority does not consider that DBP's model adequacy test is empirically true to the Black CAPM model.
The Authority notes that the unresolved issues in relation to the estimates of the zero beta premium may explain why the Black CAPM has never seen widespread adoption by financial practitioners.
DBP's model adequacy test produces nonsensical outcomes
The Authority notes that based on the findings from its model adequacy test, DBP is of the view that the bias in its Sharpe Lintner CAPM analysis is not only statistically significant, but economically significant as well, with a mean forecast error of around four percentage points per annum. DBP considers that this means that a regulator using the Authority's approach to setting prices would provide investors with returns that are four percentage points lower than they could be earning by facing similar levels of systematic risk elsewhere in the economy [DBP, Proposed Revisions DBNGP Access Arrangement, 2016 - 2020 Regulatory Period, Rate of Return, Supporting Submission: 12, 31 December 2014, p. 60].
The implication of DBP's finding is that the expected return on equity for low beta assets, such as the ATCO GDS, the GGP and the DBNGP, needs to be increased by 4 percentage points, based on DBP's analysis and conclusion. For example, DBP argues that the expected return for DBP or ATCO (a low asset beta) using historical data on DBP's model adequacy test should be 11.28 per cent.
The Authority notes that the market return on equity for a long period is approximately 10.3 per cent [See the section 'Lower bound of the MRP range' below for the Authority's estimate of the long run historic return on the market], which is lower than DBP's estimated return for low asset betas such as DBP and ATCO. DBP is therefore suggesting that its return on equity is more risky than the market as a whole. The Authority does not consider that this view is sound.
There is conceptual support for the equity beta of an infrastructure network benchmark efficient entity being less than 1:
• business risk - which may be disaggregated into intrinsic (economic) risk and operational risk - is the primary driver of systematic risk, and this risk is low for the benchmark efficient entity relative to the market average;
• despite relatively high financial leverage, the benchmark efficient entity does not have high financial risk - rather it is the intrinsic risk of the firm which is the key driver of systematic risk.
McKenzie and Partington endorse the view that the equity beta is likely to be below 1, concluding that there is [McKenzie, Partington, Report to the AER: Estimation of the Equity Beta (Conceptual and Regulatory Issues) for a Gas Regulatory Process in 2012, April 2012, p. 15]:
…evidence to suggest that the theoretical beta of the benchmark firm is very low. While it is difficult to provide a point estimate of beta, based on these considerations, it is hard to think of an industry that is more insulated from the business cycle due to inelastic demand and a fixed component to their pricing structure. In this case, one would expect the beta to be among the lowest possible and this conclusion would apply equally irrespective as to whether the benchmark firm is a regulated energy network or a regulated gas transmission pipeline.
The Authority noted these views in its Draft Decision and considered that the reasoning is relevant. This provided further support for the Authority's view that DBP's model adequacy test produces a nonsensical results.
DBP took issue with this point, submitting the Authority has ignored standard errors, has failed to take account of the expected return on debt for high risk firms in portfolio 9, and has overlooked that the return on equity can be below the return on debt for long periods [DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Access Arrangement Period Supporting Submission: 56, 24 February 2016, p. 42]. However, the fact that the return on equity can be below the return on debt, ex post, simply amplifies the point that 'no rational investor invests in shares expecting decades of negative real returns ex ante, or expecting that bonds will outperform equities, yet these were actual outcomes. Thus differences between ex ante expectations and ex post outcomes are a major problem for tests of asset pricing models' [Partington, G. and Satchell, S., Report to the ERA: The Cost of Equity and Asset Pricing Models, May 2016, p. 7]. The Authority remains of the view that the outcomes for portfolio 9 highlight the extreme empirical problems of DBP's approach.
On balance, the Authority remains of the view that the findings of DBP's analysis are not robust and the approach produces nonsensical outcomes.
185 The ERA then moved on to consider cross validation issues and the question of consistency between debt and equity.
186 At pp 93 to 96 (pars 425 to 443) of Appendix 4 the ERA expressed its conclusions with respect to relevant models and information produced to it. The ERA said that it had significant concerns with the owners' estimate for the return on equity - both conceptual and empirical. In particular, the ERA expressed the view that:
(a) the model adequacy test approach, which sought to evaluate the forecasting power of various models of the return on equity, was not appropriate for the purpose of estimating the return on equity for regulatory purposes;
(b) the 'betastar' method (which was intended to transform the results of the Black CAPM into the SL-CAPM) was fraught with conceptual and empirical problems and could not be relied upon to meet the requirements of the NGL (WA) and the NGR because it:
(i) introduced a full quantum of ex post anomalous returns into the SL-CAPM beta term;
(ii) thereby introducing an adjustment for beta in the SL-CAPM which perversely introduced significant bias into what is an unbiased beta estimate; and
(iii) is ill-posed in mathematical terms (that is, is increasingly distorted) as the ZBP/MRP ratio approaches 1, raising serious questions about the veracity of the resulting return on equity;
(c) based on the Black CAPM, the owners' proposals suffered from the fact that estimates of the zero-beta return are unstable and cannot be relied upon in the Australian context. As a consequence, the ERA was not convinced that the Black CAPM is an acceptable pricing model for estimating the return on equity, given that the empirical implementation of the Black CAPM:
(i) is not robust. In contrast to the risk-free rate, the expected return on the zero beta asset is unobservable and there is no apparent consensus on methods for estimating this return;
(ii) relies upon average zero beta return estimated over more than 20 years of data, which typically results in estimates of the zero beta return, and the imputed zero beta premium, being less reflective of prevailing market conditions than the risk free rate estimates utilised in the SL-CAPM; and
(iii) is not widely used in practice. There is little evidence that other regulators, academics or market practitioners used the Black CAPM to estimate the return on equity.
187 At pp 95 to 96 (pars 437 to 443), the ERA said:
The Authority acknowledges that there is much debate about whether an adjustment needs to be made to the SL-CAPM. This was recognised by the Authority in the Guidelines and Draft Decision, with reference to the theoretical properties of Black CAPM. However, analysis since, by the Authority and its consultants, in response to DBP, has made the Authority concerned that it would likely be making a greater error by making an adjustment to the SL-CAPM - through alpha - than by making no adjustment. The Authority is not convinced such an adjustment would meet the allowed rate of return objective, or the requirements of the NGO or the RPP.
Accordingly, the Authority has determined to retain the use of the 'vanilla' SL-CAPM for this Final Decision, with the beta parameter based on the central, best estimate. Further, in light of the foregoing, no adjustment is made for alpha.
The Authority is satisfied that the resulting return on equity derived using the SL-CAPM is consistent with the allowed rate of return objective, and with the other requirements of the NGL and NGR. The Authority considers that the resulting SL-CAPM estimate for the return on equity:
• is reflective of economic and finance principles and market information;
• is fit for purpose, which is reflected in its broad acceptance in the finance industry as a means for estimating the cost of capital;
• can be implemented in accordance with good practice;
• is parsimonious, is not unduly sensitive to errors in inputs or arbitrary filtering, and is therefore difficult to game;
• uses input data that is credible and verifiable, comparable and timely and clearly sourced;
• is sufficiently flexible to allow for changing market conditions and new information to be reflected in regulatory outcomes, as appropriate.
In summary, the Authority determines the following for the purpose of estimating a return on equity in this Final Decision:
• The SL-CAPM will be utilised to estimate the return on equity.
• The Fama French (three factor) Model is not relevant and will not be used for the purpose of estimating a return on equity.
• The Black CAPM is relevant for informing the theory of the return on equity.
- However, given it is not reliable and practical to estimate a robust return on equity using this model, the model will not be used directly.
- Neither is it used indirectly. It is only used now to inform the theory of the return on equity.
- A revised consideration of the theoretical implications of the model makes clear that no adjustment to equity beta is appropriate. In addition, the Authority considers that there is no compelling evidence to apply an alpha adjustment to the return on equity determined by the vanilla CAPM, as a means to account for 'low beta bias' observed in ex post returns, at the current time.
• The DGM is a relevant model for informing the market return on equity and also the forward looking MRP.
• Other information such as historical data on equity risk premium; surveys of market risk and other equity analysts' estimates are also relevant for the purpose of estimating the MRP and the market return on equity. In addition, DBP's primary cross-check method is also accepted. This other material will be used as a cross check for the return on equity.
The Authority remains of the view that its reasons for adopting the SL-CAPM are sound. The Authority considers that its application of the SL-CAPM meets the requirements of the NGL and NGR, including the allowed rate of return objective.
Accordingly, the Authority considers that the estimated return on equity adopted in this Final Decision is commensurate with the equity costs incurred by a benchmark efficient entity with a similar degree of risk as DBP with respect to the provision of reference services. The Authority therefore considers that the estimated rate of return meets the allowed rate of return objectives and the requirements of the NGR and NGL.
In line with the requirements of NGR 87(5), the Authority has evaluated the relevance of a broad range of material for estimating the return on equity, covering relevant estimation methods, financial models, market data and other evidence for this Final Decision.
188 At pp 97 to 98 (pars 446 to 458) of Appendix 4, the ERA set out its reasoning as to why it had come to the view in the final decision that its estimate of the equity beta for use in the SL-CAPM is not biased. Those paragraphs are in the following terms:
Estimate of the equity beta
Following further evaluation of DBP's betastar claims, set out above, the Authority has determined that its estimate of the equity beta for use in the SL-CAPM is not biased. Accordingly, the Authority has determined that it will not adjust beta in determining the return on equity for this Final Decision. The task then is to determine the best, central estimate of beta.
Under the CAPM, the total risk of an asset is divided into systematic and non-systematic risk. Systematic risk is a function of broad macroeconomic factors (such as economic growth rates) that affect all assets and cannot be eliminated by diversification of the investor's asset portfolio.
The key insight of the CAPM is that the contribution of an asset to the systematic risk of a portfolio of assets is the correct measure of the asset's risk (known as beta risk), over and above the return on a risk free asset.
In contrast, non-systematic risk relates to the attributes of a particular asset. The CAPM recognises this risk can be managed by portfolio diversification. Therefore, the investor in an asset does not require compensation for this risk.
In the CAPM, the equity beta value is a scaling factor applied to the market risk premium, to reflect the relative systematic risk for the return to equity of the firm in question, as compared to the systematic risk for all assets. Two types of risks are generally considered to determine a value of equity beta for a particular firm: (i) the type of business, and associated capital assets, that the firm operates; and (ii) the amount of financial leverage (gearing) employed by the firm.
In the Rate of Return Guidelines, the Authority considered that empirical evidence provides the best means to inform its judgment for equity beta [Economic Regulation Authority, Explanatory Statement for the Rate of Return Guidelines: Meeting the Requirements of the National Gas Rules, December 2013, p. 161].
However, as discussed above (paragraphs 378 to 380, there is conceptual support for the equity beta of an infrastructure network benchmark efficient entity being less than 1 [See for example Australian Energy Regulator, Draft Decision Jemena (NSW), Attachment 3: Rate of return, November 2014, p. 3-235]. The Authority noted these views in the Draft Decision and considered that the reasoning is relevant [In the Draft Decision, the Authority noted DBP's view that model adequacy tests suggest that application of the SL-CAPM is not estimating what low beta firms 'actually earn for their equity investors' (Dampier Bunbury Pipeline, DBP Submission to ATCO Draft Decision, 7 January 2015, p. 3). However, the Authority considers that the evidence provided by DBP does not accord with the well accepted theoretical underpinnings of the CAPM, in that it suggests that as beta (systematic risk) declines, the equity risk premium increases. This raises significant issues for the DBP empirical analysis, and the underlying quality of the data that is used for that analysis. Similarly, the Authority considers that the points made by the ENA also refer to the same matters (Energy Networks Association, WA ERA Draft Decision for ATCO Gas ENA Response, 12 January 2015, p. 4). In particular, the evidence on the performance of SL-CAPM for low beta stocks evaluated by the ENA's consultant NERA utilises the same SIRCA database which is used by DBP (see NERA Economic Consulting, Estimates of the zero-beta premium, June 2013, p. 15). Furthermore, as a related point, the Authority does not consider that the four estimates cited by ENA are robust in the Australian context. At the current time, the Authority remains of the view that the conceptual foundation of the CAPM supports the estimates of the return on equity set out in this Final Decision.].
Nonetheless, the conceptual analysis does not provide sufficient grounds to establish the point value of the equity beta. To inform its decision on the point value, the Authority conducted a detailed empirical estimation of the required equity beta as part of the development of the Rate of Return Guidelines [Econometric analysis of beta was conducted in: Economic Regulation Authority, Explanatory Statement for the Rate of Return Guidelines, December 2013, Chapter 12. Justification and explanation for econometric techniques was provided in Economic Regulation Authority, Appendices to the Explanatory Statement for the Rate of Return Guidelines, December 2013, Appendix 17, 22 and 23].
In its Guidelines, the Authority evaluated the following issues in relation to the estimates of equity beta; including:
• the level of imprecision for any empirically estimated value of the equity beta [Economic Regulation Authority, Explanatory Statement for the Rate of Return Guidelines, 16 December 2013, p. 162.];
• a range of other issues, including those relating to sampling and instability; and
• that it was inappropriate to include overseas businesses in the comparator sample which was used to estimate the required equity beta of the benchmark efficient entity [Economic Regulation Authority, Explanatory Statement for the Rate of Return Guidelines, December 2013, p. 188].
The Authority noted in the Guidelines that it would update its estimate of beta at the time of each access arrangement decision [Economic Regulation Authority, Explanatory Statement for the Rate of Return Guidelines, 16 December 2013, p. 197].
For this Final Decision, the Authority will continue to estimate beta in the way that was set out in the Guidelines, albeit updated. Given the decision not to adjust beta - as set out in 'Step 1 - Identifying relevant materials' above - the Authority adopts the best, central estimate of beta.
The Authority notes that DBP states that it has no issue with the Authority's revised estimates for beta that were set out in the Draft Decision [DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Access Arrangement Period Supporting Submission: 56, 24 February 2016, p. 63]:
In respect of beta, DBP has no issue with the estimation of beta as undertaken by the ERA; when we use five years of weekly, end-of-the-week returns, we obtain roughly the same results the ERA does.
The Authority therefore takes DBP's broad acceptance of the beta material set out in the Draft Decision, apart from a number of issues which it raised in its response to the Draft Decision. For example [ibid]:
We do, however, have two issues in respect to beta:
(a) The estimate of beta the ERA has used of 0.7 produces a result which is not consistent with the approach it has used in the past, because it has failed to take into consideration the changes in its beta estimation.
(b) The second relates to a potential issue concerning the efficiency of the market portfolio.
189 At pp 99 to 100 (pars 459 to 466), the ERA set out its response to the submissions made by the owners in respect of beta. The ERA said:
First, the issue raised by DBP that the Authority's Draft Decision estimate of beta, of 0.7, was not consistent with the approach used in the past, refers to the Authority's previous practice of adjusting beta for so-called low beta bias. However, the Authority indicated in the Rate of Return Guidelines that it would review the evidence for beta adjustment [Economic Regulation Authority, Explanatory Statement for the Rate of Return Guidelines, 16 December 2013, p. 162]. On the basis of the review set out in this Final Decision, the Authority now considers that there is no evidence to support it making an adjustment to beta. Accordingly, the Authority adopts the central, best estimate of beta for this Final Decision. DBP's point has no bearing on the central, best estimate, as no adjustment is being made. Accordingly, it is not considered further.
However, relevant to that central estimate, DBP raises the issue of a structural break in the estimate of beta, around late 2014, for rolling three year betas, and for five-yearly betas in April 2012 (value-weighted portfolios) and September 2013 (equal weighted portfolios) [DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Access Arrangement Period Supporting Submission: 56, 24 February 2016, p. 65]. DBP contends that [DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Access Arrangement Period Supporting Submission: 56, 24 February 2016, p. 64]:
The changing confidence interval in the ERA's own analysis points to a deeper issue in respect of beta. That is, beta appears to be changing, and changing substantially, over the past twelve months. Figure 5 provides a comparison of rolling three and five-year betas over the past several years.
However, it is not exactly clear what DBP's point is [ibid]:
Both beta calculations give roughly similar results until around the end of October 2014. From the end of that date, both begin to trend upwards (as do the ERA's results), but the three-year betas trend upwards much more sharply. For a value-weighted portfolio, the mean beta estimate, before an adjustment for bias (like the ERA's choice of 0.7 is around 0.95) and, as CEG points out, the lower bound of the 95 percent confidence interval for a three-year beta is, at present, above the bias-adjusted figure that the ERA uses for beta.
The Authority takes the point that the beta is changing, and that there is then a question of the appropriate averaging period. The Authority notes conflicting views on this topic. For example, SFG submitted to the Authority that it considers it 'implausible' that equity beta estimates could change over a two year period [ATCO Gas Australia, ATCO Gas Australia's Response to the ERA's Draft Decision, 22 December 2014, Appendix 9.1, p. 9]. However, the rolling beta estimates produced by the Authority in the Guidelines convinced it that, for individual firms, the relative sensitivity to systematic risk can vary quite dramatically [Only HDF falls outside the estimated range]. The Authority has no reason to believe that this does not reflect a re-rating by the market of the respective firms, in terms of risk relative to the market.
Therefore, the Authority considers there is no issue with the fact that the measure of beta does change over time - that is exactly why it undertook, in the Rate of Return Guidelines, to update its estimate of the beta just prior to its Final Decision [Economic Regulation Authority, Explanatory Statement for the Rate of Return Guidelines, 16 December 2013, p. 197]. That said, the Authority considers that five years provides an appropriate estimation period which smooths out short term fluctuations in beta. The Authority does not consider it appropriate to depart from its current practice, of placing emphasis on the five year estimates, simply because the data suggests that a higher estimate of beta could be obtained by using a shorter averaging period. Nonetheless, the Authority notes that information.
Second, the Authority dealt with the issue of the efficiency of the market portfolio at paragraph 324 to 326 above.
DBP also argued that the Authority's approach to selecting beta based on confidence intervals is flawed. DBP argued that [DBP, Proposed Revisions DBNGP Access Arrangement 2016 - 2020 Regulatory Period Supporting Submission: 60 Response to Australian Competition Tribunal Decisions, 22 March 2016, pp. 8-9]:
In essence, confidence intervals tell one something about the precision of a parameter estimate (such as of beta) within a given model. However, it tells one nothing about the performance of that model itself. A model can perform very poorly but still have very precisely estimated parameters. The issue of low-beta bias is not a problem in the estimation of beta per-se. As DAA point out, that can be improved simply by increasing sample size. The issue is rather that the outputs of the model produce results which are systematically wrong; too low where the beta of a stock is below one and too high when the beta of a stock is above one.
However, the Authority does not accept DBP's point here, for the reasons set out under 'Step 1' above. DBP is wrong to criticise the Authority's beta estimates on a basis that is, at its essence, a point about model adequacy. The Authority has rejected DBP's 'model adequacy test' approach to estimating the return on equity. However, the Authority notes that DBP considers that 'the issue of low-beta bias is not a problem in the estimation of beta per-se'.
190 At pp 100 to 103 (pars 467 to 478) of Appendix 4, the ERA set out its updated estimates of beta for the purposes of its final decision. In those paragraphs, the ERA said:
The Authority's updated estimates of beta for this Final Decision
The following Table 1 reports a range of estimates of Australian infrastructure betas from various sources, with an emphasis on the most relevant and recent.
Source: The AER's Draft Decision for ActweAGL Distribution Determination, Table 3-55, page 3-262 and the ERA's 2015 study (Economic Regulation Authority, Draft Decision on Proposed Revisions to the Access Arrangement for the Dampier to Bunbury Natural Gas Pipeline, 22 December 2016, Appendix 4Aii, p. 233 - 234.
The detail of the Authority's 2015 study set out in the table was reported at Appendix 4Aii of the Draft Decision.
The Authority noted in the Draft Decision it considered that the 95 per cent confidence interval for the beta estimate was 0.3 to 0.8 [Economic Regulation Authority, Draft Decision on Proposed Revisions to the Access Arrangement for the Dampier to Bunbury Natural Gas Pipeline, 22 December 2016, Appendix 4, p. 52]. The Authority then determined a point estimate for beta at 0.7, allowing for some adjustment towards the top end of the range to account for the theory underpinning the Black CAPM.
DBP contends that the Authority's approach to calculating averages has the effect of artificially lowering the range for beta and does not accord with the Henry approach considered by the Tribunal. DBP states [ibid]:
Specifically, the ERA makes four estimates for the individual firms and then two different portfolio estimates. The upper and lower bounds of 0.81 and 0.41 (respectively) are formed by averaging across the six upper bounds of confidence intervals for the LAD regression estimates and the six lower bounds of the LAD regression estimates (the LAD estimates exhibit the widest range - see DDA4 Table 29, page 194). By contrast, Henry (2014) does not mix portfolio and individual estimates in this way, and reports his ranges as the minimum and maximum of the confidence intervals for each set of regressions, rather than the averages across lower and upper bounds.
Three of the four firms examined by the ERA (APA, AST and SKI) give similar results to the portfolio results, generally, but one (DUET) gives results which are substantially below the other three firms, and the portfolios. By forming the averages in the way that it has, the ERA has effectively given disproportionate weight to DUET, and has dragged down its averages accordingly.
The Authority accepts these points in principle, rather than substance (there are issues with the numbers DBP quotes, and its subsequent inferences). The Authority takes account of those points in what follows.
The Authority's 2016 estimates
For this Final Decision, the Authority had Pink Lake Analytics further update the beta estimates for this Final Decision, in part to address the data issues that had been raised by HoustonKemp [Pink Lake Analytics, Variance of the ZBP estimator, June 2016, Appendix G, p. 59]. The detailed results are reported at Appendix 4A, with a key table reproduced here (Table 2).
Drawing on the results reported in Appendix 4A, the Authority considers that a 95 per cent confidence interval range of equity beta using the most recent data is from 0.479 and 0.870 based on the portfolio results (see Appendix 4A, Table 21 and Table 22). The central estimate given by the average of the portfolios is 0.699. The Authority notes that portfolio estimates have a narrower range than the individual assets.
Based on its own analysis and the other evidence before it, together with the recognition that estimates of equity beta from empirical studies exhibit a high level of imprecision, the Authority is of the view that the point estimate of equity beta of 0.7 (rounded) provides a conservative and appropriate central best estimate for beta for use in the SL-CAPM.
Conclusions with regard to equity beta
Based on the above considerations, the Authority is of the view that available Australian estimates of equity beta are reliable and that the estimates from these studies should be used to determine an appropriate equity beta for a network service provider.
The Authority considers that available estimates of equity beta in Australia, including Henry's studies and the Authority's own analyses, as presented in Table 1 and Table 2 above, as well as submission material from DBP, indicate a best equity beta estimate of (a rounded) 0.7. Rounding the estimate to one significant figure accounts for the acknowledged imprecision of the estimate.
That estimate gives greatest weight to the Authority's 2016 estimates of equity beta - using data for the most recent 5 years.
On balance, the Authority remains of the view that it is appropriate to account for a range of evidence in its determination of the equity beta point estimate. Based on its considerations outlined above, the Authority has determined to adopt the estimate of equity beta of 0.7 for this Final Decision for the DBNGP.
191 The conclusions expressed by the ERA at pp 100 to 103 (pars 467 to 478) were supported by detailed calculations in Appendix 4A. Appendix 4A reported updated estimates for beta for use in the SL-CAPM. The ERA noted (at p 190 (par 915)) that the ERA engaged Pink Lake Analytics to assist with its analysis.
192 At p 193 (par 930) of Appendix 4A, the ERA produced Table 20. That Table was in the following form:
193 At p 193 (pars 930 to 931) of Appendix 4A, the ERA said:
Table 20 reports estimates of each firm's beta across the different regression methodologies, with a data set from June 2011 to May 2016. Equally-weighted and value-weighted portfolios are also reported.
The point estimate of 𝛽 for purposes of the Authority's RoE evaluation is taken fro the mean 𝛽, averaged across the two weighted portfolios and the OLS, LAD, MM and T-S estimators. This results in a 𝛽 = 0.699, rounded to 𝛽 = 0.7 (Table 20)
194 In order to establish Ground 1, the owners must satisfy the Tribunal that the ERA has committed a reviewable error within the meaning of s 246 of the NGL(WA) and that, if that ground is made good any likely reconsidered decision would, or would be likely to, result in a material preferable designated NGO decision.
195 We now move to consider the owners' arguments in relation to Ground 1.