AUSTRALIAN COMPETITION TRIBUNAL
IN THE AUSTRALIAN COMPETITION TRIBUNAL
IN THE MATTER OF APPLICATION UNDER SECTION 71B OF THE NATIONAL ELECTRICITY LAW FOR A REVIEW OF A DISTRIBUTION DETERMINATION MADE BY THE AUSTRALIAN ENERGY REGULATOR IN RELATION TO AUSNET ELECTRICITY SERVICES PTY LTD PURSUANT TO CLAUSES 6.11.1 AND 11.60.4 OF THE NATIONAL ELECTRICITY RULES
MR RF SHOGREN (MEMBER)
DR DR ABRAHAM (MEMBER)
THE TRIBUNAL DETERMINES THAT:
THE TRIBUNAL NOTES: the AER’s error in calculation identified at  of its reasons in Application by ActewAGL Distribution  ACompT 2 and, in accordance with  of those reasons, leaves it to the AER to determine the appropriate response to its error.
1 These reasons concern two issues which involve only one distribution network service provider (DNSP), AusNet Electricity Services Pty Ltd (AusNet). They arise from the AER’s final decision and distribution determination entitled Final Decision: AusNet Services distribution determination 2016 to 2020 (Final Decision), published on 26 May 2016.
2 These reasons are to be read with the reasons in Application by ActewAGL Distribution  ACompT 2. Those reasons provide the relevant background.
3 The two issues to which we have referred are (i) return on debt – Bloomberg Valuation Service (BVAL) curve and (ii) forecasting opex – self-insurance.
4 The basis of the Tribunal’s powers of review is as follows.
5 Section 71C of the National Electricity Law (NEL) provided:
71C Grounds for review
(1) An application under section 71B(1) may be made only on 1 or more of the following grounds:
(a) the AER made an error of fact in its findings of facts, and that error of fact was material to the making of the decision;
(b) the AER made more than 1 error of fact in its findings of facts, and that those errors of fact, in combination, were material to the making of the decision;
(c) the exercise of the AER's discretion was incorrect, having regard to all the circumstances;
(d) the AER's decision was unreasonable, having regard to all the circumstances.
(1a) An application under section 71B(1) must also specify the manner in which a determination made by the Tribunal varying the reviewable regulatory decision, or setting aside the reviewable regulatory decision and a fresh decision being made by the AER following remission of the matter to the AER by the Tribunal, on the basis of 1 or more grounds raised in the application, either separately or collectively, would, or would be likely to, result in a materially preferable NEO decision.
(2) It is for the applicant to establish a ground listed in subsection (1) and the matter referred to in subsection (1a).
6 The “national electricity objective” (NEO) set out in s 7 stated as follows:
7 National electricity objective
The objective of this Law is to promote efficient investment in, and efficient operation and use of, electricity services for the long term interests of consumers of electricity with respect to--
(a) price, quality, safety, reliability and security of supply of electricity; and
(b) the reliability, safety and security of the national electricity system.
7 The expression “materially preferable NEO decision” was defined in s 71A with reference to s 71P(2a)(c) as follows:
71P Tribunal must make determination
(1) If, following an application, the Tribunal grants leave in accordance with section 71B(1), the Tribunal must make a determination in respect of the application.
See section 71Q for the time limit within which the Tribunal must make its determination.
(2) Subject to subsection (2a), a determination under this section may--
(a) affirm the reviewable regulatory decision; or
(b) vary the reviewable regulatory decision; or
(c) set aside the reviewable regulatory decision and remit the matter back to the AER to make the decision again in accordance with any direction or recommendation of the Tribunal.
(2a) Despite subsection (2), the Tribunal may only make a determination--
(a) to vary the reviewable regulatory decision under subsection (2)(b); or
(b) to set aside the reviewable regulatory decision and remit the matter back to the AER under subsection (2)(c),
(c) the Tribunal is satisfied that to do so will, or is likely to, result in a decision that is materially preferable to the reviewable regulatory decision in making a contribution to the achievement of the national electricity objective (a “materially preferable NEO decision”) (and if the Tribunal is not so satisfied the Tribunal must affirm the decision); and
(d) in the case of a determination to vary the reviewable regulatory decision--the Tribunal is satisfied that to do so will not require the Tribunal to undertake an assessment of such complexity that the preferable course of action would be to set aside the reviewable regulatory decision and remit the matter to the AER to make the decision again.
(2b) In connection with the operation of subsection (2a) (and without limiting any other matter that may be relevant under this Law)--
(a) the Tribunal must consider how the constituent components of the reviewable regulatory decision interrelate with each other and with the matters raised as a ground for review; and
(b) without limiting paragraph (a), the Tribunal must take into account the revenue and pricing principles (in the same manner in which the AER is to take into account these principles under section 16); and
(c) the Tribunal must, in assessing the extent of contribution to the achievement of the national electricity objective, consider the reviewable regulatory decision as a whole; and
(d) the following matters must not, in themselves, determine the question about whether a materially preferable NEO decision exists:
(i) the establishment of a ground for review under section 71C(1);
(ii) consequences for, or impacts on, the average annual regulated revenue of a regulated network service provider;
(iii) that the amount that is specified in or derived from the reviewable regulatory decision exceeds the amount specified in section 71F(2).
(2c) If the Tribunal makes a determination under subsection (2)(b) or (c), the Tribunal must specify in its determination--
(a) the manner in which it has taken into account the interrelationship between the constituent components of the reviewable regulatory decision and how they relate to the matters raised as a ground for review as contemplated by subsection (2b)(a); and
(b) in the case of a determination to vary the reviewable regulatory decision--the reasons why it is proceeding to make the variation in view of the requirements of subsection (2a)(d).
(3) For the purposes of making a determination of the kind in subsection (2)(a) or (b), the Tribunal may perform all the functions and exercise all the powers of the AER under this Law or the Rules.
(5) A determination by the Tribunal affirming, varying or setting aside the reviewable regulatory decision is, for the purposes of this Law (other than this Part), to be taken to be a decision of the AER.
8 We refer to the decision of the Full Court in Australian Energy Regulator v Australian Competition Tribunal (No 2)  FCAFC 79; 345 ALR 1 for a general review of the content of the grounds of (limited) merits review in s 71C of the NEL.
9 We note in particular what the Full Court said at  and - about “findings of fact”, as follows:
… the AER submitted that when the Tribunal considered whether there was an “error of fact”, the concept of “error” set a threshold that was not passed simply because there was material that could support a different finding of fact or simply because the Tribunal might, if it considered the material afresh, prefer to make a different finding of fact: DBNGP at  and WA Gas Networks at . Thus far we agree.
… we accept the submission on behalf of the electricity network respondents that the concept of ‘findings of fact’ encompasses more than ‘primary’ fact finding. In our opinion, findings of facts within s 71C would generally include inferences drawn from primary facts or conclusions drawn from primary facts. We would also observe that we would not regard ‘evaluative judgment’ as an appropriate discrimen for the purposes of s 71C. In our opinion, depending on the particular circumstances, ‘findings of fact’ may include such a finding even where it is based on an ‘evaluative judgment’.
… In our view the touchstone for s 71C material error is that the Tribunal may not act where there is no error and error is not made out by choosing one available fact or opinion over others. For example, the mere availability of a different inference from facts would not establish error within s 71C: the Tribunal must find that there was error in the AER’s decision-making before it may intervene.
10 At page 3-115 of Attachment 3 to the Final Determination, the AER said:
We are satisfied that using a third party data series (or multiple series), appropriately chosen, is commensurate with the efficient debt financing costs of a benchmark efficient entity. It is also consistent with the rule requirement that the change in revenue (resulting from the annual debt update) is effected through the automatic application of a formula that is specified in the determination. This is because:
• A third party data series can be practically applied in the annual debt update process—We discuss this point further below.
• A third party data series is independent information developed by finance experts with access to financial datasets—These experts develop this independently from the regulatory process and for the use of market practitioners.
• Using a third party data series also reduces the scope for debate on debt instrument selection and curve fitting—For instance, independent data service providers have already exercised their judgement on bond selection, curve fitting and adjusting yields. However, we still must exercise our regulatory judgement to assess which third party data series (or combination of series) is better suited for contributing to the achievement of the ARORO.
11 At pages 3-118 and following of Attachment 3 to the Final Determination, the AER said, omitting footnotes:
Choice of third party data series (including adjustments)
In the previous section, we explained our decision is to use third party published data series to estimate the allowed return on debt, rather than deriving our own data series. In this section, we explain our choice of third party data series, including adjustments we have decided to make to those data series.
Our decision is to adopt a simple average of the debt data series published by the RBA and Bloomberg that match, as close as available, our benchmarks of a BBB+ credit rating and a 10 year debt term. Specifically, our decision is to adopt a simple average of:
• the 10 year estimate from the non-financial corporate BBB rated data series published by the RBA (the RBA curve), and
• the 10 year yield estimate from the Australian corporate BBB rated Bloomberg Valuation Service (BVAL) data series published by Bloomberg (the BVAL curve).
The RBA and BVAL curves are both 'broad BBB' rated data series in that they reflect bond pricing generally across the BBB+, BBB and BBB- rated spectrum of bonds.
Our decision is also to make certain adjustments to the RBA and BVAL curves so these rates are consistent with our 10 year benchmark debt term and also so they can be applied across the dates of a service provider's averaging periods. Those adjustments are:
For the RBA curve, to extrapolate the data series from a 'target' 10 year term to an 'effective' 10 year term using the method recommended by Dr Lally (the Lally method), to interpolate the monthly data points to produce daily estimates, and to convert the estimates from a semi-annual to an effective annual rate.
• For the BVAL curve, to convert the estimates from a semi-annual to an effective annual rate.
• The above positions are consistent with the approach we adopted in the first round of decisions since the publication of the Guideline, the most recent being our decisions released in November 2015.
We are satisfied that a simple average of the two curves will result in a return on debt that contributes to the achievement of the ARORO. This is because:
• Based on analysis of the bond selection criteria (including approach for identifying outliers), we consider that both approaches employed by the RBA and Bloomberg have their unique strengths and weaknesses, but we are not satisfied that either is clearly superior.
• Based on analysis of the curve fitting (or averaging) methodologies, we consider that both approaches have their unique strengths and weaknesses, but we are not satisfied that either is clearly superior.
• Both curves require adjustments from their published form to make them fit-for-purpose, and we are not satisfied that either can be more simply or reliably adjusted to estimate the annual return on debt.
• A simple average is consistent with expert advice from Dr Lally that we adopt a simple average of the BVAL curve and the RBA curve, subject to the necessary adjustments to each curve. In particular, Lally concluded that based on analysis of the curves, it was reasonably likely that a simple average of the two curves would produce an estimator with a lower mean squared error (MSE) than using either curve in isolation. Lally also advised:
…on the question of which index better reflects the cost of debt for the efficient benchmark entity, there is no clear winner.
• The two curves have regularly produced materially different results at particular points in time. Both curves have their strengths and shortcomings, but it is not clear to us that one approach is clearly superior. Consequently, when the curves depart, we consider it is not easily discernible which curve produces estimates that better reflect the efficient financing costs of a benchmark efficient entity. We also note that the BVAL curve has produced estimates both higher than, lower than, and similar to, the RBA curve, depending on the particular point in time. So there is no clear indication that one curve produces systematically higher or lower estimates than the other.
• A simple average of two curves, in these circumstances, is consistent with the Tribunal's decision in the ActewAGL matter where the Tribunal concluded that:
…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.
• A simple average of the two curves will reduce the likely price shock if either curve becomes unavailable or produces erroneous estimates during the period.
We have assessed the new information received in current proposals from service providers who recommend that we depart from our previous position of adopting a simple average of the RBA and BVAL curves. That new information does not persuade us to depart from our position or reasons from recent decisions. We explain our reasons for this decision in the remainder of this section and in Appendix I.
We also requested Dr Lally review the recommendations from his previous report in light of the material submitted by service providers with current proposals. As part of that analysis, we requested Dr Lally review both the AER's approach and the various approaches proposed by service providers with current proposals against a set of criteria drawn from the requirements of the law and the rules, including the ARORO. After reviewing that material, Dr Lally concluded:
…the AER’s proposed approach satisfies the criteria and these criteria are not satisfied by any other proposed approach.
Finally, I have previously provided advice on these implementation issues to the AER and nothing in these submissions warrants any change in that advice.
This is by no means a complete extract of the AER’s reasoning.
12 In Appendix I, the AER set out its more detailed analysis on its responses to issues raised by key stakeholders, including: the Reuters curve; new criticisms of the BVAL curve; and other issues. It said, at page 3-320, in relation to its detailed analysis of the key issues raised by stakeholders relating to the implementation of the return on debt approach:
Overall, we consider that much of this analysis indicates the shortcomings of annual testing or re-evaluation of the choice of data series at each decision stage. Much of the new analysis submitted prior to this decision is based on ex-post analysis of the performance of yield curves over a short averaging period. In contrast, we require service providers to specify averaging periods ahead of time because this avoids the introduction of bias into the analysis that can arise where the outcome of the choice is already known. Especially where this analysis is highly qualitative, we are not satisfied that this process can be automatically implemented by application of a formula.
Importantly, we must be satisfied that the approach that we adopt will continue to contribute to estimates which achieve the allowed rate of return objective for each annual update over the five year regulatory period. For this reason, we consider it is most critical that the approach we adopt is reasonable and fit for purpose over an extended period. At any points in time, it is possible that one curve or the other will better reflect the costs of the benchmark entity. However, for the reasons set out in this and previous decisions, we are not satisfied that there is a robust means to quantitatively identify which of the curves this is using the information currently before us. For this reason, our detailed analysis and expert advice focused on the underlying characteristics of the curve, their fitness for purpose, and their representativeness of the benchmark efficient entity. Having done so, we remain satisfied that a simple average of the BVAL and RBA curves will contribute to an estimate that achieves the allowed rate of return objective.
13 The AER said that in Applications by Public Interest Advocacy Centre Ltd and Ausgrid  ACompT 1 (Ausgrid) at , the Tribunal had addressed the question whether the AER should estimate the allowed return on debt using the RBA data series alone or a simple average of the RBA and Bloomberg data series. The Tribunal upheld the AER’s decision and found that, “averaging of the two curves was an acceptable measure of the DRP”.
14 The relevant statutory provision was in the following terms:
Return on debt
6.5.2(h) The return on debt for a regulatory year must be estimated such that it contributes to the achievement of the allowed rate of return objective.
(i) The return on debt may be estimated using a methodology which results in either:
(1) the return on debt for each regulatory year in the regulatory control period being the same; or
(2) the return on debt (and consequently the allowed rate of return) being, or potentially being, different for different regulatory years in the regulatory control period.
15 AusNet submitted that in deciding to estimate the return on AusNet’s debt from an average of two yields, the yield charted by the BVAL curve at 10 years and the yield charted by the RBA curve at 10 years, the AER had made errors of fact, exercised its discretion incorrectly, and acted unreasonably.
16 In summary, AusNet submitted that under r 6.5.2(h) of the NER, properly construed, the return on AusNet’s debt must be estimated such as to be commensurate with the efficient financing costs of a pure-play, regulated energy-network business operating within Australia, which borrows money for 10 years at a time. The yield charted by the BVAL curve at 10 years was not commensurate with the efficient financing costs of such a business for three reasons.
17 First, the spread to the swap rate implied in the BVAL curve at 10 years was determined virtually exclusively by just one bond, the Asciano bond. That bond matured in 8.5 years. None of the other bonds underlying the BVAL curve matured in more than 6 years. Of the bonds underlying the RBA curve, by contrast, 11 matured in 8–12 years.
18 Secondly, the return on 80% of AusNet’s debt, over the regulatory control period 2016–20, would be extrapolated from the averaging period 25 January to 19 February 2016. During that period, Asciano was the target of two takeover bids. Those bids heavily influenced the yield on the Asciano bond, but would not have influenced the efficient financing costs of a pure-play, regulated energy-network business operating within Australia.
19 Thirdly, about 83% of the money borrowed by such a business was borrowed from overseas. Yet none of the money borrowed under any of the bonds underlying the BVAL curve was borrowed from overseas. By contrast, 74% of the money borrowed under the bonds underlying the RBA curve was borrowed from overseas.
20 In oral submissions, Senior Counsel for Ausgrid submitted as follows.
21 The evidence disclosed the Asciano bond as a disproportionate influence on the BVAL curve at 10 years. That had two important consequences. The first was that factors specific to the Asciano bond were virtually exclusively setting the level and driving the movement of the BVAL curve at 10 years. Secondly, that obscured movements in the wider financial market. In the result, the BVAL curve at 10 years did not reflect the efficient financing costs of the benchmark efficient entity. It reflected the actual financing costs of Asciano as they were perceived to be during AusNet’s averaging period of 25 January to 19 February 2016 at a time when Asciano was subject to two takeover bids by larger more diversified firms. This meant that the market perception of the risk associated with Asciano was low. Again, that was not relevant to the benchmark efficient entity.
22 The AER’s assessment of these curves was only carried out once every five years, around the beginning of the regulatory control period. If a simple average of two curves was used and one curve was performing well and one was performing badly as an indicator, then the one that was performing badly would blunt the one that was performing well, and that would produce a result which was not consistent with the allowed rate of return objective. A better outcome was to rely on the one that was performing well.
23 Curves performed differently at different points in time and could, therefore, perform differently and have a different impact depending on the averaging period. Here, AusNet had identified concerns of a different type in relation to the BVAL curve, that is, that it was lagging behind market conditions and not reflective of them. For AusNet’s averaging period, there was only one long dated bond in the BVAL curve in a relatively small sample of bonds under which all of the debt was raised in Australia, when the benchmark efficient entity raised at least 80 per cent of its funds overseas. Those concerns were, therefore, exacerbated in relation to the BVAL curve.
24 Then there was the aberrant performance of the BVAL curve during AusNet’s averaging period, attributable to the two takeover offers and the status of those offers. There were very good reasons, demonstrated by statistical analysis by CEG, for discarding the BVAL curve altogether in the case of AusNet. That did no violence to the DNSPs as a group because their averaging periods were different. And because the AER in allowing different averaging periods must shoulder the burden of considering each distributor individually.
25 Turning to the alternative position of adopting the BVAL seven year estimate extrapolated out to 10 years, the AER’s fallback position, the way in which the seven year BVAL curve was to be extrapolated out to 10 years was to use the 7 to 10 year curve from the RBA curve and adjust the BVAL curve by taking that component of the RBA curve. That in itself, it was submitted, demonstrated the inappropriateness of relying on the BVAL curve. Only one bond had a greater duration than 6.5 years, and it was, therefore, necessary to go to the RBA curve to make adjustments to it if the BVAL curve was taken to be 7 years.
26 As to the errors involved in the decision, AusNet relied on the grounds of review set out at - of its application. In relation to error of fact, AusNet’s case was that the AER found that the BVAL curve was fit for purpose when it averaged it with the RBA curve. The submission was that that was an error of fact. The AER should have found the BVAL curve was not fit for purpose and, therefore, not used it. That was the case, firstly, because of the influence of the Asciano bond; secondly, because of circumstances pertaining to the Asciano bond, being the takeover offers; and thirdly as to the source of debt, that is, 80 per cent of the funds of the benchmark efficient entity being borrowed overseas 10 years at a time, which was not reflected in the BVAL curve. AusNet also relied on the great disparity in bond population, 22 bonds in the BVAL curve versus 85 in the RBA curve. There was only one over six years in the BVAL curve, and 11 in the 8 to 12 year tenor in the RBA curve. Therefore, it was submitted, a simple average of the two curves constituted an error of fact. In relation to incorrect exercise of discretion and unreasonableness, AusNet submitted that the AER incorrectly exercised its discretion in conducting the average of the two curves for the reasons already outlined. In relation to unreasonableness, AusNet submitted that the AER’s decision lacked reason, was unreasoned, and, further, was unreasonable in the result.
27 In relation to remedy, AusNet submitted the Tribunal should set aside the decision or vary it. The Tribunal should not remit the matter to the AER but should correct the decision itself.
28 Senior Counsel for AusNet referred to Application by ActewAGL Distribution  ACompT 4 (ActewAGL) at , Application by Jemena Gas Networks (NSW) Ltd (No 5)  ACompT 10 at -,  and , and Application by Envestra Limited (No 2)  ACompT 4 at - and submitted that Ausgrid at - and Application by Jemena Gas Networks (NSW) Ltd  ACompT 5 at  were distinguishable.
29 The AER submitted, in summary, that AusNet’s application relied on a number of matters as supporting a conclusion that the AER erred in using the BVAL curve in its return on debt estimate.
30 The AER considered AusNet’s submissions, notwithstanding that the submissions were filed outside time and close to the time for the AER to make a final decision. The AER decided that each of the RBA and BVAL curves had strengths and weaknesses and that a simple average of the two curves was the preferable course. That decision was consistent with what was said in ActewAGL at  and was also endorsed in Ausgrid at .
31 In the earlier of these two decisions, ActewAGL, the Tribunal said, at :
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.
32 In the later decision, Ausgrid, the Tribunal said, at :
In the Tribunal’s view, whilst there are arguments for the sole use of the RBA curve, it has not been shown that – for the purposes of estimation of the return on debt – any ground of review has been made out. The AER had a choice to make as to what data services, or combination of data services, it should use. Its reasons for selecting the combination of data services are cogent, and reasonable. It is not shown to have misunderstood or overlooked material information. Although there are facts underlying the choice of the AER, the Tribunal is not persuaded of any particular material factual finding which is different from those made by the AER. For the purposes of the relevant Final Decisions, the AER does not positively find that the RBA curve was clearly superior to the BVAL curve, so that its averaging of the two curves was an acceptable measure of the DRP. The Tribunal is not satisfied that, on the material, the AER should have exercised its discretion to select either the RBA curve only, or some other formula for the estimation of the return on debt. Consequently, the Tribunal is not satisfied that the AER made an irrational decision in this respect.
33 As to the lack of foreign currency bonds in the BVAL curve, the AER submitted that:
(a) that matter was never raised by AusNet as an issue in the regulatory process and the issue was therefore excluded by s 71O of the NEL;
(b) regardless, the issue was considered by the AER and the AER’s decision did not involve error, particularly having regard to the fact that the RBA curve was likely to be overweight with regard to foreign currency bonds in comparison to a suitable benchmark.
34 As to the Asciano bond in the BVAL curve, the evidence did not support a conclusion that the 10 year spread to swap implied by the BVAL curve was inappropriately low and, in the relevant period, the BVAL curve generated more typical results than the RBA curve.
35 As to the other matters raised by AusNet, the AER submitted that the strengths and weaknesses of both curves were considered by the AER and it concluded that neither curve was clearly superior. That was a reasonable opinion to arrive at.
36 If the Tribunal found that there was error in the AER’s decision, it submitted:
(a) the Tribunal should nevertheless affirm the AER’s decision because the decision was already conservatively biased towards AusNet;
(b) alternatively, the Tribunal should implement the alternative approach foreshadowed by the AER, to use the BVAL 7 year estimate extrapolated to 10 years.
37 In oral submissions, Senior Counsel for the AER submitted that AusNet relied on five matters as giving rise to an error of fact. The same five matters were also identified as the matters that gave rise to an incorrect exercise of discretion and/or unreasonableness.
38 The first matter was the point relating to overseas borrowings, particularly the foreign currency of the bonds, and a proposition was put that about 80 per cent of money borrowed by a benchmark efficient entity was borrowed overseas. That fact was contentious. It was also contentious whether and when that issue was raised by AusNet in this regulatory process. The second and third points both related to the Asciano bond: the second was the proposition that the spread to swap rate implied by the BVAL curve of 10 years was determined virtually exclusively by the Asciano bond. The ancillary and third point was that Asciano had been the subject of recent takeover and, therefore, it was asserted by AusNet, the pricing of its bond would have been affected by that takeover. The fourth point was a sample size point. It was a point relating to the fact that the RBA curve was derived from 85 bonds; whereas the BVAL curve was derived from 22 bonds. AusNet’s fifth and related point was that, by reason of the underlying sample, the RBA curve was clearly superior.
39 The AER submitted the claim was that the AER’s use of the simple average of the two curves gave rise to an error of fact. The point was that the RBA curve was representative of the efficient financing costs of a benchmark efficient entity when the BVAL curve was much less so, and that was said to be an error of fact. The AER submitted that that point was not properly characterised as an error of fact. It submitted that that opinion it formed about the relative merits of the curves and the extent to which they were reflective of the efficient financing costs of a benchmark efficient entity, and whether one was superior or the other, was really a qualitative opinion and one about the satisfaction of the rule criterion. The rule criterion was the criterion of efficient financing costs of a benchmark efficient entity. The AER accepted that it may be possible to identify matters of fact that underlie the formation of the opinion, and those matters of fact could be identified to see if there was any error in an underlying finding of fact. Here there was no real relevant dispute between the parties about underlying matters of fact. The only available grounds were wrong exercise of discretion and/or unreasonableness. It was properly reviewed as a question of the reasonableness of the view that the AER came to in making its assessment.
40 The finding that the AER “still considered that the BVAL curve was fit for purpose” was, essentially, a qualitative opinion because “fit for purpose” can only take its meaning from the raw criterion. The purpose was for seeking to estimate the efficient financing costs of a benchmark efficient entity. Once one identified the sort of matters that Dr Lally identified and the AER took into account, it was very much an evaluative exercise which must be a judgement or an opinion. An assessment of fit for purpose was not properly a finding of fact but a judgement. The proper task for the Tribunal to review the AERs decision was, essentially, to assess whether it was reasonable. AusNet had not said much about incorrect exercise of discretion. Neither in its written submissions nor in its oral submissions had AusNet articulated what aspect of the exercise of discretion was incorrect.
41 As to the five points referred to at  above, the AER submitted as follows.
42 As to the first point, it was not raised before the final decision in the regulatory process. Indeed, to the contrary, AusNet endorsed using the Reuters curve in its revised regulatory proposal, knowing that the Reuters curve only included Australian currency bonds. Far from raising the issue, AusNet’s submissions to the AER were really almost to the contrary to the point it now took. That was a s 71O point. As to the substance of the point, the AER submitted the point did not identify error on the part of the AER. That was because the AER had previously given attention to this issue and had formed a view that the Bloomberg curve was, nevertheless, representative of the benchmark efficient entity. Where this was given attention by the AER was Dr Lally’s November report. In that report Dr Lally observed that Australian corporates drew debt financing from three primary sources, local bank borrowing, local bond issues and foreign bond issues. Amongst Australian regulated energy businesses a report from PwC in 2013 estimated the 2012 proportions at approximately 25 per cent local bank borrowing, 50 per cent local bond issues and 25 per cent foreign bond issues respectively. And Dr Lally then went on to explain the reasons for this mix, and there were a number of reasons why that mix was, apparently, seen as observed amongst Australian energy regulated businesses. AusNet relied for its contention of 80 per cent foreign bonds on the CEG Jemena memoranda, which made some reference to this figure. So there was no doubt CEG had a different view. The AER submitted there were both pros and cons from the inclusion of foreign currency bonds. There was material before it that supported its conclusion. It was contrary to CEG’s claim that 80 per cent of long-term debt was raised by a benchmark efficient entity and foreign currency. Nevertheless, the AER was entitled to rely on the material that it had before it on that point. The AER submitted no error arose in that respect.
43 As to the second and third points, the AER undertook an analysis of the impact of the Asciano bond and, more relevantly, analysed what the Bloomberg curve was demonstrating and showing through the relevant period. It asked whether there was a proper basis to conclude that the Bloomberg curve could not be relied upon to give an appropriate estimate of the efficient financing costs of a benchmark efficient entity, and whether it should be jettisoned. At that time there would only be one curve available to the AER, being the RBA curve with its particular features and some disadvantages. The AER, based on its analysis, came to the view that it considered the most appropriate course was still to rely upon the Bloomberg curve, which was published by a highly reputable and independent firm, Bloomberg, and could be expected to be relied upon by market participants.
44 As to the fourth and fifth points, the fact that the RBA curve had a larger sample size of bonds was uncontentious. That was common factual ground known by the AER. But that fact in and of itself was not a matter raised by AusNet as an issue during the regulatory process. The AER observed in its decision that the first significant weakness in the RBA curve by way of comparison was that while it had a larger sample size, the RBA only published estimates for one day in each month and as a consequence the AER needed to interpolate between month-end estimates in order to have a daily yield series. The impact of this was that where the AER adopted the RBA estimate, it had only 12 data points for the spread to risk-free rate, the Commonwealth Government securities that were used to produce a full year of estimates. In contrast, the Bloomberg curve was published daily, which meant there were approximately 20 times as many data points for estimating the spread to swap over the course of a month compared to the RBA curve. Dr Lally discussed the relative merits of the two curves. First, there was the discussion of the fact that the RBA curve might be considered to be overweight in 20 foreign currency bonds, and the consequences of that and some of the issues that arose from that. Also the Lally November 2014 report observed that the bonds in the RBA curve were likely to have inferior liquidity compared to the bonds in the Bloomberg curve. Of course, there were pros and cons discussed in that Lally report, but it was the AER’s view, and based on the material before it, that it would not be desirable to base the estimate of the return on debt only on the RBA curve. There were a number of reasons for that: partly that it was only an estimate once a month, partly it was overweight in foreign currency bonds compared to the material before the AER as to how the Australian regulated entity was likely to raise its debt, and inferior liquidity of the bonds. The AER considered the material about the Asciano bond, looked at a number of aspects of the Bloomberg curve, but still came to the conclusion that over any lengthy period the Bloomberg curve ought to be an appropriate basis on which to base an estimate.
45 UnitingCare Australia submitted that an important part of the 2013 guideline was the transition process from “on the day” to “trailing average”. The guideline accepted that in applying the “trailing average” approach an average of estimates from the RBA and Bloomberg valuation service would be utilised. The average from the two data sets was to minimise impacts of data fluctuations over time, providing a degree of certainty for both network businesses and end consumers. The AER had made the right decision in maintaining the guideline transitional approach to “trailing average”, which was fair for both businesses and consumers, over time.
46 We turn to consider these submissions.
47 The power of the Tribunal is first to identify any error of the specified kind. Although this is merits review, it is limited merits review.
48 In our opinion, the claim that the conclusion of the AER that the Bloomberg curve was fit for purpose, as challenged by AusNet, was not a finding of fact within s 71C of the NEL, that is, that the AER made an error of fact in its findings of facts. We find it to be a matter of choosing an available opinion over another involving no error of fact. This is not to say that a matter of opinion or judgement may never involve an erroneous finding of fact: see  above.
49 We do not accept the submission on behalf of AusNet to the effect that further fact-finding on the part of the AER was necessary once it had selected a method.
50 We reach the same conclusion in relation to AusNet’s claim that a secondary finding of fact was that a simple average of the two curves was appropriate. Again, in our view, this was a matter of opinion involving no error of fact.
51 We do not accept AusNet’s submission that, under r 6.5.2(h) of the NER, the return on AusNet’s debt must be estimated such as to be commensurate with the efficient financing costs of a pure-play, regulated energy-network business operating within Australia, which borrows money for 10 years at a time. The rule does not state that and that submission is inconsistent with a provision concerned with the concepts of “estimate”, “contribution” and “objective” with which the rule is concerned.
52 AusNet’s claim that the AER incorrectly exercised its discretion in taking the average of the two curves was put essentially on the same basis as the claimed error or errors of fact. In our view it fails for the same reasons.
53 We reach the same conclusion in relation to AusNet’s submission that the AER’s decision lacked reason, was unreasoned, and was unreasonable in the result. In our opinion the AER’s decision was one on which reasonable minds might differ, but that does not establish that the decision was unreasonable. Indeed, in our opinion, the AER’s decision was both reasoned and reasonable.
54 As a consequence, in our opinion it is not necessary to decide the questions which arose in relation to s 71O or s 71R of the NEL.
55 As to the former, it provided:
71O Matters that may and may not be raised in a review
(2) In a review under this Subdivision, the following provisions apply in relation to a person or body, other than the AER (and so apply at all stages of the proceedings before the Tribunal):
(a) a regulated network service provider to whom the reviewable regulatory decision being reviewed applies may not raise in relation to the issue of whether a ground for review exists or has been made out any matter that was not raised and maintained by the provider in submissions to the AER before the reviewable regulatory decision was made;
(b) a regulated network service provider whose commercial interests are materially affected by the reviewable regulatory decision being reviewed may not raise in relation to the issue of whether a ground for review exists or has been made out any matter that was not raised and maintained by the provider in submissions to the AER before the reviewable regulatory decision was made;
(c) an affected or interested person or body (other than a provider under paragraph (a) or (b)) may not raise in relation to the issue of whether a ground for review exists or has been made out any matter that was not raised by the person or body in a submission to the AER before the reviewable regulatory decision was made;
(d) subject to paragraphs (a), (b) and (c)--
(i) the applicant, or an intervener who has raised a new ground for review under section 71M, may raise any matter relevant to the issues to be considered under section 71P(2a) and (2b); and
(ii) any person or body, other than the applicant or an intervener who has raised a new ground for review under section 71M, may not raise any matter relevant to the issues to be considered under section 71P(2a) and (2b) unless it is in response to a matter raised by--
(A) the AER under subsection (1)(b)(iii); or
(B) the applicant under subparagraph (i); or
(C) an intervener under subparagraph (i).
(3) For the purposes of subsection (2)(d)--
(a) a reference to an applicant includes a reference to a person or body who has applied to the Tribunal for leave to apply for a review under this Subdivision; and
(b) a reference to an intervener includes a reference to a person or body who has applied to the Tribunal for leave to intervene in a review under this Subdivision.
56 Since we are of the view that this ground fails even if AusNet is permitted to raise it, we do not need to decide whether it constitutes or consists of “any matter that was not raised and maintained by the provider in submissions to the AER before the reviewable regulatory decision was made”. We refer here particularly to the overseas bond point.
57 As to the latter, it provided:
71R Matters to be considered by Tribunal in making determination
(1) Subject to this section, the Tribunal, in acting under this Division with respect to a reviewable regulatory decision--
(a) must not consider any matter other than review related matter (and any matter arising as a result of consultation under paragraph (b)); and
(b) must, before making a determination, take reasonable steps to consult with (in such manner as the Tribunal thinks appropriate)--
(i) network service users and prospective network service users of the relevant services; and
(ii) any user or consumer associations or user or consumer interest groups,
that the Tribunal considers have an interest in the determination, other than a user or consumer association or a user or consumer interest group that is a party to the review.
(3) If in a review the Tribunal is of the view that a ground for review has been made out, the Tribunal may, on application by a party to the review, allow new information or material to be submitted if the party can establish to the satisfaction of the Tribunal that the information or material--
(a) was publicly available or known to be available to the AER when it was making the reviewable regulatory decision; or
(b) would assist the Tribunal on any aspect of the determination to be made and was not unreasonably withheld from the AER when it was making the reviewable regulatory decision,
and was (in the opinion of the Tribunal) information or material that the AER would reasonably have been expected to have considered when it was making the reviewable regulatory decision.
(4) Subject to this Law, for the purpose of subsection (3)(b), information or material not provided to the AER following a request for that information or material by it under this Law or the Rules is to be taken to have been unreasonably withheld.
(5) Subsection (4) does not limit what may constitute an unreasonable withholding of information or material.
(5a) In addition, if in a review the Tribunal is of the view--
(a) that a ground for review has been made out; and
(b) that it would assist the Tribunal to obtain information or material under this subsection in order to determine whether a materially preferable NEO decision exists,
the Tribunal may, on its own initiative, take steps to obtain that information or material (including by seeking evidence from such persons as it thinks fit).
(5b) The action taken by a person acting in response to steps taken by the Tribunal under subsection (5a) must be limited to considering decision related matter under section 28ZJ.
(6) In this section—“review related matter” means--
(a) the application for review; and
(b) a notice raising new grounds for review filed by an intervener; and
(c) the submissions made to the Tribunal by the parties to the review; and
(d) decision related matter under section 28ZJ; and
(e) any other matter properly before the Tribunal in connection with the relevant proceedings.
58 Since we are of the view that this ground for review has not been made out, the occasion does not arise for allowing new information or material to be submitted.
What the AER decided
59 At pages 7-9 of the Final Decision, the AER noted that AusNet included a category specific forecast for self-insurance above actual self-insurance it incurred in 2014. This was $8.4 million ($2015) higher than in the AER’s preliminary decision. In its preliminary decision the AER did not include a category specific forecast for self-insurance.
60 At page 7-24, under the heading “7.4.4, Other costs not included in the base year”, the AER said:
We have included debt raising costs and guaranteed service level payments in our final decision opex forecast. We have not included any other category specific forecast in our final decision opex forecast.
61 The AER said AusNet proposed a category specific forecast for self-insurance; however, the AER had applied a “base-step-trend” approach to forecasting self-insurance consistent with its Guideline.
62 At pages 7-95 to 7-98, the AER said, omitting footnotes:
Consideration of self-insurance forecast
We do not consider AusNet Services provided sufficient new information for us to move from our preliminary position not to include a category specific forecast for self-insurance in the opex forecast. We prefer a 'base-step-trend' approach for the reasons discussed below.
In its revised proposal, AusNet Services proposed a category specific forecast for self-insurance of $16.9 million ($2015). This increased its total opex forecast by $7.7 million ($2015) compared to leaving self-insurance costs in the base year.
AusNet Services considered self-insurance losses are volatile and can vary markedly from year to year. For this reason, it stated the quantification of these losses is best suited to an actuarial analysis that forecasts self-insurance based on expected losses determined from historical data, rather than on actual losses in a single year. It stated the AER’s approach will not result in a more accurate forecast of total opex than such an analysis, particularly when base year opex is materially influenced by losses due to abnormal events.
Both we and AusNet Services have forecast opex for 2016–20 by predominantly using a top down revealed cost methodology. That is, both we and AusNet Services have taken the costs incurred in 2014 and used them to estimate total opex for 2016–20.
We use this approach because total opex tends to be relatively recurrent - suggesting it is a reasonable basis for forecasting total opex for the next regulatory control period. … AusNet Services’ total opex has been similar in the three most recent regulatory years. This is the case even though AusNet Services incurred a self-insurance loss of $10 million in 2015.
By using a revealed cost approach neither we nor AusNet Services are forecasting that the opex AusNet Services incurred in 2014 for each specific program or category will be the same in each year of the next regulatory control period. We have not considered what forecast opex AusNet Services will spend on each opex program and project in the next regulatory control period. The top down nature of this approach (which is consistent with the NER requirement to determine opex in total) means it is not necessary to consider exactly how AusNet Services will allocate opex to programs and projects in the next regulatory control period. AusNet Services seems to accept this given it did not present evidence analysing what it expects to spend on each program and project in the next regulatory control period.
We would generally expect that to keep opex relatively recurrent, a service provider can reallocate resources between different projects and between different categories.
We also have concerns about category specific forecasting approaches when used in conjunction with a revealed cost forecasting approach. Under such a hybrid approach, a service provider has an incentive to use a bottom-up forecasting approach for new projects or programs, or where the cost is expected to rise in the forecast period. Where a service provider expects the costs of projects or programs to decline, its incentive is to use a base year approach. Under such a hybrid forecasting approach, a service provider would be financially rewarded as a result of the costs of projects and programs that are declining but would not be penalised for the costs of the projects and programs that are increasing.
AusNet Services considered our preliminary decision failed to recognise the impact of self-insurance losses on total opex. It stated that if it paid a $10 million deductible during 2014 (the base year for the current period) rather than in 2015, its opex forecast for the current period would be at least $50 million higher under the AER’s self-insurance approach. However, AusNet Services' statement assumes we do not assess the efficiency of the base year before we use it as a starting point for our estimate of total opex. If the base year appears to be inefficient, we may choose to undertake a more detailed review or substitute it with our own estimate.
AusNet Services’ submission also assumes that if AusNet Services incurred a large self-insurance loss it would not adjust any of its costs in response. If a service provider incurred a large increase in costs in one area, it may try to reduce its costs in other areas to limit the net impact to the business. For example, AusNet Services incurred a self-insurance loss of $10 million in 2015; however, its actual total opex only increased by $5.6 million ($2015) in that year.
In any case, in the next regulatory control period, the revenue impacts of one off self-insurance losses in the base year will be addressed by including self-insurance in the EBSS. The interaction of the EBSS with a revealed cost forecasting approach means the net impact on total revenue of a self-insurance loss (or savings) in the base year is small. For example, if AusNet Services incurs a $10 million self-insurance cost in the base year, its opex forecast for the following period will be $50 million higher. However, its EBSS reward will be lower by a similar magnitude, offsetting the higher opex forecast. Similarly, if AusNet Services pays no self-insurance costs in the base year, its opex forecast going forward will be lower but it will receive an offsetting EBSS reward. When the EBSS applies, the costs of a self-insurance event will be shared between the network service provider and network consumers according to the sharing ratio in the EBSS. That is, regardless of the timing of the event the cost will be shared approximately 30:70 between the service provider and network consumers.
AusNet Services also stated our preliminary decision did not treat self-insurance consistently with insurance. It stated self-insurance is analogous to insurance, with the “premium” being a self-insurance allowance. Therefore, just like for insurance, the relevant cost in the base year is the allowance, or premium, not the self-insurance losses incurred. It stated by accepting that insurance premiums are a cost that should be allowed for in opex forecasts, but not adopting the same position on self-insurance, the preliminary decision provides a perverse incentive to rely solely on insurance, rather than self-insurance, to manage insurable risks. We disagree. Our forecasting approach does not create incentives for a network service provider to spend opex on one category rather than another. We do not forecast expenditure for individual categories of opex such as insurance and self-insurance. Rather we forecast total opex. In other words, we provide network service providers with a total opex forecast which they are free to allocate as they determine. This means AusNet Services has an incentive to choose the most efficient mix of total opex including insurance, self-insurance and risk mitigation for its circumstances.
The applicant’s submissions
63 AusNet submitted, in summary, that in its Regulatory Proposals, it proposed an actuarial forecast of the cost it would incur in self-insuring certain events over the regulatory control period 2016–20. That forecast was rejected by the AER, which substituted the self-insured losses suffered by AusNet in 2014, escalated by a rate of change.
64 In doing so, AusNet submitted, the AER equated the cost of self-insurance with self-insured losses. That was an error of fact. The cost of self-insuring an event in each of a sequence of years was the amount, equivalent to a premium, reserved in each of those years to split the risk of the event across those years. It was not the amount that must be paid out when the event occurred. That amount was equivalent to the insurance payout. That was the self-insured loss.
65 AusNet submitted it was possible that self-insured losses may, in any given year, equal the cost of self-insuring them. But, given the volatility of such losses, it was not probable. Therefore it was unreasonable to forecast that the losses will equal the cost.
66 While denying AusNet the full cost of its self-insurance, the AER did allow it the full cost of its external insurance. That gave it an incentive to insure externally when it could more cheaply insure itself. That was not an “effective incentive … to promote economic efficiency with respect to direct control network services”, as required by the revenue and pricing principle (RPP) laid down in s 7A(3) of the NEL.
67 AusNet submitted that greater external insurance would ultimately, contrary to the NEO, cost consumers more without improving either the supply of their electricity or the electricity system. Thus that incentive was the product of an incorrect exercise of discretion.
68 In oral submissions, Senior Counsel for AusNet submitted as follows.
69 In making a distribution determination the AER was required to make each of the decisions which were listed as constituent decisions in r 6.12.1. The relevant constituent decision was r 6.12.1(4), a decision in which the AER, acting in accordance with r 6.5.6(d), did not accept the total of the forecast operating expenditure for the regulatory control period that was included in the current building block proposal. In that case the AER must set out its reasons for that decision and an estimate of the total of the distribution network’s or DNSP’s required opex for the regulatory control period that the AER was satisfied reasonably reflected the opex criteria, taking into account the opex factors.
70 The statutory provisions relied on by AusNet were in the following terms:
6.5.6 Forecast operating expenditure
(a) A building block proposal must include the total forecast operating expenditure for the relevant regulatory control period which the Distribution Network Service Provider considers is required in order to achieve each of the following (the operating expenditure objectives):
(1) meet or manage the expected demand for standard control services over that period;
(2) comply with all applicable regulatory obligations or requirements associated with the provision of standard control services;
(3) to the extent that there is no applicable regulatory obligation or requirement in relation to:
(i) the quality, reliability or security of supply of standard control services; or
(ii) the reliability or security of the distribution system through the supply of standard control services, to the relevant extent:
(iii) maintain the quality, reliability and security of supply of standard control services; and
(iv) maintain the reliability and security of the distribution system through the supply of standard control services; and
(4) maintain the safety of the distribution system through the supply of standard control services.
(b) The forecast of required operating expenditure of a Distribution Network Service Provider that is included in a building block proposal must:
(1) comply with the requirements of any relevant regulatory information instrument;
(2) be for expenditure that is properly allocated to standard control services in accordance with the principles and policies set out in the Cost Allocation Method for the Distribution Network Service Provider; and
(3) include both:
(i) the total of the forecast operating expenditure for the relevant regulatory control period; and
(ii) the forecast operating expenditure for each regulatory year of the relevant regulatory control period.
(c) The AER must accept the forecast of required operating expenditure of a Distribution Network Service Provider that is included in a building block proposal if the AER is satisfied that the total of the forecast operating expenditure for the regulatory control period reasonably reflects each of the following (the operating expenditure criteria):
(1) the efficient costs of achieving the operating expenditure objectives; and
(2) the costs that a prudent operator would require to achieve the operating expenditure objectives; and
(3) a realistic expectation of the demand forecast and cost inputs required to achieve the operating expenditure objectives.
(d) If the AER is not satisfied as referred to in paragraph (c), it must not accept the forecast of required operating expenditure of a Distribution Network Service Provider that is included in a building block proposal.
(e) In deciding whether or not the AER is satisfied as referred to in paragraph (c), the AER must have regard to the following (the operating expenditure factors):
(4) the most recent annual benchmarking report that has been published under rule 6.27 and the benchmark operating expenditure that would be incurred by an efficient Distribution Network Service Provider over the relevant regulatory control period;
(5) the actual and expected operating expenditure of the Distribution Network Service Provider during any preceding regulatory control periods;
(5A) the extent to which the operating expenditure forecast includes expenditure to address the concerns of electricity consumers as identified by the Distribution Network Service Provider in the course of its engagement with electricity consumers;
(6) the relative prices of operating and capital inputs;
(7) the substitution possibilities between operating and capital expenditure;
(8) whether the operating expenditure forecast is consistent with any incentive scheme or schemes that apply to the Distribution Network Service Provider under clauses 6.5.8 or 6.6.2 to 6.6.4;
(9) the extent the operating expenditure forecast is referable to arrangements with a person other than the Distribution Network Service Provider that, in the opinion of the AER, do not reflect arm’s length terms;
(9A) whether the operating expenditure forecast includes an amount relating to a project that should more appropriately be included as a contingent project under clause 6.6A.1(b);
(10) the extent the Distribution Network Service Provider has considered, and made provision for, efficient and prudent non-network alternatives; and
(11) [Note: Clause left intentionally blank]
(12) any other factor the AER considers relevant and which the AER has notified the Distribution Network Service Provider in writing, prior to the submission of its revised regulatory proposal under clause 6.10.3, is an operating expenditure factor.
6.12 Requirements relating to draft and final distribution determinations
6.12.1 Constituent decisions
A distribution determination is predicated on the following decisions by the AER (constituent decisions):
(1) a decision on the classification of the services to be provided by the Distribution Network Service Provider during the course of the regulatory control period;
(2) a decision on the Distribution Network Service Provider's current building block proposal in which the AER either approves or refuses to approve:
(i) the annual revenue requirement for the Distribution Network Service Provider, as set out in the building block proposal, for each regulatory year of the regulatory control period; and
(ii) the commencement and length of the regulatory control period as proposed in the building block proposal;
(3) a decision in which the AER either:
(i) acting in accordance with clause 6.5.7(c), accepts the total of the forecast capital expenditure for the regulatory control period that is included in the current building block proposal; or
(ii) acting in accordance with clause 6.5.7(d), does not accept the total of the forecast capital expenditure for the regulatory control period that is included in the current building block proposal, in which case the AER must set out its reasons for that decision and an estimate of the total of the Distribution Network Service Provider’s required capital expenditure for the regulatory control period that the AER is satisfied reasonably reflects the capital expenditure criteria, taking into account the capital expenditure factors;
(4) a decision in which the AER either:
(i) acting in accordance with clause 6.5.6(c), accepts the total of the forecast operating expenditure for the regulatory control period that is included in the current building block proposal; or
(ii) acting in accordance with clause 6.5.6(d), does not accept the total of the forecast operating expenditure for the regulatory control period that is included in the current building block proposal, in which case the AER must set out its reasons for that decision and an estimate of the total of the Distribution Network Service Provider’s required operating expenditure for the regulatory control period that the AER is satisfied reasonably reflects the operating expenditure criteria, taking into account the operating expenditure factors;
71 The amount sought by AusNet for expected actuarially assessed future insurance losses was $3.4 million per annum. The amount allowed was the $1.7 million per annum that AusNet sought to have excluded. There was, in fact, some overlap with other categories but this was essentially an issue of principle to be determined by the Tribunal.
72 The net effect on AusNet was that it received $8.5 million less opex than it would have received had the actuarial assessment of the likely annual cost of self-insured losses been accepted by the AER.
73 AusNet’s distribution network was the most bushfire prone in the world. Therefore, for AusNet, appropriate risk management was a highly important issue.
74 AusNet’s complaint was that a category-specific forecast was not included as it had sought for the cost of self-insured losses. The AER did not include any other category forecasts.
75 The section dealing with consideration of self-insurance forecast disclosed the AER’s approach to be as follows. First, opex tended to be recurrent. Secondly, the concept of self-insurance being part of opex was acceptable to the AER. Thirdly, opex can be reallocated if, for example, there was a higher than provided for self-insurance loss in any year. Opex can be reallocated to cover that loss, whilst still keeping within the total allowed opex for that year. Fourthly, the AER was not prepared to permit what it regarded to be an inefficient cost in the base year to form part of its base-step-trend methodology. Fifthly, any extraordinary self-insured losses in any year were to be dealt with by the Efficiency Benefit Sharing Scheme (EBSS), under which those extraordinary self-insured losses would be shared as to 30 per cent by AusNet and 70 per cent by consumers, which carried with it the proposition that AusNet would be unfunded as to 30 per cent of those self-insured losses. Sixthly, the AER made the point that self-insurance losses in the base year were the equivalent of the cost of self-insurance to AusNet.
76 Senior Counsel for AusNet submitted the AER’s reasons did not respond to the opex criteria in the rules. The AER, having rejected AusNet’s proposal for opex, was required to estimate its own total opex and in doing so was required to make an assessment of the efficient costs of a prudent DNSP in achieving the opex objectives. The opex objectives, in r 6.5.6A(2), could be summarised as: “Maintaining the quality, reliability, security and safety of the network and supply to consumers under the network.”
77 It was the task of the AER to estimate total opex which it was satisfied reasonably reflected the opex criteria, that is, the efficient costs of a prudent DNSP in achieving the opex expenditure. That had not occurred in relation to self-insured losses. If the starting point was that self-insurance was efficient risk management, and there was no issue that the risks AusNet self-insured were appropriate risk management by AusNet, and if it was accepted that there was an opex cost to retaining self-insured losses, then it was necessary to give consideration to what was an efficient cost of doing so. That had not occurred at all in this case.
78 What the AER considered it was doing was identifying something which was likely to be relatively recurrent and, therefore, a reasonable basis for forecasting total opex, which included within it self-insured losses. It must, therefore, have assumed that self-insured losses were at least relatively recurrent, when, in fact, they were not, and the examples given by the AER in its Final Decision demonstrated that.
79 AusNet chose to self-insure as part of its risk management strategy. The wisdom of its choices had not been challenged or questioned. The issue was: what was the appropriate component that should be allowed for in opex for self-insured losses? That, in turn, threw up the question of how that should be assessed. The issue here was not the total amount of opex per se, but whether the AER should have included a category-specific forecast for self-insured losses, which it did not. That state of affairs was founded on the AER deciding that the self-insured losses incurred in the base year were efficient costs. But they could not be an efficient cost which could be used to predict the so-called smooth or recurrent efficient cost in future years.
80 If the AER had not made any other adjustments for things which could not be controlled or were not recurrent, AusNet’s case might be considerably more difficult. But the AER did, in fact, adopt the hybrid approach that it referred to in its Final Decision, and its error here was to not accept a category-specific forecast for self-insured losses. Had the AER not done a category-specific forecast for anything, AusNet’s case would be much weaker. But here it had chosen to identify costs it considered were not efficient, or were unlikely to be recurrent, or which should be assessed on a category-specific basis using a different measure, for example, debt raising costs. So it was clear from its treatment of self-insured losses that the AER did not address the opex criteria in its assessment of AusNet’s total opex.
81 The AER’s proffered reasons seemed to assume that self-insured losses would be reasonably recurrent. They were not. In addition, the AER took comfort from the fact that although there was almost six times the level of uninsured losses in 2015 when compared to 2014, the opex was almost the same. That was not a relevant consideration to the opex criteria.
82 The AER did not, in relation to self-insured losses, follow the approach it followed in relation to debt raising costs, that is, replacing the figure put forward by AusNet with the figure that was appropriate for a benchmark firm. As AusNet has a unique network in terms of risk and, therefore, had unique issues in relation to risk management, which included the question of what to self-insure and what not to self-insure, special and specific consideration should have been given to that component of AusNet’s opex. And the suggestion that notwithstanding the matters that were unique to AusNet and its network and the risks to which it was exposed, it could, like all the other DNSPs with less risk exposed networks, move its opex or capex around to cover uninsured losses, was not an appropriate application of the opex criteria.
83 The AER eschewed a hybrid approach whilst, in fact, embracing it where it saw fit. It had, in substance, accepted the base year self-insured losses as efficient. That was its answer to AusNet’s proposal. Then, in substance, the AER stated that no more detailed review or different approach was required. But because self-insured losses were not equivalent to forecast future expenditure on self-insured losses, the position taken by the AER was in error. Likewise, the impact of the EBSS was not relevant to the opex criteria and, in any event, would leave AusNet potentially exposed to 30 per cent self-insured losses for which it had no provisional cover.
84 AusNet submitted that what the AER was required to do and did, as part of the final determination, was to estimate the total of AusNet’s opex for the regulatory control period, and that in doing so it was required to be reasonably satisfied that its estimate reflected the operating expenditure criteria. That, in turn, required its estimate to reasonably reflect the efficient costs of a prudent DNSP in the context of a realistic expectation of demand and realistic cost inputs required to achieve the objectives.
85 The AERs Final Decision, AusNet submitted, consisted of making an assessment of likely future occurrences and states of affairs in connection with AusNet’s network, and doing so on the basis of a variety of evidence, which included historic facts. The issue came down to whether the AER should have removed self-insured losses from the base year and accepted a category-specific forecast, in which case it has no dispute with the AON report and the figure contended for by AusNet.
86 Turning to the question of errors of fact, four were pressed. The first was that the AER equated self-insured losses in the base year with the cost to AusNet of self-insurance, when the two were not the same. Secondly, the AER failed to find that self-insured losses were volatile and may vary considerably from year to year, but the cost of self-insurance was relatively stable. Thirdly, the AER failed to find that the cost of self-insurance to a DNSP was, in substance, a self-insurance premium conceptually equivalent to the premium for external insurance of the same risks, so that it required a category-specific allowance or forecast on an actuarial basis to be included in the base year. Fourthly, the AER found that self-insured insurance losses could be controlled or funded by reducing costs in other areas, which was an erroneous finding of fact because self-insured losses were unpredictable and volatile and highly variable.
87 AusNet submitted that in substituting its own estimate of the total opex for the regulatory period, it may be inferred that the AER accepted that the self-insurance risk management practices of AusNet were prudent and efficient, that self-insurance had a cost, that there should be an allowance for the cost of self-insurance in the substitute forecast it made, and that its task was to identify an appropriate allowance. It identified the appropriate allowance to be the equivalent of the sum paid out in 2014. That, AusNet submitted, was material to the making of the decision because, firstly, it was a significant item of opex; secondly, the total amount involved was $8.5 million; thirdly, the issue was of considerable importance to AusNet in the context of its particular network.
88 Turning to the incorrect exercise of discretion and unreasonableness, AusNet submitted as follows.
89 The discretion that the AER exercised was to make and substitute its own estimate of total opex under r 6.12.1(4) of the NER. In doing so, the AER was required to decide what opex reasonably reflected the opex criteria, taking into account the opex expenditure factors to achieve the opex objectives. It exercised the discretion to make and substitute its own estimate of total opex in the preliminary decision and varied it slightly in the Final Determination. In that exercise of discretion, it applied the base-step-trend methodology, which commenced with determining whether actual expenditure in the base year reasonably reflected the opex criteria. The common genesis of the exercise of discretion that occurred with the AER deciding to make and substituting its own estimate of total opex was in various findings of fact. Here, the findings of fact were in relation to the opex that it considered was the appropriate opex for AusNet for the regulatory period, which in turn were based on findings of fact concerning opex in the base year, and findings of fact in relation to the step and trend parts of the escalation and adjustment process. The AER weighed up the considerations that it considered to be relevant and it made decisions as to what weight to give those considerations, and it thereby determined, by that process, what it considered to be the right result, that is, the appropriate opex allowance for the regulatory period.
90 As illustrated by Envestra Limited (No. 2) at , ActewAGL at , and Application by APT Allgas Energy Limited (No 2)  ACompT 5 at , the discretion can be incorrectly exercised in a number of ways. AusNet submitted that what occurred here was that none of the reasons advanced by the AER in consideration of self-insurance forecast in the Final Decision were permitted by the rules, were relevant to the opex criteria, or relevant to the opex objectives.
91 AusNet also submitted that the exercise of discretion was in error by a failure to have regard to mandatory relevant factors. The authority for those general propositions was Australian Competition and Consumer Commission v Australian Competition Tribunal  FCAFC 83; 152 FCR 33 at , Envestra Limited (No 2) at  and APT Allgas (No. 2) at .
92 Applying those principles to the present case, the AER had taken the view that opex tended to be recurrent. It was appropriate to include self-insurance losses. Opex could be reallocated. It would not permit an inefficient cost in relation to self-insurance in the base year. Extraordinary losses could be dealt with via the EBSS. Self-insurance opex in the base year was the equivalent of self-insurance losses paid out in the base year. All of those matters were incorrect and did not engage with the opex criteria. None of them, except whether it was appropriate to include self-insurance in opex, affected the question of whether self-insured losses were efficient expenditure. Each of the reasons set out in the Final Decision between pages 95 and 98 were otherwise irrelevant to and did not discharge the AER’s obligations in estimating its own forecast total opex.
93 The errors in the exercise of the AER’s discretion were summarised as follows. First, self-insurance costs were not the equivalent of self-insurance losses and vice versa. A self-insurance premium or something analogous to a self-insurance premium actuarially calculated was the appropriate opex allowance. That was put to the AER and rejected. Secondly, the AER should have made a category-specific assessment of the appropriate allowance for self-insured losses within total opex, as it did in respect of debt raising costs, but it did not do so. Thirdly, the AER considered that the actual self-insured losses in 2014, the base year, represented the cost of self-insurance in the base year which was, in effect, the converse of AusNet’s complaint that the AER erred in considering self-insurance losses were the same as self-insurance costs. And, fourthly, in applying the base-step-trend methodology, it was an error to use self-insured losses as a revealed cost of self-insurance. AusNet further submitted that the AER failed to have regard to the high volatility and unpredictability of self-insured losses, and that was an error in the exercise of its discretion. AusNet submitted that by reason of that high volatility and unpredictability it was wrong to use self-insured losses to forecast the cost of self-insurance, the burden of self-insurance in the future. AusNet submitted the AER erred in the exercise of its discretion by regarding it as available and appropriate to a DNSP to reallocate opex to cover some or all of its uninsured liability losses.
94 In relation to unreasonableness, AusNet accepted that unreasonableness must be of the decision itself, not a step in the factual findings or the reasoning. It accepted that there must be something like logical error or irrationality in the decision. It submitted that the ground would be made out if the AER’s decision could not be justified by reference to its stated reasons, and it contended that that was, in fact, what had occurred. There was an arbitrary and capricious result which had not been determined by reference to the applicable criteria in the rules and which was unreasonable. It was unreasonable because the component of opex in respect of self-insurance, being one year’s losses, was not reflective of the actual reasonable forecast cost to AusNet of self-insured losses in the future. One year’s incurred losses could be no more than one element in a very long chain of elements required to reliably predict the cost of uninsured losses in the future. AusNet submitted that it was unreasonable for the AER, in the exercise of its discretion, to suggest the allowance for uninsured losses was the amount AusNet paid out in 2014 for uninsured losses, and that it could cover any greater losses by moving parts of its opex or capex to cover those losses because in recent years AusNet’s opex had been relatively stable. A further reason for the AER taking that position was that AusNet had the ability to choose how it dealt with its uninsured losses. It was able to choose between external insurance, self-insurance and other risk mitigation tactics. This was simply unreasonable and incorrect. What was required, and what the AER failed to do, was to determine whether or not there should be a category-specific allowance for uninsured losses, and if so on what basis. It was required to do that, firstly, because AusNet raised it, and, secondly, because the discharge of its tasks in arriving at its own estimate, which it was required to be reasonably satisfied reflected the criteria in the rules, compelled the AER to assess separately and on an actuarial basis the likely cost of uninsured losses in the future for the regulatory period. So, in conclusion, AusNet submitted that the AER deciding to substitute and making its own estimate of total opex for the regulatory period had produced an unreasonable result.
95 As to the materially preferable NEO in respect of self-insured losses, AusNet submitted it had been denied $8.46 million in total for the upcoming regulatory period of its efficient costs of self-insurance, there being no dispute that its self-insurance was an efficient cost at some level or another. That compelled AusNet to meet its costs of self-insurance by reducing expenditure in other areas where the allowance for those areas was considered to be efficient. That was inappropriate. If it was prudent and efficient for AusNet to self-insure as part of its risk management strategy, then it should have an appropriate allowance for the predicted cost of its self-insured losses for the regulatory period. This issue was of prime importance to AusNet given the nature of its network and it was an issue which had a long tail.
96 If the error in approach taken by the AER in relation to AusNet’s self-insured losses was not corrected in this review, it was likely to have significant adverse long-term consequences not only for AusNet, but in respect of all DNSPs who chose to self-insure as part of their risk management strategies. It created an artificial incentive to insure externally, regardless of the cost. If less risk was self-insured and more risk was externally insured, the obvious ultimate consequences would be that those additional costs would be passed on to consumers, and at a level which was considered by AusNet and AON to be likely to be considerably in excess of the cost of self-insurance. The reason was that, as AON had indicated, the cost of insuring externally the risks which were self-insured was considered to be uncommercial, to the extent that AusNet chose to carry that risk itself.
97 An appropriate allowance within total opex for self-insured losses was infinitely preferable to using the EBSS, and using the pass-through provisions of the legislation as a primary risk management tool. AusNet also submitted that correcting the decision in relation to self-insurance losses within opex would materially advance the revenue and pricing principles and, therefore, the NEO.
The AER’s submissions
98 The AER submitted, in summary, that insisting on the correct identification of the decision made was fundamental. In determining base opex, the AER submitted it was not required to, and did not, estimate “self-insurance costs” as a discrete subcategory. Rather, the AER did not accept AusNet’s proposal that its self-insurance costs should be excluded from the revealed base year costs, and forecast as a separate category. That was the step in the AER’s decision-making to which AusNet objected.
99 In substance, the AER submitted, AusNet’s complaint went to the method that AER used to derive base opex:
(a) Was the AER permitted simply to use the revealed costs in the base year (without applying any special approach for self-insurance)?
(b) Or was the AER required to (i) remove self-insured losses from the base year, and (ii) substitute an actuarial forecast of self-insured losses?
100 The AER submitted the NER did not require the AER to remove self-insured losses from the base year: r 6.12.3(a). It distinguished the position prior to the 2012 rule change, under former r 6.12.3(f).
101 The AER submitted it did not “equate” self-insured losses with self-insurance costs. Actuarial estimates were not self-insurance “costs” for the purposes of opex. A premium paid to an external insurer was an expenditure; an actuarial forecast was not.
102 AusNet did not maintain provisions for the self-insurance costs that it now claimed. The amounts in question were “theoretical premiums”.
103 In 2011-2015 the AER estimated self-insurance as a discrete sub-category of opex. But it was incorrect to describe that sub-estimate as an earmarked “self-insurance allowance”.
104 The AER submitted its decision did not perversely incentivise inefficient external insurance.
105 Even if error was established, any revenue impact was negligible and was in the short-term only. Varying the decision in the way proposed by AusNet would not advance the long-term interests of consumers in a materially better way.
106 If the materially preferable NEO decision threshold was satisfied, then, the AER submitted, the decision should be remitted to the AER for it to review AON’s actuarial assumptions and consider whether any consequential adjustments were required.
107 In oral address to the Tribunal, Counsel for the AER submitted as follows.
108 The premise of AusNet’s submissions was that the AER was required to make, and did make, an estimate of costs of self-insured losses for the 2016-2020 period. The AER rejected that proposition. Rule 6.12.3 dealt with the extent of the AER’s discretion. By rr 6.12.1(4)(i) and (ii), the AER must accept or reject the estimate of total forecast opex, and if it did so, it must determine its substitute forecast of the total amount of opex required in accordance with r 6.5.6. In r 6.5.6(c), where the opex criteria were stated, again, the AER’s obligation was to decide whether the total forecast of opex for the 2016-2020 period was one that it was satisfied reasonably reflected the opex criteria.
109 What the AER did was not to make any estimate of self-insured losses or costs as a category. Rather, it declined AusNet’s invitation that it should do so, for stated reasons. That was, ultimately, the question of principle. Was the AER required to make a category forecast for opex, or was it open to the AER to take the approach of applying the revealed cost method to the total base year opex as a whole by taking a (top-down) assessment of overall base year opex that included the self-insured losses that had been incurred in 2014 and using that as the base component in the AER’s base-step-trend method.
110 In making its assessment of the substitute forecast of total opex, the AER’s decision about the base component was an intermediate step in arriving at its assessment of the total forecast overall. The decision that was made under r 6.12.1(4)(ii).
111 AusNet proposed a category-specific forecast for self-insurance. However, the AER applied a base-step-trend approach to forecasting self-insurance, consistent with the guideline.
112 AusNet in its submissions referred to matters that the AER did not include in its assessment of the revealed costs for the 2014 base year. It was more helpful, the AER submitted, to address these matters in terms of category-specific treatment.
113 There were four items to address. The first was debt raising costs. The second was guaranteed service level payments. The third was the demand management incentive allowance. The fourth was movements in provisions.
114 As to the first of these items, the approach that the AER took across all networks for the costs of raising equity and the costs of raising debt was that they were done on a benchmark basis consistent with the benchmark approach that the AER applied to estimating return on equity and estimating return on debt. In the return on debt field, there was much debate about assumptions about timing and the way in which a benchmark efficient entity would go about raising debt, and the AER took the approach of attributing benchmark debt raising costs in conjunction with the benchmark approach that it took to the return on debt. At page 372, in relation to AusNet’s response to the primary decision, the AER said it was still not persuaded the proposed approach would provide a realistic estimate of the efficient costs required to meet the opex objectives. It considered that it was important to forecast debt raising costs based on a benchmarking approach rather than service providers’ actual costs. This was also a matter that was hardwired into the post-tax revenue model that the AER publishes. When a business submitted that it was required to make an input of the percentage amount of its debt raising costs, that was calculated through the model in accordance with the imputed debt raising schedule of the benchmark efficient entity. And so that applied across the networks and it was not a matter that was adjusted idiosyncratically.
115 The second item was guaranteed service level payments. These were a continuation of a Victoria-specific incentive scheme that was originally administered by the Essential Services Commission (ESC). Historically the ESC had calculated those on a historical average basis and the AER continued that historical method, inter alia because transitioning from that old method to the revealed cost method may create windfall gains and losses. No-one complained against the AER continuing to do that in relation to guaranteed service level payments.
116 The third item was the demand management incentive allowance. That referred to a revenue allowance that was given under the AER’s demand management incentive scheme. That was an incentive scheme that was in a similar category as the EBSS, the Service Target Performance Incentive Scheme and the like. Its purpose was to give networks a revenue allowance to investigate and conduct demand management projects. And so the expenditure that a DNSP used for those projects in the base year was not included because, conceptually, that allowance was funded through the incentive building block, not through the opex. To include that expenditure in opex would result in double counting.
117 The fourth item was movements in provisions. The simple point was that the AER took a standard approach in relation to its expenditure forecast of basing its forecasts exclusive of movements in provisions year to year. It was a question of accounting treatment, akin to the question whether one accounted on a cash basis or on an accruals basis. So that was not, strictly speaking, an exclusion of costs as much as just a matter of accounting approach.
118 These four matters, the AER submitted, were a matter of the regulatory framework and the evolution of regulatory practice. It was not sufficient to reason, as AusNet did, that because the AER chose not to include those amounts in the revealed cost in the base year, that therefore it was inconsistent for the AER not to have done so in relation to self-insurance.
119 At page 7-27 of the Final Decision, the AER summarised that the difference in the base opex amount for this period was attributable to differences in the allocation approach to, amongst other things, self-insurance and debt raising.
120 At page 7-92, under the heading ‘Other Costs Not Included in the Base Year’, the AER explained that it preferred a base-step-trend approach to assessing most opex categories and said:
However, when appropriate, we may assess some opex categories using other forecasting techniques.
121 It may be necessary to take category-specific forecasts, for essentially practical and common sense reasons. The contest here was whether the AER erred, within s 71C, by not making a category-specific forecast in relation to self-insurance. This, the AER submitted, was plainly about a matter of method, about which the AER had a discretion under r 6.12.3 and that was because the rules said nothing about self-insurance and did not require the AER to estimate self-insurance.
122 The Tribunal had to approach that question by reference to what was the regulator’s statutory task, which was to make an estimate of total opex, and what it had said in the guideline about the approach that it proposed to take, and take into account those matters and the submissions that it received from AusNet. The AER considered whether it should make a category-specific forecast and, for the reasons given in the final decision, decided not to adopt that approach but to use the revealed cost method, including self-insurance.
123 Counsel for the AER referred to the changes to the rules that were made in 2012 and to the deletion in 2012 of r 6.12.3(f). Formerly r 6.12.3(a) stated:
Subject to this clause and other provisions of this Chapter 6 explicitly negating or limiting the AER's discretion, the AER has a discretion to accept or approve, or to refuse to accept or approve, any element of a regulatory proposal.
That discretion was previously limited r 6.12.3(f) which read:
If the AER refuses to approve an amount or value referred to in clause 6.12.1, the substitute amount or value on which the distribution determination is based must be:
(1) determined on the basis of the current regulatory proposal; and
(2) amended from that basis only to the extent necessary to enable it to be approved in accordance with the Rules.
124 When read in conjunction with the opex criteria, the effect of the former r 6.12.3(f) was that if AusNet came with a regulatory proposal to estimate its opex with a category-specific forecast for self-insurance, then the AER was obliged to make its determination using a category-specific forecast of self-insurance, and amend it from that basis only to the extent that it was necessary for the AER to be satisfied that it reasonably reflected the opex criteria. That was the position until 2012.
125 The limitation was removed by the rule change in 2012 to make clear that the AER was free to determine an opex estimate that reasonably reflected the opex criteria without any limitation or constraint as to the basis for that forecast being imposed by the method applied in the DNSP’s regulatory proposal.
126 That change was mirrored by the guideline that the AER published in 2013, the Expenditure Forecast Assessment Guideline at page vii:
Economic benchmarking techniques will allow us to analyse the efficiency of NSPs over time and compared to their peers. They will also allow us to estimate a top down forecast of expenditure and estimate productivity change.
So in conjunction with the introduction of opex benchmarking, there was a change of approach to a more top-down method of assessing networks’ expenditure forecasts. At page 62, under the heading Assessing Base Opex it was said:
We agree with CitiPower, Powercor, SA Power Networks and the ENA that the EBSS provides a strong continuous efficiency incentive and therefore base year opex should be an efficient starting point for forecasting opex. However, this is only true if the NSP responds to that incentive.
127 That statement was an important background consideration in understanding the philosophy behind the revealed cost approach to the base component of the expenditure. Base opex was intended to reflect those categories of opex that were largely recurrent year to year. And because the EBSS was in place, and because it had now been in place for at least two previous periods, that imposed efficiency incentives consistently throughout the process, and so it was reasonable to assume that a network’s revealed expenditures in the base year – typically the penultimate year of the previous period – would have responded to those incentives and, therefore, the prima facie assumption was that those expenditures ought to reflect efficient costs.
128 That did not prevent the AER from deciding that those expenditures did not reflect efficient costs. But in this case the AER was satisfied that DNSPs had responded to the efficiency incentives and, therefore, the base component was, in aggregate, reasonably reflective of the efficient costs that a prudent operator would incur.
129 Counsel for the AER was asked by the Tribunal what was the position, on this approach, if the base year costs did not actually reflect the underlying costs faced by the firm: it could have significantly higher self-insurance costs not reflected in the losses in a particular year, or even a sequence of years associated with those categories of risk expenditure. The decision, in effect, not to include them in the base opex gave rise to the question whether that was an appropriate thing to do with costs which were not revealed in ongoing expenses for a category of cost, which would be expected to be volatile, or to vary from year to year. The problem was that the actual expense in any year could be volatile; whereas the notional costs, the actuarially assessed risk, could actually be stable. Could a large number of volatile components be added up to something which was relatively stable?
130 Counsel responded that, first, it was important that the task was to estimate aggregate opex, not to estimate expected self-insurance costs over the five years. The AER did not assume that self-insurance losses in the base year were equivalent. It simply disagreed with the proposal that it had been given to carry out a category-specific forecast for self-insurance costs. Second, as the 2015 example showed, the fact that a sizeable loss was incurred did not mean, therefore, that aggregate opex would increase correspondingly. It did not follow from the fact that a large loss was incurred in this category in relation to a particular category of opex, that aggregate opex cease to be recurrent year to year. The fact that self-insurance costs have been estimated within the revealed costs for the base year, did not mean that aggregate costs would cease to be relatively recurrent year on year such that it was inappropriate to use that measure of aggregate costs, which included self-insured losses, as the measure for the base year. Counsel submitted it was within the range of reasonable methodologies to suggest that an actuarial approach could be taken.
131 But the mere fact that there was an alternative approach open did not establish the exercise of discretion was incorrect. The mere fact that the Tribunal might, starting by itself, think that this ought to be a case for a category-specific cost, was not by itself enough to amount to an incorrect exercise of discretion. It had to be shown that the AER’s own exercise of discretion miscarried or was incorrect for a reason that rose to the level of, as AusNet put it, misapplication of principles or methodologies mandated by the rules.
132 The Expenditure Forecast Assessment Guideline at page 75 addressed the appropriateness of including these cost categories, such as self-insurance, in the overall opex estimate for the base year where the amount of the losses incurred year to year may be volatile. The topic heading was “Assessing lumpy costs” and it was said:
We agree with Incenta that caution is required were there are significant lumpy costs. However, the question is not whether individual cost categories are lumpy but whether total opex is lumpy. Forecasting some categories bottom up, for example, but others using revealed cost risks upwardly biased forecasts. This is because NSPs would have an incentive to only use an alternative forecasting approach for those lumpy cost categories where expenditure is atypically low in the base year. For categories where expenditure is higher than usual in the base year they have an incentive to forecast using revealed costs. Similarly, NSPs have an incentive to use an alternative forecasting approach where work volumes are expected to increase in the next period but use revealed cost for those categories where volumes are expected to decline. Consequently there is a risk that a 'hybrid' forecasting approach will give upwardly biased forecasts. If total opex is not materially lumpy then a revealed cost forecast is appropriate regardless of whether individual categories are lumpy or not.
133 That was the principle that the AER was giving effect to in its approach in the Final Decision. As facilitated by the change to the rules in 2012, the AER now had a clear preference for using methods of broad and general application that tried to limit the number of exclusions of specific categories that needed to be assessed individually, because it was the case, and regulatory experience taught, that there were real difficulties in the AER’s ability to meaningfully assess efficiency when everything is broken down into subcategories of subcategories.
134 In the 2013 report of the Productivity Commission, Electricity Network Regulatory Frameworks, in Volume 1 on page 189, about incentive regulation, it was said:
Another important feature of incentive regulation is that it is based on high-level outcomes, such as yearly expenditure and network reliability. It is designed to leave the day-to-day decisions, such as project choice and the timing of asset replacement, to the network business.
135 At page 303 under the subheading “(c) Benchmarking, information and efficient contracting” the following was said in relation to the pre-2012 regulatory regime:
In the current regulatory regime, network businesses have large information advantages over the regulator, which they may be able to exploit to increase regulatory revenue allowances. For example, ex ante, they may claim the need for allowances to meet additional demand or to replace specific ageing assets, and then, ex post, invest at lower levels, taking the residual as a surplus. Ex ante, the regulator may not know enough information to challenge highly detailed ‘bottom-up’ costs and demand forecasts. However, the scope for gaming the regulator can be reduced (and the scope for efficient contracts strengthened) if a benchmarking model can be built that:
• uses a small set of verifiable data
• takes account of the key operating differences of the businesses
• adequately predicts efficient costs.
136 There had been a marked change in the AER’s approach since 2012 on account of the rule change, and on account of matters such as the Productivity Commission’s investigation. That had led to the introduction of benchmarking, whatever the lived experience might have been. That was the fundamental change in approach that had been applied.
137 Turning to AusNet’s Regulatory Proposal for this period at page 183, these were the adjustments or exclusions for base year opex for which AusNet contended in its initial proposal. There were nine matters that were sought to be excluded. Some of them were negative amounts and some of them were positive amounts. That helped to underscore the fact that there was a netting off over time of these things, or from year to year. These nine categories of exclusions were the nine categories of costs that were excluded from the base year component of AusNet’s previous opex forecast for the 2011 to 2015 period. So the starting point for AusNet’s proposal was a continuation of the way in which things had been done previously.
138 Turning to the Final Decision, at page 7-95, there were approximately five categories listed, and some of them would result in increases to the base opex if they were included – if the revealed costs were included in the base opex.
139 For the insurance, the AER concluded that it would have no material impact. And for self-insurance, including the cost in the revealed base opex, the AER concluded it would result in a reduction to the base opex amount. AusNet’s response to the preliminary decision was to accept the first four categories as categories included in the base year opex, but to reject self-insurance, which was the one whose inclusion in the base year would result in a reduction.
140 At page 7-94, which preceded the direct discussion of self-insurance, the AER acknowledged that using a category-specific forecast may produce more accurate forecasts of expenditure for those categories: but that was not the exercise. The critical question was: accepting that self-insured losses was a volatile category in and of itself, did the inclusion of self-insured losses in the base year mean that total opex will, therefore, be lumpy or volatile from year to year?
141 There was an important point here of proportion. AusNet said that the AER had underfunded by $8.4 million. AusNet also said that the aggregate cost on the actuarial basis, using AON’s figures, was $17.1 million. That underfunding amount, $8.4 million, reflected 0.7 per cent of total opex. That was not total revenue but was only 0.3 per cent of total revenue. Even if the Tribunal were to accept AON’s numbers as being correct, the extent of underfunding in this case is said to be only 0.7 per cent of opex. That was not an unmanageable variability year to year.
142 Even if one made the simplifying and incorrect assumption that the aggregate of all the other categories would remain exactly identical year to year, even quite sizeable changes in self-insured losses from one year to the next had a very small overall effect on total opex.
143 When one had regard to that question of proportion here, it simply did follow that because self-insurance losses were volatile year to year, that including it in the base opex will cause total opex to be lumpy year to year. And that went directly to the passage in the guideline set out at  above.
144 The requirement of the rules was that the AER must be satisfied that its estimate of total opex reasonably reflected the opex criteria. The highest case put against the AER was that it got the true forecast of opex wrong by only 0.7 per cent. It was open to the Tribunal to find that an estimate that is within 0.7 per cent of the true answer did, in fact, reasonably reflect the opex criteria as they related to total opex. It was the total estimate of opex, not the estimate of self-insurance, that must reasonably reflect the opex criteria.
145 The effect of the change to the legislation was intended to prevent these minor or relatively trivial contests, particularly on these matters of expenditure. Here the issue was about 0.3 per cent of overall revenue. And so the complaints that AusNet made about underfunding needed to be seen in proportion. The point applied as well at the materially preferable level.
146 The AER’s primary position was that as a matter of regulatory methodology there was no error in applying the revealed cost approach.
147 The next point related to AusNet’s submission that the AER took an approach of equating self-insured losses with self-insured costs. It started life as an allegation of error of fact. In the AER’s written submissions it challenged AusNet to identify where in the reasons it had made the finding of fact, and that challenge was not taken up in AusNet’s reply submissions or in oral submissions. The first full paragraph on page 96 showed the AER did not equate the two concepts (footnote omitted):
AusNet Services considered self-insurance losses are volatile and can vary markedly from year to year. For this reason, it stated the quantification of these losses is best suited to an actuarial analysis that forecasts self-insurance based on expected losses determined from historical data, rather than on actual losses in a single year.
So that sentence showed that the AER was fully aware of the distinction between the concept of the actuarial cost on the one hand and the revealed cost of actual losses in a single year. The same point was exposed in the last paragraph on page 7-98 set out at  above.
148 Counsel for the AER submitted that if the task were to estimate self-insurance losses it accepted that the natural way to estimate those losses, because of their volatile nature, would be an actuarial forecast. But this came back to the overall regulatory exercise that was being undertaken, which was to estimate total opex starting with the main component of total opex being the revealed cost in the base year. The regulatory task in its proper frame was total opex. The question was not whether self-insurance as a category was materially lumpy from year to year but whether overall opex was materially lumpy year to year. The AER’s answer to that second question, as to overall opex, was that it was not materially lumpy year to year, even when self-insurance was included. One of the reasons why it was not was because self-insurance, even if assessed on an actuarial basis, was only a very small component of the overall total.
149 On the question whether the base opex had had swings and roundabouts, Counsel for the AER submitted that the evidence in this case was that the base opex had been consistent, even when, as was the case in 2015, quite sizeable self-insurance losses were incurred in the amount of $10 million. What resulted was a small reduction in opex. AusNet bore the onus of establishing that including the self-insured amounts in the 2014 base year, or in the base year generally, resulted in aggregate opex being lumpy to such an extent that it was not a reliable measure for the base year.
150 There was an issue that arose in the Ausgrid decision about the EBSS and whether removal of movements in provisions related to defined benefit superannuation liabilities were an actual cost for the purpose of applying the EBSS. The AER’s reasons were that movements in provisions were not actual costs for the regulatory purpose of the EBSS and the Tribunal rejected the networks’ claims of error. The critical paragraphs in the Ausgrid decision were  and , as follows:
The detailed grounds of review of Networks NSW are set out above. It appears that, at this point in the AER’s reasoning, the relevant asserted error of fact is that the AER concluded that changes to provisions were not actual costs incurred in delivering network services. However, whatever may be the available accounting methods and assumptions from time to time for the making of provisions, or the appropriateness of the reported provisioning in accordance with accounting standards, it is not the case that the provisioning of those costs meant that they were “actual costs” incurred in delivering network services; they were estimates of liabilities incurred and to be paid in due course determined in accordance with accounting standards and based upon assumptions made by each of Networks NSW about the various elements going to the making up of the amount which, ultimately, would be paid to meet those liabilities.
The AER decided to reject the provisioning approach previously adopted and exposed by the adjustments made by Networks NSW (in particular by the Essential accounts and methodology for provisioning) and it decided not to directly reflect actual efficiency gains or losses by the accounting provisioning for those future liabilities. Its use of the term “actual costs” is in that context. It does not mean that the AER simply chose to ignore the costs incurred by Networks NSW, as reflected by the payments made by the Networks NSW businesses for those liabilities, or to ignore the fact that there are future liabilities which are reflected in the provisioning estimates.
151 The AER’s submission was that for the purpose of deriving the base year component of opex, with which the EBSS worked hand in glove, the same could be said about actuarial forecasts. They were not even provisions in this case. The same point could be made in relation to the actuarial forecasts, changing the word “accounting” to “actuarial”. The fact that the AER had made the decision of regulatory approach to use a top-down approach to the base year without treating self-insurance as a special category in the special case of AusNet did not mean that it had ignored the fact that these related to future liabilities which may vary year to year.
152 Although the actuarial estimate reflected an actuarial notion of a cost, the regulatory regime was one that started from costs as revealed and they were assumed to be efficient because they were amounts paid out under the EBSS and under the incentives imposed on the EBSS and in reduction of the residual amount that was available to go to the shareholders as profit. Because of those two efficiency incentives, the ex ante allowance and the EBSS, there was an assumption that expenditures at the aggregate level – not at the category level – were likely to be efficient and that was why it was a sound starting point to use the revealed cost in the base year. And for these purposes in this regulatory context the revealed cost is the amounts paid out.
153 For good reason, the AER had included its costs in base opex and not assessed them separately. And, as a consequence it had not assessed separately what those costs might have been on a year basis. In effect, it had not assessed the efficiency of any of the individual categories of expenditure inside base opex, it had taken it as a whole.
154 What was done in the previous decision was that the AER, under its old approach, did make a category-specific forecast for self-insurance. That became a component of the overall estimate for opex which in turn became a component of the overall revenue – regulated revenue allowance for those years. It was simply incorrect to posit that, because the AER historically made an estimate of the number, that therefore that was somehow an earmarked fund out of which either self-insured losses had to be paid or provisioned into some separate account. That was not how the regime worked either in relation to self-insurance or in relation to any other category of opex or in relation to opex as a whole. The fundamental axiom of incentive regulation was that the network got an allowance overall and it was entitled to run its business however it liked with them, spending it on whichever categories of opex it considered most efficient or most advantageous. And, to the extent that it was able to reduce its expenditures below those forecast amount, that resulted in an increased return to the shareholders. That was one of the incentive effects.
155 Ultimately, the AER made a decision that resulted in the network receiving an overall revenue allowance. How the AER has arrived at that decision, whether it was by forecasting opex as a discrete category or not, had no bearing upon whether AusNet thought it more efficient to self-insure or to insure externally, insofar as there was a choice. So insofar as there was a choice available, then the incentive regime, both the ex ante allowance, revenue allowance and the EBSS operated to incentivise AusNet to manage its risks in the way that it thought most efficient.
156 As to the EBSS, under the opex factors listed in r 6.5.6(e) one of the mandatory considerations the AER was required to consider was how an opex forecast related to any applicable incentive scheme. So far from being a prohibited consideration, it was a mandatory consideration.
157 As to materially preferable, the s 71P(2)(a) question, the question was: was the putative alternative decision one that will advance the long-term interests of consumers, or one that will promote efficiency in the long-term interests of consumers in a materially-better way than the decision that the AER made? Even if one accepted the AON forecasts, the amount in issue was 0.3 per cent of AusNet’s regulated revenue. Where the dollar amount and the percentage were trivial, this was a case where even if error has been taken established, the Tribunal must not remit or vary the decision. Secondly, there was no structural underfunding, no long-term asymmetric impact on this decision one way or the other. The AER’s approach may equally result in adopting the base year in years where the self-insured losses were, to some extent, above the actuarially expected amount. Thirdly, there were sound regulatory reasons, learned regulatory experience, why the AER had changed its approach in this way. The question was not whether individual categories were lumpy, and should therefore be hived off and subject to a bottom-up interrogation which imposed additional demands on the regulator and, in practical terms, impaired its ability to conduct meaningful assessments. The AER’s approach focussed on assessing the overall opex which was not, on any of the material before the Tribunal, materially lumpy from year to year, even when, as happened in 2015, a large self-insurance event occurred. Finally, on the materially preferable point, the threshold at which that pass-through event kicked in was 1 per cent of the annual regulated revenue which, for AusNet, was $6 to $7 million, and so that mitigated the risk of underfunding.
158 The Queensland Electricity Users Network (the QEUN) submitted that self-insurance referred to the practice of insuring one’s own interests by maintaining a fund to cover possible losses, rather than by purchasing an insurance policy. Pursuant to NER r 6.4.3(a)(7), a distribution network service provider’s opex was one building block that formed part of its annual revenue requirement.
159 AusNet submitted that self-insurance costs should be accorded the same treatment for opex purposes as the costs of an insurance premium and that the AER was in error for basing its analysis on self-insurance losses rather than self-insurance costs.
160 The QEUN maintained that an insurance premium provided assurance to consumers that the network business did have the capacity to restore the network in a timely manner following a natural disaster. This minimised the financial and lifestyle hardship caused to consumers connected to that section of the national grid and to other parts of the national grid remote to the damaged section. The ability to self-insure was dependent on the profitability of the network business and the dividends declared to shareholders. Self-insurance could only provide assurance of timely restoration if the network business has a locked reserve for natural disasters.
161 If the request by AusNet for an additional $8.46 million to cover self-insurance was granted, the opex for AusNet and consequently its allowed revenue cap would rise by $8.46 million. The AER had informed the QEUN that the allowed revenue cap of a distribution network business could be spent as the business saw fit. Consequently, the $8.46 million could be provided to shareholders as a dividend and AusNet could request a positive pass through if the cost of a bushfire exceeded 1% of AusNet’s annual allowed revenue.
162 As consumers throughout the NEM vehemently believed that electricity prices were already at unsustainable levels, any further increase in power bills would further exacerbate the trend of consumers accessing hardship programs and would cause bill shock.
The Tribunal’s analysis
163 We turn to consider these submissions.
164 The Tribunal finds that none of the four claimed errors of fact is made out.
165 AusNet has not established that the AER equated self-insured losses in the base year with the cost to AusNet of self-insurance. The Tribunal agrees that the AER’s approach to assessing base year opex had the same effect as assuming that self-insurance costs were equated to self-insurance losses, but concludes that that assumption had no material consequences for the total of the opex allowance assessed by the AER. That is, because the AER derived the opex allowance from a base-year aggregate, it was only required (by r 6.12.1(4)(ii)) to be satisfied reasonably that the allowance reflected the efficient operating costs of a prudent service provider (meeting the operating expenditure objectives in r 6.5.6(a)).
166 It is clear to the Tribunal that the AER was aware of the industry-wide trends in, and relative contributions to the total of, the various separate operating cost categories. The Tribunal concludes that the AER did not err in deciding that an aggregated, “top-down” approach to opex could satisfy the opex criteria. In another case, this conclusion might be tempered by a potentially larger problem with the AER’s implementation of this approach because, as determined in Ausgrid, the AER’s benchmarking methodology for determining the efficiency of revealed expenditure is flawed. Consequently, and despite the AER’s review of trends in components of operating expenditures, its starting point for assessing opex is not necessarily as sound as the AER says. This issue was not raised by AusNet. The Tribunal in these circumstances proceeds on the basis that the AER has appropriately identified an efficient level of initial opex, necessarily including an amount that compensates AusNet for its self-insurance costs (the equivalent of a premium for actual insurance, not the expenditure on self-insured outlays).
167 Any “error” or approximation arising from the decision not to separately assess self-insurance costs is then a natural consequence of the AER’s aggregate method for assessing opex and does not constitute an error of fact.
168 Next, we do not conclude that the AER’s treatment of the volatility of self-insured losses as against the relative stability of the cost of self-insurance reveals error. The only potential source of error in the AER’s aggregate approach would be if the inclusion of self-insurance losses, instead of, say, an actuarially determined provision in lieu of a premium for actual insurance, meant the aggregate base-year opex was materially different from (understating or overstating) an amount that would be needed to satisfy the operating expenditure objectives. The AER was aware of this potential error and was reasonably able to conclude that the magnitude of any error was unlikely to be material. The AER’s decisions to assess separately some other categories of expenditure were irrelevant to this conclusion.
169 Thirdly, and consequent on the first two points, the claimed failure of the AER to find that the cost of self-insurance to a DNSP was, in substance, a self-insurance premium conceptually equivalent to the premium for external insurance of the same risks, so that it required a category-specific allowance or forecast on an actuarial basis to be included in the base year, was not in our opinion an error of fact but was a result of the methodology adopted by the AER.
170 The fourth claimed error of fact, that the AER found that self-insured insurance losses could be controlled or funded by reducing costs in other areas when self-insured losses were unpredictable and volatile and highly variable again in our opinion was not an error of fact but a choice of methodology.
171 It follows that the incorrect exercise of discretion, as put on behalf of AusNet, also fails, based as it was on the claim that none of the reasons advanced by the AER in consideration of self-insurance forecast were permitted by the rules, were relevant to the opex criteria, or relevant to the opex objectives.
172 We also reject the claim that either the reasoning or the conclusion of the AER in this respect was unreasonable. In our opinion the AER’s decision is justified by reference to its stated reasons. Further, there was not an arbitrary and capricious result which had not been determined by reference to the applicable criteria in the rules. We do not accept that the AER’s decision was unreasonable on the ground that the component of opex in respect of self-insurance, being one year’s losses, was not reflective of the actual reasonable forecast cost to AusNet of self-insured losses in the future.
173 For these reasons, the Tribunal affirms the reviewable regulatory decision, being the Final Decision: AusNet Services distribution determination 2016 to 2020.
I certify that the preceding one hundred and seventy-three (173) numbered paragraphs are a true copy of the Reasons for Determination herein of the Honourable Justice Robertson, Mr RF Shogren and Dr DR Abraham .