Solving for x the New South Wales Gas Supply Cliff

Transcription

1 Solving for x the New South Wales Gas Supply Cliff Paul Simshauser and Tim Nelson Level 6, 144 Edward Street Brisbane, QLD 4001 March 2014 Abstract On Australia s east coast over the period , we forecast that aggregate demand for natural gas will increase three-fold, from 700 PJ to 2,100 PJ per annum, while our forecast of system coincident peak demand increases 2.4 times, from 2,790 TJ to 6,690 TJ per day. This extraordinary growth is being driven by the development of three Liquefied Natural Gas plants at Gladstone, Queensland. Almost simultaneously, a non-trivial quantity of existing domestic gas contracts currently supplying NSW will mature. Much of that gas has been recontracted to LNG producers in Queensland thus creating a gas supply cliff in NSW. Compounding matters, recent policy developments have placed binding constraints over the development of new gas supplies in NSW. In this article, we present our dynamic partial equilibrium model of the interconnected gas system and produce forecasts with daily resolution. We find that absent additional supply-side development, unserved load events will remain more than a theoretical possibility due to inter-temporal spatial constraints Keywords: Gas Markets, Energy Policy, Energy Security. JEL Codes: L95, L98, Q41 and Q Introduction Australia s east coast gas market, which spans Queensland, New South Wales (NSW), the Australian Capital Territory (ACT), Victoria, South Australia and Tasmania, is undergoing a large demand-side shock. Over the past 10 years, the demand for natural gas has averaged yearon-year growth of 2.1%. In 2013, aggregate demand was about 700 peta joules per annum (PJ/a) with a coincident peak winter load of 2,690 tera joules per day (TJ/d). By 2016, just three years later, we forecast gas demand to rise to 2,100 PJ/a with a system-wide coincident peak winter load of 6,690 TJ/d. This represents a three-fold increase in aggregate demand and a 2.4 times increase in peak load driven by the development of three Liquefied Natural Gas (LNG) facilities in Gladstone comprising 6 trains, each with theoretical ex-field loads of between PJ/a. We are unaware of any mature, large-scale national energy markets experiencing a three-fold increase in aggregate demand in such a short period of time. The first of the LNG trains will be commissioned in Q4 2014, with the balance of the six trains commissioned in rapid succession thereafter. The history of LNG developments can be traced back to discoveries of large reserves of Coal Seam Gas (CSG) in Queensland. 1 Queensland has been highly successful in developing its gas industry with proved plus probable (i.e. 2P 2 ) reserves escalating dramatically from 3,400 PJ in Paul Simshauser is the Chief Economist at AGL Energy Ltd and Professor of Economics at Griffith University. Tim Nelson is Head of Economics & Sustainability at AGL Energy Ltd. The authors are indebted to Tony Stone for his work on our GPEM Model. We are also grateful for the insightful comments and criticisms on an earlier draft of this article provided by Paul Taliangis (Core Energy), Paul Hyslop (ACIL Allen), Dr Graeme Bethune (EnergyQuest), Mick McCormack (APA Group), Michael Fraser (AGL Energy), Paul Ashby (AGL Energy) and AGL Energy s Applied Economic and Policy Research Council (See Section 9). However, all views, errors and omissions are entirely the responsibility of the authors. Our declaration is contained in Section 9. 1 The CSG resources were known to exist at least as far back as the 1980s. Advances in drilling technology enabled the resource to be booked as economically recoverable reserves during the 2000s. 2 1P or proved reserves are those thought to be reasonably certain (i.e. 90% confidence limit) of being recovered (i.e. production wells have been drilled). 2P or proved plus probable reserves have a 50% confidence limit (i.e. commercial gas flow from a pilot well has been demonstrated as a stabilised flow over several months for a CSG well). 3P or proved plus probable plus possible reserves have a 10% confidence limit of being recovered. Page 1

2 2005 to 41,200 PJ by During this discovery and appraisal phase, it was evidently clear to resource owners that the east coast gas market was not sufficiently large enough to enable the monetisation of reserves in suitable timeframes and at the scale necessary to maximise profit, and so developing an export market for natural gas in the form of LNG was a logical strategic solution. Not only would it result in the rapid expansion of aggregate demand, but would also have the benefit of linking domestic gas prices, historically ca$3 3 per gigajoule (/GJ), to the north Asian export market price of ca$6-9/gj equivalent ex-field netback price over the medium term (Simshauser et al, 2011). The practical evidence is that this strategy has worked. Forward gas prices have risen beyond the top-end of this range, and aggregate demand is now trending towards 2,100 PJ/a. For Queensland, LNG developments are forecast to produce $850 million per annum in state royalties and taxes once the industry is fully operational, along with 18,000 jobs and add $3 billion per annum to Gross State Product. 4 While the gas supply industry has experienced rapid growth in Queensland, NSW, which also has natural gas resources, placed a moratorium on CSG development over an 18 month period to September Then, after releasing a comprehensive regulatory framework for natural gas development, the NSW Government declared non-scientifically based 2 km exclusion zones during 2013 in response to community concerns. The energy industry grossly underestimated community sentiment because, in our opinion, the scientific, economic and engineering evidence points towards an industry that can develop natural gas safely and economically given an appropriate regulatory framework. 5 A dispassionate assessment could only conclude that the industry s initial response to community concern was entirely inadequate. By the end of 2013, industry engagement with the community had improved dramatically. But the NSW policy was in advanced stages of being reset and ultimately the arbitrary exclusion zones were announced in February 2013 and implemented in January There is, however, an irony with the restrictive policy in NSW by comparison to the accommodative policy in Queensland. NSW is the region most vulnerable to a large demand shock because it is almost completely reliant on interstate gas supplies to satisfy local demand. Under conditions of a supply crisis, state Governments constitutionally hold emergency powers and are capable of redirecting contracted gas to stabilise the energy market or maintain system pressures, and is initiated following exhaustion of Part 20 of the National Gas Rules and curtailment by gas network operators. 6 However, as the South Australian Energy Minister recently stated: if there is some sort of state-wide emergency and we need to keep power going, we can direct gas to the power stations But unless there has been an explosion or a pipeline issue, I would not allow any commercial arrangement [in South Australia] to be broken to supply gas to Sydney; I would not let NSW break a single commercial contract I would say fix your own regulatory problems (Martin, 2013, p.4). In our view, the South Australian position is likely to be commonly held in Queensland and Victoria. To be clear, east coast Australia has vast gas resources and substantial gas reserves. However, an inter-temporal mismatch between supply and demand is predictable because gas consumption patterns display diurnal cycles with marked seasonal variation (i.e. a winter bias), 3 All financial data are expressed in Australian Dollars unless otherwise specified. 4 See Queensland Government at 5 See for example the NSW Chief Scientist at data/assets/pdf_file/0016/31246/130730_1046_cse-csg-july-report.pdf 6 Part 20 of the National Gas Rules refers to the calling of Contingency Gas by the Australia Energy Market Operator, as the financial market operator. This is the last on-market step before the distribution network system operator initiates involuntary curtailment. Jurisdictional Energy Ministers or their delegate can direct the use of on-shore gas inside their jurisdiction, and this can extend to directions which stop interstate exports. This latter aspect is crucial to NSW's supply situation. The use of such emergency powers are of course intended to deal with short term supply disruption events (e.g. Longford, Varanus Island, extreme weather events). Systemic unserved load events, in a practical sense, should be resolved through higher prices and consequent demand-side adjustment. Page 2

3 and gas pipelines face capacity constraints on technical and economic grounds. As a result, a granular analysis of the east coast gas system is required in order to examine the true extent of any potential mismatch, and more importantly, what government policy and industry development options exist to help remedy the NSW gas supply cliff. In this article we present a dynamic partial equilibrium model of the east coast gas system with daily resolution over the five-year period Our model solves for differential equilibrium conditions given binding constraints associated with maximum theoretical field production, shipper nominations, pipeline capacity and storage facility limits. Our principal focus is not to present forecast gas prices by node, but rather, to examine the issue of energy security. We focus specifically on infrastructure augmentation and field production given our aggregate load forecast thereby enabling us to identify any unserved loads. 7 Our analytical procedure is therefore consistent with what Joskow (1976) has loosely described as the British Approach to energy system modelling. 8 This article is structured as follows. In Section 2, we present aggregate gas market demand data while Section 3 outlines gas reserves and production data. In Section 4 we outline the nature of the problem facing NSW. Our gas market model is explained in Section 5 along with our detailed demand-side, supply-side and infrastructure forecast assumptions. Section 6 presents the results of our quantitative analysis. Policy recommendations and concluding remarks follow. 2. The aggregate demand for natural gas The production and consumption of natural gas on the east coast of Australia can be traced back to with the development of Victoria s Bass Strait fields, the Roma fields in Queensland and the Cooper Basin in South Australia. Figure 1 presents annual consumption for the east coast states. Growth in aggregate demand over the past decade has averaged 2.1% yearon-year. Figure 1: East Coast aggregate demand for natural gas Source: AGL Energy, Core Energy, EnergyQuest, esaa. 7 Or as Paul Hyslop (ACIL Allen) suggested to the authors unmet demand. 8 As Joskow (1976) observed, the British approach tended to focus on optimising the supply-side for a given load curve. The American approach tended to use a homogeneous technology and focus on periodic and shifting loads. The French approach focused on optimising the supply-side given stochastic demand and hence represented a combination of the British and American approach (and pre-dated both). See for example Turvey (1969), Steiner (1957) and Boiteux (1949, 1956) respectively. 9 Note that Townsville demand, which is isolated from the existing east coast interconnected grid, has been excluded from this analysis. Page 3

4 The aggregate demand for natural gas on the east coast is set to treble due to export loads associated with LNG plants in Queensland, which is illustrated in Figure 2. Figure 2: Aggregate demand f Source: AGL Energy, Core Energy, Frontier Economics, ACIL Allen, EnergyQuest, esaa, One trend worth noting is that while we project total demand to rise rapidly to 2100 PJ/a in 2016 (driven by LNG exports), our forecast of underlying domestic gas demand incorporates a material contraction. Figure 2 notes that the domestic market falls from PJ/a in 2014 to just PJ/a in 2017 that is, a 20.4% contraction in domestic gas load. Our analysis tends to indicate that domestic consumption will peak during the period, driven primarily by the rise of gas used in power generation a trend that we forecast to reverse as the market value of gas increases relative to electricity. In Table 1, we provide domestic gas consumption data by consumer segment and by region for our weather corrected base year of Table 1: East Coast annual gas consumption (TJ/a) by region and consumer segment for 2012 REGION Residential & Industrial & Power SME Commercial Generation Total ACT 7, , ,875.6 NSW 37, , , ,673.9 VIC 116, , , ,713.9 QLD* 5, , , ,138.6 SA 10, , , ,847.3 TAS , , ,722.0 TOTAL 178, , , ,971.5 QLD* (ex Townsville) Mt Isa - 15, , ,150.5 Brisbane 5, , , ,292.7 Gladstone - 33, , ,647.3 Regional Qld , ,048.2 TOTAL 5, , , ,138.6 Source: AGL Energy, AEMO, Core Energy. Page 4

5 3. Aggregate supply of natural gas Aggregate 2P reserves on the east coast of Australia have risen sharply from 2008, driven primarily by natural gas exploration and development in Queensland. The build-up in reserves from 2005 through to the time of writing in 2014 is illustrated in Figure 3. Figure 3: East Coast 2P Reserves Source: EnergyQuest A key issue facing the market for natural gas is how these 2P reserves are notionally allocated, that is, with an export-market bias or a domestic market bias. As Figure 4 highlights, ca80% of existing 2P reserves are notionally allocated for export to North Asian markets in the form of LNG cargoes as they are owned by market participants with financial obligations (or corporate strategic targets) to supply LNG. While price rather than reserve ownership will be the ultimate determinant of how gas flows within the east-coast interconnected system, Figure 4 tends to indicate that ceteris paribus, the default directional flow for a majority of gas will be to LNG loads. Page 5

6 Figure 4: Notional allocation of 2P Reserves as at 2013 Source: EnergyQuest, AGL Energy Ltd. In Figure 5, we present regional gas production along with regional gas consumption for our base year of This is important because it highlights the reliance that NSW currently has on interstate transfers to satisfy indigenous demand. Figure 5: Regional Gas Production vs. Regional Gas Consumption Source: Core Energy, EnergyQuest, AGL Energy Ltd. Note that in 2012, local gas production represented only 4% of final NSW gas demand. The majority of this supply comes from the Camden CSG field, in Western Sydney 4. Solving for x the NSW gas supply cliff In this article, our primary interest relates to how the market for natural gas will dynamically adjust during the period in which LNG facilities commence their production run-up and simultaneously various long-dated gas supply contracts in NSW mature. Historic gas supply Page 6

7 agreements vary considerably in tenor, price and quantity. At the time of writing, our database indicates that there are about 70 industrial strength gas supply agreements. 10 These supply agreements can be further broken down into reseller agreements, power station agreements and industrial supply agreements. Reseller contracts range from 2-20 years in tenor with an average term-to-maturity of 12.1 years. Pricing for this otherwise homogeneous commodity on a unity-load factor basis spans the range of ca$3/gj for contracts written in the mid-2000s all the way through to contemporary oil price-linked contracts with values as high as ca$10/gj. Annual contract quantities vary considerably, from as low as 1 PJ/a to more than 100 PJ/a with an average contract quantity of 18.4 PJ/a. However, while the average contract could be thought of as 12.1 years in tenor with a contract quantity of 18.4 PJ/a and a market value of ca$6-9/gj, the maturity profile of existing contracts is highly congested. The average term-to-maturity for reseller contracts as at 2013 was just 4.7 years on a volume-weighted basis. Our database of gas supply agreements to power stations displays pricing trends reflective of their earlier commencement, with substantially longer tenors and an average term-to-maturity of 11.7 years. This is not entirely surprising given that one of the constraints associated with the development of project financed gas-fired power stations is a secure, long-dated fuel supply contract (Nelson and Simshauser, 2013). Finally, our analysis of industrial supply agreements indicates remaining terms-to-maturity of 4.8 years. Figure 6 presents a time series which focuses on gas supply agreements relevant to the NSW region as at the end of Note the sharp run-down in contracted supply during : Figure 6: The NSW gas supply cliff: f This cliff-edge in contracted gas supply to NSW is not a unique episode. An equivalent rundown in contracts occurred during the early-2000s and was resolved without fanfare through a series of renegotiations between gas producers, resellers and industrial consumers. However, the combination of rapid LNG load growth in Queensland and more recent (and unexpected) supplyside development constraints in NSW has created uncertainty as to how the situation in Figure 6 will be resolved this time around. Figure 7 combines our data from Figures 1, 2 and 6, and is the sole reason why the NSW gas supply cliff warrants policy and quantitative analysis at all. 10 We use the term industrial strength to reflect substantive gas supply agreements with power stations, resellers and large industrial customers who have transacted directly with gas producers. 11 Note that supply agreements would technically be higher than consumption but flexibility mechanisms typically ensure supply agreements equal consumption. Page 7

8 Figure 7: Aggregate gas demand vs. aggregate gas supply f Solving for x that is, the notional contracted gas supply gap in Figure 7 does not relate to reserves. The problem that we identify is seasonal and spatial. That is, the market needs to satisfy inter-temporal gas demand while remaining within the physical, technical and economic envelope of the interconnected gas system. The issue facing policymakers is that even under ideal operating conditions, there is insufficient gas production, pipeline and storage capacity during the height of winter to meet system-wide peak gas demand, particularly during working weekdays when diurnal patterns reach their maximum. Figure 8 illustrates the seasonal and diurnal variation in aggregate gas demand by contrasting daily winter consumption (i.e. Sunday 1 July 2012 to Tuesday 31 July 2012) with summer consumption (i.e. Sunday 1 January 2012 to Tuesday 31 January 2012) and annual average consumption by day. Note that the peak winter load is 80% higher than summer load. Figure 8: Winter vs. summer gas load Page 8

9 High pressure natural gas pipelines are vitally important to the proper functioning of the interconnected gas system. They connect locationally distributed gas demand with locationally distributed gas supply sources, and are necessarily a scarce resource because of the substantial upfront investment commitment and high ongoing fixed capital and operating costs. In the absence of quantitative analysis, one might be tempted to conclude that the NSW gas supply cliff can be readily resolved by rapidly expanding existing gas pipeline capacity and storage infrastructure between NSW and Victoria, for example. This is technically possible, and as our modelling later reveals, all economic options will be absolutely crucial. But as with all network infrastructure assets there is an economic saturation point major augmentation of pipeline capacity as the sole augmentation option would be financially unviable in the long run given demand uncertainty, and because at the margins, other lower-cost supply options exist. Indeed, the financial cost of solving for x solely through pipeline augmentation would almost certainly do more damage than good to the proper functioning of the gas market, and would result in sizable welfare losses. 12 Economically efficient marginal pipeline compression and looping has been identified and we expect those expansions to be underwritten by gas shippers, and committed to by profit maximising pipeliners. We present the beneficial effects of these economic infrastructure expansion projects later in Section 6. In this respect, the market is responding as expected at least where policy constraints do not exist. From a practical perspective, to the extent that any residual (and transient) mismatch exists between supply and demand, price-sensitive gas-intensive loads may be forced to exit. However, it is worth reviewing what happens when unplanned system shortages occur, and specifically, which customer segments are exposed. There are 1.3 million residential and business connections to the reticulated gas network in NSW (esaa, 2013). Of these, 1.1 million are defined as household and commercial users, with the balance being industrial users and as IPART (2012) note, there are about 450 large industrial users. If system security is threatened due to insufficient supply being available to meet the aggregate demand of all users, even after dispatch of any contingency gas, then the gas network operator commences curtailment. The activation of curtailment procedures is authorised under the network operator s approved access arrangement, which has the status of subordinate energy legislation in that jurisdiction. Such procedures are intended for short-term imbalances. Under more severe conditions, the NSW Government will invariably be required to direct available supply (i.e. local gas production and any interstate gas imports) to some users and not others meaning that under The Energy and Utilities Administration Act 1987, and specifically s24-26, the NSW Minister for Energy has the power to override the legally binding gas supply agreements of resellers and the largest industrial consumers. Emergency services and hospitals are likely to be prioritised over households, who are in turn likely to be prioritised over industrial consumers. In short, the largest industrial users are progressively shed from the system until physical equilibrium between forecast demand and supply is restored, regardless of the existence of legally binding gas supply agreements. 13 Quantifying the economic impacts of an inherent imbalance in the market for natural gas is a difficult exercise and requires Computable General Equilibrium Modelling (and to be sure, is well beyond the scope of this article). 14 However, demonstrating that short run adverse economic consequences are possible is relatively straight forward. In his classic article on marginal cost pricing, Hotelling (1938) noted long ago that the economic way to handle scarcity situations is to charge a sufficiently high price to limit demand. And so gas-intensive industrial users in the NSW region who have supply contracts maturing over the period will face one of three general options during an episode of scarcity; (1) pay a substantially higher price for natural 12 Material pipeline expansions from Victoria to meet Sydney s winter gas load would represent substantially under-utilised assets and face perennial stranding risk from further demand destruction or expansion in indigenous supply. Such uncertainty would make pipeline augmentation (to meet peak loads) unbankable for the private sector, and too costly for a state or Federal Government to fund. 13 The NSW Government last used its emergency powers in 2007 when gas shortfalls, caused by unexpected weather-driven coincidental peaks in different jurisdictions, resulted in around 35% of large industrial consumers being curtailed for two days. 14 For an example of such CGE Modelling, see ACIL Allen (2013). Page 9

10 gas supply; (2) cease trading; or (3) reduce production. Under these general conditions, it is difficult to imagine employment levels associated with NSW manufacturers being completely unaffected in the short run. At this point we would caution readers not to draw alarmist conclusions about the likely state of the broader economy. While ABS data reveals that the NSW manufacturing sector has ca185,000 employees, not all manufacturers are gas-intensive, and manufacturers that are gas-intensive (and therefore at risk) will invariably represent a small share of total national employment. However, Burgan and Spoehr (2013) and Brain (2013) analysed the macroeconomic effects and employment cycles associated with the closure of manufacturing plants in South Australia. Key findings from their study is that while the broader macro economy may barely register the change, recessionary conditions can appear in the immediately affected local community with pronounced flow-on effects around the plant closure location. As Burgan and Spoehr (2013) and Brain (2013) explain in the manufacturing case, during the first quarter after plant closure, job losses stood at around 6,500 but over the four years following swelled to almost 12,000 with much of this occurring in three suburbs. 15 Figure 9 shows the locational relationship between manufacturing-related employment and industrial gas consumers in NSW. The line series, measured on the LHS y-axis, presents manufacturing jobs as a percentage of total employment using Australian Bureau of Statistics NSW electoral district data, assembled in descending order. The bar series, measured on the RHS y-axis, presents AGL data on the number of large gas consuming sites within that district. So for example in Smithfield, 17.3% of the workforce is estimated to be employed in manufacturing, and AGL consumer account data indicates there are at least 12 large gas consuming sites. At the other extreme, Vaucluse has just 3% of its workforce estimated to be involved in manufacturing employment with no large gas consuming sites. Figure 9: Proportion of manufacturing jobs and gas consuming sites by location in NSW Source: ABS and AGL data 15 One peer reviewer correctly pointed out that the South Australian manufacturing case involved a single industry with highly specific skills, and the economic analysis did not utilise CGE Modelling. In the case of a gas market disruption event in NSW (i.e. either demand destruction or energy shortages), employment impacts in NSW are more likely to be dispersed and transient in nature. Page 10

11 Our observation is that the overwhelming majority of public analysis and debate relating to eastcoast energy security has thus far relied on, at best, annualised modelling results (while in a surprising number of instances, such debate contains no quantitative analysis whatsoever). Annualised data gives a broad indication of the overarching pricing pressures that customers on the east-coast are facing. However, annualised analysis of the gas market, given the substantial seasonal variation in aggregate demand, is simply unable to identify binding inter-nodal pipeline constraints and intra-zonal storage limitations, and in turn, the frequency and intensity of any unserved load events. This is important given that there is at least the potential for regulatory intervention to override gas supply contracts under such conditions. In order to understand the extent of inter-temporal mismatches between aggregate gas demand and supply and the binding production, pipeline and storage constraints facing gas producers, shippers and consumers, a dynamic partial equilibrium model of the interconnected gas system with daily resolution is necessary which we present in Section Dynamic partial equilibrium model of the interconnected gas market: GPEM Model Our gas model (GPEM Model) is a template interconnected gas system model that can be modified to mimic local market conditions. The GPEM Model assumes gas can be shipped from any supplier to any consumer subject to pipeline constraints, along with gas shipper nomination constraints. A series of regions exist with each node comprising the three domestic consumer segments identified in Table 1, along with the export LNG terminal segment at Gladstone. Nodes are interconnected via a series of 19 high pressure gas pipelines. Our GPEM Model ultimately seeks to maximise welfare in the market for natural gas, and this objective is implemented formally by maximising the sum of consumer and producer surplus after satisfying differentiable equilibrium conditions. 5.1 Definition of Regions and Suppliers In the GPEM Model, Ɲ is the ordered set of regions or nodes in our interconnected gas system and Ɲ is the number of nodes in the set. Let ƞ i be node i where i {1.. Ɲ } ^ ƞ i Ɲ. (1) Let Q i be the maximum demand for all consumer segments at node ƞ i expressed in TJ/d. Let Ψ i be the set of gas suppliers at node ƞ i. Let Ṗ ψi be the maximum productive capacity of gas supplier ψ at node ƞ i. Let ρ ψi be the quantity of gas supplied at node ƞ i by supplier ψ where ψ {1.. Ψ i } (2) Let c i be the quantity of gas delivered to node ƞ i expressed in Tera Joules per day (TJ/d). 5.2 Definition of Pipelines and Transport Pathways In the GPEM Model, T is the ordered set of pipeline segments in the system and T is the number of segments in the set. Let t j be node j where j {1.. T } ^ t j T (3) Let Ʊ j and j be the two nodes that are directly connected to pipeline segment t j where (4) And let f j be gas flow on pipeline segment t j from Ʊ j to j expressed in TJ/d. Page 11

12 Let R be the ordered set of all paths. Let R k be path k between two nodes ƞ x and ƞ y. Let r kj be node j in path R k where j {1.. R k } ^ r kj R k (5) Let T r be the ordered set of pipeline segments in path R k. Let t kj be pipeline segment j in path R k where j {1..( R k -1)}. (6) Let fc j be the maximum allowed flow along pipeline segment t j. Let fm j be the minimum allowed flow along pipeline t j. Let fr r be the flow of gas along path R k. Let ᵽ k be the cost of shipping 1 unit of gas (i.e. 1 TJ of gas) along path k, subject to: k,x,y r kx r ky x y (7) and t j [Ʊ j = r ki j = r k(i+1) ] [ j = r ki Ʊ j = r k(i+1) ] (8) The purpose of equation (7) is to ensure that each node appears only once in a path, while the purpose of equation (8) is to ensure that all nodes are connected via pipelines. 5.3 Model Calculations The flow on any given pipeline is the sum of flows attributed to all paths (that is, forward flows less reverse flows), as follows: (9) The clearing vector of quantities demanded or supplied (including from storage facilities) in node i = 1 n, is given by the sum of flows in all paths starting at that node, less flows in paths ending at that node if applicable: (10) Net positive quantities at a node are considered to be net supply ρ ψi and negative quantities imply net demand c i : { (11) 5.4 Demand Functions Let C i (q) be the valuation that consumer segments at node ƞ i are willing to pay for quantity q TJ of gas. We explicitly assume that demand in each period i to be independent of other demand periods. Let P ψi (q) be the prices that supplier ψ expects to receive for supplying q TJ of gas at node ƞ i. 5.5 Objective Function: Optimal welfare will be reached by maximising the sum of producer and consumer surplus, given by the integrals of demand curves less gas production and pipeline costs. The objective function is therefore expressed as: Page 12

13 Subject to:. (12) fm i f i fc i ^ Ṗ ψi 5.6 The Netback Price of Natural Gas One of the crucial calculations within our GPEM Model is the netback price of gas as this provides guidance to the marginal valuation C i (q) of LNG producers at the Wallumbilla node ƞ w. The netback price of gas is important because it identifies the price for which an LNG exporter is financially indifferent as to whether at the margins they produce LNG and load it onto a ship, or sell their intended feedstock back to the broader gas market (i.e. to another LNG producer or to a domestic market participant). LNG exports from Australia to North Asia are generally referenced against the most popularly traded Asian energy index derivative, the Japanese Crude Cocktail or JCC 16, which is an average of the roughly top 20 crude oils by volume in that country. 17 Provided the clearing price of oil (expressed in dollars per barrel or US$/bbl) remains within a non-extreme trading range, then a simple rule-of-thumb netback price of gas (in the form of LNG fob cargo at Gladstone in US$/GJ) can be thought of as roughly 13.7% of the oil price. 18 For our purposes, we are interested in a more granular analysis referenced to our Wallumbilla node. In practice, LNG export sales contracts are often subject to an S-Curve which has the effect of muting extreme oil price movements to buyers and sellers by placing partial caps and floors on the traded LNG price. Additionally, as we are interested in the marginal price of east coast Australian gas supply, then the LNG supplier with the highest contract value is the relevant focal point for netback calculations. Our view is that this is the GLNG project, which is thought to have a netback price of ca14% of the JCC. Our netback model of gas prices ex-field, which we take to be ex-wallumbilla, is expressed as follows: [ [ ( {. }. ) (.. ) ( )] ( ). [. ]. (1 )] [ ( {.. } {. }. ) (. ) (1 )] (13) The ex-field netback price of gas ( is calculated by taking the LNG price Delivered Ex- Ship ( into North Asia. We make use of the historical relationship between the price of Brent Crude and JCC in producing our estimates via historically derived intercept and slope ( coefficients. The S-Curve which is then applied to our oil price estimate uses conventional LNG intercept terms ( and slope (, with the details of how this is applied set out in equation (14).. [. ]. [. ] { (14) Returning to equation (13), our value for is then converted from US$/MMBtu to US$/GJ by dividing this result with the appropriate conversion constant. Boil-off losses (b) are then deducted. To transform the loss-adjusted LNG export price (expressed in USD) to an AUD 16 Japanese Crude Cocktail is in fact a nickname for the JCC. The formal name is Japan Customs-cleared Crude. 17 JCC reference prices can be obtained from the Petroleum Association of Japan at 18 It is worth noting that 13.7% x JCC (expressed in US$/GJ) is the equivalent of 14.5% x JCC (expressed in US$/MMBtu) the latter being conventional units for LNG pricing. Page 13

14 LNG fob price at Gladstone, boil-off costs, shipping costs and trading costs are deducted, with the resulting outcome converted into Australian dollars at our assumed exchange rate. In order to arrive at a suitable estimate of the ex-field netback price of gas ), the cost of Liquefaction ( ) and inter-nodal pipeline tariffs ( ) need to be deducted. Cost and quantity streams associated with LNG terminals are discounted at a nominal pre-tax hurdle rate. The allin cost of liquefaction at the j th plant is derived by discounting the cost streams associated with the initial LNG plant capital and ongoing capital works, along with fixed and variable O&M costs. Cost streams are the subject to projected inflation rates, (. Plant output, which is initially expressed in Mt pa for producing the various cost streams, is subject to the constant term which in this instance will convert LNG cargo from Mt/a to GJ/a and is then escalated at the relevant revenue inflator and discounted by the cost of capital. Dividing the present value of the cost stream by the present value of the quantity stream produces the long run marginal cost of an LNG terminal, ( ). Inter-nodal pipeline tariffs ( ) are derived by discounting the cost streams associated with the initial pipeline capital costs and ongoing capital works, along with fixed operating costs, all of which are subject to projected inflation rates, (. Annual flow rates are also escalated at projected inflation rates, (. Dividing the present value of the cost streams by the present value of the quantity streams then produces our long run average cost of transportation, ( ). An issue of considerable importance during the ramp-up phase of LNG productive capacity in terms of defining the relevant value for is whether short run or long run costs are the defining variable in relation to netback calculations, and specifically, our value for. Over the long run, it seems clear enough that the appropriate value to ascribe in equation (13) is indeed the full value for. However, this is not in our view the relevant value in the short run. Utilising a short run value for the j th LNG plant, would have the effect of increasing, quite materially, the ex-field netback price of gas. Our logic here, which all peer reviewers were in strong agreement with, is as follows. Assume the j th LNG plant has productive capacity of Ṗ but in the event only has shipper nominations equal to ṗ such that Ṗ > ṗ during the initial commissioning year(s). In this instance, is it the long run average cost of liquefaction or their short run marginal cost of LNG production from the j th plant that will maximise profit under conditions of scarcity? It seems clear to us that it is the latter, and so in the short run, particularly during the period of our analysis 19, the appropriate value for will indeed be elevated through the use of a short run marginal cost for. 5.7 Input Assumptions Regional nodes ƞ i in the GPEM Model are best viewed in the context of an East Coast map, which we present in Figure 10. The major pipeline routes are also included along with the pipelines under active construction at the time of writing. 19 One reviewer noted that in the long run, during any supply-side disruption event (such as major flooding), short run dynamics would once again prevail. Page 14

15 Figure 10: GPEM Model Nodes and Pipelines Major gas pipelines (t j ) and their associated maximum flow rates (fc i ) and tariffs (ᵽ k ) are presented in Table 2. Table 2: Pipelines and Pipeline Capacity Gas Pipeline Pipeline name Length From Node To Node Max Flow Tariff (km) (TJ/d) ($/GJ) (t j ) (ƞ i ) (ƞ i ) (fc i ) (ᵽ k ) CBR Canberra to Dalton 58 Dalton Canberra 77 $0.04 CGP Carpentaria Gas Pipeline 840 Ballera Mt Isa 119 $1.44 EGP Eastern Gas Pipeline 797 Longford Sydney 294 $0.96 LMP Longford to Melbourne Pipeline 174 Longford Melbourne 1030 $0.24 MAP Moomba to Adelaide Pipeline 1185 Moomba Adelaide 253 $0.50 MSP Moomba to Sydney Pipeline 1300 Moomba Sydney 439 $0.82 NVI NSW - Victoria Interconnect 88 Culcairn Young 71 $0.15 NVI_1 NSW - Victoria Interconnect 320 Melbourne Culcairn 92 $0.32 QGP Queensland Gas Pipeline 627 Wallumbilla Gladstone 145 $0.96 RBP Roma to Brisbane Pipeline 438 Wallumbilla Brisbane 240 $0.51 SEAGas South East Australia Gas Pipeline 680 Pt Campbell Adelaide 314 $0.58 SWP South West Pipeline 150 Pt Campbell Melbourne 353 $0.27 QSN QSN Link Pipeline Ballera Moomba 527 $ SWQP South West Queensland Pipeline Wallumbilla Ballera 404 $0.85 TGP_1 Tasmanian Gas Pipeline Longford Bell Bay 129 $ TGP_2 Tasmanian Gas Pipeline Bell Bay Hobart 129 $2.00 APLNG APLNG Pipeline 530 Surat Gladstone 1560 $0.55 QCLNG QCLNG Pipeline 540 Surat Gladstone 1510 $0.55 GLNG GLNG Pipeline 435 Surat Gladstone 1429 $0.55 Sources: ACIL Allen, AEMO, AGL Energy Ltd, APA, Frontier Economics. Given the number of pipeline routes R k between the set of nodes Ɲ, there are 159 plausible supply combinations and associated constraint equations along with 570 variables in solving for each Page 15

16 period (i.e. day). Our forecast for aggregate daily gas load through to 2018 uses the weathercorrected 2012 consumption data (by regional node and by consumer segment) as outlined in Table 1. These data are projected forward at differential growth rates for each segment in each node but as noted earlier in Section 2, overall we forecast domestic gas demand to contract by more than 20%. In the Residential & SME market segment, our forecast growth rates are broadly consistent with those of the Independent Market Operator (AEMO, 2013) with Queensland, South Australia, NSW, Victoria and Tasmania at 2.4%, 0.7%, 2.3%, 1.1% and 2.7% per annum, respectively. Our Commercial & Industrial market segment growth rates for Queensland, South Australia, NSW, Victoria and Tasmania are 1.7%, 0.4%, 1.3%, 0.2% and 1.3% per annum, respectively. 20 However, to be sure, while our growth rates are similar, our industrial and mass market load is 27 PJ/a lower than AEMO (2013) by Our forecast contractions in the demand for natural gas from the power station fleet 21 can only be described as very substantial. We assume 2014 gas-fired electricity generation largely mimics 2013, but in 2015 we envisage a 41% reduction, followed by a 12% fall in 2016, and a further 20% contraction in 2017 due to substitution effects (to coal-fired power generation). 22 To put this into perspective, during 2012 gas consumed by power stations represented 30.6% of total domestic demand during the winter peak season with total annual demand of PJ/a. By 2018, we forecast that gas used in power stations will contract to just 70.8 PJ/a, representing only 9.8% of domestic demand during the winter peak, and less than 3% of total system demand. 23 Daily gas consumption by each LNG terminal is modelled discretely given their importance to overall system demand. The initial ramping phase of the first LNG train is assumed to commence in 2014 with the sixth and final LNG train commencing from early We assume each facility follows a roughly 120-day commissioning schedule to full load. 24 Initially, on-site gas turbine plants are commissioned over a 30 day period, followed by an additional 30 day period of ancillary plant commissioning. First liquefaction therefore occurs after two months of commissioning activities with the plant operating at a full and stable load two months later. We apply a 1.5% forced outage rate to each LNG facility which is randomly distributed throughout the year via a Monte Carlo simulation. Our forecast aggregate gas load curve from by consumer segment with daily resolution is set out Figure These growth rates are also broadly consistent with AEMO (2013). A number of industrial consumers may well seek to convert gas boilers to coal over the long run in response to rising gas prices. However, our view is that as this would require associated environmental permitting due to metropolitan air-shed constraints, it is unlikely to occur en-masse within our study timeframes. Implicit in our assumptions is an own price elasticity of demand for natural gas in the Commercial & Industrial market segment of , a result broadly consistent with Hill and Cao (2012). 21 Our gas-fired generation has been exogenously derived from a dynamic partial equilibrium model of the National Electricity Market for each discrete gas-fired power station on the interconnected gas system. 22 In February 2014 Stanwell announced that it would withdraw their 385MW Swanbank CCGT plant from the energy market from October 2014 and re-sell their fuel to the gas market. See We understand from discussions with the LNG producers that other gas-fired generators such as Origin Energy s 630MW Darling Downs CCGT and Alinta s 450MW Breamar OCGT have also struck contracts to on-sell their fuel. However, unlike Stanwell s decision to mothball their plant, Darling Downs CCGT and Braemar OCGT plant will remain available for emergency duties in the electricity market for a certain (and limited) number of days each year that is, they will have the option to re-call their fuel during high electricity demand/price event days. 23 The reduction in gas-fired power generation (assuming an average heat rate of 8,000kJ/kWh) equates to ca21,350 GWh/a or the equivalent of 4,850 MW of gas-fired plant running at a utilisation rate of 50%. 24 We assume the first two LNG trains are commissioned in Q and therefore take longer than 120 days to achieve full load. One reviewer noted that following LNG terminal construction, 180 days may be more appropriate to account for commissioning activities. Page 16

17 Figure 11: Aggregate gas load forecast for the east coast f Our forecast aggregate supply curve for 2018 and set of gas suppliers (Ψ i ), is presented in Figure 12. The y-axis is based on our maximum theoretical daily quantities capable of being produced by each supplier (Ṗ ψi ). After incorporating storage facilities (at Iona and Newcastle), maximum theoretical daily output in 2018 is 6,865 TJ/d on an unconstrained basis, noting that gas pipeline and gas storage constraints will place a lower practical limit on quantities produced. 25 Note that Figure 12 also includes more than 300 TJ/d of productive capacity from fields in NSW that are subject to policy uncertainty and (non-scientific) exclusion zones. Figure 12: Aggregate gas supply for the east coast for 2018f 25 There is also the 150 TJ/d Dandenong storage facility in Victoria and the MSP also has storage capacity of 150 TJ/d. We have not incorporated these facilities in our GPEM Model. Dandenong is of course on the southern constraint to NSW, and the MSP will face the same issues as our NGSF i.e. restocking economically (see Section 6.3 for details). Page 17

18 In terms of our calculation for the netback price of gas ( ) Table 3 presents our key input assumptions. Note that our LNG plant discount rate is 12% in line with EnergyQuest (2010), Core Energy (2010) and Simshauser et al.(2011) while our gas pipeline discount rate is 11%. When these inputs are combined with equations (13) and (14), they produce a long run netback price of $7.95/GJ and a short run netback price of $11.20/GJ. Table 3: LNG netback input assumptions Variable Unit Value Energy Conversion MMBtu/GJ Boil-off Losses (%) b 1.90 Boil-off Costs US$/GJ Б j 0.25 Shipping Costs US$/GJ ς j 0.80 Trading Costs US$/GJ Ṫ j 0.50 AUD Exchange Rate A$/US$ x 0.85 Brent Intercept US$/bbl a 2.77 Brent Slope US$/bbl Ƅ 0.97 Brent Oil Price US$/bbl B i S Curve Intercept US$/MMBtu α S Curve Slope US$/MMBtu β LNG Cost of Capital (%) PV LNG Capex A$M/t 1,300 LNG Capital Works (%) 0.25 LNG Fixed O&M A$Mt pa LNG Variable O&M A$t pa 5.00 LNG Production Mt pa 7.36 Energy Conversion Mtpa/GJpa Pipeline Capex $/km/inch K j Pipeline Diameter inches z j 42 Pipeline Length km d j 540 Pipeline Opex $/km o j 69,000 Pipeline Cost of Capital (%) PV Inflation Rate (%) 2.50 Sources: Simshauser, Nelson and Doan (2011), Core Energy, AGL Energy, Bloomberg, ACIL Allen. 6. GPEM Model Results We have used our GPEM Model and input assumptions from Section 5 to simulate gas market conditions on the east coast of Australia over the five-year period In our base case scenario, all demand-side and supply-side investments in Queensland are expected to proceed as envisaged (i.e. LNG terminal loads and natural gas production, respectively). 6.1 Base case scenario In our base case scenario, we assume that all existing supply contracts and associated shipper nominations relevant to NSW remain in place as outlined in Figures 6-7. However, we specifically assume no new investment in infrastructure or field development occurs within NSW which enables us to establish a suitable business as usual baseline while meeting the objective function set out in equation (12). Our aggregate system-wide results for served and unserved load (TJ/d) are presented in Figure 13. LNG fleet unserved load, which are evident from inspection of Figure 13, occurs from April 2016 and continues to occur most days of the year. These shortages coincide with the commissioning of the sixth and final LNG train The second terminal at GLNG is scheduled to ramp-up to full load over a 2-3 year window. See slide 21 at (viewed Feb 2014). Page 18

19 Figure 13: GPEM base case model results for the five year period The dynamics of the east coast grid change quite fundamentally once the LNG fleet commences operation the directional flow of gas along the QSN Pipeline 27 eventually reverses with increasing intensity as aggregate demand in Queensland begins to outstrip localised supply. In other words, the Queensland region switches from an exporting region to an importing region. Based on our aggregate supply assumptions, total LNG fleet demand is unable to be met as Figure 14 clearly illustrates. Shortfalls occur for more than 330 days per annum (i.e. 90% of the year) and by as much as 250 TJ/d with a median shortage of 130 TJ/d. However, in the context of a 2,100 PJ/a system this represents a shortfall of just 2.0%. Figure 14: GPEM base case model results LNG Fleet We noted earlier that our long run marginal value for C i (q) associated with LNG terminals at the Wallumbilla hub is ca$7.95/gj but that under scarcity conditions which based on Figure The QSN pipeline connects the Moomba Production Node with the Ballera Interconnection Node. See Figure 10 for details. Page 19

20 evidently exist our forecast of the appropriate value for C i (q) for the LNG fleet is our short run netback estimate of ca$11.20/gj at the Wallumbilla hub. Taking these two statistics with the results from Figure 14 helps to explain why our east coast gas contract database has recently registered short run contracts struck at values of $10/GJ for the period While not evident in our aggregate supply function presented in Figure 12, we assume that in Queensland, theoretical productive capacity progressively increases to just over 4,100 TJ/d by 2016 and that Cooper Basin productive capacity increases from the current level of ca275 TJ/d over the next two years to ca420 TJ/d. To the extent that field production undershoots these parameters, the supply shortages outlined in Figure 14 would be amplified. In terms of the domestic market, given our aggregate demand assumptions set out in Figure 11, unserved load events begin to appear in the most energy security-vulnerable region (i.e. NSW) from May of 2016 as Figure 15 illustrates. In the analysis presented in Section 4 and Figure 6 in particular, we highlighted that a material quantity of gas supply contracts to NSW expire during Furthermore, Figure 5 demonstrated that NSW supplies less than 4% of indigenous demand and that development constraints currently exist. These factors, along with binding constraints along the Victoria-NSW pipelines, are key drivers of the unserved winter peak load. To be clear, no energy shortages occur in the South Australian, Victorian, Tasmanian, or Queensland domestic markets so our focus hereafter will remain on the NSW region. The LNG fleet on the other hand experiences persistent shortages in a manner largely consistent with Figure 14 in each subsequent scenario we present. Figure 15: GPEM base case model results NSW and ACT In our base case scenario, the extent of shortages in NSW is very material. In all, there are 118 days of energy shortages or event days per annum with varying degrees of intensity. To be sure, these results are mild by comparison to the Independent Market Operators equivalent case (AEMO, 2013). 28 The frequency distribution of the intensity of event days is illustrated in Figure 16 and displays a median shortage of 102 TJ/d. 29 The maximum shortage is 256 TJ/d and under these conditions, 39.7% of the NSW market would technically be unserved. 28 See in particular Figure 12 in AEMO (2013) which assumes no augmentation or enhanced NSW supply for the future year 2018 and has 193 days of unserved load. AEMO (2013) did not disclose their forecast NSW imbalance for the calendar years. 29 As a number of our peer reviewers noted, a 102 TJ/d shortage represents less than 1.5% of total east coast gas system load. However, this is nonetheless a significant volume of unserved load in the NSW region. Page 20

21 Figure 16: Intensity of event days in the base case scenario in 2016 A key issue for the energy industry, policymakers and society more generally is the welfare impact of transient (but systemic) unserved load events in NSW. From a strictly theoretical perspective, what would a median unserved load event of 102 TJ/d mean for the NSW economy in terms of demand rationing? In Figure 17, we have ranked AGL s largest 100 gas consumers in NSW by annual consumption in descending order (x-axis). On the LHS y-axis, we have plotted the maximum daily demand of each customer as an index, and the RHS y-axis illustrates the Cumulative Maximum Daily Quantity Curve. Based on this analysis, a median event day of 102 TJ/d would require the 48 largest industrial consumers in NSW to be shed from the system (including 56 days during the winter season in 2016) in line with how curtailment occurs in practice. As an aside, under a 60 th percentile outage of 123 TJ/d, all 100 consumers in Figure 17 (incorporating businesses as diverse as food manufacturers to International Hotels to large suburban entertainment venues) would need to be curtailed due to the diminishing maximum daily demand of Commercial & Industrial consumers. Figure 17: Maximum Daily Quantity of Large Industrial & Commercial Customers in NSW Page 21

22 Under this scenario (which to be sure is merely our base or business as usual scenario), the NSW gas network would be operating near the edge of collapse for almost a third of the year. This is an unacceptable outcome and so our modelling efforts must turn to the economic expansion of pipeline capacity, storage infrastructure and field development options. 6.2 Pipeline expansion scenario The first expansion to infrastructure we consider assumes that gas resellers underwrite additional looping and/or compression along the Eastern Gas Pipeline (EGP) and the Culcairn Interconnect. 30 We assume that the EGP is expanded by 60 TJ/d to 354 TJ/d and the Culcairn Interconnect is increased by 50TJ/d to 120 TJ/d prior to the crucial 2016 winter season. These are vitally important augmentations of scarce pipeline resources, and are likely to represent the last of the low hanging fruit. In the base case, these two pipelines between Victoria and NSW formed binding constraints the combined maximum flow of gas from Victoria to NSW along the EGP and the Culcairn Interconnect was limited to 366 TJ/d and was reached on 188 days during 2016 (i.e. more than 50% of the year). The constraints were binding in all but three days of the peak winter season and the only reason the pipelines were not fully loaded on those three winter days was due to a major forced outage of an LNG train in Gladstone (producing a surplus market). Pipeline capacity expansion will therefore be highly beneficial to gas shippers trying to clear the market in NSW, and to (Victorian) producer and (NSW) consumer welfare. Figure 18 presents the results of the pipeline expansion scenario. Figure 18: GPEM Pipeline Expansion - NSW The extent of unserved load is, in relative terms, reduced dramatically in NSW from 118 to 55 event days with the median deficit falling from 102 TJ/d to 50TJ/d. Supply shortages to the LNG fleet continues to persist in line with Figure 14. The frequency distribution of the intensity of event days in NSW in our pipeline expansion scenario is presented in Figure 19, along with markers from our base case scenario for comparative purposes. 30 The EGP connects the Longford Production Node in Victoria to the Sydney Demand Node. The Culcairn NSW-Vic Interconnect (NVI) connects the Moomba to Sydney (MSP) pipeline in NSW to the Victorian Transmission System. See Figure 10 for details. Page 22

23 Figure 19: Intensity of event days in NSW in the pipeline expansion scenario in Newcastle Gas Storage The expansion of pipelines had a highly beneficial effect but an unacceptable level of unserved load remains. In our next expansion to infrastructure, we introduce storage in NSW. The Newcastle Gas Storage Facility is being built specifically to deal with NSW peak domestic loads at a capital cost of about $300 million. It has a maximum practical injection rate of 120 TJ/d with total storage capacity of 1.5 PJ. The GPEM Model results are presented in Figures Our modelling of storage facilities is based on the utilisation of capacity at the first sign of trouble in order to minimise unserved load events until inventories are exhausted. 31 The addition of the Newcastle Facility reduces the number of event days materially, from 55 days down to 21 days in The changing structure in NSW consumer demand means that in 2017, event days are reduced to 12 per annum, but rise back to 21 event days in Figure 20: GPEM Pipeline & Newcastle Storage results for NSW In practice, storage inventories would be dispatched to maximise profit (i.e. targeting the highest value periods). Page 23

24 Note however that the storage facility is limited by its capacity of 1.5 PJ and re-injection rate of 10 TJ/d. From inspection of Figure 20, it is evident that the storage facility has been effective in reducing the number and intensity of event days from late-autumn (i.e. first week of May) through to early-winter. However, in this scenario NSW remains short energy and so Newcastle Storage inventories are exhausted by the start of July as Figure 21 reveals. The facility is unable to commence its re-stocking process (on an economic basis) until after the winter season, with inventories not fully restored until January the following year. Figure 21: 1.5PJ Newcastle Gas Storage Facility Inventory Balance In this particular scenario, while event days have declined in 2016 from 55 to 21, there has been an increase in the intensity of events, with median unserved load of 61 TJ/d (vs. 50 TJ/d prior to adding storage). Prima facie this may seem counter-intuitive, but as Nguyen (1976, p.242) noted: It may appear at first sight that the possibility of storage removes the peak load problem. However, storage is costly so that problems associated with peak demand are modified rather than removed Storage facilities cannot resolve energy shortages. Their intended purpose is to resolve capacity shortages. The frequency distribution of the intensity of the event days is presented in Figure 22, with the prior scenarios represented by the markers for comparative purposes. Page 24

25 Figure 22: Intensity of event days in the Newcastle Gas Storage scenario in Adding to the aggregate supply function GPEM Model results thus far tend to confirm that expanding the gas infrastructure serving NSW consumers is necessary, but not sufficient. The extent and severity of unserved load is reduced but far from eliminated and remains at entirely unacceptable levels. Fundamental expansion of the aggregate supply function is therefore required. The most sobering aspect of our subsequent analysis is that we cannot identify any supply-side expansion options inside the NSW boundary constraint prior to 2017 due to moratoriums placed on natural gas developments by successive administrations in prior periods. To be perfectly clear on this, absent demand rationing in other regions (i.e. LNG producers redirecting contracted gas to NSW), material supply increases in Queensland 32 (i.e. allowing gas swaps between Wallumbilla and Moomba) or additional demand destruction in NSW, the 2016 unserved load events in Figures 20 and 22 are otherwise unavoidable i.e. 21 event days with median unserved load of 61 TJ/d. In a practical sense, and as a number of peer reviewers emphasized, the forced exit of price-sensitive, gas-intensive manufacturing loads in NSW would seem almost inevitable. If so, our NSW/ACT load forecast would require further downward adjustment (i.e. our existing NSW/ACT load forecast, which contracts by 12.4% by 2017, would need to be further downgraded). To the extent that any residual shortfall may exist in NSW, we would expect LNG producers, gasfired power stations, gas shippers and large industrial consumers to initiate a series of transactional swaps and options so that NSW demand and supply clears without the need for emergency intervention by the NSW Minister for Energy. Indeed, gas sales contracts with embedded summer season call options exist between power stations and LNG producers. 33 However, in our opinion predicting participant behaviour under extreme energy market conditions is difficult at best. And so if such transactions are insufficient, or peak demand is higher than the market anticipates due to extreme weather events, coordinated supply curtailment in NSW will be invoked to forcibly solve for x in 2016 as occurred in NSW during the winter of One reviewer observed that some major gas users have pursued additional supply options from other non-conventional sources in South Australia. This has not been accounted for given the timeframe under which we are modelling (i.e. primarily ). 33 Specifically, certain gas-fired power stations will substantially reduce their output once the LNG terminals commence their ramp-up and on-sell their fuel supplies to the LNG producers. The power station will retain an ability to re-call some component of the fuel during the Q1 Summer period each year (i.e. to meet summer peak loads in the electricity market). Due to the confidentiality of such arrangements, we are uncertain as to whether call options also cover the domestic winter peak loads of industrial gas consumers. Page 25

26 Above all, our GPEM Model cannot solve for x in NSW in the absence of demand destruction or unserved load, and so the reliability of energy supply in NSW would seem to be uncertain for the 2016 year. Under these conditions, gas prices at Wallumbilla, Moomba and Sydney will almost certainly rise towards short run opportunity marginal cost (i.e. a short run LNG netback price adjusted for varying values for ) because the market is inherently short energy. And given temporal constraints between the timeframes of commercial decisions and their implementation, regulatory intervention in the form of ordered demand rationing during the winter 2016 period will remain at least more than a theoretical possibility. Our supply-side analysis must therefore turn to what is possible inside our NSW boundary constraint from At the time of writing, the policy and regulatory environment in NSW is still not conducive to investment commitment. However, if policy uncertainty can be resolved during early 2014 then the PJ/a Gloucester project may be capable of entering into production by Gloucester Stage One has the requisite State and Federal Government approvals and production could commence once a particular pilot (known as the Waukivory Pilot) has been approved. 35 Our understanding is that the Narrabri project, which was also capable of producing from 2017, has (as with Gloucester) been adversely affected by dynamic inconsistency. Our view is that 2018 may be a more realistic date for field production from Narrabri given that, at the time of writing, requisite approvals from the State and Federal Governments are pending. In Figure 23, we have opted to illustrate the impact of Gloucester on the security of energy supply in NSW, noting that the initial phase of Narrabri (i.e. feeding into the Central Ranges Pipeline) would in theory have an equivalent effect. Figure 23: GPEM Gloucester scenario NSW The results in Figure 23 reveal that there are no shortages from 2017 due to local NSW production, which we assume to be 62 TJ/d or 22.6 PJ/a. Figure 24 provides additional insights on energy security in NSW through the inventory balances of the Newcastle Gas Storage Facility. Note in Figure 24 that inventories are exhausted during 2016 (i.e. unserved load events persist). But inventory levels are non-zero throughout 2017 and 2018 which indicates that once local NSW natural gas fields enter production, energy security in NSW is restored, albeit with no margin for error. 34 NSW productive capacity Ṗ ψi in Figure 23 is based on deterministic assumptions whereas in practice they are of course subject to uncertainty. 35 For clarity, Gloucester will supply domestic loads for the life of the project. Page 26

27 Figure 24: 1.5PJ Newcastle Gas Storage Facility Inventory Balance The margin for error with Gloucester producing, or lack thereof, is illustrated in a static analysis in Figure 25. Here, we present two supply scenarios for NSW. The LHS of Figure 25 contrasts 2018 NSW load against the notional combined capacity of the expanded EGP and Culcairn Interconnects (from Victoria) along with production from the existing Camden field and the Gloucester project. Notice that the Newcastle Gas Storage Facility is vital for ensuring energy security during winter. The RHS of Figure 25 illustrates the additional capacity that would be associated with Narrabri (at an initial 50 TJ/d), Camden North (22 TJ/d) and Hunter (55 TJ/d). Recall that NSW currently produces 4% of indigenous demand. In Figure 25, local production increases to 19% in the LHS scenario (due to the addition of Gloucester) whereas the RHS scenario, which incorporates Narrabri, Camden North and Hunter, has local production at 50% of NSW aggregate demand. And, as Narrabri reaches its total projected output (of ca200 TJ/d), NSW production would rise to almost 90% of indigenous demand. Figure 25: 2018 Victorian Interconnect & NSW Production Capacity Scenarios Page 27

28 6.5 Optimality Scenarios When initially discussing our modelling results with industry colleagues, two questions were frequently raised what supply could be brought on under conditions of fast-tracked approvals, and was the current situation avoidable? In relation to the former question, we would expect that the RHS of Figure 25 could be achievable by In terms of the latter question, we used the GPEM Model to analyse an optimal scenario. Here, we assumed away all prior policy constraints, thus enabling the Gloucester, Camden North, Hunter and Narrabri fields to enter production in 2016 and 2017 (rather than 2017 and 2018). EGP and Culcairn Interconnect expansions and Newcastle Gas Storage Facility are also assumed to enter as previously described. Figure 26 illustrates the result of this optimal scenario - no unserved load in NSW. Figure 26: Optimal scenario NSW CSG Fields enter production one year earlier ( ) Another scenario we examined incorporated supply-side policy constraints in NSW (i.e. Gloucester in 2017 and Narrabri in 2018), and delaying by one year the sixth and final LNG train in Queensland. The relevant results are illustrated in Figures 27 and 28 and show that, with the benefit of hindsight, the potential for unserved load in NSW could have been eliminated. 36 However, what we have not identified is the undoubtedly substantial economic costs associated with delaying such a large investment in an LNG terminal by one year, including those associated with interrupting the work flow of a construction workforce, lost field production, pipeline utilisation and lost export sales. These opportunity costs are no doubt material and forced delay (even by one year) must also carry at least some probability of outright project collapse due to regulatory intervention. Whether such costs in Queensland outweigh the benefits of avoiding scarcity pricing in NSW is the domain of general equilibrium modelling. 36 To be fair to the Commonwealth and Queensland Governments, it would have been impossible for them to foresee the actions of successive NSW Governments and the associated consequential lack of new supplies. Page 28

29 Figure 27: GPEM model results LNG Fleet with sixth train delayed by one year Figure 28: GPEM model results NSW (with sixth LNG train delayed) Policy Implications and Concluding Remarks When two energy economists produce long-range forecasts with transient but nonetheless apparently unavoidable unserved load events, it must be accompanied with associated caveats, hence our listed assumptions in Section 5. However, based on those assumptions, there can be no doubt that our theoretical aggregate simultaneous maximum demand of 6,690 TJ/d in 2016 will outstrip our maximum theoretical supply given transport constraints. Stochastic demand and prevailing policy uncertainty around the future of prospective indigenous supplies in NSW means it would be uneconomic to remove these constraints solely through long-dated infrastructure (i.e. pipeline) investment commitments. Additional productive field capacity can eventually be brought on stream, but not within the timeframes required for NSW to avoid an especially high risk of unserved load events, manufacturing demand destruction and associated unemployment, or both. Page 29

30 In our GPEM model, which seeks to maximise welfare, the burden of unserved load was borne by industrial consumers in NSW and the LNG fleet in Gladstone. The LNG fleet experiences considerably larger energy shortages of up to 250 TJ/d from and for 330 days each year under all scenarios we envisage. NSW on the other hand would experience about 21 days of unserved load events (61 TJ/d median) in 2016 given our aggregate demand function, which to be sure, already incorporates a 20+% contraction in domestic gas demand. With the Gloucester field producing in 2017 and Narrabri in 2018, the security of supply in NSW is capable of being restored. Our model results do not envisage households or small businesses experiencing energy shortages in any region at any time. The burden of unserved load events could, in theory, fall entirely on the LNG producers. But in our view this is unlikely. Putting to one side LNG commissioning commitments, LNG plant minimum load constraints, LNG contract obligations, the imperfect substitutability of physical and financial LNG cargo delivery and the fact that two of the three Gladstone producers (GLNG and APLNG) will be keen to establish their credentials 37 as reliable LNG suppliers the relevant price of a commodity made available during scarcity conditions will transition to short run opportunity marginal cost, as Ng (1987) explains in considerable detail. We noted earlier that our long run approximation for the LNG netback price was about $7.95. However, we also noted that in the short run our netback price, marked at the Wallumbilla Hub, is more than $11/GJ well above the commonly cited ca$6-9/gj range. This is our estimation of the relevant value for the short run opportunity marginal cost of gas supply while LNG trains remain under-utilised. 38 The practical outworking of either result will invariably be what the energy industry describes as industrial demand destruction. The transmission of transient but uncharacteristically high gas prices through the economy, for an uncertain period of time, will likely result in the forced closure of some gas-intensive manufacturing and consequent unemployment. To be sure, with the exception of coal in some limited circumstances and subject to air quality constraints, there is no (economic) substitute fuel for gas used in many industrial applications. Simshauser and Nelson (2014) noted that not all calls for economic reform need to threaten imminent disaster if the policy recommendations are not speedily adopted. Such advice does not seem appropriate in this instance. There is clearly a need for urgency if our forecast results are even half correct. 39 And, our policy prescriptions had best be right. 7.1 Diverting Supply Over the course of the past 12 months, there have been spirited calls to introduce a domestic gas reservation policy by various industrial consumers. Such a policy would see existing gas reserves directed to the domestic market. In the special case of NSW, there are virtually no existing gas supplies to reserve and so such a policy would be ineffective in any event. In practice, given that mineral rights are the domain of state governments, success would require Queensland to apply a domestic gas reservation policy. It is not immediately obvious to us that this would make any sense at all for Queensland. After all, under no rational circumstances would LNG producers allow gas shortages to arise amongst Queensland domestic loads. And, such dynamically inconsistent policy would represent an extraordinary retrospective intervention by the state. Queensland would bear the entire (and 37 Our thanks to Dr Graeme Bethune (EnergyQuest) for pointing this out to us. 38 In theory, this represents a logical cap on pricing, albeit a cap that moves with global spot LNG prices and as one peer reviewer noted, Gladstone netback prices recently touched caus$18/mmbtu. 39 The independent market operator, AEMO (2013) has identified similar gas supply shortages for NSW in their modelling of the 2018 period which tends to discount any notion that our results are at odds with market fundamentals. Indeed, our peak demand forecast for NSW in 2018 is lower than AEMO (2013) by ca25 TJ/d. Page 30

31 significant) financial, reputational and sovereign cost of such a policy, while the primary benefits would be largely extracted by NSW. As Simshauser and Nelson (2012) demonstrate in considerable quantitative detail, capital market premiums associated with perceptions of sovereign risk can very quickly overrun the expected benefits of dynamically inconsistent policymaking. Before taking such action, Queensland would no doubt focus on the consequential risks to Queensland Bond pricing (and by implication, the government s balance sheet). 40 Besides which, state royalties and taxation streams arising from natural gas production and LNG exports no doubt form an important component of the Queensland Government s forward fiscal policy strategy, and we have little doubt that policymakers in that state will be focused on achieving the forecast $850 million during the 2016 year. A gas reservation policy in Queensland would therefore seem an unlikely outcome and without constitutional amendments, would appear almost impossible to provide a resolution to transient unserved load events in NSW in Economists generally baulk at the concept of domestic reservation policies because unless it is structured carefully, deadweight losses predictably arise. From a welfare economics perspective, a reservation policy amounts to the equivalent of a production tax (i.e. forcing gas suppliers to sell some component of their product below market value which, ceteris paribus, reduces the incentive to invest in new capacity) and a simultaneous consumption subsidy (i.e. resulting in over-consumption by industrial consumers). 41 Retrospective application of a domestic reservation policy (or any policy for that matter) would have dire consequences for Australia s international reputation. Forward LNG sales contracts and the $62.5 billion 42 of private investment in LNG terminals, gas pipelines, field exploration and field development required to support them, have been committed in good faith and are irreversible. Redirecting gas supplies committed (under legal contract) from LNG export facilities to domestic markets by retrospective policy intervention would rightly result in legal challenges, raise sovereign issues with our trading partners, and result in an acute form of dynamic inconsistency. As Simshauser (2014) explains, in capital-intensive industries, few factors could be more damaging to achieving welfare maximising outcomes than a dynamic inconsistency problem. When perceptions of policy uncertainty arise due to opportunistic changes by policymakers, firms will respond rationally (Kydland and Prescott, 1977). The origins of the literature on dynamic inconsistency can be traced back to Kydland and Prescott (1977), which in turn led to an entire field of economics research, applying to issues as broad as central banking and monetary policy, through to regulated industries and investment patterns (see for example Rogoff, 1985; Taylor, 1985; Batabyal, 1996; Virag, 1999; Haubrich, 2000; Brito et al. 2011; Simshauser, 2014). But Kydland and Prescott (1977, p.486) evidently had the US energy industry (amongst others) in mind with their original contribution: rational agents are not making investments in new sources of oil in anticipation that price controls will be instituted in the future. Currently there are those who propose to tax away excessive profits of the oil companies with the correct argument that this will not affect past decisions. But rational agents anticipate that such appropriations may be made in the future, and this expectation affects their current investment decisions, thereby reducing future supplies For these reasons, a domestic reservation policy applied retrospectively seems an unhelpful policy suggestion. 7.2 Restricting Demand Policymakers could investigate the plausibility of a prospective National Benefits Test for future LNG export capacity developments with a focus on the pacing and sequencing of new LNG 40 Our thanks to the Independent Chair of AGL s Applied Economic and Policy Research Council and former Queensland Treasury Corporation Director (Elizabeth Nosworthy) for pointing out this rather obvious corollary. 41 Deloitte Access Economics (2013, p.ii) estimates that the introduction of a domestic gas reservation policy would cost the Australian economy $6 billion in foregone GDP at Deloitte Access Economics (2013). Page 31

32 capacity relative to 2P reserves, given our federal system of government. 43 We remain tentative about such a policy because only general equilibrium modelling is capable of revealing whether a case genuinely exists. 44 Errors by central planners can be substantial, and we have already noted that the opportunity cost of delaying the marginal LNG producer in Queensland would be nontrivial. However, the USA has dealt with acute energy security issues for more than 40 years (i.e. dating back to the first OPEC oil price shock in 1973) and applies a national test to developments given their experience. To be clear, the design of any such policy should not generate domestic price subsidies, but aim to minimise destructive scarcity events that arise through the asymmetric information associated with a federal system of government. 7.3 Expanding Aggregate Supply In discussing east coast interconnected gas system scenarios with the authors, Professor Paul Stevens from University College London noted that Governments cannot change the geology, but they can change policy. And so in our view the most obvious policy response that Federal and State Governments should focus on is to remove non-scientific and therefore unnecessary regulatory and policy barriers to natural gas exploration and field development. 45 To be sure, exploration and development of natural gas can result in externalities, and externalities count when the objective function is to maximise welfare. Accordingly, our comments should not be interpreted as a suggestion to by-pass due process or obviate the requirement for best-practice regulatory frameworks associated with natural gas development. Following the introduction of non-scientific policy barriers in NSW, natural gas fields that should have been producing in 2016 to maximise welfare and contribute to NSW energy security are now being written-off the Balance Sheets of project proponents. The Camden North field, adjacent to the existing Camden field in NSW (which as an aside has been producing natural gas since 2001) is a case in point. Moratoriums and arbitrary setback zones which can be traced back at least as far as 2010 and uncertainty over how new regulatory frameworks will be implemented even at the time of writing, have collectively taken their toll on the potential of the aggregate supply function. Grundoff (2013), Denniss (2013) and Ogge (2013) have argued that expanding gas supply in NSW will have no beneficial effect on gas prices. This oversimplifies energy system dynamics and is not correct. 46 Adding substantially to the aggregate supply function to remove scarcity events would have the effect of reducing clearing prices from short run opportunity marginal cost back down to long run netback prices. Additionally, Sections 2-6 have described a fundamental demand-supply imbalance, and increasing supply in our GPEM Model has two primary effects. First, unserved load is eliminated hardly a trivial enhancement to welfare. Second, and more importantly, increasing aggregate supply will eventually place sufficient downward pressure on unit prices so that they fall below long run LNG netback prices that is, once export volumes are fully subscribed. And to be perfectly clear on this, gas demand from LNG terminals on the east coast is bounded. 47 LNG production is an immensely capital intensive 43 In the case of eastern Australia, this is unlikely to be necessary given the elevated cost of marginal gas supplies and the capital intensive nature of new LNG plant capacity. Specifically, the three LNG terminals currently under construction in Gladstone were committed when the expected structure of the aggregate supply function was considerably flatter (i.e. ca$3-4/gj). In our view, as the marginal cost of gas supply has been revealed (i.e. at elevated levels) it is unlikely that new LNG plant could enter profitably. The practical evidence, at least at the time of writing, is that BG Group s plans to expand their Gladstone LNG project would be removed from their global list of potential growth opportunities, and Shell have announced further delays to their proposed LNG Terminal at Gladstone. 44 Such a policy would presumably require cooperation and coordination between the Federal Government, and state governments given they hold mineral rights. 45 The Commonwealth Government has identified that given the tight timeframes associated with bringing on new supply, it is necessary to prioritise and expedite projects that can alleviate the prospect of shortages (Commonwealth Government, 2014). 46 Such economic constructs apply neatly to commodity markets like agriculture where transport constraints are minimal and barriers to debt and equity capital are trivial (i.e. at least by comparison to those facing a $10 billion LNG terminal). 47 Once a commodity faces bounded demand due to capital and transport constraints, price separation either side of the constraint is entirely predictable, and consistent with economic theory. For example, steaming coal prices at Queensland s Meandu Mine are Page 32

33 activity, and capital is a scarce resource particularly when the prime source of funding is structured project finance as is invariably the case with east coast LNG investments. At the time of writing, the consensus view among energy industry executives and energy stock analysts is that future east-coast LNG developments are unlikely any time soon (Nelson, 2013). 48 However, it is inarguable that the underlying cost structure of new field development, conventional and unconventional, are demonstrably higher than those undertaken in prior periods as Figure 12 clearly demonstrates. 49 It does not take complex dynamic partial equilibrium modelling to demonstrate these basic constructs. Let us consider a conventional welfare analysis of increasing gas supplies following a large demand shock. Figure 29 shows an initial aggregate daily supply curve in period 1 (S 1 ) serving a domestic gas market given by domestic demand curve (D d ). This results in an equilibrium whereby the domestic gas price is P d and the domestic quantity consumed is Q d. In period 2, LNG facilities are commissioned and export quantities result in a quantum shift in the aggregate demand function, from (D d ) to the domestic plus export demand curve (D d+e ). Note that the LNG facilities have some element of must run capacity 50 which is highly price inelastic, hence the parallel-shift component of demand curve D d+e. But LNG producers also have the ability to substitute some of their contracted output from alternative LNG plants (i.e. from within their own global portfolio of plant, or from the LNG spot market if it is available at lower marginal cost) hence our introduction of the kinked demand curve denoted by the segment ď-đ. Figure 29: Increase in gas demand (LNG) and gas supply The point that Grundoff (2013) and others could reasonably make is that while the aggregate supply function crosses the kinked segment at ďđ, domestic gas prices will be somewhat unresponsive to marginal increases in the aggregate supply function. But as we noted above, capital is scarce, east coast LNG capacity is bounded and the marginal cost of gas supplies from new fields are near double the levels when the LNG fleet was first committed. It should be obvious that few firms can afford to carry under-utilised, highly capital-intensive LNG equipment purely for the convenience of creating domestic supply scarcity for small residual volumes of domestic gas. As a result, once the aggregate supply function passes the kinked component of the demand curve on the east coast (at any point to the right of đ), gas prices will fall. This is ca$1.50/gj, less than half the current value of steaming coal in the seaborne market at ca$3.50/gj; and 2013 spot electricity prices in Queensland averaged $68.41/MWh whereas NSW averaged just $53.90/MWh again due to binding capital and transport constraints. 48 See also our Footnote 43, and our thanks to Paul Hyslop (ACIL Allen) for emphasizing this point with us. 49 As one peer reviewer noted, with Coal Seam Gas playing an increasing role in satisfying aggregate demand, the absence of any associated liquids or oil means that the historical cross subsidy arising from these products (implicit or in some cases explicit) is gradually being washed out of the market. 50 The inflexible production loads of LNG trains generally varies between 20-60% of nameplate capacity. Page 33