**1. Introduction**

#### *1.1. Background to Australia's National Electricity Market*

Australia's National Electricity Market (NEM) is an energy-only gross pool in which all generators bid into a central platform and are dispatched under a uniform first-price auction clearing mechanism. Being a mandatory gross pool, all generators must sell their output in the spot market, and energy retailers must buy all load from the spot market.

The volatility that accompanies organized energy-only spot markets, particularly those with a high VoLL (NEM value of lost load at AUD \$14,500/MWh is amongs<sup>t</sup> the highest in the world) creates the conditions necessary for active trade in forward contracts. While there is an almost endless array of derivative instruments, the three most commonly traded contracts are swaps, \$300 caps and increasingly, run-of-plant power purchase agreements (PPAs). Swaps and caps are traded both on-exchange and over-the-counter, generally over a 1–3 year tenor at quarterly resolution, with market liquidity running at ~300% of physical trade and considerable variation in liquidity by season and region. PPAs on the other hand tend to be long-dated (10–15 year), structured as run-of-plant instruments and designed specifically to underwrite the entry of variable renewable energy (VRE), that is, wind or utility-scale solar PV, with the market characterized by a small number of buyers and a large number of (potential) sellers/developers.

Australia's 20% renewable energy target (RET) is a renewable portfolio standard that requires energy retailers achieve a 20% renewable market share by 2020. The broad-based market scheme is mobilized through the trade of renewable energy certificates, and has been highly successful in facilitating VRE plant entry. The 20% RET is expected to be comfortably met by 2020.

Once the 20% RET has been met, the NEM faces a virtual policy vacuum in terms of future CO2 emissions reductions. The most recent attempt (in 2018) at designing a bipartisan approach to reduce power sector CO2 emissions in a manner compatible with Australia's international obligations and the NEM's energy-only market design involved a scheme known as the *National Energy Guarantee*. This market-based scheme was comprised of a reliability obligation (plant capacity) and an emissions obligation (CO2 emissions), with energy retailers being the liable parties. But landing a united energy and climate change policy architecture to guide the NEM's ongoing transition has proven to be a complex political task [1,2]. Indeed, the politics of a two-decades-long climate change policy war has underpinned the downfall of at least three sitting Australian Prime Ministers since 2010, including Prime Minister Malcolm Turnbull (by his own party) whilst pursuing the *National Energy Guarantee*.

Policy discontinuity has led to cyclical swings in VRE plant investment commitments. The Commonwealth Government's decision to review the 20% RET in early-2014 produced su fficient policy uncertainty to induce a transient cessation in buy-side PPA activity—such instruments typically being an essential ingredient to VRE project commitment in an energy-only market [2].

One response to the Commonwealth Government's climate change policy discontinuity has been the emergence of government-initiated contracts-for-di fferences (CfDs) undertaken unilaterally by sub-national governments—first by the Australian Capital Territory (wind, 2015), then Queensland (solar PV, 2016), South Australia (semi-CfD for battery storage, 2017) and Victoria (wind and solar PV, 2018). These targeted, centrally planned CfD auctions have been successful at meeting their policy objectives.

The Commonwealth election in mid-2019 gave rise to two distinct policy choices at the national level. The incumbent conservative Liberal governmen<sup>t</sup> sought to underwrite new "dispatchable" (i.e., coal, gas, pumped-hydro) generation in order to reduce energy prices by way of government-initiated CfDs. The democratic Labor opposition on the other hand sought to revive the Liberal government's *National Energy Guarantee* in order to reduce CO2 emissions and provide the policy stability sought by market participants to guide ongoing investment in *on-market* transactions. However, if bipartisan support for the *National Energy Guarantee* could be achieved, Labor would default to a wide-ranging program of centrally planned CfDs to ensure ongoing VRE investment continuity to meet a 50% renewable target by 2030.

Government-initiated and taxpayer-funded CfDs are an interesting development [3,4]. CfDs can play a legitimate role in dealing with energy market failures relating to missing policy, missing markets and incomplete markets. As a policy mechanism, they represent a means by which to deliver generation plant capacity that, for whatever reason, the market is failing to deliver. Government-initiated CfDs have the e ffect of diversifying buy-side forward market liquidity and in doing so bring about certain short run benefits. Holding all else constant, apart from reducing CO2 emissions, VRE CfDs facilitate state/regional economic development and by adding new renewable supply can reduce spot prices, at least on a transient basis.

Used modestly, the impact of government-initiated CfDs on an energy-only market is likely to be benign at worst. However, a wide-ranging program of CfDs whereby material levels of *on-market* transactions (i.e., between generators and retailers) are crowded-out by *o*ff*-market* CfD transactions (i.e., between governments and generators) is likely to give rise to unintended side e ffects, and this forms the primary motivation of this article. To be clear, this article does not question the welfare implications of introducing VRE plant, and forcing coal plant exit. The addition of VRE plant is taken as an exogenous policy constraint to meet a legitimate and binding CO2 emissions reduction objective. The focus is

whether as a policy instrument, government-initiated CfDs are compatible with the e fficient operation of an energy-only gross pool market.

In this article, a partial equilibrium model of an energy-only market is used to analyze the effects of rising levels of VRE plant, facilitated by government-initiated CfDs rather than via *on-market* transactions typically associated with a conventional renewable portfolio standard. The power system model commences with a thermal plant stock (i.e., coal and gas plant), and then VRE plant enters continuously until a 40% market share is reached.

Key results arising from the modeling are that (1) as VRE plant enters, coal plant exits, and spot market equilibrium is maintained; however, (2) because VRE plant enters via government-initiated CfDs and displaces *on-market* transactions, when coal plant exits, the forward market experiences progressively rising shortages of "primary issuance" hedge contract capacity. This latter finding has important implications for policymakers.

Structural shortages of hedge contract capacity may not matter in an energy+capacity market in which overall price volatility remains within tight limits, or in highly meshed multi-regional power grids where inter-regional trade can "prop-up" faltering market liquidity in an adjacent region. But for an imperfectly interconnected energy-only market with an extremely high VoLL, policy-induced hedge contract shortages present certain problems because falling liquidity is unlikely to be arrested by proprietary traders. On the contrary, evidence from financial markets reveals that proprietary traders and caught with unwanted inventory and irreversible forward positions [5]. Liquidity providers can be expected to exit a market with sharply falling liquidity for fear of being.

A poorly functioning forward market in an energy-only market setting may increase operational risks of incumbent market participants, produce excess contract price premiums, force the most price-elastic (industrial) customers into unwanted spot market exposures, unintentionally *foreclose* non-integrated 2nd tier retailers, and ultimately drive investment activity above e fficient levels to address hedge shortages—all of which must ultimately harm consumer welfare. In short, extensive use of centralized CfDs appears inherently incompatible with the underlying design of an energy-only gross pool market with a high VoLL. Whether the market design requires change to accommodate CfDs, or renewable portfolio standards can be extended to avoid the need for centralized transactional intervention, is an open question.

#### *1.2. On the Importance of the Financial (Hedge Contract) Market and CfDs*

Energy markets are never complete or free of market failures [6]. One of the more prominent failures inherent in energy-only markets is their seeming inability to deliver the requisite mix of derivative instruments required to facilitate e fficient and timely plant entry [6–15].

Long-dated contracts are typically a pre-condition for the timely entry of project financed plant, and while Australia's NEM is noted for favorable forward market liquidity, the majority of activity spans only 1–3 years—well short of contracts that deliver optimal financing and facilitate timely and e fficient ex-ante investment commitment. On reason why liquid forward markets have failed to calibrate beyond 3 years is because competitive retailers cannot a fford to hold hedge portfolios dominated by inflexible long-dated contracts when large components of their customer book switch supplier every 2–3 years [8,16–22].

Government-initiated CfDs have arisen due to a combination of missing or incomplete markets, and form one of a number of policy mechanisms used by governments to meet a decarbonization objective or reliability constraint [23,24]. Government-initiated CfD's have been progressively gaining prominence amongs<sup>t</sup> policymakers [25–29] and amongs<sup>t</sup> academics [3,23,30,31].

Typically, a government-initiated CfD will attempt to minimize the levelized cost of electricity (LCoE) as a surrogate for maximizing value to taxpayers. At one level the use of LCoE as a prime metric is understandable because forecasting market outcomes 10–15 years in advance is notoriously di fficult. But as a stand-alone metric, LCoE is flawed because it treats technology output as homogeneous products as if governed by the law of one price [24,32–35]. That is, while the *physical* properties of electricity are largely homogeneous over space and time, from a *market* perspective there is rich price variation over time, space, and lead time-to-delivery, making the traded commodity a heterogeneous good (i.e., due to an inability to arbitrage, the absence of a single dominant technology, and variations in marginal costs) [6]. The economic value of plant output is not identical and assuming otherwise introduces two biases; base plant is favored over peak, and stochastic plant is favored over dispatchable plant [6].

In real-time, the law of one price applies; stochastic output from wind and solar PV are good substitutes for thermal generation. However, each year there are 105,120 NEM dispatch intervals and associated spot prices (i.e., every 5 min) and when demand is higher than forecast, all else equal, dispatchable generators increase output and receive a higher average price. Conversely, stochastic generators rarely reduce output in periods of oversupply, and hence sell disproportionately at lower prices [6,24,34,35].

Furthermore, as VRE technologies move from niche to material market shares, deployment success becomes a significant driver of market value which is amplified when thermal plant fails to exit [24,32–38]. High levels of VRE shielded by CfDs and priority-dispatched will initially place downward pressure on price [14,24,39,40]. Given negligible marginal running costs, these so-called *merit-order e*ff*ects* arising from policy-induced VRE plant entry became apparent in markets such as Germany as early as 2008 [41] and had been prominent in the SA region of the NEM [42–45]. Consequently, market values of incumbent VRE (and future) plant will be adversely a ffected from a stream of continual entry through a combination of production "correlation e ffects", "merit-order effects" and "price-impression e ffects" [24,35,37]. However, merit-order e ffects eventually unwind when thermal plant is forced to exit [39,46]. This set of market dynamics has implications for a wide-ranging program of government-initiated CfDs.

#### *1.3. Government-Initiated CfDs: Motivation and Application*

The policy objective of government-initiated CfDs is to introduce generation plant that energy markets are failing to deliver. In this sense, CfDs have the e ffect of bringing forward future power projects to today, with the benefits, costs, and risks of doing so allocated to electricity consumers, taxpayers, and incumbent rivals.

There are many reasons why governmen<sup>t</sup> intervention is legitimately required in energy markets. As is well understood in economics, organized spot markets and their associated forward contract markets fail to internalize known externalities. For example, energy-only spot markets may undervalue reserve capacity until it is actually required. In the absence of an explicit price on carbon, energy markets also undervalue CO2 emissions and will therefore only be produced at the e fficient level by chance. And as with many markets, research and development (R&D) is not valued; but this is compounded in energy markets because participants are unable to capture the benefits of a first-of-a-kind plant investment—in fact, the contrary is usually the case in that the market avoids costly mistakes of the first iteration of a new technology. Absent some form of governmen<sup>t</sup> intervention, R&D will be under-supplied by the market.

Government-initiated CfDs can have the e ffect of "priming" a market by helping emerging technologies overcome certain entry barriers. The Queensland government's Solar150 program in 2016 awarded CfDs to four solar PV projects totalling 150 MW at a time when solar PV struggled to compete with wind. The policy had the e ffect of kick-starting a wave of solar developments; by late-2018 a total of 1945 MW of solar PV had been committed *on-market* including some projects on a purely merchant basis (i.e., without a PPA). In South Australia, a policy to introduce a 100 MW utility-scale battery for system stability similarly primed the market for storage—there are now 215 MW of commissioned batteries, a further 155 MW have reached financial close, and 1897 MW are under active development across Australia. In short, while there are many policy mechanisms available to remedy energy market failures, CfD's are indeed a viable policy option.

#### *1.4. CfDs: How They Work and Why They Work*

In the classic case, a government-initiated CfD auction will specify a particular technology (e.g., solar PV) or technology set, output or rated capacity (e.g., up to 100 MW) and timing for delivery (e.g., able to reach financial close within 6 months of being awarded a long-dated CfD). The CfD is in turn a form of long-dated fixed price contract, usually expressed in \$/MWh. In application, a CfD is a derivative instrument because payouts are referenced against spot prices. In a two-way CfD with a strike price of say \$65/MWh, the contracting governmen<sup>t</sup> (i.e., taxpayers) will pay the difference to the renewable project proponent whenever spot prices fall below \$65, and the renewable project proponent pays the governmen<sup>t</sup> whenever spot prices are above \$65. CfDs are typically run-of-plant instruments such that difference payments only apply when the renewable plant is producing—and the plant's variable output is bid into the market and dispatched whenever it is profitable to do so (i.e., spot price > \$0/MWh if the CfD has a settlement floor price of \$0/MWh). Absent material plant failure, in which case some form of liquidated damages may apply, the weather-related volume risks and forward price risks associated with a VRE plant are effectively transferred to the contacting governmen<sup>t</sup> (taxpayers). After writing a CfD, plant output cannot be sold twice, and so the plant's capacity is extracted from the forward market and taxpayers in turn hold a speculative instrument (unless the governmen<sup>t</sup> chooses to on-sell the CfD in secondary markets).

A government-initiated CfD overcomes missing and incomplete markets and crucially in the context of Australia, can successfully navigate carbon policy uncertainty because CfDs provide revenue certainty (i.e., virtual market immunity) to the power project proponent.

Finally, because power projects are capital-intensive, the cost of debt and equity capital is an important driver of overall plant unit costs (\$/MWh). The direct involvement of a governmen<sup>t</sup> through long-dated CfDs enhances the credit quality of power projects, and this generally enables higher levels of debt, a lower cost of debt capital, and makes the task of equity capital raising easier. Consequently, holding all other variables constant, by transferring the price, volume, policy, and credit default risks of power projects to taxpayers, government-initiated CfDs are capable of producing a lower LCoE for entering projects.

#### *1.5. The Impact of CfDs vs. Carbon Pricing and Renewable Certificate Markets*

CO2 emissions reduction policies ultimately seek to alter the plant stock in a way that reduces output from coal plant and increase output from renewable and cleaner (e.g., gas-fired) resources. Regardless of the policy mechanism used (e.g., cap and trade emissions trading scheme, emissions intensity scheme, carbon tax, renewable portfolio standard, clean energy target or government-initiated CfD), wealth transfers amongs<sup>t</sup> producers occur. Carbon-intensive forms of generation are adversely affected, while low and zero emissions plant benefit from any explicit or implicit price on CO2 emissions.

Government-initiated CfDs differ from broad-based market schemes (e.g., carbon prices or renewable portfolio standards) because of the direct involvement of governmen<sup>t</sup> in the transaction and the reallocation of market, credit, and policy risks to taxpayers. Project bankers and the credit committees of Banks, which allocate scarce debt capital, have a strong preference for long-dated government-initiated CfDs because from a credit perspective there is negligible risk of counterparty default. By contrast to conventional NEM-based over-the-counter *on-market* transactions, a government-initiated CfD re-orientates policy and credit risk away from buy-side energy market participants, and vests this with taxpayers.

When deployed judiciously, the implications of CfDs are generally benign. In the case of the Queensland Solar150 program for example, any distortionary impacts arising from 150 MW (0.4 GWh) of solar-based CfDs in a 10,500 MW (54,000 GWh) regional market would be hard to detect. Taxpayers have a collective financial exposure to CfDs that will ultimately prove to be out-of-the-money; but this needs to be balanced with other policy objectives (e.g., state development, subsequent economic and environmental benefits of the 1945 MW of *on-market* solar PV projects that immediately followed). But what happens when CfDs are not used to "prime" a market, but rather, are used to "replace" the

market; that is, replace broad-based market mechanisms and *on-market* transactions like those used in renewable portfolio standards to drive VRE entry?

Holding all else constant, so-called *merit-order e*ff*ects* can be expected. That is, adding more supply, renewable or non-renewable, will reduce wholesale prices. But it will do this in the short- to medium run. Because the purpose, and e ffect, of the entry of VRE plant at-scale is designed to replace coal plant output, it will inevitably do so. Ultimately, the marginal coal plant will find it unprofitable, and will therefore exit. At this point, prices can be expected to rebound—and in the context of the NEM this is more than a theoretical observation. There is nothing inherently wrong with this policy objective, or the course of events that follows per se. But government-initiated CfDs undertaken at-scale may adversely impact the e fficiency of an energy market (as distinct from "priming" a market) by comparison to an *on-market* renewable porfolio standard for three reasons.

First, governments are remote from power system operations and power system contract and risk managemen<sup>t</sup> requirements. Government-initiated CfD auctions are therefore typically based on simplified metrics such as minimizing LCoE by way of open-auction, or at discriminatory price benchmark set by governmen<sup>t</sup> bureaucrats to accommodate technological variations in production or cost. But as noted in Section 1.2, LCoE is a flawed metric and an overreliance on it in CfD auctions risks introducing an ine fficient pattern of plant entry in a way that on-market transactions may have avoided. In contrast, broad-based market schemes like the *National Energy Guarantee* or a well-designed renewable portfolio standard require market participants to focus not on the LCoE, but on the timing, location, and market value of new plant output [4,24,34,35]. And to the extent that market participants introduce an ine fficient pattern of plant entry vis-à-vis timing, location and market value, the risk and consequence of such errors vest with shareholders, not taxpayers.

Second, government-initiated CfDs introduce quasi-market participants that, through the design of the CfD, are almost completely sheltered from the NEM's energy-only short and medium-run locational, spot and forward price signals—the primary signals relied upon by market institutions and policymakers to regulate system performance, system reliability, investment patterns, and long run consumer prices. In contrast, on-market transactions undertaken by profit-maximizing firms under a renewable portfolio standard requires that market participants assess the relative pattern of entry, locational considerations, and absorb the risks of inadequate or excess entry relative to policy objectives. Broad-based market schemes can therefore be expected to outperform a central buyer on a risk-adjusted basis; and I must emphasize on a *risk-adjusted* basis—as one reviewer noted government-initiated CfDs should produce better financing terms and a lower overall cost of capital by comparison to on-market transactions. But this observation ignores risks transferred to taxpayers and the opportunity cost of using scarce governmen<sup>t</sup> balance sheet resources. In comparison to a central buyer, market schemes are likely to accumulate a more optimal composition of assets and allocation of investment risks, reflecting the combination of physical power system requirements, policy-related constraints, and the risk appetite of participants to transactions.

Third, and by far the most adverse implication of a wide-ranging government-initiated CfD program is the potential distortion to forward markets and therefore market e fficiency more generally. Unlike renewable portfolio standards, a wide-ranging policy of government-initiated CfD instruments that form a progressively larger share of a forward market will ultimately damage the *primary-issuance* of hedge contracts.

Following an initial or "primary loss" of on-market contract liquidity, the exit of proprietary traders that invariably follows (i.e., through fear of being caught with an illiquid position and unwanted inventory) will drive a "secondary loss" of market depth and liquidity. Combined, this is capable of culminating in a structural shortage of hedge contracts (i.e., forward market liquidity dropping below 100%).

There is, of course, nothing preventing market participants from writing their own CfDs with VRE plant and other plant capacity in order to mitigate looming (or actual) hedge shortages. Indeed, in the long run this may be the only way by which hedge shortages can be remedied in an energy-only

market setting with government-initiated CfDs, noting that the end result is excess capacity. But a wide-ranging program of government-initiated CfDs is likely to *crowd-out* on-market transactions —neither of these alternatives would appear to produce efficient results.

In short, a well-intentioned wide-ranging program of government-initiated CfDs can be expected to create hedge contract shortages and at best raise forward prices above the efficient level, and at worst unintentionally foreclose 2nd tier retailers while replacing well-functioning forward markets with quasi-market participants who are indifferent to the outcomes facing final electricity consumers—all of which must ultimately harm welfare. Any response by market participants that might otherwise logically follow risks being *crowded-out* by ongoing governmen<sup>t</sup> CfDs, or may induce excess capacity. How these shortages emerge can be demonstrated quantitatively, and this forms the focus of Sections 2 and 3.

## **2. Materials and Methods**

In order to analyze the impact of government-initiated CfDs on forward markets, a power system simulation model (NEMESYS) has been used [24]. NEMESYS is a dynamic, security-constrained, unit-commitment model with 30-min resolution and price formation based on a uniform, first-price auction clearing mechanism consistent with the NEM design. As with [47], this partial equilibrium model assumes perfect competition, transmission and ramp-rates, free entry and exit to install any combination of (indivisible) capacity that satisfies differentiable equilibrium conditions. The modeled power system commences with a conventional thermal plant stock, with scenarios derived by exogenously determining progressively higher levels of VRE output (i.e., by way of policy). And as with [35], the focus of simulations is half-hour resolution over a single year. Model logic, drawn from [24], is as follows.
