Incentives for early adoption of carbon capture technology: further considerations from a European perspective

This note details two comments on a recent policy proposal in Comello and Reichelstein (2014) aimed at favouring the early adoption of Carbon Capture (CC) technology in the next generation of thermal-based power plants to be installed in the United States.


Introduction
The prohibitively high cost of CC technology for first-of-a-kind plants is recurrently cited as a major barrier to its large-scale deployment.To overcome this problem, Comello and Reichelstein (2014) recently articulate an innovative policy proposal to enable substantial cost reductions by leveraging the sizeable deployment of thermal-based power generation projected in the U.S. during the period 2017-2027.The proposal combines two ingredients: a binding and inflexible emission standard; and the "Accelerated Carbon Capture Deployment" (ACCD) -a preannounced schedule of Investment Tax Credits (ITC) and Production Tax Credits (PTC) -aimed at providing an incentive for newly built power plants in the U.S. to adopt CC immediately.This brief note extends the analysis by considering two issues.In a first section, we apply the framework detailed in the original article1 to generate a schedule of tax-credits that is robust to alternative scenarios for CC deployments outside the U.S. In a second section, we reflect on the possible emergence of a coordination game capable of hampering the desired early deployment of that technology and propose a modified schedule of tax-credits that is sufficient to overcome that problem.

-The role of early CC deployments outside the U.S.
Using a list of proposed but still undecided projects (GCCSI, 2013), the authors assume the installation of nearly 3 GW of foreign CC capabilities between 2014 and 2020.However, in Europe, the funding of large CC projects has recently proven to be difficult causing delays and several project cancelations (Lupion and Herzog, 2013).As early foreign projects are posited to engender international spillovers, one may wonder whether these withdrawals could undermine the proposal's success.
To render the proposal robust to the vicissitudes impacting foreign projects, we consider a 'worstcase' scenario whereby foreign deployments are restricted to the unique Canadian 130MW power plant finalized in 2014.To compensate for the absence of foreign early investments, augmented ITC and PTC schedules are needed (cf., Figure 1) but our evaluations confirm that this robust version is almost as attractive as the initial version:2 • The Levelized Cost of Electricity (LCOE) obtained with a facility that becomes operational by the end of 2027 is approximately 7.9 ¢/kWh if CC technology is consistently adopted by all the newly built U.S. thermal power plants. 3 The magnitude of the tax-credit levels remain politically acceptable (cf., Figure 1).
• Overall, the cumulated (undiscounted) foregone tax revenue to the U.S. Treasury reaches about $8.2 billion. 4This robust schedule of incentives thus represents a cost-effective solution for achieving a large scale deployment of this innovative technology.

-Strategic interactions among CC adopters
Recent European literature on CC and storage has highlighted the interactions that exist among CC adopters connected to a common infrastructure system (Mendelevitch, 2014;Massol, et al., 2015).In the present paper, infrastructure issues are neglected but the use of an experience curve de facto generates some interactions.It is instructive to examine these interactions further.
Our discussion is structured as follows.A first subsection reviews the evaluation of the schedule of tax-credits used in the original article.The second subsection focuses on CC adoption in a given year by several players and proves that the incentives offered that year may not be sufficient to rule out the possible occurrence of a coordination game with multiple equilibria.As uniqueness is not achieved, these players may select an equilibrium where some emitters rationally prefer to delay CC adoption.
As that phenomenon may jeopardize the desired policy outcomes, the third subsection identifies an appropriately augmented minimum level of incentives that is sufficient to overcome that issue.The last subsection reports the numerical results obtained using the associated ITC and PTC schedules.For the sake of brevity, all the formal proofs are provided in a Supporting Document.

A -The subsidy scheme in Comello and Reichelstein (2014)
We consider a given year t in { } 2017,..., 2027 and let: t K denote the total planned capacity of all the power plants to be installed during that year; and t CK denote the state variable that gives the cumulated CC capacity already installed during the preceding years. 5or an investor that considers installing a power plant during that year, we let: In year t , we do not model the tax-credits but simply assume that their effect is to lower the LCOE measured on a power plant that early adopts CC capabilities.We let t S be the levelized subsidy and ( ) ( ) Recall that the ACCD tax-credits are set so that, for a facility to be installed in a given year, it becomes advantageous to adopt CC capabilities immediately compared to retrofitting that plant by the end of 2027.The evaluation of the schedule of tax-credits presented in the original paper is detailed in an associated spreadsheet model: the "NGCC + CC Calculator" (Comello and Reichelstein, 2014).In this model, CC adoption at the maximum level is assumed in each year t so that the cumulated CC capacity available at the beginning of the next year is ( ) The authors underline that, by construction, the tax-credits prevent possible 'deviation' from what they name an 'equilibrium path' of early CC adoption.Indeed, if one models CC adoption as a one-shot game involving 11 players (one for each year) where each player faces a binary choice with respect to early CC adoption for the entire capacity t K , the proposed subsidy makes 'early adoption' the best response of every player when the other players also jointly decide 'early adoption'.In this game, the proposed schedule of tax-credits thus makes the decision vector stating 'early adoption' for every player a pure strategy Nash Equilibrium (NE).

B -Is that proposed subsidy sufficient?
In reality, the problem at hand involves a sequence of annual decisions and the cumulated impact of installed CC capacity on future LCOE figures generates a 'path dependency' to past CC adoption decisions.The capacity forecasts and the standard plant size used by the authors together suggest that is the cumulated CC capacity in operation at that time.
several power stations will be installed in some years (particularly during the period 2023 -2027).As these plants are likely to be owned by independent companies, one may wonder whether, in each year t , the threshold level t S is sufficient to induce the joint early adoption of CC capability by all the plants to be installed that year.
To address this issue, one has to examine the strategic interactions among the plants to be installed in a given year.To clarify the presentation, we consider a simplified case and assume that two identical plants of size 2 t K are installed in year t .These plants are controlled by two independent players labeled '1' and '2' and we posit that each player faces a binary choice with respect to the immediate installation of CC capabilities.We also assume that early CC adoption has already been decided by all predecessors so that the cumulated CC capacity available at the beginning of the year is The objective of each player is to minimize its LCOE.Absent any tax-credits, their strategic interactions can be summarized with the 'unsubsidized' normal form game in Table 1.

Table 1. The 'unsubsidized' normal form game [ INSERT TABLE 1 HERE ]
Following the authors, we posit that it pays to delay CC adoption when the rival adopts the CC technology early: ( ) As N t c is assumed to be a strictly decreasing function, it also pays to delay CC adoption when the rival does not early adopt the capture technology. 8Hence, the condition (2) indicates that 'late adoption' is a strictly dominant strategy for a player and therefore the pure strategy where both players decide 'late adoption' is the unique NE.Hence, absent any tax-credits, the two players rationally prefer to delay CC adoption.
Emergence of a coordination game 8 Because ( ) We now examine the case where tax-credits are implemented and consider the levelized subsidy level, t S .For a player that early adopts CC capabilities, the payoff is now given by the subsidized LCOE.
The 'subsidized' normal form game is thus derived from the 'unsubsidized' one by using the subsidized LCOE function N t c ɶ in lieu of the unsubsidized one.Two results can be highlighted for the 'subsidized' normal form game.
Finding 1 -For any levelized subsidy that verifies t t S S ≥ , the strategy vector stating 'early adoption' for each player is a NE of the subsidized game.
Finding 2 -Implementing a levelized subsidy t S that verifies t t

S S ≥
may not be large enough for the strategy ('early adoption', 'early adoption') to be the unique NE.In particular, a subsidy level is such that ('late adoption', 'late adoption') remains a NE.
Together, these two findings suggest that implementing a levelized subsidy that solely verifies t t

S S ≥
could lead to a coordination game with possibly several NEs.
However, the existence of a NE other than the one that provides early CC adoption at level t K in year t is a source of concern.The following finding proves that a lower-than-expected level of CC adoption in year t may undermine the ability of the proposed schedule of levelized subsidies to also achieve the early adoption of CC capabilities in subsequent years.
Finding 3 -Possible existence of a snowball effect: If delayed adoption were to be decided by some players in year t , a levelized subsidy 1 t S + that solely verifies the condition 1 1 t t S S + + ≥ can be too small to make 'early adoption' of CC capabilities the best response of an investor in year 1 t + .

C -A remedy
Because of this possible snowball effect, one may desire that the schedule of levelized subsidies rules out any possibility for the investors in any given year t to pick up a NE that does not lead to generalized early CC adoption.As there can be between-year variations in the number of players and in the sizes of their plants, we believe that it is preferable to opt for a threshold level that is large enough to be independent from these considerations.The following proposition offers a sufficient condition for generalized CC adoption to occur.

D -Application
This subsection reports the results obtained using the minimum threshold above.The numerical evaluations have been conducted assuming the robust capacity deployment scenario discussed in Section 1.9 Compared to the values in the original article, our evaluations indicate that the magnitude of the ITC levels remains similar; however, substantially augmented PTC are needed (cf., Figure 2).

Figure 2. The tax credit schedule needed to obtain a unique NE [ INSERT FIGURE 2 ABOUT HERE ]
These new levels generate a substantial increase in the cumulative (undiscounted) foregone tax revenue to the U.S. Treasury: about $14.1 billion.This is a 110% increase over the figure obtained in the original article.Nevertheless, we believe that this cost figure remains tolerable for such an ambitious policy, especially one that would now be rendered robust to both foreign adverse events and domestic gaming issues.

Conclusions
This note discusses the feasibility of the policy proposal in Comello and Reichelstein (2014).Two lines of arguments have been considered.First, we have examined the effects of early CC deployments outside the U.S. Second, we have determined that the initially proposed ACCD schedule can be insufficient to engender the desired generalized early adoption of CC capabilities because of the possible co-existence of multiple Nash equilibria.In both cases, a modified version of the policy has been proposed using the detailed cost structure developed in Comello and Reichelstein (2014).Though higher incentive levels have been obtained, our findings confirm that the cost of the proposed ACCD policy to the U.S. Treasury is not out of reach.This modified policy thus represents an interesting instrument to break the 'vicious circle' that currently hampers the deployment of CC technologies.

A.2 Methodology
We proceed as in Comello and Reichelstein (2014) and identify the minimum incentives required to bridge the gap between: (i) the Levelized Cost of Electricity (LCOE) of a facility that is retrofitted in 2027, and (ii) the LCOE of a plant that immediately adopts CC capabilities.For each year, the minimum investment tax credit (respectively production tax credits) is calibrated so that the levelized capacity (respectively variable) cost in case of immediate CC adoption 1 is equal to those incurred in case of a retrofit in 2027. 2

A.3 Results
In Table A.2, we detail the "Accelerated Carbon Capture Deployment" (ACCD) tax credits obtained in the robust scenario.These results have been generated using an adapted version of the original NGCC+CC Calculator.3 1 For the sake of comparability, this levelized cost is determined using the fiscal depreciation schedules and the 10% first-ofa-kind premium defined in Comello and Reichelstein (2014).
2 Note that these minimum tax credits schedule are not necessarily tiered.The use of different fiscal depreciation schedules between two successive years can generate a situation whereby the investment tax credit needed in a given year is larger than that needed during the previous year.To overcome that issue, we proceed as in the NGCC+CC Calculator and generate tiered schedules.In this paper, a linear programming approach is implemented to obtain these tiered schedules.The linear program is aimed at determining the tiered tax credit schedule that minimizes the cumulative foregone tax revenue to the U.S. Treasury and verifies two types of constraints: (i) the tiered tax credit in a given year must not be lower than that of the subsequent year, and (ii) the tiered tax credits must not be lower than the minimum values required for the LCOE of a new plant to be lower than the LCOE of a plant retrofitted in 2027.Note: MACRS indicates that all capital expenditures will be eligible for the five-year accelerated depreciation schedule according to MACRS; 150% DB indicates that all capital expenditures will be subject to the current 150% balance, 20-year depreciation schedule.

Appendix B -Mathematical proofs
This Appendix presents the formal proofs of the results stated in Section 2. Unless otherwise specified, the notation is based on the one introduced in the paper.
Finding 1 -For any levelized subsidy that verifies t t

S S
≥ , the strategy vector stating 'early adoption' for each player is a Nash equilibrium of the subsidized game.
Proof: Suppose both players decide 'early adoption'.Each player incurs the subsidized LCOE: is de facto verified which indicates that 'early adoption' is the best response of a player whenever its rival also picks that strategy. Q.E.D.
Finding 2 -Implementing a levelized subsidy t S that verifies t t

S S ≥
may not be large enough for the strategy ('early adoption', 'early adoption') to be the unique Nash equilibrium.In particular, a subsidy level t t

S S
= is such that ('late adoption', 'late adoption') remains a Nash equilibrium.
Proof: We have to check the conditions for the vector ('late adoption', 'late adoption') to no longer be a Nash equilibrium.We examine the best response of a player when the rival decides 'late adoption'.The condition for that player to systematically prefer 'early adoption' when the rival decides 'late adoption' is

Appendix C -A robust schedule of ITC and PTC that is immune to strategic gaming considerations
This Appendix details the assumptions and results commented in Section 2.

C.1 Assumptions and methodology
The simulation is based on the projected capacity deployments retained in Appendix A (Table A.1).
We follow the methodology in Appendix A.2 except that the tax credits are now calibrated so as to provide a levelized subsidy that is at least as large as the threshold level mentioned in the Proposition in Section 2. These tax credits thus prevent the existence of Nash equilibriums where some emitters could rationally prefer to delay CC adoption.

C.2 Results
In Table C.1, we detail the "Accelerated Carbon Capture Deployment" (ACCD) tax credits obtained in the robust scenario when the incentives are derived from the levelized subsidy discussed in the Proposition in Section 2. These results have been generated using an adapted version of the original NGCC+CC Calculator.4  Note: MACRS indicates that all capital expenditures will be eligible for the five-year accelerated depreciation schedule according to MACRS; 150% DB indicates that all capital expenditures will be subject to the current 150% balance, 20-year depreciation schedule.
The "Cahiers de l'Économie" Series The "Cahiers de l'économie" Series of occasional papers was launched in 1990 with the aim to enable scholars, researchers and practitioners to share important ideas with a broad audience of stakeholders including, academics, government departments, regulators, policy organisations and energy companies.
All these papers are available upon request at IFP School.All the papers issued after 2004 can be downloaded at: www.ifpen.fr The list of issued occasional papers includes:

Figure 1 .
Figure 1.The modified ACCD tax credits schedule under a robust scenario (a) Modified ITC schedule (b) Modified PTC schedule

Figure 2 .
Figure 2. The tax credit schedule needed to obtain a unique NE (a) Modified ITC schedule (b) Modified PTC schedule In each year t , any tax-credits yielding a levelized subsidy t S that verifies

Table A .2. The ACCD tax credits in the Robust Scenario
In each year t , any tax credits yielding a levelized subsidy t S that verifies