California’s Cap-and-Trade Market Enters its Teen-Age Years
Extending cap-and-trade could be more impactful than proposed changes to the allowance budget.
California’s cap and trade (C&T) program for greenhouse gasses, now 12 years old, is going through puberty and policymakers are trying to decide what it should be when it grows up. California established the Cap-and-Trade Program under AB 32 in the late 2000’s, and the state achieved its 2020 climate target of returning to 1990 levels years ahead of schedule. The currently legislated 2030 target is significantly more stringent at 40% below 1990 levels, and the ARB’s 2022 Scoping Plan says the state should achieve a 48% below 1990 levels by 2030 to be on track for carbon neutrality. The California Air Resources Board (ARB) is undertaking a rulemaking that will influence several aspects of the program’s future.
While much of the focus has been on how to adjust the “budget” of allowances (i.e., the cap) over the next 8 years, an even more important question is whether the program will be extended beyond 2030. As part of a modeling team of UC Davis researchers, I have been working with ARB to help flesh out what the impacts of proposed changes to the program could be. We presented our preliminary findings last week.
The C&T program has been through ups and downs, in terms of both public perception and actual carbon price. One criticism has been that it is not ambitious enough, as reflected in the fact that the price of carbon allowances from the program have periodically flirted with its price floor. The actual greenhouse gas (GHG) emissions subject to the program’s cap have run well below the program’s carbon budget over its first decade. Parties have continued to debate whether this constitutes success (“emissions are low!!”) or a failure (“the cap is too high!!”) or something in between.
My own take has largely been “thank goodness the program had a price floor, otherwise the carbon price may have been much lower.” This is one of the corollaries of the mantra that the cap is not really a cap (TCINRAC). The carbon “cap” – or more accurately the budget of allowances allocated or sold into the market – is not a fixed amount. It expands or contracts in response to price bounds set by the legislature and ARB. When the carbon price sank to its floor in the middle of the last decade, tens of millions of allowances went unsold and close to 40 million metric tons were removed from the main allowance budget, effectively tightening the cap. While this “flexible cap” framework has been a critical aspect of the program’s success to date, the current rulemaking is asking whether further adjustments need to be made.
In previous work with Severin Borenstein, Frank Wolak, and Matt Zaragoza-Watkins (BBWZ), we developed a modeling framework that deployed time-series forecasting tools to project the range of expected business as usual (BAU) emissions. Severin, Frank, and I repeated this exercise in 2017 (using data through 2015) as well. The outlook from the vantage point of 2023 is much rosier than it was in 2017. Emissions declined steadily from 2015-2019 and have not fully rebounded from the dip induced by the pandemic in 2020. Seeing as most of these declines came from the electricity sector, the new BAU forecast, which relies upon historic trends to project forward, expects continued declines in electricity emissions. But BAU industrial and transportation emissions are forecast to be essentially flat.
As with those previous studies, we apply estimates of abatement – coming from a variety of programs such as the Zero-emissions Vehicle (ZEV) policies, the Low-carbon Fuels Standard (LCFS), and residential electrification – that could be expected to further reduce emissions relative to BAU. We think of these activities as the “supply” of abatement, and the higher the carbon price, the more abatement would be expected. Also as carbon prices rise, more allowances also get supplied to the market, because…TCINRAC. We can illustrate the multi-year impact of all of this as a supply curve.
What matters for carbon prices is not the relationship between annual emissions and the annual allowance budget (i.e., the cap). Because allowances can be banked, prices are largely determined by the net surplus or deficit of allowances over a much longer time horizon, such as the lifetime of the program through 2030. We can calculate the expected “net demand” for allowances as the difference between our BAU forecast and the number of allowances available. We can plot the probability distribution of this net demand through 2030, along with our abatement supply assumptions, to get a sense of how likely it will be that the demand for allowances might exceed the supply at given price levels, as shown in the graph below.
Two things seem clear from our preliminary work. First, there is a likely surplus of allowances in the program through 2030, meaning carbon prices would probably settle at the 2030 floor price (prices are currently around the 2030 floor) if no changes are made. Note that about a third of the forecast probability distribution has even BAU emissions below the cumulative budget through 2030. Second, however, the surplus of allowances gets eaten up steadily starting late in this decade, so that the demand for allowances would exceed supply by the middle of the 2030’s. We can plot the forecast range of annual allowance surplus and deficits as well as track the running cumulative surplus (or deficit). If we exclude the allowances held back for price-containment, the median of the forecast distribution has the market going into deficit somewhere around 2033.
One of the key insights of BBWZ was that cap-and-trade markets are characterized by an inelastic supply of allowances and a relatively inelastic demand for allowances (driven by abatement costs) as well. In such an environment, prices tend towards extremes. Even a small shift in the demand for allowances (say from a pandemic, or a recovery) can drive prices down or up sharply in response. Washington State has provided the latest example of this dynamic. This is one reason why we have been strong advocates of TCINRAC, otherwise known as “price-collars,” in cap-and-trade markets.
Therefore, it wasn’t surprising to see that – simply by assuming the program continues for a few years past 2030 – our model predicts a much higher probability that carbon prices would reach one of the higher “tiers” in the price-containment system. If the program is assumed to sunset in 2030, surplus allowances would be worthless, and prices would probably be stuck at the price floor. If the program is extended into 2040 or later, the expected need for allowances in those later years increases their value and raises expected prices substantially. Beyond that, if the allowance budget is reduced (or the cap is “tightened”) it becomes quite likely that prices will reach the price ceiling, which would be around $110/ton in 2030 and $187 per ton in 2040.
There has been some debate over how aggressive California should be in tightening the cap, and ARB has considered scenarios where the cap would be reduced to 40%, 48% or even 55% below 1990 levels by 2030. However, another underappreciated corollary of TCINRAC is that once carbon prices reach the price ceiling there isn’t much effect of tightening the cap even more. At that point, abatement activities and total emissions are set by the price, not by the nominal cap, because emissions move around in response to carbon prices. So, while setting the overall cap level matters for capped emissions, it matters less than most people think. The single biggest impact California can have to encourage abatement through its cap and trade is committing to the program beyond 2030.
Bushnell, James. “California’s Cap-and-Trade Market Enters its Teen-Age Years” Energy Institute Blog, UC Berkeley, November 27, 2023, https://energyathaas.wordpress.com/2023/11/27/californias-cap-and-trade-market-enters-its-teen-age-years/
Categories
Energy Institute Blog 
C&T shares a lot of features with cryptocurrencies and the ups and downs of its “pricing” are no surprise nor is its ultimate failure. I certainly hope that it is not renewed in 2030. That gives us plenty of time to work on a carbon tax system that will ultimatley prove more robust.
I’m not sure how the anology is being made between cryptocurrencies and C&T. The former is a system of money used as a medium of exchange whereas C&T is a market platform that conducts business in U.S. dollars. That aside, while a carbon tax still has merit on the national level, that won’t likely happen in individual states. The vote threshold and political hurdles are just too high. With that in mind, if a carbon tax is so much better than C&T, why didn’t WA adopt a tax instead of C&T in 2021?
It is working. PG&E used to be over 50% carbon-based fossil fuels in 2004, now they are at just 8.5% and only natural gas. My reason for going to off-grid solar was to be as fossil fuel free as possible. After NEM3.0 and my home burning up in a fire, destroying my solar panel system and the low carbon usage of PG&E, I am debating re-building my solar panel system since the benefit would only be economic since PG&E has become so clean here in California.
How can we tell whether observed effects are due to changes in behavior vs migration of industry out of state or substitution of imports for previously made in California product?
The real eye-opener of the modeling study for CARB was its bottom-line conclusion that “Most alternative scenarios yield prices that follow the price ceiling through at least 2035.” The implication is that under most alternative scenarios, the emissions cap will be breached. Cap-and-Trade’s fundamental advantage of “environmental certainty” was expressly forfeited in favor of price certainty when the legislature mandated a price ceiling in AB 398 (Garcia 2017).
It would be very helpful if the model could provide succinct statistical metrics (e.g., mean expectation values and 1-sigma confidence intervals), not only for allowance prices but also for emissions. What is the estimated probability of achieving the 40%, 48%, or 55% reduction target in 2030, or the net-zero target in 2045? Also, in addition to annual emission rates, cumulative emissions (time-integrated rates) in 2045 should be considered. Climate impacts and long-term decarbonization costs are more directly affected by cumulative emissions than by emission rates, and policies that reduce emission rates sooner rather than later will achieve lower cumulative emissions.
The key takeaway from the preliminary modeling results is that emissions and costs are driven more by the price ceiling than by emission targets. The current price ceiling was initially set at $61/MTCO2e in 2021 (real 2018 dollars) based on a conservative estimate of the social cost of carbon (SC-CO2), but the IWG has recently increased its SC-CO2 estimate to $190/MTCO2 (in 2020, at a 2% discount rate). Also, CARB believed that a higher price ceiling “would be excessive relative to prices needed to achieve the 2030 target” because the 2017 Scoping Plan found that “there is 96 percent likelihood that the adopted Scoping Plan scenario with the existing Cap-and-Trade Program will achieve the 2030 emissions target” (from CARB’s 2018 ISOR for the C&T regulation). The UC Davis modeling work probably refutes the 96% estimate. Furthermore, CARB did not consider any need to exceed the statutory 2030 target (40% reduction) in setting the price ceiling. CARB will likely be considering an increase in the ceiling price, and modeling scenarios including alternative ceiling prices could help inform the policy choice.
A primary consideration in setting the price ceiling is the “need to avoid adverse impacts on resident households, businesses, and the state’s economy” (AB 398). Allowance prices are not necessarily the best metric for gauging such impacts. The 2022 Scoping Plan’s economic modeling results projected economic cost indicators such as household income, which are more relevant. The projected change in average annual household income for the Scoping Plan Scenario (48% reduction) relative to the Reference Scenario (40% reduction) was only -$40 in 2035 and +$2 in 2045 (Figure H-22 in Appendix H), suggesting that a higher price ceiling sufficient to raise the emission reduction expectation value from 40% to at least 48% in 2030 would not significantly impact households. Would the UC Davis model be capable of calculating economic indicators such as household income?
Another reason for disfavoring allowance prices as a cost metric is that economic impacts are influenced more by how carbon pricing revenue is spent than by the carbon price itself. Such impacts can be mitigated with a more targeted expenditure of allowance auction proceeds. For example, according to the “Summary of Price Response Assumptions” in the UC Davis workshop presentation (slide 24), a carbon price of $200/ton would translate to an electricity price increase of 3.7¢/kWh. However, electricity ratepayers currently do not pay the full cost of carbon pricing. Some of the Cap-and-Trade auction revenue is returned directly to utility ratepayers in the form of a climate credit that compensates for the compliance cost of the Cap-and-Trade Program on their electricity bills. Other alternatives include output-based allocation and direct subsidization of zero-carbon technologies. The allowance revenue allocation can have much greater influence on economic costs than the other cost-containment measures such as the APCR. Could the UC Davis model account for allowance revenue allocation in calculating economic costs?
If the model is capable of projecting long-term costs in 2045, including the social cost of remaining accumulated emissions, it would probably predict that early action to reduce GHG’s is, in the long term, less expensive and less risky than deferred action. A higher emissions reduction target would increase the odds that allowances prices will be at the ceiling limit, which will maximize the incentive for early action. The price ceiling guards against potential adverse economic impacts of an overly ambitious emission target, so there would be no harm in adopting the Governor’s 55% target as a “stretch goal”. A more optimum reduction target might be 100%, which would effectively fix the allowance price at the ceiling. The allowance auction would become a fixed-price sale without any cap, and all of the price-minimization mechanisms of Cap-and-Trade – trading, banking, offsets, the APCR – would become superfluous. That appears to be where the evolutionary path of California’s Cap-and-Trade system is heading.
Climate policy is typically framed as a constrained optimization problem: Minimize costs subject to an emissions constraint. Given the reality that we are unable or unwilling to pay the cost for any guaranteed environmental outcome, it would make more sense to minimize emissions subject to the limitation of how much we are able and willing to spend. Perhaps the UC Davis model could, with some further refinements and upgrades, do the minimization calculation and take the guesswork out of legislative and regulatory climate policy.
for policy makers outside of CnT the options should focus on what the “relatively inelastic” demand markets are. Eg large emitters with long term commitments act strategically (utilities) emitters small or large with short term commitments or uses have to cover emission cost whatever (total inelastic) think gas tank fillers or say new emission reduction technology startups or demonstrations.