Comments on the proposed emission trading rule
William M. Shobe, Professor of Public Policy
University of Virginia, April 8, 2018

Executive Summary

Executive Directive 11 essentially directs the Virginia Department of Environmental Quality to establish regulations limiting C02 emissions from larger electric power plants and to use the Regional Greenhouse Gas Initiative (RGGI) allowance trading program as a compliance flexibility mechanism for enforcing those regulations.

The overall structure of the regulation appears to be appropriate for Virginia’s case and is consistent with our using the RGGI compliance mechanism. The one key weakness of this draft rule is in the setting of the cap. DEQ has made a series of serious errors in estimating business-as-usual emissions. These errors result in a cap in 2020 that is approximately 4 million tons per year too high. The appropriate level of the initial cap should be 30 to 31 million tons rather than the proposed 33 to 34 tons. The specific errors include:

• Underestimating the likely amount of renewables in the BAU case
• Overestimating the growth in generation beyond what even the utilities are expecting
• Mistaking normal weather variation for trends in emissions

These errors are large and all operate in one direction, making it appear that Virginia should have a higher cap. DEQ needs to lower the proposed cap so that it is consistent with available data.

Output-based updating of allowance allocations (9VAC5-140-6215. CO2 allocation methodology)

• Output-based updating of allocations is appropriate and prevents emission leakage.

Model runs using both IPM (ICF) and Haiku (Resources for the Future) show that output-based updating of allowance allocations helps reduce the “leakage” of emissions from under the cap while retaining incentives to shift generation away from high-emitting sources. The mechanism is straightforward. Free allocation of allowances acts as an implicit subsidy for the generation of electricity by granting to ratepayers the market value of the stream of allowances. Generators take this grant into account when calculating their marginal cost of generation and so can maintain relative competitiveness with the generators in the rest of the PJM region. This prevents generation (along with its emissions) from migrating out of Virginia and into the uncapped portions of PJM.

Output-based allocation seems the appropriate choice given the potential for leakage of emissions into the rest of PJM.

Consignment (9VAC5-140-6430. Consignment auction)

• The consignment auction improves efficiency and fairness

Consignment auction of allowances have two key advantages over direct allocation. First, by enhancing liquidity in the auction, requiring consignment probably improves price discovery in the RGGI market. There may also be a subtler mechanism at work. The act of consignment and the resulting requirement that Virginia utilities purchase back what they need may make allowance prices more salient to market players and the generators, in particular. Professors Holt and Shobe at UVA have tested the performance of consignment auctions in laboratory experiments. Although these results are not yet published, they do provide preliminary evidence that consignment auctions improve price discovery in emission markets.

The second benefit of consignment auctions is that they monetize the value of the grant of allowances to the generators, that is, they establish a clear market value of the grant. This allows the State Corporation Commission (SCC) to easily establish whether allowance value is being transferred to ratepayers rather than being retained by generators. Given the value of the free grant of allowances, it is critical that ratepayers be protected from generators pocketing the value of allowances. The consignment auction helps make this possible.

Emission containment reserve

• The emission containment reserve helps correct over-allocation

In every emission market established to date, allowances have been over-allocated at first. In the case of RGGI, the cap has been reduced dramatically due to the initial over-allocation. Even after the initial allocation, costs often fall faster than the cap leading to lower than expected allowance prices. As I will argue in the next section, the proposed rule continues this pattern of over-allocation, since DEQ has set the initial cap too high. This makes the ECR an important backup mechanism for ensuring that our emission reductions will be greater, if the costs of achieving those reductions fall well below expectations.

Cap

• DEQ has set the cap too high due to incorrect analysis.
• The initial cap should be 30 to 31 million tons C02/year rather than 33 to 34.

DEQ has substantially overestimated business-as-usual emissions over the next 15 years. This makes achieving the reductions for a given cap level appear much more expensive than they really are. DEQ’s analysis is not off by just a little, it is grossly in error. The agency has provided an analysis that is inconsistent with facts that were readily available to the agency at the time it did its analysis. What is more, the bias is clearly in one direction, overstating the emissions that would occur in the absence of this rule. This, in turn, overstates the cost of achieving a given reduction.

1. The solar build:  DEQ’s “Reference Case 1” (RC1) assumes that Virginia will generate zero electricity with solar PV for the entire forecast horizon. This assumption is false. At the time DEQ did its analysis, Virginia had more than 100 MW (nameplate) of solar PV in operation with more than 250 MW under construction. By the end of 2017, Virginia had just more than 360 MW of solar PV capacity in operation. This capacity can be expected to generate approximately 720 GWh of electricity per year. This was known to DEQ at the time of its analysis.
    
   In addition to the solar already in operation, the PJM interconnect queue has several gigawatts of solar PV slated for Virginia - and did at the time DEQ did its modeling. Two years ago, Dominion Energy had agreed to have 400 MW in place by 2020, but in April of 2017 the company announced in its integrated resource plan its intention to build around 240 MW per year for the next 15 years because solar PV had a levelized cost of energy cheaper than new coal and competitive with new natural gas generation. This estimated solar build was for Dominion’s “No carbon regulation” case. And Dominion accounts for only about 70% of generation in Virginia. APCO and ODEC had both announced that they were adding solar capacity as well, even in the no carbon regulation case.
   
   There are currently over 700 MW of Virginia solar PV capacity in the engineering and procurement stage. The PJM interconnection queue has close to 6 GW of capacity planned for Virginia in the next few years. Much of this was already on the queue when DEQ assumed zero solar build for Virginia over the next 15 years. At the time of its analysis, DEQ had every reason to know that Virginia would probably have at least one GW of solar PV capacity by 2020. This is roughly 2,000 GWh per year of generation. And yet, the agency assumed in RC1 that there would be zero solar PV built in Virginia before 2031. This is grossly in error and erroneously inflates the appropriate level of the cap. Reference Case 2 (RC2) is only marginally better; again, understating the likely solar PV capacity and generation that would occur in the absence of the rule. The agency assumed that, by 2020, Virginia will have a capacity of 344 MW and will generate only 819 GWh of solar PV electricity. This is less than half of what would reasonably have been expected even before ED11 was announced.
   
   Taking the current 360 MW and adding 240 MW per year (just Dominion’s anticipated build), this implies solar PV generation of about 1,300 - 1,500 GWh more per year than estimated in RC2. If the solar PV displaces half coal and half natural gas, then DEQ has over-estimated C02 emissions by nearly 1.5 million tons per year just due to under-estimating Virginia solar capacity. The mistake is much greater for RC1, where solar PV is incorrectly assumed to be zero.
   
   Summary: By underestimating the amount of solar PV generation that would have occurred without the rule, DEQ has overestimated business-as-usual (BAU) emissions by around 1.5 million tons/year. RC1 is factually false and of no value in emissions estimation, because it assumes less solar than already existed or was under construction in Virginia at the time the analysis was done. Both scenarios ignore already contracted capacity increases in the short run. If DEQ’s staff in charge of estimating emissions is having trouble estimating future solar PV installations, they might call DEQ employee Beth Major, the contact point for small renewable installation permitting. She can help out.
    
2. Rate of growth in generation: Both of DEQ’s Reference Case scenarios err by assuming wildly unrealistic rates of growth in electricity generation. This, in turn, results in unrealistically high capacity factors for coal plants in Virginia and unrealistic growth in fossil fuel generation capacity (mostly natural gas). This mistake in DEQ’s analysis further inflates expected business-as-usual emissions and is used to justify a higher cap than is necessary.
   
   In April of 2017, Dominion Energy published its integrated resource plan (IRP), in which it reports its forecast of future growth in electricity consumption in its service area. This document, is public and presumably well-known to DEQ staff. In its IRP, Dominion estimates future generation growth to be 1.3% per year. (In fact, my analysis shows that Dominion has a pattern of over-forecasting demand growth, but more on this shortly.) Accepting Dominion’s own estimate for demand growth for now, it is obvious that DEQ has once again made a serious error in its modeling of reference case emissions by assuming unrealistically high growth rates in generation. DEQ’s generation scenario for RC 1 has generation growing at an average rate of 1.9% per year and RC2 has it growing at a whopping 3.4% per year.
   
   Dominion represents 70% of generation in Virginia. If Dominion estimates that its energy production will rise at 1.3%, then other parts of Virginia must be rising much faster, to make up the difference. But this is not true. The APCO region, which is the second largest in Virginia, has flat or declining demand due to out-migration from the region. The remainder of the state is too small to make up the difference, but does not have growth rates higher than Dominion’s, in any event. DEQ has made a serious error by assuming a higher growth rate for generation than the electric utilities are using in their own capacity planning. This mistake inflates the estimated need for fossil fuel combustion in future years.
   
   Furthermore, Dominion’s own forecasts are too high. My analysis of Dominion’s forecast, presented to the State Corporation Commission last year, clearly shows that Dominion has over-forecast demand every year for which we have actual observations, since at least 2012. Its forecasts of future generation have fallen dramatically over this same period but are still too high and will continue to fall in the next few years because of a flaw in its forecasting methodology.
   
   Generation has grown faster than demand since 2015 because of a Virginia state policy to repatriate generation and reduce imports of electricity. The process of repatriating imports is now essentially complete. Dominion is even anticipating small amounts of exports over the next few years, given that it has just completed, or is nearing completion of, three large new natural gas generators. Now that the process of repatriating generation is complete, generation and demand will tend to grow at the same rate.
   
   Recent growth in electricity demand in Virginia has been less than 1% per year even as the state economy has grown following the last recession. Breaking electricity demand into its component parts helps us understand why. Recent trends in both residential and industrial demand have been negative, that is negative growth in demand. In the industrial sector, this is due to a shift to less energy intensive industries. In the residential sector, this is due to the penetration of energy efficient technologies and improvements in the energy performance of the building shell.
   
   The commercial sector can be divided into two parts, server farms and other. Non-server farm demand has been declining in the past few years, while server farm demand has been rising rapidly. So, the one source of increase in electricity demand in Virginia in recent years has been server farms. But this is a small fraction of overall electricity demand in Virginia and is already accounted for in Dominion’s forecast. DEQ has no basis for its grossly overstated estimates of future demand growth in Virginia. Its assumptions are worse than arbitrary, they are perverse.

   Another fact about server farm demand is relevant. Many firms building server farms want to cover their energy demand with renewable generation and the firms are increasingly insisting that the generation be local. Google, for example, now claims to be powered by 100% renewable energy. And just last month, Microsoft announced that it has purchased about 350 MW of a 500 MW facility planned for Spotsylvania County for completion in 2019 and 2020. (Keeping in mind that this 500 MW facility doesn’t exist in any of DEQ’s baseline scenarios.) Server farm demand cannot account for the growth in fossil fuel emissions assumed in DEQ’s faulty analysis.

   A simple exercise shows how damaging this mistake is to DEQ’s analysis. DEQ’s two reference cases make different assumptions about 2017 total generation: 96,786 for RC1 and 93,305 for RC2. At the time DEQ did this analysis, there was zero chance that demand would be as high as assumed in RC1, but this is consistent with the general pattern of unsupported and erroneous assumptions in its analysis. Actual generation for 2017 was 93,500 GWh.

   To be very conservative, we take the higher of the two 2017 generation estimates from DEQ’s reference cases, 96,786 GWh (even though it didn’t actually happen) and increase it at 1% per year. The resulting generation profile shows that DEQ’s assumed generation is in excess of any reasonable expectation by 3,600 GWh per year by 2020 and 10,500 GWh by 2031. If you assume that each GWh displaces half coal and half natural gas, then each 1,000 GWh is associated with on the order of 1 million short tons of C02. In light of this, it is clear that DEQ’s analysis has grossly overestimated BAU emissions. Combined with the solar PV analysis, the 2020 emission overestimate is on the order of 4 million tons of C02 per year.
   
   The assumption of half displacement of gas and half coal is probably somewhat conservative. Chances are that more coal dispatch will be displaced. This is clear from Dominion’s 2017 IRP. The company had a BAU scenario and a scenario for operating under a cap under the Clean Power Plan (CPP). One of the major differences between these two scenarios is the retirement of significant coal capacity in 2020, when the CPP was to come into force. These coal plants were not retired under the BAU scenario. This implies that substantial reductions in coal dispatch can be anticipated under this cap, which will ultimately be tighter than what would have been true under the CPP. And coal dispatch is already falling sharply due to the addition of the new natural gas capacity. Net electricity generation from coal in Virginia fell from 15,600 GWh in 2016 to 10,110 GWh in 2017. This downward trend will continue as Dominion brings its Greensville natural gas power plant online in 2019.
    
3. Imaginary emission trends: Emblematic of DEQ’s difficulties in its modeling exercise to establish Virginia’s cap is what appears to be an amateurish mistake of finding “trends” in recent C02 emissions where there are none. In a presentation (available here) titled “Virginia Carbon Dioxide Trading Program: Proposed Regulation” [undated, unsigned], DEQ presents a slide on page 3 with the title, “Virginia Power Plant C02 Emission Trends”. The slide does not show “trends” it is simply a bar chart showing annual C02 emissions. Emissions appear to fall until 2012 and then start rising from 2012 to 2016. Presumably, this apparent u-shape is what the title of the slide is referring to. The problem here is that the apparent movement in emissions does not represent “trends”. Emissions fell rather abruptly to around 30 million tons per year during the Great Recession and then never recovered. The apparent movements around 30 million are almost all due to weather. Weather normalized emissions for 2011 and 2012 are around 30 million tons. Even 2016 is a special case, with a particularly hot summer where Virginia was generating its own electricity rather than importing it. Given the one third drop in coal dispatch in 2017 from 2016 and flat demand (see EIA data), 2017 emissions will fall back to near the 30-million ton level, leaving the “trend” in emissions completely flat since 2009.
   
   DEQ has failed to make a case for a cap greater than 30 million tons per year. In recent years, any increases in generation due to load growth (including repatriating imports) has been offset by reduced emission intensity of generation. Since nearly all increments to generation in Dominion’s IRP are solar PV, through to the end of the 15-year planning horizon, emission intensity is bound to fall further.
   
   4) Natural gas prices: In its reference cases, DEQ assumes a natural gas price of $2.83 in 2017 rising to $3.95 in 2020. In April, 2018, the spot price of natural gas is hovering around $2.75/MMBtu. To match DEQ’s assumption, natural gas prices must rise more than 30% in the next two years. And yet, the futures price for natural gas, as of April 3, 2018, is $2.70. DEQ just assumed a high rate of growth in natural gas prices and plugged that assumption into its model even though it was well known at the time that there was a substantial probability that the price would be lower. This adds just one more upward bias in the estimated business-as-usual emissions.