The Regional Greenhouse Gas Initiative (RGGI) was supposed to be nearing completion of a 2016 Program Review but the election of Donald Trump and the fate of the national Clean Power Plan has delayed that process. This is the third post in a series of posts that will discuss how RGGI has fared so far and how that could affect the program review. As noted previously, I believe that RGGI allowance prices add to the cost of doing business but because the cost of allowances can be added into the bid price it is a nuisance and not a driver of decisions. This post addresses the effect RGGI has had on other pollutants as estimated by Abt Associates: Analysis of the Public Health Impacts of the Regional Greenhouse Gas Initiative, 2009 -2014.
According to the Energy Coordinating Agency this report shows that since 2009, RGGI has significantly reduced air pollution from fossil fuel power plants. The report goes on to estimate improvements to the health of people living in the Northeast as a result of those pollution reductions. According to the study the effort to curb carbon emissions has prevented 300-830 adult deaths, avoided 13,000 – 16,000 respiratory illnesses and staved off 39,000 – 47,000 lost work days for workers.
I have been involved in the RGGI program process since its inception. Before retirement from a Non-Regulated Generating company, I was actively analyzing air quality regulations that could affect company operations and was responsible for the emissions data used for compliance. The opinions expressed in this post do not reflect the position of any of my previous employers or any other company I have been associated with, these comments are mine alone.
I will briefly describe the three steps that Abt Associates used in their analysis. This post will primarily analyze the first step in their process. I compare their predicted generation and emissions reductions relative to the total reductions observed in the measured data. For the last two steps I offer some general comments.
Abt Associates used a three-step analytic process to estimate the impacts on air quality and public health resulting from implementation of the RGGI program from 2009 to 2014. In each of these steps they used specific modeling tools and datasets to estimate the incremental impacts of RGGI on the following variables: generation (in megawatt-hours (MWh)) by power plants, air pollution emissions, air quality, and public health. The approach is sequential. Results steps one and two were input to the final step that projected health impacts. Abt Associates noted that they “reviewed draft results at a highly disaggregated level and performed quality control before using results as an input to the next analytic step. In many cases, draft results were benchmarked to results from similar analyses and studies as another cross-check.”
The three steps are described as follows:
Step 1: Estimate annual changes in electric generation and emissions of air pollutants at power plants as a result of RGGI implementation from 2009 to 2014 using electricity dispatch modeling and EPA emissions data for EGUs.
Step 2: Estimate annual changes in air quality at the county level associated with changes in SO2 and NOx emissions from power plants, by year. EPA Co-Benefits Risk Assessment (COBRA) screening model
Step 3: Assess public health impacts associated with changes in air quality due to RGGI implementation from 2009 to 2014. EPA Benefits Mapping and Analysis Program (BenMAP).
Abt Associates calculated changes in electricity generation due to RGGI first on an absolute basis and then expressed them as a percentage relative to generation that would have occurred in the No RGGI scenario. They note that:
In competitive electricity markets, changes in any number of variables, including fuel prices, weather, and plant and system operational changes, can cause variations in the level of electricity dispatched by a given power plant (or group of plants) from year to year. In this analysis, however, our modeling results account for all of these factors and thereby isolate the incremental effect of RGGI on electricity markets and dispatch. Thus, we interpret RGGI-induced changes in generation to be a result of the combination of (1) RGGI states’ investments in energy efficiency and renewable energy and (2) the effect of CO2 allowance prices on electricity dispatch.
Therefore their Table 4 results in Comparison of Abt Associates Generation Changes Due to RGGI relative to Total Generation Changes are the reductions they assume that would not have occurred were it not for RGGI. Let’s try to compare their estimates with the total generation reductions that occurred.
I used measured data for my estimate of the total load reduction and that methodology has differences relative to their approach. I assume that their annual differences in Table 4 are relative to 2008. The EPA Clean Air Markets Division website provides gross load generation data from all electric generating units that participate in RGGI. Modeling electricity generation in their approach uses net load so there is a difference there. In my analysis I did not include New Jersey and they did for the first compliance period. I did not go to the effort to manually include only those sources that participate in RGGI for the 2008 data. Instead I calculated the load from all programs that report to EPA in 2008. I believe that is a minor error because the difference in the RGGI only and “All Program” data sets in the years 2009-2014 ranged from 0.50% to 0.71%.
In my approach I calculated the total change in generation from 2008. The Abt Associates approach calculated reductions in annual generation ranging from a low of 2.0 percent in 2011 to a high of 7.0 percent in 2013. The difference in gross load between 2008 and the six years in the first two RGGI compliance periods ranged from a low of 3.5% in 2010 to a high of 18.7% in 2014. If we assume the two approaches are compatible, then we can estimate the fraction of the total reduction resulting from the Abt Associates estimates of RGGI investments and CO2 allowance prices on electricity dispatch. In 2010, this comparison suggests that the fraction of reductions induced by RGGI was over 80% of the total reduction. Even in the lowest year, 2011, the fraction of reductions induced by RGGI was 26.4%.
Disappointingly, the Abt Associates analysis does not provide an estimate of CO2 emission reductions determined by their approach so that we can directly compare alternative approaches provided in previous work. Although generation is not directly comparable to CO2 emissions it is relevant to point out the conclusions from my previous post:
The upper bound in CO2 emissions reductions due to RGGI is an econometric model that estimates that emissions would have been 24 percent higher without the program. RGGI estimates that emissions would have been 17% higher than without a program. If you assume that all the savings in fossil fuel use only displaced natural gas use instead of some other aggregation of fuels then emissions would have been only 5% higher.
These numbers are starkly different and need to be addressed before the Abt Associates analysis can be considered credible. To that end I recommend that Abt Associates provide their CO2 emission reduction estimates.
In the first step Abt Associates estimated annual changes in electric generation and emissions of air pollutants at power plants as a result of RGGI implementation from 2009 to 2014 using electricity dispatch modeling and EPA emissions data for Electric Generating Units (EGUs). Essentially they are estimating the emissions that would have occurred were it not for the RGGI investments and CO2 adder cost to the dispatch cost of the plants. Using the same methodology as for the generation estimate I calculated the total emission reductions. Sulfur dioxide cumulative reductions from a 2008 baseline totaled 1,673,601 tons from all nine RGGI states compared to the ABT Associates estimated impact of RGGI reduction of 109,000 tons so the estimated reduction of RGGI is 6.5% of the total. Nitrogen oxides cumulative reductions from a 2008 baseline totaled 335,440 tons from all nine RGGI states compared to the ABT Associates estimated impact of RGGI reduction of 46,000 tons so the estimated reduction of RGGI is 13.7% of the total. While I have reservations that RGGI had any impact other than from the load associated with RGGI investments, I can accept them as upper bound estimates of the SO2 and NOx emission reductions due to RGGI.
In the second step Abt Associates used EPA’s Co-Benefits Risk Assessment (COBRA) screening model. According to EPA, COBRA works as follows:
- COBRA contains detailed emission estimates of PM2.5, S02, NOX, NH3, and VOCs for the year 2017 as developed by the U.S. EPA. Users create their own scenario by specifying increases or decreases to the baseline emission estimates. Emission changes can be entered at the county, state, or national levels, and outcomes can be modeled nationwide or for smaller geographic areas.
- COBRA uses a reduced form air quality model, the Source-Receptor (S–R) Matrix, to estimate the effects of emission changes on ambient PM.
- Using an approach to estimating avoided health impacts and monetized benefits that is generally consistent with EPA practice, the model translates the ambient PM changes into human health effects and monetizes them.
- Users can view the results in tabular or geographic form.
There are multiple problems with this model for this application. EPA notes that it is “inflexible and simple” such that it is limited to the 2017 timeframe and there is no ability to import a different baseline. Abt Associates applied the model to different periods and used a different baseline. The RGGI sources are point sources and this model does not address their specific factors that convert emissions from these plants to particulate matter as opposed to the generic factors used. Finally, note that all the air quality impacts are a function of particulate matter.
Secondly, BenMAp “estimates the number and economic value of heath impacts resulting from changes in air quality – specifically, ground-level ozone and fine particles but note that the COBRA impacts are solely based on particulate matter. The model presumes that the health impacts of both pollutants include premature death and aggravated asthma. Ambient levels of sulfur dioxide, nitrogen oxides and particulate matter have all been reduced in recent years but at the same time asthma rates have been increasing. This inconsistency needs to be addressed before this approach can be accepted.
There also is a deeper issue associated with EPA epidemiological work with PM and health impacts that is the basis of BenMAP. EPA’s epidemiological work is based on data sets that are not available for independent review, were prepared by organizations that were being paid by EPA (and would lose funding if this were not an issue) and the relationship they claim is not present in other similar data sets. Moreover, EPA has yet to determine the specific effect of PM on humans that triggers the claimed impacts of their epidemiological studies. Given those limitations I do not accept these health effects. I encourage you read Steve Milloy’s book “Scare Pollution: Why and How to Fix the EPA” (2016) Bench Press for the complete story of this travesty.
Finally, we can make an order of magnitude estimate of the health outcomes using their methodology by simply scaling their emission reductions relative to the total emission reductions (Abt Associates Cumulative RGGI Health Benefits and Total Cumulative Health Benefits, 2009-2014). This estimate does not account for even the gross site specific implications of the Abt Associates analysis. However, because the air quality estimated concentrations are proportional to the emissions input the general concept is acceptable. There is one other gross assumption. Because fine particulates in the northeast are primarily related to SO2 and BenMAP calculates health impacts of particulates, I only scaled the SO2 emissions to predict the health impact differences. My point of these numbers is that the predicted impacts are large enough that someone should be able to prove this model works by evaluating the observed health effect data. Until that is done I will remain skeptical of this approach.
Not to be petty but in order for an analysis to be useful in public policy debate the results have to be reproducible. This analysis went out of its way to make replicating their numbers difficult. Frankly when there are obvious obfuscations you start to wonder if it was deliberate. To prevent that perception providing numbers up front is the safest course. Also note that I did ask the corresponding author for data but did not get a response. The following is a list of specific whiny issues I had with the report.
What was the baseline? I could not find this definition anywhere. I assume that because annual numbers were used that the year before the program started was the baseline. In my previous analyses of RGGI data I use the three-year period 2006 to 2008 as the baseline to compare with the three-year RGGI compliance periods.
SO2 and NOX data in metric tons. I do not understand why the emissions are reported in metric tons. In order to compare their results with any EPA report they have to be converted. Moreover, the input to the COBRA model is in short tons.
No tables with numbers. In order to compare their projections with actual emissions I had to manually scale bars on their figures. I think my estimates are good enough but this was an unnecessary hassle.
No CO2 reductions report. This was serious enough to include in the main body but bears repeating. In order for this study to be credible it needs to be compared to previous work. Without CO2 that is impossible.
Emission reductions in Figures 6 and 8 don’t add up to the emission reductions in Figure 7. After manually estimating the bars in the two figures it was very frustrating to find they didn’t match their totals. Double checking my work it is clear that there is a mistake in the Abt numbers.