NOTE

More on the Recent US Emissions Decline

March 4, 2013

Last month we published a note, Coal Claws Back, in which we observed that the recent decline in US CO2 emissions, due in large part to low-cost natural gas, appears to be bottoming out. As natural gas prices have risen in recent months, coal’s share of US power generation has recovered — from a low of 33% in April to 42% in November of last year. In that note, we looked at how much of the reduction in the carbon-intensity of US energy supply since 2005 (a major factor in the drop in CO2 emissions, along with weaker energy demand) came from natural gas, nuclear and renewables. We found that while natural gas played the leading role, wind, biomass and biofuels were also an important part of the story. The Breakthrough Institute (BTI) took issue with our analysis in a piece last week titled “Debunking Rhodium”, arguing that natural gas is responsible for at least 3-10 times the emission reductions of non-hydro renewables.

What explains the difference?  Mostly, it’s that we’re asking different questions:

Base year or business-as-usual: We analyze how much of the change in emissions trajectory since 2005 reported by the Energy Information Administration (EIA) was due to slower-than-expected economic growth vs. faster-than-expected reductions in the energy-intensity of growth vs. changes in the carbon-intensity of energy supply. To do that requires comparing actual 2012 emissions to what they would have been had the country continued along a business-as-usual trajectory. In their critique, BTI focuses on the reduction in absolute emissions since 2007. Both are reasonable approaches, they just answer different questions.

Economy-wide or power sector: In analyzing the reduction in carbon-intensity of energy supply that’s occurred since 2005 (the amount of CO2 emitted per unit of energy consumed) we look at all sectors – residential, commercial, industrial and transportation as well as the power sector. BTI focuses just on the power sector. Within the power sector, our analysis also shows natural gas accounted for the majority of the emission reductions achieved through fuel switching.

What displaces what: Figuring out which fuels displaced which other fuels in a rapidly changing US energy mix is the most challenging part of the analysis. We proportionally allocate the decline in carbon-intensity of energy supply within each sector reported by the EIA thanks to less coal and oil demand to natural gas, nuclear and renewables based on the increase in market share of each fuel and their relative emissions factors. BTI raises the important point that within the power sector, growth in renewables may not displace as much coal-fired generation as growth in natural gas. As an alternative, they use estimates from the academic literature of how much CO2 each kWh of additional wind displaces and apply this to all non-hydro renewables (as wind accounted for the vast majority of the growth in the power sector) for an estimated 30-100 million tons in CO2 reductions. This is a useful approach for assessing the impact of a single fuel, but also has weaknesses when assessing the relative weight of each fuel in delivering the total emission reductions reported by EIA (our goal). For example, in the power sector, EIA data for January-November puts CO2 emissions (on an annualized basis) at 373 million tons below 2007 levels. We estimate 62 million tons of that came from a fall in electricity demand, which leaves 311 million tons from fuel switching. BTI estimates that non-hydro renewables and natural gas combined reduced emissions by 330-600 million tons, more than the total reductions reported in the data.

Bottom line: A more robust approach to assessing the relative role natural gas, nuclear and renewables have played in the recent emissions decline than either we or BTI employed would be to use a detailed energy model to isolate the impact of the change in supply of each fuel source across sectors and CO2 emissions from the economy as a whole. Such analysis would certainly be interesting, and we may look to undertake it in the months ahead. More important than this accounting exercise, however, is the outlook for emissions going forward, which was the primary purpose of our February piece. Regardless of exactly how large natural gas’s role has been in reducing US emissions to date, recent data and industry and government forecasts suggest it will take new policy to extend those emissions cuts going forward as both the economy and natural gas prices begin to recover.


Base Year or Business-as-Usual?

Our note began with an estimate based on January-October data from the US Energy Information Administration (EIA), that US CO2 emissions from energy combustion have declined by 12.6% relative to 2005 levels[1]. This is a noteworthy development given America’s 2009 pledge in Copenhagen to reduce emissions 17% below 2005 levels by 2020, something that energy forecasters did not see coming [2]. At a macro level, three factors determine a country’s energy-related CO2 emissions trajectory: the rate of economic growth, the energy-intensity of the economy (how much primary energy is consumed per unit of GDP), and the carbon-intensity of energy supply (the amount of CO2 emitted per unit of primary energy consumption). We wanted to know which of these had changed to deliver the recent emissions reduction surprise.

As explained in our piece, the only way to do this is to create a counter-factual: what would have happened to emissions had the economy grown at its potential, and energy-intensity and carbon-intensity had continued to follow historical trends between 2005 and 2012. This is a common approach employed by analysis to assess the impact of changes in energy supply or energy policy (see here, here and here). Government and academic economists estimate current potential GDP growth at around 2.5% per year. Between 2005 and 2012, the economy grew at 1.1%. Between 1990 and 2005, the energy-intensity of the US economy declined by 1.9% a year. Between 2005 and 2012, it declined by 1.8%. And the carbon-intensity of energy supply, which had remained more or less unchanged between 1990 and 2005, fell significantly between 2005 and 2012.

So it was slower economic growth and less carbon-intensive energy supply that were the big surprises. Energy-intensity improvements were very important in keeping emissions in check, but the rate of improvement between 2005 and 2012 wasn’t any greater than between 1990 and 2005. As we pointed out, energy efficiency improvements in cars, buildings, industry and power generation were likely greater between 2005 and 2012 than in prior years, but the multi-decade structural shift in the US economy from manufacturing (more energy-intensity) to services (less energy-intensive) slowed after 2005, which offset these technical efficiency gains when measured from an economy-wide standpoint.

We then performed a crude assessment of how the change in fuel mix had impacted the carbon-intensity of energy supply. Using monthly data from the EIA, we allocated the reduction in carbon-intensity of energy consumption in each sector (residential, commercial, transportation, industry and electric power) to fuel (natural gas, nuclear and renewables), based on their change in share between 2005 and 2012 and the direct CO2 emission factors of each fuel within that sector as reported in the monthly EIA data, and sum them up across sectors.  

BTI takes issue with this approach, preferring to measure US emissions against 2007 levels, the year that US CO2 emissions peaked. This is a reasonable approach for assessing the impact of fuel switching on US CO2 emissions through lower carbon intensity, but doesn’t tell you how much of a role energy-intensity improvements or slower economic growth played in determining the recent US emissions trajectory. Those were the questions we were attempting to answer.

Economy-wide or Power Sector?

Second, in our assessment of what contributed to the decline in the carbon-intensity of US energy supply, we looked at all sectors while the BTI analysis is focused solely on power generation. Roughly half of the reduction in carbon-intensity due to increased renewables penetration in our analysis occurred outside the power sector. Of this, roughly half was from increased use of biofuels in the transportation sector. BTI rightly notes that this estimate does not account for the life-cycle impact of biofuels, something we flagged in the original report. Those emissions are not reported in the EIA data, nor are methane emissions from natural gas production, or the life-cycle emissions associated with manufacturing wind turbines and solar panels. People are talking about how the EIA is reporting 12.6% decline in emissions relative to 2005 levels and we wanted to explain why that’s happened. Had we used a different metric from the EIA one (e.g. including land use), we couldn’t have done that.

The other half of the non-power reduction in carbon-intensity due to renewables came mostly from biomass used in industrial boilers. Roof-top solar in the residential sector also played a meaningful role. Applying our methodology to the power sector alone, natural gas accounted for 51% of the reduction in carbon-intensity between 2005 and 2012 (using January-November data), while wind accounted for 38% and other renewables accounted for 5%.

BTI raises another important point in their critique: the switch from coal to natural gas improves energy efficiency in addition to reducing carbon-intensity. The reason is that natural gas-fired power plants require less primary energy (measured in BTU) to produce the same amount of electricity (measured in kWh) – known as thermal efficiency. According to the EIA data, natural gas-fired power generation had a 23% higher thermal efficiency than coal-fired power generation in the US in 2012. This is not captured in our carbon-intensity decomposition, which uses primary energy, but is included in our assessment of the reduction in the energy-intensity of the economy along with other technical efficiency improvements. We did not try to attribute fuel switching-related reductions in energy-intensity for the following reasons:

1. Our top-line goal was to assess the relative contributions of slower economic growth, reduced energy-intensity of economic growth, and reduced carbon-intensity of energy supply to recent emission reduction trends.

2. Switching from coal to non-combustible renewables also has an efficiency effect (no thermal conversion loss as its not thermal energy) but this is not captured in the EIA data. In calculating the primary energy consumption required for each kWh of non-combustible renewable generation, the EIA uses the average thermal efficiency of the fossil fuel fleet.  This differs from the International Energy Agency (IEA) which uses a 100% conversion efficiency assumption.

3. A change in the energy mix impacts energy-intensity in ways beyond thermal efficiency which are difficult to capture. For example, while lower cost natural gas might improve efficiency in power generation, it could also lead to increased overall energy demand in manufacturing as energy-intensive industries like chemicals and fertilizer expand, and in the residential and commercial sectors as households and businesses are able to get more energy for their buck.

For a power sector-only analysis, you can measure energy in kilowatt hours rather than BTU, which avoids the thermal efficiency issue mentioned above. Table 1 below lists electric power generation by fuel and resulting CO2 emissions as reported in the EIA’s February Monthly Energy Review (MER). While full-year generation data is available for 2012, CO2 emissions data is only available for January-November, so we’ve used that range for both and annualized it to estimate full-year figures. We use here a 2007 base year to be consistent with the BTI analysis. To control for emission reductions resulting from lower electricity demand over the past five years, we calculate what 2012 emissions would have been had the generation mix stayed at 2007 levels (e.g. coal at 49.7% rather than 38.3%). 

Coal, oil and nuclear all suffered a decline in generation between 2007 and 2012, with coal accounting for the vast majority of the total [3]. Using annualized January-November data, coal generation fell by 498 billion kWh, oil fell by 42 billion kWh and nuclear fell by 37 billion kWh. Natural gas and renewables increased generation by 475 billion kWh, with natural gas accounting for 70.6% of the total and wind 21.8% (Table 2). Overall power demand was down by 102 billion kwh.

Using the emission factors from the 2007 data (amount of CO2 emitted per kWh generated), the decline in coal and oil generation reduced CO2 emissions from those fuel sources by 501 million tons. To assess the contribution of each of the fuels listed in Table 2 in cutting those emissions we calculate what generation in 2012 would have been had each fuel maintained its 2007 share of generation. We then proportionally distribute the decline in emissions from reduced coal and oil-fired power generation and factor in the increase in emissions from higher natural gas generation.

We proportionally distribute the decline in electricity demand among fuels as well, which accounts for 62 of the 373 million ton reduction in power sector CO2 emissions between 2007 and 2012. Of the 311 million tons from fuel switching, 58% comes from natural gas, 30% from wind, 9% from hydro, with the rest distributed among solar, geothermal, waste and other minor sources of generation. Note that this analysis covers the electric power sector only, not end-use generation like onsite combined heat and power or roof-top solar.    

What Displaces What?

BTI’s final critique is that our market-share analysis ignores differences among fuels in what they displace. For example, in the power sector analysis above, we are assuming that the decline in coal, oil and nuclear generation not due to a decline in demand is replaced by natural gas and renewables proportional to their share of the overall generation increase. One new kilowatt hour from wind replaces the same ratio of coal, oil and nuclear generation (the fuels that have declined over the past five years) as one kilowatt hour from natural gas. We chose this approach for simplicity, and because of our economy-wide scope.   

In reality, there are important regional and sectoral differences in which fuels compete for dispatch in competitive markets, or what criteria public utility commissions apply in regulated markets in approving new capacity. BTI cites a paper from the Colorado School of Mines that estimates that new wind power competes with other renewables and natural gas, as well as coal, and that as a result, an additional megawatt hour (MWh) of wind generation displaces anywhere between 0.08 tons of CO2 and 0.92 tons of CO2, depending on where in the country it occurs. That’s less than the 0.95 emissions saving rate used in our proportional approach. And its one reason why BTI estimates non-hydro renewables reduced emissions by 30-100 million tons between 2007 and 2012 while Table 2 lists non-hydro renewables reductions of 98.5 million tons and total renewables reductions of 127.4 million tons. The Colorado School of Mines paper estimates of 0.612 tons per kWh emissions saving rate nationally. Apply that to the growth in non-hydro renewables listed in Table 2 and you get 67 million tons.

While these kind of empirically-derived estimates of emission savings rates provide more accuracy in assessing the impact of an increase in one type of generation in isolation, they don’t work well for attributing the impact of increases in multiple sources of generation at once. The same regional variation the Colorado School of Mines paper finds with wind is true with natural gas and as well, though likely to a lesser extent. If wind competes with natural gas for dispatch, then it stands to reason that natural gas competes for dispatch with wind. Indeed, the IEA’s Global Age of Gas study found that increase natural gas generation thanks to lower gas prices displaces renewables and nuclear as well as coal. Comparing the high natural gas price (called “Low Estimated Ultimate Recovery”) and low natural gas price (called “High Technically Recoverable Resources”) side-cases from the EIA’s 2012 Annual Energy Outlook shows a similar effect in a US context (though there is more renewables than nuclear displaced than in the IEA’s global assessment).

So for an apples-to-apples comparison, you would need to apply an emission savings rate less than the emissions factor of coal to additional gas generation as well as to renewables. We would guess that policy support for renewables like the Production Tax Credit and state-level renewable portfolio standards makes wind generation less sensitive to changes in natural gas generation than vice-versa and thus leaves wind with a lower emission savings rate, but you’d need to run a couple scenarios through a good dispatch model to get accurate assessment when natural gas and renewables are deployed in tandem.

The BTI analysis does not suggest what the emissions saving rate for natural gas should be. Instead, they offer a range of 300-500 million tons or total emissions reductions resulting from increased natural gas generation. We are unable to square that range with the 311 million tons of reductions in power sector emissions the EIA reports occurred between 2007 and 2012 due to fuel switching, or even the 373 million tons of reductions from the power sector overall. Even if you include BTI’s lower-end estimate of 30 million tons of reductions for non-hydro renewables, take the lower-end estimate for natural gas (300 million tons) and ignore the 29 million tons of reductions we estimate occurred from hydro all together, then you are still 19 million tons above the emissions reductions reported by the EIA.

BTI offers three citations for their 300-500 million natural gas ton estimate. The first is a November 2012 report from NREL that states “this switch from coal to natural gas, combined with growth of renewable energy generation, has led to a reduction of carbon dioxide emissions in the U.S. power sector of about 300 million tons.” That’s pretty consistent with our findings in Table 2. The second is a blog posting from former Pennsylvania DEP Director John Hanger in April 2012 who projected that the growth in natural gas’s market share would be responsible for 300 million tons of emission reductions by the end of 2012, but he uses 2000 as a base year, not 2007. And the third is an NPR interview with David Victor from July of 2012 in which he estimates that natural gas will reduce US emissions by 400-500 million tons.

David is a good friend and top-notch analyst so we asked him for background on this number. He wanted us to note that he is the first to say that his numbers are simple calculations based on monthly power shares rather than decomposing all the many different factors that affect power dispatch.  He pointed us to a ‘dot Earth’ blog posting where he first talked about these issues, in which he states “if the U.S. electric grid operated as it did in the 2000s with coal as king then annual national emissions would be 475 million metric tons of CO2 higher”, which is the total decline in emissions from falling coal-fired power generation, not necessarily not just natural gas. He also did the analysis in July, and used the most recent EIA monthly data available at that time – April 2012. As we noted in our analysis, April was the low-point for coal-fired power generation and in the months since coal’s share has recovered from 33% to 42%. David agrees that “someone should work through the details some time with a real dispatch model that is run for mid 2000s conditions and today”. He also flagged a new essay he will publish shortly in Energy & Security: Towards a New Foreign Policy Strategy on the opportunities and challenges in further scaling gas-fired power generation in the years ahead both in the US and globally, which we look forward to reading.

Bottom Line

Properly attributing the recent emissions decline is a challenging exercise, both in the relative roles played by changes in economic growth, energy-intensity and carbon-intensity, and within carbon-intensity the contribution from individual fuels. We are delighted that BTI has stress-tested our approach and offered some important qualifications to keep in mind when considering the power sector piece of the equation. As mentioned, a more robust approach to assessing the relative role natural gas, nuclear and renewables have played in the recent emissions decline than either we or BTI employed would be to use a detailed energy model to isolate the impact of the change in supply of each fuel source across sectors and CO2 emissions from the economy as a whole. Such analysis would certainly be interesting, and we may look to undertake it in the months ahead. More important than this accounting exercise, however, is the outlook for emissions going forward, which was the primary purpose of our February piece. Regardless of exactly how large natural gas’s role has been in reducing US emissions to date, recent data and industry and government forecasts suggest it will take new policy to extend those emissions cuts going forward as both the economy and natural gas prices begin to recover. 


[1] Annual CO2 emissions for 2005-2011 were taken from the EIA Annual Energy Review Table 11.1. 2012 emissions were estimated by applying the year-on-year change in January-October CO2 emissions from the EIA Monthly Energy Review and applying it to 2011 annual emissions. Note that the Annual Energy Review emissions estimates vary slightly from the Monthly Energy Review numbers. Last week EIA published November CO2 data which takes our estimate for full year emission reductions down to 12.1% below 2005 levels, due to the recovery in coal-fired power generation discussed in our original note.

[2] See for example, the EIA’s Annual Energy Outlooks from 2005-2008.

[3] The small attribution of carbon-intensity reduction attributed to nuclear in our February note despite the slight decline in generation is the result of using a primary energy, rather than kWh calculation.