Skip to content

Subsidizing renewables for the damage not done

In this divided age, few topics beyond motherhood, apple pie, and the iPhone 6 enjoy widespread public approval. So it is notable that, in a recent Gallup Poll, two out of three Americans support an increased reliance on solar and wind energy sources.


While (almost) all of us seem to agree that more is better when it comes to renewable energy, things get more complicated when it comes to determining what form that more should take.  In other words, how do we get the biggest bang for our green energy buck?

In last week’s blog post, Severin argued convincingly that the answer lies in the creation of policy and market incentives that accurately reflect the real benefits and costs of different renewable technology options. A great idea!  But messy and controversial to implement.  To design these incentives, we need to measure and monetize the various costs and benefits that alternative energy technologies incur and afford.

Some co-authors, Duncan Callaway and Gavin McCormick (who was featured in an earlier post), and I have been trying to tackle one corner of this larger valuation exercise. In some ongoing research which will be released soon as an Energy Institute working paper, we estimate the greenhouse gas emissions impacts associated with incremental increases in renewable energy generation in different parts of the country.

The basic idea is as follows: when a wind turbine or solar PV system is connected to the electricity grid, the clean energy produced will displace electricity generation at other sources. We estimate the associated emissions impacts which largely depend on the emissions intensity of the marginal production that gets crowded out.

At this point, you might be wondering why we should be so concerned with measuring the emissions damages not done. If our objective is to design policy incentives to accurately reflect emissions costs, why not penalize emissions damages directly with an emissions price?

“Tax carbon” is a hallowed refrain on this blog (and on a vanity license plate of an economist we know and love). But when it comes to designing policies to encourage renewable energy, production-based subsidies and credits (such as the production tax credit and renewable portfolio standards) are a politically preferred policy instrument in the U.S. This is true now, and as states look to leverage existing renewable energy policies to comply with the proposed Clean Power Plan, this could hold true for the forseeable future.  So long as we are in the business of subsidizing renewables for avoided emissions damages, it’s worth thinking about how to design these second-best incentives.

Measuring damages avoided

To estimate the emissions impacts of marginal changes in electricity generated by existing sources, we use hourly data from six major independent system operators (ISOs) in the United States over the years 2010-2012. We then match these estimates with simulated renewable energy production across thousands of wind and solar sites to estimate the average quantity of emissions displaced per MWh of renewable energy generated across different regions and technologies. We also consider the emissions impacts of some common energy efficiency improvements.

The figure below summarizes our estimates of avoided emissions on a per MWh basis over the 2010-2012 period. The colors denote the different technologies we consider. Technology-specific estimates are grouped by region. The bars of each box plot denote the range of/variation in our estimates due to the day-to-day variability in power system operations.


Pounds  of Carbon Dioxide displaced per MWh of renewable energy generated (or energy saved)

The graph shows lots of variation across regions in the average quantity of emissions displaced per MWh generated or saved[1]. This is not surprising given the large differences in the generating portfolios across regions. Displacing a MWh of conventional electricity production had a relatively small impact on emissions in California where the generating mix is not very carbon intensive. The largest emissions reductions are found in the Midwest (MISO) and Mid-Atlantic (PJM). In these regions, the generating units that would be crowded out when renewables kick in are often coal-fired.

There is much less variation in avoided emissions across different resources – for example solar PV versus wind – within a region. Intuitively, this is because marginal emissions rates are fairly constant within regions across hours and across seasons. One exception is New York (NYISO), where the marginal emissions rates are significantly lower on average during high-demand hours. Solar PV resources and commercial lighting retrofits, which generate electricity/savings disproportionately during daylight hours, displace fewer emissions per MWh than wind energy or residential lighting improvements.

What does this mean for subsidizing green?

If we want to design production-based credits or subsidies to accurately reflect emissions damages avoided, these results suggest that subsides should vary significantly across regions. Variation in avoided damages across technologies within a region appears less important.

To put these estimates into some kind of perspective, we assign a dollar value to each ton of CO2 displaced, $38/ton, and compare these monetized avoided damage benefits to the average wholesale electricity market value of the renewable electricity generated. The graph below summarizes our estimates for solar and wind energy for two extreme cases: California (relatively less carbon intensive on the margin) and the Mid-Atlantic (relatively more carbon intensive generation).


Marginal value per MWh of Solar and Wind Energy Generated

The blue bars show the average wholesale market value of the electricity produced by wind and solar resources, respectively, in these two regions over this period. These values reflect the fuel and operating costs avoided at marginal sources. Electricity generated by solar PV is somewhat more valuable because solar resources are disproportionately available during high demand hours when marginal operating costs are higher.

Our estimates of avoided emissions damages, measured in terms of dollars per MWh, are shown in green. In California, these avoided emissions benefits are approximately a third as large as the wholesale market value. In PJM, monetized emissions benefits and the wholesale market value are of similar magnitude.

Smart subsidies for renewable energy

Our punch line is that the marginal value of emissions displaced per MWh of renewable energy generated has been economically significant in recent years. And these values vary significantly across regions with different generation portfolios. Of course, the quantity of emissions damages truly avoided will also depend on what other policies and programs are in play. For example, if a region imposes a binding emissions cap, an incremental increase in renewable energy will not reduce overall emissions in any meaningful sense.

These estimates of avoided emissions damages capture only one dimension of the potential benefits generated by incremental increases in renewable electricity generation. But it’s an important dimension, particularly when it comes to policies that are designed to reduce the carbon intensity of the electricity sector. From an economic perspective, these policies would ideally impose a tax on emissions calibrated to the damage caused. If instead these policies take the form of renewable energy credits, these incentives should reflect the level of – and variation in- the damages avoided.

[1] Note that if a region has imposed a binding cap on emissions, increasing renewable electricity generation may affect the way the emissions target is met, but not the level of aggregate emissions. Emissions in California were not capped during our study period.



23 thoughts on “Subsidizing renewables for the damage not done Leave a comment

  1. I’m looking at your chart titled, “Pounds of Carbon Dioxide displaced per MWh of renewable energy generated (or energy saved)” and while it makes sense that energy efficiency measures would displace GHGs at levels comparable to renewables in some regions with low renewable generation, I would expect the effect of energy efficiency on GHG emissions would be lower in California, where 23% of our power supply is from renewable sources, thus efficiency is displacing both conventional and clean energy generation. As the RPS increases, wouldn’t the GHG emission reduction benefits of efficiency decrease?

    • Hi Betony

      Thanks for reading.

      Briefly, we econometrically estimate the relationship between variation in hourly emissions within an ISO footprint and variation in hourly thermal generation. Our goal is to isolate the variation in generation that most closely mimics the effect that an incremental increase in RE or EE would have on net load. We allow these marginal emissions rate estimates to vary systematically by hour of day, type of day (e.g. load shape), season, and region. Our maintained assumption is that the marginal unit that gets backed off is not a renewable/non-emitting sources (which are not included in our hourly generation data). We then match these estimated marginal emissions rates with simulated RE generation or EE savings. To your good point, if in fact the marginal unit(s) that would respond to a reduction in net load is renewable, then our estimates are over-estimates.

      Working paper will be out very soon. We are revising in light of some great comments received in response to the post. We’ll look forward to comments you have on the paper – and methods described therein- if you have any!

      • The impact of EE on generation is an interesting illustration of how short and long term emissions reductions can differ. Looking at California, in the hourly CAISO market, the reductions are in excess of 90% gas fired. However, for mid-term contracting purposes through 2020 as the IOUs go from something about 24% renewables today to 33% renewables in 2020, the renewable share is actually closer to 36% (assuming no load growth–not a bad assumption at the moment). So the incremental emissions are in fact only 64% conventional (e.g., gas) generation. That arises because of the RPS legislative requirements. On the other hand, added renewables generally do not displace other renewables in the contract queue, so it’s probably 100% conventional resources. And this may even include reduction in baseload generation resources such as coal. There’s a strong argument that the retirement of coal plants elsewhere in WECC is being accelerated by the introduction of more renewables. (Of course, more EE might be doing the same.

  2. In looking at the values in the paper, they appear to be short-run annual emission rates. Given that renewables generally are long-term energy investments, a more appropriate metric would be the long-run emission reductions for the life of the renewable plant. Yes, this is a more complex analysis and, yes, it may not differ substantially numerically (although I have my doubts for California due to its WECC interconnections), but that doesn’t mean that one shouldn’t choose the correct methodological perspective. That may mean setting up the problem and then acknowledging the difficulties and the necessary shortcuts taken.

  3. But why look only at emissions? What not analyze and advocate for the damage not done by thermal plants? These are better than intermittent renewable sources at avoiding the damage done to people due to blackouts or higher electricity prices when you are poor and barely getting by, are they not?. What about the environmental effects of killing birds or disposing of worn out wind turbines or solar panels with all their rare metals? It seems to me that if you look at only air emissions, which obviously favor renewables, you have tipped the scales in favor of renewables.

    • The problem is one of defining and enforcing property rights. Utilities are pretty well able to measure and price energy sold to prevent personal “damages” from lacking electric service. Equity issues of high resource prices are best met by using income contributions rather than muddling price signals to achieve those goals. So neither of those points are valid, at least from an economics standpoint.

      Environmental damages arise because we are NOT able to easily define and enforce property rights to clean air, thriving species or global climate conditions, at least not yet. We have charged our governments to manage those property rights for us collectively (through majority rule–sorry, but individuals don’t have individual choices–they have to persuade others to choose differently).

      As for relative damages, many (including me) have looked at the relative damages that you list. The fact is that potential damages from climate change dwarf air quality damages, and those in turn are substantially larger than the sum of the other environmental damages that you list. There is a very large environmental economics literature behind these estimates.

  4. Good analytical framework but the RTO dispatch assumptions may be a problem that may undercut conclusions.

    The relative variations between RTOs in pounds of CO2 displaced don’t makes sense except as broad annual averages – and that’s not how generation is dispatched in either bid-based RTO markets or among vertically integrated utilities (e.g., based on fuel costs, unit marginal heat rate that varies by output, unit commitment costs, run time limits and opportunity costs of selling in other markets). Based on the current generation mix and prices, gas generation is generally on the margin in PJM now. In most hours in most RTOs (except ISONE) some portion of the marginal generation above minimum unit loading levels will be coal and some will be gas. However, in most regions, the coal/gas CC/gas CT ratios vary significantly by season and time of day. In CAISO, out of state renewable generation imports into CAISO are likely to displace coal – most of the time. Within California, penetration of wind at night and solar at mid day is radically distorting unit commitment and economic dispatch (yielding the duck curve phenomenon). This is probably increasing the effective average heat rate for the gas combustion turbine generation used to chase the net load curve, so the 900 lb range in the chart seems low.

    The more interesting modeling question is the optimal generation/conservation resource mix for the longer term, which would be quite different – and that might lead to very different second best policy prescriptions, particularly since policies tend to outlast the problems they are intended to solve.

    • Thanks for your thoughts.

      Agreed on the importance of looking beyond annual averages. It’s hard to see from the figures, and this short summary only hinted at our approach, but our approach actually involved estimating marginal emissions rates that vary by season/hour of day/type of day (i.e. different load shapes) for each region. We then matched these hourly estimates with simulated hourly production. So what those graphs are summarizing is average emissions displaced across many simulated hours.

      You and others point out that looking at emissions reductions over the life of the investment is important. I agree. But our jumping off point with this paper was thinking about production-based subsidies/credits which are ubiquitous. And thinking within the context of this policy incentive structure.

      Thanks for reading

      • I’m not sure how “our jumping off point with this paper was thinking about production-based subsidies/credits which are ubiquitous. And thinking within the context of this policy incentive structure” conflicts with using long-term metrics of emission reductions (which for example would capture deeper incremental vs marginal changes)? Is that about the amount of analysis required? Or is it that current policies focus on using short-run emission reductions?

        • The latter.

          Our maintained assumption that a sustained commitment to pricing carbon emissions, a preferable approach from an economics perspective, is out of reach (at least in the near term). So we’ve been thinking more about where we may be headed with the proposed Clean Power Plan, noting that several states are looking for ways to build on existing policy infrastructures for compliance purposes.

          In ongoing discussions about “modular” or “common elements” approaches to multi-state compliance, the tracking and crediting of emissions reductions associated with RE and EE is an important issue. Within this policy context, states are concerned with demonstrating compliance with short/medium run emissions standards/targets. Thus, it seemed to us that thinking about how to credit RE *production* (versus investment) for emissions displacement over short time scales (that align with compliance time frames) is relevant (admittedly in a constrained optimization sense).

      • So are the emission reductions calculated using the reported hourly emission values from the RTOs, which are momentary values that exclude start up and other interhour actions that affect emissions, or using modeling runs that apply incremental amounts of renewables and incorporate the changes associated with multi-hour/day operational changes by power plants?

  5. Thank you for this entry. But it raises a question. Given proper pricing of environmental damages (and some potentially heroic assumptions about natural resource availability), should one not expect equi-marginal renewable generation values?

    It seems fairly obvious that efficient pricing for the environmental damage inflicted by fossil fired generation (or the compensation to renewables for the counterfactual) would change based on the nature of the fossil generation. Coal generation imposes higher environmental costs than does a CCGT, thus renewable generation (and energy efficiency, etc) would be more valuable in regions with coal on the margin as compared to regions where gas is the marginal resource fuel. We can argue about the precise costs, but the general relationship holds.

    Where one might find disagreement is in the relationship between these cost functions and the role of public opinion on their merits. Setting aside specific cost metrics for the moment, the fact remains that the history of activities such as manufacturing, electricity generation, refining, mining, etc., is characterized by manufacturers (and by extension, consumers) getting a free input into a production (consumption) activity. This cost may represent a regional, class-based or intertemporal transfer. Or it may be socialized, similar to a group averaging the check at a restaurant. But the cost exists. Until this cost is recognized, one is left curious about the point of specific incentives.

    For a producer, not paying the actual environmental cost of production is economically identical to the producer not paying its suppliers or stealing factory equipment. Certainly the legislative, regulatory and business arena is replete with examples of pseudo-logical arguments favoring one specific approach or another. But none address the core issue, that being the incredible giveaway currently taking place.

    At least the (hopefully objective) ivory tower must stick to this basic idea. If implemented, a proper pricing system will also appropriately value renewable generation. Short of actual locational resource distinctions, one should expect a move toward equimarginality.

  6. I question the author’s premise that there is widespread support for renewables, at least insofar as that support requires the public to pay for them. I provide this anecdote as food for thought.

    At my rotary club, two weeks ago, the Director of the Armand Bayou Nature Center was our guest speaker. In his presentation, he displayed a chart of charitable giving in the USA. The largest amount of donations, 24 percent of all donations, went to religious charities. The smallest percent of donations, at less than 3 percent, went to environmental/conservation charities combined.

    I thought this was interesting because it really speaks to the public’s priorities better than any survey. Economists like to make a distinction between stated preferences and revealed preferences. To whom the public donates money is a good proxy for their revealed preference on this issue. The revealed preference is that there is very little real, i.e. financial, support, for anything to do with the environment or conservation.

    Thus, if you want to promote renewables, marketing them from an environmental consequences angle, whether actual or avoided, is not going to ever gain any traction. To convince people to change their habits, you have to appeal to their fear or their greed. The environmental angle plainly does not appeal to their fear; and policy prescriptions, which equate to higher prices, do not appeal to their greed.

  7. I think we should consider that the real benefit of solar will occur when it allows for distributed, autonomous electricity. The cost/price of underground power lines in some places approaches $15,000 per household. With Tesla offering 10 kWh batteries at $3500 wholesale, with efficiency allowing households to use fewer kWh/day and with the continuing decline in solar system costs, it will soon be cheapest for households to not have over-head power lines by having autonomous solar systems in each house in a neighborhood. In addition the removal of the overhead power lines increases the proper values of the houses in the neighborhood, as will a high quality autonomous solar electric system. These inexorable economic forces/drivers will likely within a decade make is so beneficial to “urban off-grid” on electricity that the standard electric utility model will be more or less obsolete in the sunnier parts of the U.S. Many of the calculations in this analysis assume the standard electric utility cost/business model.

    What may very well happen for utilities like PG&E, is that they may have to be in some areas a gas utility that then has an operation and service contract for the autonomous solar PV/battery systems for the customers that buy its gas supply/service. The new, very different business models will make the economic calculations for the future utility very different from the current framework of the electric utility operation and costs relatively soon.

    • I’ve done some rough calculations that suggest the only places in the US where PV plus storage make economic sense are Hawaii and Alaska, and that’s without considering the amounts of storage that are required to deliver the same level of service reliability to households they’re seeing today with minimal inconvenience. Let’s also not ignore the facts that a) the installed cost of Tesla’s batteries is closer to $7,000 for a 10 kWh system (or about the same for the 7 kWh system they offer for daily use, b) those battery packs have periodic component replacement costs, c) there needs to be spare PV and storage capacity to deal with successive days of adverse weather and the effects of seasonal variations in solar production. For my household, where consumption is roughly in line with the California average, the amortized price of electricity for PV plus storage ignoring the above is about 35 cents per kWh.

    • Very soon! Within a week or two. Comments on this blog will certainly inform our final revisions.
      Thanks for reading


  8. Two comments:

    1) Your price of avoided carbon emissions, while it corresponds to the Federal government’s estimates, is probably low. Moore and Diaz show the modest and reasonable tweaking of the assumptions in a widely accepted benefit cost model yields an avoided cost of carbon over $200/ton CO2: Moore, Frances C., and Delavane B. Diaz. 2015. “Temperature impacts on economic growth warrant stringent mitigation policy.” Nature Clim. Change. vol. 5, no. 2. 02//print. pp. 127-131. []

    2) You haven’t included the other environmental costs that are avoided from renewables, mainly criteria air pollutants. Existing coal plants in particular are very dirty, and in some parts of the country those external costs are considerable. Epstein, Paul R., Jonathan J. Buonocore, Kevin Eckerle, Michael Hendryx, Benjamin M. Stout Iii, Richard Heinberg, Richard W. Clapp, Beverly May, Nancy L. Reinhart, Melissa M. Ahern, Samir K. Doshi, and Leslie Glustrom. 2011. “Full cost accounting for the life cycle of coal.” Annals of the New York Academy of Sciences. vol. 1219, no. 1. February 17. pp. 73-98. []
    Also see Muller, Nicholas Z., Robert Mendelsohn, and William Nordhaus. 2011. “Environmental Accounting for Pollution in the United States Economy.” American Economic Review vol. 101, no. 5. August. pp. 1649–1675. []

    Both items would change the results substantially.

    • Hi Jonathan:

      Social cost of carbon is subject to debate, as you note. Some argue for a higher number, others suggest $38 is too low. We elected to go with the number recommended by the inter-agency working group. Although the paper also backs into this calculation and estimates the social cost of carbon at which technologies would be cost effective, given the costs (recent estimates of investment costs) and benefits (avoided operating costs, quantity of emissions displaced) we try to account for. Your comment suggests this exercise will be preferred by some.

      On (2). We chose to focus on GHGs for two reasons. One, according to the MD estimates generated by MMN, external costs of GHGs account for a majority of the monetized value of emissions damages per marginal MWh (again we assume that $38/ton number when valuing carbon). The other issue is that key criteria pollutants – NOx and SO2 – were capped during our study period in regions where the greatest damage caused (national acid rain program and regional NOx programs). If we think these caps were binding over this period (CAIR permit prices barely above zero), then it can be argued that renewables did not displace NOX or SO2 emissions in any meaningful sense.

      Interested in your thoughts on both points as we revise the paper….

%d bloggers like this: