Adam Stein of TerraPass (a major offset provider) argues in favor of offsets on the basis of additionality. (An offset is a payment by a polluter to somebody else to reduce their emissions so the polluter does not have to. Additionality is the claim that we know how much pollution would have occurred without the payment.) It’s very generous of him to contribute to a discussion which will probably lead to a drop in sales of his company’s product.
The key points in this argument are that a) measures of additionality are imprecise, and b) that fact matters a great deal more than the same imprecision in other methods of putting a price on carbon.
Let’s examine the first point — that additionality cannot be measured with much certainty or precision.
The fundamental problem is that additionality in a project setting has to measure against “what might have been.” A consultant develops a “business as usual” (BAU) scenario, declares that it is what will happen without the project, and calculates any savings over this as an emissions reduction.
“Wait a minute”, you might say, “that doesn’t sound so hard. Look at what is happening now, and take that as the BAU case.” Unfortunately, that is not how typical economics works. Typical resource use curves show resources being used ever more efficiently, but in larger quantities. In other words, you produce more dollars worth of output from a barrel of oil, but that makes it worthwhile to use more oil, not less. So BAU scenarios should actually include a significant number of improvements in resource efficiency use.
Similarly, offset providers, at least the best ones, look at whether offsets are super-regulatory — that is, whether they go beyond what regulations require. (There are offset providers out there who don’t, but Adam’s TerraPass is not one of them.)
The problem here is twofold. One is that regulations often change. Another is that political pressure, rather than regulation, may sometimes shut down emission sources. Adam offers super-regulatory methane flaring as an example of robust and measurable additionality:
…it might be possible, for example, to calculate the amount of super-regulatory landfill methane flaring that would occur in the absence of carbon offsets (none).
In response to arguments about other ways methane flaring might end, he added, “carbon offsets are the only value stream for landfill gas flaring projects. This is as close to 100% additionality as one could hope for.”
I’ll agree that it is as close to 100% as one could hope for — not very!
Let me quote Sajida Khan via Carbon Trade Watch:
In 1980, Bisasar Road Landfill in the Indian suburb of Clare Estate officially opened its gates to rubbish-dumping trucks. I was just three years old at the time and living in the suburb next door. During the course of my entire childhood, the Bisasar Road landfill was a regular topic of discussion, as my mother and I made trips to visit my grandmother nearby.
Clare Estate was the bridge between our familial residences. I vividly remember the preparations, as we hit that short stretch. Car windows had to be rolled up. Nostrils had to be squeezed tight with tiny, pincer-gripped fingers. Breaths needed to be held. The stench was reminiscent of my public school toilet on a really hot Durban day.
A few years after the opening of the dump at the site of that quarry, I remember my mother excitedly telling me that the Bisaser Road facility would be shut and transformed into a park for the community. As a child whose life was spent riding a bike around our tar-covered parking lot the idea of a park in our vicinity was just too thrilling.
The article goes on to tell how strongly the local community fought for shutting down the dump, and how a wealthy white suburb nearby won closure of their dump:
Public reaction was swift, as people blocked the site entrance of the dump, held demonstrations and marches, and circulated a petition to council that gained 6000 signatures. But nothing worked, so Khan decided to take legal action on behalf of the residents and schools.
As the battle raged, a wealthy white-dominated suburb to the north of Durban was quickly closing its landfill. Umhlanga, situated at the shore’s edge and expanding into rolling sugarcane-covered hills, was ‘earmarked for up-market property development,’ according to Bryan Ashe.
If you are measuring additionality of the Bisasar energy project, it seems like the chance that it might have shut down without the subsidy from CDM is not such a reach after all. In fact, there is a good argument that shutdown should be the baseline scenario.
And this brings us to another problem with the offset idea. Adam and I can go back and forth on the additionality of a particular offset. The problem is, there will never be adequate data as to who is right and who is wrong. Adam could argue that the dump would have continued to operate, even without offset subsidies, but without methane flaring. I could argue that South Africa has shut down other dumps, and that the community might have won on this one as well. Does anyone think this can be settled with reasonable certainty?
So we don’t even have the feedback of being able to go back later and say, “boy did they get the additionality wrong on that one,” or alternatively, “the project was wrong for all sorts of reasons, but they were really close on the emissions reduction calculations.” We can advance reasons other than emissions reductions why it should never have been done, but in terms of whether it reduced emissions compared to what would have happened otherwise — we won’t ever know. Remember, this is a type of project a representative of TerraPass described as “as close to 100% additionality as one could hope for”.
Call this anecdotal; but remember the TerraPass rep chose the anecdote, and examine the point it illustrates.
Project-based offsets by nature depend on counterfactual BAUs. BAU scenarios are generally linear extrapolations, which are almost always wrong. (A linear extrapolation looks at the present and says “the future will be the same, only more so.” Or as the song lyric goes, “that’s just the way it is: some things will never change.”) But that is not the main problem. The main problem is lack of feedback.
Normally, when you make a prediction, you can look back later and see if it was right or wrong. But since an emissions-reduction project is a difference between what actually happens and a counter-factual, you can’t tell after the project is complete if the counter-factual is true any better than you could before the project was proposed. You may get real-world feedback on the numerator, but never on the denominator.
OK, but as one commentator mentioned, isn’t this just a male obsession with numbers and precision? If you do a lot of projects, aren’t you going to reduce emissions? Does it really matter whether you know by how much?
The problem is that in a trading system, you really do need a reasonably accurate emissions reduction calculation. If the credits are sold within Kyoto, or another system that regulates carbon, they become legal permissions to emit greenhouse gases. Outside of such systems, they become a means to emit greenhouse gases while still claiming carbon neutrality. Highly inaccurate numbers increase emissions within carbon trading systems and reduce offset purchases outside them to a PR stunt or self-deception.
Further, the project focus, issuing credits on a project-by-project basis, increases this inaccuracy.
A good informational analogy is the difference between measuring out a cup of milk with a measuring cup, and measuring out a cup of milk by pouring it into a teaspoon 48 times. For the latter, your accuracy is going to be a lot lower and your transaction costs will be a lot higher. A standard carbon tax is a rough and ready measure that can function well on the rough and ready level of information we actually have about carbon emissions. A carbon trading system, especially one with offsets, requires much more detailed information — more detailed than we have. It is a classic error every engineer has encountered: its precision is greater than its accuracy.
So an emission trading system requires more accuracy than a carbon tax system in order to correct errors, but by its nature will generate less accurate information. That is a serious informational double whammy!
If an enforced carbon tax produces less reduction than you expect, you increase the tax — politically tough, but technically simple. (I think you need old-fashioned rule-based regulation and public initiatives as well, but set that aside for the moment.) If you had already planned an increase as part of a step by step process, you increase it more than you had planned.
When a similar failure occurs under an enforced carbon trading regime, you know part of the problem has to be invalid carbon credits. Tightening the cap does not solve that problem. You have to track down (at least as a class) where invalid credits were issued. The precision problem with BAUs makes it difficult to track this down with offsets. Note that getting the sign right is not sufficient precision! You need a good approximation of how much you have reduced emissions on a project by project basis.
Lets see if we can move beyond the abstract, to examine BAU in practice. Consider the Kyoto treaty for a moment.
Kyoto is currently projected to fall short of its goal even within EU nations that actively work to comply.
Let’s focus on one part of this troubled system, the Clean Development Mechanism (CDM). The idea of CDM is that a polluter in a nation that has committed to reducing emissions can finance a project in a nation that has not, and thus reduce total world emissions at lower prices while boosting clean development in poor nations. The “development” is a key object of CDM. For the CDM to be considered successful in meeting that criteria, most emissions reduction should occur via efficiency improvements and renewable energy — things that actually transfer technology and increase clean capital stock in the poor nations. Another measure is the extent to which the carbon market encourages investment of CDM funds where they will produce the most reduction for the money. A third measure is how much many carbon credits are actually produced. (Note that this is not necessarily the same as carbon emissions reduced; that is what we are arguing about.)
Projections were made about all three of these. Projections for the first two were qualitative — CDM dollars would flow towards clean infrastructure development, and do so at the lowest possible price. In terms of the third, many actors made estimates of actual credit expected to be generated.
If we look at task difficulty we would expect the first two to be the most accurate, since they merely make qualitative guesses, where the third is quantitative. But if we look at incentives, we would expect the third to be the most accurate. There is a strong incentive with a carbon market to produce as many credits as possible. We would expect the second to less precise, because while keeping costs down is one way to be profitable, providing the most niceness for the nickel is not necessarily the best way to make the highest profits. The first criterion is where we would expect the least accuracy: there is almost nothing in practice within the CDM market that favors clean infrastructure over other kinds of investments. In winning approval for CDM, though, there was a strong incentive to sell it as a means of financing popular technologies such as renewable energy.
If we look at actual results, they line up with these expectations — people guessed wrong where they had incentives to guess wrong, somewhat right in the absence of such incentives.
In terms of clean energy, you will sometimes see misleading statistics that 60% or 70% of projects in the pipeline involve renewable energy or efficiency. But the proper way to measure this is what percent of claimed emissions improvements come from this source, and if you look at the Stanford paper I linked above, published in June 2006, you will see that fewer than a third of emissions reductions by volume are projected to be from energy related projects — including renewable energy, efficiency improvements, and cement projects. Only around 18% of reductions are projected to be from renewable energy, including hydroelectric projects. If you look at latest official CDM pipeline spreadsheet, updated in early February, you will find that omitting rejected or withdrawn projects, renewables and efficiency still represent about 36% of total emissions credits to be generated through 2012. Note that this last figure includes many projects in early stages of registration or validation — so not all will be approved.
In terms of the second, projects have often paid far more than the actual cost of the emission reductions — in spite of the fact that CDM emission reductions are far less expensive than similar reductions in rich nations. How can you overpay for cheap reductions? The problem here is that non-fossil-fuel gases (such as HFC-23) are traded against CO2. You can expensively reduce HFC-23 (which is a byproduct of HFC-22) emission more cheaply than you can inexpensively reduce CO2 emission. But since you will ultimately have to do both, overpaying for HFC-23 reductions ultimately wastes money that could have paid for additional reductions.
Worse, such overpayment artificially inflated HFC-23 production specifically to earn CDM money. The problem here:
The economics of HFC-23 projects create incentives for strategic behavior that if left unchecked, undermine the environmental efficacy of the CDM. Consider the 1 kg of HCFC-22 produced by a CDM project that the calculation above showed to be equivalent to 0.35 t CO2e or 0.35 CERs. At current market prices of â‚¬9/CER,101 the production of 1 kg of HCFC-22 will produce a subsidy of â‚¬3.15. The cost of HFC-23 abatement is estimated to be on the order of â‚¬0.09/kg HCFC-22 (see Box 1)102 Thus the net from subsidy minus abatement costs to an HCFC-22 producer is approximately â‚¬3.08/kg HCFC-22. This subsidy compares quite favorably with the wholesale price for HCFC-22,[…] â‚¬1.60/kg.
Even with these highly restrictive rules on eligibility, there is relatively strong evidence that HCFC-22 producers participating in the CDM have behaved strategically to direct a greater share of the subsidy to themselves by artificially inflating their base year production in two ways. First, the fraction of HFC-23 produced by the production of CFC-22 can be reduced by modification of the conditions under which chemical synthesis occurs. Dupont has been able to consistently produce, in its United States based HCFC-22 plant, HFC-23 byproduct percentages as low as 1.3%. The economics of HCFC-22 production in the absence of a CDM subsidy dictate that HFC-23 production be minimized because it is in effect a waste product costing both energy and materials. For this reason, almost all plants have historically monitored the HFC-23/HCFC-22 ratio in their production. The CDM methodology eventually approved for HFC-23 abatement set 3% as the maximum percentage of HFC-23 byproduct allowable in the baseline data of a participating plant. The average of all reported baseline data at the 11 participating plants is 2.99% – very close to the maximum allowable value. This suggests that even if the project participants were not actually aiming for the 3% sweet spot that would minimize their production costs (due to wasted feedstocks) but maximize their CDM subsidy (due to more CERs for a given production rate of HCFC-22, they were certainly not as concerned with minimizing this percentage as developed world producers, not eligible for the CDM subsidy, seem to be.
In addition, at least some of the HCFC-22 plants participating in the CDM appear to have ramped up production during the baseline period (2000-2004) far beyond the expected growth in the sector (15%). Figure 5 shows the baseline data supplied by plants participating in the program compared with the predicted growth rate for the industry over the 2002-2004 period. Most plants exceeded the growth rates predicted for the developing world industry as a whole. These increases in HCFC-22 production amongst the 75% of developing world producers participating in the CDM led to a CDM participant production growth rate of 50% rather than 33%, as had been predicted ex-ante by market analysts.118 Whether or not these plants increased production due to demand for HCFC-22 or in anticipation of higher CER revenue is impossible to say given publicly available information. Nevertheless, a circumstantial case exists that at the least, rather than building new plants, HCFC-22 producers elected to add capacity at existing plants during the CDM baseline period in order to take advantage of the CDM subsidy.
Remember, every one of the credits so generated allowed a unit fossil fuel to be burned in the rich world. It is a real and serious problem. Getting prices wrong, even when providing what appears to be a positive incentive, can have serious real-world consequences.
In terms of the third example, projected credits vs. actual, we don’t have figures for all CDM. But there was a partial OECD study that made predictions for a substantial percent of the market. They came up 15% short — which is fair enough. 85% accuracy is certainly more than reasonable in predicting the growth rates of a new and volatile market, though it would be pretty awful in more mature markets. Then again, there was no strong incentive to be overoptimistic.
Again, we can’t check “Business as Usual” estimates of what “would have been” directly. But what do you think BAU resembles in terms of incentives: the first two cases, where there are strong rewards for estimating in a particular direction, and which went extremely wrong? Or the last case, which lacks such incentives and produced a reasonable estimate? Especially since, unlike any of the three cases, it is somewhere between difficult and impossible to tell what would have happened after the fact.
Any system that reduces emissions has two possible points of failure. One is excessively weak requirements. (This can take the form of too many permits in an auctioned permit system, too low a tax rate under a carbon tax, or weak rules in a quantity-based system of regulation.) Offset trading adds a third failure point — excessively inaccurate BAU estimates, something extremely hard to detect and (short of deliberate fraud) impossible to detect with any greater certainty after the project is complete than before it begins.