Carbon accounting: philosophical investigations?

This short essay argues current carbon accounting frameworks may not be entirely helpful. Metrics like “Scope 1”, “Scope 2” and “Scope 3” only make sense for a small subset of companies, but they lack applicability, comparability, completeness, and reliability across the broad landscape of global emissions. We prefer granular carbon accounting which can be compared to counterfactuals. Our argument is illustrated with examples; and based on the philosophy of language, from Wittgenstein through to the present day.


Philosophy of language: an overview?

Many moons ago, I was awarded a degree in Philosophy and Neuroscience at the University of Oxford. It turns out that the philosophy of language is surprisingly helpful for carbon accounting, as explored in this short research note…

At the turn of the twentieth century came the logical positivists. They sought to codify language into symbols and logical operations. For example, to say that “my car emits CO2” can be analyzed as a proposition (P), and broken down into the words “my car” which specially references a particular hunk of metal sitting in the driveway, the property of “emitting” (which in this case applies to the car) and a type or class called “CO2” which represents all molecules with the chemical structure O=C=O. My car runs on gasoline and so P is true. All language, and all of the problems of philosophy, the logical positivists claimed, could be analyzed as logical statements like this, which will be true or false depending on the state of affairs in the world. Some went so far as to claim that these kinds of logical statements were all there was to language. 

Then came Wittgenstein (1889-1951), still to my mind the greatest philosopher of all time, who shattered logical positivism, and philosophy more broadly. To Wittgenstein, the strict rules proposed by logical positivists have nothing to do with how language is actually spoken and understood by people in the real world. Language is no more than an imprecise set of customs. These customs work because users of the language are generally like-minded, and tend to interpret utterances in similar ways. In a few circumscribed cases, utterances in a language might have a logical structure like “my car emits CO2”. But language is complex, and many other meaningful utterances totally defy this framework, like answering a question with a thumbs up. Other utterances in language appear to make sense, such as “if it’s 6am in New York, what time is it on the sun?”. But on closer inspection, all that is shown by these kinds of philosophical problems is the boundaries of our linguistic customs. No philosophical questions have proper answers.

Then came modern philosophy, reeling, and trying to justify its existence in the aftermath of Wittgenstein. Yes, many questions of philosophy are at risk of being unanswerable. But philosophy can nevertheless provide some useful ways of thinking about things. One of my favorite  writers was David Lewis (1941-2001), famous for “possible world semantics”, where language can be analyzed by considering the world around us, but also other hypothetical worlds. “If I switch my car to an EV, my CO2 emissions will fall” means that in the vast majority of possible worlds where I switch my car to an EV, my CO2 emissions fall. Then we can start considering what these possible worlds look like (including worlds where I am charging my EV from the grid, or from a diesel generator, or even the broader range of possible worlds where I sell my car altogether and take public transport). 

If you have made it this far, thank you for not giving up already. We argue below that traditional carbon accounting frameworks are like logical positivism. They are subject to holes and philosophical muddlings that could bring Wittgenstein back from the grave. But useful carbon accounting can be salvaged, especially via a “possible worlds” approach.

CO2 accounting frameworks: logical positivism revived?

The logical positivism of the CO2 accounting world is the Greenhouse Gas Protocol. It starts with Scope 1 emissions, which covers all of the CO2 emitted directly by a company, such as the fuel combusted at company facilities. It then adds Scope 2 emissions, which cover the CO2 embedded in the heat and electricity purchased by the company. Finally, it adds Scope 3 emissions, which covers all the CO2 indirectly emitted through the use of the company’s products.

CO2 accountants, like logical positivists before them, invariably seem to believe that all of the challenges of CO2 accounting can be solved by strong-arming companies into hiring CO2 accountants; moreover that climate progress can best be achieved in the form of corporate promises to reduce these emissions categories; and most brazenly, that CO2 accounting simply comes down to Scope 1, Scope 2 and Scope 3.

That’s not how companies work, any more than logical positivism is how language works. It is tempting to conceptualize ExxonMobil as a single giant furnace in Texas, and then ask the company why it can’t simply tell us how much CO2 is being emitted out of its furnace. But the reality is a Super-Major with a presence in over 200 countries globally, and many activities in each country. In many of these, it is not the owner or operator of assets, but a minority shareholder, in a joint venture operated by a third-party (which may or may not even relay the requisite data back to other JV partners). In many others, it does not do its own drilling, construction or maintenance, but hires third-party contractors (e.g, oil service companies).

That’s not how financial accounting works either. Back when I was a sell-side analyst, I remember downgrading another Oil Major to a ‘Sell’ because I discovered they had over $15bn of debt “off their balance sheet”, hidden away from their gearing calculations. As the company correctly pointed out, IFRS accounting rules do not require them to consolidate the debt in these entities, as they are minority shareholders. Only their share of income must be consolidated. Obviously, the financials of contractors and service providers are not consolidated either. And so it is with CO2 disclosures that are published by companies today. These disclosures are usually limited to consolidated assets, or even more narrowly, to consolidated AND operated assets.

Limiting CO2 disclosures to a subset of assets, however, limits the comparability of these disclosures. Companies with concentrated portfolios (few assets, large controlling stakes, operations performed in-house) will appear to have more CO2 emissions than otherwise identical companies with sparse portfolios (many assets, non-controlling stakes, operations performed by contractors).

Scope 1 & 2 emissions can also be “gamed” by giving up control of an asset, such as selling a corporate headquarters then leasing it back (it is no longer “our asset”); or by outsourcing activities that used to be performed in house (it is no longer “our maintenance crew”).

Some industries also defy measurement via the Scope 1, 2 and  3 framework. Also back when I was a sell-side analyst, I would be sent on punishing ‘marketing trips’, where over the course of ten days, I would participate in 70 meetings, in 10 different cities, going Coast to Coast in the United States. (By the end, you didn’t know which way was up or down!). But I once tried adding up the CO2 emissions from one of these trips. The calculations became pretty hairy. I would be trying to figure out what model plane or taxi I was travelling on, how to allocate its emissions between myself and my co-passengers, and whether to adjust for nuances such as business class seating. I think these are genuine challenges for carbon accounting, and they were part of my CO2 footprint as an analyst. Whereas the Greenhouse Gas Protocols effectively ignore them. Under their frameworks, these are simply the Scope 1 emissions of transportation companies, from airlines (who report them) to individual Uber drivers (who do not).

As a general rule, the smaller the entity, the less likely they will have the resources to calculate their Scope 1, 2 and 3 emissions. (I may calculate a look-through of all our CO2 emissions on behalf of the West household, but I suspect I am somewhat unusual!). Are we thereby saying that only big corporations need to reduce their emissions, while smaller corporations, and the 8bn people in the world are somehow exempted?

Scope 3 emissions can be a totally meaningless concept. Imagine you are that airline or Uber driver, trying to quantify how your service has indirectly contributed to a customer’s CO2 footprint? You have delivered me to my destination. How are you supposed to know if I am in town to discuss energy stocks in an office building (low carbon) or set fire to the local library (high carbon). If the latter were the case, then the law would not hold you responsible, but the carbon accountants might! For a more real example, consider the Tesla photographed below, towing a diesel generator, which if used to charge the Tesla would result in 30% higher emissions per mile than simply driving an ICE car (data here).

Scope 2 emissions can also be meaningless and hard to measure. For example, renewable energy credits are legal contracts where all parties agree to pretend that gas and coal electrons are wind and solar electrons and vice versa. As per our recent research, these RECs can create some very strange implications for carbon accounting (below).

Scope 1 emissions are also debatable, for example, where Scope 1 emissions sources are only inferred, not measured directly. For example, we estimate that 2% of all methane is leaked across the entire global value chain, from producer to consumer, but if a company genuinely knew whenever it was unintentionally leaking methane, it would be a lot easier to fix the leaks (note below, including some interesting new technologies to help).

Other important emissions categories are not captured at all in this framework. For example, our research also shows that almost two-thirds of CO2 emissions are embodied in materials and products (data below), not in purchased fuel and electricity. But the emissions associated with purchasing raw materials do not appear to be captured by Scope 1, 2 or 3 categories.

The destruction of nature, or CO2 fluxes caused by deforestation or environmental degradation do not seem to be captured either. Nor is there an elegant solution in these carbon accounting frameworks to capture the positive impacts from nature-based solutions that are sponsored by companies (do they offset Scope 1, 2 or 3, or go in their own separate bucket?).

Most pressingly, from our own CO2-accounting models, it is not even particularly helpful to know the absolute number of Scope 1 or 2 emissions, unless you have a “per unit” metric, on which to compare different companies. In some industries, these “units” are inherently apples and oranges (e.g., how do you compare the CO2 per iPhone to the CO2 per hamburger to determine whether Apple is a more sustainable company than McDonalds).

Even when competitors in an industry are producing highly similar products, care is warranted. We have seen this, for example, in our screen of refiners below. Some of the apparent laggards are simply producing cleaner fuels (e.g., to meet California fuel standards), or they are also co-producing petrochemicals, a fundamentally different product. Hence our analysis has been very careful to adjust for these effects, before declaring true “leaders and laggards”.

Is carbon accounting a useless endeavor: homage to Wittgenstein?

To re-iterate, carbon account frameworks seem to make sense, in a very circumscribed set of contexts, especially where a company operates a single, large asset, like a blast furnace, and we can measure the Scope 1, 2 and 3 emissions of this operation. This is a little bit reminiscent of how logical positivism seems to make sense, in a very circumscribed set of contexts, such as very simple propositions in a language.

There are also many examples of companies that seem to defy meaningful measurement via Scope 1, 2 and 3; categories of emissions that are altogether missed; reductions in Scope 1&2 emissions that do not actually have any bearing on CO2 emissions; and on closer inspection, the concepts of Scope 1, 2 and 3 are a little bit fuzzy.

To say that Scope 1, 2 and 3 emissions conceptually capture everything that needs to be captured in carbon accounting is like saying that all of language boils down to logical operations. This might all make carbon accounting questions seem unanswerable.

Our own approach to carbon accounting: homage to Lewis?

Carbon accounting is not useless. It simply needs to be done right. And this involves a mixture of data-crunching and philosophy. Below are a dozen principles we have found helpful, illustrated with examples, where possible. Our conclusion is that companies that want to promote climate progress should publish granular, asset-level emissions data, and strive to reduce the emissions of these underlying processes; rather than focusing on headline data at the corporate level.

(1) Companies do not emit CO2, processes do. To re-iterate, ExxonMobil is not a single giant furnace in the center of Texas.  It is a company that owns interests in various assets. Each asset uses various processes to make various products. It is these processes that cause CO2-equivalent fluxes and need to be decarbonized in the most cost-effective way, regardless of whether the process is undertaken in-house or by a contractor. The same goes for any company. Carbon accounting would be vastly better if numbers were published at the asset- or process-level, not the company level.

(2) CO2-equivalent fluxes should be the primary unit of account, not just emissions. Climate change is caused by the accumulation of greenhouse gases in the atmosphere.  Hence accounting for climate impacts should ideally capture the motion of all greenhouse gases into, and out of, the atmosphere (model below). These fluxes should all be converted into the common currency of CO2-equivalents (1 ton of methane is equivalent to 25 tons of CO2 and 1 ton of N2O is 298 tons of CO2).

(3) Processes can be combined like building blocks. For example, the total CO2 intensity of a gallon of gasoline can be calculated from the CO2 intensity of development, production, transport, refining and marketing (data below). The CO2 intensity of a mile of vehicle travel can then be calculated from the CO2 intensity of a gallon of gasoline, plus the process of manufacturing the vehicle and the process of ultimately scrapping the car. This allows our build-ups to assess entire value chains, and not just the companies in specific portions of the value chain.

(4) CO2 considerations are usually relative. It is tempting to dichotomize the world into “carbon emitting” and “non-carbon emitting” categories. The reality is a spectrum (below).

(5) Relativity requires a baseline. I like to think about the counterfactual to a process as the closest possible world(s) in which that process does not occur. I.e., the world(s) that are as similar as possible to the actual world, but with the one process in question changing. For a simple example, when we talk about the CO2 emissions of driving a mile, we usually mean relative to the counterfactual of not driving that mile. When we talk about the CO2 of Factory A’s product, the counterfactual might be factory B’s product. No process on Planet Earth occurs in a vacuum.

(6) Counterfactuals matter. All of this discussion of counterfactuals might seem academic. Here is an example for why they matter. Planting a biofuel crop has a totally different CO2 flux if the alternatives are (a) leaving that agricultural land fallow (b) tearing down a rainforest. Where these counterfactuals get really useful, is in considering the relative CO2 fluxes that can be achieved with 1MWH of renewables (displacing coal power is best, conversion to hydrogen is materially worse) or 1kg of waste biomass (deep burial or biochar are superior to biofuels, note below).

(7) Some processes are lower-carbon. Specifically, this means that the process results in a smaller flux of CO2 into the atmosphere, relative to the closest possible worlds in which the process is not undertaken. As an example, green plastics are generally lower-carbon than fossil plastic, (but still materially higher-carbon than not using plastic at all).

(8) Some processes are carbon-negative. Specifically, this means that the process results in a net flux of CO2 out of the atmosphere, into some other sink, relative to the closest possible worlds in which the process is not undertaken. For example, an acre of seaweed cultivation project results in 2 net tons of CO2-equivalent being sequestered in the deep ocean every year, relative to the alternative of not undertaking the seaweed cultivation project (note below).

(9) Some processes are carbon-neutral. Specifically, this means that the process results in zero net flux of CO2 into or out of the atmosphere, relative to the closest possible worlds where the process is not undertaken. This definition matters, because honestly, we see many projects claiming to be “carbon neutral” which are nowhere close on our numbers.

(10) If in doubt, build a ledger.  The best way I have found to calculate these carbon fluxes is to put two columns side-by-side in a spreadsheet, one reflecting the process, the other reflecting the counterfactual, and then go line-by-line, sub-process-by-sub-process. For a recent example, please see our data-file into biochar below.

(11) Counterfactuals will sometimes be debatable, but these are good debates to be having. By definition, a counterfactual is a fiction, a scenario that has not happened. Sometimes it is appropriate to consider a range of counterfactuals rather than a single counterfactual.  

(12) Accounting for nature-based solutions is no more complicated than accounting for any other process.  Some commentators criticize that carbon accounting is much more complicated for nature-based solutions. In our experience, it is no more complicated than the carbon accounting for any other carbon-reduction technology.

Please treat TSE data-files as building blocks. We now have over 500 research notes, data-files and models on our website. We have around 35 separate CO2 screens linked here. Often, doing the CO2 accounting for a large project is a function of combining the data from many of these pre-existing data-files. If we can help you with CO2-accounting please let us know.

Conclusions: what should companies do?

If companies want to help decision-makers understand their CO2 intensities, in our view, they should go a long way beyond blanket, company-level disclosures of Scope-1, and Scope-2 emissions. Instead, give us the underlying data: i.e., an asset-by-asset list of each facility, its output, its combustion emissions, methane emissions, electricity purchases, estimated CO2 intensity per unit of electricity, and other relevant data. This would make for vastly more meaningful comparisons.