We recently got an email from Heatmap saying “Half of companies with net-zero commitments call scope 3 their biggest barrier.” And we thought duh what? And then “Probably the other half said ‘What’s scope 3?’” It turns out to be the carbon emissions you don’t emit but get blamed for anyway. More fun from the wacky world of climate economics. Which has this regrettable feature in common with interventionist economics generally: They don’t know enough to know what they don’t know. The email also said “Most are stuck on spend-based estimates too rough to prove whether their reduction programs are working.” As in the old Soviet Union the planners basically planned to spend because unless you let consumer preferences form prices in competitive markets, you can’t tell whether you’re making something useful or just sawdust sausages and ill-fitting size 12 shoes all for left feet.

In a striking departure from modern journalistic practice, we responded by Googling “Scope 3”. And it turns out to be something very revealing in its unrevealingness. For instance, from the “Carbon Trust”:

“The Greenhouse Gas Protocol – which provides the most widely recognised accounting standards for greenhouse gas emissions – categorises GHG emissions into three ‘scopes’. Scope 1 covers direct emissions from owned or controlled sources. Scope 2 covers indirect emissions from the purchase and use of electricity, steam, heating and cooling. By using the energy, an organisation is indirectly responsible for the release of these GHG emissions. Scope 3 includes all other indirect emissions that occur in the upstream and downstream activities of an organisation.”

So here’s the problem. Scope 1 is reasonably easy to measure, at least in principle. It’s the GHGs you emit directly by doing whatever you’re doing, minus presumably staff breathing out CO2 while at work. Scope 2 isn’t too bad: it’s the same as Scope 1 except it’s from your power suppliers for whatever share of their output you acquire. But Scope 3 is a very different beast. It’s all the emissions from all the people you bought stuff from, plus all the emissions from all the people who bought your products. All of them. Every single one. Even your kid’s lemonade stand. And all based on pseudo-scientific “accounting and reporting standards” that sound good, make work, generate money and are no use in the real world.

Here we must again cite Leonard Read’s 1958 classic “I, Pencil”, a demonstration even from a simpler era when people used pencils not AI that tracing out all the steps involved in making things, was inherently a transcomputable and never-ending quest. Thus, the pencil tells the reader, “not a single person on the face of this earth knows how to make me.” Take any part of it: the wood, the paint, the eraser, the graphite, and you touch a global supply chain of near infinite complexity. Among a great many other things, that:

“My bit of metal – the ferrule – is brass. Think of all the persons who mine zinc and copper and those who have the skills to make shiny sheet brass from these products of nature. Those black rings on my ferrule are black nickel. What is black nickel and how is it applied? The complete story of why the center of my ferrule has no black nickel on it would take pages to explain.”

And to explain who made the coffee and shoes and razor blades for the miners, and who produced the fuel for the ships that carried these goods about, and who made the machine tools to create sheet brass, and on and on, would defy any central planner even before they got to the problem that all these people in turn used pencils (we said it was 1958). Which is why after an initial burst of idealism Soviet Five-Year Plans were essentially reduced to replicating last year’s plan and adding a few percent.

Now imagine trying to track all the carbon emissions. Because it’s what Scope 3 aims to do as if it were a mere detail.

Incidentally Soviet planners from the 1960s on into the 1980s actually thought that more powerful computers would overcome this problem, being unclear on the concept of “transcomputable” particularly as it applied to systems with feedback loops. Then the USSR done died. But not this hope.

Thus Heatmap insists that it can be done with their handy guide, which “lays out a four-step playbook for closing the gap between scope 3 commitments and measurable supply chain progress, using domain-specialized sustainability AI” and from whose three “What you’ll learn” gems the first is:

“How to move from spend-based estimates to product-level emissions data in hours – with AI-powered material mapping and multi-agent validation”

Oooh. AI-powered. Domain-specialized no less. And multi-agent validation. We suppose it beats getting it all from Agent 86. But not by much because the fantasy of measuring all the carbon outputs from all your inputs relies on the fantasy of measuring all your inputs and you can’t.

In a market economy, a real one, you don’t have to. As Read also said, as long as you know the price you must pay for the immediate inputs, how to assemble them into outputs, and the price you can get for those outputs, you can tell whether a purchase is rational. But if you have to know more than you can, you’re sunk. Especially if you don’t know you can’t.

Welcome to the wonderful world of carbon accounting. Do not mistake it for the real kind of accounting though.

Or indeed the government kind, as Lorrie Goldstein wrote in early October in the Calgary Sun that:

“We’ll never know the total cost Canadians paid to fund former prime minister Justin Trudeau’s failed climate change policies that Prime Minister Mark Carney has largely abandoned. That’s because in addition to the upfront costs, there were the lost opportunity costs caused by putting a damper on the expansion of Canada’s oil and natural gas sectors for a decade.”

Even the upfront costs are a bit of a mystery, being spread over 149 federal agencies and supplemented, contradicted or otherwise impinged on by 364 provincial and territorial programs. The downside is impossible to estimate. And so are the results, other than the incineration of over half a billion dollars of public money… not counting municipal follies. Now imagine trying to put a dollar value on the carbon emissions prevented via the bad weather supposedly prevented worldwide. It simply can’t be done, even for part of the public sector in one country. It certainly can’t be done for the entire private economy of the globe. Not with all the pencils in the world, or all the computers. Though at least with the latter you can program in the answer you want and get it back out.

Oh, in a related story, and an incomprehensible one for related reasons, Bloomberg Green warns that in addition to using way too much energy and way too much water, AI data centres use way too much water:

“Data center developers have started touting new technologies that slash the amount of water needed to cool hot-running servers in drought-stressed places like Arizona. Those efforts, though, are not addressing the vast majority of the water consumed by the AI buildout. Miles away from multiplying data centers, power plants are now consuming enormous volumes of water to generate the electricity required to run artificial intelligence, including those more efficient cooling systems.”

What, are they creating hydro dams? The story, amazingly, doesn’t say, while thundering that:

“New research indicates that such indirect water consumption is rapidly growing. In a recent paper, University of Michigan researchers estimated water used offsite would boost the total annual consumption of three large AI data centers planned for the state by 14 billion gallons.”

Following the links, we eventually learned from a different branch of Bloomberg that everything has costs:

“Generating the electricity used by data centers consumed another 211 billion gallons in 2023 – about 12 times the amount data centers consumed directly. That’s because power plants fueled by fossil fuels, much like data centers, require water for cooling. Nuclear, geothermal and hydropower sources of electricity also consume varying amounts of water; solar and wind sources use little or none.”

It seems the water evaporates and then leaves the solar system or something. (That something being falls back to Earth and reenters the hydrological cycle.) But why stop there?

Don’t data centre workers drink water? And then… well, never mind because it’s disgusting. And don’t the people who sell them food drink water? And mix it with antifreeze in their radiators? And water their lawns? There’s no telling where it might stop. Or whether.

P.S. If carbon accounting did work they might not like it. Tammy Nemeth recently chortled “Imagine my surprise this morning when I read “climate correspondent” Chris Hatch’s Zero Carbon newsletter grouching that emissions from Canada’s oilsands are ‘rounding errors’ in the grander scheme of global emissions.” And what if people realized their own emissions didn’t make a ding dang bit of difference even if human CO2 were causing bad weather? There’d be a run on the carbon bank.