Economic climate models are elegant, brilliant and, in the US, failing us

Imagine your own child. Perhaps she's nine this year. In 2030 she will be 21, just starting out on her own. Now: how much less do you value the quality her life two decades from now over that of your own life, today? Let's add some precision to that question: in percentage points, how much less important to you now is each new year in the future? Does importance drop by 2.5 percent per year? Or by 3 or 5 per cent? Or are you a narcissistic hedonist, a terrible parent, at 7 per cent?

Last week William Nordhaus won the Nobel prize for a genuinely magnificent achievement: an integrated assessment model of climate and growth. There are many of these models now. They place a layer of macroeconomics over atmospheric science, and produce dollar value on the damages people will have to pay for when a ton of carbon dioxide hits atmosphere. Mr. Nordhaus has refined his own model over a quarter century, and is open about the limits to certainty in any model.

The federal government of the United States does not share his caution. US law requires, in fact, that any response to climate change must be justified by an integrated assessment model. The US uses these models to launder its own choices, to pretend that decisions on climate will have outcomes that are:

• precisely measurable,
• certain, and
• purely economic.

But as with decisions on any regulation, our options on climate change are in fact:

• not measurable with any precision,
• uncertain, and
• profoundly moral.

The answer to that question about how much less you value your daughter than yourself, for example — that's in the model. It's called a "discount rate," and depending on who's modelling, the rate does in fact run between 2.5 and 7 per cent a year. If you want to lie to yourself, hide it in a model.

William Nordhaus published his first work on climate change (paywall) in February of 1977. "Unlike many of the wolf cries," he wrote, "this one, in my opinion, should be taken very seriously." (February of 1977!) The next year, Jimmy Carter signed Executive Order #12044, instructing that federal regulations should be clear, reflect public comments, consider "direct and indirect effects," and choose the "least burdensome of the acceptable alternatives."

Cost-benefit analysis had been a part of federal policy-making since 1902, when a law directed the Army Corps of Engineers to weigh the cost of a new harbour against the value of commerce. As public projects grew, and corporate management grew up, mid-century was a hopeful time for measuring and predicting. After Carter's order, regulators at least had to consider regulatory costs. Ronald Reagan signed an order to make cost-benefit analysis mandatory for major regulations. Bill Clinton refined it in 1993 with #12866, an executive order that still stands, amended by Presidents Bush and Obama.

The next year, Mr Nordhaus published his first integrated assessment model of the climate and economic growth. Again, it produces a number: the social cost of carbon, the dollar value of the damage on Earth of a ton of carbon dioxide in the air. In Europe and China, the response has been — painfully, slowly, too late — to explicitly build that cost into markets to discourage carbon-based fuels.

In the US, integrated assessment models became policy in a different way. The Environmental Protection Agency did not build a market to charge for the social cost of carbon. But: looking for a way to answer four decades of executive orders on cost-benefit analysis, it borrowed Mr Nordhaus's work, among others, to model for regulatory changes on coal-fired plants. The federal government uses models to analyse other policies — tax legislation, for example. But by elevating cost-benefit analysis — models — into a mandatory part of how it makes laws, the US has accidentally created the perfect way to evade and veil what should be moral choices.

Every regulation, every law, takes something from someone and gives it to someone else. Someone has to stop doing something. Someone else is healthier, or financially more secure. Humans built representative democracies as the least-bad way of weighing these choices. The logic of cost-benefit analysis is this: presented with serious math, serious people will make serious choices. That's not what's happening, though.

In the US, presented with math, we argue the math — we game the models to avoid hard choices. It started more than a decade ago with ten-year macroeconomic models for tax laws. It's happening now with integrated assessment models.

Let's go back to the discount rate — that awful choice over how much less you value your daughter's happiness in the future than your own happiness now. It's an idea borrowed from finance, where it makes sense. A dollar a year from now is worth less than a dollar now; you can enjoy the dollar in hand, or put it in Treasuries for a safe return of about 3 per cent. But it's hard to understand how there's a discount function for what happens if, say, the Gulf Stream stops pumping warm water into the North Atlantic.

Integrated assessment models assume that paying now to avoid devastating and irreversible changes to global weather patterns is the same thing as making a financial investment with a long time horizon. They're clearly not the same thing, though. If an investment doesn't pay off, it's literally not the end of the world. To treat them as the same thing is already inherently a moral decision.

And alone the discount function creates the single largest difference among models. Barack Obama's EPA offered results from a range of discounts: 2.5, 3 and 5 per cent. Donald Trump's EPA discounts at 3 and 7 per cent. The agency offers a range of discounts, because, according to an EPA fact sheet from 2016, "no consensus exists on the appropriate rate to use for analyses spanning multiple generations." We could perhaps agree on a discount for our own futures, but to agree on a discount for our own children is impossible.

Or absurd.

As you can see below, compounded through 2030, a 7 per cent yearly discount rate reduces the present value of the future social cost of carbon to essentially zero. This is a just mathy way of saying "we don't care."

Again: this chart illustrates what the dollar value of climate damage looks like when we change the compounded percentage by which we choose to value our children less than ourselves.

Let's say we come to an agreement on discounting, though. We're still allowing ourselves to look at a set of numbers that imply a certainty we don't have. To come up with useful values for the social cost of carbon, a working group in the Obama administration started with three different integrated assessment models. Recognising that there are some things about both climate outcomes and economic consequences that we just don't know, they ran each model through what's called a Monte Carlo method — yes, named after the casino in Monaco — where they randomly change the value of inputs, then examine the distribution of results. Then, they averaged this distribution across all three models.

Then, the administration did something interesting. They chose these averages to represent the social cost of carbon. But they also published what they call the "high impact" costs — the costs of the damage from of a ton of carbon dioxide in the 5 per cent of worst-case scenarios that came out of the simulations. As you can see, by 2030 these costs are three times the averages.

It's good and useful that the Obama administration's working group at least published the high-impact scenarios. But by choosing the averages to represent the social cost of carbon, the group already made a massively important moral and political choice. They decided before they even went public that the US might consider paying to avoid the most likely scenario, but not the worst one.

Here's how the Committee for the Prize in Economic Sciences in Memory of Alfred Nobel put it, as they praised Mr Nordhaus's work (emphasis ours):

Given the large uncertainties about future climates, thinking about appropriate policies involves – explicitly or implicitly – taking a stance on risk and uncertainty. Likewise, any policy considerations involve taking a stance on discounting. Since the effects of carbon emissions are much more long-lived than humans, it becomes critical to value the welfare of future generations. On both accounts, moral values may be necessary to complement scientific measurements. What models can do is to translate different value judgments into different paths for policy.

Alphaville keeps returning to the idea of a nine-year-old daughter because we in fact have two of them ourselves. So imagine the following:

You are about to take a car trip with your daughter. There will likely be an accident, which will hurt her. You could prevent that pain in the future, by paying now to prevent the accident. Before you decide, though, you can discount her pain in the future, by a percentage of your choice. That'll make it cheaper for you now, but what you don't pay for, she'll still feel in the future.

Hold on, though. There's also a five percent chance of a catastrophic accident that will cause even more pain. You could pay to avoid that accident, too, but you can also choose not to know the cost, as it might be too high for you to even consider paying.

Is this an economic decision, or a moral one?