Recent Posts

Wednesday, July 24, 2019

Some Comments On Economic Climate Change Models

Other than a few comments on the Green New Deal (on the basis of economics), I have shied away from discussing the issues around climate change. However, I ended up in the cross-fire of a discussion between Steve Keen and others on Twitter, as a result of tweeting a link to this article: "The Cost of Climate Change: Steve Keen Dismantles William Nordhaus." As an immediate disclaimer, I am not familiar with the literature, but one basic premise of Keen's seems correct: the time discounting methodology used by Nordhaus makes little sense. Since it follows an optimal control paradigm, it unsurprisingly attracts defenders among neoclassical economists. However, as someone with a control engineering background, it does not really work. It only makes sense if the costs of climate change are essentially trivial, which is an obviously disputed point.

A Question That Scientists Should Have Answered Decades Ago

The most elementary observation is that we should have had the scientific questions around human-caused climate change settled a long time ago, and the answers should have been presented in a convincing fashion. Unfortunately, no matter what the state of the science is, the "convincing" part is missing. We are in a situation where beliefs on the science of climate science coincide almost perfectly with peoples' leanings on political economy.

I do not have the expertise to offer a convincing analysis either way. In order for my arguments to make any sense, we have to grant the premise that there are valid reasons to worry about carbon emissions. However, I have heard the criticism that risks are being exaggerated, for any number of reasons. Since I am in the "Peak Everything" camp, I was always of the opinion that carbon emissions will be self-limiting, and so never pursued the subject in depth.

From a technical perspective: are there feedback effects that will cause a self-perpetuating run in temperature in one direction or another? For example, the paper "Trajectories of the Earth System in the Anthropocene" cited by Steve Keen discusses such temperature spirals. This is referred to as a "tipping point." The possibility of a tipping point is more worrying than the traditional way of presenting climate change - as a gradual rise in temperature over the next 50-100 years.

An Engineering Question

The issues around man-made climate change are an engineering question - if only because engineers caused the problem. There are two aspects.
  1. What is the effect of rising carbon dioxide levels on climate (which is a scientific question)?
  2. If the effects of projected carbon dioxide levels is unwanted, is there a way of reversing this?
My argument is that given the fog of uncertainty (or at least, a fog of obfuscation) around those questions, we cannot come up with any convincing answer. Furthermore, the fog also allows bad answers to flourish.

Discounting Future Costs: Questionable

Attempting to estimate costs of climate change and discounting their value to the present is arguably the wrong answer in the environment of uncertainty we face.

Very simply, everyone understands that we are potentially imposing certain costs to avert a future problem; we do not need a discount function to pretend that trade off is somehow "scientific." Furthermore, how exactly can we assign a dollar value to the prospect of losing coastal cities in the future (and thus discount to the present)?*

Why do we discount things? We are trading off the value of present resources versus future resources. Since we typically assume that we are in an economy where capital investment results in growth, it makes sense that we need a greater amount of future benefits to balance present costs.

In this context, the embedded assumption is that we can mitigate the effects of climate change in the future, instead of incurring the cost of mitigation now. Given the uncertainty around feedback effects, that is not necessarily true. In the worst case, we may have only a very limited (possibly empty) set of technically feasible paths to avoid catastrophe (which is a message that many climate scientists are arguing). That is, the future might be now (or yesterday).

What we need is an understanding of the question: what needs to be done now (or possibly later) that will have what effect on future outcomes? If we had clarity on that, the political process will tell us what the revealed preference is for the discount rate.

Aggregated "Capital" and "Technology"

William Nordhaus offers a summary of his modelling as follows in this article:
The DICE model views the economics of climate change from the perspective of neoclassical economic growth theory. In this approach, economies make investments in capital, education, and technologies, thereby reducing consumption today in order to increase consumption in the future. The DICE model extends this approach by including the "natural capital" of the climate system as an additional kind of capital stock. By devoting output to emissions reductions, economies reduce consumption today but prevent economically harmful climate change and thereby increase consumption possibilities in the future.
The DICE model has 12 behavioral equations, two variables to be optimized, and several identities. In the GAMS version, the simplest model has about 240 lines of operational code. A run of 1,000 years takes five seconds, so it can be used for projects with multiple states of the world and Monte Carlo experiments.
 I would argue that two key features will drive his analysis off in the wrong direction.
  1. The use of optimisation is an unsurprising result of a fixation on 1960s optimal control theory. As a control systems engineer educated in the 1990s, I would argue that we need to hold optimisation with skepticism. In particular, any option is optimal if you choose the right the objective function. For example, I am in the peak oil camp, and I have an extremely pessimistic view of foreign policy - views that I view as hard-headed realism. From my perspective, the objective function should emphasise energy independence, and thus something like the Green New Deal is the optimal strategy to follow. Meanwhile, Nordhaus' "optimal" strategy (discussed below) would be decidedly sub-optimal using my objective function.
  2. Lumping physical systems under aggregated "capital" and "technology" is hardly rigorous.  "Capital" is hardly fungible - a factory that produces SUV's is a liability, not an asset, if there is a necessity to cut carbon emissions. If one wants to sound like an internet Austrian, one could easily argue that a significant portion of current fixed investment is malinvestment if we are forced to change course on energy consumption in the future. "Technology" is a somewhat faith-based belief. In practice, the large leaps in productivity that came since the industrial revolution were the result of harnessing highly concentrated energy sources. We are running to physical limits on productivity and resource depletion curves. The issues around carbon emissions involve physical laws; creating a new app will not bypass those.
Nordhaus' premise in discussions is that a carbon tax can control emissions, creating an "optimal" path. The problem is straightforward: for many applications, there are no alternatives to carbon-based fuels. Meanwhile, energy costs are embedded in practically all goods, and many services (server farms devour electricity). All that the carbon tax might accomplish is raise the price level. Yes, alternative energy sources will have a competitive edge - but those sources are reliant on high energy density manufacturing processes, and face their own depletion curves for critical inputs. The simulations of policy options are entirely divorced from engineering reality.

Even if we accept the premise that carbon taxes can avert dangerous climate change, we face a few questions.
  1. What is the maximal acceptable path for carbon emissions, and/or what are the consequences of each path? (The argument I have seen made by some more alarmed scientists is that there is no feasible path for carbon taxes to be effective, which runs contrary to the assumptions in the simulations.)
  2. To what extent will a carbon tax reduce global emissions (and to what extent will this be sustained)? After all, manufacturing will just relocate to jurisdictions that are more lenient with regards to the tax. By analogy, imagine a tax on breathing. Regardless of the tax level, people will keep doing it.
  3. Fiscal policy needs to be properly simulated, which is a challenge for neoclassical growth models ("let us assume full employment"). If the tax is small, it does not change behaviour. If it changes behaviour, it would almost certainly cause a recession as a result of the annihilation of fixed investment in the affected industries. Although global recessions do a good job of lowering carbon emissions, it is unclear how fast a recovery would be with the industrial sectors smashed. Such an outcome is not politically sustainable, nor "optimal."
  4. What are the distributional effect of these carbon taxes? If the burden of the taxes falls almost entirely on the middle classes, it will be politically doomed. But if offsetting tax cuts are made elsewhere, net effect on behaviour may be a wash. Energy expenses are often viewed as necessities, and so unless household budget constraints bite, energy use might not budge.
None of those questions are going to be answered with a toy optimisation model.

The reality is this: if we do want to reduce carbon emissions, we need an actual technical plan to get from point A to point B. Say what you want about the Green New Deal, it at least attempts to grasp that nettle.

Concluding Remarks

I am obviously not an expert on the climate change modelling literature that follows the traditions set by Nordhaus. However, given the fundamental defects of the approach, I see little value in pursuing the subject.


* I was passed a reference to the paper "The Economic Effects of Climate Change." Measuring the cost of climate change as a reduction in GDP raises some issues. The most obvious one is that GDP is not a measure of well-being, as is well recognised. Instead, it a measure of work effort. A key example to keep in mind is World War II. The war did a very good job of raising economic output in the United States relative to the Great Depression, and it created a lot of jobs. However, that does not imply the war was a good thing for the United States. As such, measuring losses in terms of GDP drawdown is meaningless for policy discussions involving subjects that are not solely about economic activity. (For example, it makes sense to discuss the effects on GDP of active fiscal or monetary policy.)

(c) Brian Romanchuk 2019


  1. Isn't there a lot of controversy with the data being fudged? The "hockey stick graph."

    1. Yeah, that’s one of the claims. From my perspective, the big issue is how do distinguish man-made warming from a natural rise?

      I took basic physics, but all my advanced science was on electromagnetism and quantum physics (allegedly needed for understanding semiconductors). I’ve read pro and con arguments, and whatever I’m reading at the time seems to make sense. Like I wrote, they really dropped the ball on presenting convincing arguments for people like me who want the actual story, and not arguments about “consensus.”

  2. This comment has been removed by a blog administrator.


Note: Posts are manually moderated, with a varying delay. Some disappear.

The comment section here is largely dead. My Substack or Twitter are better places to have a conversation.

Given that this is largely a backup way to reach me, I am going to reject posts that annoy me. Please post lengthy essays elsewhere.