Is Climate Change an Economic Problem?

In Chapter 3 of the Climate Casino, William Nordhaus argues that climate change is an economic rather than scientific problem (find my reviews of other chapters here). The argument is controversial. Heat waves, melting ice sheets, droughts, rising sea level, storms and record temperatures are typically thought of as a complex scientific problem. Prof. Nordhaus argues that the economic problem is more straightforward: the people that are dumping CO2 into the atmosphere don't have to pay for their emissions. If they did, they would be more careful about what they did with the by-products of burning fossil fuels. Economists call this an externality, that is, "...a by-product of economic activity that causes damages to innocent bystanders" (p 18). Because no one owns the atmosphere, not even government, there is no one to charge for its use and no market in which to purchase such environmental services. 

Economists like this argument; it shifts the debate from "who is to blame" to "who should pay" for damages caused by climate change. Unfortunately, polluters argue that no one is actually being harmed by CO2 emissions and that there is no observable link between CO2 emissions, climate change and economic damages. 

Prof. Nordhaus confronts these arguments by presenting the data and using models to predict the future. Since 1900, global CO2 emissions have been rising linearly (on a log scale) in his Figure 2 above (p. 21). Although there have been some interesting periods of acceleration (after WWII, for example), the general trend has been continuously upward.

Over the same period, there have been technological improvements that increased the efficiency with which fossil fuels are being burned. Technical efficiency has reduced the carbon intensity of, at least, the US economy as displayed in Figure 3 above (from page 22 of the Climate Casino, again presented on a log scale). This process is called decarbonization, but "...while the carbon intensity of production is declining, it is not declining fast enough to reduce total CO2 emissions, either for the world or for the United States" (p. 23). In other words, technological change will not get us out of this dilemma and neither will the substitution of renewable resources because they are more expensive than fossil fuels--returning us to the economic problem.

Another part of the economic problem is that the negative impacts of fossil fuel burning, even with increased technical efficiency in their use, will not be felt for many years in the future. To look out past 2013 when the Climate Casino was written into the future when environmental effects could become serious, we need a mathematical model that can make quantitative predictions. Prof. Nordhaus has such a model, called the DICE model, as in "roll the dice". In Figure 3 above (from p. 33 on the Climate Casino), emissions projections out to 2100 from the DICE model are compared to projections made by eleven EMF models from the Stanford University Energy Modeling Forum. The projections from the DICE model compare quite favorably with the average of the EMF models.

Prof. Nordhaus acknowledges that there is a great deal of "uncertainty" in these estimates. CO2 emissions in 2100 range from about 40 billion to over 130 billion tons per year. The differences in predictions have nothing to do with "chance" or "random events" but rather with the different structures and assumptions of the models. The models are totally deterministic. 

In the face of uncertainty about the future (and uncertainty about the models), Prof. Nordhaus makes the following suggestion:

Start with a best-guess scenario for output, population, emissions, and climate change. Take policies that will best deal with the costs and impacts in this best-guess case. Then consider the potential for low-probability but high-consequence outcomes in the Climate Casino. Take further steps to provide insurance against these dangerous outcomes. But definitely do not assume that the problems will just disappear. (p. 34-35)

This sounds reasonable, but the arguments in Chapter 3 of the Climate Casino do not seem well connected to the suggestion. Is Prof. Nordhaus saying that the DICE and EMF-22 models are necessary to making judgments about either the future or about policy actions that should be taken in the present? Are the DICE and EMF-22 models an accurate simplification of how the world economy works? How do we know that? Have the models been somehow validated? Is there a way we can check the projections being made by the models? If we are going to run policy experiments on the models, how do we know the world economy will respond in the same way? How would we even think about implementing policies that would have an impact on the world economy? Is there any way for us to evaluate these models as systems? What kind of confidence would we have to have in models to make predictions almost eighty years into the future? Is there any reason to believe that we can really predict that far into the future? And, how do the models prove that climate change is really an economic problem? As I will argue later, it seems more realistic to view it as a "system" problem, specifically a problem of the Capitalist world-system, a system that is great at exploiting environmental systems for profit. Is it possible to change this system? How have macro-societal systems changed in the past?

So many questions! They will all have to be topics for future posts, keeping in mind that this blog is about causal macrosystems and not about climate change. Unfortunately, it is in the area of climate change that the stakes are highest and to which, hopefully, researchers are bringing their best models!

EXERCISES

1. Download the DICE users manual (here). How well was the model validated? What evidence was offered to support using the model for policy experiments? How were the model parameters determined? Evaluate Ackerman's (2007: 1657 here) concern that "...the optimistic projections and modest optimal policies often attributes to models such as DICE may be artifacts of parameter choices, rather than robust forecasts about an uncertain future" For more topics to pursue in the DICE manual, see below.
2. Follow up on the references offered in the DICE manual, particularly the Ramsey model (the one that DICE is based on here) and versions of DICE that are not deterministic (here). Try to find a reference that validates the model on historical data.
3. Download the "Simple Excel" version of DICE (here and here). Try to understand the model and explore different runs. Make simple changes to the alternative assumptions. NOTE: the Excel model seems "under development" and I have not been able to get it running. Skip to Exercise 4 if you're having trouble.
4. Download the free educational version VENSIM PLE (here) and the version of DICE in the VENSIM language from Tom Fiddaman's System Dynamics Model Library, specifically here. Become familiar with the VENSIM language. Most of the causal macrosystems I will review have been implemented in VENSIM and it is free.

Using the DICE manual, evaluate and critique the following statements:

1. DICE is an optimization model. "...the use of optimization can be interpreted in two ways: First, from a positive point of view, optimization is a means of simulating the behavior of a system of competitive markets; and, second, from a normative point of view, it is a possible approach to comparing the impact of alternative paths or policies on economic welfare" (p. 6)
2. "In the DICE and RICE models, the world or individual regions are assumed to have well-defined preferences, represented by a social welfare function, which ranks different paths of consumption" (p. 6).
3. "Technological change takes two forms: economy-wide technological change and carbon-saving technological change. The level of total factor productivity ... is a logistic equation similar to that of population...TFP growth declines over time" (p. 9). This is an interesting change of specification. In the neoclassical economic growth model, TFPG is simply an exponential function, meaning it increases forever since there are no inherent limits to the growth of knowledge (see my earlier post here).
4. "...providing reliable estimates of the damages from climate change over the long run has proven extremely difficult" (p. 10 and following).
5. "The DICE-2013R model explicitly includes a backstop technology, which is a technology that can replace all fossil fuels" (p. 13).
6. "The model assumes that incremental extraction costs are zero and that carbon fuels are efficiently allocated over time by the market" (p. 14)
7. "As with the economics, the modeling philosophy for the geophysical relationships has been to use parsimonious specifications so that the theoretical model is transparent and so that the optimization model is empirically and computationally tractable" (p. 15).
8. "This discussion implies that we can interpret optimization models as a device for estimating the equilibrium of a market economy. As such, it does not necessarily have a normative interpretation. Rather, the maximization is an algorithm for finding the outcome of efficient competitive markets" (p. 22). How well does this interpretation fit the world system?
9. "The real interest rate is a critical variable for determining climate policy" (p. 25).
10. "Earlier versions of DICE and other IAMs tended to have a stagnationist bias, with the growth rate of total factor productivity declining rapidly in the coming decades. The current version assumes continued rapid total factor productivity growth over the next century, particularly for developing countries" (p. 36).
11. "The 2007 model over predicted output by about 6 percent, primarily because of failure to see the Great Recession. We have reduced output to put the new path on the current starting point" (p. 41).
12. "Since many computerized climate and integrated assessment models contain between 10,000 and 1 million SLOC...[source lines of code]..., there is the prospect of many bugs contained in our code" (p. 52).
13. "...there is a tendency to develop models that increase in parallel with the rapidly expanding frontier of computational abilities. This leads to increasingly large and complex models. We need also to ask, do we fully understand the implication of our assumptions? Is disaggregation really helping or hurting?" (p. 53)
14. "The properties of linear stochastic systems are moderately well-understood, but that is not the case of all non-linear stochastic systems" (p. 54).