Good point about everyone using exponential discounting. I see Stern’s low rate as a quiet way to de-emphasize this model, but I must concede that when both sides of the argument buy into the same equation and argue only about one parameter, it weakens my position.

A revised, weaker version of my argument would ask whether, accepting exponential discounting and arguing about rates, the short-term rate (around 3%) is empirically justified for long-term intergenerational calculations.

I’d agree at the outset that there is MORE empirical basis for a rate around 3%, as Stern’s critics assert, than for a rate less than 1%, as Stern uses—because a few percent more accurately reflects short-term (decadal) preferences while we don’t see sub-percent rates anywhere in practice—but I’m concerned that we don’t know very much at all about whether this short-term rate actually reflects people’s long-term (200 years or more) preferences. In other words, there’s more evidence in favor of a more conventional discount rate, but not enough to convince me that the higher rate is actually even approximately correct for this exercise.

In searching for empirical evidence for intergenerational time-preferences, is it valid to apply our equations retrospectively? Just as we ask climate models to account for ice ages we might ask whether it truly reflects people’s preferences to say that one death in the year 1307 would be equivalent to one billion deaths today (3% time-discount rate). Or is this line of inquiry misleading?

Stern and his critics all use exponential discounting, so the internal discussion is about the discount rate.

But yes, perhaps exponential discount rates should be abandoned.

Mark:

People do discount future health risks, so it is not unreasonable that an analyst assumes the same.

“The fundamental flaw is that climatesceptics think that 5 degrees is a preposterous claim in the first place.”

Is that 5 degrees Fahrenheit (2.8 deg C)? If so, I’d say it’s unlikely, but not “preposterous.” On the other hand, I WOULD agree that 5 degrees Celsius is “preposterous.”

No, his critics value a death today from any cause almost twice as much as ten POSSIBLE deaths 100 years from now from any cause.

You have no idea whether or not people will be immortal 100 years from now. In fact, there is a reasonable possibility that people will be.

Who says it doesn’t? What IS the price of a futures contract to deliver 10 billion barrels of oil in 1000 years?

Thanks for your thoughtful response to my comment. I agree with much that you say. However, I assert that the critics of Stern also apply a discounting model that’s got little empirical basis.

I agree with your assessment that Stern values climate-related deaths far more than those caused by other, more common hazards, but his critics value a death today from any cause almost twice as much as ten deaths 100 years from now from any cause. Unlike money, lives are not fungible so there’s invariably a moral aspect to choosing a discount model.

You seem to agree with this, but argue that there’s an empirical basis for saying that people’s actual behavior reflects exponential time-discounting of lives and I dispute this.

Taking a model (exponential discounting) that works well for certain types of market transactions across years or decades and applying it to completely different sorts of transactions across centuries for which markets don’t exist seems rather Procrustean to me. As a normative exercise (we should discount future deaths exponentially at current rates of return), it’s as good a starting point as any other, but I’m not convinced that choosing such a model is anything more than a statement of moral preference.

Even on the short term, people’s behavior regarding smoking, wearing seat belts, and purchasing insurance, or saving for retirement do not come close to exhibiting exponential discounting. Moreover, people’s preferences depend more on how the issue is framed (in Kahneman/Tversky’s sense of the word, not Lakoff’s) so any effort to quantify time preferences ends up rather forced because the data are so dependent upon the cognitive context in which the decision is taken.

I agree that there’s a strong empirical basis for saying that markets for money or commodities exhibit exponential time-discounting over periods of years to decades.

What I dispute is the notion that there is an empirical foundation for saying that people’s preferences for intergenerational transactions in human lives exhibit similar exponential discounting.

The discount rate is a mix of ethical and empirical considerations.

On the ethical side, there is the question how much we should care about the future.

On the empirical side, there is the question how much we do care about the future.

The answers are very different.

Stern came down on the ethical side, without sensitivity analysis, and without consideration of the wider implications. Dasgupta, for instance, computed that Stern argues that we all should save 97.5% of our income.

Besides that, Stern although a civil servant at the time, ignored the discounting guidelines of his own ministry. That implies, for instance, that Stern argues that a death caused by climate change is more important than a death caused by unsafe roads or substandard medical care.

The fundamental flaw is that climatesceptics think that 5 degrees is a preposterous claim in the first place.

This has problems both because people’s actual behavior demonstrates hyperbolic rather than exponential time-discounting and because long-term consequences are qualitatively different from short-term ones: Exponential time-discounting suggests that a futures contract for the delivery of 10 billion barrels of crude oil in one thousand years should cost less than a dime today. The fact that it doesn’t shows that markets don’t blindly apply exponential discounting over multi-generational time spans. Also, the fact that people have nonlinear utility curves means that the value of (A+B) may be either much greater or much less than the value of A plus the value of B considered individually [see, e.g., D. Kahneman and A. Tversky, "Choices, Values, and Frames," (Cambridge, 2000) or P. Slovic, "Perception of Risk" (Earthscan, 2000)].

Ignoring these real-world complications in order to produce simple economic models frequently leads economists and policy analysts to conduct normative arguments in the disguise of objective quantitative analysis. This is particularly evident when we combine monetizing lives saved or lost with time-discounting monetized values. For instance, we run into the well-known and prima facie absurd conclusion that saving one life today outweighs saving 6 trillion lives from a global catastrophe in the year 3007.

The lesson I take away from the detailed technical arguments about costs and benefits between Stern and his interlocutors is not that these things prove that Stern is right or wrong to reject discounting, but that the same sorts of cautions that apply to taking scientific models too seriously apply as well to economic analyses of costs and benefits. When the policy question is reduced to technical economics of which discounting model to use, we know that we’re seeing the same sorts of meaningless precision that plagues arguments that start from the premise that knowing exactly how high the sea level will rise by 2100 is somehow going to settle normative policy disputes about adaptation vs. mitigation.

Harvey Brooks’s very wise essay, “The Resolution of Technically Intensive Public Policy Disputes” [Science, Tech., and Human Values 9, 39-50 (1984)], points out that when there are not universally acknowledged hard facts to test models, scientists and engineers (and by extension economists) tend (perhaps unconsciously) to bias their estimates of uncertain quantities in favor of the policy outcome they support: “The more an issue is in the public eye, the more expert judgments are likely to be influenced unconsciously by pre-existing policy preferences or by supposedly unrelated factors such as media presentations, the opinions of colleagues or friends, or even the emotional overtones of certain words used in the debate.” [p. 40]

So I agree with the Stern critics that Stern seems to have crafted his analysis to conform with his prejudices, but I would add that in choosing to advocate the use of exponential time-discounting using current rates of return, his critics are doing the same thing. Choosing your time-discount model is a normative, not an analytic or descriptive action.

All of which puts me squarely in line with Pielke’s assessment: “we may indeed be in a situation where analysis is viewed as being more useful as a tool of persuasion than clarifying the consequences of a wide range of alternative courses of action. In such a situation policy analyses will be far less important than the political dynamics.”

If we are indeed moving to a “post-analysis” phase, it can easily be successfully argued that no scientifically valid analysis has ever been performed. So it’s more like we’re continuing on without any valid scientific analysis.

The temperature projections in AR4 (Asssessment Report #4) have no scientific validity, since they don’t estimate the probabilities for each of the scenarios, or the probability of scenarios resulting in temperatures above or below the scenarios given.

This is exactly like the temperature projections in the TAR (Third Assessment Report).

So we’ll have decisions based not on science, but non-science…aka, nonsense. This doesn’t seem like a step forward.