"Our greatest responsibility is to be good ancestors."

-Jonas Salk

Friday, October 15, 2010

Jim Manzi's Cherry Pick

This follows on to previous discussion of Manzi here. Manzi's original piece is at The New Republic.

First off, what are integrated assessment models?
The DICE model, developed by William Nordhaus, is a dynamic integrated model of climate change in which a single world producer-consumer makes choices between current consumption, investing in productive capital, and reducing emissions to slow climate change.

... Current carbon emissions add to atmospheric concentrations via a fixed retention ratio, and realized temperature change is modeled by a three-box model representing the atmosphere, mixed-layer upper ocean, and deep ocean. Damage from climate change is a quadratic function of realized temperature change with a
3-degree change calibrated to cause a 1.3 percent world GNP loss.
(emphasis added)

Begging the question, wouldn't you say?

Jim Manzi bases his approach to the future of the world on the fact that IPCC WG II quotes the results of such models as showing modest impacts of climate change, but those modest impacts are built in. These are not simulations. These are nothing like the physics based models we have in physical climatology. These are guesses.

What does IPCC really have to say about them, other than the graph which Manzi ultimately references?
"It is likely that the globally aggregated figures from integrated assessment models underestimate climate costs because they do not include significant impacts that have not yet been monetised. It is virtually certain that aggregate estimates mask significant differences in impacts across sectors and across regions, countries, and locally. It is virtually certain that the real social cost of carbon and other greenhouse gases will rise over time; it is very likely that the rate of increase will be 2% to 4% per year. By 2080 it is likely that 1.1 to 3.2 billion people will be experiencing water scarcity; 200 to 600 million, hunger; 2 to 7 million more per year, coastal flooding."
I am no great enthusiast for the AR4 WG II report, and they are not entirely free from responsibility for Manzi's grossly dangerous conclusion. But a fair reading of the executive summary ought to have been enough to dissuade Manzi from using the figure in question. The text which he advises the world to base its entire future is an egregious cherry pick.

Nobody who calls GCMs into question should pay the least attention to the Nordhaus type model, which is grossly underconstrained by theory or observation, and instead is constrained by guesswork.

Manzi takes a different tack, treating IPCC as authoritative. Thus he accepts or claims to accept the WG I sensitivity spectrum, and uses this as cover for picking a single chart out of WG II as representative of IPCC's impact assessment. But that choice is not representative of IPCC's impact assessment at all, and is explicitly disavowed in the chapter executive summary.

I hope Manzi will acknowledge his error and change his position.


crf said...

No link to Manzi?

Michael Tobis said...

Good point. Fixed.

Unknown said...


You know GCMs well (almost better than I do, in fact, at a level of computational detail), but you're mixing apples and oranges. There is a difference between (1) the estimate of the costs in year X of a temperature increase path up to that point built up from detailed sub-estimates, and (2) the discounting calculations used to estimate the PV impacts of various alternative courses of actions ASSUMING these underlying analyses (among many others). You know complex software development well. There is a code object within (2) that is called and is calibrated based on analysis (1). That's all the quote from Nordhaus that you cite means.

The second quote refers to the higher-level integrated climate-economics models, and the resulting Social Cost of Carbon; it doesn't refer to the calibration of (1), which is what I've repeatedly referenced.

Look, if the IPCC doesn't want to be cited as making this specific projection, they shouldn't put it in their Summary for Policymakers.


Unknown said...

Sorry, in the prior comment "almost better than I" was supposed to be "almost certainly better than I"

Quick posting :)

Anonymous said...

Nordhaus had established the 1.3% figure in a 1992 discussion paper, already.
The 'DICE' Model: Background and Structure of a Dynamic Integrated Climate-Economy Model of the Economics of Global Warming

Relevant pages are 22 (section A9), 44 and 45. According to my cursory reading Nordhaus seems to have looked only into impacts on agriculture, coasts and the energy sector and to have extrapolated from the US to the rest of the world. The impact on global GNP is also small, as the impact of agriculture on GNP is seen as shrinking. Caveats on pages 44 and 45 don't instill a high level of confidence:

This estimate depends heavily on the estimates for the United States and the method of extending these estimates to other countries. Both of these are subject to large margins of errors


It should be noted, again, that, because of infirmities in the underlying estimates of damage, this result does not inspire high levels of confidence.

Nordhaus also estimates impacts on the agriculture to be 4.92% and on costal activities to be 6.91% (table III-C1). I marvel at the three significant digits.

Hank Roberts said...

I'll ask again: what's the GDP of the oceans? What was it a century ago?

David B. Benson said...

Might care to look at what Barton Paul has been posting regarding a drought index, which is growing, and thus the serious impact upon agricultural production.

We have recently been through a year of lowish agricultural production and that starting with low reserve food stocks. Look at the (permanent) damage done to some of the world's poorest peoples.

Agriculture is central; computers for blogging are not. GDP is a meaningless measure of reality.

Michael Tobis said...

David, I think you have a good point. The idea that agriculture is "a small part of the economy" is based on the same sort of small-signal thinking that the rest of economics is. Demand for food is pretty much "inelastic" in the economists' sense. If supply goes down, the percentage of the economy that goes to food skyrockets. In the end an iPad is worth less than a stale peanut butter sandwich. The economists will call this "inflation" of food prices, but it may just as well be considered a more realistic valuation of iPads and everything else.

On the other hand, we have the inequitable distribution of "capital", which is based on how many fingers your country gets on the iPad before it is shipped. China's share is going up, and now China is demanding meat. Consequently Haitians are starving.

Is this any way to run a railroad?

Steve Bloom said...

While we're speaking of drought, the NYT had a OKish article on the current very nasty drought in and around Syria. Not good.

David, BPL's paper is basically a straight extrapolation of current trends. While it's a calculation that might make for a good sidebar in a paper exploring modeled and/or measured drought trends, the problem is that such a thing is impossible. There's a limit to how much drought there can be without a reduction in the hydrological cycle (which consistent with projections the cycle is on the increase), and in addition non-linear effects are bound to kick in. Have a look at this fresh paper, which is quite bad enough.

David B. Benson said...

MT --- Haitians are in big trouble for other reasons, methinks.

As for reforming economics, the old idea was to convert $$ in utils, ala utility theory. I don't think that works although some labor along. For instance, attempting to be precise about satiation (of desire for a good).

I opine more progress could be made by considering ecology and economy to be a constraint satisfaction problem (CSP). Considerable progress, both in the theory and in the practice of CSPs, has been made.

Part of all the difficulties, IMHO, is there now are so many constraints that just mere human reasoning, together with old-fashined linear approximations, no longer arrives at solutions.

[And finally, the way to run a railroad is to move lotsa coal on it, every day.]

David B. Benson said...

Steve Bloom --- I agree that BPL's analysis is overly linear. However, despite the increase in the hydrological cycle (or perhaps because of it), a higher fraction of the total falls into the oceans (I think). In any case, the drought index is increasing and at the same time too much rain falls all at once in various episodes; also bad for agriculture. (So for that reason, Indonesia is going to have to import rice for at least six months.)

[Unfortunately, my .pdf reader croaks on that paper from Nature, claiming it contains an unrecoverable error.]

Michael Tobis said...

Steve, your link points here, not to the Times.

Michael Tobis said...

NYT Syria story.


Steve Bloom said...

OK, the NYT article link (which was empty).

For the second link, try going here and clicking through; it's the second paper listed. Also, the first paper shows matching plant productivity trends.

See also this paper showing a recent increase in river flows globally.

Not a pretty picture, on the whole.

Steve Bloom said...

Yow, indeed. Also see Rami Zurayk's blog, which is IMHO a stupendous resource.

Michael Tobis said...

I appreciate the butter-up but I don't understand Jim Manzi's claim:

"There is a difference between (1) the estimate of the costs in year X of a temperature increase path up to that point built up from detailed sub-estimates, and (2) the discounting calculations used to estimate the PV impacts of various alternative courses of actions ASSUMING these underlying analyses (among many others). You know complex software development well. There is a code object within (2) that is called and is calibrated based on analysis (1). That's all the quote from Nordhaus that you cite means."

The quote says DICE gets temperature from a simple box model and that "Damage from climate change is a quadratic function of realized temperature change with a 3-degree change calibrated to cause a 1.3 percent world GNP loss." Try as I may I can't make that consistent with Jim's reading. Can anyone bridge the gap?

David B. Benson said...

Russia, Ukraine and one of the 'stans have all stopped wheat exports. Locally, the wheat ranchers have been receiving record prices, measured in unadjusted $$. The only time when prices skyrocketed more, using $$ adjusted for inflation, was during the great USSR wheat debacle when the US bought wheat to send to the USSR (don't remember the year).

Steve Bloom said...

Here's an article describing what happens when a similar-sounding drought happens in a country (Australia) with enough resources to soften the blow.

Also, there's a new paper (press release) describing the negative impact of floods and droughts (mainly the latter) on river ecologies. Check out the dark humor in the headline.

Brian said...

The way I put it all together from bluegrue and Jim's comments is that a sketchy and avowedly tentative, 1992 paper on costs for certain parts of the US economy was assumed to cover all costs for the entire US economy and then extrapolated somehow to apply to the global economy. The IPCC equation calibrates based on the assumption that the 1.3% figure from the 1992 paper is correct.

In other words, the 1.3% isn't drawn from thin air, but nor is it drawn from something giving great confidence, either. Who knows, though, maybe as with everything else climate-related, we can just hope that we'll get lucky.

Maybe there's more to this story that others can add.

Anonymous said...

I'm not sure if I understand Jim Manzi at all?
Can someone say it in some other way?

This is not sarcasm, I hate when I feel that there is a misunderstanding and people generate conflict from that.

James Annan said...

I think Jim is saying that although DICE uses a trivial quadratic, this was derived as a fit to a more complex model. Much the same as how us climate scientists often use rather simple energy balance models (eg MAGICC and Bern model, which both feature heavily in IPCC assessments) which have been calibrated to GCMs.

I suspect that even the complex economics models are much more conceptual and far less constrained by reality than the GCMs though (inasmuch as one can make meaningful comparisons, which may not be that easy). Which is probably what mt meant to say, even if the example of DICE isn't necessarily the best one to use.

Michael Tobis said...

No, a 5% loss of capital is much worse than a 5% decline in GDP. I used the arbitrary multiplier of 100, which perhaps is confusing, but it seemed to me like a conservative (in the sense of generous-to-the-opposing position) estimate. That is, a gradual loss of 5% of, um, everything, would be equivalent to the loss of 500% of annual production.

The fact that the factor of 100 appears in two places in this analysis is confusing, but it really seemed like a fair number to me.

As an aside, presuming the above is clear: Some interesting questions about linearity are raised. If 5% of everything went away suddenly, it seems like it might be Game Over for economics. The multiplier, like many other multipliers, may have something to do with rates.