« The intergenerational transmission of IQ | Main | Powerful Women: Does Exposure Reduce Bias? »
Are climate models like economics models?
Here's another reader request:
You've spent a lot of time studying economic models. You probably have an opinion about their overall reliability.
How should that opinion influence your view of the issue of environmental change, given that many of the inferences about such change come from general climate models that are, in some ways, very similar to economic models?
I would prefer betting markets, but I don't think they would suggest something much different from the current scientific consensus. Economic models aside, economic empirics give us every reason to believe that (apart perhaps from environmental issues) today's mixed economies with democratic capitalism have produced, and will continue to produce, entirely satisfactory outcomes. Make of climate models what you may, there is lots of evidence that a) biodiversity is being hammered, and b) climate change will bring desertification, drought, and possibly coastal flooding to many parts of the world, among other dilemmas. I don't have a lot of faith in the exact predictive powers of climate models, or for that matter economic models, but uncertainty about outcomes should make us worry more not less. Uncertainty usually has two tails, not just one.
Posted by Tyler Cowen on August 30, 2008 at 08:37 AM in Economics | Permalink
Comments
The key difference between climate models and economic models is that economic models include volition, human choice.
CO2 molecules cannot change their minds in response to higher prices or different social norms, and act differently.
I'll keep this short, though I've thought about this a bit: the best boundary between science and social science is the boundary between the inanimate and the animate and cognizant player.
Posted by: odograph at Aug 30, 2008 8:56:02 AM
Maybe I should continue ;-), and say that I think the best, most reliable, models are the simple ones involving inanimate objects. We can calculate the speed of a falling body (or pool shots) based on simple reliable physics.
Climate models can be complex, but at least they grow out of that, and layer more and more "pool balls" into the equation.
As soon as we jump to economic models we are doing something very different. We do not have a reliable base physics upon with to expand. Instead economic, and other social models, attempt to average out (and guess at) future human choices.
I consider those two schools night and day, YMMV.
Posted by: odograph at Aug 30, 2008 9:02:37 AM
It is difficult to make a valid blanket statement about 'models' in general. Some models are more sensitive to their parameters than others. (Indeed, some models one could construct would be inherently unstable / ill-posed.)
The criticism of climate models as I understand it is not that they are getting the physics (that they actually model) wrong, which as other commenters are alluding is hard to do. The criticism is that they are so sensitive to parameter estimates with huge error bars (or parameters that are assumed to be constants when they may be dynamic, or...etc) that their signal-noise ratio is meaninglessly small.
I'm not sure I can agree that with Tyler that this sort of uncertainty should "make us worry more". We need a valid reason to "worry" in the first place and if the models are GIGO then we would not really have one.
Posted by: Sonic Charmer at Aug 30, 2008 9:57:15 AM
odograph,
"The key difference between climate models and economic models is that economic models include volition, human choice."
No, the key difference is that the conclusions reached by climate modellers have much more significant impacts on their economic and intellectual self-interests. There are no possible datasets of experimental climate futures that can be used to 'improve' a climate model. Climate models only tend to 'improve' to the extent that they produce results that better align with the expectations and the self-interests of the modeller.
Should a modeller actually publish a contrary result, he is halfway to an inferior career change.
Regards, Don
Posted by: Don Lloyd at Aug 30, 2008 10:04:34 AM
Sonic, wasn't the important mental leap from Conrad Lorenz's weather models that "extreme sensitivity to initial conditions" was not just in the models, it was in nature too?
Posted by: odograph at Aug 30, 2008 10:06:38 AM
Back atcha' Don. I suspect your post has "economic and intellectual self-interests."
How would you disprove that?
Posted by: odograph at Aug 30, 2008 10:08:30 AM
Uncertainty about outcomes should make us more concerned about possible bad outcomes, but should make us less confident in any particular policy's ability to prevent them.
Posted by: Cyrus at Aug 30, 2008 10:36:30 AM
Cyrus, I think it should make us investigate low-cost, high-leverage, responses first. As an example, bike lanes are easy and inexpensive, requiring SUVs to get 40 MPG is hard and expensive.
Posted by: odograph at Aug 30, 2008 10:50:36 AM
Yes, obviously the path to fame and riches in America for the past eight years has been to predict dire consequences arising from global climate change.
It worked for one guy. One guy! For the rest, not so much. I haven't noticed any huge amounts of government funding, for example, being ladled out to climate change researchers. Corporate America not exactly embracing them either.
Either climate change researchers are very poor at predicting what will bring them economic gain, or they're following where the data leads them.
Posted by: Lindemann at Aug 30, 2008 11:23:34 AM
Tyler why don't you call it climate empirics? Or perhaps better, complex physical system empirics: The classroom assignment would be to try to find one complex physical system which can receive a continuous extra input without flipping into a new dynamic state. Are there any at all? Ecological food webs, astronomical solar systems?
On the other hand, economic empirics could be stated as the theory that human creativity, aided by human freedom, will overcome adversity and lead to a better material future. That's what we've been seeing for three hundred years -- not the "successes" of economic models, they're just hitching a ride.
So then, the outcome is: take care of climate empirics NOW, and economic empirics will do the rest. There will be no net cost. You don't even have to learn math!
Posted by: Lee A. Arnold at Aug 30, 2008 12:11:55 PM
"uncertainty about outcomes should make us worry more not less."
This is a key point and one that climate skeptics seem to misunderstand.
Marty Weitzman has shown that there is a small, but real, chance of a true catastrophe and that chance only grows if we think the climate models are noisy predictions of the future.
Posted by: a student of economics at Aug 30, 2008 12:45:08 PM
Someone is taking a look at the accuracy of the models...
http://www.atypon-link.com/IAHS/doi/pdf/10.1623/hysj.53.4.671?cookieSet=1
Posted by: Fact Checking at Aug 30, 2008 1:28:49 PM
"Economic models aside, economic empirics give us every reason to believe that (apart perhaps from environmental issues) today's mixed economies with democratic capitalism have produced, and will continue to produce, entirely satisfactory outcomes."
It is rare to see such a strikingly clear statement of how we got into our current mess, and how easy it is to rationalize policies that will just keep pushing us to the brink.
Rephrased, what Tyler is saying is “So long as we disregard the predictions of the imminent collapse of the global systems that sustain our basic needs and provide the free services that our extractive economy depends upon, everything is working fine! Carry on.”
I appreciate the concern in the following sentences, and it is too easy to dismiss this posting a platitude. I just want to know how the vast majority of the Earth that is not benefiting from participation in the democratic capitalist system can be dismissed with a parenthetical statement.
Posted by: Alger at Aug 30, 2008 2:50:52 PM
Tyler,
You say you assume that betting markets predict about the same global warming outcome as scientists do, on average. Perhaps this is true of the formal betting markets -- but not in other markets where much larger bets are effectively being placed on the climate.
For example, if you believe warming is real, you probably wouldn't buy an ocean-front home right now, but I haven't noticed them getting cheaper of late, even with the down real estate market. For a people who claim a pretty broad consensus on this issue, we don't bet like we do.
Posted by: Scoop at Aug 30, 2008 3:53:01 PM
Climate models are really simulations. The equations for fluid motion and energy transfer are solved numerically on a grid using finite difference methods to move forward in time. Some of uncertainty arises from inadequate resolution but the most important source of uncertainties is from is specification of things like cloud formation, ocean temperatures, etc. Most predictive economic models are statistical fits to data with some assumptions about the relationship between variables. I think the assuming that climate models are "very similar to economic models" leads to misunderstanding of what the uncertainty in climate models mean.
Posted by: joan at Aug 30, 2008 5:31:00 PM
climate change will bring desertification, drought, and possibly coastal flooding to many parts of the world
I've never understood the global warming = more drought connection as it's my understanding that the AGW theory requires significant feedback in the context of added water vapor in the air, which seems to imply more precipitation. Then you would have to add in melting ice which would increase the amount of water in the percipitation / evaporation loop.
Posted by: Jody at Aug 30, 2008 7:21:06 PM
As a high-school dropout who has more intelligence than ambition, I'll put it to you this way: both the "science" of climate change and the "science" of economics have failed to impress me with the rigor of their empirics.
At a scholarly site such as Climate of the Past http://www.clim-past-discuss.net/papers_in_open_discussion.html it appears to me that half the papers are summaries of what happens when you tweak computer programs and datasets this way or that. Past posts on this forum of similar tendencies in the sphere of economics speak for themselves.
Should serious policy makers stake their careers on such thin reeds? It seems foolish to me.
Posted by: Bob Knaus at Aug 30, 2008 8:20:41 PM
Model-building in physics is a management career path.
Basically you make up equations that are not real physics equations.
Posted by: Ron Hardin at Aug 30, 2008 9:40:33 PM
When Rich posted this question on the other thread, I was ecstatic, and hoped that Tyler or Alex would answer him. Alas, I think Tyler misses (what I take to be) Rich's main point.
Tyler (if you're still reading), I know you are not as rabid a free marketeer as some of us, but I think you would generally endorse this history:
In the 1950s, the "consensus" among the most credentialed economists--the ones who taught at the best schools, the ones advising presidents, and the ones with the most prestigious publications--was that free markets, left to their own devices, could lead to human misery. At the very least, fine-tuning by government could greatly improve living conditions for future generations. Indeed, Noble Laureate Paul Samuelson declared as late as the late 1970s (right?) that the Soviet Union might soon overtake the US in economic output.
It's true, there were some "deniers." But most of them were cranks like Ludwig von Mises and Friedrich Hayek who couldn't even put their crotchety objections into a model. In fact, that was part of their whole point--they objected to the whole enterprise of modeling, saying it couldn't capture all the relevant variables, blah blah blah. Sooo 19th century.
Admittedly, there was a big gun here or there who came out against the consensus, like Milton Friedman. But even though he did publish some smart critiques in the peer-reviewed literature, it was soon apparent that he was just trying to be rich and famous by appealing to conservatives. He spent most of his career writing pop books and giving speeches to the choir, rather than engaging in serious research like the true scholars at MIT and Harvard. (I hear that Friedman took money from the denier Volcker Fund, proof that he wasn't objective.)
When you boiled down the rhetoric of these planning deniers, it usually wasn't a scientific objection to the rigorous mathematical approach of Samuelson et al. Rather, it was a knee jerk political reaction. "Waaa, I don't want to pay higher taxes. Waaa, I'm afraid of jack-booted thugs running the economy." Oh please.
[/analogy not too far from reality]
==============
One final point: I think you are totally dodging the important question here, Tyler, when you say:
I don't have a lot of faith in the exact predictive powers of climate models, or for that matter economic models, but uncertainty about outcomes should make us worry more not less. Uncertainty usually has two tails, not just one.
I think what Rich may have had in mind is this: If we are uncertain about macro models, should that make us more willing to give Paulson et al. the power to run things? I say no, do you?
If you agree the answer is no, then why would our uncertainty about the effects of humans on the climate, cause us to give more power to the government to regulate our lives? Since when do we give politicians the benefit of the doubt?
Posted by: Bob Murphy at Aug 30, 2008 10:29:14 PM
Tyler,
I think Bob Murphy makes a good point. And, if I recall correctly (and I think I do because I reviewed the book in Fortune in 1985), Samuelson (with Nordhaus) was making the statement about the Soviet Union in the 1985 edition. Nordhaus, hmmm. Seems as if I've heard that name before. Oh, yes, he's one of the worriers about global warming.
Best,
David
Posted by: David R. Henderson at Aug 30, 2008 11:21:58 PM
One thing held in common by all computer models: They rely on data. A reliable model would produce an accurate forecast with good data.
It seems to me that, given good data, economic models would could be useful but still prone to go awry because of human elements, but I could be wrong.
Climate models are, in fact, extremely difficult to produce because of the complexity of the system. As I have surveyed the literature available online I have discovered that some think there is a simple linkage between the temperature of the Sun at any given time and the climate on earth. Other models give little weight to Sun temperature. I have also read of a theory that the total rainfall on the face of the earth is nearly constant, but that there is no data available to prove or disprove this theory. It seems like total rainfall on the planed might be an important statistic to any climate model, but we are incapable, so far, of measuring this.
I have to vote for economic models over climate models. Economic systems are complex, but we seem to have learned a lot about collecting good data and determining which numbers are salient to the model.
Posted by: Rob at Aug 30, 2008 11:25:26 PM
Precision is not in the cards, so we have to aim at a general direction. I think in most of academia, at least in engineering, the publishing of a model is important because of the model. Noone expects the model to prove anything, at least for a long time. They use experiments or experience to prove the model.
But it's not the uncertainty that worries me. It's the certainty.
On the one hand we have the uncertainty of human-caused climate, and the uncertainty as to how bad it would be, the uncertainty as to whether it may be at least a partial improvement, the uncertainty as to what we can do about it, and the uncertainty about the idea that when it became more obvious we could fix it quickly.
On the other hand we have the absolute certainty of the economic damage and hindered progress damage that would be done by the only ideas offered by the people who only understand political "solutions."
Posted by: Andrew at Aug 31, 2008 4:47:45 AM
I embarked on a 'Who is really right" campaign when I was director of public policy analysis back at Enron and had the advantage of getting the alarmists to cooperate because Enron was "green." (The skeptics are always ready to cooperate....) I worked with probably 15 scientists and have over 150 emails for when the history of the debate is written some years from now. One big name middle-of-the-roader/alarmist was particularly helpful, and he was a paid consultant.)
Maybe the most important thing I learned is the gulf between the modelers and the empiricists. The modelers have much higher sensititity estimates than the empiricists. Indeed, before slicing and dicing the explanations, we do have the fact that the world has warmed remarkably little since the end of the Little Ice Age in the mid-19th century compared to what models predict given the buildup of GHGs in the atmosphere today. Some say ocean delay and aerosol offsets and such, but the other explanation is that the positive feedbacks that get the predicted 2x equilibrium warming from 1.2C (neutral feedbacks) to the predicted 3C (IPCC mid-range model average) are overstated badly.
So here is the crux of the matter. There is no "skeptic" model and there cannot be because the microphysics of climate that MIT's Richard Lindzen and some others believe to be reality cannot be incorporated into the macro models. Models simply do not have the resolution and computing power to have a "Lindzen" version.
What is fascinating to me about this debate is the micro-macro dichotomy. It might turn out that Lindzen is the "F.A. Hayek" of the debate because he was right and he was micro oriented.
By the way, Lindzen is the top theoretician in the debate. My middle-of-the-roader often remarked how Lindzen was the smartest guy of them all but was a theoretical risk taker too.
Hope this helps....
Posted by: Rob Bradley at Aug 31, 2008 9:51:59 AM
Actually climatology has proceeded much farther since the day Enron failed due to mismanagement and fraud.
Joan is the only one who answered the top headline question correctly.
I would add that the implication is that economic models are scientifically weaker than climate models.
Posted by: Lee A. Arnold at Aug 31, 2008 11:37:04 AM
For those interested in how climate models work, and how they are different to economic models may be interested in The Physics of Climate Modeling or A new kind of scientist.
Posted by: Gavin at Aug 31, 2008 1:29:56 PM
