25 July 2017

What To Do About the Immaturity of Systems Modeling

Sam Harris recently had a conversation with Scott Adams (Dilbert creator and author of How to Fail at Everything and Still Win Big) about Trump. Adams predicted Trump's victory because he sees Trump as a master persuader. There's a lot to say about that but Adams made a really useful distinction about what he saw as the three stages of climate change policy. He distinguishes between:
1) The reality of climate change as an ongoing phenomenon that seems to be man made;
2) The ability to simulate future climate change with good models; and,
3) An appreciation of the economic policy implications of the above.

A decade or three ago, it was fashionable to dismiss climate change. This has become problematic for at least two reasons. One, the science is not that sophisticated. Certain industrial activity releases greenhouse gases. These gases - as the name suggests - work like a greenhouse and trap heat. That science is not exactly quantum entanglement and the data for greenhouse gas emissions and resultant warming seems to track pretty well to the theory. So Adams cedes this point and allows that climate change is probably real and ongoing.

Adams worked as a financial analyst at a bank, though, and challenges the second point: the ability to simulate future climate change. He said that it reminds him of the financial models he ran as an analyst that - should they reveal something his boss didn't like - could readily be changed with just a tweak of a few variables. Given we can't really forecast accurately what might happen, it is good to be skeptical, he says.

In this he has sort of put his finger on something really important and is sort of missing the point (probably intentionally).

What is really important is that systems define so much about what does or does not go well in our world - systems as varied as the economy and financial markets, energy systems, ecosystems and school systems - and yet we really don't understand system dynamics that well. Systems are tough to model and our models are not great. This is reason to be skeptical about any predictions but it also suggests that systems modeling deserves a massive infusion of research money. A crowd was gathered to watch a hot air balloon ascend and some woman said, "What is the use of all this new technology." Benjamin Franklin answered, "Madam, what is the use of a new born infant?" Systems simulation matters a great deal and is not that mature. Better to invest more heavily in it than to walk away from it. (And I think that computers' ability to simulate systems is maturing just when that capability is most needed for shaping policy dependent on such systems.)

And even with admitted limitations of models for any systems, it is worth asking whether even the models Adams was tweaking for his boss were all that bad. Once you understand a model for an economy or business, you articulate risks, a range of outcomes, and important variables. With good models you learn what factors they are most sensitive to (housing mortgages are sensitive to widespread economic downturns or refinancing from a drop in interest rates, for instance) and even spotty historical data can give you some sense of the probability of those events. (Yes. Nassim Taleb has rightfully pointed out that markets can be rocked by unpredictable events but risk mitigation can protect you from some of these rare events. A person who has saved three years of salary is better prepared for an event "they never could have predicted" than is someone with only three months of salary.)  Financial models are a little sketchy in prediction but there are ways to gauge their efficacy in spite of a large margin of error. (For instance, only about 20% of businesses succeed past 5 years. If your bank lending model assumes that is going to raise to 50%, it will probably be wrong; if it assumes that it will raise to 25% or drop to 15%, it could be right but done properly even that should require a coherent explanation that tracks to the numbers rather than arbitrary tweaks.) Further, to the extent that Adam's boss was unique in cheating the models so that they showed what he wanted, his bank would suffer. There is a drive to make models more accurate and - within the financial world at least - big rewards for such accuracy.

Models force questions and conversations about what variables matter and they bound reasonable outcomes. It is true that climate change models will be wrong but the simplest truth is pretty easy to predict: we will emit more greenhouse gases and temperatures will be higher than they would have been without these emissions. There are a host of unknowns that come with that (will particular regions benefit or lose, will changes in wind or sea currents result in unexpected cooling in certain regions, might unforeseen natural phenomenon or new technology absorb these gases, etc.) but the general story is known. If you invest in stocks over a 25 year period you can't be sure of when your portfolio will drop by half or raise 20% a year for successive years but you can reasonably guess that over your lifetime you'll be a better shape for having saved 10% of your income than not. Same with greenhouse gases; reducing emissions will drive less uncertainty, disruption and climate change.

Denying climate change is in a long tradition of denying scientific results like the health hazards of tobacco or the notion that we orbit the sun. Climate change deniers are traditionalists who conflate market economies with oil and gas and see an admission of climate change as a threat to those forces. (It does seem like climate change will threaten oil and gas. The possibility of oil and gas being displaced by alternative energy is not a refutation of markets that periodically unleash gales of creative destruction, though, but is instead an affirmation. Markets are no more dependent on oil than horses and markets don't treat fossil fuel industries as sacred.)

As to Adams' third point about evaluating the economics of climate change policy, I'll just say this. If it is inevitable that we'll adapt new energy technologies, there is less likely to be a penalty for rushing into creating and then converting to alternative energies than there is to be a penalty for delaying that change.

No comments: