Search This Blog

Thursday, September 13, 2012

Climate and Financial "Models" - The New Pornography?

Yesterday I wrote about pornography and climate change.

One of the often cited reasons that we should all "believe" is that the climate models tell us things are our fault.  (This aside from the fact that the mere fact climate disaster is presented in video form is suspect.)

Climate models are computer programs into which people put all the data they have as well as various computational rules to apply to that data.  The model then produces "predictions" based on the data.

These models are always present to the great unwashed as "supremely accurate" in their predictions.

But are they?

It turns out they are not as accurate as we might think or believe.

First off there are many climate models, in 2007 as many as twenty or more.

Secondly, and more importantly, there are questions being raised about the accuracy of these models (see this as an example).

From this site for example:


Now these types of plots showing the discrepancy between the model's prediction and reality have been around for some time.

Yesterday I found this interesting quote: "...that models can capture all necessary information and then accurately predict future risk."

However, this is not from a climate discussion but from a financial discussion (see this).

It turns out that government uses models for other things besides the climate.  In this case predicting the soundness of financial institutions.

In a discussion about the Basil II banking standards designed to protect investors and the public from bank failures the notion of "models" is discussed: "A deadly flaw in financial regulation is the assumption that a few years or even a few decades of market data can allow models to accurately predict worst-case scenarios. The authors suggest that hundreds or even a thousand years of data might be needed before we could trust the Basel machinery."

Now financial markets are not as complex and planetary weather.  After all, at least theoretically one could capture (in today's computer driven trading world) every transaction associated with a given market.

However, my guess is that government regulators cannot manage this amount of data, nor can industry provide it.

So instead, like climate scientists, they use a model.

A flawed model.

Why is it flawed?  Who knows.

But the issue is that the modeling process is a failure: To wit, the investment bank Bear Stearns, which was in compliance with all regulatory models, failed during the financial crisis of 2008.

Now a financial system has a finite number of data points, unlike a climate system which has an infinite number.  Similarly since all the trading transactions occur on computers one imagines that there is a complete set of written rules to describe that system.

A planetary climate system, on the other hand, has an unmeasurable number of interactive rules and elements: plants, animals, humans, the sun, the orbit of the planet, gravity, the makeup of the planet, the atmosphere, solar flares, and on and on.

And these things do not occur in one spot where they can be accurately measured.  Instead there are some 510 million square kilometers of earth surface to think about and measure.

So even measuring each thing once per square kilometer requires 510 million measurements.

Take each measurement once per hour and you have 12 billion measurements per day.

But we don't have sensors all over the ocean, for example, so we can't accurately do even this much.

So we have to estimate.

No doubt like they have to do on the financial side...

And things break down.

But today's folks get locked into "believing in the model" - whether it works or not apparently.


Like pornography the model creates a focus point where believers can find comfort.  It takes their minds away from the reality that have a real scientific theory of weather with measurable prediction is well beyond what humanity is capable of at this point.

But models, like pornography, appear to be nice substitutes for the "real thing" that are easy for people to believe in.

The only problem would appear to be that they don't always work...

No comments:

Post a Comment