“The essential problem is that our models – both risk models and econometric models – as complex as they have become, are still too simple to capture the full array of governing variables that drive global economic reality.[...] But risk management can never reach perfection. It will eventually fail and a disturbing reality will be laid bare, prompting an unexpected and sharp discontinuous response..” Alan Greenspan, former Governor of the US Federal Reserve, writing in the "Opinion" column of the FTMarch 16 2008.
The current financial meltdown has exposed the excessive risk-taking behavior of banks and investment firms over the last decade in the United States and much of the Western World. This has led to much soul-searching. A recent NYT article ( “Risk Mismanagement” by Joe Nocera, New York Times, January 4 2009) argues that the meltdown can be blamed on the way financial sector firms (and regulators) calculated exposure. Most relied on a class of models called VaR — Value at Risk. These very complex mathematical models express risk over an entire portfolio as a single number. For example, a $50 million weekly VAR means there is a 99% chance the portfolio will not lose more than $50 million over the course of the coming week.
The problem with VaR is that it assumes a normal market. And it says nothing about what would happen in the remaining 1% -- the tail of the distribution. Some argue that VaR is a poor risk-management tool, even a dangerous one if its use creates a false sense of security. Others, however, argue that there is much to be said for a measure that captures 99% of the risk. And still others claim that models cannot be blamed -- only the naïve use that was made of them. This group maintains that VaR models are useful when combined with professional expertise and good judgment. Indeed, Goldman Sachs--the one firm that came out relatively unscathed from the meltdown--used both mathematical models and good judgment to pull out early from the more toxic markets.
So what does this have to do with climate change? Well, here again we have a situation where economists (and their clients) want a quantification of risk. This is needed because decisions to abate climate change have costs -- potentially high costs. And most integrated assessment models (IAMs which are the climate change economists’ equivalent to the financiers’ VaRs) show that the potential cost of climate change (the “value at risk”) is relatively modest in present discounted value terms. The optimal path, from this perspective, may be to abate modestly now and increase efforts later when the world is (hopefully) richer and has access to better technology.
The problem of course is that most IAMs suffer from the same limitations as the VAR. They typically don’t go into the “tails” of the distribution, where catastrophes potentially lie. Large, irreversible catastrophes of the kind we had not even envisaged 15 years ago such as the shutting down of the North Atlantic current, which could bring a new ice age to Northern Europe and trigger untold disasters.
More generally, if we agree with Alan Greenspan’s view that “our models – both risk models and econometric models – as complex as they have become, are still too simple to capture the full array of governing variables that drive global economic reality.[...]” what should we think of the limitations of models that hope to add the full array of variables governing climate change and associated physical impacts?