Next project? CDOs squared.

Photographer: Rae Russel/Getty Images

Crisis of Confidence in Financial Engineering

Noah Smith is a Bloomberg View columnist. He was an assistant professor of finance at Stony Brook University, and he blogs at Noahpinion.
Read More.
a | A

Lots of people say that economics was discredited by the financial crisis. That has a grain of truth to it -- most of the leading macroeconomics theories were more or less discredited by the Great Recession. But economics is much more than just macro, and the rest of the profession has actually been increasing in credibility as it gets better data. 

If you really want to see a field that took a hit from 2008, look at financial engineering. 

Lots of people think that financial engineering is part of economics. That’s not really true. Financial engineering is usually taught in business schools, or occasionally in applied math departments. Although some financial-engineering models have won the Nobel Prize in economics, the two disciplines really do different things. Econ usually focuses on individual decision-making, and individual preferences take center stage. Financial products such as stocks or bonds are treated as just another kind of commodity that people want to buy, like bread or video games. Financial engineering, on the other hand, focuses on purely mathematical or statistical relationships between different financial products. 

For example, a financial economist might make models of stock prices or interest rates, based on assumptions about investors’ risk aversion and on various macroeconomic processes such as business cycles and inflation. A financial engineer, on the other hand, might try to determine the right price for options, taking stock prices and interest rates as a given. Instead of human behavior, these models typically rely on the assumption of efficient markets -- the idea that although prices might deviate from the model for a while, they will eventually come back to the “right” level. 

The difference is important. Economics, with its focus on human behavior, usually doesn’t do a very good job of explaining financial markets -- human behavior, after all, is a very hard thing to model. But economists realize this, and so they don’t put too much faith in their models -- and neither do people in the private sector. The models are loose explanations, suggestions of possible mechanisms that might be at work in the markets. You wouldn’t place trades based on an economic model -- or if you did, you would make sure the bet was very small. 

The mathematical models of financial engineering, however, are expected to hold with much greater precision. For decades, traders placed trades -- big, highly leveraged ones -- that market prices would move toward the level implied by derivative pricing models. 

Those bets paid off…until they didn’t. Until 2008, the most famous example of financial engineering spectacularly failing was Long-Term Capital Management, the titanic hedge fund that blew up and had to be bailed out in 1998. LTCM’s traders made models of the proper relationships between bond prices, and bet on these models with huge amounts of leverage. The models worked like a dream before  they suddenly and catastrophically failed, driving the firm into a death spiral. 

Of course, the failure of LTCM was a small hiccup compared with the disaster in 2008. The U.S. financial system created enormous amounts of derivative assets -- the CDOs and CDSs that you read about in the news -- all based, directly or indirectly, on housing prices. When housing prices didn’t obey the statistical relationships that the models relied on, the value of those assets became impossible to estimate. So they crashed, taking down our big banks. 

Why did financial engineering models fail so disastrously? Essentially, people had too much confidence in them. Whereas the private sector basically didn’t trust economists’ models at all, traders trusted financial engineers’ models so much that they were willing to bet that they would hold true down to razor-thin precision. 

Since then, people have called for less leverage in the financial system. Leverage amplifies bets, so this is essentially just a call to hedge all bets based on financial engineering models. 

But how much to hedge? There’s no clear answer. Many financial engineers have tried to correct their models by increasing the number of tail events -- huge, unusual losses -- that might be expected to occur. The problem is that these models are very difficult to measure against the data. Tail events are, by definition, very rare, and financial history doesn’t contain enough of them to give us a good idea of how often they happen. 

Frustrated financial engineers have suggested more qualitative approaches. In his 2012 book, "Models Behaving Badly," pioneering financial engineer Emanuel Derman suggests that we treat all models as approximations, and use judgment to decide how much to believe in each. Andrew Lo, director of Massachusetts Institute of Technology’s Laboratory for Financial Engineering, says that mathematical models will always be plagued by some amount of “uncertainty” -- statistics slang for randomness that can’t be measured with a probability distribution. 

What this really means is that everyone should simply put much less faith in financial engineering. Leverage should be reduced, and making bets on derivatives should simply become a much, much smaller part of what the financial industry does. 

So it wasn’t economics that took the biggest hit from the crisis. Financial engineering, once revered and relied upon by many of the smartest people on the planet, has been humbled.

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

To contact the author of this story:
Noah Smith at nsmith150@bloomberg.net

To contact the editor responsible for this story:
James Greiff at jgreiff@bloomberg.net