One of the current debates going in finance is about models and the role they played in the current problems ie. CDO disasters and ratings flaws. To paraphrase a popular expression, models don't shoot holes in balance sheets, overly confident bankers shoot holes in balance sheets.
The models are merely one of the tools used to establish the false beliefs (over confidences) that collectively lead to many financial actors taking bad actions.
Many have debated that we should get away from "fancy" risk models and others argue that if people just understood more about how to use or not use models we would be OK. Unfortunately the debate is a little stale, so I will throw in my bit. With these 2 fundamental rules:
- Models aren't dangerous, misapplied models are dangerous.
- The likely misapplication of a model on a large scale is a function of a groups confidence level.
Financial knowledge is not a linear thing, it is like chess. You learn it in levels and at each level the skills required change and the understanding of the game changes. Consider the following levels of knowledge in chess:
- What is the objective of the game?
- Which way does the bishop move?
- What is castling?
- How many points is a rook worth?
- What is En Passent
- What is the Spanish opening?
- What is it like to "see the board" like a grand master
Each level of knowledge has different rules and new ways to appreciate the game. The application of math in finance and risk is the same way. The difference is that instead of losing the game, if one overly relies on a piece of knowledge it can be disastrous. Over confidence kills and a little bit of knowledge can lead to over confidence.
In formal learning there is a taxonomy of knowledge known as Blooms Taxonomy. I submit the following diagram below to show an example of Bloom's taxonomy laid onto the learning of mathematical finance.
Each practitioner will proceed up the chain with a piece of knowledge such as how Value at Risk VaR for example works until they have either mastery or have reached their personal peak capability for comprehension. We can't all be Gary Kasparov.
Now for the Dangerous part. There comes a point in our learning when we believe we understand enough to act on our knowledge. Confidence can be a very dangerous thing in finance or knowledge as it may lead to the application of knowledge in a flawed way. Belief may rush ahead of skill. The graph below highlights the case for the individual financial modeling practitioner.
As one progresses along from ignorance(1) to mastery (6) our ability screw things up is not a flat line. In the beginning a person wouldn't act on a financial model or approach out of ignorance or unfamiliarity. As we learn a few things we get confident enough "to give it a shot". Ideally people learn the limitation of a model or approach and move on.
VaR for example stops being the Silver bullet of market risk once one appreciates the flaw of averages, extreme leptokurtosis of financial data etc. or the whole silliness of a the EMH and most modern finance in general.
This mismatch of confidence and application occurs on the group level as well. Consider a firm engaging in a new initiative (FAD) such as CDS's, Value at Risk or moving into the hot new Phoenix condo market. They will move slowly at first until a few $'s hit the balance sheet and the organization runs into the new goldmine and no one wants to get between the rapacious banker and their bonus. Sr. management definitely doesn't want to kill the new golden goose. Organizationally groups go through the same phases.
Now for the final point. An entire industry can go through these phases which causes huge potential for disaster as the industry "learns" its way forward. If you imagine the individual actors (banks) and an industry going through an evolution and were to plot the to hills as 2 axis other you may get something like the chart below.
The failure of acting without full competence or knowledge now becomes a much larger issue. CDSs were a classic case of this. My guess is that 1-3% of the participants new about the convexity risk inherent, but the industry was right in the middle of the learning curve and confidence was extremely high as mark to market balance sheets made many confuse skill (applied knowledge) with luck.
I mention this because of a personal concern with Basel 2. Individual firms will each be acting and learning how apply and game a new unified risk approach. They will be doing this more or less simultaneously throughout the world. As each bank starts applying its "skill" the Basel 2 framework will get more tightly coupled.
My concern is that this tightly coupled risk framework will amplify systemic risk creating something that no single participant (central bank) is experienced enough to understand much less control.
A banking system filled with ignorant participants with low confidence is safer than a group of highly confident mediocre participants. Ignorance and fear, keeps many from acting to boldly.
As an anthropologist, I view money as a social phenomenon, a belief system that sometimes gets ahead of itself. Understanding behavior, overconfidence and group motivations of various actors in finance can be a helpful tool in the world of money.