World’s first model-driven recession
Blog: Bridgeland and Zahavi on Business Modeling
We are in the midst of the world’s first recession caused by modeling, a model-driven recession.
We are certainly in the middle of a recession, a declining economy. Some think this downturn is the beginning of a depression, comparable to the depressions of the 1930s or the 1870s. Others think this will be more like the typical post-WWII recessions, lasting 1 or 2 years.
Every recession is different, with different causes, different dynamics, and different outcomes. For example, the 1990-1991 recession was triggered by the oil price shocks of the first Gulf War, and by the failure of 747 savings and loan associations. (I oversimplify. At the risk of stating the obvious, the American economy is enormously complex, the most complex human-constructed artifact in world history. There are no single causes for anything.) How did today’s recession happen? This recession started with the big banks and other financial institutions in New York.
In the course of their everyday business, big banks buy securities and take other financial positions. All of these positions carry risks, risks that the positions will decline in value, become worthless, or even lead to large liabilities. Banks attempt to manage their risks, both by limiting the positions they take and by hedging their positions with other positions, positions that will gain value when the former positions lose it.
Starting in the mid-1990s many banks began to use a new form of model to manage their risks: value at risk. VaR models attempted to measure the riskiness of a position in terms of a 1% worst-case. A VaR value of $1 million means that there is a 1% chance of losing more than $1 million on a particular position. VaR values are computed both for individual positions and for portfolios of thousands of positions taken together. Banks used VaR models to determine how much risk they were taking, and how they should hedge that risk. (The business journalist Joe Nocera explains the history behind value at risk, how the banks adopted these models, and what happened next.)
Unfortunately VaR models are fundamentally flawed. There are several problems with value at risk, but the biggest problem is the implicit statistical assumption that financial markets behave as normal distributions. The normal distribution is of course just the good old bell curve from elementary statistics, an accurate model for some phenomena: for example, men’s heights. But normal distributions are known to be a poor model for the price fluctuations of financial instruments. Extreme movements are much more likely on Wall Street than in a normal distribution. (Power law distributions are a better model of financial markets.)
So VaR works fine, except when the markets change quickly. David Einhorn compares VaR to “an airbag that works all the time, except when you have a car accident”.
The consequence of banks using VaR models to manage risk is that they systematically underestimated the risk of big losses. They saw their portfolios as substantially less risky than they turned out to be. Based on the model-driven overoptimistic assessment of the portfolios, the banks acquired more risky securities than they otherwise would have. Then when things turned, everything declined quickly. Once a bank started to experience a larger loss than it thought likely under its VaR models, it sold the assets, leading to lower prices for those assets, and bigger losses for everyone else who held those assets, or similar ones. The losses fed on themselves.
To some extent, this is a typical pattern for a speculative bubble and its subsequent burst. The railroad bubble of the 1870s and the following Panic of 1873 followed essentially the same steps. People were overoptimistic about the value of the railroads that many were building. Then when things turned, everyone tried to sell at the same time, prices declined, and panic ensured. But the bad models that drove the Panic of 1873 were merely in people’s heads; they were poor mental models, not poor computer models. Today’s recession is the first caused by computer models.
We business modelers can take some small comfort here. It was not business models that drove this recession. VaR is a model discipline of capital market valuation. VaR is about the chance that a given security will decline in value. VaR is not about models of corporate goals or processes or business rules, not the kind of business models we describe in the book. Their models were at fault, not ours.
And if their models got us into this fix, maybe our models have a role in getting us out. Phil Gilbert argues that we need to improve the productivity of white collar work to turn the corner on this recession. I think he is right. Historically the greatest increases in productivity have come during recessions. When times are good, everyone focuses on meeting the topline numbers, on selling more and then somehow delivering on what has been sold. Costs and productivity are less important. When times are bad, everyone focuses on the bottomline numbers, on somehow generating more earnings from the meager sales they are able to make. Costs and productivity rule. And since white collar work is vastly larger than blue collar work these days, increasing the productivity of white collar work is essential to improving overall productivity.
We can use business modeling to increase the productivity of white collar work, so fewer people create more value. For example, when business process models are executed in a BPMS, the work can be done faster, better, and more consistently. Our book is about the business value of business modeling, and much of that business value show up as white collar productivity.
In addition to improving white collar productivity, Phil also talks about increasing the transparency of white collar work. He says: “There needs to be a revolution in implementation of processes that bring greater visibility and less risk to all aspects of our businesses. It is no longer acceptable that senior management remain ignorant of the goings-on at even the deepest depths of the organization.” This identification of visibility and risk is new to me, but perhaps it should not be. We all accept the importance of visibility in the public sector, as the premier means of preventing corruption. In fact the most influential anti-corruption NGO is called Transparency International—their very name reflects the link between visibility and the biggest public sector risk, corruption.
The growth of systemic risk in Wall Street over the last 20 years is not the same as public sector corruption of course. In fact, it is closer to incompetence than corruption. As Michael Lewis says “[Meredith Whitney] just expressed most clearly and loudly a view that was, in retrospect, far more seditious to the financial order than, say, Eliot Spitzer’s campaign against Wall Street corruption. If mere scandal could have destroyed the big Wall Street investment banks, they’d have vanished long ago. This woman wasn’t saying that Wall Street bankers were corrupt. She was saying they were stupid. These people whose job it was to allocate capital apparently didn’t even know how to manage their own.”
But there is a similarity between this model-driven stupidity about risk and simple corruption. Both rely on secrecy. Corruption relies on secret payments from (for example) a favored firm to a government official. The big banks relied on secrecy about the risks they were taking to achieve the profits they reported. If the risks had been visible to everyone—shareholders, other lenders, analysts, regulators—the banks might have avoided this mess.
Perhaps the mess was unavoidable, even with better visibility. VaR might have driven the banks to make the same blunders, and just made those blunders more rapidly obvious to everyone outside. The increased visibility from business modeling can still help us grow the rest of the economy again. Most of the economy is outside Wall Street, outside of the financial service industry entirely. Those businesses need better visibility of their business processes, the business rules they are using, and the goals and objectives they are trying to achieve. Our business models can deliver that kind of visibility. With that visibility we can turn the corner on this model-driven recession.