Monday, September 19, 2005
Natural Disasters: Risk Management and Incentives
Another natural disaster, another round of recriminations. When the tsunami hit southeast Asia last December, the talk was all about the absence of early-detection systems and/or effective communication between various entities, leading to a complete lack of warning given to those living in the danger zone. When Mumbai was lashed by rains in July, the knives were out for the authorities who failed to recognize how overloaded the city’s power and transport infrastructure was prior to the event, and how vulnerable it would be to an external disruption. And now, in the aftermath of Katrina, there’s been an orgy of finger-pointing and I-told-you-so-ing from all sides of the political spectrum. The Economist summarizes the situation well:
All these criticisms are fine and dandy when made with hindsight, but they miss a fundamental point: democratic societies can never be fully prepared for extreme events. This statement may seem like a reckless generalization, but I believe I can back it up. Read on.
Let’s start by putting the situation into a risk-management framework. It’s axiomatic that in a given situation, and without infinite resources, you cannot guard against each and every negative event. So there’s a tradeoff; you have to choose which events you want to guard against, and which ones you’ll leave ‘unhedged’. How do you make this choice? The usual method used in financial risk management is to estimate the loss arising from a particular event, multiply it by the probability of that event, and then compare that figure with the cost of guarding against that event.
A numerical example might make this clearer. Consider two events: event A leads to a loss of $100, while event B leads to a loss of $20. But there’s only a 1% chance that event A will occur, while event B has a 10% chance of occurring. The ‘expected value’ for A is thus $100 * 1% = $1 (loss), while the E.V. for B is $20 * 10% = $2 (loss). If it costs the same amount of money to prevent A and B, then clearly you’re better off guarding against B. On the other hand if B costs $1 to prevent while A costs $0.01 to prevent, then (assuming perfect scalability) you should be guarding against A.
Now, a natural disaster like Katrina is a perfect instance of a low-probability, high-loss event - call it a type A event. But the risk management framework above should be perfectly capable of including such events in its calculations. (One could argue, of course, that the authorities grossly underestimated the ‘true’ probability of massive damage to New Orleans from a hurricane, but that’s a highly posterior argument). The fault in the system is subtler.
The problem is that the decision of which types of events (A or B) to guard against is made by elected officials. And elected officials do not have the same incentives as society as a whole. For an elected official, the payoff structure is strictly binary: either you get re-elected, or you don’t. As an elected official, you can guard against a type A event (like a hurricane), and on the extremely rare (say 1 in 1000) chance of the event happening, you’re a prescient genius. Or you can spend the same amount of money on schools or hospitals or tax cuts, which may have a worse E.V. for society as a whole, but will, 999 times out of 1000, result in your re-election. Which would you choose?
And there you have it. Politicians will always prefer a small chance of a huge loss, to the certainty of a small loss - even though in E.V. terms such a choice might be sub-optimal for their constituents. And the incompetence on display after Katrina, or for that matter after the tsunami, was thus not the product of political or bureaucratic misfeasance or malfeasance, easy as that explanation would be. No; it was merely the consequence of systematic under-preparation for low-probability outcomes, by rational politicians with short-term electoral goals. A triumph for the dismal science!
Pundits explained the government's failure in every way they pleased. Anti-war types blamed Iraq, particularly the fact that thousands of National Guard troops had been sent there. Environmental types blamed Mr Bush's lackadaisical attitude to wetlands. Many Democrats saw it as proof that Mr Bush and the Republicans cared nothing for America's poor and black. Liberals argued that Katrina showed why, as James Galbraith, a vocal leftist economist at the University of Texas, put it, the “government of the United States must be big, demanding, ambitious and expensive.” A Wall Street Journal column, in contrast, argued that the hurricane showed the danger of relying too heavily on inefficient government.
All these criticisms are fine and dandy when made with hindsight, but they miss a fundamental point: democratic societies can never be fully prepared for extreme events. This statement may seem like a reckless generalization, but I believe I can back it up. Read on.
Let’s start by putting the situation into a risk-management framework. It’s axiomatic that in a given situation, and without infinite resources, you cannot guard against each and every negative event. So there’s a tradeoff; you have to choose which events you want to guard against, and which ones you’ll leave ‘unhedged’. How do you make this choice? The usual method used in financial risk management is to estimate the loss arising from a particular event, multiply it by the probability of that event, and then compare that figure with the cost of guarding against that event.
A numerical example might make this clearer. Consider two events: event A leads to a loss of $100, while event B leads to a loss of $20. But there’s only a 1% chance that event A will occur, while event B has a 10% chance of occurring. The ‘expected value’ for A is thus $100 * 1% = $1 (loss), while the E.V. for B is $20 * 10% = $2 (loss). If it costs the same amount of money to prevent A and B, then clearly you’re better off guarding against B. On the other hand if B costs $1 to prevent while A costs $0.01 to prevent, then (assuming perfect scalability) you should be guarding against A.
Now, a natural disaster like Katrina is a perfect instance of a low-probability, high-loss event - call it a type A event. But the risk management framework above should be perfectly capable of including such events in its calculations. (One could argue, of course, that the authorities grossly underestimated the ‘true’ probability of massive damage to New Orleans from a hurricane, but that’s a highly posterior argument). The fault in the system is subtler.
The problem is that the decision of which types of events (A or B) to guard against is made by elected officials. And elected officials do not have the same incentives as society as a whole. For an elected official, the payoff structure is strictly binary: either you get re-elected, or you don’t. As an elected official, you can guard against a type A event (like a hurricane), and on the extremely rare (say 1 in 1000) chance of the event happening, you’re a prescient genius. Or you can spend the same amount of money on schools or hospitals or tax cuts, which may have a worse E.V. for society as a whole, but will, 999 times out of 1000, result in your re-election. Which would you choose?
And there you have it. Politicians will always prefer a small chance of a huge loss, to the certainty of a small loss - even though in E.V. terms such a choice might be sub-optimal for their constituents. And the incompetence on display after Katrina, or for that matter after the tsunami, was thus not the product of political or bureaucratic misfeasance or malfeasance, easy as that explanation would be. No; it was merely the consequence of systematic under-preparation for low-probability outcomes, by rational politicians with short-term electoral goals. A triumph for the dismal science!
