When Insurance Fails - Reflecting on AIG

This Planet Money episode on homeowners insurance reminded me of when I first learned about insurance market failures in college. Insurance markets tend to fail when at least one of three factors is present:

• Moral hazard The insured takes on more risk or incurs costly activities because the insurer is going to pick up the tab. Phrases that I have heard, half serious and half in jest: "Sure I can go skiing! I have hit my max deductible on my health insurance." "I don't mind if my glasses break. Insurance covers another pair this year anyway." "I need to make sure I get my doctor checkups, eyes checked, and dental cleaning. I don't need it, but I'll pay loads more when I switch plans next month."
• Adverse selection A narrow market emerges because the insured possesses much more information than the insurer. My wife and I learned about this in 2008 while evaluating insurance for pregnancy coverage. When we looked at the cost of coverage we discovered that because of the premiums and caps, on an NPV basis the coverage made sense to purchase only if my wife became pregnant within the next six months. A baby was not in the immediate plan, but the insurers did not know that. While the point of insurance is to transfer the risk for a price, a six month breakeven NPV is far outside the range of most people's risk appetites. This limits the number of people applying for coverage to the truly risk averse and those who plan on getting pregnant in the very near term.
• Correlated risks Large-scale meteor strikes, war, and 1918 Spanish flu-style outbreaks do not have insurance products. Either the event never occurs and the insurer walks away with the money or the event occurs and the insurer must pays off everyone at the same time and the insurer does not have sufficient assets to pay out. Accordingly, neither a rational buyer nor the rational seller should enter into the contract.

As I reflected on the podcast and some long ago college lectures, I couldn't help but see the parallels with this and AIG's Financial Products division. AIG wrote credit default swaps to insure collateralized debt obligations. In layman's terms, if a pool of mortgages started to default above a certain rate, AIG would be forced to pay the losses. Companies would approach AIG and offer to pay premiums if AIG were willing to assume the risk. To some degree I see all three forms:

• Moral hazard Anything that could make it through the AIG filters would be pushed through. If the price is identical, the first CDOs in line for credit default swap are the lowest quality relative to the perceived risk. 
• Adverse selection In late 2006, I spoke with a regional mortgage broker who referred to a section of his loan portfolio as "hot potatoes." All the due diligence boxes were checked, but because he was closer to the market, he was more familiar with the underlying risk than the standard scoring mechanisms would show. He was in a hot market, and he knew that the buyers who purchased, rebundled, and securitized would be none the wiser. One fascinating fact here according to a comprehensive FDIC report is that while it is true that, "the loans small lenders sold into the secondary market had both higher default risk and higher prepayment risk," the same was not true for the larger banks. Fortunately for AIG, the larger banks kept the loans with higher default risk, but lower prepayment risk. The data on adverse selection in the collateralized debt obligation market is a mixed bag and AIG ended up lucking out as it could have been much worse.
• Correlated risks While I recognize that many believed that real estate movements were not correlated on a national level because historically that was not the case in the United States, nationally correlated real estate prices had occurred at different intervals across many countries over the last fifty years. A friend once told me of a financial risk conference he attended in 2006, and while "across the board decline in real estate prices" was on the board of one of the primary presentations, it was listed in the "very low probability" bucket. While hindsight is 20-20, this was considered a known risk, and insuring it on this scale is like betting all-in against a pop star becoming addicted to cocaine. Sure, it hasn't happened yet to this person, but are you ready to bet the company, the largest insurer in the world, on it not happening?

While adverse selection and moral hazard can run losses for a company, if there are no correlated risks and the ramp up in underwriting volume is not large, the losses will start to pile up and corrections will be made before it gets too bad. Correlated risks, however, were the doozy that sank the company.