In September 2009, the always thoughtful Society of Actuaries released a fine paper entitled “The Financial Crisis and Lessons for Insurers” by a talented team of authors (Robert W. Klein, Gang Ma, Eric R. Ulm, Shaun Wang, Xiangjing Wei, George Zanjani). We think that all readers of this blog should read the full paper, because it’s excellent and filled with wisdom. In just one respect, however, using an American football analogy, the paper leaves the ball just short of the end zone and fails to score a touchdown. This blog extols the virtues of the paper and tries to convert an additional 7 points from that touchdown.
The full text of this nice paper is available via this link to the Society of Actuaries website:
http://soa.org/files/pdf/research-2009-fin-crisis.pdf
The paper gets off to a very good start in my mind by laying out “home prices” as the number one cause of what’s been called by various names that the authors list: “subprime crisis, credit crisis, financial crisis.” This isn’t a credit crisis—the credit problems are “derivatives” from another fundamental cause. It’s a “home price crisis” as the authors note on page 4:
The authors correctly cite New York University’s Nouriel Roubini and Robert Shiller of Yale University as two people who warned of the home price bubble. My favorite early warning of home prices as a risk factor, however, is the December 10, 2003 “Loss Distribution Model” by my partner Robert A. Jarrow and co-authors for the Federal Deposit Insurance Corporation: In the appendix of the paper itself, three macro economic factors that drive the correlated default of U.S. banks are explicitly identified: home prices, interest rates and bank stock prices. For an introduction to the FDIC Loss Distribution Model, see this link to the FDIC website:
http://www.fdic.gov/bank/analytical/fyi/2003/121003fyi.html
Professor Jarrow recently noted that, even in 2003, early simulations showed what we now know with hindsight was true: the FDIC was heavily underfunded with respect to its deposit insurance obligations.
My other favorite “early warning” of the home price crisis is from the Economist, which published this graphic on June 16, 2005 showing that the home price bubble in the U.S., the United Kingdom, and Australia had far outstripped the bubble in Japanese home prices, which ultimately saw home prices drop 60% in major cities:
The full text of this fine article is available by subscription on this web page:
http://www.economist.com/printedition/displayStory.cfm?Story_id=4079027
Returning to the Society of Actuaries paper, we’re off to a great start. This is a home price crisis, and we’re focusing on the right ball.
On page 5 of the paper, the authors correctly note that traditional fixed income analytics haven’t been as helpful as one would have wanted in the crisis:
The authors elaborate on the shortcomings of traditional duration measures of risk later in the paper. Page 45 describes why effective duration measures fell short of the current crisis:
This is one of the areas where the authors fell short of the “end zone” that represents a score in American football. What they could have said, and almost said, was something like this:
“The current crisis makes it quite clear that securities values are affected by changes in many macro economic factors. When these factors change, they change the value of the security by causing changes in the default probability, credit spread, prepayment probabilities, and bid-offered spread on the security. ‘Duration’ and related measures of price changes (‘price elasticity’ to be more precise) stem from changes in one obvious macro factor: interest rates. What the current crisis reveals clearly is that analytics have to be supplemented with additional measures of price elasticity like these:
- The percentage change in the security’s price in response to an x% change in home prices
- The percentage change in the security’s price in response to an x% change in commercial real estate prices
- The percentage change in the security’s price in response to an x% change in oil prices
- The percentage change in the security’s price in response to an x% change in foreign exchange rates
- The percentage change in the security’s price in response to an x% change in the stock price index
and so on for other relevant macro factors.”
The single minded focus of most risk managers on interest rate risk to the exclusion of almost all other macro factors is one of the prime reasons why the market value of so many large institutions was devastated when the home price drop came about—they’d been operating with no direct calculations of their exposure to home prices and other macro factors. We also note that when we say “security” above, we are including every transaction on the balance sheet of insurance companies and other financial institutions, because as we’ve noted in other blogs these macro factors affect mortality rates and the supply of other liabilities to the institutions. We also note that stress testing default probabilities, credit spreads, and prepayment rates is no substitute for stress testing the macro factors directly. If we break the link between changes in the macro factor and the valuation of the security, we obscure the risk.
The Society of Actuaries paper continues on page 7 by noting that there is a need for improved early warning systems for the potential failure of regulated insurance firms:
When the authors return to this topic on page 72, they note that the current techniques used by insurance regulators are backward looking and dependent on financial ratios to the exclusion of almost all else:
Among U.S. bank regulators, for example, the Office of the Comptroller of the Currency recently announced that it has subscribed to daily updates of a default probability service that is driven not only by financial ratios but also by macro economic factors and returns in the common stock market. For details, contact us at info@kamakuraco.com. Using a similar approach would allow the National Association of Insurance Commissioners and other insurance regulators to follow the evolution of default probabilities over time for stock insurance companies. A similar methodology, which explicitly incorporates home prices and other macro factors, can be employed for mutual insurance companies. The example below shows the evolution of 1 year default probabilities (blue) and 5 year default probabilities (yellow) for MetLife:
MetLife clearly navigated the credit crisis much more successfully than AIG, Bear Stearns, Lehman Brothers, and Washington Mutual.
The authors note that the role of enterprise risk management has been growing in the insurance industry on page 55:
The Society of Actuaries authors turn next to the key role of enterprise risk management in the prevention of future crises in summarizing their conclusions on page 8:
On page 14 and again later on page 61 the authors come tantalizingly close to insisting on stress testing with respect to a long list of macro factors, rather than interest rates alone, but they fall short. They instead mention the symptoms of macro factor movements (ratings downgrades, failure of liquidity suppliers, increase in correlations in asset returns):
They come close again to mentioning the magic words “home prices” or “macro factors” on page 70, where they discuss “dynamic financial analysis,” in insurance industry lingo, known as stochastic multiperiod simulation to the rest of the world. They contrast this approach to “RBC,” risk based capital in the insurance industry:
They come tantalizingly close to saying this:
“The best approach to preventing future crises driven by macro economic factors like interest rates, home prices, commercial real estate prices, stock prices, oil prices and foreign exchange rates is stochastic, dynamic financial analysis and stress testing of those factors. This Monte Carlo simulation and related stress tests would link changes in macro factors directly to default probabilities, credit spreads, prepayment rates, bid-offered spreads and liability-side considerations like mortality rates and the probability of other insurance events.”
Ambac’s failure, for example, to quantify its probability of paying off on insurance contracts related to the default of mortgage backed securities and collateralized debt obligations is at the heart of the crisis there.
It was only on page 77 on macro factor stress tests that the authors mention “stress tests” and “a major contracting in the housing market” in adjacent sentences. We know that this was implicit in many of their comments noted above, but we would argue that, since home prices were correctly identified as the root cause of the crisis, stress testing and simulation of these macro factors and their impacts on the firm should have been the major theme of the paper. They were oh-so-close to saying this explicitly, but they didn’t quite get credit for a touchdown.
Finally, the authors on page 76 make the very appropriate point that reliance on ratings has greatly magnified the repercussions of the drop in home prices:
Another implication of a reliance on ratings is even more important: since the mechanism that connects movements in macro factors to changes in ratings and explicit default probabilities at distinct maturities has never been clearly articulated, the use of ratings makes the analysis that we describe above impossible to do correctly. That’s why ratings have to go. For more on how to do that, see our May 12, 2009 blog “A Ratings Neutral Investment Policy”
The six co-authors of “The Financial Crisis and Lessons for Insurers” have done a fine job for the Society of Actuaries. Our only real critique is the need to say more clearly, loudly and simply
“Risk analysis should have macro factors like home prices and other factors at the very heart of the enterprise wide risk management calculations. To do otherwise is to sow the seeds of the next crisis.”
Donald R. van Deventer
Kamakura Corporation
Honolulu, October 9, 2009