Lessons From the Japanese Earthquake

Transcription

1 Lessons From the Japanese Earthquake Why the U.S. Should Use International Reinsurance Markets By Ed Hochberg and François Morin The catastrophic March earthquake in Japan had many in the United States wondering how the country would handle a similar calamity. The international reinsurance market can relieve U.S. taxpayers of unintended financial costs tied to such a disaster. What if It Happened Here? On March 11, 2011, a magnitude 9.0 earthquake struck off the Pacific coast of Tōhoku, Japan. The earthquake and resulting tsunami caused thousands of human casualties, widespread damage and the worst nuclear accident in a generation. In the U.S., many wondered what would happen if a similar event were to occur, particularly in one of its earthquake-prone areas. It is likely that the human and economic toll from a similar event would be substantial. However, this article examines a less-well-understood aspect of a U.S. earthquake the amount of the economic risk that has been de facto transferred to the U.S. taxpayer. In the present environment of substantial federal deficits and ballooning debt, this article questions whether this hidden risk transfer is the most efficient and appropriate means of risk financing, given the low-frequency, high-severity nature of these events. Because most mortgagees do not require earthquake insurance, the current system presents a substantial downside risk to the economy. This risk is especially serious, given the potential impact on the banking system and the U.S. government s grappling with deficits, debt limits and a recent rating agency downgrade. If earthquake insurance were required to obtain a mortgage, similar to windstorm and other natural hazards, such economic risk could be mitigated, in combination with the appropriate participation from the international reinsurance markets. Japan and New Zealand Earthquakes In 2010 and 2011, several very expensive earthquake events affected the Pacific Ring of Fire, an area prone to earthquakes, underscoring some of the issues associated with this loss exposure. As presented in our Insights article, the Japanese earthquake caused an estimated $300 billion of economic loss. Because a relatively small percentage was insured and an even smaller percentage was insured in the international market, much of that exposure will ultimately be absorbed by the Japanese government and its people. According to a May 19, 2011 Reuters report, the Japanese GDP contracted 0.9% in the first quarter versus a median economic forecast of 0.5% contraction, and on an annualized basis the contraction was 3.7% versus the median forecast of 2.0%. Even though the earthquake occurred only three weeks before the end of the quarter, it is blamed for much of the Japanese economy s underperformance. Though the Nikkei 225 has recovered somewhat from this initial drop, it is still down approximately 6% from its pre-quake levels. In a country already struggling with major deficits and whose debt was recently downgraded by major rating agencies, this economic impact is clearly unwelcome. Because most mortgagees do not require earthquake insurance, the current system presents a substantial downside risk to the economy. Emphasis 2011/3 7

2 Perhaps most importantly, there is the recognition that a healthy economy requires a healthy banking system. Meanwhile, Christchurch, New Zealand, was the scene of three devastating earthquakes between September 2010 and June As noted in our Insights article, in contrast to Japan, a much greater percentage of the economic loss was insured and a greater percentage of this risk was transferred to the international markets. Although it is expected that the cumulative effect of these earthquakes will reduce the GDP growth rate by 1% to 1.5%, as reported by the International Business Times, New Zealand Taxation reports that New Zealand is not contemplating a special levy to pay for the damage. In fact, the NZX 50 has recovered most of what was lost following the February 2011 earthquake. Relative to the two countries economies, the Japanese and New Zealand earthquakes were both very significant events. The contrasting outcomes are attributable to the percentage of their population purchasing earthquake coverage and the use of international risk transfer markets in these economies. As a percentage of the economic loss, New Zealand used the risk transfer markets to a far greater extent than Japan. In fact, the absolute dollar amounts transferred may turn out to be about the same, even though Japan s economy is many times the size of New Zealand s. Interestingly, many economists predicted rebuilding activities in both countries would spur economic growth in subsequent quarters. However, Japan s deficit may have tempered some of that enthusiasm. With a relatively small contribution from international reinsurance markets, the vast majority of the resulting costs will very likely add to Japan s debt and taxpayer burden. It is a conventional view of Japan that while it has a very large economy, it is fairly closed to external capital providers for its overall risk transfer needs. In contrast, as a smaller economy, New Zealand has historically transferred a much more significant portion of this risk to international risk transfer markets. What may come as a surprise to many is that the U.S. may be far more like Japan than New Zealand in how it transfers its risk exposure from earthquakes. Private Sector Versus Public Sector Debate Generally, the U.S. tends to rely on the private sector to provide most risk capital. However, there is considerable debate about whether this approach is optimal for certain types of peak exposures. Discussions at the federal level have focused on the appropriate degree of government participation in the key area of flood insurance. In Florida and some other coastal states, there has been a lengthy discussion about the appropriateness of the states involvement with hurricane risk and whether post-event funding is the proper means by which to finance such risk. Without opining on the appropriateness of these risk financing methods, the current techniques are clearly different from the present handling of much of the U.S. earthquake exposure: By and large, there is an explicit requirement to insure these exposures (where applicable), especially in the residential markets. There is typically a risk premium charged for the exposure, although some will debate its adequacy. There is more reliance on the international risk transfer markets, rather than an implicit or explicit guarantee from the federal government, particularly for wind exposure. There are several important reasons why the risk financing mechanisms have these characteristics. Perhaps most importantly, there is the recognition that a healthy economy requires a healthy banking system. If these exposures were not insured, they would potentially put the banking industry at undue risk. In addition, failure to require insurance, when appropriate, creates an economic distortion encouraging people who are not exposed to a particular peril to effectively subsidize those who are exposed. We have these characteristics for most insurable perils in spite of the debate regarding windstorm and even flood insurance in places like Florida. 8 towerswatson.com

3 But even though earthquake peril is a major exposure to homeowners and businesses, it is not treated this way. Major modeling firms estimate billions of dollars in annual average loss, with catastrophic events expected to be in the hundreds of billions of dollars. For these reasons, it is all the more perplexing that mortgagees generally do not require earthquake insurance. Insurance is required for fire, hail, lightning and, of course, windstorm, yet it is not required for this major exposure. Consequently, earthquake insurance, particularly for residential exposures, has a very low take-up rate, and the banking system and the federal government absorb a significant amount of earthquake risk. California In order to illustrate this point, the most logical place to examine is California. From an economic standpoint, it is home to over half of the U.S. earthquake risk. In addition, residential mortgage loans in California are nonrecourse. In most cases, this means that a lender cannot pursue the assets of a defaulting mortgagor outside of the mortgaged property. In effect, the nonrecourse creates a put option, whereby a borrower could make an economic assessment that the value of his or her home is less than the mortgage and it is in his or her best economic interest to simply walk away from the mortgage. Many economists believe that, even when it may be in their best economic interest to put the house to the lender, most people choose not to for two reasons: The stigma of defaulting on a mortgage The impact on the defaulter s credit report However, researchers at Deutsche Bank did a study that concluded that the propensity to default increases significantly as the loan-to-value (LTV) ratio increases above 100%, particularly for nonrecourse states. In fact, once the LTV ratio exceeds 130%, Deutsche Bank estimated over 80% of homeowners in that situation would default on their mortgage. Many economists believe that the propensity to default on negative- mortgages is increased in situations of significant economic and emotional stress such as significant damage to a home and belongings from an earthquake. California s residential property values have been hit especially hard since the real-estate collapse. According to First American CoreLogic, although California s overall residential LTV ratio is approximately 81%, roughly 35% to 40% of residential mortgages are underwater. So it is reasonable to assume that significant uninsured damage, particularly where the homeowner does not have the means to fund major repairs, would further reduce the value of homes in an affected area. LTV ratios would be driven higher, dramatically increasing the propensity to default. It is not hard to fathom that mortgagees (e.g., banks, Fannie Mae, Freddie Mac) would have significant financial exposure to such a situation. California s present situation makes it much more troubling that the take-up rate for residential earthquake insurance in California is extremely low. According to the California Earthquake Authority (CEA), the take-up rate is approximately 12%. Several primary factors appear to be influencing the low take-up rate: The expense of the premium The perceived value of the policy (especially given deductibles and sub-limits on most policies) The length of time since the last major earthquake in the U.S., making the risk less tangible The fact that mortgagees do not require earthquake insurance The CEA also suggests that the take-up rate is even lower in some higher-risk areas due to the higher premium levels. In effect, most homeowners choose to self-insure this exposure, a risk financing method that is not usually optimal for low-frequency, highseverity exposures. So the residential mortgage market (and the mortgagees, by extension) faces a very significant uninsured exposure to earthquakes. Even though earthquake peril is a major exposure to homeowners and businesses, it is not treated this way. Emphasis 2011/3 9

4 Ed Hochberg Specializes in brokerage and reinsurance. Towers Watson, Philadelphia François Morin Specializes in property & casualty risk software and consulting. Towers Watson, Hartford California Case Study The following scenario offers context to what a major California earthquake could mean to the banking industry and, by extension, the U.S. government and taxpayers. This example is illustrative and is not a prediction of what would happen, but rather, what could happen. In fact, there are a lot of reasons to believe the impact could be potentially worse than what is described here. In this example, we select one particular event that approximates a one-in-250-year (0.4% occurrence exceedance probability) return period event within the Risk Management Solutions database. Our focus is on residential exposures because there is a more direct link to mortgage issues, but clearly commercial risks would also create exposures for banks. Within the approximate boundaries of secondary uncertainty, we assume the overall insured loss would be approximately $40 billion of loss to buildings (versus contents), of which $18 billion is attributable to residential exposures. However, that $18 billion figure is only the insured loss. Given the low take-up rate for residential earthquake insurance and the amount of insured co-participations such as deductibles, the uninsured loss could be closer to $210 billion, or about 92% of the total. Based upon the following assumptions, we can estimate the amount of potential loss to mortgagees: Approximately two-thirds of homeowners have mortgages. The overall LTV ratio is approximately 81%, but approximately 40% of the mortgages are in a negative- position (allowing for an approximate distribution of loans among LTV ratios). The decline in property value is equal to the uninsured building damage. Default rates are based upon the Deutsche Bank estimates. Figure 1 sets out the $386 billion of residential property values prior to the loss that becomes approximately $222 billion after the impact of the uninsured losses. Figure 2 shows that approximately $72 billion of aggregate pre-loss becomes $92 billion of aggregate negative. Importantly, approximately $156 billion worth of mortgages pre-loss with negative becomes $273 billion, a $117 billion increase. Furthermore, the vast majority of such negative mortgages become seriously negative (LTV ratio greater than 125%), which dramatically increases the propensity to default. In Figure 3, we compute the impact on mortgagees. The default probabilities are applied to the negative, which results in losses to mortgagees of approximately $97 billion. If you add in another 10% of the property values for transaction and holding costs, the mortgagees incur a total loss of about $112 billion. Based upon Federal Reserve data, approximately 50% of the mortgages are held or guaranteed by Fannie Mae or Freddie Mac. Since these entities are effectively insolvent, it is likely around $50 billion to $60 billion would become a direct obligation of U.S. taxpayers through additional capital infusions. Another 20% is held by commercial banks. While most large money center banks likely have a sufficient capital cushion to absorb their share of $20 billion to $25 billion, invariably, some smaller banks would become insolvent. Those insolvencies would add pressure to the Federal Deposit Insurance Corporation (FDIC) and potentially require taxpayer funds or increases in FDIC rates that would be passed on to depositors. Earthquake insurance, particularly for residential exposures, has a very low take-up rate, and the banking system and the federal government absorb a significant amount of earthquake risk. 10 towerswatson.com

5 Figure 1. Loss in market value of residential properties due to uninsured earthquake losses Loan-to-value ratio Beginning Percent of mortgages Estimated residential property value Implied loan amount Uninsured loss New residential property value () 85% or lower 59.0% % to 100% 11.0% % to 125% 12.0% Over 125% 18.0% Figure 2. Revised loan-to-value ratios of negative- mortgages Loan-to-value ratio Beginning Original New New LTV Value of mortgages with negative Percent with negative LTV of loans with negative 85% or lower % % 120.6% 86% to 100% 1.2 (16.8) 168.9% % 168.9% 101% to 125% (7.5) (27.6) 201.9% % 201.9% Over 125% (33.6) (63.0) 257.6% % 257.6% 72.1 (91.5) Figure 3. Potential loss to the federal government from defaulted mortgages Loan-to-value ratio Beginning Value of underwater mortgages LTV of negative loans Implied property value Negative Default probability (per Deutsche Bank) Direct loss on foreclosure sale Transaction and holding costs at 10% Total loss to mortgagee 85% or lower % 62.2 (12.9) 76% (9.7) (6.2) (15.9) 86% to 100% % 24.4 (16.9) 81% (13.7) (2.4) (16.1) 101% to 125% % 26.7 (27.2) 81% (22.1) (2.7) (24.7) Over 125% % 40.0 (63.0) 81% (51.2) (4.0) (55.2) (119.8) (96.7) (15.3) (112.1) Interpretation One could look at those figures in the context of the U.S. economy and budget, and be relatively unimpressed. Budget wonks throw around numbers that measure in the trillions. However, in the current environment, where the U.S. is struggling with budget deficits and a recent downgrade from a rating agency, an additional $50 billion to $75 billion of unanticipated cost would be unwelcome at best. That said, there are several arguments suggesting the impact would not be nearly as bad as anticipated: LTV ratios used in the analysis inherently ignored land values. The probability of this sort of an event is very low. The local economy tends to expand after a major loss, as infrastructure is reconstructed. The first point is technically true. However, we would argue several offsetting factors. An area beset by massive damage and mortgage default as described by this scenario would experience a vast increase in defaulted (and damaged) housing stock inventory, which would reduce property values to a point substantially worse than just the amount of Emphasis 2011/3 11

6 The imprecise nature of catastrophe modeling and the potential add-on of commercial exposure make it quite possible that the impact on mortgage holders could be far more substantial. uninsured damage. Along similar lines, this increase in vacant and damaged housing stock could easily result in far more significant holding and transaction costs than the 10% assumed in this example. It is very possible that the land values would not provide much of a buffer. While it is true that most models would suggest that there is only a 0.4% annual chance of an event like this one occurring, a review of probability needs to offer some context. Assuming statistical independence among years, there would be an approximate 10% chance of an event of this magnitude sometime in the next 25 years. For an event with a 1% annual probability of exceedance, there is about a 22% chance of such an event over the next 25 years. Of course, these probability estimates also assume a level of precision in catastrophe modeling that likely does not exist in real life (as discussed in more detail below). Lastly, conventional economic wisdom suggests that after an initial shock, the local economy will begin to expand at a faster rate. Economists Eduardo Cavallo and Ilan Noy have postulated that in developed countries, even major disasters are unlikely to affect economic growth in the long run. Prime examples of the resiliency of the modern economies are Japan after the Kobe earthquake and the U.S. after the Northridge earthquake. In both instances, the local economies appeared to grow above expectations very shortly after the event. While that may be true, empirical evidence is always difficult to interpret with events that, by their nature, are infrequent, with a resulting very small sample size. We would also argue that these events occurred in a very different economic time. Downgrades from rating agencies were not an issue for the U.S. or Japan, and the U.S. was not in the aftermath of a major banking crisis. In both instances, local and federal governments had the wherewithal to directly or indirectly help with the rebuilding and set the economies back on track in a reasonable period of time. Today, however, it is not clear that we would see the same sort of resiliency. Specifically, a major earthquake in California would now come at a time when the U.S. federal government is grappling with massive deficits and debt, and is also absorbing a major hit from mortgage losses. In contrast with 1994, California is in very difficult financial shape and may be unable to contribute to the recovery in the manner that would be required. And while major banks have recovered from 2008 and 2009 lows, they are still generally reeling from the recent financial crisis. To count on a quick economic recovery to prevent a related mortgage and banking crisis may not be prudent under present circumstances. In fact, there are many reasons to believe the actual impact of an event like this one could be far worse. For one thing, to suggest we know the exact probability of the size and economic impact of an event would ascribe a level of precision in catastrophe modeling that does not exist in reality. While very valuable, by their nature, these models estimate the impact of events that have not happened. For windstorm, the major modeling firms have been honing their estimation methods largely by looking at historical events. For earthquake, there are far fewer data points with which to work. It is very possible that either the frequency assumptions and/or severity assumptions could be worse than what are suggested here. Additionally, we have not calculated the impact of uninsured losses on commercial mortgagees. There would be a considerable impact, which would be additive to the residential exposure. The imprecise nature of catastrophe modeling and the potential add-on of commercial exposure make it quite possible that the impact on mortgage holders could be far more substantial. 12 towerswatson.com

7 Risk Transfer As we have discussed, earthquake insurance has a very low take-up rate, particularly in the residential sector, creating significant exposure to the mortgage industry and the federal government. In effect, there has been a conscious or unconscious decision to retain this exposure on the part of homeowners (who risk the in their home, to the extent they have any), mortgagees and the federal government (which ultimately stands behind the mortgagees). This decision to self-insure leaves homeowners, banks and the federal government with considerable exposure to earthquake risk. Generally, in risk financing, low-frequency, highseverity risks are prime candidates to be transferred when possible. It appears earthquakes are such an exposure since the main financial effects are near the tail of the distribution. If it were not possible to transfer such an exposure in a significant way, it would nullify this argument. However, the international reinsurance market alone has over $300 billion of capital, and potentially billions more could be accessed through the capital markets use of catastrophe securitization transactions. If the federal government were to implicitly or explicitly require the exposure to be insured (at least in earthquake-prone areas), it would have several significant advantages: It would mitigate the risk of a severe financial blow to the federal government at a time when it can arguably not easily afford it and when it could be most disruptive. It would provide a more explicit and potentially fair allocation of the cost of this risk than presently exists. Though there may be initial resistance to this requirement, as noted above, there are many reasons why treating earthquake differently from other major hazards does not appear to make much sense. Conclusion Many might like to think the U.S. would fare better financially than Japan if it suffered a major earthquake. Unfortunately, there are many reasons to believe that may not be the case. With such a low take-up rate for earthquake insurance, homeowners in the U.S. have not availed themselves of the capital of the international reinsurance market to finance such a high-severity, low-frequency risk. In effect, both the federal government and U.S. taxpayers are unwittingly absorbing that risk, in the event of negative impact on mortgagees, led by Fannie Mae and Freddie Mac. Given this potential risk, it is very difficult to understand why earthquakes essentially get a pass compared with other hazards that potentially threaten property that is accepted as collateral by U.S. banks and mortgagees. For comments or questions, call or Ed Hochberg at , ed.hochberg@towerswatson.com; or François Morin at , francois.morin@towerswatson.com. With such a low take-up rate for earthquake insurance, homeowners in the U.S. have not availed themselves of the capital of the international reinsurance market to finance such a highseverity, low-frequency risk. Emphasis 2011/3 13