The Safety of State Bonds: A Historical Perspective

By Marc Joffe

The last state general obligation bond default occurred in 1933. Yet many state GOs yield significantly more than US Treasury bonds, reflecting investor fears of future defaults. While past results are no guarantee of future performance, history does offer investors valuable insights into our present situation.

As Table 1 shows, state and territorial bond defaults were relatively frequent during the 19th century.

Table 1. List of State Bond Defaults.

State	Default	Cure	Source	Notes
Alabama	c. 1870		New York Times (1930)
Arkansas	1/1841	7/1869	English (1996)	Some bonds repudiated in 1884
Arkansas	3/1/1933	1941	KBRA (2011)
Florida	1/1841	2/1842	English (1996)	Territory; Debt repudiated
Florida	c. 1870		Ratchford (1941)
Georgia	c. 1870		New York Times (1930)
Illiniois	1/1842	7/1846	English (1996)
Indiana	7/1841	7/1847	English (1996)
Louisana	2/1843	1844	English (1996)	Some bonds repudiated
Louisiana	3/1/1933	5/20/1933	KBRA (2011)	Due to failure of Hibernia Bank
Maryland	1/1842	1/1848	English (1996)
Michigan	7/1841	7/1849	English (1996)	Some bonds redeemed at 30%
Minnesota	1860		Ratchford (1941)
Mississippi	3/1841	11/1852	English (1996)	Mostly repudiated
Missouri	c. 1870		Smythe (1904)
North Carolina	c. 1870		New York Times (1930)
Pennsylvania	8/1842	2/1845	English (1996)
South Carolina	c. 1870		New York Times (1930)
South Carolina	7/15/1932		KBRA (2011)	No loss of principal or interest; maturing bonds were redeemed with new bonds rather than cash.
Tennessee	c. 1870		Ratchford (1941)
Texas	c. 1930		KBRA (2011)	Interest and principal not remitted to certain state-controlled funds; individual investors do not appear to have been impacted.
Virginia	c. 1870		Ratchford (1941)
West Virginia	c. 1870		New York Times (1930)

Sources:

English, W. B. (1996). Understanding the Costs of Sovereign Default: American State Debts in the 1840's. The American Economic Review, 86, 259-275.

Kroll Bond Rating Agency (2011). An Analysis of Historical Municipal Bond Defaults Lessons Learned – The Past as Prologue.

New York Times. Old Repudiated Debts that Stir the British. August 10, 1930. Page X12.

Ratchford, B. U. (1941). American State Debts. Durham, NC: Duke University Press.

Smythe, R. M. (1904). Obsolete American Securities and Corporations. New York: R. M. Smythe.

Most of the 19th century defaults occurred in two waves: one that followed the Panic of 1837 and the other following the Civil War. The first wave of defaults happened after a number of states made heavy investments in canals and state-chartered banks. In the deflationary years that followed the 1837 financial collapse, eight states and the territory of Florida failed to service their obligations. A recent article by Jeff Hummel provides an excellent summary of this situation, along with references for further reading.

The post-Civil War defaults were concentrated in ravaged Southern states, many of which took on substantial loads of debt under “carpetbagger” controlled governments. The carpetbaggers – northerners who came down south with their possessions wrapped in old carpets – seized control of some state governments, issued bonds and then kept much of the proceeds. When local politicians regained control of state governments, they deemed the carpetbagger-incurred debt as illegitimate and repudiated it.

Since the 19th century, all states with the exception of Vermont have implemented balanced budget requirements and other restrictions on debt issuance. As a result, state bonded indebtedness as a percentage of GDP has remained relatively low, while the national debt has skyrocketed.

During the 20th century, there was only one case in which a state bond default resulted in losses for individual investors: the Arkansas default of 1933. Arkansas got into trouble after assuming a large volume of road bonds issued by local governments during the 1920s. This heavy debt load combined with sharply decreased property tax collections (the result of falling property values during the Depression) rendered the state insolvent.

Many readers will undoubtedly be skeptical of state balanced budget requirements given the many news reports of politicians circumventing these restrictions. While elected officials do employ many gimmicks, they also impose real spending cuts and revenue enhancements when closing budget gaps. The result is some growth in debt burdens, but not nearly enough to trigger a solvency crisis. Professors Daniel Bergstresser and Randolph Cohen provide a detailed discussion of constitutional balanced budget rules, methods used to circumvent them and their overall restraining influence in a recent Harvard Business School paper.

Evaluating a Government’s Debt Burden

A government’s debt burden is often stated in terms of a Debt-to-GDP ratio. This ratio scales debt to the size of the economy, thus providing a more consistent measure across political subdivisions with varying populations and wealth. The Bureau of Economic Analysis reports US GDP by State. This BEA measure is called Gross State Product (or GSP). The Census bureau collects state and local government financial data (including indebtedness) each year, with the most recent data being available for Fiscal 2009. By combining the BEA and Census data set, we can measure Debt/GSP by State, as we do in Table 2.

Table 2. Debt/GSP Ratio By State, 2009

State	Total Debt	Gross State Product	Debt/GSP Ratio
Alabama	8,155,943	166,819,000	4.89%
Alaska	6,589,698	45,861,000	14.37%
Arizona	12,324,879	249,711,000	4.94%
Arkansas	4,135,051	98,795,000	4.19%
California	134,571,934	1,847,048,000	7.29%
Colorado	17,202,374	250,664,000	6.86%
Connecticut	28,394,151	227,550,000	12.48%
Delaware	5,984,645	60,660,000	9.87%
Florida	38,885,422	732,782,000	5.31%
Georgia	13,455,164	394,117,000	3.41%
Hawaii	6,880,242	65,428,000	10.52%
Idaho	3,501,676	53,661,000	6.53%
Illinois	56,962,364	631,970,000	9.01%
Indiana	23,711,889	259,894,000	9.12%
Iowa	6,353,306	136,062,000	4.67%
Kansas	5,857,295	122,544,000	4.78%
Kentucky	13,364,138	155,789,000	8.58%
Louisiana	17,504,772	205,117,000	8.53%
Maine	5,297,276	50,039,000	10.59%
Maryland	23,472,579	285,116,000	8.23%
Massachusetts	74,597,901	360,538,000	20.69%
Michigan	29,591,278	369,671,000	8.00%
Minnesota	10,524,424	258,499,000	4.07%
Mississippi	6,208,639	94,406,000	6.58%
Missouri	19,217,206	237,955,000	8.08%
Montana	4,723,765	34,999,000	13.50%
Nebraska	2,516,775	86,411,000	2.91%
Nevada	4,444,804	125,037,000	3.55%
New Hampshire	8,411,660	59,086,000	14.24%
New Jersey	56,897,866	471,946,000	12.06%
New Mexico	8,001,721	76,871,000	10.41%
New York	122,651,630	1,094,104,000	11.21%
North Carolina	19,910,714	407,032,000	4.89%
North Dakota	1,888,148	31,626,000	5.97%
Ohio	27,949,184	462,015,000	6.05%
Oklahoma	9,855,393	142,388,000	6.92%
Oregon	12,678,820	167,481,000	7.57%
Pennsylvania	41,924,042	546,538,000	7.67%
Rhode Island	9,180,938	47,470,000	19.34%
South Carolina	15,313,021	158,786,000	9.64%
South Dakota	3,626,024	38,255,000	9.48%
Tennessee	4,847,786	243,849,000	1.99%
Texas	30,438,160	1,146,647,000	2.65%
Utah	6,267,888	111,301,000	5.63%
Vermont	3,426,670	24,625,000	13.92%
Virginia	24,301,179	409,732,000	5.93%
Washington	24,603,219	331,639,000	7.42%
West Virginia	6,501,995	61,043,000	10.65%
Wisconsin	20,913,355	239,613,000	8.73%
Wyoming	1,320,852	36,760,000	3.59%
50 State Total	1,045,339,855	13,915,950,000	7.51%

By international standards, these ratios are quite low – well below those of the US federal government and other major Western countries. They also compare favorably to large Canadian provinces such as Quebec and Ontario – which had bonded debt to gross product ratios of 44% and 30% respectively in 2009 (according to public accounts documents filed by each province).

Also, it is worth noting that the Census debt totals include much more than a state’s general obligation bonds. These totals also incorporate revenue bonds, industrial revenue bonds, pollution control bonds, special assessment bonds, certificates of participation (COPs), judgments, mortgages and construction loan notes (CLNs) – which are generally junior to general obligations.

When assessing states and national governments, rating agencies often consider the ratio of interest expense to revenue. This ratio is more useful that Debt/GDP because it also reflects the impact of interest rates and the government’s ability to derive revenue from the economy. Japan, for example, can sustain very high Debt/GDP ratios (above 200% by some measures) because it faces very low interest rates. Governments that have relatively limited ability to extract revenue like Greece (due to tax evasion) and the US federal government (due to difficulty of passing tax increases) may face crises at lower levels of Debt/GDP. American states also face lower ceilings on their Debt/GSP ratios because their ability to raise tax rates is limited by the relative ease of relocating to another state (as opposed to another country).

Table 3 shows Interest Expense to Revenue ratios based on Census data.

Table 3. Interest Expense to Total Revenue, 2009

State	Interest Expense	Revenue	Interest/Revenue Ratio
Alabama	340,732	20,504,479	1.66%
Alaska	310,326	9,001,893	3.45%
Arizona	508,006	23,232,724	2.19%
Arkansas	161,048	12,879,574	1.25%
California	6,220,851	113,389,307	5.49%
Colorado	805,668	10,336,795	7.79%
Connecticut	1,432,900	20,931,946	6.85%
Delaware	284,055	5,787,487	4.91%
Florida	1,480,969	45,602,974	3.25%
Georgia	660,288	33,614,408	1.96%
Hawaii	432,964	6,751,116	6.41%
Idaho	180,803	5,537,466	3.27%
Illinois	2,963,785	40,530,099	7.31%
Indiana	899,488	27,948,706	3.22%
Iowa	252,713	13,207,715	1.91%
Kansas	346,754	11,654,910	2.98%
Kentucky	525,344	18,993,131	2.77%
Louisiana	954,107	23,100,188	4.13%
Maine	268,987	6,468,113	4.16%
Maryland	1,044,928	24,071,623	4.34%
Massachusetts	3,736,377	37,260,774	10.03%
Michigan	1,147,573	47,714,878	2.41%
Minnesota	526,994	22,781,153	2.31%
Mississippi	218,614	14,375,357	1.52%
Missouri	847,712	17,937,884	4.73%
Montana	167,993	4,828,033	3.48%
Nebraska	108,050	7,380,731	1.46%
Nevada	212,010	7,531,884	2.81%
New Hampshire	391,573	5,639,852	6.94%
New Jersey	2,139,595	42,946,396	4.98%
New Mexico	324,507	9,643,066	3.37%
New York	5,505,131	92,112,356	5.98%
North Carolina	637,401	30,150,407	2.11%
North Dakota	146,457	4,372,683	3.35%
Ohio	1,445,927	24,961,252	5.79%
Oklahoma	490,185	17,428,959	2.81%
Oregon	456,106	7,507,145	6.08%
Pennsylvania	1,823,759	38,795,484	4.70%
Rhode Island	437,293	4,710,553	9.28%
South Carolina	780,304	19,897,729	3.92%
South Dakota	135,135	2,436,934	5.55%
Tennessee	236,118	18,750,827	1.26%
Texas	1,242,925	82,053,694	1.51%
Utah	243,708	8,783,428	2.77%
Vermont	164,354	4,559,365	3.60%
Virginia	919,732	26,224,905	3.51%
Washington	1,110,674	24,476,808	4.54%
West Virginia	246,458	11,112,547	2.22%
Wisconsin	1,262,365	8,518,436	14.82%
Wyoming	64,221	4,787,884	1.34%
50 State Total	47,243,967	1,123,226,058	4.21%

These ratios reflect each state’s total revenue – not just its general fund revenue – and its interest expense on all bonds – not just general obligations.

When Arkansas defaulted in 1933, its interest to revenue ratio was about 30% - well above the level in any other state at the time and well above present levels. The only two other defaults during the last century by governments similar to US states in comparable nations (New South Wales, Australia in 1931 and Alberta, Canada in 1936) also occurred after the interest to revenue ratio exceeded 30%.

Applicability to Today

Could a state default with an interest to revenue ratio significantly below 30%?  While any financial result is possible, it is highly unlikely. Default is a political decision in which elected officials balance two considerations: (1) inability to spend money on programs due to debt service expenses versus (2) embarrassment and loss of bond market access by defaulting.

When interest expenses are a relatively low proportion of revenue, defaulting does not make any political sense. Indeed, defaulting makes much less political sense now than it did in the 19th century or during the Great Depression.

Most state debt in the 19th century was held in Europe. In 1933, most Arkansas debt was held by investors in other states - in New York and elsewhere. With the inception of state income taxes and their exclusion of most in-state municipal bond interest, a very large proportion of state bonds are now owned by high income residents. Since the bonds are in the hands of voters and campaign contributors, the choice to default is even more politically suicidal than in was in the 1930s.

Pensions and Retiree Benefits

Fears about current state credit quality center around public employee pensions and other employee benefits.  The size of this problem – from a credit standpoint – is often exaggerated because the unfunded liability is juxtaposed with the state’s annual budget. Since the unfunded liability is payable over many years, this comparison is faulty. Further, unfunded liabilities quoted in the media often apply to an entire pension system, whose costs are only partially borne by the state. For example, the State of California is responsible for only about 36% of the beneficiaries in the CalPERS system (according to page 151 of the CalPERS 2011 Comprehensive Annual Financial Report.

Underfunded state employee pensions are nothing new. The problem was also common in the 1970s and early 1980s – the last period of extended poor stock market performance – and did not produce any general obligation defaults. Statistics published by Alicia Munnell and her colleagues at the Boston College Center for Retirement Research show that pension contributions as a proportion of state budgets have yet to return to the peaks reached thirty years ago.

Unfunded retiree health benefits have also been a constant. While it is true that health care costs have risen substantially in recent decades, much of this extra cost is borne by Medicare.

Conclusion

When considering the risk of a state G.O. bond default, investors should be careful to separate their political views from actionable investment information. Inadequately funded retirement plans and political evasions of balanced budget requirements are bad public policies – worthy of criticism in the court of political opinion. Whether these issues are serious enough to trigger an outright default on principal and interest payments, is a very different question, and one that is best answered through historical research and quantitative analysis.

---------------------------------------------------------------------

Marc Joffe is a consultant with PF2 Securities Evaluations, which recently published an open source Public Sector Credit Framework. In 2011, he researched and co-authored Kroll Bond Rating Agency’s municipal bond study. Prior to that, Marc was a Senior Director at Moody’s Analytics. Marc owns State of California bonds.

PF2 Launches Open-Source Sovereign and Muni Rating Tool

Hi everyone

For those of you who have been following our last few posts, you'll be happy to know that we've launched PSCF today.

In conjunction with the launch, we're making available sample models for the United States and California.

We've included a set of slides to help you get through this. There's also a white paper taking you through the construction and describing our approach. It's open-source, so feel free to have a bash.

Click around at http://www.publicsectorcredit.org/pscf.html and share your thoughts.

 ~ PF2

PSCF - Press Release

Substandard and Porous? A Belated Response to Nate Silver

After S&P downgraded the US last August, Nate Silver analyzed the agency’s record on sovereign debt and found it wanting (See Why S&P’s Ratings Are Substandard and Porous). Silver ran a number of statistical tests, and determined that S&P’s ratings were serially correlated, highly related to the Corruption Perceptions Index and less predictive of default than simple quantitative measures.

Silver’s analysis is impressive for an industry outsider, but it suffered some deficiencies. For example, he applied a linear scale (AAA = 9, AA = 8, A = 7, etc.) when mapping ratings to decimal values. Given the structure of historic default rates, some sort of geometric scaling would have been more appropriate.

Our criticism, however, should not compromise Silver’s core point: that statistical analysis can probably do a better job of telling us about sovereign default probabilities than the traditional rating agency approach.

An Argument for a Model-Based Approach

Certainly, an intensive statistical analysis avoids several of the pitfalls facing sovereign ratings, as currently implemented.

First, a rating methodology that relies heavily on qualitative techniques is vulnerable to bias. The serial correlation Silver found stems from a natural bias at rating agencies against extreme actions. Rather than imposing a large-scale, multiple-notch downgrade, rating committees may be predisposed to implementing a lesser, often single-notch change in the hope that subsequent events will obviate further action. While the biased rater might thus apply a rating inconsistent with the methodology, a computer model, lacking the capacity to “hope,” reverts directly to the honest, brutal truth.

There’s also a more fundamental objection to the rating agency model. Their qualitative approach, requiring the human touch, is labor intensive. But for-profit rating agencies are oddly notorious for understaffing their sovereign rating groups.

A model-based approach enables more frequent, more intensive analysis – as opposed to the infrequent reviews sovereigns now receive. New data can be loaded into the model at regular intervals and new results calculated. Analysts should still oversee the model parameters and check any results that may look suspicious.

The Ingredients of a Sovereign Debt Model

Model-based approaches to sovereign risk often involve credit default swap spreads, as the independent or dependent variable. A model can either extract default probabilities from CDS spreads, or attempt to predict those spreads on the grounds that they are a proxy for actual risk. Silver takes this latter approach in his piece.

The use of market inputs in credit models is fairly common. Such a modeling choice often implicitly or explicitly relies on the Efficient Market Hypothesis – the idea that market prices incorporate all relevant information and are thus the best available estimate of value.

Since the financial crisis, critics of EMH have sharpened their attacks. But whether or not you subscribe to rational expectations, the use of sovereign CDS is hard to defend. Most EMH advocates recognize that only liquid markets are efficient. Since liquid markets have numerous participants, their equilibrium prices incorporate substantial amounts of information.

This is not the case with sovereign CDS markets. Kamakura Corporation examined sovereign trading volumes reported by DTCC for late 2009 and 2010, and found that the vast majority of sovereign CDS contracts were traded fewer than five times per day (excluding inter-dealer trades). Five transactions per day falls well short of a liquid market, and thus the information content of sovereign CDS spreads is doubtful at best.

Absent meaningful CDS spread data, what else can a government credit model rely upon?

While one might look at Corruption Perception Indices, per capita GDP and/or terms of trade, it is not clear that these inputs will differentiate between advanced economy sovereigns and sub-sovereigns. Fortunately, government issuers produce reams of actual and projected fiscal data. This information, combined with demographic inputs and economic forecasts, can take us a long way.

When we suggest that budget forecasts can be employed in government credit modeling, skeptics point out accuracy issues with government forecasters.

The most famous forecasting error is attributed to the US CBO, which predicted trillions of surpluses for the first decade of the 21st century, instead of the trillions in deficits that actually appeared.

CBO forecasts are usually published in the form of point estimates. To be reliable, they have to reflect accurate forecasts of interest rates, GDP, tax levels and a host of other macroeconomic and policy variables. Given the number of variables and our (collective) limited capacity to predict, the point estimate is bound to be wrong. That notwithstanding, we can be pretty certain that these variables will fall within a given range. For example, it is almost certain that US GDP growth will be somewhere between -3% and +6% next year (2013). If we run a large number of scenarios with different GDP growth rates within this range, it is likely that some of the trials will closely approximate the ultimate fiscal outcome.

We can run a large number of budget scenarios by using a Monte Carlo simulation – in which scenarios are created by generating random numbers. Budget simulation forms the basis for PF2’s Public Sector Credit Framework that we will release next week. The tool allows the user to enter a default threshold in the form of a fiscal ratio; create macroeconomic series that vary with each trial through linkages to random numbers; and design fiscal series that rely on one or more of these macroeconomic elements. If you would like to learn more about this technology, please contact us at info@pf2se.com, or call +1 212-797-0215.
------------------------
Contributed by PF2 consultant Marc Joffe. Marc previously researched and co-authored Kroll Bond Rating Agency’s Municipal Default Study. This is the last of four blog posts introducing PF2’s Public Sector Credit Framework. Previous posts on this topic may be found here, here and here.

Pro Bono Finance

Lawyers fight to save death row inmates. Doctors provide charity care – treating the indigent, often without government reimbursement. In the financial services industry, volunteer work usually takes the form of pitching in at schools and cleaning parks. We finance folks lack a tradition of using our skills for public service. With our reputation in tatters, perhaps it is time to begin such a tradition.

Fears of sovereign and municipal debt crises offer a worthy volunteer opportunity. Government debt problems can easily become matters of life and death. Argentina’s sovereign debt crisis claimed 24 lives in December 2001. In Greece, crisis-related protests have claimed at least 5 lives and caused over 300 injuries. Annual suicide rates are up about 20 percent, as people despair over their diminished circumstances.

A US federal debt crisis could similarly lead to violent protests, fatalities and widespread psychological damage; it could also be accompanied by high levels of inflation and sudden, sharp cuts in benefits. Older people dependent on savings and social insurance payments would be especially hard hit.

Because hurricanes and tornados kill and injure, scientists have invested substantial time and effort in forecasting these natural disasters and helping members of the public avoid them. Fiscal crises – a type of human-made disaster – can also be anticipated and potentially curbed, or even avoided.

Last summer’s debt ceiling debate was a failed opportunity to avoid a US fiscal crisis. As we look back on the debate, it becomes evident that false and misleading rhetoric frequently crowded out accurate information. Among the myths that plagued last year’s discussion:

• Failing to raise the debt ceiling would have inevitably triggered a default.¹
  
  
• The US government has never defaulted in its entire history (see our earlier blog post for the facts).
  
  
• The nation’s long term budget imbalance can be resolved simply by controlling domestic discretionary spending or allowing the Bush tax cuts on high earners to expire (neither of these steps generate enough savings to avoid future problems).<
  
  
• A 90% Debt-to-GDP ratio will trigger a fiscal crisis (see Japan).
  
  
• We need to balance the budget to avoid a fiscal crisis (the Debt-to-GDP ratio will improve as long as the stock of debt grows more slowly than GDP).
  
  
• Rapid economic growth is impossible if the federal government spends more than 20% of GDP (see 1999 economic and fiscal statistics for a refutation of this contention).

The financial community can provide a useful community service by educating the public about sovereign, state and municipal credit issues. And when I say educate, I don’t mean pontificate. Many of us - this writer included - hold views about what should be done about taxes and spending. Mixing these opinions with facts is not an unambiguous public service. Just as we strive to dispassionately evaluate credit and select investment opportunities, we can and should separate fact from opinion when informing voters about their fiscal options.

Mary Meeker and her colleagues issued a free report entitled “USA, Inc.” that provided the type of service I am suggesting. The report, which received substantial publicity, analyzed the nation’s fiscal position in a manner similar to that of an equity investor analyzing a business – with hundreds of slides describing revenue and expense drivers.

The next challenge is to broaden the scope of analysis to provide a credit perspective with its focus on default risk and recovery. Also, rather than misapply corporate or structured debt analytics, we need a fresh approach that directly addresses the unique challenges of assessing government debt. We’ll also be better positioned if the analysis is ongoing, or even real-time, rather than a “snapshot” analysis provided in the Meeker report.

Ideally, having a well-structured, up-to-date model on which to base policy positions seems enviable. If would assist Congress and the Administration in defining the task at hand and dispelling myths surrounding the task – and it would encourage an environment in which commentators would be required to support their opinions with quantifiable data, not simply foggy criteria.

At PF2, we will kick-start the effort to dispel the fog of opinion by offering a free, open source Public Sector Credit Framework (PSCF). Our framework will be accompanied by a timely, transparent US federal budget simulation model, which we’ve designed to estimate the likelihood of a fiscal crisis in each of the next thirty years. Although we don’t know everything about this topic, we do know that many financial professionals are equipped to improve the software and the model. We encourage you all to join us in enhancing the analysis.

Few can afford to provide pro bono services exclusively – and we are no exception. If the framework generates interest, we may use it as a platform for valuing sovereign and municipal bonds – a service we hope to monetize. But that notwithstanding, the software and model are being supplied at no charge under GNU’s Lesser General Public License, for anyone to use and improve. Assuming interest is sufficient, we will regularly update the US federal model as a public service.

Many of us in the financial service industry have done quite well. Having reaped some of the rewards, we think we have found a great way to give back. We look forward to your collaboration.

------------------------------------
¹ Treasury could have avoided a default through some combination of asset sales and spending reductions. The President could have invoked a clause in the 14^th Amendment of the Constitution to mandate principal and interest payments that would have exceeded the debt ceiling. Congress would then have had to file a legal case to overturn the President’s order.

Contributed by PF2 consultant Marc Joffe. Marc previously researched and co-authored Kroll Bond Rating Agency’s Municipal Default Study. This posting is the third in a series of posts leading up to May 2nd. The prior pieces can be accessed by clicking here and here.

Credit Rating Agency Models and Open Source

When S&P downgraded the US from AAA to AA+, the US Treasury accused the rating agency of making a $2 trillion mathematical error. S&P initially denied this accusation, but adjusted some of its estimates in a subsequent press release. Economist John Taylor defended S&P, contending that its calculations were based on a defensible set of assumptions, and thus could not be categorized as a mistake. S&P’s model, which projected future debt-to-GDP ratios, has not been made public. As a result, it is difficult for outside observers to decide whom to believe: the rater or the rated.

There are at least three ways a model’s results can be wrong: if the model’s code itself doesn’t function as intended; if the known inputs are incorrectly entered, and if the assumptions are misapplied. In cases as important as the evaluation of US sovereign debt, we think rating agencies and the investing public would be better off if the relevant models were publicly available. Some may argue that the inputs to the models are proprietary or that they reflect qualitative assumptions valuable to the ratings agencies – i.e., that they are a “secret sauce.” But, even if rating agencies want to keep their assumptions proprietary, making the models themselves available would decrease the likelihood of rating errors arising from software defects.

Keeping one’s internal processes internal is the traditional way. Manufacturers assume that consumers don’t want to see how the sausages are made. In the internet era, it is now much easier to produce the intellectual equivalent of sausages in public – and, as it happens, many consumers are interested in the production process and even want to get involved. Wikipedia provides an excellent example of the open, collaborative production of intellectual content: articles are edited in public and the results are often subject to dispute. Writers get almost instantaneous peer review and the outcome is often rapid iteration moving toward the truth. In their books, Wikinomics and Macrowikinomics, Dan Tapscott and Anthony Williams suggest that Wikipedia’s mass collaboration style is the wave of the future for many industries – including computer software.

Many rating methodologies, especially in the area of structured finance, rely upon computer software. At the height of the last cycle, tools that implemented rating methodologies such as Moody’s CDOROM^TM, were popular with both issuers and investors wondering how agencies might look at a given transaction. While the algorithms used by these programs are often well documented, the computer source code is usually not released into the public domain.

Over the last two decades, the software industry has seen a growing trend toward open source technology, in which all of a system’s underlying program code is made public. The best known example of open source system is Linux, a computer operating system used by most servers on the internet. Other examples of popular open source programs include Mozilla’s Firefox web browser, the WordPress content management system and the MySQL database.

In financial services, the Quantlib project has created a comprehensive open source framework for quantitative finance. The library, which has been available for more than 11 years, includes a wide array of engines for pricing options and other derivatives.

Open source allows users to see how programs work and with the help of developers, fully customize software to meet their specific needs. Open source communities such as those hosted on GitHub and SourceForge, enable users and programmers from all over the world to participate in the process of debugging and enhancing the software.

So how about credit rating methodologies? Open source seems especially appropriate for rating models. Rating agencies realize relatively little revenue from selling rating models; they are more likely to be used to facilitate revenue generation through issuer-paid ratings.

Open source enables a larger community to identify and fix bugs. If rating model source code were in the public domain, investors and issuers would have a greater chance to spot issues. Rating agencies would be prevented from covering up modeling errors by surreptitiously changing their methodologies. In 2008, The Financial Times reported that Moody’s errantly awarded Aaa credit ratings to a number of Constant Proportion Debt Obligations (CPDOs) due to a software glitch. The error was fixed, but the incorrectly rated securities were not immediately downgraded according to the FT report. Had the rating software been open source, it would not have been much more difficult to conceal this error, and it would have offered the possibility for a positive feedback loop – an investor or other interested party could have found and fixed the bug on Moody’s behalf.

Not only do open source rating models promote quality, they may also reduce litigation. The SEC issued Moody’s a Wells Notice in respect of the above mentioned CPDO issue, and may well have brought suit. (A Wells Notice is a notification of intent to recommend that the US government pursue enforcement proceedings, and is sent by regulators to a company or a person.) Investors have brought suit against the rating agencies to the extent they felt the ratings were inappropriate, for model-related errors or otherwise. By unveiling the black box, the rating agencies would be taking an active approach in buffering against litigation, and enjoy the material defense that, “yes we may have erred, but you were afforded the opportunity to catch our error – and didn’t.”

Unlike the CPDO model employed by Moody’s, the S&P US sovereign "model" likely took the form of a simple spreadsheet containing adjusted forecasts from the Congressional Budget Office. In contrast to the structured and corporate sectors, there are relatively few computer models for estimating sovereign and municipal default probabilities. While little modeling software is available for this sector, accurate modeling of government credit can be seen as a public good. Bond investors, policy makers and citizens themselves could all benefit from more systematic analysis of government solvency.

Open source communities are a private response to public goods problems: individuals collaborate to provide tools that might otherwise appear in the realm of licensed software. Thus open source government default models populated with crowd-sourced data maybe the best way to fill an apparent gap in the bond analytics market.

On May 2^nd, PF2 will contribute an open source Public Sector Credit Framework, which is aimed at filling this analytical gap, while demonstrating how future rating models can be distributed and improved in an iterative, transparent manner. If you wish to participate in beta testing or learn more about this technology please contact us at info@pf2se.com, or call +1 212-797-0215.

--------------------------------------------
Contributed by PF2 consultant Marc Joffe. Marc previously researched and co-authored Kroll Bond Rating Agency’s Municipal Default Study. This posting is the second in a series of posts leading up to May 2nd. The prior piece can be accessed by clicking here.

Multiple Rating Scales: When A Isn’t A

Philosophers from Aristotle to Ayn Rand have contended that “A is A.” Apparently none of these thinkers worked at a credit rating agency - in which “A” in one department may actually mean AA or even BBB in another. While the uninitiated might naively assume that various types of bonds carrying the same rating have the same level of credit risk, history shows otherwise.

During the credit crisis, AAA RMBS and ABS CDO tranches experienced far higher default rates than similarly rated corporate and government securities. Less well known is the fact that municipal bonds have for decades experienced substantially lower default rates than identically rated corporate securities – and that the rating agencies never assumed that a single A-rated issuer ought to carry the same credit risk in both sectors. This discrepancy was noted in Fitch’s 1999 municipal bond study and confirmed by Moody’s executive Laura Levenstein in 2008 Congressional testimony on the topic. Later in 2008, the Connecticut attorney general sued the three major rating agencies for under-rating municipal bond issues relative to other asset categories. (The suit was recently settled for $900,000 in credits for future rating services, but without any admission of responsibility). Last year, three economists – Cornaggia, Cornaggia and Hund – reported that government credit ratings were harsher than those assigned to corporates, which, in turn, were more severe than those assigned to structured finance issues.

One might ask why it is important for ratings in different credit classes to carry the same expectation in terms of either default probability or expected loss? Perhaps we should accept the argument that ratings are intended to simply provide a relative measure of risk among bonds within a given asset class.

There are at least two problems with this approach. First, it is unnecessarily confusing to the majority of the population that is unaware of technical distinctions in the ratings world. Second, it creates counterproductive arbitrage opportunities.

If an insurer is rated AAA on a more lenient scale than insurable entities in another asset class, the insurer can profitably "sell" its AAA rating to those entities without creating any real value in the process.

Municipal bond insurance is a great example. Monoline bond insurers like AAA-rated Ambac, FGIC and MBIA insured bonds issued by states, cities, counties and other municipal issuers for three decades prior to the 2008 financial crisis. In some cases, the entities paying for insurance were of a stronger credit quality than the insurers. As it happened, the insurers often failed while the issuers survived, leaving one to wonder why the insurance was necessary.

During this period, general obligation bonds had very low overall default rates. According to Kroll Bond Rating Agency’s Municipal Default Study, estimated annual municipal bond default rates by issuer count have been consistently below 0.4% since 1941. Similar findings for the period 1970-2010 are reported in The Bloomberg Visual Guide to Municipal Bonds by Robert Doty. This 0.4% annual rate applies to all municipal debt issues, including unrated issues and revenue bonds. The annual default rate for rated, general obligation bonds is less than 0.1%.

Given this long period of excellent performance, one might reasonably expect that most states and other large municipal issuers with diversified revenue bases to be rated AAA. No state has defaulted on its general obligation issues since 1933, and most have relatively low debt burdens when compared to their tax base. Despite these facts, the modal rating for states is typically AA/Aa with several in the A range. (This remains the case despite certain rating agencies’ claims that they have recently scaled up their municipal bond ratings to place them on a par with corporate ratings).

The depressed ratings created an opportunity for municipal bond insurers to sell policies to states that did not really need them. For example, the State of California paid $102 million for municipal bond insurance between 2003 and 2007. Negative publicity notwithstanding, the facts are that single A rated California has a Debt to Gross State Product ratio of 5% (in contrast to a 70% Debt/GDP ratio for the federal government) and that interest costs represent less than 5% of the state’s overall expenditures. While pension costs are a concern, they are unlikely to consume more than 12.5% of the state’s budget over the long term – not nearly enough to crowd out debt service.

California provides but one example. The Connecticut lawsuit mentioned above also cited unnecessary bond insurance payments on the part of cities, towns, school districts, and sewer and water districts.

Meanwhile, AAA-rated municipal bond insurers carried substantial risks, evident to many not working at rating agencies. For example, Bill Ackman found in 2002 that MBIA was 139 times leveraged. As reported in Christine Richard’s book Confidence Game, Ackman repeatedly shared his research with rating agencies – to no avail.

This imbalance between the ratings of risky bond insurers and those of relatively safe municipal issuers essentially created the monoline insurance business – a business that largely disappeared with the mass bankruptcy and downgrading of insurers during the 2008 crisis.

Inconsistent ratings across asset classes thus do have real world costs. In the US, taxpayers across the country paid billions of dollars over three decades for unneeded bond insurance. Individual municipal bond investors, often directed by their advisors to focus on AAA securities only, missed opportunities to invest in tens of thousands of bonds that should credibly have carried AAA ratings, but were depressed by the raters’ inopportune choice of scale.

We believe that one reason for the persistent imbalance between municipal, corporate and structured ratings is the dearth of analytics directed at government securities. Rating agencies and analytic firms offer models (and attendant data sets) that estimate default probabilities and expected losses for corporate and structured bonds. Such tools are relatively rare for government bonds. Consequently, the market lacks independent, quantitatively-based analytics that compute credit risks for these instruments. This lack of alternative, rigorously researched opinions allows the incorrect rating of US municipal bonds to continue, without the alleviation of a positive feedback loop.

Next month, PF2 will do its part to address this gap in the marketplace with the release of a free, open source Public Sector Credit Framework, designed to enable users to estimate government default probabilities through the use of a multi-period budget simulation. The framework allows a wide range of parameterizations, so you may find it useful even if you disagree with the characterization of municipal bond risk offered above. If you wish to participate in beta testing or learn more about this technology please contact us at info@pf2se.com, or call +1 212-797-0215.

--------------------------------------------
Contributed by PF2 consultant Marc Joffe. Marc previously researched and co-authored Kroll Bond Rating Agency’s Municipal Default Study.

Withdrawing, Confidently

As structured finance deals wind down and the asset pools grow smaller, the situation often arises that the effectiveness of outstanding tranche ratings – previously based on a portfolio-level diversification – can hinge on the performance of one or two bonds. The problem is compounded, of course, in that many of the models work best for large, diverse, portfolios and often break down when the portfolios become arbitrarily small.

The question then becomes, if a rated tranche can just as easily be rated AAA or D, what does one do?

This tricky situation, now part skill, part luck, calls into question the predictive content of highly sophisticated ratings models when the outcome is really not a model-driven result, but simply a short-term occurrence (e.g., a payoff or a default) or the lack of an occurrence, in a credit-default swap environment.

Moody’s and S&P suffered severe blushes in January when a well-structured CDO, backed heavily by other CDOs and RMBS (including substantial subprime, yes subprime), paid off in full ­– with their outstanding ratings on all tranches having been in the CC to CCC range. What was interesting was that both ratings agencies had visited this deal as recently as June of last year.
From our conversation with analysts at one of the agencies, what happened here was simply that as the deal was winding down, the manager was able to sell the few remaining assets at prices high enough to pay down all the notes, rendering irrelevant the Monte Carlo default simulation trials being run by the raters. In other words, the model let them down.

In an interesting, perhaps prudent decision, S&P took a different course in an announcement they made earlier today, entitled “S&P Takes Various Ratings Actions on 30 U.S. RMBS Deals.” As certain deals dwindled down, compromising the predictive content of their ratings, they chose to simply withdraw the ratings.

“We subsequently withdrew our ratings on certain affected classes that are backed by a pool with a small number of remaining loans. If any of the remaining loans in these pools default, the resulting loss could have a greater effect on the pool's performance than if the pool consisted of a larger number of loans. Because this performance volatility may have an adverse affect on our outstanding ratings, we withdrew our ratings on the related transactions.”

While it may cause frustration to note holders to see the ratings withdrawn, it augurs well that a rating agency is able and willing to say that it cannot have confidence in the outcome, and therefore chooses to withdraw its rating rather than have investors rely, perhaps falsely, on a rating in which it does not have confidence.

Analysis of The Shortcomings of Statistical Sampling in the Mortgage Loan Due Diligence Process

This is a popular litigation-related piece on our website we thought we'd share through this post (pdf version available here) - enjoy the read.



Introduction

Financial institutions, when assembling mortgage pools for the purpose of inclusion in residential mortgage-backed securities (RMBS), often hire independent analytical companies, like Clayton Holdings LLC (“Clayton”), to perform due diligence on the loans and flag any that are problematic.

Leading up to the financial downturn, Clayton reviewed mortgages for its clients - investment and commercial banks and lending platforms, including those of Bear Stearns, Barclays, Bank of America, C-Bass, Countrywide, Credit Suisse, Citigroup, Deutsche Bank, Doral, Ellington, Freddie Mac, Greenwich, Goldman, HSBC, JP Morgan, Lehman, Merrill Lynch, Morgan Stanley, Nomura, Société Générale, UBS and Washington Mutual (the “Issuers”). As such Clayton was purportedly one of the larger due diligence companies that analyzed whether these loans met specifications like loan-to-value ratios, credit scores and the income levels of borrowers.

Clayton describes, in the presentation it provided to the Financial Crisis Inquiry Commission (“FCIC”), the results of its review of a total of 911,039 mortgage loans between Q1 2006 and Q2 2007 ₁. As can be seen from the chart, of the loans shown to Clayton, Clayton determined approximately 72% of them to be in compliance, and 28% of them to be out of compliance with the standards tested, or “non-conforming.”

Upon determining that a loan failed to meet its guidelines, an Issuer (i.e., Clayton’s client) would have the ability to exercise their contractual right in “putting back” these non-conforming loans to the mortgage lenders – New Century, Fremont, Countrywide, Decision One Mortgage – rather than include them in securitizations.

The regulatory bodies, and the media, have concentrated heavily on the sizeable portions of non-conforming loans, and the lowering of underwriting standards throughout this period; but for this analysis, we concentrate on a more illuminating aspect of the way in which non-conforming loans ultimately found their way into the securitized RMBS pools.

There are at least two ways that non-conforming loans can find their way into the securitizations:

• First, the Issuer may choose to waive the loan back into the pool, despite its being originally rejected by Clayton.

• Second, a more overwhelming mechanism, is to not show the loan to Clayton.

The Intricacies of Loan Sampling

Importantly, it seems to have been common practice for Issuers to show only a sample of the loans to Clayton. A sample risks being unreflective of the population of loans, but random sampling can provide an effective statistical approximation under very strict conditions. It can be a cheaper process and, if the sample is well chosen, can accurately reflect the pool.

The objective of sampling is satisfied if the randomly-selected sample is sufficiently large, and is deemed to be in order. Alternatively, if the sample fails to meet expectations, the entire portfolio ought to be revisited. However, in the mortgage due diligence process the samples were often deemed to be problematic – they resulted in an average of 28% of loans failing their criteria. Importantly, the samples were then adjusted, as we understand it, but the original portfolios were not: the Issuers would only put back certain non-conforming loans from that sample.

In this case, the resulting sample, after throwing out certain non-conforming loans, fails to accurately depict the remaining portfolio of loans it was chosen to represent.



How the Sampling Process Worked, and Difficulties Therewith

Former President and COO of Clayton, D. Keith Johnson, explained to the FCIC committee, during their hearing of September 2010, that in the 2004 to 2006 time period, sample sizes went down to the region of two to three percent₂. As the sample size decreases, which it did, the effect of the sampling process alone begins to undermine the effectiveness of the due diligence process.

The media have focused their attentions on what happened to the 28% non-conforming loans – the slices in red in the associated charts. Indeed, many of these non-conforming loans, approximately 39%, were not “kicked out” or put back to the mortgage lenders, but were “waived” back in to the to-be-securitized portfolio. This 39% is substantial, and a factor worthy of the media’s attentions.

But the game-changing fact is not among these 28% non-conformers, or the 39% of them which remained in the securitized pool. These are only part of a sample, and when the sample becomes insignificantly small, its overall contribution to the portfolio as a whole is rendered less meaningful. Rather, it is more prudent to consider the composition of the pool as a whole.

For illustrative purposes, let us assume that the sample loans shown to Clayton represented 3% of the pools, on the higher end of those referred to in the abovementioned Johnson hearing. Let us conservatively assume that the sample provided to Clayton was truly randomly selected.₃

For a pool of 10,000 loans, Clayton would have been presented with approximately 300 loans, or 3%.

As we can see from the analysis performed, the effect of “throwing out” 61% of all non-conforming loans is marginal: the pool’s overall composition decreased only from 28% non-conforming to 27.67% non-conforming thanks to the due diligence process. Even had the Issuers returned all 84 non-conforming loans, the overall portfolio would not have been greatly altered –non-conforming loans would have declined from 28% to 27.16%.

When a random sample is tampered with, the final product, by definition, no longer represents the original pool. Here, the sample reflects that ultimately 89% of the pool is conforming and 11% are non-conforming (11% = 28% x 39%). But given the reality of the situation, with the original pool remaining status quo, in fact 27.76%, not 11%, of the overall pool was non-conforming, even after the put backs administered as part of the due diligence process.

A well-selected random sample can effectively capture the characteristics of a pool under certain conditions. But an altered sample seldom accurately reflects the original pool.



1 http://fcic-static.law.stanford.edu/cdn_media/fcic-testimony/2010-0923-Clayton-All-Trending-Report.pdf
2 http://fcic.law.stanford.edu/resource/interviews#J
3 course, the sample sizes used and the percentages rejected by Clayton will differ from Issuer to Issuer. So too will the waiver rate.