May 21, 2013
Commissioners, Mr. Butler, Office of Credit Ratings Staff: Thank you for the opportunity to share some ideas for improving credit ratings.
Globally, over $100 trillion is invested in bonds. Accurate credit assessments are important because they help ensure that interest rates for these securities properly reflect their risks. Interest rates, like prices in any other industry, provide signals that affect the supply of and demand for new bonds. If these price signals are wrong, our society's savings will be improperly invested and our prospects for much needed economic growth will be reduced.
Rating agencies and alternate providers of credit insight thus serve a crucial role in our economy. As a former executive at a rating agency, I know that these companies can and should do better. Indeed, the credit rating industry, ossified by decades of inertia and well-intended but status quo promoting regulation, is desperately in need of creative destruction.
Before I offer my solution, I need to explain what best practice in credit assessment looks like. Methods that rely solely or primarily on the subjective evaluations of human inspectors may be appropriate for grading eggs Double A or Single A, and they made sense when John Moody created the credit rating industry a century ago. But in todays world of big data, abundant computer processing power and advanced social science research techniques, we should be relying more heavily on models to assess the millions of bonds available to investors. Such models would ideally churn out numerical default probability or expected loss estimates rather than letter grades that dont have clear and consistent meanings.
While certain credit models failed during the financial crisis, this fact does not imply that the application of computer modeling to credit assessment is inappropriate. Instead, it implies that we need better models populated with more complete, more accurate and more objectively derived data.
Credit modeling, like most intellectual processes, can benefit from peer review. Indeed, the progression of corporate credit modeling from Altman's Z Score in the 1960s to Merton's option based approach in the 1970s to the more recent innovations of Duffie, Jarrow and others has taken place in peer reviewed academic journals. While much of this research is now embedded in proprietary commercial models not subject to peer review, one implementation is free, transparent and open to feedback from other researchers. This open corporate credit model is maintained by the National University of Singapore's Risk Management Institute. RMIs Credit Risk Initiative publishes model-derived default probabilities on tens of thousands of firms each day.
Transparent credit modeling tools and software have been announced – but not yet implemented – by a number of analytics firms in structured finance. This writer has developed open source models for sovereign and municipal bonds. In 2012, I published a Public Sector Credit Framework on GitHub (an open source repository) that enables analysts to create and run government bond rating models using a multi-year fiscal simulation. More recently, I developed a logistic model of municipal bond risk for California cities in response to a request from a unit of the State Treasurers Office. This fully documented model is now under peer review. Meanwhile, Wikirating – a Swiss-based start-up – is applying open models and polling techniques to elicit greater public participation on the ratings process.
These early initiatives show the potential for open source credit models to take their place next to proprietary software and agency ratings in the panoply of tools available to fixed income investors. These alternatives can all improve through competition with one another.
The SEC and other regulatory bodies can take a major step to encourage the development of open source, transparent, academically derived credit models and methodologies. Section 939A of the Dodd-Frank Act requires regulatory agencies to replace references to NRSRO ratings in their regulations with alternative standards of credit-worthiness.
I suggest that the output of a certified, open source credit model be included in regulations as a standard of credit-worthiness.
Thus far, regulatory changes have involved substituting a fixed formula or the judgment of the regulated entity for NRSRO ratings as credit-worthiness standards. Fixed formulae are readily circumvented (or gamed). For example, a regulator may define an investment grade structured finance security as one with a minimum percentage of overcollateralization. An issuer can defeat the intent of this regulation by providing the mandated overcollateralization but trapping excess spread in a reserve account for later distribution to subordinated bondholders. An open source methodology can evolve more quickly than regulations to address such structural innovations.
The danger of relying upon the judgment of the regulated entity is demonstrated by the spate of financial industry bailouts starting in 2008. Some of the institutions requiring bailouts were able to reduce their capital requirements by availing themselves of Basel IIs Internal Rating Based approach. While financial institutions can increase profits by taking on additional risk, allowing them to self-regulate involves conflicts of interest every bit as serious as those that afflict the credit rating agencies.
Transparent models maintained by university research centers and other neutral parties can remedy this conflict of interest. To be effective, these models must undergo a rigorous certification process designed by the SEC. The certification process itself should be administered by an independent body – perhaps by the rating assignment entity created under the Franken Amendment, if implemented.
The past two decades have demonstrated the incredible power of mass collaboration techniques such as open source software and Wikis to better our lives. Coming from nowhere in 1992, the open source Linux operating system is now the dominant program for running Internet servers. Wikipedia has replaced encyclopedias as the worlds repository of accessible knowledge.
By empowering open source credit assessment methodologies, regulators have the power to bring the benefits of mass collaboration to the world of credit ratings. As we look back from 2013, we can see the great wisdom of government officials who, in the early 1990s, opened up the Internet to users outside the defense industry and universities. Now financial regulators have an opportunity to make a similar enlightened decision.