An under-reported part of the 2008 financial crisis was the collapse of municipal bond insurers, like AMBAC and MBIA. These AAA rated companies were all downgraded or dissolved during the meltdown, after having collected billions of dollars from state and local governments. These insurance premia could have been spent on social services, but instead went to Wall Street.

According to Marc Joffe, a former rating agency employee, this situation occurred because rating agencies use different scales to rate governments and bond insurers. Because bond insurers were rated on an easier scale, they were able to get AAA ratings which they then “sold” to underrated governments. Marc believes that the solution to problems like this in the future is to have more transparent government bond ratings. To this end, he recently published an open source credit rating model.

Michel Bauwens: What is wrong with the current system of credit rating — why are they making so many mistakes?

Marc Joffe: Over the last decade or so, credit rating agencies have come under attack a number of times. First, in 2001, we had the problem with Enron, a large company which committed serious accounting fraud and had to be liquidated (i.e., shut down). While most of the criticism for that went to the company itself and its auditor, Arthur Andersen, rating agencies were also castigated for downgrading the company to junk status too late — just a few days before it declared bankruptcy. In 2007, we had the situation with mortgage-backed securities. Many AAA-rated mortgage-backed securities, especially those backed by U.S. subprime mortgages, experienced losses — with some trading at only a few cents on the dollar.

Then, in 2008, in a less well-known situation, the Attorney General of Connecticut sued rating agencies for systematically assigning lower ratings to state and local governments than to other bond issuers. And, finally, we have the European debt crisis, in which the agencies have been criticized by both investors and by governments over the timing of downgrades. In this last case, it is often hard to distinguish between legitimate criticism and “shooting the messenger.” Clearly, some countries needed to be downgraded since they are in serious danger of defaulting; but, in certain cases, rating agencies could have provided better and more timely analysis.

There are three general explanations for these problems: (1) conflicts of interest, (2) insufficient analysis, and (3) lack of transparency. The fact that credit rating agencies are paid by bond issuers, in most cases, creates a perception that the agencies are serving these issuers rather than bond investors. However, many of the problems I just listed seem to be due to incomplete work, rather than outright dishonesty. For example, the Connecticut municipal bond suit appears to have been caused by the fact that the agencies failed to adjust ratings to reflect improving credit performance by state and local governments after World War II. During the Depression, thousands of municipal bond issuers defaulted, causing Moody’s — the only active agency in the municipal market at the time — to slash most of the ratings.

In the 70 years that followed, municipal defaults have been relatively rare (despite the publicity you see for the occasional problem), but ratings were not fully restored to their 1929 levels. As a result, a state with a single-A rating is much safer than a corporation- or asset-backed security carrying the same rating. This is confusing to investors and needs to be adjusted. These various problems have been hard to identify and correct because rating agency procedures are not fully public. Rating agencies do publish methodology documents, but minutes of rating committee meetings — at which ratings are determined — are not published. Thus, we cannot be sure whether the rating agency followed its published methodology or how it interpreted that methodology in any given case.

MB: Why do you think an open-source approach would solve these issues and how does your system work?

MJ: My open-source approach directly addresses the transparency concern. In the Public Sector Credit Framework, the rating is produced by a quantitative model that can take into account budgetary, economic, and policy factors affecting the rated government. The framework actually calculates a probability of default from which a rating is inferred. All model inputs and outputs are visible in an Excel workbook. (We hope to support open-source alternatives to Excel in the future, but Excel was chosen as the initial platform because of its pervasiveness within the financial community.) Also, all the VBA and C source code driving the PSCF models are published on GitHub. In fact, I encourage any developers visiting Shareable to fork the distribution and help us improve the software.

The system calculates default probabilities by executing a multi-year fiscal simulation. A large number of economic scenarios are run through the model and the proportion of these scenarios that trip a user-defined threshold are classified as default cases. For example, if you assume that Spain will default once its debt-to-GDP ratio reaches 150%, PSCF will calculate the likelihood that Spain will exceed this threshold each year and characterize that as the annual default probability.

MB: What has been your personal motivation and path that led you to this solution?

MJ: I worked at a rating firm for nine years and went through the 2007-2008 crisis. I saw some of my colleagues leave the organization, and some testify against it at government hearings. While many people remaining at the company dismissed these critics and tried to deny everything to themselves, I believed that many of the criticisms were true. In 2009 and 2010, as the focus started shifting to government debt (both sovereign and municipal), I concluded that rating agencies were not performing the best possible analysis for this important asset class.

By 2011, I came to realize that my best opportunity to raise the level of analysis would be from outside the major rating agencies. I hope the release of this open-source software will start a conversation about best practices for government bond rating and, indeed, all types of bond rating. A number of European organizations have been taking steps toward creating not-for-profit alternatives to rating agencies. These include the Bertelsmann Foundation, Roland Berger, and Wikirating. I hope these players will consider open source, in general, and PSCF, in particular, to maximize the transparency of their ratings. Already, the National University of Singapore’s Risk Management Institute has been applying a Wiki-style, not-for-profit model to corporate credit assessments. We need to do much the same thing for government bond ratings.

MB: Can you give us an example of how the new system gives different results?

MJ: The State of California is rated single-A by all three agencies. This is despite the fact that state debt only accounts for about 6 percent of Gross State Product and costs for interests and pensions use up only about 5 percent of all state revenues. Our California model indicated that the state has virtually zero chance of defaulting over the next five years — an analysis window sometimes cited by credit rating agencies — so we think California should actually be rated AAA.

MB: What is the status and traction of your project? How do you see your next steps?

MJ: We have been getting some publicity — most notably an interview on Canadian Broadcasting Corporation’s Lang and O’Leary Exchange and a write-up in the popular Financial Times Alphaville blog. I have also been asked to speak at a couple of academic conferences, but we still have a very long way to go. We especially need quantitative analysts and software developers to get involved so that we can have a robust open source community that continually improves our technology.

MB: How can open-source credit help with the fiscal crises in the United States, Europe, and around the world?

MJ: Accurate government bond ratings are very important because ratings are supposed to act as a signal to the bond markets. If these markets get the wrong signals and fail to set bond yields in proportion to each government’s risk, they, in turn, send incorrect signals to policymakers and the general public. As we have seen in Argentina and Greece, government debt crises can cause mass despair, violence, and even death. We need to do the best possible job of assessing the risk of these crises occurring, so that politicians get the proper signals from government bond markets.

Right now, it seems to me that the markets are treating several Eurozone countries the same, even though there are differences. In the case of Italy, the country is running a relatively small deficit and has implemented a long-term solution to escalating retirement costs, yet its bond rates have been spiraling upwards with the others (as we speak in late July 2012). The conclusion might be that tight fiscal management has no benefit. If, however, Italy’s reforms have actually reduced its default risk, such a conclusion would be unfortunate. Our analysis of Italy was recently published on our website.

MB: Does this apply to personal credit, and, if so, how would that work?

MJ: In the U.S. at least, most people have a credit score. The algorithm for determining those scores is proprietary, but the inputs are generally known. While I could see the advantages of an open-source personal credit scoring model, I think the vast majority of people would prefer that their credit data remain private. By contrast, governments are supposed to be transparent. Thus, my ultra-transparent approach is most applicable to government bond ratings because there is no good reason for the inputs, calculations, or outputs to remain secret.


Michel Bauwens


Michel Bauwens

Bauwens is founder of the Foundation for Peer-to-Peer Alternatives and works in collaboration with a global group of researchers in the exploration of peer production, governance, and property. He has