Making Credit Ratings Data Publicly Available

Since the early 1970s, credit rating agencies have generally employed an “issuer pays” model under which they charge debt issuers for the cost of their services. As discussed in a recent paper, under this model, one might expect ratings to be provided to investors for free – to maximize their market impact – but the reality is more complex.

Issuers may cite the rating on any given debt issue, and investors may view individual ratings at rating agency websites. However, rating agencies take measures to prevent uncompensated bulk download and redistribution of their credit ratings. For example, the terms of use on Moody’s web site state:

“[Y]ou may not modify, create derivatives of, copy, distribute, broadcast, transmit, reproduce, publish, license, transfer, sell, mirror, frame, “deep link”, “scrape”, data mine, or otherwise use any information or material obtained from or through this Site.”

Instead, they encourage institutional investors to purchase subscription ratings feeds, such as Moody’s Ratings Delivery Service and Standard & Poor’s RatingsXPress. The prices of these services are not published, but the authors believe (based on previous industry experience) that annual subscriptions cost more than $100,000 for all asset classes.

The lack of inexpensive access to bulk rating data hinders the ability of academics and other third parties to research the behavior and accuracy of credit ratings. After the well-publicized failure of structured finance ratings during the 2007-08 global financial crisis, regulators and policy makers recognized the importance of making bond ratings publicly available.

The Securities and Exchange Commission adopted the first publication rule in February 2009, amending Regulation 17g-2 to impose a modest requirement on NRSROs to publish a 10% sample of their ratings on a six-month delayed basis in XBRL (eXtensible Business Reporting Language) format. A few months later, the SEC strengthened this requirement by adding Regulation 17g-7, a more comprehensive rule that requires NRSROs to publish their full rating histories on a 12-month delay (or a 24-month delay for investor paid ratings). The rating history files must be updated monthly, formatted in XBRL and posted to each rating agency's website.

Unfortunately, the ways in which the major NRSROs have responded to these rules have not made credit rating data available in an easily accessed or comprehensive way and have instead hindered academic and think-tank research into credit ratings. Financial researchers who lack the funds required to purchase bulk ratings must use a variety of ad hoc methods to obtain rating data or limit their studies of credit ratings. Many researchers are unfamiliar with XBRL. It can also be difficult to locate historical files.

Because of the problems associated with NRSRO ratings history disclosure, online ratings data has been of only limited use. At a recent conference session on public finance, one of us found that none of the academic researchers in attendance were even aware of the availability of this data, while one smaller NRSRO reported that downloads of its credit rating history file averaged only one per month.

In late 2017, the Center for Corporate and Securities Law at the University of San Diego School of Law commissioned a software project to make the SEC-mandated credit rating data more accessible. Developers created a software tool written in Python that crawls credit rating agency websites, downloads the XRBL files, and converts them to Comma Separated Value (CSV) format.

This software tool is open source and freely available on GitHub. The most recently processed ratings data are available at www.ratingshistory.info. This web page contains a set of ratings history CSV files broken down by agency and asset category. Most of these files can be directly loaded into Microsoft Excel; some of them are large, with more than one million rows, requiring the use of database tools. Finally, the entire universe of rating actions is available on Data.World, where it can be queried online. The total universe includes over eight million rating actions, which include assignments, upgrades, downgrades, and withdrawals.

Marc Joffe is a Senior Policy Analyst at the Reason Foundation and a guest contributor to the Oxford Business Law Blog.

Frank Partnoy is the George E. Barrett Professor of Law and Finance at the University of San Diego School of Law and a guest contributor to the Oxford Business Law Blog.