A recent article in Wall Street & Technology “Focus on Collateral: Optimizing ISDA CSA Document Metadata” by Mahesh Muthu caught our eye. We look at the often daunting task of accurately capturing all the terms in a CSA.
The article said,
“…In the OTC derivatives market, collateral management terms and conditions are largely driven by the framework within the ISDA Credit Support Annex (CSA) document. ISDA CSAs can be heavily negotiated documents, but at a minimum contain terms covering eligible collateral, how collateral can be held and used, as well as terms covering margin calculation and posting. Although the information embedded within these documents seemed of little importance prior to the 2008 financial crisis, many institutions are now investing heavily in people, processes, and technology to mine and actively manage such document metadata…”
and
“…A CSA can be decomposed into more than 200 data points. However, many firms have only been capturing approximately 70 data points, on average, from each document. This impacts the ability to perform deeper, more sophisticated analytics and scenario analysis. From those 70 data points, an alarming 65% of agreements have at least one attribute from the CSA coded incorrectly, and 5% of CSAs are unenforceable due to missing pages, agreements, and/or amendments, pointing to a lack of a controlled operation. Compounding this issue, new terms and conditions are appearing in CSAs, further emphasizing the need to have a granular, structured data model and a strong process to ensure there is not basis risk between what is legally agreed and represented within systems…”
We should note that the author works for eClerx, a firm in the business of managing all that CSA data accurately. This article feels a bit like PR spin. But for our purposes, let’s assume the statistics are reasonable. Capturing all the terms in a CSA is critical to insure you know what exposure you are taking. Even worse is when agreements are amended and no one tells the collateral management people. Lots of breaks and frustration to follow. In a crisis environment, no one is going to care that your collateral system didn’t filter something out that it should have — only that what you have doesn’t cover the exposure.
Getting a CSA right also allows for accurate optimization of collateral. Most collateral management systems use some flavor of a multivariable programming to determine the best (read: cheapest) collateral to deliver or substitute. And without an accurate list of constraints, the process isn’t worth much. For example, if you don’t have a system which can match CSA rules to proposed collateral, the CSA rules you do have are incomplete or just plain wrong, you may never know you have a problem if collateral received is outside of what your agreement (and by extension your risk people) prescribe. Imagine if the thresholds were screwed up, currencies didn’t match, or calls were missed altogether. If, as Mathu says, 65% of agreements have errors, that is a disturbingly high number. Not all CSA rules are of equal importance, but there is no excuse for being sloppy about this. With centralized clearing, the problem is only going to get worse and less forgiving.