By Anne Friberg and Mary Ann Dowling
Members of The NeuGroup’s FX Managers’ Peer Groups tinker with counterparty assessment models to find a proper role for CDS spreads in the analysis.
During the financial crisis, companies sought alternatives to credit ratings to assess the counterparty risk of their banks. The idea was that agency ratings aren’t forward-looking enough and don’t give the whole picture of the financial health of any given institution. After all, Lehman Brothers had an A rating before imploding; meanwhile its CDS spreads were through the roof. Since then, CDS spreads have emerged as a metric used in various models for risk assessment, and increasingly monitored and incorporated as early-warning signals to supplement old-fashioned credit ratings. The question is, how can companies make the most appropriate use of this metric?
In recent meetings of The NeuGroup’s FX Managers’ Peer Groups 1 and 2, members heard from peers about different methods of using CDS spreads in bank counterparty-risk assessments. What follows are a few examples.
Company 1: A 3-sigma approach with CDS spreads. Company 1 used to rely on credit ratings for counterparty risk and would only trade with counterparties that were rated A-/A-/A3, or better. Post-financial crisis it wanted to augment minimum credit rating requirements with CDS spreads, but struggled with volatility and relativity. To use them in its model, Company 1 considered (1) the current absolute value of the counterparty’s CDS spreads; (2) the CDS spread averages over various periods of time; (3) current CDS spreads relative to the standard deviation (σ) from various means; and (4) a combination of standard deviation and absolute values.
After much research and scenario analysis, Company 1 selected a model that considers a 30-day CDS spread average relative to a longer standard-deviation period, with trading suspended if the counterparty’s CDS spread moves more than 3σ, or if the counterparty has a net CDS spread of greater than 400bps. Counterparty risk is reviewed weekly and before any large trade is transacted. To override a bank’s suspension, approvals must be obtained from treasury and the CFO.
Company 2: average CDS spread changes. Company 2 is in the process of reviewing its current “one-dimensional” approach to global counterparty limits, which is based on credit ratings, book value of equity and the needs of the business. In the event of a split agency rating, the lowest credit rating applies. Credit ratings are manually updated each month in an in-house web application and limits are automatically adjusted based on the limits established by credit rating. However, the company also generates daily CDS reports and monitors CDS spreads, and realized there might be something more optimal—which incorporates CDS spread movements—to determine limits.
The alternative approach under review is that counterparty limits will still be based on credit ratings but individual bank limits will be adjusted by comparing their individual CDS spread to a peer average determined by grouping the CDS spreads of financial companies by credit rating (using regression analysis). The company would determine the level of acceptable total bank exposure by first taking the total counterparty exposure for the most recent quarter, subtract fixed-income investment in sovereigns and supranational debt, and then add a provision for cash growth and balance flexibility. Based on a CDS index from a subset of the counterparty banks, it can then calculate individual limits:
Index = (100 + avg. CDS spread) / (100 + CDS difference from lowest CDS spread); Limit = total acceptable exposure / number of banks x CDS index.
The company would test quarterly for any significant changes in CDS spreads and adjust accordingly. One challenge is that the subset of counterparty banks might have the same or similar credit ratings, meaning there wouldn’t be enough data points for robust regression analysis. The model might also be hard to sustain.
Company 3: CDS in an “all encompassing” approach. Company 3 also uses a standard deviation methodology that plots each member of the bank group relative to the mean with a view to identifying the strongest bank partners. Risk measurements are broken into three categories: CDS/equity, debt spreads, and static data. Individual bank data is weighted within each category and at total category level to arrive at a total weighted score. A scorecard is then generated reflecting the weighted totals that are divided into three categories:
Green: better than bank group average; Yellow: <1 standard deviation above bank group average; Red: >1 standard deviation above bank group average.
The scorecard model is then used to assign lines to counterparties. Banks are grouped into four categories: favored, acceptable, watch and probation. The methodology mostly applies to new business allocation but existing business adjustments are considered under extreme circumstances. This approach allows for gradual changes to the portfolio based on counterparty health over time and prevents abrupt changes to existing relationships. The assigned lines are only for FX, investments, and IR swaps business.
Refining the CDS model
Counterparty risk assessment has taken on greater meaning since the crisis and the follow-on downgrades—both in sovereign and institution. As a result, experimentation with more granular data and more frequent review has taken hold and become more refined in the last couple of years. Those who have yet to incorporate a closer examination of market data like CDS spreads should seriously consider it, starting with spreads relative to a peer group and spread averages over different time frames, and the like.