CubeLogic and PFE – Stress Testing in Volatile Markets
By Dr. Gary M. Vasey, originally posted on CTRM Center
I recently talked to CubeLogic’s Thomas Lloyd and Kieren Child about Potential Future Exposure (PFE). PFE measures how much expected income is tied to a particular counterparty and what exposure might reasonably become in the future and, according to Thomas, “the exposure has two dimensions: money expected to be received that may not arrive if the counterparty goes bankrupt, and having to pay significantly more at market prices if the counterparty can’t deliver the needed commodities” He told me that “PFE differs from walk-forward exposure by modelling realistic market movements rather than just projecting current positions forward while walk-forward shows current position developing over time according to existing trades and expected payments. Essentially, PFE acts as a stress test that respects correlations between the different commodity curves to show reasonable worst-case scenarios.
As you might expect in an era of geopolitically-driven uncertainty and volatility, stress testing of almost any sort is rising in popularity. Firms want to get a better handle on their potential exposure to sudden events. Thomas told me that PFE initially gained in popularity after the financial crash but that the recent market volatility has increased demand further. Events like the power market bankruptcies of a few years ago (Robin Hood Energy and others), Russia’s invasion of Ukraine causing gas price spikes and Houthi attacks on shipping in Strait of Hormuz, have all served to increase interest in using PFE. Thomas also highlighted concentration risk concerns where single events can snowball catastrophically—“a sunken oil tanker could cause $10-50 million loss plus delivery failures and lawsuits creating liquidity crunches,” he said. European government storage mandates and political volatility have also made forward-looking risk assessment more critical and he believes that companies that lack PFE modelling have been more severely impacted by unexpected market swings than those with plans for handling market changes.
CubeLogic’s CCFE (CubeCalc Future Exposure) uses a Monte Carlo simulation with a two-factor model to calculate potential future exposure across different granularity levels. It can run standalone or may be integrated with CubeLogic’s suite and requires only annual calibration, Kieren said. “The key differentiator is the model’s ability to simulate new instruments without historical data by using curve-based analysis that incorporates the Samuelson effect and seasonality,” Kieren told me. “The product includes CSA modelling for margin forecasting and provides price statistics for model validation as well.” Around half a dozen customers are implementing the product right now.
Kieren was also keen to stress that its PFE approach provides realistic simulations resecting correlations between curves rather than simple stress tests. “Walk-forward exposure shows current position projected forward with payments and expected payments incorporated and the figures can change completely if market shifts happen tomorrow or within a week,” he noted. For example, a simple stress test might show the impact of a 10% gas price increase but completely miss the fact that that oil and other commodities would likely move together. “PFE captures how hedging positions (like gas hedged against oil) affect overall exposure across entire portfolio,” he said.
The model foundation uses a two-factor approach analyzing curves individually then modelling correlations between them and incorporates the Samuelson effect – trades become more volatile as they approach maturity. It also respects seasonality as winter and summer trades have distinct price evolution patterns compared to one another – Kieren told me. Furthermore, the model can simulate instruments without historical data by understanding curve behavior generally and “the approach solves the problem of instruments that have been years from maturity in historical data but are now weeks from maturity,” he said. The model only requires calibration once annually and then runs daily for the entire year while new instruments can be incorporated from day one without any historical data and still produce realistic PFE immediately.
Thomas and Kieren are proud of CubeLogic’s CCFE (CubeCalc Future Exposure) product, which is available standalone with its own APIs or integrated with CubeLogic suite having had a hand in designing and building it. Some of the features that he mentioned include;
- Three integration options: completely standalone, fed by Cube data without feeding back, or full integration with loader feeding results back into Cube,
- PFE calculated at multiple granularity levels: netting set level, counterparty level, groups of counterparties, and enterprise-wide,
- Multi-level calculation captures hedging effects between counterparties that can’t be seen by simply adding individual PFE numbers,
- Individual trade-level “what-if” analysis available on ad-hoc basis for evaluating new trade impacts,
- CSA modeling included for margin forecasting, providing liquidity visibility at portfolio level,
- Price statistics provided alongside results show percentile distributions (e.g., 50% of time gas didn’t exceed certain price, 95% reached higher threshold).
Implementing the software can produce PFE figures quickly, but the UAT cycle and particularly building comfort with the figures can take longer before customers use them for limits or board reporting, they told me. “People need to trust the models they use and it is the building trust and confidence that can sometimes take a little time.” Despite that, CubeLogic is seeing a lot of interest in the product and expect it to be widely adopted given the circumstances in the market.