Monte Carlo Valuation of Future Exposure and XVA

Summary

The document raises a practical question about calculating potential future exposure (PFE) in an XVA pricing engine. It distinguishes the valuation of a derivative at a future horizon from the market inputs used to obtain that valuation. Curves and implied quantities may be bootstrapped at present, but future exposure depends on how market factors and credit adjustments evolve across possible future states.

The central issue is whether simulation is needed for every product or mainly for instruments without convenient closed-form future valuations. The question points toward Monte Carlo as a way to generate future market scenarios and revalue a portfolio at each horizon, then incorporate CVA in the exposure measure. It does not provide an answer, numerical example, or comparison of simulation with analytic methods. The discussion therefore frames the modeling distinction but leaves open implementation choices, including when simpler valuation formulas can be used within a scenario-based exposure calculation.

Key ideas

• PFE is framed as positive derivative value after subtracting CVA at a future horizon.
• Present market curves and implied parameters do not by themselves determine derivative values across future scenarios.
• Monte Carlo can support future exposure estimates by simulating market states and revaluing instruments at each horizon.
• Closed-form pricing may still be applicable for some products, while complex exotics often motivate simulation.
• The document poses these issues but does not resolve them or provide empirical comparisons.

Tags

# Confusion About PFE Calculation and XVA Pricing Engine's Exclusive Reliance on Parameter Simulation


# Confusion About PFE Calculation and XVA Pricing Engine's Exclusive Reliance on Parameter Simulation












Potential Future Exposure (a credit risk metric) is calculated using $$PFE(\tau) = \text{max}\Big(0, \mathcal{P}_{derivative}(\tau) - CVA(\tau)\Big)$$, where $\mathcal{P}$ is the price / fair value / TV / MtM of derivative instrument; and $CVA$ is Credit Valuation Adjustment.

I am looking inside an XVA Pricing Engine (that does the derivative pricing and then spits out PFE), and all the parameters are obtained by simulation; when I search PFE calculations material online, all also imply that pricing of $\mathcal{P}$ has to be done using Monte Carlo simulation (which to me means the same as simulating your parameters).

I don't fully get why -- you can have market implied / bootstrapped forward curves for rates, inflation, volatility, default probability -- so you can calculate $\mathcal{P}$ (and by extension $PFE$ not just for $t=0$, but any time period in the future. The only circumstances simulating parameters using Heston, Hull-White, Garman-Kohlagen, Jarrow-Yildirim woul kind of make sense to me is if you needed to price some exotic derivatives that don't have closed for solutions -- and then -- said -- eh, I might as well just have consistency and price everything by simulation in my C++ Pricing Engine.

Could someone please help explain.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.