Bermudan Swaptions, Trees and Price Bounds
A Bermudan swaption can be exercised on specified dates. On each date the holder compares exercise with the value of waiting. An independently evaluated exercise policy gives a lower bound, while a valid dual construction gives an upper bound. The gap helps distinguish pricing uncertainty from simulation error and makes the accuracy of a numerical method easier to assess.
How it works
A Bermudan swaption can be exercised on several specified dates. On each date the holder compares the swap’s exercise value with the value of keeping the option alive. Trees and simulation regression estimate this decision. Evaluating a feasible policy independently gives a lower bound, and a valid dual construction can supply an upper bound.
Worked example
At an illustrative first exercise date the underlying swap is worth 4 on immediate exercise. If the discounted modelled value of keeping the option is 5, wait. If it is only 3, exercise. An out-of-sample simulation can then apply the fixed rule to both favourable and adverse paths to estimate its value.
Validation and limitations
Check exercise schedules, curve conventions and the regression basis or tree refinement. Compare results across independent training and evaluation sets. Where available, quote the lower and upper bounds with simulation error. See Andersen–Broadie.
Further reading
- Primal–Dual Simulation Algorithm for Pricing Multidimensional American Options
Leif Andersen and Mark Broadie · PDF
- Valuing American Options by Simulation
Francis Longstaff and Eduardo Schwartz · PDF