SOFR Compounding and Time Discretization in Derivatives Models

Summary

The document raises a numerical modeling concern about the transition from LIBOR-linked instruments to SOFR-linked instruments. Since SOFR compounds and resets daily, a model that previously used monthly time steps for multi-year instruments may introduce greater pricing error when applied to SOFR products.

It identifies a computational tradeoff: using finer time steps may reduce discretization error but can substantially increase evaluation time. The document asks whether alternative approaches exist, but contains no proposed solution, comparison, numerical evidence, or model-specific conclusions. The issue applies broadly to lattice, Monte Carlo, and finite-difference methods, though the appropriate treatment may depend on the instrument and how its cash flows reflect daily compounding.

Key ideas

• SOFR’s daily compounding and reset schedule can make coarse monthly time grids less suitable.
• Finer time discretization may reduce pricing error while increasing computation time.
• The concern applies to lattice, Monte Carlo, and finite-difference methods.
• The document poses the modeling question but provides no tested alternative or quantitative evidence.

Tags

# SOFR transition impact on numerical efficiency


# SOFR transition impact on numerical efficiency












As the SOFR transition kick in, the libor insturments and derivatives link to SOFR which poses natural challenge for numerical methods such as lattice, monte carlo and finite difference, where people used to (at least some people) discretize the time steps to be month basis if the security expired in years.

As the nature of SOFR of daily compounding and daily reset, this discretization scheme would likely pose higher mispricing error. If reducing the discretization scheme, it may pose much longer evaluation time.

What people think of this? Any alternate way of handling this?

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.