Stochastic experiments in stabilisation of money market benchmarks

The main input of this research is a stochastic model of a theoretical panel of contributors (banks) to a money market index. The model proved to constitute a useful environment for testing various index formulae, their characteristics and some trade-offs that may arise while deciding on the particular benchmark’s design.  It may be also used to evaluate indices without historical data or stress them against different scenarios of adverse changes in market conditions or panellists’ behaviour. The hypothetical problems with: changes in the panel’s composition as well as the irregularity of daily contributions may strongly influence the utility of a final benchmark to be used in medium and long term loan contracts, especially with retail clients. Our focus is on several selected classes of benchmarks’ formulae that are derived from the raw index and allow for some confinement of the mentioned drawbacks while decreasing quality measured by other criteria (goodness of fit). The set of classes include: the geometric time weights with different smoothing parameters and observation window’s length used on the original raw index, stabilisation of the raw index in bands, rolling window volume weights rebalancing and finally the geometric time weights performed on log-volume transformed index. The potential trade-offs in such a benchmark’s stabilisation efforts are shown.

Published in Bezpieczny Bank (Safe Bank) No 4 (73) 2018 pp 44-61

Additional files

optimalSetsComparedMAC_MAE

optimalSetsComparedSD_MAE