Risk managers and research analysts often struggle with developing models that would capture fair liquidity measures of fixed income and exchange-traded instruments.
The reason is that there is no unified approach; hence some banks use a parametric multi-factor approach, while others seek more simplicity.
In this article we would like to outline the key pros and cons of parametric modelling.
Parametric liquidity models can be advantageous when the analyzed instrument samples possess similar characteristics that can be considered as inputs, and hence the model can be applicable to a wide range of various security types.
The benefit is that financial engineers will not need to reinvent the model’s logic, but rather alter some imposed constraints and factor weights.
Empirical observations demonstrate that parametric modelling can generate accurate results, which makes it popular.
On the other hand parametric models require constant recalibration.
It is a common practice to use principal component analysis (PCA) to determine the weighting coefficients so that the models would fit a given snapshot of market microstructure or conditions.
But such recalibrations can be good enough for a certain time period or a particular situation and be inefficient in the long run.
The embedded constraints can impose biases that will distort the output data.
Therefore it is sometimes beneficial to implement simple single-factor models.
Since there is a finite number of liquidity characteristics it could be feasible to select 1 – 2 elements in the model design, such as bid/ask spreads and daily volumes to generate acceptable results. We provide an example of results generated by such a simple model (bubble size corresponds to bond modified duration). We can see that the regression line demonstrates a fairly strong relationship between the bid/ask spread and liquidity scores.