What are some good ways to assess the concentration risk for a rates curve or by currency when volumes traded by instrument are not easily available? For ex, if for some currencies, the PV01 at certain tenors is very high or the sum of absolute risks for a curve is very high, how can we quantify on whether we can exit the positions in a given number of days?
I was asked this question numerous times, often by clearing houses as a measure to gauge their initial margins and concentration fees.
As a market maker you may have limited information about the volumes traded, i.e. you know your own volumes. Our strategy was to first infer expectations about the market as a whole, and ask brokers to corroborate this view or provide some generic, non-sensitive information that would be useful.
Once you have a range of tradeable volumes you can consider market impact of various sizes. Numerous empirical studies often find there is a square root relationship between market impact and volume of order (sorry I dont have time to look up some references right now (perhaps someone can edit them to my answer)). I.e. consider the hypothetical market impact of various notionals of 30Y IRS:
10mm 0.08bps 100mm 0.25bps 1000mm 0.80bps 5000mm 1.79bps 10000mm 2.53bps
Market impact is inevitably related to execution time provided, for example if you are asked to 100mm in 5days as opposed to 1 second the cost due to market impact will be different: in one second you would obviously have to lift all available prices at that time, regardless of what they are.
When it comes to considering the volumes of instruments such as 10Y/30Y spread, where the information might be mixed with data on outright 10Y or outright 30Y trades. The recommendation is to build a model with some sensible parameters.
If your concern is solely risk then you will inevitably have to make the decision about how much residual risk you will tolerate upon liquidation:
- Will you be content being delta neutral measured over any tenors?
- Will you be content if major macros buckets are allocate to zero, eg 2Y 5Y 10Y or 30Y?
- Will you only be content if every yearly bucket is hedged, i.e. 15Y 16Y 17Y cannot be +1mm, -2mm, +1mm pv01 for example.
- Will you only be content if every position is exited with every counterparty, so you book is essentially closed?
The levels of tolerance also greatly impact the answer. Practically it is probably only worth considering upto about the second level of the above bullet points if you are talking about an execution timeframe of days.