In the summer of 2021, COVID was still very much seen as a problem, and most people didn’t take the holiday they normally would. Neither did I. I was sitting on a Greek island and working long hours helping a foreign consulting company run the analysis for a client bank, of how to best de-risk or exit their exotic derivatives positions.
The bank had built up a “client-facing” book, and client-facing, as so often with exotics, meant retail-distribution-led. This has led to certain one-sided exposures of an exotic (non-vanilla) nature, that were frankly difficult to hedge outright (except back-to-back, which the bank had also done on occasion).
In this situation, you only have a number of choices:
- Sell the whole book: it will still be slightly diversified (especially in relation to pin-risk, digital risks, and gamma concentrations near expiries, barriers, or trigger dates), and so fetch — in theory — a better price than if you had hedged every position back-to-back by itself. This is the only way to get rid of the “exotic” risks (like vomma, vanna, and other higher-order risks) completely.
- Over-hedge the book with vanilla options. This is very complex in practice, as you are trying to construct an overall book which “cannot lose”, but you leave exotic risks open. In general, this is impossible to do perfectly, although there may be individual cases where one can come close to a risk-less combination of exotic risks and vanilla hedges. Great care must be taken that the over-hedge performance is in principle model-independent! I.e. the over-hedge must be a hedge against model risk as well.
- Run the full model-risk, and delta (and possibly gamma and vega-) hedge day to day. This is often very risky, and exposes model risks which practitioners often don’t even know exist until they experience them first-hand by the P&L performance of their options books.
Like many (typicall French 😉 ) banks, the bank had decided to go for option 3 in the past, as well as back-to-back heding some new issues if the margins allowed for it. In doing that on a non-diversidied book (led, as it was, by retail-distribution demand, which was very one-sided in certain risks), the model risks were obviously at an unusually high level. The challenge now was how to (a) explain this to higher management and (b) convince the traders to own up to the issue, and the nature of the issue.
GENConsult, over the summer, implemented a partial replica of the bank’s models in the Matematica [TM] framework, with only the salient features of the bank’s model present (ignoring for example interest rate risk, forward-spread risk, borrow-costs, and other factors which were extremely unlikely to account for significant portions of the exotic risks). This allowed the bank to distill the crucial risks without the full complexity of their models. The Mathematica model was also handed over to the client.
Also, with the benefit of distance between my own time as a quant analyst and then later as a global head of exotic derivatives trading (which ended in 2005), I think I was able to communicate difficult concepts a lot more clearly than I could have ever done while deep in the weeds as it were. I was surprised I still had it in me to write fully workable mathematical derivatives models (in this case, a local volatility model), and the whole calibration mechanism which brings prices in line with market prices, and to be able to run my own model on an entire trading book as well. With this model, we were able to forward-simulate the entire book under different future scenarios. With the bank’s own model, that woud have been impossibly manual, and impossibly time-consuming (but more precise, to be fair).
As it happened, on this occasion, the bank also had a significant piece of luck on its side, as it was actually possible to cleanly switch to approach 2, somethign which is not at all guaranteed to work in general, but which worked in this case. In exchange for taking a small additional model reserve, it was possible to put on vanilla hedges which would ensure that the book has positive P&L going forward, in virtually any conceivable scenario (in other words, the reserve would always be more than what’s needed), as long as they also managed to forget delta-heding. This would be a significant deviation from the bank’s standard model, but in pricinple safe to do, and would require a side pocket from a risk (and capital requirements) perspective. The reserve required was approximately 0.5% of the total book notional, so really amazingly cheap, and certainly good compared to the daily P&L swings the book had encountered in the past.
I think the bank was happy when they digested my report, my briefings, and took over my model for their own use. The front office would have liked to continue with the business as is in some fashion (as “on the face of it”, at issue, it produced nice paper profit). But in this case, this was not what I advised, as the P&L was difficult to defend with the non-divsified exotic-risk “shape” of the book. My recommendation was to take the additional reserve, put in the over-hedge, and thereby put the book into a passive run-off mode and watch what happens (not even delta-hedge!). It was in an incredibly lucky position that this solution even existed. When my work was done, the decision hadn’t been made, because I guess sometimes old habits (being emotionally wedded to the upfront paper-profits) die hard. I do believe they will have gone with the recommendation eventually. I also know and understand that they would never tell me, because it’s crucial the market doesn’t know or guess what they did or who they are. So, what solution they implemented in the end, I do not know. I guess one day, when it’s all over, I’ll find out.
genconsult.me 