This is a review of the Global Derivatives 2017 Conference I was fortunate to attend in Barcelona.
The conference was very diverse and covered more than just derivatives - overall there were more than 100 presentations so it was impossible to see more than a fraction of what was on offer. What follows is mainly based on what I managed to attend, and no omission is intended to be deliberate. During the main conference I concentrated on the interest rate, XVA ('X' Valuation Adjustments), hall of fame, and regulation streams.
Pre-conference – Buy-side and Quant Tech Summits:
I split my time on the Monday between these two summits as both were interesting.
At the Quant Tech Summit, Massimo Morini of Banca Imi led a panel discussion on Blockchain. It was noted that despite the hype (e.g. self-executing smart trades) there had been some high profile bitcoin theft and therefore some challenges to be met before it’s ready for prime time. We learned that bitcoin exchanges are under various forms of cyber-attack 24/7. Also it was noted that some models of crypto-currency would represent fundamental shifts that society may not be ready for, as most countries do not disclose individual’s financial affairs to the extent a public ledger could.
I also heard fascinating views about the state of play of quantum-vs-classical computatational technologies. Helmut Katzgraber of Texas A&M University suggested it could be 5 years before we get a firmer view of the potential of quantum (based partly on trends in early devices like D:Wave and quantum annealers), while Vasil Denchev of Google was more optimistic.
"Overall it was one of the most interesting and best-executed conferences I've attended."
At the Buyside summit, Rick Lacaille outlined how a large buyside player (State Street) was approaching liquidity risk management. He highlighted the different asset and funding liquidity risk challenges of the asset manager (which contrast with those in the maturity transformation and derivatives business of banks). Modelling is in early stages and framed in terms of time, quantity and discount.
Erik Vynckier noted how much central banks are driving global markets given their growing balance sheets. He also mentioned the large number of robo-adviser investment strategies now on offer, and the large number of start-ups attempting to dis-intermediate banks. (Read the session coverage here). Buyside firms also have large Monte Carlo challenges for asset/liability modelling (like banks do for XVA).
Given the operation of some of these buyside players at the intersection of financial and insurance markets I felt they could use people with combined actuarial and quantitative skillsets.
I felt very privileged to present to such well-known experts as Fabio Mercurio and Chris Kenyon, and also to be able to discuss my research with Damiano Brigo and Andrew Green. I'm also grateful to Julien Hok and Colin Turfus for their thoughts.
I showed an approach to interest rate modelling with stochastic volatility (and co-jumps) that extends the square root approach and makes it more robust while giving lower Bermudan prices and exposure profiles.
For credit I showed how to integrate the grade migration matrix approach directly into the intensity approach while still being able to calibrate to CDS (Credit Default Swap) and CDS options quasi-analytically. This enables linking counterparty survival and forward credit grade for KVA (Capital Valuation Adjustment) - a new source of right/wrong way risk. It also provides a direct and very natural way to model credit mitigants such as CSA (Credit Support Annex) downgrade triggers consistent with survival probability in the risk-neutral XVA simulation.
At what other finance conference could one bounce ideas off so many experts, and make serendipitous connections with like minds?
The large European banks are still thinking about FRTB (Fundamental Review of the Trading Book) despite delays. In a panel discussion on regulation, significant budgets for FRTB programs (in come cases multiple tens of millions stg. indicative) were mentioned.
Jesper Andreasen highlighted that banks may need a suite of FRTB models rather than just one; on one hand sacrificing some accuracy for speed on the trading side while using fewer approximations and more expensive calculations for regulatory purposes. (Read session coverage here)
One question that arose was whether regulation is now relying too much on re-pricing.
Alexander Sokol of CompatibL is proposing to ISDA that CSA language is amended to state that known trade payments due to arrive before the margin is due to arrive are excluded (subtracted) from valuation for margin purposes. This is intended to reduce exposure spikes that can occur even with IM (Initial Margin). These spikes were also discussed in Michael Pykhtin’s presentation.
Forward SIMM and Dynamic Initial Margin
Given the difficulty of dealing with this efficiently, this was unsurprisingly a hot topic at the conference. In the discussions this topic was also intertwined with AAD methods.
Just a few of the relevant talks included:
- Massimo Morini pointed out that there is no conceptual issue in blending historical data for market shocks with risk-neutral data in Monte Carlo as it is similar to what is often done in pricing path-dependent products. However he also noted that the simple diffusion models typically used in XVA may fail to capture extreme historical risks when they roll off in the forward simulation.
- Conti and Dubrovich of UniCredit discussed the brute force (nested Monte Carlo) as well as various proxies for both CCP (Centralised Counterparty) and ISDA SIMM (Standardised IM Method), coming down in favour of calculating forward IM using AAD methods instead of regression proxies.
- Youssef Elouerkhaoui of Citi gave a fascinating a deep dive into XVA with gap risk for credit products and associated sensitivities.
All of these speakers stressed the significance of AAD methods in calculating sensitivities for forward/dynamic SIMM given the heavy computational burden.
Adjoint Algorithmic Differentiation (AAD)
It seemed clear at the conference this is an idea gaining much popularity. Luca Capriotti (Credit Suisse) pointed out that “bump and reval" for XVA is "so 90s”.
As an indication of the state of the art, two presentations covered use of AAD methods for XVA sensitivities, for path dependent derivatives with collateral – one by Alexandre Antonov of Numerix and another from Brian Huge and Antoine Savine of Danske Bank. This is in addition to the emphasis others placed on AAD within forward SIMM.
Andrew Green discussed combining GPU with AAD for XVA sensitivity speedups by parallelisation and showed how they can be seen as complimentary not alternatives.
I felt very fortunate to also attend a hands-on AAD coding workshop with Mike Giles, Luca Capriotti and Uwe Naumann; combining practitioner, mathematician, and computer scientist perspectives. For many, it is not easy to get an intuitive feel for these techniques but this tutorial showed how to. It is clear that AAD can provide a big speedup but may not be easy to integrate with other software technologies such as OpenMP and can be a challenge for third party vendors given issues like unpredictable memory overheads in some contexts. We learned that the most useful supplementary tools apart from AAD fundamentals were payoff smoothing (requiring a bias/variance trade-off) and Capriotti’s “binning” method (for confidence interval estimation for sensitivities). Participants were given access to various C++ source code examples and a trial of the Numerical Algorithm Group’s DCO (Derivative Computation by Overloading) AAD library.
Capriotti gave fascinating insights into the increased deployment of AAD techniques at Credit Suisse over time, as management realised the benefits of this technology .
Coherent XVA and KVA
Damiano Brigo gave a colourful and entertaining view on the past and present of valuation. This questioned whether the industry would ever be ready to tackle non-linearities inherent in XVA.
Jon Gregory covered the overall XVA journey and the emphasised the advantages of bringing more rigour to the VAs, particular in areas like "hard" wrong-way risk. (This is an area where the modelling approach I developed can definitely help).
Brigo also showed how to construct a simple alternative to CRA (Credit Rating Agency) market-implied ratings by combining ratings and CDS data. His application was to monitor Euro area exit probabilities. It would be very interesting to apply this model if another exit concern arises in future. (Read session coverage here)
Smart Beta Fixed Income
EDHEC's Riccardo Rebonato demonstrated some early research into fixed income smart beta; the potential is high but so are the data challenges. (In a separate talk he outlined a useful technique based on normalised principal components with several applications including yield curve deformation and stress testing. This included fascinating insights into his time at Pimco).
Interest Rate Risk in the Banking Book
Sebastian Schlenkrich of D-Fine gave a very interesting presentation on the valuation of loan options subject to non-rational exercise. His ideas would be of interest to banks grappling with the behavioural aspects of loan option valuation.
Australian Bank Representation at the conference
Panel Discussion on the Future of Quants
We were treated to a fascinating debate on this topic with John Hull, Damiano Brigo, Jesper Andreasen and Matthew Sargaison. XVA and regulation is creating a lot of work at the moment but will eventually become business as usual, what then for the industry? (Read session coverage here)
As someone who is both an actuary and quantitative analyst, I see one aspect of this being that quantitative finance, while a fascinating career, is not a profession in the truest sense. As an actuary I'm subject to a professional code of conduct, CPD requirements, and professional standards, and the possibility of sanction if requirements are not met. I also see increasing co-operation between the profession's associations around the globe. None of this exists for quantitative finance practitioners. This is interesting given that there are some parallels in the services quants in banks and actuaries in insurers now provide for pricing and capital. For quants the situation means more freedom of practice and more educational variety among practitioners than for actuaries. On the flipside they do not have the statutory roles that actuaries have - which can be a double-edged sword. While actuarial has a reputation of being difficult, the body of knowledge to acquire to be successful in quantitative finance is vastly more challenging to attain.
Apart from these differences, quants and actuaries have some similar challenges as technical experts in different areas of financial services - such as professional development to maintain relevance, and development of good communication skills.
John Hulsman of the US Council on Foreign Relations was fascinating as a Washington insider. He opined that 2% growth is the baseline to stave off populism, and it’s a mistake to think it will fade away any time soon despite the likes of Le Pen and Wilders not winning recently.
I also heard that ISDA's next risk weight calibration may lead to increased risk weights for SIMM.
Overall it was one of the most interesting and best-executed conferences I've attended.