Podcast: Artur Sepp on rates volatility and decentralised finance
Quant says high volatility requires pricing and risk management models to be revisited
The volatility in interest rates observed in 2022 caught some pricing and risk management models off-guard.
The three-month rate went from an average of three basis points in 2021 to 200bp last year. The 10-year rate, its implied volatility and realised volatility approximately doubled.
Many quant models are implemented with caps and floors on admissible values of their parameters and variables. In high volatility regimes these bounds may be breached, and models could not return reliable outputs, making them unusable. When that happens, quants need to step in and patch them up by widening the caps and reworking numerical solvers to fix the problem.
Artur Sepp, head of quantitative strategies at Clearstar Labs and the guest in the latest edition of Quantcast, has seen this all before. “In 2008, when I worked at Merrill Lynch and we had the market crash, for some of the local volatility models … volatility was capped at 80%,” he says. “And, of course, when we hit 80%, that model would not fit the market. It could not produce any reasonable risk.” The equity quants team had to adjust the model on the fly.
By the time volatility had spiked in the rates market in 2022, Sepp had long left banking. However, his interest in solving problems in that space was revived when Parviz Rakhmonov, a quantitative analyst at Citibank, proposed that they work together on a rates model capable of handling extreme market moves.
The result of the collaboration is an extension of the popular one-factor Cheyette model that is extended to include the lognormal stochastic volatility, which originated for equity derivatives in Sepp's earlier paper with Piotr Karasinski, and takes into account the correlation between rates and volatility.
“The objective of the model is to feature the positive volatility skew for swaptions [because] local volatility models are not flexible enough to produce positive skew,” he explains.
That makes the model, according to Sepp, suitable for unconventional regimes – such as the one in 2022 that was characterised by high rates, their volatilities, and positive implied volatility skews.
He says that higher levels of volatility and the need to deal with black swan events – be they in rates, equities or commodities, as happened in 2020 when Covid hit – are pushing quants to reconsider volatility models. That explains, he says, the unusually large quantity of published research on volatility, including on Risk.net, over the past two years.
Hello, buy side
After leaving the banking sector, Sepp moved to the buy side and joined Julius Baer, where he was responsible for developing trading strategies. In this episode of Quantcast, he shares how his knowledge of pricing models and his C++ coding skills helped him tackle problems in trading.
His current role at Clearstar Labs is to develop trading strategies in decentralised finance. This involves providing liquidity to DeFi protocols and taking market-neutral on-chain and off-chain positions in digital currencies using methods adapted from the asset management toolbox.
In 2022, Sepp and Alex Lipton, global head of quantitative research and development at the Abu Dhabi Investment Authority, proposed an automated market-making framework for currencies to be exchanged on blockchain consistently with off-chain markets. “This framework is very suitable for central bank digital currencies,” he says.
Something closely resembling their approach was adopted as part of the Bank of International Settlements’ Project Mariana, and Sepp expects more institutions to follow.
A paper Sepp has been working on more recently, and which will soon be published on Risk.net (stay tuned!), looks at two questions: whether cryptocurrencies can be considered an asset class; and, if so, how they should be included in a diversified portfolio.
His answer to the first question is that cryptocurrencies are indeed an asset class, mainly because they show an expected positive risk premium and contribute to a portfolio’s diversification. To answer the second, he adopts four distinct approaches, and, strikingly, finds very similar results: on average, the optimal allocation is between 2% and 3%.
Sepp is currently working on a diversification model for a portfolio of DeFi staking strategies that would account for the so-called protocol risk – the risk that a given protocol fails to deliver on its mission for any reason, because it is hacked, destabilised by market agents or because it turns out to be fraudulent. A critical aspect here is that DeFi protocols form linked networks where a failure in a network's node would inadvertently impact all other protocols in the network. The portfolio diversification problem can be elegantly formulated as a linear program with control for interactions between protocols.
Index
00:00 Intro and motivation of the model
06:45 Model description and novelty
09:45 Comparison with Lyashenko and Goncharov’s model
13:35 Results and applicability
20:42 Banking experience in a buy-side role
25:52 Automated market-making on blockchain
31:20 Allocation of cryptocurrencies in a diversified portfolio
38:15 Open quant questions on crypto and future research projects
To hear the full interview, listen in the player above, or download. Future podcasts in our Quantcast series will be uploaded to Risk.net. You can also visit the main page here to access all tracks, or go to the iTunes store or Google Podcasts to listen and subscribe.
Now also available on Spotify and Amazon Music
Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.
To access these options, along with all other subscription benefits, please contact info@risk.net or view our subscription options here: http://subscriptions.risk.net/subscribe
You are currently unable to print this content. Please contact info@risk.net to find out more.
You are currently unable to copy this content. Please contact info@risk.net to find out more.
Copyright Infopro Digital Limited. All rights reserved.
As outlined in our terms and conditions, https://www.infopro-digital.com/terms-and-conditions/subscriptions/ (point 2.4), printing is limited to a single copy.
If you would like to purchase additional rights please email info@risk.net
Copyright Infopro Digital Limited. All rights reserved.
You may share this content using our article tools. As outlined in our terms and conditions, https://www.infopro-digital.com/terms-and-conditions/subscriptions/ (clause 2.4), an Authorised User may only make one copy of the materials for their own personal use. You must also comply with the restrictions in clause 2.5.
If you would like to purchase additional rights please email info@risk.net
More on Cutting Edge
Market-making in spot precious metals
A market-making framework is extended to account for metal markets’ liquidity constraints
Choosing trading strategies using importance sampling
The sampling technique is more efficient than A-B testing at comparing decision rules
A comparison of FX fixing methodologies
FX fixing outcomes are mostly driven by length of calculation window
Quantum cognition machine learning: financial forecasting
A new paradigm for training machine learning algorithms based on quantum cognition is presented
Backtesting correlated quantities
A technique to decorrelate samples and reach higher discriminatory power is presented
A hard exit threshold strategy for market-makers
A closed-form solution to derive optimal stop-loss and profit-taking levels is presented
Pricing share buy-backs: an alternative to optimal control
A new method applies optimised heuristic strategies to maximise share buy-back contracts’ value
CVA sensitivities, hedging and risk
A probabilistic machine learning approach to CVA calculations is proposed