Many fields of finance involve complex optimization problems under strict time constraints — problems where even marginal improvements could generate immense value for firms. Could quantum computing bring such improvements? Two recent IBM studies with major financial industry participants have explored potential use cases for quantum computing in finance.
Bond Pricing
For several reasons, bonds primarily trade over-the-counter (“OTC”). The way that this is practically accomplished is that a potential buyer or seller submits a request for quote (“RFQ”) to various liquidity providers (banks or dealers) for a given quantity of a particular bond. Liquidity providers compete among themselves in a blind auction structure – they cannot see or otherwise directly respond to competitor quotes in response to an RFQ. Therefore, dealers must balance competing goals when quoting a bond: if a dealer offers a quote which is too favorable to the client, they are likely to win the trade but reap a lesser reward for doing so. If the quote is too conservative, the client will likely go elsewhere.
Because of the volume and competitive structure of RFQs, small improvements in pricing accuracy and response time can be worth huge amounts for liquidity providers. However, the “true price” of a bond at any time is determined by an immense number of variables, many of which may be structurally interrelated. Moreover, optimizing pricing additionally depends on inventory levels and associated carrying costs. These attributes, when combined with the potential gains from even marginal improvements, make it an appealing candidate for quantum computing.
Accordingly, an HSBC and IBM study recently investigated whether a mixed quantum-classical system utilizing IBM’s Heron quantum processor could yield improvements in bond pricing versus existing classical-only systems. The study is inconclusive and is mainly focused on additional areas of exploration rather than immediately-applicable performance gains. For example, the study indicates that certain performance improvements may have been attributable to quantum noise. “Quantum noise” describes interference with the optimal operation of a quantum computer, whether caused by environmental effects or the inherent uncertainty of quantum mechanics. Quantum noise is generally viewed as a source of errors rather than encoding information. While noise can reduce overfitting issues with back-tested models (as were used here), the same effect should be obtained with conventional noise (and in this study a noiseless simulation performed worse). So, the published results are not a “Sputnik moment” for quantum computing in finance, but instead represent a study that can be built upon.
Portfolio Optimization
A separate problem which lends itself to quantum computing is that of portfolio optimization. In a sense, portfolio optimization is the basic question of asset management — given a set of constraints, what is the optimal composition of assets in a portfolio? As one might expect, determining optimal allocations from the full menu of financial instruments on a continuous basis is an extraordinarily difficult task, and one whose difficulty increases exponentially as the number of variables increases. Accordingly, working approaches to portfolio optimization depend on simplifying assumptions of varying suitability, such as normal distributions on returns and static correlations between assets. The traditional model is easily solvable by classical computing, but adding additional constraints such as lot sizes or maximum portfolio sizes (e.g., cardinality constraints – pick 50 from 500) can easily tip the optimization problem from being trivial to being extremely difficult. Because a quantum computer can encode exponentially many states, it can (theoretically) explore the solution space in a more efficient manner. Quantum computing may also be able to avoid “local minima” — portfolio combinations which cannot be improved with small changes but which are not optimal across the full solution space. As with bond pricing, small improvements in accuracy or efficiency could translate to significant gains for market participants.
A recent study by Vanguard and IBM, again using IBM’s Heron quantum processor, aimed to benchmark a mixed classical-quantum system against current classical portfolio optimization techniques. The study used a relatively small pool of assets (109 bonds, within the 133-qubit limit of the Heron r1 chip) and benchmarked for time and accuracy against IBM’s conventional CPLEX solver. As the team acknowledged, this is a classically easy problem at this scale and could be solved within a few seconds by CPLEX or similar. However, the team found that a mixed classical-quantum approach, run both on a simulated basis and on local hardware, could perform within an acceptable level of accuracy to the classical baseline. Intriguingly, the results also indicated that harder-to-simulate ansätze (essentially quantum circuit architecture) may perform better, further supporting potential quantum advantage at higher levels of complexity. As the team acknowledged, true quantum advantage will only be possible at a level of complexity where classical solvers fail, which again will require more sophisticated quantum computers than are currently available.
Conclusion
The HSBC and Vanguard collaborations with IBM are early steps in the exploration of how quantum processors might be used to address practical challenges in finance. Although they do not represent validation of the hypothesis, their results can be used for developing other studies for assessing the feasibility of quantum approaches to complex financial problems. The results also illustrate that (i) quantum computing will have the most impact in addressing extremely complex and difficult computational challenges, so the optimal use cases for quantum computing are more likely targeted rather than universal and (ii) research results are not always positive or conclusive, but those results are still worth sharing and can be the basis for further studies.
Covington is monitoring developments globally in this fast-growing area.
Visit Covington’s Quantum Computing web page for additional updates. Please reach out to a member of the team with any inquiries.