Optimisation is a term that is used fairly loosely for a broad range of activities, particularly these days when the topic of collateral management is high on the agenda. There are many different aspects to a collateral optimisation programme, but taken as a holistic concept we can consider it to be a process whereby an institution can attempt to minimise the cost of collateral that its business activity incurs and to maximise the return on its assets.
This has become an absolute necessity for many institutions, both on the buy side and the sell side in the new regulatory environment. Required collateral volumes are increasing hugely in proportion to the size of the outstanding positions. The impact of collateral terms for any given trade is a key determinant of how profitable that trade will be and so there is a strong incentive for institutions to operate their collateral programmes as optimally as possible. This changing environment has brought the distinct but related disciplines of margin optimisation and collateral optimisation to the forefront.
There are a number of different dimensions to the collateral optimisation problem. Any new trade will affect overall collateral requirements. If a trade is centrally cleared, as the majority of OTC derivatives soon will be, it will attract initial margin and variation margin requirements according to the rules and models of the central counterparty (CCP) concerned. The initial margin component will be calculated according to the exchange’s methodology as approved by its regulator and will likely be based on a VaR calculation. If a trade is bilateral, the terms of the collateral agreement in place with that counterparty and the existing portfolio will determine the cost. Even if the trade is bilateral and there is no collateral agreement in place, it will be hedged and that hedge will attract collateral requirements.
That the transformation of the OTC market into a centrally cleared model is a game-changing event is well documented. There are various estimates of the amount of additional collateral that will be required overall, which run into the trillions of dollars. We have also seen recently the BIS / IOSCO consultative document proposing broadly similar provisions for initial margin on bilateral trades. The impact of this, if it were to become a reality, will also be huge and will pose new and interesting problems to the market.
In this environment, it is imperative to minimise the impact of this burden of extra collateral in terms of the amount of collateral that is required and the cost of funding that collateral. To state the optimisation problem succinctly, we should look at two key questions:
- How can I minimise the overall cost of collateral that I put up?
- How can I make the best use of my pool of available assets?
We can then summarise the answers to these questions in two rather simple statements:
- I need to optimise the settlement location of new trades. We might call this margin optimisation
- I need to optimise how I allocate my assets to my collateral requirements. We might call this collateral optimisation
Let us consider the first statement regarding margin optimisation. When I decide to put on a new trade, there are decisions to me made about where and how to trade. There are CCPs operating in various locations and more entering into the market. Assuming that I am a member or have access to more than one CCP through clearing brokers, I need to understand the cost of transacting at each one. Each CCP has its own margin terms and the amount of margin that is required will depend on its model and the trades that I already have there. In many instances, I may also have the choice of doing the trade with a bilateral counterparty. I will have negotiated bilateral collateral agreements with my direct counterparties and each of those will have specific rules governing eligible collateral and haircuts, and the collateral requirement calculation terms (thresholds, minimum transfer amounts, and so on). In addition, there may well also be the newly proposed initial margin requirements as proposed in the BIS / IOSCO paper. These agreements may have been negotiated some years previously and there may be some scope for renegotiation of these terms (which in itself can be regarded as a form of optimisation), but in the short term they can be regarded as fixed.
The key to solving the margin optimisation problem lies in working out how best to balance the portfolio of trades across the various counterparties and possible settlement locations. The goal is to minimise the amount of collateral that is required across all of the obligations, but also to take advantage of preferential eligibility and haircut rules for various collateral types that will match my portfolio of available assets. These potential offsets also need to be taken into account in the decision making process around deal allocation. We know from optimisation theory that there must be a single objective function to any optimisation, and so these rules
and constraints must be presented to a margin optimisation engine as a cost parameter to determine how to maximise revenue from the asset portfolio. To achieve margin optimisation, we need to calculate the funding costs of collateral for a new trade with any of the possible counterparties or settlement locations and choose the best one.
For each counterparty or CCP, we can simulate the impact of the new deal by adding it to the existing portfolio and running through an approximation of the relevant initial margin calculation for each potential counterparty or CCP. This way, I can estimate the amount of extra collateral that will be required and the funding cost of that collateral. This calculation is often referred to as the Funding Value Adjustment (FVA). One methodology that is used for calculating FVA is to perform a Monte Carlo simulation on the underlying portfolio and also all the collateral assets using the funding or collateral yield curves for each currency or asset. Along each scenario, the collateral balance across each time-step is integrated with respect to the spread between the funding and the collateral rates using the simulated yield curves, and this can be averaged across all scenarios to recover the total cost of collateral. The inputs to each calculation are therefore the underlying portfolio and market data, the current collateral balance and the assets that it comprises, a funding rate per currency or per asset where non-cash collateral is used, and a collateral rate per asset representing the contractual rate that is paid when the asset is held as collateral (for example, the interest on cash). The calculation also needs to include the collateral requirement terms, the applicable posted and received haircuts per asset and so forth. For the calculation to be meaningful, we may also want to assume some time band for the funding requirement to be considered rather than the entire length of the deal.
We will then want to identify which is the most advantageous settlement location. The result could vary widely given the portfolio effects of the existing population of deals at the various locations. So a deal placed at one CCP may add significantly to the initial margin requirements there, whereas the same deal may have an offsetting effect at another CCP and would actually decrease the initial margin requirements. These calculations look difficult and operationally intensive at first, but in fact if you are already performing CVA calculations, there should not be much incremental effort. However, you do need to perform the calculation for each potential settlement location and identify the optimal counterparty or CCP fast enough for the result to be useful pre-deal.
It is important to note that the optimal location will not always be the one that gives the best absolute result in terms of the value of initial margin that is required; you will also need to take into account the profile of eligible collateral and how that matches your funding profile in the various asset classes. For centrally cleared transactions, some CCPs are now offering the capability for members to define the set of futures that they may wish to offset their swaps in the VaR margined portfolio and those that they wish to keep in the SPAN margined portfolio. I have left out the cost of capital to support the trade, which will differ potentially greatly depending on whether the trade is bilaterally or centrally cleared, but that is also clearly an important component of the final result.
Ultimately, the profitability of a deal must take into account the cost of the collateral that is required to support it. If this can be measured and estimated pre-deal, then it must factor in to the deal pricing and whether or not to enter into the deal in in the first place. This is a part of the process for the pre-deal decision support and a more incentivising tool to the front office than charging back actual costs of collateral to the desk on a historic basis, which is perhaps a more traditional method.
It is immediately apparent when we look at the challenges of collateral optimisation that we will get better results if we cast a wider net. We need to run our optimisation algorithms across the broadest set of requirements possible and with a single consolidated view of the available inventory. When we consider the new collateral landscape, it is clear that the old model of business-level silos does not cut it any more. Many institutions have adapted and have brought those silos together into a single enterprise collateral management ecosystem. Getting different business lines to buy into a single collateral organisation and centralised decision-making process for collateral allocation can be difficult in some institutions. Of course, there is still scope for optimising within product silos, optimisation tools and costs can be shared, and this is better than no optimisation at all. Nevertheless, the benefits of a centralised inventory and of centralised allocation decisions are potentially significant and certainly measurable.
Once we have the centralised global inventory of assets and the associated eligibility and haircut rules, we then need to determine the optimal way to allocate these assets to the collateral requirements resulting from the margin optimisation exercise above. There is the temptation to take a ‘fire and forget’ approach to collateral allocation. That is the way that it was traditionally performed–—make a decision on the cheapest-to-deliver collateral at the time that a call is received, post out those assets, and forget about them until the exposure drops and they can be recalled. Even refinements of this approach, whereby you have some kind of ranking of agreements and assets and you allocate them sequentially, is demonstrably not the way to solve the collateral optimisation problem.
Optimisation algorithms work differently and much more effectively. They have a single objective function, which is to minimise the overall opportunity cost of the pledged collateral assets, or in other words, maximise the revenue from the overall collateral asset pool. This is an important distinction. A crucial aspect in the context of collateral optimisation is the distinction between single requirement-based optimisation and overall optimisation. The first type is to optimise the allocation of collateral assets for a single requirement in isolation, ie, find the lowest quality of accepted collateral for a single margin call and do this sequentially or by a ranking. The latter is working across the global set of requirements to find the cheapest overall combination of assets that are allocated to the various collateral requirements. This is how a true optimisation algorithm will work and it will yield significantly better results.
The algorithm must also consider not only new pledges of collateral in performing the allocations; it must consider that previously posted collateral may be substituted and redeployed elsewhere. There is something of an art to the calibration of these algorithms. The costs of use of different assets must be determined, including movement costs, and they must be tailored to understand the constraints of a feasible solution (eligibility rules, haircuts, concentration limits, and so on), and they must take into account operational constraints such as the number of substitutions that you can physically perform in any one optimisation run. The next part of the optimisation process is to automate the collateral trade generation to cope with the increased number of movements that will occur once optimising the allocation is started. Such a collateral optimisation solution, if correctly deployed, represents a significant competitive advantage and constraint on the costs of doing business.
Margin and collateral optimisation are relatively new disciplines that are gaining traction as institutions formulate their responses to the new regulations. When central clearing kicks in, these activities will no longer be a luxury; they will be a necessary tool for institutions to deploy their capital most efficiently and to retain their competitive edge.