Navigating the complexities of derivative valuation

Essential market information for valuations isn’t always readily available without high subscription costs from market data vendors. Proprietary economic data such as inflation rates and lending rates, or private market indices are linked to interest rate derivatives or total return derivatives and valuation of such instruments rely heavily on up-to-date information readily available.

Limitations in or lack of access restricts an entity’s capacity to value derivative positions on their own reliance and would rely on counterparty valuations.

Often, there is a vast array of information and data points out there that it becomes a challenge to able to identify the data that is most relevant for the type of derivative valued. There could be multiple interest rates published by different entities in the same market, for example Secured Overnight Funding Rate and the Federal Funds Rates, which could become an obstacle in deciding what is most appropriate.

Furthermore, the question of whether the input data in the valuation is reliable also arises. Therefore, it is imperative that the valuation input used such as FX or Interest rates are benchmarked in the overall market environment. Another issue noticed is the volatility of inputs which usually cluster around crucial market events as evidenced by the charts below. 

Derivatives are inherently complex instruments to value because their value is derived from an underlying asset. Certain derivative structures introduce contingent payoffs, multiple asset classes, or credit support annexes, the valuation of which requires in-depth understanding of path dependencies, correlation between asset classes, payoff structure, and complex pricing models.

Derivative valuation models such as Black-Scholes model (and their adaptations such as Garman-Kohlhagen), stochastic volatility models, ISDA Standard Models, and Discounted Cashflow are commonly used by services such as Bloomberg and ICE Data Derivatives. The ability to use such models effectively requires specialised knowledge of derivative pricing models which is at times a challenge.

In addition to data accessibility constraints, we observe data limitations where specific data points are not easily available and require bridging techniques such as bootstrapping, extrapolation, and parametric modelling. Such techniques rely on existing market information to derive the appropriate data required for valuation.

Bespoke or custom-built valuation models introduce model risk. Model risks tend to make the valuation unreliable if not addressed appropriately. Such risk comes in the form of inappropriate methodology being used, methodology being applied incorrectly, operational oversight, misunderstanding of input data points etc. 

Valuation data inputs can be classified as observable and unobservable inputs. Observable inputs are straight forward and readily available, use of which leads to less complications. Unobservable inputs on the other hand bring subjectivity and estimation uncertainty. Unobservable inputs are often constructed based on observable data.

However, it is important to ensure that accurate and appropriate observable data are used otherwise, valuation results become unreliable.