Avoiding circularity in valuation models

Your model is only as good as the data that feeds it

Models play an essential role in valuing instruments where observable market activity is limited. They help firms interpolate between known points, estimate missing values and produce consistent pricing across complex portfolios.

However, when market data is used to calibrate or validate a model, firms also need to understand how those inputs were created.

Some pricing sources may already incorporate aggregation, interpolation, modelling or evaluated inputs. These sources can provide valuable coverage and consistency, particularly in less liquid markets. But if their construction is not fully understood, a model may be calibrated or tested using inputs that contain assumptions similar to its own.

This can introduce circularity into the valuation process, weakening the independence of the evidence used to challenge the model.

How circularity can arise

A valuation model uses market inputs to estimate the value of an instrument. Depending on the product, these inputs might include curves, spreads, volatility surfaces or prices for comparable instruments.

In many OTC markets, transactional activity is concentrated in a relatively small number of liquid or semi-liquid instruments. The available orders and trades may therefore cover only a subset of the market points required for valuation and risk workflows.

Producing complete and consistent datasets often requires interpolation, modelling or other forms of transformation. This is not inherently problematic. Modelled and interpolated data can play an important role where direct market observations are sparse.

The risk arises when an input has been derived using assumptions similar to those embedded in the firm’s own model. In this situation, the resulting valuation may reflect related assumptions at more than one stage. If those inputs are then used to validate the output, the process may confirm the underlying methodology rather than provide a sufficiently independent challenge.

The challenge is understanding where transformation has taken place and how the resulting input relates to the model being calibrated or tested.

Without sufficient transparency, it may be difficult to determine:

• Which inputs are based on direct market observations
• Where aggregation, interpolation or modelling has been applied
• Whether related assumptions are being introduced more than once
• Whether calibration and validation inputs are sufficiently independent
• Whether the results reflect additional market evidence or reproduce an existing methodology

Understanding what your inputs represent

Different data sources provide different forms of market evidence and involve different degrees of transformation.

Trade data records completed transactions, while order data can provide evidence of firm market interest at a particular point in time. Indicative, composite and evaluated pricing can provide broader and more consistent coverage, but may incorporate aggregation, interpolation, modelling or a defined evaluation methodology.

For example, broker-sourced indicative pricing, like Parameta’s indicative pricing, can transform sparse market observations into consistent curves, surfaces or instrument-level datasets. This can provide the coverage required for valuation and risk workflows where transactional observations alone are insufficient.

These approaches can play an important role in supporting complex OTC workflows. However, firms still need to understand what an input represents, how it has been constructed and whether it is appropriate for its intended use.

The question is not which source is universally best. It is whether firms understand the assumptions and transformations within an input before using it for calibration, valuation or independent validation.

Preserving independence in model validation

Model validation should provide an independent assessment of whether a model’s outputs remain aligned with available market information.

If the data used to test a model contains assumptions or methodologies closely related to those embedded in the model, the validation process may confirm those assumptions rather than challenge them.

Additional forms of market evidence can provide distinct points of comparison. Trade and order data, for example, can help firms assess model outputs against completed activity or firm market interest where available.

Broader pricing sources can provide more consistent coverage across a curve or volatility surface. However, their methodology and relationship to the model being tested should be understood when assessing the independence of the validation process.

The aim is not to identify a single source of truth. It is to combine complementary sources while maintaining an appropriate degree of separation between:

• The evidence used to calibrate a model
• The assumptions embedded within the model
• The data used to validate its outputs

Why data lineage matters

Data lineage shows where an input originated and how it was transformed before reaching a model or valuation process.

Clear data lineage can help firms identify whether an input is based on a trade, an order, indicative pricing or a calculated value. It can also show where aggregation, interpolation, modelling or a defined methodology has contributed to the resulting price.

This visibility is particularly important when valuations are reviewed by independent control functions, model validation teams, auditors or regulators.

A clearer understanding of data provenance can help firms:

• Separate direct market observations from calculated or model-derived values
• Assess the independence of calibration and validation inputs
• Identify where assumptions, aggregation or interpolation have been introduced
• Compare model outputs with distinct forms of market evidence
• Strengthen model governance and valuation controls

The objective is not to eliminate modelling. It is to understand where modelling has been applied and ensure that the evidence used to challenge a model is sufficiently distinct from the assumptions being tested.

Building a stronger evidence base

Composite, evaluated, indicative and transactional data each provide a different view of the market. Their suitability depends on the instrument, available liquidity and intended workflow.

Where consistent market coverage is required, modelled and interpolated data can play an important role. Transactional observations alone may not provide the complete curves, surfaces or instrument coverage required for valuation and risk processes.

Where firms need to calibrate, challenge or independently validate a model, access to additional forms of market evidence can provide separate points of comparison and help reduce the risk of circularity.

Understanding how each input has been created allows firms to combine these sources more effectively, balancing consistent market coverage with transparency and independent challenge.

Bringing greater transparency to valuation workflows

Parameta Solutions provides indicative pricing, trade and order data across global OTC markets, giving firms access to different layers of broker-sourced market information.

Trade and order data can provide evidence of observable market activity where available. Parameta’s indicative pricing transforms available market information into broader, consistent datasets that can support valuation, model calibration and risk workflows across instruments for which transactional observations may be sparse.

Used alongside composite, evaluated and internally modelled pricing, these datasets can help firms examine the provenance and construction of their inputs, compare outputs against different forms of market evidence and bring greater transparency to valuation and validation workflows

Request a data sample tailored to the asset classes and instruments relevant to your modelling and valuation workflows.