9.3 Model Calibration

ASC 820-10

Valuation Techniques

35-24C If the transaction price is fair value at initial recognition and a valuation technique that uses unobservable inputs will be used to measure fair value in subsequent periods, the valuation technique shall be calibrated so that at initial recognition the result of the valuation technique equals the transaction price. Calibration ensures that the valuation technique reflects current market conditions, and it helps a reporting entity to determine whether an adjustment to the valuation technique is necessary (for example, there might be a characteristic of the asset or liability that is not captured by the valuation technique). After initial recognition, when measuring fair value using a valuation technique or techniques that use unobservable inputs, a reporting entity shall ensure that those valuation techniques reflect observable market data (for example, the price for a similar asset or liability) at the measurement date.

When an entity determines that a transaction price represents fair value on initial recognition, but the valuation model that the entity expects to use for subsequent measurement yields an initial estimate of fair value that differs from the transaction price, the valuation model should be calibrated in such a way that it yields the same estimate of fair value as the transaction price. An entity should not recognize a gain or loss at inception because of a difference between the valuation model and the transaction price. An inception gain or loss would only be recognized as a result of subsequent changes in circumstances.

ASC 820-10-35-24C does not specify how to calibrate a model or model inputs. Accordingly, various calibration methods may be acceptable under ASC 820-10-35-24C depending on the valuation technique used, the availability of information about market-participant assumptions (e.g., relevant observable inputs), the terms of the instrument, and the nature of the entity’s portfolio. However, regardless of the method an entity uses, if the transaction price represents the fair value of a contract at inception and the initial fair value differs from the entity’s inception-date model value, the inception difference should not be recognized (1) upon the initial recognition of the transaction or (2) after initial recognition unless there has been a subsequent change in circumstances. Instead, the inception difference should be recognized as the uncertainty in the inputs, the uncertainty in the model, or both are eliminated. The calibration method chosen should be applied systematically, rationally, and consistently.

As noted above, the appropriate method for calibrating a model depends on an entity’s circumstances. Entities often employ valuation adjustments to reflect the factors that market participants would consider in setting a price when those factors are not otherwise captured in the model (e.g., adjustments for uncertainty in model inputs or model complexity). Appropriate use of valuation adjustments should result in an estimate of fair value that reflects the price at which market participants would transact as of the reporting date. Similarly, to comply, at initial recognition, with the fair value measurement objective prescribed by ASC 820-10-30-3A and ASC 820-10-35-24C, an entity may determine that it needs to make a separate valuation adjustment to calibrate its model (and to ensure that the inception difference is not recognized). In these situations, the entity should review the valuation adjustment periodically to ensure that it reflects any new information and is consistent with ASC 820’s exit price notion. ASC 820-10-35-24C indicates that “[a]fter initial recognition, when measuring fair value using a valuation technique or techniques that use unobservable inputs, a reporting entity shall ensure that those valuation techniques reflect observable market data (for example, the price for a similar asset or liability) at the measurement date.”

The example below illustrates a model calibration approach that may be acceptable when the inception difference can be isolated to a particular unobservable input. Entities should follow the steps in the example below regardless of whether the calibration approach results in direct adjustments to model inputs or valuation adjustments to the modeled estimate of fair value.

Example 9-3

Model Calibration for Electricity Contract

Entity A enters into an electricity forward contract to purchase 100 MW of electricity (daily) at Location X for a 10-year term. The forward contract is accounted for at fair value because it meets the definition of a derivative and does not qualify for (or the entity did not elect) the scope exception for normal purchases and normal sales under ASC 815-10-15-22 through 15-26. There are three years of observable forward prices at Location X. Entity A’s model value at inception is a gain of $1 million; however, no cash was exchanged between the parties. On the basis of an analysis under ASC 820-10-30-3A, A determines that the transaction price equals the exit price (i.e., fair value equals zero at inception) and that it must apply ASC 820-10-35-24C. The only unobservable input with a significant effect on the model value at inception is the forward electricity price for Location X.

In this scenario, A might apply ASC 820’s calibration guidance as follows:

Step 1: Identify the source of the inception difference — Entity A has established that the only unobservable input that significantly affects the model value is forward electricity prices.
Step 2: Adjust the unobservable inputs or establish valuation adjustments in such a way that the adjusted model value equals the transaction price at inception — In a systematic, rational, and consistently applied manner, adjust the unobservable forward price points (i.e., years 4–10) on the forward price curve in such a way that the model produces a fair value equal to zero (i.e., the exit price equals the transaction price).

When calibration results in direct adjustment to model inputs, entities should consider the impact of such adjustments on the valuation of other instruments in their portfolio (e.g., instruments valued under different models but by using the same or similar pricing inputs). The calibration may affect the valuation of other instruments because the calibrated inputs may replace or supersede the assumptions previously used for unobservable inputs. Recall that in step 1, A determined that the inception difference was driven solely by the unobservable Location X electricity prices; therefore, any resulting “calibration adjustment” represents a calibration of these unobservable electricity prices to the most recent available information (the forward contract’s transaction price). As a result, if A derives the fair value of a portfolio of contracts by using long-dated (in this example, beyond three years) Location X electricity prices, the calibration adjustment to the electricity forward prices would most likely affect the fair value of the other long-dated Location X contracts. The considerations noted above for calibration techniques that result in direct adjustment to model inputs may also apply when a calibration results in valuation adjustments to the modeled estimate of fair value. In other words, calibration adjustments result in updated information about assumptions market participants use in assessing unobservable inputs or valuation adjustments and may have relevance beyond the recently executed transaction.
Step 3: Recognize subsequent changes in model value — The inception difference will be recognized when (1) unobservable inputs become observable or (2) unobservable inputs or valuation adjustments are adjusted to reflect new information (e.g., through calibration of the model or model inputs to reflect new transaction data or through the passage of time). In this example, in the absence of further calibration or changes in input observability, one would expect the inception difference to be recognized over the first seven years of the contract (as years 4–10 become observable). Note that no inception difference should remain for a period in which settlement has occurred. In addition, no portion of the inception difference should remain once all the unobservable inputs become observable.

Confidential and Proprietary — for Use Solely by Authorized Personnel

This publication provides comprehensive guidance; however, it does not address all possible fact patterns, and the guidance is subject to change. Consult a Deloitte & Touche LLP professional regarding your specific issues and questions.