Level 3 valuations: test the inputs behind unrealized gains
An unrealized gain on an illiquid investment is only as interpretable as its valuation method and significant inputs. Reconcile the Level 3 balance, isolate gains from new investment, and test the assumptions that matter. The output should explain the gain’s mechanism without pretending an analyst can recreate a private transaction price from a short note.
Define what Level 3 means
IFRS 13 provides the IFRS fair-value measurement framework. The hierarchy concerns the inputs to valuation; Level 3 involves significant unobservable inputs. It is not a quality score or proof of manipulation. US GAAP ASC 820 has related requirements, but scope and detailed presentation still need framework-specific review.
Identify the unit of account, valuation technique, measurement date and whether a recent transaction was orderly and relevant. An issuer’s plan to hold an asset indefinitely does not itself determine fair value. A funding round price can require adjustments for different rights, seniority or transaction circumstances.
Hypothetical private investment gain
Assume opening Level 3 assets are 100 million. Purchases add 20, sales remove carrying value of 10, unrealized gains add 18 and currency adds 2. Closing assets are 130. The 30 balance increase contains 20 of fresh capital; it is not a 30 investment return.
Suppose 12 of the gain relates to a revenue-multiple model. Forecast revenue increases from 40 to 45 while the multiple rises from 4 to 4.4. Simplified enterprise value moves from 160 to 198 before debt, ownership rights and other adjustments. This illustrates two distinct drivers, not a complete valuation of the investor’s stake.
Holding the multiple at 4 produces 180, while holding revenue at 40 and using 4.4 produces 176. The separate changes should not be added without accounting for the interaction. All amounts are hypothetical, and the exercise is a sensitivity map rather than an independent appraisal.
Copy the input worksheet
Record asset, method, significant input, prior input, current input, disclosed range, sensitivity and evidence date. Add purchases, disposals, transfers between hierarchy levels and gains recognized in profit or OCI. Reconcile the opening and closing population before comparing input averages.
Where an input is a weighted average, identify whether the weights changed because assets were bought or sold. A higher average multiple may reflect portfolio composition rather than revaluation of unchanged investments. Do not attribute a group-level input change to a named holding unless the disclosure permits it.
Challenge calibration and correlation
Look for subsequent financing rounds, secondary transactions, operating results and changes in preferred rights. A later transaction can inform review without proving what fair value should have been at an earlier measurement date. Preserve the information available at each date.
Read sensitivities as conditional calculations. Lower revenue and a lower multiple may be economically correlated; testing them separately understates a combined adverse scenario. Conversely, combining the largest adverse input from unrelated assets can create an incoherent portfolio stress. State which relationships are assumed and which cannot be observed publicly.
The final note should identify the portion of reported profit linked to these valuations, the significant inputs and the evidence needed to reassess them. It should not turn an unobservable input into a fabricated precise market quote.
FAQ
Does a transfer into Level 3 mean the asset lost value?
Not necessarily. It means the relevant input observability changed under the applicable hierarchy. Examine the transfer reason and valuation movement separately. Reduced market evidence can matter even if the reported value is unchanged.
Can unrealized gains be ignored completely?
They can materially affect reported earnings and capital. Show a clearly labeled earnings sensitivity if useful, while retaining the reported result and examining the underlying assumptions rather than assuming every unrealized gain is unreal.