Technical Article · Editorial Line 01 · Credit & Structured Products
When face value is worth half.
The fair value of credit rights in FIDCs under Level 3.
Technical Article · Editorial Line 01 · Credit & Structured Products
The fair value of credit rights in FIDCs under Level 3.
A portfolio of credit rights with a face value of seventy-eight million reais is measured at fair value at around fifteen million. The layperson reads this as loss, deterioration of patrimony. The technician reads it as measurement. The difference between the two numbers does not translate pessimism or fraud; it translates, in market money on today's date, a future payment promise whose probability of realization and whose timing of realization must be estimated with method. Understanding why face value can be worth half, or much less, is to understand the core of what makes private credit a market as promising as it is dangerous when measurement discipline does not keep up with growth.
And growth has been vertiginous. The credit-rights fund industry has expanded and become popular, and regulation has followed: CVM Resolution 175, of December 2022, reorganized the entire fund framework in Brazil, and its Normative Annex II now specifically governs FIDCs, replacing the old rule that had been in force since 2001. More recently, the regulator's technical area returned to the subject with clarifications on the application of these provisions. The regulatory framework has matured. The question is whether fair value measurement practice has matured at the same pace, because it is on this that the number appearing in the investor's quota depends.
When a credit right has an active market and an observable price, measuring it is trivial. The problem is that the overwhelming majority of credit rights in FIDC portfolios has none of that. There is no quotation screen, no recent transaction of identical instrument, no liquidity. Under the fair value hierarchy of the accounting standard, this pushes measurement to Level 3, the level of non-observable inputs, where value depends on assumptions the valuer constructs rather than prices the valuer reads.
Level 3 is not a gray zone where anything goes. It is precisely the opposite: it is the level that requires the highest degree of substantiation, documentation, and auditability, precisely because the absence of an observable price removes the natural market contradiction. When the number depends on assumptions, the assumptions must be explicit, defensible, and testable. This is where several recent episodes of abrupt repricing of credit fund quotas originated: not in fraud necessarily, but in optimistic assumptions that did not survive the first reality test.
Measuring the fair value of a credit right is bringing to present value the flow expected to actually be received, discounted at a rate that remunerates the risk of not receiving it. This simple sentence hides three technical decisions that determine everything.
The first is the estimate of expected flow, not contracted flow. Face value is what the debtor promised to pay. Expected flow is what is projected to be received after incorporating the probability of default and the loss incurred when default occurs. The technical language that organizes this comes from the credit risk framework: probability of default, loss given default, and exposure at default. Expected loss has little of a gloomy worldview: it is the arithmetic of a portfolio that, by statistical construction, will not receive everything owed to it.
The second decision is the discount rate. It must remunerate the time value of money and the risk that remains after having already adjusted the flow for expected loss. There is more than one legitimate technical path, and coherence between the treatment of risk in the numerator and the denominator is what avoids both double-counting and omission of risk.
The third decision is timing. Defaulted or recovering credit rights do not pay on the contractual date; they pay, when they pay, after a collection or enforcement process that has its own and uncertain duration. Extending the expected receipt period reduces present value as much as raising the probability of loss, and estimating that period based on real historical recovery curves, rather than the manager's expectations, is an essential part of discipline.
A round-number example shows the combined force of the three decisions. A defaulted portfolio of R$ 100 million in face value, with an expectation of recovering 30%, or R$ 30 million, not on the contractual date but over three years of collection, and discounted at a rate that remunerates that risk, arrives at a present value well below the nominal R$ 30 million of recovery, and at a fraction of face. Extending recovery from three to five years, with everything else constant, brings present value down again. Timing and probability pull in the same direction, and it is the combination of the two, not either in isolation, that explains the distance between face and fair value.
It is useful to separate two accounting regimes that tend to be conflated. When the asset is measured at fair value through profit or loss, which is the framework for a Level 3 marked portfolio, the expected loss does not generate a separate provision: it is already incorporated in fair value itself, because the price a market participant would pay for the credit right discounts the probability of not receiving. The expected loss model of the financial instruments standard, which mandates anticipating loss based on current expectation rather than waiting for materialization, governs assets measured at amortized cost, typical of a bank portfolio, and there translates into provision.
The practical consequence unites both regimes at a single point of discipline. The loss that would feed a provision model and the loss embedded in fair value arise from the same reading of the portfolio, and an entity cannot sustain an optimistic fair value while its own default estimates point elsewhere. It is this internal coherence, not the existence of two separate accounts, that an attentive audit demands.
There is a line that the independent valuer cannot cross, and it defines the very reason for the existence of independent valuation in this market. The fund manager has a model, has assumptions, and has a number. The role of the independent valuer is not to endorse that number nor to produce, by principle, a different one, but to submit the assumptions to technical contradiction, verify that non-observable inputs rest on real recovery evidence, test internal model coherence, and document a trail a third party can walk and reconstruct. Independence here is measured by the integrity of the process, not by the obligation to diverge from the manager. When the process is sound, converging with the manager's number is good news, not failure.
A FIDC rarely has a single class of quotas, and this structure is essential to what is being valued. The portfolio of credit rights is one, but the right over the cash flows it generates is divided into layers: senior quotas receive first, with a defined target return and protection; mezzanine quotas come next; and subordinated quotas absorb the first losses, functioning as the cushion protecting the upper classes. This subordination, plus overcollateralization where the value of credit rights exceeds that of senior and mezzanine quotas, is the principal mechanism of credit enhancement in the structure.
The consequence for valuation is direct and echoes what is seen in venture capital capital structures: the value of each class of quota depends on that class's position in the waterfall, not on a proportional slice of portfolio value. The same deterioration in quality of credit rights affects classes unequally. A loss that barely scratches the senior quota, protected by a thick subordinated layer, may consume the subordinated quota entirely. Valuing the subordinated quota without modeling the waterfall, as if it tracked portfolio value linearly, ignores precisely the risk it exists to absorb. Pricing each class requires projecting portfolio cash flow, applying priority order, and checking what is left, or not, for each layer under different loss scenarios.
Estimating how much and when one receives cannot rest on the manager's expectation or on a single comfortable assumption. Discipline requires looking at actual historical behavior of the portfolio and of similar portfolios, and the tool for this is vintage analysis, which tracks groups of credit rights originated in the same period over their lives, measuring how default and recovery evolve. From these curves the empirical basis is extracted for probability of default, for actual loss incurred, and above all for recovery period, which tends to be the most underestimated and most sensitive parameter in present value.
The quality of this estimate depends on a governance chain that regulation has reinforced. CVM Resolution 175 brought the requirement of registering credit rights composing the portfolio, attributing responsibilities to administrator and manager and placing the underlying assets under scrutiny, and the regulator's technical area returned to the topic to clarify the duty of verifying the underlying according to the nature of the credit. Performed credit rights, arising from goods already delivered or services already rendered, carry different risk from to-be-performed rights, which depend on a future delivery, and valuation must distinguish the two. A fair value model that does not see this difference treats as equivalent a promise already fulfilled by the originator and a promise still conditioned on a future event, and the private credit market has shown, more than once, the cost of treating these as equal.
"A FIDC quota is worth what its portfolio is worth, and the portfolio is worth what can be defended with method when the market stops believing in comfortable assumptions."
The Brazilian private credit market is asking for, more than return, measurement governance. Recognizing early that face can be worth half is the work done well itself. The real risk lies in sustaining that face is worth face until the quota discovers otherwise at the worst possible moment.
References