Skip to content

New to Omni?·Get 20% off your first purchase with code WELCOME20 at checkout.See plans

CAIA® Level II: when do smooth returns hide portfolio risk?

Smooth illiquid returns may hide delayed valuation. Learn to test the risk comparison with unsmoothing, interpret its assumptions, and avoid false certainty.

ConceptOmni Curriculum Team6 min read

A neat stack of blank white pages rests over crumpled blank sheets beside a pencil and ruler on a navy desk.
On this page

A smooth reported return series can make an illiquid strategy look less risky and more diversifying than a liquid strategy, even when the underlying economics are no calmer. For a CAIA® Level II candidate, the useful question is whether the apparent stability comes from economic exposure or from valuations that incorporate information slowly. An unsmoothing calculation is a sensitivity check on reported risk, not a way to discover an exact tradable price.

Why a calm return series can mislead

A quoted security can be repriced whenever buyers and sellers trade. An illiquid holding may be marked from an appraisal, a valuation model or an old dealer indication. Its reported value can therefore move in stages as evidence arrives, although the economic news arrived earlier. CAIA Association explains that appraisal anchoring can create serial dependence in returns and suppress conventional risk measures. The problem concerns the measurement process: a calm series does not itself establish that cash flows, exit values or the investor’s exposure are stable. This distinction matters whenever a candidate compares a private asset with a frequently traded index.

Look carefully at the unit of comparison. A fund index can average across managers, so its stability may partly reflect diversification across constituents as well as valuation lag. A single holding has another risk profile. CAIA Association also notes that many alternative investment benchmarks are not directly investable. A candidate should identify whether the question describes a fund, a manager, an index or a property before interpreting the reported standard deviation. The same apparent return pattern can arise from different mechanisms; the correct answer should name the mechanism the facts support.

A practical diagnostic is positive correlation between a reported return and the previous period’s reported return. That pattern is consistent with information being incorporated over time. It is not proof: genuine economic momentum, contractual cash flows, manager strategy and changes in market conditions can also produce serial dependence. Thin trading deserves investigation, not an automatic accusation of mispricing. CAIA’s Level II curriculum companion specifically places first-order autocorrelation and unsmoothing among the learning objectives. The exam therefore asks candidates to move from a symptom to a conditional adjustment and then to an interpretation.

What the adjustment assumes

In a simple model, the observed return combines some of the current underlying return with a carry-over from the previous observed return. Call the carry-over weight phi. When phi is positive and below one, a new shock is only partly recognised in the current mark. Solving that model for the current underlying return removes the inherited component and scales up the remainder. The previous value in this expression is the previous reported return. Substituting a previously reconstructed return changes the assumed model and can produce a different sequence.

rtobs=(1−ϕ)rtunderlying+ϕrt−1obsr_t^{obs}=(1-\phi)r_t^{underlying}+\phi r_{t-1}^{obs}
A simple model of current reported return as current underlying return plus a lagged reported component.
r^tunderlying=rtobs−ϕrt−1obs1−ϕ\widehat r_t^{underlying}=\frac{r_t^{obs}-\phi r_{t-1}^{obs}}{1-\phi}
Estimated underlying return from the current and previous reported returns and the assumed smoothing weight.

Estimating phi from a short history is uncertain. The adjustment becomes especially sensitive when phi is close to one because the denominator becomes small. A model that assumes one lag may also miss valuation effects spread across several periods. Moreover, an adjusted series is still an estimate based on marks, not a record of prices at which an investor could have sold. These are reasons to treat an adjusted volatility or beta as a model output with assumptions attached. They do not make the reported series a better measure by default.

An illustrative comparison that changes the answer

Consider a wholly invented sequence of quarterly reported returns for a private credit portfolio: a starting return of zero, followed by one percent, two percent, one percent and zero. Assume, solely for illustration, that half of each reported return reflects the previous reported return. The current reported return of two percent is partly an echo of the prior one percent. Removing that echo leaves one and a half percentage points, and dividing by the remaining half gives an estimated underlying return of three percent for that quarter.

Apply the same calculation consistently to the other quarters. The reconstructed returns are two percent, three percent, zero and minus one percent. Their average equals the reported series’ average over those four quarters, but their path is less placid: the reported sequence of one, two, one and zero percent ranges from zero to two percent, whereas the reconstructed sequence spans minus one to three percent. These figures are invented arithmetic, not observed performance. The lesson is that a similar average can coexist with a materially different pattern of risk.

Suppose a liquid benchmark also delivered an average near one percent per quarter but showed abrupt losses in the same environment. Comparing only reported volatility could favour the private series. The adjustment asks whether that advantage survives a plausible lag assumption. It can also change estimated correlation with a benchmark, because two assets may react to the same news in different reporting quarters. A naive same-quarter correlation may miss the relationship. The direction and size of a correlation change depend on the actual paired data and model; they cannot be asserted from the private series alone.

The decision also depends on when the valuation catches up. If a private mark recognises a market shock in a later quarter, an investor comparing contemporaneous observations might see an apparently weak link between the private portfolio and listed assets. Shifting or modelling that delayed recognition can reveal a different dependence. Yet a lag adjustment cannot tell an investor how much cash could be raised immediately, what a transaction discount would be, or whether the fund’s redemption terms permit an exit. CAIA’s treatment of illiquidity makes those separate questions important. Keep valuation smoothing, market liquidity and contractual liquidity as distinct concepts when explaining what the reconstructed series does and does not solve.

A robust comparison should also use like-for-like frequency and the same observation window. Quarterly private marks should not be compared mechanically with a daily-traded benchmark’s raw daily volatility. Aggregating the liquid series to quarters makes the periods comparable, while a consistent time window avoids attributing a change in market regime to an asset class. Then test more than one plausible smoothing coefficient if the estimate is uncertain. If the conclusion changes entirely across reasonable assumptions, report that sensitivity rather than a single polished risk number. This is useful decision support because it shows which part of the apparent diversification depends on the marking model.

How to answer the exam question

First, state what is observed: a low-volatility series, positive lagged dependence, or a delayed response to market news. Second, state what is inferred: valuation lag may be smoothing the reported returns. Third, identify the model and compute using the prior observed value if the coefficient is supplied. Fourth, interpret the adjusted result as an estimate of economic risk, with its assumptions. This sequence is stronger than merely writing that unsmoothing raises volatility. The numerical outcome only answers a question after the candidate has identified which returns are observed and what the coefficient represents.

The common errors are consequential. Dividing by phi rather than one minus phi misstates the shock. Subtracting the previous underlying estimate uses the wrong lag. Treating positive autocorrelation as proof of superior skill mistakes a reporting pattern for performance. Calling an adjusted series the actual transaction record overstates what the method can deliver. Assuming correlation must rise in every sample similarly outruns the evidence. CAIA Association’s discussion of alternative manager indices shows why the interpretation of smoothing should sit alongside questions about manager dispersion, index construction and whether a benchmark can be invested in.

For a candidate deciding what to practise, use the CAIA program overview to place the topic within alternative-investment risk, then compare how the Omni learning method moves from a concept to a decision. A concise self-check is to take a reported sequence, label each observation by period, choose a stated phi, reconstruct the series, and explain in words why the risk comparison changed. If the question supplies no coefficient or reliable history, give a qualitative warning rather than fabricating a precise adjusted statistic.

Sources

  1. Level II Curriculum Companion 2025 | CAIA Association · caia.org · Retrieved
  2. An Alternative View of Manager Selection Risk | CAIA Association · caia.org · Retrieved
  3. Introduction to Alternative Investments | CAIA Association · caia.org · Retrieved

Explore CAIA study resources

Educational content, not investment advice. This article is written to help candidates prepare for professional exams. It is not a recommendation to buy, sell or hold any security, and it does not take your personal circumstances into account.

Omni Finance Academy is an independent exam-preparation provider and is not affiliated with or endorsed by CFA Institute, GARP, CAIA Association or CFP Board. CFA®, FRM®, CAIA® and CFP® are trademarks of their respective owners. Full disclaimer.