How do survivorship bias and appraisal smoothing distort CME inputs?

Question

AcadiFi · Accepted Answer

These are two of the most impactful data biases for asset allocation, and they both make certain investments look *artificially* attractive. Let's break them down:

## Survivorship Bias

**What it is:** When a database only includes entities that currently exist (or existed through the end of the sample), the average performance is overstated because failures have been removed.

**Hedge Fund Example:**
A hedge fund database reports that the average hedge fund returned 8% over the last decade. But during that decade, 40% of the funds that existed at the start shut down — many because of poor performance. The 8% average only reflects the survivors. The true average, including the failed funds, might be 4%.

**Impact on CME:**
- Overstated expected returns for the asset class
- Understated risk (the worst outcomes are removed)
- Higher Sharpe ratios than are realistically achievable
- MVO will over-allocate to the asset class because it looks artificially attractive

**How to adjust:** Use databases that track dead/liquidated funds (e.g., Hedge Fund Research includes non-reporting funds). Apply a survivorship bias adjustment — typically 1–3% per year for hedge fund indices.

## Appraisal (Smoothed) Data

**What it is:** Real estate, private equity, and other illiquid assets are valued through periodic appraisals rather than market transactions. Appraisers tend to anchor to previous valuations and adjust gradually, creating artificially smooth return series.

**Real Estate Example:**
A commercial property is appraised at $10M in Q1, $10.2M in Q2, $10.1M in Q3, and $10.3M in Q4. These small, gradual changes suggest low volatility. But if the property were traded on an exchange, the quarterly prices might swing between $9M and $11M. The appraisal-based returns dramatically understate the true economic volatility.

**Impact on CME:**

```mermaid
flowchart LR
    A[Appraisal-Based Returns] --> B["Understated Volatility
(appears low-risk)"]
    A --> C["Understated Correlations
(appears diversifying)"]
    B --> D["MVO Over-Allocates
to Real Estate/PE"]
    C --> D
    D --> E["Portfolio Risk
Higher Than Expected"]
    style A fill:#1a1a2e,stroke:#c9a84c,color:#e5e7eb
    style D fill:#8b0000,stroke:#c9a84c,color:#e5e7eb
    style E fill:#8b0000,stroke:#c9a84c,color:#e5e7eb
```

- **Understated volatility:** Quarterly standard deviation might show 4% vs. true economic volatility of 12%+
- **Understated correlations:** Smooth returns appear uncorrelated with public equity markets, but the underlying economic exposure is actually quite correlated
- **Overstated Sharpe ratio:** Same return with lower apparent risk = inflated risk-adjusted return
- **MVO distortion:** The optimizer sees a high-return, low-volatility, low-correlation asset and allocates heavily to it — sometimes 30–50% in unconstrained optimization

**How to adjust (unsmoothing):**
Apply a statistical unsmoothing technique (like the Geltner adjustment) that de-lags the appraisal-based returns. This produces return series with more realistic volatility and correlation estimates. After unsmoothing, real estate volatility typically doubles or triples, and correlations with equities increase substantially.

**Exam Tip:** If an MVO produces an unrealistically large allocation to real estate or PE, the first thing to check is whether the inputs used appraisal-based data without unsmoothing.

Practice identifying these biases in our CFA Level III question bank.

How do survivorship bias and appraisal smoothing distort CME inputs?

Master Level III with our CFA Course

Related Questions

Related Articles

Practice Questions