Capital Market Expectations Framework for CFA Level II

Capital market expectations (CME) are your views on the expected returns, risks and correlations of asset classes. They are inputs to the strategic asset allocation. They are not forecasts of single securities. Weak CME give weak portfolios, however good the optimizer is.

The framework is a seven-step sequence:

• Step 1: Specify the set of expectations needed, including the time horizon.
• Step 2: Research the historical record.
• Step 3: Specify the method and/or model to use and its information requirements.
• Step 4: Determine the best sources for the information.
• Step 5: Interpret the current investment environment using the selected data and methods.
• Step 6: Provide the set of expectations and document the results.
• Step 7: Monitor performance and use it to refine the process.

The forecaster's tools fall into three groups. Statistical methods use historical data: sample means, shrinkage estimators and time-series models. Discounted cash flow (DCF) methods estimate the expected return from current prices and expected cash flows. The Gordon growth model is a DCF model: expected equity return ≈ the dividend yield plus the long-run growth rate. For bonds, the yield to maturity plays a similar role. The Grinold-Kroner model is a DCF-style decomposition of the expected equity return into income, earnings growth and repricing. Risk premium approaches build expected return as a risk-free rate plus premiums. The Singer-Terhaar approach is a risk premium approach built on a global (international) CAPM. Many vignettes also mention economic analysis and survey views, so know that these are judgment-based.

The exam focuses on what goes wrong. Data limitations include: historical estimates that are subject to regime change, so past data may not hold in the future; a short history or data that is too infrequent; non-independence of data (such as serial correlation); and inconsistent definitions of indexes or asset classes. Be cautious when using ex post (historical) risk to forecast ex ante risk. Historical risk can mis-forecast future risk, for example when a rare, risky event did or did not occur in the sample period.

Data measurement errors and biases are a separate group: transcription errors, survivorship bias (failed funds are removed from the database), and appraisal (smoothed) data, which understates volatility and correlation with other assets. Backfill bias is a related selection issue: funds enter a database after good performance and their earlier history is added.

The curriculum's forecasting pitfalls are:

• Failure to account for the economic cycle.
• Historical estimates not conditioned on the current state: using unconditional averages when current conditions differ.
• Inappropriate use of historical estimates: treating a sample from one period or regime as if it describes the future.
• Inappropriate analytical methods: for example data-mining bias (searching until something fits) and time-period bias (results depend on the sample period).
• Failure to account for conditioning information: ignoring information that changes the expected distribution.
• Misinterpretation of correlation: for example, treating correlations as stable when they change across periods or in stress.
• Psychological traps: anchoring, status quo, confirmation, overconfidence, prudence and recallability.
• Model and input uncertainty.

Supplementary points that often appear in vignettes: inconsistent time horizons (using a forecast built for one horizon in a decision with another). A forecast is best seen as a probability distribution, not a single point.

Most questions give a vignette describing how an analyst built forecasts. Your job is to name the step, limitation or bias, and then state its effect.