The Ultimate Objective: Getting Long-Run Returns Right
Before examining the framework itself, we must be clear about what it needs to accomplish. The ultimate objective is to develop a set of projections with which to make informed asset allocation decisions. Since asset allocation is the primary determinant of long-run portfolio performance, the projections underlying these decisions are among the most important determinants of whether investors achieve their long-term goals.
It follows that getting the long-run level of returns approximately right must be a top priority. Even sophisticated methods will produce frustratingly large forecast errors over relevant horizons. We should seek to limit those errors, but we should not put undue emphasis on the precision of projections for individual asset classes.
A Cautionary Tale: The Technology Bubble
Until the late 1990s, many institutional investors simply extrapolated historical return data into forecasts. During the technology bubble, this practice led many to project double-digit portfolio returns indefinitely. Such inflated projections allowed institutions to underfund their obligations and set unrealistic goals — many of which had to be painfully scaled back when the bubble burst.
Since that time, most institutions have adopted explicitly forward-looking methods. Return projections have declined sharply. As a practical illustration, US private foundations (which must distribute at least 5% of assets annually) may struggle to generate long-run returns sufficient to cover their required distributions, expenses, and inflation under realistic projections.
Consistency Over Precision
Far more important than individual precision are two forms of internal consistency:
Cross-sectional consistency ensures that estimates for all asset classes reflect the same underlying assumptions. If your macro view assumes 2.5% real growth, every asset class projection — equities, bonds, real estate — should be anchored to that same assumption. Inconsistency across asset classes leads to portfolios with poor risk-return characteristics over any horizon.
Intertemporal consistency ensures that estimates over different time horizons connect through a plausible path. If you project a recession next year but 8% annualized equity returns over the next decade, the implied recovery path must be realistic. Inconsistency across horizons distorts the connection between portfolio decisions and investment time frame.
The 7-Step CME Framework
The following flowchart shows the complete disciplined process:
Let us examine each step in detail.
Step 1: Specify the Asset Classes and Time Horizons
The analyst must define the universe of asset classes for which expectations will be developed. This universe should include all asset classes that will typically receive a distinct allocation in client portfolios — it needs to reflect the key dimensions of decision-making in the firm's investment process.
At the same time, the universe should be kept as small as possible. Even a minimal set of asset classes creates a challenging expectations-setting task. For six asset classes, you need 27 quantitative inputs (6 returns, 6 volatilities, 15 correlations). For ten classes, that number rises to 65.
The decision hierarchy matters. One firm may prioritize segmenting global equities by sector, with geographic distinctions as secondary. Another firm prioritizes geography, treating sector breakdowns as secondary. The asset class universe should align with whichever hierarchy the firm actually uses.
Step 2: Research the Historical Record
Most forecasts have some connection to the past. The historical record contains useful information on investment characteristics, suggesting at least some possible ranges for future results. Beyond raw historical facts, the analyst should identify and understand the factors that affect asset class returns.
The information must be sliced across multiple dimensions: geography (global, regional, domestic versus non-domestic), major asset classes (equity, fixed-income, real assets), and sub-asset classes (equity styles and sizes, fixed-income maturities and credit qualities, real estate and commodities).
Step 3: Select Methods and Models
The analyst must be explicit about which forecasting methods will be used and must justify the selection. Critically, the effectiveness of different approaches varies with the investment time horizon.
A discounted cash flow approach is usually most appropriate for long-range equity forecasting. If shorter-horizon forecasts are also needed, intertemporal consistency requires that the short-term method be calibrated so its projections converge to the long-range forecast as the horizon extends.
| Method | Best Horizon | Typical Application |
|---|---|---|
| Discounted Cash Flow | Long-term (10-20 yr) | Equity fair value |
| Grinold-Kroner | Long-term (10-20 yr) | Equity return decomposition |
| Building Blocks | Long-term (10-20 yr) | Fixed-income expected returns |
| Business Cycle Analysis | Medium-term (1-3 yr) | Tactical tilts |
| Momentum/Technical | Short-term (1-12 mo) | Tactical trading |
| Econometric Models | Variable | Multi-factor forecasting |
Step 4: Determine the Best Data Sources
Using flawed or misunderstood data is a recipe for faulty analysis. The analyst must research the quality of alternative data sources and strive to fully understand the data.
- Broad coverage: Large, commercially available databases (Bloomberg, FactSet) and reputable financial publications provide widely disseminated information covering the broad spectrum of asset classes and geographies.
- Specialized sources: Trade publications, academic studies, government and central bank reports, corporate filings, and broker/dealer research provide more specialized information.
- Data frequency: Daily series are more useful for shorter-term expectations. Monthly, quarterly, or annual data series are useful for longer-term CME.
Analysts should remain alert to new, superior data sources as they become available.
Step 5: Interpret the Current Environment
This is the heart of the process. The analyst applies selected models to current data, using experience and judgment to interpret results. A common set of assumptions, compatible methodologies, and consistent judgments must be applied to ensure mutually consistent projections across asset classes and over time horizons.
This step could be described as implementing your investment and research process. The analyst should be doing this work every day, continuously updating their understanding of the economic and market environment.
Step 6: Document Conclusions
At designated intervals, the analyst must synthesize, document, and defend their views. The projections should be accompanied by the reasoning and assumptions behind them. What distinguishes this step from day-to-day research is that the analyst must make simultaneous projections for all asset classes and all designated horizons.
This documentation creates accountability and enables retrospective evaluation of what worked and what did not.
Step 7: Monitor, Review, and Improve
The final step uses experience to improve the expectations-setting process. Previously formed expectations are measured against actual results to assess accuracy. Good forecasts are:
- Unbiased — objective and well-researched
- Efficient — minimizing the size of forecast errors
- Internally consistent — both cross-sectionally and intertemporally
A standard rule of thumb in statistics is that at least 30 observations are needed to meaningfully test a hypothesis. Since CME projections are typically updated quarterly or annually, quantitative evaluation of forecast errors in real time may be of limited value for a process that is already reasonably well constructed.
The most valuable part of the feedback loop is often qualitative and judgmental — understanding why certain forecasts missed and how the process can be refined.
The Centralized Approach
As the setting of explicit capital market expectations has become both more common and more sophisticated, many asset managers have adopted a centralized approach. A dedicated analyst or team is responsible for developing projections used by the firm's investment professionals across all client portfolios.
This centralization enables the firm to leverage the requisite expertise and deliver more consistent advice to all clients. Without it, different portfolio managers might use different macro assumptions, leading to inconsistent recommendations across the firm.
Key Takeaways for CFA Level III
- Getting the long-run return level approximately right is the top priority
- Cross-sectional and intertemporal consistency matter more than individual precision
- The asset class universe should mirror the firm's decision hierarchy
- Short-term models must converge to long-run projections as the horizon extends
- The feedback loop is often qualitative rather than purely statistical
- Forward-looking methods have replaced naive historical extrapolation
Test your understanding of the CME framework with our practice questions, or explore related topics in our CFA Level III community Q&A.