Expected returns and risk metrics are foundational inputs for constructing asset allocations. Investors must compare potential returns across asset classes with clarity and precision. In this article, we outline a handful of checks that help ensure your Capital Market Assumptions are fit for this purpose.
A quick note on definitions — the most commonly used definition for CMAs is long-term expected returns by asset class, typically projected over a 10-year horizon. As we’ve noted previously, expected returns in isolation only tell part of the story. Full CMAs must incorporate risk inputs too — in this article, we’ll treat “CMAs” as referring to both expected return and risk assumptions.
There is a broad landscape of providers publishing long-term expected returns. A smaller subset also include risk metrics (though rarely on a true long-term basis). These providers tend to be large asset managers, investment consultants or specialist research houses, and their assumptions are generally updated annually, at best monthly.
The ubiquity of CMA publications can be misleading — glossy PDFs may stray into marketing territory, and it is surprisingly difficult to build a coherent, comprehensive and practically useful set of CMAs.
Key checks for your CMA estimates
Many providers do not deliver a framework that is robust enough for strategic and tactical asset allocation processes: they may omit key asset classes, mis-specify horizons or apply conceptual shortcuts that undermine the quality of the assumptions. To help you evaluate your CMA source, here are some important questions to ask.
1. Is there evidence the CMAs “work”?
This is perhaps the most intuitive question — can these estimates reliably predict future returns or measure future risk?
In practice the answer is often “no” — partly because it’s inherently difficult. For example, if your expected return estimates are for a 10-year horizon, you may only have a few decades of data, which leaves you with very few independent observations (e.g., 30 years of data gives effectively only three 10-year periods). Using overlapping return periods in regressions can help, but the numbers are against you.
One practical workaround: build short-term forecasts (e.g., one-year ahead) within the same modelling framework as the long-term forecasts. With, say, 1-year forecasts, you can then test predictive power more meaningfully. We use a framework where short- and long-term forecasts are tightly linked – the long-term is a series of short-term intervals after all. This gives us reassurance that long-term expected returns are reasonable on the basis that of evidence attached to short-term expected returns. In addition, short-term estimates of expected returns, at high frequencies, for a full set of investment opportunities, is also prerequisite for navigating market opportunities (irrespective of the way they are implemented e.g. TPA vs TAA/SAA).
2. Do the CMAs include risk estimates, and how is long-term risk modelled?
As mentioned, many CMA sets omit risk entirely — which means you’re missing a major pillar of asset allocation design. Even when risk metrics are included, we regularly encounter conceptual issues.
If expected returns are referring to a 10-year horizon, then presumably risk estimates are too, and this is where things often go wrong. A common example: even sophisticated investment houses project the volatility of cash to be near 0% over a 10-year horizon. The issue is that, while cash may be near risk-free over a short-horizon, over longer horizons an investor must re-invest multiple times. This exposes investors to rollover risk, where re-investment will be subject to higher or lower rates on cash investments, which are unknown in advance. When modelling this effect properly, returns over a 10-year period can have a volatility (annualised) above 6%.
More broadly: modelling long-term volatility, correlations and other risk metrics is inherently more complex than short-term risk modelling. Many financial modelling frameworks are built for horizons of days, weeks or months — not decades. We have written before on how we address some of the factors to think about over the long-term, and there is plenty of further uncharted territory to cover going forward. The essence of our thinking is that risk modelling requires forward-looking macro-finance simulation models to make sense of complex interactions over long horizons.
3. Are the CMAs tailored to your circumstances, and do they respond quickly when the market changes?
Generic assumptions can be useful, but strategic asset allocation decisions often require adjustments for your own context: currency base (USD vs GBP vs EUR), horizon (5 years vs 20 years), real vs nominal returns, and the specific asset-classes you invest in (emerging single-country equities, private markets, hedge-fund strategies). Many CMAs are implicitly treating all investors as one size fits all.
Additionally, market conditions evolve. Some providers may revise only annually even when fundamentals have shifted, especially in today’s environment where macro volatility is high and regimes are shifting. CMA frameworks need to be more nimble, offering short- and long horizon expected returns, at a higher frequency – ideally updated every day.
4. Are CMAs jointly determined, accounting for common macro drivers?
A further area worth examining is whether your CMA estimates are consistently defined across a consistent set of macro and market drivers. Many CMA frameworks are constructed asset-class-by-asset-class, each with its own assumptions and its own calibration history, that lead to internal contradictions.
Expected returns and risk across asset classes are connected through a set of underlying macro present value drivers. A coherent CMA framework reflects that structure. If expected returns shift because real rates rise, you should see bond, equity, and private market expectations adjust in a consistent way, with magnitudes that make intuitive and quantitative sense.
Final thoughts
Building your own CMA framework is not a trivial task. Under the hood it requires substantial data inputs, sophisticated modelling, and ongoing maintenance — as we previously discussed in this article.
That’s why many rely on external CMA sources — yet because of the variation in methodological quality and contextual fit, it is vital to validate and challenge the estimates rather than accept them at face value.
At Allocation Strategy, we have designed a CMA framework to address the key pain-points identified above: we integrate short-term forecasts, include robust long-term risk modelling, and offer client-specific tailoring for currency, horizon and asset-class coverage.