• CMAs seek to help allocators with their asset allocation decisions by establishing long-term forecasts for different financial assets. The problem is that they either fail to do this with much accuracy or make their forecasts with such a degree of caution that the output is of limited practical use.
• In this paper, we propose a framework to help allocators to calculate these assumptions for themselves, as well as explore some practical applications for the resulting data.
• A critical question that is not given sufficient consideration in our view is: what are you proposing to use CMAs for? There seem to be two options. First, allocators may have a liability window over the next 10 years, and want a base case for how assets might match these liabilities. Or secondly, they may want an unconditional base rate which serves as the foundation for ultra-long portfolio allocations. If the latter, and we suspect this is the case for most users, then the timeframe consideration becomes particularly important.
• In Part One, we discuss forecast timeframes. In Part Two, we describe four credible methods for determining CMAs. In Part Three, we use findings from the prior two sections to calculate a range of CMAs for US stocks and 10-year Treasuries (UST10). These assumptions allow us, in Part Four, to derive strategic asset allocations (SAA), based on optimisation parameters such as return, volatility drawdown and Sharpe ratio.
• In this paper we focus only on US equities and intermediate duration, as the two most consequential portfolio building blocks.
  
  
  Part One: CMA timeframes

Inherent in CMAs is the idea that asset returns are more predictable on a long-term horizon than on a short-term view. In Figure 1, we show the distribution of one- and 10-year rolling returns to US stocks and bonds, over the ultra-long-term.

• Figure 1. Distribution of returns for US stocks (top) and bonds (bottom), 1800-2024

• The conditional historic method takes the same historic data, but conditions it based on starting valuation. It employs the apparent truism that a higher starting valuation should result in a lower forward return. If we take the smoothed earnings yield on equities as a proxy for return, and the same for the yield-to-maturity on US Treasuries, we see an attractive fit between starting valuation and forward return, at least on an overlapping basis
  
  
  Figure 3. Relationship between starting valuation and 10-year forward return for US stocks (top) and bonds (bottom), 1881-2024