Yesterday, I wrote a post on how asset managers and investment consultants seem to use mostly a CAPM with some bells and whistles to come up with long-term asset class forecasts. These forecasts are the key input to creating portfolio allocations and getting them right is of crucial importance. Yet, as I discussed yesterday, to come up with these long-term forecasts, most investors use a model that is empirically wrong. But what can you do if for some reason, all you have at your disposal is the CAPM and you still need to come up with long-term forecasts?
A team of researchers from Australia and New Zealand compared different simple methodologies to come up with return forecasts for the S&P 500 and US Treasuries for the next 10 years and then used these return forecasts to create stock/bond portfolios. To do this, they used the CAPM to come up with an expected equity risk premium for stocks, but they estimated the risk premium in a CAPM using different valuation methods.
The simplest approach is to use the earnings yield derived from R. Shiller’s CAPE as an estimate for the risk premium of the S&P 500. As I have shown yesterday, this is essentially what you get if you calculate the average return forecasts of asset managers and investment consultants. Hence, I will use the CAPE methodology to derive a risk premium that is used in the CAPM as shorthand for what is the consensus approach in asset allocation today.
This approach to estimating the risk premium of equities in a CAPM is then compared with four alternative approaches to estimating the risk premium:
The historic average risk premium (no CAPM estimation needed here)
The risk premium for stocks as derived from dividend yields and their relationship with future returns
The risk premium for stocks as derived from a Gordon Growth Model (GGM) using dividend yields and historic dividend growth as input
The risk premium derived from the combination of a Gordon Growth model and the CAPE valuation approach
Using these expected risk premia for stocks vs. bonds, one can create optimal portfolios using mean-variance optimisation and check how they performed in the 10 years after the portfolios were formed. Going back to 1891, the chart below shows the realised alpha of these optimised portfolios vs. the expected return as predicted by the CAPM.
Realised alpha from different forecast methodologies
Source: Ma et al. (2023)
Three things stand out.
First, using historic average returns is really not a good way to construct portfolios. Do you know that commonplace legal disclaimer “past returns are no indication for the future”? This is why that disclaimer exists.
Second, using the CAPE as a way to estimate future equity returns is much better than using historic average returns and can enhance portfolio returns quite a bit.
Third, the CAPE approach is pretty lousy when compared to a Gordon Growth Model or a combination of different return predictors. Compared to using the CAPE, you can typically increase your returns by some 0.5% or more by creating return forecasts from a collection of models, some of which include estimates for future growth. Note that the CAPE does not make any assumptions about future growth. Instead, it is purely valuation based and backward looking. But markets reward growth and by relying on valuation metrics like CAPE alone, you tend to ignore the growth potential of investments, and this can make you overly pessimistic about future returns.
Even if you are a value investor you need to assess future growth potential to be successful.
Standing back, do these precise valuation models give a false sense of ‘scientific’ accuracy? I can see that the Schiller CAPE is useful to assess the market - S&P500 over valued, FT-AS fair value. The Gordon GM is useful for stable Div paying companies. Otherwise is not the progress of the stock price down to ‘circumstances’ - for an individual company, and Macro changes for the market. We must always assess an individual stock by fundamentals; the flow of the market depends on Fed Reserve, economic environment, war/ pandemic, and so on.
2023 Charles H. Dow Award Winner:
https://cmtassociation.org/wp-content/uploads/2023/05/The-5-Percent-Canary.pdf
your comments...