Research Affiliates: Starting Yields Matter To Future Returns

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Random Walker
Posts: 2120
Joined: Fri Feb 23, 2007 8:21 pm

Research Affiliates: Starting Yields Matter To Future Returns

Post by Random Walker » Fri Oct 06, 2017 12:45 am

http://www.etf.com/sections/features-an ... nopaging=1

I came across this Research Affiliates article on etf.com. I think using historical average returns to make asset allocation decisions is a mistake. But I do believe historical standard deviations are useful. Expected returns depend tremendously on current valuations. The two most popular ways to estimate returns are the Gordon Discount Dividend Model (I believe what Bogle uses ) and 1/CAPE10. This article describes the problems with historical returns and how to better estimate future returns. Eager to hear comments.

Dave

Always passive
Posts: 149
Joined: Fri Apr 14, 2017 4:25 am
Location: Israel

Re: Research Affiliates: Starting Yields Matter To Future Returns

Post by Always passive » Fri Oct 06, 2017 1:01 am

Random Walker wrote:
Fri Oct 06, 2017 12:45 am
http://www.etf.com/sections/features-an ... nopaging=1

I came across this Research Affiliates article on etf.com. I think using historical average returns to make asset allocation decisions is a mistake. But I do believe historical standard deviations are useful. Expected returns depend tremendously on current valuations. The two most popular ways to estimate returns are the Gordon Discount Dividend Model (I believe what Bogle uses ) and 1/CAPE10. This article describes the problems with historical returns and how to better estimate future returns. Eager to hear comments.

Dave
So you agree with the paper, right? If that is the case, have you changed the geographic equity allocation to reflect the current 1/PE10s?

Random Walker
Posts: 2120
Joined: Fri Feb 23, 2007 8:21 pm

Re: Research Affiliates: Starting Yields Matter To Future Returns

Post by Random Walker » Fri Oct 06, 2017 7:54 am

Always Passive,
Yes I have. BUT only because it was a convenient move towards more efficient world market cap allocation. About two years ago my Equity was roughly 60/40 US/Int. I took advantage of the difference in valuations and thus expected returns to move to 50/50 US/Int on the equity side.
I would have made the above change at any point of my investing lifetime. I also started to decrease my overall equity allocation at about the same time. This decision was based on my place in the investing lifecycle, current valuations, expected returns. I know this smells of market timing to some people, but I view it more as customizing one's glide path towards a retirement portfolio.

Dave

User avatar
in_reality
Posts: 4320
Joined: Fri Jul 12, 2013 6:13 am

Re: Research Affiliates: Starting Yields Matter To Future Returns

Post by in_reality » Fri Oct 06, 2017 8:10 am

Always passive wrote:
Fri Oct 06, 2017 1:01 am
Random Walker wrote:
Fri Oct 06, 2017 12:45 am
http://www.etf.com/sections/features-an ... nopaging=1

I came across this Research Affiliates article on etf.com. I think using historical average returns to make asset allocation decisions is a mistake. But I do believe historical standard deviations are useful. Expected returns depend tremendously on current valuations. The two most popular ways to estimate returns are the Gordon Discount Dividend Model (I believe what Bogle uses ) and 1/CAPE10. This article describes the problems with historical returns and how to better estimate future returns. Eager to hear comments.

Dave
So you agree with the paper, right? If that is the case, have you changed the geographic equity allocation to reflect the current 1/PE10s?
Not the OP, but I am moving more toward global cap weighting by reducing US holdings.

In any case, the Research Affiliates Asset Allocation tool is interesting in that you can chart this for various asset classes and see their calculations for their expected returns. You can also choose the return model to use valuations or not (and just use Dividend and Growth). Not everyone agrees that US valuations are as high as they appear (due to accounting changes). Anyway, if you want, you can even create and account and create a blended model say 50% Valuation based and 50% (only Growth + Yield).

The thing that confuses me about their predictions is how much lower developed markets are forecast in their smart beta tool. It's somewhat similar to their Asset Allocation tool but uses factors as asset classes. Looking at the Fundamental Index for Developed markets, you can select to see expected real returns in excess of the index as well as realized 5 year real returns in excess of the index (hint: lower than expected and not very different). In any case, even using the rosier excess expected returns - the total real expected returns (of the "superior" fundamental index) is under 4% (no exact figure is given but I'll guess 3.8%). By comparison, the Asset Allocation tool has expected real returns of the EAFE at 4.8%. Maybe Canada is not in the EAFE but is in the Fundamental Index, but surely there is some other difference in the tools. Why is the fundamental index showing lower expected returns than the market in the asset allocation tool if it's also showing it's expected to out do the benchmark at current valuations.

Anyway, their smart beta tool also includes a number to indicate how expensive a factor is relative to it's own history which could be useful.

So they have a lot of information there, but the tool takes some looking at to figure out. And there are so many calculation methods to understand (for example multiple types of value, multiple of momentum etc etc).

Anyway, their article points out that Bogle is saying the same thing - valuations are likely to reduce future returns. RA points to Vanguard's tool which uses 4% real (higher than Bogle's estimate as he apparently included valuations by subtracting 2% from the expected returns of growth and yield ). In truth though, Vanguard does include a hefty allocation to international which has expected returns in their target retirement and advisory offerings.

I think it's not impossible for US valuations to stay high for a long time. US expected returns from Growth + Yield (excluding valuations) are not so different than international (~3.5% US vs 4.4% DEV vs 4% EM). So while there is a possibility that the US sees some reversion towards historic valuations, I am not going to make a big bet on it happening. The US could see a 2% lowered return in the next 10 years from valuation, but there are many things in play here and if the US is viewed as a better market to invest in, valuations could stay where they are. I can't predict what people will do that well.

Always passive
Posts: 149
Joined: Fri Apr 14, 2017 4:25 am
Location: Israel

Re: Research Affiliates: Starting Yields Matter To Future Returns

Post by Always passive » Fri Oct 06, 2017 10:38 am

in_reality wrote:
Fri Oct 06, 2017 8:10 am
Always passive wrote:
Fri Oct 06, 2017 1:01 am
Random Walker wrote:
Fri Oct 06, 2017 12:45 am
http://www.etf.com/sections/features-an ... nopaging=1

I came across this Research Affiliates article on etf.com. I think using historical average returns to make asset allocation decisions is a mistake. But I do believe historical standard deviations are useful. Expected returns depend tremendously on current valuations. The two most popular ways to estimate returns are the Gordon Discount Dividend Model (I believe what Bogle uses ) and 1/CAPE10. This article describes the problems with historical returns and how to better estimate future returns. Eager to hear comments.

Dave
So you agree with the paper, right? If that is the case, have you changed the geographic equity allocation to reflect the current 1/PE10s?
Not the OP, but I am moving more toward global cap weighting by reducing US holdings.

In any case, the Research Affiliates Asset Allocation tool is interesting in that you can chart this for various asset classes and see their calculations for their expected returns. You can also choose the return model to use valuations or not (and just use Dividend and Growth). Not everyone agrees that US valuations are as high as they appear (due to accounting changes). Anyway, if you want, you can even create and account and create a blended model say 50% Valuation based and 50% (only Growth + Yield).

The thing that confuses me about their predictions is how much lower developed markets are forecast in their smart beta tool. It's somewhat similar to their Asset Allocation tool but uses factors as asset classes. Looking at the Fundamental Index for Developed markets, you can select to see expected real returns in excess of the index as well as realized 5 year real returns in excess of the index (hint: lower than expected and not very different). In any case, even using the rosier excess expected returns - the total real expected returns (of the "superior" fundamental index) is under 4% (no exact figure is given but I'll guess 3.8%). By comparison, the Asset Allocation tool has expected real returns of the EAFE at 4.8%. Maybe Canada is not in the EAFE but is in the Fundamental Index, but surely there is some other difference in the tools. Why is the fundamental index showing lower expected returns than the market in the asset allocation tool if it's also showing it's expected to out do the benchmark at current valuations.

Anyway, their smart beta tool also includes a number to indicate how expensive a factor is relative to it's own history which could be useful.

So they have a lot of information there, but the tool takes some looking at to figure out. And there are so many calculation methods to understand (for example multiple types of value, multiple of momentum etc etc).

Anyway, their article points out that Bogle is saying the same thing - valuations are likely to reduce future returns. RA points to Vanguard's tool which uses 4% real (higher than Bogle's estimate as he apparently included valuations by subtracting 2% from the expected returns of growth and yield ). In truth though, Vanguard does include a hefty allocation to international which has expected returns in their target retirement and advisory offerings.

I think it's not impossible for US valuations to stay high for a long time. US expected returns from Growth + Yield (excluding valuations) are not so different than international (~3.5% US vs 4.4% DEV vs 4% EM). So while there is a possibility that the US sees some reversion towards historic valuations, I am not going to make a big bet on it happening. The US could see a 2% lowered return in the next 10 years from valuation, but there are many things in play here and if the US is viewed as a better market to invest in, valuations could stay where they are. I can't predict what people will do that well.
If you look at the RA efficient model portfolios, none of them include US stocks.

User avatar
in_reality
Posts: 4320
Joined: Fri Jul 12, 2013 6:13 am

Re: Research Affiliates: Starting Yields Matter To Future Returns

Post by in_reality » Fri Oct 06, 2017 11:50 pm

Always passive wrote:
Fri Oct 06, 2017 10:38 am
in_reality wrote:
Fri Oct 06, 2017 8:10 am
Always passive wrote:
Fri Oct 06, 2017 1:01 am
Random Walker wrote:
Fri Oct 06, 2017 12:45 am
http://www.etf.com/sections/features-an ... nopaging=1

I came across this Research Affiliates article on etf.com. I think using historical average returns to make asset allocation decisions is a mistake. But I do believe historical standard deviations are useful. Expected returns depend tremendously on current valuations. The two most popular ways to estimate returns are the Gordon Discount Dividend Model (I believe what Bogle uses ) and 1/CAPE10. This article describes the problems with historical returns and how to better estimate future returns. Eager to hear comments.

Dave
So you agree with the paper, right? If that is the case, have you changed the geographic equity allocation to reflect the current 1/PE10s?
Not the OP, but I am moving more toward global cap weighting by reducing US holdings.

In any case, the Research Affiliates Asset Allocation tool is interesting in that you can chart this for various asset classes and see their calculations for their expected returns. You can also choose the return model to use valuations or not (and just use Dividend and Growth). Not everyone agrees that US valuations are as high as they appear (due to accounting changes). Anyway, if you want, you can even create and account and create a blended model say 50% Valuation based and 50% (only Growth + Yield).

The thing that confuses me about their predictions is how much lower developed markets are forecast in their smart beta tool. It's somewhat similar to their Asset Allocation tool but uses factors as asset classes. Looking at the Fundamental Index for Developed markets, you can select to see expected real returns in excess of the index as well as realized 5 year real returns in excess of the index (hint: lower than expected and not very different). In any case, even using the rosier excess expected returns - the total real expected returns (of the "superior" fundamental index) is under 4% (no exact figure is given but I'll guess 3.8%). By comparison, the Asset Allocation tool has expected real returns of the EAFE at 4.8%. Maybe Canada is not in the EAFE but is in the Fundamental Index, but surely there is some other difference in the tools. Why is the fundamental index showing lower expected returns than the market in the asset allocation tool if it's also showing it's expected to out do the benchmark at current valuations.

Anyway, their smart beta tool also includes a number to indicate how expensive a factor is relative to it's own history which could be useful.

So they have a lot of information there, but the tool takes some looking at to figure out. And there are so many calculation methods to understand (for example multiple types of value, multiple of momentum etc etc).

Anyway, their article points out that Bogle is saying the same thing - valuations are likely to reduce future returns. RA points to Vanguard's tool which uses 4% real (higher than Bogle's estimate as he apparently included valuations by subtracting 2% from the expected returns of growth and yield ). In truth though, Vanguard does include a hefty allocation to international which has expected returns in their target retirement and advisory offerings.

I think it's not impossible for US valuations to stay high for a long time. US expected returns from Growth + Yield (excluding valuations) are not so different than international (~3.5% US vs 4.4% DEV vs 4% EM). So while there is a possibility that the US sees some reversion towards historic valuations, I am not going to make a big bet on it happening. The US could see a 2% lowered return in the next 10 years from valuation, but there are many things in play here and if the US is viewed as a better market to invest in, valuations could stay where they are. I can't predict what people will do that well.
If you look at the RA efficient model portfolios, none of them include US stocks.
That's not entirely true.

As I mentioned, you can also select "Dividend and Growth only" (and not use valuation measures) for future returns. Using that method of future returns, the lowest Volatility point of 4% in the RA efficient model portfolios actually has about 50% of stocks in US small caps. The highest Volatility point of 14% has about 27% of stocks in the US.

Valuations returning to a historical mean are a possibility, and in that model perhaps there are no US stocks. Such an event is not a given though, and even if it does occur, there is nothing to say it couldn't happen gradually over years or decades or several decades. Again, accounting changes could mean the historical level of valuations is too low for the world today.

Also, note that:
When excluding FANG stocks (Facebook, Amazon, Netflix, and Google parent Alphabet) plus the two largest technology companies (Apple and Microsoft), the U.S. does not look as overvalued as it may first appear. In fact, its price-to-earnings ratio becomes comparable to international markets, which have been touted for having more bargain opportunities during the recent expansion.
So perhaps there is some wisdom in using a US value fund and avoiding Banco Sabadell SA and CaixaBank SA (Spanish bank stocks) or other companies that could be disrupted by potential European fragmentation (UK, Catalonia)

User avatar
packer16
Posts: 890
Joined: Sat Jan 04, 2014 2:28 pm

Re: Research Affiliates: Starting Yields Matter To Future Returns

Post by packer16 » Sat Oct 07, 2017 12:53 pm

The biggest issue here is they exclude any data to the contrary. The primary data I see that contradicts this is that put together by Damodaran. He estimates, based upon current prices & forward looking data that the current ERP is 4 to 5%. So if you assume that the bonds will return 1% real you get 5 to 6%. The biggest issue with RAs methodology is that implicitly assumes the average historical interest rate which generate the average historical P/Es (which is probably like 5%+) will happen again in the future. IMO this assumption is wrong! The factors that drove the 5%+ interest rates is higher inflation which is driven by shortages of labor & collective effort (in the case of the communist countries) which we have no sign of today. Investing is relative returns game & today the premium for owning stocks vs. bonds is still quite high.

The other issue they miss the one of growth versus starting price. Over time faster growth will overcome cheapness. This is like comparing a compounder to declining revenue stock. Buying the decliner because it is cheaper is silly because it may be more overvalued than the compounder. I look at the US as a compounder some EMs lets say Russia for instance like a declining revenue stock. Should we invest in Russia vs. US because it is cheaper today? It depends. If the US can grow faster than Russia then the US may cheaper even though the multiple today is higher.

Another more cynical point of view is that since RA is compensated based upon a % of AUM, they have a vested interest in keeping future growth potential low as they will collect more fees.

Packer
Buy cheap and something good might happen

Post Reply