Exposure Distributions in Equivalent-Risk Portfolios

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
User avatar
TD2626
Posts: 431
Joined: Thu Mar 16, 2017 3:40 pm

Exposure Distributions in Equivalent-Risk Portfolios

Post by TD2626 » Tue Oct 03, 2017 11:38 pm

Note: This is a theoretical post intended to examine where along a risk-return spectrum, for a constant, pre-defined level of desired overall portfolio risk/return, one should attempt to get one’s return. There are a lot of assumptions about market efficiency and which risk/return levels assets should go into – so this is more of a theory-based discussion.


Assume that we have a market that over the long run (many decades) is efficient. That means that an asset can get you more return for more risk, or less return for less risk, but not more return for less risk. Assume the investor has a long time horizon.

Say the market’s assets and their risk/return levels look like this:

(Bonds)
Risk/return level 1 – Lower-risk, lower return bonds (Short-term Gov’t Bond)
Risk/return level 2 - Total Bond Index Fund (or inter-term bonds)
Risk/return level 3 - Higher risk, higher return bonds (Long-term bonds, junk bonds, convertible bonds, Emerging Market bonds)
(Stocks)
Risk/return level 4 - Lower-risk, lower-return stocks (dividend stocks, preferred stocks, utilities, etc)
Risk/return level 5 - Total World Stock Market Index Fund (or stocks)
Risk/return level 6 – Higher-risk, higher return* stocks (Emerging Markets, Small Caps, Value Stocks)


*Note that “Higher risk, higher return” [stocks/bonds] indicates that after many decades of efficient market investment, these investments would, based on general consensus and prior data, be seen as reasonably likely to generate higher returns in exchange for higher risk (as measured by standard deviation). “Higher” is relative to the base case of total market funds of the relevant kind (stocks or bonds). All investing involves risk and history is no guarantee.



Which of the following portfolios would be best? Assume that the exact size of the tilts (generally 5-10% tilts) are adjusted so that each portfolio can be reasonably thought to have roughly the same expected risk and return of the standard 60/40 portfolio.

Portfolio A: Plain Vanilla Portfolio
Uses Total World Stock and Total Bond. A relatively standard 60/40 portfolio, say.

Portfolio B: Larry-like Portfolio
Starts with portfolio A but adds tilts towards Risk/Return level 1 assets and Risk/Return level 6 assets.

Portfolio C: Wellington-like,“income” or “inverse Larry” portfolio
Starts with portfolio A but adds tilts towards Risk/Return level 3 assets and Risk/Return level 4 assets.

Portfolio D: Double Barbell (barbell in bonds and barbell in stocks)
Starts with portfolio A but adds tilts towards Risk/Return levels 1, 3, 4, and 6.

snarlyjack
Posts: 431
Joined: Fri Aug 28, 2015 12:44 pm
Location: Montana

Re: Exposure Distributions in Equivalent-Risk Portfolios

Post by snarlyjack » Wed Oct 04, 2017 7:10 am

TD2626,

Since your asking about "theoretical" market efficiency.

Here is a article about the "efficient frontier" by Millennial Revolution.
Called "let's math this stuff up".

Enjoy...

http://www.millennial-revolution.com/in ... io-retire/

rkhusky
Posts: 4313
Joined: Thu Aug 18, 2011 8:09 pm

Re: Exposure Distributions in Equivalent-Risk Portfolios

Post by rkhusky » Wed Oct 04, 2017 7:24 am

TD2626 wrote:
Tue Oct 03, 2017 11:38 pm
Which of the following portfolios would be best? Assume that the exact size of the tilts (generally 5-10% tilts) are adjusted so that each portfolio can be reasonably thought to have roughly the same expected risk and return of the standard 60/40 portfolio.
I don't think "best" is knowable when pertaining to the future. You can do Monte Carlo or Portfolio Visualizer-type analysis to see what would have worked best in the past under various market conditions.

User avatar
TD2626
Posts: 431
Joined: Thu Mar 16, 2017 3:40 pm

Re: Exposure Distributions in Equivalent-Risk Portfolios

Post by TD2626 » Fri Oct 06, 2017 6:56 pm

rkhusky wrote:
Wed Oct 04, 2017 7:24 am
TD2626 wrote:
Tue Oct 03, 2017 11:38 pm
Which of the following portfolios would be best? Assume that the exact size of the tilts (generally 5-10% tilts) are adjusted so that each portfolio can be reasonably thought to have roughly the same expected risk and return of the standard 60/40 portfolio.
I don't think "best" is knowable when pertaining to the future. You can do Monte Carlo or Portfolio Visualizer-type analysis to see what would have worked best in the past under various market conditions.
I did some backtesting. I found nothing clear. Backtesting on Portfolio Visualizer usually only yields 10-20 years of data, which isn't long enough to evaluate super-long-term strategies.

These backtests are as follows:

Portfolio A: Plain Vanilla Portfolio
Uses Total World Stock and Total Bond. A relatively standard 60/40 portfolio, say.
---Approximated as Vanguard Life Strategy Moderate Growth and put into the "benchmark".

Portfolio B: Larry-like Portfolio
Starts with portfolio A but adds tilts towards Risk/Return level 1 assets and Risk/Return level 6 assets.
---A rough, hypothetical possibility was put into Portfolio 1 in PortfolioVisualizer.

Portfolio C: Wellington-like,“income” or “inverse Larry” portfolio
Starts with portfolio A but adds tilts towards Risk/Return level 3 assets and Risk/Return level 4 assets.
---A rough, hypothetical possibility was put into Portfolio 2 in PortfolioVisualizer.

Portfolio D: Double Barbell (barbell in bonds and barbell in stocks)
Starts with portfolio A but adds tilts towards Risk/Return levels 1, 3, 4, and 6.
---A rough, hypothetical possibility was put into Portfolio 3 in PortfolioVisualizer.

A backtest using asset classes:
https://www.portfoliovisualizer.com/bac ... Yield3=2.5

A somewhat different backtest using assets/funds:
https://www.portfoliovisualizer.com/bac ... tion12_3=3

Note that this is probably what people refer to as excessive backtesting, which is subject to biases and issues.. History is no guarantee. Also, these are non-optimized theoretical illustrations for discussion only.

I think that the main takeaway is that it would be hard to prove from backtesting alone which options are better.

rkhusky
Posts: 4313
Joined: Thu Aug 18, 2011 8:09 pm

Re: Exposure Distributions in Equivalent-Risk Portfolios

Post by rkhusky » Sat Oct 07, 2017 8:33 am

TD2626 wrote:
Fri Oct 06, 2017 6:56 pm
I think that the main takeaway is that it would be hard to prove from backtesting alone which options are better.
I think you are right.
Note how the difference in start dates affected the results between your two charts. Putting the start date for the first chart to 1999 makes them look very similar.
The smaller the tilt, the less likely you will be able to discern a difference. The Larry portfolio would go 100% to the most risky assets (like small value for both US and Int'l), with an increase in bonds to make the risk equivalent.
To see a longer time frame in Portfolio Visualizer, you have to choose investments that were around for a long time. Otherwise, you have to go to someplace like Ken French's site to get data to put into a spreadsheet. Morningstar has general categories that go back, but you can't combine them like you can on Portfolio Visualizer.

User avatar
TD2626
Posts: 431
Joined: Thu Mar 16, 2017 3:40 pm

Re: Exposure Distributions in Equivalent-Risk Portfolios

Post by TD2626 » Sat Oct 07, 2017 11:00 pm

Here are my opinions on this subject:

Portfolio A: Plain Vanilla Portfolio
Uses Total World Stock and Total Bond. A relatively standard 60/40 portfolio, say.
---This portfolio has simplicity and little complexity. It has substantial theoretical merit. However, from an intellectual perspective I wonder if it is possible to do better? This portfolio invests in relatively few asset classes and is exposed primarily to market beta risk.

Portfolio B: Larry-like Portfolio
Starts with portfolio A but adds tilts towards Risk/Return level 1 assets and Risk/Return level 6 assets.
---This portfolio may reduce kurtosis risk. However, there is a substantial risk of underperformance if the factor tilts don't work out.

Portfolio C: Wellington-like,“income” or “inverse Larry” portfolio
Starts with portfolio A but adds tilts towards Risk/Return level 3 assets and Risk/Return level 4 assets.
---This portfolio may have fairly high income, aiding those with a "spend the income" policy. The assets tilted towards could in my opinion have a relatively high Sharpe ratio to make up for a high correlation to the 60/40 portfolio. (History doesn't necessarily support this,though... it's just an opinion.) (Is this reasonable or did I make a mistake with the logic?)

Portfolio D: Double Barbell (barbell in bonds and barbell in stocks)
Starts with portfolio A but adds tilts towards Risk/Return levels 1, 3, 4, and 6.
---This may be reasonable for someone with a high complexity tolerance and with a deep understanding of all assets involved. It combines some of the potential benefits of B and C (and the potential drawbacks). It is likely difficult to implement, though. Also, if you try to "tilt toward everything interesting" you end up underweighting the plain, boring things (like "blend" stocks and intermediate-term bonds) that could arguably be the best value over the long run.

Any ideas? I am particularly interested on feedback regarding portfolios C and D.

snarlyjack
Posts: 431
Joined: Fri Aug 28, 2015 12:44 pm
Location: Montana

Re: Exposure Distributions in Equivalent-Risk Portfolios

Post by snarlyjack » Sun Oct 08, 2017 12:00 am

TD2626,

I kind of like simple portfolios. They have lower ER's.
I did your back test with just 1 fund. The TSM fund
(by itself) blew all of your portfolios out of the water,
by $Thousands of dollars. Give it a try.

Post Reply