Engineered Portfolio

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
dowse
Posts: 349
Joined: Tue Sep 10, 2013 3:10 pm

Engineered Portfolio

Post by dowse » Tue Mar 27, 2018 9:38 am

I'm curious what folks here think about the blog "Engineered Portfolio", https://engineeredportfolio.com/ mentioned in another thread. I'm intrigued by the portfolios they came up with based on extensive numerical analysis. Of course, it's all in the underlying assumptions and choice of data set. Interestingly, their portfolios only include bond funds if the time horizon is 2-3 years.

Valuethinker
Posts: 36681
Joined: Fri May 11, 2007 11:07 am

Re: Engineered Portfolio

Post by Valuethinker » Tue Mar 27, 2018 10:02 am

dowse wrote:
Tue Mar 27, 2018 9:38 am
I'm curious what folks here think about the blog "Engineered Portfolio", https://engineeredportfolio.com/ mentioned in another thread. I'm intrigued by the portfolios they came up with based on extensive numerical analysis. Of course, it's all in the underlying assumptions and choice of data set. Interestingly, their portfolios only include bond funds if the time horizon is 2-3 years.
"Skating to where the puck was".

The problem with Markowitz Optimizers, as Bill Bernstein points out, is that you drop in historical correlations and returns and you get a portfolio designed to outperform in the past. The future optimal portfolio will be different.

South Africa & Sweden? Really. Huge overweightings. If you want Gold & Precious Metal mining stocks, Vanguard has a fund for that. Sweden? It's a handful of multinationals plus some banks.

Like many here, one of the effects of spending time on this Forum is to tend to increase your bond weightings. Historically I have held almost no bonds. But as I age, and as it becomes clear just how bad the 2008-09 Crash was, the extra risk of being nearly 100% equities no longer seems worth it. Talking to the many learned people here has suggested to me that one should hold more bonds.

123
Posts: 3900
Joined: Fri Oct 12, 2012 3:55 pm

Re: Engineered Portfolio

Post by 123 » Tue Mar 27, 2018 10:06 am

They're a little quirky since they seem to suggest South Africa equity, Sweden equity, and gold for 10% each (30% combined) for a number of portfolios on the time horizon. It would seem that they are ignoring the old saying that "past performance does not indicate future results". On the surface it looks like they just cooked up some suggestions based on past performance. They are no different than any financial adviser in that respect. Look at the great results I could have gotten if I could predict the future. Well, yeah, anyone can say that.
The closest helping hand is at the end of your own arm.

User avatar
patrick013
Posts: 2427
Joined: Mon Jul 13, 2015 7:49 pm

Re: Engineered Portfolio

Post by patrick013 » Tue Mar 27, 2018 12:02 pm

I think they're doing 5 times more work than needed but the
data seems reliable, etc.. Expected returns are after inflation
which adds another estimate to the mix.
age in bonds, buy-and-hold, 10 year business cycle

User avatar
iceport
Posts: 4008
Joined: Sat Apr 07, 2007 4:29 pm

Re: Engineered Portfolio

Post by iceport » Tue Mar 27, 2018 12:29 pm

Valuethinker wrote:
Tue Mar 27, 2018 10:02 am
dowse wrote:
Tue Mar 27, 2018 9:38 am
I'm curious what folks here think about the blog "Engineered Portfolio", https://engineeredportfolio.com/ mentioned in another thread. I'm intrigued by the portfolios they came up with based on extensive numerical analysis. Of course, it's all in the underlying assumptions and choice of data set. Interestingly, their portfolios only include bond funds if the time horizon is 2-3 years.
"Skating to where the puck was".

The problem with Markowitz Optimizers, as Bill Bernstein points out, is that you drop in historical correlations and returns and you get a portfolio designed to outperform in the past. The future optimal portfolio will be different.
That is my take as well. Several of Rodc's old posts, summarizing his own analyses with lots of efficient frontier plots, make that point convincingly.

It's not clear whether the "engineers" know that, or they're just making a good pitch. Either way, the back-tested portfolios seem odd. Do they propose changing asset allocations (moving in and then back out of some asset types) as one transitions from, say, a 10-year horizon to a 5-year horizon? All based on back-testing?

I'll stick with Bernstein's advice.

"Discipline matters more than allocation.”
"Discipline matters more than allocation.” ─William Bernstein

User avatar
TD2626
Posts: 637
Joined: Thu Mar 16, 2017 3:40 pm

Re: Engineered Portfolio

Post by TD2626 » Tue Mar 27, 2018 3:21 pm

It's probably what happens if you do excessive backtesting and go with it. Sometimes the outputs of backtests are a quirk of history and not all that likely to work well in the future. It reminds me of this viewtopic.php?t=5410

Excessively unconventional and highly risky, in my opinion.

User avatar
patrick013
Posts: 2427
Joined: Mon Jul 13, 2015 7:49 pm

Re: Engineered Portfolio

Post by patrick013 » Tue Mar 27, 2018 4:11 pm

Image

Their work is kinda impressive even tho they are averaging around
alot of numbers and if they are really using an optimization config.
The above is alot more basic.

Their upper bounds are 20% for each asset class.
The above can go from 10-90% for each asset class.
The above uses beta instead of four risk metrics and
using beta only should have 20% less of most market
volatility/risk. The above also uses nominal expected return
instead of four return metrics leading to real returns.

Funds chosen above are based on some economic concerns.
Large cap underperformance in rising rate scenario. Short
term bonds being strategic for later move to long term bond
fund. So it looks very aggressive for the next 5-10 years or so
but not as refined as the engineered portfolios to get a blend of
8 metrics.
age in bonds, buy-and-hold, 10 year business cycle

pascalwager
Posts: 1272
Joined: Mon Oct 31, 2011 8:36 pm

Re: Engineered Portfolio

Post by pascalwager » Wed Mar 28, 2018 2:52 am

The engineers' "savings" portfolio was interesting. Maybe not sufficiently liquid under all circumstances to be called "cash", but intended for up to one-year time horizon:

US mid-cap value stocks 10%
Gold 10%
Total int'l bonds 10%
US TIPS 10%
US T-bills 60%

pascalwager
Posts: 1272
Joined: Mon Oct 31, 2011 8:36 pm

Re: Engineered Portfolio

Post by pascalwager » Thu Mar 29, 2018 12:57 am

Valuethinker wrote:
Tue Mar 27, 2018 10:02 am
dowse wrote:
Tue Mar 27, 2018 9:38 am
I'm curious what folks here think about the blog "Engineered Portfolio", https://engineeredportfolio.com/ mentioned in another thread. I'm intrigued by the portfolios they came up with based on extensive numerical analysis. Of course, it's all in the underlying assumptions and choice of data set. Interestingly, their portfolios only include bond funds if the time horizon is 2-3 years.
"Skating to where the puck was".

The problem with Markowitz Optimizers, as Bill Bernstein points out, is that you drop in historical correlations and returns and you get a portfolio designed to outperform in the past. The future optimal portfolio will be different.

South Africa & Sweden? Really. Huge overweightings. If you want Gold & Precious Metal mining stocks, Vanguard has a fund for that. Sweden? It's a handful of multinationals plus some banks.

Like many here, one of the effects of spending time on this Forum is to tend to increase your bond weightings. Historically I have held almost no bonds. But as I age, and as it becomes clear just how bad the 2008-09 Crash was, the extra risk of being nearly 100% equities no longer seems worth it. Talking to the many learned people here has suggested to me that one should hold more bonds.
They have no use for the "miners", only actual gold. So the Vanguard fund would not apply.

As for Sweden and S. Africa, they consider Australia to be, perhaps, the world's best stock market. But it's too large a market for their purposes--they had concerns about a higher correlation with EAFE because of the market size and low correlations are paramount in their choice of assets.

For example, S. Africa has correlations of only 0.29 and 0.17 respectively with US mid-cap value and US consumer staples--the main US stock holdings.

HanMan
Posts: 6
Joined: Tue Sep 13, 2016 4:14 pm

Re: Engineered Portfolio

Post by HanMan » Thu Mar 29, 2018 10:03 pm

Thanks for starting this topic dowse! My name is Stephen Hanly, one of the authors of the engineeredportfolio blog (here's the about page https://engineeredportfolio.com/about/). Chris and I are both members of the Bogleheads community so we always appreciate when others here find something interesting from our posts! There are a couple things I want to address/clarify.

Unconventional Portfolios
I'm the first to admit that our portfolios are unconventional and frankly a bit shocking. But they are the result of detailed historical analysis where we removed the lense of trying to meet the norm or predict the future. We are very frank (I believe and hope) that our portfolios are solely the result of the historical analysis. We make absolutely no guarantees about the future and I try to take every opportunity I can to say the future is unknown. But if we can't predict the future, where else are we to look than the past? Much of our analysis came about after coming across the Bogleheads backtesting spreadsheet viewtopic.php?f=10&t=2520&start=700#p2753812 and spans a diverse and long time range where there were periods of economic expansion, recessions, high inflation, and deflation. Our goal was to find the portfolios that worked best with the data we have, which is the past.

Motivation for our Blog
We're not trying to sell an investment service, our blog is aimed to be a portal where Chris and I can just share some information we've learned that may get others thinking about their own investments.
iceport wrote:
Tue Mar 27, 2018 12:29 pm
"Discipline matters more than allocation.”
I wholeheartedly agree with Bernstein here (and on many of his points). For me personally, I couldn't do the more traditionally recommended portfolios due to either arrogance or curiosity or both. But I wanted to have discipline in my investing strategy so I'd stick with it - so this meant I needed to develop my own. I'm the type of person that can't blindly follow someone else's advice - I need to develop my own conclusions in order to have conviction in them. Chris and I are not selling anything. All I really want to do is to encourage others to maybe develop their own portfolios that can have a bit of their own personality in it if it may help them stick with it. Readers of our blogs are likely people like myself who want something a bit different, to other friends and family members I talk to who give me blank stares when it comes to investing - I very strongly recommend they buy a Vanguard target date fund.

Bonds
I just want to correct the original post in this thread
dowse wrote:
Tue Mar 27, 2018 9:38 am
Interestingly, their portfolios only include bond funds if the time horizon is 2-3 years.
That's not true - our 5 year portfolio has 30% bonds and 10% gold. It's only when someone is able to truly put the blinders on for a full 10 years where bonds may not be necessary, but our 10-year portfolio still has 10% gold. Going into the analysis we performed I assumed bonds were only for really safe portfolios - after that analysis, I am a big fan of bonds. But I do like the bond class that is least correlated to stocks which is long-term treasuries. Our safer, more bond-heavy portfolios have no long-term treasuries however, it's only helpful for stock-heavy portfolios. Here's a more detailed analysis of bond returns and various asset classes: https://engineeredportfolio.com/2017/06 ... rformance/

Gold vs Gold Miners
Valuethinker wrote:
Tue Mar 27, 2018 10:02 am
If you want Gold & Precious Metal mining stocks, Vanguard has a fund for that.
Gold and precious mining stocks should be avoided like the plague. I hated the idea of gold before analyzing the historical data, I liked Buffet's quotes on gold (here are some: http://commodityhq.com/education/top-se ... investing/). Gold is a stupid investment... on its own. Gold has a very unique combination of being both a "safe haven asset" and a hedge against inflation. So it helps when stocks are doing really bad (and people are piling money into "safer" assets) and it helps when bonds are doing really bad (due to inflation). Gold mining stocks on the other hand are correlated to the price of gold (which is volatile) and are also businesses that can lose a lot of money when stocks in general are doing poorly. So it provides the bad part of gold which is the volatility without providing the good part which is its negative correlation to stocks and bonds. This is the subject of a future blog but I'll repeat myself... gold mining stocks should be avoided like the plague.

For more see the comparison of stocks, to gold miners, to gold. In 2008 stocks lost 37%, gold was up 5%, meanwhile gold miners lost 56%. https://www.portfoliovisualizer.com/bac ... etals2=100

Sweden and South Africa
Yeah these are unique. My interest in individual country performance came about after looking at global stocks in general (https://engineeredportfolio.com/2017/05 ... -equities/). I found that the pacific stocks did quite a bit differently than europe and wondered if we are shortchanging ourselves to look as everything else besides the US so coarsely. In the US we look at value, small, sectors, factors etc. while internationally it seems to be a bit more of a catch-all. After coming across the credit suisse report I dug into individual country stock market returns (https://engineeredportfolio.com/2017/07 ... ck-market/).

South Africa and Sweden both have very long stock market history and track record. These countries are unique and not that well represented in cap weighted international funds. But not only have they offered historically better returns, they have some interesting advantages. Sweden is a bit out of the way compared to the rest of Europe so they don't get involved in disputes as much, they also have an abundance of natural resources. South Africa has even more natural resources and provides some geographic diversification/exposure to Africa in general. We had to explicitely limit our portfolios to "only" 10% exposures to Sweden and South Africa. Based on the 40 years of historical data, the optimization routine we had wanted 20% allocations for a number of portfolios - we had to step in a bit then (same goes with gold BTW).

Correlations
pascalwager wrote:
Thu Mar 29, 2018 12:57 am
low correlations are paramount in their choice of assets
Thank you pascalwager, yes low correlations meant everything to us (our optimization techniques). This is largely what led us to such unconventional portfolios.

Do your own research!
Again, my aim in sharing what we have on our blog (and will continue to) is to get people thinking. I share the data (and code in Quantopian) when I can to help encourage others to draw their own conclusions. Our portfolios are probably a little too out there for many, but for some it may open your eyes to interesting asset classes that historically have offered unique advantages to more traditional portfolio constructions.

sandramjet
Posts: 231
Joined: Thu Oct 23, 2014 11:28 pm

Re: Engineered Portfolio

Post by sandramjet » Thu Mar 29, 2018 11:30 pm

Maybe I missed it, but why did you have to
HanMan wrote:
Thu Mar 29, 2018 10:03 pm
explicitly limit our portfolios to "only" 10% exposures to Sweden and South Africa. Based on the 40 years of historical data, the optimization routine we had wanted 20% allocations for a number of portfolios - we had to step in a bit then (same goes with gold BTW).
It seemed like you say that you are trying to construct portfolios that are based on the data ... but here you seem to be saying you don't believe it enough to really follow the data? Why is that? I can see where it may be interesting to see what the data would "truly" say, from a historical standpoint, but this seems to go against that. What was your rationale?

MoneyMarathon
Posts: 161
Joined: Sun Sep 30, 2012 3:38 am

Re: Engineered Portfolio

Post by MoneyMarathon » Fri Mar 30, 2018 1:10 am

When I see EWD and EZA, what I see is the concentration of the top holdings -- and the fees.

56% of the South Africa fund is in the top ten holdings.

54% of the Sweden fund is in the top ten holdings.

If trying to understand this, I would want to know why these 20 companies offer most of the diversification or return I'm buying. Then I'm wondering whether I could get that exposure more efficiently -- if there's a theory behind it. I don't know -- financials stocks in countries that have their own sovereign currency outside of the Euro and dollar? Or companies that trade on exchanges in those currencies? (Or whatever it is.)

Valuethinker
Posts: 36681
Joined: Fri May 11, 2007 11:07 am

Re: Engineered Portfolio

Post by Valuethinker » Fri Mar 30, 2018 7:16 am

HanMan wrote:
Thu Mar 29, 2018 10:03 pm


Gold vs Gold Miners
Valuethinker wrote:
Tue Mar 27, 2018 10:02 am
If you want Gold & Precious Metal mining stocks, Vanguard has a fund for that.
Gold and precious mining stocks should be avoided like the plague. I hated the idea of gold before analyzing the historical data, I liked Buffet's quotes on gold (here are some: http://commodityhq.com/education/top-se ... investing/). Gold is a stupid investment... on its own. Gold has a very unique combination of being both a "safe haven asset" and a hedge against inflation. So it helps when stocks are doing really bad (and people are piling money into "safer" assets) and it helps when bonds are doing really bad (due to inflation). Gold mining stocks on the other hand are correlated to the price of gold (which is volatile) and are also businesses that can lose a lot of money when stocks in general are doing poorly. So it provides the bad part of gold which is the volatility without providing the good part which is its negative correlation to stocks and bonds. This is the subject of a future blog but I'll repeat myself... gold mining stocks should be avoided like the plague.

For more see the comparison of stocks, to gold miners, to gold. In 2008 stocks lost 37%, gold was up 5%, meanwhile gold miners lost 56%. https://www.portfoliovisualizer.com/bac ... etals2=100
This tells me that gold mining cos are highly geared plays on the gold price. The sector has a number of issues:

- managements tend to be piratical - the nature of gold mining (highly polluting) and where you mine it (corruption etc.)
- historically the companies "sold forward" production (gold hedging) thus limiting the upside of price moves for investors (I think they are less likely to do this, now)
- Barrick in particular, a big chunk of the sector, has been managed very poorly (historically)
- cost of producing gold is rising steadily due to reserve exhaustion (South Africa and other estabilished markets) and rising environmental restrictions/ greater royalties (newer territories)
Sweden and South Africa
The general issue here is that past lack of correlation may not reflect future.
South Africa and Sweden both have very long stock market history and track record. These countries are unique and not that well represented in cap weighted international funds. But not only have they offered historically better returns, they have some interesting advantages. Sweden is a bit out of the way compared to the rest of Europe so they don't get involved in disputes as much, they also have an abundance of natural resources.
"not involved in disputes as much" - you wouldn't have said that in the 1600s ;-).

Sweden is not a member of NATO but is, I believe, reintroducing military conscription because of what is going on across the Baltic.

Natural resources. Trees yes (but so do Canada and Finland). Hydropower yes (also Brasil, Canada, Finland, Norway). Metals mostly worked out, I believe. I wouldn't call Sweden a natural resources play.

One issue with the Swedish index is the importance of holding companies - the Investor AB group in particular, which owns chunks of many of the major Swedish companies. The index is even more concentrated than it looks. You've got quite a big bet on one pharma there (Astra Zeneca).

The other is Financial Services and Swedish financial cos are not immune to international forces. They've also got their own crazy little housing bubble going on (negative interest rates driving soaring Stockholm housing prices). And in the early 1990s they had to restructure their whole banking system-- Sweden was uncorrelated with the rest of markets, in the worst possible way.

The basic thing though is Sweden, but also Finland and Norway offer these characteristics. However of course Finland had the rise and fall of Nokia. And Norway has Norsk Hydro (fertilizer, aluminium) as huge stocks. But if you are seeking diversification into Scandinavia, you probably want all 4 countries (ie including Denmark, which gives you the world's largest diabetes treatment company, the world's largest (?) standalone wind power company, the world's largest standalone offshore wind developer, etc.
South Africa has even more natural resources and provides some geographic diversification/exposure to Africa in general.
Again financial services. And that's not immune to world market conditions. South African, like Australian, stock market is really a sort of geared play on world economic growth (natural resources demand) + a big financial services sector. (Canada has a similar structure, but is somewhat less correlated).
We had to explicitely limit our portfolios to "only" 10% exposures to Sweden and South Africa. Based on the 40 years of historical data, the optimization routine we had wanted 20% allocations for a number of portfolios - we had to step in a bit then (same goes with gold BTW).
Aye, therein lies the rub. A quant strategy but if the quant strategy "feels" wrong, you adjust it?

In addition, 40 years of data is not that long. For SA & Sweden you've got 100+ years of data, so you might as well use it? However it does include the periods in which the 2 countries went from underdeveloped (incredible as it may seem, around 1900 Sweden was one of the poorer countries in Europe) to developed (SA only for a percentage of the population, but the stock market has tended to reflect that).

SA is always seen as the country which manages to have both the problems of an advanced economy (wage rigidity, loss of manufacturing etc.) AND the problems of an underdeveloped country.
Correlations
pascalwager wrote:
Thu Mar 29, 2018 12:57 am
low correlations are paramount in their choice of assets
Thank you pascalwager, yes low correlations meant everything to us (our optimization techniques). This is largely what led us to such unconventional portfolios.

Do your own research!
Again, my aim in sharing what we have on our blog (and will continue to) is to get people thinking. I share the data (and code in Quantopian) when I can to help encourage others to draw their own conclusions. Our portfolios are probably a little too out there for many, but for some it may open your eyes to interesting asset classes that historically have offered unique advantages to more traditional portfolio constructions.
[/quote]

Fair enough. It's a desktop study?

The problem is the "skating to where the puck was" which is true of Markowitz portfolio analyzers. William Bernstein tackled that in "the Intelligent Asset Allocator" (?title) the book which eventually led me to this Forum.

So that would be my (our?) response.

It's nice to know what historically worked, but big bets on what worked historically which are far off the "market" portfolio give you a lot of risk of tracking error.

If you look at the Swedish and South African stock markets, there's little reason to think low correlations will persist. And you have big concentration risk. Any set of impressionistic facts + perceptions about those countries & economies can be used to make the opposite argument.

I take the point gold v. PM & M stocks. You seem to taking the William Bernstein line "the most patient asset" re the former, and that's the argument for them. That there is the cost of expected real returns which are negative in the long run BUT you do have a low correlation/ negative correlation with stock markets.

If I read Campbell Harvey (the best academic pieces on gold that I know about) then that may not be so, may not be as true as it was. But it's where the bull case comes from?

Valuethinker
Posts: 36681
Joined: Fri May 11, 2007 11:07 am

Re: Engineered Portfolio

Post by Valuethinker » Fri Mar 30, 2018 7:56 am

MoneyMarathon wrote:
Fri Mar 30, 2018 1:10 am
When I see EWD and EZA, what I see is the concentration of the top holdings -- and the fees.

56% of the South Africa fund is in the top ten holdings.

54% of the Sweden fund is in the top ten holdings.

If trying to understand this, I would want to know why these 20 companies offer most of the diversification or return I'm buying. Then I'm wondering whether I could get that exposure more efficiently -- if there's a theory behind it. I don't know -- financials stocks in countries that have their own sovereign currency outside of the Euro and dollar? Or companies that trade on exchanges in those currencies? (Or whatever it is.)
Quick eyeball

- more Industrials than I had imagined in Swedish index. Remember though that Investor AB owns stakes in most of the major Swedish companies (so there's double concentration there). The industrials explains partly a lower correlation because industrials are not as large a part of most countries' indices. It also means greater correlation with world economy - Atlas Copco for example would be a play on China (construction machinery)?

- South Africa it really is all financials. Good news, that probably means you are buying exposure to all of Sub Saharan Africa. For demographic reasons, despite the horrendous problems, that is something I am bullish about. Bad news, you are vulnerable to their own various ructions re politics driving economics and exchange rates

dowse
Posts: 349
Joined: Tue Sep 10, 2013 3:10 pm

Re: Engineered Portfolio

Post by dowse » Fri Mar 30, 2018 9:28 am

HanMan wrote:
Thu Mar 29, 2018 10:03 pm

Bonds
I just want to correct the original post in this thread
dowse wrote:
Tue Mar 27, 2018 9:38 am
Interestingly, their portfolios only include bond funds if the time horizon is 2-3 years.
That's not true - our 5 year portfolio has 30% bonds and 10% gold. It's only when someone is able to truly put the blinders on for a full 10 years where bonds may not be necessary, but our 10-year portfolio still has 10% gold. Going into the analysis we performed I assumed bonds were only for really safe portfolios - after that analysis, I am a big fan of bonds. But I do like the bond class that is least correlated to stocks which is long-term treasuries. Our safer, more bond-heavy portfolios have no long-term treasuries however, it's only helpful for stock-heavy portfolios. Here's a more detailed analysis of bond returns and various asset classes: https://engineeredportfolio.com/2017/06 ... rformance/
Hi Stephen. Thanks for correcting me. I misread the chart at first (kind of missed the whole bottom section). Sorry for that. Retired engineer here, so I am analytical to the core. Your findings do lead to portfolios that seem to run against conventional wisdom here on this forum, so I'm happy to start a discussion on it. We are already getting an interesting exchange of ideas. Keep up the good work!

Post Reply