Backtesting some robo-advisers asset allocations

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

jbolden1517 wrote:Well LadyGeek there you go. SIP with those adjustments now outperformed 3-fund by 204 basis points on average when backtested using your chosen methodology.
Er, I am not sure to understand what you mean. The 'SIP EMB' in my latest experiment displayed a CAGR of 10.3% in the time period being studied (1985-2016), while the 70/30 3-Funds portfolio displayed a CAGR of 9.3%.

When running my data table on all periods of ten years, then the SIP EMB advantage grows a bit (average CAGR 9.4% compared to 8.1% for the 3-funds).
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

siamond wrote:
jbolden1517 wrote:That's going to be too low a correlation. What about 50% EM stock, 50% INTL bonds.
[...]
I do have data for the last 20 years. The green box JPM-EMBI is the index that Schwab's fund (VanEck) is tracking.
http://www.lazardnet.com/us/docs/sp0/20 ... nvestmentF
Thank you, great pointer, much better than Barclays or Citi. JPM-EMBI matches reasonably well the (very few!) years of the corresponding Vanguard fund (VGOVX), and this index does provide pretty good EM Bonds history, starting from 1994.

As to the previous years, trying to fill the gap 1985 to 1993, the best I found is 1/3rd EM, 1/3rd HY, 1/3rd Int'l Bonds.
That's interesting. That makes sense. Sort of captures all the aspects of EM bonds, nicely. Good to see that worked. Definitely better than Intl bonds.
siamond wrote: Yes, VERY clunky, but definitely a better match for std-deviation and correlation with JPM EMBI for known years than other things I tried. => side note: this made me understand why SIP (and other folks) are eager to use EM Bonds, as this asset class has an interestingly low correlation (e.g. less than 0.5) with pretty much everything else. I learn something every day with this forum!
That is interesting. I would have figured it would be a higher correlation given the liquidity requirements. I might want to pick up some of that VanEck fund for myself.
siamond wrote: So... I cobbled together an extended JPM EMBI data series starting from 1985. And assigned the 7% EM Bonds to it. And lo and behold, the SIP results are indeed improved, and now it's roughly on par with my Tilt2 portfolio. And I guess that if I were to address the more minor issues (e.g. better MBS mapping, etc), this would improve a tad further. Then the big thing remains 'fundamental' vs. 'classic' value, but let's stop short of re-opening this discussion (please!). Note that one could of course also use 'fundamental' funds in a regular tilted asset allocation (some Bogleheads definitely do that).
Agreed. There is a discussion about market weight vs. strong value tilt (basically avoid crappy stocks) and there is a discussion about how best to implement strong value tilts. The discussion about how best to implement strong value tilts is rather complex and likely to get more complex as value tilting methodologies get more sophisticated. I wouldn't want to trust backtesting here. My personal opinion is for passive use a variety of passive methods or just pay the 100 basis points and go with a quality value house mutual fund. I like Schwabs 3 factor model (sales, cash flow, dividends) as being a nice mixture which is why they are the largest chunk of my portfolio. Goldman Sachs multi-factor funds are really interesting for forward going portfolios. OTOH Goldman Sachs has a well deserved reputation for screwing over their clients, so while I hold some I don't recommend them for value investors who might not be paying attention nor for a core.
siamond wrote: I also tried to replace the 9% money market 'cash' by TBM, which makes the whole thing even better - just saying! This is the part of the SIP portfolio I really don't understand.
I can explain and it presents a good example of where Vanguard is outstanding relative to the competition. Vanguard may be incompetent but they never have mixed motives. Bogleheads tends to focus on mutual fund fees because Vanguard literature focuses so much on those. But those have come down a lot. The only people making huge profits running mutual funds are those companies driving their funds into the ground with 2.2% ERs and declining assets. No one is really making much money anymore on new funds as ETFs eat the industry. The deep discounters are now mainstream so there is no money to be made on transactions. Fidelity and Schwab now have to offer $5 stock trades and live on 25 basis point 12b-1 fees and even those margins are under increasing pressure. Being a brokerage is a bad business to be in right now which is why so many of them are diversifying into value adds, being sold off to get themselves value adds or just merging into successful brokerages. For the big houses the profits in the brokerage industry are much larger on the sellside than the buyside. And even for them the profits on the buyside are mostly from things like mid-business lending. Everyone talks about the 1% account management fees but the big brokerages end up having to provide so many services and have such a high cost of client acquisition that they don't make huge profits there. (I know this is shocking for Bogleheads but stay with me on this).

The way the big houses really make money from their buyside clients is using them as a way to offload securities from the sellside clients and pick up the 6% commission on a success for an issuance. Think about how much 30 basis points annually on the entire USA corporate bond market is. And then remember that many of these big houses also make a sizable chunk (though a much lower percentage) of the global bond market as well for countries without a sophisticated sellside.

Charles Schwab has gotten to the size that they can start thinking of competing on the sellside and is restructuring themselves slowly to be more like the big brokerage houses. Just as Charles Schwab was an early innovator in discount buyside brokerage they hope to be a discounter (but not deep discounter) for sellside brokering. To do that though they need to be easily able to deploy hundreds of millions if not billions of dollars into assets of their choosing for their clients on a regular basis. Charles Schwab built their business on DIY investors. They don't have pension funds under management, they aren't managing endowments, they don't have wealthy investors.

SIP and the ETFs are I believe a strategic product. They do a few core things to advance this goal.
1) It brings complex investment management in house. Schwab has to run funds. That means expensive software and the in house expertise to write it, especially big data and quantitative expertise. If forces them to make market in a high volume security (their ETFs) and gain the experience in doing this kind of billion dollar selling. They are practicing with the ETFs learning the skills they will need to be a big sellside house when mistakes are cheap not billions of dollars in loses.
2) It gets Schwabs desks doing heavy volume. Charles Schwab is learning how to handle billions in daily transactions without say calling Merrill and asking their desks to handle it.
3) It gets Schwab having large quantities of lendable funds so they can fund their in house desk.
4) It brings in a client base who is not DIY and as their assets grow might become good candidates for their full service financial advice (who of course is going to advise them to buy securities Schwab's trading desk is having trouble offloading).

The cash position isn't invested in bonds because Schwab wants it invested in their trading desk. They need their desk to be able to comfortably hold a $3b bond issue for 10 days. Right now they would have to borrow. But if they get say $500b in SIP of that gets them up to $30b in floating funds for their desk.

SIP is unique among the robo-advisors in that it doesn't charge a fee. The Schwab ETFs are cheap there is no money to be made there directly. Their own trading desk can make profits from rebalancing but that's a small percentage of assets under management. The cash on the other hand.

They still don't have a big derivatives business which is the last component for a sellside house. But they have a neat subsidiary: http://www.optionsxpress.com/ . What do you think all those 20 year olds whom Schwab is teaching to short options will want to buy in 20 years? It ain't gonna be TSM. At least a good fraction of them are going to want to move up to more complex and sophisticated products and Schwab can move up to complex and more sophisticated products right along with them. And now they have the cash to be able to write those sorts of complex derivatives.

So IMHO that's the reason for the cash position. A good chunk of the reason SIP exists at all is for that cash position. SIP is a great product. But it is a great product because its maker has another agenda.
siamond wrote: Bottomline: SIP is basically a somewhat complicated value-y asset allocation, with a few interesting subtleties in terms of diversification and play on correlations. Although one can construct a simpler tilted portfolio with regular Bogleheads principles that would have performed more or less the same. As Ladygeek said, the Schwab robo-adviser essentially acted as an adviser with its own (possibly good) ideas on how to construct an asset allocation. While the 'robo' automation makes the relative complexity of the portfolio much more bearable.
Agree with all.
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

siamond wrote:
jbolden1517 wrote:Well LadyGeek there you go. SIP with those adjustments now outperformed 3-fund by 204 basis points on average when backtested using your chosen methodology.
Er, I am not sure to understand what you mean. The 'SIP EMB' in my latest experiment displayed a CAGR of 10.3% in the time period being studied (1985-2016), while the 70/30 3-Funds portfolio displayed a CAGR of 9.3%.

When running my data table on all periods of ten years, then the SIP EMB advantage grows a bit (average CAGR 9.4% compared to 8.1% for the 3-funds).
Oh sorry I did that calculation of the SIP TBM which of course is about 100 basis points better. I stand corrected.

I'm still not seeing how that's possible though even when I drop to the right line on the graph. If A outperforms B by 1.4x in a 20 year period the spread has to be 1.7%.
rkhusky
Posts: 10909
Joined: Thu Aug 18, 2011 8:09 pm

Re: Backtesting some robo-advisers asset allocations

Post by rkhusky »

jbolden1517 wrote: I think Tilt2 demonstrates that the more you move away from large cap growth and into value the better.
Vanguard's Growth Index (large growth) has returned 9.44% over the last 25 years. Value Index (large value) has returned 9.50% over the last 25 years. Not much difference to my untrained eye.

Interestingly, Morningstar's lists of top 25 holdings shows Apple (P/E=18), Walt Disney (P/E=19), and Gilead Sciences (P/E=8) in Growth Index and Microsoft (P/E=32), Exxon Mobil (P/E=34), Chevron (P/E=72), and Abbot Laboratories (P/E=55) in the Value Index.
Last edited by rkhusky on Thu Aug 03, 2017 10:59 am, edited 1 time in total.
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

rkhusky wrote:
jbolden1517 wrote: I think Tilt2 demonstrates that the more you move away from large cap growth and into value the better.
Vanguard's Growth Index (large growth) has returned 9.44% over the last 25 years. Value Index (large value) has returned 9.50% over the last 25 years. Not much difference to my untrained eye.
Notice how much lower the base 70/30 portfolio is. That's what 1/2 vanguard growth, 1/2 vanguard value does to you. It going to be much worse if you go 100% growth and then cut with 30% bonds. These value portfolios are doing a percent better with less volatility before the cut. After the cut well ... siamond what happens to a 70% Vanguard Growth Index / 30% total bond portfolio for the test period?
rkhusky
Posts: 10909
Joined: Thu Aug 18, 2011 8:09 pm

Re: Backtesting some robo-advisers asset allocations

Post by rkhusky »

jbolden1517 wrote: Notice how much lower the base 70/30 portfolio is. That's what 1/2 vanguard growth, 1/2 vanguard value does to you.
How sure are you that it is large growth causing the difference? There are tilts to emerging markets, small value, small international, and REIT's thrown into the mix.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

jbolden1517 wrote:
rkhusky wrote:
jbolden1517 wrote: I think Tilt2 demonstrates that the more you move away from large cap growth and into value the better.
Vanguard's Growth Index (large growth) has returned 9.44% over the last 25 years. Value Index (large value) has returned 9.50% over the last 25 years. Not much difference to my untrained eye.
Yes, Large Value didn't display any premium in a long, long time. Tilt2 worked better because this is tilting towards small(er) value, which is an entirely different story. Note that REITs largely intersects with small/mid value.
rkhusky wrote:
jbolden1517 wrote: I think Tilt2 demonstrates that the more you move away from large cap growth and into Notice how much lower the base 70/30 portfolio is. That's what 1/2 vanguard growth, 1/2 vanguard value does to you. It going to be much worse if you go 100% growth and then cut with 30% bonds. These value portfolios are doing a percent better with less volatility before the cut. After the cut well ... siamond what happens to a 70% Vanguard Growth Index / 30% total bond portfolio for the test period?
Doesn't make much of a difference. 70% large growth was actually (surprisingly) a smidge better. But again, large value is just not effective AT ALL, doesn't hurt, doesn't help. And the cap-weighing math totally waters down small/mid value when combining with large-caps. Only a specialized size+value tilt worked well in the past.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

jbolden1517 wrote:I'm still not seeing how that's possible though even when I drop to the right line on the graph. If A outperforms B by 1.4x in a 20 year period the spread has to be 1.7%.
1986 to 2016 is 30+ years... Yes, time flies... :wink:
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

siamond wrote:
jbolden1517 wrote:I'm still not seeing how that's possible though even when I drop to the right line on the graph. If A outperforms B by 1.4x in a 20 year period the spread has to be 1.7%.
1986 to 2016 is 30+ years... Yes, time flies... :wink:
Yep that would do it. Now I'm getting 112 basis points. :idea:
User avatar
LadyGeek
Site Admin
Posts: 68701
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Backtesting some robo-advisers asset allocations

Post by LadyGeek »

FYI - The discussion was starting to stray in several different directions. I moved posts discussing asset allocation and topics not related to backtesting into the first thread: Re: Backtesting some robo-advisers asset allocations
Wiki To some, the glass is half full. To others, the glass is half empty. To an engineer, it's twice the size it needs to be.
User avatar
Tyler9000
Posts: 577
Joined: Fri Aug 21, 2015 11:57 am

Re: Backtesting some robo-advisers asset allocations

Post by Tyler9000 »

This is a fascinating topic. I'm still getting caught up to the various threads, but it seems like a very worthwhile effort. Special thanks to Siamond for the legwork and to LadyGeek for the wiki info.
User avatar
LadyGeek
Site Admin
Posts: 68701
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Backtesting some robo-advisers asset allocations

Post by LadyGeek »

You might find this article interesting: Robo-Adviser Asset Allocation and Securities, which I used to address some concerns on wording in the wiki article: Re: Backtesting some robo-advisers asset allocations
Wiki To some, the glass is half full. To others, the glass is half empty. To an engineer, it's twice the size it needs to be.
User avatar
Tyler9000
Posts: 577
Joined: Fri Aug 21, 2015 11:57 am

Re: Backtesting some robo-advisers asset allocations

Post by Tyler9000 »

That's very helpful. Thanks!
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

jbolden1517 wrote:[...] SIP is unique among the robo-advisors in that it doesn't charge a fee. The Schwab ETFs are cheap there is no money to be made there directly. Their own trading desk can make profits from rebalancing but that's a small percentage of assets under management. The cash on the other hand.

They still don't have a big derivatives business which is the last component for a sellside house. But they have a neat subsidiary: http://www.optionsxpress.com/ . What do you think all those 20 year olds whom Schwab is teaching to short options will want to buy in 20 years? It ain't gonna be TSM. At least a good fraction of them are going to want to move up to more complex and sophisticated products and Schwab can move up to complex and more sophisticated products right along with them. And now they have the cash to be able to write those sorts of complex derivatives.

So IMHO that's the reason for the cash position. A good chunk of the reason SIP exists at all is for that cash position. SIP is a great product. But it is a great product because its maker has another agenda.
That is a fascinating explanation. Thank you so much for sharing your expertise. Hm, I would then observe that this 9% cash appears to be beneficial for Schwab, but quite a drag for the SIP customers. If I were looking for a robo-adviser, I am not too sure I would be happy about that. Maybe something to capture in the corresponding review.

This also means that we have to be careful to compare apples to apples, and properly factor in cash positions vs. robo-adviser extra layer of fees. My next step is Betterment -- I already drafted it in my spreadsheet, it seems to be quite a reasonable one, will refine tomorrow and share the results.

Any more views on Schwab SIP, before we move on?
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

One more thought I'd like to share. This type of backtesting obviously suffers from various limitations, the past isn't the future, yadiyah, yadiyah.

One special limitation in this case is that we use the Schwab SIP AA for a 70/30 case, and used it as a fixed allocation to backtest over some 30+ years. The detailed AA starts by the split between domestic (US) and international, which we observed to be 50/50. Well that happens to be roughly the respective market weights for now. Now check the following chart, and look at the 90s. How things change... I strongly suspect that if the conditions of the 90s would come back, the Schwab SIP AA split would NO LONGER be 50/50. Hence backtesting with 50/50 all the way was, er, a tad dubious.

Image

If I were to make my life difficult, I could probably run a more precise backtest. But I won't. I do remember a key study from Dr. Bernstein showing that keeping a fixed split domestic/international actually improved total returns a tad (not a lot) compared to dynamically following market weights (I actually think I reproduced this study by then, but didn't keep a proper archive). So... I don't think it's worth it in this case to try to be overly precise. After all, we're mostly trying to sanity check those 'robo-advised' AAs, not much more.
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

siamond wrote:
jbolden1517 wrote:[...] SIP is unique among the robo-advisors in that it doesn't charge a fee. The Schwab ETFs are cheap there is no money to be made there directly. Their own trading desk can make profits from rebalancing but that's a small percentage of assets under management. The cash on the other hand.

They still don't have a big derivatives business which is the last component for a sellside house. But they have a neat subsidiary: http://www.optionsxpress.com/ . What do you think all those 20 year olds whom Schwab is teaching to short options will want to buy in 20 years? It ain't gonna be TSM. At least a good fraction of them are going to want to move up to more complex and sophisticated products and Schwab can move up to complex and more sophisticated products right along with them. And now they have the cash to be able to write those sorts of complex derivatives.

So IMHO that's the reason for the cash position. A good chunk of the reason SIP exists at all is for that cash position. SIP is a great product. But it is a great product because its maker has another agenda.
That is a fascinating explanation. Thank you so much for sharing your expertise. Hm, I would then observe that this 9% cash appears to be beneficial for Schwab, but quite a drag for the SIP customers. If I were looking for a robo-adviser, I am not too sure I would be happy about that. Maybe something to capture in the corresponding review. My next step is Betterment -- I already drafted it in my spreadsheet, it seems to be quite a reasonable one, will refine tomorrow and share the results. This also means that we have to be careful to compare apples to apples, and properly factor in cash positions vs. robo-adviser extra layer of fees.
I'm going to give a really long answer here. I really don't know of a better alternative. 9% cash vs. short bonds is probably about 18 basis points at worst. Which is less than the fees of any of the robos.

I've had a lot of verbal conversations about this. I have a close friend who often gets the "what should I invest in" question from his friends and family. I do as well. So for 2 decades we've bounced ideas around. For a long time our simple portfolio (outside 401k, 40sb...) was Oakmark Global + (possibly) short term bonds. The robos came out and at the time Betterment didn't have the value tilt so 50% Betterment / 50% Oakmark Global. Then Schwab came out with SIP. The portfolio was clearly better. And from an expensive perspective the cash vs. short bonds (say 18 basis points max) did less damage than the 35 basis points (at the time) for Betterment and the 100+ for Oakmark global. Also remember I weigh the smart beta aspects of the fundamentals more positively than you are weighing them.

So I think Schwab has conflicts of interest and at the same time despite this I think SIP is so far the best of the mainstream bunch, and that's the reason I reviewed them first, I have a clear favorite. At the same time I will not deny that brokerages that develop sell side interests have a long history of not successfully negotiating the complexity of balancing their sell side interests and their buy side interest. I can't think of one that hasn't had a major ethical scandal in the last 50 years where they acted against their buyside client's best interest. Schwab isn't off to a great start with that cash position. My personal opinion is that the law is asking of brokerages something that is simply impossible and every brokerage should have to be one and only one of:
a) Market operations (clearing house, flow trader, margin bank, money market fund, operating exchange, trading from their own desk, derivative writing...)
b) Buy side (mutual funds, brokerage, ETF operators, CEFs, financial advisor, annuity / insurance issuer...)
c) Sell side

So I do see serious conflicts of interest. I do give big bonus points to companies like Fidelity, Vanguard, Ally that are exclusively on the buyside. There is a safety in that. I told my daughter, the recommendation I care most about (I assume Bogleheads will disagree with this), "Vanguard because of an idealogical obsession with low fees has a scary lack of depth and competence for a house of their side. As a result they may accidentally screw you over. But they will never deliberately screw you over. There are few other players in the finance industry I'm reasonably confident that will be true of for the next 80 years.".

If you are the sort of person who benefits from a robo-advisor you are unlikely to be the sort of person to be able to properly weigh the plusses and minuses of Schwab's slide into a sellside house. I wish that weren't so and disclosure would matter but it just isn't. I think this is a call that we as the people on this thread just have to make one way or another. The situation with Betterment is easier to understand, so I did include it. The way I see it, at the end of the day this is Bogleheads not Schwabheads. While my recommendation may be SIP, Boglehead's recommendation will never be SIP.

So even with my strong beliefs and understandings of conflicts of interest my personal accounts are with BofA/Merrill Edge who are as conflicted as it is possible to be. Those conflicts of interest allow them to offer a richer range of services to me at a lower cost than would otherwise be possible. I understand this stuff far more than people who use robos and frankly I think much better than most. And yet still I often have no idea what their interests are in offering me these services. Why do I have hundreds of quite good loaded mutual funds being offered to me with no load charges? Why is Merrill waiving margin interest so aggressively between buys and settlements and settling cash late allowing me to collect extra interest? Why did Merrill go out of their way to cut my options commissions below their posted rates? Why is Bank of America lending me money below their cost because I have a large balance at Merrill? I have benefited from all these policies. I can speculate about possible angles for why they are losing money on so many of those things but I have no idea why. Those sorts of things don't happen for Vanguard Brokerage customers because Vanguard's business model is straight forward. Is that to your net benefit or net harm?

This issue of the complex conflicts between sellside and buyside is a general problem that effects the entire finance industry and effects a huge range of products. It even indirectly does effect Bogleheads. For example Blackrock right now doesn't have enough issuance of new bonds for their ETFs. As a consequence they are guaranteeing sales for sell side houses below prevalent rates if they can drum up corporate interest in borrowing. Vanguard of course indexes sales so they blindly buy these bonds at Blackrock's below fair value rates. How do we want to cover that on 3-fund discussions? What if it gets to be 1/2 the portfolio?

So that's my $.02 on conflicts of interest. I think it deserves its own threads. This is a big problem that effects all Americans in every financial product.
____

Now onto the little problem. The intermediate bonds vs. short bond thread I'm still not completely sold on. I'm getting more positive to it since I've been here, Bogleheads are changing my mind on this one. But I'd really love to backtest those two through a bad cycle for bonds (say 1932-1982). So I'm not sure for the "safe bonds" what my recommendation is today. I'm liking all the extra yield form the intermediate bonds that are popular here, the extra risk doesn't seem so bad. But I want to put it under more stress before I switch to agreeing with 3-funders that intermediates are the way to go.

siamond wrote: Any more views on Schwab SIP, before we move on?
Nope. I think we did what we are going to do. Until you have a better way to model smart beta, and can backtest further I'm satisfied we did the best Simba can do for Schwab. I think you did a great job.
Last edited by jbolden1517 on Fri Aug 04, 2017 7:33 pm, edited 3 times in total.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

jbolden1517 wrote:I'm going to give a really long answer here. I really don't know of a better alternative. 9% cash vs. short bonds is probably about 18 basis points at worst. Which is less than the fees of any of the robos.

Understood - I wasn't dissing SIP here, I was just pointing out that we should carefully weigh such cash drag against fees, and be clear about such issues in reviews. All the extra color you provided is very interesting though (would be worth a blog entry! :wink:)

My "SIP TBM" above was based on re-assigning the 9% money market / cash to a TBM position (modeled after VBMFX), which is diversified and displays an intermediate-term on average. I ran a quick test replacing TBM by STB (modeled after VBISX). The former displays roughly a 30 basis point advantage over cash, while the latter displays roughly a 20 basis point advantage (you were right on the money!). The delta is VERY consistent among periods of 10 years, always in favor of TBM.
jbolden1517 wrote:The intermediate bonds vs. short bond thread I'm still not completely sold on. I'm getting more positive to it since I've been here, Bogleheads are changing my mind on this one. But I'd really love to backtest those two through a bad cycle for bonds (say 1932-1982). So I'm not sure for the "safe bonds" what my recommendation is today. I'm liking all the extra yield form the intermediate bonds that are popular here, the extra risk doesn't seem so bad. But I want to put it under more stress before I switch to agreeing with 3-funders that intermediates are the way to go.
I am happy to run a more extensive backtest on the matter, just open a new thread, share your views and questions in the OP, and I can provide the same type of charts and metrics as we used here. Oh, and I share your concern about recency bias with bonds, the post WW-II period is downright frightening.
longinvest
Posts: 4535
Joined: Sat Aug 11, 2012 8:44 am

Re: Backtesting some robo-advisers asset allocations

Post by longinvest »

siamond wrote: Oh, and I share your concern about recency bias with bonds, the post WW-II period is downright frightening.
Any such statement should be qualified to say that it concerns the performance of nominal bonds relative to inflation in an era when the Gold Standard was abandoned (Only Two Centuries of Data). We have no data to backtest the behavior inflation-indexed bonds such as Treasury Inflation Protected Securities (TIPS) over the last century.

Oh, and by the way, the Gold Standard cannot be abandoned again.
Last edited by longinvest on Fri Aug 04, 2017 10:07 am, edited 1 time in total.
Bogleheads investment philosophy | One-ETF global balanced index portfolio | VPW
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

siamond wrote: I am happy to run a more extensive backtest on the matter, just open a new thread, share your views and questions in the OP, and I can provide the same type of charts and metrics as we used here. Oh, and I share your concern about recency bias with bonds, the post WW-II period is downright frightening.
Thanks for the offer. Done. viewtopic.php?f=10&t=224979
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

Let's take a look at Betterment now. Here is an extract of the review the all-knowledgeable jbolden1517 provided here:
jbolden1517 wrote:Betterment is today the most popular robo-advisor out their among mainstream choices, they were also one of the first 2. They use a value tilt domestically and a pure indexing approach internationally. They also have a heavy Vanguard focus.
[...]
Core funds are funds most Bogleheads will be familiar with (similar to what Bogleheads who do a little slice and dice + value tilt would have). The bond allocation is more complex than most Bogleheads and seems to be dilutive. Probably somewhat better than 3-fund for rebalancing but not much better. Here is a web application giving the allocations as you dynamically move the stock up or down: https://www.betterment.com/resources/re ... tion-chart

They currently offer up to 1 year with no fees (at $500k, starts at 1mo free at $10k) The fee structure after is:
25 basis points -- robo advisor only
40 basis points -- One FA call + robo
50 basis points -- Unlimited FA + robo
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

Thanks to this link, one can fairly easily figure out what exact Asset Allocation Betterment would recommend for a 70/30 portfolio (I focused on the tax-preferred case, as it seemed more typical). And then we can proceed with comparing to a 3-funds portfolios, to more tilted portfolios, and while we're at it, let's include the SIP 'EMB' portfolio that was discussed in previous posts. While not forgetting that Betterment charges an additional 0.25% fee for their services.

Image
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

Here is the usual risk (equated to volatility) / reward (equated to total-returns) chart for the 1985-2016 time period.

Image

For those of you who care about this type of metrics, I ran a test for all decades in the time period, looking at Sharpe and Sortino ratios. Betterment is doing a tad better on average than the 3-funds portfolio, but SIP EMB and the other tilted portfolios clearly did better. I re-ran the test while forcing the Betterment add'l fee to zero, this fixed the Sharpe ratio, but Sortino remained sub-par.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

Finally, let's look at the Telltale chart, using the 3-funds 70/30 as benchmark. Remember, this is the relative performance of a given portfolio (aggregate growth of the portfolio divided by the cumulative growth of the benchmark). So this basically shows the periods of time where a given portfolio does better (or worse) than TSM, and where we end up at the end, relatively speaking.

Image

Here we can see that Betterment didn't fall as hard as the others (compared to TSM) during the deep crisis, and overall displayed a fairly sustained advantage compared to TSM. Still, it just doesn't compare to the performance of SIP or the Tilt2 portfolio.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

Overall, the Betterment choices seem quite reasonable, and displayed a very decent track record in the period of time we studied (1985-2016). The portfolio definition is also fairly straightforward and reasonably easy to understand (certainly easier than SIP). The thing that puzzles me the most is the emphasis on Large-Cap Value and Bonds-Int'l, neither of which adding much to the portfolio, if anything (at least in this time period). The absence of a cash position is also appreciated, but this comes with a 0.25% additional fee, which is quite significant.

What do you guys think?
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

siamond wrote:Overall, the Betterment choices seem quite reasonable, and displayed a very decent track record in the period of time we studied (1985-2016). The portfolio definition is also fairly straightforward and reasonably easy to understand (certainly easier than SIP). The thing that puzzles me the most is the emphasis on Large-Cap Value and Bonds-Int'l, neither of which adding much to the portfolio, if anything (at least in this time period). The absence of a cash position is also appreciated, but this comes with a 0.25% additional fee, which is quite significant.

What do you guys think?
I'm a little surprised that Betterment ended up that volatile an extra 1/2 percent isn't huge but I would have figured it would be lower. I also am very impressed at the sort of consistent outperformance. 1989, 1993, 1997 are the only years a Betterment investor would have felt they weren't "keeping up" with their friends and even then not by much. There is a lot of talk about "tracking error". I don't care much, frankly I think its a good thing for Americans to have tracking error, their lives correlate with the market. But some people are concerned and its nice to see the negative tracking error was so infrequent. Obviously this is a worse portfolio than SIP, Tilt1 or Tilt2 in terms of meeting retirement goals. But I'm glad they did as well as they did consistently relative to the 70/30. I guess my original point about being boring stands. Even in comparison they are kinda boring. Boring is probably a good thing in investing. If someone is looking for a portfolio that's performs a lot like 3-fund but just a touch better, my intuition was Betterment made sense. But I didn't quite expect to see this level of consistency.

OK. Now some questions for the next round. Wealthfront is the other big player that uses CRAAL (there are many more robos. I could do a review and then we could do another comparison. Problem is for taxable they include: Natural Resources Index (DJP). Do you have any way of handling that? Would be curious to see how that plays as a diversifier.

The other performance test that would be way cool if we can handle it is Blackrock's robo. It isn't available in the United States but there are European Bogleheads. Problem is it uses weekly rebalancing based on risk. Basically the idea is that the FTSE has a slight negative week to week correlation (-1.81%) but a strong week to week volatility correlation (66.19%). I don't know if this is true for the SP500 but I suspect it likely is (though the numbers would obviously be slightly different). So to hold risk constant the portfolio needs to adjust. So the customer picks their 95% drawdown (i.e. there is a 95% chance you will not drawdown below X amount in the next year) and the portfolio allocates around that risk. Obviously we can't do that. But we likely could treat is like a CRAAL portfolio and see what it would do without adjustment. I think I can put together a sort of median 70/30ish portfolio to test. These are the indexes though, very European.

Deutsche Bank SONIA Total Return Index
Barclays EuroAgg Treasury Total Return Index
JP Morgan Emerging Markets Bond Index
Barclays US Treasury 10 Year Term Index
FTSE Actuaries Government Securities UK Gilts Under 5 Years Total Return Index
FTSE Actuaries UK Conventional Gilts All Stocks Total Return Index
Barclays EuroAgg Corporate Total Return Index
Markit iBoxx USD Liquid High Yield Total Return Index
FTSE Developed Europe Index
FTSE All-World Developed Asia Pacific ex Japan Net Total Return Index
MSCI Emerging Markets Investable Market Index
S&P 500 Net Total Return Index
FTSE Japan Net Total Return Index
FTSE EPRA/NAREIT Developed Dividend+Net Total Return Index
Bloomberg Commodity Index 3 Month Forward Total Return Index

Can we do anything with those?
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

jbolden1517 wrote:I also am very impressed at the sort of consistent outperformance. 1989, 1993, 1997 are the only years a Betterment investor would have felt they weren't "keeping up" with their friends and even then not by much. [...] If someone is looking for a portfolio that's performs a lot like 3-fund but just a touch better, my intuition was Betterment made sense. But I didn't quite expect to see this level of consistency.
Well, large-cap value (LCV) really doesn't make much of a difference compared to large-cap-blend (LCB) or TSM. The combined LCV/TSM position in Betterment is significantly higher than the LCB position in SIP. EAFE at 30% is also obviously very strongly correlated to the Int'l position we have in the 70/30 portfolio. The bonds side is also not that different, in truth (Int'l Bonds is surprisingly aligned with TBM - not sure this will last, but at least so far). I think this explains the consistency.
jbolden1517 wrote:Now some questions for the next round. Wealthfront is the other big player that uses CRAAL (there are many more robos. I could do a review and then we could do another comparison. Problem is for taxable they include: Natural Resources Index (DJP). Do you have any way of handling that? Would be curious to see how that plays as a diversifier.
That one should be fine. We have an iShares GSG date series tracking commodities, relying on the S&P GSCI for the earlier years, starting in 1970.
jbolden1517 wrote:The other performance test that would be way cool if we can handle it is Blackrock's robo. It isn't available in the United States but there are European Bogleheads. [...] I think I can put together a sort of median 70/30ish portfolio to test. These are the indexes though, very European. [...]
Hm. This seems more tricky. That would imply more research to assemble appropriate data series, and it isn't too clear we could go back in time enough. Let me think a bit more about it.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

jbolden1517 wrote:I'm a little surprised that Betterment ended up that volatile an extra 1/2 percent isn't huge but I would have figured it would be lower.
I looked at that a little more closely, as this puzzled me a bit too. Turns out that the 1985-2016 entire time period gives one result (a bit less than 50 basis points), while my data table (looking at all decades, and I also did all 15 years periods) indicates an average delta which is a bit smaller, between 20 and 30 basis points. Looking at each smaller period (10 or 15 years), the delta is actually quite variable, Betterment being sometimes *less* volatile than the 3-funds portfolio. As to max. drawdowns, they are very similar.
=> Bottomline: if I were a Betterment investor, I wouldn't worry much about this...
LocusCoeruleus
Posts: 47
Joined: Wed May 24, 2017 10:00 pm

Re: Backtesting some robo-advisers asset allocations

Post by LocusCoeruleus »

In terms of baseline studies, we have looked at Base 70/30 (with 2 tilts) and Base 80/20. Based on the great updates by our simba expert siamond, may I propose a Base 75/25 with a slightly different Tilt (call it tilt 3):

Image

Image

Image

Image
User avatar
Ethelred
Posts: 379
Joined: Sun Oct 30, 2016 9:38 am

Re: Backtesting some robo-advisers asset allocations

Post by Ethelred »

siamond wrote:Overall, the Betterment choices seem quite reasonable, and displayed a very decent track record in the period of time we studied (1985-2016). The portfolio definition is also fairly straightforward and reasonably easy to understand (certainly easier than SIP). The thing that puzzles me the most is the emphasis on Large-Cap Value and Bonds-Int'l, neither of which adding much to the portfolio, if anything (at least in this time period). The absence of a cash position is also appreciated, but this comes with a 0.25% additional fee, which is quite significant.

What do you guys think?
Almost all of the under-performance between Betterment and SIP plus the two tilted portfolios happens in the period 2000 to 2006. What is driving that? The main differences are the lack of I-Small, I-Value and US REITs.
Avo
Posts: 1099
Joined: Wed Jun 11, 2008 2:21 am
Location: California

Re: Backtesting some robo-advisers asset allocations

Post by Avo »

I would like to see the results with a "classic" bogleheads tilt: equal parts TSM, SCV, Total Int'l, and Int'l Small for equities (which is implementable with Vanguard index funds/ETS) and TBM for bonds.
Avo
Posts: 1099
Joined: Wed Jun 11, 2008 2:21 am
Location: California

Re: Backtesting some robo-advisers asset allocations

Post by Avo »

Ethelred wrote:Almost all of the under-performance between Betterment and SIP plus the two tilted portfolios happens in the period 2000 to 2006. What is driving that?
More MCV and SCV, and a big does of Health Care.
User avatar
Ethelred
Posts: 379
Joined: Sun Oct 30, 2016 9:38 am

Re: Backtesting some robo-advisers asset allocations

Post by Ethelred »

Avo wrote:More MCV and SCV, and a big does of Health Care.
Sorry, you misunderstand. I was asking Siamond about the portfolios he tested earlier.
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

LocusCoeruleus wrote:In terms of baseline studies, we have looked at Base 70/30 (with 2 tilts) and Base 80/20. Based on the great updates by our simba expert siamond, may I propose a Base 75/25 with a slightly different Tilt (call it tilt 3):
I think that's a good analysis if you want to argue healthcare as a diversifier. Which it might be since it is going to be extremely non-cyclical and correlate more with legal changes, government funding including pseudo-funding through mandates and population demographics. That being said I'd be cautious about seeing those returns as typical because not only was healthcare a rebalancer it also experienced tremendous growth. Healthcare was allowed to massive expand as part of America's GDP. And not only was the massive expansion permitted but it was allowed inefficiencies unique almost globally. And not only that the government guaranteed many healthcare providers protection from margin pressure. I'm going to sound like a typical Boglehead here. Picking the right industry for a 30 year run is awesome for your returns. All you have to do now to capture that return is know what are the hot industries are in 2047 that the market of 2017 is failing to appreciate the value of.
Last edited by jbolden1517 on Sat Aug 05, 2017 9:27 am, edited 1 time in total.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

Avo wrote:
Ethelred wrote:Almost all of the under-performance between Betterment and SIP plus the two tilted portfolios happens in the period 2000 to 2006. What is driving that?
More MCV and SCV, and a big dose of Health Care.
2000 to 2006 saw a really big surge in small/mid value premium, and REITs went with it. SIP being significantly more 'value-y' than Betterment (the LCV position in Betterment doesn't do much, only small/mid value makes a difference), this explains what Ethelred noticed.

Avo, I am not sure what you mentioned Health Care though? Why do you think that SIP is different from Betterment in this respect?
Last edited by siamond on Sat Aug 05, 2017 9:32 am, edited 1 time in total.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

LocusCoeruleus wrote:In terms of baseline studies, we have looked at Base 70/30 (with 2 tilts) and Base 80/20. Based on the great updates by our simba expert siamond, may I propose a Base 75/25 with a slightly different Tilt (call it tilt 3):
Well, I'm glad to see that somebody got the hang of using the Simba spreadsheet! Nicely done! :happy
jbolden1517 wrote:I think that's a good analysis if you want to argue healthcare as a diversifier. Which it might be since it is going to be extremely non-cyclical and correlate more with legal changes, government funding including pseudo-funding through mandates and population demographics. That being said I'd be cautious about seeing those returns as typical because not only was healthcare a rebalancer. Healthcare was allowed to massive expand as part of America's GDP. And not only was the massive expansion permitted but it was allowed enormous inefficiencies unique almost globally. And not only that the government guaranteed protection from margin pressure. I'm going to sound like a typical Boglehead here. Picking the right industry for a 30 year run is awesome for your returns. All you have to do now to capture that return is know what are the hot industries are in 2047 that the market of 2017 is failing to appreciate the value of.
Yes, I would agree with jbolden1517. There had been numerous studies about sectors (I remember a great one from Credit Suisse, can dig up the pointer if interested),and they are indeed randomly cyclical. And even the naive observer (that would me!) can see that the health-care run-up of the last couple of decades appears unlikely to repeat itself. Aside from that, the 'Tilt3' portfolio is VERY tilted, and it isn't surprising that it outperformed the others. I would be wary about making such an extreme bet though. Draw a telltale chart with 100% SCV, 100% MCV, 100% REITs and compare to TSM (start earlier than 1985), and you'll see occasional spurts of big (relative) growth, and then multiple decades during which no premium whatsoever materialized (or a negative premium did!). Staying the course in those conditions would have been terribly hard. Don't know what the future will bring us, but I'd be wary about such observations of the past, and not go too far in the tilting direction.
User avatar
LadyGeek
Site Admin
Posts: 68701
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Backtesting some robo-advisers asset allocations

Post by LadyGeek »

LadyGeek wrote:
LocusCoeruleus wrote:Siamond- a request to do similar analysis for wealthfront & betterment. I love that telltale chart. Thanks!
I also like that telltale chart, which deserves more explanation than what's in the backtesting spreadsheet.

Next up for the wiki, a deep-dive into the details of the telltale chart. siamond has given me some pointers.
We now have a new wiki page: Telltale chart

The discussion thread is here: New Wiki page: the Telltale Chart => feedback?
Wiki To some, the glass is half full. To others, the glass is half empty. To an engineer, it's twice the size it needs to be.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

Let's give it a try with Wealthfront now. You can find an extensive review here.

Here are the two asset allocations we'll backtest. One is recommended for tax-sheltered accounts (we'll call it WealthfrontSh), and the other is recommended for taxable account (we'll call it WealthfrontTx).
jbolden1517 wrote:
siamond wrote:Could you please document a Wealthfront portfolio that maps to 70/30 according to such criteria?
The fund is dividend growth. That ain't a bond, agree completely.

Funds:
* US Total Stock Index (VTI)
* Developed Foreign Stock Index (VEA)
* Emerging Markets Index (VWO)
* Dividend Yield Stock Index (VIG)
* Natural Resources Index (DJP)
* Municipal Bonds Index (MUB)
* Emerging Market Bond Index (EMB)
* Corporate Bond Index (LQD)
* Real Estate Index (VNQ)

Retirement (funds as above):
19% US stock
15% Dividend stock
15% Foreign stock
11% EM stock
10% Real estate
21% corp bond
9% EM bond

taxable: (note I had to lower my risk tolerance below what I used for retirement to get the portfolio to 70/30; this corresponds to the 65/35 non-taxable)
33% US stock
6% Dividend
15% Foreign stock
12% EM
5% natural resources
29% munis
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

The Wealthfront Asset Allocations are a bit peculiar by their use of natural resources (commodities) and high-yielding dividend funds.

We have a data series in the Simba spreadsheet for commodities, based on the iShares GSG fund and for early years the S&P GSCI index. The returns were pretty poor (actually truly abysmal in recent years): 2.8% CAGR for the 1985-2016 time period (was significantly better between 1970 and 1985). The only point of using commodities might be because of their remarkably low correlation (close to zero!) with pretty much everything else. Still, the atrocious CAGR would have been a big drag.

Now the high-yielding dividend fund is a whole different game. Weathfront uses Vanguard VIG, the corresponding mutual fund investor being VDAIX. Vanguard tracks a Nasdaq "Dividend Achievers" index. Unfortunately, both the fund and the index started in 2006, so we only have 10 years of history. A quick comparison with the S&P 500 shows that this fund didn't make much of a difference with the S&P 500 (click here for the usual Morningstar graph). Which matches my intuition about such dividends-oriented funds.

Then I looked for indices providing more history... There is an iShares ETF (DVY) similar to the Vanguard one, which tracks the Dow Jones US Select Dividend index, which has solid history (back to 1992). It displays amazing performance though, and while digging a bit more, it appears to be much more in the Mid-Cap-Value realm, while Vanguard and its index are more in the Large-Cap-Blend realm. Then there is the grand-daddy of such indices, the amusingly named S&P 500 Dividend Aristocrats index, allowing to go back to 1990. This one seems more compatible with the Vanguard large-cap optic, although it uses an equal-weight technique while Vanguard/Nasdaq are cap-weighted.

What troubles me a lot though is that the S&P Aristocrats index displays a monster 11% return since 1990, while the S&P 500 only returned 9%. That is a little hard to believe, unless somehow the S&P Aristocrats filtering procedure ends up creating a strong factor tilt (well, maybe it does). If anybody has more views to share on this, please enlighten us.

Anyhoo, I ended assembling the following data series (adjusting with the Vanguard ER 0.17% for the index years):
- 1985-1989: S&P 500 (by lack of a better source)
- 1990-2006: S&P 500 Dividend Aristocrats
- 2007-2016: Vanguard VDAIX
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

And now for the usual charts. That is... interesting!

Image

Image
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

Well, the charts from the last post are a little eye-popping, right?

- The Wealthfront "taxable" portfolio isn't terribly attractive. After some initial hiccups, it basically stayed on par with the 70/30 3-funds portfolio. It seems to me that those commodities really did no good, while the EM and Div-Yielding (low) exposure helped balance things out.

- Now as the Wealthfront "sheltered" portfolio, this one rocked the boat. Excellent CAGR (10.8%), plus Sharpe and Sortino ratios are the highest I've seen so far. No commodities in there, but a strong exposure to both EM and Div-Yielding (and also EM bonds) apparently did the trick. I remain skeptical about the Div-Yielding data series and the S&P 500 Aristocrats though, and I strongly suspect the Wealthfront "sheltered" backtesting is artificially generous.

I ran one last test, replacing the Div-yielding data series by a simple mapping to TSM. This brought the overall performance on par with SIP EMB, but the trajectory gets steadier, much less of a relative drop at the end of the 90s (and a better Sharpe ratio than SIP). So I think we can say that even if the high-yielding dividends data series is a bit of a fantasy, Wealthfront (tax-)Sheltered remains a quite remarkable candidate, which apparently displayed excellent performance. While Wealthfront Taxable is much less remarkable...
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

siamond wrote:Well, the charts from the last post are a little eye-popping, right?

- The Wealthfront "taxable" portfolio isn't terribly attractive. After some initial hiccups, it basically stayed on par with the 70/30 3-funds portfolio. It seems to me that those commodities really did no good, while the EM and Div-Yielding (low) exposure helped balance things out.
Question. Did you tax adjust for the munis obviously with such a heavy weight in munis this becomes critical.
siamond wrote: - Now as the Wealthfront "sheltered" portfolio, this one rocked the boat. Excellent CAGR (10.8%), plus Sharpe and Sortino ratios are the highest I've seen so far. No commodities in there, but a strong exposure to both EM and Div-Yielding (and also EM bonds) apparently did the trick. I remain skeptical about the Div-Yielding data series and the S&P 500 Aristocrats though, and I strongly suspect the Wealthfront "sheltered" backtesting is artificially generous.
I would agree the results are going to be good. I suspect a dividend yield fund would have done well for the same reason the Healthcare "tilt3" fund did well. It pays a lot to have a big tilt where in sectors and industries that won. Dividend growth was clearly a good strategy to be in. Some outperformance doesn't surprise me.

As bond yields have cratered investors who want yield have had to turn from bonds (funds and munis) to stocks. Index funds sell concentration so buybacks have been less effective since 2009 in elevating stock prices and dividends more effective in elevating prices. Since 2012 the payout ratio between dividends and earnings for the SP500 has been rising quite rapidly reversing a multi-decade trend towards buybacks.
Image

As dividend stocks have appreciated more than the market, traditional growth investors performance chasers have been moving into dividend growth stocks giving them yet another boost. The chart reflects what happens when a strategy that would backtest really well is used.

Now to get out my crystal ball. If bond yields remain artificially depressed due to central bank buying (and forced financial institution purchases in Europe) it is not hard to imagine that investors turn to equity income and dividends become a key component of return. I think it is entirely possible that a dividend growth strategy continues to work for a very long time. Investors who have been through the 2000 bear may be becoming more skeptical of earnings growth estimates and assigning a higher risk premium to earnings growth as opposed to dividend growth. In 2008 dividend yield up held up much better than share prices (-21% vs. -57%). Dividends did provide portfolio safety as for people needing to draw income. Money invested from dividends will experience the same dollar cost averaging advantages as any other periodic money invested coming in at the harmonic mean (a lower share price). With fully invested strategies, indexing and passive, becoming normal dividends should be expected to compound slightly faster. In other words all other things being equal dividends are preferable to retained earnings. If dividend growth is rewarded more strongly than earnings growth companies will start structuring themselves to provide a stable dividend growth. Those that already have done this will have a slight advantage.

So this is one strategy I don't see any reason for a mean reversion in coming decades. If anything I would lean a bit towards it not reverting. But the one time boom from having been there early, that I don't see repeating.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

jbolden1517 wrote:If bond yields remain artificially depressed due to central bank buying (and forced financial institution purchases in Europe) it is not hard to imagine that investors turn to equity income and dividends become a key component of return. [...]
So this is one strategy I don't see any reason for a mean reversion in coming decades. If anything I would lean a bit towards it not reverting.
Well, I understand your point, but it doesn't seem to match the reality of the past 10 years with real-life funds, please check this Morningstar comparison chart. This compares the Vanguard Dividend Appreciation Index Fund (VDAIX), the iShares Select Dividend ETF (DVY) and the regular S&P 500 Vanguard fund (VFINX). Hardly convincing!
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

siamond wrote:Well, I understand your point, but it doesn't seem to match the reality of the past 10 years with real-life funds, please check this Morningstar comparison chart.
Interesting Dividend Aristocrats crushes 2008-present, with slightly better performance than the SP500 2002-2008.
Image
DVY isn't much different than SP500. Not sure what to make of this. Anyone want to weigh in?
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

jbolden1517 wrote:DVY isn't much different than SP500. Not sure what to make of this. Anyone want to weigh in?
I would just observe that, contrary to common wisdom, there might be something to an equal-weighting approach (here is an interesting write-up from Bogleheads AlohaJoe, and another one on the same topic).

This being said, I believe it is costly to operate such a fund in real life, due to the constant need for rebalancing and other considerations. There is a fund which tracks the S&P 500 Aristocrats, dubbed ProShares S&P 500 Dividend Aristocrats (NOBL). Love the stock ticker! :D Doesn't have enough history to tell us anything though.
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

siamond wrote:
jbolden1517 wrote:DVY isn't much different than SP500. Not sure what to make of this. Anyone want to weigh in?
I would just observe that, contrary to common wisdom, there might be something to an equal-weighting approach (here is an interesting write-up from Bogleheads AlohaJoe, and another one on the same topic).

This being said, I believe it is costly to operate such a fund in real life, due to the constant need for rebalancing and other considerations. There is a fund which tracks the S&P 500 Aristocrats, dubbed ProShares S&P 500 Dividend Aristocrats (NOBL). Love the stock ticker! :D Doesn't have enough history to tell us anything though.
Guggenheim has been playing with equal weighted sectors. I'm starting to eye those funds for myself for next big swing into the USA. I'm glad people are trying this strategy out. Seems to implement a sort of automatic value tilt without totally throwing away growth. I'd assume as managers get experience in running ETFs this way costs of doing it will decline, unless it turns out to be a terrible idea. The index I've had my eyes on for 2 decades is the Value Line index. Right now First Trust is trying to sort of do it (100 stocks) and pretty much failing: http://www.ftportfolios.com/retail/etf/ ... Ticker=FVL Will be interesting now that people are trying.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

jbolden1517 wrote:
siamond wrote:Well, the charts from the last post are a little eye-popping, right?

- The Wealthfront "taxable" portfolio isn't terribly attractive. After some initial hiccups, it basically stayed on par with the 70/30 3-funds portfolio. It seems to me that those commodities really did no good, while the EM and Div-Yielding (low) exposure helped balance things out.
Question. Did you tax adjust for the munis obviously with such a heavy weight in munis this becomes critical.
I didn't tax-adjust anything. Simba's computations are performed pre-tax. Everybody's situation is different post-tax...
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

siamond wrote:
jbolden1517 wrote:
siamond wrote:Well, the charts from the last post are a little eye-popping, right?

- The Wealthfront "taxable" portfolio isn't terribly attractive. After some initial hiccups, it basically stayed on par with the 70/30 3-funds portfolio. It seems to me that those commodities really did no good, while the EM and Div-Yielding (low) exposure helped balance things out.
Question. Did you tax adjust for the munis obviously with such a heavy weight in munis this becomes critical.
I didn't tax-adjust anything. Simba's computations are performed pre-tax. Everybody's situation is different post-tax...
Can we up the returns on the munis by 40% and rerun? (or at least 30%). Given a 29% position in munis I think we are being hugely unfair to Wealthfront taxable by ignore the higher tax efficiency. This is not a small issue for them. We could be talking 150 basis points especially for those early years.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

jbolden1517 wrote:
siamond wrote:I didn't tax-adjust anything. Simba's computations are performed pre-tax. Everybody's situation is different post-tax...
Can we up the returns on the munis by 40% and rerun? (or at least 30%). Given a 29% position in munis I think we are being hugely unfair to Wealthfront taxable by ignore the higher tax efficiency. This is not a small issue for them. We could be talking 150 basis points especially for those early years.
Well, it would be inconsistent to adjust the Munis for tax savings, while staying pre-tax for the other assets (e.g. foreign tax credit for international, etc). Plus such adjustment might fit one personal tax situation, and not another. Better stay pre-tax and stay consistent. I see your point though, so I ran a test replacing the 29% Munis by 20% TBM and 9% Corp of the same duration (no particular reason for the split, I just picked some numbers!). This should be an indirect way to capture your point, as the munis are supposed to be priced based on their average tax savings compared to TBM (I am repeating common wisdom here - I suspect the reality is probably more subtle). WealthfrontT2 is the new AA, and sure enough, it does look better.

Image
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

siamond wrote:
jbolden1517 wrote:
siamond wrote:I didn't tax-adjust anything. Simba's computations are performed pre-tax. Everybody's situation is different post-tax...
Can we up the returns on the munis by 40% and rerun? (or at least 30%). Given a 29% position in munis I think we are being hugely unfair to Wealthfront taxable by ignore the higher tax efficiency. This is not a small issue for them. We could be talking 150 basis points especially for those early years.
Well, it would be inconsistent to adjust the Munis for tax savings, while staying pre-tax for the other assets (e.g. foreign tax credit for international, etc). Plus such adjustment might fit one personal tax situation, and not another. Better stay pre-tax and stay consistent. I see your point though, so I ran a test replacing the 29% Munis by 20% TBM and 9% Corp of the same duration (no particular reason for the split, I just picked some numbers!). This should be an indirect way to capture your point, as the munis are supposed to be priced based on their average tax savings compared to TBM (I am repeating common wisdom here - I suspect the reality is probably more subtle). WealthfrontT2 is the new AA, and sure enough, it does look better.

Image
That's about right. It gets the rebalancing advantages but doesn't collect the value premium. I think those are good estimates. After all back testing in general needs to be taken with a huge grain of salt.

As for the stuff about personal you actually led me to find an interesting point. I changed the portfolio to one for someone earning $!5k / yr with a $20k portfolio and it kept the municipal bonds! That is just a bug in the algorithm. That portfolio makes no sense at all at those numbers.
User avatar
Topic Author
siamond
Posts: 5726
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

Just to provide a few numbers for more context behind the TBM/Munis reasoning, here are a few US bonds CAGRs for the 1985-2016 time period (incl. the Vanguard ER, but not the robo additional ER - if any-):
- TBM: 6.8%
- IT Treasuries: 6.9%
- ST Treasuries: 5.3%
- IT Corp: 7.9%
- IT Munis: 6.1%
- TIPS: 6.9%
- HY Corps: 8.0%
- T-Bills: 3.5%

Remember, this is ONE time period though, where bonds did pretty good overall. If there is one type of assets where past returns can be especially misleading when trying to guesstimate the future, this would be bonds...
Post Reply