What we know about robo advisors

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

siamond wrote:I have a broader question though. I find myself quite unconvinced by the value-added of the 'robot' doing the TLH and/or the rebalancing on my behalf. And I don't see that the two portfolios documented so far are that special, aside from what I would call 'fairly meaningless hair splitting' in some categories. :wink:

Where I would see true value-added would be a truly useful process requiring more computational and modeling power. Optimizing correlations to death ala MPT might be one (and yes, a 3 fund portfolio isn't entirely optimal in this respect), but it seems a fool's errand to me, first because it's an always moving target (correlation patterns keep changing - just ask questions about gold miners to MPT supporters before/after 2008!),
This whole thread you are making the argument for Hedgeable. The use dynamic correlations weighing intra day trades, daily correlations, weekly, monthly along with longer term. So as correlations shift dynamically they adjust the portfolio targets.
siamond wrote: but more importantly because it equates risk with volatility, and personally, volatility is like the 5th or 6th item of my list of risks to put under control... I actually view long-term risks as much more important than short-term risks, and defined my AA accordingly. And trying to explain a lot of shades of grey about risks to a computer, good luck with that.
Volatility is sort of the standard definition of risk. I personally lean towards the Kondratiev model: how does the portfolio do in inflation recessions, deflationary depressions, reflationary growth surges, disinflationary asset booms? So right now for example I think we are either near the end of deflation or the beginning of reflation (depending on choices not yet made). Which means I'm worried about default risk. I'm also worried about sharp sudden inflation as monetary velocity picks up. I'm not worried about gradually falling interest rates or stagflation. Though I think any "good for all seasons" portfolio needs to be (one of the reasons I think TSM + intermediate high quality bonds is terrible for income investors).

So I agree with your sentiment for myself but I don't see how to define that rigorously in a way that could get broad agreement in this society.
siamond wrote: Where I would see true value-added would be tax optimization, and that goes way beyond TLH. Full tax planning requires analyzing decades ahead to what will happen (e.g. RMDs, SSA/Pension) and what could happen (asset location selection and changes, things like Roth conversion, states when you work/retire, etc). THAT is complicated. Personally, I did my own spreadsheet for this, but I know it will be a pain to maintain, and that I am not taking all factors in account, very far from it, and not everybody is as geeky as I am. It appears that there are few financial advisors capable of doing something like that, actually. Heck, because it's complicated!
That starts getting into wealth management not financial advisors. You can get that and much more but it ain't cheap. Expect something like a min fee of $20k / yr. As far as robos I think robos are a classic disruptive technology for financial advice / wealth management Image
They are working towards green line customers. What you want is the purple line. Add in things like intergenerational transfer of wealth and strategic philanthropy and you have the purple line. Give them another decade or two. They'll get there to the green and start challenging the purple.
siamond wrote: jbolden1517, are you aware of any robo-adviser going in such direction, some form of tax-optimized long-term planning?
Sort of. The cyborg-advisor model I mentioned robos have that sort of service. Basically the robo does the day to day implementation of the advisor's plan. I was thinking of skipping robos like Envestnet that aim themselves at the cyborg-advisor relationship because Bogleheads is a very do-it-yourself kind of crowd.

It wouldn't surprise me if Hedgeable in say 2030 is moving in that direction providing they can stay in business. They are having trouble growing fast right now. The next bear I think will be make or break for them. If they do as well as they claim they will, I suspect they quickly rise to billions in assets. If they don't then there is no more Hedgeable. Motif is already massive and their platform already has most of the customization to allow for what you are talking about in a cyborg advisor role. The money behind them is Goldman Sachs, JP Morgan, Foundation (Uber, Netflix, Lending Club)... Certainly Goldman and JPMorgan offer those services. Foundation has the pockets and technical depth to handle anything. So there is hope. But I have no inside info on their direction. The other one that I think might go that far near term is Macroaxis. They have $60b (not a typo) under management, don't seem to sell stuff at all as far as I can tell, and gets talked about in the financial press all the time. The are already doing rather sophisticated stuff with portfolios and this is before they even consider themselves to be a product.

So there is hope but no options I know of today.
User avatar
siamond
Posts: 6008
Joined: Mon May 28, 2012 5:50 am

Re: What we know about robo advisors

Post by siamond »

jbolden1517 wrote:This whole thread you are making the argument for Hedgeable. The use dynamic correlations weighing intra day trades, daily correlations, weekly, monthly along with longer term. So as correlations shift dynamically they adjust the portfolio targets.
Ok, but... stepping back a bit, this should show that the whole thing is rather unstable, hence false precision is just basically pointless. Anyhoo, this point is completely overridden by the next one (what is a useful definition of risk?) imho. Researchers and quants jumped on MPT because you can do a ton of great math about it. I don't believe that much complexity is actually needed, once you get solid diversification, you get the primary effect, and then other considerations should take over. Anyhoo, Ladygeek wanted some discussion about MPT, so I rambled a bit! I will provide Sharpe/Sortino ratios for the backtesting later tonight, to satisfy the MPT-oriented thinkers! :wink:
jbolden1517 wrote:Volatility is sort of the standard definition of risk. [...] So I agree with your sentiment for myself but I don't see how to define that rigorously in a way that could get broad agreement in this society.
Well, this is the standard definition of researchers because it's a neat mathematical formula, and you can do plenty of great math about it. This doesn't make it meaningful in real life for you and me. Personally, what matters the most is the trajectory of my income (the sum of SSA/Pension with the withdrawals from my portfolio), the higher the better (top-level goal), with some level of stability (2nd level goal) AND adaptiveness (top-level goal). As Michael McClung beautifully put it, "loss of income" should matter much more than "loss of value". I do agree this is hard to put in a neat little formula, but then, that's why I defined my AA by myself, a very manual process. This being said, I can see that robo-advisers are mostly for people who spend much less time thinking about those things than we do. Just bugs me that this silly 'volatility==risk' assumption finds its way at the core of some robo-advisers... :(
jbolden1517 wrote:
siamond wrote: Where I would see true value-added would be tax optimization, and that goes way beyond TLH. Full tax planning [...]
That starts getting into wealth management not financial advisors. You can get that and much more but it ain't cheap. I think robos are a classic disruptive technology for financial advice / wealth management
Yup, the latter (wealth management) seems much more promising to me. And will definitely require a 2nd generation of products, and I guess we're not there yet.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

siamond wrote:
jbolden1517 wrote:This whole thread you are making the argument for Hedgeable. The use dynamic correlations weighing intra day trades, daily correlations, weekly, monthly along with longer term. So as correlations shift dynamically they adjust the portfolio targets.
Ok, but... stepping back a bit, this should show that the whole thing is rather unstable, hence false precision is just basically pointless.
I assume you mean backtesting with Simba? In which case I agree. They (Hedgeable) make no claim they are going to "buy and hold" any particular portfolio at any particular time. There is no reason to evaluate them as if they were. Depending on what the math says they might put an accumulation investor in 30% gold, 15% EMs, 45% cash and 10% REIT. Ultimately the brave souls in it now are the ones testing it out. The next bear is when we really get to see how things pan out. If Hedgeable gets them out early and back into a mainstream portfolio either late in the bear or early in the next bull then I think its wise to give them part of the portfolio. We've entered the age of bigdata dynamic MPT and classic asset allocation is going to be in a defensive position.
siamond wrote: Anyhoo, this point is completely overridden by the next one (what is a useful definition of risk?) imho.
jbolden1517 wrote:Volatility is sort of the standard definition of risk. [...] So I agree with your sentiment for myself but I don't see how to define that rigorously in a way that could get broad agreement in this society.
Well, this is the standard definition of researchers because it's a neat mathematical formula, and you can do plenty of great math about it. This doesn't make it meaningful in real life for you and me. Personally, what matters the most is the trajectory of my income (the sum of SSA/Pension with the withdrawals from my portfolio), the higher the better (top-level goal), with some level of stability (2nd level goal) AND adaptiveness (top-level goal).
That you can define mathematically. Its utility. Which is the log of your real annual spending (equivalent to the geometric mean of your spending). For an income oriented investor the goal is to reduce the volatility of the outflows not portfolio volatility. That's one of the reasons I'm a big fan of these new inflation adjusted annuities. They aren't great in terms of average annual return, but they cover the tail risk you care about most. Similarly my argument for equity income (income investors being an area where I think 3-fund goes from highly suboptimal to reckless). That's why I advocate so strongly for these diversifiers. Making sure the early years of retirement are good is the key to getting a reliable income. Once you get past the early years safely you are in the clear. If not then the rest of retirement becomes touch and go.
siamond wrote: This being said, I can see that robo-advisers are mostly for people who spend much less time thinking about those things than we do. Just bugs me that this silly 'volatility==risk' assumption finds its way at the core of some robo-advisers... :(
I think it makes sense for accumulation investors. They care about compounding. The two things that kill compounding are low returns and high volatility.
siamond wrote: Yup, the latter (wealth management) seems much more promising to me. And will definitely require a 2nd generation of products, and I guess we're not there yet.
Actually the 3rd (or more likely 6th). We've already had the 2nd generation with Motif, Covester, Hedgeable...
User avatar
siamond
Posts: 6008
Joined: Mon May 28, 2012 5:50 am

Re: What we know about robo advisors

Post by siamond »

jbolden1517 wrote:
siamond wrote:
jbolden1517 wrote:This whole thread you are making the argument for Hedgeable. The use dynamic correlations weighing intra day trades, daily correlations, weekly, monthly along with longer term. So as correlations shift dynamically they adjust the portfolio targets.
Ok, but... stepping back a bit, this should show that the whole thing is rather unstable, hence false precision is just basically pointless.
I assume you mean backtesting with Simba? In which case I agree. They (Hedgeable) make no claim they are going to "buy and hold" any particular portfolio at any particular time. There is no reason to evaluate them as if they were. [...] We've entered the age of bigdata dynamic MPT and classic asset allocation is going to be in a defensive position.
No, I meant the dynamic MPT optimizers will be a somewhat unstable engine, and then its false precision seems rather pointless. At least, this is my gut feeling (and also what I read about past MPT experiments), I didn't work that much on MPT optimization myself because... the definition of risk is totally wrong imho. My own definition of risk is more complicated than a utility function, by the way, notably the adaptiveness side of things, but let's not side-track this thread with my own considerations! Would be cool if those robo-thingies could have a few utility functions handy, though.

I was mostly trying to point out that those robo-advisers seem very MPT-centric, and this could be quite a cause of concern to some investors (starting by myself) who view risk with different lens. Maybe something to point out in the wiki.

PS. and yes, you're right, the backtesting I am assembling (almost done) does assume a fixed allocation. It's not quite the right framework for a variable AA (tactical AA, dynamic MPT optimization, whatever).
User avatar
siamond
Posts: 6008
Joined: Mon May 28, 2012 5:50 am

Re: What we know about robo advisors

Post by siamond »

siamond wrote:
LadyGeek wrote:I'd also like to take a counterpoint to your most interesting comparison of SIP vs. 3-fund. For me, I'm not convinced until I see some data. In this case, it should be possible to backtest* the SIP portfolio against several "conventional" lazy portfolios.

We've got a spreadsheet for that: Simba's backtesting spreadsheet

I asked the guys in the support thread to see what they think: Re: Simba's backtesting spreadsheet [a Bogleheads community project]

* Backtesting - Using historical data to predict future performance. (Caveat: Tread carefully, past performance does not predict future performance.)
Hi there. I am doing most of the maintenance on the Simba spreadsheet nowadays. I can certainly run and document a backtest trying to approximate the SIP portfolio and compare it to what a typical (human being!) Bogleheads would do. I would suggest to select the closest 3-funds portfolio we can find (the 'classic' Bogleheads choice), but also to add a fairly typical tilted portfolio (as many Bogleheads do exactly that, starting by myself). Then compare.

This wouldn't be a perfect simulation, but at the first glance, I think we can get pretty close. I can also share some thoughts about rebalancing techniques. Now things like TLH, we can't model, as this *highly* dependent on one's personal circumstances. I'll work more on this later today, and open a dedicated thread on the matter. Overall, this robo-adviser discussion is very interesting, thank you for your illuminating inputs, jbolden1517.
Here you are, I ran a full backtest for the time period 1985-2016, comparing SIP to a 3 funds portfolio and more.

Please check this new thread, and please provide feedback out there.
User avatar
LadyGeek
Site Admin
Posts: 95686
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: What we know about robo advisors

Post by LadyGeek »

^^^ We can continue the general discussion here.

The wiki page has been updated with jbolden1517's Betterment review. See: Robo-adviser (Member reviews)
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
billthecat
Posts: 1052
Joined: Tue Jan 24, 2017 1:50 pm
Location: USA

Re: What we know about robo advisors

Post by billthecat »

jbolden1517 wrote:Schwab Intelligent Portfolios
Image
How/where did you generate that allocation summary?
We cannot direct the winds but we can adjust our sails • It's later than you think • Ack! Thbbft!
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

Acorns

I think Acorns (https://www.acorns.com/) deserves a review because they are targeting a customer the typical Boglehead knows but doesn't know how to help. This is a robo-advisor aimed at poor people or people who are bad at saving and irresponsible with their money. The core of the investing program is that credit cards are linked to Acorns. When you make a purchase Acorns rounds up the next dollar and puts the difference in escrow (so a $12.61 purchase is charged $13 putting $.39 in escrow). When the escrow hits $5 it is all invested. The other way money is entered into the account is their promotional partnerships. Their mobile app is a mobile shopping application. Vendors offer promotions with often quite large coupons (buy breakfast at Starbucks, get 30% of the price back to your Acorns account). They also offer automated or manual withdrawals from most bank accounts and these can be timed for when the paycheck hits (so some money is saved before it is spent).

Acorns is free for college students for their first 4 years from signing up. After that, and for anyone else they charge $1/mo or for accounts under $5k and 25 basis points for those above $5k.

The portfolios are a mixture of low cost ETFs from Vanguard, Blackrock and Pimco keeping underlying fees in the 5-15 basis point range. The portfolios themselves were developed by Harry Markowitz (the large REIT component demonstrates his belief in correlation diversification), and investments are overseen by Christopher Jones from BlackRock. So big names are helping out here. They have 5 degrees of aggressiveness from conservative to aggressive with no other variations. They involve the same 7 ETFs

USA large: 15-20% (VOO, Vanguard S&P 500)
USA small: 15-35% (VB, Vanguard Small Cap ETF)
EM: 0-15% (VWO, Vanguard Emerging Markets ETF)
REIT: 5-10% (VNQ, Vanguard REIT ETF)
International large: 5-15% (VEA, Vanguard FTSE Developed Markets ETF)
Government bond: 0-35% (SHY, Blackrock iShares 1-3 Year Treasury Bond ETF)
Corporate bond: 0-25% (CORP, PIMCO Investment Grade Corp Bond ETF)

with bonds being 60% of the conservative portfolio and 0% of the most aggressive. Acorns is a fractional share brokerage (obvious given the amounts).

They offer taxable only. There are competitors now who do the same thing. The most well known one is Stash Invest: https://www.stashinvest.com/ which does mostly the same thing with a slightly different portfolio and different coupon deals. Reviews of Acorns from normal investors have been positive. Their portfolio in head to head comparison has generally performed as well or better than most other robos. End user reviews are extremely positive for the phone application, the promotions and rounding up do really seem to work. Many are saving $500-2000 per year vs. the $0 they were saving / investing previously. Normal users were mostly extremely unhappy with only getting 10+% return per year, evidently they were expecting more like 100%+.

For this crowd I'm more OK with large growth that I would be with any other robo-advisor. Having mostly positive 3 year returns are probably much more important than maximizing total return. The $1/mo fee can as a percentage be quite high, but no one else is even interested in helping these sorts of customers at all. I have to assume this service is likely for the current period being operated at a loss, and the profits will come later as a fraction of their user base starts accumulating large accounts.

I think most Bogleheads know someone (or a lot of someones) who is the right target for this robo-advisor. If the feedback is typical they may have to spend some time explaining the importance of compounding but at least on a day to day basis Acorns will help get some money actually into savings and investment. After a decade on Acorns and suddenly discovering they actually have equity and not debt they probably won't need this service anymore. At least that's the hope.
peter3901
Posts: 12
Joined: Sat Aug 13, 2016 7:05 pm
Location: Chicago, IL

Re: What we know about robo advisors

Post by peter3901 »

jbolden1517 wrote:
siamond wrote:Well, I could easily add Betterment to the backtesting simulation, this is actually simpler than SIP. You're right, I could see Bogleheads folks going for a portfolio like that, except for the relative complexity of the bonds side of the equation.

This would be comparing a bit apples to oranges though, given Betterment's 90/10 AA vs. SIP 70/30 (if we assimilate cash to bonds, and commodities to stocks - at the macro level) AA. Is there a Betterment variation that would be closer to 70/30?
I could probably dig up a 70/30 for Betterment. You would just run it an force a 70/30 result by changing your risk tolerance. The only thing you are testing is the effect of the value premium over 3-fund which of course is going to be good. The rest of it is just a more complex version of 3-fund (the bonds are especially silly). FWIW on the comparison with SIP I agree with nedsaid the SIP 70/30 is taking substantially more risk on the bond side than 3-fund does. You might want to compare to both 70/30 and 80/20 3-fund.
Use this to find the Betterment portfolio at any stocks/bonds combination: https://www.betterment.com/resources/re ... tion-chart
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

peter3901 wrote:
jbolden1517 wrote:
siamond wrote:Well, I could easily add Betterment to the backtesting simulation, this is actually simpler than SIP. You're right, I could see Bogleheads folks going for a portfolio like that, except for the relative complexity of the bonds side of the equation.

This would be comparing a bit apples to oranges though, given Betterment's 90/10 AA vs. SIP 70/30 (if we assimilate cash to bonds, and commodities to stocks - at the macro level) AA. Is there a Betterment variation that would be closer to 70/30?
I could probably dig up a 70/30 for Betterment. You would just run it an force a 70/30 result by changing your risk tolerance. The only thing you are testing is the effect of the value premium over 3-fund which of course is going to be good. The rest of it is just a more complex version of 3-fund (the bonds are especially silly). FWIW on the comparison with SIP I agree with nedsaid the SIP 70/30 is taking substantially more risk on the bond side than 3-fund does. You might want to compare to both 70/30 and 80/20 3-fund.
Use this to find the Betterment portfolio at any stocks/bonds combination: https://www.betterment.com/resources/re ... tion-chart
Excellent link I'm moving it to the review.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

[Posts merged into here from: Backtesting some robo-advisers asset allocations, see below. --admin LadyGeek]
LadyGeek wrote:I think this discussion provides a necessary perspective in that robo-advisers are just that - advisers. Portfolio performance is no different for a robo-adviser than anyone else.

I've added a new section in the wiki to address this. See: Robo-adviser (Portfolio performance)

Comments / questions / concerns are welcome.
I'd be a bit hesitant about this line, "The selection process is no different with a robo-adviser portfolio than one based on a "conventional" portfolio from, for example, Fidelity, Schwab, or Vanguard." I'd revised to something a bit more nuanced like, "The selection process for those robo-advisors who employ a fixed allocation is no different with a robo-adviser portfolio than one based on a "conventional" portfolio from, for example, Fidelity, Schwab, or Vanguard. For those who employ a dynamic allocation the selection process will emulate a global tactical asset allocation mutual fund." Huygens, Hedgeable... will change allocations radically regularly. Both Motif and Coverster can.

I guess as long as we are discussing this. Should I do anything with robos that are only open to accredited investors? I'm thinking they deserve a post or two.
User avatar
LadyGeek
Site Admin
Posts: 95686
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: Backtesting some robo-advisers asset allocations

Post by LadyGeek »

^^^ The first step is to pick an asset allocation - which is the ability, willingness, and need to take risk. For example, 60% stocks / 40% bonds.

Are you saying that portfolios which implement Tactical asset allocation don't follow that process? If not, how would one select a portfolio?
jbolden1517 wrote:I guess as long as we are discussing this. Should I do anything with robos that are only open to accredited investors? I'm thinking they deserve a post or two.
You can post what you wish, but I think accredited investors might be down in the weeds in terms of the overall discussion.
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.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

LadyGeek wrote:^^^ The first step is to pick an asset allocation - which is the ability, willingness, and need to take risk. For example, 60% stocks / 40% bonds.

Are you saying that portfolios which implement Tactical asset allocation don't follow that process? If not, how would one select a portfolio?
A 60/40 stock bond portfolio gives you a standard deviation. Given 18 asset classes (USA stocks, commodities, limited partnership, foreign stock...) there is something on the order of (50)^18 possible portfolios. Each of those portfolios has a standard deviation and an expected return. By diluting with cash (or concentrating with leverage for the few who allow that) the standard deviation can be fixed to match a 60/40 portfolio. Once diluted that portfolio has an expected return. The system looks among all possible portfolios for the one with the highest expected return given current pricing / valuation, current correlations between assets, current sentiment (momentum)... and chooses the one with the highest expected return. Of course as the inputs change the optimal portfolio changes.

You get to determine your degree of risk. But you have no idea what specifically you will be holding. Now of course for taxable those changes need to be made more gradually than for tax free accounts. But the idea is the same.
jbolden1517 wrote:I guess as long as we are discussing this. Should I do anything with robos that are only open to accredited investors? I'm thinking they deserve a post or two.
You can post what you wish, but I think accredited investors might be down in the weeds in terms of the overall discussion.[/quote]

I thought so too. BTW did you notice the Acorns review?
User avatar
siamond
Posts: 6008
Joined: Mon May 28, 2012 5:50 am

Re: What we know about robo advisors

Post by siamond »

peter3901 wrote:
jbolden1517 wrote:I could probably dig up a 70/30 for Betterment. You would just run it an force a 70/30 result by changing your risk tolerance. The only thing you are testing is the effect of the value premium over 3-fund which of course is going to be good. The rest of it is just a more complex version of 3-fund (the bonds are especially silly). FWIW on the comparison with SIP I agree with nedsaid the SIP 70/30 is taking substantially more risk on the bond side than 3-fund does. You might want to compare to both 70/30 and 80/20 3-fund.
Use this to find the Betterment portfolio at any stocks/bonds combination: https://www.betterment.com/resources/re ... tion-chart
I followed the link, focused on the case of a 401k/IRA (otherwise, the system loads up on munis, and this seems a tad atypical to me). Here is the Asset Allocation that is suggested by Betterment. Strong (domestic) value tilts and high exposure to Int'l bonds.

Code: Select all

EM         6.30%
Int'l Dev 30.10%
US SCV     3.60%
US MCV     4.10%
US LCV    12.90%
TSM	    12.90%
EMB	     4.20%
Int'l Bd  10.30%
Corp Bd    5.20%
TBM	    10.40%
	
Total    100.00%
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

siamond wrote:
peter3901 wrote:
jbolden1517 wrote:I could probably dig up a 70/30 for Betterment. You would just run it an force a 70/30 result by changing your risk tolerance. The only thing you are testing is the effect of the value premium over 3-fund which of course is going to be good. The rest of it is just a more complex version of 3-fund (the bonds are especially silly). FWIW on the comparison with SIP I agree with nedsaid the SIP 70/30 is taking substantially more risk on the bond side than 3-fund does. You might want to compare to both 70/30 and 80/20 3-fund.
Use this to find the Betterment portfolio at any stocks/bonds combination: https://www.betterment.com/resources/re ... tion-chart
I followed the link, focused on the case of a 401k/IRA (otherwise, the system loads up on munis, and this seems a tad atypical to me). Here is the Asset Allocation that is suggested by Betterment. Strong (domestic) value tilts and high exposure to Int'l bonds.

Code: Select all

EM         6.30%
Int'l Dev 30.10%
US SCV     3.60%
US MCV     4.10%
US LCV    12.90%
TSM	    12.90%
EMB	     4.20%
Int'l Bd  10.30%
Corp Bd    5.20%
TBM	    10.40%
	
Total    100.00%
That looks right to me. 1/2 value 1/2 CAP domestic and index foreign. BTW don't know if they gave you the funds but I list them in the review. Mostly Vanguard so this one should be very easy.
4nwestsaylng
Posts: 516
Joined: Thu Jun 15, 2017 2:03 am

Re: What we know about robo advisors

Post by 4nwestsaylng »

Very interested that you all, especially jbolden, are looking at the robo advisors, backtesting, etc..I decided to open an SIP account as I was already with Schwab and easy to transfer funds, but will be following this thread. Concerned about possibiity that there is "substantial risk" in the SIP bonds vs 3 fund, I will have to look at those bond ETFs more carefully, but at this point, since I have Vanguard ST and Int tax exempt, the SIP bond is not a large portion of my AA,however it bears watching.

Whether I stay with SIP long term or go to another roboadvisor, I have at least crossed the threshold for a portion of my AA to be in a robo fund, so will follow your analyses and really appreciate that the subject is getting attention. It may be that in the end, the 3 fund is equal or better, but the comments about how the market is now savvy to the huge index influence, keeping a fixed proportion of a company in the index, was something I did not consider
Also wonder if the fundamental indexes may be ultimately a more effective counter to this than cap weighted indexes. Not clear to me how the current cap weighted indexes end up with index behavior that maintains a fixed proportion of a company in the index. You would think if the market cap of a company drops, the proportion of that company in the index would also drop. Maybe this could be clarified.
User avatar
siamond
Posts: 6008
Joined: Mon May 28, 2012 5:50 am

Re: What we know about robo advisors

Post by siamond »

jbolden1517 wrote:That looks right to me. 1/2 value 1/2 CAP domestic and index foreign. BTW don't know if they gave you the funds but I list them in the review. Mostly Vanguard so this one should be very easy.
Yes, it is indeed very easy. I just gave it a quick run, and this appears to be a fairly simple (for a robo-adviser!) and fairly effective allocation (at least, it would have been in the 1985-2016 time period). I didn't factor their additional ER yet though. Will let the discussion go on about SIP tonight on the other thread, then contribute a detailed Betterment analysis tomorrow.
User avatar
LadyGeek
Site Admin
Posts: 95686
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: What we know about robo advisors

Post by LadyGeek »

On an administrative note, the discussion in Backtesting some robo-advisers asset allocations was losing focus. I moved posts discussing asset allocation and other topics not related to backtesting into here.
jbolden1517 wrote:I thought so too. BTW did you notice the Acorns review?
Yes, I just got around to it. See: Robo-adviser (Member reviews).
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
LadyGeek
Site Admin
Posts: 95686
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: Backtesting some robo-advisers asset allocations

Post by LadyGeek »

jbolden1517 wrote:
LadyGeek wrote:^^^ The first step is to pick an asset allocation - which is the ability, willingness, and need to take risk. For example, 60% stocks / 40% bonds.

Are you saying that portfolios which implement Tactical asset allocation don't follow that process? If not, how would one select a portfolio?
A 60/40 stock bond portfolio gives you a standard deviation. Given 18 asset classes (USA stocks, commodities, limited partnership, foreign stock...) there is something on the order of (50)^18 possible portfolios. Each of those portfolios has a standard deviation and an expected return. By diluting with cash (or concentrating with leverage for the few who allow that) the standard deviation can be fixed to match a 60/40 portfolio. Once diluted that portfolio has an expected return. The system looks among all possible portfolios for the one with the highest expected return given current pricing / valuation, current correlations between assets, current sentiment (momentum)... and chooses the one with the highest expected return. Of course as the inputs change the optimal portfolio changes.
I understand what you mean, but I think you're missing the "big picture".

This article has a more comprehensive perspective: Robo-Adviser Asset Allocation and Securities, by AAII (American Association of Individual Investors), November 19, 2016. Backtesting is discussed. Be sure to read the comments section.

From the conclusion:
Most robo-advisers aren’t designed to “beat the market,” so if that’s your goal you will likely be disappointed.
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
siamond
Posts: 6008
Joined: Mon May 28, 2012 5:50 am

Re: Backtesting some robo-advisers asset allocations

Post by siamond »

LadyGeek wrote:From the conclusion:
Most robo-advisers aren’t designed to “beat the market,” so if that’s your goal you will likely be disappointed.
That certainly makes a lot of sense. If the various hedge funds and quants at large never succeeded to design algorithms that reliably and sustainably make a big difference, there is no reason those robo-advisers would do any better.

For me, the point of backtesting the recommendations of a few of them is primarily to check that the 'robo-advised' AA isn't silly and is at least decent, and then to develop a better understanding of those engines, so that we can make recommendations about them in a reasonably factual basis.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: Backtesting some robo-advisers asset allocations

Post by jbolden1517 »

LadyGeek wrote:
jbolden1517 wrote:
LadyGeek wrote:^^^ The first step is to pick an asset allocation - which is the ability, willingness, and need to take risk. For example, 60% stocks / 40% bonds.

Are you saying that portfolios which implement Tactical asset allocation don't follow that process? If not, how would one select a portfolio?
A 60/40 stock bond portfolio gives you a standard deviation. Given 18 asset classes (USA stocks, commodities, limited partnership, foreign stock...) there is something on the order of (50)^18 possible portfolios. Each of those portfolios has a standard deviation and an expected return. By diluting with cash (or concentrating with leverage for the few who allow that) the standard deviation can be fixed to match a 60/40 portfolio. Once diluted that portfolio has an expected return. The system looks among all possible portfolios for the one with the highest expected return given current pricing / valuation, current correlations between assets, current sentiment (momentum)... and chooses the one with the highest expected return. Of course as the inputs change the optimal portfolio changes.
I understand what you mean, but I think you're missing the "big picture".

This article has a more comprehensive perspective: Robo-Adviser Asset Allocation and Securities, by AAII (American Association of Individual Investors), November 19, 2016. Backtesting is discussed. Be sure to read the comments section.

From the conclusion:
Most robo-advisers aren’t designed to “beat the market,” so if that’s your goal you will likely be disappointed.
Right but most robo-advisors use a static asset allocation not a tactical one. Those may be aiming to beat the market through tilting (SIP being a good example reviewed). But there are robo-advisors that do use tactical allocations: Hedgeable being the most obvious example. Huygens does as well. I should mention one of the ones they cover in that article that I haven't mentioned: Alpha Architect tries to beat the market, as their name implies (my understanding is so far it has not worked out that way). They clearly state their 4 means to beat the market are: value investing portfolio, momentum, downside protection (tactical allocation) and active management.

That article gives a good example of tactical allocation. Note the two portfolios Hedgeable has the taxable and non-taxable accounts in. For taxable it is a non-diversified USA stock fund. Clearly aiming for market timing (though they don't call it that) to capture the gains and jump in time:

Code: Select all

Med/High Risk - Taxable
Cash	2.0%
Currencies	2.0%
Emerging Mkt Stocks	2.0%
Fixed Income	5.6%
Int'l Stocks	1.9%
U.S. Stocks	86.5%
Conversely the retirement allocation has already moved out of USA stocks since the tax implications are lower and is now extremely defensive waiting for a good chance to get back in:

Code: Select all

Med/High Risk - Retirement
Cash	2.0%
Emerging Mkt Fixed Inc	7.5%
Fixed Income	48.3%
MLP	6.0%
Real Estate	9.9%
U.S. Stocks	26.3%
Many of the strategies from Covester similarly are "market beaters". For example if you select the "Global Downside Protected" strategy. Then the objectives are:
1) Invest Globally:
-Diversification across geographic markets provides the opportunity to benefit when there are attractive markets outside the home country.

2) Create alpha from beta management:
-Return of any equity market index (the equity beta) can be achieved dynamically and cheaply via ETFs or futures
-Additional return (alpha) can be generated from systematic beta management – informed choices about which stock markets to own-Generating alpha from beta management is often under-appreciated and under-utilized by investors

3)Take risk only when rewarded:
-Determine “bad neighborhoods”: Examine valuation, momentum, fundamentals, sentiment, risk and currency impacts for each market
-"Water the flowers, not weeds”: Spread the risk budget to good neighborhoods only
-“Correlation matters”: When too many neighborhoods turn bad, we will move assets to short-term fixed income to avoid losses for investors.
(bold mine)

There is today as much diversity among robo-advisors as among human advisors. Your line about a robo-advisor being like an automated human advisor is correct. Betterment and Wealthfront both had static allocations so robos started with very much a CRAAL philosophy but the situation has changed. And seems to be continuing to change faster. So I would disagree strongly that robos overall aren't trying to beat the market. Some are, many aren't.
User avatar
Ethelred
Posts: 384
Joined: Sun Oct 30, 2016 9:38 am

Re: What we know about robo advisors

Post by Ethelred »

So, I went onto the Schwab SIP to investigate. More for interest than anything else. A few things I found interesting:
- One significant disadvantage I see is that you only open one type of account, meaning that they cannot help with asset location (taxable vs. traditional IRA/401k vs Roth). Without this, it's difficult to see any robo advisor account being fully automated, unless you have no taxable investments. It loooks like Betterment offer this, called "Tax-Coordinated Portfolio (tm)", but I can't see any sign of it for Schwab or Wealthfront.

- As an extension of the point above, many investors, probably the majority of those still employed, will be required to keep their retirement accounts with their employer's provider. For those investors, fully automating asset allocation requires the robo to somehow work with the 401k asset allocation. I'm not sure this is ever likely to happen.

- I'm sure there is more nuance to what they do with it, but the required cash position reminds me of all those HSA accounts that don't charge administration fees, but instead require you to hold a minimum cash balance in order to invest in mutual funds, often about $2500.

- I ran my details through their questionnaire to generate an asset allocation, and I was surprised by how conservative it was. This is what it produced:

Stocks 61%
US Large Company Stocks - Fundamental 11%
US Large Company Stocks 8%
US Small Company Stocks - Fundamental 7%
International Developed Large Company Stocks - Fundamental 7%
International Developed Large Company Stocks 5%
US Small Company Stocks 4%
International Developed Small Company Stocks - Fundamental 4%
International Emerging Market Stocks - Fundamental 4%
International Developed Small Company Stocks 3%
International Emerging Market Stocks 3%
US Exchange-Traded REITs 3%
International Exchange-Traded REITs 2%
Fixed Income 23.5%
US Corporate High Yield Bonds 8%
Investment Grade Municipal Bonds 7.5%
International Emerging Market Bonds 5%
International Developed Country Bonds 3%
Commodities 5%
Gold and Other Precious Metals 5%
Cash 10.5%
Cash 10.5%

- 61% stock surprised me a little, given that I'm currently at 70%. The main inputs were age 42, retire in 10 years, with high but not very high risk tolerance.
- Why is it recommending pairs of ETFs with both fundamental and market cap weights? This seems odd.
- Notice the cash percentage is higher than the more aggressive asset allocation that was back-tested. If we assume that the cash position makes them more profit than the ETF fees, then Schwab make more money from conservative investors.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

Ethelred wrote:So, I went onto the Schwab SIP to investigate. More for interest than anything else. A few things I found interesting:
- One significant disadvantage I see is that you only open one type of account, meaning that they cannot help with asset location (taxable vs. traditional IRA/401k vs Roth).
I'm not sure what you mean. They all offer taxable and tax exempt. Schwab handles more advanced accounts like trusts, annuities... than most.
Ethelred wrote: Without this, it's difficult to see any robo advisor account being fully automated, unless you have no taxable investments. It loooks like Betterment offer this, called "Tax-Coordinated Portfolio (tm)", but I can't see any sign of it for Schwab or Wealthfront.
So you mean hold more bonds in your tax-exempt and less in your taxable?
Ethelred wrote: - As an extension of the point above, many investors, probably the majority of those still employed, will be required to keep their retirement accounts with their employer's provider. For those investors, fully automating asset allocation requires the robo to somehow work with the 401k asset allocation. I'm not sure this is ever likely to happen.
They are working on it. The 401k allocation they can assume is just a normal retirement allocation and they just pick a normal taxable and... mostly it works out ok. The real complex problem is investors with a huge percentage in company stocks. Wealthfront and Hedgeable have good programs for that. They have to know the company though to know how to build a portfolio that anti-correlates with the company and then do selling. Hedgeable creates lots of paired trades that create synthetic losses. Wealthfront gradually sells. https://www.wealthfront.com/selling-plan
Ethelred wrote: - I'm sure there is more nuance to what they do with it, but the required cash position reminds me of all those HSA accounts that don't charge administration fees, but instead require you to hold a minimum cash balance in order to invest in mutual funds, often about $2500.
Nope your analogy is correct, same idea.
Ethelred wrote:
- 61% stock surprised me a little, given that I'm currently at 70%. The main inputs were age 42, retire in 10 years, with high but not very high risk tolerance.
- Why is it recommending pairs of ETFs with both fundamental and market cap weights? This seems odd.
- Notice the cash percentage is higher than the more aggressive asset allocation that was back-tested. If we assume that the cash position makes them more profit than the ETF fees, then Schwab make more money from conservative investors.
61 and 70 are pretty close. I guess my standards for automated advice are lower given how badly many naive investors do on their own. Glass half empty / half full.

The fundamental are value non cap weighted (sales, cashflow and dividend weighted). So for each stock asset class they are holding 2/3rds value weighted, 1/3 cap weighted. They are going for a strong value tilt with some growth exposure.
User avatar
Ethelred
Posts: 384
Joined: Sun Oct 30, 2016 9:38 am

Re: What we know about robo advisors

Post by Ethelred »

jbolden1517 wrote:
Ethelred wrote:So, I went onto the Schwab SIP to investigate. More for interest than anything else. A few things I found interesting:
- One significant disadvantage I see is that you only open one type of account, meaning that they cannot help with asset location (taxable vs. traditional IRA/401k vs Roth).
I'm not sure what you mean. They all offer taxable and tax exempt. Schwab handles more advanced accounts like trusts, annuities... than most.
Ethelred wrote: Without this, it's difficult to see any robo advisor account being fully automated, unless you have no taxable investments. It loooks like Betterment offer this, called "Tax-Coordinated Portfolio (tm)", but I can't see any sign of it for Schwab or Wealthfront.
So you mean hold more bonds in your tax-exempt and less in your taxable?
Ethelred wrote: - As an extension of the point above, many investors, probably the majority of those still employed, will be required to keep their retirement accounts with their employer's provider. For those investors, fully automating asset allocation requires the robo to somehow work with the 401k asset allocation. I'm not sure this is ever likely to happen.
They are working on it. The 401k allocation they can assume is just a normal retirement allocation and they just pick a normal taxable and... mostly it works out ok. The real complex problem is investors with a huge percentage in company stocks. Wealthfront and Hedgeable have good programs for that. They have to know the company though to know how to build a portfolio that anti-correlates with the company and then do selling. Hedgeable creates lots of paired trades that create synthetic losses. Wealthfront gradually sells. https://www.wealthfront.com/selling-plan
Yes, I am talking about optimizing asset location, not asset allocation. Often that means more bonds in tax-advantaged accounts, but doesn't always.

This is the Bogleheads wiki page that talks about tax-efficient fund placement (essentially the same thing): https://www.bogleheads.org/wiki/Tax-eff ... _placement

This is the Betterment page that talks about their tax-efficient option: https://www.betterment.com/tax-coordinated-portfolio/

So, yes, you can simply keep the same asset allocation in both your taxable and tax-advantaged accounts, but that is not optimal for taxes. My first point is that few of the robos seem to be set up to solve this problem. My second point is that solving this problem requires the robo advisor having access to retirement accounts, which is impractical for many employed investors.

But I do agree that company restricted stock and options are potentially a larger problem for those investors who have them.
The fundamental are value non cap weighted (sales, cashflow and dividend weighted). So for each stock asset class they are holding 2/3rds value weighted, 1/3 cap weighted. They are going for a strong value tilt with some growth exposure.
That makes sense, value versus blend. I'd missed that the fundamental index is value-based. Thanks.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

Ethelred wrote: Yes, I am talking about optimizing asset location, not asset allocation. Often that means more bonds in tax-advantaged accounts, but doesn't always.[

This is the Bogleheads wiki page that talks about tax-efficient fund placement (essentially the same thing): https://www.bogleheads.org/wiki/Tax-eff ... _placement
I get that. Think of mainstream robos as replacing a pretty good but not great financial advisor who tend to put all their clients in pretty similar asset allocations. Like Vanguard's FA.

Ultimately you can't just treat this as a static asset allocation because you can't rebalance between taxable and non taxable money. For purposes of management they are two portfolios. I'm a big believer in determining your asset allocation you count everything: home as a non-diversified real estate investment, mortgage as a short bond position, pension as an inflation (or not) bond position, social security as an annuity... And that's how you would want to determine your investment asset allocation, which then gets divided into to your various accounts.

But for this customer base you assume someone who can't or won't do that. Most robos are going to give them a more tax efficient portfolio that's fairly close to what you want in the IRA, and less tax efficient in the taxable money. It won't be an ideal asset allocation. I agree. At a certain point, yes a financial planner or wealth manager is needed or you just need to know what you are doing.

But I certainly think they seem to be comparable to most FAs.
Ethelred wrote: So, yes, you can simply keep the same asset allocation in both your taxable and tax-advantaged accounts, but that is not optimal for taxes.
It won't be the same. Assume you have Alan the investor who has risk tolerance X. He has 1/2 his money for retirement taxable and 1/2 his money for retirement non-taxable. The after tax returns on bonds vs. high dividend stocks vs. international vs.... change in the computation. Which means when the system optimizes it is going to naturally end up with more bonds in the retirement and less in the taxable. Capital gains taxes matter which means it is going to assign a lower weight to funds that incur high capital gains vs. those that don't. Etc...

And this works regardless of how complex you make it. If Bill wants his taxable half for a 10 year expense and 1/2 for a retirement portfolio. The robo builds him an allocation (XT) which is basically the some of two portfolios a (T10) ten year portfolio and a (TR) taxable retirement portfolio.

Remember in general the risk tolerance not the asset allocation is what is being fixed here. With the more mainstream robos that feels like the same thing, but with the less mainstream ones the portfolios can swing wildly while having the same risk tolerance.
Ethelred wrote: My first point is that few of the robos seem to be set up to solve this problem. My second point is that solving this problem requires the robo advisor having access to retirement accounts, which is impractical for many employed investors.
How much harm to you think it does to have two portfolios. Betterment is getting into 401ks. Merrill is already in them.

So let's assume: Carl has a 401K with Betterment. His wife Diane has her 401K with Merrill Guided. They have an HSA invested with Blackrock global asset allocation. And their taxable money is with Schwab.

We have Ed who has a so/so financial advisor handling his 401K, his wife's, their HSA and their taxable. How much better does Ed do?
Avo
Posts: 1099
Joined: Wed Jun 11, 2008 2:21 am
Location: California

Re: What we know about robo advisors

Post by Avo »

Depends on the fees of the advisor, and what "so-so" means in practice.

If Ed's advisor is Vanguard, and guides Ed into tax-efficient asset location, and after a few years Ed figures out he can do all this himself (because it's really not that hard to implement a Vanguard portfolio yourself) and stops paying the advisor fees for the next 30 years ...

While Carl's asset location is tax inefficient, and Carl is getting very different portfolios from his four different robos and worries that some of them are bad, and feels confused as to whether his overall strategy is right or not, and really wishes he had someone he could talk it over with, but doesn't, and the market looks frothy, so maybe he should go all cash now ...
User avatar
Ethelred
Posts: 384
Joined: Sun Oct 30, 2016 9:38 am

Re: What we know about robo advisors

Post by Ethelred »

jbolden1517 wrote:I get that. Think of mainstream robos as replacing a pretty good but not great financial advisor who tend to put all their clients in pretty similar asset allocations. Like Vanguard's FA.

Ultimately you can't just treat this as a static asset allocation because you can't rebalance between taxable and non taxable money. For purposes of management they are two portfolios. I'm a big believer in determining your asset allocation you count everything: home as a non-diversified real estate investment, mortgage as a short bond position, pension as an inflation (or not) bond position, social security as an annuity... And that's how you would want to determine your investment asset allocation, which then gets divided into to your various accounts.

But for this customer base you assume someone who can't or won't do that. Most robos are going to give them a more tax efficient portfolio that's fairly close to what you want in the IRA, and less tax efficient in the taxable money. It won't be an ideal asset allocation. I agree. At a certain point, yes a financial planner or wealth manager is needed or you just need to know what you are doing.

But I certainly think they seem to be comparable to most FAs.
I feel like we're not disagreeing, but we're talking past each other. I realize that you can't rebalance directly between taxable and tax-advantaged accounts, but you can manage asset location in a way that is more tax-efficient.
It won't be the same. Assume you have Alan the investor who has risk tolerance X. He has 1/2 his money for retirement taxable and 1/2 his money for retirement non-taxable. The after tax returns on bonds vs. high dividend stocks vs. international vs.... change in the computation. Which means when the system optimizes it is going to naturally end up with more bonds in the retirement and less in the taxable. Capital gains taxes matter which means it is going to assign a lower weight to funds that incur high capital gains vs. those that don't. Etc...

And this works regardless of how complex you make it. If Bill wants his taxable half for a 10 year expense and 1/2 for a retirement portfolio. The robo builds him an allocation (XT) which is basically the some of two portfolios a (T10) ten year portfolio and a (TR) taxable retirement portfolio.

Remember in general the risk tolerance not the asset allocation is what is being fixed here. With the more mainstream robos that feels like the same thing, but with the less mainstream ones the portfolios can swing wildly while having the same risk tolerance.
What's interesting here, now I look at the Schwab website, is that your thinking precisely matches theirs. Their solution for optimizing tax-efficiency through asset location for SIP is part of their Schwab Intelligent Advisory service. For a fee of 0.28%, a Schwab CFP will discuss the investor's needs and then recommend separate portfolios for each separate SIP account, taxable or tax-advantaged. At the annual check-in, the advisor could recommend changes in details of these separate portfolios. However, the robo advisor is not involved in asset location, so there is no real-time adjustment between the two.

But it looks like Betterment is going beyond this, and involving the robo advisor in asset location decisions. Obviously this requires your 401k to be held with them to work. This is the blog post where they explain how it works in more detail: https://www.betterment.com/resources/in ... portfolio/

This is the sort of service I was talking about.
How much harm to you think it does to have two portfolios. Betterment is getting into 401ks. Merrill is already in them.

So let's assume: Carl has a 401K with Betterment. His wife Diane has her 401K with Merrill Guided. They have an HSA invested with Blackrock global asset allocation. And their taxable money is with Schwab.

We have Ed who has a so/so financial advisor handling his 401K, his wife's, their HSA and their taxable. How much better does Ed do?
This question that has so many unknowns as to be akin to "how long is a piece of string". For a knowledgeable investor, self-guided (or maybe robo) is better. For a confused investor, the financial advisor may add enough value to justify their fees. The likelihood of an employed couple having four separate accounts, all of which are individually managed (robo or in person), is small.

If the point you are trying to make is that the value of asset location decisions are exceeded by any fees involved, then it depends. Betterment's blog post claims that in the mid case they tested the increase is 0.48%. It looks like their fees for this service are zero, I'd assume because the value to them is that you have to hold both your taxable and tax-advantaged accounts with them to make it work.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

Avo wrote:Depends on the fees of the advisor, and what "so-so" means in practice.

If Ed's advisor is Vanguard, and guides Ed into tax-efficient asset location, and after a few years Ed figures out he can do all this himself (because it's really not that hard to implement a Vanguard portfolio yourself) and stops paying the advisor fees for the next 30 years ...
No question. 35 basis points * 30 years = 10%. It is better if he does it himself. That being said if you believe in MPT (which has a lot of evidence) the Vanguard portfolio underperforms most of these robos by way more than 35 basis points.
Avo wrote: While Carl's asset location is tax inefficient, and Carl is getting very different portfolios from his four different robos and worries that some of them are bad, and feels confused as to whether his overall strategy is right or not, and really wishes he had someone he could talk it over with, but doesn't, and the market looks frothy, so maybe he should go all cash now ...
I get your point. The robos seem to be working well for a group of millennials. But when I was reading the Acorn comments and realizing how wildly off base their customers were, in terms of market expectations it was kinda scary. That being said, you can look at the glass half full and say they are taking customers this wildly off base and putting them in a sensible portfolio.

Anyway most of the robos have someone you can talk to. Just to go with the scenario. Diane (Carl's wife) is with Merrill Guided, she can talk it over with someone for free, that's included in the 45 basis points. Of course if Carl and Diane qualify for Merrill Lynch that may not be to her advantage :) Betterment has a cheap upgrade for a one time a year conversation and a very reasonable one for unlimited. Schwab has a more expensive upgrade for their internal team plus one of the best external teams around. Blackrock I'm not sure what happens. They are mostly focused on advisor only robos (which I haven't covered yet) and these kinds of services for the Germans (https://de.scalable.capital/ ).
Avo
Posts: 1099
Joined: Wed Jun 11, 2008 2:21 am
Location: California

Re: What we know about robo advisors

Post by Avo »

jbolden1517 wrote:That being said if you believe in MPT (which has a lot of evidence) the Vanguard portfolio underperforms most of these robos by way more than 35 basis points.
MPT is 65 years old. Where are the mutual funds that have out-performed by running an MPT strategy for the past, say, 20 years?
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

Avo wrote:
jbolden1517 wrote:That being said if you believe in MPT (which has a lot of evidence) the Vanguard portfolio underperforms most of these robos by way more than 35 basis points.
MPT is 65 years old. Where are the mutual funds that have out-performed by running an MPT strategy for the past, say, 20 years?
I'm not sure what you mean by "out performed". Out performed what? From my comment I'd assume you mean the Vanguard portfolio. But those are portfolios not funds. There are mutual funds that are portfolios but most are not.

In any case I'll just refer you to the thread where we looked at SIP and various value tilted portfolios vs. 3-fund (and that wasn't even including smart beta advantages), you will see a pretty consistent advantage especially if you remove the cash drag. There was a classic study done in 2000 which Larry used to like to talk about (so I'm assuming the people here are familiar). Of the 198 mutual funds which had survived to 2000 vs. the SP500.
Between 1971 to 1997 I found that out of the 198 SURVIVING mutual funds about 20% beat the SP500
+1% - 39 Funds (20%)
+2% - 24 Funds (12%)
+3% - 15 Funds (08%)
+4% - 06 Funds (03%)
+5% - 02 Funds(01%)
+6% - 01 Fund (less than 1%)


There are funds that beat the index, the number is not 0. I'm not a mutual fund guru. You would have found more of those on this board's grandfather.

If you want data more current and has actual historical record, I'll pick a good candidate that has nifty tools as part of their marketing. Blackrock global. There is no 10 year period where it has underperformed global stocks: https://www.blackrock.com/tools/global-allocation/chart . And for most shorter periods it has beaten them as well. You can play around with various portfolios for risk and return and see how much it helps all sorts of portfolios: https://www.blackrock.com/tools/global- ... /challenge. BTW that's with a 1% ER dragging down returns.

Now you might say they are using TAA not simply CRAAL. But then you can think of "blackrock global" as an asset and see how adding additional assets diversify in and boost whatever portfolio you had in mind using the same tool. But regardless the tool will work in demonstrating how clear cut portfolio design is in boosting return while holding risk constant.

Play around with the allocation tool very cool.
Avo
Posts: 1099
Joined: Wed Jun 11, 2008 2:21 am
Location: California

Re: What we know about robo advisors

Post by Avo »

As of today, MDLOX (Blackrock Global Allocation, A Shares) has slightly underperformed VSMGX (Vanguard LIfe Strategy Moderate Growth, a classic 60/40 split, currently 40% intl stocks) over the past 10 years, and that's before the 5.25% load on MDLOX. Volatility looks a little higher on the Vanguard fund, and it did worse in the crash (-36% from 8/4/07 to the bottom vs -25%).

MDLOX is not something I would pay ongoing fees for.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

Avo wrote:As of today, MDLOX (Blackrock Global Allocation, A Shares) has slightly underperformed VSMGX (Vanguard LIfe Strategy Moderate Growth, a classic 60/40 split, currently 40% intl stocks) over the past 10 years, and that's before the 5.25% load on MDLOX. Volatility looks a little higher on the Vanguard fund, and it did worse in the crash (-36% from 8/4/07 to the bottom vs -25%).

MDLOX is not something I would pay ongoing fees for.

MDLOX has a heavier international. Its a global allocation fund. The USA recovered much more than international did after 2008. But I'll use that allocation to prove the point about MPT, which was the topic not whether you like Blackrock Global.

A 40% core bond / 35% USA stock / 25% international stock over the last 29 years (the life of the Blackrock fund)
Annualized Return: 6.52% Risk (Standard Deviation): 8.96% Growth of $10k: $48,484.52 Max Drawdown: -37.70% Sharpe Ratio: 0.43

Blackrock fund:
Annualized Return: 9.26% Risk (Standard Deviation): 9.71% Growth of $10k: $91,451.58 Max Drawdown: -29.37% Sharpe Ratio: 0.68

So Blackrock is riskier (in standard deviation terms) and has better returns. Funds beat each other all the time. But that's not what's interesting in terms of MPT (which if you remember was the question). What is interesting and proves the point about diversification (MPT) nicely is if you take the the above fund (your choice BTW) and add 20% Blackrock Global to it (a diversifying asset):

20% Blackrock + 80% VSMGX
Annualized Return: 7.07% Risk (Standard Deviation): 8.97% Growth of $10k: $55,232.39 Max Drawdown: -36.04% Sharpe Ratio: 0.49

That portfolio of 20% Blackrock + 80% VSMGX has exactly the same risk as 100% VSMGX. It has a slightly smaller max drawdown. But the return is 55 basis points higher. You just got better return for no extra risk. And that is what MPT promises that through diversification you can increase return without needing to take on extra risk.

The same argument you all make against individual stocks applies to asset classes.

___

As an aside regarding the snide comments about Blackrock. The load is quite often waived (for example I didn't pay it). Most people get it in a 401K or as part of an FA relationship. The load exists as an option for those brokers who want it. Also the Blackrock fund is riskier (higher SD) but it had a lower max drawdown (in 2008) even compared to 60/40. That's because of the TAA component. That's part of what cut the return the last 10 years. The fund sat out the start of the bull in exchange for having missed quite a bit of the bear. The start of the bull had great returns. TAA is not in and of itself a good strategy. But it is extremely effective at diversifying CRAAL portfolios because it handles well there biggest problem severe drawdowns. This fact that TAA goes well with CRAAL is why Vanguard originally had a 25% TAA component in their life strategy funds. They did it because it was the right thing to do for their customers. The problem politically and ideologically for Vanguard is that they are opposed to precisely the sort of market timing that good TAA funds engage in. Given the popularity of the Life Strategy funds they couldn't for political/idealogical reasons have the TAA fund actually do the market timing it was supposed to be doing. So it was dropped.
Avo
Posts: 1099
Joined: Wed Jun 11, 2008 2:21 am
Location: California

Re: What we know about robo advisors

Post by Avo »

jbolden1517 wrote:over the last 29 years (the life of the Blackrock fund)
Way too long. The fund outperformed spectacularly early on, with low assets. Over the past ten years, though, it significantly trails the plain-vanilla M* 60/40 global index, CAGR 4.8% for MDLOX vs 5.7% for the index.

This is a common phenomenon in actively managed funds.

I don't dispute the basic point of MPT, that adding uncorrelated assets reduces portfolio volatility. What I dispute is that additional active management of the sort that you champion is worth the cost and complexity.
User avatar
siamond
Posts: 6008
Joined: Mon May 28, 2012 5:50 am

Re: What we know about robo advisors

Post by siamond »

I have a question about robo-advisers. In a way, they act as advisors. In another way, one could view them as acting a fund (of funds) manager. The latter could be troublesome though.

Say I am a newbie investor, I like the hand-holding of an advisor and (foolish me!) defer to his/her expertise, but I don't like the cost, so I go with a robo-adviser as a middle ground. Ten years later, I get closer to retirement, I wake up, read all about passive investing and decide to take things in my own hands. How does that work, notably for the assets located in taxable accounts? In other words, can I easily 'fire' the robo-adviser, and take full control?
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

Avo wrote:
jbolden1517 wrote:over the last 29 years (the life of the Blackrock fund)
Way too long. The fund outperformed spectacularly early on, with low assets. Over the past ten years, though, it significantly trails the plain-vanilla M* 60/40 global index, CAGR 4.8% for MDLOX vs 5.7% for the index.

This is a common phenomenon in actively managed funds.

I don't dispute the basic point of MPT, that adding uncorrelated assets reduces portfolio volatility. What I dispute is that additional active management of the sort that you champion is worth the cost and complexity.
I don't know how you can say you weren't disputing MPT, "MPT is 65 years old. Where are the mutual funds that have out-performed by running an MPT strategy for the past, say, 20 years?" Either you believe MPT is false or obviously mutual funds using MPT have out-performed. Lots of mutual funds have outperformed their indexes.

I don't know where you are getting these numbers, are you sure you are catching distributions (dividends). You might also be using a calculator that's subtracting off the load. The global stock index over the last 10 years returned 54.2%. Blackrock returned 60.8%. This BTW agrees with Morningstar which has 10k growing to 16,478. http://beta.morningstar.com/funds/xnas/malox/quote.html (my ER is a little higher but this way it doesn't include load. More importantly the volatility on global stock was 16.3% while Blackrock was 10.3%. So Blackrock had the returns of a 110/-10 portfoloio while having the volatility of a 70/30 portfolio. Right now it is only 55% stock with 45% commodities, bonds, cash...

MPT doesn't require active management. This whole thread is about passive vehicles (a few of the robos like Alpha Architect have human intervention). The backtesting thread (viewtopic.php?f=10&t=224764&start=50) shows those more complicated portfolios, using nothing but index funds crushing 3-fund. What I'm championing broadly is dealing with reality.

a) Valuations matter. Buying stocks at too high prices lead to bad returns. Buying stocks at high prices passively doesn't change that. Similarly for bonds.
b) No divine prices. Stock price is a single number. It can at best only be the solution to one possible valuation equation. So if you hypothesize the price of assets are set perfectly with respect to one variable it must be set imperfectly with respect to all others.
c) History matters. We have 1000 years of investing history. The last 10 in the USA are not representative of the only possible market conditions.
d) Risk matters so diversification matters. Diversification means holding non correlating assets in quantities large enough to diversify away risk. Not holding large numbers of assets in such small quantities that they have little or no impact. Hence cap weighting does not guarantee diversification.
e) Costs matter. This one generally goes over well here when it comes to ERs.
f) Indexers mutual funds are market participants just like any other mutual fund. Other players in the market are not idiots and can respond to your strategies with counter strategies designed to exploit your inefficiencies. Their counter strategies are a serious threat.
g) No more propaganda. Be honest with yourselves about the strengths and weaknesses of your approach. Truthfully and accurately consider other approaches.
h) The USA represents post facto selection bias. The TSM argument is no different then any other "hot fund" type argument.
j) Math matters. Stop making up fake equations and numbers.
k) Depletion matters. Income investing is very different than accumulation investing.

I don't think those are extreme views.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

siamond wrote:I have a question about robo-advisers. In a way, they act as advisors. In another way, one could view them as acting a fund (of funds) manager. The latter could be troublesome though.

Say I am a newbie investor, I like the hand-holding of an advisor and (foolish me!) defer to his/her expertise, but I don't like the cost, so I go with a robo-adviser as a middle ground. Ten years later, I get closer to retirement, I wake up, read all about passive investing and decide to take things in my own hands. How does that work, notably for the assets located in taxable accounts? In other words, can I easily 'fire' the robo-adviser, and take full control?
simple answer: By law they all allow transfer in kind to another broker. Not much different than any broker to broker transfer.

addendums:
The robos at brokers make this super easy. Fidelity Go is an extreme example it is literally a mouse click. The robo investor is fired and no longer manages your account at all. Whatever positions you are in are under your control. Merrill Edge Guided similarly, TD Ameritrade Essential Portfolios same thing.

Betterment is obnoxious about it (medallion guarantee over $250k...) but most brokers can handle that for you. Some of the others also get a bit more troublesome as asset side goes up.

Asset builder you are going to have problems because they use DFA. The asset builder account is Schwab institutional not individual. You can fire asset builder same as firing any other advisor. But you probably can't leave Schwab with the shares unless you are going to one to another FA institutional account. The institutional account would be linked and Schwab would then only allow minor adjustments like directing or not directing distriuctions or selling.

In addition some of the assets get more troublesome as asset side goes up. Hedgeable for example will start putting you in private partnership and specialty assets (bitcoin for example) so the receiving broker is going to have to be able to handle that. You might have some problems but even then you'd only have to sell 2% of the portfolio.

Covester for an accredited investor is going to be far and away the worst here, depending on the portfolio they choose. Interactive is one of (if not the most) diverse discount brokerages around. They try and offer the full suite of buy side services like prime brokers. For example they have a money market option in even units of $100m. Their robo (Covester) reflects that desire. Most (all?) other discount USA brokers probably couldn't handle the kinds of assets an accredited investor in Covester would have in their account: directly holding foreign stock, foreign currency, hedge fund partnerships, non-market derivatives... Obviously if you wanted to move to Goldman, JPMorgan, Merrill, Credit Suisse... you'll be fine. But you definitely ain't transferring that to Vanguard which was your question.
slinky$
Posts: 28
Joined: Mon Aug 18, 2014 3:33 am

Re: What we know about robo advisors

Post by slinky$ »

jbolden1517 wrote:Schwab Intelligent Portfolios


At this point Schwab Intelligent Portfolios is my recommended robo-advisor
Interesting and thanks for doing the reviews. It would be cool if you could put together a ranking of the common robos in a simplified chart.

Just saying :sharebeer

Please do one for Wealthfront if you get chance.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

slinky$ wrote:
jbolden1517 wrote:Schwab Intelligent Portfolios


At this point Schwab Intelligent Portfolios is my recommended robo-advisor
Interesting and thanks for doing the reviews. It would be cool if you could put together a ranking of the common robos in a simplified chart.

Just saying :sharebeer

Please do one for Wealthfront if you get chance.
I agree on Wealthfront. There are one of the big 2, they are a must. I was thinking them for next. Either that or a post on advisor only robos.

There are pretty good charts for the common robos. The robo report I would subscribe to. I'd love to post it here but it is copyright. If I did do a chart what Boglehead specific information would you want? I was thinking of ER for the robo and description in Boglehead language. The other question is who is the target? If it is naive investor those are all over the web. The other option is a Boglehead reader who is helping a friend. So for example I could use shorthand like (DFA based) and assume the reader knows DFA style allocations.

Part of my problem is that while I know this boards father, mother and grandfather I don't know this board's readership too well. I'm still learning what people here know and don't know. I'm still learning what mainstream finance ideas are accepted and which are not. To complicate this for an investing community this board is rather unique in the level of conflicting ideas that are accepted but only in some contexts. So the same concept with slightly different language is either strongly rejected or considered so benign as to not be worth mentioning. That makes creating a standard chart more complex.
Avo
Posts: 1099
Joined: Wed Jun 11, 2008 2:21 am
Location: California

Re: What we know about robo advisors

Post by Avo »

jbolden1517 wrote:I don't know how you can say you weren't disputing MPT, "MPT is 65 years old. Where are the mutual funds that have out-performed by running an MPT strategy for the past, say, 20 years?" Either you believe MPT is false or obviously mutual funds using MPT have out-performed.
My claim is that there are no such funds.

MDLOX, your example, isn't one: it uses all sorts of stuff, including a big dose of human judgment: "Seasoned team of 50+ professionals have navigated diverse markets for over 28 years"; "the importance of differentiating between securities within an index has become more meaningful"; blah blah. They're not just choosing index funds in different asset classes according to an MPT correlation matrix, which is what SIP claims to do. (But what happens when the correlation matrix changes, which it does daily? How often should adjustments be made? And that matrix was determined in the first place by averaging over what time frame?)

And, over the last 10 years, MDLOX has underperformed both the relevant index (60/40 global stock/bond, not global stock; orange line) and the comparable Vanguard fund, Life Strategy Moderate Growth (green line):

Image

Here I am using MDLOX, which is the lowest ER share class that a retail investor can buy; 5.25% front-end load waived in the chart. (The institutional-class MALOX has a $2M minimum.)

The composition of the M* 60/40 global index is detailed here:
http://portfolios.morningstar.com/fund/ ... ture=en-US

My point is that a "pure" MPT strategy (one that does not include stock selection etc) is untested in running actual money for retail investors.

And your hindsight-based choice of an active fund that uses multiple asset classes has underperformed its plain-vanilla Vanguard equivalent over the past 10 years, as well as the relevant index.

Caveat emptor.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

Avo wrote:
jbolden1517 wrote:I don't know how you can say you weren't disputing MPT, "MPT is 65 years old. Where are the mutual funds that have out-performed by running an MPT strategy for the past, say, 20 years?" Either you believe MPT is false or obviously mutual funds using MPT have out-performed.
My claim is that there are no such funds.
I think you need to qualify that claim a bit. There are three possible claims.
1) Given any fund there exists some index which outperforms it over some period of time. That's true for any fund including any index fund.
2) There does not exist any open ended mutual fund using an MPT approach which as a result of that approach beat its index.
3) There does not exist any open ended mutual fund using a passive balanced approach which beat an index using the same balanced approach for 10 years.

(3) may be true because there haven't been passive multi asset funds until recently. There certainly were pension funds and retirement accounts using MPT passive approaches. But for most of them some of the vehicles were active. So (3) may be true because it hasn't been tested. In general the move to passive investing in a widespread way is new. Going forward it could be a disaster. VPACX was a disaster for Asian investors. As you know I have serious concerns already about USA SP500.

(2) I gave you an example of a fund that does that. And has done it this decade as well (which was a point in dispute). And then I showed the evidence regarding MPT by backtesting random portfolios you picked and constructing new portfolios which when you added a low volatility global stock fund (though I'll agree active) to them reduced volatility.

As for your dismissal of volatility there we disagree. You want return and don't care about volatility just use leverage. A good asset on leverage becomes a better asset. The reason I'm holding Blackrock global is not return. It does OK for return, slightly outperforming global stock. But I'm not having any problem with return. I'm getting terrific returns. But I'm getting terrific returns at the expense of ending up on too much effective leverage. My volatiles were getting way out of control (once I realized the problem I started correcting). And there low volatility sucks up that leverage like a sponge. If you have two assets A and B with both having the same return and B being 1/2 as volatile by adding B to the portfolio at X% you suck up X/2% leverage from other assets relative to holding A. In low interest rate environments return and low volatility are interchangeable.

Bogleheads talk about risk a lot and hold a lot of cashish sorts of bonds to bring their volatility down. But they don't describe what they are doing as deleveraging their portfolios. I think that's a mistake in the culture here.
Avo wrote:They're not just choosing index funds in different asset classes according to an MPT correlation matrix, which is what SIP claims to do. (But what happens when the correlation matrix changes, which it does daily? How often should adjustments be made? And that matrix was determined in the first place by averaging over what time frame?)
Now those are good questions. The answer is that most of the mainstream robos use long term charts like Ibbotson's which are reviewed every decade. So the asset allocation might change about every decade. To pick the other extreme Hedgeable uses tick to tick, hourly, daily, monthly and annually. They can and will flip a portfolio hard as correlations dynamically change. They make no claim you as an investor with them will be holding a "standardish" kind of portfolio. Hyugens does the same thing.

Wealthfront being one of the earliest in this space talked about this issue most directly in their early discussion.

Here is what the mainstream guys do (Wealthfront is explicit they do this, the rest pretty likely followed their lead). To get the MPT equations to work you need covariances and expected returns. Covariances if you intend to change your model infrequently are easy to get from historical data. Expected returns you can use historical data. But then often you end up with portfolios that are really weird. (sort of like hedgeable portfolios).
Black-Litterman solved this problem by asserting something like CAPM you use the covariance matrix and then set the expected returns of all asset classes to match their market weights. Which means an average investor with average risk tolerance holds a market weight portfolio. As you increase risk the portfolio tilts but it tilts in a way that looks very much like the market, that is what a mainstream advisor would put a client in.
Avo wrote: And, over the last 10 years, MDLOX has underperformed both the relevant index (60/40 global stock/bond, not global stock; orange line) and the comparable Vanguard fund, Life Strategy Moderate Growth (green line):
Why is that the relevant index? How did you pick that index say vs. any number of other indexes? That index is 40% USA stock, 20% international stock with 65% of the bonds being USA domestic. Why would that be the relevant index for a global fund? Blackrock is selling a global fund. Part of the reason I know about Blackrock is the 50/50 allocation. When I buy a global fund I want 50/50 at least. Blackrock right now is positioned rather bearishly, but the stocks are 50/50 not 60/40. Similarly the bonds are 55/45 (a slight tilt), not 65/35.

I will agree there are lots of indexes that Blackrock underperformed. There are lots of indexes every fund underperforms there are a lot of indexes. But if I'm going to compare a fund to an index, I'm going to compare it to an index tracking the assets it holds.
Avo wrote:Here I am using MDLOX, which is the lowest ER share class that a retail investor can buy; 5.25% front-end load waived in the chart. (The institutional-class MALOX has a $2M minimum.)
Most people get it in a 401K. I didn't.
Avo wrote: My point is that a "pure" MPT strategy (one that does not include stock selection etc) is untested in running actual money for retail investors.
True. The passive revolution affects all sorts of things. But what is the theory you are proposing? That MPT will work worse with passive vehicles that track an index that it builds its correlation matrixes on more closely than it did with active funds that deviate from those indexes in unpredictable ways?
Avo wrote: And your hindsight-based choice of an active fund that uses multiple asset classes has underperformed its plain-vanilla Vanguard equivalent over the past 10 years, as well as the relevant index.
Well we already discussed it did better than its relevant index. It did pretty close to but slightly underperformed an entirely unrelated index you picked. As for "hindsight-based", that's fair. I'm a recent supporter of Blackrock Global. But I've been an advocate for Oakmark global as a simple portfolio for people starting out for 2 decades. If I had known about Blackrock Global that would have been a better fit. But I didn't. My affection for the robos is recent. My dislike of the Life Strategy funds is also not recent. I've been critical of their weak international exposure and cap weighting since the 1990s. When they dropped TAA and Vanguard got more fundamentalist about indexing they got dropped entirely as way too undiversified to be safe. So those aren't hindsight-based choices. Those were choices made in real time.
Oakmark Global beats MSCI world by 5.6%.
Of course that's active and not really a portfolio at all. It was just a safe mutual fund to put people in who didn't know anything and weren't going to look when there weren't good options. But I certainly don't think that 2 decades later I was wrong.
Avo
Posts: 1099
Joined: Wed Jun 11, 2008 2:21 am
Location: California

Re: What we know about robo advisors

Post by Avo »

You sure write a lot.

M* chose that 60/40 stock/bond index for MDLOX, not me.

It matches the current allocation of MDLOX, which is 39.96% bonds and cash, 60.04% stock and "other" (according to M*).
User avatar
LadyGeek
Site Admin
Posts: 95686
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: What we know about robo advisors

Post by LadyGeek »

jbolden1517 wrote:...As for your dismissal of volatility there we disagree. You want return and don't care about volatility just use leverage. A good asset on leverage becomes a better asset. The reason I'm holding Blackrock global is not return. It does OK for return, slightly outperforming global stock. But I'm not having any problem with return. I'm getting terrific returns. But I'm getting terrific returns at the expense of ending up on too much effective leverage. My volatiles were getting way out of control (once I realized the problem I started correcting). And there low volatility sucks up that leverage like a sponge. If you have two assets A and B with both having the same return and B being 1/2 as volatile by adding B to the portfolio at X% you suck up X/2% leverage from other assets relative to holding A. In low interest rate environments return and low volatility are interchangeable.
Defining your terminology is important to avoid misunderstandings. Could you please explain what you mean by "leverage"?

I'm familiar with leverage from this perspective: Leverage, which is something you want to avoid.
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.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

LadyGeek wrote:
jbolden1517 wrote:...As for your dismissal of volatility there we disagree. You want return and don't care about volatility just use leverage. A good asset on leverage becomes a better asset. The reason I'm holding Blackrock global is not return. It does OK for return, slightly outperforming global stock. But I'm not having any problem with return. I'm getting terrific returns. But I'm getting terrific returns at the expense of ending up on too much effective leverage. My volatiles were getting way out of control (once I realized the problem I started correcting). And there low volatility sucks up that leverage like a sponge. If you have two assets A and B with both having the same return and B being 1/2 as volatile by adding B to the portfolio at X% you suck up X/2% leverage from other assets relative to holding A. In low interest rate environments return and low volatility are interchangeable.
Defining your terminology is important to avoid misunderstandings. Could you please explain what you mean by "leverage"?

I'm familiar with leverage from this perspective: Leverage, which is something you want to avoid.
Nope the kind from the article is the kind of leverage I meant. Here is what I meant by the comment.
Assume you can borrow at 3%.
Asset A on heads returns +30% and on tails -10% (Bernstein's stock model). This asset has a +10% expected return and a 8.16% serial return (this is a risky asset there is a pretty big spread).
Asset B on heads returns +5% and on tails returns +4%. This asset has a 4.5% expected return and a 4.49% serial return (safe means the serial return is close to the expected return)
(note this sort of asset would never exist in the real world).

B is unquestionably a "worse fund" using just a strictly return definition. It returns less than A.
However B is much less volatile it can be held safely on any amount leverage. Say for example you hold B at 10::1 leverage. Your portfolio is 1000% B, -900% debt. The debt is costing you -27%.
Heads you get +50% on B -27% interest = +23%. Tails you get +40% -27% interest = +13%. Exepected 18% Serial 17.89
First off this way better than A. It is both safer and has a higher return. Let's go crazy high
B on 1000::1 leverage. Your portfolio is 100,000% B. -99,900% debt
Heads you get +5000-2997 = +2003%, tails you get = +4000-2997=1003%. Expected return +1503%. Serial return +1423%
OK so now it is pretty risky :D

So you can see there are assets, at least in theory, you can't hold raw but because of low volatility it makes sense to hold on leverage. Given any asset you can either dilute with cash (and intermediate bonds act enough like cash to count here) or concentrate with debt to create a higher returning more volatile version of the asset.

Now B is a stupid theoretical asset. There rarely (never) are assets that safe that high returning. But often you can synthetically using leverage create assets that look like B. For example a 10 year corporate bond is often funded by shorting the corresponding 10 year treasury and collecting the 30 basis points in interest. Doing 1 unit of this is a terrible returning asset (you get 30 basis points max on your money). Doing this at 40::1 leverage however is a 15% reasonably safe returning asset. You get totally mauled if the corporation gets into trouble and otherwise collect 16% (39*30 = 11.7 on the short/long play plus say 4.3% on 1 copy of the bond held with actual money)
So your return looks something like: +15%, 95% of the time when everything goes well, 0% 4% of the time when there is a little trouble and the bond drops a notch, -40% 1/2% of the time (when the company suddenly has serious problems and drops two notch suddenly), -150% 1/2% of the time when the bond suddenly falls into junk status.
And when you hold bank stocks as part of TSM that's the kind of asset they are holding from which you are getting your returns.

So let's look at A which is a fairly good model for stocks.
A 3% interest rate your best possible serial return is at 2.1::1 leverage. Close enough to 2::1 so we'll do the math
200% A, -100% debt returns +57% on heads, -23% on tails. Expected :17% but serial return of just 9.94%. This is the thing your article was talking about. You can see how huge the spread is (stocks on 2::1 leverage are very risky) and in exchange for all that risk i boosted my serial return just an extra 1.79%. In fact stocks are so risky that if I increase my leverage much more than this my serial return starts going down
so for example if I tried to hold this portfolio at 4.5::1 leverage I'd actually have a negative serial return.

You all have probably noted in discussion of 3fund that 80/20 stock isn't that much better than 60/40 stock. That's because of the problem of stock risk. Diluting the risk of the stock boosts returns and helps make up for the lower returns of stocks to bonds. That same argument applies to 200/-100. Via leverage low volatility can be converted to return and via. dilution return can be converted to lower volatility. Serial return which is what you really care about depends on keeping your volatility under control while boosting your expected return. The fund we were discussing has an expected return close to a global stock index but a volatility much lower. So it reduces portfolio leverage, the same role that bonds play in 3fund.
User avatar
LadyGeek
Site Admin
Posts: 95686
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: What we know about robo advisors

Post by LadyGeek »

That helps, but I'm still missing a few pieces. To create leverage, I assume that something needs to be borrowed. Where does that come from? Shorting is one method.

I'm also looking at metrics. Wouldn't the Sharpe and Sortino ratios be used to measure this effect? Regardless of how you get there, these ratios provide a standardized way to characterize performance.

Simba's backtesting spreadsheet (look inside the spreadsheet's README tab) links to a good tutorial: [url=chrome-extension://mhjfbmdgcfjbbpaeojofohoefgiehjai/index.html]Sortino A 'Sharper' Ratio[/url]
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.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

LadyGeek wrote:That helps, but I'm still missing a few pieces. To create leverage, I assume that something needs to be borrowed. Where does that come from? Shorting is one method.
Yes when you short you borrow stock and sell it. Traditionally when you leverage up (like in your article) you borrow money and buy stuff. Or you can take on synthetic leverage through derivatives (which is actually how my portfolio volatility got too high).
LadyGeek wrote: I'm also looking at metrics. Wouldn't the Sharpe and Sortino ratios be used to measure this effect?
Not quite. In fact the Sharpe ratio uses the same ideas I was talking about. The Sharpe ratio is computed by subtracting off the risk free rate of return from asset return and then dividing by volatility. So essentially it asks the question, "if you were borrowing the money to buy this asset what would your volatility adjusted return be"? The Sharpe ratio is going to be the same for 10% TSM, 90% Cash and 200% TSM, -100% Cash. It factors out leverage, so it doesn't measure it.

Sortino I'm much less familiar with. While searching I did find a good article on it as a comparison to Sharpe: https://www.sunrisecapital.com/wp-conte ... o_0213.pdf . I don't have a good intuition about Sortino. Looking at the formula I kinda agree with the typical usage. Its a good way to measure a portfolio or a mutual fund or an advisor against a relevant index. So for example on this thread measure SIP against 3Fund 70/30 . Or measuring portfolios against their index. I suspect though it wouldn't work well. Sortino is trying to penalize the "large loss" type problem. My portfolio is way above average shielded against that though lots of diversification and anti-correlation; I practice what I preach. The market opens up Monday down 50%, I'm likely up a lot but even if I'm down I'm down 15% tops. The market opens up Monday up 50% I'm definitely up at least 15%. So the Sortino ratio would be very high. It just isn't capturing the right measure of risk I was experiencing.

Standard measures like "portfolio value at risk" and "worst case loss during defined times" are probably the right measures for what I'm calling daily crazy. You might be onto something though that I should start measuring using standard metrics. This isn't really a mutual fund investor type problem. Individual stocks much less large baskets don't move the way derivatives swing. Imagine if your stock portfolio went randomly between +25/-25 off a nice +10% annual line every day. Not compounded just swinging around. You are getting your 10% from your stocks and all those swings are going to balance out. But day to day your portfolio would feel completely out of control. That's a bad analogy but sort of what was happening.
User avatar
siamond
Posts: 6008
Joined: Mon May 28, 2012 5:50 am

Re: What we know about robo advisors

Post by siamond »

jbolden1517 wrote:Standard measures like "portfolio value at risk" and "worst case loss during defined times" are probably the right measures for what I'm calling daily crazy. You might be onto something though that I should start measuring using standard metrics.
We made quite an effort to compute various fairly standard metrics in the Simba spreadsheet (many thanks to MachineGhost, who provided a lot of expertise in the process), and try to get their formulas properly implemented (I hope we did!). Sharpe and Sortino (in their recent forms), Max Drawdown, Alpha/Beta, and then less known (but very sensible) metrics like the Ulcer Index. There is also a short glossary in the spreadsheet, providing a few pointers to useful references providing more information. This post in the backtesting thread is a good example of what is computed.

Personally, I think that the Sharpe ratio is a weird metric, and Sortino/Ulcer are much more meaningful, but that's just me.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

Wealthfront: https://www.wealthfront.com

Wealthfront is the second of the 2 big names that started the robo-advisor service. The design came from Burton Malkiel (Chief Investment Officer at Wealhfront). They hit growth problems a little earlier than Betterment changed direction in 2016 and have expanded their offerings since then. There are tons of reviews on the web of their basic service but I'd comment that Wealthfront is adding services regularly so please check the dates on reviews.

Wealthfront unlike most other robo-advisors does not offer a paid upgrade to a human advisor. They very much stick with the vision of the early movement to replace advisors. Their website and newsletter are unquestionably advisor hostile talking about how advisor favor funds with large advertising and marketing budgets resulting in fee structures 5x that of passive managed. They even are critical of the more full featured advisory services. For example they attacked wealth managers noting what 1% of $10m, and noting that services like setting up a trust or establishing family governance guidelines... should just be handled by an attorney with no need for FA advice. There is no question their DIY and anti-advisor sentiment is very much in alignment with Boglehead philosophy.

They tend to do more stock in taxable and more bonds in non taxable. Their sample non-taxable is a 70/30 portfolio (assuming you count REITs as bonds):
Image
The list of Vanguard ETFs I assume are familiar to most readers here. Their taxable accounts are differentiated by including natural resources as a diversifier. And again note the heavy Vanguard focus. As an aside they have modeled their investor education after Vanguard's so many of the materials have a similar feel. They are noted for their heavy EM focus for diversification.
Image

Their depletion yield oriented portfolio (a late in life income investor) looks quite good. It creates a reasonably high current income, stability for depletion and has an appropriate and heavy attention to avoiding inflation risk. The core of the portfolio is a municipal (EM and corporate for non taxable) and TIPS mixed with some equity income. US, foreign and EM are used as satellites. I normally would want a higher foreign allocation but with this much in TIPS the portfolio is heavily inflation protected.
Here is an example for taxable and in parenthesis tax free yield fund
8 (6)% US stock
5 (5)% foreign stock
5 (5) % EM
15 (5)% dividend stock
0 (5)% real estate
7 (0)% natural resources
25 (25)% TIPS
35 (0)% municipal bonds
0 (35)% corporate bonds
0 (14)% EM bonds

Wealthfront has the lowest fee structure. Their ETFs average about 5 basis points. The fee for the serive is 25 basis points for the portfolio above first $10k (supposedly if you use this link they will increase the amount to $15k: https://www.wealthfront.com/invited/AFF ... sVBsXPeVg0), If you refer a friend who signs up this waiver goes up by another $5k.
On the high end as well as the low the fees go down somewhat with size. At $100k they start introducing "direct investing" which is holding a portfolio of stocks to replicate the index rather than holding and ETF. This substantially increases the possibilities for tax loss harvesting. At the $100k level they are targeting 30% of the portfolio. As the portfolio gets larger bonds and foreign stocks get added so at $500k they are targeting 80% held directly (the highest number I've seen). They don't charge for stock trades and as the portfolio shifts towards directly held stocks they reduce fees below the full 25 basis points. With the tax loss harvesting on direct holding low base fees to start and decreasing fees a larger portfolio at Wealthfront would have an a large effective negative cost basis. Wealthfront was the earliest innovator in tax lost harvesting starting the program in 2014, and they take justifiable pride in it. Their daily tax lost harvesting engine has boosted to returns on clients has averaged an extra 1.55% annually since then. For customers with direct holding this average hits almost 3%. They do not offer a service where they created paired investments to create synthetic losses like some robos-advisors however.

In terms of extra they have quite a few. They offer a total financial health and management application called Path. Path if you set it up tracks your actual spending and savings and allows you to see the impact of adjustments. So for example if you want to see the impact on retirement of taking extra expensive vacations it will show the decreasing in savings leading to retirement 2 years later. Path also monitors external accounts and flags investment options where fees are high, the portfolio is tax inefficient, there is too much cash drag, or there is a lack of diversification. They have recently expanded their integration with colleges to assist in saving for college tuition

Unusually for a pure robo service they have a margin program, they offer a portfolio line of credit where you can borrow up to 30% of portfolio at the effective federal funds rate +3.60%. This margin rate declines down to federal funds +2.35% at a $1m account balance (note account balance not margin balance). This is better than most brokers (Vanguard who leans high for example is federal funds +6.5%) but not in line with the very lowest cost margin accounts.

For people with company stock they have company specific selling programs and rebalancing (i.e. it is tailored to specific companies). They are continuously expanding the list of companies and ESPP (employee stock purchase plans) they integrate with.

Reviews of their customer service desk are uniformally excellent. This was a strong point in customer reviews. Wealthfront may not offer human advisors but they clearly spend on training their staff and don't understaff even at their very low fee structure.

There are several downsides. They have no cash program which can complicate using Wealthfront for shorter term spending needs. The don't offer fractional shares of ETFs which bothered reviews but not their customers. They do not offer customization beyond risk based weightings. Customization was the most common complaint which is odd because few of the mainstream robo-advisors offer any more customization than Wealthfront. For some reason their customer base seems to be attracted to Motif (a broker with thousands of robo style portfolios that can be set for investment) and compares them to Motif rather than the other mainstream robo choices like Betterment or Wise Banyon.

I'll start the editorial part of this review with this comment. Compared to Motif's interface technologically Wealthfront lags, but so does every other broker out there. Motif has great programmers. Wealthfront is trying to steer people towards a mainstream portfolio filled with broadly diversified standard set of asset classes based on an analysis of their returns and correlations. Motif certainly allows for that sort of portfolio and has hundreds of good ones to choose from. But their focus is on being a really cool and fun broker for millennials. Their unquestionable programming talent is directed at assisting their clients to giving full rein to all sorts of wild investing impulses with highly customized portfolios and advice on portfolio design coming in from all angles the mainstream, the offbeat and the downright wacky. Wealthfront's portfolio is one of their Motifs. I'm sure if I picked a random person holding the Wealthfront motif they would be holding the Wealthfront portfolio right alongside their Atlanta sports franchises motif and technology companies with a focus on watches motif. The two companies just aren't selling the same product.

I love the in your face attitude of the website towards financial advisors, though I do disagree a bit. But it is nice to hear the gospel of first generation robo investing preached this unapologetically. Wealthfront lacks a value tilt in their portfolios, and they cap weight everything. For me that portfolio design is a deal breaker. I can't recommend a portfolio I disagree with this strongly despite how much else I like about this robo-advisor, when better (non-cap weighted, value tilted) portfolios exist. Of course both these objections apply to 3fund as well. So for someone who disagrees with me then compared to 3fund the sophisticated TLH, automatic management, Path and portfolio leverage provide real advantages, without having to take on a value tilt. The high EM exposure is great. The Natural Resources seems quite creative and given their investment advisor I suspect that works like a classical commodities position, possibly with a higher return. I will say I am really excited for the backtest with the natural resources and heavier EM mixed in to see if there are any unexpected benefits, I'll post if here as an edit. The depletion income portfolio I think is quite good and that one I can recommend.

The fee structure is excellent for small accounts and with a minimum of $500 they clearly play well for people starting out (Schwab for example has a $5k minimum). I'm not sure Acorns though isn't a better fit with the coupons and forced savings for low income investors. The portfolio lending is a huge plus that doesn't get enough credit, especially at the quite reasonable interest rates. If you disagree with Schwab's value tilt and care greatly for tax lost harvesting (which you should) they seem like a solid choice. Their product is solid in every way but it is hard to imagine in the diversified robo market that exists today that they would be many people's first choice. And that I think is the problem I see for them going forward. They are a terrific robo to be the point of comparison with the robo you eventually pick. For Bogleheads they embrace Vanguard funds, very low cost investing (as I've argued probably negative cost in practice), DIY with a strong dose of hostility to financial advisors. They might be a good choice for this board's default choice as well because of that strong philosophical alignment.
Last edited by jbolden1517 on Sun Aug 06, 2017 10:11 am, edited 4 times in total.
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

siamond wrote:
jbolden1517 wrote:Standard measures like "portfolio value at risk" and "worst case loss during defined times" are probably the right measures for what I'm calling daily crazy. You might be onto something though that I should start measuring using standard metrics.
We made quite an effort to compute various fairly standard metrics in the Simba spreadsheet (many thanks to MachineGhost, who provided a lot of expertise in the process), and try to get their formulas properly implemented (I hope we did!). Sharpe and Sortino (in their recent forms), Max Drawdown, Alpha/Beta, and then less known (but very sensible) metrics like the Ulcer Index. There is also a short glossary in the spreadsheet, providing a few pointers to useful references providing more information. This post in the backtesting thread is a good example of what is computed.

Personally, I think that the Sharpe ratio is a weird metric, and Sortino/Ulcer are much more meaningful, but that's just me.
I got a good laught from the ulcer index. I'd still do great there though. Max downturn time was a week. Maybe I was being a crybaby? :confused
User avatar
siamond
Posts: 6008
Joined: Mon May 28, 2012 5:50 am

Re: What we know about robo advisors

Post by siamond »

jbolden1517 wrote:Their sample non-taxable is a 70/30 portfolio (assuming you count REITs as bonds) [...]
In the various backtests we did so far, we counted REITs as stocks (as Bogleheads typically do). And of course, whatever Prof. Malkiel says, dividend stocks are stocks too... Could you please document a Wealthfront portfolio that maps to 70/30 according to such criteria?
Topic Author
jbolden1517
Posts: 868
Joined: Sun Jul 09, 2017 2:53 pm

Re: What we know about robo advisors

Post by jbolden1517 »

siamond wrote:
jbolden1517 wrote:Their sample non-taxable is a 70/30 portfolio (assuming you count REITs as bonds) [...]
In the various backtests we did so far, we counted REITs as stocks (as Bogleheads typically do). And of course, whatever Prof. Malkiel says, dividend stocks are stocks too... 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
Post Reply