optimal strategy for exploiting robo-advisors

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Topic Author
ksJoe
Posts: 32
Joined: Wed Jul 11, 2007 9:40 pm

optimal strategy for exploiting robo-advisors

Post by ksJoe » Fri Nov 08, 2019 10:04 pm

I had a random thought last night, tell me what I'm missing here. I think with a little bit of hassle I can get the aggressive tax loss harvesting benefits from a robo-advisor, without paying them for managing a large amount of money.

Assumptions:
  • While a portfolio is rapidly growing, robos can save you a lot of money with tax loss harvesting (wealthfront has done great for me).
  • While a portfolio is rapidly being sold off during retirement, robos can likely save some money by micromanaging which lots to sell.
  • On a large portfolio (i.e. when new contributions are a small percentage of account balance), TLH can't do much for the portfolio overall
  • The 1/4% fee is outrageous on large balances

The basic issue is that their fee assessment is disconnected from their value creation.

So here's the strategy:
  • Only use a robo to manage the new investment dollars
  • Every time the robos balance gets big enough that new contributions are a fairly small as a percentage of balance, and the annual fee is in the hundreds, transfer the balance to a low cost brokerage.
  • Repeat this cycle: new $ go to the robo for micromanagement the first 1-2 years, then get transferred elsewhere for maintenance holding.
  • During retirement, old $ could be gradually transferred back to the robo gradually to micromanage the sell off, if desired.
  • If there is market goes down a lot, the whole balance could be briefly transferred back to the robo for some mid-life TLH.
  • In general, don't rebalance the funds you've transferred out of the robo. Instead adjust holdings in tax advantaged account to compensate.

Since uncle Sam mandated they transfer cost basis info with the assets, you can transfer the balance away from the robo, then bring it back years later, and never have to compute any cost basis yourself

Minor details:
  • To avoid wash sale rule violations, dividends should not be reinvested on the balance parked at a normal brokerage (its preferable let the robo do TLH on those $ anyway).
  • The process can be enhanced by creating more than 1 account at the robo (as a workaround for them not allowing partial in-kind transfers out).

What this accomplishes:
  • When TLH is attainable, we get it automated and micromanaged at an economical cost
  • Stagnant portfolio dollars hang out for free at a low cost brokerage
  • Think of it like this: wouldn't it be nice if the robo would wave the fees on those big old holdings that will never see more TLH? That's basically what I'm proposing, by transferring them to a generic brokerage to hang out for a few years.

In order to be worth the trouble, a person would need to be avoiding their fees on a sizeable balance (e.g. avoiding hundreds per year in fees).

They could try to block this strategy by imposing transfer out (or other) fees, but it would take big fees to negate the strategy. This is unlikely to happen since the industry trend has been rapidly falling fees. Adding punitive fees would really make them stand out.

The best way to block this strategy, would be for them to align their fee to the value they create. That is, have a steeply regressive fee. Of course, that's unlikely since those are presumably the accounts where they make most of their revenue.

So.... what's wrong with this strategy?

User avatar
Silly Wabbit
Posts: 103
Joined: Sat Mar 25, 2017 9:54 pm

Re: optimal strategy for exploiting robo-advisors

Post by Silly Wabbit » Fri Nov 08, 2019 10:16 pm

What sort of micromanaging would be happening when selling during retirement? Most lots will be long term gains. Sell the ones with the highest basis.

Topic Author
ksJoe
Posts: 32
Joined: Wed Jul 11, 2007 9:40 pm

Re: optimal strategy for exploiting robo-advisors

Post by ksJoe » Fri Nov 08, 2019 10:22 pm

Silly Wabbit wrote:
Fri Nov 08, 2019 10:16 pm
What sort of micromanaging would be happening when selling during retirement? Most lots will be long term gains. Sell the ones with the highest basis.
Yeah, that's the part I'd thought about the least. That's a ways off for me. There may not be much point in that.

Because of their aggressive TLH and managing all taxible events on a "per lot" basis, I'm assuming the cost basis will be highly fractured (like thousands of lots). Paying attention to that might be painful. If so, transfer it back and let their software deal with it.

Or if its no big deal, never transfer the $ back.

HomeStretch
Posts: 2629
Joined: Thu Dec 27, 2018 3:06 pm

Re: optimal strategy for exploiting robo-advisors

Post by HomeStretch » Sat Nov 09, 2019 12:13 am

Some of the robo-advisor portfolios (including Wealthfront) that have been posted in forum threads have included a high number of holdings. I can envision a possible complex portfolio of hundreds of holdings using your strategy of transferring holdings periodically to a low-cost brokerage. You also wouldn’t want to automatically reinvest dividends and capital gains at the second brokerage in order to avoid inadvertent wash sales in the robo-advisor account.

jocdoc
Posts: 77
Joined: Wed Oct 30, 2013 5:29 am

Re: optimal strategy for exploiting robo-advisors

Post by jocdoc » Sat Nov 09, 2019 7:06 am

A Somewhat simpler strategy is to have the robo platform manage the stock portion asthe bond portion of the portfolio is likely in a tax advantaged account. As you get older the bond portion will likely be a more significant % of your portfolio and doesn't need managing.

Lee_WSP
Posts: 1167
Joined: Fri Apr 19, 2019 5:15 pm
Location: Arizona

Re: optimal strategy for exploiting robo-advisors

Post by Lee_WSP » Sat Nov 09, 2019 9:27 am

I fail to see what the robo would do that 15 minutes of my time & a calendar reminder would not be able to do.

User avatar
Watty
Posts: 17437
Joined: Wed Oct 10, 2007 3:55 pm

Re: optimal strategy for exploiting robo-advisors

Post by Watty » Sat Nov 09, 2019 9:58 am

ksJoe wrote:
Fri Nov 08, 2019 10:04 pm
So.... what's wrong with this strategy?
The big problem is that you will end up with a portfolio of possibly hundreds of stocks that you never be able to get rid because of the tax hit.

Topic Author
ksJoe
Posts: 32
Joined: Wed Jul 11, 2007 9:40 pm

Re: optimal strategy for exploiting robo-advisors

Post by ksJoe » Sat Nov 09, 2019 10:42 am

Lee_WSP wrote:
Sat Nov 09, 2019 9:27 am
I fail to see what the robo would do that 15 minutes of my time & a calendar reminder would not be able to do.
Automated software can do it much more aggresively and thoroughly.

I opened an account at Wealthfront in July 2018 with $5k, and contributed $1k weekly since. Just in 2018 they harvested $1594 in short term losses.

I've seen TLH events as low as $6. That's what I mean by micromanaging the TLH while accumulating.

Edit to add: Also I assume they are tracking every purchase as an individual tax lot, rather than average cost or FIFO, and always chosing the most adventagious individual lot for each taxible event.
Last edited by ksJoe on Sat Nov 09, 2019 10:49 am, edited 1 time in total.

Lee_WSP
Posts: 1167
Joined: Fri Apr 19, 2019 5:15 pm
Location: Arizona

Re: optimal strategy for exploiting robo-advisors

Post by Lee_WSP » Sat Nov 09, 2019 10:49 am

ksJoe wrote:
Sat Nov 09, 2019 10:42 am
Lee_WSP wrote:
Sat Nov 09, 2019 9:27 am
I fail to see what the robo would do that 15 minutes of my time & a calendar reminder would not be able to do.
Automated software can do it much more aggresively and thoroughly.

I opened an account at Wealthfront in July 2018 with $5k, and contributed $1k weekly since. Just in 2018 they harvested $1594 in short term losses.

I've seen TLH events as low as $6. That's what I mean by micromanaging the TLH while accumulating.
And that's compared to what? How do you know you wouldn't have achieved similar results?

alex_686
Posts: 5097
Joined: Mon Feb 09, 2015 2:39 pm

Re: optimal strategy for exploiting robo-advisors

Post by alex_686 » Sat Nov 09, 2019 10:55 am

ksJoe wrote:
Fri Nov 08, 2019 10:04 pm
  • On a large portfolio (i.e. when new contributions are a small percentage of account balance), TLH can't do much for the portfolio overall
  • The 1/4% fee is outrageous on large balances
The first assumption is wrong. TLH can reduce or eliminate unrealized long term capital gains. It does not matter if you are contributing or withdrawing funds. What matters is the volatility and correlation of the underlying securities.

The second statement is not a assumption but a assertion - and perhaps incorrect. Does it make sense to pay a .25% annual fee if it reduces the tax drag from long term capital gains by .5% a year? Of course, this rests on the assumption on long term returns, future LTCG tax rate, and the effectiveness of the TLH.

Topic Author
ksJoe
Posts: 32
Joined: Wed Jul 11, 2007 9:40 pm

Re: optimal strategy for exploiting robo-advisors

Post by ksJoe » Sat Nov 09, 2019 11:03 am

Lee_WSP wrote:
Sat Nov 09, 2019 10:49 am
ksJoe wrote:
Sat Nov 09, 2019 10:42 am
Lee_WSP wrote:
Sat Nov 09, 2019 9:27 am
I fail to see what the robo would do that 15 minutes of my time & a calendar reminder would not be able to do.
Automated software can do it much more aggresively and thoroughly.

I opened an account at Wealthfront in July 2018 with $5k, and contributed $1k weekly since. Just in 2018 they harvested $1594 in short term losses.

I've seen TLH events as low as $6. That's what I mean by micromanaging the TLH while accumulating.
And that's compared to what? How do you know you wouldn't have achieved similar results?
I don't know that - but this reminds me of the basic argument for or against indexing :P

I don't have the time or skill to try to beat everyone else. (on stock selection or taxes)
I assume the consensus view of the experts is the best and most efficient assesment (on stock price and TLH).
I like the idea of paying a very low cost and accepting the results of others research (on index funds and TLH).

Its not a perfect analogy, but I see some similarity there.

Also, by letting the robo do it, I will never have to explain myself or justify my actions to the IRS man (on wash sales). They can be as aggressive as they want to be, and I just enter the number they give me on the tax form. Worst case, Wealthfront issues a correction and I go with that.

They create the most value when the account balance (and their fee) is the lowest. I'd like to make use of that, but when the account value is well into the 6 figures IMHO they're just not worth it. So I'm looking to capture the value they offer on new investment dollars and avoid the long term cost.

User avatar
nedsaid
Posts: 12668
Joined: Fri Nov 23, 2012 12:33 pm

Re: optimal strategy for exploiting robo-advisors

Post by nedsaid » Sat Nov 09, 2019 11:06 am

ksJoe wrote:
Sat Nov 09, 2019 10:42 am
Lee_WSP wrote:
Sat Nov 09, 2019 9:27 am
I fail to see what the robo would do that 15 minutes of my time & a calendar reminder would not be able to do.
Automated software can do it much more aggresively and thoroughly.

I opened an account at Wealthfront in July 2018 with $5k, and contributed $1k weekly since. Just in 2018 they harvested $1594 in short term losses.

I've seen TLH events as low as $6. That's what I mean by micromanaging the TLH while accumulating.

Edit to add: Also I assume they are tracking every purchase as an individual tax lot, rather than average cost or FIFO, and always chosing the most adventagious individual lot for each taxible event.
I wonder if there are any academic studies regarding aggressive tax loss harvesting and investment returns. This strategy involves lots of turnover to generate the tax losses and I am wondering if this is creating a friction with bid/ask spreads and incorrect sell/buy decisions. In other words, I am wondering if the "Nedsaid effect", the ratio of incorrect sell/buy decisions being 2:1 or even 3:1 to correct sell/buy decisions. In other words, what you sell tends to do better than what you buy to replace, this is a big problem for amateur stock pickers and less so for investment professionals. I do know that this effect hits the professionals too. I am wondering if the algorithms and Artificial Intelligence have overcome this problem. This is one big reason that I warn against trading your portfolio a lot. Any research on this?
A fool and his money are good for business.

Topic Author
ksJoe
Posts: 32
Joined: Wed Jul 11, 2007 9:40 pm

Re: optimal strategy for exploiting robo-advisors

Post by ksJoe » Sat Nov 09, 2019 11:13 am

alex_686 wrote:
Sat Nov 09, 2019 10:55 am
The first assumption is wrong. TLH can reduce or eliminate unrealized long term capital gains. It does not matter if you are contributing or withdrawing funds. What matters is the volatility and correlation of the underlying securities.
Yes, poorly worded on my part. I should have said there is less opportunity when not contributing (because over time the market tends to go up, so older dollars in general would have a lower cost basis, and so it would generally take a more significant market decline to have TLH on older investments. But all that actually matters is having a significant enough decline). But I did mention in the event of a steep decline, assets could be transferred back temporarily for TLH.
alex_686 wrote:
Sat Nov 09, 2019 10:55 am
The second statement is not a assumption but a assertion - and perhaps incorrect. Does it make sense to pay a .25% annual fee if it reduces the tax drag from long term capital gains by .5% a year? Of course, this rests on the assumption on long term returns, future LTCG tax rate, and the effectiveness of the TLH.
True. Its an assertion based on my assumption, which is not necessarily correct.

Topic Author
ksJoe
Posts: 32
Joined: Wed Jul 11, 2007 9:40 pm

Re: optimal strategy for exploiting robo-advisors

Post by ksJoe » Sat Nov 09, 2019 11:22 am

nedsaid wrote:
Sat Nov 09, 2019 11:06 am
I wonder if there are any academic studies regarding aggressive tax loss harvesting and investment returns. This strategy involves lots of turnover to generate the tax losses and I am wondering if this is creating a friction with bid/ask spreads and incorrect sell/buy decisions. In other words, I am wondering if the "Nedsaid effect", the ratio of incorrect sell/buy decisions being 2:1 or even 3:1 to correct sell/buy decisions. In other words, what you sell tends to do better than what you buy to replace, this is a big problem for amateur stock pickers and less so for investment professionals. I do know that this effect hits the professionals too. I am wondering if the algorithms and Artificial Intelligence have overcome this problem. This is one big reason that I warn against trading your portfolio a lot. Any research on this?
I would love to see that! I've wondered similar things.

The robos maintain a particular asset allocation, so I doubt there is much chance of replacing with a something that performs significantly different. I'm under the impresion (but may be wrong) that they snuggle up to the line on wash sale rules and what is "substantially identical" assets. I personally don't want flirt with that.

On friction - they hold so many accounts, I wonder if they have to put the transactions out on the open market? Or can they "buy" and "sell" directly between their customer's accounts with no friction? (I have no info on this, just wondering)

Lee_WSP
Posts: 1167
Joined: Fri Apr 19, 2019 5:15 pm
Location: Arizona

Re: optimal strategy for exploiting robo-advisors

Post by Lee_WSP » Sat Nov 09, 2019 11:25 am

ksJoe wrote:
Sat Nov 09, 2019 11:03 am
Lee_WSP wrote:
Sat Nov 09, 2019 10:49 am
ksJoe wrote:
Sat Nov 09, 2019 10:42 am
Lee_WSP wrote:
Sat Nov 09, 2019 9:27 am
I fail to see what the robo would do that 15 minutes of my time & a calendar reminder would not be able to do.
Automated software can do it much more aggresively and thoroughly.

I opened an account at Wealthfront in July 2018 with $5k, and contributed $1k weekly since. Just in 2018 they harvested $1594 in short term losses.

I've seen TLH events as low as $6. That's what I mean by micromanaging the TLH while accumulating.
And that's compared to what? How do you know you wouldn't have achieved similar results?
I don't know that - but this reminds me of the basic argument for or against indexing :P

I don't have the time or skill to try to beat everyone else. (on stock selection or taxes)
I assume the consensus view of the experts is the best and most efficient assesment (on stock price and TLH).
I like the idea of paying a very low cost and accepting the results of others research (on index funds and TLH).

Its not a perfect analogy, but I see some similarity there.

Also, by letting the robo do it, I will never have to explain myself or justify my actions to the IRS man (on wash sales). They can be as aggressive as they want to be, and I just enter the number they give me on the tax form. Worst case, Wealthfront issues a correction and I go with that.

They create the most value when the account balance (and their fee) is the lowest. I'd like to make use of that, but when the account value is well into the 6 figures IMHO they're just not worth it. So I'm looking to capture the value they offer on new investment dollars and avoid the long term cost.
I want to know how much better it is than just doing it on December 15th.

alex_686
Posts: 5097
Joined: Mon Feb 09, 2015 2:39 pm

Re: optimal strategy for exploiting robo-advisors

Post by alex_686 » Sat Nov 09, 2019 11:26 am

ksJoe wrote:
Sat Nov 09, 2019 11:13 am
Yes, poorly worded on my part. I should have said there is less opportunity when not contributing (because over time the market tends to go up, so older dollars in general would have a lower cost basis, and so it would generally take a more significant market decline to have TLH on older investments. But all that actually matters is having a significant enough decline). But I did mention in the event of a steep decline, assets could be transferred back temporarily for TLH.
Still not true. For context I was part of a team that did this stuff back in the 90s. We were able to reduce long term gains in a rising market while tracking the market. It did not matter if the clients were contributing or withdrawing from the account. The key bits that drove this strategy were volatility and correlation of assets.

I kind of wish that I knew of a good white paper on this subject because there is such a huge misconception of this strategy here on Bogleheads and my 20 year memory is not gong to do justice to this subject. IIRC, if a stock went up by 10% in a year, there was a 50% chance that it would drop below 20% its purchase price. That is in a up market. It is that volatility you exploit

alex_686
Posts: 5097
Joined: Mon Feb 09, 2015 2:39 pm

Re: optimal strategy for exploiting robo-advisors

Post by alex_686 » Sat Nov 09, 2019 11:47 am

nedsaid wrote:
Sat Nov 09, 2019 11:06 am
I wonder if there are any academic studies regarding aggressive tax loss harvesting and investment returns. This strategy involves lots of turnover to generate the tax losses and I am wondering if this is creating a friction with bid/ask spreads and incorrect sell/buy decisions.
This strategy should be "passively trading", and not doing "information trading". In theory you can make money this way. In reality, a 200% to 300% turnover would generate a drag of about 10 to 20 bps.
nedsaid wrote:
Sat Nov 09, 2019 11:06 am
In other words, I am wondering if the "Nedsaid effect", the ratio of incorrect sell/buy decisions being 2:1 or even 3:1 to correct sell/buy decisions. In other words, what you sell tends to do better than what you buy to replace, this is a big problem for amateur stock pickers and less so for investment professionals. I do know that this effect hits the professionals too. I am wondering if the algorithms and Artificial Intelligence have overcome this problem. This is one big reason that I warn against trading your portfolio a lot. Any research on this?
I think we have danced this dance before? I said your concerns were based on Fear, Uncertain, and Doubt. You said no. Could you please be specific on what type of incorrect sell/buy decisions you are talking about. This is a index strategy.

Let me be specific. The manager has no opinion on the direction of the market, of any individual security, and is trying to match the market's return. No position can be under/over weighted by 10%. So a index position of 1% could be no more than 1.1% to 09% This off-weighting would last 31 days and then may well be inverse. The portfolio is neutrally loaded on factors such as size, sector, value etc. The portfolio also has a tracking error of under 150 bps, which includes the bid/ask drag from the frequent trading. There is just not a lot of room for decisions here.

There has been plenty of academic studies on this. Most of the foundational stuff was written 30 years ago. Like I said in a earlier comment, I really wish I had a good non-proprietary source.

Topic Author
ksJoe
Posts: 32
Joined: Wed Jul 11, 2007 9:40 pm

Re: optimal strategy for exploiting robo-advisors

Post by ksJoe » Sat Nov 09, 2019 12:06 pm

alex_686 wrote:
Sat Nov 09, 2019 11:26 am
ksJoe wrote:
Sat Nov 09, 2019 11:13 am
Yes, poorly worded on my part. I should have said there is less opportunity when not contributing (because over time the market tends to go up, so older dollars in general would have a lower cost basis, and so it would generally take a more significant market decline to have TLH on older investments. But all that actually matters is having a significant enough decline). But I did mention in the event of a steep decline, assets could be transferred back temporarily for TLH.
Still not true. For context I was part of a team that did this stuff back in the 90s. We were able to reduce long term gains in a rising market while tracking the market. It did not matter if the clients were contributing or withdrawing from the account. The key bits that drove this strategy were volatility and correlation of assets.

I kind of wish that I knew of a good white paper on this subject because there is such a huge misconception of this strategy here on Bogleheads and my 20 year memory is not gong to do justice to this subject. IIRC, if a stock went up by 10% in a year, there was a 50% chance that it would drop below 20% its purchase price. That is in a up market. It is that volatility you exploit
Lets assume an account exists for 30 years, and has regular contributions. After 30 years, wouldn't it be the case that some of the holdings have such a low cost basis that its highly unlikely that particular purchase lot would be a part of TLH?

With individual stocks, anyone can go bankrupt at any time, but if we assume its index funds and the lottery effect is diversified away....

alex_686
Posts: 5097
Joined: Mon Feb 09, 2015 2:39 pm

Re: optimal strategy for exploiting robo-advisors

Post by alex_686 » Sat Nov 09, 2019 12:19 pm

ksJoe wrote:
Sat Nov 09, 2019 12:06 pm
Lets assume an account exists for 30 years, and has regular contributions. After 30 years, wouldn't it be the case that some of the holdings have such a low cost basis that its highly unlikely that particular purchase lot would be a part of TLH?

With individual stocks, anyone can go bankrupt at any time, but if we assume its index funds and the lottery effect is diversified away....
First, we can junk the "regular contributions". Regular, irregular, contribution, withdraw, etc. These just do not matter.

Second, we can junk the "index funds and lottery effect". This is a index strategy that tries to match the index. It is firmly rooted in the passive investment theory. We are just looking to exploit the tax code, not the market.

So, to answer your question: No - there should not be any holdings with a low cost basis. This is one of the basic premises of the strategy. Over 30 years there should have been enough volatility and low correlation stocks to rotate out all of the low cost basis stocks and reset the cost basis to a higher level. Strategies like this can have very high turnover - often over 100%.

User avatar
nedsaid
Posts: 12668
Joined: Fri Nov 23, 2012 12:33 pm

Re: optimal strategy for exploiting robo-advisors

Post by nedsaid » Sat Nov 09, 2019 4:48 pm

alex_686 wrote:
Sat Nov 09, 2019 11:47 am
nedsaid wrote:
Sat Nov 09, 2019 11:06 am
I wonder if there are any academic studies regarding aggressive tax loss harvesting and investment returns. This strategy involves lots of turnover to generate the tax losses and I am wondering if this is creating a friction with bid/ask spreads and incorrect sell/buy decisions.
This strategy should be "passively trading", and not doing "information trading". In theory you can make money this way. In reality, a 200% to 300% turnover would generate a drag of about 10 to 20 bps.
nedsaid wrote:
Sat Nov 09, 2019 11:06 am
In other words, I am wondering if the "Nedsaid effect", the ratio of incorrect sell/buy decisions being 2:1 or even 3:1 to correct sell/buy decisions. In other words, what you sell tends to do better than what you buy to replace, this is a big problem for amateur stock pickers and less so for investment professionals. I do know that this effect hits the professionals too. I am wondering if the algorithms and Artificial Intelligence have overcome this problem. This is one big reason that I warn against trading your portfolio a lot. Any research on this?
I think we have danced this dance before? I said your concerns were based on Fear, Uncertain, and Doubt. You said no. Could you please be specific on what type of incorrect sell/buy decisions you are talking about. This is a index strategy.

Let me be specific. The manager has no opinion on the direction of the market, of any individual security, and is trying to match the market's return. No position can be under/over weighted by 10%. So a index position of 1% could be no more than 1.1% to 09% This off-weighting would last 31 days and then may well be inverse. The portfolio is neutrally loaded on factors such as size, sector, value etc. The portfolio also has a tracking error of under 150 bps, which includes the bid/ask drag from the frequent trading. There is just not a lot of room for decisions here.

There has been plenty of academic studies on this. Most of the foundational stuff was written 30 years ago. Like I said in a earlier comment, I really wish I had a good non-proprietary source.
The thing is, if you are maintaining an index and selling the losers within that index for tax loss harvesting, it seems that you have to replace the losers with SOMETHING. The algorithms can work around the wash sale rules no problem, you can sell the losers and buy them back in 30 days. But what do you do in the interim? This isn't fear, uncertainty, and doubt. Just wondering what they do. I suppose they do something like sell Exxon and buy Chevron. It seems you would have friction from bid/ask spreads probably not a huge deal, as someone else said, maybe 10 to 20 basis points. You could use sampling to mostly replicate index performance, I suppose you could replicate the S&P 500 with 200 stocks. The thing is the trading must create some tracking error because the performance of what you sell isn't going to be exactly the performance of what you buy to replace. I don't think that what I am saying is illogical at all. You could get a slight performance benefit from the trading or a slight drag. My guess based upon experience is that over time it would be a drag.

The strategy as explained above is to sell losers to offset winners and over time get the tax basis in the portfolio higher and higher. The big savings are when the investors sells out because the stocks at that point should have relatively high basis. I get that.

If you are getting portfolio drag of under 150 basis points, that is still close to 1 1/2 percent a year. That is not indexing if you have that much drag. I think I read somewhere that the tax drag on the S&P 500 or the US Total Stock Market Index is 0.50% a year.

One reason that I am skeptical about this is that I owned two Quantitative Funds for years that were supposed to "improve" on the S&P 500. They were American Century Equity Growth, which tried to be a bit growthier than the Index and American Century Income and Growth, which tried to generate a dividend yield higher than the Index. Both funds kept the same sector weights as the index and both sampled with about 200 stocks and often with lower market caps than the Index. At one time, they were trying to capitalize a bit on the Size factor. I noticed that they now are doing this with fewer stocks 109 for Income and Growth and 123 for Equity Growth. For years, these funds did well, reliable outperforming the benchmark by about 1% a year. The last decade has been tough on these funds and they both have been trailing the S&P 500. They did this at 0.67% fees. The Vanguard S&P 500 Admiral Shares have returned 13.51% a year vs. 11.96% for Income and Growth vs. 12.42% annually for Equity Growth.

Granted, American Century was trying to beat the index and they were not investing for tax efficiency. Their failure to meet the benchmarks was probably due to factors, their fees, and increasing efficiency of the markets. It was hard enough to beat the index with quantitative strategies. You can see my skepticism over a higher turnover tax loss harvesting/offsetting realized capital gains from winners. The more trading, the more friction costs and those add up over time.

For the tax managed strategy to work, the benefit from realized capital losses and the resulting tax savings has to outweigh the management fee and the performance drag from all the trading.

Of course, the lower the fees to do this, the better. The robots charge 0.25% and in another thread someone was hiring a firm that was charging 0.70%. I would go for the robots. I also know that Vanguard used to have tax managed products but they are managed for all the investors in the fund and not for just one individual.

I saw an article in Advisor Perspectives that explains this. The question is how has this worked in actual practice? The article said that tax drag can be 1% to 3% a year. On an unmanaged index it is probably about 0.58% a year ((2% dividend yield x (.20 Capital Gains + .05 State Income Tax + .038 Net Investment Income Tax)) for wealthy individuals.

https://www.advisorperspectives.com/com ... d-indexing
A fool and his money are good for business.

User avatar
nedsaid
Posts: 12668
Joined: Fri Nov 23, 2012 12:33 pm

Re: optimal strategy for exploiting robo-advisors

Post by nedsaid » Sat Nov 09, 2019 5:34 pm

The question I have really boils down to this:

1) How much performance drag is there on a portfolio from the management fees and from the trading?

2) How much per year in basis points are the tax savings?

If 2 is bigger than 1, you have a winning strategy. Pretty simple question. Not sure there is a clear answer, my guess is that it is an "it depends" kind of answer.
A fool and his money are good for business.

Topic Author
ksJoe
Posts: 32
Joined: Wed Jul 11, 2007 9:40 pm

Re: optimal strategy for exploiting robo-advisors

Post by ksJoe » Sat Nov 09, 2019 5:40 pm

alex_686 wrote:
Sat Nov 09, 2019 12:19 pm
So, to answer your question: No - there should not be any holdings with a low cost basis. This is one of the basic premises of the strategy. Over 30 years there should have been enough volatility and low correlation stocks to rotate out all of the low cost basis stocks and reset the cost basis to a higher level. Strategies like this can have very high turnover - often over 100%.
I think our disconnect is that we are assuming very different approaches to investing.

I'm not interested in investing in individual stocks, and I'm not interested in having 100% turnover. I chose passive investing based on asset alloction, where sales are either for rebalancing, TLH, or (eventually) because I need the money. Within the context of this investment approach, I expect there to eventually be considerable assets that are highly unlikely to participate in additional TLH. These are the assets I prefer not to pay the robo fee on, which is the motivation for the strategy I posted.

illumination
Posts: 230
Joined: Tue Apr 02, 2019 6:13 pm

Re: optimal strategy for exploiting robo-advisors

Post by illumination » Sat Nov 09, 2019 6:04 pm

I just don't really see the value in this in terms of paying for it. Say you own 10 index funds (which is probably more than recommended) how often are you really tax loss harvesting and how long does this take to where you need a service for it?

As an example, I TLH'ed from VIOV to SLYV. I could easily see what lots I purchased were losses (I don't even have to do the math, Schwab can break it out by lot sale with a mouse click and tell me the exact loss from current price) I sold VIOV and purchased SLYV. It's like 5 minutes. And you probably aren't doing this constantly with all your positions. This is probably something you can spend a few minutes per year doing. Lower fees long terms is more important than tax loss harvesting anyway.

Where I could see robo-advising paying off is more when you have a huge amount of positions, which isn't really the Boglehead way. I don't want to go near that because you would have a mess if you ever wanted to stop (at least in a taxable account)

Where I will concede I have made a mistake before is when you have several accounts, like say say an HSA somewhere else and you hit a wash sale rule because of a reinvested dividend or something. But you would have to sign up for a roboadvisor for all of your accounts to properly manage that (unless you can just tell them?) I'd rather just not have the ongoing fees than squeeze out the last cent on TLH.

User avatar
nedsaid
Posts: 12668
Joined: Fri Nov 23, 2012 12:33 pm

Re: optimal strategy for exploiting robo-advisors

Post by nedsaid » Sun Nov 10, 2019 12:48 pm

I want to make it clear that I am in favor of tax loss harvesting as a tool though not as an investment strategy. By all means, if you have a big realized capital gain in one place sell another investment with a big unrealized loss to offset it. That is good tax management. I just wonder about letting the tax tail wag the investment dog. I also favor good tax managed mutual funds but over time these funds tend to be discontinued which defeats the purpose of owning them in the first place.

The strategy of a high turnover indexing strategy, aggressively trading to capture capital losses while tracking an index, seems like a laudable goal but I am skeptical it can actually be achieved. Another poster said that the performance drag pursuing such a strategy was less than 150 basis points or 1.5% a year but as I pointed out about the tax drag on a passive S&P 500 or Total Stock Market Index for wealthy individuals is probably 0.58% a year or so, less so for those of us in lower tax brackets. Another thing to consider is that the goal of a high turnover strategy is to eliminate the low tax basis stocks in your portfolio over time but this is likely at the cost of performance drag. Problem is that performance drag compounds fairly quickly and at some point might outweigh whatever tax savings you would get. It would depend upon how long one plans to hold on to the portfolio.

Again, Nirvana is achieved if you can track the index with little or no performance drag and generate a very tax efficient portfolio at the same time. Is this achievable?

John Bogle used to say that the drag of 100% portfolio turnover in a portfolio was 50 basis points or 0.50% a year, this included commissions, bid/ask spreads, and market impact. Trading costs are probably lower today and trading an individual's portfolio would have almost no market impact cost, today it would be mostly bid/ask spreads. Seeing that a firm would charge for this service, you have to get the performance drag to be less than the tax drag on a passive S&P 500 or US Total Stock Market Index which is 0.58% or so for those in high tax brackets.

You also have to take into account the tax basis of the portfolio, if you buy and hold an S&P 500 or Total Market Index for many years, it will be pretty tax efficient as long as you hold it but there will be large capital gains when you sell. Presumably using the high turnover indexing strategy would give you tax efficiency year after year as well as a much smaller capital gain when you liquidate the portfolio after many years. If the performance drag is minimal, this strategy would be a big winner over both the short and long term. If you can keep the performance drag from all the trading down to 20 basis points or 0.20% and if the robot charges you 0.25% a year, you have a shot. Problem is that when you turn over a portfolio, what you sell isn't going to perform exactly as what you buy to replace, my thesis is that over time that this will be a drag no matter how closely one matches the correlation and other characteristics of the stocks you are buying and the stocks you are selling. I wonder if the robots can overcome the "Nedsaid effect" that I describe above, even a minimal effect from incorrect sell/buy decisions add up over time. You have to pretty much eliminate this effect to make all this work.
Last edited by nedsaid on Sun Nov 10, 2019 1:14 pm, edited 2 times in total.
A fool and his money are good for business.

Lee_WSP
Posts: 1167
Joined: Fri Apr 19, 2019 5:15 pm
Location: Arizona

Re: optimal strategy for exploiting robo-advisors

Post by Lee_WSP » Sun Nov 10, 2019 1:00 pm

Couldn't the robots and the individual schedule both harvesting and contributions every thirty days? If there are any lots that are under, you sell that index and swap it for another index. It cannot be that hard to do.

Previously we were prevented from doing this because of trading fees, but now that they're gone....

kevinpet
Posts: 76
Joined: Fri Jan 13, 2012 1:27 pm
Location: CA
Contact:

Re: optimal strategy for exploiting robo-advisors

Post by kevinpet » Wed Nov 13, 2019 12:39 am

nedsaid wrote:
Sat Nov 09, 2019 4:48 pm
The thing is, if you are maintaining an index and selling the losers within that index for tax loss harvesting, it seems that you have to replace the losers with SOMETHING. The algorithms can work around the wash sale rules no problem, you can sell the losers and buy them back in 30 days. But what do you do in the interim? This isn't fear, uncertainty, and doubt. Just wondering what they do. I suppose they do something like sell Exxon and buy Chevron.
I have no knowledge of exactly how other firms do it, but the individual stock portion of Wealthfront accounts is based on minimizing tracking error to the index while maximizing tax-loss harvesting within a framework of quadratic optimization. This is covered in the Methodology section of https://research.wealthfront.com/whitep ... arvesting/ A later section covers some actual observed results.

The notion of selling Exxon and buying Chevron is accurate but the correlations don't always match our intuitions. If you have any specifics you are interested in, I can see if I can find FAQs or similar that address them. I generally avoid Wealthfront threads because I don't think anyone wants a company shill chiming in, but it seems appropriate here.

User avatar
nedsaid
Posts: 12668
Joined: Fri Nov 23, 2012 12:33 pm

Re: optimal strategy for exploiting robo-advisors

Post by nedsaid » Wed Nov 13, 2019 12:49 am

kevinpet wrote:
Wed Nov 13, 2019 12:39 am
nedsaid wrote:
Sat Nov 09, 2019 4:48 pm
The thing is, if you are maintaining an index and selling the losers within that index for tax loss harvesting, it seems that you have to replace the losers with SOMETHING. The algorithms can work around the wash sale rules no problem, you can sell the losers and buy them back in 30 days. But what do you do in the interim? This isn't fear, uncertainty, and doubt. Just wondering what they do. I suppose they do something like sell Exxon and buy Chevron.
I have no knowledge of exactly how other firms do it, but the individual stock portion of Wealthfront accounts is based on minimizing tracking error to the index while maximizing tax-loss harvesting within a framework of quadratic optimization. This is covered in the Methodology section of https://research.wealthfront.com/whitep ... arvesting/ A later section covers some actual observed results.

The notion of selling Exxon and buying Chevron is accurate but the correlations don't always match our intuitions. If you have any specifics you are interested in, I can see if I can find FAQs or similar that address them. I generally avoid Wealthfront threads because I don't think anyone wants a company shill chiming in, but it seems appropriate here.
In theory, what you described should work beautifully. You ought to be able to match the characteristics of the stocks you are selling with the characteristics of the stocks you are buying to replace. My thesis, is that over time you will get some drag doing this. Quants probably do a good job here, my off the wall guess is that the drag will be 20 to 30 basis points, a reduced "Nedsaid effect" but an effect nonetheless. Will add that to the 20 or so basis point drag from bid/ask spreads. I will read your white paper and see if my suspicions are correct.

I used the imperfect analogy of the American Century Quantitative Funds which try to "improve" the indexes. Pretty much saying that if quantitative analysis can't find stocks with better characteristics than the index as a whole, I am skeptical that you can exactly match the characteristics of stocks so that when you sell/buy that you keep the exact characteristics of the index. Plus I always think of Buffett's comment that as motion increases that returns decrease.

Reading the article, it does seem that the tax benefits from the Wealthfront strategy are more than I had realized. Ready to hit the sack, the brain is conking out, I will look at it tomorrow. Something for folks, particularly in higher tax brackets to consider. I like the idea of individual stock tax loss harvesting. It seems the idea does work, but again I need to read the article thoroughly. Some backtesting here and also some historical data. Thanks, this is what is needed.
A fool and his money are good for business.

User avatar
unclescrooge
Posts: 3893
Joined: Thu Jun 07, 2012 7:00 pm

Re: optimal strategy for exploiting robo-advisors

Post by unclescrooge » Wed Nov 13, 2019 10:57 am

nedsaid wrote:
Sat Nov 09, 2019 4:48 pm
alex_686 wrote:
Sat Nov 09, 2019 11:47 am
nedsaid wrote:
Sat Nov 09, 2019 11:06 am
I wonder if there are any academic studies regarding aggressive tax loss harvesting and investment returns. This strategy involves lots of turnover to generate the tax losses and I am wondering if this is creating a friction with bid/ask spreads and incorrect sell/buy decisions.
This strategy should be "passively trading", and not doing "information trading". In theory you can make money this way. In reality, a 200% to 300% turnover would generate a drag of about 10 to 20 bps.
nedsaid wrote:
Sat Nov 09, 2019 11:06 am
In other words, I am wondering if the "Nedsaid effect", the ratio of incorrect sell/buy decisions being 2:1 or even 3:1 to correct sell/buy decisions. In other words, what you sell tends to do better than what you buy to replace, this is a big problem for amateur stock pickers and less so for investment professionals. I do know that this effect hits the professionals too. I am wondering if the algorithms and Artificial Intelligence have overcome this problem. This is one big reason that I warn against trading your portfolio a lot. Any research on this?
I think we have danced this dance before? I said your concerns were based on Fear, Uncertain, and Doubt. You said no. Could you please be specific on what type of incorrect sell/buy decisions you are talking about. This is a index strategy.

Let me be specific. The manager has no opinion on the direction of the market, of any individual security, and is trying to match the market's return. No position can be under/over weighted by 10%. So a index position of 1% could be no more than 1.1% to 09% This off-weighting would last 31 days and then may well be inverse. The portfolio is neutrally loaded on factors such as size, sector, value etc. The portfolio also has a tracking error of under 150 bps, which includes the bid/ask drag from the frequent trading. There is just not a lot of room for decisions here.

There has been plenty of academic studies on this. Most of the foundational stuff was written 30 years ago. Like I said in a earlier comment, I really wish I had a good non-proprietary source.
The thing is, if you are maintaining an index and selling the losers within that index for tax loss harvesting, it seems that you have to replace the losers with SOMETHING. The algorithms can work around the wash sale rules no problem, you can sell the losers and buy them back in 30 days. But what do you do in the interim? This isn't fear, uncertainty, and doubt. Just wondering what they do. I suppose they do something like sell Exxon and buy Chevron. It seems you would have friction from bid/ask spreads probably not a huge deal, as someone else said, maybe 10 to 20 basis points. You could use sampling to mostly replicate index performance, I suppose you could replicate the S&P 500 with 200 stocks. The thing is the trading must create some tracking error because the performance of what you sell isn't going to be exactly the performance of what you buy to replace. I don't think that what I am saying is illogical at all. You could get a slight performance benefit from the trading or a slight drag. My guess based upon experience is that over time it would be a drag.

The strategy as explained above is to sell losers to offset winners and over time get the tax basis in the portfolio higher and higher. The big savings are when the investors sells out because the stocks at that point should have relatively high basis. I get that.

If you are getting portfolio drag of under 150 basis points, that is still close to 1 1/2 percent a year. That is not indexing if you have that much drag. I think I read somewhere that the tax drag on the S&P 500 or the US Total Stock Market Index is 0.50% a year.

One reason that I am skeptical about this is that I owned two Quantitative Funds for years that were supposed to "improve" on the S&P 500. They were American Century Equity Growth, which tried to be a bit growthier than the Index and American Century Income and Growth, which tried to generate a dividend yield higher than the Index. Both funds kept the same sector weights as the index and both sampled with about 200 stocks and often with lower market caps than the Index. At one time, they were trying to capitalize a bit on the Size factor. I noticed that they now are doing this with fewer stocks 109 for Income and Growth and 123 for Equity Growth. For years, these funds did well, reliable outperforming the benchmark by about 1% a year. The last decade has been tough on these funds and they both have been trailing the S&P 500. They did this at 0.67% fees. The Vanguard S&P 500 Admiral Shares have returned 13.51% a year vs. 11.96% for Income and Growth vs. 12.42% annually for Equity Growth.

Granted, American Century was trying to beat the index and they were not investing for tax efficiency. Their failure to meet the benchmarks was probably due to factors, their fees, and increasing efficiency of the markets. It was hard enough to beat the index with quantitative strategies. You can see my skepticism over a higher turnover tax loss harvesting/offsetting realized capital gains from winners. The more trading, the more friction costs and those add up over time.

For the tax managed strategy to work, the benefit from realized capital losses and the resulting tax savings has to outweigh the management fee and the performance drag from all the trading.

Of course, the lower the fees to do this, the better. The robots charge 0.25% and in another thread someone was hiring a firm that was charging 0.70%. I would go for the robots. I also know that Vanguard used to have tax managed products but they are managed for all the investors in the fund and not for just one individual.

I saw an article in Advisor Perspectives that explains this. The question is how has this worked in actual practice? The article said that tax drag can be 1% to 3% a year. On an unmanaged index it is probably about 0.58% a year ((2% dividend yield x (.20 Capital Gains + .05 State Income Tax + .038 Net Investment Income Tax)) for wealthy individuals.

https://www.advisorperspectives.com/com ... d-indexing
Is it possible this sort of of TLH worked better before the tech industry arrived with it's 80% gross margins?

I heard personal capital's spiel about TLH and it sounded similar, although they stopped short of 100% turn over.

lazydavid
Posts: 2547
Joined: Wed Apr 06, 2016 1:37 pm

Re: optimal strategy for exploiting robo-advisors

Post by lazydavid » Wed Nov 13, 2019 12:52 pm

I think this could have worked two years ago, when Wealthfront used Apex Holdings as its underlying broker/trading platform. When they migrated to maintaining their own accounts, I could no longer transfer in-kind. So now when I have RSUs vest, I have to sell them at UBS, ACH to my bank, and then ACH to Wealthfront.

That eliminates the option of carrying basis out of Wealthfront, and then back in. But even if it worked, now you'd be talking about potentially thousands of tax lots being transferred. That sounds like a massive pain, even if it supposedly works correctly.

User avatar
nedsaid
Posts: 12668
Joined: Fri Nov 23, 2012 12:33 pm

Re: optimal strategy for exploiting robo-advisors

Post by nedsaid » Wed Nov 13, 2019 11:15 pm

I took time to read the Weathfront article on their tax loss selling strategies. It seems that there is a drag on performance from all of the trading but that the tax savings outweighs the performance drag.

The performance information was presented in a confusing manner, I wish they would have presented the returns of their tax loss portfolios side by side with the Total Stock Market Index. It would have been great if they had then presented the after tax returns year by year. In their effort to simplify the data it tended to confuse me. The sense that I got was that you got a tracking error from the index of 1.72% to 1.82% but that your after tax returns were boosted from 2.15% to 2.66%. This was from backtested data.

They did have actual performance numbers but I was left pretty confused. What was clear was that most of the time you could harvest more tax losses from a portfolio of individual stocks than you could with a portfolio of ETFs.

The information confused me and I was confused by the after tax performance they were claiming. They were assuming a big initial deposit to get the account started and then that the clients would add to the account. If you are buying and holding a market index, there would be very little portfolio turnover thus most all of the taxes would be generated from dividends. My estimate is that the tax drag on a buy and hold of a Total Stock Market ETF would be 0.58% from a dividend yield of 2%. How can this strategy give you an increase after tax return of over 2% a year when the tax drag is only 0.58% a year? I just don't see where the big increases in after-tax returns are coming from as their example assumed a big initial investment and additions to the account. In the example, no one sold down their account but just added to it.

There was no discussion about the tax basis of the account and how a high turnover strategy would eventually get rid of the low basis stocks in the portfolio. The really big tax savings would come when you sell the portfolio after many years, buying and holding an Index Fund with low turnover would have a pretty large unrealized capital gains. If you sold a tax managed portfolio of individual stocks mimicking the index after many years, your unrealized capital gains within the portfolio would be relatively small compared to the Index Fund.

The problem is that they admitted that there would be a drag on portfolio performance. That annual performance drag of 1.7% to 1.8% a year also compounds and I suppose could over time wipe out the effect of the harvested capital losses. The problem is that if all you do is create capital losses, the tax code only allows you to take $3,000 in losses a year. You can offset capital gains with capital losses but you have to have realized capital gains to do this. You cannot offset unrealized capital gains. There was no discussing of selling winners within the portfolio that the losses could offset.

The information was unclear to me and presented in a confusing manner. It was clear to me that this could work. The problem is that this example didn't seem to be thought all the way through. There just seems to be something missing in the logic. They just say that they are generating this tax alpha but it isn't clear to me how they are doing this. Again, you are limited by the tax code in taking $3,000 in net capital losses a year. This isn't going to work if the client never sells or if you never have realized capital gains within the portfolio. Doesn't make sense to do tax loss selling if all you do is take your $3,000 capital losses a year and just build up your capital loss carryforward year after year. Don't you at some point want to use all that capital loss carryover? A performance drag of 1.7% to 1.8% a year seems like a pretty high price to pay.

I am not the smartest guy in the world but I am not the dumbest either. I just don't think their examples were very good and the data was not presented very well. I was left confused, it was like trying to keep track of which shell out of three has the peanut underneath. Very confusing.
A fool and his money are good for business.

Murgatroyd
Posts: 282
Joined: Sun Jan 21, 2018 8:23 pm

Re: optimal strategy for exploiting robo-advisors

Post by Murgatroyd » Thu Nov 14, 2019 3:38 pm

nedsaid wrote:
Wed Nov 13, 2019 11:15 pm
I took time to read the Weathfront article on their tax loss selling strategies. It seems that there is a drag on performance from all of the trading but that the tax savings outweighs the performance drag.

The performance information was presented in a confusing manner, I wish they would have presented the returns of their tax loss portfolios side by side with the Total Stock Market Index. It would have been great if they had then presented the after tax returns year by year. In their effort to simplify the data it tended to confuse me. The sense that I got was that you got a tracking error from the index of 1.72% to 1.82% but that your after tax returns were boosted from 2.15% to 2.66%. This was from backtested data.

They did have actual performance numbers but I was left pretty confused. What was clear was that most of the time you could harvest more tax losses from a portfolio of individual stocks than you could with a portfolio of ETFs.

The information confused me and I was confused by the after tax performance they were claiming. They were assuming a big initial deposit to get the account started and then that the clients would add to the account. If you are buying and holding a market index, there would be very little portfolio turnover thus most all of the taxes would be generated from dividends. My estimate is that the tax drag on a buy and hold of a Total Stock Market ETF would be 0.58% from a dividend yield of 2%. How can this strategy give you an increase after tax return of over 2% a year when the tax drag is only 0.58% a year? I just don't see where the big increases in after-tax returns are coming from as their example assumed a big initial investment and additions to the account. In the example, no one sold down their account but just added to it.

There was no discussion about the tax basis of the account and how a high turnover strategy would eventually get rid of the low basis stocks in the portfolio. The really big tax savings would come when you sell the portfolio after many years, buying and holding an Index Fund with low turnover would have a pretty large unrealized capital gains. If you sold a tax managed portfolio of individual stocks mimicking the index after many years, your unrealized capital gains within the portfolio would be relatively small compared to the Index Fund.

The problem is that they admitted that there would be a drag on portfolio performance. That annual performance drag of 1.7% to 1.8% a year also compounds and I suppose could over time wipe out the effect of the harvested capital losses. The problem is that if all you do is create capital losses, the tax code only allows you to take $3,000 in losses a year. You can offset capital gains with capital losses but you have to have realized capital gains to do this. You cannot offset unrealized capital gains. There was no discussing of selling winners within the portfolio that the losses could offset.

The information was unclear to me and presented in a confusing manner. It was clear to me that this could work. The problem is that this example didn't seem to be thought all the way through. There just seems to be something missing in the logic. They just say that they are generating this tax alpha but it isn't clear to me how they are doing this. Again, you are limited by the tax code in taking $3,000 in net capital losses a year. This isn't going to work if the client never sells or if you never have realized capital gains within the portfolio. Doesn't make sense to do tax loss selling if all you do is take your $3,000 capital losses a year and just build up your capital loss carryforward year after year. Don't you at some point want to use all that capital loss carryover? A performance drag of 1.7% to 1.8% a year seems like a pretty high price to pay.

I am not the smartest guy in the world but I am not the dumbest either. I just don't think their examples were very good and the data was not presented very well. I was left confused, it was like trying to keep track of which shell out of three has the peanut underneath. Very confusing.
Glad to see you're coming around to the opinion that this can work. I'm going to again try to put some meat on this bone as perhaps the only Bogleheads participant that owns this strategy via over 3 years in the Fidelity Tax Advantaged SMA. I'm not as technically adept as you and others, but I will try....

Here's actual results: This account tracks the S&P 500. Through October, the S&P 3 year pre-tax annualized return is up +14.91%. The 3 year pre-tax annualized return for my account AFTER FEE is +14.89%. Adding back the average fee of .52% the return would be +15.41%. The Harvested yield ( I like that term from the Wealthfront White Paper) has averaged 3.53%. In other words, if I were to do my taxes today, I have taken a cumulative 3.53% of the invested amount in net capital losses to my tax return. I am not a fan of the tax alpha statistic and interestingly, Fidelity and Wealthfront calculate it differently. My actual after tax return is not impressive because I'm only in the 12% bracket right now due to early retirement and only living off the dividends, that said, it is 15.68% versus the S&P at 14.6%. As a comparison to the Wealthfront white paper harvested yields we can compare 2017 and 2018 (because I only had a partial year in 2016) for their stock segment versus my account which is all stocks. For 2017 WF harvested 2.17%, Fido was 2.12%. For 2018 WF was 3.38%, Fido was 5.17%. I suspect the 2018 difference is because of human intervention at Fido.

Obviously, I am delighted that the investment return is strong enough to offset the fees to only .02% below the target. My sense is Fidelity has made the determination that customers would be happier with investment return than forgoing harvested tax losses which would be unknown. So, apparently, a focus on return can eliminate tracking error. That said, beating the market could be an anomaly for this 3-year period. We opened the account expecting to incur the fee and are prepared to do just that if this positive performance trend disappears.

Fidelity informs that an important part of the process is identifying groups of stocks that behave the same way. This is how they can sell and buy and maintain performance while avoiding wash sale rules. I believe this is continually updated because I've observed different stock groups going in and out over time. That's about all they say other than the ability to harvest is almost totally dependent on market volatility and that some techniques are proprietary.

The account holds around 250 stocks at any time. What folks might not appreciate is there are typically 3 to 6 tax lots of each stock so they can trade on small movements in price. Its not just holding one tranche each of the 250. There is also far more turnover than folks can imagine, in excess of 200%. I can certainly appreciate Wealthfront letting the computer do all the work. As you might think per the amount of holdings, there can be periods of significant momentum selling. When the market swooned late in 2018 the amount of activity was head spinning. It also resulted in large tax losses. This year has been tough because the market has been sideways to up. I noticed in the Wealthfront paper significant tax harvesting differences by year over the longer term. To be specific, this year has dragged down my cumulative harvested losses which was above 4%. I have little doubt that when the market burps they will take advantage. When we were discussing this SMA with our advisor he was clear that we need to think about this account long term.

So far the only downside I have found is there are some unicorn stocks with long term gains that do nothing but appreciate and are hard to sell. As one might expect they are FAANGS and a couple of other high flyers. These tranches are less than 2% of the account. They may account for some of the strong performance. Even so, short term lots of Amazon, Microsoft and Apple have been bought and sold numerous times. While we have used some of the excess losses above the $3,000 annual deduction to offset other cap gains in the self directed portion of our portfolio, I simply told my advisor that we can work together on a time to sell the unicorns and balance against the accumulated tax loss carry-forwards that we have. This would be hard to do at Wealthfront.

I know this sounds like a pitch for the account but we have just been delighted with it. I've said in other threads, I would not buy any of the other SMA's Fidelity offers.

Regards,
Murgatroyd

User avatar
nedsaid
Posts: 12668
Joined: Fri Nov 23, 2012 12:33 pm

Re: optimal strategy for exploiting robo-advisors

Post by nedsaid » Thu Nov 14, 2019 7:55 pm

Murgatroyd wrote:
Thu Nov 14, 2019 3:38 pm
nedsaid wrote:
Wed Nov 13, 2019 11:15 pm
I took time to read the Weathfront article on their tax loss selling strategies. It seems that there is a drag on performance from all of the trading but that the tax savings outweighs the performance drag.

The performance information was presented in a confusing manner, I wish they would have presented the returns of their tax loss portfolios side by side with the Total Stock Market Index. It would have been great if they had then presented the after tax returns year by year. In their effort to simplify the data it tended to confuse me. The sense that I got was that you got a tracking error from the index of 1.72% to 1.82% but that your after tax returns were boosted from 2.15% to 2.66%. This was from backtested data.

They did have actual performance numbers but I was left pretty confused. What was clear was that most of the time you could harvest more tax losses from a portfolio of individual stocks than you could with a portfolio of ETFs.

The information confused me and I was confused by the after tax performance they were claiming. They were assuming a big initial deposit to get the account started and then that the clients would add to the account. If you are buying and holding a market index, there would be very little portfolio turnover thus most all of the taxes would be generated from dividends. My estimate is that the tax drag on a buy and hold of a Total Stock Market ETF would be 0.58% from a dividend yield of 2%. How can this strategy give you an increase after tax return of over 2% a year when the tax drag is only 0.58% a year? I just don't see where the big increases in after-tax returns are coming from as their example assumed a big initial investment and additions to the account. In the example, no one sold down their account but just added to it.

There was no discussion about the tax basis of the account and how a high turnover strategy would eventually get rid of the low basis stocks in the portfolio. The really big tax savings would come when you sell the portfolio after many years, buying and holding an Index Fund with low turnover would have a pretty large unrealized capital gains. If you sold a tax managed portfolio of individual stocks mimicking the index after many years, your unrealized capital gains within the portfolio would be relatively small compared to the Index Fund.

The problem is that they admitted that there would be a drag on portfolio performance. That annual performance drag of 1.7% to 1.8% a year also compounds and I suppose could over time wipe out the effect of the harvested capital losses. The problem is that if all you do is create capital losses, the tax code only allows you to take $3,000 in losses a year. You can offset capital gains with capital losses but you have to have realized capital gains to do this. You cannot offset unrealized capital gains. There was no discussing of selling winners within the portfolio that the losses could offset.

The information was unclear to me and presented in a confusing manner. It was clear to me that this could work. The problem is that this example didn't seem to be thought all the way through. There just seems to be something missing in the logic. They just say that they are generating this tax alpha but it isn't clear to me how they are doing this. Again, you are limited by the tax code in taking $3,000 in net capital losses a year. This isn't going to work if the client never sells or if you never have realized capital gains within the portfolio. Doesn't make sense to do tax loss selling if all you do is take your $3,000 capital losses a year and just build up your capital loss carryforward year after year. Don't you at some point want to use all that capital loss carryover? A performance drag of 1.7% to 1.8% a year seems like a pretty high price to pay.

I am not the smartest guy in the world but I am not the dumbest either. I just don't think their examples were very good and the data was not presented very well. I was left confused, it was like trying to keep track of which shell out of three has the peanut underneath. Very confusing.
Glad to see you're coming around to the opinion that this can work. I'm going to again try to put some meat on this bone as perhaps the only Bogleheads participant that owns this strategy via over 3 years in the Fidelity Tax Advantaged SMA. I'm not as technically adept as you and others, but I will try....

Here's actual results: This account tracks the S&P 500. Through October, the S&P 3 year pre-tax annualized return is up +14.91%. The 3 year pre-tax annualized return for my account AFTER FEE is +14.89%. Adding back the average fee of .52% the return would be +15.41%. The Harvested yield ( I like that term from the Wealthfront White Paper) has averaged 3.53%. In other words, if I were to do my taxes today, I have taken a cumulative 3.53% of the invested amount in net capital losses to my tax return. I am not a fan of the tax alpha statistic and interestingly, Fidelity and Wealthfront calculate it differently. My actual after tax return is not impressive because I'm only in the 12% bracket right now due to early retirement and only living off the dividends, that said, it is 15.68% versus the S&P at 14.6%. As a comparison to the Wealthfront white paper harvested yields we can compare 2017 and 2018 (because I only had a partial year in 2016) for their stock segment versus my account which is all stocks. For 2017 WF harvested 2.17%, Fido was 2.12%. For 2018 WF was 3.38%, Fido was 5.17%. I suspect the 2018 difference is because of human intervention at Fido.

Obviously, I am delighted that the investment return is strong enough to offset the fees to only .02% below the target. My sense is Fidelity has made the determination that customers would be happier with investment return than forgoing harvested tax losses which would be unknown. So, apparently, a focus on return can eliminate tracking error. That said, beating the market could be an anomaly for this 3-year period. We opened the account expecting to incur the fee and are prepared to do just that if this positive performance trend disappears.

Fidelity informs that an important part of the process is identifying groups of stocks that behave the same way. This is how they can sell and buy and maintain performance while avoiding wash sale rules. I believe this is continually updated because I've observed different stock groups going in and out over time. That's about all they say other than the ability to harvest is almost totally dependent on market volatility and that some techniques are proprietary.

The account holds around 250 stocks at any time. What folks might not appreciate is there are typically 3 to 6 tax lots of each stock so they can trade on small movements in price. Its not just holding one tranche each of the 250. There is also far more turnover than folks can imagine, in excess of 200%. I can certainly appreciate Wealthfront letting the computer do all the work. As you might think per the amount of holdings, there can be periods of significant momentum selling. When the market swooned late in 2018 the amount of activity was head spinning. It also resulted in large tax losses. This year has been tough because the market has been sideways to up. I noticed in the Wealthfront paper significant tax harvesting differences by year over the longer term. To be specific, this year has dragged down my cumulative harvested losses which was above 4%. I have little doubt that when the market burps they will take advantage. When we were discussing this SMA with our advisor he was clear that we need to think about this account long term.

So far the only downside I have found is there are some unicorn stocks with long term gains that do nothing but appreciate and are hard to sell. As one might expect they are FAANGS and a couple of other high flyers. These tranches are less than 2% of the account. They may account for some of the strong performance. Even so, short term lots of Amazon, Microsoft and Apple have been bought and sold numerous times. While we have used some of the excess losses above the $3,000 annual deduction to offset other cap gains in the self directed portion of our portfolio, I simply told my advisor that we can work together on a time to sell the unicorns and balance against the accumulated tax loss carry-forwards that we have. This would be hard to do at Wealthfront.

I know this sounds like a pitch for the account but we have just been delighted with it. I've said in other threads, I would not buy any of the other SMA's Fidelity offers.

Regards,
Murgatroyd
Here's the deal, tax loss yield is a bogus concept in my opinion. Reason being is that let's say $3,000 in capital losses will result in maybe $864 in tax savings when offsetting long term capital gains. I take the $3,000 x (.20 capital gains tax rate + 0.038 Net Investment Income Tax + .05 State Income Tax Rate) to get the $864. What I am trying to say is that you don't get dollar for dollar in tax savings from a realized capital loss. In my example, you get 28.8 cents for every dollar of capital loss. Bogus concept.

The other thing I am trying to get across is that net capital losses by themselves are capped at $3,000 that you can take each year per the tax code. To really get the benefit, they need to be offset by realized Capital Gains elsewhere.

What you need to do is calculate your after tax return and it varies for everyone. Some folks are in the 0% Long Term Capital Gains Tax Bracket, the 15% LTCG Tax Bracket, and the 20% LTCG Tax Bracket. Also, lower incomes don't have to worry about the 3.8% Net Investment Income Tax. Income tax rates among the states vary, I use 5% as sort of an average. A few states have no State Income Tax, most have lower rates, my state tops out at a 9.9% tax bracket. In addition, some cities and local governments have their own income tax. So the after tax return is going to vary for everyone.

What I would say is that the higher income earner you are, the more you would benefit from this strategy. Were I a wealthy billionaire, I would give lots of consideration. People of more modest means and who are in lower tax brackets probably ought not to bother. You just have to work out the numbers. As I said, this CAN work and I thank you for providing real life numbers. If you can get the realized capital losses and yet have little tracking error from an index, you have a winning strategy.

The other thing I would say is you need to focus on the tax basis or cost for your account. What you want is for the tax basis of the portfolio to keep going up because presumably your low cost stocks will be sold and replaced with stocks with a higher basis. It sounds like Fidelity has a focus of generating capital losses and not on the overall tax basis of the portfolio. Again, the capital losses do limited good until you have realized Long Term Capital Gains elsewhere to offset.
A fool and his money are good for business.

Lee_WSP
Posts: 1167
Joined: Fri Apr 19, 2019 5:15 pm
Location: Arizona

Re: optimal strategy for exploiting robo-advisors

Post by Lee_WSP » Thu Nov 14, 2019 10:27 pm

nedsaid wrote:
Thu Nov 14, 2019 7:55 pm
The other thing I am trying to get across is that net capital losses by themselves are capped at $3,000 that you can take each year per the tax code. To really get the benefit, they need to be offset by realized Capital Gains elsewhere.
I believe the rule is $3,000 of your regular income per year, but you can bank it for future years and can offset unlimited cap gains with losses.

User avatar
nedsaid
Posts: 12668
Joined: Fri Nov 23, 2012 12:33 pm

Re: optimal strategy for exploiting robo-advisors

Post by nedsaid » Thu Nov 14, 2019 10:34 pm

Lee_WSP wrote:
Thu Nov 14, 2019 10:27 pm
nedsaid wrote:
Thu Nov 14, 2019 7:55 pm
The other thing I am trying to get across is that net capital losses by themselves are capped at $3,000 that you can take each year per the tax code. To really get the benefit, they need to be offset by realized Capital Gains elsewhere.
I believe the rule is $3,000 of your regular income per year, but you can bank it for future years and can offset unlimited cap gains with losses.
That is correct.
A fool and his money are good for business.

Post Reply