Factor investing portfolio construction advice

Have a question about your personal investments? No matter how simple or complex, you can ask it here.
Post Reply
Topic Author
Morik
Posts: 984
Joined: Tue Nov 25, 2014 12:26 pm

Factor investing portfolio construction advice

Post by Morik »

I would like to build a factor investing portfolio.

Please keep discussion of whether or not factor investing is a reasonable strategy to other threads (e.g., viewtopic.php?t=379541).

I have two constraints due to fund availability in retirement accounts.
1. The 401k only allows 99% of the balance to be sent to a self directed brokerage account.
Fund options in the 401k are S&P500, extended market index, intl developed markets index, emerging markets index.
1% of the balance is something like half a percent of the overall portfolio. I plan to just hold the extended market index here.

2. The 403b has the following available:

Contract 1: TIAA traditional, liquid version. I plan to keep this as part of my fixed income allocation.
Contract 2: TIAA traditional, illiquid, in transfer payout annuity being moved over to contract 3. Something like 5 years left here. Also counted towards fixed income.

Contract 3 relevant funds:
DFCEX (DFA EM Core), 0.39% ER
DFIEX (DFA Intl core), 0.24% ER
DFISX (DFA Intl small), 0.39% ER
DFQTX (DFA US Core 2), 0.19% ER
DFLVX (DFA US LCV), 0.22% ER
DFSTX (DFA US small), 0.27% ER
TISBX (TIAA small cap blend), 0.05% ER
VIMAX (Vanguard mid-cap index), 0.05% ER


The 403b contract 3 is ~10% of the entire portfolio.


I am interested in targeting size, value, and momentum, though I am open to switching this around. My main goal is to load up on risk factors; I chose these as it seems momentum pairs well with value and size/value are very easy to invest in.

I prefer to hold international at roughly market cap. For simplicity I'll plan to maintain a 60/40 US/intl split. If the ratio changes significantly (e.g., becomes 70/30 or 50/50), I would update to match.

Desired stock/bond ratio: 75/25. I realize that this is going to be significantly more volatile than a 75/25 holding VT (whole world market) as its equity.

US equities are 45% of the portfolio.
I like the idea of using XSVM (US SCV + Momentum) + VFMO (momentum).
VFMO has a lot of size factor exposure due to holding a wide range of market caps compared to MTUM.

XSVM over the recent past has had negative momentum loading though when I look at regressions... I am guessing this is because it is primarily a SCV fund that tries to pick up momentum within that space?

I think using VFMO get me more reliable momentum loading--something like:
30% VFMO
15% XSVM

Over past year, the pure XSVM portfolio would have ~0 momentum, vs 0.25 or so for the VFMO portfolio. Though the XSVM portfolio also has stronger size & value loading over this period.

International equities are 30% of the portfolio.

I am currently thinking I would use DFCEX in the 403b, plus Avantis funds to get small & value loads. They screen for profitability and they can also use discretion on when to trade which can help cut trading costs in less liquid markets.

Option 1:
15% AVDV (Intl SCV)
5% AVES (EM value)
10% DFCEX (EM value)

Option 2--add international momentum (vs just getting it all on the US side):
10% IMTM
10% AVDV (Intl SCV)
10% DFCEX (EM value)

Note that option 2 lowers EM exposure. EEMO (EM momentum) is the only EM momentum-primary fund I have found. There are some other EM funds that don't primarily target momentum but do consider it, such as UEVM.
But these funds appear to have a lot of negative alpha (EEMO had -6.66% annual alpha over a 10 year period, -10% over the past 4 years or so) when I look at regressions, which makes me less inclined to use them.

I could instead drop XSVM to 5% and not take away from the EM allocation, using DFSTX (US SCV) in the 403b.

But in any case, there is no option like XSVM or VFMO in the international space that I can see--all of the momentum funds will come with negative size & ~0 value loading.


My current thinking is to just take momentum on the US side. I also think I'd rather use DFA emerging markets than take up space in the US allocation.

So the overall portfolio then would look like this:
30% VFMO (US momentum)
15% XSVM (US SCV + momentum)
15% AVDV (Intl SCV)
10% DFCEX (EM value)
5% AVES (EM value)
20% Nominal fixed income
5% inflation-linked fixed income

The fixed income portion consists of TIAA traditional, I-bonds, EE-bonds, EDV, and LTPZ. I will glide to lower bond durations as my investment horizon shrinks.


Most of the portfolio is in tax advantaged accounts. The taxable portion of the portfolio is currently ~6% of the overall portfolio. If I use VFMO I would hold that in taxable. If I don't use VFMO, I would probably hold XSVM in taxable--on a brief look it looks more tax efficient than AVDV, which would be the other option to hold in taxable.


Questions:
1. Any major issues that I'm not seeing here?

2. Should I try to get momentum loading internationally? Or just do it on the US side to minimize offsetting value/size loading?

3. Best option to hold in the 403b? Replace some AVES with DFA EM core? Or hold DFSTX for US small/value?
typical.investor
Posts: 3268
Joined: Mon Jun 11, 2018 3:17 am

Re: Factor investing portfolio construction advice

Post by typical.investor »

Morik wrote: Mon Jun 20, 2022 12:11 pm XSVM over the recent past has had negative momentum loading though when I look at regressions... I am guessing this is because it is primarily a SCV fund that tries to pick up momentum within that space?

I think using VFMO get me more reliable momentum loading--something like:
30% VFMO
15% XSVM

Over past year, the pure XSVM portfolio would have ~0 momentum, vs 0.25 or so for the VFMO portfolio. Though the XSVM portfolio also has stronger size & value loading over this period.
Well, what is to say that a higher mom loading will result in better capturing of the momentum premium? I ask because it's kinda assumed here that it will which is not I think correct.

Also, one year seems really short to get an accurate picture of a fund.

Finally, I know momentum tests well, but what funds have really successfully captured mom. I see knowledgeable firms like AQR and Alpha Architect not seemingly doing well in the momentum space. It just seems difficult for a real fund to do. That was my own conclusion anyway.

I think you should read through the fund construction of XSVM and your other options to see what makes sense to you.

Back to my first question, in a period where the value premium is positive, you might assume a find with a higher value loading will have higher returns. It's empirically disproved though. The most common value definition is book/market. And if that is your definition then funds whose performance is explained by that will load high on value. But if your fund screens by 1/price to earnings, then it won't have as strong a loading using a b/m loading. Yet, the 1/price to earnings may actually have better performance due to its 'value' exposure. This is true over a long period. I believe the 1/price to earnings fund will have a higher alpha (unexplained) loading but if you regressed by a 1/price to earnings measure, the opposite would be true. It would have a higher 'value' loading than a b/m fund.

So to me, loadings are information that help explain fund performance, but since we don't know what the best construction of value is, we can't really use loadings to pick a best fund.
Topic Author
Morik
Posts: 984
Joined: Tue Nov 25, 2014 12:26 pm

Re: Factor investing portfolio construction advice

Post by Morik »

typical.investor wrote: Tue Jun 21, 2022 3:56 am
Morik wrote: Mon Jun 20, 2022 12:11 pm XSVM over the recent past has had negative momentum loading though when I look at regressions... I am guessing this is because it is primarily a SCV fund that tries to pick up momentum within that space?

I think using VFMO get me more reliable momentum loading--something like:
30% VFMO
15% XSVM

Over past year, the pure XSVM portfolio would have ~0 momentum, vs 0.25 or so for the VFMO portfolio. Though the XSVM portfolio also has stronger size & value loading over this period.
Well, what is to say that a higher mom loading will result in better capturing of the momentum premium? I ask because it's kinda assumed here that it will which is not I think correct.

Also, one year seems really short to get an accurate picture of a fund.

Finally, I know momentum tests well, but what funds have really successfully captured mom. I see knowledgeable firms like AQR and Alpha Architect not seemingly doing well in the momentum space. It just seems difficult for a real fund to do. That was my own conclusion anyway.

I think you should read through the fund construction of XSVM and your other options to see what makes sense to you.

Back to my first question, in a period where the value premium is positive, you might assume a find with a higher value loading will have higher returns. It's empirically disproved though. The most common value definition is book/market. And if that is your definition then funds whose performance is explained by that will load high on value. But if your fund screens by 1/price to earnings, then it won't have as strong a loading using a b/m loading. Yet, the 1/price to earnings may actually have better performance due to its 'value' exposure. This is true over a long period. I believe the 1/price to earnings fund will have a higher alpha (unexplained) loading but if you regressed by a 1/price to earnings measure, the opposite would be true. It would have a higher 'value' loading than a b/m fund.

So to me, loadings are information that help explain fund performance, but since we don't know what the best construction of value is, we can't really use loadings to pick a best fund.
Right, so I understand that getting a higher loading on a factor won't necessarily capture the 'true' underlying premium perfectly (assuming there is a real underlying premium); it depends on the particular construction of the factor which can deviate as you said from different constructions.

I've read Larry's book on factor investing--there was a chapter where he talked about the value thing you are talking about. But while the premium actually captured by any particular construction will vary, a reasonable construction should still capture that premium (at least, a chunk of it).

Going back to momentum--sure different constructions will vary in their premium/explanatory power and I'll see various levels of alpha depending on exactly how the factor is constructed. But looking at the data, the FF momentum factor construction does capture a momentum premium over time. Sure there are other constructions that may end up working better, but as you said that isn't something we can identify apriori.

But surely this doesn't mean you can't target a factor--there should be a value premium on reasonable constructions of value, whether that is b/m, 1/price, etc.

If I look back over the history of VFMO (march 2018-present) using FF 4-factor + term & credit, the annual alpha is very small (-0.12%) and it had a 0.3 load on the FF 4-factor momentum.
And sure different constructions work differently--if I look at Alpha Architect 5-factor + term & credit, the annual alpha is -0.45% and momentum loading on the AA momentum factor was 0.4.

If I look at FF statistics from that same period, the momentum premium was 0.15%. AA factors over the exact same period say momentum returned 2.64%. So it happened to be that AA momentum factor construction shows a premium, while the FF momentum construction doesn't over that period.
But if the momentum factor is real, reasonable constructions should be capturing some of it, and increasing my loading on a reasonable construction of it should allow me to capture some of that return.

Looking at Alpha Architect's QMOM, over its life (jan 2016-present) it had a 0.42 load on the AA momentum factor, and the return of the AA momentum factor over that time period was 0.7%.
FF momentum returned -1.33% that period, and QMOM had a load of 0.59 on that.
And QMOM also had negative alpha of -1.5 (FF) or -2 (AA).
I would say QMOM hasn't done well because the momentum factor (at least, those two constructions of it) didn't do well over that period. I wouldn't say QMOM didn't capture any of it...
typical.investor
Posts: 3268
Joined: Mon Jun 11, 2018 3:17 am

Re: Factor investing portfolio construction advice

Post by typical.investor »

Morik wrote: Tue Jun 21, 2022 10:25 am If I look at FF statistics from that same period, the momentum premium was 0.15%. AA factors over the exact same period say momentum returned 2.64%. So it happened to be that AA momentum factor construction shows a premium, while the FF momentum construction doesn't over that period.
But if the momentum factor is real, reasonable constructions should be capturing some of it, and increasing my loading on a reasonable construction of it should allow me to capture some of that return.
Yes, I would agree that in general that a reasonable construction should capture some of the premium. I believe this definitely for value but less so for momentum.

Anyway, this might be worth a read to help in the decision on how much XSVM to use.
https://alphaarchitect.com/2015/03/the- ... trategies/
and updated
https://alphaarchitect.com/2021/05/valu ... -separate/

That research would perhaps suggest holding US SCV in a separate fund.

For me, I am a little skeptical on how well momentum funds can, after trading costs, capture the premium.

I don't see major issues in your portfolios though. They look fine assuming you can tolerate the difference in return from the market when your chosen factors are not doing so well.
Topic Author
Morik
Posts: 984
Joined: Tue Nov 25, 2014 12:26 pm

Re: Factor investing portfolio construction advice

Post by Morik »

typical.investor wrote: Tue Jun 21, 2022 6:50 pm
Morik wrote: Tue Jun 21, 2022 10:25 am If I look at FF statistics from that same period, the momentum premium was 0.15%. AA factors over the exact same period say momentum returned 2.64%. So it happened to be that AA momentum factor construction shows a premium, while the FF momentum construction doesn't over that period.
But if the momentum factor is real, reasonable constructions should be capturing some of it, and increasing my loading on a reasonable construction of it should allow me to capture some of that return.
Yes, I would agree that in general that a reasonable construction should capture some of the premium. I believe this definitely for value but less so for momentum.

Anyway, this might be worth a read to help in the decision on how much XSVM to use.
https://alphaarchitect.com/2015/03/the- ... trategies/
and updated
https://alphaarchitect.com/2021/05/valu ... -separate/

That research would perhaps suggest holding US SCV in a separate fund.

For me, I am a little skeptical on how well momentum funds can, after trading costs, capture the premium.

I don't see major issues in your portfolios though. They look fine assuming you can tolerate the difference in return from the market when your chosen factors are not doing so well.
I think the article is interesting but the methodology they used is hard to compare with the funds that are in the wild.
I cannot seem to go prior to 1/1/1992 when looking at the Alpha Architect factor returns on PV (I was trying to find the returns over the periods they used in their various charts in those articles).
I'd be interested in what the various factor loadings at for those different constructions...

I am not too concerned about the XSVM # of holdings--sure 120 companies is a lot fewer than the ~400-600 held by other SCV funds I was looking at, but it should be sufficient to avoid much idiosyncratic risk.

After giving things some more thought and poking around a lot more in portfolio visualizer, I think I do actually want to pick up some momentum internationally as well.

My current thinking is:
15% VFMO
30% XSVM
10% DFCEX
10% AVDV
10% IMTM
(25% fixed income)

Or up to 25% VFMO/20% XSVM.

XSVM seems like a nice way to get loadings on size/value/momentum to tie the other assets together. I expect it to generally always be loading on size/value, while it seems to fluctuate a bit on whether it has positive or negative momentum loading--but I expect that over longer periods of time it will have higher momentum loading than using AVUV or VIOV, without sacrificing much (if anything) on the size/value side.

I would hold VFMO in taxable.
Topic Author
Morik
Posts: 984
Joined: Tue Nov 25, 2014 12:26 pm

Re: Factor investing portfolio construction advice

Post by Morik »

Ok, so I'm going with:

15% VFMO
30% XSVM
10% DFCEX
10% AVDV
10% IMTM
20% Nominal fixed income (duration matched)
5% inflation-linked fixed income (duration matched)


If I can find a reasonable way to invest in certain alternatives I would shave off some of the equity allocation and put it there. Instead of 75/25, I'd do 60/25/15 (equities/fixed income/alternatives). Specifically alternatives that have reasonable expected returns net of costs, and are less correlated with market returns. Variance risk premium, carry factor, reinsurance, etc.

I am still looking into this. The investments listed in Larry Swedroe's Black Swans book are not available to me as I don't use an advisor, and even if I had access they have high expenses which makes me cautious regarding how much that digs into the expected return.

The Vanguard alternative strategies fund looks interesting: Some of its portfolio is targeting the carry premium, and various other things that are less correlated with market risk. But it is expensive at 1.28% ER, and I am not sure how to evaluate how well it has met its goals over time.

Another interesting one is CCRV (Commodity Curve Carry Strategy). It's net ER is 0.4% which is a bit more manageable than 1.28%. It hasn't been around very long and the very high returns over the past year make me think its return is driven more by commodity returns than capturing the carry premium... and its investment objective seems to indicate this as well: "seeks to track the investment results of an index composed of commodities with the top ten highest ranking roll yields".
I haven't learned very much about commodities in terms of investing, and it isn't clear to me how the expected volatility/return of this fund would differ from a commodities fund that wasn't using a carry strategy.

So while I am interested in potentially adding alternatives to this portfolio, I need to learn more about them and investigate available options further.



Prior to using a factor tilted portfolio I was holding 75/25 VT/fixed income. My thought at that time was that I'd stay at this allocation probably forever. With the factor tilts added I am giving some thought to eventually dropping to 60/40. If I also had an alternatives allocation I'd probably do something like 50/35/15 (equities/fixed income/alternatives).
quietseas
Posts: 640
Joined: Fri Dec 27, 2013 4:43 pm

Re: Factor investing portfolio construction advice

Post by quietseas »

Personally I'd be careful about holding VFMO in taxable. It still only has $198.5M in assets in over 4 years of operation. Vanguard closed out similar funds in Europe due to low asset levels/low demand and in the US has merged mutual funds with 10X that level into other funds. At current level of assets it certainly is not what Vanguard was hoping for. I don't think Vanguard has any interest in funds that long term have less than $5B of assets in them these days. I'd have to guess there is a decent probability that Vanguard will close this ETF and liquidate it, or merge it into something else. No one knows but you can bet internal discussions about this have happened at Vanguard. If you buy it in taxable you have to be prepared for this possibility which is much more likely than for example iShares doing the same to MTUM which has $10B in it.
Topic Author
Morik
Posts: 984
Joined: Tue Nov 25, 2014 12:26 pm

Re: Factor investing portfolio construction advice

Post by Morik »

quietseas wrote: Sat Jun 25, 2022 1:37 pm Personally I'd be careful about holding VFMO in taxable. It still only has $198.5M in assets in over 4 years of operation. Vanguard closed out similar funds in Europe due to low asset levels/low demand and in the US has merged mutual funds with 10X that level into other funds. At current level of assets it certainly is not what Vanguard was hoping for. I don't think Vanguard has any interest in funds that long term have less than $5B of assets in them these days. I'd have to guess there is a decent probability that Vanguard will close this ETF and liquidate it, or merge it into something else. No one knows but you can bet internal discussions about this have happened at Vanguard. If you buy it in taxable you have to be prepared for this possibility which is much more likely than for example iShares doing the same to MTUM which has $10B in it.
I am not familiar with the effects of ETF closure. I assume at the very least there would be a forced sale which would realize any unrealized capital gains. Are there other effects as well? (E.g., a capital gains distribution at closure time?)

For my personal tax situation and analyzing the 2021 distribution data, the tax drag of each fund I would be able to hold in taxable is:
- VFMO: 19 bp
- XSVM: 38 bp
- AVDV: 60 bp
- IMTM: 119 bp
- EDV: 72 bp (I think? 0% qualified dividends AFAICT. This is calculating state taxes though, I think I wouldn't pay those as EDV is all treasuries?)
- LTPZ: Didn't look it up, I assume it is similarly not very efficient.
(I used the Tax efficiency 2021 sheet & added the data for these funds to it)

So I guess the question is whether to hold VFMO to save ~20bp of tax drag and hope that they don't close the fund & force me to realize gains, or hold XSVM and accept the higher tax drag.


The other question would be what to do if VFMO goes away--MTUM would cancel out a lot of size/value factor compared to VFMO... I guess I could just move it all into XSVM, or re-examine which factors I want to target.
Last edited by Morik on Sat Jun 25, 2022 4:53 pm, edited 1 time in total.
rkhusky
Posts: 12796
Joined: Thu Aug 18, 2011 8:09 pm

Re: Factor investing portfolio construction advice

Post by rkhusky »

I doubt Vanguard would liquidate the ETF. More likely would be a rollover into another ETF or change the investment strategy to gain more AUM. If you didn't like the resulting ETF, you would have to sell and realize the gains, if any.
quietseas
Posts: 640
Joined: Fri Dec 27, 2013 4:43 pm

Re: Factor investing portfolio construction advice

Post by quietseas »

Morik wrote: Sat Jun 25, 2022 3:08 pm
For my personal tax situation and analyzing the 2021 distribution data, the tax drag of each fund I would be able to hold in taxable is:
- VFMO: 19 bp
- XSVM: 38 bp
- AVDV: 60 bp
- IMTM: 119 bp
- EDV: 72 bp (I think? 0% qualified dividends AFAICT. This is calculating state taxes though, I think I wouldn't pay those as EDV is all treasuries?)
- LTPZ: Didn't look it up, I assume it is similarly not very efficient.
(I used the Tax efficiency 2021 sheet & added the data for these funds to it)
What's in tax advantaged and what's in taxable? Where are your new investments going to go over the remainder of your income earning career?

There's a reason total market equity indexed investments are usually recommended as the predominant choice in taxable accounts. If you plan to hold this investment for decades I would not rely on the whims of the fund manager or the IRS to give favor to the ETF structure. I would put investments in taxable that are inherently more tax efficient.

I'm NOT trying to talk you out of your portfolio but rather suggesting you consider the "known unknowns" like tax changes and fund manager changes before you commit to it especially in a taxable account where you have a big cost with making changes later. I'm old enough to remember back in the 1990s when active funds like Fidelity Magellan or Vanguard PRIMECAP were often recommended for taxable accounts. Whatever benefits this portfolio might have from a factor theory perspective could easily be wiped out by tax changes you have no control over.
Topic Author
Morik
Posts: 984
Joined: Tue Nov 25, 2014 12:26 pm

Re: Factor investing portfolio construction advice

Post by Morik »

quietseas wrote: Sat Jun 25, 2022 5:30 pm
Morik wrote: Sat Jun 25, 2022 3:08 pm
For my personal tax situation and analyzing the 2021 distribution data, the tax drag of each fund I would be able to hold in taxable is:
- VFMO: 19 bp
- XSVM: 38 bp
- AVDV: 60 bp
- IMTM: 119 bp
- EDV: 72 bp (I think? 0% qualified dividends AFAICT. This is calculating state taxes though, I think I wouldn't pay those as EDV is all treasuries?)
- LTPZ: Didn't look it up, I assume it is similarly not very efficient.
(I used the Tax efficiency 2021 sheet & added the data for these funds to it)
What's in tax advantaged and what's in taxable? Where are your new investments going to go over the remainder of your income earning career?

There's a reason total market equity indexed investments are usually recommended as the predominant choice in taxable accounts. If you plan to hold this investment for decades I would not rely on the whims of the fund manager or the IRS to give favor to the ETF structure.

I'm NOT trying to talk you out of your portfolio but rather suggesting you consider the "known unknowns" like tax changes and fund manager changes before you commit to it especially in a taxable account where you have a big cost with making changes later. I'm old enough to remember back in the 1990s when active funds like Fidelity Magellan or Vanguard PRIMECAP were often recommended for taxable accounts. Whatever benefits this portfolio might have from a factor theory perspective could easily be wiped out by tax changes you have no control over.
Portfolio is ~$1.8m, taxable is $120k so about 6.7%. We have had most of our savings in retirement accounts, and used existing taxable balances over the past several years on purchasing our current home/home improvements/etc. These expenses are mostly wrapped up now, but I do anticipate another $30k-$40k spent over the next few years on a few remaining items we want to change.

I'm not including series I & series EE bonds as taxable since they grow tax deferred.

A large component of my income (nearly half) is linked to RSUs so I can't precisely predict savings amounts, but our savings going forward look something like this:
- Around $100k to tax advantaged accounts annually. This includes traditional 401k + employer match, traditional 403b + employer match, after-tax 401k, backdoor IRAs for both of us, and an HSA that we max out.
- Likely $30k to i-bonds (me, spouse, trust).
- Additional savings past that would go into the taxable brokerage account. If I didn't do $30k of i-bonds that money would go into taxable brokerage instead. Assuming $30k goes to ibonds, I expect something like $50k-100k to be available here annually.

So the taxable account growth from contributions is going to be something like 25-45% of the total portfolio growth from contributions, and the growth of the portfolio from contributions is ~10-13%. If the market falls further this percentage would rise, but since half my income is in vesting RSUs, my savings rate would also fall a little.

So my plan, assuming VFMO doesn't get the axe, would be to hold VFMO in taxable. If taxable rises to more than 15% of the portfolio, I'd start adding XSVM to taxable. I don't anticipate the taxable portion will grow to more than 45% of the portfolio, though I suppose depending on how I structure withdrawals in retirement that it could potentially happen then.
User avatar
drumboy256
Posts: 415
Joined: Sat Jun 06, 2020 2:21 pm

Re: Factor investing portfolio construction advice

Post by drumboy256 »

Morik wrote: Sat Jun 25, 2022 11:44 am Ok, so I'm going with:

15% VFMO
30% XSVM
10% DFCEX
10% AVDV
10% IMTM
20% Nominal fixed income (duration matched)
5% inflation-linked fixed income (duration matched)


If I can find a reasonable way to invest in certain alternatives I would shave off some of the equity allocation and put it there. Instead of 75/25, I'd do 60/25/15 (equities/fixed income/alternatives). Specifically alternatives that have reasonable expected returns net of costs, and are less correlated with market returns. Variance risk premium, carry factor, reinsurance, etc.

I am still looking into this. The investments listed in Larry Swedroe's Black Swans book are not available to me as I don't use an advisor, and even if I had access they have high expenses which makes me cautious regarding how much that digs into the expected return.

The Vanguard alternative strategies fund looks interesting: Some of its portfolio is targeting the carry premium, and various other things that are less correlated with market risk. But it is expensive at 1.28% ER, and I am not sure how to evaluate how well it has met its goals over time.

Another interesting one is CCRV (Commodity Curve Carry Strategy). It's net ER is 0.4% which is a bit more manageable than 1.28%. It hasn't been around very long and the very high returns over the past year make me think its return is driven more by commodity returns than capturing the carry premium... and its investment objective seems to indicate this as well: "seeks to track the investment results of an index composed of commodities with the top ten highest ranking roll yields".
I haven't learned very much about commodities in terms of investing, and it isn't clear to me how the expected volatility/return of this fund would differ from a commodities fund that wasn't using a carry strategy.

So while I am interested in potentially adding alternatives to this portfolio, I need to learn more about them and investigate available options further.



Prior to using a factor tilted portfolio I was holding 75/25 VT/fixed income. My thought at that time was that I'd stay at this allocation probably forever. With the factor tilts added I am giving some thought to eventually dropping to 60/40. If I also had an alternatives allocation I'd probably do something like 50/35/15 (equities/fixed income/alternatives).
This was an interesting read. I'm guessing for duration matched bonds you're using ZROZ/EDV/VGLT and LTPZ for Tips? Again, assumptions. Composition wise, are you wanting US momentum + international value? Also, the EM segment, any aversion to the Avantis funds? Both AVEM and AVES are value based but have different spreads in terms of stocks. Obviously the biggest component for EM is value + currency hedge from US markets but not sure if the DFA fund is more or less worth it. (Depends on your preference on composition).

My only other comment is on VFMO. It's interesting because its newer and most of the "active" funds either skew really heavily concentrated and of course, carry a premium for expense ratio. Vanguard, oddly, does neither of those things (which is interesting) and I would almost say it's a weirdly equal weighted fund that feels more or less like the fund managers have FOMO across all cap types. The one other fund that (ironically) is passively managed but more concentrated is is XMMO (INVESCO S AND P MIDCAP MOMENTUM ETF). It has a has a higher AUM and volume trades but as mentioned is more concentrated than VFMO. However, if you look at the overlap (https://www.etfrc.com/funds/overlap.php) between XSVM and VFMO, you get about 40%. If you pair XVSM + XMMO you get 0%. For me, it checks a couple of boxes: 1) is passively managed against an SP index that you can track and/or anticipate a bit better 2) AUM is higher and 3) industry weights (for me) are more favorable. You also mentioned running VFMO in taxable, XMMO is about on par or the same as Vanguards fund but for me, is a bit more "factor'ish" so to speak.

As to the commodities, there is a great post by HedgeFundie (link: viewtopic.php?t=275899). I do not actively recommend people to own commodities as they (imo) as they are as volatile as Crypto is. That said, depending on personal risk tolerance, preference and timeline, they could be a useful tool. Since I don't want to get messaged, the only commodity that I do have in my portfolio (that is rather small at this point) is FIW - FIRST TRUST WATER ETF, that to me, is more or less "fun money" of which I sell profits from a mHFEA strategy into that fund. Overall, aside from the boilerplate "commodities are in VTSAX...." statement, overweighting them based on the portfolio is an interesting proposition.

Hope the dialogue was useful.
Promise is one thing. Fulfilling that promise is quite another. - Sir Alex Ferguson | // Merri-Bogle WW SCV + Chill
Topic Author
Morik
Posts: 984
Joined: Tue Nov 25, 2014 12:26 pm

Re: Factor investing portfolio construction advice

Post by Morik »

drumboy256 wrote: Sat Jun 25, 2022 6:48 pm This was an interesting read. I'm guessing for duration matched bonds you're using ZROZ/EDV/VGLT and LTPZ for Tips? Again, assumptions.
For nominal I use EDV + series EE bonds + the liquid version of TIAA traditional (+ temporarily the illiquid version of TIAA traditional, currently in the middle of a transfer payout annuity, something like 5 years left on that). I'd also consider using GOVZ/ZROZ.
Once my investing horizon shrinks enough that I need shorter durations I'd start blending in VGLT/etc that are < 20 years duration.

For inflation-linked I'm using series I bonds + LTPZ. My LTPZ position is really small and is likely to disappear soon as I add $30k of i-bonds per year (and right now I have a little less than $30k in LTPZ).
I will either:
- end up increasing my bond allocation (keeping 20% in nominal, increasing the inflation-linked allocation)
- end up shifting more from nominal to inflation-linked, though my impression is that nominals are better diversifiers for an equity-heavy portfolio.
- stop purchasing i-bonds so my inflation-linked fixed income doesn't exceed 5% of the portfolio.
drumboy256 wrote: Sat Jun 25, 2022 6:48 pm Composition wise, are you wanting US momentum + international value? Also, the EM segment, any aversion to the Avantis funds? Both AVEM and AVES are value based but have different spreads in terms of stocks. Obviously the biggest component for EM is value + currency hedge from US markets but not sure if the DFA fund is more or less worth it. (Depends on your preference on composition).
I am interested in targeting size + value + momentum globally, but it seems it is easier to get a good amount of each in the US with a fund like XSVM + a momentum fund that isn't negatively loaded on size/value. Internationally I did opt to add IMTM. It does offset some of the size/value loading, but I do want the momentum load it gives me. This is not a clear cut decision for me though.
No intl momentum: https://www.portfoliovisualizer.com/fac ... ation3_1=0

Intl size: 0.84
Intl value: 0.46
Intl mom: 0.03

With momentum: https://www.portfoliovisualizer.com/fac ... tion3_1=34

Lower EM allocation...
Intl size: 0.27
Intl value: 0.16
Intl mom: 0.22

I get momentum loading, but I do give up a lot of size & value loading for it... not sure whether its worth the tradeoff. If there were a smaller cap and/or more valuey momentum fund internationally that had a reasonable expense ratio that would definitely interest me. I may end up not using momentum internationally unless I can find such a fund...
That would make my ex-US allocation 10% DFCEX, 5% AVES, 15% AVDV. (See below for why DFCEX.)
I think I'll actually drop IMTM--I've gone back and forth several times in my mind on this, so I think I'll go with the simpler option of not including it.
If I find a reasonably priced international momentum or multifactor ETF that has neutral or positive size/value loading & doesn't show significant negative alpha, I'd use that here.

As for the DFCEX instead of an Avantis fund--10% of the portfolio is in a 403b where the available relevant fund options are:
DFCEX (DFA EM Core), 0.39% ER
DFIEX (DFA Intl core), 0.24% ER
DFISX (DFA Intl small), 0.39% ER
DFQTX (DFA US Core 2), 0.19% ER
DFLVX (DFA US LCV), 0.22% ER
DFSTX (DFA US small), 0.27% ER
TISBX (TIAA small cap blend), 0.05% ER
VIMAX (Vanguard mid-cap index), 0.05% ER

So my choices were:

Hold one of the US funds replacing 10% of XSVM or VFMO:
- Hold TISBX or VIMAX and mostly just get the size + market load for that holding
- Hold DFLVX and mostly get a value + market factor load
- Hold DFQTX with a small bit of size, value, profitability + market loading.
- Hold DFSTX displacing some XSVM--this gets me size & a little value loading, but isn't as valuey as XSVM, and also doesn't have the momentum load.

I'd have to give up some momentum loading in the US if I went with a US fund in the 403b.

Hold one of the Intl funds replacing the 10% AVDV:
- Hold DFIEX -- seems to load on market & credit factors, negative loading on momentum, term, and size factors. (FF 4 factor model + term & credit)
- Hold DFISX -- this does load on size but doesn't really load value.

Hold one of the EM funds replacing 10% AVES:
- Hold DFCEX -- this is what I considered my best option out of all of these. It has a little bit of value loading in addition to market beta, but I feel like I give up the least here. AVES does load more value, and also has some size loading... but the difference between DFISX & AVDV is bigger I think. So on balance this option minimizes the loss of factor loads compared to the other options.


One other option would be TIAA traditional, illiquid version. I really don't like the illiquidity here. I feel like taking DFCEX instead of AVES is a better trade-off than adding to the illiquid version of TRAD available in this contract.
drumboy256 wrote: Sat Jun 25, 2022 6:48 pm My only other comment is on VFMO. It's interesting because its newer and most of the "active" funds either skew really heavily concentrated and of course, carry a premium for expense ratio. Vanguard, oddly, does neither of those things (which is interesting) and I would almost say it's a weirdly equal weighted fund that feels more or less like the fund managers have FOMO across all cap types. The one other fund that (ironically) is passively managed but more concentrated is is XMMO (INVESCO S AND P MIDCAP MOMENTUM ETF). It has a has a higher AUM and volume trades but as mentioned is more concentrated than VFMO. However, if you look at the overlap (https://www.etfrc.com/funds/overlap.php) between XSVM and VFMO, you get about 40%. If you pair XVSM + XMMO you get 0%. For me, it checks a couple of boxes: 1) is passively managed against an SP index that you can track and/or anticipate a bit better 2) AUM is higher and 3) industry weights (for me) are more favorable. You also mentioned running VFMO in taxable, XMMO is about on par or the same as Vanguards fund but for me, is a bit more "factor'ish" so to speak.
Ah I somehow missed XMMO when I was looking for momentum ETFs. In the period since VFMO came around (march 2018 to present), it does look fairly similar in terms of factor loads; a little less size & value load (by .08/.09), a little more momentum (by 0.04), negative credit factor compared to VFMO, and then some positive alpha.
I think I still prefer VFMO to it but its a close call. XMMO could certainly replace VFMO if it ends up closing.

And FDMO is another that might work (Fidelity momentum ETF). It's mostly LCG, but the factor regression shows neutral size & value loading. The rolling regression shows that the neutral size & value loadings has been pretty stable for the life of the fund.

FDEV also looks like it got reasonable momentum loading as well as size/value in the international space, but it has had a lot of negative alpha (> 6% annual)...
drumboy256 wrote: Sat Jun 25, 2022 6:48 pm As to the commodities, there is a great post by HedgeFundie (link: viewtopic.php?t=275899). I do not actively recommend people to own commodities as they (imo) as they are as volatile as Crypto is. That said, depending on personal risk tolerance, preference and timeline, they could be a useful tool. Since I don't want to get messaged, the only commodity that I do have in my portfolio (that is rather small at this point) is FIW - FIRST TRUST WATER ETF, that to me, is more or less "fun money" of which I sell profits from a mHFEA strategy into that fund. Overall, aside from the boilerplate "commodities are in VTSAX...." statement, overweighting them based on the portfolio is an interesting proposition.

Hope the dialogue was useful.
Thanks--I did find that thread while poking around trying to learn more about alternatives.
Adding a utilities position is an interesting proposition.
Looking at the factor loads on XLU from 2010-2022, it had only 0.5 market loading (so indeed should be less correlated with market returns), -0.37 size, 0.31 value, 0.25 momentum, 0.42 term, and 0.18 credit.
Though when I look at the rolling regression it has had generally negative size load, generally positive momentum term & value loading.
I'll investigate this further.

I do get a good bit of term factor from the long bonds, value & momentum from my factor tilts. My portfolio doesn't have much of a credit factor load. I wonder what other alternatives might be exposed to credit factor risk...

There are several utilities ETFs that look interesting:
JXI (global utilities etf, no EM exposure though), JHMU (US utilities with some factor filtration/weighting), or the cheaper XLU/VPU/FUTY (US utilities).
I do find it interesting that the factor regression models (I tried several) show low R squared for utilities; usually < 50%. This was also pointed out in that thread you linked.
Topic Author
Morik
Posts: 984
Joined: Tue Nov 25, 2014 12:26 pm

Re: Factor investing portfolio construction advice

Post by Morik »

I do find the idea of adding a sector fund for Utilities to this portfolio interesting.

Just to recap, currently the portfolio I have in mind looks like this:
15% VFMO
30% XSVM
15% AVDV
15% DFCEX/AVES (DFCEX due to fund availability in 403b, outside of 403b using AVES)
20% nominal fixed income
5% inflation-linked fixed income

I looked at rolling correlations between US utilities (XLU | VPU) and global utilities (JXI, which is ~2/3 US utilities and 1/3 ex-US developed market utilities).
Image

I'm comparing with VBR and VSS because the Avantis ETFs don't have enough history.

I found it interesting that US utilities had a lower correlation to US small cap value than international utilities did, at least over some of these periods.
Image

Looking at correlations with intl small caps, again US utilities had a lower correlation overall than international utilities did. For this period .61 correlation of VSS & JXI, vs 0.35 for VSS & VPU.
Image

My takeaways:
- Utilities had relatively low correlation with US & international small caps.
- Adding international utilities dampened the effect. Either there is something special going on with US utilities compared to international ones, or the variation in correlation just happened to favor US utilities over this period.

Assuming US utilities are NOT special in some way, this implies to me that the range of the correlation going forward is probably fairly wide. E.g., note in 2014 for the comparison with VBR that the correlation with US utilities was 0.2, while correlation with (66% us, 34% intl) utilities was 0.6. I don't know of a good way to check whether the US holdings in JXI were similar to VPU or not... so either shifting 1/3 of the utilities to be international caused the correlation to be a lot higher for whatever reason, or the US holdings of JXI were sufficiently different from VPU's holdings to increase the correlation by that much, or some combination of the two.

In any case, despite international utilities having higher correlations with VSS/VBR than US utilities did, the correlation was still low (< 0.8).
Just for kicks I looked at correlations with EDV (vs utilities, and vs vss/vbr).
Image

Image

So utilities had a much higher correlation with long bonds than vbr/vss did, though still low (<0.5).


In the ex-US space I have only found JXI as a global utilities etf. Other ex-US etfs are 'infrastructure' rather than broad utilities. E.g., IGF, PAVE, EMIF, etc.
These infrastructure ETFs have had higher correlations to other equities than the broad utilities have... E.g.,
Image

I compared US-only utilities to US-only infrastructure--I was wondering whether the higher correlation of the ex-US infra funds was just due to international utility-like things having higher correlations than the US ones. I.e., maybe the reason JXI had higher correlation to VBR/VSS is the same reason these ex-US infrastructure ETFs are showing higher correlations. But it looks like the correlation between utilities & infrastructure isn't super high:
Image

This makes me think infrastructure is not a reasonable proxy for utilities.


Things I'm thinking about:
- If this phenomenon continues, adding a utilities position to this portfolio would be adding an asset class that has somewhat low correlation with my other equity assets.

- On the other hand, they are more correlated with long bonds than my other equity assets would be, though the correlation is still fairly low.

- JXI is the only international broad utilities sector ETF I could find, and it includes US. So it seems my options are either:
a) Just use JXI, pay ~0.4% ER. Global utilities exposure. E.g.,

12% VFMO
24% XSVM
15% JXI
12% AVDV
12% DFCEX/AVES (DFCEX due to fund availability in 403b, outside of 403b using AVES)
20% nominal fixed income
5% inflation-linked fixed income

This maintains 60/40 US/intl exposure.

b) Just use VPU/XLU, pay ~0.08% ER. US only utilities exposure. With a 10-15% position in the portfolio carved out from the other equities, this would shift my AA to a bit of a home country bias.
E.g.,

12% VFMO
24% XSVM
15% VPU
12% AVDV
12% DFCEX/AVES (DFCEX due to fund availability in 403b, outside of 403b using AVES)
20% nominal fixed income
5% inflation-linked fixed income

This shifts US/Intl to ~66/34 instead of 60/40.

c) Take a US utilities position out of just my US equities and leave international alone:

10% VFMO
20% XSVM
15% VPU
15% AVDV
15% DFCEX/AVES (DFCEX due to fund availability in 403b, outside of 403b using AVES)
20% nominal fixed income
5% inflation-linked fixed income

This keeps the US/Intl exposure where it was.

d) Don't add a utilities position to this portfolio.


After writing this out I'm leaning towards adding a utilities position... but I'm not sure whether to go US-only and save ~33 bp on the ER, or go intl with the assumption that the higher correlation of ex-US utilities was just luck of the draw, and going forward if I just go with US only it could easily go the other way with US utilities having a higher correlation than the intl ones do.
My instinct is to say 'US only is good enough', save the 33 bp.
Regarding introducing a home country bias, I'm thinking I'd rather do that (have the utility position take equally from my other equities) than to take it all from VFMO/XSVM...



EDIT: And some portfolio backtests.

https://www.portfoliovisualizer.com/bac ... ation8_3=5

Only 2 years of history there though, swapping out AVDV for DFA's intl small value fund gets a little more:

https://www.portfoliovisualizer.com/bac ... ation8_3=5

Still only a few years though--swapping out VFMO for MTUM gets a longer period:
https://www.portfoliovisualizer.com/bac ... ation8_3=5

In general it looks like adding JXI or VPU to the portfolio improved risk adjusted returns.

EDIT 2: I also looked at some utilities ETFs that apply factor screening (PUI, JHMU). Compared to the non-factor US utilities (XLU/VPU) they had higher correlations with funds like VIOV:
Image
Post Reply