Four-Factor-Only Portfolio?

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Topic Author
duffer
Posts: 400
Joined: Tue Sep 09, 2014 8:18 pm

Four-Factor-Only Portfolio?

Post by duffer »

I am thinking about converting the equity portion of my portfolio into a Four-Factor-Only portfolio, i.e., in equities, I would hold only funds or ETFs representing the factors of size, value, momentum, and profitability (or quality). The percentage in each might be 25%, but I need to know more about that issue also.

So, my question(s):

Can folks here help me with the pros and cons of such a Four-Factor-Only portfolio and also the question of how I should approach the question of what percentage should be in each factor?

(If there were funds or ETFs for illiquidity, I might add that and make it a five-factor-only portfolio.)
stlutz
Posts: 5585
Joined: Fri Jan 02, 2009 12:08 am

Re: Four-Factor-Only Portfolio?

Post by stlutz »

I think the answer depends in part on what you want the portfolio to do. Is the goal more to:

a) invest in funds that will outperform the market over time?
b) invest in funds that have low correlation with each other thus providing lower portfolio volatility and a rebalancing bonus even though each ETF may or may not beat the market over the long haul.

I think the answers may be different depending upon your underlying assumptions.
randomguy
Posts: 11285
Joined: Wed Sep 17, 2014 9:00 am

Re: Four-Factor-Only Portfolio?

Post by randomguy »

It is pretty much impossible to get rid of beta exposure without some speciality products. Just finding funds that don't counter act each other ( getting momentum from one fund and avoiding negative momentum in another) is pretty darn hard also.
Topic Author
duffer
Posts: 400
Joined: Tue Sep 09, 2014 8:18 pm

Re: Four-Factor-Only Portfolio?

Post by duffer »

stlutz wrote:I think the answer depends in part on what you want the portfolio to do. Is the goal more to:

a) invest in funds that will outperform the market over time?
b) invest in funds that have low correlation with each other thus providing lower portfolio volatility and a rebalancing bonus even though each ETF may or may not beat the market over the long haul.

I think the answers may be different depending upon your underlying assumptions.
Can you expand on this a bit more? It would help my understanding a lot if you were to say what one might do with a 4-factor portfolio if a) were the goal or if b) were the goal.

Thanks for your help.
Topic Author
duffer
Posts: 400
Joined: Tue Sep 09, 2014 8:18 pm

Re: Four-Factor-Only Portfolio?

Post by duffer »

randomguy wrote:It is pretty much impossible to get rid of beta exposure without some speciality products. Just finding funds that don't counter act each other ( getting momentum from one fund and avoiding negative momentum in another) is pretty darn hard also.
Can you explain a bit more?

I was thinking of using the iShares factor ETFs: MTUM, QUAL, VLUE, and SIZE. What variations in balancing these funds would produce different diversification/volatility outcomes?

Thanks for your help.
stlutz
Posts: 5585
Joined: Fri Jan 02, 2009 12:08 am

Re: Four-Factor-Only Portfolio?

Post by stlutz »

In terms of the robustness of the historical return data, momentum would be #1, low volatility (not one of your suggested options) would be #2, value would be #3, size is a distant #4 and quality #5.

Let's assume you were after "beating the market". I'd throw quality out altogether. Size would only be useful in conjunction with value (in historical backtests, the value premium is much larger with small stocks than large stocks). And low-volatility is about performance on a risk-adjusted basis, not absolute, so you might toss that out as well. So, at that point you'd probably make momentum your largest holding, small value as #2, and maybe a small holding in quality.

Suppose you were instead focused more on correlation. In that case, I would pick funds for all 5 factors I listed and just weight them equally. At any given time, some of you funds will be outperforming and others will not.
I was thinking of using the iShares factor ETFs: MTUM, QUAL, VLUE, and SIZE.
A couple of notes on VLUE and SIZE. VLUE doesn't have that big of a value orientation to it. IUSV might be a better choice for overall value. For small value, I'd consult other threads here for discussions on the options. VBR is the best starting point, however. "SIZE" is not really a smallcap fund--it's a mislabled lower volatility fund (securities are weighted in reverse volatility order. It coincidentally tilts away from megacap stocks as a result. If you want a smallcap fund, I think VB or IJR are much better places to start.

Hope that's all at least slightly helpful. :happy
randomguy
Posts: 11285
Joined: Wed Sep 17, 2014 9:00 am

Re: Four-Factor-Only Portfolio?

Post by randomguy »

duffer wrote:
randomguy wrote:It is pretty much impossible to get rid of beta exposure without some speciality products. Just finding funds that don't counter act each other ( getting momentum from one fund and avoiding negative momentum in another) is pretty darn hard also.
Can you explain a bit more?

I was thinking of using the iShares factor ETFs: MTUM, QUAL, VLUE, and SIZE. What variations in balancing these funds would produce different diversification/volatility outcomes?

Thanks for your help.
If you buy a fund with +.5 momentum and another fund with -.5 momentum, you end up with 0 load of momentum. Those funds have a short history but you can go to portfolio visualizer to pull up some numbers.
Topic Author
duffer
Posts: 400
Joined: Tue Sep 09, 2014 8:18 pm

Re: Four-Factor-Only Portfolio?

Post by duffer »

stlutz wrote:In terms of the robustness of the historical return data, momentum would be #1, low volatility (not one of your suggested options) would be #2, value would be #3, size is a distant #4 and quality #5.

Let's assume you were after "beating the market". I'd throw quality out altogether. Size would only be useful in conjunction with value (in historical backtests, the value premium is much larger with small stocks than large stocks). And low-volatility is about performance on a risk-adjusted basis, not absolute, so you might toss that out as well. So, at that point you'd probably make momentum your largest holding, small value as #2, and maybe a small holding in quality.

Suppose you were instead focused more on correlation. In that case, I would pick funds for all 5 factors I listed and just weight them equally. At any given time, some of you funds will be outperforming and others will not.
I was thinking of using the iShares factor ETFs: MTUM, QUAL, VLUE, and SIZE.
A couple of notes on VLUE and SIZE. VLUE doesn't have that big of a value orientation to it. IUSV might be a better choice for overall value. For small value, I'd consult other threads here for discussions on the options. VBR is the best starting point, however. "SIZE" is not really a smallcap fund--it's a mislabled lower volatility fund (securities are weighted in reverse volatility order. It coincidentally tilts away from megacap stocks as a result. If you want a smallcap fund, I think VB or IJR are much better places to start.

Hope that's all at least slightly helpful. :happy
This is very helpful, and gives me some good homework to do. How do you get the info you have about the composition of the various IShares factor funds?

Also, Larry Swedroe had an interesting post in another thread about factor investing. He seems to be pointing to the beat-the-market vs. non-correlatation dimension that you were citing, but he doesn't say--at least in the post--what the variations might be to get different effects. How would one find out how the various factors are correlated? I have ordered his book which is cited in his post. The quote from him is below:

Larry Swedroe wrote:
" Modern financial theory demonstrates that there are a variety of identifiable factors that exploit almost all the risk and return of diversifiable portfolios.
The leading/big ones are beta, size, value, momentum and profitability (quality) on the stocks side (you can add liquidity) and term and credit on bond side.
Each of them are persistent and pervasive across markets and time and even across asset classes in many cases (like value and momentum)

They also have low correlation with each other, so the whole is greater than sum of parts (Markowitz). So the question then is how much exposure do you want to each of the factors.

There is no one right portfolio, IMO but there is one right for you and perhaps the most important issue is choosing to only take risks you understand because otherwise you'll likely not be able to adhere to the plan when the factors perform poorly for extended periods, which will likely happen to each of them at some point (which is why IMO you diversify across them). So choosing one that you can stick with is more important than choosing some "perfect" plan.

If you want to see the evidence and the logic for the way I personally invest then read Reducing the Risk of Black Swans---it's only about a 2-3 hour read."
User avatar
backpacker
Posts: 1620
Joined: Mon Sep 22, 2014 2:17 pm

Re: Four-Factor-Only Portfolio?

Post by backpacker »

duffer wrote:
randomguy wrote:It is pretty much impossible to get rid of beta exposure without some speciality products. Just finding funds that don't counter act each other ( getting momentum from one fund and avoiding negative momentum in another) is pretty darn hard also.
Can you explain a bit more?

I was thinking of using the iShares factor ETFs: MTUM, QUAL, VLUE, and SIZE. What variations in balancing these funds would produce different diversification/volatility outcomes?

Thanks for your help.
Hi duffer! You probably want to read through Robert T's Collective Thoughts if you haven't already. Tilting can be intellectually and financially rewarding (though there are of course no guarantees). It's easy to set the portfolio up, but understanding what the various parts are doing is important for staying the course through tough times.

VLUE is the same thing as Vanguard's Large Value fund. SIZE is a midcap fund. The size premium, if there is one, is finicky. It seems to work when combined with value (smaller value has historically done better than larger value). It also seems to work at the extremes, i.e. in the very smallest of the micro caps. The iShares fund doesn't get you either small value or extreme micro caps, so it's not worth the extra expense IMO. I'm not convinced that there is any such thing as a "quality" factor. And in any case, adding more factors can often end up just diluting your tilt rather than improving it. So I wouldn't bother with QUAL.

My own portfolio is a value portfolio with a mild tilt towards smaller stocks. MTUM is an interesting fund. I could eventually be convinced to replace a bit of my mid cap value with MTUM.

Domestic:
25% VOE (mid cap value)
20% PSXV (small value)
5% BRSIX (micro)

International:
25% PXF (large value)
10% VSS (small)
15% PXH (emerging large value)
mickens16
Posts: 340
Joined: Wed Apr 06, 2011 8:52 am

Re: Four-Factor-Only Portfolio?

Post by mickens16 »

Another excellent post by Robert T.

http://www.bogleheads.org/forum/viewtop ... 8#p2206358
stlutz
Posts: 5585
Joined: Fri Jan 02, 2009 12:08 am

Re: Four-Factor-Only Portfolio?

Post by stlutz »

How do you get the info you have about the composition of the various IShares factor funds?
I look for information from the index provider. For example, here is a brief facts sheet on the underlying index for SIZE: http://www.ishares.com/us/products/2514 ... factor-etf. It states, "Each MSCI Risk Weighted Index is constructed by reweighting the constituents of its market cap weighted parent index. To derive a risk weight for each security in the parent index, the security's 3-year weekly historical local return is calculated first. The risk weight is then computed as the ratio of the inverse of the security variance to the sum of the inverse of the security variances of all constituents in the parent index." So, the index doesn't explicitly take size into consideration at all. It's just that the result is that it ends up with mostly large-to-midcap stocks (see the M* stylebox: http://etfs.morningstar.com/quote?t=size).
stlutz
Posts: 5585
Joined: Fri Jan 02, 2009 12:08 am

Re: Four-Factor-Only Portfolio?

Post by stlutz »

Also, Larry Swedroe had an interesting post in another thread about factor investing. He seems to be pointing to the beat-the-market vs. non-correlatation dimension that you were citing, but he doesn't say--at least in the post--what the variations might be to get different effects. How would one find out how the various factors are correlated?
My take on all of this is different from Larry's and the other "expert tilters" on this board.

My own thoughts: All of these factors have backtesting going for them. Various researchers ran a test on a bunch of different strategies and came up with certain ones that would have worked well. For a few of them (momentum, low vol, value), the results of research are so consistent across time, countries, and researchers that one can reliably say they did work well. For others (size, quality), the research results are much more inconsistent.

Problems arise when people then go searching for after-the-fact "explanations" as to why these results occur, as they then conclude makes them in to a near guarantee that they will work in the future. I don't find this approach persuasive. If these researchers had started out by saying, "We think stocks with characteristics X, Y, and Z should outperform because of these reasons", and then tested to see if that was the case--well then I would find the "explanations" to be much more persuasive. Making up reasons after the fact moves more into the realm of marketing and away from understanding how the world actually works.

So, for me we are really just saying that because factors A, B, and C worked over the long-term historically, they will likely work again in the future. Nothing more, nothing less.

A lot of people here are very wedded to making big investments in small-cap value stocks. The problem is that this is <5% of the market, and there is a constant pursuit to find funds that are concentrated in ever smaller pieces of this already small piece. There is a lot of risk in making this kind of super-focused bet which is generally under-appreciated.

I'm okay with people having their favorite factors (mine is low vol). However, I think one should approach factor investing as being a 60/40 proposition more than a 90/10 one in terms of providing added performance. That is, there is a reasonable chance that what worked in the past will continue to in the future, but there is certainly no guarantee. With this understanding, I believe that your approach of selecting on a variety of factors makes a lot of sense.

In terms of correlations, as noted above value and momentum are often negatively correlated with each other. So, using a value fund and momentum fund is always a good idea. Quality and value tend to be more positively correlated. Low vol is somewhat correlated to value, but not as much as quality. Again, these are averages across time; things always vary from year to year and market cycle to market cycle. For example, low vol is frequently more correlated to growth than value. That is why I suggest just equal weighting all of your options is a great starting, and possibly finishing, point.
Topic Author
duffer
Posts: 400
Joined: Tue Sep 09, 2014 8:18 pm

Re: Four-Factor-Only Portfolio?

Post by duffer »

Thank you very much stlutz and backpckaer, for your help. You have certainly given me a lot to read and think about!!!

I am likely to move toward an an approximately equal weight allocation for the time being. How to get there is a little complicated and may take me a long while I have been overweight small caps for several years, am up nicely in those funds, but exactly how to migrate those allocations while they are in a correction is a trick. I would like to wait a while for them to fall closer in line with the S&P and then dollar cost-average I suppose? Maybe I will post this as a problem in the personal advice forum.

Thank you again for your excellent help.

duffer
AviN
Posts: 479
Joined: Mon Mar 10, 2014 8:14 am

Re: Four-Factor-Only Portfolio?

Post by AviN »

Beta is one of the four factors so I don't think the OP wants to get rid of it.
randomguy wrote:It is pretty much impossible to get rid of beta exposure without some speciality products. Just finding funds that don't counter act each other ( getting momentum from one fund and avoiding negative momentum in another) is pretty darn hard also.
JohnnyFive
Posts: 181
Joined: Mon Sep 08, 2014 1:14 pm

Re: Four-Factor-Only Portfolio?

Post by JohnnyFive »

If this is your goal, likely PXSV and MTUM would be all you would need. PXSV takes care of the small and value, as well as incorporating some quality and avoiding some negative momentum and MTUM would give you some momentum tilt.
randomguy
Posts: 11285
Joined: Wed Sep 17, 2014 9:00 am

Re: Four-Factor-Only Portfolio?

Post by randomguy »

AviN wrote:Beta is one of the four factors so I don't think the OP wants to get rid of it.
randomguy wrote:It is pretty much impossible to get rid of beta exposure without some speciality products. Just finding funds that don't counter act each other ( getting momentum from one fund and avoiding negative momentum in another) is pretty darn hard also.
It is a factor. It wasn't one of the 4 that the OP was interested in investing in. I don't know if that was intentional or not. The general point is that when you buy a momentum fund, the odds of having zero loading (plus or minus) on size, value, and beta is pretty much impossible. When you mix a couple of funds together it is really easy to cancel out your loadings.
daffyd
Posts: 199
Joined: Sun Sep 29, 2013 11:51 pm
Location: Australia

Re: Four-Factor-Only Portfolio?

Post by daffyd »

mickens16 wrote:Another excellent post by Robert T.

http://www.bogleheads.org/forum/viewtop ... 8#p2206358
In case you haven't read it, I'd reiterate reviewing the above thread. A combination of PXSV and MTUM should be sufficient for domestic equity.

Depending on your tax brackets if you have existing ETFs/funds in taxable you could use new money and whatever options you have in your 401k/IRA to gradually move towards the allocation you want. It's probably not worth incurring large capital gains taxes, for instance, to move from VBR to PXSV or from VTI to MTUM.

Incidentally, if you believe that valuation is of any use, your weight between domestic and foreign stocks will likely have a much larger impact on your returns than which particular domestic funds you use.
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Four-Factor-Only Portfolio?

Post by Robert T »

.
Few thoughts ...

I think its best to rely on analysis of the underlying factor exposures of a funds/series, rather than relying on a name. For example:

25% MSCI Momentum (tracked by MTUM)
25% MSCI Quality (tracked by QUAL)
25% MSCI Risk Weighted (tracked by SIZE)
25% MSCI Value Weighted (tracked by VLUE)

This does not imply that the combination will capure 25% of any momentum, quality, size and value premium. The underlying factor exposure of the 25x4 combination - based on analysis of the backtested series (using mkt, size, value, and momentum factors from Ken French's website, and the quality factor (QmJ) from Asness et al.) has been:
  • 1976 - 2013

    Alpha = -0.04
    Mkt = +1.03
    Size = -0.11
    Value = +0.12
    Momentum = +0.03
    Quality = +0.18
    R squared = 0.98
So a close to market beta, with a modest tilt to value, momentum, and quality; and a larger than market cap tilt (despite adding a smaller cap fund). Despite the name of the funds, the underlying series, in combination, only provide a modest factor tilt. This is okay, if this is the extent of exposure sought after.

Compare this to the factor exposure from 65% RAFI Pure Small Value:35% MSCI Momentum (PXSV tracks RAFI Pure Small Value, and MTUM tracks MSCI Momentum). Here are the factor loads of the backtested combination over the same time period
  • 1976-2013

    Alpha = +0.03
    Mkt = +1.07
    Size = +0.57
    Value = +0.47
    Momentum = -0.03
    Quality = +0.19
    R squared = 0.94
Larger factor exposure across beta, size, value, and quality. Despite adding a positive allocation to momentum, the combination had a slight negative exposure to momentum (value funds tend to have exposure to negative momentum so adding an explicit allocation to momentum can partly offset this effect - as in above example).

IMO a factor analysis (as above) is the best way to understand the factor expoure of a particular portfolio, at least historically.We each have to decide what factor exposure we want, and then construct a portfolio in the best way to get it (for some it is simply beta). My preference is, in addition to beta, to target positive size and value exposure, with a zero momentum (and quality) load (i.e. zero alpha after accounting for exposure to beta, size, and value). A value tilt can often add both a negative momentum and negative quality tilt (as with the Russell 1000 Value). The RAFI series methodology tends to reduce the negative quality loads (adds a positive load), and adding the momentum fund can reduce the negative momentum load. Of all factor exposures, I am least confident about 'quality' given the wide variation in definitions and associated performance over time (unlike say value). For "minimum volatility" simply add bonds (I don't see 'minimum volatility' as a specific factor exposure beyond adding bonds). I have not found much (any) value in adding MSCI Minimum Volatility to a portfolio (or holding it one its own), compared to adding bonds to reduce volatility. For example:
  • 1989-2013 - annualized return/standard deviation

    10.7%/14.9 = MSCI Minimum volatility
    10.7%/14.7 = 80% to the 4x25% portfolio above: 20% T-bills
    12.7%/14.4 = 45% RAFI Pure Small Value:25% MSCI Momentum:30% T-bills
As with the above analysis, my preference is to focus on the underying factor exposure of a particular fund/series, rather than on its name. IMO this provides a more direct way of targeting a particular factor exposure that you want.

Obviously no guarantees.

Robert
.
Last edited by Robert T on Mon Oct 06, 2014 4:45 pm, edited 2 times in total.
jaab
Posts: 112
Joined: Thu Aug 07, 2014 12:52 pm

Re: Four-Factor-Only Portfolio?

Post by jaab »

Robert T wrote:
  • 1989-2013 - annualized return/standard deviation

    10.2%/14.9 = MSCI Minimum volatility
    10.3%/14.7 = 80% to the 4x25% portfolio above: 20% T-bills
    12.1%/14.4 = 45% RAFI Pure Small Value:25% MSCI Momentum:30% T-bills
Robert, can you compare some measure of downside tail risk for the three options? Worst year, worst three years, maximum drawdown or something along these lines? Would be interesting to see how the minimum volatility portfolio fits in. Only if not much work is needed, of course. :happy Another point I wonder about is what happens with longer duration than T-Bills, because of the supposed term/duration exposure of minimum volatility.
YDNAL
Posts: 13774
Joined: Tue Apr 10, 2007 4:04 pm
Location: Biscayne Bay

Rhttp://www.bogleheads.org/forum/ve: Four-Factor-Only Portfo

Post by YDNAL »

duffer wrote:I am thinking about converting the equity portion of my portfolio into a Four-Factor-Only portfolio, i.e., in equities, I would hold only funds or ETFs representing the factors of size, value, momentum, and profitability (or quality).
The four factors - and any future after-the-fact* "discovery" - are held in something like Vanguard VTSMX. :)

* borrowed from stlutz.
Landy | Be yourself, everyone else is already taken -- Oscar Wilde
randomguy
Posts: 11285
Joined: Wed Sep 17, 2014 9:00 am

Re: Rhttp://www.bogleheads.org/forum/ve: Four-Factor-Only Po

Post by randomguy »

YDNAL wrote:
duffer wrote:I am thinking about converting the equity portion of my portfolio into a Four-Factor-Only portfolio, i.e., in equities, I would hold only funds or ETFs representing the factors of size, value, momentum, and profitability (or quality).
The four factors - and any future after-the-fact* "discovery" - are held in something like Vanguard VTSMX. :)

* borrowed from stlutz.
Problem of course is that VTSMX only has loading on Beta. All the other factors cancel themselves out. Good news if factor investing is an artifact of backtesting. Not so good if it is useful tool to predicate future performance.
AviN
Posts: 479
Joined: Mon Mar 10, 2014 8:14 am

Re: Four-Factor-Only Portfolio?

Post by AviN »

randomguy wrote:
AviN wrote:Beta is one of the four factors so I don't think the OP wants to get rid of it.
It is a factor. It wasn't one of the 4 that the OP was interested in investing in. I don't know if that was intentional or not. The general point is that when you buy a momentum fund, the odds of having zero loading (plus or minus) on size, value, and beta is pretty much impossible. When you mix a couple of funds together it is really easy to cancel out your loadings.
Sorry, my mistake.
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Four-Factor-Only Portfolio?

Post by Robert T »

jaab wrote:
Robert T wrote:
  • 1989-2013 - annualized return/standard deviation

    10.2%/14.9 = MSCI Minimum volatility
    10.3%/14.7 = 80% to the 4x25% portfolio above: 20% T-bills
    12.1%/14.4 = 45% RAFI Pure Small Value:25% MSCI Momentum:30% T-bills
Robert, can you compare some measure of downside tail risk for the three options? Worst year, worst three years, maximum drawdown or something along these lines? Would be interesting to see how the minimum volatility portfolio fits in. Only if not much work is needed, of course. :happy Another point I wonder about is what happens with longer duration than T-Bills, because of the supposed term/duration exposure of minimum volatility.
Jaab,

Made correction to earlier annualized return, added 5 yr T-notes, and 2001-2002, 2008 downside.
  • 1989-2013 - annualized return/standard deviation

    10.7%/14.9 = MSCI Minimum volatility
    10.7%/14.7 = 80% to the 4x25% portfolio above: 20% T-bills
    11.1%/14.5 = 80% to the 4x25% portfolio above: 20% 5 yr T-notes
    12.7%/14.4 = 45% RAFI Pure Small Value:25% MSCI Momentum:30% T-bills
    13.7%/14.1 = 45% RAFI Pure Small Value:25% MSCI Momentum:30% 5 yrs T-notes

    2001-2002/2008 returns

    -21.9%/-25.7% = MSCI Minimum volatility
    -17.7%/-28.6% = 80% to the 4x25% portfolio above: 20% T-bills
    -14.8%/-26.3% = 80% to the 4x25% portfolio above: 20% 5 yr T-notes
    +10.1%/-26.6% = 45% RAFI Pure Small Value:25% MSCI Momentum:30% T-bills
    +15.1%/-23.1% = 45% RAFI Pure Small Value:25% MSCI Momentum:30% 5 yrs T-notes
Robert
.
YDNAL
Posts: 13774
Joined: Tue Apr 10, 2007 4:04 pm
Location: Biscayne Bay

Re: Rhttp://www.bogleheads.org/forum/ve: Four-Factor-Only Po

Post by YDNAL »

randomguy wrote:
YDNAL wrote:
duffer wrote:I am thinking about converting the equity portion of my portfolio into a Four-Factor-Only portfolio, i.e., in equities, I would hold only funds or ETFs representing the factors of size, value, momentum, and profitability (or quality).
The four factors - and any future after-the-fact* "discovery" - are held in something like Vanguard VTSMX. :)

* borrowed from stlutz.
Problem of course is that VTSMX only has loading on Beta.
"Problem" is in the eye of the beholder!
randomguy wrote:All the other factors cancel themselves out. Good news if factor investing is an artifact of backtesting. Not so good if it is useful tool to predicate future performance.
Problem^2, when Fama/French published their "research" (The Cross-Section of Expected Stock Returns), for instance, the momentum or profitability factors were not "expected (even recognized)" to do anything. Today, the drum beat sounds different, so do they "predicate" anything, or is it just more after-the-fact "research"?
Landy | Be yourself, everyone else is already taken -- Oscar Wilde
stlutz
Posts: 5585
Joined: Fri Jan 02, 2009 12:08 am

Re: Four-Factor-Only Portfolio?

Post by stlutz »

I think its best to rely on analysis of the underlying factor exposures of a funds/series, rather than relying on a name. For example:

25% MSCI Momentum (tracked by MTUM)
25% MSCI Quality (tracked by QUAL)
25% MSCI Risk Weighted (tracked by SIZE)
25% MSCI Value Weighted (tracked by VLUE)

This does not imply that the combination will capure 25% of any momentum, quality, size and value premium. The underlying factor exposure of the 25x4 combination - based on analysis of the backtested series (using mkt, size, value, and momentum factors from Ken French's website, and the quality factor (QmJ) from Asness et al.) has been:

1976 - 2013

Alpha = -0.07
Mkt = +0.93
Size = -0.13
Value = +0.08
Momentum = +0.03
Quality = +0.15
R squared = 0.98


So a slightly below market beta, with a modest tilt to value, momentum, and quality; and a larger than market cap tilt (despite adding a smaller cap fund). Despite the name of the funds, the underlying series, in combination, only provide a modest factor tilt. This is okay, if this is the extent of exposure sought after.
Hi Robert,

Thanks for running these numbers. I think they are about what you would logically expect if you put together a portfolio that pursues a bunch of different factors. And the large-cap tilt would also be expected as the base population of securities for each of these indices is a large-mid index (i.e. no small caps).

One might then say that the OP's proposal is pretty much the same as holding the entire market.

However, that's actually not true and it brings up a point of confusion on my part. All of these MSCI indices outperformed the market for period that they were back-constructed for. For example, the quality index beats its base index by about 1% per year since inception, momentum by 3.4%, the value index by .4%, and risk-weighted by about 1.7%. Where does that get reflected in this type of analysis? Or, as I have understood in the past, factor analysis is mostly about analyzing the fluctuations over time as opposed to where you actually end up?

Thanks,
Topic Author
duffer
Posts: 400
Joined: Tue Sep 09, 2014 8:18 pm

Re: Four-Factor-Only Portfolio?

Post by duffer »

daffyd wrote:
mickens16 wrote:Another excellent post by Robert T.

http://www.bogleheads.org/forum/viewtop ... 8#p2206358
In case you haven't read it, I'd reiterate reviewing the above thread. A combination of PXSV and MTUM should be sufficient for domestic equity.

Depending on your tax brackets if you have existing ETFs/funds in taxable you could use new money and whatever options you have in your 401k/IRA to gradually move towards the allocation you want. It's probably not worth incurring large capital gains taxes, for instance, to move from VBR to PXSV or from VTI to MTUM.

Incidentally, if you believe that valuation is of any use, your weight between domestic and foreign stocks will likely have a much larger impact on your returns than which particular domestic funds you use.
Thanks for this. i do think valuation matters, but I am reluctant to jump into international just now, though I guess that is the point of taking advantage of valuations.
Topic Author
duffer
Posts: 400
Joined: Tue Sep 09, 2014 8:18 pm

Re: Four-Factor-Only Portfolio?

Post by duffer »

Robert T wrote:.
Few thoughts ...

I think its best to rely on analysis of the underlying factor exposures of a funds/series, rather than relying on a name. For example:

25% MSCI Momentum (tracked by MTUM)
25% MSCI Quality (tracked by QUAL)
25% MSCI Risk Weighted (tracked by SIZE)
25% MSCI Value Weighted (tracked by VLUE)

This does not imply that the combination will capure 25% of any momentum, quality, size and value premium. The underlying factor exposure of the 25x4 combination - based on analysis of the backtested series (using mkt, size, value, and momentum factors from Ken French's website, and the quality factor (QmJ) from Asness et al.) has been:
  • 1976 - 2013

    Alpha = -0.04
    Mkt = +1.03
    Size = -0.11
    Value = +0.12
    Momentum = +0.03
    Quality = +0.18
    R squared = 0.98
So a close to market beta, with a modest tilt to value, momentum, and quality; and a larger than market cap tilt (despite adding a smaller cap fund). Despite the name of the funds, the underlying series, in combination, only provide a modest factor tilt. This is okay, if this is the extent of exposure sought after.

Compare this to the factor exposure from 65% RAFI Pure Small Value:35% MSCI Momentum (PXSV tracks RAFI Pure Small Value, and MTUM tracks MSCI Momentum). Here are the factor loads of the backtested combination over the same time period
  • 1976-2013

    Alpha = +0.03
    Mkt = +1.07
    Size = +0.57
    Value = +0.47
    Momentum = -0.03
    Quality = +0.19
    R squared = 0.94
Larger factor exposure across beta, size, value, and quality. Despite adding a positive allocation to momentum, the combination had a slight negative exposure to momentum (value funds tend to have exposure to negative momentum so adding an explicit allocation to momentum can partly offset this effect - as in above example).

IMO a factor analysis (as above) is the best way to understand the factor expoure of a particular portfolio, at least historically.We each have to decide what factor exposure we want, and then construct a portfolio in the best way to get it (for some it is simply beta). My preference is, in addition to beta, to target positive size and value exposure, with a zero momentum (and quality) load (i.e. zero alpha after accounting for exposure to beta, size, and value). A value tilt can often add both a negative momentum and negative quality tilt (as with the Russell 1000 Value). The RAFI series methodology tends to reduce the negative quality loads (adds a positive load), and adding the momentum fund can reduce the negative momentum load. Of all factor exposures, I am least confident about 'quality' given the wide variation in definitions and associated performance over time (unlike say value). For "minimum volatility" simply add bonds (I don't see 'minimum volatility' as a specific factor exposure beyond adding bonds). I have not found much (any) value in adding MSCI Minimum Volatility to a portfolio (or holding it one its own), compared to adding bonds to reduce volatility. For example:
  • 1989-2013 - annualized return/standard deviation

    10.7%/14.9 = MSCI Minimum volatility
    10.7%/14.7 = 80% to the 4x25% portfolio above: 20% T-bills
    12.7%/14.4 = 45% RAFI Pure Small Value:25% MSCI Momentum:30% T-bills
As with the above analysis, my preference is to focus on the underying factor exposure of a particular fund/series, rather than on its name. IMO this provides a more direct way of targeting a particular factor exposure that you want.

Obviously no guarantees.

Robert
.
Thank you for this very interesting and detailed post. It will take me a while to mull the implications, which are especially complicated since I have small and value holdings in a taxable account that I don't want to take a tax hit on. So figuring out whta the factor loads of my actual holdings would be gets very complicated and is probably beyond my capacity.

How did you derive factor loadings? Is it doable by someone with no background in financial math?

Thanks again.

duffer
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Four-Factor-Only Portfolio?

Post by Robert T »

stlutz wrote:Thanks for running these numbers. I think they are about what you would logically expect if you put together a portfolio that pursues a bunch of different factors. And the large-cap tilt would also be expected as the base population of securities for each of these indices is a large-mid index (i.e. no small caps).

One might then say that the OP's proposal is pretty much the same as holding the entire market.

However, that's actually not true and it brings up a point of confusion on my part. All of these MSCI indices outperformed the market for period that they were back-constructed for. For example, the quality index beats its base index by about 1% per year since inception, momentum by 3.4%, the value index by .4%, and risk-weighted by about 1.7%. Where does that get reflected in this type of analysis? Or, as I have understood in the past, factor analysis is mostly about analyzing the fluctuations over time as opposed to where you actually end up?
Stlutz,

Its reflected in the factor loads.

For example over the 1976-2013 the average annual factor premiums were the following (derived from monthly data):

7.6% = Market (Mkt-Rf)
3.4% = Size
4.1% = Value
8.2% = Momentum
4.7% = Quality

Just to note the annual average T-bill returns over this period = 4.9%. Using the factor loads from above analysis multipled by the above listed average annual premiums (+ the estimated annual alpha) gives the average return for the fund/series being used in the analysis. i.e. in the case of the 25x4: monthly alpha = -0.04, annual alpha = -0.44. so -0.44 + 7.6*1.03 - 0.12*3.4 + 0.12*4.1 + 0.03*8.2 + 0.18*4.7 + 4.9 (adding back Rf which was substracted from the fund/series for the analysis) = 13.4%. This is 1.4% higher than the average 12% return for MSCI US equity market (standard) over the same time period. The draw back with these estimates is they are average returns, not annualized, but the differences with 'market' returns are fairly similar (if using monthly returns). For example. a comparison of annualized returns of the 4x25 series gives a 1.6% difference with the market (13.0% vs. 11.4%), compared to 1.4% with the average returns.

Hope this helps.

Robert

PS: Just to note the slight revision in the earlier factor load estimates (corrected in earlier post).
.
Last edited by Robert T on Tue Oct 07, 2014 1:19 am, edited 4 times in total.
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Four-Factor-Only Portfolio?

Post by Robert T »

duffer wrote:How did you derive factor loadings? Is it doable by someone with no background in financial math?
This may help http://www.efficientfrontier.com/ef/101/roll101.htm , or for funds with a long history of live returns http://www.portfoliovisualizer.com/factor-analysis

Robert
.
SnowSkier
Posts: 268
Joined: Fri Mar 16, 2012 12:21 pm
Location: Austin

Re: Four-Factor-Only Portfolio?

Post by SnowSkier »

Robert T wrote:
Its reflected in the factor loads.

For example over the 1976-2013 the average annual factor premiums were the following (derived from monthly data):

3.4% = Size
4.1% = Value
8.2% = Momentum
4.7% = Quality
...
Robert,
Are these all long minus short? (i.e. SMB, HML, UMD, QMJ)
If so, would it be possible to derive the equivalent historical premiums for long-only? (If this is the correct terminology? ... Perhaps I should say long minus market? ... not totally sure if my question makes any sense, so feel free to educate)
Does long-only tend to be about half of the long minus short premiums?
SnowSkier
Posts: 268
Joined: Fri Mar 16, 2012 12:21 pm
Location: Austin

Re: Four-Factor-Only Portfolio?

Post by SnowSkier »

I think this separate post from Larry answers most of my questions about long-only versus long-minus-short.

A Close Look At Factors in the Real World, by Larry Swedroe:
http://www.bogleheads.org/forum/viewtop ... 0&t=148464
Post Reply