DFA - Core vs. Components

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Topic Author
Amishman
Posts: 310
Joined: Mon Jun 11, 2007 8:08 pm

DFA - Core vs. Components

Post by Amishman »

Why would an advisor recommend component strategies over core strategies other than for marketing purposes (i.e. selling complexity)?
SmallHi
Posts: 1718
Joined: Wed Feb 21, 2007 5:11 pm

Post by SmallHi »

Here are some reasons:

1) they want an "all value portfolio" globally, or at least in Int'l or Emerging Markets. In that case, the value funds would be used exclusively. In the case of EM, you could make a case that EM Value is the best strategy anyway (with no EM Vector fund yet)

2) they want a tilt higher than can be achieved by even the most aggressive Core equity fund (Vector)

3) an investor has low cost total market funds available in their company sponsored retirement plan, and need to augment those with small/value components in "outside" accounts

4) they believe a "Core/sattelite" TSM/SV structure has higher expected returns

5) more specifically, they don't believe the return dimensions are linear (they believe the small cap premium is only present in small/micro cap stocks, and the value premium is only present in the highest 10% or 20% of BtM stocks)...at which case you would want to tilt with the remote sattelite asset class funds.

sh
Derek Tinnin
Posts: 864
Joined: Tue Jan 08, 2008 6:16 pm
Location: cincinnati, oh

Post by Derek Tinnin »

IMO, a Core 2 and/or Vector approach is a logical solution for most tilted portfolios, but as SH points out, components have a place here and there. The key is to not recommend a one-size-fits all strategy.

Other examples I can think of to use components include:

You may have a small Roth IRA where you want maximum expected return, so you use a SV component.

You may want to set aside a portion of your portfolio for charitable gifts to institutions that have a hard time accepting mutual funds (logistically more difficult to accept), so you use component ETF's (treated like a gift of "stock") for that purpose.

You may believe that greater tax efficiency can result from using TM Targeted Value (component) plus TM Core 2 instead of TM Core 2 plus Vector. (splitting hairs on that one...)

Overall though, component slice and dice isn't attractive as it used to be...
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T »

.
FWIW, yesterday I set-up the following portfolios in the M* tracker. They cover (i) some options available for a 60/40 one fund portfolio, or there abouts (each has about a 60:40 benchmark), (ii) a simple core/vector portfolio, and (iii) a component portfolio.
By my estimate, the latter two portfolios have similar expected returns (using assumption that the DFA US and Intl vector funds both have a size and value load of 0.4 and 0.4 respectively, and the EM core size and value loads are 0.2 and 0.2 respectively).

Will leave them in the tracker, and check back in periodically to see how they are doing.

Robert
.
SmallHi
Posts: 1718
Joined: Wed Feb 21, 2007 5:11 pm

Post by SmallHi »

Robert,

The DFA Emerging Market Core fund may not be the best compliment to a US/Int'l Vector foundation. EM Value is actually more in line with Vector than EM Core is. EM Core is very market oriented, with just a very modest size/value tilt. In terms of multifactor exposure, its a step backwards from Vector.

EM Core is actually kind of a hybrid, and in terms of US exposure, is closer to a 50% US Core 1, 50% US Core 2 (or 50% Large, 25% Value, 25% Small). On the other hand, there isn't a lot of difference between what a DFA EM Vector would look like, and what DFA EM Value does look like (Vector would be a bit smaller cap, a bit less value oriented, but because of the Core technology, it'd likely be a bit more tax efficient and hold a few hundred extra names).

If you are going to dial up a factor focused DFA allocation, US Vector, Int'l Vector, and EM Value are probably the best pieces.

As for your ETF portfolio...why not just stick with US MCV and Micro? Is the 600VL necessary? Further, are the loadings on #2 and #3 the same? #3 seems a bit light on SC?

just ramblings...

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

Post by Robert T »

Sh,

Don't have much time now, but just a few (partial) answers.
  • - the two portfolios have similar expected return, by my estimate (the second gets more of it from a value relative to size tilt).
    - so loadings aren't the same, for the first I get a size and value of around 0.37 each, for the second at about 0.2 and 0.4 respectively. Add in slight differences in fixed income loads, and the expected returns come out quite close.
    - yes DFA EM value could be used, but in my simple example was just trying to match my (needed) return of about 7.5% (after costs), with two examples. Adding EM value would increase expected return (and risk) above this level.
    - 600VL isn't necessary, but gives some exposure to the 6-9 decile gap not provided by Russell MCV and the Bridgeway fund (if worried about decile tracking error), and helps with the 0.2 and 0.4 overall factor match.
Robert
.
User avatar
altruistguy
Posts: 111
Joined: Tue Jun 12, 2007 8:55 am

Hard to fit in Core funds when there are lots of "Bins&

Post by altruistguy »

Hello Folks,

The question was about why one might NOT use DFA core-type funds.

The core funds are quite attractive for folks who have all their money in one "bin". For example, if all your money is in a taxable account (and you have no existing holdings with large unrealized capital gains), or all your money is in an IRA.

HOWEVER, almost none of my clients portfolios are that way. Virtually all of my clients have 401(k) or 403(b) accounts, as well as 529 accounts, possibly 457 accounts, Traditional IRAs, Roth IRAs, taxable accounts, trust accounts, etc. If managing a large portfolio built from all those things, it gets complicated. For example, most 401(k) plans have REALLY BAD investment choices. But often there is one index fund (typically an S&P 500 index fund) that is clearly "least bad". In that sort of circumstance, we'd try to load up on that least bad investing option in the 401(k) plan.

In order to take advantage of the foreign tax credit, we'd try to get as much as possible of foreign stock exposure in a taxable account. We'd try to get as much as possible of bond, REIT, commodities future, and other tax-inefficient asset classes in tax-preferred accounts. We'd try to get as much as possible of the portfolio's small and value exposure domestically (its a more cost-effective way to get exposure to those risk factors). All of the above is part of optimizing the design/deployment of the overall portfolio. Further, you may already have stuff in a taxable account with large unrealized gains that you just can't justify selling. And after you've done all that optimizing and taking into account which asset classes are where, it is difficult to fit in a core fund without really screwing things up, from the overall portfolio perspective.

When does a core fund make sense? Again, when/if you just have one "bin" in your portfolio. Or, perhaps for simplicity's sake, you bypass the optimization of your overall portfolio in the manner described above and you make all bins the same. But bypassing the above portfolio optimizations isn't something that a good investment advisor would be doing.

We like the core funds. We like them a lot. But we haven't found many opportunities to use them. To date, we have recommended them to (only) two clients, but neither has accepted our recommendation.

I hope that the above elucidates on how a portfolio is prudently designed and the limitations on using the core funds in a typical well-optimized overall portfolio.

Eric E. Haas
Derek Tinnin
Posts: 864
Joined: Tue Jan 08, 2008 6:16 pm
Location: cincinnati, oh

Re: Hard to fit in Core funds when there are lots of "B

Post by Derek Tinnin »

altruistguy wrote:HOWEVER, almost none of my clients portfolios are that way. Virtually all of my clients have 401(k) or 403(b) accounts, as well as 529 accounts, possibly 457 accounts, Traditional IRAs, Roth IRAs, taxable accounts, trust accounts, etc. If managing a large portfolio built from all those things, it gets complicated. For example, most 401(k) plans have REALLY BAD investment choices. But often there is one index fund (typically an S&P 500 index fund) that is clearly "least bad". In that sort of circumstance, we'd try to load up on that least bad investing option in the 401(k) plan.
I know where you are coming from Eric, but I have to disagree unless most of your clients have most of their money in 401k's or other outside accounts.

Assuming a moderate to moderately aggressive tilter that stays around .2 to .3 loads, you can use an S&P 500 fund in the 401k, but still stay within the Core family to achieve your desired tilt. Just use more Vector.

For portfolios where most of the assets are in traditional IRA's, Roth IRA's, SEP IRA's, joint accts, trust accounts, individual accounts, etc., all with access to core funds (and more representative of my typical client - 401k's have smaller weights), I find no problems implementing core funds.

The 529 plans will typically have a separate objective and different allocation, so integrating those is more rare.
altruistguy wrote:In order to take advantage of the foreign tax credit, we'd try to get as much as possible of foreign stock exposure in a taxable account.
Agree. but why can't this be done with International Core or International Vector? No need to use components just to capture the FTC.
altruistguy wrote:We'd try to get as much as possible of bond, REIT, commodities future, and other tax-inefficient asset classes in tax-preferred accounts.
Agree, but this has nothing to do with core vs. component. You are going to use components for those classes anyway. And soaking up IRA assets with those components means holding more Core in the taxable accounts, where their inherent tax efficiency really shines.
altruistguy wrote:We'd try to get as much as possible of the portfolio's small and value exposure domestically (its a more cost-effective way to get exposure to those risk factors).
Not sure about that comment. Yes, domestic funds are cheaper to own, but I wouldn't avoid a huge portion of the investable universe of small and value stocks globally to achieve that goal. After all, you need to own as much of that universe as you can to make sure you actually capture the size/value premiums (driven by the small number of companies that are extreme outperformers relative to the pack). The diversification impact of non-US multi-factor exposure goes beyond the relative expense differences IMO.

altruistguy wrote:Further, you may already have stuff in a taxable account with large unrealized gains that you just can't justify selling. And after you've done all that optimizing and taking into account which asset classes are where, it is difficult to fit in a core fund without really screwing things up, from the overall portfolio perspective.
I agree. Cap gains can get in the way. In the current market environment, however, cap gains are rare. What a great time to move to Core! :D

I have not really found any difficulty in achieving the various goals of asset location (tax efficiency, FTC credit, integration with outside accts, etc.) by using Core funds. In fact, I think it actually improves the process. I still need to add components from time to time, but I can usually find an optimal combo of Core and component, with the primary benefit being less required component. Your practice may have a different client profile, but with a client base having modest outside account holdings, I find the Core approach makes a heck of a lot of sense.
SmallHi
Posts: 1718
Joined: Wed Feb 21, 2007 5:11 pm

Post by SmallHi »

Robert T wrote:Sh,

Don't have much time now, but just a few (partial) answers.
  • - the two portfolios have similar expected return, by my estimate (the second gets more of it from a value relative to size tilt).
    - so loadings aren't the same, for the first I get a size and value of around 0.37 each, for the second at about 0.2 and 0.4 respectively. Add in slight differences in fixed income loads, and the expected returns come out quite close.
    - yes DFA EM value could be used, but in my simple example was just trying to match my (needed) return of about 7.5% (after costs), with two examples. Adding EM value would increase expected return (and risk) above this level.
    - 600VL isn't necessary, but gives some exposure to the 6-9 decile gap not provided by Russell MCV and the Bridgeway fund (if worried about decile tracking error), and helps with the 0.2 and 0.4 overall factor match.
Robert
.
By my rough calculations, the ETF and DFA portfolios aren't really that close. Your ETF allocation has an exposure to SmB of about 0.20, and HmL of about 0.45. The DFA allocation is 0.37 SmB and 0.37 HmL.

If you instead drew up the following DFA allocation (equity only):

25% DFA US Vector
25% DFA US Marketwide Value
13.5% DFA Int'l Vector
13.5% DFA Int'l Value
13% DFA EM Value

..you'd be in the neighboorhood of the 0.20/0.45. Although, from an investability standpoint, I doubt there is much benefit to the 5 fund allocation over the previous 3 fund allocation I suggested (US Vector, Int'l Vector, and EM Value).

Also, because 5YR Global has only about 1/2 the exposure to TERM as 3-7YR Treasury (DFGBX is only about +0.25), and only modest (+0.01) exposure to DEFAULT, I think a 37.5% US Vector, 28% Int'l Vector, 9.5% EM Value, and 25% 5YR Global portfolio is about similar in total factor risk to the ETF portfolio you built (a bit more size risk, a bit less fixed income risk -- consistent with keeping fixed income safe with low exposure to DEF/TERM and instead spending savings on equities).

I know its just for tracking purposes, so I don't think it matters much...

sh
User avatar
Leif
Posts: 3698
Joined: Wed Sep 19, 2007 4:15 pm

Post by Leif »

I'm not an advisor, but I like the DFA TM Core funds in taxable, with tilts in my tax deferred accounts. I did have EM Value in my taxable, and it has been very tax efficient until this year. With the gains I had in that fund I never expected to be able to move it, or have a need to. However, with some big CG distributions that appear to be coming in EM Value, I see it may not be that great of a fund for a taxable account.

The recent market has given me the "opportunity" to sell my EM Value, and move it to my tax deferred. Also, the new TM Core funds, which must have some tremendous losses on the books now, as well as being naturally tax efficient, appear to be an excellent set of funds US & Intl. to be in going forward, for a tax sensitive investor.

So, I'm thinking TM US & Intl. Core in my taxable, and all others EM Value, LV, SV, and REITS in my tax deferred.
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T »

.
Sh,

This is what I get. As mentioned above, I wasn’t trying to match factor loads, just expected return (my estimate in the table below has expected return to be the same across both portfolios). FWIW the size/value/default/term factor loads within the stock and bond allocations for the two portfolios are: 0.37/0.37/0/0.20 for the DFA portfolio (assumed), and 0.21/0.40/0.0/0.5 for the ETF portfolio (mostly estimated). There will likely be annual 'tracking' error between the two portfolios given the differences in their size and value exposure.

Code: Select all

                         Stock & Bond Fund Factor Loads * Expected Factor Premiums[1]
                         -----------------------------------------------------------   
PORTFOLIO 1 [2] [DFA core/vectore portfolio]
-----------
DFA US Vector                0.375 * (1.00 * 6.5) + (0.40 * 2.0) + (0.40 * 4.0) 
DFA Intl. Vector           + 0.280 * (1.00 * 7.0) + (0.40 * 2.0) + (0.40 * 4.0) 
DFA EM Core                + 0.095 * (1.00 * 9.0) + (0.20 * 2.0) + (0.20 * 4.0) 
DFA 5 yr Global            + 0.250 * (1.00 * 3.7) + (0.00 * 0.5) + (0.20 * 1.6)

EXPECTED RETURN            = 7.9% 


PORTFOLIO 2 [3] [ETF Portfolio]
----------- 
Vanguard Midcap Value        0.234 * (1.00 * 6.5) + (0.10 * 2.0) + (0.65 * 4.0)
iShares S&P600 Value       + 0.094 * (1.00 * 6.5) + (0.65 * 2.0) + (0.60 * 4.0)
Bridgeway Ultra Small Co.  + 0.047 * (1.00 * 6.5) + (1.10 * 2.0) + (0.10 * 4.0)
iShares EAFE Value         + 0.234 * (1.00 * 7.0) + (-0.1 * 2.0) + (0.35 * 4.0)
iShares EAFE Small Cap     + 0.047 * (1.00 * 7.0) + (0.90 * 2.0) + (0.10 * 4.0)
Vanguard EM                + 0.094 * (1.00 * 9.0) + (0.00 * 2.0) + (0.00 * 4.0)      
iShares Lehman 3-7 Trsy    + 0.250 * (1.00 * 3.7) + (0.00 * 0.5) + (0.50 * 1.6) 

EXPECTED RETURN [4]        = 7.9%


PORTFOLIO 3 [DFA Tax-Managed Component Portfolio] 
----------- 
DFA TM US MktWide Value      0.170 * (1.00 * 6.5) + (0.00 * 2.0) + (0.70 * 4.0) 
DFA TM US Targeted Value   + 0.070 * (1.00 * 6.5) + (0.50 * 2.0) + (0.80 * 4.0) 
DFA TM US Small            + 0.135 * (1.00 * 6.5) + (0.85 * 2.0) + (0.30 * 4.0) 
DFA TM Intl. Value         + 0.170 * (1.00 * 7.0) + (0.10 * 2.0) + (0.50 * 4.0) 
DFA Lntl. Large cap        + 0.110 * (1.00 * 7.0) + (-0.1 * 2.0) + (0.00 * 4.0) 
DFA EM                     + 0.095 * (1.00 * 9.0) + (0.00 * 2.0) + (0.00 * 4.0)      
DFA Int. Government        + 0.250 * (1.00 * 3.7) + (0.00 * 0.5) + (0.60 * 1.6) 

EXPECTED RETURN            = 7.9% 


  • [1] For all stock funds the above lines reflect: Allocation share * (Market load * expected market premium) + (Size load * expected size premium) + (Value load * expected value premium). For all bond funds the above lines reflect: Allocation share * (Expected T-bill return) + (Default load * expected default premium) + (Term load * expected term premium). For the calculations about the expected size, value, default and term premium were taken to be 2.0, 4.0, 0.5, and 1.6% respectively. The ‘market’ loads on all funds were taken to be 1.00, and the expected T-bill returns to be 3.7%.

    [2]For Portfolio 1, all DFA vector/core factor loads used were not estimated from historical data. The above reflect what I understand them to be from previous threads which had the vector value loads ranging from 0.40 to 0.45.

    [3]For Portfolio 2, all factor loads reflected above are 'rounded up' estimates from monthly data from July 1995 to Feb 2009 for the underlying indexes tracked by the funds with two exceptions. The first is the iShares Lehman 3-7 Treasury fund which is assumed to have a zero default and 0.5 term load (the Vanguard intermediate treasury fund has had a term load of about 0.55). The second is the iShares EAFE Small cap fund for which historical data of the underlying index is limited. A 0.9 and 0.1 size and value load is used based on earlier DFA research. The factor loads for the Bridgeway fund are those of its benchmark, the CRSP10 over the same time period.

    [4] If average expected returns are used instead of annualized (which is what should really be used), I get portfolio 1 to have a marginally higher expected average return (0.07% higher) than portfolio 2. All estimates are before costs/taxes etc.
The above estimates assume the characteristics of the underlying indexes from July 1995 to Feb 2009 represent the long-term characteristics of the funds. Time will tell.

Robert

Edited to update the factor load estimates and to add a DFA TM portfolio that matches the US, Intl, and fixed income factor loads of the ETF portfolio - at least as close as I could get them. Expected return numbers remain unchanged.
Last edited by Robert T on Mon May 25, 2009 3:08 pm, edited 2 times in total.
SmallHi
Posts: 1718
Joined: Wed Feb 21, 2007 5:11 pm

Post by SmallHi »

Robert,

Just an FYI, I have found that the EAFE Small Index has had about the same loading to small cap as the DFA Int'l Small Cap fund since 1999. So assuming this was also the case from 95-99, then you are looking at an Int'l SmB of about 0.90. I also find EAFE Value with a higher HmL that you report from 95-08, about 0.47.

Both portfolios look good.

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

Post by Robert T »

.
Sh,

Here are my earlier estimates FWIW:

Value load on EAFE Value: I used the Ind_all file on Ken French’s website to calculate an Intl HmL factor (High BE/ME minus Low BE/ME). Using these data, the annual Intl HmL premium from 1975-2007 was 6.7% while the corresponding US HmL premium was 6.1%. The Intl factors are calculated in the same way as the US factors (i.e. the 30% BtM sorts) as I understand, and if I recall the associated country HmL factors were used by FF in their study of international value premiums. Using these data over the same time period as before (July 1995 – April 2007) I get a value load on EAFE value of 0.32. Using data from 1975 gives the same value load.

Code: Select all

July 1995 – April 2007

                  Alpha     Beta     Size     Value        R^2 
EAFE Value        
Coefficient       -0.08     0.97    -0.07      0.32       0.97 
t-statistic       -1.21    58.24    -2.57     12.56

Size load on EAFE Small: Using data from April 2001 to Sept 2006 (a fairly short series given EAFE small cap data availability), with an SmB series derived from the S&P/citigroup developed market data set (Small [less than 1 billon] minus Big [greater than 5 billion]), I earlier got a size load in the 0.90s (and a value load of less than 0.10). However, I don’t have much faith in the estimate given the short time period and as the Intl. SmB series used was not constructed in the same way as the US SmB series. As an alternative, I also looked at average weighted market caps which, for better or worse, led me to a size load similar to the S&P600. Of all the series listed above this is the factor load estimate on which I have the least confidence and could be wrong (however using a 0.9 size load and 0.1 value load for the EAFE small cap series gives the same portfolio expected return).

Robert

PS: Just dug up this earlier DFA presentation. They seem to have the following factor load estimates for EAFE Value and EAFE small cap (see pg. 10 & 18.). So I may be off on the EAFE small cap size load, but not too far off on their estimate on the value load of EAFE value.

http://www.ifa.com/Media/Images/PDF%20f ... alysis.pdf

Code: Select all

                     Factor loads
                     --------------------------------
                     Mkt      Size     Value      R^2

EAFE Value*          0.99    -0.05     0.38      0.96
EAFE Small**         1.11     0.91    -0.02      0.97

*  Jan 1985 - Feb 2005
** Jan 1993 - Feb 2005

Source: http://www.ifa.com/Media/Images/PDF%20files/DimensionalVs.VanguardAnalysis.pdf

.
User avatar
Sammy_M
Posts: 1935
Joined: Sun Nov 25, 2007 7:30 am

Post by Sammy_M »

Robert,
Thank you for the very thorough and educational explanation of your expected return calculations shown above. I'm finally getting around to calculating expected returns on my portfolio and have a quick question -- are the expected returns you mention above real or nominal? I'm trying to figure out what to use for a TIPS fund. Thank you.
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T »

.
Sammy_M,
a quick question -- are the expected returns you mention above real or nominal? I'm trying to figure out what to use for a TIPS fund.
  • - The expected return calculations are in nominal terms.
    - For a TIPS fund I would use an expected return lower than the nominal bond return (of similar maturity), as the latter should receive an inflation risk premium over the former. Its hard to know the size of this premium – Larry’s book suggest the “the premium might be 0.25%”. In this earlier speech Bernanke used 0.5%.
    - For individual TIPS held to maturity, IMO the expected return is expected inflation + the coupon rate of the individual issue (but realize that not everyone agrees) .
Robert
.
Post Reply