Bogleheads

[Robert T]

.
The Dimensions of Stock Returns: 2007

It compares the traditional, core and vector approaches which may be of interest.

Robert
.

[SmallHi]

I find those "market grids" are very helpful in visualizing a Core or component funds tilt away from Large/Growth stocks...

Thanks for the link.

SH

[Robert T]

.
One interesting statistic from the article is that when an 'apples to apples' comparison is used, the US value and small cap effect has - on average - historically been similar (4.79% vs. 4.68%) (although as demonstrated earlier the value effect has been more persistent than the size effect).

The US Vector portfolio map looks more even across the small-large and value-growth dimensions than a conventional tilt. It holds 3,404 stocks (about 75% of TSM), has an average weighted market cap half the size of TSM, and a significantly lower P/B (1.7 versus 2.6) [from the fact sheet]. Its portfolio turnover over the six months prior to May 2007 was 2% - so looks good so far.

A few questions for the DFA gurus (any updates on the following?):

(i) DFA now has a Tax-Aware US Core 2 fund. What’s the difference between DFA’s Tax-Aware strategies and their Tax-Managed Strategies? I presume there must be a difference given it was not called a Tax-Managed US Core 2.
(ii) Have there been any estimates of the tax-savings from core 2 vs. its TA equivalent?
(iii) Does DFA plan to bring out a TA-US Vector fund and if so, when?
(iv) Does DFA plan to bring out an Intl Vector and EM Vector fund and if so when? And are TA-versions likely?

The case for an EM vector fund may be lower than for Intl. Developed. As the EM vaue fund already has a small cap tilt (its already about half the size [weighted average market cap] of the EM fund), and the difference in the weighted average book-to-market with the EM fund is 0.53 vs. 0.36, which is similar to the US vector comparison to TSM which is 0.58 vs. 0.39. So perhaps the next step is just a TM EM value fund.

Thanks.

Robert
.

[larryswedroe]

no plans for any TA vector funds
Should have TA int' core out next year, when not sure.

I would think, no that the savings are more in the trading costs and that is due to lower turnover of course, and that also results in more tax efficiency. but the differences not likely to be too large. But with lower OER and lower internal trading costs and lower turnover producing more tax efficiency perhaps 25bp or so.

[SmallHi]

<post deleted> for inaccuracy.

SH

[larryswedroe]

SH you are definitely wrong on the dividend issue. Whoever told you that is simply wrong. If the dividend taxation treatment changed they would turn on div management (not div minimization) on for all TM and TA funds.

[Rick Ferri]

Vector funds are fine. There is nothing wrong with the concept in that it does follow a FF strategy, and more important, it allows DFA to continue to gain assets from advisors without creating a worse overcapacity issue in the overstuffed small value and microcap funds.

However, I still would rather approach the markets as three separate risks. That is the way Fama/French designed the three-factor model, and it is how I can directly control the exposure to each risk. As such, I favor a total stock market index fund as the core (VTI; 0.07% fee), then directly control value and size exposure through a DFA fund or some other value index fund and through a micro stock index fund.

Rick Ferri

[Robert T]

.
Larry and SH,

Thanks for the info. I will continue to monitor performance – particularly tax efficiency but also try to understand more the size of the transaction cost savings.

I am not sure how significant transaction/trade costs are in the conventional component approach versus the integrated approach. I understand the arguments but don’t really know how much transaction costs eat into returns. For example one would expect these to show up in the difference between a fund return and the index it tracks (at least my understanding). Here’s an example:

[SmallHi]

[SmallHi]

[Rick Ferri]

Robert T,

I would recommend splitting your Europe and Pacific allocations 50/50 rather than 'swinging with the EAFE'. In my book, All About Asset Allocation, I show that there has been about a 1% advantage to splitting those allocations are rebalancing.

SmallHi,

[SmallHi]

[larryswedroe]

saying that the core funds have everything to do with capacity is utter nonsense.Without question.
Now was it actually driven because capacity was becoming an issue in some cases. YES
But core funds are a CLEAR improvement on the component approach. More tax efficient and lower turnover meaning more tax efficiency. Which of course has nothing to do with capacity. And of course they have lower ERs because it costs less to run them--lower turnover.

[Robert T]

.
Rick,

I read your book and noted the 0.9% per year higher annualized return of the 50:50 Europe:Pacific combination over EAFE from 1970-2004. If we assume no tax consequences from rebalancing then the after tax gain would be about 0.75% given the 0.15% superior tax efficiency of Vanguard Intl. Tax Managed versus the Europe:Pacific combination (at least this is what it was over last five years). This declines to about 0.68% due to the lower expense ratio of Vanguard Intl. Tax Managed versus the regional funds (0.20 vs. 0.27, lower if ETFs are used 0.15 vs. 0.18.). So if the tax consequence of rebalancing is lower than about 0.7% then the 50:50 combo may come out ahead. At a portfolio level (at least for me) this translates into about a 0.04% higher return (low, but I agree not zero).

It seems that the higher the number of asset classes included in a portfolio, the higher the level of funds needed to rebalance (if rebalancing with new money). (e.g. 5 x 25 rebalancing bands are smaller with 10 asset classes than with 2 implying potentially higher portfolio tax consequences from rebalancing by buying and selling).

Another important statistic from your book is that EAFE value had and annualized return of 2.2% more than EAFE over 1975-2004 and as a result I give more focus to value diversification than beta diversification (i.e. have substantial value allocation but don't have a separate allocation to Europe and Pacific). And as my preference is to try to rebalance with new money and not have to buy and sell to rebalance I am currently comfortable reducing the number of asset classes by holding one fund that tracks EAFE than two which track a Europe and Pacific index. Not sure if this makes sense…

SH,

I don’t have access to the long time series for the core strategies, but get the 70% VTI, 20% IJJ (instead of IJS) and 10% BRSIX closer to Core 1 than 2 and over the past year to end Sept 2007 Core 1 returned 15.39% while the combination returned 15.88% - so fairly close. While we can’t make judgments on one observation I think the two strategies will come out fairly close with existing instruments (Core and US vector) – IMO the difference isn’t larger than the cost of access (if we were to simply make an access comparison). This may change as the DFA line up expands with more TA and vector funds (or as the track record becomes clearer).

Robert
.

[SmallHi]

Robert,

All excellent points. For a value tilted investor, including EAFE Value provides valuable diversification from the standpoint that you have spread your tilt across continents. (its safter to assume risk premiums are global instead of country specific)

I think too much is made of historically low correlations between asset classes/regions. Relying on an individual country rebalancing bonus seems far more fickle than a global risk premium that adds meaningful long term diversification benefits to a domestic portfolio. Investors comfortable with the risk would be better off skipping EAFE, or equal weighted regional allocations and using Int'l Value and Small cap classes to add diversification to a US portfolio.

As for the long term comparisons between the portfolios I linked above, you are correct...there shouldn't be dramatic return disparities net of fees. That was kinda my point. There is validity in either approach. To say that the advent of the Core strategies was based purely on capacity is very short sighted, as is a comment that would claim that access to Core strategies (alone) is worth the fee of an advisor.

I know Rick prefers the component approach, and Larry prefers the Cores...you can make a case for both.

On one hand, you can argue that the fees are slightly lower on the component portfolio, and you have more control over your risk factors.

On the other, you can point to the convenience, internal efficiencies of packaging those risks in one integrated portfolio -- especially for the taxable investor without significant ongoing $ inflows, and slightly higher historical returns.

FWIW, the closest I could come to US Core 2 with a combo of TSM, CRSP 10, and S&P 600 Value was: 55%/10%/35% (based on montly regressions since 1995)

Since 10/05, that portfolio underperformed the net of fee return of DFQTX by about 0.35% annually throught September.

(although, going back to the 79, a 68% Russell 3000, 32% Russell 2000 Value mix would have most approximately matched the Core 2's factor exposure)

SH

PS -- I updated my post above to include a factor tracking portfolio.

[hafis50]

[Kenster1]

Robert,

I know that you've done your own little informal study to show that the S&P indexes capture the "value" factors "consistently" but I'm still trying to get a grasp or stronghold on trying to understand that.

To me, the S&P indexes (since their change) are still watered down versions with respect to value tilt. So I'm trying to understand how the value consistency fits into the fact that these indexes are just not as valuey.

For example YTD numbers:

Large Value:
IVE: +0.86% (iShares S&P500V)
IWD: -1.57% (iShares R1000V)
VTV: -0.76% (Vanguard LV)
DFLVX: -3.57% (DFA LV)

Mid Value:
IJJ: +1.51% (iShares S&P400V)
IWS: -1.74% (iShares Russell MV)
VOE: -2.79% (Vanguard MV)

Small Value:
IJS: -5.09% (iShares S&P600V)
IWN: -9.20% (iShares R2000V)
VBR: -5.96% (Vanguard SV)
DFSVX: -8.02% (DFA SV)

In a marketplace this year where Value has really slumped and Growth stocks have really turned up the heat, the S&P VALUE indexes (as shown above) YTD have been outperforming the other indexes because they're just not dipping as much into the bargain value stocks and are more blendy in nature relative to the other index/passive funds.

How do you see or interpret this?
_________________
SURGEON GENERAL'S WARNING: Any overconfidence in your ability, willingness and need to take risk may be hazardous to your health.

[SmallHi]

Kenster,

You have about nailed it. The S&P/Citigroup Indexes have the lowest value orientation of all the strategies you mentioned.

Don't read too much into quarter to quarter movements, however. Differences in index construction and small idiosincrasies tend to have a large impact over short periods.

SH

[SmallHi]

Kenster,

Below is a summary of 7 different ways you could arrive at an allocation that is significantly tilted to small cap and value (0.4 tilt to small cap; 0.6 tilt to value). In () is the annualized returns of the portfolios from 1995-2007.

RUSSELL (13.2%)
24% TSM
26% Russell Mid Cap Value
50% Russell Small Cap Value

MSCI (13.7%)
10% TSM
15% 1750 Growth
75% 1750 Value

S&P/Citigroup (13.0%)
60% 400 Value
40% 600 Value

S&P Pure Value (13.1%)
40% TSM
60% 600 Pure Value

DFA Component (14.3%)
20% TSM
30% Large Value (net of fees)
50% Small Value

DFA Core (14.3%)
30% Core 2
28% Large Value
42% Small Value

DFA Vector (14.5%)
60% Vector
22% Large Value
18% Small Value

Hope this helps.

SH

EDIT: added returns. (backed out DFA expense ratios so all holdings were gross of fees)

Also, the reason for the 0.60 Value tilt was it was the maximum tilt I could achieve with the S&P/Citigroup Indexes, and I wanted to include as many different index families as possible.

[Robert T]

.
Hafis50,

[Kenster1]

Thanks SH.

What about International?

In the developed markets, would you personally stick with overweighting the MSCI EAFE Value ETF and just add International Smallcap Blend and be done with it ... or do you think it would be worthwhile to also include the value-tilted DLS--WisdomTree International Smallcap Dividend ETF?

So in Developed International, an investor could do:

LB (MSCI EAFE)
SV (WisdomTree Int'l Smallcap--DLS)

===

LB (MSCI EAFE)
SB ((A few choices including a couple of new ones coming)
SV (WisdomTree Int'l Smallcap--DLS)

===

LV (MSCI EAFE Value)
SV (WisdomTree Int'l Smallcap--DLS)

===

LV (MSCI EAFE Value)
SB (A few choices including a couple of new ones coming)
SV (WisdomTree Int'l Smallcap--DLS)
_________________
SURGEON GENERAL'S WARNING: Any overconfidence in your ability, willingness and need to take risk may be hazardous to your health.

[Rodc]

[SmallHi]

[SmallHi]

[Robert T]

.
SH,

Very interesting analysis!!

I replicated the analysis and got similar index combinations and the same ordering of returns for1996-2006. I was surprised at the S&P and Russell performance which seems a little counter to the earlier FF3F results. The MSCI indexes do well. I also used a 0.2 and 0.4 load combination which produced a similar ordering.

Another interesting result is the corresponding FF factor benchmark (with the same 0.4 size and 0.6 value loads) had annualized returns of 14.4% for the 1996-2006 period - very similar to the DFA portfolios.

Robert
.

[hafis50]

[Robert T]

.
Hafis50,

I agree – patience is key even in the US.

Jan 1982 – Dec 1998 (17 years)
S&P500 Index 18.4%
Russell 2000 Index 13.0%

Jan 1965 – Dec 1981 (17 years)
S&P500 Index 6.3%
T-Bills 6.7%

From Random Drift in Asset Allocation

Robert
.

[Kenster1]

By the way Robert, thanks for pointing out that DFA article in the OP.

It offered a nice overview along with the pictured graphs which helped to visualize the value and size tilts.

Damn, the DFA Core 2 and Vector funds are nice -- invest and forget -- let the DFA fund managers continually manage and balance the targeted value and size tilts for you. Otherwise with some variations of TSM+LV+MV+SV+Microcap you're on your own! -- and can get tricky with extra taxes in a taxable account.

The Tax-Aware Core 2 Fund also looks interesting as well.

Tax-Aware Core 2 + International Core + EM Core -- and that's it, you can be done.

Where can I find DFA access with low minimums (e.g. $100k - $500k) and only 0.25% AUM fee?

SH -- you pulled together some sample portfolios above with some return values and they were based on ... (0.4 tilt to small cap; 0.6 tilt to value).

As a comparison, what is the size & value tilt for Core 2 and Vector?
_________________
SURGEON GENERAL'S WARNING: Any overconfidence in your ability, willingness and need to take risk may be hazardous to your health.

[SmallHi]

C2 and Vector (over the period of 7/95 through 9/07) were roughly 0.2/0.3 and 0.4/0.5 (SmB and HmL) respectively.

If you can save the management fee, wouldn't the time invested to build your own retail tilted equity portfolio be worth it?

If you want a 4 fund portfolio targeted to 0.3/0.4 (given that you like Core 2 and Vector), just use the following:

30% Citigroup 400 Value
30% Citigroup 600 Value
20% EAFE Value
10% EAFE Small
10% EM Index

...its pretty close.

I wouldn't pay for access or portfolio engineering. I would pay for ongoing advice and council, disciplined support and aid in other aspects of financial life. In this regard, I view the $2,000 to $3,000 fixed fee advisors as overpriced.

SH

[Kenster1]

FYI news for anyone using the Russell indexes...

http://www.russell.com/Indexes....entral.asp

[Kenster1]

Russell FAQ

http://www.russell.com/indexes....ion_qa.asp

[Robert T]

.
The earlier numbers posted by SH were a little surprising. The S&P Citigroup series performed slightly worse than expected (based on earlier FF3F regression analysis) and the Russell indexes performed slightly better than expected. I tried to understand why this was the case – here is what some of my subsequent analysis suggests (some of it posted below):

1. While the S&P600 value is relatively style consistent (higher R^2 and a closer to zero alpha than MSCI and Russell indexes in the FF3F regressions), the S&P400 value is less so relative to the other two indexes. And as the S&P 400 value accounted for 60% of the portfolio in the earlier analysis, overall portfolio returns were slightly lower than the MSCI and Russell portfolios.

2. Returns along the value and size dimension were not linear. Performance of a portfolio relative to its corresponding FF factor benchmark, between 1996-2006, depended on how value and size were targeted (table 1). Portfolios which used mid-cap value for value exposure and micro caps (CRSP10) for ‘small-cap’ exposure outperformed other combinations – even though the portfolios had the same size and value loadings.

3. The excess returns from the DFA portfolios posted by SH may be explained by point 2 above. i.e. DFA small value has a high size loading (closer to micro-caps) and so outperformed over the period.

4. The relative returns of microcaps (CRSP10) over the 1996-2006 period was slightly higher than the historical average (table 2), but the relative returns of mid-cap value seems much higher. As I recall, the value premium has historically been similar for mid and small caps (need to check).

5. The implications of the analysis seems to be (at least my reading): (i) for highest return for a given factor loading – take all size exposure in micro-caps (CRSP10), and (ii) as there are no CRSP10 value funds, the ‘most efficient’ way to get a value tilt is to select funds with a zero size loading (these are often mid-cap value funds) so as not to take any non-CRSP 10 ‘size’ exposure in the fund (which will ‘dilute’ a portfolios small cap exposure). (The analysis does not factor in taxes). In this respect a component approach with more control over how to achieve factor loading targets may have an advantage.

Table 1 below presents a set of portfolios with the same size and value loadings (0.43 and 0.40 respectively) but constructed in three different ways:(i) by adding a small value and small cap index to TSM; (ii) by adding a mid-cap value and small cap index to TSM; and (iii) by adding a mid-cap value and micro-cap index to TSM. The results show the third approach resulted in the highest portfolio return (and Sharpe ratio - except for the DFA portfolio) over the period.

[Rick Ferri]

[Robert T]

.

[SmallHi]

[SmallHi]

Rick,

Just curious, did you get Fama and French to admit they prefer tilting away from the Market with a combo of Bridgeway Ultra Small (or some other micro cap holding) and a SV ETF vs. their Core equity portfolios?

I am just teasing....(hard to effectively joke around in cyberspace)

SH

[Robert T]

.
SH,

[SmallHi]

Robert --

[Robert T]

.
SH,

Time period. I agree, almost all results are time period dependant. My preference is to use the longest period we have data available for both stocks and bonds. If we go back another 10 years the story seems more consistent with the last 10. The longest Mid-Cap Value series I have is from the Ibbotson Yearbook which has data from 1969. I used a 65:35 Mid-Cap Value:CRSP10 to approximate the factor loadings of the DFA vector fund (this is what my earlier numbers suggested, the results don’t change much if a 60:40 combination is used). The Vector returns data are from an earlier post of yours.

[SmallHi]

[Rick Ferri]

.

[Robert T]

.
SH,

The 1968-74 period seems to be the same story – small/micro-cap underperformance seems to have driven the relative result and the reverse again would likely be true if we had MCV back to 1927 – over the 1927-2006 period I would expect a P1-type portfolio to come out ahead - simply because of the non-linearity in the size premium (or the large relative CRSP10 premium). While the two portfolios may have the same factor loadings, a P1-type portfolio likely has higher risk (with all small-cap exposure in CRSP10), which is not captured by the factor loadings (at least this is my understanding). So no free lunch – but perhaps one way to beat a FF factor benchmark (although matching them is hard enough…).

[SmallHi]

Robert,

Have you ever checked the long term SmB/HmL loadings of a CRSP 10 portfolio?

Going back much more than 50 years, and a MCV/CRSP 10 portfolio is in a whole nother ball park from a Vector like allocation, and really no longer an apples to apples comparison. Even my 1968 - 2006 example began to drift away from target, as the HmL exposure of the component portfolio began to exceed Vector pretty noticeably (to the extent that the midcap allocation should probably be split 50% Mid Blend, 50% Mid Value)

Remember, Vector is going to stay pretty consistent around the 0.4/0.4 SmB/HmL range (usually +/- 0.1 or so). From 1928 - 2006, we are looking at an average CRSP 10 factor profile of 1.60/0.90!

As a matter of fact, going back to 1928, you could have gotten your Vector allocation out of a 50% US Large Growth, 50% CRSP 10 allocation. (And both had identical returns )

I guess this is one of the issues with using components to target risk factors...even their exposures can drift from one decade to the next pretty considerably. Of course CRSP 10 is much more guilty of this than any other portfolio I am familar with.

Here is an example:

[Robert T]

SH,

Thanks for the info. Yes components do seem to drift - but the portfolio drift may be smaller. The table compares the vector portfolio with individual components so maybe not a direct comparison. Vector may well have more constant factor loadings over time - but it would be interesting to see the fund turnover estimates from 1928 to 68. I would expect it to be much higher than from 1969 to present.

Last post for a while - Happy Thanksgiving to you all.

Robert
.

[paulob]

[Kenster1]

EM Value can come from WisdomTree's High-Yield Emerging Markets ETF.

Powershares RAFI EM ETF also appears to be somewhat value-tilted.

Although not released yet, iShares will be coming out with an MSCI EM Smallcap ETF.

There's also some actively-managed options as well.
_________________
SURGEON GENERAL'S WARNING: Any overconfidence in your ability, willingness and need to take risk may be hazardous to your health.