Who's the fairest of them all?

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Who's the fairest of them all?

Post by Robert T » Tue Dec 01, 2009 3:30 pm

.
There has been much debate on this forum on the role of asset allocation (including on the value of tilting away from the market and on which passive funds to use), security selection (active management), and market timing (specifically on moving average and inverted yield curve timing schemes).

Personally, I don’t think security selection or market timing add much, if anything, after all costs over the long-term. However, in the interests of learning (including to confirm or refute my priors), and to contribute to some of the debate, I have constructed as best I can, a set of low-maintenance portfolios. In each portfolio, one of the three sources of return will dominate: asset allocation, security selection, or market timing (table below). There are three portfolios in each of these three categories. Tracking performance of these over time may provide answers to the following questions:
  • 1. Will a DFA portfolio (of similar expected tax efficiency) add 1 to 2% [to 3%] (in alpha) over the same (very similar) factor loaded ETF portfolio?

    2. Is it possible to select a future “winning active management” fund following Swensen’s criteria (from his books)? [re: the Fairholme fund]

    3. Will a highly concentrated portfolio of the best “public equity” investment ideas of highly regarded “hedge fund” managers outperform?

    4. Will the commonly advocated moving average, inverted yield curve, or valuation based market timing schemes add value?

    5. Will any of the above outperform a simple total market portfolio?
The composition of each of the portfolios is described in more detail below. I will track performance over time, and periodically update the performance chart below. FWIW - my portfolio is similar to the ETF portfolio (has the same factor load targets).
  • Image
  • Before tax returns are reported above. All portfolios are expected to have an average long-term bond holding of 20-25%,
    although with wide variation over time in the case of the market timing portfolios
1. ASSET ALLOCATION: The DFA Tax-managed and ETF portfolios have very similar estimated factor loads across their US:Non-US developed:EM equity and fixed income allocations – based on this earlier analysis. Both portfolios have a 75:25 target stock:bond allocation, with a global small cap and value tilt. To ensure replicability of results, the portfolios will in future be rebalanced annually. The three asset allocation portfolios are:
  • (a) The DFA TM portfolio: 17% DFA TM Mktwide Value, 7% DFA TM Targeted Value, 13% DFA TM Small Cap, 17% DFA TM Intl. Value, 11% DFA Intl Large Co., 10% DFA EM, 25% DFA Intermediate Government. Here it is.
    (b) The ETF portfolio: 23% Vanguard Midcap Value, 9% iShares S&P600 Value, 5% Bridgeway Ultra-Small Co., 23% iShares EAFE Value, 5% iShares EAFE Small, 10% Vanguard EM, 25% iShares 3-7 yr Treasury. Here it is.
    (c) The Total market portfolio: 75% Vanguard Total World Stock Market: 25% Vanguard Total Bond Market. Here it is.
2. SECURITY SELECTION: The first two portfolios are highly concentrated or focused on a small number of stocks (to try to ensure security selection drives portfolio returns) – based on this earlier analysis, the third portfolio is slightly less concentrated.
  • (a) The Fairholme portfolio (FAIRX) seems to pass Swensen’s ‘active manager’ selection criteria with fairly high marks (see previous link) and has about a 20% bond allocation. Here it is.
    (b) The Cummings-Ackman-Lampert portfolio includes the ‘best ideas’ (top two stock holdings) of three highly concentrated ‘hedge fund’ managers (on the assumption they know more about the businesses than any other investors). The resulting portfolio has 6 stocks, following Buffets view that :“If you really know businesses, you probably shouldn’t own more than 6 of them”…”If you can identify six wonderful businesses that is all the diversification you need”. Here is more information and public stock holdings of these managers: Ian Cummings: public stock holdings.Bill Ackman: public stock holdings. Eddie Lampert: public stock holdings. The stock holdings of the Cummings-Ackman-Lampert portfolio will be reviewed quarterly against the the respective 13F SEC submissions, selecting the top two holdings of each. These don’t seem to change much, so turnover is expected to be low. Portfolio changes will be made accordingly. The portfolio has a 75:25 stock:bond target for comparability purposes. Here it is.
    (c) The Dodge & Cox portfolio is 75% Dodge and Cox Global Stock:25% iShares 3-7 yr Treasury. While the portfolio is less concentrated than the previous two, it currently only holds 88 stocks compared to the 2,741 stocks in the Vanguard Total World Stock fund. And if earlier analysis is any guide, it will also likely have a value tilt (around a 0.4 value load). In addition to stock selection, the fund allocations across US:EAFE:EM may also vary over time which may be another factor impacting returns.Here it is.
3. MARKET TIMING: Potfolio details.
  • (a) The 16 month moving average portfolio will be 100% DFA Global Equity [DGEIX] (as a proxy for a globally diversified small cap and value tilted portfolio) when the current ‘price’ is above the 16 mth moving average of the MSCI AC World Value (net) Index & 100% iShares 3-7 yr Treasuries [IEI] when the current ‘price’ is below the 16 mth moving average. The assessment is done at the end of each month with any change starting on the 1st of the month following the cross-over. A 16 month moving average is used (as opposed to the more common 200 day moving average) as it was the best performing moving average period using data from 1926 – based on this earlier analysis. Over time, the portfolio is expected to have about a 25% average bond allocation (as it did in the backtests from 1926). Here it is.
    (b) The inverted yield curve portfolio is based on this earlier analysis. The portfolio will be 100% bonds (iShares 3-7 yr Treasury [IEI]) from month 11 to 18 after yield curve inversion, and will have a 83:17 stock:bond target allocation for the rest of the time. As the earlier linked back-test from 1959-2009 had an average 10% bond holding, the average bond holding of the inverted yield curve portfolio is expected to be 25% (17*0.9+100*0.1=25.3 i.e. 17% bond allocation for 90% of the time, then 100% bond holding for 10% of the time, gives an average bond holding on 25.3%) As before, DGEIX is used as a proxy for a global small cap and value tilted portfolio. Again the portfolio is expected to have about a 25% average bond allocation over time. The market timing portfolio back-tests (in the table above) from 2003-08 used the MSCI AC World (Net) Value Index for the equity allocation and the Barclays Capital 3-7 yr Treasury for the bond allocation. Here it is.
    (c) The GMO valuation/forecasts portfolio comprises 10 ‘index’ funds aligned to most of the asset classes that GMO provides in its forecasts. It weights each fund/asset class by the expected real return from the GMO 7 yr asset class forecasts at the beginning of each year (equivalent to annual rebalancing to new GMO forecasted real return weights at the beginning of each year). The baseline was based on 30-yr expected returns from the start of 2003 for about a 75:25 stock:bond buy-and-hold portfolio (this influenced the number of asset classes to include. Here it is. Backtests worked just as well with fewer funds.
Robert
Last edited by Robert T on Sat Jan 30, 2010 6:11 am, edited 6 times in total.

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Post by Rodc » Tue Dec 01, 2009 4:44 pm

Will be interesting to watch.

Could you add standard deviation or Sharpe ratio or other risk measure of your choice?

Thanks
We live a world with knowledge of the future markets has less than one significant figure. And people will still and always demand answers to three significant digits.

medicevans
Posts: 122
Joined: Sun Oct 18, 2009 8:37 am

Post by medicevans » Tue Dec 01, 2009 4:52 pm

Interested to see how this comes out.

Posting to subscribe to post.

User avatar
sergeant
Posts: 1028
Joined: Tue Dec 04, 2007 11:13 pm

Post by sergeant » Tue Dec 01, 2009 6:27 pm

Thanks for doing this. I am pulling for the IYC to get clobbered so as to quiet Adrian. :D
Lincoln 3 EOW!

User avatar
jeff mc
Moderator
Posts: 2720
Joined: Mon Feb 19, 2007 9:42 pm
Location: minnesota
Contact:

Post by jeff mc » Tue Dec 01, 2009 7:24 pm

medicevans wrote:...Posting to subscribe to post.
tip: you can always simply click the link at the bottom of the thread that says "Watch this topic for replies" without needed to actually reply to subscribe to any thread

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Wed Dec 02, 2009 5:41 am

Rodc wrote:Could you add standard deviation or Sharpe ratio or other risk measure of your choice?
Rodc,

I have added the 2008 performance of each of the portfolios to the above table (as one measure of 'downside' risk). May also add standard deviation to annual updates (as number of observations increase).

Robert
.

User avatar
jeffyscott
Posts: 7587
Joined: Tue Feb 27, 2007 9:12 am
Location: Wisconsin

Post by jeffyscott » Wed Dec 02, 2009 8:35 am

How about a strategic asset allocation (timing) scheme based on valuations? Maybe whatever GMO fund it is that has the most flexibility to move money around could serve as a simple proxy for this?
press on, regardless - John C. Bogle

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Wed Dec 02, 2009 11:21 am

jeffyscott wrote:How about a strategic asset allocation (timing) scheme based on valuations? Maybe whatever GMO fund it is that has the most flexibility to move money around could serve as a simple proxy for this?
I don’t think the GMO funds adhere too well to GMOs forecasts. Here’s an alternative portfolio. It comprises 10 ‘index’ funds aligned to most of the ‘asset class’ forecasts that GMO provides in its forecasts. It weights each fund/asset class by the expected return from the GMO 7 yr asset class forecasts at the beginning of each year (annual rebalancing to new GMO forecasted return weights at the beginning of each year). I back-tested the portfolio to the start of 2003. The baseline at the start of 2003 based on 30-yr expected returns was about a 75:25 stock:bond buy-and-hold portfolio (getting it close to the 75:25 stock:bond allocation also determined the number of asset classes to include). Here’s how the portfolio performed over the comparable time period used for the results in the OP.

2003 to end Nov. 2009
Annualized return = 14.6%
Loss in 2008 = -20.3%

The asset allocation changed quite dramatically over the period – from 37% bonds in 2003, to 66% in 2007, to 18% in 2009. So tax costs would be high, but nevertheless and interesting result. May add this to the above table – FWIW last month I set up a portfolio in M* based on this approach: here it is. The Bridgeway Blue Chip fund was the closest 'High Quality" index type fund I could find. Perhaps not a good match.

For possible rebalancing considerations?

Robert
.

User avatar
dphmd
Posts: 369
Joined: Thu Mar 01, 2007 10:19 pm
Location: Georgia

Post by dphmd » Wed Dec 02, 2009 11:33 am

Good questions, and very relevant to many discussions on the Board.
Discipline is freedom.

User avatar
jeffyscott
Posts: 7587
Joined: Tue Feb 27, 2007 9:12 am
Location: Wisconsin

Post by jeffyscott » Wed Dec 02, 2009 11:43 am

Robert T wrote:I don’t think the GMO funds adhere too well to GMOs forecasts.
I think you are correct, I was trying to make it as simple as possible.
The Bridgeway Blue Chip fund was the closest 'High Quality" index type fund I could find. Perhaps not a good match.
The high quality thing is a difficulty as it is not as well defined as their other categories. I recently began allocating some money to Vg dividend growth to try to capture at least some of this. That is not an index, but there is a similar index fund VDAIX. I have no idea if this is a better or worse proxy than the Bridgeway fund.
press on, regardless - John C. Bogle

User avatar
BlueEars
Posts: 3657
Joined: Sat Mar 10, 2007 12:15 am
Location: West Coast

Post by BlueEars » Wed Dec 02, 2009 1:04 pm

Hi Robert, this should be a good exercise over the years. I have one other MT approach which you might consider though I know you've bitten off quite a chunk of work already. It involves equity rotation between a broad US portfolio and a broad international one. I believe it captures multi-year currency moves in the dollar. Here are the rules:

1) buy the equity that has the highest 12 month return at the first trading day of each month
2) stay in that position for 3 months before switching

The CAGR's for 38.25 years (Jan 1971 to Mar 2009) using EAFE and SP500:

Code: Select all

buy-hold EAFE:   6.8%
buy-hold SP500:  5.8%
equity rotate:   8.9%  (0.7 switches/year with 51% of time in EAFE)
Note SP500 index used here does not include dividends. Not sure if the EAFE includes dividends.
Last edited by BlueEars on Wed Dec 02, 2009 1:06 pm, edited 1 time in total.

the fixer
Posts: 75
Joined: Sat Jul 25, 2009 6:19 pm

Post by the fixer » Wed Dec 02, 2009 1:06 pm

1. Will a DFA portfolio (of similar expected tax efficiency) add 1 to 2% [to 3%] (in alpha) over the same (very similar) factor loaded ETF portfolio?
It’s really quite straightforward to compare factor matched index/ETF portfolios to DFAs underlying indexes over longer periods of time to avoid the flaws in 5-7 year time dependent observations. Doing so produces the 1-3% outperformance mentioned above:

1986-2008 annualized returns (inception for Russell Mid Value)
• Both portfolios have an identical SmB/HmL tilt

12% Russell 3000 Index
44% Russell Mid Value
44% CRSP 10 Index
+9.6%

100% DFA US Vector Index
+10.6%

1999-2008 annualized returns (inception for EAFE Small Cap (gross div.))
• Both portfolios have an identical SmB/HmL tilt

40% EAFE Value (net dividends)
60% EAFE Small Cap (gross dividends)
+4.0%

100% DFA Int’l Vector Index
+6.0%

1988-2008 annualized returns (inception for both indexes)
• Both portfolios are market oriented and have no dedicated size or value tilts

100% MSCI Emerging Markets Index (gross div.)
+11.4%

100% DFA Equally Weighted Emerging Markets Index
+14.4%

On another note, to take this exercise to the logical extreme, lets compare the most small cap value (and therefore, highest expected returning) index used in these examples, the S&P 600 Value, with a less small/value oriented US Vector index over the longest period available:

1995-2009 annualized returns (7/95 inception for S&P 600 Value)
• The S&P 600 Value index has a 20% to 30% stronger tilt to small cap and value

100% S&P 600 Value
+9.0%

100% DFA US Vector Index
+9.7%

This example shows that even more torqued index portfolios may not achieve the returns available to a less tilted, but better constructed, DFA strategy. In my opinion, the proposed ETF/DFA comparison above in the original post is another example of a more tilted ETF portfolio only matching the returns of a less tilted DFA allocation.

And of course these are just index returns I've shown above, and on the DFA front, there are various real world trading applications (securities lending, momentum, etc.) that add additional value when it comes to implementing their indexes in live portfolios. These have been disputed (in my opinion, incorrectly), and I’ll follow up in this thread with further thoughts if time permits (assuming I don't lose interest :lol: ).

peter71
Posts: 3768
Joined: Tue Jul 24, 2007 8:28 pm

Post by peter71 » Wed Dec 02, 2009 1:30 pm

Hi Robert,

Over the long term it shouldn't much matter whether you look at 200-day or 16-month moving averages, whether you look at switching out of stocks at 6, 12 or 18 months after an inversion and so on, but a natural question at this stage would be whether, e.g., using the 200-day moving average would have made market-timing look better or worse over the period of analysis?

Another thing that seems a little strange to me is the choice of an (almost) 7-year study period, coincidentally commencing around the bottom of a bear market. All else equal, I think you'll have more persuasive evidence against market timing and avoid criticisms of "endpoint bias" if you also look at more standard periods of 1,3,5 and 10 years.

Last but not least, any thoughts of doing something with valuations-based timing?

All best,
Pete

the fixer
Posts: 75
Joined: Sat Jul 25, 2009 6:19 pm

Post by the fixer » Wed Dec 02, 2009 6:22 pm

Along the lines of my US example above, using the Russell 3000 Value and CRSP 10 indexes, we can stretch the performance observation period back a full 30 years to 1979, which enhances the confidence and expected persistence of the findings. Conveniently, a 50% Russell 3000 Value, 50% CRSP 10 component index also has (over this period) the exact same small cap and value loadings as the DFA US Vector index, but a risk/return comparison sheds some light on certain component index fund disadvantages:

• Over the entire period, despite identical factor loadings, the annually rebalanced component index portfolio returned +11.7% annually (ignoring rebalancing costs). The US Vector index delivered +13.4%, about a 1.7% per year advantage.

• Because of the remote decile 10 exposure, the component portfolio tended to have more extreme short term returns, losing money during 8 of the 30 years, vs. only 5 yearly losses for US Vector.

• Also, because of the more diversified total market structure of the Vector portfolio relative to an extreme component allocation, you see greater tracking error for the later during periods when small cap and value stocks underperform: from 1984 to 1990, the Russell 3000 compounded at +13% per year, the US Vector index had an annualized +9% return, whereas the Russell 3000 Value/CRSP 10 combo only returned +4.8%, or about twice the tracking error.

Also note, during this period, international stocks would have eased this regret, depending on your allocation. The MSCI EAFE Value Index returned +23.2% per year, while the DFA Int’l Vector index compounded at about 4% more per year, or 27.5% annually, and the DFA Int’l Small Value index topped them all at +32.6% per year (data for MSCI EAFE Small not available).

The same portfolio comparison above could have instead been made between the DFA US Vector Index and a 3 index component combo of 35% Russell 3000, 55% Russell 2000 Value, and 10% CRSP 10 Index, another mix with identical factor loadings over a satisfactory 30 year period beginning in 1979. This mix bested the Russell 3000 Value/CRSP 10 blend by +.7% per year, annualizing at +12.4% per year, or only 1% less than the DFA US Vector index (before costs associated with maintaining and rebalancing 3 separate funds).

There are many ways to look at this, and the 1% to 2% seems to be fairly common, with that number growing for non-US allocations.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Wed Dec 02, 2009 6:38 pm

.
Jeffy, Not sure on the closest ‘High Quality” index match (re: VDIAX). The Bridgeway fund has similar sector weights as the GMO high quality fund (large for health and consumer stables if I recall). But that's only one measure, not sure if its the most important in this regard.

Les, There are likely many more alternatives for market timing. I will just try to stick to one’s I have, as they seem to get the most air-time here.

Pete, All I can say is the 16 month moving average was the best performing back-test using the longest data I have, and testing averages from 2 to 24 months. The 7 year back-test was purely for interest sake, the reason: that is when I started my current investment policy. Nothing more. The interest is how will it perform in the future. The GMO portfolio (posted earlier) is valuation-based. Grantham uses estimates of valuation to calculated his asset class forecasts.

On the earlier reported DFA back-tests:

We only have readily available data for one of the comparisons presented in the earlier post - namely EM – which had the largest difference of all the comparisons – the increasingly mentioned (claimed) 3% DFA out-performance (11.4 vs. 14.4). Together with this:
the fixer wrote:And of course these are just index returns I've shown above, and on the DFA front, there are various real world trading applications (securities lending, momentum, etc.) that add additional value when it comes to implementing their indexes in live portfolios.
Here are the actual returns of the DFA EM and Vanguard EM funds from 1995 to date. The annualized return difference was 0.6%!. And arguably over time a more equal weighted country series (as in the DFA fund) will have a greater value and small-cap tilt relative to a simple country cap weighted series (as seems to be suggested by simple comparisons of the cap and equal weighted portfolios on Ken French’s website).

Reasonable doubt?

Code: Select all


Returns (%)
                        DFA EM    Vanguard EM
		
             1995         2.2          0.6
             1996        11.4         15.8
             1997       -18.9        -16.8
             1998        -9.4        -18.1
             1999        71.7         61.6
             2000       -29.2        -27.6
             2001        -6.8         -2.9
             2002        -9.5         -7.4
             2003        60.2         57.7
             2004        29.9         26.1
             2005        29.9         32.1
             2006        29.2         29.4
             2007        36.0         38.9
             2008       -49.2        -52.8
     End Nov 2009        65.8         69.9
				
Annualized return         8.6          8.0

Robert
.
Last edited by Robert T on Wed Dec 02, 2009 6:48 pm, edited 1 time in total.

peter71
Posts: 3768
Joined: Tue Jul 24, 2007 8:28 pm

Post by peter71 » Wed Dec 02, 2009 6:48 pm

Robert T wrote:.
Pete, All I can say, is the 16 month moving average was the best performing back-test using the longest data I have, and testing averages from 2 to 24 months. The 7 year back-test was purely for interest sake, the reason: that is when I started my current investment policy. Nothing more. The interest is how will it perform in the future. The GMO portfolio (posted earlier) is valuation-based. Grantham uses estimates of valuation to calculated his asset class forecasts.
.
Hi Robert,

On GMO forecasts, I'm sorry I missed that above, though I must have written 500 posts on P/E 10 based forecasts at this point and I can't say I have any idea what a GMO forecast is . . .:D

On the other operationalization choices, I can't say I'm really persuaded there either, but a) it's your thread and b) there's no shortage of published articles out there so I'll stop giving you crap here and just focus on the latter . . . :D

All best,
Pete

User avatar
BlueEars
Posts: 3657
Joined: Sat Mar 10, 2007 12:15 am
Location: West Coast

Post by BlueEars » Wed Dec 02, 2009 7:16 pm

Robert T wrote:.
...
Les, There are likely many more alternatives for market timing. I will just try to stick to one’s I have, as they seem to get the most air-time here.
.
Hi Robert, I agree the equity/short_term_bonds mix is the most talked about here in reference to market timing. I was pointing out that one could time the equity_US/equity_international. Other timings that I've looked at are value/growth (SV/SG/Treasury, MV/MG/Treasury, LV/LG/Treasury). Then there are the closet emotional timers we see a lot on Bogleheads: REITs, gold, etc. when those markets heat up :wink: !!

User avatar
stratton
Posts: 11082
Joined: Sun Mar 04, 2007 5:05 pm
Location: Puget Sound

Post by stratton » Wed Dec 02, 2009 7:40 pm

Robert T wrote:.
Jeffy, Not sure on the closest ‘High Quality” index match (re: VDIAX). The Bridgeway fund has similar sector weights as the GMO high quality fund (large for health and consumer stables if I recall). But that's only one measure, not sure if its the most important in this regard.
Morningstar did a wide moat and quality measiring article. Sorry no link.

The Bridgeway fund like you mention is one of the ones that came up. A couple of Vanguard funds would be Dividend Growth (VDIGX) and Dividend Appreciation etf (VIG). Companies with growing dividends, but not neccessarily high yields appear to fit the high quality conditions.

Paul

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Thu Dec 03, 2009 3:11 am

.
Pete,

Here’s a bit more on GMO’s 7 year asset class forecasts. You may have to register on the GMO site to access the links – but its free.
Bill Bernstein on Grantham [GMO] forecasts: “…one of the few folks who arrives at expected asset-class returns the right way—by looking at current valuations laced with realistic analyses of per-share growth metrics, viewed on a background of historical pricing.”

Robert
.

Rodc
Posts: 13601
Joined: Tue Jun 26, 2007 9:46 am

Post by Rodc » Thu Dec 03, 2009 9:04 am

Aren't the Russell indexes known to be the worst indexes to use as the basis for an index fund, due to front running issues and maybe other factors?

If so, beating a Russell index by a point or two would not mean much.

It would seem more useful to compare to either real funds based by other index families, or to other indexes if lacking sufficient history for real funds.
We live a world with knowledge of the future markets has less than one significant figure. And people will still and always demand answers to three significant digits.

User avatar
jeffyscott
Posts: 7587
Joined: Tue Feb 27, 2007 9:12 am
Location: Wisconsin

Post by jeffyscott » Thu Dec 03, 2009 9:37 am

Robert T wrote:Jeffy, Not sure on the closest ‘High Quality” index match (re: VDIAX). The Bridgeway fund has similar sector weights as the GMO high quality fund (large for health and consumer stables if I recall).
Yesterday, I used m* instant x-ray to look at what they call "stock type". I could be mistaken, but I was thinking that may be more important than the sectors in defining "quality". Comparing one of GMO's quality funds to Bridgeway Blue Chip and VDAIX, relative to VTSMX did not really tell me much, though. The one thing in VDAIX favor was that GMO and it both had fairly similar overweights of "classic growth". But other things went the opposite way. The Bridway fund is certainly far more similar in terms of the size of companies included.

GMO's fund has about 75% of assets in it's top 25 holdings. Bridgeway's has about 46% in these same stocks and VDAIX has 39%.

However, in the m* analyst report for VDAIX it does say:
About 90% of the fund's assets are invested in firms with economic moats, according to Morningstar equity research (70% have wide moats).

Thanks to investors' preference for lower-quality fare, the fund has lagged this year.
...this fund's high-quality portfolio...
Meanwhile the Brideway fund is simply an equal weighted index of the 35 largest companies. Perhaps those do typically tend to be in the "high quality" category, though?

Finally, looking at the m* growth of $10,000 chart. The three funds all are fairly close to each other and the S&P 500 until some separation occurs in the decline starting in late 2008. In the decline and recovery, VDAIX tracked a bit closer to GMO quality than Bridgway BC 35 did.
press on, regardless - John C. Bogle

User avatar
Maestro G
Posts: 83
Joined: Fri Aug 03, 2007 7:55 pm
Location: San Francisco

Post by Maestro G » Thu Dec 03, 2009 12:32 pm

Hi Robert,

This should prove to be a very interesting comparison. Would you consider adding the Harry Browne Permanent Portfolio philosophy to the allocation category of the analysis?

If so, for purposes of a more apples to apples comparison, I would suggest the following vehicles:

VT - 25%
GLD - 25%
TLT - 25%
TUZ or SHY or I-series Savings Bonds - 25% or some combination of all for the cash portion might be interesting.

Thanks much for setting this up and your consideration of my suggestion :beer

Maestro G
Yesterday is history, tomorrow is a mystery, today is a gift, that's why it's called the present. Most daily market noise is "a tale told by an idiot, full of sound and fury, signifying nothing."

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Thu Dec 03, 2009 6:44 pm

.
Maestro,

I think a maximum of three per category is about my limit! Maybe for another day.

Robert
.

User avatar
Random Musings
Posts: 5339
Joined: Thu Feb 22, 2007 4:24 pm
Location: Pennsylvania

Post by Random Musings » Fri Dec 04, 2009 10:59 am

Robert,

With respect to your calculation of GMO - how is the "weight" of each asset class calculated from the expected returns tables? Is it independent of the baseline or a function of? Not sure how calcs are determined?
I don’t think the GMO funds adhere too well to GMOs forecasts. Here’s an alternative portfolio. It comprises 10 ‘index’ funds aligned to most of the ‘asset class’ forecasts that GMO provides in its forecasts. It weights each fund/asset class by the expected return from the GMO 7 yr asset class forecasts at the beginning of each year (annual rebalancing to new GMO forecasted return weights at the beginning of each year). I back-tested the portfolio to the start of 2003. The baseline at the start of 2003 based on 30-yr expected returns was about a 75:25 stock:bond buy-and-hold portfolio (getting it close to the 75:25 stock:bond allocation also determined the number of asset classes to include). Here’s how the portfolio performed over the comparable time period used for the results in the OP.
Regards,

RM

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Fri Dec 04, 2009 12:09 pm

.
RM,

The weight is simply, using US Large cap as an example: expected return US Large cap/ (sum of expected return of all asset classes in the portfolio) - where the expected return = the GMO 7yr asset class return forecasts. I used nominal returns for the weighting calculation (real returns + the inflation number given at the bottom of the GMO real return sheet, mainly to reduce the number of negative weigths, and remaining negative numbers were assigned a zero allocation). Here are the annual allocations resulting from this methodology for the 10 funds included in the above linked portfolio:

Code: Select all

Asset Allocation (%) for the GMO valuation/forecast portfolio 

                                        2003   2004   2005   2006   2007   2008  2009
 
Vanguard 500 Index                         5      0      2      4      2      5    11
Bridgeway Blue Chip                        5      0      2      4     17     18    16
Vanguard Small Cap                         9      2      0      3      0      3    11
Vanguard Europe Pacific                   13     14     15     11      8      9    13
Vanguard FTSE All Wld ex-US Small Cap     18     17     13      9      2      6    16
Vanguard EM                               17     19     23     18     14      9    14
Vanguard Short Term Treasury               6      9     11     13     14     14     3
Vanguard Intermediate Term Treasury        7     11     10     12     15     13     2
Vanguard Inflation Protected Secs          8     10     10     12     16     11     5
Fidelity New Markets Income               13     17     15     13     12     12     7


SUMMARY
US equity                                 18      3      3     12     20     26    38  
Non-US Developed equity                   31     31     27     20     10     15    29 
EM equity                                 17     19     23     18     14      9    14
Fixed income                              34     47     46     50     56     49    18
Fixed income (excluding EM Debt)          21     30     31     37     45     38    11


Resulting Annual Returns  (%)           37.6   16.9   15.2   14.2   12.5  -20.3  33.9  

So very high turnover (tax inefficient). Also just used the GMO numbers for Bridgeway large cap to 2007, then used their 'high quality' numbers for 2008/09. I also tested the approach with less asset classes - just S&P500, EAFE, EM, Int. Treasury, and EM debt, with very similar results. Perhaps instructive for rebalancing?

Robert
Last edited by Robert T on Sun Dec 06, 2009 2:04 pm, edited 1 time in total.

DP
Posts: 659
Joined: Thu Apr 17, 2008 5:19 pm

Post by DP » Fri Dec 04, 2009 4:25 pm

Hi,
Thanks for putting this together. Next step might be to try combining the strategies? :D. You might also be interested in Alphaclone.com. They have a number of ways of combining top hedge fund selections, both in live portfolio's and in backtests. A few combinations are free.
tip: you can always simply click the link at the bottom of the thread that says "Watch this topic for replies" without needed to actually reply to subscribe to any thread
In this regard this forum is not so well behaved. It constantly stops sending me updates on threads I subscribe to. The best work around that has been suggested is to comment in the post, then I can always search for posts that I have commented on.

Don

User avatar
Adrian Nenu
Posts: 5228
Joined: Thu Apr 12, 2007 6:27 pm

Post by Adrian Nenu » Fri Dec 04, 2009 5:05 pm

Thanks for doing this. I am pulling for the IYC to get clobbered so as to quiet Adrian.
The ones who did not pay attention to the YIC and took on too much risk are the ones who got clobbered and will get clobbered again. It's all about reducing risk of loss, not maximizing return. The IYC has been 100% accurate as a recession indicator. Those who disagree should ask themselves what do they know about the economy that the bond market doesn't know.

Adrian
anenu@tampabay.rr.com

User avatar
BlueEars
Posts: 3657
Joined: Sat Mar 10, 2007 12:15 am
Location: West Coast

Post by BlueEars » Fri Dec 04, 2009 6:34 pm

Adrian Nenu wrote:...
It's all about reducing risk of loss, not maximizing return. The IYC has been 100% accurate as a recession indicator. Those who disagree should ask themselves what do they know about the economy that the bond market doesn't know.
Hi Adrian, I'm not trying to refute your thesis but am a bit puzzled. Robert's data above showed fairly poor results in 2008 for his version of IYC. Do you recommend a different version? If so can you point me to your discussion on this, thanks.

User avatar
BlueEars
Posts: 3657
Joined: Sat Mar 10, 2007 12:15 am
Location: West Coast

Post by BlueEars » Sat Dec 05, 2009 12:23 am

Hi Robert, just wondering how you define the date of the yield curve inversion i.e. what Treasury's do you use? Thanks in advance.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Sat Dec 05, 2009 7:19 am

Les wrote:Hi Robert, just wondering how you define the date of the yield curve inversion i.e. what Treasury's do you use? Thanks in advance.
When 3 month treasury yields exceed 10 year treasury yields as presented on the website of the Federal Reserve Bank of New York (see linked spreadsheet).

Robert
.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Sat Dec 05, 2009 10:39 am

Adrian Nenu wrote:The ones who did not pay attention to the YIC and took on too much risk are the ones who got clobbered and will get clobbered again. It's all about reducing risk of loss, not maximizing return. The IYC has been 100% accurate as a recession indicator. Those who disagree should ask themselves what do they know about the economy that the bond market doesn't know.
Its about having enough at and during retirement. If someone wants to avoid losses then they should own t-bills or individual TIPS, and save more. Market timing around the yield curve to reduce loss, as a substitute for a constant exposure to bonds, has historically proven to be highly inconsistent. The 20% market decline from end of 1961 to mid 1962, the 30% market decline in 1987 (Sep to Nov), the 20% decline in 1998 (July/August) were not proceeded by an inverted yield curve. Investors who thought that risks only show up when the yield curve inverts (and had shifted to higher stock allocations) would have got “clobbered”. Similarly investors being excessively cautious (shifting to a bond portfolio) once the yield curve inverted would have missed out on the 15 to 30% market gains for the year after the Sept 1996 and Oct. 1980 inversions, and the 40 to 60% gains for the two years after those inversions. These are simply the historical facts. But we each have to do our own analysis, and make our own decisions.

Will the future provide more consistent results? Don’t know.

Robert
.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Sun Dec 06, 2009 1:54 pm

.
Out of interest, I just looked at using real return weights for the GMO valuation/forecast portfolio. Here are the results compare to the earlier nominal return weighted portfolio.

Code: Select all

2003 to end Nov. 2009
                                            Annualized     Decline 
                                             return, %     in 2008
                       
GMO portfolio (Real return weighted)            17.5        -13.0      
GMO portfolio (Nominal return weighted)         14.4        -20.3      

ETF (buy-and-hold – similar to mine)            11.0        -27.3

Here are the resulting asset allocations using the weights derived from the real returns (table below). The average annual bond allocation (excluding EM) was 29%. The backtests are indeed fairly impressive! But with substantial turnover (tax inefficiency). I will use the real return weights in future or the GMO valuation/forecast portfolio. Will also likely give more consideration to these weights in rebalancing decisions of my own portfolio.

Code: Select all


Asset Allocation (%) for the GMO valuation/forecast portfolio 

                                        2003   2004   2005   2006   2007   2008  2009 
  
Vanguard 500 Index                         2      0      0      0      0      0    12 
Bridgeway Blue Chip                        2      0      0      0     24     39    18 
Vanguard Small Cap                         9      0      0      0      0      0    11 
Vanguard Europe Pacific                   15     15     16     10      0      3    14 
Vanguard FTSE All Wld ex-US Small Cap     22     20     12      4      0      0    19 
Vanguard EM                               21     23     33     29     14      0    16 
Vanguard Short Term Treasury               5      6      8     15     16     20     0 
Vanguard Intermediate Term Treasury        5      9      7     12     17     17     0 
Vanguard Inflation Protected Secs          6      9      6     14     21      9     3 
Fidelity New Markets Income               15     19     17     15      8     13     6 


SUMMARY 
US equity                                 12      0      0      0     24     36    41  
Non-US Developed equity                   36     34     28     15      0      3    33 
EM equity                                 21     23     33     29     14      0    16 
Fixed income                              31     43     39     56     62     59    10 
Fixed income (excluding EM Debt)          16     24     21     41     53     46     4 


Resulting Annual Returns  (%)           41.3   18.4   18.4   14.9   12.7  -13.0  36.3     

Robert
.

dbonnett
Posts: 726
Joined: Tue Jun 10, 2008 3:59 pm

Post by dbonnett » Sun Dec 06, 2009 2:07 pm

Isn't there a lot of recency involved in the above asset class weightings?

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Sun Dec 06, 2009 2:25 pm

dbonnett wrote:Isn't there a lot of recency involved in the above asset class weightings?
The weights are based on the GMO 7-year Asset Class Return Forecasts (posted on the GMO site). For example, their 7-year asset class return forecasts as of December 31, 2002 were used as the asset class weights for the asset allocation in 2003, the asset class return forecasets as of December 31, 2003 were used as the weights for the asset allocation in 2004 etc. So not sure recency plays much of a role (only to the extent that it affects the valuation estimates used for the GMO forecasts).
.

User avatar
jeffyscott
Posts: 7587
Joined: Tue Feb 27, 2007 9:12 am
Location: Wisconsin

Post by jeffyscott » Sun Dec 06, 2009 3:26 pm

You know, since they do not publish the forecasts until 2-3 weeks after the end of the month, I am wondering if you should either use the November forecasts (which are not available for past years on their web site) or make the portfolio moves as of, say, Feb 1?

(Although, it is actually possible to adjust in real time, by looking at the change in prices since the last forecast was published and adjusting the forecast returns accordingly.)
press on, regardless - John C. Bogle

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Sun Dec 06, 2009 5:49 pm

jeffyscott wrote:You know, since they do not publish the forecasts until 2-3 weeks after the end of the month, I am wondering if you should either use the November forecasts (which are not available for past years on their web site) or make the portfolio moves as of, say, Feb 1?

(Although, it is actually possible to adjust in real time, by looking at the change in prices since the last forecast was published and adjusting the forecast returns accordingly.)
Yes, that’s right. For the GMO portfolio, I will use the November GMO forecasts, which will likely be published towards the end of this month, to determine the allocations for start of Jan. I don’t expect them to change dramatically from month to month bar another October 2008 month, but will also look at the December estimates published towards the end of January.
.

User avatar
BlueEars
Posts: 3657
Joined: Sat Mar 10, 2007 12:15 am
Location: West Coast

Post by BlueEars » Sun Dec 06, 2009 5:51 pm

Hi Robert, does GMO precisely describe how they go about forcasting? If not then why would one give any more weighting to this versus using something like a moving average or perhaps 12 month performance to manage rebalancing? I personally do not like black boxes no matter how good they were in the past.

User avatar
jeffyscott
Posts: 7587
Joined: Tue Feb 27, 2007 9:12 am
Location: Wisconsin

Post by jeffyscott » Sun Dec 06, 2009 5:57 pm

Les,

Scroll up to Robert's post from Thu Dec 03, 2009, 2:11 am
press on, regardless - John C. Bogle

User avatar
BlueEars
Posts: 3657
Joined: Sat Mar 10, 2007 12:15 am
Location: West Coast

Post by BlueEars » Sun Dec 06, 2009 6:23 pm

Jeffy thanks. It's interesting that GMO thinks Treasurys fair value is 3% real (not sure what maturity this is) and maybe 2.75% for TIPS. Sounds high by the standards of recent years.

the fixer
Posts: 75
Joined: Sat Jul 25, 2009 6:19 pm

Update:

Post by the fixer » Wed Dec 23, 2009 11:53 am

This conversation moved in another direction, so I just wanted to address a few comments made above that I disagree with:

Robert,

Here are the actual returns of the DFA EM and Vanguard EM funds from 1995 to date. The annualized return difference was 0.6%!. And arguably over time a more equal weighted country series (as in the DFA fund) will have a greater value and small-cap tilt relative to a simple country cap weighted series (as seems to be suggested by simple comparisons of the cap and equal weighted portfolios on Ken French’s website).
Its unlikely there is a major 'size' difference between Vanguard and DFA EM funds, given that DFA still cap weights stocks within each country. Even though some countries have less companies and smaller markets, the biggest still rise to the top. Morningstar calculates the average weighted market cap of the Vanguard EM Index at $17.6B and the DFA EM fund at $17.3B. No difference there.

Also, it does appear that more of that DFA index outperformance occured in the questionable early years from 88-93. From DFEMX inception in 94, it has only outpaced the MSCI Emerging Markets Index (gross div.) by +1.5% per year. DFEMX has actually underperformed the DFA Equally Weighted Emerging Markets Index over this entire period by 0.2% per year, but that has all occured this year and is likely to reverse (thru 12/08, DFEMX was +.2% per year ahead of the DFA index). Not a lot of value added above the DFA index no matter how you slice it, but thats not unexpected, as LG is the area of least value added from DFAs process.


Rod,
Aren't the Russell indexes known to be the worst indexes to use as the basis for an index fund, due to front running issues and maybe other factors?

If so, beating a Russell index by a point or two would not mean much
The Russell 2000 Index has had its share of problems in the last decade or so with its popularity (a risk to all indexes as they gain traction), but that has little to no impact on the Russell 3000 or 3000 Value. The Russell 2000 Value has felt some of it's core index pain, but that is easily rectifiable, as it has produced returns more in line with the S&P 600 Value despite a higher load.

Making that correction (adjusting the 2000 Value size/value loads down to 600 Value levels) puts it on a level playing field. Put another way, the biggest problem with Russell 2000 Value is it has size and value loads similar to DFA Small Value, with returns closer to the less tilted S&P 600 Value. Adjust your load assumptions, and you have fixed the problem.

Russell is beneficial to use because we have many decades of realistic data. As a matter of fact, I was under the impression that the 'go-to' mid value index was the Russell Mid Value, which in truth is no different from the 400 Value Index. Without looking, my guess is that the MSCI Mid Value index (not surprising, the newest of the bunch) has the best simulated returns?
It would seem more useful to compare to either real funds based by other index families, or to other indexes if lacking sufficient history for real funds.
We can do that, using a variety of different tilting approaches with various popular/common index fund or ETF strategies in out of sample (albiet short) timeframes to make comparisons. Last 12 months:

US Index Portfolios (with DFA US Vector-like size/value tilts)
(1) +27.1%
(2) +34.9%
(3) +34.8%
(4) +36.6%
(5) +29.3%
(6) +30.7%
Ave. (1) - (6) = +32.2%
DFVEX = +36.7%
"DFA Advantage" = +4.5%

International Index Portfolios (with DFA Int'l Vector-like size/value tilts)
(1) +42.1%
(2) +40.6%
(3) +39.5%
Ave. (1) - (3)= +40.7%
DFVQX = +48.6%
"DFA Adv" = +7.9%

Based on a simple 70/30 US/Foreign stock split, if you had the foresight to choose the single best ETF allocation (#4 and #1 respectively), you would have trailed a simple DFA Vector allocation by 2%, the midpoint between my 1% and 3% comment above.

Reasonable confirmation?

NOTE:
US Portfolio Allocations (ETF except where noted)
(1) 50% Russell 3000 Value/ 50% Bridgeway Ultra Small
(2) 12% Russell 3000/44% Russell Mid Value/44% Bridgeway Ultra Small
(3) 12% Russell 3000/44% S&P 400 Value/44% Bridgeway Ultra Small
(4) 12% Russell 3000/44% Vanguard Mid Value/44% Bridgeway Ultra Small
(5) 33% Russell 3000/67% Russell 2000 Value
(6) 33% Russell 3000/67% S&P 600 Value

Int'l Portfolio Allocations (all ETFs)
(1) 40% EAFE Value/60% EAFE Small Cap
(2) 12% EAFE/44% Wisdomtree Int'l Mid Cap Div./44% EAFE Small Cap
(3) 33% EAFE/67% Wisdomtree Int'l Small Cap Div.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Wed Dec 23, 2009 8:53 pm

.
the fixer,

On EM: The greater the weighting of a country in the Vanguard EM fund above the 14% (if I recall) EM country cap that DFA uses, the greater the relative small cap and value tilt of the DFA fund. Now I agree that this may not be much of an issue today as there are only two countries – China and Brazil – with greater than 14% country weights in the Vanguard fund (18.3 and 17.3% respectively). But in past years when there were less countries in the MSCI EM index this was likely a much more significant issue – and perhaps the source of the DFA EM fund 0.6% outperformance (which mostly occurred in the earlier years). For example, even in 2003 (according to M* Principia) the indication was for a greater value and small cap orientation. 2003 P/B: DFA EM = 1.1, Vanguard EM = 1.5, Av. Market Cap: DFA EM = 2.6 bn, Vanguard EM = 7.1 bn [if we believe the M* nos.].

On the ETF-DFA US comparison: From your comparisons:
- Portfolio 6 (with Vanguard MidCap Value and Bridgeway) return = 36.6%
- DFA US Vector = 36.7%
So not the 1 to 3+% out-performance above factor exposure (at least not for the last 12 months).

On the ETF-DFA Intl comparison: I don’t think it's possible to replicate a vector tilt with the EAFE Value and Small indexes – at least from my earlier factor tilt analysis. So IMO its not an alpha above factor exposure argument (the focus of the earlier posts) but an ‘available instrument’ argument. i.e. comparison not based on matching factors loads, but what’s possible to achieve with existing EAFE funds (at least my take), again a different argument IMO (and one I don't have much of an issue with).

In any case, I don't put much faith in one year results comparisons (but nevertheless interesting). If so, we would need to make similar outperformance inferences of the RAFI indexes.

Last 12 months returns (%) (according to M*):

P’Shares RAFI US...................49.53
P’Shares RAFI US Small..........69.63
P’Shares RAFI Intl..................43.24
P’Shares RAFI Intl. Small........56.90

[Bernstein’s earlier analysis suggested a size and value factor load on the RAFI US ‘index’ of about –0.10 and 0.35 respectively, and as I recall, Arnott’s estimates for RAFI US small were size and value loads of about 0.8 and 0.35 respectively. So a 50:50 RAFI US:US small combined loads of 0.30 and 0.35 respectively…]

Robert
.

bhzmark
Posts: 168
Joined: Mon Feb 02, 2009 8:48 pm

Re: Who's the fairest of them all?

Post by bhzmark » Thu Dec 24, 2009 10:30 am

Robert T wrote:.
The three asset allocation portfolios are:
  • (a) The DFA TM portfolio: 17% DFA TM Mktwide Value, 7% DFA TM Targeted Value, 13% DFA TM Small Cap, 17% DFA TM Intl. Value, 11% DFA Intl Large Co., 10% DFA EM, 25% DFA Intermediate Government.

For the DFA Portfolio consider adding, the DFA Core/vector funds with the recommended equity allocation of:
9% DFA US Core Equity 1 Portfolio (DFEOX)
36% DFA US Vector Equity Portfolio (DFVEX)
10% DFA Global Real Estate Securities (DFGEX)
35% DFA International Vector Equity Portfolio (DFVQX)
10% DFA Emerging Markets Core Equity (DFCEX)

It won't work to backtest it as these are too new, but would still be the best measure of the latest technology in DFA funds.
Last edited by bhzmark on Thu Dec 24, 2009 1:27 pm, edited 1 time in total.

User avatar
Hexdump
Posts: 1616
Joined: Fri Oct 10, 2008 7:28 am
Location: Houston, Texas

Lol, and here I thought

Post by Hexdump » Thu Dec 24, 2009 10:54 am

you were talking about our own Laura, based upon her recent, excellent Forbes articles.

Color me embarassed :oops:

the fixer
Posts: 75
Joined: Sat Jul 25, 2009 6:19 pm

Post by the fixer » Thu Dec 24, 2009 8:34 pm

Robert,

On EM:

Your theory doesn't stand up to the evidence. From 1995-2000, using DFA indexes, there was a -9.8% per year EM size premium (-1.3% vs. -11.1%). Over this stretch, the DFA EM Fund outperformed the MSCI EM Index by 4.6% per year. From 2001-2008, there was a +1.2% EM size premium (+9.7% vs. +10.9%), yet the DFA fund trailed the MSCI index by -0.5% per year.

On international factor loads:

I think a 40/60 EAFE Value & EAFE Small portfolio has a very close size and value load to DFA Int'l Vector -- at least dating back to 1999, when the MSCI EAFE Small Index (gross div.) was incepted. I show the combo with a 0.59 SmB coef, and a 0.04 HmL coef, while the DFA Int'l Vector index came in at 0.60 and 0.08. Because the Int'l Value premium is calculated using Large Value minus Large Growth, the HmL loadings for smaller cap funds and indexes are a bit skewed, but I still don't think they are that far off in reality. Through 11/09, here are the annualized returns for the two strategies:

EAFE Mix = +7.0%
EAFE Small = +8.0% (the highest return of the two MSCI indexes)
DFA Int'l Vector = +9.0%

On RFAI:

The returns have been very strong this year, boosted by March reconstitution which turned out to be perfect timing with hindsight. Over a longer period, I did not find that a combo of RFAI Int'l Large + Int'l Small Indexes outperformed the Int'l Vector index based on simulations from the Fundamental Index book, but it was closer than the EAFE Value/EAFE Small mix (only 0.8% DFA advantage through '07 if memory serves ?). Can't say much more than that...

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Thu Dec 24, 2009 11:56 pm

.
the fixer,

On EM: Largest outperformance of DFA EM vs. Vanguard EM was 1999 (71.7 vs 61.6).

1999 returns
DFA EM Value = 84.3
DFA EM Small = 85.4
DFA EM = 71.7
Vanguard EM = 61.6

On International Factor loads: The low HmL loads look a bit weird. What HmL series are you using?

On RAFI: Using the 1999-2007 data from Arnott’s book (the start date for the Intl Small series) gives the following result FWIW.

1999-2007.................Return/SD
RAFI Intl......................13.5/21.0
RAFI Intl Small.............17.1/22.4
RAFI 50:50 combo.......15.4/21.1
DFA Intl Vector............14.1/22.7

Robert
.

the fixer
Posts: 75
Joined: Sat Jul 25, 2009 6:19 pm

Post by the fixer » Sun Dec 27, 2009 3:05 pm

Robert T wrote:.
the fixer,

On EM: Largest outperformance of DFA EM vs. Vanguard EM was 1999 (71.7 vs 61.6).

1999 returns
DFA EM Value = 84.3
DFA EM Small = 85.4
DFA EM = 71.7
Vanguard EM = 61.6

On International Factor loads: The low HmL loads look a bit weird. What HmL series are you using?

On RAFI: Using the 1999-2007 data from Arnott’s book (the start date for the Intl Small series) gives the following result FWIW.

1999-2007.................Return/SD
RAFI Intl......................13.5/21.0
RAFI Intl Small.............17.1/22.4
RAFI 50:50 combo.......15.4/21.1
DFA Intl Vector............14.1/22.7

Robert
.
Robert,

On EM:

YTD Returns (thru 11/30):

DFA EM Large = +73.6%
DFA EM Small = +100.7%
MSCI EM Index (gross div.) = +80.9%

So, during an 11 month period with an EM size premium of over 25%, the DFA fund is lagging the MSCI index by over 7%. This, combined with my sub-period study of strong vs. non-existent size premiums in EM (showing that DFA EM underperformed MSCI when the size premium was positive, outperforming when the size premium was negative) casts considerable doubt (IMO) about a meaningful size tilt that has any impact on long run returns.

On Int'l factor loads:

Yes, loads are a bit lower because I used "Value minus Market" instead of "Value minus Growth" (just the easiest series for me to tabulate). So while the absolute numbers may differ, as long as we are making consistent comparisons, the relative results should be dependable

On RFAI:

My mistake, I was going from memory, and you correctly pointed out that the DFA Int'l Vector simulation has trailed the RFAI 50/50 simulation. It was actually a 50/50 split of DFA Int'l Large/Small Value indexes that outpaced the RFAI mix by 0.7% per year over this period (and the DFA Int'l Large Value index outpaced the RFAI Int'l Index by 1% or so going back to the early 80s (?) when its longer simulated time series was incepted). The live funds, however, trailed their index over this period, and was 0.2% behind the RFAI mix. This was an 'unlucky' stretch for the live funds, as moving the 9 year period 12 months forward (00-08 ) has the live funds ahead of their DFA indexes by as much as they trailed from 99-07.

My guess is that the RFAI Int'l funds are a bit more value tilted than the DFA Int'l Vector strategy, however. If we assume a 30% EAFE/ 70% RFAI (large/small) split from 99-07, we find a 1% advantage for Vector. If we assume a 20% EAFE/ 80% RFAI split, the DFA advantage falls to 0.3%.

Clearly, small differences in value exposure for non-US markets was very important over this stretch. Haven't studied the RFAI international strategies fully, but your research (including the individual country rebalancing benefits) seem to be spot on. Unfortunately, with a very small window of live trading data to measure, we cannot be sure about their real world 'investability', a consideration that should be made before abandoning the proven EAFE Value and EAFE Small iShares strategies.

Nothing more to add...

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Sun Dec 27, 2009 6:41 pm

.
The fixer,

Not a big deal on EM as we are only talking about a 0.6% difference. As indicated in my earlier post – I don’t expect significant size-value tilt differences between the DFA and Vanguard EM funds today, given that the largest regional weighting in the Vanguard funds are no too far off the 14% (if I recall) regional caps of the DFA fund. I still don’t think this was the case when there was a smaller number of countries in the MSCI index (consistent with the large 1999 return differences), but again, not a big deal.

A potential risk in not using value-minus-growth for international HmL is that what appears to be small absolute differences e.g. 0.04 vs. 0.08 value loads on the EAFE combo vs. DFA vector – may in fact turn out to be large relative differences (the difference between 0.04/0.08 = 0.04, between 0.08/0.16 = 0.08, between 0.16/0.32 = 0.16 etc…). Just a thought.

On the RAFI funds, I agree the backtests look okay (and I like their ‘zero’ alphas – according to Bernstein and Arnott). If we believe the factor loads on the Intl RAFI indexes are the same as the US RAFI funds then they should be close to a 0.35 value load. Not sure about Powershares implementation… Time will tell.

Robert
.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Fri Jan 01, 2010 8:10 am

.
Here are the 2009 returns for the portfolios listed in the OP:

Image

All the above portfolios have the same long-term expected average 75:25 stock:bond allocation

Asset allocation: The ETF and DFA portfolio returns were again relatively close reflecting their common factor loads, and both were slightly ahead of the Total Market Portfolio (as global midcap value and small caps outperformed global TSM, offsetting intermediate Treasury/Gov. underperformance of the US Aggregate bond index in the recovery).

MSCI ACWI.............................34.63%
MSCI ACWI Mid-cap Value.......42.67%
MSCI ACWI Small....................50.67%
BC 3-7yr Treasury...................-1.56%
BC Aggregate Bond Index........5.93%

Security Selection: The security selection portfolios did well, particularly the 'hedge-fund' trio (Cummings, Ackman, and Lampert portfolio) who's returns were double the asset allocation portfolios. There were no changes to the 6 stocks in the Cummings-Ackman-Lampert portfolio over the course of the year. The Fairholme fund also did fairly well. As did the Dodge & Cox Global fund which returned 49.2% in 2009 (compared to the above MSCI ACWI+ returns).

Market timing: The 16 month moving average portfolio lagged, being out of equities until September 2009. The inverted yield curve portfolio higher return reflects the lower bond allocation during 2009 (83:17 stock:bond at start of the year). The GMO portfolio did relatively well with only a 10 percent bond allocation in 2009 (a decline from about a 60% bond allocation in 2008).

Best for the New Year,

Robert
.

User avatar
Topic Author
Robert T
Posts: 2555
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Post by Robert T » Sun Aug 29, 2010 8:38 am


User avatar
jeffyscott
Posts: 7587
Joined: Tue Feb 27, 2007 9:12 am
Location: Wisconsin

Post by jeffyscott » Sun Aug 29, 2010 10:04 am

I wonder if that fund will close before the asset bloat gets to be a problem?

Looking forward to seeing how all these strategies have fared in 2010.
press on, regardless - John C. Bogle

Post Reply