DFA - Where's the alpha?

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: DFA - Where's the alpha?

Post by larryswedroe »

Epsilon
That kind of advisor no one should be working with--I'm referring only to an advisory who provides a fiduciary standard of care-an RIA, not a stock broker. There is simply no reason other than ignorance for working with the other kind (suitability standard)

Matjen
I have no idea where people get this 1-3% figure quoted (like the 1% annual fee when there are many with much lower fees if all you want is access, or minimal help) . The higher end is only true where you are looking at an individual asset class where Vanguard doesn't have a fund like EMV or EMS, where the premiums are large. And no one I know holds a large percentage of their portfolio there. In US the biggest outperformance is in SV where you have last 15 years difference for DFA SV and Vanguard SV of about 2%, and it other asset classes gaps are much smaller. So how much outperformance you can get from the higher loadings to the factors depends heavily on how heavily you load yourself. How much you are willing to tilt. Tilt highly and you can gain significant premiums---- and that allows you to hold less equity. Finally, when looking at the extra expense it should be the expense per unit of expected return----not just expense. And by allowing you to hold less equity you can cut portfolio costs elsewhere by avoiding manager fee on the bond side which now has higher allocation. So want to look at total cost impact.

Best wishes
Larry
User avatar
DueDiligence
Posts: 101
Joined: Mon Sep 09, 2013 3:57 pm

Re: DFA - Where's the alpha?

Post by DueDiligence »

FWIW here is a historical performance comparison of "modified ultimate buy and hold" portfolios using DFA compared with Vanguard funds (plus DLS). The comparison is consistent with the conclusion of others in the thread - differences between DFA and Vanguard portfolios are small and likely within the uncertainty. If anything the DFA portfolio shows an average advantage of 10-20 bps for ~70% equity portfolios.

Image
Results are shown for 5, 10, 15, and 20 years ending June 30, 2013. The solid lines are Vanguard and dashed lines DFA portfolios.

All portfolios use Vanguard TBM VBMFX with 70% of the equities being US. TBM varies between 0 and 100% to capture the range of standard deviations. The US equities are equal amounts of LB, LV, SB, and SV. International equities are equal amounts of ILB, ILV, ISB, and ISV. Yahoo was used for historical prices. The Simba dataset approach was used for funds and prices earlier than fund inception dates. DISVX was used for ISV performance prior to DLS inception.
"Personal preferences, circumstances, and abilities affect portfolio construction in a profound manner..." David Swensen
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: DFA - Where's the alpha?

Post by larryswedroe »

Due diligence
keep in mind that for the large blends the portfolios are the same basically, that's a commodity, a pure index. So the bigger the percentage you have of those the smaller the differences will be.
The more you tilt a portfolio to small value and then include allocations to assets like EMV (instead of EM) the bigger the advantage. And of course on the bond side the funds are likely to be similar. So you're talking 50% or more of the fund allocations are the same--hard to come up with much difference in those cases.
vesalius
Posts: 794
Joined: Tue Jul 13, 2010 7:00 pm
Location: Texas

Re: DFA - Where's the alpha?

Post by vesalius »

larryswedroe wrote:Due diligence
keep in mind that for the large blends the portfolios are the same basically, that's a commodity, a pure index. So the bigger the percentage you have of those the smaller the differences will be.
The more you tilt a portfolio to small value and then include allocations to assets like EMV (instead of EM) the bigger the advantage. And of course on the bond side the funds are likely to be similar. So you're talking 50% or more of the fund allocations are the same--hard to come up with much difference in those cases.
Larry

While I am completely on board with extreme tilts to SV, EMV and International SV in my own portfolio, I think it reasonable to make the comparison Due Diligence highlighted because that is the type of portfolio many people have and that is what they are comfortable sticking with going forward. Portfolios with 50% or so bonds and at best slight tilts to SV and EMV. With those constraints (and taking your oft given advice of looking at the portfolio in total and not each investment in isolation) it seems we all agree that the differences will be minor at best.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: DFA - Where's the alpha?

Post by larryswedroe »

vesalius
I don't disagree --just pointing out the advantages of funds like those run by DFA/Bridgeway don't come from market like portfolios/asset classes because they are commodities if you will. All S&P and EAFE funds will be very similar. So no reason to expect big differences in returns. Similarly with large value where the value premium is relatively small. But the differences can be much larger in small and especially small value, and then of course in asset classes where there may not be a good alternative.
Larry
Jebediah
Posts: 805
Joined: Tue Aug 28, 2012 9:19 pm
Location: Austin TX

Re: DFA - Where's the alpha?

Post by Jebediah »

Bradley wrote:I don't believe anyone has disputed the benefits of DFA’s core funds only the idea that they have significant advantage or certain advantage over separate non core holdings which are rebalanced according to an individual IPS.
I'm willing to say, and I think this thread has concluded that DFA's core funds (on the domestic side) provide only one benefit over a similarly loaded Vanguard or ETF portfolio: FF factor loadings that are more stable. I'm not sure I'd consider this a tangible benefit. I don't know how to quantify this feature.

But besides that, there are no benefits to DFA's core funds. As several of us have shown, the core funds have identical returns to the competing portfolios, meaning that all of the nice sounding features only amount to covering the excess management fees.

Regarding taxes, DFA Core 2 Tax Managed, since inception, has had the same tax cost as the similarly-loaded Vanguard SCV/TSM combo-- so no benefit there.

There is an unlikely, but possible benefit to not having to rebalance/pay K gains on the your solitary DFA Core fund in a year when you would have to rebalance SCV vs TSM. But this seems remote as most years SCV and TSM will not diverge by much and when you do rebalance them, it's rebalancing them both together vs. bonds, thus no different than what the DFA Core holder would have to do.
User avatar
Topic Author
Blue
Posts: 1178
Joined: Sat Jul 12, 2008 10:18 pm

Re: DFA - Where's the alpha?

Post by Blue »

Jebediah wrote:I'm willing to say, and I think this thread has concluded that DFA's core funds (on the domestic side) provide only one benefit over a similarly loaded Vanguard or ETF portfolio: FF factor loadings that are more stable. I'm not sure I'd consider this a tangible benefit. I don't know how to quantify this feature.
Agree.

Robert T's approach to quantifying impact of FF factor loading stability seems like a reasonable approach to me.
Lower factor loading drift. Assumes a 10% drift in value and size loadings of a non-DFA index fund portfolio relative to a DFA fund portfolio. This translates into actual portfolio loadings for size and value for a non-DFA fund portfolio to be 0.18 and 0.36, instead of 0.2 and 0.4 due to downside style drift. The calculation is based on an expected size and value premium of 2 and 4 respectively. The DFA style consistency advantages was estimated as 0.75* ((0.2-0.18)*2 + (0.4-0.36*4)) = 0.15. Just an assumptions - other can be used if more appropriate.
Extending this thought a bit, ERP over the long-term can somewhat be forecasted by Shiller PE 10, the Gordon equation, etc. Are there analagous measures for forecasting Small-Value premium.... are there periods of time ex ante where an investor should "tune-up" their factor drift?
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: DFA - Where's the alpha?

Post by Robert T »

.
Just did some analysis to see what alpha I would have gained beyond factor exposure by using the DFA tax-managed funds instead of available alternatives on US side (ishares, Bridgeway) [for similar tax efficiency]. Analysis is from start of 2003 when I set up my portfolio. Few observations:
  • 1. There would have been no additional alpha beyond factor exposure. If anything, the alphas were more negative on the DFA TM funds over this period.

    2. The betas on the DFA funds are all larger than 1 likely due to the exclusion of REITs and utilities which have betas significantly below 1. Higher return from more beta exposure is from higher risk. In this respect, targeting a beta exposure of 1 in a DFA portfolio requires slightly more bonds than in an ishare/Bridgeway portfolio.

    3. Interestingly the size loads of both the DFA and ishares funds are larger than longer term averages, and the value loads are smaller. This may be a function of the magnitude of the size and value premiums. Over this periods the size premium was higher than the historical average (0.37% vs 0.24% per month), and the value premium was smaller than the historical average (0.15% vs. 0.40% per month).

    4. How do you target a specific long-term portfolio size and value load if these loads vary over time? My approach is to use a time period for the portfolio factor load estimates where the size and value premiums are similar in magnitude to the long-term historical average/or expected factor premiums. This is one advantage of the vector/core type funds – that as their factor loads seem to vary less over time, its perhaps easier to target a specific portfolio factor load (lower risk of errors). Obviously no guarantee.

    Image
Robert
.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: DFA - Where's the alpha?

Post by larryswedroe »

Robert
Keep in mind that during the period you studied DFA used momentum screens to lower the negative exposure to MOM---that is going to lower the value loading --that happened in 2003 if memory serves.
That would lower the negative loading on MOM--if memory serves was typically about -0.4 to much closer to 0 now (think slightly negative)
Larry
caklim00
Posts: 2420
Joined: Mon May 26, 2008 10:09 am

Re: DFA - Where's the alpha?

Post by caklim00 »

I asked a few weeks ago about
DFFVX US Targeted Value ER .38 (.02 extra fee for my old 401k makes it .40 for me) vs VSIAX Vanguard SCV Admiral ER .10
and got no responses. DFFVX and VSIAX should have similar loads.
Jebediah
Posts: 805
Joined: Tue Aug 28, 2012 9:19 pm
Location: Austin TX

Re: DFA - Where's the alpha?

Post by Jebediah »

larryswedroe wrote:Robert
Keep in mind that during the period you studied DFA used momentum screens to lower the negative exposure to MOM---that is going to lower the value loading --that happened in 2003 if memory serves.
That would lower the negative loading on MOM--if memory serves was typically about -0.4 to much closer to 0 now (think slightly negative)
Larry
I don't see evidence of this. The momentum loadings for all the aforementioned DFA funds were, if anything, slightly better from 2003-2008 than they have been since. Here is the MOM data for Robert's chosen funds/time period (Jan 2003 - Sept 2013):

Code: Select all

Fund                           MOM factor (1/2003 - 9/2013)

DFA TM Marketwide Value       -0.06
DFA TM Targeted Value         -0.02
DFA TM Small Cap               0.02
DFA US Micro                  -0.02

iShares Russell MidCap Value  -0.05
iShares S&P 600 Value         -0.03
Bridgeway Ultra-small Co Mkt.  0.01
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: DFA - Where's the alpha?

Post by Robert T »

caklim00 wrote:I asked a few weeks ago about
DFFVX US Targeted Value ER .38 (.02 extra fee for my old 401k makes it .40 for me) vs VSIAX Vanguard SCV Admiral ER .10
and got no responses. DFFVX and VSIAX should have similar loads.
Similar value loads but different size loads. Here are the factor loads from July 2001 - September 2013 the time period for which there is available data for the CRSP Small Cap Value Index, the index tracked by the Vanguard Small Cap Value fund.

CRSP Small Value Index
Alpha.......0.13 (1.57)
Market......0.96
Size.........0.57
Value.......0.48
R^2.........0.97

DFA US Targeted Value
Alpha.......0.02 (0.29)
Market......1.06
Size.........0.79
Value.......0.51
R^2.........0.98

(t-stat in parenthesis).

So no alpha on the DFA US Targeted Value fund, and the alpha on the CRSP Small Value index is probably random drift rather than anything to rely on. Main difference is in the factor loads. DFA US Targeted Value has higher market and size loads.

Robert
.
afan
Posts: 8169
Joined: Sun Jul 25, 2010 4:01 pm

Re: DFA - Where's the alpha?

Post by afan »

Has anyone looked at the stability of the factor loads, rather than reporting just the estimates at one point in time? What is the variance in each factor load for a given fund, and are the differences in loads between two funds significant? Anyone know of published research, or working papers on the subject?

Thanks
We don't know how to beat the market on a risk-adjusted basis, and we don't know anyone that does know either | --Swedroe | We assume that markets are efficient, that prices are right | --Fama
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: DFA - Where's the alpha?

Post by larryswedroe »

Jeb
My memory may be faulty --your current data shows what I expected, close to zero MOM loading (which is good), but before screening if my memory serves it was significantly negative--like -0.4 (but like I said my memory could be wrong)
Larry
Jebediah
Posts: 805
Joined: Tue Aug 28, 2012 9:19 pm
Location: Austin TX

Re: DFA - Where's the alpha?

Post by Jebediah »

Close to zero is good. But DFA doesn't seem to be outdoing anyone else in this regard.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: DFA - Where's the alpha?

Post by larryswedroe »

Jeb
Here is the issue--the deeper the value the more negative the MOM loading unless you screen for it. Whether screening helps or not will then depend on the MOM premium, and remember if they own stocks with negative MOM they would not sell them, they just would not buy more. And they don't go long stocks with positive MOM either--just might be more patient selling them than others. In other words they might have 50 stocks trying to patiently sell at any one point. Patient trader doesn't care which of the 50 they sell today, so using algo type trading they feed small lots into the system---with a preference for the ones with least positive MOM. That is what AQR does for example.
Note it's a real balance as if you screen for MOM you lower value loading, so attempt to find the right balance.
Hope that helps
Larry
berntson
Posts: 1366
Joined: Mon Oct 29, 2012 12:10 pm

Re: DFA - Where's the alpha?

Post by berntson »

Jebediah wrote:Close to zero is good. But DFA doesn't seem to be outdoing anyone else in this regard.
Right. DFA momentum screens provide no advantage over and above allowing DFA funds to have higher value loadings without also generating more negative momentum. The loadings are all that matters.

There is no magic DFA pixie dust. But Bogleheads don't believe in pixie as a general rule, so this is really just telling us what we already knew. :happy
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: DFA - Where's the alpha?

Post by larryswedroe »

Bernston, that's correct. The MOM screens simply shift the loadings to be more favorable. And it's basically almost all about loadings--FACTOR INVESTING, not asset class investing. You can have two funds from the same asset class with very different loadings and thus very different risks and expected returns. BTW-it's why I'm now invested in Bridgeway's Omni SV funds
Larry
Jebediah
Posts: 805
Joined: Tue Aug 28, 2012 9:19 pm
Location: Austin TX

Re: DFA - Where's the alpha?

Post by Jebediah »

berntson wrote:
Jebediah wrote:Close to zero is good. But DFA doesn't seem to be outdoing anyone else in this regard.
Right. DFA momentum screens provide no advantage over and above allowing DFA funds to have higher value loadings without also generating more negative momentum.
Near as I can tell, DFA funds do not have higher value loadings.
berntson
Posts: 1366
Joined: Mon Oct 29, 2012 12:10 pm

Re: DFA - Where's the alpha?

Post by berntson »

Jebediah wrote:
berntson wrote:
Jebediah wrote:Close to zero is good. But DFA doesn't seem to be outdoing anyone else in this regard.
Right. DFA momentum screens provide no advantage over and above allowing DFA funds to have higher value loadings without also generating more negative momentum.
Near as I can tell, DFA funds do not have higher value loadings.
Actually, good point. There seems to be no particular advantage to using DFA funds for domestic value. And Larry's Bridgeway OMNI funds don't have a long enough track record to tell whether or not they have stronger value loadings. More reason to think that there is little advantage to "exclusive" funds. The only exception being, perhaps, emerging market value funds. If PowerShares ever gets around to fixing the PXH tracking issues, though, we'll have a proper alternative to DFEMX.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: DFA - Where's the alpha?

Post by larryswedroe »

jeb and Bernston

given issues with loadings (long short portfolios and how momentum impacts value) I suggest that you look at the actual valuation metrics as a better judge of exposures. So look at things like market caps, p/b,p/s,p/cf, p/e.
Note you don't need live data with Bridgeway vs DFA, the construction rules tell you that you will be both smaller and more valuey. And that shows up looking at current metrics. Just check M* data and it will be obvious.

Larry
Jebediah
Posts: 805
Joined: Tue Aug 28, 2012 9:19 pm
Location: Austin TX

Re: DFA - Where's the alpha?

Post by Jebediah »

DFA's funds do tend to be 'cheaper' (lower p/e, p/b, etc) than the alternatives. One thing I notice when I look at the current valuation metrics is that small caps look a lot more expensive than large caps. Is this often or usually the case?

Another question- Hope this isn't getting too off topic, but when is a value fund considered cheap or expensive? Are they always cheap because they hold cheap stocks and thus always have higher expected returns than TSM? Or should one think of value funds relative to their own average-- in which case, at times value is expensive and growth is cheap (even though value has the lower P/E)?

For example, say we look at DFA Vector Equity (DFVEX) and see that it has a P/E of 13.9. Do we say, wow what a bargain because the historical mean PE of the market is 15 or do we say no, it's expensive because value funds usually have a P/E of around 11 (I just made that number up) ?
Last edited by Jebediah on Mon Nov 04, 2013 2:49 pm, edited 1 time in total.
User avatar
JoMoney
Posts: 16260
Joined: Tue Jul 23, 2013 5:31 am

Re: DFA - Where's the alpha?

Post by JoMoney »

Jebediah wrote:...
Another question- Hope this isn't getting too off topic, but when is a value fund considered cheap or expensive? Are they always cheap because they hold cheap stocks and thus always have higher expected returns than TSM? Or should one think of value funds relative to their own average-- in which case, at times value is expensive and growth is cheap (even though value has the lower P/E)?
I've been trying to bring up more discussion on that... http://www.bogleheads.org/forum/viewtop ... 0#p1826558

It's seems strange that during recent years, the average "Small Value" fund has a higher P/E then the average "Large Growth" fund.
Vanguard Mega-Cap Growth P/E: 20.8x (09/30/2013 )
Vanguard Small-Cap Value P/E: 21.9x (09/30/2013 )

And somehow I doubt the higher P/E in the SV funds is due to a higher expectation of growing earnings.
"To achieve satisfactory investment results is easier than most people realize; to achieve superior results is harder than it looks." - Benjamin Graham
berntson
Posts: 1366
Joined: Mon Oct 29, 2012 12:10 pm

Re: DFA - Where's the alpha?

Post by berntson »

larryswedroe wrote:jeb and Bernston

given issues with loadings (long short portfolios and how momentum impacts value) I suggest that you look at the actual valuation metrics as a better judge of exposures. So look at things like market caps, p/b,p/s,p/cf, p/e.
Note you don't need live data with Bridgeway vs DFA, the construction rules tell you that you will be both smaller and more valuey. And that shows up looking at current metrics. Just check M* data and it will be obvious.

Larry
(1) The connection between things like average market caps, p/b,p/s,p/cf, p/e and Fama French loadings can be unpredictable. BRSVX should have stronger small and value loadings than IJS if one is just looking at the fundamentals. But if we do FF regressions going back to the beginning of BRSVX, this is not what we find. IJS had smb=.85 and hml=.33, whereas BRSVX had only smb=.76 and hml=.04.

(2) The OMNI funds have to prove that they not only have stronger small and value loadings, but also that they can do so without generating significant negative alpha. Given the performance of Bridgeway's other funds, I wouldn't take their word for it. BRSIX has generated 1.2% yearly negative alpha over the lifetime of the fund, for example.
User avatar
JoMoney
Posts: 16260
Joined: Tue Jul 23, 2013 5:31 am

Re: DFA - Where's the alpha?

Post by JoMoney »

If the average S/V funds have fundamentals rising faster than the market average, eventually it gets to the point where the only way you can get lower fundamental values is to concentrate in the true dogs that even other SV funds wouldn't be willing to hold.
"To achieve satisfactory investment results is easier than most people realize; to achieve superior results is harder than it looks." - Benjamin Graham
User avatar
DueDiligence
Posts: 101
Joined: Mon Sep 09, 2013 3:57 pm

Re: DFA - Where's the alpha?

Post by DueDiligence »

berntson wrote:
larryswedroe wrote:jeb and Bernston

given issues with loadings (long short portfolios and how momentum impacts value) I suggest that you look at the actual valuation metrics as a better judge of exposures. So look at things like market caps, p/b,p/s,p/cf, p/e.
Note you don't need live data with Bridgeway vs DFA, the construction rules tell you that you will be both smaller and more valuey. And that shows up looking at current metrics. Just check M* data and it will be obvious.

Larry
(1) The connection between things like average market caps, p/b,p/s,p/cf, p/e and Fama French loadings can be unpredictable. BRSVX should have stronger small and value loadings than IJS if one is just looking at the fundamentals. But if we do FF regressions going back to the beginning of BRSVX, this is not what we find. IJS had smb=.85 and hml=.33, whereas BRSVX had only smb=.76 and hml=.04.
How does one know which estimate is better If fundamentals and FF regressions do not agree?
"Personal preferences, circumstances, and abilities affect portfolio construction in a profound manner..." David Swensen
User avatar
Topic Author
Blue
Posts: 1178
Joined: Sat Jul 12, 2008 10:18 pm

Re: DFA - Where's the alpha?

Post by Blue »

JoMoney wrote:
Jebediah wrote:...
Another question- Hope this isn't getting too off topic, but when is a value fund considered cheap or expensive? Are they always cheap because they hold cheap stocks and thus always have higher expected returns than TSM? Or should one think of value funds relative to their own average-- in which case, at times value is expensive and growth is cheap (even though value has the lower P/E)?
I've been trying to bring up more discussion on that... http://www.bogleheads.org/forum/viewtop ... 0#p1826558

It's seems strange that during recent years, the average "Small Value" fund has a higher P/E then the average "Large Growth" fund.
Vanguard Mega-Cap Growth P/E: 20.8x (09/30/2013 )
Vanguard Small-Cap Value P/E: 21.9x (09/30/2013 )

And somehow I doubt the higher P/E in the SV funds is due to a higher expectation of growing earnings.
Larry had an interesting article from 2007 discussing Jim Davis's paper "Does Predicting the Value Premium Earn Abnormal Returns?" I couldn't find the original paper online or on ssrn and am curious if Larry's thoughts have been enhanced or nuanced since the 2007 article?
......since studies have found that when the spread in book-to-market (BtM) ratios between value stocks and growth stocks is high, the subsequent value premium tends to be high. The reverse is also true. Based on that information, if next year’s value premium is expected to be high, it would seem logical to own value stocks. If it were expected to be low, then growth stocks would seem to become the logical choice. Is it really that simple to earn abnormal returns? Does a statistical relation always translate into a viable portfolio strategy? These are the questions Jim Davis asked and answered in his study “Does Predicting the Value Premium Earn Abnormal Returns?”2 The study covered the period July 1927–June 2005. Davis found that style-timing rules did not generate high average returns despite being able to use future information about BtM spreads. In fact, he concluded that that the expected excess return of style timing is probably negative—for the same reasons that efforts to try to time the overall market are likely to fail.
berntson
Posts: 1366
Joined: Mon Oct 29, 2012 12:10 pm

Re: DFA - Where's the alpha?

Post by berntson »

DueDiligence wrote: How does one know which estimate is better If fundamentals and FF regressions do not agree?
For me, I just have no idea how to estimate future returns by looking at fundamentals, especially when the fund in question is not just buying all the stocks available in a particular index. With the FF factors, we have a corresponding portfolios with historical returns. If we plug in the historic returns for those portfolios and the factor loadings for a fund, that at least gives us some kind of estimate of future performance. On the other hand, if just hand me the fundamentals for OMNI funds, with there various screens and trading schemes, I have no idea what sort of historical portfolio I should compare them with.

Regressions also help to answer the question how much value is the manager destroying? How much negative alpha is a fund generating? In the case of many Bridgeway funds, the answer is: quite a lot. I don't know how to answer this question by looking at fundamentals.
Jebediah
Posts: 805
Joined: Tue Aug 28, 2012 9:19 pm
Location: Austin TX

Re: DFA - Where's the alpha?

Post by Jebediah »

True about alpha and momentum, but aren't today's fundamentals telling you indirectly what the fund's size-adjusted value loading is right now. Not numerically, but at least relatively?
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: DFA - Where's the alpha?

Post by larryswedroe »

Bernston
First When a fund is relatively small/new it will have lots of random tracking error. And funds like DFA and Bridgeway that are patient traders and use screens and MOM filters and algo type trading will have more random error than others. So I would not draw the conclusions you have drawn.
Second, the fund construction rules of Bridgeway's Omni funds are very transparent--simply weight the four factors (not equally though). Note that gives them exposure to the Profitability factor that a simple BtM screen would not provide.
Third, the actual current metrics tell you what a fund is more likely to do than any historical loading on a factor. Simple-smaller market cap and deeper value metrics predict higher returns. And a perfect example of that is the RELATIVE performance this year. Exactly what you would expect based on the purity hypothesis.

Jeb
Value stocks are always relatively cheaper, thus they have higher expected returns. But that is relative. The size of the premium depends as noted above on the spread between value and growth metrics. Another simple intuitive answer. The wider the spread, the bigger the expected premium and vice versa. Having said that both valuations can be expensive relative to historical averages of course.

Hope that helps
Larry
berntson
Posts: 1366
Joined: Mon Oct 29, 2012 12:10 pm

Re: DFA - Where's the alpha?

Post by berntson »

Jebediah wrote: Not numerically, but at least relatively?
Relatively, perhaps. But relative loadings will in most cases not answer the questions that an investor needs to answer. Are the higher loadings of fund A worth the extra expense and tracking error compared to fund B? Will most of the extra returns from factor tilts be destroyed by negative alpha?

Fundamentals indicate the microcap funds have a stronger small tilt than VB. Regressions indicate that the extra returns are almost always destroyed by negative alpha (from fees, front-running in the case of index funds, spreads, and who knows what else).
larryswedroe wrote: When a fund is relatively small/new it will have lots of random tracking error. And funds like DFA and Bridgeway that are patient traders and use screens and MOM filters and algo type trading will have more random error than others.
We have more than ten years of data for BRSIX and it has underperform its factor loadings by 1.2% annually. This isn't too surprising, given the fund has an expense ratio of .75. This helps to explain why garden-variety small value index funds (VB for instance) beat BRSIX (a microcap fund) over the last decade, a decade where we have had a strong small premium. (And I'm just going to ignore the 1.45 average tax ratio for BRSIX over last five years.)

Cost matters, and Bridgeway has not shown that it can deliver funds with factor tilts that justify its higher costs. Why should we think it will be different this time?
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: DFA - Where's the alpha?

Post by larryswedroe »

Bernston, you comparing a micro cap fund to a small value fund?
And with a retail fund like BRSIX they had large market impact costs due to investor withdrawals during bear markets which is why we would not invest in such funds, for that risk. The fund we use doesn't allow retail money so doesn't have that risk. And the fund we use also incorporates MOM screens. And we like the multiple value screens which actually give you more exposure to the profitability factor.
Again, check returns of the OMNI funds this year relative to DFA and Vanguard and you get exactly what you would expect to see given size and value premiums this year and metrics of the funds.

And the purity hypothesis does a great job of explaining the differences in returns. Example I looked at every year from 2009-2013 for DFA vs Vanguard SV and found the relative returns to be exactly what the metrics said they should be.

Larry
User avatar
Topic Author
Blue
Posts: 1178
Joined: Sat Jul 12, 2008 10:18 pm

Re: DFA - Where's the alpha?

Post by Blue »

larryswedroe wrote:Bernston, you comparing a micro cap fund to a small value fund?
And with a retail fund like BRSIX they had large market impact costs due to investor withdrawals during bear markets which is why we would not invest in such funds, for that risk. The fund we use doesn't allow retail money so doesn't have that risk. And the fund we use also incorporates MOM screens. And we like the multiple value screens which actually give you more exposure to the profitability factor.
Again, check returns of the OMNI funds this year relative to DFA and Vanguard and you get exactly what you would expect to see given size and value premiums this year and metrics of the funds.

And the purity hypothesis does a great job of explaining the differences in returns. Example I looked at every year from 2009-2013 for DFA vs Vanguard SV and found the relative returns to be exactly what the metrics said they should be.

Larry
Larry - It looks like on Bridgeway's website this fund (and the tax managed version) is only available through your advisory firm BAM? Is it safe to assume you had an integral part in its development/criterion rules for the fund? Curious to learn more -

link to Bridgeway Omni Funds
*These Funds are available to approved advisors through BAM Advisory Services. Please contact Bridgeway's Marketing and Client Service Team for more information. 713-661-3500, Option 1
User avatar
Topic Author
Blue
Posts: 1178
Joined: Sat Jul 12, 2008 10:18 pm

Re: DFA - Where's the alpha?

Post by Blue »

Love the google.... July 2012 WSJ Article

Bridgeway, which focuses mostly on tiny stocks and a quantitative—that is, data-driven—strategy, has retooled its computer models, merged funds, and recently launched two new funds in partnership with Buckingham Asset Management LLC in St. Louis ....

.......The two funds, Omni Small-Cap Value and Omni Tax-Managed Small-Cap Value, are managed by Bridgeway and open only to clients of approved investment advisers, such as the 130 firms in Buckingham's BAM Advisor Services network, which including Buckingham has more than $15 billion under management. Buckingham, which has some $3.5 billion under management itself, was looking for an ultrasmall-company fund for its network and liked Bridgeway's data-driven approach and low costs, says Larry Swedroe, director of research for Buckingham and BAM Advisor Services. One fund was launched at year-end 2010 and the other last August.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: DFA - Where's the alpha?

Post by larryswedroe »

Blue yes and we were heavily involved in the design--wanting to make sure it was deeper value and smaller cap and included MOM screening. And then balancing that with enough diversification (we like to see say 700 stocks) This is a part of the market with definite capacity constraints, perhaps $5b can be run effectively without running into capacity problems. Hence the limited access. With now $21 billion in AUM in the BAM community we have enough size to have fund family interested in designing a fund for our exclusive use as we can bring enough assets to the table. And we also don't have to then worry about hot money.
Bridgeway has a very strong research team especially with the recent addition of Andrew Berkin (a name those who read the financial literature will recognize).

Hope that helps

Larry
berntson
Posts: 1366
Joined: Mon Oct 29, 2012 12:10 pm

Re: DFA - Where's the alpha?

Post by berntson »

larryswedroe wrote: Again, check returns of the OMNI funds this year relative to DFA and Vanguard and you get exactly what you would expect to see given size and value premiums this year and metrics of the funds.
One year returns does not a statically significant sample make. :D

The omni funds may very well be fine funds. Investors should just be cautious: Bridgeway funds have higher expenses than garden variety index funds and Bridgeway has a history of translating those fees into negative alpha.
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: DFA - Where's the alpha?

Post by Robert T »

.
Good to see Bridgeway increasing its research capacity and good to see advisors taking the initiative to develop what they feel are the best products for their clients
http://bridgeway_web.s3.amazonaws.com/cms/documents/e7d2ff116dbd01f5/2013.09.20_Berkin_Announcement_FINAL_v1.pdf

Personally I think Bridgeway are one of the better fund management companies, and have done a relatively good job at its 'asset class' funds. For example, since inception of the Bridgeway Blue Chip 35 fund, its alpha has been zero i.e. delivering pure factor exposure.

Bridgeway Blue Chip 35: Aug. 1997- Dec 2012

Alpha............0.00
Market.........+0.98
Size.............-0.23
Value...........-0.05
Profitability...+0.23 (t-stat 6.45)

The alpha on the Bridgeway Ultra-small Co. Market will likely vary over time in sync with the (non) linearity of CRSP10 performance relative to the market. For example here are the 5-year decile returns through June 30, 2013

CRSP declies
1....6.62%
2....8.49%
3....8.74%
4...10.84%
5...11.50%
6...10.73%
7...13.20%
8...13.74%
9...13.03%
10..12.93%

The non-linearity in decile 10 will likely show up as negative alpha as SmB and HmL are more dominated by decile 7 and 8 than decile 10. So the expectation should be for a more variable alpha on micro-cap funds, by the nature of the variability of CRSP10 relative to other deciles. If we believe that the long-term performance of CRSP 10 is similar to its performance over the last 87.5 years (as below), then the long-term alpha should be positive (due to the nonlinearity).

Past 87.5 years through June 2013

CRSP declies
1....9.16%
2...10.50%
3...10.89%
4...10.91%
5...11.55%
6...11.37%
7...11.62%
8...11.64%
9...11.60%
10..13.28%

Over the life of the Bridgeway Ultra-small company market fund (BRSIX), through September 2013, annualized returns = 11.34% compared to CRSP10 returns of 11.84% over the same time period, so a 0.5% difference. While there have been huge - eye-brow raising tracking error - as in 2009, over the long-term Bridgeway has got fairly close to CRSP10.

If history is anything to go by - the trade-off in BRSIX (CRSP10) is more variable short-term alpha (sometimes significantly negative), and potentially long periods of underperformance of CRSP10, but expectations of a longer-term positive alpha simply due to the non-linearity in CRSP10. Obviously no guarantees.

Robert
.
Last edited by Robert T on Wed Nov 06, 2013 8:51 am, edited 1 time in total.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: DFA - Where's the alpha?

Post by larryswedroe »

Bernston
Of course one year doesn't tell you that much. I showed it as perfect example of the purity hypothesis---the more exposure to the factors the better or worse the fund does depending on how the asset class does. And it works virtually all the time.
And Bridgeway has done a excellent job on it's microcap fund getting returns of the specific asset class it's in.
Yes it's funds are bit more expensive even than DFA, but that is function partly of the size of the fund (DFA has lowered fees over the years as their AUM rises), but this is good example of not focusing on ER alone but the ER relative to the value add--the bit higher fee is offset we believe by significantly more exposure to the size/value premium in the sector of the market where the value premium is largest---and the multiple value screens and MOM screens.
Best wishes
Larry
tryingmybest
Posts: 49
Joined: Tue Nov 04, 2008 11:23 am

Re: DFA - Where's the alpha?

Post by tryingmybest »

The back and forth is very educational.

Robert T from http://www.bridgeway.com/strategies/sma ... rformance/ I see since inception the annualized performance of Bridgeway Ultra-small Company Market (BRSIX), is 12.64% vs. CRSP 10 which is 14.84%. Am I looking at the wrong figures?
Bradley
Posts: 486
Joined: Tue May 01, 2007 2:41 pm

Re: DFA - Where's the alpha?

Post by Bradley »

Those are the correct figure through 9/30/2013. Bridgeway has listed two different inception dates for the two different references?
Last edited by Bradley on Wed Nov 06, 2013 2:11 pm, edited 1 time in total.
You can sum up any active fund manager’s presentation at an investor conference in one sentence: “We’re doing well, all things considered.”
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: DFA - Where's the alpha?

Post by Robert T »

.
Here's the link
http://www.bridgeway.com/strategies/mut ... rformance/

Since inception 7/31/1997
BRSIX = 11.34%
CRSP10 = 11.84%

Your earlier link was from institutional investor site - inception 1/1/01
.
berntson
Posts: 1366
Joined: Mon Oct 29, 2012 12:10 pm

Re: DFA - Where's the alpha?

Post by berntson »

Robert T wrote: The alpha on the Bridgeway Ultra-small Co. Market will likely vary over time in sync with the (non) linearity of CRSP10 performance relative to the market. For example, here are the 5-year decile returns through June 30, 2013.
This is actually something I've been trying to figure out. There must be a white paper somewhere that works out the relation between the non-linearity of CRSP10 and the FF smb portfolio? Here are two hypotheses: (1) CRSP10 performance is non-linear because its smb loadings are non-linear. CRSP10 companies have smb loadings that are not only higher than CRSP8 and CRSP9, but non-linear. (2) CRSP10 performance is non-linear because the smallest companies are capturing returns from some other factor. I have been working on the assumption that (1) is the right hypotheses.

If (1) is right, then there is no particular reason to expect CRSP10 underperformance relative to CRSP8 and CRSP9 to show up as negative alpha. It could show up as smb drift. The question is: To what extent are the results for CRSP10 correlated with the smb portfolio when CRSP10 underperforms (or overperforms) relative to CRSP8 and CRSP9? I have no idea. Maybe someone else does?
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: DFA - Where's the alpha?

Post by Robert T »

berntson wrote:This is actually something I've been trying to figure out. There must be a white paper somewhere that works out the relation between the non-linearity of CRSP10 and the FF smb portfolio? Here are two hypotheses: (1) CRSP10 performance is non-linear because its smb loadings are non-linear. CRSP10 companies have smb loadings that are not only higher than CRSP8 and CRSP9, but non-linear. (2) CRSP10 performance is non-linear because the smallest companies are capturing returns from some other factor. I have been working on the assumption that (1) is the right hypotheses.

If (1) is right, then there is no particular reason to expect CRSP10 underperformance relative to CRSP8 and CRSP9 to show up as negative alpha. It could show up as smb drift. The question is: To what extent are the results for CRSP10 correlated with the smb portfolio when CRSP10 underperforms (or overperforms) relative to CRSP8 and CRSP9? I have no idea. Maybe someone else does?
The 1980s is one of the most interesting periods to look at. According to FF SmB data, from 1980-1989 the average size premium was close to zero, yet when you look at the annualized returns by decile you get the following:

1980-1989 annualized returns

Decile
1.....17.2%
2.....17.1%
3.....17.2%
4.....17.8%
5.....17.4%
6.....17.3%
7.....17.1%
8.....16.5%
9.....14.2%
10... 10.5%

Over this period, the size loads were fairly linear, while the alphas on the smallest stocks - decile 9 and 10 - were negative and statistically significant

1980-1989 - monthly data

Decile.....Size load.....Alpha
1............-0.38.........0.04
2............-0.01.........0.00
3............+0.21.........0.02
4............+0.36.........0.12
5............+0.53.........0.09
6............+0.69.........0.06
7............+0.81.........0.10
8............+0.82.........0.06
9............+0.97.......-0.14
10..........+1.12........-0.43

So when using the 3 factor model, alphas on decile 10 tend to vary due to its non-linearity in returns (as far as I can tell), rather than SmB loads, at least over the 1980s. The key factor explaining the difference over this period seems to be the 'profitability' factor. When this factor is added the alpha on decile 10 declines from -0.43 to -0.17. The profitability loads was negative (-0.30) and significant.

Here are the profitability loads by decile:

1980-1989 - monthly data

Decile....Profitability load*
1.............-0.02
2..............0.09
3..............0.02
4..............0.05
5..............0.09
6..............0.09
7..............0.11
8.............-0.02
9............-0.10
10...........-0.30

* These profitability loads are results from regressions including the mkt, size, value, momentum, and profitability factors.
Data are from Ken French and Novy-Marx websites.

My take aways:

1. Additional factors explain differences in decile 10 returns (re: its apparent non-linearity). These include the profitability factor (particular the underperformance in 1980s), momentum (although profitability seems to have a larger effect at least from 1963-2012), and default risk ( http://pages.nes.ru/agoriaev/Papers/Vas ... 20JF04.pdf ). Another factor that could also possibly contribute is liquidity risk (although have not done any analysis on this but here in an interesting paper by Ibbotson - http://www.ibbotson.com/us/documents/Me ... tStyle.pdf ).

2. I also note that the prominence of the non-linearity in the Ken French decile 10 data is not as large as the CRSP10 data, the former is rebalanced annually, the latter is rebalanced quarterly. The seems to suggest that some of the 'non-linearity' may come from more frequent rebalancing.

Possible implications for managing a micro-cap (decile 10 type) portfolio - screen for profitability and rebalance fairly frequently.

Robert
.
berntson
Posts: 1366
Joined: Mon Oct 29, 2012 12:10 pm

Re: DFA - Where's the alpha?

Post by berntson »

Thanks Robert for for the exceptionally helpful reply. So it looks as though CSRP10 stocks behave somewhat differently than other small caps, not just that they have higher returns. I'm a bit confused though. Shouldn't negative profitability and momentum loads decrease returns? If anything, this makes the historic higher returns of CSRP10 stocks more of a puzzle. Perhaps liquidity and default risk can make up the difference, but it's a big difference to make up.

I'm looking forward to reading the papers you linked to--this is something that I want to sort out before seriously considering a microcap fund for my portfolio.
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: DFA - Where's the alpha?

Post by Robert T »

berntson wrote:Thanks Robert for for the exceptionally helpful reply. So it looks as though CSRP10 stocks behave somewhat differently than other small caps, not just that they have higher returns. I'm a bit confused though. Shouldn't negative profitability and momentum loads decrease returns? If anything, this makes the historic higher returns of CSRP10 stocks more of a puzzle. Perhaps liquidity and default risk can make up the difference, but it's a big difference to make up.

I'm looking forward to reading the papers you linked to--this is something that I want to sort out before seriously considering a microcap fund for my portfolio.
Berntson,

Here's an extract from an earlier thread on CRSP10 non-linearity, with a better formatted table
Robert T wrote:What’s the source of the CRSP10 non-linearity? Transaction costs indeed play a role but IMO CRSP10 stocks have fundamentally higher risk. Consider the following table. The first column is the average company size by decile (derived from Ibbotson Yearbook), the second column is a transpose of work by Gutierrez that estimates the average company size by Moody credit rating. When transposed onto the deciles it largely maps out as in the second column below. The third column is the percentage default rates by Moody rating. Default rates in decile 10 appear to be substantially higher than the other deciles reflecting much higher risk. While this refers to bond default, the higher fundamental risk of these companies IMO is also reflected in the cost of equity capital and hence equity returns – leading to the CRSP10 non-linearity relative to the other deciles.
  • Image
On profitability and momentum loads: After accounting for "profitability", the momentum load on CRSP10 is close to zero, but I expect there is some correlation between the two i.e. those companies just passing through to bankruptcy are likely to both be the least profitable and the ones with largest negative momentum. Bridgeway tries to screen out the companies passing through to bankruptcy - and since inception the "profitability" load has been positive, not negative, so there is some apparent benefit. But this sometimes comes at a cost as in 2009 when stocks screened out were the best performers - so some sampling error risk - although since inception Bridgeway has got close to CRSP10 returns. Not perfect, but I think Bridgeway does one of the best jobs in this space. Are micro-caps essential? No. But IMO, BRSIX provides a good option for those who want exposure along the size spectrum through to decile 10, with potential 'non-linear' effects (concentrated risk). Obviously no guarantees.

Robert
.
User avatar
Topic Author
Blue
Posts: 1178
Joined: Sat Jul 12, 2008 10:18 pm

Re: DFA - Where's the alpha?

Post by Blue »

Robert T wrote:Another factor that could also possibly contribute is liquidity risk (although have not done any analysis on this but here in an interesting paper by Ibbotson - http://www.ibbotson.com/us/documents/Me ... tStyle.pdf ).
Interesting paper -
The liquidity effect, however, is strongest among microcap stocks and declines from micro- to small- to mid- to large-cap stocks. The microcap geometric mean row contains both the highest- and the lowest-return cells in the matrix.

Quartile Low Liquidity Mid-Low Liquidity Mid-High Liquidity High Liquidity
Microcap
Geometric mean 15.36% 16.21% 9.94% 1.32%
Against the backdrop of Rick Ferri's Forbes Article on microcap fund underperformance in an outsized year for microcaps,
You may think that with micro-cap returns like these, micro-cap index fund investors would be jumping for joy. But they’re not. Why? Because the returns of all micro-cap index funds were well below these returns, so low that they didn’t even reach the Wilshire 5000 return.

What is the problem with micro-cap index funds and ETFs? The problem is that there are no true micro-cap index funds or ETFs. They don’t exist. They can’t exist because most micro-cap stocks are not investable.

Many companies in the micro-cap segment have tiny floats and no trading volume and that makes them very difficult for funds to buy. Other stocks have market prices less than $1 per share and are off-limits to fund managers due to company policy. Yet in 2009, these non-investable segments produced superior returns in the micro-cap segment.
Which leads me to this question, Robert T - Can I sign up for your private newsletter of microcap value stocks with favorable momentum, profitability, and liquidity factors? :)
berntson
Posts: 1366
Joined: Mon Oct 29, 2012 12:10 pm

Re: DFA - Where's the alpha?

Post by berntson »

Robert T wrote: On profitability and momentum loads: After accounting for "profitability", the momentum load on CRSP10 is close to zero, but I expect there is some correlation between the two i.e. those companies just passing through to bankruptcy are likely to both be the least profitable and the ones with largest negative momentum. Bridgeway tries to screen out the companies passing through to bankruptcy - and since inception the "profitability" load has been positive, not negative, so there is some apparent benefit. But this sometimes comes at a cost as in 2009 when stocks screened out were the best performers - so some sampling error risk - although since inception Bridgeway has got close to CRSP10 returns. Not perfect, but I think Bridgeway does one of the best jobs in this space. Are micro-caps essential? No. But IMO, BRSIX provides a good option for those who want exposure along the size spectrum through to decile 10, with potential 'non-linear' effects (concentrated risk). Obviously no guarantees.
Thanks again Robert--I'm now more confident in the management of BRSIX after this thread. I don't particularly mind alpha, I just want to make sure the fund is not consistently losing performance to management fees and/or ill conceived screens and trading strategies. The fact the CRSP10 performance is driven by factors other than smb makes the fund especially attractive for diversification purposes. Adding a microcap fund isn't (as I assumed) just a way to add more smb.
Post Reply