DFA vs Vanguard 10-year vs. 5 and 3 yr.
DFA vs Vanguard 10-year vs. 5 and 3 yr.
DFA out performance appears to be attributed to factors occurring over 5 years ago. Am I reading the data correctly?
- ddb
- Posts: 5511
- Joined: Mon Feb 26, 2007 11:37 am
- Location: American Gardens Building, West 81st St.
If you compare apples to apples, you should find that the performance of DFA is the same as Vanguard's index funds over basically any time period.dbonnett wrote:10 year performance of DFA funds versus Vanguard was significantly better. Five year returns are almost identical. Did one or the other change the composition of their funds?
Doing a basic comparison, though, will make DFA look better when small and/or value show large outperformance (2000-2005), and will make DFA look worse when small and/or value show large underperformance (1995-1999). When value performs similarly to growth and when small performs similarly to large (2006-2009), the two fund families will look similar.
In other words, performance differences are almost always explained by the varying degrees of exposure to the Fama-French risk factors, most notably SmB and HmL factors.
- DDB
"We have to encourage a return to traditional moral values. Most importantly, we have to promote general social concern, and less materialism in young people." - PB
DFA is not going to outperform every year, and during some stretches where growth and large stocks outpace small and value stocks, Vanguard will outpace DFA. That is not always the case, however. The last year is a vivid illustration of that. Consider the following Russell index returns for the last 12 months:
Russell 1000 Growth = +40.8%
Russell 1000 Value = +29.2%
Russell 2000 Growth = +43.3%
Russell 2000 Value = +32.3%
Russell Int’l LG = +41.9%
Russell Int’l LV = +45.8%
Russell Int’l SG = +64.5%
Russell Int’l SV = +40.8%
In every asset class besides international LC, growth has beaten value. But when we compare DFAs 4 core value strategies to the closest Vanguard fund (Value Index, Small Value Index, Int’l Value, Int’l Explorer), we see the following DFA advantages over the last year:
DFA Large Value: +13.4%
DFA Small Value: +4.3%
DFA Int’l Large Value: +7.3%
DFA Int’l Small Value: -8.2%
Int’l Explorer has benefited in the last 12 months not only from a greater growth tilt (and an international small growth advantage of almost 14%), but also a huge slug of emerging markets stocks. In every other category, Vanguard lagged pretty noticeably, despite the fact that growth has strongly outpaced value.
DDB is correct, DFA does have a stronger and more consistent tilt to value stocks, but that is a benefit, not a drawback. Furthermore, their exposure to value stocks is consistent across the globe, so you don’t have to hold allocations with mismatched tilts (heavier US small/value allocation due to better options), so your portfolio structure is more uniform. DFAs funds are also better managed. They don’t arbitrarily reconstitute them one a year, but instead use cash flows to ensure they own as close to their target universe as possible on a daily basis, while patiently trading out of stocks that no longer meet the buy criteria. If anything, DFAs process has become better and more refined over the last 3-5 years.
Finally, don’t forget that while DFA funds are more tilted to riskier small cap and value stocks (thanks to MPT, not all of that risk shows up on a portfolio level), Vanguards international funds are more tilted to riskier emerging markets stocks. Depending on how you look at it, the two risks cancel each other out, leaving DFA with the risk adjusted net return advantage.
Of course, beauty is in the eye of the beholder, and Vanguard offers some excellent alternatives to DFAs funds.
Russell 1000 Growth = +40.8%
Russell 1000 Value = +29.2%
Russell 2000 Growth = +43.3%
Russell 2000 Value = +32.3%
Russell Int’l LG = +41.9%
Russell Int’l LV = +45.8%
Russell Int’l SG = +64.5%
Russell Int’l SV = +40.8%
In every asset class besides international LC, growth has beaten value. But when we compare DFAs 4 core value strategies to the closest Vanguard fund (Value Index, Small Value Index, Int’l Value, Int’l Explorer), we see the following DFA advantages over the last year:
DFA Large Value: +13.4%
DFA Small Value: +4.3%
DFA Int’l Large Value: +7.3%
DFA Int’l Small Value: -8.2%
Int’l Explorer has benefited in the last 12 months not only from a greater growth tilt (and an international small growth advantage of almost 14%), but also a huge slug of emerging markets stocks. In every other category, Vanguard lagged pretty noticeably, despite the fact that growth has strongly outpaced value.
DDB is correct, DFA does have a stronger and more consistent tilt to value stocks, but that is a benefit, not a drawback. Furthermore, their exposure to value stocks is consistent across the globe, so you don’t have to hold allocations with mismatched tilts (heavier US small/value allocation due to better options), so your portfolio structure is more uniform. DFAs funds are also better managed. They don’t arbitrarily reconstitute them one a year, but instead use cash flows to ensure they own as close to their target universe as possible on a daily basis, while patiently trading out of stocks that no longer meet the buy criteria. If anything, DFAs process has become better and more refined over the last 3-5 years.
Finally, don’t forget that while DFA funds are more tilted to riskier small cap and value stocks (thanks to MPT, not all of that risk shows up on a portfolio level), Vanguards international funds are more tilted to riskier emerging markets stocks. Depending on how you look at it, the two risks cancel each other out, leaving DFA with the risk adjusted net return advantage.
Of course, beauty is in the eye of the beholder, and Vanguard offers some excellent alternatives to DFAs funds.
- ddb
- Posts: 5511
- Joined: Mon Feb 26, 2007 11:37 am
- Location: American Gardens Building, West 81st St.
Disagree. It's neither a benefit nor a drawback, just a different methodology. There is no "correct" level of tilting towards (or away from) small and/or value stocks in a portfolio.the fixer wrote:DFA does have a stronger and more consistent tilt to value stocks, but that is a benefit, not a drawback
Dubious claim, IMO. Do you think DFA is really the only company that uses cash flows to rebalance? Vanguard isn't reconstituting funds once per year, even if the underlying index is doing so. I'd say that the jury is still out regarding whether passive non-index is better than passive index.DFAs funds are also better managed. They don’t arbitrarily reconstitute them one a year, but instead use cash flows to ensure they own as close to their target universe as possible on a daily basis, while patiently trading out of stocks that no longer meet the buy criteria. If anything, DFAs process has become better and more refined over the last 3-5 years.
Depends on the fund. Not all Vanguard foreign stock funds include exposure to emerging markets.Finally, don’t forget that while DFA funds are more tilted to riskier small cap and value stocks (thanks to MPT, not all of that risk shows up on a portfolio level), Vanguards international funds are more tilted to riskier emerging markets stocks. Depending on how you look at it, the two risks cancel each other out, leaving DFA with the risk adjusted net return advantage.
Not trying to sound like a Vanguard shill here; I don't even own a single Vanguard fund or ETF. DFA and Vanguard are both great fund companies, but neither has stumbled onto some secret that makes them or any of their funds automatically better than anybody else.
- DDB
"We have to encourage a return to traditional moral values. Most importantly, we have to promote general social concern, and less materialism in young people." - PB
DDB, as I said, beauty is in the eye of the beholder. However, I have a hard time believing any of my comments below could be disputed?
You also want to be as diversified as possible, to ensure that you own the small number of stocks that historically have driven the return premium. A great example of this is when DFA changed (2006 I believe) the value sort on their US Large Value fund from the bottom 10% of book to market to the bottom 20%. The later doubled the amount of diversification of the fund, while historically providing a more consistent (similar magnitude, but much higher statistical significance) degree of outperformance over both the S&P 500 and the Russell 1000 Value
At the same time, whether you want a value tilt or an all-value portfolio, you want a balance between broad diversification and deep value exposure, where most of the outperformance occurs. For all-value investors, this is fairly intuitive. But for tilters, all things being equal, a deeper small or value fund allows you to hold the minimal amount of portfolio dollars in non market portfolios, reducing tracking error regret during periods of underperformance.
Of course, too concentrated and you add unsystematic (3 factor) risk, as well as higher volatility and risk of loss. If you look at the FF 5X5 matrix (instead of the 2X3), you find that the biggest size cohort of the deepest value stocks actually had 100% losses in 1930.
I didn’t say DFA is the only company to use cash flows advantageously. I did say, or imply, that DFA does it to a much greater extent than Vanguard does. Their index funds are reconstituted once or twice a year, and the stock weights determined at that time are maintained until the next reconstitution date. Yes, Vanguard has incorporated some hold ranges in their MSCI indexes, but where did they get this idea from? DFA had been doing it for 20 years before Vanguard stumbled on it.
It’s a hard argument to make that when it comes to capturing an asset class, a structured approach that:
a)uses cash flows to rebalance
b)trades incoming/outgoing stocks patiently while being mindful of impact of momentum
c)targets the entire universe of stocks in the asset class (including microcap stocks)
d)screens out stocks that look like, but do not behave as the asset class should (utilities, REITS)
...does not do a better job than an index fund simply designed as a subset of a sampled total market universe where every stock must be assigned to some sub-index, regardless of whether or not the research indicates they are beneficial to driving the multifactor premium return which is then traded mechanically with minimal consideration about real-world trading issues (some, but not much).
And, it is clearly a better approach than the active funds Vanguard employs that adds the unwanted element of active manager risk.
No, this is very clearly a benefit. Smaller and more value oriented stocks have a higher expected return than they market, so you want consistent exposure to the dimension to ensure you are there when the return premium shows up. That’s fairly obvious.the fixer wrote:
DFA does have a stronger and more consistent tilt to value stocks, but that is a benefit, not a drawback
Disagree. It's neither a benefit nor a drawback, just a different methodology. There is no "correct" level of tilting towards (or away from) small and/or value stocks in a portfolio.
You also want to be as diversified as possible, to ensure that you own the small number of stocks that historically have driven the return premium. A great example of this is when DFA changed (2006 I believe) the value sort on their US Large Value fund from the bottom 10% of book to market to the bottom 20%. The later doubled the amount of diversification of the fund, while historically providing a more consistent (similar magnitude, but much higher statistical significance) degree of outperformance over both the S&P 500 and the Russell 1000 Value
At the same time, whether you want a value tilt or an all-value portfolio, you want a balance between broad diversification and deep value exposure, where most of the outperformance occurs. For all-value investors, this is fairly intuitive. But for tilters, all things being equal, a deeper small or value fund allows you to hold the minimal amount of portfolio dollars in non market portfolios, reducing tracking error regret during periods of underperformance.
Of course, too concentrated and you add unsystematic (3 factor) risk, as well as higher volatility and risk of loss. If you look at the FF 5X5 matrix (instead of the 2X3), you find that the biggest size cohort of the deepest value stocks actually had 100% losses in 1930.
Quote:
DFAs funds are also better managed. They don’t arbitrarily reconstitute them one a year, but instead use cash flows to ensure they own as close to their target universe as possible on a daily basis, while patiently trading out of stocks that no longer meet the buy criteria. If anything, DFAs process has become better and more refined over the last 3-5 years.
Dubious claim, IMO. Do you think DFA is really the only company that uses cash flows to rebalance? Vanguard isn't reconstituting funds once per year, even if the underlying index is doing so. I'd say that the jury is still out regarding whether passive non-index is better than passive index.
I didn’t say DFA is the only company to use cash flows advantageously. I did say, or imply, that DFA does it to a much greater extent than Vanguard does. Their index funds are reconstituted once or twice a year, and the stock weights determined at that time are maintained until the next reconstitution date. Yes, Vanguard has incorporated some hold ranges in their MSCI indexes, but where did they get this idea from? DFA had been doing it for 20 years before Vanguard stumbled on it.
It’s a hard argument to make that when it comes to capturing an asset class, a structured approach that:
a)uses cash flows to rebalance
b)trades incoming/outgoing stocks patiently while being mindful of impact of momentum
c)targets the entire universe of stocks in the asset class (including microcap stocks)
d)screens out stocks that look like, but do not behave as the asset class should (utilities, REITS)
...does not do a better job than an index fund simply designed as a subset of a sampled total market universe where every stock must be assigned to some sub-index, regardless of whether or not the research indicates they are beneficial to driving the multifactor premium return which is then traded mechanically with minimal consideration about real-world trading issues (some, but not much).
And, it is clearly a better approach than the active funds Vanguard employs that adds the unwanted element of active manager risk.
Other than the regional indexes and TM and taxable EAFE funds, almost every Vanguard index and actively managed International fund holds emerging markets stocks (Emerging Markets Index, FTSE All World xUS, FTSE All World xUS Small, Global Equity, Int’l Explorer, Int’l Growth, Int’l Value, Total International, and Total World).Quote:
Finally, don’t forget that while DFA funds are more tilted to riskier small cap and value stocks (thanks to MPT, not all of that risk shows up on a portfolio level), Vanguards international funds are more tilted to riskier emerging markets stocks. Depending on how you look at it, the two risks cancel each other out, leaving DFA with the risk adjusted net return advantage.
Depends on the fund. Not all Vanguard foreign stock funds include exposure to emerging markets.
- ddb
- Posts: 5511
- Joined: Mon Feb 26, 2007 11:37 am
- Location: American Gardens Building, West 81st St.
Okay, so what do all of these wonderful benefits add up to? Are DFA funds delivering consistent positive alpha after adjusting for the risks?
- DDB
- DDB
"We have to encourage a return to traditional moral values. Most importantly, we have to promote general social concern, and less materialism in young people." - PB
A 4X25 monthly rebalanced portfolio of the above value funds generated a +0.03/month positive alpha net of fees based on a 9 factor global factor model (5 regional market risks, as well as US and non-US developed size and value risks) over the last decade.ddb wrote:Okay, so what do all of these wonderful benefits add up to? Are DFA funds delivering consistent positive alpha after adjusting for the risks?
- DDB
The 5YR Global fund had a +0.09/month positive 5 factor alpha (market, size, price, term, and default).
For reference, the only index series I am aware of that was around (and not completely changed/updated) over the last 10 years was Russell on the US side and MSCI on the non-US side. A 4X25 portfolio of the Russell 1000 Value, Russell 2000 Value, MSCI EAFE Value, and MSCI EAFE Small had a -0.03/month negative 9 factor alpha, before associated ETF fees.
Is that +0.03%/month?the fixer wrote:A 4X25 monthly rebalanced portfolio of the above value funds generated a +0.03/month positive alpha net of fees based on a 9 factor global factor model (5 regional market risks, as well as US and non-US developed size and value risks) over the last decade.ddb wrote:Okay, so what do all of these wonderful benefits add up to? Are DFA funds delivering consistent positive alpha after adjusting for the risks?
- DDB
The 5YR Global fund had a +0.09/month positive 5 factor alpha (market, size, price, term, and default).
For reference, the only index series I am aware of that was around (and not completely changed/updated) over the last 10 years was Russell on the US side and MSCI on the non-US side. A 4X25 portfolio of the Russell 1000 Value, Russell 2000 Value, MSCI EAFE Value, and MSCI EAFE Small had a -0.03/month negative 9 factor alpha, before associated ETF fees.
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.
the fixer wrote:Smaller and more value oriented stocks have a higher expected return than they market, so you want consistent exposure to the dimension to ensure you are there when the return premium shows up. That’s fairly obvious.
I am more familiar hearing it referred to as a risk premium rather than a return premium - the risk being that the return never shows up in your or any lifetime.
Some would argue that the value risk premium is a free lunch or at least a free dessert because the market as a whole over estimates the risk of value, and so, returns in excess of the risk taken can be earned. If you believe this, then DFA's greater value tilt is a benefit.
Others disagree, saying that the value risk will show up someday and perhaps at the worst possible time. Like 1930 right after you just lost your job. If you believe this, then it's not a benefit, it's a feature.
Sometimes when thinking of value premiums, I am reminded of Eliza Doolittle of My Fair Lady singing "Just you wait, Henry Higgins, just you wait...".
Ken
.
1. Historically, the annually rebalanced FF portfolios have had higher returns than the quarterly rebalanced FF portfolio equivalents (FF research vs. benchmark value portfolios).
2. DFA funds have historically had a higher negative momentum load than the common index funds (likely due to the more frequent rebalancing).
3. HmL and SmB constructs are based on simple stock ‘annually rebalanced index’ assignments which generated the historical premiums.
4. I do not recall any DFA funds which I have looked at, which if assessed on an individual fund basis had a positive alpha (including the four mentioned above). If there are, I would certainly like to see the data series (evidence).
.
1. Historically, the annually rebalanced FF portfolios have had higher returns than the quarterly rebalanced FF portfolio equivalents (FF research vs. benchmark value portfolios).
2. DFA funds have historically had a higher negative momentum load than the common index funds (likely due to the more frequent rebalancing).
3. HmL and SmB constructs are based on simple stock ‘annually rebalanced index’ assignments which generated the historical premiums.
4. I do not recall any DFA funds which I have looked at, which if assessed on an individual fund basis had a positive alpha (including the four mentioned above). If there are, I would certainly like to see the data series (evidence).
.
- speedbump101
- Posts: 999
- Joined: Thu Oct 18, 2007 10:54 pm
- Location: Alberta Canada
Sightly off topic, however the turnover rate of the DFA Large Value, compared to Vanguard has DFA reporting one tenth of Vanguards turnover rate... Not sure if this is typical, however it sure got my attention when I saw it...
DFA 7% turnover -DFCVX https://personal.vanguard.com/us/funds/ ... IntExt=EXT...
Vanguard 86.3% turnover -VUVLX https://personal.vanguard.com/us/funds/ ... st=tab%3A2
SB...
DFA 7% turnover -DFCVX https://personal.vanguard.com/us/funds/ ... IntExt=EXT...
Vanguard 86.3% turnover -VUVLX https://personal.vanguard.com/us/funds/ ... st=tab%3A2
SB...
Last edited by speedbump101 on Wed Jan 13, 2010 5:12 pm, edited 1 time in total.
"Man is not a rational animal, he is a rationalizing animal" -Robert A. Heinlein
In answer to:Robert T wrote:.
1. Historically, the annually rebalanced FF portfolios have had higher returns than the quarterly rebalanced FF portfolio equivalents (FF research vs. benchmark value portfolios).
2. DFA funds have historically had a higher negative momentum load than the common index funds (likely due to the more frequent rebalancing).
3. HmL and SmB constructs are based on simple stock ‘annually rebalanced index’ assignments which generated the historical premiums.
4. I do not recall any DFA funds which I have looked at, which if assessed on an individual fund basis had a positive alpha (including the four mentioned above). If there are, I would certainly like to see the data series (evidence).
.
1) if you rebalance more frequently with no consideration of momentum characteristics (as the FF Indexes reconstituted quarterly do), then you will load more heavily on momentum (negatively), and therefore have lower returns. If you rebalance more frequently while being mindful of momentum (mainly by avoiding it in buy and sell decisions), you will benefit via more consistent value exposure without all of the negative costs of being over exposed to negative momentum.
2) DFA only started incorporating momentum screens in their funds in '05-'06. Prior to that it was just patient trading, which are not the same. Since '06, DFA LV and SV still have some negative momentum exposure, but that is less than half (coef of about 0.05) compared to the 93-05 period (coef of over 0.15)
Also, retail indexes also have exposure to negative momentum. All but the S&P 600 Value, that is. The Russell 2000 Value and MSCI 1750 Value all have exposure to negative momentum. The later two are only reconstituted once per year, so you have less consistent exposure to the value premium, and the 600 Value index is only reconstituted when the S&P committee gets around to it -- not exactly optimal for factor exposure. Of course all of these suboptimal approaches will naturally have lower (or 0) negative momentum exposure, but also lower value exposure and therefore lower expected total returns.
The 600 Pure Value doesn't need to be part of this discussion, as it is woefully underdiversified, and has had a whopping -0.3%/month negative momentum coef.
3. This changed a few years ago to avoid as much issue with negative momentum in the index series, without putting a momentum filter on the FF data series (according to FF themselves via DFA report)
4. since 2000, DFLVX has a -0.01%/month alpha, DFSVX has a -0.02%/month, DFFVX (US Targeted Value) has a +0.13%/month positive alpha, DFIVX has a -0.01%/month alpha, and DISVX has a +0.14/month positive alpha
Almost any combination of global large/small value will produce a positive 3F alpha net of fees since the Lost Decade of '00 began.
Also, the more recent Vector strategies (index data prior to inception), have also generated positive 3F alphas of 0.05% per month (both US and Int'l versions) since 2000.
Hope this helps.
No, thats low for a value fund unless you have a year where very few stocks migrated. Fama/French research indicates that 15% to 25% of stocks migrate out of their small/value segments each year, so on average, that is how much turnover you should see.speedbump101 wrote:Sightly off topic, however the turnover rate of the DFA Large Value, compared to Vanguard has DFA reporting one tenth of Vanguards turnover rate... Not sure if this is typical, however it sure got my attention when I saw it...
DFA 7% turnover -DFCVX https://personal.vanguard.com/us/funds/ ... IntExt=EXT...
Vanguard 86.3% turnover -VUVLX https://personal.vanguard.com/us/funds/ ... st=tab%3A2
SB...
This number will be lower (say 1/2 that much) if you get your value tilts through a Core fund like the Vector strategies (10% turnover last year). Compare that with 30% to 50% turnover from typical value and micro cap indexes/asset class funds, as well as portfolio turnover from rebalancing, and you are starting to talk about real expense savings.
I don't know what alpha was generated or not, but I have this, a slice & dice portfolio of Vanguard index funds compared to DFA funds for last 10 years.
lb 15.00%
lv 15.00%
sb 15.00%
sv 15.00%
re 10.00%
ilv 20.00%
isc 10.00%
Vanguard 4.63%
DFA 5.97%
Above portfolio re-balanced annually to original allocation and without cost of access to DFA.
After adding cost of DFA, performance is only slightly better. But one can say that DFA funds paid for the cost of advice over last 10 years, and an investor in them is breakeven with a DIY investor using same Vanguard funds. Now, I know for a fact that most DIY investors don't have the discipline to stay the course and most didn't have a slice & dice portfolio going into 2000. Therefore actual returns for a DIY investor could be anywhere. I know this, my own personal portfolio performed in that range, and I didn't stay still all along, I had bought and sold many funds in between, but somehow manage to come out on par with a buy & hold slice & dice portfolio.
Important lesson here, buy & hold, keep costs low, invest in index funds or passive funds, don't time the market. If you follow all this, then DFA or Vanguard, or ETFs from iShares/SSgA etc will be just fine.
lb 15.00%
lv 15.00%
sb 15.00%
sv 15.00%
re 10.00%
ilv 20.00%
isc 10.00%
Vanguard 4.63%
DFA 5.97%
Above portfolio re-balanced annually to original allocation and without cost of access to DFA.
After adding cost of DFA, performance is only slightly better. But one can say that DFA funds paid for the cost of advice over last 10 years, and an investor in them is breakeven with a DIY investor using same Vanguard funds. Now, I know for a fact that most DIY investors don't have the discipline to stay the course and most didn't have a slice & dice portfolio going into 2000. Therefore actual returns for a DIY investor could be anywhere. I know this, my own personal portfolio performed in that range, and I didn't stay still all along, I had bought and sold many funds in between, but somehow manage to come out on par with a buy & hold slice & dice portfolio.
Important lesson here, buy & hold, keep costs low, invest in index funds or passive funds, don't time the market. If you follow all this, then DFA or Vanguard, or ETFs from iShares/SSgA etc will be just fine.
Time will tell whether DFA can reduce/eliminate the negative momentum from more frequent rebalancing. And no matter which way we want to cut it, value exposure shows up in average value loads i.e. the share of average monthly value premiums captured whether rebalanced annually or more frequently. That’s the basis of determining expected returns (selecting funds/indexes with as close to a zero alpha as possible).the fixer wrote: 1) if you rebalance more frequently with no consideration of momentum characteristics (as the FF Indexes reconstituted quarterly do), then you will load more heavily on momentum (negatively), and therefore have lower returns. If you rebalance more frequently while being mindful of momentum (mainly by avoiding it in buy and sell decisions), you will benefit via more consistent value exposure without all of the negative costs of being over exposed to negative momentum.
2) DFA only started incorporating momentum screens in their funds in '05-'06. Prior to that it was just patient trading, which are not the same. Since '06, DFA LV and SV still have some negative momentum exposure, but that is less than half (coef of about 0.05) compared to the 93-05 period (coef of over 0.15)
Also, retail indexes also have exposure to negative momentum. All but the S&P 600 Value, that is. The Russell 2000 Value and MSCI 1750 Value all have exposure to negative momentum. The later two are only reconstituted once per year, so you have less consistent exposure to the value premium, and the 600 Value index is only reconstituted when the S&P committee gets around to it -- not exactly optimal for factor exposure. Of course all of these suboptimal approaches will naturally have lower (or 0) negative momentum exposure, but also lower value exposure and therefore lower expected total returns.
The 600 Pure Value doesn't need to be part of this discussion, as it is woefully underdiversified, and has had a whopping -0.3%/month negative momentum coef.
3. This changed a few years ago to avoid as much issue with negative momentum in the index series, without putting a momentum filter on the FF data series (according to FF themselves via DFA report)
4. since 2000, DFLVX has a -0.01%/month alpha, DFSVX has a -0.02%/month, DFFVX (US Targeted Value) has a +0.13%/month positive alpha, DFIVX has a -0.01%/month alpha, and DISVX has a +0.14/month positive alpha
Almost any combination of global large/small value will produce a positive 3F alpha net of fees since the Lost Decade of '00 began.
Also, the more recent Vector strategies (index data prior to inception), have also generated positive 3F alphas of 0.05% per month (both US and Int'l versions) since 2000.
On alphas since 2000 (the earlier period used), the MSCI Midcap value had an alpha of +0.22/month. Over a longer period, since June 1992 (start of the MSCI data), the alpha disappears to almost exactly zero (0.02). For earlier data I have on the DFA large value from May 1993 to Sept 2006, alpha = -0.24/month, from Jan 2000 to Sept 2006 alpha = -0.09.
.
Yes,Robert T wrote:Time will tell whether DFA can reduce/eliminate the negative momentum from more frequent rebalancing. And no matter which way we want to cut it, value exposure shows up in average value loads i.e. the share of average monthly value premiums captured whether rebalanced annually or more frequently. That’s the basis of determining expected returns (selecting funds/indexes with as close to a zero alpha as possible).the fixer wrote: 1) if you rebalance more frequently with no consideration of momentum characteristics (as the FF Indexes reconstituted quarterly do), then you will load more heavily on momentum (negatively), and therefore have lower returns. If you rebalance more frequently while being mindful of momentum (mainly by avoiding it in buy and sell decisions), you will benefit via more consistent value exposure without all of the negative costs of being over exposed to negative momentum.
2) DFA only started incorporating momentum screens in their funds in '05-'06. Prior to that it was just patient trading, which are not the same. Since '06, DFA LV and SV still have some negative momentum exposure, but that is less than half (coef of about 0.05) compared to the 93-05 period (coef of over 0.15)
Also, retail indexes also have exposure to negative momentum. All but the S&P 600 Value, that is. The Russell 2000 Value and MSCI 1750 Value all have exposure to negative momentum. The later two are only reconstituted once per year, so you have less consistent exposure to the value premium, and the 600 Value index is only reconstituted when the S&P committee gets around to it -- not exactly optimal for factor exposure. Of course all of these suboptimal approaches will naturally have lower (or 0) negative momentum exposure, but also lower value exposure and therefore lower expected total returns.
The 600 Pure Value doesn't need to be part of this discussion, as it is woefully underdiversified, and has had a whopping -0.3%/month negative momentum coef.
3. This changed a few years ago to avoid as much issue with negative momentum in the index series, without putting a momentum filter on the FF data series (according to FF themselves via DFA report)
4. since 2000, DFLVX has a -0.01%/month alpha, DFSVX has a -0.02%/month, DFFVX (US Targeted Value) has a +0.13%/month positive alpha, DFIVX has a -0.01%/month alpha, and DISVX has a +0.14/month positive alpha
Almost any combination of global large/small value will produce a positive 3F alpha net of fees since the Lost Decade of '00 began.
Also, the more recent Vector strategies (index data prior to inception), have also generated positive 3F alphas of 0.05% per month (both US and Int'l versions) since 2000.
On alphas since 2000 (the earlier period used), the MSCI Midcap value had an alpha of +0.22/month. Over a longer period, since June 1992 (start of the MSCI data), the alpha disappears to almost exactly zero (0.02). For earlier data I have on the DFA large value from May 1993 to Sept 2006, alpha = -0.24/month, from Jan 2000 to Sept 2006 alpha = -0.09.
.
Mid Value has had strong positive alpha's since 2000 (any mid value index, not just MSCI). But this is just a reversion to the mean after the 15 previous years where mid value was the worst performing of the 3 value cohorts, and had strong negative 3F alpha (based on Russell):
1986-1999 Annualized Returns (3F monthly alphas in paranthesis, statistical significance in bold)
Large Value = +17.5% (-0.08%)
Mid Value = +14.3% (-0.21%)
Small Value = +14.6% (+0.02%)
Obviously, this was a strong period for large cap stocks, which should probably serve as a warning for those using a mid value index as their large value proxy. The factor loads are in the same ballpark as a deep value large cap fund like DFLVX, but there is a significant difference in stock characteristics and expected results during periods when large beats small. The weighted average market cap of the Vanguard Mid Value Index is less than $5B. The weighted average market cap of DFA US Large Value is over $40B.
Fixed said
"A 4X25 monthly rebalanced portfolio of the above value funds generated a +0.03/month positive alpha net of fees based on a 9 factor global factor model (5 regional market risks, as well as US and non-US developed size and value risks) over the last decade. "
When you say "net of fees", which fees are you referring to? Is it the advisor fee..? If so, what did you assume to be the advisor fee
1210
"A 4X25 monthly rebalanced portfolio of the above value funds generated a +0.03/month positive alpha net of fees based on a 9 factor global factor model (5 regional market risks, as well as US and non-US developed size and value risks) over the last decade. "
When you say "net of fees", which fees are you referring to? Is it the advisor fee..? If so, what did you assume to be the advisor fee
1210
Yes, exactly we shouldn’t confuse ‘brains’ (re: DFAs active management) with alpha drift (the caution in my earlier post on the reported performance since 2000). And on the 1986-1999 performance, we also shouldn’t confuse Russell LV performance with DFA LV. The former has a larger cap weight, and outperformed the FF LV(xU), by about 1.5% annualized over this period (similar to the 1.4% outperformance of the DFA LV (FF LV(xU) over Vanguard Mid Value [proxied by Ibbotson MCV] over the same period). IMO the criteria for fund choice should be factor load driven, not large cap vs. midcap, there is a broad range of factor loads across these two categories.the fixer wrote:Mid Value has had strong positive alpha's since 2000 (any mid value index, not just MSCI). But this is just a reversion to the mean after the 15 previous years where mid value was the worst performing of the 3 value cohorts, and had strong negative 3F alpha (based on Russell):
1986-1999 Annualized Returns (3F monthly alphas in paranthesis, statistical significance in bold)
Large Value = +17.5% (-0.08%)
Mid Value = +14.3% (-0.21%)
Small Value = +14.6% (+0.02%)
Obviously, this was a strong period for large cap stocks, which should probably serve as a warning for those using a mid value index as their large value proxy. The factor loads are in the same ballpark as a deep value large cap fund like DFLVX, but there is a significant difference in stock characteristics and expected results during periods when large beats small. The weighted average market cap of the Vanguard Mid Value Index is less than $5B. The weighted average market cap of DFA US Large Value is over $40B.
.
Robert,
a) there is a difference between just happing to be in the right area of the market (ie. Mid Value) that registers a period dependent positive alpha, and engineering a portfolio in a less favorable area (mostly LV) to deliver almost 0 alpha net of fund fees.
b) your data on the FF LV xUtility index is not accurate. From 86-99, it slightly outpaced the Russell 1000 Value Index by 0.1% per year. Not sure where you are coming up with -1.4% per year?
c) with hindsight, you can cobble together a mix of mid value indexes that delievered the best returns (Ibbotson + MSCI), but to my knowledge, Russell is the only one that was actually around in the 90s, so its far more relevant. To that extent, the Russell Mid Value index was outpaced by the Dimensional Large Value Index by over 3% per year (when large was beating small) and almost 2% vs. the Fama/French Large Value xUtility Index. Both FF/DFA Large Value choices had positive, not negative tracking error relative to their Russell benchmark.
d) factor loads aside, looking at pure portfolio characteristics, the Russell 1000 Value or MSCI 750 Value is a much more appropriate benchmark for DFA Large Value than a midcap fund. Currently, for example, DFAs buy universe consists of the largest 1000 stocks in the market (or biggest 90% of market, whichever is larger), and DFA reports that the fund is only 20% in mid cap stocks. This is the exact same buy universe as the Russell 1000. Yes, it excludes a few of the mega cap stocks held in relative value indexes because of its "bottom 20% BtM sort", but its a huge leap to jump from that to a Mid Cap index with a average market cap as small as the smallest stocks in the LV fund. Now, on average, you need a mid cap value index to achieve the returns of DFAs deep value large cap fund, but thats a whole different issue.
e) the FF LV xU index has never been a great bogey for the DFA Large Value fund, as DFLVX has consistently outpaced it. Since 2000, for example, DFLVX is 7% per year ahead of FF LV xU. The Dimensional LV Index is the best bogey that exists, but even that isn't perfect.
f) for some, focusing soley on factor loads might be appropriate, but for reasons I have highlighted before, along with the huge tracking error issue I raised above (3% per year underperformance for almost 15 years) means for most investors, they need to consider analysis outside of just the 3F model outputs.
I think we will just agree to disagree on a lot of these details.
I need to make some corrections here.Yes, exactly we shouldn’t confuse ‘brains’ (re: DFAs active management) with alpha drift (the caution in my earlier post on the reported performance since 2000). And on the 1986-1999 performance, we also shouldn’t confuse Russell LV performance with DFA LV. The former has a larger cap weight, and outperformed the FF LV(xU), by about 1.5% annualized over this period (similar to the 1.4% outperformance of the DFA LV (FF LV(xU) over Vanguard Mid Value [proxied by Ibbotson MCV] over the same period). IMO the criteria for fund choice should be factor load driven, not large cap vs. midcap, there is a broad range of factor loads across these two categories.
a) there is a difference between just happing to be in the right area of the market (ie. Mid Value) that registers a period dependent positive alpha, and engineering a portfolio in a less favorable area (mostly LV) to deliver almost 0 alpha net of fund fees.
b) your data on the FF LV xUtility index is not accurate. From 86-99, it slightly outpaced the Russell 1000 Value Index by 0.1% per year. Not sure where you are coming up with -1.4% per year?
c) with hindsight, you can cobble together a mix of mid value indexes that delievered the best returns (Ibbotson + MSCI), but to my knowledge, Russell is the only one that was actually around in the 90s, so its far more relevant. To that extent, the Russell Mid Value index was outpaced by the Dimensional Large Value Index by over 3% per year (when large was beating small) and almost 2% vs. the Fama/French Large Value xUtility Index. Both FF/DFA Large Value choices had positive, not negative tracking error relative to their Russell benchmark.
d) factor loads aside, looking at pure portfolio characteristics, the Russell 1000 Value or MSCI 750 Value is a much more appropriate benchmark for DFA Large Value than a midcap fund. Currently, for example, DFAs buy universe consists of the largest 1000 stocks in the market (or biggest 90% of market, whichever is larger), and DFA reports that the fund is only 20% in mid cap stocks. This is the exact same buy universe as the Russell 1000. Yes, it excludes a few of the mega cap stocks held in relative value indexes because of its "bottom 20% BtM sort", but its a huge leap to jump from that to a Mid Cap index with a average market cap as small as the smallest stocks in the LV fund. Now, on average, you need a mid cap value index to achieve the returns of DFAs deep value large cap fund, but thats a whole different issue.
e) the FF LV xU index has never been a great bogey for the DFA Large Value fund, as DFLVX has consistently outpaced it. Since 2000, for example, DFLVX is 7% per year ahead of FF LV xU. The Dimensional LV Index is the best bogey that exists, but even that isn't perfect.
f) for some, focusing soley on factor loads might be appropriate, but for reasons I have highlighted before, along with the huge tracking error issue I raised above (3% per year underperformance for almost 15 years) means for most investors, they need to consider analysis outside of just the 3F model outputs.
I think we will just agree to disagree on a lot of these details.
Last edited by the fixer on Thu Jan 14, 2010 2:15 pm, edited 2 times in total.
No, net of fund fees. Obviously, advisor fees would reduce the net benefits, as the lack of decent options in important asset classes like Int'l Small Value results in a "hidden cost" for index/ETF based plans that doesn't show up in 3F alphas.1210sda wrote:Fixed said
"A 4X25 monthly rebalanced portfolio of the above value funds generated a +0.03/month positive alpha net of fees based on a 9 factor global factor model (5 regional market risks, as well as US and non-US developed size and value risks) over the last decade. "
When you say "net of fees", which fees are you referring to? Is it the advisor fee..? If so, what did you assume to be the advisor fee
1210
The point above was simply to show that DFA was in fact elminating negative 3F alphas net of fees (which is no easy task for deep value strategies) through their combined management techniques, contrary to what was mentioned elsewhere.
the fixer,
DFA large value
May 1993 to Sept 2006, alpha = -0.24/month
Jan 2000 to Sept 2006 alpha = -0.09/month
I took the Russell LV returns from your post 17.5%: 17.5 – 16.0 = 1.5%. Which was inaccurate?
We seem to mainly disagree on the 1-3% DFA outperformance beyond factor exposure. I simply don't see evidence of this claim. Just presenting the other side of your arguments to maintain some balance in the analysis. That's about all I have.
Robert
.
I think you missed my point. Lets take the numbers below. Is this alpha drift or “lack of brains” from DFA, despite all the 'engineering' (obviously if it was positive the question would be is it ‘DFA brains’ or alpha drift). My take is its probably alpha drift.the fixer wrote: a) there is a difference between just happing to be in the right area of the market (ie. Mid Value) that registers a period dependent positive alpha, and engineering a portfolio in a less favorable area (mostly LV) to deliver almost 0 alpha net of fund fees.
DFA large value
May 1993 to Sept 2006, alpha = -0.24/month
Jan 2000 to Sept 2006 alpha = -0.09/month
Well I got the data from Jeff’s earlier linked spreadsheet when posted on the M* site.b) your data on the FF LV xUtility index is not accurate. From 86-99, it slightly outpaced the Russell 1000 Value Index by 0.1% per year. Not sure where you are coming up with -1.4% per year?
Code: Select all
1986 20.3
1987 3.9
1988 24.1
1989 27.5
1990 -22.5
1991 34.8
1992 16.0
1993 24.5
1994 -0.3
1995 40.1
1996 20.0
1997 33.7
1998 11.9
1999 7.0
Return 16.0
Yes, we can each use what we think is most relevant.c) with hindsight, you can cobble together a mix of mid value indexes that delievered the best returns (Ibbotson + MSCI), but to my knowledge, Russell is the only one that was actually around in the 90s, so its far more relevant. To that extent, the Russell Mid Value index was outpaced by the Dimensional Large Value Index by over 3% per year (when large was beating small) and almost 2% vs. the Fama/French Large Value xUtility Index. Both FF/DFA Large Value choices had positive, not negative tracking error relative to their Russell benchmark.
Well, if we believe the FF research, it’s the factor loads that matter over the long haul. So perhaps a different starting point.d) factor loads aside, looking at pure portfolio characteristics, the Russell 1000 Value or MSCI 750 Value is a much more appropriate benchmark for DFA Large Value than a midcap fund. Currently, for example, DFAs buy universe consists of the largest 1000 stocks in the market (or biggest 90% of market, whichever is larger), and DFA reports that the fund is only 20% in mid cap stocks. This is the exact same buy universe as the Russell 1000. Yes, it excludes a few of the mega cap stocks held in relative value indexes because of its "bottom 20% BtM sort", but its a huge leap to jump from that to a Mid Cap index with a average market cap as small as the smallest stocks in the LV fund. Now, on average, you need a mid cap value index to achieve the returns of DFAs deep value large cap fund, but thats a whole different issue.
Yes I think that is the case anyway – almost all individual investors don’t use the 3F model in allocation decisions. Only a small fraction on this forum do, which is a very small fraction of all investors. We each have to decide.f) for some, focusing soley on factor loads might be appropriate, but for reasons I have highlighted before, along with the huge tracking error issue I raised above (3% per year underperformance for almost 15 years) means for most investors, they need to consider analysis outside of just the 3F model outputs.
We seem to mainly disagree on the 1-3% DFA outperformance beyond factor exposure. I simply don't see evidence of this claim. Just presenting the other side of your arguments to maintain some balance in the analysis. That's about all I have.
Robert
.
Robert,
The 17.5% was for the Dimensional US Large Value Index (the closest proxy for DFLVX). The Russell 1000 Value Index did +15.94% over this period, the FF LV xU Index did +16.01%, exactly as you reported.
Take for example (along with my Dimensional Large Value Index v. Russell Mid Value Index comparison above) the Russell Mid Value Index (smb = .18; hml = .54) and Dimensional Marketwide Value Index (smb = .15; hml = .58 ) results over a strong large cap led market from 86-99:
DFA Mktwide Value Index = +17.3%
Russell Mid Value Index = +14.3%
So the vast difference in the average size of stock in the two strategies was hidden if we only focused on "factor tilts". Different approaches can lead to remarkably different results over short/intermediate periods, and is not well explained other than simply random +/- alpha.
The 86-99 period was simply a very poor relative period for mid cap stocks compared to large/small, as 00-09 was a very good relative period for mid cap stocks.
Or an example over this same stretch comparing the truely marketwide US Vector Index with the "polar component" TSM/Mid Value/CRPS 10 mix:
US Vector Index = +15.4%
20/40/40 Index = +13.4%
Well I got the data from Jeff’s earlier linked spreadsheet when posted on the M* site.
Code:
1986 20.3
1987 3.9
1988 24.1
1989 27.5
1990 -22.5
1991 34.8
1992 16.0
1993 24.5
1994 -0.3
1995 40.1
1996 20.0
1997 33.7
1998 11.9
1999 7.0
Return 16.0
I took the Russell LV returns from your post 17.5%: 17.5 – 16.0 = 1.5%. Which was inaccurate?
The 17.5% was for the Dimensional US Large Value Index (the closest proxy for DFLVX). The Russell 1000 Value Index did +15.94% over this period, the FF LV xU Index did +16.01%, exactly as you reported.
OK, So long as we realize it is a very noisy measure and prone to mask the true characteristics and results of a portfolio over some stretches.Well, if we believe the FF research, it’s the factor loads that matter over the long haul. So perhaps a different starting point.
Take for example (along with my Dimensional Large Value Index v. Russell Mid Value Index comparison above) the Russell Mid Value Index (smb = .18; hml = .54) and Dimensional Marketwide Value Index (smb = .15; hml = .58 ) results over a strong large cap led market from 86-99:
DFA Mktwide Value Index = +17.3%
Russell Mid Value Index = +14.3%
So the vast difference in the average size of stock in the two strategies was hidden if we only focused on "factor tilts". Different approaches can lead to remarkably different results over short/intermediate periods, and is not well explained other than simply random +/- alpha.
The 86-99 period was simply a very poor relative period for mid cap stocks compared to large/small, as 00-09 was a very good relative period for mid cap stocks.
Or an example over this same stretch comparing the truely marketwide US Vector Index with the "polar component" TSM/Mid Value/CRPS 10 mix:
US Vector Index = +15.4%
20/40/40 Index = +13.4%
Robert,
Lets take 5/93 through 11/09, when the alpha on DFLVX was an insignificant -0.15% per month. FF HmL was +2.9% annualized per year, and the HmL coef of DFLVX was +0.6 (2.9 * 0.6 = +1.7%). So you would expect DFLVX to beat the market by +1.7% per year (ignoring for a minute its higher beta). But it only beat it by +0.8% per year. What happened? Well, HmL hides the fact that FF Large Growth actually beat FF Large Value -- all of the positive HmL came from the strong performance of FF Small Value over FF Small Growth. Yet the model assumes a constant return/distribution of HmL, it is unable to detect where HmL was higher (small cap) or lower (large cap) in the marketplace.
Based on a Largecap HmL of -0.2% per year over this stretch, I'd say the negative 3F alpha is misleading and not very useful. DFLVX was supposed to underperform the market by -0.1% per year before fees, not outpace it by +0.8%. Also, DFLVX has a higher beta because it screens out utilities and REITS. To the extent that these stocks go through a stretch where they outperform the market, it will contribute negatively (and randomly) to the portfolio. And over this stretch, that is exactly what happened (REITS and Utilities outpaced the market).
If you instead look at the 00-09 period, where HmL was also positive, but more evenly distributed amongst large and small stocks, then the negative alpha on DFLVX falls to an insignificant -0.01% per month.
This actually doesn't have anything to do with DFA management, it has to do with the presence of (or lack of) a value premium in different segments of the market, and the regression inputs simply taking a "marketwide average". Or model misspecification to put it simply.I think you missed my point. Lets take the numbers below. Is this alpha drift or “lack of brains” from DFA, despite all the 'engineering' (obviously if it was positive the question would be is it ‘DFA brains’ or alpha drift). My take is its probably alpha drift.
DFA large value
May 1993 to Sept 2006, alpha = -0.24/month
Jan 2000 to Sept 2006 alpha = -0.09/month
Lets take 5/93 through 11/09, when the alpha on DFLVX was an insignificant -0.15% per month. FF HmL was +2.9% annualized per year, and the HmL coef of DFLVX was +0.6 (2.9 * 0.6 = +1.7%). So you would expect DFLVX to beat the market by +1.7% per year (ignoring for a minute its higher beta). But it only beat it by +0.8% per year. What happened? Well, HmL hides the fact that FF Large Growth actually beat FF Large Value -- all of the positive HmL came from the strong performance of FF Small Value over FF Small Growth. Yet the model assumes a constant return/distribution of HmL, it is unable to detect where HmL was higher (small cap) or lower (large cap) in the marketplace.
Based on a Largecap HmL of -0.2% per year over this stretch, I'd say the negative 3F alpha is misleading and not very useful. DFLVX was supposed to underperform the market by -0.1% per year before fees, not outpace it by +0.8%. Also, DFLVX has a higher beta because it screens out utilities and REITS. To the extent that these stocks go through a stretch where they outperform the market, it will contribute negatively (and randomly) to the portfolio. And over this stretch, that is exactly what happened (REITS and Utilities outpaced the market).
If you instead look at the 00-09 period, where HmL was also positive, but more evenly distributed amongst large and small stocks, then the negative alpha on DFLVX falls to an insignificant -0.01% per month.
.
the fixer,
(1) On the factor comparison between DFA Marketwide Value and Russell Midcap Value you are also missing the most important factor load comparison over this period – the ‘market’ loading. The market premium was large over this period (the S&P500 annualized return = 18.1%). Average premiums: 1986-99: Mkt-Rf = 12.1, SmB = -2.7, HmL = 0.9, Rf = 5.4.
(2) On alpha links with the HmL decomposition. The FF3F regression does not assume constant return/distribution of HmL. It is simply an OLS parameter calculation of a variable regressed on DFLVX in this case. It is the person who uses these parameters for estimating portfolio expected return that makes the assumptions. IMO the objective is to capture (as precisely as possible) the value premium (as defined historical by FF), which makes alpha estimates important (the closer to zero the better IMO).
How has large value done historically at capturing these premiums, relative to small value? Using data from 7/26 – 12/09 here are the alphas and R^2 from the three factor regressions (and putting aside concerns of simultaneity biases).
So on average, large value has underperformed what the FF factor loads suggest it should had done by about 1.3% per year (-0.11 x 12). That’s simply what the FF regression tells us. Now we can adjust the definition of HmL to simply be LV-LG, and the alpha on the FF large Value become 0.04 with an R^2 of 0.99. So the negative alpha has been eliminated, but the value premium that’s now more accurately captured (with a close to zero alpha) is much smaller than the FF value premium (0.24 vs. 0.40). This is fine of course, but downward adjustments will also need to be made to the size of the value premium used in expected return calculations (assuming the relative value premiums since 1926 between small and large will be similar going forward). Just seems to me that large value is not a very effective/efficient vehicle for capturing target FF value premiums.
Robert
.
the fixer,
(1) On the factor comparison between DFA Marketwide Value and Russell Midcap Value you are also missing the most important factor load comparison over this period – the ‘market’ loading. The market premium was large over this period (the S&P500 annualized return = 18.1%). Average premiums: 1986-99: Mkt-Rf = 12.1, SmB = -2.7, HmL = 0.9, Rf = 5.4.
(2) On alpha links with the HmL decomposition. The FF3F regression does not assume constant return/distribution of HmL. It is simply an OLS parameter calculation of a variable regressed on DFLVX in this case. It is the person who uses these parameters for estimating portfolio expected return that makes the assumptions. IMO the objective is to capture (as precisely as possible) the value premium (as defined historical by FF), which makes alpha estimates important (the closer to zero the better IMO).
How has large value done historically at capturing these premiums, relative to small value? Using data from 7/26 – 12/09 here are the alphas and R^2 from the three factor regressions (and putting aside concerns of simultaneity biases).
Code: Select all
FF3F regressions
Alpha R^2
FF Small Value 0.02 0.99
FF Large Value -0.11 0.97
Robert
.