Best Vanguard value factor exposure?

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Topic Author
hungrywave
Posts: 90
Joined: Tue Apr 09, 2019 7:48 pm

Best Vanguard value factor exposure?

Post by hungrywave » Tue Jun 11, 2019 7:17 pm

Dear BogleBelievers,

Which Vanguard ETF or fund is best for exposure to the value factor?

I am looking to specifically incorporate both value and momentum factors into my portfolio. (Shout out to Larry Swedroe and "Reducing the Risk of Black Swans" - you've made me a believer - at least in momentum, particularly paired with value).

I would like to stick to Vanguard as much as possible because I think it is the least likely to screw me in the long run. The two Vanguard funds in particular I am considering are VBR and VFVA.

- Vanguard Small-Cap Value ETF (VBR): It has been around a long time, has net assets of $31,000M, has a P/B 1.7, and provides exposure to the (now out of favor) small cap factor (median market cap $3.8B).
- Vanguard U.S. Value Factor ETF (VFVA): It is relatively new (established 2018), has net assets of $57M, has a P/B of 1.3, and is more mid-cap centric (median market cap of $7.1B).

Turnover and expense ratios are similar for both (0.07% v 0.13%, 15.5% v 17.6%, VBR v VFVA). Available tax costs are also similar (-0.61% v -0.41%, VBR v VFVA).

VBR is indexed (though interestingly, according to Morningstar, it has drifted from their small cap category to mid cap over the last couple years). In contrast, VFVA is "active", though it says it is "rules-based" so I am not sure whether it has a normal amount of manager risk.

Of note, other potentially relevant Bogleheads posts have included:
Value ETF - VFVA [Vanguard U.S. Value Factor ETF] viewtopic.php?t=260337
Will Vanguard Factor Funds Survive? viewtopic.php?t=273466
Vanguard New Factor Funds Portfolio Statistics viewtopic.php?t=243750
Vanguard US Multifactor Funds Launch viewtopic.php?t=241691
Looking for a value index fund (low-cost) that spans all market caps viewtopic.php?t=255594
New Factor Funds vs Regular SCV tilting viewtopic.php?t=262872

What do you think? What are you doing and why?

:beer

MotoTrojan
Posts: 4514
Joined: Wed Feb 01, 2017 8:39 pm

Re: Best Vanguard value factor exposure?

Post by MotoTrojan » Tue Jun 11, 2019 7:30 pm

I prefer the S&P600 flavor for small at least. Vanguard’s CRSP based VBR is about double the market cap of VIOV (IJS is another equivalent ETF). Most metrics would call it closer to 50/50 mid/small. I like the deeper small premium exposure. Having said that I currently only hold the VBR equivalent in my 401k and no VIOV as I’m allocated differently in Roth right now.

As to more advanced multi factor funds, I’m not sure.

Topic Author
hungrywave
Posts: 90
Joined: Tue Apr 09, 2019 7:48 pm

Re: Best Vanguard value factor exposure?

Post by hungrywave » Tue Jun 11, 2019 7:38 pm

MotoTrojan wrote:
Tue Jun 11, 2019 7:30 pm
I like the deeper small premium exposure.
Thank you, MotoTrojan! Do you believe the value factor has a greater premium within the small cap segment? If so, that would be a reason to choose VBR over VFVA.

fennewaldaj
Posts: 677
Joined: Sun Oct 22, 2017 11:30 pm

Re: Best Vanguard value factor exposure?

Post by fennewaldaj » Tue Jun 11, 2019 7:39 pm

If you are targeting positive momentum in the whole portfolio VFVA is likely not a good choice. So far it has very negative momentum. VFMF (the multifactor fund) might be more what you are looking for. VBR or VIOV have much closer to zero momentum so could be combined with a specific momentum fund if that is your preference.

Topic Author
hungrywave
Posts: 90
Joined: Tue Apr 09, 2019 7:48 pm

Re: Best Vanguard value factor exposure?

Post by hungrywave » Tue Jun 11, 2019 7:43 pm

fennewaldaj wrote:
Tue Jun 11, 2019 7:39 pm
If you are targeting positive momentum in the whole portfolio VFVA is likely not a good choice. So far it has very negative momentum. VFMF (the multifactor fund) might be more what you are looking for. VBR or VIOV have much closer to zero momentum so could be combined with a specific momentum fund if that is your preference.
Thank you, fennewaldaj!

How do you know the estimated factor contributions of different ETFs and funds?

Is it reasonable to accept some negative momentum in VFVA (and some negative value in VFMO) in order to obtain overall positive momentum and value in the portfolio AND be able to rebalance the two?

columbia
Posts: 1556
Joined: Tue Aug 27, 2013 5:30 am

Re: Best Vanguard value factor exposure?

Post by columbia » Tue Jun 11, 2019 7:50 pm

The thread title is about value factor, not small value factor (which doesn’t interest me).

Is there a large cap value factor ETF out there, other than VLUE (which hasn’t managed to keep up with the large cap value index, ie VTV)?

gtwhitegold
Posts: 423
Joined: Fri Sep 21, 2012 1:55 pm

Re: Best Vanguard value factor exposure?

Post by gtwhitegold » Tue Jun 11, 2019 8:32 pm

hungrywave wrote:
Tue Jun 11, 2019 7:43 pm
fennewaldaj wrote:
Tue Jun 11, 2019 7:39 pm
If you are targeting positive momentum in the whole portfolio VFVA is likely not a good choice. So far it has very negative momentum. VFMF (the multifactor fund) might be more what you are looking for. VBR or VIOV have much closer to zero momentum so could be combined with a specific momentum fund if that is your preference.
Thank you, fennewaldaj!

How do you know the estimated factor contributions of different ETFs and funds?

Is it reasonable to accept some negative momentum in VFVA (and some negative value in VFMO) in order to obtain overall positive momentum and value in the portfolio AND be able to rebalance the two?
If you are going to target both, then you I would definitely recommend that you stick with a multi-factor fund that targets what you want. VFMF is a good place to start.

User avatar
grabiner
Advisory Board
Posts: 24366
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: Best Vanguard value factor exposure?

Post by grabiner » Tue Jun 11, 2019 9:16 pm

I use VFVA for the deeper value exposure. It covers both my large-cap and small-cap value needs, because it happens to hold the large/small ratio I want (50/50, counting small as the bottom 20% of the market).

If I wanted more small-cap value, I would probably use RZV (Invesco Small-Cap 600 Pure Value), which is micro-cap deep-value; I used to hold this ETF in my HSA before VFVA was created.
Wiki David Grabiner

fennewaldaj
Posts: 677
Joined: Sun Oct 22, 2017 11:30 pm

Re: Best Vanguard value factor exposure?

Post by fennewaldaj » Tue Jun 11, 2019 9:44 pm

hungrywave wrote:
Tue Jun 11, 2019 7:43 pm
fennewaldaj wrote:
Tue Jun 11, 2019 7:39 pm
If you are targeting positive momentum in the whole portfolio VFVA is likely not a good choice. So far it has very negative momentum. VFMF (the multifactor fund) might be more what you are looking for. VBR or VIOV have much closer to zero momentum so could be combined with a specific momentum fund if that is your preference.
Thank you, fennewaldaj!

How do you know the estimated factor contributions of different ETFs and funds?

Is it reasonable to accept some negative momentum in VFVA (and some negative value in VFMO) in order to obtain overall positive momentum and value in the portfolio AND be able to rebalance the two?
portfolio visualizer is the easiest place to estimate to find factor loads. As far as separate momentum and value fund vs multifactor the consensus seems to be with multifactor though there are some who dissent from that view.

Northern Flicker
Posts: 4212
Joined: Fri Apr 10, 2015 12:29 am
Location: Taking a break from Bogleheads

Re: Best Vanguard value factor exposure?

Post by Northern Flicker » Tue Jun 11, 2019 11:25 pm

It is not possible to provide a good answer without knowing what the rest of the OP’s portfolio looks like.

XacTactX
Posts: 6
Joined: Sat May 11, 2019 5:46 pm

Re: Best Vanguard value factor exposure?

Post by XacTactX » Wed Jun 12, 2019 12:26 am

Link to Portfolio Visualizer Stats

Link to iShares Factor Box

Analysis of Vanguard Quantitative Equity Funds

OP my suggestion would be to pick a single fund that has positive exposure to both value and momentum, do not pick two funds, one with + value and - momentum, then another fund with - value and + momentum. The reason being, the net exposure of your portfolio will be diluted to almost zero, and 1/2 of your portfolio will be cheap companies with negative recent earnings, and other half will be expensive companies with high recent earnings.

Looking at portfolio visualizer, there are two funds that might work for you, VUVLX and VSTCX. There is also VFMF, I didn't include it because it only has one year of earnings data and the factor exposures are not significant yet

Topic Author
hungrywave
Posts: 90
Joined: Tue Apr 09, 2019 7:48 pm

Re: Best Vanguard value factor exposure?

Post by hungrywave » Wed Jun 12, 2019 1:48 am

XacTactX wrote:
Wed Jun 12, 2019 12:26 am
OP my suggestion would be to pick a single fund that has positive exposure to both value and momentum, do not pick two funds, one with + value and - momentum, then another fund with - value and + momentum. The reason being, the net exposure of your portfolio will be diluted to almost zero, and 1/2 of your portfolio will be cheap companies with negative recent earnings, and other half will be expensive companies with high recent earnings.
Thank you, XacTactX!

Perhaps my thinking is flawed but here is how my mental model is currently operating:
- Both value and momentum appear to have produced premiums over long periods.
- Value and momentum are anti-correlated.
- Thus, it would make sense to hold BOTH value and momentum positions - even if their anti correlation decreased any overall portfolio tilt toward either.
- Since value and momentum are anti correlated, they may be particularly valuable during rebalancing.

Based on Simba's backtesting spreadsheet, combining momentum and value factors has an overall result intermediate to the 2 factors (which is what I would expect from diversification).

What is wrong with my line of reasoning? Why not include 2 useful factors in my portfolio even though they are anti-correlated?

klaus14
Posts: 161
Joined: Sun Nov 25, 2018 7:43 pm

Re: Best Vanguard value factor exposure?

Post by klaus14 » Wed Jun 12, 2019 1:59 am

hungrywave wrote:
Wed Jun 12, 2019 1:48 am
XacTactX wrote:
Wed Jun 12, 2019 12:26 am
OP my suggestion would be to pick a single fund that has positive exposure to both value and momentum, do not pick two funds, one with + value and - momentum, then another fund with - value and + momentum. The reason being, the net exposure of your portfolio will be diluted to almost zero, and 1/2 of your portfolio will be cheap companies with negative recent earnings, and other half will be expensive companies with high recent earnings.
Thank you, XacTactX!

Perhaps my thinking is flawed but here is how my mental model is currently operating:
- Both value and momentum appear to have produced premiums over long periods.
- Value and momentum are anti-correlated.
- Thus, it would make sense to hold BOTH value and momentum positions - even if their anti correlation decreased any overall portfolio tilt toward either.
- Since value and momentum are anti correlated, they may be particularly valuable during rebalancing.

Based on Simba's backtesting spreadsheet, combining momentum and value factors has an overall result intermediate to the 2 factors (which is what I would expect from diversification).

What is wrong with my line of reasoning? Why not include 2 useful factors in my portfolio even though they are anti-correlated?
You don't want to end up like below:
Loadings:
Value Fund A: VAL: 0.5, MOM: -0.5
Momentum Fund B: MOM: 0.5, VAL: -0.5
A+B: MOM: 0, VAL: 0

Now you are paying extra fees for basically zero factor exposure.

What you want is:
Multifactor Fund C: VAL: 0.25, MOM:0.25

Multifactor fund achieves this by (basically) ignoring value stocks if they have negative momentum and ignoring momentum stocks if they have negative value loading. they also ignore high volatility and low quality stocks. Remaining list of stocks won't have the highest value or momentum load though.

Topic Author
hungrywave
Posts: 90
Joined: Tue Apr 09, 2019 7:48 pm

Re: Best Vanguard value factor exposure?

Post by hungrywave » Wed Jun 12, 2019 2:06 am

klaus14 wrote:
Wed Jun 12, 2019 1:59 am
What you want is:
Multifactor Fund C: VAL: 0.25, MOM:0.25

Multifactor fund achieves this by (basically) ignoring value stocks if they have negative momentum and ignoring momentum stocks if they have negative value loading. they also ignore high volatility and low quality stocks. Remaining list of stocks won't have the highest value or momentum load though.
That makes sense, klaus14! I think an error in my thinking was that the Simba backtesting factors perfectly cancel when, perhaps when combined, they actually create something like the hypothetical multi factor fund you propose.

:sharebeer

User avatar
vineviz
Posts: 4012
Joined: Tue May 15, 2018 1:55 pm

Re: Best Vanguard value factor exposure?

Post by vineviz » Wed Jun 12, 2019 6:55 am

klaus14 wrote:
Wed Jun 12, 2019 1:59 am
hungrywave wrote:
Wed Jun 12, 2019 1:48 am
XacTactX wrote:
Wed Jun 12, 2019 12:26 am
OP my suggestion would be to pick a single fund that has positive exposure to both value and momentum, do not pick two funds, one with + value and - momentum, then another fund with - value and + momentum. The reason being, the net exposure of your portfolio will be diluted to almost zero, and 1/2 of your portfolio will be cheap companies with negative recent earnings, and other half will be expensive companies with high recent earnings.
Thank you, XacTactX!

Perhaps my thinking is flawed but here is how my mental model is currently operating:
- Both value and momentum appear to have produced premiums over long periods.
- Value and momentum are anti-correlated.
- Thus, it would make sense to hold BOTH value and momentum positions - even if their anti correlation decreased any overall portfolio tilt toward either.
- Since value and momentum are anti correlated, they may be particularly valuable during rebalancing.

Based on Simba's backtesting spreadsheet, combining momentum and value factors has an overall result intermediate to the 2 factors (which is what I would expect from diversification).

What is wrong with my line of reasoning? Why not include 2 useful factors in my portfolio even though they are anti-correlated?
You don't want to end up like below:
Loadings:
Value Fund A: VAL: 0.5, MOM: -0.5
Momentum Fund B: MOM: 0.5, VAL: -0.5
A+B: MOM: 0, VAL: 0

Now you are paying extra fees for basically zero factor exposure.

What you want is:
Multifactor Fund C: VAL: 0.25, MOM:0.25

Multifactor fund achieves this by (basically) ignoring value stocks if they have negative momentum and ignoring momentum stocks if they have negative value loading. they also ignore high volatility and low quality stocks. Remaining list of stocks won't have the highest value or momentum load though.
+1
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

pdavi21
Posts: 1057
Joined: Sat Jan 30, 2016 4:04 pm

Re: Best Vanguard value factor exposure?

Post by pdavi21 » Wed Jun 12, 2019 9:22 am

Small cap value funds are too small to avoid random performance. You better hope S&P 600 continues to be a good index while Russell 2000 continues to be a bad index.

The standard deviation of total returns for small cap value funds (because they have very low market capitalization) is extremely high.

Some inappropriately label this extra risk as boosting expected return. However, it is simply a result of picking a very small slice of the market and is functionally similar (though less risky) than tilting towards one megacap.

Imagine a world in which the entire stock market is companies that are the same size. If you pick 1-2% (whether at random or not), volatility and expected departure from total market returns rises significantly.

If you pick the right 1-2%, backtesters declare a good index. If you pick the wrong 3-4%, backtesters declare a bad index.

Long story short, the standard deviation of total returns for the most popular small value ETFs has made their underperformance (value underperformed over time period), statistically insignificant, while the standard deviation of large cap value ETFs returns was not enough to make their underperformance statistically insignificant (probably because they have a lot of overlap holding at least ~35% of the market). This standard deviation actually rises over time (faster than total return). This suggests it can be expected that performance may further diverge based on random effects.

Value being stronger in small caps is a logical fallacy that combines survivorship bias with random luck.
"We spend a great deal of time studying history, which, let's face it, is mostly the history of stupidity." -Stephen Hawking

pdavi21
Posts: 1057
Joined: Sat Jan 30, 2016 4:04 pm

Re: Best Vanguard value factor exposure?

Post by pdavi21 » Wed Jun 12, 2019 10:54 am

Does anyone actually have some sort of defense (academic, logical, or back-testing) indicating that a multi-factor fund is likely to outperform a mixture two factor funds?

Has anyone actually studied the factors to determine that the best momentum stocks won't be filtered out by a value screen and the best value stocks won't be filtered out by a momentum screen, or that if this wouldn't have occurred, the results were statistically meaningful and not random?

My guess is that you all are assuming, because it worked for small value for 100 years, it is likely to work for value/momentum even though the interaction between any two factors should theoretically be different, unless the factors are independent.

EDIT: If the factors are independent, the multi-fund should theoretically have higher gains if they perform and lower gains if they under-perform because the returns would be multiplied instead of averaged. This would yield much higher returns when the factors do well and much lower returns when they do poorly. If the assumptions are made:
1. Factors behave independently
2. Fund performance is mostly factor driven
3. Factor returns are going to be positive on average

Then the multi-factor fund is the superior choice, I suppose. But those are big assumptions.
"We spend a great deal of time studying history, which, let's face it, is mostly the history of stupidity." -Stephen Hawking

User avatar
Dialectical Investor
Posts: 443
Joined: Mon Dec 03, 2018 11:41 pm

Re: Best Vanguard value factor exposure?

Post by Dialectical Investor » Wed Jun 12, 2019 11:21 am

pdavi21 wrote:
Wed Jun 12, 2019 9:22 am

The standard deviation of total returns for small cap value funds (because they have very low market capitalization) is extremely high.

Some inappropriately label this extra risk as boosting expected return. However, it is simply a result of picking a very small slice of the market and is functionally similar (though less risky) than tilting towards one megacap.

Imagine a world in which the entire stock market is companies that are the same size. If you pick 1-2% (whether at random or not), volatility and expected departure from total market returns rises significantly.
I think this is pretty far off. The standard deviation of a group of small-cap stocks and a single megacap stock should not be expected to be the same simply because they have the same market capitalization. That fact tells you nothing about what to expect regarding their relative standard deviations. Also, the standard deviation of a random sample of 1-2% of a population could indeed provide a reasonable estimate of the standard deviation of a population.

pdavi21
Posts: 1057
Joined: Sat Jan 30, 2016 4:04 pm

Re: Best Vanguard value factor exposure?

Post by pdavi21 » Wed Jun 12, 2019 11:31 am

Dialectical Investor wrote:
Wed Jun 12, 2019 11:21 am
pdavi21 wrote:
Wed Jun 12, 2019 9:22 am

The standard deviation of total returns for small cap value funds (because they have very low market capitalization) is extremely high.

Some inappropriately label this extra risk as boosting expected return. However, it is simply a result of picking a very small slice of the market and is functionally similar (though less risky) than tilting towards one megacap.

Imagine a world in which the entire stock market is companies that are the same size. If you pick 1-2% (whether at random or not), volatility and expected departure from total market returns rises significantly.
I think this is pretty far off. The standard deviation of a group of small-cap stocks and a single megacap stock should not be expected to be the same simply because they have the same market capitalization. That fact tells you nothing about what to expect regarding their relative standard deviations. Also, the standard deviation of a random sample of 1-2% of a population could indeed provide a reasonable estimate of the standard deviation of a population.
That's why I said "(though less risky)". 450 different (EDIT: Management structures/sector exposures/factor exposures/differences in random factors) reduces the risk, but it is functionally equivalent if one assumes a market cap weight portfolio is a logical starting point. If 1-2% provides a reasonable estimate, then why not invest in 100% Johnson & Johnson? or 40 $10B companies? Where is the cut off that magically makes it okay to only invest in 1-2% of the market without having random effects that detract from factor performance?

EDIT: Back-testing tells us that 450-1400 companies representing 1-5% of the US stock market is not enough after just 5-10 years. One was the worst value fund, and one was the best.
"We spend a great deal of time studying history, which, let's face it, is mostly the history of stupidity." -Stephen Hawking

User avatar
Dialectical Investor
Posts: 443
Joined: Mon Dec 03, 2018 11:41 pm

Re: Best Vanguard value factor exposure?

Post by Dialectical Investor » Wed Jun 12, 2019 12:03 pm

pdavi21 wrote:
Wed Jun 12, 2019 11:31 am
Dialectical Investor wrote:
Wed Jun 12, 2019 11:21 am
pdavi21 wrote:
Wed Jun 12, 2019 9:22 am

The standard deviation of total returns for small cap value funds (because they have very low market capitalization) is extremely high.

Some inappropriately label this extra risk as boosting expected return. However, it is simply a result of picking a very small slice of the market and is functionally similar (though less risky) than tilting towards one megacap.

Imagine a world in which the entire stock market is companies that are the same size. If you pick 1-2% (whether at random or not), volatility and expected departure from total market returns rises significantly.
I think this is pretty far off. The standard deviation of a group of small-cap stocks and a single megacap stock should not be expected to be the same simply because they have the same market capitalization. That fact tells you nothing about what to expect regarding their relative standard deviations. Also, the standard deviation of a random sample of 1-2% of a population could indeed provide a reasonable estimate of the standard deviation of a population.
That's why I said "(though less risky)". 450 different (EDIT: Management structures/sector exposures/factor exposures/differences in random factors) reduces the risk, but it is functionally equivalent if one assumes a market cap weight portfolio is a logical starting point. If 1-2% provides a reasonable estimate, then why not invest in 100% Johnson & Johnson? or 40 $10B companies? Where is the cut off that magically makes it okay to only invest in 1-2% of the market without having random effects that detract from factor performance?

EDIT: Back-testing tells us that 450-1400 companies representing 1-4% of the US stock market is not enough after just 5-10 years. One was the worst value fund, and one was the best.
The 1-2% is not in reference to market cap but rather the members of the population. The market cap does not come into play. Your backtesting results are not surprising because the sample is not random.

pdavi21
Posts: 1057
Joined: Sat Jan 30, 2016 4:04 pm

Re: Best Vanguard value factor exposure?

Post by pdavi21 » Wed Jun 12, 2019 12:14 pm

Dialectical Investor wrote:
Wed Jun 12, 2019 12:03 pm
pdavi21 wrote:
Wed Jun 12, 2019 11:31 am
Dialectical Investor wrote:
Wed Jun 12, 2019 11:21 am
pdavi21 wrote:
Wed Jun 12, 2019 9:22 am

The standard deviation of total returns for small cap value funds (because they have very low market capitalization) is extremely high.

Some inappropriately label this extra risk as boosting expected return. However, it is simply a result of picking a very small slice of the market and is functionally similar (though less risky) than tilting towards one megacap.

Imagine a world in which the entire stock market is companies that are the same size. If you pick 1-2% (whether at random or not), volatility and expected departure from total market returns rises significantly.
I think this is pretty far off. The standard deviation of a group of small-cap stocks and a single megacap stock should not be expected to be the same simply because they have the same market capitalization. That fact tells you nothing about what to expect regarding their relative standard deviations. Also, the standard deviation of a random sample of 1-2% of a population could indeed provide a reasonable estimate of the standard deviation of a population.
That's why I said "(though less risky)". 450 different (EDIT: Management structures/sector exposures/factor exposures/differences in random factors) reduces the risk, but it is functionally equivalent if one assumes a market cap weight portfolio is a logical starting point. If 1-2% provides a reasonable estimate, then why not invest in 100% Johnson & Johnson? or 40 $10B companies? Where is the cut off that magically makes it okay to only invest in 1-2% of the market without having random effects that detract from factor performance?

EDIT: Back-testing tells us that 450-1400 companies representing 1-4% of the US stock market is not enough after just 5-10 years. One was the worst value fund, and one was the best.
The 1-2% is not in reference to market cap but rather the members of the population. The market cap does not come into play. Your backtesting results are not surprising because the sample is not random.
The two samples are chosen based on similar criteria. Both indices have median market caps between 1.5 -2.0 billion with a value screen. If the samples were chosen randomly, their performances would diverge more, not less.

EDIT: Don't worry about it too much. It's a fact. Investors took an uncompensated risk by choosing a small cap value fund because performance was more heavily based on fund choice than factor performance. The two most popular small cap value indices (excluding the mid cap ones) tracked by passive funds had the highest difference in performance than any other pair of similar size value funds.

So investor A picks Russell 2000 Value, loses, investor B picks S&P 600 Value, wins. How they would know which one to pick in 2010 is anyone's guess. How they would know picking one or the other could matter immensely, is by looking backwards.
"We spend a great deal of time studying history, which, let's face it, is mostly the history of stupidity." -Stephen Hawking

User avatar
Dialectical Investor
Posts: 443
Joined: Mon Dec 03, 2018 11:41 pm

Re: Best Vanguard value factor exposure?

Post by Dialectical Investor » Wed Jun 12, 2019 12:30 pm

pdavi21 wrote:
Wed Jun 12, 2019 12:14 pm
Dialectical Investor wrote:
Wed Jun 12, 2019 12:03 pm
pdavi21 wrote:
Wed Jun 12, 2019 11:31 am
Dialectical Investor wrote:
Wed Jun 12, 2019 11:21 am
pdavi21 wrote:
Wed Jun 12, 2019 9:22 am

The standard deviation of total returns for small cap value funds (because they have very low market capitalization) is extremely high.

Some inappropriately label this extra risk as boosting expected return. However, it is simply a result of picking a very small slice of the market and is functionally similar (though less risky) than tilting towards one megacap.

Imagine a world in which the entire stock market is companies that are the same size. If you pick 1-2% (whether at random or not), volatility and expected departure from total market returns rises significantly.
I think this is pretty far off. The standard deviation of a group of small-cap stocks and a single megacap stock should not be expected to be the same simply because they have the same market capitalization. That fact tells you nothing about what to expect regarding their relative standard deviations. Also, the standard deviation of a random sample of 1-2% of a population could indeed provide a reasonable estimate of the standard deviation of a population.
That's why I said "(though less risky)". 450 different (EDIT: Management structures/sector exposures/factor exposures/differences in random factors) reduces the risk, but it is functionally equivalent if one assumes a market cap weight portfolio is a logical starting point. If 1-2% provides a reasonable estimate, then why not invest in 100% Johnson & Johnson? or 40 $10B companies? Where is the cut off that magically makes it okay to only invest in 1-2% of the market without having random effects that detract from factor performance?

EDIT: Back-testing tells us that 450-1400 companies representing 1-4% of the US stock market is not enough after just 5-10 years. One was the worst value fund, and one was the best.
The 1-2% is not in reference to market cap but rather the members of the population. The market cap does not come into play. Your backtesting results are not surprising because the sample is not random.
The two samples are chosen based on similar criteria. Both indices have market caps between 1.5 -2.0 billion with a value screen. If the samples were chosen randomly, their performances would diverge more, not less.
I wouldn't be certain of that, however, I was referring to your scenario regarding companies that are the same size and the irrelevance of market cap in general when assessing what proportion of the population you are drawing from. It's like adding the net worth of people in a population, and thinking a large number of people with a small amount of wealth would behave like a single person with the same amount of wealth. The amount of wealth is just not relevant to how to go about sampling and what you can estimate about the population using the sample.

To explain the difference between two funds that are supposed to have the same criteria is a separate issue and would require further analysis that might not have anything to do with sampling. However, I myself would not use a multi-factor fund in part because of the ambiguous strategy and rules as mentioned in this thread and would instead omit one or more factors.

pdavi21
Posts: 1057
Joined: Sat Jan 30, 2016 4:04 pm

Re: Best Vanguard value factor exposure?

Post by pdavi21 » Wed Jun 12, 2019 12:34 pm

Dialectical Investor wrote:
Wed Jun 12, 2019 12:30 pm
pdavi21 wrote:
Wed Jun 12, 2019 12:14 pm
Dialectical Investor wrote:
Wed Jun 12, 2019 12:03 pm
pdavi21 wrote:
Wed Jun 12, 2019 11:31 am
Dialectical Investor wrote:
Wed Jun 12, 2019 11:21 am


I think this is pretty far off. The standard deviation of a group of small-cap stocks and a single megacap stock should not be expected to be the same simply because they have the same market capitalization. That fact tells you nothing about what to expect regarding their relative standard deviations. Also, the standard deviation of a random sample of 1-2% of a population could indeed provide a reasonable estimate of the standard deviation of a population.
That's why I said "(though less risky)". 450 different (EDIT: Management structures/sector exposures/factor exposures/differences in random factors) reduces the risk, but it is functionally equivalent if one assumes a market cap weight portfolio is a logical starting point. If 1-2% provides a reasonable estimate, then why not invest in 100% Johnson & Johnson? or 40 $10B companies? Where is the cut off that magically makes it okay to only invest in 1-2% of the market without having random effects that detract from factor performance?

EDIT: Back-testing tells us that 450-1400 companies representing 1-4% of the US stock market is not enough after just 5-10 years. One was the worst value fund, and one was the best.
The 1-2% is not in reference to market cap but rather the members of the population. The market cap does not come into play. Your backtesting results are not surprising because the sample is not random.
The two samples are chosen based on similar criteria. Both indices have market caps between 1.5 -2.0 billion with a value screen. If the samples were chosen randomly, their performances would diverge more, not less.
I wouldn't be certain of that, however, I was referring to your scenario regarding companies that are the same size and the irrelevance of market cap in general when assessing what proportion of the population you are drawing from. It's like adding the net worth of people in a population, and thinking a large number of people with a small amount of wealth would behave like a single person with the same amount of wealth. The amount of wealth is just not relevant to how to go about sampling and what you can estimate about the population using the sample.

To explain the difference between two funds that are supposed to have the same criteria is a separate issue and would require further analysis that might not have anything to do with sampling. However, I myself would not use a multi-factor fund in part because of the ambiguous strategy and rules as mentioned in this thread and would instead omit one or more factors.
You are living in a world where the logical starting point is an equal weight Index. Otherwise, you would understand that picking 50 large companies gets you closer to total market return than picking 2000 small companies (Because total market return is market cap weighted) (EDIT: I'm assuming they would be market cap weighted, while you must be assuming they would be equal weight which invalidates this example).

Really picking any 50 large companies will likely yield similar results while picking any 2000 small companies yields differing results. Because when those companies are weighted by market cap, they closely match each other, while the 2000 companies are likely to be different sets of companies.
"We spend a great deal of time studying history, which, let's face it, is mostly the history of stupidity." -Stephen Hawking

User avatar
Dialectical Investor
Posts: 443
Joined: Mon Dec 03, 2018 11:41 pm

Re: Best Vanguard value factor exposure?

Post by Dialectical Investor » Wed Jun 12, 2019 12:42 pm

pdavi21 wrote:
Wed Jun 12, 2019 12:34 pm
Dialectical Investor wrote:
Wed Jun 12, 2019 12:30 pm
pdavi21 wrote:
Wed Jun 12, 2019 12:14 pm
Dialectical Investor wrote:
Wed Jun 12, 2019 12:03 pm
pdavi21 wrote:
Wed Jun 12, 2019 11:31 am


That's why I said "(though less risky)". 450 different (EDIT: Management structures/sector exposures/factor exposures/differences in random factors) reduces the risk, but it is functionally equivalent if one assumes a market cap weight portfolio is a logical starting point. If 1-2% provides a reasonable estimate, then why not invest in 100% Johnson & Johnson? or 40 $10B companies? Where is the cut off that magically makes it okay to only invest in 1-2% of the market without having random effects that detract from factor performance?

EDIT: Back-testing tells us that 450-1400 companies representing 1-4% of the US stock market is not enough after just 5-10 years. One was the worst value fund, and one was the best.
The 1-2% is not in reference to market cap but rather the members of the population. The market cap does not come into play. Your backtesting results are not surprising because the sample is not random.
The two samples are chosen based on similar criteria. Both indices have market caps between 1.5 -2.0 billion with a value screen. If the samples were chosen randomly, their performances would diverge more, not less.
I wouldn't be certain of that, however, I was referring to your scenario regarding companies that are the same size and the irrelevance of market cap in general when assessing what proportion of the population you are drawing from. It's like adding the net worth of people in a population, and thinking a large number of people with a small amount of wealth would behave like a single person with the same amount of wealth. The amount of wealth is just not relevant to how to go about sampling and what you can estimate about the population using the sample.

To explain the difference between two funds that are supposed to have the same criteria is a separate issue and would require further analysis that might not have anything to do with sampling. However, I myself would not use a multi-factor fund in part because of the ambiguous strategy and rules as mentioned in this thread and would instead omit one or more factors.
You are living in a world where the logical starting point is an equal weight Index. Otherwise, you would understand that picking 50 large companies gets you closer to total market return than picking 2000 small companies.
That is not my world; it is the world you created in the post I responded to, which, I believe contained some statistically incorrect assumptions. But it seems we are talking past one another, and I've lost your main point.

pdavi21
Posts: 1057
Joined: Sat Jan 30, 2016 4:04 pm

Re: Best Vanguard value factor exposure?

Post by pdavi21 » Wed Jun 12, 2019 12:57 pm

Dialectical Investor wrote:
Wed Jun 12, 2019 12:42 pm
pdavi21 wrote:
Wed Jun 12, 2019 12:34 pm
Dialectical Investor wrote:
Wed Jun 12, 2019 12:30 pm
pdavi21 wrote:
Wed Jun 12, 2019 12:14 pm
Dialectical Investor wrote:
Wed Jun 12, 2019 12:03 pm


The 1-2% is not in reference to market cap but rather the members of the population. The market cap does not come into play. Your backtesting results are not surprising because the sample is not random.
The two samples are chosen based on similar criteria. Both indices have market caps between 1.5 -2.0 billion with a value screen. If the samples were chosen randomly, their performances would diverge more, not less.
I wouldn't be certain of that, however, I was referring to your scenario regarding companies that are the same size and the irrelevance of market cap in general when assessing what proportion of the population you are drawing from. It's like adding the net worth of people in a population, and thinking a large number of people with a small amount of wealth would behave like a single person with the same amount of wealth. The amount of wealth is just not relevant to how to go about sampling and what you can estimate about the population using the sample.

To explain the difference between two funds that are supposed to have the same criteria is a separate issue and would require further analysis that might not have anything to do with sampling. However, I myself would not use a multi-factor fund in part because of the ambiguous strategy and rules as mentioned in this thread and would instead omit one or more factors.
You are living in a world where the logical starting point is an equal weight Index. Otherwise, you would understand that picking 50 large companies gets you closer to total market return than picking 2000 small companies.
That is not my world; it is the world you created in the post I responded to, which, I believe contained some statistically incorrect assumptions. But it seems we are talking past one another, and I've lost your main point.
I did not intend to suggest that investing in 450 small companies is the same as investing in one large company. That's why I added the caveat that it would be less risky. You seemed to take offense to that one statement, but my main point was that picking a small sample of the US stock market introduces an uncompensated risk that is often misinterpreted as "the higher risk-higher expected return of small cap value" and may explain (after survivor-ship bias) "the out-performance of small cap value". One could dubiously suggest that a "Microsoft" tilt introduced the "Microsoft" factor which outperformed, and if it is assumed that a market cap weighted index is logical, they would be behaving just as logically as a person holding 450 small companies. Of course, we know that that is not true-that different companies in different industries add some form of valuable diversification, but it is not far off...and indeed, if there were "Microsoft" funds, rest assured, the performances would all be quite similar, while small cap value funds continued to vary in performance.
"We spend a great deal of time studying history, which, let's face it, is mostly the history of stupidity." -Stephen Hawking

User avatar
Dialectical Investor
Posts: 443
Joined: Mon Dec 03, 2018 11:41 pm

Re: Best Vanguard value factor exposure?

Post by Dialectical Investor » Wed Jun 12, 2019 1:24 pm

pdavi21 wrote:
Wed Jun 12, 2019 12:57 pm

I did not intend to suggest that investing in 450 small companies is the same as investing in one large company. That's why I added the caveat that it would be less risky. You seemed to take offense to that one statement, but my main point was that picking a small sample of the US stock market introduces an uncompensated risk that is often misinterpreted as "the higher risk-higher expected return of small cap value" and may explain (after survivor-ship bias) "the out-performance of small cap value". One could dubiously suggest that a "Microsoft" tilt introduced the "Microsoft" factor which outperformed, and if it is assumed that a market cap weighted index is logical, they would be behaving just as logically as a person holding 450 small companies. Of course, we know that that is not true-that different companies in different industries add some form of valuable diversification, but it is not far off...and indeed, if there were "Microsoft" funds, rest assured, the performances would all be quite similar, while small cap value funds continued to vary in performance.
A single stock cannot be a factor. The portfolios need to be diversified. How diversified, I don't know, but one stock is very far off. So there cannot be a Microsoft factor, and it would not be logical for a person to hold one stock and think they are behaving as logically as a person holding 450 small stocks. Some of these definitions seem a little "soft," but I believe it is by definition that a factor has had compensated risk. If the risk has not been compensated, it has not been a factor. So what you're really saying, it seems, is that SMB is not a factor. You could have just said that. :happy

maj
Posts: 425
Joined: Wed Jun 25, 2008 2:58 pm

Re: Best Vanguard value factor exposure?

Post by maj » Wed Jun 12, 2019 1:58 pm

Value factor? The topic line question.

Ever think of Vhdyx - vym?

peace

pdavi21
Posts: 1057
Joined: Sat Jan 30, 2016 4:04 pm

Re: Best Vanguard value factor exposure?

Post by pdavi21 » Wed Jun 12, 2019 3:10 pm

Dialectical Investor wrote:
Wed Jun 12, 2019 1:24 pm
pdavi21 wrote:
Wed Jun 12, 2019 12:57 pm

I did not intend to suggest that investing in 450 small companies is the same as investing in one large company. That's why I added the caveat that it would be less risky. You seemed to take offense to that one statement, but my main point was that picking a small sample of the US stock market introduces an uncompensated risk that is often misinterpreted as "the higher risk-higher expected return of small cap value" and may explain (after survivor-ship bias) "the out-performance of small cap value". One could dubiously suggest that a "Microsoft" tilt introduced the "Microsoft" factor which outperformed, and if it is assumed that a market cap weighted index is logical, they would be behaving just as logically as a person holding 450 small companies. Of course, we know that that is not true-that different companies in different industries add some form of valuable diversification, but it is not far off...and indeed, if there were "Microsoft" funds, rest assured, the performances would all be quite similar, while small cap value funds continued to vary in performance.
A single stock cannot be a factor. The portfolios need to be diversified. How diversified, I don't know, but one stock is very far off. So there cannot be a Microsoft factor, and it would not be logical for a person to hold one stock and think they are behaving as logically as a person holding 450 small stocks. Some of these definitions seem a little "soft," but I believe it is by definition that a factor has had compensated risk. If the risk has not been compensated, it has not been a factor. So what you're really saying, it seems, is that SMB is not a factor. You could have just said that. :happy
No I am not SMB does not exist, although I am somewhat factor agnostic. I am saying HML and SMB (if they exist) do not have enough of an impact on "small value" mutual funds to overcome the random (or misunderstood) performance discrepancy introduced when choosing a very small slice of the market.
"We spend a great deal of time studying history, which, let's face it, is mostly the history of stupidity." -Stephen Hawking

User avatar
vineviz
Posts: 4012
Joined: Tue May 15, 2018 1:55 pm

Re: Best Vanguard value factor exposure?

Post by vineviz » Wed Jun 12, 2019 3:32 pm

pdavi21 wrote:
Wed Jun 12, 2019 3:10 pm
I am saying HML and SMB (if they exist) do not have enough of an impact on "small value" mutual funds to overcome the random (or misunderstood) performance discrepancy introduced when choosing a very small slice of the market.
This is a claim with enough specificity to test, and it has been repeatedly tested and is demonstrably untrue.

The performance of small cap value funds differs from a random fund by more than is plausibly attributable to random chance.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

pdavi21
Posts: 1057
Joined: Sat Jan 30, 2016 4:04 pm

Re: Best Vanguard value factor exposure?

Post by pdavi21 » Wed Jun 12, 2019 3:41 pm

vineviz wrote:
Wed Jun 12, 2019 3:32 pm
pdavi21 wrote:
Wed Jun 12, 2019 3:10 pm
I am saying HML and SMB (if they exist) do not have enough of an impact on "small value" mutual funds to overcome the random (or misunderstood) performance discrepancy introduced when choosing a very small slice of the market.
This is a claim with enough specificity to test, and it has been repeatedly tested and is demonstrably untrue.

The performance of small cap value funds differs from a random fund by more than is plausibly attributable to random chance.
Not what I am saying, not even remotely close. The fund isn't random. The performance is random (or misunderstood) to the level that a group of small cap value funds have so much variation in performance, that that variation eclipses the effect of the targeted factors on their performance.

EDIT: And it's impossible that it has been tested because what passive small value indexes have there been for more than a few decades? SP 600 Value, Russell 2000 Value, and?

Perhaps you can give me two very commonly used small cap value mutual funds or ETFs that have followed different indices over their lifetime and have had nearly identical performance (EDIT: Long term) that adequately distinguishes itself from the total market?

EDIT: You could easily by starting 2000, but probably not any other time. Also, clearly the short term interval correlations are very high, so if you are trying to disprove my assertion based on short term interval correlations, you will be falling on deaf ears. Index risk seems to manifest over longer term intervals.
"We spend a great deal of time studying history, which, let's face it, is mostly the history of stupidity." -Stephen Hawking

cheezit
Posts: 196
Joined: Sat Jul 14, 2018 7:28 pm

Re: Best Vanguard value factor exposure?

Post by cheezit » Wed Jun 12, 2019 4:06 pm

pdavi21 wrote:
Wed Jun 12, 2019 3:41 pm

Perhaps you can give me two very commonly used small cap value mutual funds or ETFs that have followed different indices over their lifetime and have had nearly identical performance (EDIT: Long term) that adequately distinguishes itself from the total market?
Here you go

pdavi21
Posts: 1057
Joined: Sat Jan 30, 2016 4:04 pm

Re: Best Vanguard value factor exposure?

Post by pdavi21 » Wed Jun 12, 2019 4:17 pm

cheezit wrote:
Wed Jun 12, 2019 4:06 pm
pdavi21 wrote:
Wed Jun 12, 2019 3:41 pm

Perhaps you can give me two very commonly used small cap value mutual funds or ETFs that have followed different indices over their lifetime and have had nearly identical performance (EDIT: Long term) that adequately distinguishes itself from the total market?
Here you go
That's not close enough to be statistically relevant and DFSVX is actively managed with a high enough expense ratio that I don't think it should really be considered anyway.

EDIT: Try either:
1 Adding a small cap blend fund and a large cap value fund. While large value under-performed, and small blend greatly outperformed, small value crushed everything. Why if value is stronger in small value?
2. Start in 2003 or later. I would say 1990's or earlier, but there was only DFA before 1998, I believe.
"We spend a great deal of time studying history, which, let's face it, is mostly the history of stupidity." -Stephen Hawking

User avatar
vineviz
Posts: 4012
Joined: Tue May 15, 2018 1:55 pm

Re: Best Vanguard value factor exposure?

Post by vineviz » Wed Jun 12, 2019 6:34 pm

pdavi21 wrote:
Wed Jun 12, 2019 3:41 pm
vineviz wrote:
Wed Jun 12, 2019 3:32 pm
pdavi21 wrote:
Wed Jun 12, 2019 3:10 pm
I am saying HML and SMB (if they exist) do not have enough of an impact on "small value" mutual funds to overcome the random (or misunderstood) performance discrepancy introduced when choosing a very small slice of the market.
This is a claim with enough specificity to test, and it has been repeatedly tested and is demonstrably untrue.

The performance of small cap value funds differs from a random fund by more than is plausibly attributable to random chance.
Not what I am saying, not even remotely close. The fund isn't random. The performance is random (or misunderstood) to the level that a group of small cap value funds have so much variation in performance, that that variation eclipses the effect of the targeted factors on their performance.
This is the part of your claim that I'm addressing. It's possible, even easy, to test whether that performance is random or not.

And it's not.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

User avatar
vineviz
Posts: 4012
Joined: Tue May 15, 2018 1:55 pm

Re: Best Vanguard value factor exposure?

Post by vineviz » Wed Jun 12, 2019 7:06 pm

pdavi21 wrote:
Wed Jun 12, 2019 12:34 pm
You are living in a world where the logical starting point is an equal weight Index. Otherwise, you would understand that picking 50 large companies gets you closer to total market return than picking 2000 small companies (Because total market return is market cap weighted) (EDIT: I'm assuming they would be market cap weighted, while you must be assuming they would be equal weight which invalidates this example).

Really picking any 50 large companies will likely yield similar results while picking any 2000 small companies yields differing results. Because when those companies are weighted by market cap, they closely match each other, while the 2000 companies are likely to be different sets of companies.
It's really hard to pin down what your trying to establish here, in part because the goalpost seems to be shifting and in part because language isn't very precise.

There are fairly straightforward methods for establishing how many samples you need to draw from a population in order to get statistically significant results, given a desired margin of error and confidence interval.

Assuming a 5% margin of error and 95% CI is desired, you'd need to sample 218 out of the top 500 firms and 323 out of the next 2,000 firms in order to put equal-weight portfolios on the same footing. If the portfolios are market cap-weighted then I suspect those numbers would both be a little smaller, but not by much.

But much of this depends on how the definitions of large cap and small cap are defined. If for small caps we're taking about the S&P 600 or the Russell 2000, then 200 to 300 random stocks from those universes would be track the index incredibly tightly.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

gtwhitegold
Posts: 423
Joined: Fri Sep 21, 2012 1:55 pm

Re: Best Vanguard value factor exposure?

Post by gtwhitegold » Thu Jun 13, 2019 5:31 am

klaus14 wrote:
Wed Jun 12, 2019 1:59 am
hungrywave wrote:
Wed Jun 12, 2019 1:48 am
XacTactX wrote:
Wed Jun 12, 2019 12:26 am
OP my suggestion would be to pick a single fund that has positive exposure to both value and momentum, do not pick two funds, one with + value and - momentum, then another fund with - value and + momentum. The reason being, the net exposure of your portfolio will be diluted to almost zero, and 1/2 of your portfolio will be cheap companies with negative recent earnings, and other half will be expensive companies with high recent earnings.
Thank you, XacTactX!

Perhaps my thinking is flawed but here is how my mental model is currently operating:
- Both value and momentum appear to have produced premiums over long periods.
- Value and momentum are anti-correlated.
- Thus, it would make sense to hold BOTH value and momentum positions - even if their anti correlation decreased any overall portfolio tilt toward either.
- Since value and momentum are anti correlated, they may be particularly valuable during rebalancing.

Based on Simba's backtesting spreadsheet, combining momentum and value factors has an overall result intermediate to the 2 factors (which is what I would expect from diversification).

What is wrong with my line of reasoning? Why not include 2 useful factors in my portfolio even though they are anti-correlated?
You don't want to end up like below:
Loadings:
Value Fund A: VAL: 0.5, MOM: -0.5
Momentum Fund B: MOM: 0.5, VAL: -0.5
A+B: MOM: 0, VAL: 0

Now you are paying extra fees for basically zero factor exposure.

What you want is:
Multifactor Fund C: VAL: 0.25, MOM:0.25

Multifactor fund achieves this by (basically) ignoring value stocks if they have negative momentum and ignoring momentum stocks if they have negative value loading. they also ignore high volatility and low quality stocks. Remaining list of stocks won't have the highest value or momentum load though.
Not necessarily. Actually, in most cases this definitely isn't the case. For example, with both the AQR Multi-Style Funds and First Trust's AlphaDEX funds, they rely on a composite score where the highest scoring stocks are selected. I'm pretty sure that iShares and Vanguard both use similar methodology for stock selection. So, stocks that have a high measure for the Value factor and a low measure for the Momentum factor could still be included as long as it meets the overall threshold for inclusion.

Also, index providers can equal weight the factor sleeves, which is what J. P. Morgan and Goldman Sachs do for their factor funds.

Vanguard and iShares are the two providers that I know of that screen for volatility in their multi-factor products, and Vanguard is the only one that sounds like the methodology that you mentioned.

pdavi21
Posts: 1057
Joined: Sat Jan 30, 2016 4:04 pm

Re: Best Vanguard value factor exposure?

Post by pdavi21 » Thu Jun 13, 2019 9:07 am

vineviz wrote:
Wed Jun 12, 2019 6:34 pm
pdavi21 wrote:
Wed Jun 12, 2019 3:41 pm
vineviz wrote:
Wed Jun 12, 2019 3:32 pm
pdavi21 wrote:
Wed Jun 12, 2019 3:10 pm
I am saying HML and SMB (if they exist) do not have enough of an impact on "small value" mutual funds to overcome the random (or misunderstood) performance discrepancy introduced when choosing a very small slice of the market.
This is a claim with enough specificity to test, and it has been repeatedly tested and is demonstrably untrue.

The performance of small cap value funds differs from a random fund by more than is plausibly attributable to random chance.
Not what I am saying, not even remotely close. The fund isn't random. The performance is random (or misunderstood) to the level that a group of small cap value funds have so much variation in performance, that that variation eclipses the effect of the targeted factors on their performance.
This is the part of your claim that I'm addressing. It's possible, even easy, to test whether that performance is random or not.

And it's not.
If the difference in performance is explained by factors not known before, it is essentially random to the investor. Much like factors found recently will not be able to fully explain the performance of a fund of few hundred small securities.

I get to excited and often exaggerate my claims, but no one can deny that two large value funds track much closer than any two small value funds. It's fact. No one can deny that total market funds track closer than any other factor based subset of funds.

This is an uncompensated (and random until explained) risk. When the investor takes this risk in 2003, one small value fund does well and one does poorly. Even if they take in 1998 or August 2000, the discrepancies are notable.

For you to claim that, using the factors known at the time, or even factors known today, that difference in performance between two small value funds is less statistically significant than the factor based performance is irresponsible with a small data set. Obviously, small funds did well in 2000-2003 (some did better than others), large value did decent, and small value did very well. But is there really definitive proof that this performance can be attributed ONLY to SMB and HML (EDIT: and Beta-Oops) and not other factors?

EDIT: I should ask if there is evidence that the performance can be attributed mostly (with confidence) to the fund's goals to generally maximize SMB, HML, and Beta. If the fund fails to do so, the investor is randomly SOL as well.

EDIT: One way to prove the performance can be random is to make several total market funds that start in the middle or bottom of the index. A total market fund that doesn't include the top 5, 20, 50 holdings, would probably have very different performance than one that does. In small cap funds this could be lower because they have smaller top ten weightings typically. However, they tend to start in very different places with different criteria. This is avoidable in total and large indices, but not small or mid cap indices. Unless factors are determining the top 5, 20, 50 securities' performance, the results are arbitrary and random. This would be an even larger problem for funds that deviate from market cap weight. As soon as they do that, some factor, or random effect, is going to take over.
"We spend a great deal of time studying history, which, let's face it, is mostly the history of stupidity." -Stephen Hawking

User avatar
vineviz
Posts: 4012
Joined: Tue May 15, 2018 1:55 pm

Re: Best Vanguard value factor exposure?

Post by vineviz » Thu Jun 13, 2019 9:34 am

pdavi21 wrote:
Thu Jun 13, 2019 9:07 am
I get to excited and often exaggerate my claims, but no one can deny that two large value funds track much closer than any two small value funds. It's fact.
Track WHAT much closer?

If you mean track each other more closely, the only thing that would cause that would be to use improper sampling.

If you mean track the broad market more costly, then obviously the large cap funds will do so because the broad market consists mostly of large cap funds.
pdavi21 wrote:
Thu Jun 13, 2019 9:07 am
But is there really definitive proof that this performance can be attributed ONLY to SMB and HML (EDIT: and Beta-Oops) and not other factors?
Yes. There is.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

pdavi21
Posts: 1057
Joined: Sat Jan 30, 2016 4:04 pm

Re: Best Vanguard value factor exposure?

Post by pdavi21 » Thu Jun 13, 2019 9:55 am

vineviz wrote:
Thu Jun 13, 2019 9:34 am
pdavi21 wrote:
Thu Jun 13, 2019 9:07 am
I get to excited and often exaggerate my claims, but no one can deny that two large value funds track much closer than any two small value funds. It's fact.
Track WHAT much closer?

If you mean track each other more closely, the only thing that would cause that would be to use improper sampling.

If you mean track the broad market more costly, then obviously the large cap funds will do so because the broad market consists mostly of large cap funds.
pdavi21 wrote:
Thu Jun 13, 2019 9:07 am
But is there really definitive proof that this performance can be attributed ONLY to SMB and HML (EDIT: and Beta-Oops) and not other factors?
Yes. There is.
I mean track each other much closer to the point that a funds targeted factors can (with confidence) explain its performance. If you are suggesting that that is an arbitrary target, I agree with you.

If you are suggesting that it is not a risk for an investor, I disagree. Any fund/factor can go in any direction (including total market/beta). However, as arbitrary as it may be to track performance against a beta fund, or a HML+Beta Fund, or a SMB + HML + Beta targeting fund, the overwheliming majority of investors will be invested in market beta (in aggregate) for their stock holdings. The majority of small value investors will be invested in some form of SMB+HML+Beta targeting funds in agggregate. So the risk that you underperform or outperform peers on the largest and smallest scales is very real, considering you are already taking risks that factors underperform. When you target just Beta, this uncompensated risk disappears. As does the, believed to be compensated risk, of SMB and HML. When you target SMB, this uncompensated risk increases greatly, and when you target HML, it increases slightly. So one needs to take into account that there are uncompensated risks of underperform stock investors in aggregate (even if those investors are receiving arbitrary performance from and arbitrary target).
"We spend a great deal of time studying history, which, let's face it, is mostly the history of stupidity." -Stephen Hawking

Angst
Posts: 1968
Joined: Sat Jun 09, 2007 11:31 am

Re: Best Vanguard value factor exposure?

Post by Angst » Thu Jun 13, 2019 10:02 am

vineviz wrote:
Thu Jun 13, 2019 9:34 am
pdavi21 wrote:
Thu Jun 13, 2019 9:07 am
I get to excited and often exaggerate my claims, but no one can deny that two large value funds track much closer than any two small value funds. It's fact.
Track WHAT much closer?

If you mean track each other more closely, the only thing that would cause that would be to use improper sampling.

If you mean track the broad market more costly, then obviously the large cap funds will do so because the broad market consists mostly of large cap funds.
pdavi21 wrote:
Thu Jun 13, 2019 9:07 am
But is there really definitive proof that this performance can be attributed ONLY to SMB and HML (EDIT: and Beta-Oops) and not other factors?
Yes. There is.
I really don't understand the need for the screaming CAPS here. Based on what pdavi21 has been saying it seemed self evident to me he was referring to the "two large value funds" behaving more similar to each other than the "any two small value funds". This link is always worth keeping in mind:

https://www.bogleheads.org/forum/rules#section-2

User avatar
vineviz
Posts: 4012
Joined: Tue May 15, 2018 1:55 pm

Re: Best Vanguard value factor exposure?

Post by vineviz » Thu Jun 13, 2019 10:37 am

pdavi21 wrote:
Thu Jun 13, 2019 9:55 am
I mean track each other much closer to the point that a funds targeted factors can (with confidence) explain its performance. If you are suggesting that that is an arbitrary target, I agree with you.
I mean this respectfully, but it feels to me that you are still struggling to ask a question that you can't quite articulate.

If there is some universe of stocks, it's pretty straightforward to know how likely a portfolio containing some of those stocks will match either an index replicating the entire universe of those stocks OR another portfolio containing some of those stocks.

Once you decide on how you define "close" and how sure you want to be, it's just a matter of calculating a sample size. And that sample size isn't going to depend in any way on the market cap of those stocks.

pdavi21 wrote:
Thu Jun 13, 2019 9:55 am
So the risk that you underperform or outperform peers on the largest and smallest scales is very real, considering you are already taking risks that factors underperform. When you target just Beta, this uncompensated risk disappears. As does the, believed to be compensated risk, of SMB and HML. When you target SMB, this uncompensated risk increases greatly, and when you target HML, it increases slightly. So one needs to take into account that there are uncompensated risks of underperform stock investors in aggregate (even if those investors are receiving arbitrary performance from and arbitrary target).
Again, it's not entirely clear to me what you are asking.

Increasing the number of stocks in a portfolio generally reduces the amount of idiosyncratic (i.e. stock-specific) risk, and it doesn't actually take very many stocks to achieve a near-total reduction of this sort. This goes back to the sample size calculation I mentioned before, and it doesn't vary in any appreciable way on the market cap of the stocks in question.

But SMB and HML are NOT idiosyncratic risks: they are systematic risks. They can be avoided (by having zero exposure to the in the portfolio), but they cannot be diversified away. And because they are systematic risks, they are compensated.

So there is no sensible way for me to interpret this sentence: "when you target SMB, this uncompensated risk increases greatly". Targeting SMB does not, per se, affect uncompensated risk. Having an inadequately small portfolio of stocks that target SMB does, but then again the same would be true of RMF.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

fennewaldaj
Posts: 677
Joined: Sun Oct 22, 2017 11:30 pm

Re: Best Vanguard value factor exposure?

Post by fennewaldaj » Fri Jun 14, 2019 12:01 am

vineviz wrote:
Thu Jun 13, 2019 10:37 am


So there is no sensible way for me to interpret this sentence: "when you target SMB, this uncompensated risk increases greatly". Targeting SMB does not, per se, affect uncompensated risk. Having an inadequately small portfolio of stocks that target SMB does, but then again the same would be true of RMF.
I think what he is getting at is it is unclear to him how he was supposed to pick the S+P 600 value as the best choice in the year 2000. And how is he supposed to know that it is the best choice going forward. And picking the S+P 600 value over the R2k value made a big difference where as the difference between large value indexes is less important.

comeinvest
Posts: 198
Joined: Mon Mar 12, 2012 6:57 pm

Re: Best Vanguard value factor exposure?

Post by comeinvest » Fri Jun 14, 2019 1:39 am

vineviz wrote:
Wed Jun 12, 2019 7:06 pm
pdavi21 wrote:
Wed Jun 12, 2019 12:34 pm
You are living in a world where the logical starting point is an equal weight Index. Otherwise, you would understand that picking 50 large companies gets you closer to total market return than picking 2000 small companies (Because total market return is market cap weighted) (EDIT: I'm assuming they would be market cap weighted, while you must be assuming they would be equal weight which invalidates this example).

Really picking any 50 large companies will likely yield similar results while picking any 2000 small companies yields differing results. Because when those companies are weighted by market cap, they closely match each other, while the 2000 companies are likely to be different sets of companies.
It's really hard to pin down what your trying to establish here, in part because the goalpost seems to be shifting and in part because language isn't very precise.

There are fairly straightforward methods for establishing how many samples you need to draw from a population in order to get statistically significant results, given a desired margin of error and confidence interval.

Assuming a 5% margin of error and 95% CI is desired, you'd need to sample 218 out of the top 500 firms and 323 out of the next 2,000 firms in order to put equal-weight portfolios on the same footing. If the portfolios are market cap-weighted then I suspect those numbers would both be a little smaller, but not by much.

But much of this depends on how the definitions of large cap and small cap are defined. If for small caps we're taking about the S&P 600 or the Russell 2000, then 200 to 300 random stocks from those universes would be track the index incredibly tightly.
I'm curious what is the source of your numbers (218 out of 500 and 323 out of 2000)?
Also, what timeframe do those numbers refer to? 1 year? I believe the longer the timeframes, much less stocks are needed to achieve very small shortfall risk vs. an index. Research has shown (to make a long story short) that the performance of any one stock during a chain of timeframes (e.g. 1 year periods) are not very correlated, which means that the longer the total time, the more largely uncorrelated samples of (stock; time period) samples we have, and the less likely a shortfall / deviation from the index.

I believe the question of portfolio size (number of stocks in a portfolio) and the question of why the S&P 600 and Russell 2000 value indexes had greatly diverging performance were mixed together here, but the divergence probably came from the diverging behavior of the different market segments that these indexes target, not the sample size. As as matter of fact, there are many examples where slightly different definitions of "value" or other factors resulted in greatly diverging results that I have seen no conclusive explanations for, for example price/book vs. price/earnings vs. dividend yield, and that I attribute simply to the risk of factors not performing as expected when attempting factor investing; but hopefully over larger periods and investing in several different styles, the differences even out.

User avatar
vineviz
Posts: 4012
Joined: Tue May 15, 2018 1:55 pm

Re: Best Vanguard value factor exposure?

Post by vineviz » Fri Jun 14, 2019 7:04 am

fennewaldaj wrote:
Fri Jun 14, 2019 12:01 am
vineviz wrote:
Thu Jun 13, 2019 10:37 am


So there is no sensible way for me to interpret this sentence: "when you target SMB, this uncompensated risk increases greatly". Targeting SMB does not, per se, affect uncompensated risk. Having an inadequately small portfolio of stocks that target SMB does, but then again the same would be true of RMF.
I think what he is getting at is it is unclear to him how he was supposed to pick the S+P 600 value as the best choice in the year 2000. And how is he supposed to know that it is the best choice going forward. And picking the S+P 600 value over the R2k value made a big difference where as the difference between large value indexes is less important.
If the question really is "how do I pick the investment strategy that will have the highest return in some future period" then the answer is pretty straightforward: you don't.

Although the advantages of the S&P 600 relative to the Russell 2000 weren't invisible nineteen years ago (S&P specifically designed their index to avoid some of the well-known problems with the Russell 2000, after all), the same rules apply here as everywhere else: the future is unknown. Investors should have expected funds tracking the S&P indexes to have an easier go of it than funds tracking the Russell indexes, but it's true that it didn't have to turn out that way.

Besides, ultimately the Russell 2000 vs S&P 600 index comparison over any particular time period is more anecdote than evidence. There are over 50 different ways to invest in S&P 500 stocks using ETFs, and each one will produce different results over the next 10 years. The odds of picking the right one aren't very good, actually. And even looking backwards, it's not hard to see that the differences between large cap value indexes can be just as significant as those between small cap value indexes.

Image

Image
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

Post Reply