Why don't you factor tilt?

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
dbr
Posts: 46137
Joined: Sun Mar 04, 2007 8:50 am

Re: Why don't you factor tilt?

Post by dbr »

Apathizer wrote: Mon Sep 26, 2022 3:55 pm
burritoLover wrote: Mon Sep 26, 2022 9:40 am I'm not sure how anyone can argue that adding an SCV tilt is less diversified and more concentrated than holding only TSM. The top 20 stocks in TSM make up 30+% of the entire index and the top stock (AAPL) is double that of ALL the SCV stocks in TSM. Adding an SCV tilt reduces your concentration in the mega caps that overwhelm the TSM index even if you don't have a strong conviction with small/value factors.

And we know that anything outside of large caps is practically irrelevant in TSM - a difference in return in the hundredths of a percent over 50 years. So to say you already own SCV in TSM is just nice little meaningless anecdote.
I've made this point repeatedly. The cap weight TSM SV allocation is only nominal. A higher SV slant is necessary to provide significant diversification benefit.
What are the terms of the benefit and how do you measure it? Given a definition and a way to calculate does anyone know what the results are?

I did a very naive calculation of that using diversification ratio as a measure for holding various allocations to VSIAX and VTSMX between 2012 and 2022 and got that the diversification ratio between 100/0 and 0/100 stayed between 1.00 at the ends and 1.03 maximum at 60% VSIAX, but I am not sure I know what I am doing. That is essentially no diversification at any level of VSIAX.

In this paper https://dial.uclouvain.be/memoire/ucl/e ... is%3A14352 the following definitions of diversification are explored:

1. The Shannon’s Entropy. Originally coming from the Information Theory and developed
by Claude Shannon (1948) to solve communication problems, Shannon’s Entropy has
later been applied in finance to measure the amount of information given by observing
the market.
2. The Diversification Delta. This measure introduced by Vermorken et al. (2012) is based
on empirical entropy. However, due to several drawbacks, we will use an alternative
version of it that has been developed by Salazar et al. (2014).
3. The Diversification Ratio. This measure proposed by Choueifaty (2006) is defined as
the ratio of the portfolio’s weighted average volatility to its overall volatility.
4. The Marginal Risk Contributions. This measure allows to decompose the total risk of
a portfolio into the contributions of each individual assets.
5. The Portfolio Diversification Index. This measure indicates the number of unique
investments in a portfolio and is useful to assess marginal and cumulative diversification
benefits across asset classes.
6. The Effective Number of Bets. This measure uses the entropy of some factors (of risk)
exposure distributions to indicate the diversification of a portfolio.

I used choice 3 on a very small data set. Choice 5 does not exactly mean what it appears to mean as it presumes that first of all one has done a principal components analysis, sometimes called factor analysis, but is not the same thing as factors in a Fama-French regression.

I admit to not being well grounded in this area.

As an example of entropy I tried a very simple minded example that might not actually be correct. The example is given 9 numbers between 1 and 9, which is more diversified 1,2,3,4,5,6,7,8,9 or 5,5,5,5,5,5,5,5,5 or 1,1,1,1,5,9,9,9,9? Note they all have a mean of 5. The respective entropies are 2.20, 0.00, and 0.96 so the first list is more "spread out" over values meaning more diversified.

Here is a paper that uses the number of bets analysis: https://portfoliooptimizer.io/blog/the- ... ification/ The example they cite in the paper is for a portfolio of bond funds and not for stock factor investing. Even in that case the author's conclusion might be unexpected.

Here is a typical Fama-French paper on risk factors but the terms diversify or diversification do not appear in the paper: https://www.bauer.uh.edu/rsusmel/phd/Fa ... _JFE93.pdf Indeed the term "risk" is not actually defined in the paper though it is clear that risk factor is a term that explains returns in a regression model.

In the paper originally quoted above risk is offered as either volatility (SD) or value at risk (VaR).
rkhusky
Posts: 17654
Joined: Thu Aug 18, 2011 8:09 pm

Re: Why don't you factor tilt?

Post by rkhusky »

dbr wrote: Mon Sep 26, 2022 4:19 pm I did a very naive calculation of that using diversification ratio as a measure for holding various allocations to VSIAX and VTSMX between 2012 and 2022 and got that the diversification ratio between 100/0 and 0/100 stayed between 1.00 at the ends and 1.03 maximum at 60% VSIAX, but I am not sure I know what I am doing. That is essentially no diversification at any level of VSIAX.
The correlation between VTSAX and VSIAX from 2011 - 2022 was 0.92. For the same period, the correlation between VTSAX and DFSAX was 0.86.

The correlation between VTSMX and VISVX from 1998 - 2022 was 0.88. The correlation between VTSMX and DFSAX was 0.85 between 1993 and 2022.
dcabler
Posts: 4482
Joined: Wed Feb 19, 2014 10:30 am

Re: Why don't you factor tilt?

Post by dcabler »

dbr wrote: Mon Sep 26, 2022 4:19 pm
Apathizer wrote: Mon Sep 26, 2022 3:55 pm
burritoLover wrote: Mon Sep 26, 2022 9:40 am I'm not sure how anyone can argue that adding an SCV tilt is less diversified and more concentrated than holding only TSM. The top 20 stocks in TSM make up 30+% of the entire index and the top stock (AAPL) is double that of ALL the SCV stocks in TSM. Adding an SCV tilt reduces your concentration in the mega caps that overwhelm the TSM index even if you don't have a strong conviction with small/value factors.

And we know that anything outside of large caps is practically irrelevant in TSM - a difference in return in the hundredths of a percent over 50 years. So to say you already own SCV in TSM is just nice little meaningless anecdote.
I've made this point repeatedly. The cap weight TSM SV allocation is only nominal. A higher SV slant is necessary to provide significant diversification benefit.
What are the terms of the benefit and how do you measure it? Given a definition and a way to calculate does anyone know what the results are?

I did a very naive calculation of that using diversification ratio as a measure for holding various allocations to VSIAX and VTSMX between 2012 and 2022 and got that the diversification ratio between 100/0 and 0/100 stayed between 1.00 at the ends and 1.03 maximum at 60% VSIAX, but I am not sure I know what I am doing. That is essentially no diversification at any level of VSIAX.

In this paper https://dial.uclouvain.be/memoire/ucl/e ... is%3A14352 the following definitions of diversification are explored:

1. The Shannon’s Entropy. Originally coming from the Information Theory and developed
by Claude Shannon (1948) to solve communication problems, Shannon’s Entropy has
later been applied in finance to measure the amount of information given by observing
the market.
2. The Diversification Delta. This measure introduced by Vermorken et al. (2012) is based
on empirical entropy. However, due to several drawbacks, we will use an alternative
version of it that has been developed by Salazar et al. (2014).
3. The Diversification Ratio. This measure proposed by Choueifaty (2006) is defined as
the ratio of the portfolio’s weighted average volatility to its overall volatility.
4. The Marginal Risk Contributions. This measure allows to decompose the total risk of
a portfolio into the contributions of each individual assets.
5. The Portfolio Diversification Index. This measure indicates the number of unique
investments in a portfolio and is useful to assess marginal and cumulative diversification
benefits across asset classes.
6. The Effective Number of Bets. This measure uses the entropy of some factors (of risk)
exposure distributions to indicate the diversification of a portfolio.

I used choice 3 on a very small data set. Choice 5 does not exactly mean what it appears to mean as it presumes that first of all one has done a principal components analysis, sometimes called factor analysis, but is not the same thing as factors in a Fama-French regression.

I admit to not being well grounded in this area.

As an example of entropy I tried a very simple minded example that might not actually be correct. The example is given 9 numbers between 1 and 9, which is more diversified 1,2,3,4,5,6,7,8,9 or 5,5,5,5,5,5,5,5,5 or 1,1,1,1,5,9,9,9,9? Note they all have a mean of 5. The respective entropies are 2.20, 0.00, and 0.96 so the first list is more "spread out" over values meaning more diversified.

Here is a paper that uses the number of bets analysis: https://portfoliooptimizer.io/blog/the- ... ification/ The example they cite in the paper is for a portfolio of bond funds and not for stock factor investing. Even in that case the author's conclusion might be unexpected.

Here is a typical Fama-French paper on risk factors but the terms diversify or diversification do not appear in the paper: https://www.bauer.uh.edu/rsusmel/phd/Fa ... _JFE93.pdf Indeed the term "risk" is not actually defined in the paper though it is clear that risk factor is a term that explains returns in a regression model.

In the paper originally quoted above risk is offered as either volatility (SD) or value at risk (VaR).
vineviz's calculator can be found here: https://www.dropbox.com/s/n5vnejnvilfha ... .xlsx?dl=0

What I've seen playing with my own version is that diversification increases with the more independent sources of volatility you add. As vineviz points out elsewhere, anything with a correlation of <1.0 will help contribute to an improved diversification ratio. Just don't expect it to be much if the correlation is close to 1.0. Adding bonds increases diversification quite a bit for example. It's not my cup of tea, but Golden Butterfly goes even further. Increasing diversification by this measure is not guaranteed to increase returns as been noted elsewhere. The Holy Grail would be to find a bunch of near zero correlated assets each of which has had high historic returns in proportions that give both fantastic historic returns and high diversification. Good luck with that. :D
dbr
Posts: 46137
Joined: Sun Mar 04, 2007 8:50 am

Re: Why don't you factor tilt?

Post by dbr »

dcabler wrote: Mon Sep 26, 2022 5:17 pm

vineviz's calculator can be found here: https://www.dropbox.com/s/n5vnejnvilfha ... .xlsx?dl=0

What I've seen playing with my own version is that diversification increases with the more independent sources of volatility you add. As vineviz points out elsewhere, anything with a correlation of <1.0 will help contribute to an improved diversification ratio. Just don't expect it to be much if the correlation is close to 1.0. Adding bonds increases diversification quite a bit for example. It's not my cup of tea, but Golden Butterfly goes even further. Increasing diversification by this measure is not guaranteed to increase returns as been noted elsewhere. The Holy Grail would be to find a bunch of near zero correlated assets each of which has had high historic returns in proportions that give both fantastic historic returns and high diversification. Good luck with that. :D
Thanks for link and comments.
dbr
Posts: 46137
Joined: Sun Mar 04, 2007 8:50 am

Re: Why don't you factor tilt?

Post by dbr »

rkhusky wrote: Mon Sep 26, 2022 5:00 pm
dbr wrote: Mon Sep 26, 2022 4:19 pm I did a very naive calculation of that using diversification ratio as a measure for holding various allocations to VSIAX and VTSMX between 2012 and 2022 and got that the diversification ratio between 100/0 and 0/100 stayed between 1.00 at the ends and 1.03 maximum at 60% VSIAX, but I am not sure I know what I am doing. That is essentially no diversification at any level of VSIAX.
The correlation between VTSAX and VSIAX from 2011 - 2022 was 0.92. For the same period, the correlation between VTSAX and DFSAX was 0.86.

The correlation between VTSMX and VISVX from 1998 - 2022 was 0.88. The correlation between VTSMX and DFSAX was 0.85 between 1993 and 2022.
Thanks. Hard to get a high diversification ratio from high correlations.
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Why don't you factor tilt?

Post by vineviz »

dbr wrote: Mon Sep 26, 2022 5:57 pm
rkhusky wrote: Mon Sep 26, 2022 5:00 pm
dbr wrote: Mon Sep 26, 2022 4:19 pm I did a very naive calculation of that using diversification ratio as a measure for holding various allocations to VSIAX and VTSMX between 2012 and 2022 and got that the diversification ratio between 100/0 and 0/100 stayed between 1.00 at the ends and 1.03 maximum at 60% VSIAX, but I am not sure I know what I am doing. That is essentially no diversification at any level of VSIAX.
The correlation between VTSAX and VSIAX from 2011 - 2022 was 0.92. For the same period, the correlation between VTSAX and DFSAX was 0.86.

The correlation between VTSMX and VISVX from 1998 - 2022 was 0.88. The correlation between VTSMX and DFSAX was 0.85 between 1993 and 2022.
Thanks. Hard to get a high diversification ratio from high correlations.
It is, but a small improvement in the diversification ratio can have a surprisingly large benefit in portfolio performance. 1.03 is enough to have an impact.

Another wrinkle is the time interval used to measure the correlation. Most tools, and authors, use monthly returns but this is a problem when the typical investment horizon for stocks is between two and three orders of magnitude greater than that.

Try measuring the correlation using 5-year returns instead of monthly or even annual returns, and you'll see correlations lower.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
nedsaid
Posts: 19249
Joined: Fri Nov 23, 2012 11:33 am

Re: Why don't you factor tilt?

Post by nedsaid »

abc132 wrote: Sun Sep 25, 2022 11:06 pm
nedsaid wrote: Sun Sep 25, 2022 10:06 pm
The thing is, there are varying degrees of understanding related to these topics. Sometimes people talk past each other, sometimes people disagree because they don't quite understand each other's arguments, and sometimes people are in violent agreement. These discussions get to be pretty nuanced. Sometimes a lot of sound and fury over not much.

People are free to either accept or reject the academic research. There are smart people, here and elsewhere, who disagree whether factor premiums exist and if they do, if those factor premiums are investable.
It frustrating when you get challenged by someone for saying people predicted a 2.6% SWR and within 24 hours that same person posts that they calculated the SWR at 2.7%. It's very difficult to converse and make progress in that context. Sometimes it's more about listening than level of knowledge.
If you choose not to tilt, that is A-OK. I respect intelligent people here who factor tilt and other intelligent, well informed people who choose not to. Don't get too frustrated.
A fool and his money are good for business.
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Why don't you factor tilt?

Post by Apathizer »

vineviz wrote: Mon Sep 26, 2022 7:17 pm
dbr wrote: Mon Sep 26, 2022 5:57 pm
rkhusky wrote: Mon Sep 26, 2022 5:00 pm
dbr wrote: Mon Sep 26, 2022 4:19 pm I did a very naive calculation of that using diversification ratio as a measure for holding various allocations to VSIAX and VTSMX between 2012 and 2022 and got that the diversification ratio between 100/0 and 0/100 stayed between 1.00 at the ends and 1.03 maximum at 60% VSIAX, but I am not sure I know what I am doing. That is essentially no diversification at any level of VSIAX.
The correlation between VTSAX and VSIAX from 2011 - 2022 was 0.92. For the same period, the correlation between VTSAX and DFSAX was 0.86.

The correlation between VTSMX and VISVX from 1998 - 2022 was 0.88. The correlation between VTSMX and DFSAX was 0.85 between 1993 and 2022.
Thanks. Hard to get a high diversification ratio from high correlations.
It is, but a small improvement in the diversification ratio can have a surprisingly large benefit in portfolio performance. 1.03 is enough to have an impact.

Another wrinkle is the time interval used to measure the correlation. Most tools, and authors, use monthly returns but this is a problem when the typical investment horizon for stocks is between two and three orders of magnitude greater than that.

Try measuring the correlation using 5-year returns instead of monthly or even annual returns, and you'll see correlations lower.
Even when different things are highly correlated there can be significant dispersion. For instance over the last decade or so US and international markets have been more correlated but there's still significant dispersion. This is also a likely benefit of factors. Even if and when they're highly correlated they's a dispersion benefit is likely.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
fisher0815
Posts: 155
Joined: Tue Jun 08, 2021 3:10 am

Re: Why don't you factor tilt?

Post by fisher0815 »

I don't factor tilt because it's just stock picking based on a few simple screens.

DonIce agrees:
DonIce wrote: Sat Apr 25, 2020 5:42 pm I had never heard of factor investing until I started reading Bogleheads. Shortly after I joined the forum, I ran into factors. Factor proponents here seemed almost religious in their adherence, with admonitions like "you have to believe in them" and "never sell out even if the factor premium doesn't appear for decades, because it's a long term commitment". Then I looked at what is actually in factor funds and realized it is just nothing more than stock picking based on a few simple screens. There are countless studies robustly demonstrating that stock picking, statistically, is unlikely to outperform the market.
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Why don't you factor tilt?

Post by Apathizer »

fisher0815 wrote: Tue Sep 27, 2022 2:24 am I don't factor tilt because it's just stock picking based on a few simple screens.

DonIce agrees:
DonIce wrote: Sat Apr 25, 2020 5:42 pm I had never heard of factor investing until I started reading Bogleheads. Shortly after I joined the forum, I ran into factors. Factor proponents here seemed almost religious in their adherence, with admonitions like "you have to believe in them" and "never sell out even if the factor premium doesn't appear for decades, because it's a long term commitment". Then I looked at what is actually in factor funds and realized it is just nothing more than stock picking based on a few simple screens. There are countless studies robustly demonstrating that stock picking, statistically, is unlikely to outperform the market.
I actually see more quasi pious fervor from cap weight TSM advocates who consider factor-slants a blasphemous insult to St Bogle.
A cap weight TSM index is generally fine, but it's dominated by large growth. Factor slants are based on similar fundamentals; they're just diversifying into other aspects of the market: value, size, profitability, and reinvestment.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
User avatar
steve r
Posts: 1299
Joined: Mon Feb 13, 2012 7:34 pm
Location: Connecticut

Re: Why don't you factor tilt?

Post by steve r »

It is complicated.

I do not factor tilt because to me staying the course is more important. I doubt very seriously I could stay the course when things underperformed for a decade or whatever (I do not know the exact amount, which varies based how you play the tilts, particularly SCV.

That said, I have stayed the course with my international holdings. This suggests I do no fully buy into the theory of tilting. I see some giant names in finance being fans of tilting, but Bill Sharpe is not. I do see the very long run back test tilting outperformances.

I am tempted by Avantis worldwide that is coming -- AVGE, not so much because of the theory of factor tilting, but because I do not like the extent of some mega large cap (growth) funds on my overall portfolio. My work fund is index, I cannot reasonably change that. But reducing LCG elsewhere seems like something I can stay the course with.
"Owning the stock market over the long term is a winner's game. Attempting to beat the market is a loser's game. ..Don't look for the needle in the haystack. Just buy the haystack." Jack Bogle
rkhusky
Posts: 17654
Joined: Thu Aug 18, 2011 8:09 pm

Re: Why don't you factor tilt?

Post by rkhusky »

Apathizer wrote: Tue Sep 27, 2022 3:13 am
fisher0815 wrote: Tue Sep 27, 2022 2:24 am I don't factor tilt because it's just stock picking based on a few simple screens.

DonIce agrees:
DonIce wrote: Sat Apr 25, 2020 5:42 pm I had never heard of factor investing until I started reading Bogleheads. Shortly after I joined the forum, I ran into factors. Factor proponents here seemed almost religious in their adherence, with admonitions like "you have to believe in them" and "never sell out even if the factor premium doesn't appear for decades, because it's a long term commitment". Then I looked at what is actually in factor funds and realized it is just nothing more than stock picking based on a few simple screens. There are countless studies robustly demonstrating that stock picking, statistically, is unlikely to outperform the market.
I actually see more quasi pious fervor from cap weight TSM advocates who consider factor-slants a blasphemous insult to St Bogle.
A cap weight TSM index is generally fine, but it's dominated by large growth. Factor slants are based on similar fundamentals; they're just diversifying into other aspects of the market: value, size, profitability, and reinvestment.
fisher0815 and DonIce have it right. Factor investing is not diversifying, it is filtering out unwanted market components, reducing the portfolio to stocks that all have the same factor characteristics, like value, size, profitability, and reinvestment, thereby increasing risk with the hope of increasing return. That's the opposite of diversification.
dbr
Posts: 46137
Joined: Sun Mar 04, 2007 8:50 am

Re: Why don't you factor tilt?

Post by dbr »

vineviz wrote: Mon Sep 26, 2022 7:17 pm

It is, but a small improvement in the diversification ratio can have a surprisingly large benefit in portfolio performance. 1.03 is enough to have an impact.
When you say that my understanding of the word "portfolio performance" is expected return.* When I ran some data in different allocations between stocks and bonds the diversification ratio was 1 at 100/0 and 0/100 but rose to 1.6 at 20/80. On the other hand the return for the period in question (Portfolio Visualizer) rises monotonically from 5.23% at 0/100 to 9.67% at 100/0 and is not related to diversification. At the same time the risk rises from 3.96% to 15.22%. What does track the diversification ratio is the Sharpe ratio, which rises to .81 at 20/80 but is below .6 at 0/100 and 100/0. So in the case of blending a bond portfolio with a stock portfolio what is the meaning of saying diversification ratio is related to performance? This is not a trick question. I am genuinely baffled by the whole concept of diversification and may lack the background to figure out what on earth people are talking about. Later on we can try to understand the words "Factor investing is a risk story involving diversification over risk factors." So far in trying to read academic papers about diversification I can't find any that specifically address how one can get from the Fama-French values of the factor loadings, which describe the expected return of the portfolio, to any measure of "diversification over risk factors" that anyone can define.

*Note here expected return comes out of Portfolio Visualizer via a value for CAGR for a period. Technically that is not a pure performance measure as an estimate of the mean value of the return distribution would come from the arithmetic average of the returns sample and the CAGR is depressed from that when the risk is higher. But that does not seem to be a nuance that has any effect on how performance is related to some specific measure of diversification.

Thanks
rkhusky
Posts: 17654
Joined: Thu Aug 18, 2011 8:09 pm

Re: Why don't you factor tilt?

Post by rkhusky »

Random Walker wrote: Mon Sep 26, 2022 9:09 am You sort of brought up a dichotomy above in the two value scenarios. The struggling value company that needs a turn around versus the consistent profitable value company with not a lot of growth prospect. Factor investors believe profitability screen helps determine the better stocks for investment in the low P/B realm. I’m a bit confused by the rationale. Larry Swedroe says profitability screen screens out falling knife “value traps”. I think Cliff Asness says profitability screen identifies the low P/B stocks that are truly risky for perhaps not so obvious reasons.
Can profitability then be rightly called a risk factor? Are any of the other common factors of this type, that actually screen for less risk?
Random Walker
Posts: 5561
Joined: Fri Feb 23, 2007 7:21 pm

Re: Why don't you factor tilt?

Post by Random Walker »

dbr wrote: Tue Sep 27, 2022 8:00 am
vineviz wrote: Mon Sep 26, 2022 7:17 pm

It is, but a small improvement in the diversification ratio can have a surprisingly large benefit in portfolio performance. 1.03 is enough to have an impact.
When you say that my understanding of the word "portfolio performance" is expected return.* When I ran some data in different allocations between stocks and bonds the diversification ratio was 1 at 100/0 and 0/100 but rose to 1.6 at 20/80. On the other hand the return for the period in question (Portfolio Visualizer) rises monotonically from 5.23% at 0/100 to 9.67% at 100/0 and is not related to diversification. At the same time the risk rises from 3.96% to 15.22%. What does track the diversification ratio is the Sharpe ratio, which rises to .81 at 20/80 but is below .6 at 0/100 and 100/0. So in the case of blending a bond portfolio with a stock portfolio what is the meaning of saying diversification ratio is related to performance? This is not a trick question. I am genuinely baffled by the whole concept of diversification and may lack the background to figure out what on earth people are talking about. Later on we can try to understand the words "Factor investing is a risk story involving diversification over risk factors." So far in trying to read academic papers about diversification I can't find any that specifically address how one can get from the Fama-French values of the factor loadings, which describe the expected return of the portfolio, to any measure of "diversification over risk factors" that anyone can define.

*Note here expected return comes out of Portfolio Visualizer via a value for CAGR for a period. Technically that is not a pure performance measure as an estimate of the mean value of the return distribution would come from the arithmetic average of the returns sample and the CAGR is depressed from that when the risk is higher. But that does not seem to be a nuance that has any effect on how performance is related to some specific measure of diversification.

Thanks
We evaluate potential portfolio additions partly based on their expected return, which is an arithmetic mean. We actually eat the compounded returns of our portfolio, which is a geometric mean. Portfolio volatility causes the Geo Mean to always be less than Ari Mean. An equation that approximate this very well is Geo Mean = Ari Mean -(0.5*SD^2). Anything that can keep portfolio expected return about constant but decrease portfolio volatility will likely bring the Geo Mean closer to the Ari Mean. Productive diversification can accomplish this.

Dave
dbr
Posts: 46137
Joined: Sun Mar 04, 2007 8:50 am

Re: Why don't you factor tilt?

Post by dbr »

Random Walker wrote: Tue Sep 27, 2022 8:27 am
dbr wrote: Tue Sep 27, 2022 8:00 am
vineviz wrote: Mon Sep 26, 2022 7:17 pm

It is, but a small improvement in the diversification ratio can have a surprisingly large benefit in portfolio performance. 1.03 is enough to have an impact.
When you say that my understanding of the word "portfolio performance" is expected return.* When I ran some data in different allocations between stocks and bonds the diversification ratio was 1 at 100/0 and 0/100 but rose to 1.6 at 20/80. On the other hand the return for the period in question (Portfolio Visualizer) rises monotonically from 5.23% at 0/100 to 9.67% at 100/0 and is not related to diversification. At the same time the risk rises from 3.96% to 15.22%. What does track the diversification ratio is the Sharpe ratio, which rises to .81 at 20/80 but is below .6 at 0/100 and 100/0. So in the case of blending a bond portfolio with a stock portfolio what is the meaning of saying diversification ratio is related to performance? This is not a trick question. I am genuinely baffled by the whole concept of diversification and may lack the background to figure out what on earth people are talking about. Later on we can try to understand the words "Factor investing is a risk story involving diversification over risk factors." So far in trying to read academic papers about diversification I can't find any that specifically address how one can get from the Fama-French values of the factor loadings, which describe the expected return of the portfolio, to any measure of "diversification over risk factors" that anyone can define.

*Note here expected return comes out of Portfolio Visualizer via a value for CAGR for a period. Technically that is not a pure performance measure as an estimate of the mean value of the return distribution would come from the arithmetic average of the returns sample and the CAGR is depressed from that when the risk is higher. But that does not seem to be a nuance that has any effect on how performance is related to some specific measure of diversification.

Thanks
We evaluate potential portfolio additions partly based on their expected return, which is an arithmetic mean. We actually eat the compounded returns of our portfolio, which is a geometric mean. Portfolio volatility causes the Geo Mean to always be less than Ari Mean. An equation that approximate this very well is Geo Mean = Ari Mean -(0.5*SD^2). Anything that can keep portfolio expected return about constant but decrease portfolio volatility will likely bring the Geo Mean closer to the Ari Mean. Productive diversification can accomplish this.

Dave
Yes, I know that. In my example of the stock/bond portfolio I should add a column to display the arithmetic mean, perhaps using that formula. The result will still be that the most diversified portfolio does not have the highest arithmetic return. That value occurs at 100% stocks as far as I know, but I'll go look at it. Yep, it is still true that the return is monotonic with stock allocation and a maximum at 100/0. The mean return increased there from 9.67% to 10.83% from CAGR to arithmetic return by formula. The value at 0/100 increased from 5.23% to 5.31%.
Last edited by dbr on Tue Sep 27, 2022 8:37 am, edited 1 time in total.
Random Walker
Posts: 5561
Joined: Fri Feb 23, 2007 7:21 pm

Re: Why don't you factor tilt?

Post by Random Walker »

dbr wrote: Tue Sep 27, 2022 8:28 am
Random Walker wrote: Tue Sep 27, 2022 8:27 am
dbr wrote: Tue Sep 27, 2022 8:00 am
vineviz wrote: Mon Sep 26, 2022 7:17 pm

It is, but a small improvement in the diversification ratio can have a surprisingly large benefit in portfolio performance. 1.03 is enough to have an impact.
When you say that my understanding of the word "portfolio performance" is expected return.* When I ran some data in different allocations between stocks and bonds the diversification ratio was 1 at 100/0 and 0/100 but rose to 1.6 at 20/80. On the other hand the return for the period in question (Portfolio Visualizer) rises monotonically from 5.23% at 0/100 to 9.67% at 100/0 and is not related to diversification. At the same time the risk rises from 3.96% to 15.22%. What does track the diversification ratio is the Sharpe ratio, which rises to .81 at 20/80 but is below .6 at 0/100 and 100/0. So in the case of blending a bond portfolio with a stock portfolio what is the meaning of saying diversification ratio is related to performance? This is not a trick question. I am genuinely baffled by the whole concept of diversification and may lack the background to figure out what on earth people are talking about. Later on we can try to understand the words "Factor investing is a risk story involving diversification over risk factors." So far in trying to read academic papers about diversification I can't find any that specifically address how one can get from the Fama-French values of the factor loadings, which describe the expected return of the portfolio, to any measure of "diversification over risk factors" that anyone can define.

*Note here expected return comes out of Portfolio Visualizer via a value for CAGR for a period. Technically that is not a pure performance measure as an estimate of the mean value of the return distribution would come from the arithmetic average of the returns sample and the CAGR is depressed from that when the risk is higher. But that does not seem to be a nuance that has any effect on how performance is related to some specific measure of diversification.

Thanks
We evaluate potential portfolio additions partly based on their expected return, which is an arithmetic mean. We actually eat the compounded returns of our portfolio, which is a geometric mean. Portfolio volatility causes the Geo Mean to always be less than Ari Mean. An equation that approximate this very well is Geo Mean = Ari Mean -(0.5*SD^2). Anything that can keep portfolio expected return about constant but decrease portfolio volatility will likely bring the Geo Mean closer to the Ari Mean. Productive diversification can accomplish this.

Dave
Yes, I know that.
I don’t think a lot of people appreciate it. How much emphasis do you place on that thinking in your portfolio construction? Important? Not important?

Dave
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Why don't you factor tilt?

Post by vineviz »

dbr wrote: Tue Sep 27, 2022 8:00 am
vineviz wrote: Mon Sep 26, 2022 7:17 pm

It is, but a small improvement in the diversification ratio can have a surprisingly large benefit in portfolio performance. 1.03 is enough to have an impact.
When you say that my understanding of the word "portfolio performance" is expected return.*
That's not, IMHO, what you should interpret "portfolio performance" to mean.

The goal of diversification is not (necessarily) to increase expected return. It might, but then again it might not. There are many benefits of diversification that have nothing to do with changing expected return.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
dbr
Posts: 46137
Joined: Sun Mar 04, 2007 8:50 am

Re: Why don't you factor tilt?

Post by dbr »

vineviz wrote: Tue Sep 27, 2022 8:46 am
dbr wrote: Tue Sep 27, 2022 8:00 am
vineviz wrote: Mon Sep 26, 2022 7:17 pm

It is, but a small improvement in the diversification ratio can have a surprisingly large benefit in portfolio performance. 1.03 is enough to have an impact.
When you say that my understanding of the word "portfolio performance" is expected return.*
That's not, IMHO, what you should interpret "portfolio performance" to mean.

The goal of diversification is not (necessarily) to increase expected return. It might, but then again it might not. There are many benefits of diversification that have nothing to do with changing expected return.
Well, the whole point is what are we talking about then. I respect that I can't ask people to write essays in replies to a forum, but with all due respect "There are many benefits . . ." is not a very helpful answer.

Just as an exploration, reduction of risk would be an obvious benefit. The equally obvious answer in my data set is that minimum risk occurs at 0/100, the same place as minimum return. The one parameter that benefits from diversification is the Sharpe ratio. Can you help me understand why I should care about Sharpe ratio? Is there a list somewhere of the things an individual investor would care about that are maximized when the diversification ratio is maximized? I am pretty sure other things people talk about such as max drawdown are still least for all bonds and most for all stocks, etc.

Do you have a suggestion where a person who wants to understand what we are talking about should start to learn this stuff?

I really do appreciate your comments.
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Why don't you factor tilt?

Post by Apathizer »

rkhusky wrote: Tue Sep 27, 2022 6:45 am
Apathizer wrote: Tue Sep 27, 2022 3:13 am
fisher0815 wrote: Tue Sep 27, 2022 2:24 am I don't factor tilt because it's just stock picking based on a few simple screens.

DonIce agrees:
DonIce wrote: Sat Apr 25, 2020 5:42 pm I had never heard of factor investing until I started reading Bogleheads. Shortly after I joined the forum, I ran into factors. Factor proponents here seemed almost religious in their adherence, with admonitions like "you have to believe in them" and "never sell out even if the factor premium doesn't appear for decades, because it's a long term commitment". Then I looked at what is actually in factor funds and realized it is just nothing more than stock picking based on a few simple screens. There are countless studies robustly demonstrating that stock picking, statistically, is unlikely to outperform the market.
I actually see more quasi pious fervor from cap weight TSM advocates who consider factor-slants a blasphemous insult to St Bogle.
A cap weight TSM index is generally fine, but it's dominated by large growth. Factor slants are based on similar fundamentals; they're just diversifying into other aspects of the market: value, size, profitability, and reinvestment.
fisher0815 and DonIce have it right. Factor investing is not diversifying, it is filtering out unwanted market components, reducing the portfolio to stocks that all have the same factor characteristics, like value, size, profitability, and reinvestment, thereby increasing risk with the hope of increasing return. That's the opposite of diversification.
You keep repeating that same fallacy. If we were only investing in those for non-market factors that would be the case, but almost no one, if anyone, advocates that approach. The cap weight TSM is dominated by large growth companies. We're arguing for shifting some of that massive allocation to smaller, less expensive, more profitable companies. Light-moderate factor slants are about having a more balanced, diversified portfolio.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
Random Walker
Posts: 5561
Joined: Fri Feb 23, 2007 7:21 pm

Re: Why don't you factor tilt?

Post by Random Walker »

dbr wrote: Tue Sep 27, 2022 8:58 am The one parameter that benefits from diversification is the Sharpe ratio. Can you help me understand why I should care about Sharpe ratio? Is there a list somewhere of the things an individual investor would care about that are maximized when the diversification ratio is maximized? I am pretty sure other things people talk about such as max drawdown are still least for all bonds and most for all stocks, etc.

Do you have a suggestion where a person who wants to understand what we are talking about should start to learn this stuff?

I really do appreciate your comments.
The denominator in Sharpe ratio is SD. So improved Sharpe ratio lessens the variance drain I described above. I recommend Larry Swedroe’s books, especially Factor Book and Black Swan Book. If you want more of a textbook that is pretty readable, I’d recommend Successful Investing Is A Process by Jacques Lussier. I haven’t read it cover to cover, but I reread a couple of chapters on diversification periodically. I use it more as a reference
https://www.amazon.com/Successful-Inves ... 1118459903
Andrew Ang wrote a book as well, but I thought Lussier was more readable.

Dave
dbr
Posts: 46137
Joined: Sun Mar 04, 2007 8:50 am

Re: Why don't you factor tilt?

Post by dbr »

Random Walker wrote: Tue Sep 27, 2022 9:10 am
dbr wrote: Tue Sep 27, 2022 8:58 am The one parameter that benefits from diversification is the Sharpe ratio. Can you help me understand why I should care about Sharpe ratio? Is there a list somewhere of the things an individual investor would care about that are maximized when the diversification ratio is maximized? I am pretty sure other things people talk about such as max drawdown are still least for all bonds and most for all stocks, etc.

Do you have a suggestion where a person who wants to understand what we are talking about should start to learn this stuff?

I really do appreciate your comments.
The denominator in Sharpe ratio is SD. So improved Sharpe ratio lessens the variance drain I described above. I recommend Larry Swedroe’s books, especially Factor Book and Black Swan Book. If you want more of a textbook that is pretty readable, I’d recommend Successful Investing Is A Process by Jacques Lussier. I haven’t read it cover to cover, but I reread a couple of chapters on diversification periodically. I use it more as a reference
https://www.amazon.com/Successful-Inves ... 1118459903
Andrew Ang wrote a book as well, but I thought Lussier was more readable.

Dave
I'll have a look.
rkhusky
Posts: 17654
Joined: Thu Aug 18, 2011 8:09 pm

Re: Why don't you factor tilt?

Post by rkhusky »

Apathizer wrote: Tue Sep 27, 2022 9:08 am
rkhusky wrote: Tue Sep 27, 2022 6:45 am
Apathizer wrote: Tue Sep 27, 2022 3:13 am
fisher0815 wrote: Tue Sep 27, 2022 2:24 am I don't factor tilt because it's just stock picking based on a few simple screens.

DonIce agrees:
DonIce wrote: Sat Apr 25, 2020 5:42 pm I had never heard of factor investing until I started reading Bogleheads. Shortly after I joined the forum, I ran into factors. Factor proponents here seemed almost religious in their adherence, with admonitions like "you have to believe in them" and "never sell out even if the factor premium doesn't appear for decades, because it's a long term commitment". Then I looked at what is actually in factor funds and realized it is just nothing more than stock picking based on a few simple screens. There are countless studies robustly demonstrating that stock picking, statistically, is unlikely to outperform the market.
I actually see more quasi pious fervor from cap weight TSM advocates who consider factor-slants a blasphemous insult to St Bogle.
A cap weight TSM index is generally fine, but it's dominated by large growth. Factor slants are based on similar fundamentals; they're just diversifying into other aspects of the market: value, size, profitability, and reinvestment.
fisher0815 and DonIce have it right. Factor investing is not diversifying, it is filtering out unwanted market components, reducing the portfolio to stocks that all have the same factor characteristics, like value, size, profitability, and reinvestment, thereby increasing risk with the hope of increasing return. That's the opposite of diversification.
You keep repeating that same fallacy. If we were only investing in those for non-market factors that would be the case, but almost no one, if anyone, advocates that approach. The cap weight TSM is dominated by large growth companies. We're arguing for shifting some of that massive allocation to smaller, less expensive, more profitable companies. Light-moderate factor slants are about having a more balanced, diversified portfolio.
Long-only factor funds typically have a market loading near 1. Why would you add further market loading through TSM to that?
DaufuskieNate
Posts: 605
Joined: Wed May 28, 2014 11:53 am

Re: Why don't you factor tilt?

Post by DaufuskieNate »

rkhusky wrote: Tue Sep 27, 2022 9:24 am Long-only factor funds typically have a market loading near 1. Why would you add further market loading through TSM to that?
I can think of 2 possible reasons: to hit a targeted factor loading and to manage overall costs.

Let's say you want to target .30 loadings on both SmB and HmL. You can do a 50:50 mix of a SCV fund, with .60 loadings on SmB and HmL, and TSM. Given the low ER of TSM funds, this may be the most cost efficient way to achieve your targeted factor loadings.
Random Walker
Posts: 5561
Joined: Fri Feb 23, 2007 7:21 pm

Re: Why don't you factor tilt?

Post by Random Walker »

DaufuskieNate wrote: Tue Sep 27, 2022 9:32 am
rkhusky wrote: Tue Sep 27, 2022 9:24 am Long-only factor funds typically have a market loading near 1. Why would you add further market loading through TSM to that?
I can think of 2 possible reasons: to hit a targeted factor loading and to manage overall costs.

Let's say you want to target .30 loadings on both SmB and HmL. You can do a 50:50 mix of a SCV fund, with .60 loadings on SmB and HmL, and TSM. Given the low ER of TSM funds, this may be the most cost efficient way to achieve your targeted factor loadings.
This is an example where the cheapest fund may well not be the best. One can achieve his desired factor loadings with a cheap core TSM fund and then tilt to taste with a factor fund. If one uses a perhaps slightly more expensive factor fund with deeper factor loadings, he will need less of it to achieve his factor targets. So the additional cost is less than on first thought. The fund with deeper exposures allows the investor to use less of the more expensive factor fund. And because of deeper factor exposures, the investor takes on less additional market factor risk to achieve the loadings on size and value he desires. And it’s market beta the factor investor is trying to diversify away from anyways.

Dave
dbr
Posts: 46137
Joined: Sun Mar 04, 2007 8:50 am

Re: Why don't you factor tilt?

Post by dbr »

DaufuskieNate wrote: Tue Sep 27, 2022 9:32 am
rkhusky wrote: Tue Sep 27, 2022 9:24 am Long-only factor funds typically have a market loading near 1. Why would you add further market loading through TSM to that?
I can think of 2 possible reasons: to hit a targeted factor loading and to manage overall costs.

Let's say you want to target .30 loadings on both SmB and HmL. You can do a 50:50 mix of a SCV fund, with .60 loadings on SmB and HmL, and TSM. Given the low ER of TSM funds, this may be the most cost efficient way to achieve your targeted factor loadings.
The question constantly asked but seemingly seldom answered is how would a factor investor decide what factor loading to target and why.
Random Walker
Posts: 5561
Joined: Fri Feb 23, 2007 7:21 pm

Re: Why don't you factor tilt?

Post by Random Walker »

dbr wrote: Tue Sep 27, 2022 9:17 am
Random Walker wrote: Tue Sep 27, 2022 9:10 am
dbr wrote: Tue Sep 27, 2022 8:58 am The one parameter that benefits from diversification is the Sharpe ratio. Can you help me understand why I should care about Sharpe ratio? Is there a list somewhere of the things an individual investor would care about that are maximized when the diversification ratio is maximized? I am pretty sure other things people talk about such as max drawdown are still least for all bonds and most for all stocks, etc.

Do you have a suggestion where a person who wants to understand what we are talking about should start to learn this stuff?

I really do appreciate your comments.
The denominator in Sharpe ratio is SD. So improved Sharpe ratio lessens the variance drain I described above. I recommend Larry Swedroe’s books, especially Factor Book and Black Swan Book. If you want more of a textbook that is pretty readable, I’d recommend Successful Investing Is A Process by Jacques Lussier. I haven’t read it cover to cover, but I reread a couple of chapters on diversification periodically. I use it more as a reference
https://www.amazon.com/Successful-Inves ... 1118459903
Andrew Ang wrote a book as well, but I thought Lussier was more readable.

Dave
I'll have a look.
Also, forgot one of my all time favorite books. Asset Allocation: Balancing Financial Risk by Roger Gibson. It’s worth the price for just a couple of its graphs. In the book he makes his point with I think 4 or 5 asset classes. The specific asset classes are not important. What is important is seeing the benefits of how different uncorrelated assets mix in a portfolio. I would learn the framework and apply it to your consideration of any potential portfolio addition in the future. He uses Efficient frontier charts to show what Larry does in his books with more narrow distributions: two sides of same coin.

https://www.amazon.com/Asset-Allocation ... 7893&psc=1

Dave
dbr
Posts: 46137
Joined: Sun Mar 04, 2007 8:50 am

Re: Why don't you factor tilt?

Post by dbr »

Random Walker wrote: Tue Sep 27, 2022 10:03 am
dbr wrote: Tue Sep 27, 2022 9:17 am
Random Walker wrote: Tue Sep 27, 2022 9:10 am
dbr wrote: Tue Sep 27, 2022 8:58 am The one parameter that benefits from diversification is the Sharpe ratio. Can you help me understand why I should care about Sharpe ratio? Is there a list somewhere of the things an individual investor would care about that are maximized when the diversification ratio is maximized? I am pretty sure other things people talk about such as max drawdown are still least for all bonds and most for all stocks, etc.

Do you have a suggestion where a person who wants to understand what we are talking about should start to learn this stuff?

I really do appreciate your comments.
The denominator in Sharpe ratio is SD. So improved Sharpe ratio lessens the variance drain I described above. I recommend Larry Swedroe’s books, especially Factor Book and Black Swan Book. If you want more of a textbook that is pretty readable, I’d recommend Successful Investing Is A Process by Jacques Lussier. I haven’t read it cover to cover, but I reread a couple of chapters on diversification periodically. I use it more as a reference
https://www.amazon.com/Successful-Inves ... 1118459903
Andrew Ang wrote a book as well, but I thought Lussier was more readable.

Dave
I'll have a look.
Also, forgot one of my all time favorite books. Asset Allocation: Balancing Financial Risk by Roger Gibson. It’s worth the price for just a couple of its graphs. In the book he makes his point with I think 4 or 5 asset classes. The specific asset classes are not important. What is important is seeing the benefits of how different uncorrelated assets mix in a portfolio. I would learn the framework and apply it to your consideration of any potential portfolio addition in the future. He uses Efficient frontier charts to show what Larry does in his books with more narrow distributions: two sides of same coin.

https://www.amazon.com/Asset-Allocation ... 7893&psc=1

Dave
Thanks again. Now to read.

I noted already one of the things you have to adapt to in this reading. Lussier prints a definition of GEO mean, referring in fact to CAGR, as we have been discussing, and explicitly posts the disconnect that in finance GEO mean is the return in the geometric mean of gains (1+r) and is not the geometric mean used by mathematicians, which is, for example, not defined when negative or zero values of the return are in the list. A person can easily cope with that, but it is certainly disconcerting and energy consuming to have to constantly allow for all those little disconnects. Note there is a GEOMEAN function in Excel but that is not what Lussier's definition is. Strangely out of all the financial functions in Excel CAGR does not appear, not that one does not know how to compute it with formulas: https://support.microsoft.com/en-us/off ... f124c2b1d8

Just quirks, but it is fascinating.
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Why don't you factor tilt?

Post by vineviz »

dbr wrote: Tue Sep 27, 2022 8:58 am
Do you have a suggestion where a person who wants to understand what we are talking about should start to learn this stuff?
I think the simplest way of thinking about the benefits of diversifications is that it reduces the dispersion of outcomes.

In the simplest case where there are just two asset, investing in both of them will give you neither the best nor the worst available outcome in any discrete period of time. If one is up 5% today and the other is down 1%, then a combination of the two must give you something in between.

As I know you are aware, the expected (arithmetic) return of the combination of the two assets is simply the weighted average of the two expected returns. But the expected volatility of the combination is LESS than the weighted average of the two expected volatilities as long as the two assets are not perfectly positively correlated.

This reduction of portfolio volatility relative to the weighted average volatility of the assets is the so-called "free lunch" of diversification. It's not magic or myth. It's just math.

Where things get tricky is the movement from the simple example to the complexities of our actual portfolios.

For instance, the math above is neatly resolved when portfolios don't have any cash flows in or out. But that's almost never the case with retirement portfolios. We are adding to them during our working years and taking income from them during the retirement years.

Also, the math above represents a fairly complete description of the problem if the ONLY things an investor cares about are return and single-period (i.e. short-term) price volatility. We know these are not the only things investors care about and often are not even the most important. An investor who is withdrawing from a retirement portfolio over a period of 25-30 years is arguably not well-served by using daily or monthly volatility/correlations as a proxy for their risk.

For instance, compare Vanguard Small Cap Value Index (VISVX) and Vanguard Total Stock Market Index (VTSMX) as an example. I'm choosing ("cherry-picking") the years 2000 to 2016 merely to illustrate the kind of differences I'm talking about.

Take a conventional calculation of correlation between the returns of these funds and you might get what appears to be a "high" correlation. PortfolioVisualizer calculates it as 0.93. You might use "common sense" and conclude from that correlation that there is not much diversification benefit to be had.

But take a look at two hypothetical portfolios in decumulation, each starting with $100k and withdrawing $400/month. One is 50% VISVX + 50% VTSMX, and the other is 100% VTSMX.

Image

Obviously we could argue about the allocation, the withdrawal rate, the timeframe, etc. but I'd prefer to not get dragged into that.

I think most people would not look at two US equity funds with a correlation of 0.93 and expect that kind of performance difference is even possible.


I don't expect that kind of difference to happen over the next 20 years, but people should be aware - IMHO - that it is possible. Investors in general have a strong knee-jerk aversion to diversification, fueled by biases that anchor on recent performance and/or a desire to own the "best" asset class/fund/stock. I understand that, and am sympathetic to the impulses, but we shouldn't let those biases lead us to blithely ridicule investment concepts like "be diversified" and "stay the course".

I'm honestly not sure an individual investor needs a complete theoretical understanding of all these concepts, but there are textbooks on it. Portfolio Diversification by Francois-Serge Lhabitant is probably the most on-point but it still omits some important practical considerations.

Hope that helps some....
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
DaufuskieNate
Posts: 605
Joined: Wed May 28, 2014 11:53 am

Re: Why don't you factor tilt?

Post by DaufuskieNate »

dbr wrote: Tue Sep 27, 2022 9:59 am The question constantly asked but seemingly seldom answered is how would a factor investor decide what factor loading to target and why.
It's a great question and in my opinion the answer gets into personal preferences and appetite for risk. One person may want the maximum possible tilt in pursuit of the possibility of a high return. Another person may mix a high tilt equity portfolio with a larger allocation to safe fixed income. Others may not be able to tolerate a significant tracking error relative to TSM in their equity allocation and opt for a mild factor loading. Just like the overall stock/bond mix, there is no one right answer for everyone.

Factor investing should not be taken lightly. In my opinion, one needs to study the evidence and theory carefully in order to understand what they are getting into. Looking at how various factor allocations have performed in a variety of market conditions in the past is also important as an indication of what could happen in the future. A long-term commitment is essential, and having a solid understanding is an important part of being able to stay the course in the inevitable tough times.
dbr
Posts: 46137
Joined: Sun Mar 04, 2007 8:50 am

Re: Why don't you factor tilt?

Post by dbr »

vineviz wrote: Tue Sep 27, 2022 10:30 am
dbr wrote: Tue Sep 27, 2022 8:58 am
Do you have a suggestion where a person who wants to understand what we are talking about should start to learn this stuff?
I think the simplest way of thinking about the benefits of diversifications is that it reduces the dispersion of outcomes.

In the simplest case where there are just two asset, investing in both of them will give you neither the best nor the worst available outcome in any discrete period of time. If one is up 5% today and the other is down 1%, then a combination of the two must give you something in between.

As I know you are aware, the expected (arithmetic) return of the combination of the two assets is simply the weighted average of the two expected returns. But the expected volatility of the combination is LESS than the weighted average of the two expected volatilities as long as the two assets are not perfectly positively correlated.

This reduction of portfolio volatility relative to the weighted average volatility of the assets is the so-called "free lunch" of diversification. It's not magic or myth. It's just math.

Where things get tricky is the movement from the simple example to the complexities of our actual portfolios.

For instance, the math above is neatly resolved when portfolios don't have any cash flows in or out. But that's almost never the case with retirement portfolios. We are adding to them during our working years and taking income from them during the retirement years.

Also, the math above represents a fairly complete description of the problem if the ONLY things an investor cares about are return and single-period (i.e. short-term) price volatility. We know these are not the only things investors care about and often are not even the most important. An investor who is withdrawing from a retirement portfolio over a period of 25-30 years is arguably not well-served by using daily or monthly volatility/correlations as a proxy for their risk.

For instance, compare Vanguard Small Cap Value Index (VISVX) and Vanguard Total Stock Market Index (VTSMX) as an example. I'm choosing ("cherry-picking") the years 2000 to 2016 merely to illustrate the kind of differences I'm talking about.

Take a conventional calculation of correlation between the returns of these funds and you might get what appears to be a "high" correlation. PortfolioVisualizer calculates it as 0.93. You might use "common sense" and conclude from that correlation that there is not much diversification benefit to be had.

But take a look at two hypothetical portfolios in decumulation, each starting with $100k and withdrawing $400/month. One is 50% VISVX + 50% VTSMX, and the other is 100% VTSMX.

Image

Obviously we could argue about the allocation, the withdrawal rate, the timeframe, etc. but I'd prefer to not get dragged into that.

I think most people would not look at two US equity funds with a correlation of 0.93 and expect that kind of performance difference is even possible.


I don't expect that kind of difference to happen over the next 20 years, but people should be aware - IMHO - that it is possible. Investors in general have a strong knee-jerk aversion to diversification, fueled by biases that anchor on recent performance and/or a desire to own the "best" asset class/fund/stock. I understand that, and am sympathetic to the impulses, but we shouldn't let those biases lead us to blithely ridicule investment concepts like "be diversified" and "stay the course".

I'm honestly not sure an individual investor needs a complete theoretical understanding of all these concepts, but there are textbooks on it. Portfolio Diversification by Francois-Serge Lhabitant is probably the most on-point but it still omits some important practical considerations.

Hope that helps some....
Yes, this is helpful. It gives a person an example of the kinds of things to try to look at that quantify the outcome.
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Why don't you factor tilt?

Post by Apathizer »

DaufuskieNate wrote: Tue Sep 27, 2022 10:58 am
dbr wrote: Tue Sep 27, 2022 9:59 am The question constantly asked but seemingly seldom answered is how would a factor investor decide what factor loading to target and why.
It's a great question and in my opinion the answer gets into personal preferences and appetite for risk. One person may want the maximum possible tilt in pursuit of the possibility of a high return. Another person may mix a high tilt equity portfolio with a larger allocation to safe fixed income. Others may not be able to tolerate a significant tracking error relative to TSM in their equity allocation and opt for a mild factor loading. Just like the overall stock/bond mix, there is no one right answer for everyone.

Factor investing should not be taken lightly. In my opinion, one needs to study the evidence and theory carefully in order to understand what they are getting into. Looking at how various factor allocations have performed in a variety of market conditions in the past is also important as an indication of what could happen in the future. A long-term commitment is essential, and having a solid understanding is an important part of being able to stay the course in the inevitable tough times.
That's why I use Avantis and DFA. They have extensive academic and practical implementation of factor slants, so I defer to them since they understand it at least as well as anyone does. Certainly better than I ever will.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
rkhusky
Posts: 17654
Joined: Thu Aug 18, 2011 8:09 pm

Re: Why don't you factor tilt?

Post by rkhusky »

vineviz wrote: Tue Sep 27, 2022 10:30 am For instance, compare Vanguard Small Cap Value Index (VISVX) and Vanguard Total Stock Market Index (VTSMX) as an example. I'm choosing ("cherry-picking") the years 2000 to 2016 merely to illustrate the kind of differences I'm talking about.

Take a conventional calculation of correlation between the returns of these funds and you might get what appears to be a "high" correlation. PortfolioVisualizer calculates it as 0.93. You might use "common sense" and conclude from that correlation that there is not much diversification benefit to be had.

But take a look at two hypothetical portfolios in decumulation, each starting with $100k and withdrawing $400/month. One is 50% VISVX + 50% VTSMX, and the other is 100% VTSMX.

Image

Obviously we could argue about the allocation, the withdrawal rate, the timeframe, etc. but I'd prefer to not get dragged into that.

I think most people would not look at two US equity funds with a correlation of 0.93 and expect that kind of performance difference is even possible.
The difference is due to VISVX out-performing VTSMX during that particular period, not to the correlation.

https://www.portfoliovisualizer.com/bac ... ion2_2=100
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Why don't you factor tilt?

Post by vineviz »

rkhusky wrote: Tue Sep 27, 2022 7:16 pm The difference is due to VISVX out-performing VTSMX during that particular period, not to the correlation.
No kidding.

That's exactly my point: correlations don't tell the whole diversification story. And, in fact, can be outright misleading if not measured and interpreted correctly.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
rkhusky
Posts: 17654
Joined: Thu Aug 18, 2011 8:09 pm

Re: Why don't you factor tilt?

Post by rkhusky »

vineviz wrote: Tue Sep 27, 2022 7:24 pm
rkhusky wrote: Tue Sep 27, 2022 7:16 pm The difference is due to VISVX out-performing VTSMX during that particular period, not to the correlation.
No kidding.

That's exactly my point: correlations don't tell the whole diversification story. And, in fact, can be outright misleading if not measured and interpreted correctly.
The same is true with low correlations. They aren't worth much if not accompanied by good returns.
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: Why don't you factor tilt?

Post by vineviz »

rkhusky wrote: Tue Sep 27, 2022 7:27 pm
vineviz wrote: Tue Sep 27, 2022 7:24 pm
rkhusky wrote: Tue Sep 27, 2022 7:16 pm The difference is due to VISVX out-performing VTSMX during that particular period, not to the correlation.
No kidding.

That's exactly my point: correlations don't tell the whole diversification story. And, in fact, can be outright misleading if not measured and interpreted correctly.
The same is true with low correlations. They aren't worth much if not accompanied by good returns.
The entire point of diversification is to protect the investor from uncertainty about future returns. If we knew ahead of time which assets would have the best returns there'd be no point in diversification because there'd be no risk of underperformance.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
rkhusky
Posts: 17654
Joined: Thu Aug 18, 2011 8:09 pm

Re: Why don't you factor tilt?

Post by rkhusky »

vineviz wrote: Tue Sep 27, 2022 7:33 pm The entire point of diversification is to protect the investor from uncertainty about future returns. If we knew ahead of time which assets would have the best returns there'd be no point in diversification because there'd be no risk of underperformance.
A good argument for using a total market fund that is well diversified with virtually all the stock assets.

Note that the above example also works when replacing the SCV fund with a SCG fund:
https://www.portfoliovisualizer.com/bac ... ion2_2=100
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Why don't you factor tilt?

Post by Apathizer »

rkhusky wrote: Tue Sep 27, 2022 7:36 pm
vineviz wrote: Tue Sep 27, 2022 7:33 pm The entire point of diversification is to protect the investor from uncertainty about future returns. If we knew ahead of time which assets would have the best returns there'd be no point in diversification because there'd be no risk of underperformance.
A good argument for using a total market fund that is well diversified with virtually all the stock assets.

Note that the above example also works when replacing the SCV fund with a SCG fund:
https://www.portfoliovisualizer.com/bac ... ion2_2=100
Small growth is closely correlated with and has lower risk adjusted returns than large growth, so there's no practical benefit. Small caps as a whole closely mirror the total market. The size premium mainly originated from value and core stocks, so including growth is counterproductive.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
rkhusky
Posts: 17654
Joined: Thu Aug 18, 2011 8:09 pm

Re: Why don't you factor tilt?

Post by rkhusky »

Apathizer wrote: Tue Sep 27, 2022 7:53 pm
rkhusky wrote: Tue Sep 27, 2022 7:36 pm
vineviz wrote: Tue Sep 27, 2022 7:33 pm The entire point of diversification is to protect the investor from uncertainty about future returns. If we knew ahead of time which assets would have the best returns there'd be no point in diversification because there'd be no risk of underperformance.
A good argument for using a total market fund that is well diversified with virtually all the stock assets.

Note that the above example also works when replacing the SCV fund with a SCG fund:
https://www.portfoliovisualizer.com/bac ... ion2_2=100
Small growth is closely correlated with and has lower risk adjusted returns than large growth, so there's no practical benefit. Small caps as a whole closely mirror the total market. The size premium mainly originated from value and core stocks, so including growth is counterproductive.
But no one knows what the future holds. Better to be well diversified.
stan1
Posts: 14235
Joined: Mon Oct 08, 2007 4:35 pm

Re: Why don't you factor tilt?

Post by stan1 »

rkhusky wrote: Tue Sep 27, 2022 8:01 pm But no one knows what the future holds. Better to be well diversified.
We all agree with that statement. Maybe we leave it at that and move on to a more straight forward international or mortgage pay-off thread?
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Why don't you factor tilt?

Post by Apathizer »

rkhusky wrote: Tue Sep 27, 2022 8:01 pm
Apathizer wrote: Tue Sep 27, 2022 7:53 pm
rkhusky wrote: Tue Sep 27, 2022 7:36 pm
vineviz wrote: Tue Sep 27, 2022 7:33 pm The entire point of diversification is to protect the investor from uncertainty about future returns. If we knew ahead of time which assets would have the best returns there'd be no point in diversification because there'd be no risk of underperformance.
A good argument for using a total market fund that is well diversified with virtually all the stock assets.

Note that the above example also works when replacing the SCV fund with a SCG fund:
https://www.portfoliovisualizer.com/bac ... ion2_2=100
Small growth is closely correlated with and has lower risk adjusted returns than large growth, so there's no practical benefit. Small caps as a whole closely mirror the total market. The size premium mainly originated from value and core stocks, so including growth is counterproductive.
But no one knows what the future holds. Better to be well diversified.
There's no reason to expect most highly priced low profitability stocks to have good returns. The vast majority almost certainly won't.

I agree it's better to stay well diversified with Factor slants. That way you're adding more diversity to a cap weight TSM portfolio.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
rkhusky
Posts: 17654
Joined: Thu Aug 18, 2011 8:09 pm

Re: Why don't you factor tilt?

Post by rkhusky »

Apathizer wrote: Tue Sep 27, 2022 8:46 pm There's no reason to expect most highly priced low profitability stocks to have good returns. The vast majority almost certainly won't.
If that is true, everyone knows it and could make money by investing opposite to the market.
Random Walker
Posts: 5561
Joined: Fri Feb 23, 2007 7:21 pm

Re: Why don't you factor tilt?

Post by Random Walker »

rkhusky wrote: Tue Sep 27, 2022 9:21 pm
Apathizer wrote: Tue Sep 27, 2022 8:46 pm There's no reason to expect most highly priced low profitability stocks to have good returns. The vast majority almost certainly won't.
If that is true, everyone knows it and could make money by investing opposite to the market.
I have a strong tendency to agree with you on this. I’m a huge believer in market efficiency. BUT it seems us humans are tenaciously hard wired with a gambling instinct; we tend to overpay for growth, hoping to get a piece of a huge right tail 6 SD event. There are big limits to arbitrage and it’s way easier for the market to correct underpricing than overpricing. As I progress through my investing experience, my belief in behavioral errors gets stronger. I’m coming to believe that all the new technology serves to amplify human behavioral errors rather than eliminate them. I recommend Ilmanen’s recent book Investing Amid Low Expected Returns. It’s shorter and significantly more readable than his first.

Dave
JohnFromPNW
Posts: 194
Joined: Tue Apr 26, 2022 11:58 pm

Re: Why don't you factor tilt?

Post by JohnFromPNW »

I guess one of the reasons I don’t have a tilt as of this moment is the practical implementation of a tilt. Sure you could add small, value, or small value. But how much small, how much value? I could pick a percent but is that the right amount?

And what about profitability, investment, momentum, and volatility? Do I need those too? And how much? And how to organize those within the traditional domestic/foreign/emerging framework? One fund for each factor both domestic and international? One fund for each factor for each cap size both domestic and international?

And if I actually buy all of them will they all just offset each other and I end up back at the total market fund?
rkhusky
Posts: 17654
Joined: Thu Aug 18, 2011 8:09 pm

Re: Why don't you factor tilt?

Post by rkhusky »

Random Walker wrote: Tue Sep 27, 2022 9:50 pm
rkhusky wrote: Tue Sep 27, 2022 9:21 pm
Apathizer wrote: Tue Sep 27, 2022 8:46 pm There's no reason to expect most highly priced low profitability stocks to have good returns. The vast majority almost certainly won't.
If that is true, everyone knows it and could make money by investing opposite to the market.
I have a strong tendency to agree with you on this. I’m a huge believer in market efficiency. BUT it seems us humans are tenaciously hard wired with a gambling instinct; we tend to overpay for growth, hoping to get a piece of a huge right tail 6 SD event. There are big limits to arbitrage and it’s way easier for the market to correct underpricing than overpricing. As I progress through my investing experience, my belief in behavioral errors gets stronger. I’m coming to believe that all the new technology serves to amplify human behavioral errors rather than eliminate them. I recommend Ilmanen’s recent book Investing Amid Low Expected Returns. It’s shorter and significantly more readable than his first.

Dave
Are computer trading algorithms susceptible to human foibles? Or do they take advantage of human foibles and make the market even more efficient?
Random Walker
Posts: 5561
Joined: Fri Feb 23, 2007 7:21 pm

Re: Why don't you factor tilt?

Post by Random Walker »

JohnFromPNW wrote: Tue Sep 27, 2022 11:28 pm I guess one of the reasons I don’t have a tilt as of this moment is the practical implementation of a tilt. Sure you could add small, value, or small value. But how much small, how much value? I could pick a percent but is that the right amount?

And what about profitability, investment, momentum, and volatility? Do I need those too? And how much? And how to organize those within the traditional domestic/foreign/emerging framework? One fund for each factor both domestic and international? One fund for each factor for each cap size both domestic and international?

And if I actually buy all of them will they all just offset each other and I end up back at the total market fund?
A couple thoughts. First, makes sense to invest in each factor proportionate to your belief in each. Second, you don’t want to own an individual fund for each factor; you don’t want one fund buying a stock and another one selling the same stock. So it’s better to buy integrated funds that incorporate multiple factors. For example there are SV funds that have profitability and momentum screens and buy/hold ranges to screen out negative momentum and take a little advantage of positive momentum. Our long only portfolios are so overwhelmingly dominated by market beta, that any move towards risk parity can potentially be helpful. I wouldn’t worry so much about getting factor exposures quantitatively right. I would recommend qualitatively moving in the tilt direction to a degree you believe your tolerance for tracking variance will handle.

Dave
abc132
Posts: 2411
Joined: Thu Oct 18, 2018 1:11 am

Re: Why don't you factor tilt?

Post by abc132 »

Random Walker wrote: Tue Sep 27, 2022 9:50 pm
rkhusky wrote: Tue Sep 27, 2022 9:21 pm
Apathizer wrote: Tue Sep 27, 2022 8:46 pm There's no reason to expect most highly priced low profitability stocks to have good returns. The vast majority almost certainly won't.
If that is true, everyone knows it and could make money by investing opposite to the market.
I have a strong tendency to agree with you on this. I’m a huge believer in market efficiency. BUT it seems us humans are tenaciously hard wired with a gambling instinct; we tend to overpay for growth, hoping to get a piece of a huge right tail 6 SD event. There are big limits to arbitrage and it’s way easier for the market to correct underpricing than overpricing. As I progress through my investing experience, my belief in behavioral errors gets stronger. I’m coming to believe that all the new technology serves to amplify human behavioral errors rather than eliminate them. I recommend Ilmanen’s recent book Investing Amid Low Expected Returns. It’s shorter and significantly more readable than his first.

Dave
You have two problems here:
1) the majority of trades are being done by computers without respect to what a human would do
2) there are now other a better speculative assets not in the total stock market if you want to bet on that kind of thing
Random Walker
Posts: 5561
Joined: Fri Feb 23, 2007 7:21 pm

Re: Why don't you factor tilt?

Post by Random Walker »

rkhusky wrote: Wed Sep 28, 2022 7:06 am
Random Walker wrote: Tue Sep 27, 2022 9:50 pm
rkhusky wrote: Tue Sep 27, 2022 9:21 pm
Apathizer wrote: Tue Sep 27, 2022 8:46 pm There's no reason to expect most highly priced low profitability stocks to have good returns. The vast majority almost certainly won't.
If that is true, everyone knows it and could make money by investing opposite to the market.
I have a strong tendency to agree with you on this. I’m a huge believer in market efficiency. BUT it seems us humans are tenaciously hard wired with a gambling instinct; we tend to overpay for growth, hoping to get a piece of a huge right tail 6 SD event. There are big limits to arbitrage and it’s way easier for the market to correct underpricing than overpricing. As I progress through my investing experience, my belief in behavioral errors gets stronger. I’m coming to believe that all the new technology serves to amplify human behavioral errors rather than eliminate them. I recommend Ilmanen’s recent book Investing Amid Low Expected Returns. It’s shorter and significantly more readable than his first.

Dave
Are computer trading algorithms susceptible to human foibles? Or do they take advantage of human foibles and make the market even more efficient?
I don’t know! Fascinating question. Initial thought would be more efficient. But second thought? For example, could computers amplify momentum effects?

Dave
Random Walker
Posts: 5561
Joined: Fri Feb 23, 2007 7:21 pm

Re: Why don't you factor tilt?

Post by Random Walker »

Somehow with something like 90% of all trades between huge institutions, each backed by mega data, PhDs, computer, we got to P/E 30 range again, got huge value spread based mainly P/E expansion, and might be in process of seeing big contraction now. Year to year returns seem dominated by Bogle’s Speculative Component. Is it oxymoronic that the market is so tenaciously efficient and irrational at the same time?

Dave
rkhusky
Posts: 17654
Joined: Thu Aug 18, 2011 8:09 pm

Re: Why don't you factor tilt?

Post by rkhusky »

Random Walker wrote: Wed Sep 28, 2022 8:00 am
rkhusky wrote: Wed Sep 28, 2022 7:06 am Are computer trading algorithms susceptible to human foibles? Or do they take advantage of human foibles and make the market even more efficient?
I don’t know! Fascinating question. Initial thought would be more efficient. But second thought? For example, could computers amplify momentum effects?
On that thought, there is the voting machine (short term) versus weighing machine (long term) analogy. Perhaps the computers and day traders make the market efficient in the short term, but have little effect on the long term results of the market. Therefore, buy/hold/rebalance long term investors are not affected by computer trading.
Post Reply