Why is Large Growth disliked?

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Day9
Posts: 988
Joined: Mon Jun 11, 2012 6:22 pm

Re: Why is Large Growth disliked?

Post by Day9 »

Alchemist wrote: Wed Oct 14, 2020 5:20 am ...the case of commodities futures via PCRIX being a complete dumpster fire...
I recently re-read Larry Swedroe's last post on this forum and he clarifies that all along he recommended that if you add a commodity-linked fund like PCRIX to also increase the duration of your bond portfolio, and if you had done that you would have come out the same because duration risk has done well since he made those recommendations. Furthermore he clarified that all along he recommended the small allocation to commodity-linked investments as a kind of "insurance" in cases where both stocks and bonds do poorly (e.g. stagflation) and thankfully we have not had that kind of environment since he was recommending it. You don't consider it a mistake if you buy home insurance but your home never burned down.

Like uncorrelated says, you miss a lot by looking at an investment in isolation rather than its role in a portfolio.
I'm just a fan of the person I got my user name from
3funder
Posts: 1449
Joined: Sun Oct 15, 2017 9:35 pm

Re: Why is Large Growth disliked?

Post by 3funder »

I disagree with your premise that they're disliked. Large growth stocks have seen a nice runup. I wouldn't invest a large lump sum of money in them right now (I think international stocks are a better value if one wishes to invest a large lump sum), but that doesn't mean I dislike them.
ArmchairArchitect
Posts: 321
Joined: Thu Nov 05, 2015 10:37 am

Re: Why is Large Growth disliked?

Post by ArmchairArchitect »

I switched a lot of my holdings from VTSAX to a few Vanguard large growth funds after I updated and solicited responses to this old thread, right after I saw how resilient they were in the COVID market crash:

viewtopic.php?f=10&t=1542&p=5238866#p5238866

I was attracted by the higher upside in a good market, long-term performance (10 year and "since inception") but more importantly, the resiliency as compared to VTSAX during the COVID crash/bad times.

Keep in mind too that VTSAX includes all those Chinese stocks (eg. Luckin Coffee) that don't have to meet PCAOB auditing requirements like everyone else; that's reason enough alone to avoid it (even if they don't make up a huge portion yet).
User avatar
nedsaid
Posts: 13815
Joined: Fri Nov 23, 2012 12:33 pm

Re: Why is Large Growth disliked?

Post by nedsaid »

whereskyle wrote: Sat Oct 10, 2020 7:36 pm

For some reason factor investing gets a ton of attention here even though Jack openly criticized it and never recommended it. I have no idea why it excites people, and I wish people would just listen to Jack.
A gentle reminder that Vanguard Value Index and Vanguard Growth Index were started when Bogle ran things at Vanguard.
A fool and his money are good for business.
whereskyle
Posts: 1252
Joined: Wed Jan 29, 2020 10:29 am

Re: Why is Large Growth disliked?

Post by whereskyle »

nedsaid wrote: Thu Oct 15, 2020 4:48 pm
whereskyle wrote: Sat Oct 10, 2020 7:36 pm

For some reason factor investing gets a ton of attention here even though Jack openly criticized it and never recommended it. I have no idea why it excites people, and I wish people would just listen to Jack.
A gentle reminder that Vanguard Value Index and Vanguard Growth Index were started when Bogle ran things at Vanguard.
And he expressly regretted it.
"I am better off than he is – for he knows nothing and thinks that he knows. I neither know nor think that I know." - Socrates. "Nobody knows nothing." - Jack Bogle
User avatar
nedsaid
Posts: 13815
Joined: Fri Nov 23, 2012 12:33 pm

Re: Why is Large Growth disliked?

Post by nedsaid »

whereskyle wrote: Thu Oct 15, 2020 5:30 pm
nedsaid wrote: Thu Oct 15, 2020 4:48 pm
whereskyle wrote: Sat Oct 10, 2020 7:36 pm

For some reason factor investing gets a ton of attention here even though Jack openly criticized it and never recommended it. I have no idea why it excites people, and I wish people would just listen to Jack.
A gentle reminder that Vanguard Value Index and Vanguard Growth Index were started when Bogle ran things at Vanguard.
And he expressly regretted it.
Well Bogle did a lot of surprising things. He envisioned that people would invest in Growth Index during their working years and switch to Value Index when they retired, one reason is that Value has the higher income from dividends. He told people not to trade and that these were long term investments. His view is that Value and Growth even out over long periods of time.
A fool and his money are good for business.
Day9
Posts: 988
Joined: Mon Jun 11, 2012 6:22 pm

Re: Why is Large Growth disliked?

Post by Day9 »

nedsaid wrote: Thu Oct 15, 2020 4:48 pm
whereskyle wrote: Sat Oct 10, 2020 7:36 pm

For some reason factor investing gets a ton of attention here even though Jack openly criticized it and never recommended it. I have no idea why it excites people, and I wish people would just listen to Jack.
A gentle reminder that Vanguard Value Index and Vanguard Growth Index were started when Bogle ran things at Vanguard.
In an interview from around 20 years ago Bogle was asked how the "new" Small Growth Index Fund is consistent with his investing philosophy. He gave the example of someone who had a portfolio of blue chip stocks in a taxable account with high capital gains. Instead of realizing the capital gains and paying taxes, they could put new contributions toward the Small Growth Index fund to make their allocation more like the total market.
I'm just a fan of the person I got my user name from
kabob
Posts: 110
Joined: Wed Oct 16, 2019 9:01 am
Location: Loudon, Tn

Re: Why is Large Growth disliked?

Post by kabob »

Of all 302 Vanguard Funds Growth is KING this year... No Doubt about it.
Image
Growth, including International Growth, MegaCap, MidCap, SmallCap - just dominates the best preforming funds YTD.
With Tech & Consumer Discretionary too.
Wowsa, Yowsa, No complaints here...
Carol88888
Posts: 301
Joined: Wed Jan 24, 2018 2:24 am

Re: Why is Large Growth disliked?

Post by Carol88888 »

nisiprius wrote: Sat Oct 10, 2020 1:29 pm
Doctor Rhythm wrote: Sat Oct 10, 2020 1:20 pm...Also, are “small” and “value” each individually predictive or is the combination somehow synergistic?...
Well, I'm trying to present points of view that I don't agree with, but I would say that factor mavens suggest they are synergistic. For example, the size factor itself--the idea that small-caps have a higher return (possibly) or a higher risk-adjusted return (much more dubious)--is now acknowledged to be weak and one of the less important factors, but small-cap value is well-regarded. Cliff Asness of AQR has essays in which he says that the size factor itself is practically non-existent, yet the combination of size and "quality" is highly valuable. Really, though, they need to be explained by an advocate, and that isn't me.

I don't know if it is worse in finance than in other scientific fields, but it is very noticeable that a fair number of people who publish attention-getting academic papers either are, or soon become, professionally involved with fund companies that offer products exploiting their discoveries. Fama and French are connected with Dimensional Fund Advisors (DFA), for example.
And once the new products have arrived to take advantage of these discoveries is about the time they stop working. Knowledge that everyone has (small cap value produces highest return) is knowledge that no longer works.
User avatar
nedsaid
Posts: 13815
Joined: Fri Nov 23, 2012 12:33 pm

Re: Why is Large Growth disliked?

Post by nedsaid »

Day9 wrote: Thu Oct 15, 2020 6:45 pm
nedsaid wrote: Thu Oct 15, 2020 4:48 pm
whereskyle wrote: Sat Oct 10, 2020 7:36 pm

For some reason factor investing gets a ton of attention here even though Jack openly criticized it and never recommended it. I have no idea why it excites people, and I wish people would just listen to Jack.
A gentle reminder that Vanguard Value Index and Vanguard Growth Index were started when Bogle ran things at Vanguard.
In an interview from around 20 years ago Bogle was asked how the "new" Small Growth Index Fund is consistent with his investing philosophy. He gave the example of someone who had a portfolio of blue chip stocks in a taxable account with high capital gains. Instead of realizing the capital gains and paying taxes, they could put new contributions toward the Small Growth Index fund to make their allocation more like the total market.
Yes, there are a number of good reasons why someone would want to invest in these style based Indexes. We have lots of tools and financial products now to do such things as balancing out our portfolios. We really are in a golden era of investing.
A fool and his money are good for business.
heyyou
Posts: 3796
Joined: Tue Feb 20, 2007 4:58 pm

Re: Why is Large Growth disliked?

Post by heyyou »

Why is Large Growth disliked?
The 2000 Crash delayed my retirement by several years and I was only in broad stock and bond index funds. The LG losses were so large, I no longer trusted the next recovery to be sustainable. Thus I needed more assets after the crash than before the crash. I now have more pension, more SS from both longer contributions and delayed initiation, and a full decade of annual spending in bond index funds earning zilch right now.

Owning stock index funds is good for growth, but said growth is not a reliable source of retirement income. You have to sell the growth only when it is doing well, to then buy something else to have steady income during retirement, so there are the aggravations of imperfect timing, both selling too soon and selling too late.

Notice how LG problems influence your other stock investments in different size and value categories, so you can't hide anywhere else in other stock categories to avoid the effects of LG's losses. Small and value stock losses do not always spread their contagion to LG prices.
Alchemist
Posts: 558
Joined: Sat Aug 30, 2014 6:35 am

Re: Why is Large Growth disliked?

Post by Alchemist »

Uncorrelated wrote: Wed Oct 14, 2020 5:51 am DFSVX realized a gain of 0.79% per year relative to TSM, after expenses.
This is false. DFSVX grew $10k to $126,893 from inception until 30 Sept 2020. VTSMX grew $10k into $127,765 over that same time period. There was *no* premium. They tied in performance, but DFSVX was far more volatile with 20% deeper drawdown. VTSMX returned the same at lower levels of risk.

The story is the same if you 'slice and dice' with just a tilt. DFSVX added no more return, but increased volatility. The predictions of a SCV premium harvestable by DFSVX were wrong. It is that simple.

There is no purpose in continuing a conversation with someone who wants to re-define basic tenets of investing and alter the goal post location at supersonic speeds. Speaking of goal post movement...
Day9 wrote: Wed Oct 14, 2020 10:01 am I recently re-read Larry Swedroe's last post on this forum and he clarifies that all along he recommended that if you add a commodity-linked fund like PCRIX to also increase the duration of your bond portfolio, and if you had done that you would have come out the same because duration risk has done well since he made those recommendations. Furthermore he clarified that all along he recommended the small allocation to commodity-linked investments as a kind of "insurance" in cases where both stocks and bonds do poorly (e.g. stagflation) and thankfully we have not had that kind of environment since he was recommending it. You don't consider it a mistake if you buy home insurance but your home never burned down.
PCRIX was supposed to protect against inflation. It has a negative real return of 27 basis points since inception 18 years ago. It has a nominal return of 1.77% which underperformed FDIC insured cash. It is currently more than 60% below its peak (when Swedroe was recommending it..).

It was a bad recommendation by any measure. If you consider it 'insurance' it didn't even pay out on the claim (failed to beat inflation). TIPs are a far simpler, reliable, and costless way to insure against inflation. PCRIX was sold to make PIMCO money, it succeeded in that goal thanks to its 1.33% expense ratio.

PV link: https://www.portfoliovisualizer.com/bac ... ion1_1=100
Day9 wrote:Like uncorrelated says, you miss a lot by looking at an investment in isolation rather than its role in a portfolio.
A simple Jack Bogle two-fund, or Taylor Larimore Three fund portfolio has done better than many complicated portfolio's put together by expensive AUM advisors. Adding multi-factor tilts, commodities, Alt Funds, ect have not added safety or return but they have added complexity and cost. I am not accusing Swedroe or others like him of nefarious intent. They have been seduced by an illusion that building complex models based off the past will provide accurate predictions of the future. Those models sure feel scientific with the complicated mathematical algorithms. But they cannot account for all variables and they certainly cannot account for variables that do not exist in the past such as the development of new technology such as hydraulic fracturing (fracking) which cratered PCRIX.

"Past return does not predict future results" is simple and well known yet constantly tossed aside by very intelligent, well meaning people in the financial world.

It should not be ignored in the case of SCV or Large Growth.
langlands
Posts: 563
Joined: Wed Apr 03, 2019 10:05 pm

Re: Why is Large Growth disliked?

Post by langlands »

Alchemist wrote: Fri Oct 16, 2020 8:32 pm
Uncorrelated wrote: Wed Oct 14, 2020 5:51 am DFSVX realized a gain of 0.79% per year relative to TSM, after expenses.
This is false. DFSVX grew $10k to $126,893 from inception until 30 Sept 2020. VTSMX grew $10k into $127,765 over that same time period. There was *no* premium. They tied in performance, but DFSVX was far more volatile with 20% deeper drawdown. VTSMX returned the same at lower levels of risk.

The story is the same if you 'slice and dice' with just a tilt. DFSVX added no more return, but increased volatility. The predictions of a SCV premium harvestable by DFSVX were wrong. It is that simple.
What you posted and what Uncorrelated posted are not contradictory. When Uncorrelated says that DFSVX has a gain of 0.79% per year relative to TSM, he meant the arithmetic return, not the geometric return. (Reread his post in which he explictly details the difference between the two). Your comparison of DFSVX and VTSMX involving a fixed initial outlay and seeing the ending performance of the two without any further contributions is exactly measuring the geometric return.

Uncorrelated is saying that DFSVX has a 0.79% higher arithmetic return than VTSMX. You are saying that DFSVX has a lower geometric return than VTSMX. Both are true. The reason as you point out is the volatility. In precise mathematical returns, remember that

geometric return = arithmetic return - volatility^2/2

If the only two choices you had were 100% VTSMX or 100% DFSVX, I'd agree with you that 100% VTSMX is probably the better choice. But those aren't the only two choices. If you trust that the higher arithmetic return of DFSVX over TSM will continue into the future, there can possibly be a place for DFSVX in your portfolio. You would need estimates of the correlation between DFSVX and VTSMX and the volatility of DFSVX and VTSMX as well to figure out if it makes sense. As others have mentioned, it isn't that simple. Whether a particular asset can be a valuable addition to a portfolio cannot be determined based solely on a head to head portfolio visualizer battle.
User avatar
Uncorrelated
Posts: 1039
Joined: Sun Oct 13, 2019 3:16 pm

Re: Why is Large Growth disliked?

Post by Uncorrelated »

Alchemist wrote: Fri Oct 16, 2020 8:32 pm
Uncorrelated wrote: Wed Oct 14, 2020 5:51 am DFSVX realized a gain of 0.79% per year relative to TSM, after expenses.
This is false. DFSVX grew $10k to $126,893 from inception until 30 Sept 2020. VTSMX grew $10k into $127,765 over that same time period. There was *no* premium. They tied in performance, but DFSVX was far more volatile with 20% deeper drawdown. VTSMX returned the same at lower levels of risk.

The story is the same if you 'slice and dice' with just a tilt. DFSVX added no more return, but increased volatility. The predictions of a SCV premium harvestable by DFSVX were wrong. It is that simple.

There is no purpose in continuing a conversation with someone who wants to re-define basic tenets of investing and alter the goal post location at supersonic speeds. Speaking of goal post movement...
It's not my fault the academic literature defines premium as the difference in arithmetic returns.

If you disagree with that definition, just state so. But remember that the geometric return is a non-linear combination of the funds in a portfolio, if you want to show that DSFVS did not offer higher geometric returns, you have to show that is the case for every possible portfolio. If you use arithmetic returns, you have to show it for just one (that is where the 0.79% figure comes from, and that's after expenses).

To illustrate:
Image
100% total stock market vs DFSVX, total stock market wins. link

Image
60/40 portfolio with either total stock market or DFSVX, DFSVX wins. link
Alchemist
Posts: 558
Joined: Sat Aug 30, 2014 6:35 am

Re: Why is Large Growth disliked?

Post by Alchemist »

Uncorrelated wrote: Sat Oct 17, 2020 1:39 am It's not my fault the academic literature defines premium as the difference in arithmetic returns.
If that is the case, then the factor research is even more useless than I had originally thought.

To illustrate:
Image
100% total stock market vs DFSVX, total stock market wins. link

Image
60/40 portfolio with either total stock market or DFSVX, DFSVX wins. link
Any reasonable definition would be a tie. If you think a difference of 0.14% is meaningful over 27 years then go bananas. I see increased cost, risk, and complexity with no benefit.
garlandwhizzer
Posts: 2987
Joined: Fri Aug 06, 2010 3:42 pm

Re: Why is Large Growth disliked?

Post by garlandwhizzer »

If you are rebalancing annually over a long period of time there's a benefit to having a volatile asset like DFSVX. When DFSVX is suffering badly, its prices crashing, you move money from a stable assets like TBM into something that's on sale. When the market turns around as it always does, the volatile asset rather than suffering the downside of increased volatility produces portfolio benefits from the upside of volatility. In the example given above comparing the 60 TSM/40 TBM to 60 DFSVX/40 TBM over a 27 year period starting with the inception of DFSVX, the balanced DFSVX does outperform reflecting this annual rebalancing benefit between a stable value asset and a very volatile one. What is remarkable however, is how small the difference in CAGR is over this long time period (27 years), 0.14%, not exactly a huge difference. What is remarkable however is the huge differences in favor of the TSM balanced portfolio in terms of lower maximal portfolio drawdown (40% vs 32.4%), portfolio volatility (12.1% versus 9.3%). Consequently the TSM balanced portfolio produces a much higher Sharpe ratio (.64 vs .53), indication that for a given level of portfolio risk and volatility it produced significantly higher returns. If we were to to increase equity % in the TSM balanced portfolio to the level where it equated with the volatility and risk of the DFSVX balanced portfolio, the former would have hugely outperformed the latter.

Based on this example using real funds, not artificial models designed to magnify factor results, and choosing a SCV fund run by acknowledged experts over a long time frame, it seems reasonable to conclude that if there is any SCV outperformance at all, it comes from rebalancing from a stable bond asset into a volatile equity asset and further that the degree of that outperformance is rather pathetic relative to the increased risk and volatility required to achieve it. It would have been much more effective to simply increase TSM % relative to bonds in the original balanced portfolio which would have produced much better returns than the SCV balanced portfolio in relation to risk taken.

Garland Whizzer
000
Posts: 2716
Joined: Thu Jul 23, 2020 12:04 am

Re: Why is Large Growth disliked?

Post by 000 »

Uncorrelated wrote: Sat Oct 17, 2020 1:39 am
Alchemist wrote: Fri Oct 16, 2020 8:32 pm
Uncorrelated wrote: Wed Oct 14, 2020 5:51 am DFSVX realized a gain of 0.79% per year relative to TSM, after expenses.
This is false. DFSVX grew $10k to $126,893 from inception until 30 Sept 2020. VTSMX grew $10k into $127,765 over that same time period. There was *no* premium. They tied in performance, but DFSVX was far more volatile with 20% deeper drawdown. VTSMX returned the same at lower levels of risk.

The story is the same if you 'slice and dice' with just a tilt. DFSVX added no more return, but increased volatility. The predictions of a SCV premium harvestable by DFSVX were wrong. It is that simple.

There is no purpose in continuing a conversation with someone who wants to re-define basic tenets of investing and alter the goal post location at supersonic speeds. Speaking of goal post movement...
It's not my fault the academic literature defines premium as the difference in arithmetic returns.

If you disagree with that definition, just state so. But remember that the geometric return is a non-linear combination of the funds in a portfolio, if you want to show that DSFVS did not offer higher geometric returns, you have to show that is the case for every possible portfolio. If you use arithmetic returns, you have to show it for just one (that is where the 0.79% figure comes from, and that's after expenses).

To illustrate:
Image
100% total stock market vs DFSVX, total stock market wins. link

Image
60/40 portfolio with either total stock market or DFSVX, DFSVX wins. link
I see you are "forgetting" about volatility-adjusted returns...
User avatar
Uncorrelated
Posts: 1039
Joined: Sun Oct 13, 2019 3:16 pm

Re: Why is Large Growth disliked?

Post by Uncorrelated »

000 wrote: Sat Oct 17, 2020 4:31 pm
I see you are "forgetting" about volatility-adjusted returns...
I purposefully omitted them because I was responding to the claim that there was no outperformance.

From my point of view risk refers to the dispersion in possible outcomes. When you run a backtest, you get one outcome and therefore there is no risk. This makes backtests fundamentally unsuited to asses risk adjusted performance, a monte-carlo simulation would be more useful. I purposefully selected simple (non-optimal) portfolio's because that wasn't integral to the point. Of course I'm not suggesting anyone to use 60% DFSVX and 40% bonds.

My mean-variance analyzer is capable of estimating the efficient frontier assuming future factor returns have the same mean and variance as past returns. According to this method portfolio's with value/scv/bonds dominate portfolio's with only TSM/bonds in the time period 1993-2020ish.

It would be nice if we could elevate this discussion to something more substantial. If you want to know whether a factor has a positive return, read an academic paper. If you want to know whether fund X results in improved risk adjusted returns (err, certainty equivalent return), run a mean-variance analysis, monte-carlo simulation, or something like that. Backtests are the wrong tool for the job.
000
Posts: 2716
Joined: Thu Jul 23, 2020 12:04 am

Re: Why is Large Growth disliked?

Post by 000 »

Uncorrelated wrote: Sat Oct 17, 2020 5:46 pm I purposefully omitted them because I was responding to the claim that there was no outperformance.

From my point of view risk refers to the dispersion in possible outcomes. When you run a backtest, you get one outcome and therefore there is no risk. This makes backtests fundamentally unsuited to asses risk adjusted performance, a monte-carlo simulation would be more useful. I purposefully selected simple (non-optimal) portfolio's because that wasn't integral to the point. Of course I'm not suggesting anyone to use 60% DFSVX and 40% bonds.

My mean-variance analyzer is capable of estimating the efficient frontier assuming future factor returns have the same mean and variance as past returns. According to this method portfolio's with value/scv/bonds dominate portfolio's with only TSM/bonds in the time period 1993-2020ish.

It would be nice if we could elevate this discussion to something more substantial. If you want to know whether something outperformed, read an academic paper [studying past data]. If you want to know whether fund X results in improved risk adjusted returns (err, certainty equivalent return), run a mean-variance analysis, monte-carlo simulation, or something like that. Backtests are the wrong tool for the job.
Contradiction detected.

Academic backtesting may be more sophisticated, but it has not had and seems unlikely to have strong predictive power.

SCV was discovered and marketed solely based on past data, not based on a fundamental analysis of why it ought to be that way, like the fundamental analysis concerning the different nature of stocks and bonds.

If you want to elevate the discussion, explain why applying a few simple factor screens should provide a free lunch.
langlands
Posts: 563
Joined: Wed Apr 03, 2019 10:05 pm

Re: Why is Large Growth disliked?

Post by langlands »

Uncorrelated wrote: Sat Oct 17, 2020 5:46 pm
000 wrote: Sat Oct 17, 2020 4:31 pm
I see you are "forgetting" about volatility-adjusted returns...
From my point of view risk refers to the dispersion in possible outcomes. When you run a backtest, you get one outcome and therefore there is no risk. This makes backtests fundamentally unsuited to asses risk adjusted performance, a monte-carlo simulation would be more useful. I purposefully selected simple (non-optimal) portfolio's because that wasn't integral to the point. Of course I'm not suggesting anyone to use 60% DFSVX and 40% bonds.

My mean-variance analyzer is capable of estimating the efficient frontier assuming future factor returns have the same mean and variance as past returns. According to this method portfolio's with value/scv/bonds dominate portfolio's with only TSM/bonds in the time period 1993-2020ish.
Something doesn't quite add up. If you don't mind, could you compute the optimal gamma=1 portfolio with value/scv/bonds vs. with TSM/bonds? If what you're saying is true, the former should beat the latter in a straight up portfolio visualizer simulation from 1993-2020. That would also be much more likely to satisfy Alchemist and 000 and most people who intuitively understand CAGR (i.e. gamma = 1 utility) better than any another utility, although I know you think gamma = 1 is way more risky than the utility function most people actually have.
User avatar
Uncorrelated
Posts: 1039
Joined: Sun Oct 13, 2019 3:16 pm

Re: Why is Large Growth disliked?

Post by Uncorrelated »

000 wrote: Sat Oct 17, 2020 5:54 pm
Uncorrelated wrote: Sat Oct 17, 2020 5:46 pm I purposefully omitted them because I was responding to the claim that there was no outperformance.

From my point of view risk refers to the dispersion in possible outcomes. When you run a backtest, you get one outcome and therefore there is no risk. This makes backtests fundamentally unsuited to asses risk adjusted performance, a monte-carlo simulation would be more useful. I purposefully selected simple (non-optimal) portfolio's because that wasn't integral to the point. Of course I'm not suggesting anyone to use 60% DFSVX and 40% bonds.

My mean-variance analyzer is capable of estimating the efficient frontier assuming future factor returns have the same mean and variance as past returns. According to this method portfolio's with value/scv/bonds dominate portfolio's with only TSM/bonds in the time period 1993-2020ish.

It would be nice if we could elevate this discussion to something more substantial. If you want to know whether something outperformed, read an academic paper [studying past data]. If you want to know whether fund X results in improved risk adjusted returns (err, certainty equivalent return), run a mean-variance analysis, monte-carlo simulation, or something like that. Backtests are the wrong tool for the job.
Contradiction detected.

Academic backtesting may be more sophisticated, but it has not had and seems unlikely to have strong predictive power.

SCV was discovered and marketed solely based on past data, not based on a fundamental analysis of why it ought to be that way, like the fundamental analysis concerning the different nature of stocks and bonds.

Academics don't backtest, they create models and test them for validity. A backtest isn't the only way to evaluate past data.

If you don't believe in backtesting, why do you time the market based on recent performance? Can you show me (with statistical tests) that "emerging markets is uncompensated risk" is stronger than "hml is uncompensated risk"?

You state that academic backtesting had not and seems unlikely to have predictive power. But meanwhile, the out-of-sample HmL premium is stronger than the market premium in 3 of 4 geographic regions and Fama/French state that even in the worst geographic region, they were unable to show that the out-of-sample HmL premium was lower than the in-sample HmL premium. I can understand if you think there is not enough evidence, but there is simply no evidence that the models are wrong or not predictive. See:

Image
i.e. there is a 87% chance the expected return pre-publication and post-publication is identical. https://papers.ssrn.com/sol3/papers.cfm ... id=3525096

The equity risk premium is just as data mined as HmL. The outperformance was there first, then people started making up models that explains this outperformance, and then we had a model (CAPM) that can explain the equity risk premium. This is the same process that created Fama/French 3 and Newton's laws of motion.
If you want to elevate the discussion, explain why applying a few simple factor screens should provide a free lunch.
My personal opinion is that HmL is partially a risk story, and partially a behavioral story. The statistical evidence for HmL is stronger than that for any other active bet (except for the equity risk premium itself and BAB, which are comparable). Take note that if HmL is 100% behavioral, it's a free lunch. If HmL is 100% a risk story, the risk is compensated. This is in great contrast to other active bets such as market timing and country betting, which (apart from being statistically non-existent) very likely represent uncompensated risks.

Fama says determining whether HmL is a risk or a behavioral story is unresolvable. I like to discuss about things that are solvable, testable hypothesis such as "there was no value premium in time period X". Many people have claimed that there was no premium post-publication, the statistical tests do not indicate that.

langlands wrote: Sat Oct 17, 2020 5:57 pm
Uncorrelated wrote: Sat Oct 17, 2020 5:46 pm
000 wrote: Sat Oct 17, 2020 4:31 pm
I see you are "forgetting" about volatility-adjusted returns...
From my point of view risk refers to the dispersion in possible outcomes. When you run a backtest, you get one outcome and therefore there is no risk. This makes backtests fundamentally unsuited to asses risk adjusted performance, a monte-carlo simulation would be more useful. I purposefully selected simple (non-optimal) portfolio's because that wasn't integral to the point. Of course I'm not suggesting anyone to use 60% DFSVX and 40% bonds.

My mean-variance analyzer is capable of estimating the efficient frontier assuming future factor returns have the same mean and variance as past returns. According to this method portfolio's with value/scv/bonds dominate portfolio's with only TSM/bonds in the time period 1993-2020ish.
Something doesn't quite add up. If you don't mind, could you compute the optimal gamma=1 portfolio with value/scv/bonds vs. with TSM/bonds? If what you're saying is true, the former should beat the latter in a straight up portfolio visualizer simulation from 1993-2020. That would also be much more likely to satisfy Alchemist and 000 and most people who intuitively understand CAGR (i.e. gamma = 1 utility) better than any another utility, although I know you think gamma = 1 is way more risky than the utility function most people actually have.
I can't show that directly because my mean variance analyzer uses a factor clone of IJS (S&P600 small cap value), that fund only goes back to 2000. If we substitute DFSVX and add the difference in expense ratio, small cap value wins over total stock market (in terms of CAGR). This exact benchmark was already rejected multiple times.

Among every other part of the efficient frontier, my analyzer recommends value instead of small cap value. There are no decent value funds that have existed for as long as DFSVX, but there are some actively managed funds with a value tilt, such as AWSHX (Washington Mutual Investors Fund, 0.59% ER, large cap, positive loadings in hml/CmA/RmW, not statistically significant alpha). If we use this fund with PV's efficient frontier, we find:

Image
at the 10% stddev efficient frontier pv link. The portfolio with value has 0.4% higher expense ratio, you can definitely do better with today's funds.

Image
At the 15% stddev efficient frontier. pv link, The portfolio with value has a 0.5% higher expense ratio, you can definitely do better with today's funds.

There are more problems with backtest than gamma = 1. Backtests are very sensitive to overfitting, they can be made to show almost anything. Maybe some of you will see these backtests and think "value did indeed outperform post-publication", but realistically speaking, it's very likely the backtest played out in this way because of some unknown confounding variable. Even though this backtest supports my point, I believe the only correct thing to do here is to throw it away and demand actual evidence in the form of statistical tests. Preferable something based on more than 27 years of history.
acegolfer
Posts: 2281
Joined: Tue Aug 25, 2009 9:40 am

Re: Why is Large Growth disliked?

Post by acegolfer »

nisiprius wrote: Sat Oct 10, 2020 6:31 am I don't really understand them, but the parts that I do understand I find unsatisfying. For example, the factors are defined a priori, apparently based on intuition and common knowledge, and there are some weird things like dividing the universe of stocks into two size categories (large-cap and small-cap), but three value categories:
We also break NYSE, Amex, and NASDAQ stocks into three book-to- market equity groups based on the breakpoints for the bottom 30% (Low), middle 40% (Mediurn).and top 30% (High) of the ranked values of BE/ME for NYSE stocks.
It's not too clear why they do this--neither why they have two categories for size and three for book-to-market nor why they choose to split book-to-market 30/40/30 rather than in some other way.

There's no theoretical reason behind "2" by "3" portfolios. FF had the Merton's fishing license to find the best factors that explain the cross section returns. Once they found portfolios with explanatory power, FF could claim they are factors. (Nevertheless, FF were careful in not claiming them "risk factors" because they didn't provide the definitions for "risk". FF called them "factor mimicking portfolios.")
User avatar
Taylor Larimore
Advisory Board
Posts: 29946
Joined: Tue Feb 27, 2007 8:09 pm
Location: Miami FL

Re: Why is Large Growth disliked?

Post by Taylor Larimore »

Day9:

It is a mistake to invest in any one part of the market. The Callan Table of Periodic Returns shows this clearly:

https://www.callan.com/wp-content/uploa ... -Table.pdf

Lesson learned: Invest in a simple, tax-efficient, very diversified, Total Market Index Fund.

Best wishes.
Taylor
Jack Bogle's Words of Wisdom: "Never think you know more than the market. Nobody does."
"Simplicity is the master key to financial success." -- Jack Bogle
User avatar
Uncorrelated
Posts: 1039
Joined: Sun Oct 13, 2019 3:16 pm

Re: Why is Large Growth disliked?

Post by Uncorrelated »

acegolfer wrote: Sun Oct 18, 2020 11:52 am
nisiprius wrote: Sat Oct 10, 2020 6:31 am I don't really understand them, but the parts that I do understand I find unsatisfying. For example, the factors are defined a priori, apparently based on intuition and common knowledge, and there are some weird things like dividing the universe of stocks into two size categories (large-cap and small-cap), but three value categories:
We also break NYSE, Amex, and NASDAQ stocks into three book-to- market equity groups based on the breakpoints for the bottom 30% (Low), middle 40% (Mediurn).and top 30% (High) of the ranked values of BE/ME for NYSE stocks.
It's not too clear why they do this--neither why they have two categories for size and three for book-to-market nor why they choose to split book-to-market 30/40/30 rather than in some other way.

There's no theoretical reason behind "2" by "3" portfolios. FF had the Merton's fishing license to find the best factors that explain the cross section returns. Once they found portfolios with explanatory power, FF could claim they are factors. (Nevertheless, FF were careful in not claiming them "risk factors" because they didn't provide the definitions for "risk". FF called them "factor mimicking portfolios.")
Indeed. In A Five-Factor Asset Pricing Model (2015), they test both 2x2 and 2x3 sorts (and 2x2x2x2 sorts...) and find that the confidence intervals are basically identical.

Image

It appears that the exact sorting is unimportant, although I'm sure they choose 2x3 instead of 2x2 because the absolute numbers look at bit more impressive.

The sorts are defined as (table 3):
Image
000
Posts: 2716
Joined: Thu Jul 23, 2020 12:04 am

Re: Why is Large Growth disliked?

Post by 000 »

Uncorrelated wrote: Sun Oct 18, 2020 7:17 am Academics don't backtest, they create models and test them for validity. A backtest isn't the only way to evaluate past data.
lol

A model based on past data is a backtest. It may be more sophisticated than a simple backtest with two endpoints, but it too is dependent on the future looking like the past. Almost any data from before the widespread use of the internet (c. 1995) seems basically pointless to me. The volatility-adjusted underperformance of SCV to TSM includes the bursting of the dotcom bubble.
Uncorrelated wrote: Sun Oct 18, 2020 7:17 am If you don't believe in backtesting, why do you time the market based on recent performance? Can you show me (with statistical tests) that "emerging markets is uncompensated risk" is stronger than "hml is uncompensated risk"?
How is that relevant to this discussion? Moreover, how do you know my decision was based on market timing? But... if you desire knowledge, I already posted an answer to that thread.
Uncorrelated wrote: Sun Oct 18, 2020 7:17 am My personal opinion is that HmL is partially a risk story, and partially a behavioral story. The statistical evidence for HmL is stronger than that for any other active bet (except for the equity risk premium itself and BAB, which are comparable). Take note that if HmL is 100% behavioral, it's a free lunch. If HmL is 100% a risk story, the risk is compensated. This is in great contrast to other active bets such as market timing and country betting, which (apart from being statistically non-existent) very likely represent uncompensated risks.

Fama says determining whether HmL is a risk or a behavioral story is unresolvable. I like to discuss about things that are solvable, testable hypothesis such as "there was no value premium in time period X". Many people have claimed that there was no premium post-publication, the statistical tests do not indicate that.
You can't solve or test for the future. I care about future investment performance. I am not an investing historian.
Adfmacro
Posts: 4
Joined: Fri May 17, 2019 8:12 pm

Re: Why is Large Growth disliked?

Post by Adfmacro »

The short answer is VTXAS is a cap weighted index fund so currently very much tilted to large cap growth. Per Portfolio Visualizer large cap growth since 2017 is over 52% of the fund and large cap overall is 76%. Between 2014 and 2017 the fund was about 45% Large cap growth. By the nature of cap weight indexing. It is automatically riding the large cap growth wave already.
Post Reply