SCV tilt research data

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Topic Author
Rocinante Rider
Posts: 63
Joined: Fri Aug 19, 2022 12:52 pm

SCV tilt research data

Post by Rocinante Rider »

Is anyone aware of credible critiques of the small-cap value analysis offered by Paul Merriman and Richard Buck (e.g., in their publication "We're Talking Millions")? They contend that despite volatility and potential underperformance over shorter time periods, a significant and sustained tilt toward SCV has resulted in superior returns over any rolling 40-year period between 1970 and at least 2019. Extending even further back, they state, "Over the past 92 years, small-company value has been a great addition to any equity portfolio." Does anyone have references to analyses that dispute their data or contentions?
User avatar
Nathan Drake
Posts: 6107
Joined: Mon Apr 11, 2011 12:28 am

Re: SCV tilt research data

Post by Nathan Drake »

The data is what it is, and backed by academic finance

The refutation - you can't stay the course or believe the future won't be like the past
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
omedus82
Posts: 26
Joined: Mon Feb 13, 2012 6:29 pm

Re: SCV tilt research data

Post by omedus82 »

At the 2023 Bogleheads conference, Paul Merriman, who you referred to, gave a presentation titled "The Case for Factor Investing" and Rick Ferri gave a presentation titled "The Case Against Factor Investing." You can easily find these videos on the Bogleheads Youtube channel. They will probably help you.

Some refutations I can think of include:

1) SCV funds have higher cost
2) Tracking error vs broad index funds
3) Value is defined differently by different SCV fund providers resulting in sometimes vastly different performance results. You will see this at the end of Mr. Merriman's presentation.
4) You have to come up with a percentage to tilt to a SCV fund and stick with it for ideally the rest of your life.
5) SCV has provided higher returns but is not expected to provider high risk-adjusted returns. There are many roads to Dublin, meaning there are many different portfolios that will allow you to reach your financial goals, so sticking with a more simple portfolio makes more sense.
livesoft
Posts: 85856
Joined: Thu Mar 01, 2007 7:00 pm

Re: SCV tilt research data

Post by livesoft »

I'm more interested in what happens in the next 40 years.
Wiki This signature message sponsored by sscritic: Learn to fish.
User avatar
nisiprius
Advisory Board
Posts: 51940
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: SCV tilt research data

Post by nisiprius »

1) This gets discussed in this forum a lot.

2) Paul Merriman is clearly an enthusiast and is grinding an axe. A sore point with me is that he never discusses risk. Small-cap value is considerably riskier than the whole market. It's not surprising that there is a risk premium. The question is whether the premium more than compensates for the risk.

To put it another way, it is easy to point to higher return, much harder to show that small-cap value is different from and much better than simply increasing stock allocation, or leveraging stocks.

If you compare portfolios with equal risks, the backtested benefits of a small-cap value tilt may not go away completely, but they certainly become less robust and less convincing than comparisons in which the tilt is allowed to add to the risk.

3) For a critique, and a harsh one:Akey, Pat and Robertson, Adriana and Simutin, Mikhail(2023), Noisy Factors. Credible? Well, there's a seven-page discussion of that paper in this forum here. They are legitimate academics but the paper hasn't appeared in a peer-reviewed journal yet.

Going beyond what Akey & all actually said, the gist of the paper as I understand it is not that the factor work is wrong--but that they suspect Ken French, who seems to be the source of all factor definitions and data used by researchers, of tweaking and buffing up the methodology, not disclosing the details fully, and adjusting it every few years--in ways that have tended to make the value factor look better than it actually is.
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.
Topic Author
Rocinante Rider
Posts: 63
Joined: Fri Aug 19, 2022 12:52 pm

Re: SCV tilt research data

Post by Rocinante Rider »

Thankyou all, esp. Omedus82 and nisprius!

I'll check out the links and watch the mentioned videos. I understand well Omedus82's points 1,2, and 4 (and of course Livesoft's implication that the next 40 years might not be c/w the past). Point 5 I assume essentially means that the higher returns may not adequately compensate for the higher risk, as nisiprius elaborates upon. Point 3 definitely gets at my original question about potential challenges to the data and conclusions, including how one defines terms such as "small cap" or "value." Despite its name, VBR, for example, is arguably skewed more toward mid-cap and blend than many other SCV funds.

I do appreciate everyone's time in responding. Bogleheads rocks!
Logan Roy
Posts: 1703
Joined: Sun May 29, 2022 10:15 am

Re: SCV tilt research data

Post by Logan Roy »

I think there have been a lot of interesting papers that should at least teach caution on taking these kind of researching findings at face value:
There is no Size Effect
https://www.aqr.com/Insights/Perspectiv ... ly-Edition

I came from active trading and stock picking and algorithmic trading and What Works on Wall Street, etc. so this whole thing with 'factors' is old hat (they're rebranded stock screens) .. I was won over by it 10 years ago, because it seemed like credible academics had separated the wheat from the chaff – but that's not held up (imo).

What really killed factors for me was realising how easy it is to invent them, and how you draw a line between refining and back-fitting? I can generate far superior backtests using sectors (rather than factors), but I still don't go all-in on those, because you're only really charting market history. It's very different from identifying a clear, persistent principle or anomaly. And with 1% of stocks responsible for half the market return, any arbitrary grouping is at risk of just slightly overweighting some of those (inverted strategies, monkey portfolios, etc.).
valueinvestor
Posts: 76
Joined: Mon Jun 11, 2007 7:34 am

Re: SCV tilt research data

Post by valueinvestor »

There are practitioners like Ben Graham, Buffett, Bogle and there are theorists (who should more accurately described as equation solvers). Theorists say things like "equity premium puzzle" which is not a puzzle for just about any normal person not living in ivory towers, "rebalancing bonus" - a tooth fairy for adults and to the point of this thread "small cap value premium" which can more accurately described as “Many shall be restored that now are fallen, and many shall fall that are now in honour.”.

It is not that SCV or value did not do well. Internet disrupted many business models which happen to be in value space (like retailers) or first moved into value (like newspapers from growth space) and got decimated. The disrupters tended to be in growth space. So compared to value, growth did really well over the last several years as business models got disrupted. So the mean reversion that used to previously occur got sidetracked.

Going forward, how this plays out is anyone's guess.
User avatar
Nathan Drake
Posts: 6107
Joined: Mon Apr 11, 2011 12:28 am

Re: SCV tilt research data

Post by Nathan Drake »

nisiprius wrote: Sun Feb 11, 2024 1:39 pm
To put it another way, it is easy to point to higher return, much harder to show that small-cap value is different from and much better than simply increasing stock allocation, or leveraging stocks.
It's actually quite easy - the premium adds diversification because it's not as strongly correlated to the market.

Why would I want to take on additional cost with leverage by doubling down on the same source of risk, rather than diversifying sources of risk?
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
jbowman
Posts: 195
Joined: Sat Jan 09, 2021 4:34 pm

Re: SCV tilt research data

Post by jbowman »

Nathan Drake wrote: Sun Feb 11, 2024 3:06 pm
nisiprius wrote: Sun Feb 11, 2024 1:39 pm
To put it another way, it is easy to point to higher return, much harder to show that small-cap value is different from and much better than simply increasing stock allocation, or leveraging stocks.
It's actually quite easy - the premium adds diversification because it's not as strongly correlated to the market.

Why would I want to take on additional cost with leverage by doubling down on the same source of risk, rather than diversifying sources of risk?
I've seen you post this argument a fair number of times now. I'd like your insight on the following question if I may...

How is this not a form of special pleading? Can I not construct any number of subsets of "the market" with varying degrees of correlation to "the market"? Why not utilities, or finance stocks, or companies starting with the letter Q if the backtesting happened to show they're less correlated to VT than SCV is? In other words, what makes this subset so special?

edit: the reason for the question is that it seems to me I'd want to diversify into as many of these sub categories as possible for maximum correlation diversification... which seems like it'd come full circle and end up approximating the full market again.
adave
Posts: 539
Joined: Thu Sep 13, 2007 7:17 pm
Location: Houston

Re: SCV tilt research data

Post by adave »

Is factor investing really worth the squeeze? It could be decades before any expected return premium shows up if at all. Total market approach just seems more sensible IMO.
User avatar
Nathan Drake
Posts: 6107
Joined: Mon Apr 11, 2011 12:28 am

Re: SCV tilt research data

Post by Nathan Drake »

jbowman wrote: Sun Feb 11, 2024 3:39 pm
Nathan Drake wrote: Sun Feb 11, 2024 3:06 pm
nisiprius wrote: Sun Feb 11, 2024 1:39 pm
To put it another way, it is easy to point to higher return, much harder to show that small-cap value is different from and much better than simply increasing stock allocation, or leveraging stocks.
It's actually quite easy - the premium adds diversification because it's not as strongly correlated to the market.

Why would I want to take on additional cost with leverage by doubling down on the same source of risk, rather than diversifying sources of risk?
I've seen you post this argument a fair number of times now. I'd like your insight on the following question if I may...

How is this not a form of special pleading? Can I not construct any number of subsets of "the market" with varying degrees of correlation to "the market"? Why not utilities, or finance stocks, or companies starting with the letter Q if the backtesting happened to show they're less correlated to VT than SCV is? In other words, what makes this subset so special?

edit: the reason for the question is that it seems to me I'd want to diversify into as many of these sub categories as possible for maximum correlation diversification... which seems like it'd come full circle and end up approximating the full market again.
Because Value funds are not concentrated sector bets. They are diversified across the entire market and seeking out higher discount rates
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
User avatar
Nathan Drake
Posts: 6107
Joined: Mon Apr 11, 2011 12:28 am

Re: SCV tilt research data

Post by Nathan Drake »

adave wrote: Sun Feb 11, 2024 3:51 pm Is factor investing really worth the squeeze? It could be decades before any expected return premium shows up if at all. Total market approach just seems more sensible IMO.
It hasn’t been decades for a value investor the last 4 years

Could be decades for the market to return better than bonds as well..
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
jbowman
Posts: 195
Joined: Sat Jan 09, 2021 4:34 pm

Re: SCV tilt research data

Post by jbowman »

Nathan Drake wrote: Sun Feb 11, 2024 3:53 pm
jbowman wrote: Sun Feb 11, 2024 3:39 pm
Nathan Drake wrote: Sun Feb 11, 2024 3:06 pm
nisiprius wrote: Sun Feb 11, 2024 1:39 pm
To put it another way, it is easy to point to higher return, much harder to show that small-cap value is different from and much better than simply increasing stock allocation, or leveraging stocks.
It's actually quite easy - the premium adds diversification because it's not as strongly correlated to the market.

Why would I want to take on additional cost with leverage by doubling down on the same source of risk, rather than diversifying sources of risk?
I've seen you post this argument a fair number of times now. I'd like your insight on the following question if I may...

How is this not a form of special pleading? Can I not construct any number of subsets of "the market" with varying degrees of correlation to "the market"? Why not utilities, or finance stocks, or companies starting with the letter Q if the backtesting happened to show they're less correlated to VT than SCV is? In other words, what makes this subset so special?

edit: the reason for the question is that it seems to me I'd want to diversify into as many of these sub categories as possible for maximum correlation diversification... which seems like it'd come full circle and end up approximating the full market again.
Because Value funds are not concentrated sector bets. They are diversified across the entire market and seeking out higher discount rates
Ok, but you can still construct any number of subsets that aren't concentrated in a single sector. Why not any of those subsets if they offer meaningful diversification of risk via lower correlation with the total market? I still don't understand what sets SCV apart in this context.
User avatar
Nathan Drake
Posts: 6107
Joined: Mon Apr 11, 2011 12:28 am

Re: SCV tilt research data

Post by Nathan Drake »

jbowman wrote: Sun Feb 11, 2024 4:08 pm
Nathan Drake wrote: Sun Feb 11, 2024 3:53 pm
jbowman wrote: Sun Feb 11, 2024 3:39 pm
Nathan Drake wrote: Sun Feb 11, 2024 3:06 pm
nisiprius wrote: Sun Feb 11, 2024 1:39 pm
To put it another way, it is easy to point to higher return, much harder to show that small-cap value is different from and much better than simply increasing stock allocation, or leveraging stocks.
It's actually quite easy - the premium adds diversification because it's not as strongly correlated to the market.

Why would I want to take on additional cost with leverage by doubling down on the same source of risk, rather than diversifying sources of risk?
I've seen you post this argument a fair number of times now. I'd like your insight on the following question if I may...

How is this not a form of special pleading? Can I not construct any number of subsets of "the market" with varying degrees of correlation to "the market"? Why not utilities, or finance stocks, or companies starting with the letter Q if the backtesting happened to show they're less correlated to VT than SCV is? In other words, what makes this subset so special?

edit: the reason for the question is that it seems to me I'd want to diversify into as many of these sub categories as possible for maximum correlation diversification... which seems like it'd come full circle and end up approximating the full market again.
Because Value funds are not concentrated sector bets. They are diversified across the entire market and seeking out higher discount rates
Ok, but you can still construct any number of subsets that aren't concentrated in a single sector. Why not any of those subsets if they offer meaningful diversification of risk via lower correlation with the total market? I still don't understand what sets SCV apart in this context.
The “subsets” you discuss are basically factors, but only a few of them are completely independent and stand up to academic rigor, such as Value
Last edited by Nathan Drake on Sun Feb 11, 2024 4:35 pm, edited 1 time in total.
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
User avatar
TimeIsYourFriend
Posts: 393
Joined: Sun Dec 17, 2023 11:19 am

Re: SCV tilt research data

Post by TimeIsYourFriend »

Actual funds to invest in are key here. The market has about 20 years more data in an actual investible S&P 500 fund than small value and the index itself goes far back in various forms. No one was investing in hundreds of small value stocks diversified with industry caps in the 70’s and it was not common knowledge among investors that they outperformed the market as a whole. There wasn’t any small value index either. These stocks were expensive to trade and not remotely as liquid like today. You have to take comparisons with the past with a grain of salt including the S&P 500 and small value which is even much less certain.
"Time is your friend; impulse is your enemy." - John C. Bogle
Gaston
Posts: 1177
Joined: Wed Aug 21, 2013 7:12 pm

Re: SCV tilt research data

Post by Gaston »

TimeIsYourFriend wrote: Sun Feb 11, 2024 4:20 pm Actual funds to invest in are key here. The market has about 20 years more data in an actual investible S&P 500 fund than small value and the index itself goes far back in various forms. No one was investing in hundreds of small value stocks diversified with industry caps in the 70’s and it was not common knowledge among investors that they outperformed the market as a whole. There wasn’t any small value index either. These stocks were expensive to trade and not remotely as liquid like today. You have to take comparisons with the past with a grain of salt including the S&P 500 and small value which is even much less certain.
I think you touch on a key point. All the back-tested data in the world might help you form an investing hypothesis, but the true test begins once the hypothesis is established: that of examining forward-looking data in actual, investable products. In the case of low-cost, broad-based index funds, Ben Graham, Paul Samuelson, Edward Renshaw and Mac McQuown (among others) established the hypothesis back in the 1950s and 1960s. John Bogle then put the hypothesis to the test starting in 1976 with the launch of the first retail index fund. Many others followed. So we Boglehead's have nearly 50 years of post-hypothesis data upon which to base our actions.
“My opinions are just that - opinions.”
User avatar
Nathan Drake
Posts: 6107
Joined: Mon Apr 11, 2011 12:28 am

Re: SCV tilt research data

Post by Nathan Drake »

Gaston wrote: Sun Feb 11, 2024 4:33 pm
TimeIsYourFriend wrote: Sun Feb 11, 2024 4:20 pm Actual funds to invest in are key here. The market has about 20 years more data in an actual investible S&P 500 fund than small value and the index itself goes far back in various forms. No one was investing in hundreds of small value stocks diversified with industry caps in the 70’s and it was not common knowledge among investors that they outperformed the market as a whole. There wasn’t any small value index either. These stocks were expensive to trade and not remotely as liquid like today. You have to take comparisons with the past with a grain of salt including the S&P 500 and small value which is even much less certain.
I think you touch on a key point. All the back-tested data in the world might help you form an investing hypothesis, but the true test begins once the hypothesis is established: that of examining forward-looking data in actual, investable products. In the case of low-cost, broad-based index funds, Ben Graham, Paul Samuelson, Edward Renshaw and Mac McQuown (among others) established the hypothesis back in the 1950s and 1960s. John Bogle then put the hypothesis to the test starting in 1976 with the launch of the first retail index fund. Many others followed. So we Boglehead's have nearly 50 years of post-hypothesis data upon which to base our actions.
And 50 years isn’t very long for a single market

It’s why we also can’t only look at live funds to drive AA decisions
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
User avatar
familythriftmd
Posts: 743
Joined: Fri Sep 18, 2020 10:15 am
Location: Wisconsin

Re: SCV tilt research data

Post by familythriftmd »

nisiprius wrote: Sun Feb 11, 2024 1:39 pm
... Paul Merriman is clearly an enthusiast and is grinding an axe.

...

It's not surprising that there is a risk premium. The question is whether the premium more than compensates for the risk.
I really like how you explain it there with respect to value of risk premium over projected or actual risk! And definitely I agree that he doesn't seem to talk about a whole lot more (or at least it isn't easy to remember).
User avatar
nisiprius
Advisory Board
Posts: 51940
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: SCV tilt research data

Post by nisiprius »

Actually, the biggest puzzle for me about factor research is where the factors came from in the first place. Coming to investing theory from a background of life sciences, I was familiar with principal components analysis--having written computer programs to do it--and vaguely familiar with factor analysis (in the sense that I'm not quite sure what the difference between it and principal components analysis is).

Wikipedia describes factor analysis:
Factor analysis is a statistical method used to describe variability among observed, correlated variables in terms of a potentially lower number of unobserved variables called factors. For example, it is possible that variations in six observed variables mainly reflect the variations in two unobserved (underlying) variables. Factor analysis searches for such joint variations in response to unobserved latent variables. The observed variables are modelled as linear combinations of the potential factors plus "error" terms, hence factor analysis can be thought of as a special case of errors-in-variables models.

Simply put, the factor loading of a variable quantifies the extent to which the variable is related to a given factor.
Naturally, I just assumed that investing "factors" mean "the result of performing factor analysis on financial variables."

But in factor analysis, you don't start with the factors, you start with the variables. The factors emerge from the data. And by definition and construction, they are orthogonal, i.e. have zero correlation with each other. Having discovered an X factor which has the largest explanatory power, you don't want your other factors to contain any mixture of X in them. You construct the next factor, Y, from the set of possible factors that do not contain any X--are orthogonal to X--have zero correlation with X.

I was therefore surprised to discover that the Fama-French and other factors do not have zero correlation. Not high correlation, but not zero. Value and momentum even have negative correlation, meaning that although these are both supposed to be "good," you can't have a high loading on both of them at the same time--they are partial opposites of each other.

It seems as if the "factors" were pulled out of the air, a priori, based on intuition. It seems as if book-to-market ought to be relevant, so they just defined that as a factor and tested how well it fit the data... instead of starting with the data and finding the factor that had the best fit.

I don't understand why they did it that way. It seems as if they had a strong bias for wanting the factors to be simple accounting ratios.

It seems as if it would be much more scientific to do an actual factor analysis, and the results would quite possibly be explaining more of stock behavior using a small number of factors. Of course it would be harder to associate these factors with intuitive names.

Regardless of why Fama and French did what they did, has anyone else tried applying actual factor analysis to stock data, to extract the natural factors from within the data itself?
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.
jbowman
Posts: 195
Joined: Sat Jan 09, 2021 4:34 pm

Re: SCV tilt research data

Post by jbowman »

Nathan Drake wrote: Sun Feb 11, 2024 4:10 pm
jbowman wrote: Sun Feb 11, 2024 4:08 pm
Nathan Drake wrote: Sun Feb 11, 2024 3:53 pm
jbowman wrote: Sun Feb 11, 2024 3:39 pm
Nathan Drake wrote: Sun Feb 11, 2024 3:06 pm

It's actually quite easy - the premium adds diversification because it's not as strongly correlated to the market.

Why would I want to take on additional cost with leverage by doubling down on the same source of risk, rather than diversifying sources of risk?
I've seen you post this argument a fair number of times now. I'd like your insight on the following question if I may...

How is this not a form of special pleading? Can I not construct any number of subsets of "the market" with varying degrees of correlation to "the market"? Why not utilities, or finance stocks, or companies starting with the letter Q if the backtesting happened to show they're less correlated to VT than SCV is? In other words, what makes this subset so special?

edit: the reason for the question is that it seems to me I'd want to diversify into as many of these sub categories as possible for maximum correlation diversification... which seems like it'd come full circle and end up approximating the full market again.
Because Value funds are not concentrated sector bets. They are diversified across the entire market and seeking out higher discount rates
Ok, but you can still construct any number of subsets that aren't concentrated in a single sector. Why not any of those subsets if they offer meaningful diversification of risk via lower correlation with the total market? I still don't understand what sets SCV apart in this context.
The “subsets” you discuss are basically factors, but only a few of them are completely independent and stand up to academic rigor, such as Value
And SCV is the only one that satisfies all criteria?

I'm not trying to be argumentative here. I'm trying to figure out how ones arrived at this specific factor tilt out of the universe of all possible tilts. Based on the criteria (academic rigor, independent, exhibits some deviation from the overall market that will allow us to diversify) then only one of the following can be true. a) It's the only one that satisfies, b) it's not the only one but it's "special" somehow or c) it's not the only one and we attempt to tilt towards all possibilities that satisfy the criteria and I have a hard time seeing how that ends up anywhere besides back at some approximation of the total market.
folkher0
Posts: 850
Joined: Fri Dec 09, 2016 1:48 pm

Re: SCV tilt research data

Post by folkher0 »

nisiprius wrote: Sun Feb 11, 2024 4:48 pm Coming to investing theory from a background of life sciences, I was familiar with principal components analysis--having written computer programs to do it--and vaguely familiar with factor analysis (in the sense that I'm not quite sure what the difference between it and principal components analysis is).
Like you I have a background in biological sciences. Whenever I read the finance literature I am a bit surprised by how casual and un-rigorous much of the modeling is. Conclusions reached by financial academics seem to be weakly supported in general compared to the standards I’m used to.

I’m sure a physicist or engineer would say the same about my papers. But part of my skepticism of factor research comes from the knowledge that the methods used would not produce statistically significant results in my own field.

That’s my bias, of course. But it’s a hang up I have that I just can’t get past.
Northern Flicker
Posts: 15159
Joined: Fri Apr 10, 2015 12:29 am

Re: SCV tilt research data

Post by Northern Flicker »

Rocinante Rider wrote: Sun Feb 11, 2024 12:33 pm Is anyone aware of credible critiques of the small-cap value analysis offered by Paul Merriman and Richard Buck (e.g., in their publication "We're Talking Millions")? They contend that despite volatility and potential underperformance over shorter time periods, a significant and sustained tilt toward SCV has resulted in superior returns over any rolling 40-year period between 1970 and at least 2019. Extending even further back, they state, "Over the past 92 years, small-company value has been a great addition to any equity portfolio." Does anyone have references to analyses that dispute their data or contentions?
How many distinct 40-year periods are there between 1970 and 2019? Even up to the present day, there are only 14 of them, and they are not independent-- many include the period 1977-1983 when SCV exploded.

Why did they pick 1970? Could it be that going back 92 years instead there are 40-year periods in which SCV did not overperform?

There is historical evidence for overperformance of SCV. I've not read the article, but if your summary is representative of it, I'm not inclined to believe that their historical account is being presented in a scientific or generalizable manner.
Northern Flicker
Posts: 15159
Joined: Fri Apr 10, 2015 12:29 am

Re: SCV tilt research data

Post by Northern Flicker »

nisiprius wrote: Actually, the biggest puzzle for me about factor research is where the factors came from in the first place.
Value, size, momentum, dividend yield, and quality were all techniques used less formally by active managers before factor theory was formalized.
BitTooAggressive
Posts: 1048
Joined: Tue Jul 13, 2021 3:15 pm

Re: SCV tilt research data

Post by BitTooAggressive »

omedus82 wrote: Sun Feb 11, 2024 1:27 pm At the 2023 Bogleheads conference, Paul Merriman, who you referred to, gave a presentation titled "The Case for Factor Investing" and Rick Ferri gave a presentation titled "The Case Against Factor Investing." You can easily find these videos on the Bogleheads Youtube channel. They will probably help you.

Some refutations I can think of include:

1) SCV funds have higher cost
2) Tracking error vs broad index funds
3) Value is defined differently by different SCV fund providers resulting in sometimes vastly different performance results. You will see this at the end of Mr. Merriman's presentation.
4) You have to come up with a percentage to tilt to a SCV fund and stick with it for ideally the rest of your life.
5) SCV has provided higher returns but is not expected to provider high risk-adjusted returns. There are many roads to Dublin, meaning there are many different portfolios that will allow you to reach your financial goals, so sticking with a more simple portfolio makes more sense.
All risk adjusted returns should be the same not unique to small cap value.
User avatar
nisiprius
Advisory Board
Posts: 51940
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: SCV tilt research data

Post by nisiprius »

Northern Flicker wrote: Sun Feb 11, 2024 6:57 pm
nisiprius wrote: Actually, the biggest puzzle for me about factor research is where the factors came from in the first place.
Value, size, momentum, dividend yield, and quality were all techniques used less formally by active managers before factor theory was formalized.
Understood.

But it seems to me that if you went to the data and let the data tell you what the factors are, rather than starting with the assumption that some a priori vector is a factor, you might get bigger factors and they would be guaranteed to be uncorrelated. If it turned out that (say) the biggest factor actually had a high correlation with book-to-market, that would be a strong indication of the "reality" of book-to-market as a primary factor.
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.
User avatar
arcticpineapplecorp.
Posts: 14916
Joined: Tue Mar 06, 2012 8:22 pm

Re: SCV tilt research data

Post by arcticpineapplecorp. »

nisiprius wrote: Sun Feb 11, 2024 1:39 pm 1) This gets discussed in this forum a lot.

2) Paul Merriman is clearly an enthusiast and is grinding an axe. A sore point with me is that he never discusses risk. Small-cap value is considerably riskier than the whole market. It's not surprising that there is a risk premium. The question is whether the premium more than compensates for the risk.

To put it another way, it is easy to point to higher return, much harder to show that small-cap value is different from and much better than simply increasing stock allocation, or leveraging stocks.

If you compare portfolios with equal risks, the backtested benefits of a small-cap value tilt may not go away completely, but they certainly become less robust and less convincing than comparisons in which the tilt is allowed to add to the risk.
I thought Merriman has shown that while the risk of small cap value is greater, a tilt to small cap value has had higher returns with about the same amount of risk? That seems to be Paul's shtick, sort of arguing for a free lunch (at least in the past)? I don't tilt myself, but isn't this similar to Markowitz's work showing you can reduce volatility by holding assets that on their own may have high volatility but you decrease the volatility if the assets are non-correlated? I know this didn't work in 2008.

I think for the OP one question is that even though back testing has shown improved returns, who actually got those returns in real life (if there weren't small cap value index funds to invest in until much later after the research came out)? And we know that many things that showed improved returns in backtests, fail to live up once everybody knows about it. Not saying that's the case with small cap value, but it could be. I know Swedroe says the factor is robust and persistent. So who knows?
Last edited by arcticpineapplecorp. on Sun Feb 11, 2024 7:48 pm, edited 1 time in total.
It's hard to accept the truth when the lies were exactly what you wanted to hear. Investing is simple, but not easy. Buy, hold & rebalance low cost index funds & manage taxable events. Asking Portfolio Questions | Wiki
Gaston
Posts: 1177
Joined: Wed Aug 21, 2013 7:12 pm

Re: SCV tilt research data

Post by Gaston »

Rocinante Rider wrote: Sun Feb 11, 2024 12:33 pm Is anyone aware of credible critiques of the small-cap value analysis offered by Paul Merriman and Richard Buck (e.g., in their publication "We're Talking Millions")? They contend that despite volatility and potential underperformance over shorter time periods, a significant and sustained tilt toward SCV has resulted in superior returns over any rolling 40-year period between 1970 and at least 2019. Extending even further back, they state, "Over the past 92 years, small-company value has been a great addition to any equity portfolio." Does anyone have references to analyses that dispute their data or contentions?
You asked about creditable critiques. I think the guys who run the Rational Reminder podcast are pretty good at presenting both sides of the debate on factor investing. Among their guests who challenge factor investing, see episode 220 (Jonathan Berk and Jules van Binsbergen) and episode 151 (Brad Cornell).
“My opinions are just that - opinions.”
Northern Flicker
Posts: 15159
Joined: Fri Apr 10, 2015 12:29 am

Re: SCV tilt research data

Post by Northern Flicker »

nisiprius wrote: Sun Feb 11, 2024 7:36 pm
Northern Flicker wrote: Sun Feb 11, 2024 6:57 pm
nisiprius wrote: Actually, the biggest puzzle for me about factor research is where the factors came from in the first place.
Value, size, momentum, dividend yield, and quality were all techniques used less formally by active managers before factor theory was formalized.
Understood.

But it seems to me that if you went to the data and let the data tell you what the factors are, rather than starting with the assumption that some a priori vector is a factor, you might get bigger factors and they would be guaranteed to be uncorrelated. If it turned out that (say) the biggest factor actually had a high correlation with book-to-market, that would be a strong indication of the "reality" of book-to-market as a primary factor.
You can do that, but then a different data set must be used to test hypotheses about which principal components may be associated with higher expected return.
User avatar
Nathan Drake
Posts: 6107
Joined: Mon Apr 11, 2011 12:28 am

Re: SCV tilt research data

Post by Nathan Drake »

jbowman wrote: Sun Feb 11, 2024 6:40 pm
Nathan Drake wrote: Sun Feb 11, 2024 4:10 pm
jbowman wrote: Sun Feb 11, 2024 4:08 pm
Nathan Drake wrote: Sun Feb 11, 2024 3:53 pm
jbowman wrote: Sun Feb 11, 2024 3:39 pm

I've seen you post this argument a fair number of times now. I'd like your insight on the following question if I may...

How is this not a form of special pleading? Can I not construct any number of subsets of "the market" with varying degrees of correlation to "the market"? Why not utilities, or finance stocks, or companies starting with the letter Q if the backtesting happened to show they're less correlated to VT than SCV is? In other words, what makes this subset so special?

edit: the reason for the question is that it seems to me I'd want to diversify into as many of these sub categories as possible for maximum correlation diversification... which seems like it'd come full circle and end up approximating the full market again.
Because Value funds are not concentrated sector bets. They are diversified across the entire market and seeking out higher discount rates
Ok, but you can still construct any number of subsets that aren't concentrated in a single sector. Why not any of those subsets if they offer meaningful diversification of risk via lower correlation with the total market? I still don't understand what sets SCV apart in this context.
The “subsets” you discuss are basically factors, but only a few of them are completely independent and stand up to academic rigor, such as Value
And SCV is the only one that satisfies all criteria?

I'm not trying to be argumentative here. I'm trying to figure out how ones arrived at this specific factor tilt out of the universe of all possible tilts. Based on the criteria (academic rigor, independent, exhibits some deviation from the overall market that will allow us to diversify) then only one of the following can be true. a) It's the only one that satisfies, b) it's not the only one but it's "special" somehow or c) it's not the only one and we attempt to tilt towards all possibilities that satisfy the criteria and I have a hard time seeing how that ends up anywhere besides back at some approximation of the total market.
SCV is simply the way to extract the highest loadings on Value. And Value can have multiple accounting measures and all work (P/B, P/Cash Flows, EV/EBITDA, etc).

There are other "factors" that do have significance in the data, but none as strong as Value and with a theoretical framework.
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
User avatar
nisiprius
Advisory Board
Posts: 51940
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: SCV tilt research data

Post by nisiprius »

Nathan Drake wrote: Mon Feb 12, 2024 1:02 am ...There are other "factors" that do have significance in the data, but none as strong as Value and with a theoretical framework...
There is a theoretical framework that says you can calculate how much of the value factor a portfolio has, and that if, going forward, value does better than market, then those portfolios will do better than the market, and that if, going forward, value does worse than the market, those portfolios will do worse than the market. Similarly, if the stock market in general goes up, portfolios with high beta will go up more than the market, and if the stock market goes down, portfolios with high beta will sink lower than the market. And similarly for size. Thus, the combination of beta, size, and value explains more than beta alone. And if the efficient market hypothesis is true, they explain most of what can be explained, the residuum being an unpredictable random walk.

Fine.

It also suggests that alpha for a value portfolio should be measured against a value benchmark rather than a total market benchmark... and so on an so on.

But that does not help predict how value will do going forward.

Is there any theoretical framework that explains why the value factor should have a higher risk-adjusted return than the market?
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.
Northern Flicker
Posts: 15159
Joined: Fri Apr 10, 2015 12:29 am

Re: SCV tilt research data

Post by Northern Flicker »

nisiprius wrote: Is there any theoretical framework that explains why the value factor should have a higher risk-adjusted return than the market?
Should a value-tilted portfolio have a higher risk-adjusted return than the market? An efficient market would arbitrage that away, establishing an expected return commensurate with risk.
Logan Roy
Posts: 1703
Joined: Sun May 29, 2022 10:15 am

Re: SCV tilt research data

Post by Logan Roy »

Nathan Drake wrote: Mon Feb 12, 2024 1:02 am
jbowman wrote: Sun Feb 11, 2024 6:40 pm
Nathan Drake wrote: Sun Feb 11, 2024 4:10 pm
jbowman wrote: Sun Feb 11, 2024 4:08 pm
Nathan Drake wrote: Sun Feb 11, 2024 3:53 pm

Because Value funds are not concentrated sector bets. They are diversified across the entire market and seeking out higher discount rates
Ok, but you can still construct any number of subsets that aren't concentrated in a single sector. Why not any of those subsets if they offer meaningful diversification of risk via lower correlation with the total market? I still don't understand what sets SCV apart in this context.
The “subsets” you discuss are basically factors, but only a few of them are completely independent and stand up to academic rigor, such as Value
And SCV is the only one that satisfies all criteria?

I'm not trying to be argumentative here. I'm trying to figure out how ones arrived at this specific factor tilt out of the universe of all possible tilts. Based on the criteria (academic rigor, independent, exhibits some deviation from the overall market that will allow us to diversify) then only one of the following can be true. a) It's the only one that satisfies, b) it's not the only one but it's "special" somehow or c) it's not the only one and we attempt to tilt towards all possibilities that satisfy the criteria and I have a hard time seeing how that ends up anywhere besides back at some approximation of the total market.
SCV is simply the way to extract the highest loadings on Value. And Value can have multiple accounting measures and all work (P/B, P/Cash Flows, EV/EBITDA, etc).

There are other "factors" that do have significance in the data, but none as strong as Value and with a theoretical framework.
I think it depends what a particular academic wants to call a 'factor'.

Momentum is a more persistent market anomaly. And in the Malkiel's Monkey paper, all sorts of crazy stock screens outperform value – I think the random monkey portfolios come pretty close. Just stocks starting with the letters B-U-F-F-E-T-T works well historically – I think it beats BRK.
User avatar
Nathan Drake
Posts: 6107
Joined: Mon Apr 11, 2011 12:28 am

Re: SCV tilt research data

Post by Nathan Drake »

Logan Roy wrote: Mon Feb 12, 2024 12:49 pm
Nathan Drake wrote: Mon Feb 12, 2024 1:02 am
jbowman wrote: Sun Feb 11, 2024 6:40 pm
Nathan Drake wrote: Sun Feb 11, 2024 4:10 pm
jbowman wrote: Sun Feb 11, 2024 4:08 pm

Ok, but you can still construct any number of subsets that aren't concentrated in a single sector. Why not any of those subsets if they offer meaningful diversification of risk via lower correlation with the total market? I still don't understand what sets SCV apart in this context.
The “subsets” you discuss are basically factors, but only a few of them are completely independent and stand up to academic rigor, such as Value
And SCV is the only one that satisfies all criteria?

I'm not trying to be argumentative here. I'm trying to figure out how ones arrived at this specific factor tilt out of the universe of all possible tilts. Based on the criteria (academic rigor, independent, exhibits some deviation from the overall market that will allow us to diversify) then only one of the following can be true. a) It's the only one that satisfies, b) it's not the only one but it's "special" somehow or c) it's not the only one and we attempt to tilt towards all possibilities that satisfy the criteria and I have a hard time seeing how that ends up anywhere besides back at some approximation of the total market.
SCV is simply the way to extract the highest loadings on Value. And Value can have multiple accounting measures and all work (P/B, P/Cash Flows, EV/EBITDA, etc).

There are other "factors" that do have significance in the data, but none as strong as Value and with a theoretical framework.
I think it depends what a particular academic wants to call a 'factor'.

Momentum is a more persistent market anomaly. And in the Malkiel's Monkey paper, all sorts of crazy stock screens outperform value – I think the random monkey portfolios come pretty close. Just stocks starting with the letters B-U-F-F-E-T-T works well historically – I think it beats BRK.
The 'all sorts of crazy stock screens' likely does not stand up to any sort of robustness tests, otherwise it would have been taken more seriously.

Momentum is shown as a factor, but is prone to crazy turnabouts and faces implementation cost challenges, and does not have a strong theoretical framework. Its inclusion does seem useful as a means of timing individual components in a given fund, which is part of the trading strategies implemented by funds from DFA and Avantis.
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
User avatar
Taylor Larimore
Posts: 32822
Joined: Tue Feb 27, 2007 7:09 pm
Location: Miami FL

Re: SCV tilt research data

Post by Taylor Larimore »

Rocinante Rider wrote: Sun Feb 11, 2024 12:33 pm Is anyone aware of credible critiques of the small-cap value analysis offered by Paul Merriman and Richard Buck (e.g., in their publication "We're Talking Millions")? They contend that despite volatility and potential underperformance over shorter time periods, a significant and sustained tilt toward SCV has resulted in superior returns over any rolling 40-year period between 1970 and at least 2019. Extending even further back, they state, "Over the past 92 years, small-company value has been a great addition to any equity portfolio." Does anyone have references to analyses that dispute their data or contentions?
Mr. Merriman is a very nice man and has been a long-time advocate of small-cap value stocks. Unfortunately, his recommended portfolio had the lowest return of eight professional portfolios:

https://www.marketwatch.com/lazyportfolio%20

The 3-Fund Portfolio had the highest returns.

Best wishes.
Taylor
Jack Bogle's Words of Wisdom: "One of the seemingly indestructible myths of investing is that stocks with small market capitalizations outpace stocks with large market capitalizations over time."
"Simplicity is the master key to financial success." -- Jack Bogle
User avatar
Nathan Drake
Posts: 6107
Joined: Mon Apr 11, 2011 12:28 am

Re: SCV tilt research data

Post by Nathan Drake »

Taylor Larimore wrote: Mon Feb 12, 2024 7:39 pm
Rocinante Rider wrote: Sun Feb 11, 2024 12:33 pm Is anyone aware of credible critiques of the small-cap value analysis offered by Paul Merriman and Richard Buck (e.g., in their publication "We're Talking Millions")? They contend that despite volatility and potential underperformance over shorter time periods, a significant and sustained tilt toward SCV has resulted in superior returns over any rolling 40-year period between 1970 and at least 2019. Extending even further back, they state, "Over the past 92 years, small-company value has been a great addition to any equity portfolio." Does anyone have references to analyses that dispute their data or contentions?
Mr. Merriman is a very nice man and has been a long-time advocate of small-cap value stocks. Unfortunately, his recommended portfolio had the lowest return of eight professional portfolios:

https://www.marketwatch.com/lazyportfolio%20

The 3-Fund Portfolio had the highest returns.

Best wishes.
Taylor
Jack Bogle's Words of Wisdom: "One of the seemingly indestructible myths of investing is that stocks with small market capitalizations outpace stocks with large market capitalizations over time."
Well unfortunately this is only the last 10 years. And we know that the prior 10 years, the 3-fund portfolio had some of the WORST returns.

Paul Merriman has always been one to quickly highlight how important it is to have a long term allocation, and 10 years is nowhere near sufficient to draw any conclusions.
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
Logan Roy
Posts: 1703
Joined: Sun May 29, 2022 10:15 am

Re: SCV tilt research data

Post by Logan Roy »

Nathan Drake wrote: Mon Feb 12, 2024 6:47 pm
Logan Roy wrote: Mon Feb 12, 2024 12:49 pm
Nathan Drake wrote: Mon Feb 12, 2024 1:02 am
jbowman wrote: Sun Feb 11, 2024 6:40 pm
Nathan Drake wrote: Sun Feb 11, 2024 4:10 pm

The “subsets” you discuss are basically factors, but only a few of them are completely independent and stand up to academic rigor, such as Value
And SCV is the only one that satisfies all criteria?

I'm not trying to be argumentative here. I'm trying to figure out how ones arrived at this specific factor tilt out of the universe of all possible tilts. Based on the criteria (academic rigor, independent, exhibits some deviation from the overall market that will allow us to diversify) then only one of the following can be true. a) It's the only one that satisfies, b) it's not the only one but it's "special" somehow or c) it's not the only one and we attempt to tilt towards all possibilities that satisfy the criteria and I have a hard time seeing how that ends up anywhere besides back at some approximation of the total market.
SCV is simply the way to extract the highest loadings on Value. And Value can have multiple accounting measures and all work (P/B, P/Cash Flows, EV/EBITDA, etc).

There are other "factors" that do have significance in the data, but none as strong as Value and with a theoretical framework.
I think it depends what a particular academic wants to call a 'factor'.

Momentum is a more persistent market anomaly. And in the Malkiel's Monkey paper, all sorts of crazy stock screens outperform value – I think the random monkey portfolios come pretty close. Just stocks starting with the letters B-U-F-F-E-T-T works well historically – I think it beats BRK.
The 'all sorts of crazy stock screens' likely does not stand up to any sort of robustness tests, otherwise it would have been taken more seriously.

Momentum is shown as a factor, but is prone to crazy turnabouts and faces implementation cost challenges, and does not have a strong theoretical framework. Its inclusion does seem useful as a means of timing individual components in a given fund, which is part of the trading strategies implemented by funds from DFA and Avantis.
I don't think FF recognise momentum as a factor(?). But I think the theory's far more robust. It's that information travels through markets with latency – so it starts with very high information investors, and eventually filters through to 'mom & pop' retail investors, via the media .. i.e. The institutions start selling when Ted spreads go one way; while the 'mom & pop' investors are selling when the market's already 50% down and CNBC's forecasting the end of stocks.

The problem with value as a 'risk premium' is that there's no example of a market compensating risk with a higher return (after that risk has played out). We can say longer-dated bonds typically have higher yields, and more volatility, but that's a term premium; not necessarily a risk premium (Treasuries are zero risk). And other factors, like Quality, don't fit that framework at all. Others, like Size, don't exist (but should, if the market was paying investors a premium for investing in higher risk assets).
User avatar
Nathan Drake
Posts: 6107
Joined: Mon Apr 11, 2011 12:28 am

Re: SCV tilt research data

Post by Nathan Drake »

Logan Roy wrote: Tue Feb 13, 2024 1:44 pm
Nathan Drake wrote: Mon Feb 12, 2024 6:47 pm
Logan Roy wrote: Mon Feb 12, 2024 12:49 pm
Nathan Drake wrote: Mon Feb 12, 2024 1:02 am
jbowman wrote: Sun Feb 11, 2024 6:40 pm

And SCV is the only one that satisfies all criteria?

I'm not trying to be argumentative here. I'm trying to figure out how ones arrived at this specific factor tilt out of the universe of all possible tilts. Based on the criteria (academic rigor, independent, exhibits some deviation from the overall market that will allow us to diversify) then only one of the following can be true. a) It's the only one that satisfies, b) it's not the only one but it's "special" somehow or c) it's not the only one and we attempt to tilt towards all possibilities that satisfy the criteria and I have a hard time seeing how that ends up anywhere besides back at some approximation of the total market.
SCV is simply the way to extract the highest loadings on Value. And Value can have multiple accounting measures and all work (P/B, P/Cash Flows, EV/EBITDA, etc).

There are other "factors" that do have significance in the data, but none as strong as Value and with a theoretical framework.
I think it depends what a particular academic wants to call a 'factor'.

Momentum is a more persistent market anomaly. And in the Malkiel's Monkey paper, all sorts of crazy stock screens outperform value – I think the random monkey portfolios come pretty close. Just stocks starting with the letters B-U-F-F-E-T-T works well historically – I think it beats BRK.
The 'all sorts of crazy stock screens' likely does not stand up to any sort of robustness tests, otherwise it would have been taken more seriously.

Momentum is shown as a factor, but is prone to crazy turnabouts and faces implementation cost challenges, and does not have a strong theoretical framework. Its inclusion does seem useful as a means of timing individual components in a given fund, which is part of the trading strategies implemented by funds from DFA and Avantis.
I don't think FF recognise momentum as a factor(?). But I think the theory's far more robust. It's that information travels through markets with latency – so it starts with very high information investors, and eventually filters through to 'mom & pop' retail investors, via the media .. i.e. The institutions start selling when Ted spreads go one way; while the 'mom & pop' investors are selling when the market's already 50% down and CNBC's forecasting the end of stocks.

The problem with value as a 'risk premium' is that there's no example of a market compensating risk with a higher return (after that risk has played out). We can say longer-dated bonds typically have higher yields, and more volatility, but that's a term premium; not necessarily a risk premium (Treasuries are zero risk). And other factors, like Quality, don't fit that framework at all. Others, like Size, don't exist (but should, if the market was paying investors a premium for investing in higher risk assets).
It’s not necessary for there to be a premium “after the risk is played out”. That’s consistent and central to the concept of risk itself. If that weren’t a potential outcome over any undetermined period of time, there would be no premium at all because there would be no risk
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
Logan Roy
Posts: 1703
Joined: Sun May 29, 2022 10:15 am

Re: SCV tilt research data

Post by Logan Roy »

Nathan Drake wrote: Tue Feb 13, 2024 2:10 pm
Logan Roy wrote: Tue Feb 13, 2024 1:44 pm
Nathan Drake wrote: Mon Feb 12, 2024 6:47 pm
Logan Roy wrote: Mon Feb 12, 2024 12:49 pm
Nathan Drake wrote: Mon Feb 12, 2024 1:02 am

SCV is simply the way to extract the highest loadings on Value. And Value can have multiple accounting measures and all work (P/B, P/Cash Flows, EV/EBITDA, etc).

There are other "factors" that do have significance in the data, but none as strong as Value and with a theoretical framework.
I think it depends what a particular academic wants to call a 'factor'.

Momentum is a more persistent market anomaly. And in the Malkiel's Monkey paper, all sorts of crazy stock screens outperform value – I think the random monkey portfolios come pretty close. Just stocks starting with the letters B-U-F-F-E-T-T works well historically – I think it beats BRK.
The 'all sorts of crazy stock screens' likely does not stand up to any sort of robustness tests, otherwise it would have been taken more seriously.

Momentum is shown as a factor, but is prone to crazy turnabouts and faces implementation cost challenges, and does not have a strong theoretical framework. Its inclusion does seem useful as a means of timing individual components in a given fund, which is part of the trading strategies implemented by funds from DFA and Avantis.
I don't think FF recognise momentum as a factor(?). But I think the theory's far more robust. It's that information travels through markets with latency – so it starts with very high information investors, and eventually filters through to 'mom & pop' retail investors, via the media .. i.e. The institutions start selling when Ted spreads go one way; while the 'mom & pop' investors are selling when the market's already 50% down and CNBC's forecasting the end of stocks.

The problem with value as a 'risk premium' is that there's no example of a market compensating risk with a higher return (after that risk has played out). We can say longer-dated bonds typically have higher yields, and more volatility, but that's a term premium; not necessarily a risk premium (Treasuries are zero risk). And other factors, like Quality, don't fit that framework at all. Others, like Size, don't exist (but should, if the market was paying investors a premium for investing in higher risk assets).
It’s not necessary for there to be a premium “after the risk is played out”. That’s consistent and central to the concept of risk itself. If that weren’t a potential outcome over any undetermined period of time, there would be no premium at all because there would be no risk
I don't get what you're saying there. Even with a fixed outcome, there would be a 'term premium'. But my point is that there's no advantage to going into a betting shop and just picking 100:1 shots. The net return should be the same, whatever level of risk you're betting on.
User avatar
Nathan Drake
Posts: 6107
Joined: Mon Apr 11, 2011 12:28 am

Re: SCV tilt research data

Post by Nathan Drake »

Logan Roy wrote: Tue Feb 13, 2024 2:18 pm
Nathan Drake wrote: Tue Feb 13, 2024 2:10 pm
Logan Roy wrote: Tue Feb 13, 2024 1:44 pm
Nathan Drake wrote: Mon Feb 12, 2024 6:47 pm
Logan Roy wrote: Mon Feb 12, 2024 12:49 pm

I think it depends what a particular academic wants to call a 'factor'.

Momentum is a more persistent market anomaly. And in the Malkiel's Monkey paper, all sorts of crazy stock screens outperform value – I think the random monkey portfolios come pretty close. Just stocks starting with the letters B-U-F-F-E-T-T works well historically – I think it beats BRK.
The 'all sorts of crazy stock screens' likely does not stand up to any sort of robustness tests, otherwise it would have been taken more seriously.

Momentum is shown as a factor, but is prone to crazy turnabouts and faces implementation cost challenges, and does not have a strong theoretical framework. Its inclusion does seem useful as a means of timing individual components in a given fund, which is part of the trading strategies implemented by funds from DFA and Avantis.
I don't think FF recognise momentum as a factor(?). But I think the theory's far more robust. It's that information travels through markets with latency – so it starts with very high information investors, and eventually filters through to 'mom & pop' retail investors, via the media .. i.e. The institutions start selling when Ted spreads go one way; while the 'mom & pop' investors are selling when the market's already 50% down and CNBC's forecasting the end of stocks.

The problem with value as a 'risk premium' is that there's no example of a market compensating risk with a higher return (after that risk has played out). We can say longer-dated bonds typically have higher yields, and more volatility, but that's a term premium; not necessarily a risk premium (Treasuries are zero risk). And other factors, like Quality, don't fit that framework at all. Others, like Size, don't exist (but should, if the market was paying investors a premium for investing in higher risk assets).
It’s not necessary for there to be a premium “after the risk is played out”. That’s consistent and central to the concept of risk itself. If that weren’t a potential outcome over any undetermined period of time, there would be no premium at all because there would be no risk
I don't get what you're saying there. Even with a fixed outcome, there would be a 'term premium'. But my point is that there's no advantage to going into a betting shop and just picking 100:1 shots. The net return should be the same, whatever level of risk you're betting on.
The uncertainties with inflation mean that a term premium is still a risk premium despite being “fixed” in nominal terms, that’s why there is a premium.

This isn’t going in and randomly selecting 100 stocks; it’s selecting stocks with higher discount rates across the market, a proxy for risk and an additional premium embedded into stock prices for taking such risk
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
Logan Roy
Posts: 1703
Joined: Sun May 29, 2022 10:15 am

Re: SCV tilt research data

Post by Logan Roy »

Nathan Drake wrote: Tue Feb 13, 2024 2:24 pm
Logan Roy wrote: Tue Feb 13, 2024 2:18 pm
Nathan Drake wrote: Tue Feb 13, 2024 2:10 pm
Logan Roy wrote: Tue Feb 13, 2024 1:44 pm
Nathan Drake wrote: Mon Feb 12, 2024 6:47 pm

The 'all sorts of crazy stock screens' likely does not stand up to any sort of robustness tests, otherwise it would have been taken more seriously.

Momentum is shown as a factor, but is prone to crazy turnabouts and faces implementation cost challenges, and does not have a strong theoretical framework. Its inclusion does seem useful as a means of timing individual components in a given fund, which is part of the trading strategies implemented by funds from DFA and Avantis.
I don't think FF recognise momentum as a factor(?). But I think the theory's far more robust. It's that information travels through markets with latency – so it starts with very high information investors, and eventually filters through to 'mom & pop' retail investors, via the media .. i.e. The institutions start selling when Ted spreads go one way; while the 'mom & pop' investors are selling when the market's already 50% down and CNBC's forecasting the end of stocks.

The problem with value as a 'risk premium' is that there's no example of a market compensating risk with a higher return (after that risk has played out). We can say longer-dated bonds typically have higher yields, and more volatility, but that's a term premium; not necessarily a risk premium (Treasuries are zero risk). And other factors, like Quality, don't fit that framework at all. Others, like Size, don't exist (but should, if the market was paying investors a premium for investing in higher risk assets).
It’s not necessary for there to be a premium “after the risk is played out”. That’s consistent and central to the concept of risk itself. If that weren’t a potential outcome over any undetermined period of time, there would be no premium at all because there would be no risk
I don't get what you're saying there. Even with a fixed outcome, there would be a 'term premium'. But my point is that there's no advantage to going into a betting shop and just picking 100:1 shots. The net return should be the same, whatever level of risk you're betting on.
The uncertainties with inflation mean that a term premium is still a risk premium despite being “fixed” in nominal terms, that’s why there is a premium.

This isn’t going in and randomly selecting 100 stocks; it’s selecting stocks with higher discount rates across the market, a proxy for risk and an additional premium embedded into stock prices for taking such risk
I get the premise. So why would Quality stocks have a premium?
User avatar
TimeIsYourFriend
Posts: 393
Joined: Sun Dec 17, 2023 11:19 am

Re: SCV tilt research data

Post by TimeIsYourFriend »

Logan Roy wrote: Tue Feb 13, 2024 4:22 pm
Nathan Drake wrote: Tue Feb 13, 2024 2:24 pm
Logan Roy wrote: Tue Feb 13, 2024 2:18 pm
Nathan Drake wrote: Tue Feb 13, 2024 2:10 pm
Logan Roy wrote: Tue Feb 13, 2024 1:44 pm

I don't think FF recognise momentum as a factor(?). But I think the theory's far more robust. It's that information travels through markets with latency – so it starts with very high information investors, and eventually filters through to 'mom & pop' retail investors, via the media .. i.e. The institutions start selling when Ted spreads go one way; while the 'mom & pop' investors are selling when the market's already 50% down and CNBC's forecasting the end of stocks.

The problem with value as a 'risk premium' is that there's no example of a market compensating risk with a higher return (after that risk has played out). We can say longer-dated bonds typically have higher yields, and more volatility, but that's a term premium; not necessarily a risk premium (Treasuries are zero risk). And other factors, like Quality, don't fit that framework at all. Others, like Size, don't exist (but should, if the market was paying investors a premium for investing in higher risk assets).
It’s not necessary for there to be a premium “after the risk is played out”. That’s consistent and central to the concept of risk itself. If that weren’t a potential outcome over any undetermined period of time, there would be no premium at all because there would be no risk
I don't get what you're saying there. Even with a fixed outcome, there would be a 'term premium'. But my point is that there's no advantage to going into a betting shop and just picking 100:1 shots. The net return should be the same, whatever level of risk you're betting on.
The uncertainties with inflation mean that a term premium is still a risk premium despite being “fixed” in nominal terms, that’s why there is a premium.

This isn’t going in and randomly selecting 100 stocks; it’s selecting stocks with higher discount rates across the market, a proxy for risk and an additional premium embedded into stock prices for taking such risk
I get the premise. So why would Quality stocks have a premium?
I never got this claim. That you can use simple profitability metrics to target stocks with higher profitability and expect a premium. Any half-drunk investor can do the same with simple metrics and it presumes that the market doesn't like companies that are profitable and therefore misprices them - what?
"Time is your friend; impulse is your enemy." - John C. Bogle
User avatar
Nathan Drake
Posts: 6107
Joined: Mon Apr 11, 2011 12:28 am

Re: SCV tilt research data

Post by Nathan Drake »

Logan Roy wrote: Tue Feb 13, 2024 4:22 pm
Nathan Drake wrote: Tue Feb 13, 2024 2:24 pm
Logan Roy wrote: Tue Feb 13, 2024 2:18 pm
Nathan Drake wrote: Tue Feb 13, 2024 2:10 pm
Logan Roy wrote: Tue Feb 13, 2024 1:44 pm

I don't think FF recognise momentum as a factor(?). But I think the theory's far more robust. It's that information travels through markets with latency – so it starts with very high information investors, and eventually filters through to 'mom & pop' retail investors, via the media .. i.e. The institutions start selling when Ted spreads go one way; while the 'mom & pop' investors are selling when the market's already 50% down and CNBC's forecasting the end of stocks.

The problem with value as a 'risk premium' is that there's no example of a market compensating risk with a higher return (after that risk has played out). We can say longer-dated bonds typically have higher yields, and more volatility, but that's a term premium; not necessarily a risk premium (Treasuries are zero risk). And other factors, like Quality, don't fit that framework at all. Others, like Size, don't exist (but should, if the market was paying investors a premium for investing in higher risk assets).
It’s not necessary for there to be a premium “after the risk is played out”. That’s consistent and central to the concept of risk itself. If that weren’t a potential outcome over any undetermined period of time, there would be no premium at all because there would be no risk
I don't get what you're saying there. Even with a fixed outcome, there would be a 'term premium'. But my point is that there's no advantage to going into a betting shop and just picking 100:1 shots. The net return should be the same, whatever level of risk you're betting on.
The uncertainties with inflation mean that a term premium is still a risk premium despite being “fixed” in nominal terms, that’s why there is a premium.

This isn’t going in and randomly selecting 100 stocks; it’s selecting stocks with higher discount rates across the market, a proxy for risk and an additional premium embedded into stock prices for taking such risk
I get the premise. So why would Quality stocks have a premium?
Could be entirely behavioral. These stocks are generally quite boring (not all), and not perceived as glamorous as most their stocks with higher growth potential, so investors may be systematically underpricing them due to preference

Will it be arbitraged away? Hard to tell. If every actor in the market had the same long term goals in mind, maybe, but if it’s intrinsic to human nature, maybe not.
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
User avatar
nisiprius
Advisory Board
Posts: 51940
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: SCV tilt research data

Post by nisiprius »

Logan Roy wrote: Tue Feb 13, 2024 1:44 pm ...I don't think FF recognise momentum as a factor(?)...
They don't. Their three-factor theory recognized size and value. Their five-factor theory recognizes size, value, profitability, and investment.

And yet momentum is widely accepted within the factor community. I'm not in a position to judge but this inconsistency in the basic list of factors from experts is one of the reasons for my skepticism.

It's not as if all experts listed the same factors in about the same order of size and some of them just went a little farther down the list than others. So most astronomers agree that Pluto shouldn't be called a planet; fine. This is more like Fama and French not having Neptune on their list of planets.
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.
dcabler
Posts: 4409
Joined: Wed Feb 19, 2014 10:30 am

Re: SCV tilt research data

Post by dcabler »

nisiprius wrote: Tue Feb 13, 2024 5:08 pm
Logan Roy wrote: Tue Feb 13, 2024 1:44 pm ...I don't think FF recognise momentum as a factor(?)...
They don't. Their three-factor theory recognized size and value. Their five-factor theory recognizes size, value, profitability, and investment.

And yet momentum is widely accepted within the factor community. I'm not in a position to judge but this inconsistency in the basic list of factors from experts is one of the reasons for my skepticism.

It's not as if all experts listed the same factors in about the same order of size and some of them just went a little farther down the list than others. So most astronomers agree that Pluto shouldn't be called a planet; fine. This is more like Fama and French not having Neptune on their list of planets.
About halfway down Fama French's website is Momentum Factor
https://mba.tuck.dartmouth.edu/pages/fa ... brary.html
search the page for "Momentum Factor"

Cheers.
Northern Flicker
Posts: 15159
Joined: Fri Apr 10, 2015 12:29 am

Re: SCV tilt research data

Post by Northern Flicker »

What FF traditionally lacked was a model with both momentum and quality. Momentum is the factor least compatible with the efficient market hypothesis.
Last edited by Northern Flicker on Tue Feb 13, 2024 7:34 pm, edited 1 time in total.
Logan Roy
Posts: 1703
Joined: Sun May 29, 2022 10:15 am

Re: SCV tilt research data

Post by Logan Roy »

Nathan Drake wrote: Tue Feb 13, 2024 4:48 pm
Logan Roy wrote: Tue Feb 13, 2024 4:22 pm
Nathan Drake wrote: Tue Feb 13, 2024 2:24 pm
Logan Roy wrote: Tue Feb 13, 2024 2:18 pm
Nathan Drake wrote: Tue Feb 13, 2024 2:10 pm

It’s not necessary for there to be a premium “after the risk is played out”. That’s consistent and central to the concept of risk itself. If that weren’t a potential outcome over any undetermined period of time, there would be no premium at all because there would be no risk
I don't get what you're saying there. Even with a fixed outcome, there would be a 'term premium'. But my point is that there's no advantage to going into a betting shop and just picking 100:1 shots. The net return should be the same, whatever level of risk you're betting on.
The uncertainties with inflation mean that a term premium is still a risk premium despite being “fixed” in nominal terms, that’s why there is a premium.

This isn’t going in and randomly selecting 100 stocks; it’s selecting stocks with higher discount rates across the market, a proxy for risk and an additional premium embedded into stock prices for taking such risk
I get the premise. So why would Quality stocks have a premium?
Could be entirely behavioral. These stocks are generally quite boring (not all), and not perceived as glamorous as most their stocks with higher growth potential, so investors may be systematically underpricing them due to preference

Will it be arbitraged away? Hard to tell. If every actor in the market had the same long term goals in mind, maybe, but if it’s intrinsic to human nature, maybe not.
The problem is that this doesn't lend any weight to the theory that the market's paying you a premium for you taking more risk. (rather the opposite)

Which sounds like we're just changing the theory to fit the result. The market pays you for taking more risk (when it's value) and less risk (when it's quality).
User avatar
Nathan Drake
Posts: 6107
Joined: Mon Apr 11, 2011 12:28 am

Re: SCV tilt research data

Post by Nathan Drake »

Logan Roy wrote: Tue Feb 13, 2024 6:36 pm
Nathan Drake wrote: Tue Feb 13, 2024 4:48 pm
Logan Roy wrote: Tue Feb 13, 2024 4:22 pm
Nathan Drake wrote: Tue Feb 13, 2024 2:24 pm
Logan Roy wrote: Tue Feb 13, 2024 2:18 pm

I don't get what you're saying there. Even with a fixed outcome, there would be a 'term premium'. But my point is that there's no advantage to going into a betting shop and just picking 100:1 shots. The net return should be the same, whatever level of risk you're betting on.
The uncertainties with inflation mean that a term premium is still a risk premium despite being “fixed” in nominal terms, that’s why there is a premium.

This isn’t going in and randomly selecting 100 stocks; it’s selecting stocks with higher discount rates across the market, a proxy for risk and an additional premium embedded into stock prices for taking such risk
I get the premise. So why would Quality stocks have a premium?
Could be entirely behavioral. These stocks are generally quite boring (not all), and not perceived as glamorous as most their stocks with higher growth potential, so investors may be systematically underpricing them due to preference

Will it be arbitraged away? Hard to tell. If every actor in the market had the same long term goals in mind, maybe, but if it’s intrinsic to human nature, maybe not.
The problem is that this doesn't lend any weight to the theory that the market's paying you a premium for you taking more risk. (rather the opposite)

Which sounds like we're just changing the theory to fit the result. The market pays you for taking more risk (when it's value) and less risk (when it's quality).
Right, the market may be paying you for having more for having "boring, slow and steady" returns rather than moonshots and lottery tickets. I don't consider this a premium, more of a behavioral quirk that seems to find itself into market pricing and be difficult to arbitrage away.

Value I consider a genuine risk-based premium with some behavioral aspect as well, but quality seems solely behavioral. As an investor in these products, I expect a Value premium based on risk but also hope (but don't count on) receiving one for any of the other factors with a behavioral explanation.
20% VOO | 20% VXUS | 20% AVUV | 20% AVDV | 20% AVES
Logan Roy
Posts: 1703
Joined: Sun May 29, 2022 10:15 am

Re: SCV tilt research data

Post by Logan Roy »

dcabler wrote: Tue Feb 13, 2024 5:21 pm
nisiprius wrote: Tue Feb 13, 2024 5:08 pm
Logan Roy wrote: Tue Feb 13, 2024 1:44 pm ...I don't think FF recognise momentum as a factor(?)...
They don't. Their three-factor theory recognized size and value. Their five-factor theory recognizes size, value, profitability, and investment.

And yet momentum is widely accepted within the factor community. I'm not in a position to judge but this inconsistency in the basic list of factors from experts is one of the reasons for my skepticism.

It's not as if all experts listed the same factors in about the same order of size and some of them just went a little farther down the list than others. So most astronomers agree that Pluto shouldn't be called a planet; fine. This is more like Fama and French not having Neptune on their list of planets.
About halfway down Fama French's website is Momentum Factor
https://mba.tuck.dartmouth.edu/pages/fa ... brary.html
search the page for "Momentum Factor"

Cheers.
There's an article here that includes Fama's views on momentum:

"In fact in the video he says he’s still “hoping it goes away.” This makes sense because, as Fama discusses, it’s one of the harder factors to reconcile with the Efficient Markets Hypothesis (EMH), a contribution for which he is justifiably acclaimed. In fact, he calls momentum the “biggest embarrassment to the theory” out there."

https://www.aqr.com/Insights/Perspectiv ... n-Momentum
Logan Roy
Posts: 1703
Joined: Sun May 29, 2022 10:15 am

Re: SCV tilt research data

Post by Logan Roy »

Nathan Drake wrote: Tue Feb 13, 2024 6:43 pm
Logan Roy wrote: Tue Feb 13, 2024 6:36 pm
Nathan Drake wrote: Tue Feb 13, 2024 4:48 pm
Logan Roy wrote: Tue Feb 13, 2024 4:22 pm
Nathan Drake wrote: Tue Feb 13, 2024 2:24 pm

The uncertainties with inflation mean that a term premium is still a risk premium despite being “fixed” in nominal terms, that’s why there is a premium.

This isn’t going in and randomly selecting 100 stocks; it’s selecting stocks with higher discount rates across the market, a proxy for risk and an additional premium embedded into stock prices for taking such risk
I get the premise. So why would Quality stocks have a premium?
Could be entirely behavioral. These stocks are generally quite boring (not all), and not perceived as glamorous as most their stocks with higher growth potential, so investors may be systematically underpricing them due to preference

Will it be arbitraged away? Hard to tell. If every actor in the market had the same long term goals in mind, maybe, but if it’s intrinsic to human nature, maybe not.
The problem is that this doesn't lend any weight to the theory that the market's paying you a premium for you taking more risk. (rather the opposite)

Which sounds like we're just changing the theory to fit the result. The market pays you for taking more risk (when it's value) and less risk (when it's quality).
Right, the market may be paying you for having more for having "boring, slow and steady" returns rather than moonshots and lottery tickets. I don't consider this a premium, more of a behavioral quirk that seems to find itself into market pricing and be difficult to arbitrage away.

Value I consider a genuine risk-based premium with some behavioral aspect as well, but quality seems solely behavioral. As an investor in these products, I expect a Value premium based on risk but also hope (but don't count on) receiving one for any of the other factors with a behavioral explanation.
Then we've got a market that's both efficient and inefficient. And we're still changing the theory to fit the result – when the purpose of the theory should be to validate that the result isn't just data-mining.
Post Reply