Has SCV underperformed TSM since 1984?

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Topic Author
skierincolorado
Posts: 2377
Joined: Sat Mar 21, 2020 10:56 am

Has SCV underperformed TSM since 1984?

Post by skierincolorado »

Per Simba's spreadsheet SCV has underperformed TSM since 1984. First of all, I'm wondering if I am missing anything in that statement. Second, if SCV has underperformed for 38 years, how reliable is the SCV premium? I knew it was variable and one needed to hold for a very long time horizon. But 38 years is longer than most people's investment horizon. And even if one had a 38 year investment horizon before withdrawals, 38 years is a long enough period to argue something fundamental has changed and the premium no longer exists.
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Has SCV underperformed TSM since 1984?

Post by Apathizer »

I compared VTI and DFA US SV. Over the last 20 years DFA has out-performed by about 1% annually, and that's even after 2020 which was brutal for SV. If we cut it off at 2019, DFA out-performed by about 1.75% annually. To me that seems like compelling evidence a SV slant is worthwhile. If nothing else SV has a fairly low correlation with the TSM, so I think it provides beneficial diversification.
https://www.portfoliovisualizer.com/bac ... ion2_2=100
https://www.portfoliovisualizer.com/bac ... ion2_2=100
Last edited by Apathizer on Wed Jul 06, 2022 1:54 am, edited 4 times in total.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
Alpha4
Posts: 176
Joined: Tue Apr 17, 2012 8:47 pm

Re: Has SCV underperformed TSM since 1984?

Post by Alpha4 »

A couple of things to note:

One, as Apathizer pointed out, even if SCV had a slightly smaller expected return than TSM it would still--in terms of diversification and risk-adjusted returns--make holding it worthwhile so long as it wasn't at a 100% correlation with TSM (which it isn't).

Two, the Simba spreadsheet doesn't include returns so far in 2022 (to 6-30-2022). if you do that the returns starting at the beginning of 1984 and going to the end of June 2022 are virtually dead even; SCV turns $10K on 1-1-84 into $502K today and TSM turns that same $10K into $505.6K today....that's maybe a few basis points difference at most in annual CAGR.

Three, I might be wrong here but from my understanding the Simba spreadhseet uses VISVX/VSIAX/and the ETF versions thereof as its SCV standard for use in backtesting rather than a truly small valuey choice like DFSVX (or AVUV, or RZV, etc). Compared to a "purer" SCV offering such as those three funds, VISVX is kind of like a "sorta SCV if you squint hard enough" choice but is really more a quasi-MCV quasi-SCV "small cap value lite" fund with lower loadings/exposures to the value and small factors/premiums than DFSVX/AVUV/RZV. Lower loadings to value and small lead to lower expected returns (and indeed, to lower actual realized returns).
rkhusky
Posts: 17763
Joined: Thu Aug 18, 2011 8:09 pm

Re: Has SCV underperformed TSM since 1984?

Post by rkhusky »

Apathizer wrote: Wed Jul 06, 2022 1:09 am. If nothing else SV has a fairly low correlation with the TSM, so I think it provides beneficial diversification.
Low correlation is not enough, you also need sufficiently good returns to be beneficial. Cash has low correlation with stocks, but you probably don’t want a large cache of $100 bills in your portfolio.
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Has SCV underperformed TSM since 1984?

Post by Apathizer »

rkhusky wrote: Wed Jul 06, 2022 12:00 pm
Apathizer wrote: Wed Jul 06, 2022 1:09 am. If nothing else SV has a fairly low correlation with the TSM, so I think it provides beneficial diversification.
Low correlation is not enough, you also need sufficiently good returns to be beneficial. Cash has low correlation with stocks, but you probably don’t want a large cache of $100 bills in your portfolio.
I have presented evidence SV has had better than and is fairly uncorrelated with the TSM. I'd say that's 'sufficiently good'.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
Marseille07
Posts: 16054
Joined: Fri Nov 06, 2020 12:41 pm

Re: Has SCV underperformed TSM since 1984?

Post by Marseille07 »

skierincolorado wrote: Wed Jul 06, 2022 1:00 am Per Simba's spreadsheet SCV has underperformed TSM since 1984. First of all, I'm wondering if I am missing anything in that statement. Second, if SCV has underperformed for 38 years, how reliable is the SCV premium? I knew it was variable and one needed to hold for a very long time horizon. But 38 years is longer than most people's investment horizon. And even if one had a 38 year investment horizon before withdrawals, 38 years is a long enough period to argue something fundamental has changed and the premium no longer exists.
I don't think you are missing anything. I remember seeing some data which showed enormous performance of SCV during the 1930s and 1940s. We know SCV is underperforming since 2009 or so. 38-year underperformance does not surprise me at all.

That said, you might find pockets of SCV outperformance still, for example between 2000~2022.
rkhusky
Posts: 17763
Joined: Thu Aug 18, 2011 8:09 pm

Re: Has SCV underperformed TSM since 1984?

Post by rkhusky »

Apathizer wrote: Wed Jul 06, 2022 12:18 pm
rkhusky wrote: Wed Jul 06, 2022 12:00 pm
Apathizer wrote: Wed Jul 06, 2022 1:09 am. If nothing else SV has a fairly low correlation with the TSM, so I think it provides beneficial diversification.
Low correlation is not enough, you also need sufficiently good returns to be beneficial. Cash has low correlation with stocks, but you probably don’t want a large cache of $100 bills in your portfolio.
I have presented evidence SV has had better than and is fairly uncorrelated with the TSM. I'd say that's 'sufficiently good'.
The last time I looked SCV had an 80-90% correlation with TSM.
User avatar
nisiprius
Advisory Board
Posts: 52211
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: Has SCV underperformed TSM since 1984?

Post by nisiprius »

Alpha4 wrote: Wed Jul 06, 2022 1:19 am...One, as Apathizer pointed out, even if SCV had a slightly smaller expected return than TSM it would still--in terms of diversification and risk-adjusted returns--make holding it worthwhile so long as it wasn't at a 100% correlation with TSM (which it isn't)...
No, that's not true. And it is a theoretical point which is worth pinning down.

You do not necessarily get a benefit just because the correlation is lower than 100%.

Let's consider an initial portfolio P and a proposed diversifier D, and for the sake of argument assume that our figure of merit is "the Sharpe ratio over some specified time period." So, we will say D is "just as good as" P if it has just as high a Sharpe ratio. And we will say "D improves P" if adding D to P raises the Sharpe ratio of P.

Suppose that P and D have correlation that is less than 100% but one that would be called "high" in a statistics class. For example, let's say the correlation coefficient ρ = 80%.

It is true that if P and D are equally "good," have the same Sharpe ratio over some time period of interest, that D will improve P as long as p < 100%.

But it is not true if D is not a good investment in itself.

Specifically, if we take the Sharpe ratios of P and D and divide the smaller by the larger to get R... D will only improve P if p < P

If the Sharpe ratio of the diversifier is lower than that of the portfolio, then it is not good enough to have "not 100% correlation." The correlation has to be lower than the ratio we've just defined.

In other words: in order to improve a portfolio, the diversifier need not have negative correlation and need not have zero correlation--but it not necessarily good enough to have a correlation of less than 100%.

If the portfolio has a Sharpe ratio of 0.5, and the correlation with the diversifier is 80%, then the diversifier will improve the portfolio if it has a Sharpe ratio of better than 0.4 by itself. But if has a Sharpe ratio of less than 0.4, adding it to the portfolio will make it worse.

(An interest special case is assets with a Sharpe ratio of zero... for example, long-short alt funds that don't beat their benchmark of Treasury bills. In order for them to improve a portfolio, it is not sufficient for them to have low correlation or even zero correlation, they will only improve the portfolio if their correlation is negative.)

A diversifier whose Sharpe ratio is lower than the portfolio creates a balance of effects. Low correlation tends to increase the portfolio, but the lower Sharpe ratio tends to drag it down, and you don't know automatically in advance which effect predominates. It is a balancing act. It is theoretically possible for a lousy investment to improve a portfolio through low correlation, but that doesn't usually happen in the real world.
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.
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Has SCV underperformed TSM since 1984?

Post by Apathizer »

rkhusky wrote: Wed Jul 06, 2022 12:29 pm
Apathizer wrote: Wed Jul 06, 2022 12:18 pm
rkhusky wrote: Wed Jul 06, 2022 12:00 pm
Apathizer wrote: Wed Jul 06, 2022 1:09 am. If nothing else SV has a fairly low correlation with the TSM, so I think it provides beneficial diversification.
Low correlation is not enough, you also need sufficiently good returns to be beneficial. Cash has low correlation with stocks, but you probably don’t want a large cache of $100 bills in your portfolio.
I have presented evidence SV has had better than and is fairly uncorrelated with the TSM. I'd say that's 'sufficiently good'.
The last time I looked SCV had an 80-90% correlation with TSM.
Yes, that's about right, but it's only part of a well-diversified portfolio. I'm not arguing it should be the only diversifier.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
User avatar
nisiprius
Advisory Board
Posts: 52211
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: Has SCV underperformed TSM since 1984?

Post by nisiprius »

Let me use one of Paul Merriman's charts to illustrate a different point from the point it is often used to illustrate:

Image

The point I want to make is that, like so many financial assets, small-cap value's outperformance is very bursty. Taking his arrows at face value, there have been four "bursts" in the period 1928-2020. (We might or might not be in another one right now, but I think it's too early to tell).

What this means is that any comparison of small-cap value with the S&P 500 is highly end-point dependent. It also means that we can almost predict how the comparison will turn out simply by knowing the timing of those bursts and asking how many of them the time period brackets.

Thus, the thread title asks about "since 1984," and on the evidence of the chart, 1984 to the present has included only one burst.

For about 1983-present, the way any comparison turns out depends almost entirely on one single fact: did the time period involved include 2000-2004, or not?

There might be something to it, but finding out whether or not there is rendered very difficult by the burstiness of the data and by the unavailability of systematic, passive-ish ways to invest in the small and value factors prior to about 1993.

Data quality is a problem, too. It is now generally acknowledged that the paper that brought attention to the size effect, Rolf Banz' 1981 paper, was founded on faulty data and greatly exaggerated the size of the effect--if, indeed, it exists at all. And a researcher, Mathias Hasler, interviewed here, found that "the construction of the original HMR portfolio includes six seemingly innocuous decisions," and it turned out that they actually matter and that by chance all of these decisions tended to lead to larger estimates for the size of the value premium. Averaging over all the ways those choices could have been made suggested to Hasler that the size of the value premium could have been exaggerated by a factor of four. Or, at least, that the data isn't adequate to estimate the size of the value premium with any precision at all.
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
skierincolorado
Posts: 2377
Joined: Sat Mar 21, 2020 10:56 am

Re: Has SCV underperformed TSM since 1984?

Post by skierincolorado »

nisiprius wrote: Wed Jul 06, 2022 2:21 pm Let me use one of Paul Merriman's charts to illustrate a different point from the point it is often used to illustrate:

Image

The point I want to make is that, like so many financial assets, small-cap value's outperformance is very bursty. Taking his arrows at face value, there have been four "bursts" in the period 1928-2020. (We might or might not be in another one right now, but I think it's too early to tell).

What this means is that any comparison of small-cap value with the S&P 500 is highly end-point dependent. It also means that we can almost predict how the comparison will turn out simply by knowing the timing of those bursts and asking how many of them the time period brackets.

Thus, the thread title asks about "since 1984," and on the evidence of the chart, 1984 to the present has included only one burst.

For about 1983-present, the way any comparison turns out depends almost entirely on one single fact: did the time period involved include 2000-2004, or not?

There might be something to it, but finding out whether or not there is rendered very difficult by the burstiness of the data and by the unavailability of systematic, passive-ish ways to invest in the small and value factors prior to about 1993.

Data quality is a problem, too. It is now generally acknowledged that the paper that brought attention to the size effect, Rolf Banz' 1981 paper, was founded on faulty data and greatly exaggerated the size of the effect--if, indeed, it exists at all. And a researcher, Mathias Hasler, interviewed here, found that "the construction of the original HMR portfolio includes six seemingly innocuous decisions," and it turned out that they actually matter and that by chance all of these decisions tended to lead to larger estimates for the size of the value premium. Averaging over all the ways those choices could have been made suggested to Hasler that the size of the value premium could have been exaggerated by a factor of four. Or, at least, that the data isn't adequate to estimate the size of the value premium with any precision at all.
I get that it can be bursty. But 38 years is longer than any other period without a premium by far. I don't like charts like the one that you posted because the annotations can be used to highlight features in ways that make them appear more persistent or reliable than they actually are. But if you look closely the last two horizonol annotated lines are two of the longest horizontal periods on the chart. Not only that but they their vertical separation is the smallest of any pair on the graph. In other words the current plateau is barely higher than the last plateau and both plateaus are quite long. Not only that but continuing the data to present makes the current plateau even lower. And the horizontal lines help the eyes gloss over important details that occur within the last 38 years. It doesn't do us much good that scv was near an all time high vs sp500 in 2003. It doesn't change the fact that 38 years is an exceptionally long period of underperformance... especially when the volatility is higher and sharpe ratio lower.

Like I said I get that it is bursty. But 38 years is too long for most investors horizons. And it's a long enough period to throw into doubt the whole premium. Something could have fundamentally changed.

I do hold a small stake in scv.
Call_Me_Op
Posts: 9881
Joined: Mon Sep 07, 2009 2:57 pm
Location: Milky Way

Re: Has SCV underperformed TSM since 1984?

Post by Call_Me_Op »

Even though you picked a relatively long period of time, it was still cherry-picked. Extend it back to 1972 (the earliest data Portfolio Vusualizer uses) and the premium rises to 3%.
Best regards, -Op | | "In the middle of difficulty lies opportunity." Einstein
Topic Author
skierincolorado
Posts: 2377
Joined: Sat Mar 21, 2020 10:56 am

Re: Has SCV underperformed TSM since 1984?

Post by skierincolorado »

Call_Me_Op wrote: Wed Jul 06, 2022 4:38 pm Even though you picked a relatively long period of time, it was still cherry-picked. Extend it back to 1972 (the earliest data Portfolio Vusualizer uses) and the premium rises to 3%.
I would not call it cherry picked. It answers a meaningful question which is how long has it been since the premium showed up. I think it is fair and meaningful to ask how long you have to go back to see a statistically significant premium. It tells you how long you might have to hold and it also provides evidence on whether the premium still exists or not.
garlandwhizzer
Posts: 3565
Joined: Fri Aug 06, 2010 3:42 pm

Re: Has SCV underperformed TSM since 1984?

Post by garlandwhizzer »

Factor model returns make no difference. The models are unrealistic and and make assumptions that have never existed in the real world like cost-free, fiction-free long/short. All factors except beta (TSM) get long short. Beta doesn't get to add onto its long return a cost-free short which in effect doubles the returns. It would be easy to add to TSM's long position the negative of a cost-fee index that shorts TSM. That would in effect double TSM's return, but factor models don't do that. Instead they subtract from TSM's long only returns the risk free rate of T bond yields. The models claim that beta is "just another factor,'' but quite clearly there are different rules for beta than the other factors. These rules magnify factor returns and minimize by at least half beta returns.

TSM in reality, as opposed to factor theory, is a multi-factor fund whose relative factors are defined and whose relative factor weights are set by the aggregate judgement of market participants. They are highly educated, very experienced professionals making real time decisions. TSM is by far the best multi-factor fund as its outperformance relative factor maven multi-factor funds clearly demonstrates.

If beta (TSM) is just another factor, as has been claimed numerous times, it should be treated on a level playing field with other factors (cost-free, long-short) which for starters doubles beta's return. The next thing to level the playing field is to subtract from each non-beta factor returns the estimated trading costs (which are big in MOM) and losses due to trading frictions (which are huge in the SC space). TSM does essentially no trading so it is almost immune from trading costs and frictions. Hence returns from all non-beta factors should be reduced by a fair and just estimate of these costs. Then we'd have a level playing field. Good luck on getting that. Looking at the returns as presented in factor models (S + V = about 9%) is simply a joke.

The more accurate way to measure the difference in SCV and TSM in the real world is to follow real funds over long time periods. There were no SCV funds that I'm aware of prior to 1993 when the factors were described. We can't go further back than that. I think it reasonable to pick a SCV fund that is run by state-of-the-art Nobel Prize winning experts. That fund over this time frame (1993 - present) is DFSVX. Comparing the two we find that over a 29 year period SCV outperformed TSM by 1.3% but it did so with significantly greater volatility as measured by standard deviation and with greater risk as measured by maximum drawdown. Actually, TSM had a higher Sharpe ratio (return per unit of risk). Actually each of the two go through long multi-year periods of outperformance alternating with multi-year periods of underperformance.

https://www.portfoliovisualizer.com/bac ... ion2_2=100


Long term SCV is expected to produce slightly increased return if you can can emotionally handle the increased risk and volatility in bear markets without panic selling. A SCV fund run by experts has not located some secret spot on the risk/return scale where increased returns come without increased risk. Interestingly, the same results over this time frame could have been achieved by modestly increasing TSM exposure relative to bonds. Both increase risk and volatility to produce increased returns. It's really very simple.

Personally, I hold 75 TSM/25 SCV in an equity dominated portfolio which smoothes out the SCV ride and may slightly increase overall portfolio returns above 100% TSM. TSM only I believe to be a solid choice, ultra low cost and wide diversification. Sometimes however LC growth stocks get so overvalued that they carry bubble risk. That happened in the late 1990s and IMO it happened in 2020 - 2021 with skyrocketing valuations in many tech darlings. SCV exposure comes in handy at such times. Nobody ever gets overly excited and euphoric about small struggling companies.

Garland Whizzer
User avatar
nisiprius
Advisory Board
Posts: 52211
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: Has SCV underperformed TSM since 1984?

Post by nisiprius »

skierincolorado wrote: Wed Jul 06, 2022 4:15 pm...I get that it can be bursty. But 38 years is longer than any other period without a premium by far. I don't like charts like the one that you posted because the annotations can be used to highlight features in ways that make them appear more persistent or reliable than they actually are. But if you look closely the last two horizonol annotated lines are two of the longest horizontal periods on the chart. Not only that but they their vertical separation is the smallest of any pair on the graph. In other words the current plateau is barely higher than the last plateau and both plateaus are quite long. Not only that but continuing the data to present makes the current plateau even lower. And the horizontal lines help the eyes gloss over important details that occur within the last 38 years. It doesn't do us much good that scv was near an all time high vs sp500 in 2003. It doesn't change the fact that 38 years is an exceptionally long period of underperformance... especially when the volatility is higher and sharpe ratio lower.

Like I said I get that it is bursty. But 38 years is too long for most investors horizons. And it's a long enough period to throw into doubt the whole premium. Something could have fundamentally changed.

I do hold a small stake in scv.
I was trying to be careful to be objective, and I guess failed to make it clear that I'm personally a SCV skeptic. My point is that the nature of the data is such that it is pretty darned hard to tell if there is really a benefit or not. I was not arguing that you should hold tight because patient investors are certain to collect some of those sure-thing bursts.

When the structure of the returns is like that--short bursts and long intervals in between them--the bursts dominate any medium-term time interval. And as you extend the time period back into the past, you start getting into data problems, and survivorship issues in the corpus of data, and serious reservations about whether it was really the same thing, with the same quantitative behavior, fifty years ago.

A great deal of discussions of various strategies suffer from what I call the "one great shining moment" problem. Very often some fund or portfolio with impressive outperformance owes it all to one great shining moment. Or, less obviously, some strategy that has had some long good periods and some long bad periods and it depends if your 25-year-long period encompassed two good and one bad, or two bad and one good.

Another big problem with Paul Merriman's presentations in general, and this chart in particular, is that he does not even try to account for risk at all.
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
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: Has SCV underperformed TSM since 1984?

Post by willthrill81 »

Alpha4 wrote: Wed Jul 06, 2022 1:19 am A couple of things to note:

One, as Apathizer pointed out, even if SCV had a slightly smaller expected return than TSM it would still--in terms of diversification and risk-adjusted returns--make holding it worthwhile so long as it wasn't at a 100% correlation with TSM (which it isn't).

Two, the Simba spreadsheet doesn't include returns so far in 2022 (to 6-30-2022). if you do that the returns starting at the beginning of 1984 and going to the end of June 2022 are virtually dead even; SCV turns $10K on 1-1-84 into $502K today and TSM turns that same $10K into $505.6K today....that's maybe a few basis points difference at most in annual CAGR.

Three, I might be wrong here but from my understanding the Simba spreadhseet uses VISVX/VSIAX/and the ETF versions thereof as its SCV standard for use in backtesting rather than a truly small valuey choice like DFSVX (or AVUV, or RZV, etc). Compared to a "purer" SCV offering such as those three funds, VISVX is kind of like a "sorta SCV if you squint hard enough" choice but is really more a quasi-MCV quasi-SCV "small cap value lite" fund with lower loadings/exposures to the value and small factors/premiums than DFSVX/AVUV/RZV. Lower loadings to value and small lead to lower expected returns (and indeed, to lower actual realized returns).
Another issue that needs to be considered is the consistency of the returns. Tyler9000, one of our posters here who runs the fantastic site Portfolio Charts, created a metric for this which he refers to as start-date sensitivity, discussed in this thread. Using that metric, since 1970, SCV has had lower start-date sensitivity than a 60/40 AA. And this makes sense from a factor standpoint because SCV is impacted by three factors rather than only one (maybe 'one and a half' for a 60/40).

Note that start-date sensitivity is not the same as 'volatility' as measured by standard deviation. Standard deviation explains less than half the variation in start-date sensitivity.
The Sensible Steward
heyyou
Posts: 4461
Joined: Tue Feb 20, 2007 3:58 pm

Re: Has SCV underperformed TSM since 1984?

Post by heyyou »

Tyler does show domestic Large caps, SCV, LTT, and gold, as the top portfolio holdings from a long list of possible choices. Those are more about their combined inverse correlations, than their separate individual performances, as seen by LTT making the list. I wonder if SCV recovers sooner sometimes, or maybe looks better because it recovered further (having fallen further).

These days, my preference is to mostly just buy and hold with less emphasis on watching for every minor rebalancing opportunity, with the current situation being an exception.
Massdriver
Posts: 233
Joined: Tue Jan 02, 2018 11:05 am

Re: Has SCV underperformed TSM since 1984?

Post by Massdriver »

Nisiprius used data going back further than the 90s in a SCV thread. :sharebeer
Marseille07
Posts: 16054
Joined: Fri Nov 06, 2020 12:41 pm

Re: Has SCV underperformed TSM since 1984?

Post by Marseille07 »

Massdriver wrote: Wed Jul 06, 2022 11:29 pm Nisiprius used data going back further than the 90s in a SCV thread. :sharebeer
Iirc the conclusion was there isn't much to see?
abc132
Posts: 2435
Joined: Thu Oct 18, 2018 1:11 am

Re: Has SCV underperformed TSM since 1984?

Post by abc132 »

Apathizer wrote: Wed Jul 06, 2022 1:09 am I compared VTI and DFA US SV. Over the last 20 years DFA has out-performed by about 1% annually, and that's even after 2020 which was brutal for SV. If we cut it off at 2019, DFA out-performed by about 1.75% annually. To me that seems like compelling evidence a SV slant is worthwhile. If nothing else SV has a fairly low correlation with the TSM, so I think it provides beneficial diversification.
If you change that to the last 3-17 years VTI will win by a good 1% annually. The answer should be fairly obvious from this example: it depends on your start/end date and not everyone will capture a premium if they suddenly decide to change from total market to a tilt. Even waiting a long time may not help because it depends quite a bit on your start date. A good example is that someone who started investing 20-25 years ago and contributed periodically will still be behind with SCV even in the time periods where a fixed amount of SCV outperformed by 1%+ annually --> they didn't have enough invested to even capture the premium when it did exist.

I would advocate for more realistic data sets that include accumulation and decumulation. I think the chance of realizing a SCV premium are less than most people think, even when it does exist. It can come down to the luck of what overperforms when your assets are significant rather than what outperformed over a longer period of time.

I suggest modelling a 1$ portfolio with $1 monthly contributions for the last 20 years and seeing for yourself that SCV loses despite having the better 20 year performance. Anyone that believes they will get the SCV premium should do this.
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Has SCV underperformed TSM since 1984?

Post by Apathizer »

abc132 wrote: Thu Jul 07, 2022 12:37 am
Apathizer wrote: Wed Jul 06, 2022 1:09 am I compared VTI and DFA US SV. Over the last 20 years DFA has out-performed by about 1% annually, and that's even after 2020 which was brutal for SV. If we cut it off at 2019, DFA out-performed by about 1.75% annually. To me that seems like compelling evidence a SV slant is worthwhile. If nothing else SV has a fairly low correlation with the TSM, so I think it provides beneficial diversification.
If you change that to the last 3-17 years VTI will win by a good 1% annually. The answer should be fairly obvious from this example: it depends on your start/end date and not everyone will capture a premium if they suddenly decide to change from total market to a tilt. Even waiting a long time may not help because it depends quite a bit on your start date. A good example is that someone who started investing 20-25 years ago and contributed periodically will still be behind with SCV even in the time periods where a fixed amount of SCV outperformed by 1%+ annually --> they didn't have enough invested to even capture the premium when it did exist.

I would advocate for more realistic data sets that include accumulation and decumulation. I think the chance of realizing a SCV premium are less than most people think, even when it does exist. It can come down to the luck of what overperforms when your assets are significant rather than what outperformed over a longer period of time.

I suggest modelling a 1$ portfolio with $1 monthly contributions for the last 20 years and seeing for yourself that SCV loses despite having the better 20 year performance. Anyone that believes they will get the SCV premium should do this.
Again, size is only one factor. Yes, it may under-perform for several years, but the odds of all 5 factors under-performing for an extended period are low. That's why it's important to include all of them.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
Call_Me_Op
Posts: 9881
Joined: Mon Sep 07, 2009 2:57 pm
Location: Milky Way

Re: Has SCV underperformed TSM since 1984?

Post by Call_Me_Op »

skierincolorado wrote: Wed Jul 06, 2022 5:48 pm
Call_Me_Op wrote: Wed Jul 06, 2022 4:38 pm Even though you picked a relatively long period of time, it was still cherry-picked. Extend it back to 1972 (the earliest data Portfolio Vusualizer uses) and the premium rises to 3%.
I would not call it cherry picked. It answers a meaningful question which is how long has it been since the premium showed up. I think it is fair and meaningful to ask how long you have to go back to see a statistically significant premium. It tells you how long you might have to hold and it also provides evidence on whether the premium still exists or not.
Even over the 38-year period that you chose, there was a 1% premium for SCV. Perhaps as important, SCV is a pretty good diversifier for Total Market - so I think it is a good idea t hold some SCV even of you think your time horizon is limited. This argument (about whether to invest in SCV) will continue for a long time - but to me it's a no-brainer. Unfortunately, it is not available in my retirement account.

Of course not everyone should do it. Many people get spooked when there is negative tracking error - and that is the biggest danger.
Best regards, -Op | | "In the middle of difficulty lies opportunity." Einstein
User avatar
nisiprius
Advisory Board
Posts: 52211
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: Has SCV underperformed TSM since 1984?

Post by nisiprius »

I believe it is troublesome that there are, as far as I know, no mutual funds or ETFs that actually hold the Fama-French research portfolios. To some extent there is a bait-and-switch. The bait is presentations based on one set of stocks in the research data, and the switch is funds based on something different. If we can have weird things like an ETF that tracks the DJIA, why can't we have an ETF that tracks the Fama-French research portfolios (specifically the "Size and Book-to-Market Portfolios?")

I believe that what Mathias Hasler found is troublesome if true. (I listened to a detailed interview but haven't actually read the paper). The abstract of the paper is:
The construction of the original HML portfolio (Fama and French, 1993) includes six seemingly innocuous decisions that could easily have been replaced with alternatives that are just as reasonable. I propose such alternatives and construct HML portfolios. In sample, the average estimate of the value premium is dramatically smaller than the original estimate of the value premium. The difference is 0.09% per month and statistically significant. Out of sample, this difference is statistically insignificant. The results suggest that the original value premium estimate is upward biased due to a chance result in the original research decisions.
The issue here is not so much the bias, but the fragility.

You can define the universe of large-caps as either the CRSP Large-Cap Index (tracked by the Vanguard Large-Cap index Fund) or the Russell 1000 or the S&P 500, and you are talking about many decisions, some of which don't seem "innocuous" at all--yet there is very difference in their behavior or in any investing decisions you might make based on them:

Source

Image

In other words, the behavior of "large-cap stocks" is robust with regard to different definitions of "large-cap stocks." All sensible definitions give very similar results.

Yet the Fama-French definitions of the other categories are so fragile that the tiny, "innocuous" choices mentioned by Hasler can make up to a factor of four difference in the estimated size of the value premium.

To me, this means that the data that are used to guide factor investing is not at all robust.
Last edited by nisiprius on Thu Jul 07, 2022 7:25 am, edited 2 times in total.
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
burritoLover
Posts: 4097
Joined: Sun Jul 05, 2020 12:13 pm

Re: Has SCV underperformed TSM since 1984?

Post by burritoLover »

SCV can underperform the market for long periods 10-20 years at times. If your start/end dates in your backtest include a large portion of bad period(s), there doesn't appear to be a premium (or even negative) or it is greatly reduced. SCV is riskier than the market - if it were to deliver the premium consistently every 10 year period, then it wouldn't be as risky and likely wouldn't have the same premium to begin with. What will happen to the premium going forward - no one knows.

Image
User avatar
nisiprius
Advisory Board
Posts: 52211
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: Has SCV underperformed TSM since 1984?

Post by nisiprius »

burritoLover wrote: Thu Jul 07, 2022 7:22 am SCV can underperform the market for long periods 10-20 years at times. If your start/end dates in your backtest include a large portion of bad period(s), there doesn't appear to be a premium (or even negative) or it is greatly reduced. SCV is riskier than the market - if it were to deliver the premium consistently every 10 year period, then it wouldn't be as risky and likely wouldn't have the same premium to begin with. What will happen to the premium going forward - no one knows.
[image]
Nicely stated.

OK, I'd add the problem that extending the data to reduce the sampling error--to capture three or four bursts instead of zero or one--pushes you back into years when you have less and less confidence in the data itself. Even more seriously, you have less and less confidence that the whole period actually represents a time over which "small-cap value" has behaved in, quantitatively, the same way. Before and after the publication of Graham and Dodd's "Security Analysis," for example
Last edited by nisiprius on Thu Jul 07, 2022 7:31 am, edited 1 time in total.
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.
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Has SCV underperformed TSM since 1984?

Post by YRT70 »

skierincolorado wrote: Wed Jul 06, 2022 1:00 am Per Simba's spreadsheet SCV has underperformed TSM since 1984.
I don't use his spreadsheet but IFA puts SCV over that period (1984-now) on 11.65%, TSM 10.84%.
https://www.ifa.com/portfolios/100/#4

Portfoliovisualizer: SCV 11.64%, TSM 10.53%
https://www.portfoliovisualizer.com/bac ... ion2_2=100
Marseille07
Posts: 16054
Joined: Fri Nov 06, 2020 12:41 pm

Re: Has SCV underperformed TSM since 1984?

Post by Marseille07 »

burritoLover wrote: Thu Jul 07, 2022 7:22 am SCV can underperform the market for long periods 10-20 years at times. If your start/end dates in your backtest include a large portion of bad period(s), there doesn't appear to be a premium (or even negative) or it is greatly reduced. SCV is riskier than the market - if it were to deliver the premium consistently every 10 year period, then it wouldn't be as risky and likely wouldn't have the same premium to begin with. What will happen to the premium going forward - no one knows.

Image
This chart is nonsense. With hindsight knowledge that SCV's all-time CAGR has been higher than the S&P, you can always pencil in a chart like this (otherwise CAGR won't be higher than the S&P).
User avatar
burritoLover
Posts: 4097
Joined: Sun Jul 05, 2020 12:13 pm

Re: Has SCV underperformed TSM since 1984?

Post by burritoLover »

Marseille07 wrote: Thu Jul 07, 2022 10:10 am
burritoLover wrote: Thu Jul 07, 2022 7:22 am SCV can underperform the market for long periods 10-20 years at times. If your start/end dates in your backtest include a large portion of bad period(s), there doesn't appear to be a premium (or even negative) or it is greatly reduced. SCV is riskier than the market - if it were to deliver the premium consistently every 10 year period, then it wouldn't be as risky and likely wouldn't have the same premium to begin with. What will happen to the premium going forward - no one knows.

Image
This chart is nonsense. With hindsight knowledge that SCV's all-time CAGR has been higher than the S&P, you can always pencil in a chart like this (otherwise CAGR won't be higher than the S&P).
Not sure what you mean. This is a 90-year chart indicating that you have to suffer long periods of underperformance vs. the S&P 500 (or some estimation of the S&P 500 in earlier periods) to stick with small-cap value. Not sure why that is non-sense.
abc132
Posts: 2435
Joined: Thu Oct 18, 2018 1:11 am

Re: Has SCV underperformed TSM since 1984?

Post by abc132 »

Apathizer wrote: Thu Jul 07, 2022 1:38 am
abc132 wrote: Thu Jul 07, 2022 12:37 am
Apathizer wrote: Wed Jul 06, 2022 1:09 am I compared VTI and DFA US SV. Over the last 20 years DFA has out-performed by about 1% annually, and that's even after 2020 which was brutal for SV. If we cut it off at 2019, DFA out-performed by about 1.75% annually. To me that seems like compelling evidence a SV slant is worthwhile. If nothing else SV has a fairly low correlation with the TSM, so I think it provides beneficial diversification.
If you change that to the last 3-17 years VTI will win by a good 1% annually. The answer should be fairly obvious from this example: it depends on your start/end date and not everyone will capture a premium if they suddenly decide to change from total market to a tilt. Even waiting a long time may not help because it depends quite a bit on your start date. A good example is that someone who started investing 20-25 years ago and contributed periodically will still be behind with SCV even in the time periods where a fixed amount of SCV outperformed by 1%+ annually --> they didn't have enough invested to even capture the premium when it did exist.

I would advocate for more realistic data sets that include accumulation and decumulation. I think the chance of realizing a SCV premium are less than most people think, even when it does exist. It can come down to the luck of what overperforms when your assets are significant rather than what outperformed over a longer period of time.

I suggest modelling a 1$ portfolio with $1 monthly contributions for the last 20 years and seeing for yourself that SCV loses despite having the better 20 year performance. Anyone that believes they will get the SCV premium should do this.
Again, size is only one factor. Yes, it may under-perform for several years, but the odds of all 5 factors under-performing for an extended period are low. That's why it's important to include all of them.
I would be happy to respond to a post of performance with all 5 factors if/when you provide a similar PV link with all 5 factors.

I was responding to your post which only included SCV. Your choice of combination of small and value DFSVX unperformed for the last 17 years, and underperforms with periodic additions for more than the past 25 years. In the context of what actually happens I found the choice of words about a 20 year premium misleading even though a fixed amount of DFSVX does outperform for an exactly 20 year period.

My comments have nothing to do with several years of underperformance - I started in the years (early 2000's) you suggested with SCV overperformance and looked at what happens when you accumulate through them. Counting the number of years we could have started accumulating SCV and captured a premium, I believe it may be 0 of the last 25 years. My conclusion would be that capturing any SCV premium is more difficult than your comments suggest. I will also add a new comment that volatility is harmful during decumulation and extracting a premium there could also be more difficult than is suggested by returns without decumulation.

I have no comments about whether one should invest in SCV or even 5 factors at once. Those are personal choices and preferences.
Last edited by abc132 on Thu Jul 07, 2022 10:42 am, edited 1 time in total.
Marseille07
Posts: 16054
Joined: Fri Nov 06, 2020 12:41 pm

Re: Has SCV underperformed TSM since 1984?

Post by Marseille07 »

burritoLover wrote: Thu Jul 07, 2022 10:29 am Not sure what you mean. This is a 90-year chart indicating that you have to suffer long periods of underperformance vs. the S&P 500 (or some estimation of the S&P 500 in earlier periods) to stick with small-cap value. Not sure why that is non-sense.
SCV has averaged something like 12%/year. S&P, 10%/year. With this information, you can always pencil in a chart like this.

It's kind of like tracking the position of a racer in Formula 1 after the race is over. With knowledge that this racer won the race, the graph would always show them coming out on top.
User avatar
burritoLover
Posts: 4097
Joined: Sun Jul 05, 2020 12:13 pm

Re: Has SCV underperformed TSM since 1984?

Post by burritoLover »

Marseille07 wrote: Thu Jul 07, 2022 10:40 am
burritoLover wrote: Thu Jul 07, 2022 10:29 am Not sure what you mean. This is a 90-year chart indicating that you have to suffer long periods of underperformance vs. the S&P 500 (or some estimation of the S&P 500 in earlier periods) to stick with small-cap value. Not sure why that is non-sense.
SCV has averaged something like 12%/year. S&P, 10%/year. With this information, you can always pencil in a chart like this.

It's kind of like tracking the position of a racer in Formula 1 after the race is over. With knowledge that this racer won the race, the graph would always show them coming out on top.
This is a graph of SCV / S&P 500 performance - if anything it shows how painful holding SCV can be - do you see that 19.5 year period where SCV underperformed? A graph of SCV and S&P 500 separately would look world's better but that is not what this is.
Marseille07
Posts: 16054
Joined: Fri Nov 06, 2020 12:41 pm

Re: Has SCV underperformed TSM since 1984?

Post by Marseille07 »

burritoLover wrote: Thu Jul 07, 2022 10:45 am This is a graph of SCV / S&P 500 performance - if anything it shows how painful holding SCV can be - do you see that 19.5 year period where SCV underperformed? A graph of SCV and S&P 500 separately would look world's better but that is not what this is.
That's fair, but my point is that if you use hindsight, then you can draw a chart like that which appears like SCV eventually comes out on top.

Basically the chart is misleading because it looks as though SCV is sure to eventually come out on top, when there's no such guarantee. This thread is all about the OP potentially discovering 38 years of underperformance, hidden away from Merriman's chart.
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Has SCV underperformed TSM since 1984?

Post by YRT70 »

burritoLover wrote: Thu Jul 07, 2022 10:45 am This is a graph of SCV / S&P 500 performance - if anything it shows how painful holding SCV can be - do you see that 19.5 year period where SCV underperformed? A graph of SCV and S&P 500 separately would look world's better but that is not what this is.
Was it really that painful? I don't have access to the oldest data but If I look at the 17 years of underperformance from 1984 - 2001 TSM returned 15.5% while SCV returned 14.5% (IFA data). I would not find that painful but that is subjective. The decade after when TSM returned 0% would have been painful though if I held TSM.
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Has SCV underperformed TSM since 1984?

Post by YRT70 »

Marseille07 wrote: Thu Jul 07, 2022 10:55 am
burritoLover wrote: Thu Jul 07, 2022 10:45 am This is a graph of SCV / S&P 500 performance - if anything it shows how painful holding SCV can be - do you see that 19.5 year period where SCV underperformed? A graph of SCV and S&P 500 separately would look world's better but that is not what this is.
That's fair, but my point is that if you use hindsight, then you can draw a chart like that which appears like SCV eventually comes out on top.

Basically the chart is misleading because it looks as though SCV is sure to eventually come out on top, when there's no such guarantee. This thread is all about the OP potentially discovering 38 years of underperformance, hidden away from Merriman's chart.
To me the chart only looks at the past and makes no future predictions. The future might be different, true.

PS. I don't think there was actually 38 years of underperformance. See Portfoliovisualizer or IFA data here: posting.php?mode=quote&p=6763474
Marseille07
Posts: 16054
Joined: Fri Nov 06, 2020 12:41 pm

Re: Has SCV underperformed TSM since 1984?

Post by Marseille07 »

YRT70 wrote: Thu Jul 07, 2022 11:13 am To me the chart only looks at the past and makes no future predictions. The future might be different, true.

PS. I don't think there was actually 38 years of underperformance. See Portfoliovisualizer or IFA data here: posting.php?mode=quote&p=6763474
Right, I've seen that PV before, maybe something doesn't line up between PV and Simba's backtest, which the OP looked at.
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: Has SCV underperformed TSM since 1984?

Post by willthrill81 »

Marseille07 wrote: Thu Jul 07, 2022 11:18 am
YRT70 wrote: Thu Jul 07, 2022 11:13 am To me the chart only looks at the past and makes no future predictions. The future might be different, true.

PS. I don't think there was actually 38 years of underperformance. See Portfoliovisualizer or IFA data here: posting.php?mode=quote&p=6763474
Right, I've seen that PV before, maybe something doesn't line up between PV and Simba's backtest, which the OP looked at.
PV uses the Fama-French dataset for SCV until 1998, where it then uses Vanguard's VISVX fund.

I'm looking at the data sources for the Simba backtesting spreadsheet, but it doesn't seem to specify where the SCV data came from.
The Sensible Steward
Topic Author
skierincolorado
Posts: 2377
Joined: Sat Mar 21, 2020 10:56 am

Re: Has SCV underperformed TSM since 1984?

Post by skierincolorado »

burritoLover wrote: Thu Jul 07, 2022 7:22 am SCV can underperform the market for long periods 10-20 years at times. If your start/end dates in your backtest include a large portion of bad period(s), there doesn't appear to be a premium (or even negative) or it is greatly reduced. SCV is riskier than the market - if it were to deliver the premium consistently every 10 year period, then it wouldn't be as risky and likely wouldn't have the same premium to begin with. What will happen to the premium going forward - no one knows.

Image
I've always understood there can be long periods of underperfmance. That's necessarily the case given the variance of assets is very high even when one has a higher long term return.

But it's not just 20 years. It's 38, at least according to simba spreadsheet methodology. That raises two problems for me. One is that it is longer than most peoples investment horizon. The other is that it is such a long period it becomes harder to chaulk it up to variance. Statistically, the longer a period of underperformance, the less likely it is due to variance. With high variance, 10 or 20 year periods of underperformance are somewhat likely to occur occasionally. But a 38 year period is much less likely. Statistically it raises the probability that something fundamental has changed and the premium no longer exists.

All that said I still own some scv but am more cautious than I was before. I haven't sold any but won't increase in the future. I think there is still some theoretical and empirical support for scv premium and 38 years is not enough to totally bury the theory. But personally I would not hold more than 20% scv. There is also still the diversification benefit.
User avatar
burritoLover
Posts: 4097
Joined: Sun Jul 05, 2020 12:13 pm

Re: Has SCV underperformed TSM since 1984?

Post by burritoLover »

willthrill81 wrote: Thu Jul 07, 2022 12:27 pm
Marseille07 wrote: Thu Jul 07, 2022 11:18 am
YRT70 wrote: Thu Jul 07, 2022 11:13 am To me the chart only looks at the past and makes no future predictions. The future might be different, true.

PS. I don't think there was actually 38 years of underperformance. See Portfoliovisualizer or IFA data here: posting.php?mode=quote&p=6763474
Right, I've seen that PV before, maybe something doesn't line up between PV and Simba's backtest, which the OP looked at.
PV uses the Fama-French dataset for SCV until 1998, where it then uses Vanguard's VISVX fund.

I'm looking at the data sources for the Simba backtesting spreadsheet, but it doesn't seem to specify where the SCV data came from.
So does that mean portfolio visualizer is using long-short Fama-French returns for small value before 1998?
abc132
Posts: 2435
Joined: Thu Oct 18, 2018 1:11 am

Re: Has SCV underperformed TSM since 1984?

Post by abc132 »

YRT70 wrote: Thu Jul 07, 2022 11:08 am
burritoLover wrote: Thu Jul 07, 2022 10:45 am This is a graph of SCV / S&P 500 performance - if anything it shows how painful holding SCV can be - do you see that 19.5 year period where SCV underperformed? A graph of SCV and S&P 500 separately would look world's better but that is not what this is.
Was it really that painful? I don't have access to the oldest data but If I look at the 17 years of underperformance from 1984 - 2001 TSM returned 15.5% while SCV returned 14.5% (IFA data). I would not find that painful but that is subjective. The decade after when TSM returned 0% would have been painful though if I held TSM.
I think the idea would be that if you picked SCV in hopes of +1% delta annually you certainly would not have picked it for -1% annually. If an expected +1% difference was enough to include it, then a -1% difference is enough to not have wanted to include it. You lost by a margin as big as what you hoped to gain.

Those that don't care about 1% annually can safely ignore any SCV vs market threads. Lots of opinions but either is likely to succeed.
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: Has SCV underperformed TSM since 1984?

Post by willthrill81 »

burritoLover wrote: Thu Jul 07, 2022 3:41 pm
willthrill81 wrote: Thu Jul 07, 2022 12:27 pm
Marseille07 wrote: Thu Jul 07, 2022 11:18 am
YRT70 wrote: Thu Jul 07, 2022 11:13 am To me the chart only looks at the past and makes no future predictions. The future might be different, true.

PS. I don't think there was actually 38 years of underperformance. See Portfoliovisualizer or IFA data here: posting.php?mode=quote&p=6763474
Right, I've seen that PV before, maybe something doesn't line up between PV and Simba's backtest, which the OP looked at.
PV uses the Fama-French dataset for SCV until 1998, where it then uses Vanguard's VISVX fund.

I'm looking at the data sources for the Simba backtesting spreadsheet, but it doesn't seem to specify where the SCV data came from.
So does that mean portfolio visualizer is using long-short Fama-French returns for small value before 1998?
The FAQ at PV says this in a note about asset class returns.
The "100 Portfolios Formed on Size and Book-to-Market" data from Professor Kenneth French's research data library is used to derive synthetic index data for US market equities. The methodology used is identical to the one discussed in this online article at Portfolio Charts.
Also, I must correct myself in that PV uses VISVX from 1999 and forward, not 1998.
The Sensible Steward
Marseille07
Posts: 16054
Joined: Fri Nov 06, 2020 12:41 pm

Re: Has SCV underperformed TSM since 1984?

Post by Marseille07 »

abc132 wrote: Thu Jul 07, 2022 4:02 pm Those that don't care about 1% annually can safely ignore any SCV vs market threads. Lots of opinions but either is likely to succeed.
Isn't this pretty much everyone? If one really cares about 1% then holding 40% in bonds would be a ridiculous idea.
User avatar
burritoLover
Posts: 4097
Joined: Sun Jul 05, 2020 12:13 pm

Re: Has SCV underperformed TSM since 1984?

Post by burritoLover »

willthrill81 wrote: Thu Jul 07, 2022 4:29 pm
burritoLover wrote: Thu Jul 07, 2022 3:41 pm
willthrill81 wrote: Thu Jul 07, 2022 12:27 pm
Marseille07 wrote: Thu Jul 07, 2022 11:18 am
YRT70 wrote: Thu Jul 07, 2022 11:13 am To me the chart only looks at the past and makes no future predictions. The future might be different, true.

PS. I don't think there was actually 38 years of underperformance. See Portfoliovisualizer or IFA data here: posting.php?mode=quote&p=6763474
Right, I've seen that PV before, maybe something doesn't line up between PV and Simba's backtest, which the OP looked at.
PV uses the Fama-French dataset for SCV until 1998, where it then uses Vanguard's VISVX fund.

I'm looking at the data sources for the Simba backtesting spreadsheet, but it doesn't seem to specify where the SCV data came from.
So does that mean portfolio visualizer is using long-short Fama-French returns for small value before 1998?
The FAQ at PV says this in a note about asset class returns.
The "100 Portfolios Formed on Size and Book-to-Market" data from Professor Kenneth French's research data library is used to derive synthetic index data for US market equities. The methodology used is identical to the one discussed in this online article at Portfolio Charts.
Interesting - thanks.
abc132
Posts: 2435
Joined: Thu Oct 18, 2018 1:11 am

Re: Has SCV underperformed TSM since 1984?

Post by abc132 »

Marseille07 wrote: Thu Jul 07, 2022 4:32 pm
abc132 wrote: Thu Jul 07, 2022 4:02 pm Those that don't care about 1% annually can safely ignore any SCV vs market threads. Lots of opinions but either is likely to succeed.
Isn't this pretty much everyone? If one really cares about 1% then holding 40% in bonds would be a ridiculous idea.
No, because holding bonds is not a choice about performance gains. Equating stocks and bonds and choosing in terms of performance - I hope nobody invests this way.

Cutting coupons to save $100 and instead being charge an extra $100 is a more meaningful example.
Marseille07
Posts: 16054
Joined: Fri Nov 06, 2020 12:41 pm

Re: Has SCV underperformed TSM since 1984?

Post by Marseille07 »

abc132 wrote: Thu Jul 07, 2022 4:58 pm No, because holding bonds is not a choice about performance gains. Equating stocks and bonds and choosing in terms of performance - I hope nobody invests this way.

Cutting coupons to save $100 and instead being charge an extra $100 is a more meaningful example.
That's fair, although chasing the extra +1% becomes +0.6% when your AA is 60/40, assuming you gain the +0.6% at all (which you might not). I'm really not sure if it is worth it at that point.
User avatar
nisiprius
Advisory Board
Posts: 52211
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: Has SCV underperformed TSM since 1984?

Post by nisiprius »

To me, the most salient point in this discussion is that due to issues of definition and data, nobody is really sure of the answer to the question "Has SCV underperformed TSM since 1984?" The data is too fragile to make an unequivocal statement.
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.
abc132
Posts: 2435
Joined: Thu Oct 18, 2018 1:11 am

Re: Has SCV underperformed TSM since 1984?

Post by abc132 »

Marseille07 wrote: Thu Jul 07, 2022 5:01 pm That's fair, although chasing the extra +1% becomes +0.6% when your AA is 60/40, assuming you gain the +0.6% at all (which you might not). I'm really not sure if it is worth it at that point.
I would go for it if I was sure the premium would show up and that I would be able to extract any premium. I'm happy when my fixed income gains a +1% delta and that is only 30% of my portfolio (+0.3% annual returns). I haven't found the SCV arguments convincing enough personally to make a big move although it seems like a reasonable enough investment choice in that it is unlikely to impact financial success or failure. I have a little bit of SCV that won't move the needle regardless of what happens to SCV.

I do feel that the inability to define what is SCV and what is has done is likely to lead to investing errors. There is a whole language built up around behavioral errors, "you have to invest in international SCV", or "you have to screen for xxx". The first is just what backtested better and the second leaves an infinite amount of filtering to make it impossible for SCV performance to ever be reasonably evaluated. Constructing an argument that can't lose regardless of future performance is a behavioral trap. If you tack on high enough fees to the above statements you do put some risk into financial success.
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: Has SCV underperformed TSM since 1984?

Post by willthrill81 »

I always find it interesting in these threads where so many question the premia for small and value companies but have no trouble with the premium for market beta, which has also been flat or negative for decades at a time.
The Sensible Steward
Northern Flicker
Posts: 15363
Joined: Fri Apr 10, 2015 12:29 am

Re: Has SCV underperformed TSM since 1984?

Post by Northern Flicker »

garlandwhizzer wrote: TSM in reality, as opposed to factor theory, is a multi-factor fund whose relative factors are defined and whose relative factor weights are set by the aggregate judgement of market participants.
This is conceptually incorrect. All factors other than the market factor are diversified away by the market portfolio, so it lacks multifactor exposure.

If you try to say that SCV is part of TSM and therefore TSM includes exposure to the size and value factors, you will double count the size and value factors when you model the return of a SCV stock.

In the 3 factor model, the return of a stock is a linear combination of exposures to the market, size, and value factors and the stocks residual alpha. If there were day value factor exposures in the market portfolio, the value factor would be counted twice, once in the value factor term and again in the market factor term.

Factors other than the market factor are modeled as offsets from the market return. The market portfolio has no such offset from the market return.
garlandwhizzer wrote: The models are unrealistic and and make assumptions that have never existed in the real world like cost-free, fiction-free long/short.
This is also incorrect. Factor models are defined as a linear model of components of return. Long-short portfolios are used to try to measure factor premia, and as a basis for some alt investments that are available, but they are not necessary to harvest whatever premia for size, value, or quality factors are delivered. You need the long-short portfolio to harvest only a factor premium without also harvesting the market factor. This is a dubious goal anyway, so long-short portfolios are unnecessary.
User avatar
PicassoSparks
Posts: 417
Joined: Tue Apr 28, 2020 5:41 am

Re: Has SCV underperformed TSM since 1984?

Post by PicassoSparks »

Thank you for this discussion.

As a person not sophisticated enough in stats to be able to understand the factor analysis, here’s what it has felt like,

1: As the value factor underperformed for a longer and longer time, we got a lot of articles from researchers finding new ways to define the value factor in order to find it again. “We’re certain the factor exists, we must be looking for it wrong.” This made me suspicious.

2: The academic literature has thrown off a zoo of factors. This makes me more suspicious.

3: Risk factors seem to indicate that it’s impossible for markets to learn (in some ways) because there is the stable set of excess returns that can’t be arbitraged away. It seems to be that assuming stationarity in a complex system that has seen considerable regulatory transformation over time is a very strange assumption. So much suspicion!
Apathizer
Posts: 2507
Joined: Sun Sep 26, 2021 2:56 pm

Re: Has SCV underperformed TSM since 1984?

Post by Apathizer »

willthrill81 wrote: Thu Jul 07, 2022 10:40 pm I always find it interesting in these threads where so many question the premia for small and value companies but have no trouble with the premium for market beta, which has also been flat or negative for decades at a time.
Exactly. Ben Felix has discussed and written about this. The market factor was determined by the same research that found other significant factors. Specifically value, size, profitability, and reinvestment.

It seems like psychological or ideological bias to only recognize the market factor while dismissing the other four significant factors. Again, the existence of all five factors was determined by the same research so recognizing one and dismissing the others doesn't make sense.
ROTH: 50% AVGE, 10% DFAX, 40% BNDW. Taxable: 50% BNDW, 40% AVGE, 10% DFAX.
Post Reply