Larry Swedroe: Rediscovering The Size Effect

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Random Walker
Posts: 2870
Joined: Fri Feb 23, 2007 8:21 pm

Larry Swedroe: Rediscovering The Size Effect

Post by Random Walker » Fri Jun 15, 2018 12:29 am

http://www.etf.com/sections/index-inves ... nopaging=1

I didn’t realize that the size premium had been on such shaky ground. But looks like it has been saved by accounting for other factors: profitability, quality, defensive. In the presence of these factors, the size effect is made much stronger. Larry reviews a couple of papers from the AQR people on the subject.
The Q Factor model (market, size, profitability, investment) explains almost all the anomalies of the Fama-French 4 Factor model (market, size, value, momentum) with the exception of small growth stocks with low profitability. The size premium is diluted by junk. There is likely a behavioral explanation for this: investor preference for small stocks with lottery ticket like characteristics. The performance of a passive small cap fund can depend tremendously on how it defines it's universe of small cap stocks. Avoiding small growth stocks with high investment and low profitability, IPOs, and stocks in bankruptcy, rather than blindly following a popular public index has been profitable.

Dave

User avatar
JoMoney
Posts: 5367
Joined: Tue Jul 23, 2013 5:31 am

Re: Larry Swedroe: Rediscovering The Size Effect

Post by JoMoney » Fri Jun 15, 2018 3:36 am

Interesting. The S&P indices also have a bit of light touch screening with regard to profitability, and seasoning requirements before an IPO is eligible for inclusion in the index. There are studies that point to S&P's indices and that the 'Value' factor doesn't show up.
Also, when looking at the available history, the 'Size' factor looks at bit odd when seeing that larger Mid-Caps outperformed the Small-Caps.
Seems to me this idea of chasing 'factors' is less about "risk premiums" and more about chasing past returns (that are likely to be self defeating when investors actually implement)
Image
"To achieve satisfactory investment results is easier than most people realize; to achieve superior results is harder than it looks." - Benjamin Graham

User avatar
nisiprius
Advisory Board
Posts: 36179
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: Larry Swedroe: Rediscovering The Size Effect

Post by nisiprius » Fri Jun 15, 2018 7:30 am

Point #1: Resetting the clock

Notice that Larry Swedroe puts the word "saved" within quotation marks:
However, they do “save” the size effect by demonstrating it is made much stronger (and implementation costs are reduced) when size is combined with the newer common factors of profitability, quality and defensive (low beta)—the return premium is greater for other factors in small stocks.
But the thing is, you shouldn't get to do this kind of "reset" by saying that something that has failed in its original form has been "saved" (or "rediscovered" or "resurrected") if you redefine it into something different, using 20/20 hindsight and backtesting on combinations, adjustments, modifications, and variations of the original premise. Or rather, you can, if you like, but you need to reset the clock on your testing again. You haven't "resurrected" anything, you are growing something new on the decaying remains of the old.

The size effect in its original form was just a size effect. It was based on data from 1926-1975 inclusive, i.e. fifty years of backtesting. Investment gurus at the time said that it was robust; for example, in 1983 Robert Arnott said "A growing body of evidence exists that small-capitalization stocks significantly outperform large-capitalization stocks. This effect is so strong and so consistent that even advocates of the Efficient Market Hypothesis have found no refutation of this effect." Dimensional Fund Advisors offered its first fund on the strength of it: DFSCX, now called the "DFA US Micro Cap Portfolio."

Yet Larry Swedroe notes:
Unfortunately, the size premium basically disappeared in the U.S. after the publication of Banz’s work. Using data from Dimensional Fund Advisors, from 1984 through 2017, the annual size premium in U.S. stocks was just 0.3% on an annual average basis and a negative 0.1% on an annualized basis.
It has since been found that Banz's original work suffered from survivorship bias in the data he used.

The effect in its original form, based on 50 years of backtesting, looked good and was said to be strong and consistent. 36 years later, it is so sick that it needs to be "saved." But the only way to "save" it is, in fact, to create something new, with fresh 20/20 hindsight backtesting. Certainly, you can try out various combinations of the small factor with other things to see if it potentiates or synergizes with them. But once again you have 20/20 hindsight and bias. What is the starting date for the claimed "resurrection" (Asness' choice of words)? Possibly 2015, the publication date of Size Matters, If You Control Your Junk? But it's not in any sense a resurrection of the size factor in its original form, it's a new thing which (for some reason) they choose to describe by an old name. In any case, the resurrected size factor, which Asness describes as "spectacular," is now about three years old.

Well, Banz' paper was published in 1981, DFSCX had its inception on 12/23/1981, Arnott was describing the effect as "strong and consistent" in 1983, and DFSCX did very well for just about three years... before fading and underperforming the S&P 500 for the next seventeen.

It did, however, finally make up the lost ground--but not after adjusting for risk, which is the topic of my next posting.
Last edited by nisiprius on Fri Jun 15, 2018 7:46 am, edited 3 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
nisiprius
Advisory Board
Posts: 36179
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: Larry Swedroe: Rediscovering The Size Effect

Post by nisiprius » Fri Jun 15, 2018 7:36 am

Point #2: Adjusting for risk

Discussions of the size effect always get slippery if we are not clear on whether we just mean "return" or whether we mean "risk adjusted return." Recall that the original statement for the size effect in its original form, in Rolf Banz words, was "It is found that smaller firms have had higher risk adjusted returns."

Larry Swedroe, while noting that "the size premium basically disappeared in the U.S. after the publication of Banz’s work," notes that DFSCX, the DFA US Micro Cap Portfolio Fund, did outperform the S&P 500--presumably in part because of DFSCX investing in "micro caps," i.e. smaller-cap stocks than the Fama-French "small" universe.

The problem here is the usual problem that occurs in all such discussions: do we choose to take any account of the extra risk involved in small caps? If we don't, then, sure, there is a fairly robust "premium."

But the extra risk is not trivial. Over the lifetime of DFSCX (whole years 1982-2017 inclusive) it had a standard deviation of 20.02%, compared to 16.30% for the S&P 500. The result is that DFSCX had a lifetime Sharpe ratio of 0.50, compared to 0.56 for the S&P 500 (0.55 for the real world of VFINX, the S&P 500 index fund). The Sharpe ratio was lower.*

Why does this matter when we eat return, not Sharpe ratio? Because it means that (in theory) you would be better off leveraging the S&P 500 than using DFSCX. In real life, and assuming you do not want lower risk than 100% stocks, you might have been better off increasing your broad-stocks (S&P 500 or total market) than by allocating something to DFSCX. The correlation of DFSCX and VFINX is imperfect, 0.71, so there is a diversification effect but it is not strong. The effect in an actual portfolio with other assets would need to be tested specifically.

But this does mean that the supposed superiority of DFSCX over the S&P 500 or the total market fades into the land of "maybe" if we take risk into account. The original statement for the size effect in its original form was about "risk adjusted return." If we don't have higher risk adjusted return, then we are left with yet another "correlation" story.


*You can see similar results in PortfolioVisualizer here. PortfolioVisualizer only goes back to 1985. I didn't want to restrict the comparison to 1985 and beyond, because that would be unfavorable to DFSCX and its first few excellent years. I used PortfolioVisualizer's CASHX asset to calculate the Sharpe ratio. I validated my calculation procedures by applying them to PortfolioVisualizer's monthly data and got exactly the same numbers they get. I thought it was important to go back to 1982 because DFSCX performed well during the first few years and omitting the earliest data is unfavorable to DFSCX. I only used annual data, for the whole years 1982-2017 inclusive. A quick test comparing annual and monthly data over the time period showed little difference in results.
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.

Random Walker
Posts: 2870
Joined: Fri Feb 23, 2007 8:21 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by Random Walker » Fri Jun 15, 2018 9:29 am

Obviously returns are important. And they get discussed here lots. But correlations are important as well, and how asset classes, factors, styles mix in a portfolio is very important. What matters is the portfolio as a whole. I’m gaining an appreciation for the Sharpe ratio of an investment in isolation, but I think what an investment does to the Sharpe ratio of a portfolio is what really counts. I recall the correlations are about as below:
Size to market 0.4
Value to market 0.1
Size to value 0.1
Also, I believe the effect of other factors are more pronounced in the small realm.

Dave

User avatar
nisiprius
Advisory Board
Posts: 36179
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: Larry Swedroe: Rediscovering The Size Effect

Post by nisiprius » Fri Jun 15, 2018 10:07 am

Here's the thing about correlations and "the portfolio as a whole." It is a robust and compelling effect when we talk about diversifying stocks by adding bonds.

It is tenuous and evasive for subclasses of stocks mixed with other subclasses of stocks, and for subclasses of bonds mixed with other subclasses of bonds. By "tenuous and evasive" I mean that the effect is small, that it is very sensitive to choices of endpoints, and it is sensitive to what particular index or mutual fund is used to represent the class. Hence the endless debates about "but that's only true if you use the Russell 2000" etc.

If there is no superior Sharpe ratio for small-caps, then the argument for including them must be that that they have low enough correlations with large-caps to improve the Sharpe ratio of the portfolio as a whole.

If we look at the SBBI data from 1926 through 2017, inclusive, this is what we see:

1) The Sharpe ratio was 0.437 for large company stocks, 0.414 for small company stocks. So small company stocks did not have superior risk-adjusted return.

2) The correlation between large company stocks and small company stocks was 0.793. That was not perfect correlation, but it was hardly low correlation, and doesn't suggest any powerful diversification effect. In contrast, the correlations of the three different SBBI bond series with stocks were 0.149, -0.002, and -0.033.

Now, for what happens in the portfolio as a whole. For our baseline portfolio we will use 60% large company stocks, 40% intermediate term government bonds. The Sharpe ratio of the 60/40 portfolio is 0.494, which exceeds the Sharpe ratio of either of its two constituents (0.437 for stocks, 0.328 or bonds). That is the low correlations between stocks and bonds produced the hoped-for MPT diversification effect. In fact, adding bonds added 0.062 to the Sharpe ratio even though the bonds alone had a lower Sharpe ratio than stocks. In other words, exactly the things people talk about.

So the magic happens when we diversify stocks by adding bonds. But will it happen when we diversify stocks by adding a different subclass of stocks? Will adding small caps to the portfolio--so that we have both large caps and small caps, not just large caps--improve the portfolio further, and, if so, by how much?

3) The answer is that a portfolio of 30% large company stocks, 30% small company stocks, and 40% intermediate-term government bonds had a Sharpe ratio of 0.496. So, yes, the Sharpe ratio was improved from 0.494 without small-caps to 0.496 with small caps, an increase of 0.002.

That is a very small difference, and it would almost certainly change depending on choices of endpoints or exact choices of asset classes.

The diversification benefit--the extra Sharpe ratio in the portfolio as a whole--for adding bonds to stocks was thirty times larger than the incremental effect of further diversification by adding small-cap stocks to large-cap stocks.
Last edited by nisiprius on Fri Jun 15, 2018 10:27 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
nisiprius
Advisory Board
Posts: 36179
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: Larry Swedroe: Rediscovering The Size Effect

Post by nisiprius » Fri Jun 15, 2018 10:13 am

Larry Swedroe has sent me three PMs, which he has given me permission to quote. I will quote them verbatim with slight trimming, and without further comment.
First, the extra risk is the compensation for the premium, eliminate the risk adjustment and there is no market beta premium either, so that is bogus argument. What's also important is that it's a different risk, adding diversification benefits.

Second, the "saving" isn't backtesting, DFA "saved" the size premium 35 years ago by screening out such junk long ago as this has been well documented in the literature with the high beta anomaly being known in the 1970s
Why does this matter when we eat return, not Sharpe ratio? Because it means that (in theory) you would be better off leveraging the S&P 500 than using DFSCX

In real world leverage has high costs and that means your idea doesn't work
In addition it would dramatically increase tail risks...
Simplest story for small is liquidity costs is risk and should be compensated, then you have whole bunch of other macro ideas, such as fewer sources of capital and first to be cut off from capital in crises.

Here's simple example in my industry RIAs. Have 50mm in AUM sell about 4x EBITDA (all else equal, meaning margins, age of clients, size of clients), but at about $100mm it becomes say 5x and at $350mm maybe 6x and at $1b 7x, and so on. That's simple cost of capital story.
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
vineviz
Posts: 985
Joined: Tue May 15, 2018 1:55 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by vineviz » Fri Jun 15, 2018 1:53 pm

nisiprius wrote:
Fri Jun 15, 2018 10:07 am
The answer is that a portfolio of 30% large company stocks, 30% small company stocks, and 40% intermediate-term government bonds had a Sharpe ratio of 0.496. So, yes, the Sharpe ratio was improved from 0.494 without small-caps to 0.496 with small caps, an increase of 0.002.

That is a very small difference, and it would almost certainly change depending on choices of endpoints or exact choices of asset classes.

The diversification benefit--the extra Sharpe ratio in the portfolio as a whole--for adding bonds to stocks was thirty times larger than the incremental effect of further diversification by adding small-cap stocks to large-cap stocks.
That change from 60/0/40 (large/small/bonds) to 30/30/40 improved the Sharpe ratio by increasing returns MORE than enough to compensate for the any extra portfolio volatility they might have generated.

But remember that simply replacing some of the large cap stocks with small cap stocks isn't our only option. If, as you insisted we should, we care mostly about risk-adjusted portfolio returns then have more decisions to make.

If we increase our bond holdings to 50%, to help offset the portfolio volatility induced by including, small caps then we can improve our overall risk-adjusted return even further.

I'm perfectly happy to concede that other portfolio choices can have a bigger marginal impact on the portfolio than replacing some large caps with small caps. But happily we don't have to choose to do just one or the other.

Plus even if the marginal benefit is small we can be motivated by the fact that marginal cost of including small caps is even smaller, especially today. Excellent small cap index funds like (VIOO) Vanguard S&P Small-Cap 600 ETF and (IJS) iShares S&P Small-Cap 600 Value ETF are available with annual expense ratios of 15bps or less. Even during the worst 15-year period of relative performance for small caps vs large caps (1984 to 1998) in history, small cap stocks had a CAGR of over 11%. In other words, maintaining a reasonable allocation to small cap stocks even during their WORST years of relative performance is unlikely to devastating to an investor's portfolio.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

Random Walker
Posts: 2870
Joined: Fri Feb 23, 2007 8:21 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by Random Walker » Fri Jun 15, 2018 2:32 pm

I went back and took a look at the appendix on size in Larry’s factor book. I was reminded that there is a bit of a definitional problem potentially clouding the issue. All of the factors are defined as long-short portfolios. Generally all the stocks in a given universe are ranked according to the metric, then the premium is defined as the premium for the top 30% minus the premium for the bottom 30%. But the FF definition of size is top half minus bottom half. This dilutes the effect. If one uses the same 30/30 definition used for value instead of the 50/50 definition, the effect is bigger. If one moves to 20/20 or 10/10, it’s that much more pronounced.

Dave

snarlyjack
Posts: 780
Joined: Fri Aug 28, 2015 12:44 pm
Location: Montana

Re: Larry Swedroe: Rediscovering The Size Effect

Post by snarlyjack » Fri Jun 15, 2018 2:39 pm

How about a real life example.
(you guy's are kind of talking above my head...)

I read all of Paul Merrimans articles & agree
with his statement that a 1/2% or 1% can change
your life.

With that in mind I changed my Roth IRA to 100%
Vanguard Small Cap Value Fund (VSIAX). My broker
account is 100% Vanguard High Dividend Yield Index Fund
(VHDYX). I can't touch the Roth account for 36 more years.

The Small Cap Value Fund is about 60% Mid Caps & 40% Small Caps.
The High Dividend Yield Fund is 100% Large Cap. Value Fund.
So, Large, Medium & Small Cap value stocks. 100% stocks, no bonds.

How does the numbers look for this portfolio? Any comments
would be appreciated...long term buy & hold with Vanguard Funds.

User avatar
vineviz
Posts: 985
Joined: Tue May 15, 2018 1:55 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by vineviz » Fri Jun 15, 2018 2:49 pm

snarlyjack wrote:
Fri Jun 15, 2018 2:39 pm
How does the numbers look for this portfolio? Any comments
would be appreciated...long term buy & hold with Vanguard Funds.
What percent of the combined portfolio is each investment (VBR and VHDYX)?

Is it 50%/50% or something else?
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

snarlyjack
Posts: 780
Joined: Fri Aug 28, 2015 12:44 pm
Location: Montana

Re: Larry Swedroe: Rediscovering The Size Effect

Post by snarlyjack » Fri Jun 15, 2018 3:03 pm

Vineviz,

Right now the % is really small about 10%
small caps. That's because the brokerage account
is so large.


It's about $300,000. High Dividend Yield &
$30,000. Small Cap Value. What I would like to do
is not make any more contributions to the High Dividend
Yield & concentrate 100% on Small Cap Value. It might
take another 5-6 years to build up the Small Cap Value
to a meaningful amount. Adding to both the Roth SCV & SCV
in the brokerage account.

Thanks for your help!

jalbert
Posts: 3423
Joined: Fri Apr 10, 2015 12:29 am

Re: Larry Swedroe: Rediscovering The Size Effect

Post by jalbert » Fri Jun 15, 2018 3:03 pm

It is always a good idea to retain some skepticism about research results when the researcher takes the liberty of deciding which well understood experimental design principles they will follow and which they will abandon.
Index fund investor since 1987.

User avatar
vineviz
Posts: 985
Joined: Tue May 15, 2018 1:55 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by vineviz » Fri Jun 15, 2018 4:14 pm

snarlyjack wrote:
Fri Jun 15, 2018 3:03 pm
Right now the % is really small about 10% small caps. That's because the brokerage account is so large.
My overall opinion is that if you're young and are confident in your ability to weather a year in which your accounts are down 30% to 40% that a portfolio that is entirely in U.S stocks isn't crazy but it is unnecessarily risky. You are well-diversified among U.S. stocks, but not having any holdings of bonds or foreign stocks leaves the portfolio less diversified than it could (should) be.

I think adding to the small cap index fund is great, but I'd also think about moving some of the Vanguard High Dividend Yield Index to increase your diversification.

The two funds I'd consider to have the most diversification impact would be Vanguard Long-Term Treasury Index Fund Admiral Shares (VLGSX) as a bond holding and Vanguard Emerging Markets Stock Index Fund Admiral Shares (VEMAX) as an international holding. Neither are particularly well diversified within their respective asset class, but each would have a powerful diversification benefit when added to your VHDYX even with a relatively small investment.

Moving towards having just 15% of your portfolio in each of VSIAX, VEMAX, and VEMAX would be a lot healthier IMHO than the current allocation.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

snarlyjack
Posts: 780
Joined: Fri Aug 28, 2015 12:44 pm
Location: Montana

Re: Larry Swedroe: Rediscovering The Size Effect

Post by snarlyjack » Fri Jun 15, 2018 4:57 pm

Thank You Vineviz,

I think I' am to young for bonds. I' am comfortable with 100% stocks.
I was thinking of the Total International Stock Fund.

I keep learning. The Bogleheads have helped a lot.
It takes a long time to put together a well thought out portfolio.

Once again...Thank You, Vineviz.

stlutz
Posts: 4606
Joined: Fri Jan 02, 2009 1:08 am

Re: Larry Swedroe: Rediscovering The Size Effect

Post by stlutz » Fri Jun 15, 2018 6:23 pm

nisiprius: Back in this thread Small-cap tilt, 1925-2004: good enough to make you switch?, you concluded that the the small cap effect is there but that it is small. (That is my favorite BH thread ever, BTW). I'm curious if that is still your view or if you have evolved a bit and concluded the that small cap effect never was there and it was never spectacular?

snarlyjack
Posts: 780
Joined: Fri Aug 28, 2015 12:44 pm
Location: Montana

Re: Larry Swedroe: Rediscovering The Size Effect

Post by snarlyjack » Fri Jun 15, 2018 8:12 pm

Nisiprius:

I did read the blog "small-cap tilt, 1925-2004: good enough to make you switch?"
Which I wanted to comment on.

It was a very interesting challenge & story. However, going through what I went
through settling my Mom's estate. I would not suggest to anyone to have 10
different funds. It makes the estate settling to hard & complex.

My Mom used 3 different banks, plus a safety deposit box, plus assorted
cd's, plus all the other stuff (cars, house, etc.). I do disagree with Paul Merriman
on the use of 10 different funds. In my opinion it's just to much. 2 or 3 funds is ok,
4 maybe, 10 is just to many.

My whole thought was...what are you doing Mom...You need to simplify.

david1082b
Posts: 367
Joined: Fri Jun 09, 2017 12:35 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by david1082b » Fri Jun 15, 2018 10:05 pm

The article says DFCSX when I think it means DFSCX. DFCSX is a different DFA fund, a continental European fund apparently, that started in 1988.

User avatar
Robert T
Posts: 2521
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Larry Swedroe: Rediscovering The Size Effect

Post by Robert T » Sat Jun 16, 2018 12:58 pm

.
Using 1927-2017 data from Ken French’s website.

On the magnitude of small cap return differences

Annualized return / Growth in $1
..9.8% / $4,849 = Large caps (Largest 30%)
12.0% / $30,484 = Small Caps (Smallest 30%)

Small cap stocks outperformed large cap stocks. The average return difference was statistically significant, and the implied $1 growth in small caps was over 6 times greater than large caps over this period.

On risk-adjusted return

Annualized return / Standard deviation / Sharpe Ratio
..9.8% / 19.3 / 0.42 = Large caps (Largest 30%)
12.0% / 33.3 / 0.40 = Small caps (Smallest 30%)

A common interpretation of this type of result (including in the earlier AQR paper) is that even though small caps had higher return, there was no risk-adjusted return (Sharpe Ratio) benefit of small caps. i.e. small caps were simply the equivalent to leveraged large caps (‘beta’).

I would caution such a simplistic interpretation based on the realism of the underlying assumptions of the Sharpe Ratio:
  • Unrestricted risk-free borrowing. This clearly is not a realistic assumption.
    Normally distributed returns (often returns are skewed, not normally distributed). The Sharpe Ratio, through use of standard deviation, assumes that returns are normally distributed. However, the annual return distribution of small caps was positively skewed (larger right tail), while for large caps it was negatively skewed (larger left tail) – so even if you did have unrestricted risk-free borrowing, simply leveraging large caps would not result in the same return distribution as small caps (it would have a larger left tail).
Comparisons that somewhat address both of these issues are that following (these examples require no borrowing).

Let’s start with a 60:40 Small Cap:T-bill portfolio. From the Sharpe ratio above (0.40 and 0.42), one could achieve a similar result by simply owning large caps (in place of small caps), and less T-bills. The two portfolios below demonstrate this effect. They had very similar return, and as expected, had the same Sharpe ratios as in the above example. What is different is the max calendar year loss (due to differences in the skewness of returns).

Annualized return (%) / Standard Deviation / Sharpe Ratio / Max calendar year loss

..9.7 / 19.9 / 0.40 / -32.0 = 60% Small Caps (Smallest 30%):40% T-bills
..9.8 / 19.3 / 0.42 / -43.9 = 100% Large Caps (Largest 30%)

But very few people seem to hold t-bills for fixed income. If we start with a portfolio of 50:50 Small Caps:5-yr T-notes, we get a higher Sharpe Ratio and lower max calendar year loss.

Annualized return (%) / Standard Deviation / Sharpe Ratio / Max calendar year loss

..9.8 / 16.6 / 0.46 / -26.0 = 50% Small Caps (Smallest 30%):50% 5-yr T-notes
..9.8 / 19.3 / 0.42 / -43.9 = 100% Large Caps (Largest 30%)

Which would have been the preferable portfolio?

Most of the higher small cap returns came in the early part of the series (1927 to 1981), while in the latter part of the series (1982-2017) there has essentially been no return difference.

1927-1981: Annualized return
..8.5% = Large caps (Largest 30%)
12.2% = Small Caps (Smallest 30%)

1982-2018: Annualized return
11.8% = Large caps (Largest 30%)
11.6% = Small caps (Smallest 30%)

Some suggest the latter period is much more relevant as the earlier period was plagued by data (measurement) issues. I think much of the data issues have been cleared up in the Ken French data set, and I am less convinced that the earlier period is less relevant/irrelevant. Even in the latter period there seems to be two very distinct sub-periods, the first 18 years – 1982-1999 when large caps outperformed, and the second 18 years, 2000-2017 when small caps outperformed. The table below shows the annualized returns of small caps and large caps for each 18 year period from 1928.

The variation in the 18 year annualized returns was much wider for large caps (ranging from 18.5% to 5.4%), and narrower from small caps (ranging from 13.8% to 9.8%).

Which would you have preferred?

Annualized Returns (%)

…………………….Small Caps…….Large Caps

1928-1945………..10.1………….……4.7
1946-1963….…….11.8……………..13.2
1964-1981…………13.8…………….…6.5
1982-1999…………13.4……………..18.5
2000-2017………….9.8………………. 5.4

Max………………….13.8……………….18.5
Min…………………….9.8………………..5.4
Range…………………4.0……………….13.7

Obviously no guarantees.

Robert
.
Last edited by Robert T on Sat Jun 16, 2018 1:40 pm, edited 1 time in total.

snarlyjack
Posts: 780
Joined: Fri Aug 28, 2015 12:44 pm
Location: Montana

Re: Larry Swedroe: Rediscovering The Size Effect

Post by snarlyjack » Sat Jun 16, 2018 1:11 pm

Here is a short video by Kevin O'Leary talking about
why he thinks small caps are going up...interesting conversation.

Dated May 2018. Enjoy...

https://www.youtube.com/watch?v=DEfHnhq_HeM

Random Walker
Posts: 2870
Joined: Fri Feb 23, 2007 8:21 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by Random Walker » Sat Jun 16, 2018 1:53 pm

Robert’s data above serves as an excellent introduction to the benefit of diversifying across uncorrelated sources of return. The small cap / bonds portfolio had exposure to market beta, size, term. The large growth portfolio was effectively market beta alone. There is a significant positive effect on annualized return by avoiding big drawdowns.

Dave

User avatar
nisiprius
Advisory Board
Posts: 36179
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: Larry Swedroe: Rediscovering The Size Effect

Post by nisiprius » Sat Jun 16, 2018 3:49 pm

stlutz wrote:
Fri Jun 15, 2018 6:23 pm
nisiprius: Back in this thread Small-cap tilt, 1925-2004: good enough to make you switch?, you concluded that the the small cap effect is there but that it is small. (That is my favorite BH thread ever, BTW). I'm curious if that is still your view or if you have evolved a bit and concluded the that small cap effect never was there and it was never spectacular?
Thank you, I'm delighted that you remembered that one. I was trying to decide whether to update it. I did hedge, I wrote, "it's there, probably, but it's small." I think you're right: I was skeptical then, but I've gotten more skeptical.

I find it suspicious that as time goes on, the evidence for the size effect (in its original form) keeps getting weaker. If it were real, I would expect it to get stronger. It seems to me that the evidence for the value effect has also been getting weaker rather than stronger.
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.

MotoTrojan
Posts: 1890
Joined: Wed Feb 01, 2017 8:39 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by MotoTrojan » Sat Jun 16, 2018 6:29 pm

snarlyjack wrote:
Fri Jun 15, 2018 2:39 pm
How about a real life example.
(you guy's are kind of talking above my head...)

I read all of Paul Merrimans articles & agree
with his statement that a 1/2% or 1% can change
your life.

With that in mind I changed my Roth IRA to 100%
Vanguard Small Cap Value Fund (VSIAX). My broker
account is 100% Vanguard High Dividend Yield Index Fund
(VHDYX). I can't touch the Roth account for 36 more years.

The Small Cap Value Fund is about 60% Mid Caps & 40% Small Caps.
The High Dividend Yield Fund is 100% Large Cap. Value Fund.
So, Large, Medium & Small Cap value stocks. 100% stocks, no bonds.

How does the numbers look for this portfolio? Any comments
would be appreciated...long term buy & hold with Vanguard Funds.
Dividend fund in taxable is very inefficient, but I know you are a firm believer in dividends and many have expressed to you why they aren’t ideal.

snarlyjack
Posts: 780
Joined: Fri Aug 28, 2015 12:44 pm
Location: Montana

Re: Larry Swedroe: Rediscovering The Size Effect

Post by snarlyjack » Sat Jun 16, 2018 7:42 pm

Thank You, Moto Trojan.

The dividend fund is good. It gives me all the emergency funds
that I might need. I' am no longer contributing to the fund &
it is passively rolling along (my emergency expenses could
include property taxes, new roof, car maintenance, etc.).

Currently, I' am 100% focused on long term growth. I think
small cap value funds might help. All of my funds are large cap.
or small cap value funds. As a investor I prefer value funds.

Thank You for your help.

Dead Man Walking
Posts: 660
Joined: Wed Nov 07, 2007 6:51 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by Dead Man Walking » Sun Jun 17, 2018 1:31 am

I've followed the 15 year returns of funds at Morningstar for over a decade and most of the time small caps have outperformed mid caps and large caps. Consequently, I have tilted toward small caps. From my unscientific comparisons, the argument for value over growth depends more on which of them is in vogue at the time. Blend funds tend to be in the race no matter which happens to be in vogue. From my perspective, small is more significant than growth or value. I am an unsophisticated investor who relies on performance statistics. I believe in the aphorism that money talks and bs walks.

DMW

snarlyjack
Posts: 780
Joined: Fri Aug 28, 2015 12:44 pm
Location: Montana

Re: Larry Swedroe: Rediscovering The Size Effect

Post by snarlyjack » Sun Jun 17, 2018 7:27 am

Dead Man Walking,

That was great! I love your aphorism
"money talks & b.s. walks". I'll have to
remember that...

I totality agree with you. I thought Kevin O'Leary's
video was pretty informative for different assorted
reasons.

Thank You, Dead Man Walking :happy

https://www.youtube.com/watch?v=DEfHnhq_HeM

WhyNotUs
Posts: 1299
Joined: Sun Apr 14, 2013 11:38 am

Re: Larry Swedroe: Rediscovering The Size Effect

Post by WhyNotUs » Sun Jun 17, 2018 7:30 am

It must be exhausting coming up with all of these topics to write about. Larry, give yourself some time off. Set it and forget it.
I own the next hot stock- VTSAX

david1082b
Posts: 367
Joined: Fri Jun 09, 2017 12:35 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by david1082b » Sun Jun 17, 2018 12:10 pm

Dead Man Walking wrote:
Sun Jun 17, 2018 1:31 am
I've followed the 15 year returns of funds at Morningstar for over a decade and most of the time small caps have outperformed mid caps and large caps. Consequently, I have tilted toward small caps. From my unscientific comparisons, the argument for value over growth depends more on which of them is in vogue at the time. Blend funds tend to be in the race no matter which happens to be in vogue. From my perspective, small is more significant than growth or value. I am an unsophisticated investor who relies on performance statistics. I believe in the aphorism that money talks and bs walks.

DMW
It depends on your start date. Earlier in the thread someone posted a chart showing the S&P mid-400 outperformed the small 600 from early 1995 to May this year:

Image

1998 seemed to be a distinct break-point for the two indexes, with the small 600 falling away more and not having made the ground back as of now

garlandwhizzer
Posts: 1909
Joined: Fri Aug 06, 2010 3:42 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by garlandwhizzer » Sun Jun 17, 2018 1:41 pm

Robert T wrote:
1982-2018: Annualized return
11.8% = Large caps (Largest 30%)
11.6% = Small caps (Smallest 30%)
Factor investing theory suggests that over sufficiently long periods of time, factors outperform beta consistently and reliably. Otherwise why hold onto a factor portfolio? The above is 36 year period in which small caps as measured by the smallest cap 30% underperformed large caps as measured by the largest cap 30%. Since in theory size is a very easy thing to measure, cap weight alone, it is clear that the large caps here have considerably negative size exposure relative to the small caps which have positive exposure. So if the small factor is persistent and reliable over long time spans (36 years is a long time span in anyone's estimation) where is the outperformance? Some factor apologists argue that one must get rid of junk or negative MOM in the SC space to get the numbers to work. They don't suggest getting rid of junk with quality screens or getting rid of negative MOM in the LC space to make that modified LC versus SC a valid comparison. Anything can be shown to outperform on backtesting if you discard the parts of it that underperformed. It should be apparent that size alone has not worked to produce outperformance without tweaking it using 20/20 backtesting hindsight in the rearview mirror for the last 36 years. Therefore its persistence and reliability for outperformance are clearly demonstrable in the last 36 years and in serious question going forward.

I respectfully disagree with Robert T and believe that long term factor backtesting from 1927 to 1982 is inherently flawed for multiple reasons including wide bid/ask spreads in the SC space, trading frictions and illiquidity in the SC space, failure to account for SCV firms that failed, and the fact that during that long time period 1927 - 82 the overwhelming majority of investors completely ignored investing in the SC space which created abundant deep value in the SCV space that does not exist today at all. Cost free index comparison between beta and SCV does not IMO necessarily define reality going forward as this illustration clearly demonstrates, massive outperformance of SCV for 55 years (1927-82) followed by slight underperformance for the next 36 years. The most obvious conclusion to me is junk-in, junk-out in the underlying model.

Confession: I actually tilt modestly to SCV myself but do so mostly to increase diversification, adding SCV, the opposite of LCG which tends to dominate TSM, for improved diversification. I am more certain that there will be diversification benefit to the overall portfolio than that there will be significant outperformance from SCV.

Garland Whizzer

User avatar
vineviz
Posts: 985
Joined: Tue May 15, 2018 1:55 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by vineviz » Sun Jun 17, 2018 2:12 pm

garlandwhizzer wrote:
Sun Jun 17, 2018 1:41 pm
Factor investing theory suggests that over sufficiently long periods of time, factors outperform beta consistently and reliably. Otherwise why hold onto a factor portfolio? The above is 36 year period in which small caps as measured by the smallest cap 30% underperformed large caps as measured by the largest cap 30%.
I think anyone sincerely interested in understanding this question should at the very least be aware that a comparison starting arbitrarily in 1982 is isolating the very worst period of relative performance for small caps. In fact, 1984 to the oil is the most extreme such period in history.

In fact, small caps have outperformed large caps in most of the 36 years mentioned, and from 1991 to present have outperformed by 300 bps per year on average.

I might even argue that cherry picking a period that includes the worst available returns and STILL producing a draw amounts to faint praise more than a strong condemnation of the effect
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

snarlyjack
Posts: 780
Joined: Fri Aug 28, 2015 12:44 pm
Location: Montana

Re: Larry Swedroe: Rediscovering The Size Effect

Post by snarlyjack » Sun Jun 17, 2018 2:42 pm

Here is a new analysis put together by a engineer
going back to 1926-2017 that I thought was interesting.

Enjoy...

https://paulmerriman.com/wp-content/upl ... risons.pdf

https://paulmerriman.com/the-case-for-s ... -retirees/

stlutz
Posts: 4606
Joined: Fri Jan 02, 2009 1:08 am

Re: Larry Swedroe: Rediscovering The Size Effect

Post by stlutz » Sun Jun 17, 2018 3:56 pm

Annualized return (%) / Standard Deviation / Sharpe Ratio / Max calendar year loss

..9.8 / 16.6 / 0.46 / -26.0 = 50% Small Caps (Smallest 30%):50% 5-yr T-notes
..9.8 / 19.3 / 0.42 / -43.9 = 100% Large Caps (Largest 30%)
Hi Robert: I'm afraid I don't understand how one portfolio with 100% stocks is a useful comparison to one with 50% stocks.

I used the Simba/Siamond spreadsheet to compare two balanced portfolios from 1927ff with the weighting adjusted to provide equal standard deviation. This was what I ended up with:

Portfolio 1: 60% Total Stock / 40% Int. Treasuries
Portfolio 2: 50% Small Cap Stocks / 50% Intermediate Treasuries.

And the later one does produce a slightly higher Sharpe Ratio (.49 vs. .46), although the second also had the highest drawdown. This would be consistent with nisiprius' prior analysis that there could be something there, but the effect is small and it's very period dependent as to whether you'd have gotten any benefit.

When you leverage a portfolio by reducing bond allocations, what actually ends up happening is that you have to take a lot of additional risk to get a little bit of additional return. The 100/0 TSM portfolio returned exactly 1% more per year than the 60/40 allocation, but was 60% more volatile. Now that extra return isn't nothing, but it's clear that whether you lever up a total market portfolio or a smallcap-only portfolio, the Sharpe ratio goes down as you increase the equity component as opposed to staying constant.
I think anyone sincerely interested in understanding this question should at the very least be aware that a comparison starting arbitrarily in 1982 is isolating the very worst period of relative performance for small caps. In fact, 1984 to the oil is the most extreme such period in history.

In fact, small caps have outperformed large caps in most of the 36 years mentioned, and from 1991 to present have outperformed by 300 bps per year on average.

I might even argue that cherry picking a period that includes the worst available returns and STILL producing a draw amounts to faint praise more than a strong condemnation of the effect
I'm going to call LOL on this one. Just did some checking and it appears that 1991 is a cherry picked date to maximize outperformance numbers. Switch to 1990 or 1992 and the numbers get much smaller. :happy

But even if we go with 1991, the outperformance is half what you stated and it's still worse from a Sharpe ratio perspective.

https://www.portfoliovisualizer.com/bac ... Blend2=100

But, more generally, let's keep in mind that it is the smallcap factor advocates from AQR who advocated that the factor needing "saving". When the ones promoting a strategy question its effectiveness in its most simple form, then reconsideration is warranted.

User avatar
vineviz
Posts: 985
Joined: Tue May 15, 2018 1:55 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by vineviz » Sun Jun 17, 2018 4:06 pm

stlutz wrote:
Sun Jun 17, 2018 3:56 pm

I'm going to call LOL on this one. Just did some checking and it appears that 1991 is a cherry picked date to maximize outperformance numbers. Switch to 1990 or 1992 and the numbers get much smaller. :happy

But even if we go with 1991, the outperformance is half what you stated and it's still worse from a Sharpe ratio perspective.

https://www.portfoliovisualizer.com/bac ... Blend2=100
My point was that the period of severe large cap outperformance is quite distinct: 1984 to 1990 is an anomaly in the historical record.

And if you want to see what, if any, difference there is between large cap and small cap returns, you should probably compare those asset classes. A total stock market is a blend of both.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

stlutz
Posts: 4606
Joined: Fri Jan 02, 2009 1:08 am

Re: Larry Swedroe: Rediscovering The Size Effect

Post by stlutz » Sun Jun 17, 2018 4:16 pm

And if you want to see what, if any, difference there is between large cap and small cap returns, you should probably compare those asset classes. A total stock market is a blend of both.
As an explanation, the reason I use the total market is that this is the "base" starting point for most folks here. I advocate owning all parts the market as opposed to picking only certain slices, whether that be large caps, smallcap, companies with ticker symbol starting with the letter A etc. So, I'm interested in comparing this standard recommendation vs. the various recommendations that get promoted here to focus on sub-sections of the market.

If I was evaluating some type of long/short strategy, comparing once slice vs. another slice is more relevant.

MIretired
Posts: 712
Joined: Fri Sep 06, 2013 12:35 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by MIretired » Sun Jun 17, 2018 9:40 pm

Thanks for posting some of the distribution of returns data.

User avatar
siamond
Posts: 4032
Joined: Mon May 28, 2012 5:50 am

Re: Larry Swedroe: Rediscovering The Size Effect

Post by siamond » Sun Jun 17, 2018 10:05 pm

If we just stick to a simple definition of "small" consistent with CRSP indices, then I ran the numbers recently in this blog entry.

Here is the corresponding Telltale chart (i.e. cumulative growth relative to Total-Market's cumulative growth). Click to see a bigger version. The case for (US) small-caps has never been convincing and remains so. My primary comment about SCB (small-cap-blend) was:
  • SCB displayed a nice premium, but mostly due to two time periods (early 40s and late 70s). It stayed roughy on par with TSM for two periods of more than 30 years.

Image

I don't care about Sharpe ratios or some other not-so-meaningful metric, the premium is just not established well enough, period. Now if we play around and add other filters (whatever AQR came up with to make it work more convincingly), then well, maybe this is something real or maybe they just tortured the numbers enough until they liked it, I don't know. All I know is that I won't bet on an ever-shifting 'rediscovered' definition.

PS for Nisiprius: small-caps as defined by SBBI is micro-caps, remember? VERY misleading.

Solo Prosperity
Posts: 366
Joined: Mon May 11, 2015 2:53 pm

Re: Larry Swedroe: Rediscovering The Size Effect

Post by Solo Prosperity » Mon Jun 18, 2018 3:40 pm

Robert T wrote:
Sat Jun 16, 2018 12:58 pm
.
Using 1927-2017 data from Ken French’s website...
Wanted to add 3 things:

1. The size effect is not linear like most other commonly accepted factors. From 1927-2017, the best decile measured by performance was the 10th (Smallest) decile (Good start). The second best decile was the 8th (3rd smallest) decile (Not bad). The third best decile was the 5th (6th smallest) decile (Uh oh).
In fact, the 8th, 7th, 6th and 5th deciles all performed better than the 9th decile. To me, that is a tough pill to swallow in accepting the size effect as a premier anomaly.

2. Another issue with the size effect is how challenging and costly it can be to invest in those smallest percentiles of micro-cap firms. Liquidity and Volume concerns can create quite a spread in B/A spreads for that bottom decile.

A small DIY investor could do it with more ease but trying to replicate the size effect in an ETF seems to be challenging.

For example, the average weighted market cap of IWC (ishares Micro-Cap ETF) is $680 million. The average firm size in the smallest decile is $127 million. How many names could the index hold in that smallest decile without large costs incurred to performance? I don't know, but it seems clear that investing in the size effect presents unique liquidity and volume challenges a lot sooner than other factors.

3. AQR just published a great paper revisiting the size effect. The most interesting chart to me was when they looked at the "January" effect of small-caps. Essentially, the entire premium has occurred in the month of January.

Now I know any investment strategy will historically have better months than others due to chance, but the size effect has had an ~0 premium historically outside of January. That is bizarre.

All three of those factors are responsible for my lack of explicit size exposure in my portfolios. I still have some passive exposure to "size" via small-cap value funds and my personal small-cap value strategy, but I do not invest in small-cap blend funds, and I do not plan to.

Post Reply