Small company effect is overstated due to flawed dataset

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
carolinaman
Posts: 3289
Joined: Wed Dec 28, 2011 9:56 am
Location: North Carolina

Small company effect is overstated due to flawed dataset

Post by carolinaman » Sat Jun 23, 2018 6:22 am

An interesting article by John Rekenthaler of Morningstar. He interviewed Cliff Asness of AQR who claims that the small company effect is overstated due to flawed data.

https://www.morningstar.com/articles/87 ... esize.html

Key excerpt:
"He is considerably more certain about the small-company effect. "There isn't one … In fact, there never was a size effect. Among other issues, the data used to discover it was flawed (though no fault of the author, that was the data back then) in a way that favored small stocks."

"Fact, Fiction, and the Size Effect," by three AQR employees, reports that from 1936 through 1975, there was a "weak" size effect. The effect did not deliver much, and it wasn't statistically significant by all measures, but one could reasonably argue that it existed.

But not in the strong form that was originally promulgated. A major reason for the discrepancy, suggest the paper's authors, is that the early studies neglected companies that were delisted--that is, stocks that dropped off the exchanges. Such was not their intent, but the database that they used (CRSP) was not then complete. When the true, lower returns for the delisted stocks are considered, what once appeared to be a fat advantage for small firms dwindles to slender."

ping1050
Posts: 10
Joined: Thu Jul 17, 2014 4:55 pm

Re: Small company effect is overstated due to flawed dataset

Post by ping1050 » Sat Jun 23, 2018 6:46 am

I find it hard to believe that when calculating factor premium such as value and small, they would have overlooked this within the datat set.

And using portfoliovisualizer during his stated timeframe. 1986 to present, there appears to be a significant premium for SCV.

Would be interesting to hear Paul Merrimans thoughts on this article.

MikeG62
Posts: 1175
Joined: Tue Nov 15, 2016 3:20 pm
Location: New Jersey

Re: Small company effect is overstated due to flawed dataset

Post by MikeG62 » Sat Jun 23, 2018 6:59 am

At least for the last ten year period there is (using iShares ETF's as proxies with trailing 10 year returns through March 30, 2018):

S&P 500 = 9.44% (IVV)
S&P Mid Cap Core = 10.80% (IJH)
S&P Small Cap Core = 11.30% (IJR)

My US equity exposure has long been ~50% LC, 25% MC and 25% SC.
Real Knowledge Comes Only From Experience

carolinaman
Posts: 3289
Joined: Wed Dec 28, 2011 9:56 am
Location: North Carolina

Re: Small company effect is overstated due to flawed dataset

Post by carolinaman » Sat Jun 23, 2018 8:54 am

ping1050 wrote:
Sat Jun 23, 2018 6:46 am
I find it hard to believe that when calculating factor premium such as value and small, they would have overlooked this within the datat set.

And using portfoliovisualizer during his stated timeframe. 1986 to present, there appears to be a significant premium for SCV.

Would be interesting to hear Paul Merrimans thoughts on this article.
You make a good point and I was surprised by that assertion as well. It will good to hear from other experts on this.

User avatar
vineviz
Posts: 2062
Joined: Tue May 15, 2018 1:55 pm

Re: Small company effect is overstated due to flawed dataset

Post by vineviz » Sat Jun 23, 2018 9:11 am

ping1050 wrote:
Sat Jun 23, 2018 6:46 am
I find it hard to believe that when calculating factor premium such as value and small, they would have overlooked this within the datat set.
One assertion that Asness makes that I think is actually true relates to this. In the early days of assembling the CRSP database, researchers had no good way to adding companies that had been delisted (e.g. gone bankrupt, failed reporting criteria, etc.) or even knowing how many of those companies there were. For this reason, the smaller cap deciles (where most of those failed companies would have existed) accidentally overstated the returns because they weren't accounting for stocks that went to $0.00, especially for the earliest periods of data (1926 to 1970). Researchers in the 1970s and early 1980s wouldn't have necessarily known this was a problem and couldn't have corrected for it anyway.

CRSP subsequently tracked down those delistings and added them in, so the data we have today (which still show a small cap premium, by the way) are much cleaner.
ping1050 wrote:
Sat Jun 23, 2018 6:46 am
And using portfoliovisualizer during his stated timeframe. 1986 to present, there appears to be a significant premium for SCV.
The claim that Asness makes, that the small cap premium is dead, really should have been qualified more carefully. The pure size premium is smaller than we thought AND once you account for other risk factors (value and quality, especially) is statistically very small (maybe virtually zero).

On the other hand, factor investors already kind of knew this which is why you see them talking about SCV a lot more than SCG and why they prefer the S&P 600 small cap index (which screens for quality, and launched in 1994) over the CRSP and Russell indexes (which don't).
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

User avatar
Svensk Anga
Posts: 438
Joined: Sun Dec 23, 2012 5:16 pm

Re: Small company effect is overstated due to flawed dataset

Post by Svensk Anga » Sat Jun 23, 2018 11:34 am

A delisted small company might well have gone bust, but it may have been bought out/merged with another company. Did the CRSP database account for buy-outs as poorly as it did for bankruptcies? If so, there may be more small cap premium in the history than Asness asserts.

User avatar
LadyGeek
Site Admin
Posts: 48632
Joined: Sat Dec 20, 2008 5:34 pm
Location: Philadelphia
Contact:

Re: Small company effect is overstated due to flawed dataset

Post by LadyGeek » Sun Jun 24, 2018 2:33 pm

This thread is now in the Investing - Theory, News & General forum (theory).
Wiki To some, the glass is half full. To others, the glass is half empty. To an engineer, it's twice the size it needs to be.

longinvest
Posts: 3042
Joined: Sat Aug 11, 2012 8:44 am

Re: Small company effect is overstated due to flawed dataset

Post by longinvest » Sun Jun 24, 2018 5:49 pm

If the "Small Company Effect" and "premium" are really robust, they should work in Canada too. Let's look:

Source: Portfolio Visualizer

Portfolio 1 (blue): iShares Core S&P/TSX Capped Composite Index ETF (XIC)
Portfolio 2 (red): iShares S&P/TSX SmallCap Index ETF (XCS)

The time period was automatically adjusted based on the available data (Jun 2007 - May 2018) for the selected asset: iShares S&P/TSX SmallCap Index ETF (XCS)

Code: Select all

Portfolio performance statistics
                                                                                                                   
Portfolio     Initial Balance   Final Balance   CAGR   Stdev   Best Year  Worst Year  Max. Drawdown   Sharpe Ratio  Sortino Ratio  
Portfolio 1       $10,000          $15,455      4.04%  12.91%   34.54%     -33.34%       -43.58%          0.33           0.45      
Portfolio 2       $10,000          $10,799      0.70%  18.68%   58.30%     -43.88%       -55.41%          0.10           0.14      
Image

The small-cap ETF underperformed the total Canadian stock market by a wide margin and with much higher volatility. Small caps didn't even keep up with Canadian inflation which was 1.7% (annual) during the above period.

That's the kind of chart that factor peddlers don't show us.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VLB/ZRR

User avatar
nisiprius
Advisory Board
Posts: 36879
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: Small company effect is overstated due to flawed dataset

Post by nisiprius » Sun Jun 24, 2018 8:58 pm

ping1050 wrote:
Sat Jun 23, 2018 6:46 am
I find it hard to believe that when calculating factor premium such as value and small, they would have overlooked this within the datat set.

And using portfoliovisualizer during his stated timeframe. 1986 to present, there appears to be a significant premium for SCV.

Would be interesting to hear Paul Merrimans thoughts on this article.
There are two major sources of confusion in these discussions.

1) Are we talking about a "size effect" aka "small firm effect" aka "size factor," or are we talking about some second-order effect it has when combined with value (e.g. small-cap value) and/or other factors?

2) Are we merely talking about a "premium," i.e. higher return, which might just be reward for additional risk, or are we actually talking about higher risk-adjusted return?

A 1981 paper by Rolf Banz, here, which sort of started the whole factor thing--although they weren't called factors then--asserted that
It is found that smaller firms have had htgher risk adjusted returns, on average, than larger lirms. This ‘size effect’ has been in existence for at least forty years and is evidence that the capital asset pricing model is misspecified.
Notice that he was talking about higher risk-adjusted return. Also notice that the thrust of the paper was not improving portfolios or beating the market, it was about attacking CAPM and/or the efficient market hypothesis.

The data Banz used were flawed by delisting bias, and the effect isn't as strong as he said it was.

It is questionable whether small-cap stocks actually have had higher risk-adjusted returns. And there is at least some doubt about the weaker statement, that there is even a small-cap premium, because it has strange characteristics, like being entirely attributable to one single period of about nine years.

Since the existence of a size factor has been asserted for a long time, factor mavens are very reluctant to give it up. However, even they talk in terms of "resurrecting it" or "saving it" or "rescuing it," which is a polite way of saying that the size factor as initially presented in 1981 is weak to nonexistent, but that they think they have discovered some new hybrid factor in which size plays some role.
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.

Valuethinker
Posts: 36348
Joined: Fri May 11, 2007 11:07 am

Re: Small company effect is overstated due to flawed dataset

Post by Valuethinker » Mon Jun 25, 2018 7:31 am

longinvest wrote:
Sun Jun 24, 2018 5:49 pm
If the "Small Company Effect" and "premium" are really robust, they should work in Canada too. Let's look:

Source: Portfolio Visualizer

Portfolio 1 (blue): iShares Core S&P/TSX Capped Composite Index ETF (XIC)
Portfolio 2 (red): iShares S&P/TSX SmallCap Index ETF (XCS)

The time period was automatically adjusted based on the available data (Jun 2007 - May 2018) for the selected asset: iShares S&P/TSX SmallCap Index ETF (XCS)

Code: Select all

Portfolio performance statistics
                                                                                                                   
Portfolio     Initial Balance   Final Balance   CAGR   Stdev   Best Year  Worst Year  Max. Drawdown   Sharpe Ratio  Sortino Ratio  
Portfolio 1       $10,000          $15,455      4.04%  12.91%   34.54%     -33.34%       -43.58%          0.33           0.45      
Portfolio 2       $10,000          $10,799      0.70%  18.68%   58.30%     -43.88%       -55.41%          0.10           0.14      
Image

The small-cap ETF underperformed the total Canadian stock market by a wide margin and with much higher volatility. Small caps didn't even keep up with Canadian inflation which was 1.7% (annual) during the above period.

That's the kind of chart that factor peddlers don't show us.
There's all kinds of reasons why Canada might be a poor example to choose-- although the point is well made. It's a tiny stock market compared to the US one, and it's heavily tilted:

- on the financials to super caps - the Canadian banking and insurance industry, where the individual cos are of similar size bracket to all but the American superfinancials

- natural resources stocks - there, you get a lot of very speculative exploration companies, wildcatters etc. Those companies should be heavily tilted towards the "small cap growth" segment (i.e. hope over earnings, lottery ticket effects (right skew returns))

Many of us lived through the long, long period of Canadian market underperformance against the US and Europe in roughly 1987-1997, so we are unpleasantly aware of these truths. And then there was 2000 (the Nortel & JDS collapses).

What we've really caught on these graphs is that big Canadian financials have done a lot better than small Canadian resource stocks over the past 10 years. i.e. it looks like a sector story as much as a size story?

You'd need to run it on global portfolios, and you could look at Japan, UK etc in isolation- ie markets which are quite large in a world context and have more diversity in their components.

You'd also need a much longer sample period. I would say 50+ years.

I'd also have to check re definition of "small" in a Canadian context over an American. That's also an issue in the UK market. Roughly speaking we think "small cap" in the UK is less than £100m (say USD 130m on average), whereas in the US it was something like less than $500m?

pascalwager
Posts: 1247
Joined: Mon Oct 31, 2011 8:36 pm

Re: Small company effect is overstated due to flawed dataset

Post by pascalwager » Mon Jun 25, 2018 7:03 pm

I have a DFA portfolio with US and int'l large value, US deciles 6-10 value, US microcap and int'l small company. I guess the int'l small company and US microcap do provide diversification, but no substantial premium. (But I think they're also filtering for some other factors.)

The various small company funds do have higher ERs than the large value, but I'm not really impelled to do any exchanges right now.
Preferred AA: Total US and foreign stock markets and short-term Treasury fixed income

Jeff Albertson
Posts: 570
Joined: Sat Apr 06, 2013 7:11 pm
Location: Springfield

Re: Small company effect is overstated due to flawed dataset

Post by Jeff Albertson » Mon Jun 25, 2018 8:02 pm

In a related post,
The key point with this chart is that US small cap stocks are trading at the most expensive valuations on record (and materially higher vs large caps).
Everyone, join the crowd & let's factor!
Image
https://www.topdowncharts.com/single-po ... Valuations

AlohaJoe
Posts: 3833
Joined: Mon Nov 26, 2007 2:00 pm
Location: Saigon, Vietnam

Re: Small company effect is overstated due to flawed dataset

Post by AlohaJoe » Mon Jun 25, 2018 8:49 pm

ping1050 wrote:
Sat Jun 23, 2018 6:46 am
And using portfoliovisualizer during his stated timeframe. 1986 to present, there appears to be a significant premium for SCV.
Are Small Cap Value and Small Cap the same thing or are they different things?

Post Reply