a look at the size premium and BtM

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Topic Author
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

a look at the size premium and BtM

Post by larryswedroe »

http://www.etf.com/sections/index-inves ... =1&start=4

Thought this would be of interest, especially in light of the discussions on the size premium
Best wishes
Larry
User avatar
Blue
Posts: 1178
Joined: Sat Jul 12, 2008 10:18 pm

Re: a look at the size premium and BtM

Post by Blue »

Always enjoy your articles Larry. Nicely written as usual.

Interesting terminology, "economically significant" vs "statistically significant". I interpret the former as, "would have made the investor a significant return but could have been because of random luck that may not be repeatable"? Is this a reasonable interpretation or does "economically significant" have some specific other meaning?
Topic Author
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: a look at the size premium and BtM

Post by larryswedroe »

Blue
Yes, exactly right. The higher the t-stat the less likelihood it was a random outcome.

Larry
lazyday
Posts: 3849
Joined: Wed Mar 14, 2007 10:27 pm

Re: a look at the size premium and BtM

Post by lazyday »

Value stocks are considered the top 30 percent of equities when ranked by BtM (or other value metrics).
On the other hand, small stocks are considered the bottom 50 percent when ranked by market cap.
Is this just to make the H in HML noticeably different from the L?

Is it true that using portfolio weighted averages, companies in the bottom half by size are much smaller than the top half, while companies in the bottom half by BTM are only moderately lower in BTM?

Did FF provide a reason for using 30% vs 50%?
Topic Author
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 7:28 am
Location: St Louis MO

Re: a look at the size premium and BtM

Post by larryswedroe »

lazyday
Sorry but I have no idea why that convention about 30% value, 40% core and 30% growth is used, but small and large and 50/50.

As to your question on differences in HML. First the difference goes up and down, it's not constant. The difference greatly widened in late 90s in the growth bubble and then narrowed sharply in the aftermath of the bubble bursting. Logically the premium is larger when the difference is wider and vice versa (valuations matter). I don't know what you would consider moderately lower and it depends on how one defines the universe. Small value stocks tend to have lower btm than large value. For example, DFA LV has BTM of 1.6 while DFA SV has BTM of 1.3. Now can compare that to that of VFINX (S&P 500) 2.4. So VFINX BTM is about 50% higher than DFA LV and 85% higher than DFA SV.

Hope that helps
Larry
afan
Posts: 8191
Joined: Sun Jul 25, 2010 4:01 pm

Re: a look at the size premium and BtM

Post by afan »

Interesting results, but I have to question whether the t stats mean anything at all. The t stat would make sense for pre specified comparisons, based on theory. They mean considerably less, perhaps nothing, when applied to data that has already be analyzed for a size effect, over the same periods of time. In short, they are not "testing" anything, they are restating the observation of a size premium for part of this period.

But it gets much worse when they slice and dice the data by time period and by BTM, then pick out a few combinations that produce high t stats. This is a meaningless exercise. Retrospective searches through data, generating a large number of comparisons and then picking out the small minority that appear "significant" is a time dishonored technique. The results tell us nothing. They would tell us nothing even if this were new data. Since it is the same data that has been combed over for decades, the likelihood of doing a meaningful test here is near zero.

Had they done some correction for multiple comparisons then they would have addressed at least one of the issues. The repeat trip through the same data would remain a problem.

Still, interesting that the size premium showed up more strongly in some conditions than in others. Now run that forward for 80 years and let's see whether the observation holds up for new data.
We don't know how to beat the market on a risk-adjusted basis, and we don't know anyone that does know either | --Swedroe | We assume that markets are efficient, that prices are right | --Fama
Post Reply