Recent factor loadings?

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Topic Author
mrwhitepe
Posts: 26
Joined: Tue Mar 14, 2017 11:17 pm

Recent factor loadings?

Post by mrwhitepe »

I noticed something strange regarding factor loadings - when you go to portfoliovisualizer and check out lets say WisdomTree International SmallCap Fund (DLS) which should have nice size and value loadings, then you can see that for the date range 2006-2017 they are fine (size loading 0.34 value loading 0.11) but when you take a shorter date range, for example 2015-2017, then the loadings change to size loading 0.42 and value loading -0.2! Certainly not valuey.

I have found the same pattern with other supposedly size and value tilted ETFs, like the iShares Edge MSCI Multifactor Intl Small-Cap ETF (ISCF) which has the loadings 0.74 for size and -0.51 for value for the date range 2015-2017.

Whats your opinion about this, am I missing something? As it is pretty confusing, I cant see why anyone would use these ETFs for factor tilting when they obviously are pretty bad in capturing proper factor loadings during a several year span?
User avatar
in_reality
Posts: 4529
Joined: Fri Jul 12, 2013 6:13 am

Re: Recent factor loadings?

Post by in_reality »

mrwhitepe wrote:I noticed something strange regarding factor loadings - when you go to portfoliovisualizer and check out lets say WisdomTree International SmallCap Fund (DLS) which should have nice size and value loadings, then you can see that for the date range 2006-2017 they are fine (size loading 0.34 value loading 0.11) but when you take a shorter date range, for example 2015-2017, then the loadings change to size loading 0.42 and value loading -0.2! Certainly not valuey.
That's the problem I think with measuring factor loads.

For example if you are using Global EX-US, French-Fama research factors for their three factor model from 1/1/2015 - you have:

DLS HML -0.21
FNDC HML -0.13 (I use this fund so included it - it's a fundamental index so should have a value load)
VSS HML -0.12 (Vanguard FTSE All Wd Ex US Small Cap ETF)

So it looks like a cap weighted fund actually has the highest value loading.

Keep in mind though that the model attributes anything that is unexplained to alpha.

DLS Annual Alpha 3.93%
FNDC Annual Alpha 3.69%
VSS Annual Alpha -2.12%

DLS and FNDC both outperformed VSS quite handily in that time period (1/2015- now). ISCF performed like DLS and FNDC did.

Now switch to a four or five factor model. FNDC jumps in size and value loadings but lowers in Annual Alpha. DLS jumps to 6.61% Annual Alpha.

So either international data is wrong (and I have to believe it's wrong for VSS which included emerging small caps and emerging in particular is not great data), or the model is wonky. Probably both. In any case, the model seems unable to explain returns precisely and makes use of Alpha to make the math work.
mrwhitepe wrote:I have found the same pattern with other supposedly size and value tilted ETFs, like the iShares Edge MSCI Multifactor Intl Small-Cap ETF (ISCF) which has the loadings 0.74 for size and -0.51 for value for the date range 2015-2017.

Whats your opinion about this, am I missing something? As it is pretty confusing, I cant see why anyone would use these ETFs for factor tilting when they obviously are pretty bad in capturing proper factor loadings during a several year span?
I am not sure on ISCF methodology and what loadings you would expect. If you are picking a stock for multiple factors, can you really get holdings that have high value loads and stronger momentum, and lower volatility and what ever else it's designed to load on, or do you get a composite? Performance wise, it's seems to behave like other small cap international funds in that time.

I've read here that one would do best to look at the loadings of the index that a fund follows over it's entire lifetime to get the best idea for what loadings to expect. That is certainly true for a fund like FNDC whose value exposure will increase when the value-growth spread is large. I wonder if ISCF would be the same.

On a side note, it doesn't appear like the French-Fama standard to measure is actually that good. Having a SMB >0 doesn't actually mean you have exposure to small stocks.
We consider a reference portfolio consisting of 80% BIG (B) and 20% SMALL (S) where B and S are the same as in SMB; that is, returns are .8B  .2S . Table 1 presents sample statistics for this portfolio. It shows, for example, that the average monthly return for small stocks (1.26) has exceeded the return (1.02) to big stocks. But already anticipating our findings, notice that in spite of the 80% weight on B, the portfolio’s correlation with SMB is a positive,0.39.
What Does ßSMB>0 Really Mean?
Hsiu-lang Chen and Gilbert Bassett
University of Illinois at Chicago
August 2014
User avatar
grabiner
Advisory Board
Posts: 35265
Joined: Tue Feb 20, 2007 10:58 pm
Location: Columbia, MD

Re: Recent factor loadings?

Post by grabiner »

Part of the problem with factor loadings is that they measure correlation with a factor, but funds differ in ways unrelated to the factor. If Fund A outperforms fund B when value outperforms growth, then Fund A will have a higher value loading, even if the reason for the outperformance has nothing to do with value. You can see this because the funds show a poor correlation with the predicted returns given the factor weights; the R^2 will be low.

In theory, this should be less of a problem with indexes, but index inclusion rules still make a difference. If one emerging markets index includes South Korea and another doesn't, and the Korean market performs well over a period in which growth outperforms value, the index excluding South Korea will have a larger value loading.
Wiki David Grabiner
rkhusky
Posts: 17654
Joined: Thu Aug 18, 2011 8:09 pm

Re: Recent factor loadings?

Post by rkhusky »

grabiner wrote: You can see this because the funds show a poor correlation with the predicted returns given the factor weights; the R^2 will be low.
+1. The factor loadings don't mean much when the R^2 is low. The zigs and zags of the fund's returns have to match those of the references' returns.
sean.mcgrath
Posts: 786
Joined: Thu Dec 29, 2016 5:15 am
Location: US in NL

Re: Recent factor loadings?

Post by sean.mcgrath »

mrwhitepe wrote:I noticed something strange regarding factor loadings - when you go to portfoliovisualizer and check out lets say WisdomTree International SmallCap Fund (DLS) which should have nice size and value loadings, then you can see that for the date range 2006-2017 they are fine (size loading 0.34 value loading 0.11) but when you take a shorter date range, for example 2015-2017, then the loadings change to size loading 0.42 and value loading -0.2! Certainly not valuey.
I noticed something similar and posted about it in this post. I never did get a satisfying reply. My view is either the factors have decreased over time, or they have long cycles and it matters when you measure. And to the posts above, my R2s where in the range of 97%, so definitely not that.

Sean
Post Reply