How to get Fama-French EAFE Factors, with results

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

Back in 2007, Robert T wrote a helpful post about how one can go about calculating the International (EAFE) Fama-French research factors. (The factors are useful for doing regressions to determine the factor weighting of EAFE funds such as EFV or PXF. For all you ever wanted to know about factor analysis, see here.)

Things have changed a bit since then, so I figured I would update his info with what I've found. If you haven't done any regressions yourself yet, you might want to get some practice on the US side before reading on. Here's a great tutorial to start with. I suggest starting by running a few US funds, since the factors are much easier to come up with, so there are less ways to make a mistake. OK, moving on:

Basically, the difficulty with EAFE is that Prof. French gives us research factors for the US, and more recently for several regions: Global, North America, Europe, Asia-Pacific ex-Japan, and Japan. But if you want EAFE, it's problematic trying to derive it from the above. That's because the size and value thresholds for Europe, Asia, and Japan are different and vary with respect to each other, so you can't simply do a market-cap-weighted average or some such to get EAFE factors.

So, what do we do instead? You can get the Mkt and HmL factors from Kenneth French's website:
Robert T wrote:For Non-US Developed Market - Mkt factor:
  • I use the Mkt data from the Ind_all file under Intl. Index Portfolios from from Ken French’s website
Robert T wrote:For Non-US Development Mkt - Risk Free Rate [Rf]:
  • I use US T-bills as the risk free rate for all international FF3F regressions as did Sinquefield in this article and Fama-French in this one.
(Among other places, you can get the risk-free rate from French's website as well. They're in the list of US Fama-French factors, or in the developed market factors.)

The tricky one is SmB:
Robert T wrote:For Non-US Developed Market - SmB factor
  • We're on our own:

    Here are two options: From the MSCI website just use EAFE small cap index minus EAFE or from the S&P/citigroup website use Small (less than 1 billon) minus Big (greate than 5 billion). The latter has a longer total return series of monthly returns – back to 1989. The magnitude of the small cap premium was similar using the two measures (larger with the S&P/citigroup data) and they have a fairly high correlation coefficient (0.91).
I wasn't able to access the S&P citigroup data via Robert's link, and couldn't find it elsewhere on their site; I assume it is no longer public, although if I'm wrong and someone can point me to it, that'd be great. (Edit: found it still up on their old site, but apparently it's going to be pay-only as of the end of this month. Anyway, the MSCI method works fine.)

Fortunately, the MSCI method worked pretty well. The Fama-French methodology for calculating SmB is to take the returns of the smallest 10% of the given region by market cap and subtract the returns of the largest 90%. We can't do exactly that, but we can come close.

The MSCI EAFE Std index (the one tracked by VEA and EFA) covers the top 84% by market cap, and the small-cap index (tracked by SCZ) covers the next 14%. (There is now a micro-cap index as well, which will allow a slight improvement on this technique in the future, but it only has a few years' data right now.) So if you take the returns of the small-cap index minus those of the EAFE index, you get something close to the definition of SmB.

You can get the MSCI index performance data here.

You'll want to choose Developed Markets, USD. For size, you want the standard, which will show up in the list as "EAFE", and the small-cap, which will be "EAFE SC". Run the search and click on the name in the list. When the chart comes up, set the term to Full History, the Index Level to 'gross' (which means total returns including dividends, gross of withholding taxes,) and the frequency to monthly. Update, then click 'Download Data' at the top.

Once you've done that for both the standard and small indices, you can crunch SmB. For each month, you want to calculate the change from the previous month and express it as a percentage. ie (C/P - 1)*100, where C=current month value and P=previous month value. Do that for the 'small' and 'big' indices, then subtract the values in your big list from the ones in your small list. The result is your SmB!

Note: The MSCI small index has gross returns data back to Dec. 2000. If you want to go earlier than that you can set it to 'price' to get earlier values instead. It's not technically perfect, but we're already saying 14% = 10%, so whatever, right? ;) I actually did it both ways to check, and the SmB results for the two sets had a correllation of 0.998 - so if you want to use price it doesn't really matter.

These are the correlations between the SmB calculated as above and the various global SmB factors on Prof. French's site, since Jan. 2001:

Code: Select all

Global   Europe   Asia-Pac Ex-J  Japan    North America
0.845    0.816    0.428          0.570    0.370
So our calculated SmB factor for EAFE has a strong correlation to the global and European SmB factors, weaker to Asia and Japan, and weakest to North America. Sounds about right.

So, here's what I came up with:

Code: Select all

        Mkt-Rf   HmL    SmB	     RF
199007	 1.52	-0.15		       0.68
199008	-10.94	-1.02		      0.66
199009	-14.42	 1.87		      0.60
199010	 15.63	-3.77		      0.68
199011	-6.62	 1.00		       0.57
199012	 0.79	-2.23		       0.60
199101	 2.72	-0.07		       0.52
199102	 9.48	 5.37		       0.48
199103	-5.69	-0.66		       0.44
199104	 1.10	-0.10		       0.53
199105	-1.00	-0.57		       0.47
199106	-7.33	 0.17		       0.42
199107	 4.45	 1.40		       0.49
199108	-2.57	-0.99		       0.46
199109	 5.05	 1.13		       0.46
199110	 1.33	-3.02		       0.42
199111	-4.76	-2.12		       0.39
199112	 4.70	-1.04		       0.38
199201	-3.03	 3.95		       0.34
199202	-3.61	 0.31		       0.28
199203	-6.79	 1.12		       0.34
199204	-0.03	 7.25		       0.32
199205	 6.28	-0.88		       0.28
199206	-4.78	 0.10		       0.32
199207	-2.96	-3.82		       0.31
199208	 6.05	-5.59		       0.26
199209	-2.69	 0.30		       0.26
199210	-5.80	 0.24		       0.23
199211	 1.11	-0.45		       0.23
199212	 0.14	 0.91		       0.28
199301	-0.02	 3.38	 1.869	 0.23
199302	 2.85	 0.27	 1.459	 0.22
199303	 9.57	 5.72	 1.804	 0.25
199304	 8.79	 3.61	 2.185	 0.24
199305	 2.33	-1.60	 4.530	 0.22
199306	-3.36	-1.27	-3.848	 0.25
199307	 4.48	 1.43	 0.813	 0.24
199308	 5.03	 0.86	 0.351	 0.25
199309	-3.13	 0.50	 0.019	 0.26
199310	 2.15	-0.23	-3.384	 0.22
199311	-8.65	 1.62	-1.241	 0.25
199312	 6.56	 3.53	 1.450	 0.23
199401	 9.90	 3.21	 2.745	 0.25
199402	-1.08	 1.68	 2.294	 0.21
199403	-3.70	 1.79	 2.170	 0.27
199404	 3.59	 0.77	 0.094	 0.27
199405	-1.06	 1.79	-0.573	 0.32
199406	 1.55	 1.03	 0.757	 0.31
199407	 0.45	-0.51	-0.654	 0.28
199408	 1.75	-0.19	-1.246	 0.37
199409	-3.51	 0.28	-0.326	 0.37
199410	 3.28	 0.65	-1.788	 0.38
199411	-5.64	-0.71	-2.048	 0.37
199412	 0.63	-1.06	 0.796	 0.44
199501	-4.44	-0.04	 0.133	 0.42
199502	-0.78	-0.49	-0.906	 0.40
199503	 5.20	-0.57	-4.033	 0.46
199504	 3.80	-2.06	-1.397	 0.44
199505	-1.84	 0.70	-0.984	 0.54
199506	-1.99	 0.30	-0.335	 0.47
199507	 5.36	-0.46	 0.322	 0.45
199508	-3.55	-0.08	 1.133	 0.47
199509	 0.65	-0.90	-2.092	 0.43
199510	-2.13	-0.15	-1.181	 0.47
199511	 1.83	 0.28	-1.424	 0.42
199512	 3.37	 0.85	-0.485	 0.49
199601	-0.14	 1.79	 2.737	 0.43
199602	 0.17	 0.89	 0.726	 0.39
199603	 1.52	-0.72	-0.010	 0.39
199604	 2.43	 1.01	 2.467	 0.46
199605	-1.48	-0.73	-0.064	 0.42
199606	 0.16	-0.61	-0.559	 0.40
199607	-3.37	 0.23	-2.481	 0.45
199608	-0.49	 0.22	 0.250	 0.41
199609	 2.56	-0.17	-2.057	 0.44
199610	-1.75	-0.27	-0.628	 0.42
199611	 3.63	 2.57	-2.566	 0.41
199612	-1.46	 1.37	-2.035	 0.46
199701	-4.22	 0.94	 1.608	 0.45
199702	 1.17	 0.13	 0.362	 0.39
199703	 0.22	-0.80	-2.752	 0.43
199704	 0.05	-2.62	-3.732	 0.43
199705	 6.71	 0.99	 1.204	 0.49
199706	 5.00	-1.87	-3.588	 0.37
199707	 1.30	 0.10	-5.020	 0.43
199708	-7.53	 1.77	 1.202	 0.41
199709	 5.17	-1.78	-7.586	 0.44
199710	-7.61	 3.15	 2.269	 0.42
199711	-1.40	-4.04	-5.823	 0.39
199712	-0.07	-0.44	-6.573	 0.48
199801	 4.47	 6.91	 1.807	 0.43
199802	 5.97	 4.37	 2.673	 0.39
199803	 2.66	 1.64	-0.747	 0.39
199804	 0.30	-0.50	-0.864	 0.43
199805	-0.15	-0.24	 0.930	 0.40
199806	-0.16	-2.77	-4.617	 0.41
199807	 0.56	-0.62	-2.262	 0.40
199808	-12.44	-3.04	-0.799	 0.43
199809	-2.83	-1.99	-1.394	 0.46
199810	 9.79	-0.79	-2.270	 0.32
199811	 4.75	 0.23	-0.218	 0.31
199812	 3.27	-2.44	-3.438	 0.38
199901	-0.60	-1.58	-0.846	 0.35
199902	-2.81	 2.51	 1.240	 0.35
199903	 3.90	 5.98	 1.363	 0.43
199904	 4.12	 7.63	 3.500	 0.37
199905	-4.89	-0.93	 1.636	 0.34
199906	 3.42	 2.99	 1.535	 0.40
199907	 2.96	 2.44	 0.203	 0.38
199908	 0.87	 2.12	 1.874	 0.39
199909	 0.71	-0.34	-1.205	 0.39
199910	 3.60	-3.86	-5.004	 0.39
199911	 4.38	-7.91	-2.826	 0.36
199912	 8.86	-5.39	-7.951	 0.44
200001	-6.98	 0.72	 6.564	 0.41
200002	 2.44	-12.76	-0.138	 0.43
200003	 3.17	 6.85	-1.409	 0.47
200004	-5.89	 3.82	-2.158	 0.46
200005	-3.66	 8.29	 2.335	 0.50
200006	 3.64	 3.60	 4.994	 0.40
200007	-4.55	 1.55	-2.002	 0.48
200008	 0.88	-0.78	 3.958	 0.50
200009	-5.11	 3.53	-0.780	 0.51
200010	-3.40	 3.27	-4.492	 0.56
200011	-3.59	 7.59	 3.553	 0.51
200012	 2.55	 3.20	-3.120	 0.50
200101	-0.18	 3.25	 3.061	 0.54
200102	-7.92	 4.67	 5.499	 0.38
200103	-7.57	 0.15	-0.926	 0.44
200104	 7.02	-0.06	 1.575	 0.39
200105	-3.71	-0.03	 3.294	 0.32
200106	-4.28	 0.31	 1.124	 0.28
200107	-2.50	 1.50	-2.292	 0.30
200108	-2.82	 2.99	 3.606	 0.31
200109	-10.35	-4.85	-3.237	 0.28
200110	 2.78	 0.84	 2.560	 0.22
200111	 3.35	 1.46	 0.168	 0.17
200112	 0.06	-0.68	-3.050	 0.15
200201	-5.04	-1.71	 2.958	 0.14
200202	 0.64	-1.52	 0.962	 0.13
200203	 5.66	 1.40	 1.195	 0.13
200204	 0.61	-0.61	 2.946	 0.15
200205	 1.42	 4.09	 3.128	 0.14
200206	-4.21	-2.75	 0.235	 0.13
200207	-9.09	 1.62	 2.154	 0.15
200208	-0.53	 0.19	-1.073	 0.14
200209	-10.70	-2.89	 2.947	 0.14
200210	 5.35	-1.07	-6.617	 0.14
200211	 4.54	 3.82	-1.669	 0.12
200212	-3.14	 0.31	 1.498	 0.11
200301	-3.82	 3.61	 2.755	 0.10
200302	-2.37	-0.15	 1.493	 0.09
200303	-1.96	-1.94	 1.377	 0.10
200304	 9.56	 3.53	-0.425	 0.10
200305	 6.34	 4.81	 2.466	 0.09
200306	 2.26	 1.61	 2.894	 0.10
200307	 2.43	 3.64	 0.678	 0.07
200308	 2.41	 2.04	 4.177	 0.07
200309	 3.45	 3.44	 2.902	 0.08
200310	 6.26	 3.90	 1.749	 0.07
200311	 2.08	-0.09	-2.644	 0.07
200312	 7.51	 0.71	-1.716	 0.08
200401	 1.71	 2.07	 3.663	 0.07
200402	 2.36	 1.30	 0.524	 0.06
200403	 0.52	 0.68	 3.543	 0.09
200404	-2.39	 0.51	-0.821	 0.08
200405	 0.12	 0.17	-1.993	 0.06
200406	 2.01	 1.44	 3.186	 0.08
200407	-3.25	 0.81	-1.278	 0.10
200408	 0.14	 0.87	 0.265	 0.11
200409	 2.42	 1.24	-0.235	 0.11
200410	 3.39	 0.64	 0.346	 0.11
200411	 6.74	 1.62	 1.114	 0.15
200412	 4.46	 0.90	 0.608	 0.16
200501	-1.57	 1.10	 3.805	 0.16
200502	 3.88	 0.45	-0.316	 0.16
200503	-2.76	 0.94	 0.895	 0.21
200504	-2.71	-2.00	-0.060	 0.21
200505	-0.71	 0.08	-0.102	 0.24
200506	 1.57	-0.42	 1.350	 0.23
200507	 2.97	 1.87	 1.451	 0.24
200508	 2.73	 0.08	 0.353	 0.30
200509	 3.37	-0.17	-0.341	 0.29
200510	-2.61	 1.62	-0.349	 0.27
200511	 1.80	-1.10	 0.757	 0.31
200512	 4.82	-1.09	 3.345	 0.32
200601	 5.75	-0.34	 0.683	 0.35
200602	-0.46	 3.04	-1.008	 0.34
200603	 2.94	-0.42	 1.520	 0.37
200604	 4.40	-0.20	-0.461	 0.36
200605	-4.85	-0.10	-1.922	 0.43
200606	-0.33	-1.08	-2.198	 0.40
200607	 0.71	 1.26	-3.697	 0.40
200608	 2.14	 0.26	 0.034	 0.42
200609	-0.06	 2.06	 0.246	 0.41
200610	 3.63	 1.26	-0.226	 0.41
200611	 2.52	 0.65	 1.283	 0.42
200612	 2.97	 1.73	 0.198	 0.40
200701	 0.34	 1.11	 1.410	 0.44
200702	 0.50	-0.51	 0.860	 0.38
200703	 2.42	-0.77	 0.660	 0.43
200704	 3.68	-0.04	-0.763	 0.44
200705	 1.42	 0.72	-1.122	 0.41
200706	-0.06	-0.53	-0.339	 0.40
200707	-2.03	-2.26	 1.083	 0.40
200708	-2.03	-1.15	-3.605	 0.42
200709	 5.08	-1.73	-4.233	 0.32
200710	 3.73	 0.14	 2.302	 0.32
200711	-3.92	-3.91	-3.924	 0.34
200712	-2.59	 1.07	-1.213	 0.27
200801	-8.76	 1.50	-0.528	 0.21
200802	 1.77	-2.93	 2.798	 0.13
200803	-1.40	 2.21	 0.739	 0.17
200804	 5.08	-1.20	-3.140	 0.18
200805	 1.22	-3.91	 0.601	 0.18
200806	-7.96	-3.73	-0.016	 0.17
200807	-3.67	 2.59	-1.358	 0.15
200808	-4.42	 1.02	 0.200	 0.13
200809	-14.05	 3.18	-2.692	 0.15
200810	-20.09	-2.03	-3.561	 0.08
200811	-5.43	-1.61	 0.945	 0.03
200812	 6.52	-1.01	 0.823	 0.09
200901	-9.69	-3.18	 3.402	 0.00
200902	-9.93	-1.93	 0.981	 0.01
200903	 5.82	 5.38	 0.177	 0.02
200904	 12.34	 17.03	 2.477	 0.01
200905	 12.77	 2.62	 2.277	 0.00
200906	-0.20	-0.41	 2.514	 0.01
200907	 9.03	 4.84	-1.501	 0.01
200908	 4.77	 6.01	 2.643	 0.01
200909	 4.16	 0.11	 1.158	 0.01
200910	-1.42	-2.03	-0.520	 0.00
200911	 2.08	-1.54	-2.013	 0.00
200912	 1.22	-0.12	-0.689	 0.01
201001	-3.66	 0.49	 3.419	 0.00
201002	-0.96	-0.18	-0.711	 0.00
201003	 6.36	 1.37	 1.045	 0.01
201004	-0.94	 1.94	 3.479	 0.00
201005	-11.22	-1.70	-1.007	 0.01
201006	-0.97	-1.41	 0.579	 0.01
201007	 9.57	 2.13	-0.853	 0.01
201008	-3.31	-1.01	 0.208	 0.01
201009	 10.12	-0.06	 1.602	 0.01
201010	 4.06	 0.04	 0.249	 0.01
201011	-4.88	-1.92	 1.223	 0.01
201012	 8.39	 1.89	 3.547	 0.01
201101	 2.21	 4.74	-1.578	 0.01
201102	 3.05	 0.20	-1.048	 0.01
201103	-1.69	-3.07	 2.133	 0.01
201104	 6.16	-1.09	-0.761	 0.00
201105	-2.65	-2.03	 0.336	 0.00
201106	-1.20	-0.37	-0.408	 0.00
201107	-1.34	-2.78	 0.535	 0.00
201108	-8.89	-3.12	 0.808	 0.01
201109	-9.72	 0.73	-0.835	 0.00
201110	 8.51	-1.05	-2.446	 0.00
201111	-4.69	-1.93	-0.575	 0.00
201112	-1.25	 0.30	-0.967	 0.00
201201			           2.925	
201202			           0.247	
201203			           0.533	
201204			           1.409	
201205			          -0.329	
201206			          -2.966	
201207			          -0.759	
201208			          -0.013	
And finally, a few regressions using these factors:

Code: Select all

First, a sanity check:
                                          Factor Loadings
    Fund                        Mkt-Rf    SmB     HmL     Alpha

    MSCI EAFE All-Cap Index       1.01    0.03    0.03    -0.50%
       t-values                 130.40    1.13    1.94    -0.86

    R^2 = 0.998     (2007/12 - 2011/12)
Looks pretty good. Bit of negative alpha there, but not statistically significant.


    Fund                        Mkt-Rf    SmB     HmL     Alpha     MER

    VEA                          1.05   -0.34    0.08    -0.10%    0.12%
       t-values                 39.55   -3.88    1.52    -0.05

    R^2 = 0.976     (2007/08 - 2011/12)
(Note VEA tracks the MSCI EAFE 'standard' AKA 'large + mid' index, so the negative size loading is expected.)


    Fund                        Mkt-Rf    SmB     HmL     Alpha     MER

    SCZ                          1.07    0.50    0.05    -0.45%    0.40%
       t-values                 32.93    4.31    0.84    -0.18

    R^2 = 0.973     (2008/02 - 2011/12)
	

    Fund                        Mkt-Rf    SmB     HmL     Alpha     MER

    PXF                          1.07   -0.30    0.48    -1.23%    0.75%
       t-values                 33.94   -2.96    7.57    -0.53

    R^2 = 0.972     (2007/08 - 2011/12)


    Fund                        Mkt-Rf    SmB     HmL     Alpha     MER

    EFV                          1.04   -0.30    0.35    -3.26%    0.40%
       t-values                 39.71   -3.86    6.67    -1.95

    R^2 = 0.968     (2005/09 - 2011/12)
Now there's some negative alpha...
	
	
	And again over a shorter time period to compare with PXF:
    Fund                        Mkt-Rf    SmB     HmL     Alpha     MER

    EFV                          1.03   -0.35    0.38    -3.39%    0.40%
       t-values                 32.62   -3.51    5.96    -1.46

    R^2 = 0.969     (2007/08 - 2011/12)


Edit: Forgot DFA!
                                          Factor Loadings
    Fund                        Mkt-Rf    SmB     HmL     Alpha     MER

    DFA Int'l Vector             1.08    0.22    0.19    -1.03%    0.54%
       t-values                 34.48    2.16    3.42    -0.49

    R^2 = 0.984
(Includes some EM which may throw it off a BIT, but R^2 still very high.)

Hope that's useful to someone! If you see room for improvement of the method, please let me know. And if there's an EAFE fund you'd like analyzed but don't feel like doing the leg work, let me know and I might be in a helpful mood. ;)

Edit: Note: these factors are only appropriate for profiling funds that invest across EAFE. For specific country or smaller regional funds (like Europe), it's better to use the regional or country data from Prof. French's site. For World Ex-US or Emerging, Jason Hsu provides factors that should be fairly comparable to the FF ones here: http://www.jasonhsu.org/research-data.html

Edit 2: it should be noted that making the approximations described here does have some effects on the eventual results, which I discuss in a post below.
Last edited by ClosetIndexer on Mon Sep 10, 2012 10:33 pm, edited 2 times in total.
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

Update: this test is incomplete. For a better analysis, see my post below.

Just for fun, I figured I'd test how close the SmB factors we get with this MSCI index method are to the official Fama-French ones. As mentioned above, we're using MSCI's indices, which cover the top 84% and near-bottom 14% of the market, rather than the exact specification of the research factor as per FF, which is the bottom 10% minus the top 90%.

To test, we can use the same MSCI index-based technique to calculate a SmB factor for the united states, and compare it to the 'official' FF research factor. In the US, the large+mid index is the top 86% of the market by market cap, rather than the top 84% for international, and the small-cap index is the next 12% instead of the next 14%. Still, this should give us a rough idea of how different the resulting factor ends up being by making this kind of approximation.


Over the past 10 years, the monthly SmB factor calculated in this way has a 93% correlation with the official research factor. What does that mean as far as results in our regressions? To get an idea, I did regressions on a few funds over the same time periods, using the two different sets of SmB data. Here are the results for VBR and IJS:

Code: Select all

    Using research SmB factor:
                                Mkt-Rf    SmB     HmL     Alpha     

    VBR                          0.98    0.63    0.43    -0.54%    R^2 = 0.970
       t-values                 36.32   11.92    9.24    -0.41

    Using estimated SmB factor:
    VBR                          0.92    0.75    0.34    -0.89%    R^2 = 0.975
       t-values                 34.98   13.65    7.84    -0.74

	   
    Using research SmB factor:
                                Mkt-Rf    SmB     HmL     Alpha     

    IJS                          0.92    0.85    0.40    -1.72%    R^2 = 0.962
       t-values                 37.55   18.11    9.10    -1.32

    Using estimated SmB factor:
                                Mkt-Rf    SmB     HmL     Alpha     

    IJS                          0.89    0.95    0.25    -1.38%    R^2 = 0.959
       t-values                 33.55   17.15    5.18    -1.02

So... we do still get something roughly representative. The SmB exposure is overestimated compared to the research factors, which is expected, since we're using a 'watered down' version of the factor definition. Unfortunately, the change to SmB also reduces the results for Mkt-Rf and HmL, to a greater degree than I would have expected.

Conclusions? Using the above technique will give you a fair estimate of the FF factors for EAFE funds. However, it will likely overestimate the SmB factor somewhat, and underestimate the Mkt and HmL factors, compared with what one would get using the precise official definitions of the FF factors. It will still be perfectly appropriate for comparisons between EAFE funds. However, one should be wary of making direct comparisons between factors on the US and international sides of a portfolio, since they were arrived at in different manners and will not be directly comparable. (Or course, this would certainly be the case even if we had perfect international SmB data, due to differences between the two markets.)
Last edited by ClosetIndexer on Mon Sep 10, 2012 10:32 pm, edited 1 time in total.
staythecourse
Posts: 6993
Joined: Mon Jan 03, 2011 8:40 am

Re: How to get Fama-French EAFE Factors, with results

Post by staythecourse »

Can you tell what the correlation of U.S. SMB and HML vs. EAFE SMB and HML? I am trying to figure out if the useful of the value and small premium extend into the added diversification return of low correlations?

U.S. and EAFE market premiums seem to be highly correlated as one would expect since the euro and the globalization, but wondered about if it is the same for the small and value premiums.

Thanks.
"The stock market [fluctuation], therefore, is noise. A giant distraction from the business of investing.” | -Jack Bogle
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

staythecourse wrote:Can you tell what the correlation of U.S. SMB and HML vs. EAFE SMB and HML? I am trying to figure out if the useful of the value and small premium extend into the added diversification return of low correlations?

U.S. and EAFE market premiums seem to be highly correlated as one would expect since the euro and the globalization, but wondered about if it is the same for the small and value premiums.

Thanks.
Yep, just take the EAFE SmB and HmL lists from my post above and pop them into excel or openoffice calc next to the corresponding lists from the US research factors from Ken French's website, and use the 'correl' function to determine the correlation over the period you're interested in. I created a quick and dirty spreadsheet where you can enter the start and end months and get the correlation. It's here if you want to play with it.

Over the entire period for which I have international data (Jan. 1993 - Dec. 2011), with the factors calculated as described above, I get these correlations:

Mkt-Rf Correlation: 0.8190808285
HmL Correlation: 0.2750288622
SmB Correlation: 0.4616633636

So short answer, it looks like it's much less true for SmB and HmL than it is for Mkt. Good news I guess!
staythecourse
Posts: 6993
Joined: Mon Jan 03, 2011 8:40 am

Re: How to get Fama-French EAFE Factors, with results

Post by staythecourse »

ClosetIndexer wrote:
staythecourse wrote:Can you tell what the correlation of U.S. SMB and HML vs. EAFE SMB and HML? I am trying to figure out if the useful of the value and small premium extend into the added diversification return of low correlations?

U.S. and EAFE market premiums seem to be highly correlated as one would expect since the euro and the globalization, but wondered about if it is the same for the small and value premiums.

Thanks.
Yep, just take the EAFE SmB and HmL lists from my post above and pop them into excel or openoffice calc next to the corresponding lists from the US research factors from Ken French's website, and use the 'correl' function to determine the correlation over the period you're interested in. I created a quick and dirty spreadsheet where you can enter the start and end months and get the correlation. It's here if you want to play with it.

Over the entire period for which I have international data (Jan. 1993 - Dec. 2011), with the factors calculated as described above, I get these correlations:

Mkt-Rf Correlation: 0.8190808285
HmL Correlation: 0.2750288622
SmB Correlation: 0.4616633636

So short answer, it looks like it's much less true for SmB and HmL than it is for Mkt. Good news I guess!
Interesting. That is the reason I have felt supports Mr. Swedroe's low beta highly tilted portfolio. Looks like the data supports that.

Good luck.
"The stock market [fluctuation], therefore, is noise. A giant distraction from the business of investing.” | -Jack Bogle
Jerry_lee
Posts: 915
Joined: Tue Mar 06, 2012 12:55 pm

Re: How to get Fama-French EAFE Factors, with results

Post by Jerry_lee »

staythecourse,

From 1975-2011:

Code: Select all

Data Series	US Market	US Size	US Value	World Size 	World Value
US Market	   1.00000				
US Size	     0.09517	 1.00000			
US Value	   -0.30296	 0.10873	1.00000		
World Size	 -0.04677	 0.72708	0.19391	 1.00000	
World Value	-0.23852	 0.22489	0.87614	 0.46875	    1.00000
The most disciplined investor in the world.
staythecourse
Posts: 6993
Joined: Mon Jan 03, 2011 8:40 am

Re: How to get Fama-French EAFE Factors, with results

Post by staythecourse »

Jerry_lee wrote:staythecourse,

From 1975-2011:

Code: Select all

Data Series	US Market	US Size	US Value	World Size 	World Value
US Market	   1.00000				
US Size	     0.09517	 1.00000			
US Value	   -0.30296	 0.10873	1.00000		
World Size	 -0.04677	 0.72708	0.19391	 1.00000	
World Value	-0.23852	 0.22489	0.87614	 0.46875	    1.00000
That's awesome. I wonder if we should have another thread to discuss the implications of this asset allocation and portfolio management?

Good luck.
"The stock market [fluctuation], therefore, is noise. A giant distraction from the business of investing.” | -Jack Bogle
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

Jerry_lee wrote:staythecourse,

From 1975-2011:

Code: Select all

Data Series	US Market	US Size	US Value	World Size 	World Value
US Market	   1.00000				
US Size	     0.09517	 1.00000			
US Value	   -0.30296	 0.10873	1.00000		
World Size	 -0.04677	 0.72708	0.19391	 1.00000	
World Value	-0.23852	 0.22489	0.87614	 0.46875	    1.00000
The nice thing about these numbers is that they can be arrived at using the official FF research data, as opposed to my estimated EAFE data above. Don't suppose you feel like crunching the correlations for the various developed factors as well and adding them! I've got too much other number-crunching on my to-do list already! :)

One thing to note looking at this data is that 'World' includes USA, so these correlations are going to be higher than the ones between US/international, which I estimated above. Due to my necessary estimates of the international factors though, my correlation numbers will be a bit lower than the 'true' correlations.

The best picture would be painted by checking the correlations between the US, NA, Europe, Asia ex-Japan, and Japan factors from French's site. staythecourse...? ;)
Jerry_lee
Posts: 915
Joined: Tue Mar 06, 2012 12:55 pm

Re: How to get Fama-French EAFE Factors, with results

Post by Jerry_lee »

ClosetIndexer wrote:
Jerry_lee wrote:staythecourse,

From 1975-2011:

Code: Select all

Data Series	US Market	US Size	US Value	World Size 	World Value
US Market	   1.00000				
US Size	     0.09517	 1.00000			
US Value	   -0.30296	 0.10873	1.00000		
World Size	 -0.04677	 0.72708	0.19391	 1.00000	
World Value	-0.23852	 0.22489	0.87614	 0.46875	    1.00000
The nice thing about these numbers is that they can be arrived at using the official FF research data, as opposed to my estimated EAFE data above. Don't suppose you feel like crunching the correlations for the various developed factors as well and adding them! I've got too much other number-crunching on my to-do list already! :)

One thing to note looking at this data is that 'World' includes USA, so these correlations are going to be higher than the ones between US/international, which I estimated above. Due to my necessary estimates of the international factors though, my correlation numbers will be a bit lower than the 'true' correlations.

The best picture would be painted by checking the correlations between the US, NA, Europe, Asia ex-Japan, and Japan factors from French's site. staythecourse...? ;)
Closet, sorry I abbreviated World ex US to just "world". I know that makes no sense, as World implies US + foreign, I was more worried about formatting :happy
The most disciplined investor in the world.
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

Ah, cool. Sorry, the quote above confused me - thought that data came from staythecourse.

Sooo now I'm curious where you got the world ex-us data. Best source I've found for that so far is Jason Hsu's site: http://www.jasonhsu.org/research-data.html

But the results I get with that data are pretty different from what you presented. Comparing the US HmL factor as given by FF to Hsu's calculation of the Global ex-US HmL factor for instance, I get correlations of

1982-Present: 0.51

2002-Present: 0.64

(Your result: 0.87 - that's what made me assume your world data included US, as does the Global Developed data on French's site.)

Again, for a relatively definitive answer, the best thing would be to do correlations between the US and international factors provided by Prof. French. If someone else doesn't beat me to it, I'll do that once my to-do list gets a bit shorter!
staythecourse
Posts: 6993
Joined: Mon Jan 03, 2011 8:40 am

Re: How to get Fama-French EAFE Factors, with results

Post by staythecourse »

ClosetIndexer wrote:Again, for a relatively definitive answer, the best thing would be to do correlations between the US and international factors provided by Prof. French. If someone else doesn't beat me to it, I'll do that once my to-do list gets a bit shorter!
I would agree. It will be interesting to see the correlations of the small and value vs. market premiums from U.S. and foreign. This may be the best way to show that diversifying is best done at the small and value premiums vs. geographical (market weights) when one is talking about diversifying under equities. Especially if one shows a continued lower correlation of small and value during the rising market premiums from the time the euro became a mainstay use to current.

I would love to help, but have two thumbs about how to do it.

Good luck.
"The stock market [fluctuation], therefore, is noise. A giant distraction from the business of investing.” | -Jack Bogle
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

My test above of this methodology for getting EAFE factors was incomplete. I should really use the identical methodology on the US data - calculating both HmL and SmB factors in the same manner as for EAFE, and see how the total results compare to the official US FF research factors.

So I did, and here's what I got:

Code: Select all

HmL Correlations:
1992-2011: 0.90
2002-2011: 0.91

SmB Correlations:
1992-2011: 0.83
2002-2011: 0.93

(Data begins in July 1992 and ends Dec. 2011.)
Again, these are the correlations between the United States Fama-French research factors published on Prof. French's site, and a set of United States factors that I came up with, using the exact method I describe in the OP above for calculating EAFE factors. The idea is to illustrate how much of a difference our approximations above make.

So, those correlations are reasonably high, especially for the past decade. So how much of a difference do we see in real-world results? I used both sets of factors to profile several funds from SV to LV. For each fund I list the regression results with the official factors followed by the estimated factors:

Code: Select all

                                Mkt-Rf    SmB     HmL     Alpha     
    VBR official                 0.98    0.63    0.43    -0.54%    R^2 = 0.970
       t-values                 36.32   11.92    9.24    -0.41

    VBR estimate                 0.92    0.69    0.30    -0.55%    R^2 = 0.972
       t-values                 31.86   11.12    6.75    -0.42

                                Mkt-Rf    SmB     HmL     Alpha     
    IJS official                 0.92    0.85    0.40    -1.72%    R^2 = 0.962
       t-values                 37.55   18.11    9.10    -1.32

    IJS estimate                 0.87    0.93    0.19    -1.11%    R^2 = 0.955
       t-values                 30.36   14.80    3.72    -0.77

                                Mkt-Rf    SmB     HmL     Alpha     
    VOE official                 1.01    0.25    0.27     0.23%    R^2 = 0.969
       t-values                 33.25    3.54    4.62     0.13

    VOE estimate                 0.97    0.34    0.17     0.09%    R^2 = 0.970
       t-values                 28.83    3.90    2.88     0.05
	   
                                Mkt-Rf    SmB     HmL     Alpha     
    VTV official                 0.93   -0.21    0.33    -0.60%    R^2 = 0.972
       t-values                 46.11   -5.43    9.51    -0.61

    VTV estimate                 0.94   -0.34    0.33     0.10%    R^2 = 0.972
       t-values                 42.38   -7.09    9.73     0.10
And there you have it. Now, do these numbers make sense? Yes, they do. :) For SmB, our 'small' market chunk from the MSCI fund covers the 2 to 16 percentiles of the market. aka the near-bottom 14% of total market cap. The Fama-French data uses a different technique for the US, sorting all stocks by market cap and taking everything below the median. This results in less stocks in the small bucket, which causes their SmB factor to be more 'pure'. So, so since we're doing regressions against a less-small SmB in our estimates, we end up overestimating our SmB exposure across the board.

For HmL, the official definition is:

HML = 1/2 (Small Value + Big Value) - 1/2 (Small Growth + Big Growth).

In our estimate, what we're effectively doing is

HML = Total Value - Total Growth

But since 'large' stocks make up the majority of the total BtM-sorted portfolios, our HmL factor is primarily tracking the movements of large value stocks, rather than equally considering small and large value stocks as the official research factor does. For this reason, you get very accurate HmL factors for large value funds like VTV, underestimate HmL somewhat for mid-cap value funds like VOE, and underestimate it more significantly for small-cap value funds like VBR and IJS.

TL;DR:
While the method in the OP produces factors that can give good regression results (high R-squared) for EAFE funds, it is important to be aware of how the approximations used will affect the SmB and HmL factors, compared to the results one would theoretically get using the research factors:
It will slightly overestimate the magnitude* of the SmB factor for all funds.
It should be accurate for the HmL factor of large-cap funds, but will underestimate the magnitude* of HmL more and more as the cap-size of the funds being profiled decreases (aka as their SmB increases).

*This means the magnitude in either direction. So a fund with a 'true' SmB of -0.2 would be overestimated to, say, -0.3. One with a true SmB of 0.3 might be overestimated to 0.42.

Now if only someone with access to the underlying data would produce a set of research factors for EAFE (and Canada while they're at it), none of this would be necessary! Until then, hope this helps somebody. (It at least helped me understand the model better, going through the process!)
Last edited by ClosetIndexer on Sat Sep 15, 2012 2:23 pm, edited 2 times in total.
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

staythecourse wrote:
ClosetIndexer wrote:Again, for a relatively definitive answer, the best thing would be to do correlations between the US and international factors provided by Prof. French. If someone else doesn't beat me to it, I'll do that once my to-do list gets a bit shorter!
I would agree. It will be interesting to see the correlations of the small and value vs. market premiums from U.S. and foreign. This may be the best way to show that diversifying is best done at the small and value premiums vs. geographical (market weights) when one is talking about diversifying under equities. Especially if one shows a continued lower correlation of small and value during the rising market premiums from the time the euro became a mainstay use to current.

I would love to help, but have two thumbs about how to do it.

Good luck.
I just can't resist a good challenge... Here you go: http://www.bogleheads.org/forum/viewtop ... 4#p1486141
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by LadyGeek »

As mentioned on another thread, I'm going to put this tutorial in the wiki. I've been doing a lot of reading and have spotted a few areas in the wiki that could use an update.

So far, I've added Robert T's post in the wiki: Fama and French Three-Factor Model (External links)

From various forum posts and in Malkiel's Random Walk Down Wall Street (what I'm reading now, excellent), I think it would be a good idea to include the transaction costs - a major influence on the success of this technique.

It's not part of the regression analysis per se, perhaps as another step in the process.
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.
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

LadyGeek wrote:From various forum posts and in Malkiel's Random Walk Down Wall Street (what I'm reading now, excellent), I think it would be a good idea to include the transaction costs - a major influence on the success of this technique.
By transaction costs, do you mean the costs paid by the funds? Or the commissions paid by the investor? I'm thinking the former.

Transaction costs of the funds should show up in the fund alphas, as should MER. However, alpha does tend to be the least precise of the figures generated by these regressions, so it is handy to look at these figures separately too, to make sure the alpha makes sense. Similarly a good idea to look at tracking error of a fund vs its index.

I do agree that it's a good idea to look at trading costs as part of the whole picture. If you consider only trading expenses of a fund, one may be higher - but the increased trading might be for the purpose of maintaining more consistent factor exposure, or avoiding negative momentum, etc, so the resulting alpha could still be closer to zero (less negative).

(As far as an individual's trading costs, those are certainly also a factor, but one that differs based on re-balancing choices, portfolio size, nationality, etc.)

Of course, just my opinions. I've still got plenty to learn! :)
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by LadyGeek »

Actually, I'm not quite sure. Even more opportunity to learn. :)

Here's one version: Mutual Funds: Additional Costs (under Transaction costs) It seems to include both.

I also see forum posts that refer to transaction costs as the "cost to do a transaction," i.e. trading costs. Then there's taxes, but perhaps that's stretching things too far.

Can someone help answer?
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.
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

LadyGeek wrote:Actually, I'm not quite sure. Even more opportunity to learn. :)

Here's one version: Mutual Funds: Additional Costs (under Transaction costs) It seems to include both.

I also see forum posts that refer to transaction costs as the "cost to do a transaction," i.e. trading costs. Then there's taxes, but perhaps that's stretching things too far.

Can someone help answer?
Well, as far as trading costs paid by the funds, IMO the key is to look at alphas compared to the 3F (or 4F) model, as well as tracking error with respect to the funds' underlying indexes. Both of these metrics should include the effects of both management expenses and any other costs of the funds, including trading. Tracking error will be more accurate, but alpha will also include how efficiently the index itself captures its factor exposure.

For an individual investor's transaction costs, I feel that's a separate issue that has more to do with minimizing unnecessary trading, choosing funds with tight spreads, paying attention to tax efficiency, and paying as little as possible in brokerage fees.
staythecourse
Posts: 6993
Joined: Mon Jan 03, 2011 8:40 am

Re: How to get Fama-French EAFE Factors, with results

Post by staythecourse »

LadyGeek wrote:Can someone help answer?
Different transaction costs that would need to be answered:

1. brokerage costs
2. bid/ask spreads
3. taxation on dividends

If your backtesting made up funds that might have existed at the time you would need to look at:
1. Loads
2. ER and 12b-1 fees (whenever they were started)
3. Turnover
4. Taxation

If you can figure this out you will have done more then DFA, Fama and French, and nearly every acadamecian working until now. That work would be worth MORE then the bogleheads wiki.

Great idea, but since no one has EVER gotten close to computing it I would doubt anyone here will figure it out.
"The stock market [fluctuation], therefore, is noise. A giant distraction from the business of investing.” | -Jack Bogle
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by LadyGeek »

:shock: So much for reading an academic text. I was hoping to include a one-line entry for "enter expenses here" and then do the math. I guess not.

Approximating expenses might not be productive, as the amount of uncertainty has to be less than the uncertainties used in the regression model. (Thinking like an engineer here.)

Could anything be done with trading costs, e.g. what you pay to buy/sell the funds?

I don't think I fully understand how this plays together yet, so my questions might be off.

I've got a page started in the wiki, but it's not ready for review. (Hint: Wiki activity can be found by clicking on the Wiki Content link in the left sidebar, then Recent changes. Wiki editors may edit directly.)
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.
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by LadyGeek »

Next question: Can someone please explain the T-Statistics? I see the math, but I don't have an understanding of how the numbers apply to the results. What's important to know?

Along a similar path, how precise do these results have to be? In the quote below, I see "-0.2" overestimated to "-.3," so I think to the nearest 0.1 is sufficient.
ClosetIndexer wrote:TD;DR:
While the method in the OP produces factors that can give good regression results (high R-squared) for EAFE funds, it is important to be aware of how the approximations used will affect the SmB and HmL factors, compared to the results one would theoretically get using the research factors:
It will slightly overestimate the magnitude* of the SmB factor for all funds.
It should be accurate for the HmL factor of large-cap funds, but will underestimate the magnitude* of HmL more and more as the cap-size of the funds being profiled decreases (aka as their SmB increases).

*This means the magnitude in either direction. So a fund with a 'true' SmB of -0.2 would be overestimated to, say, -0.3. One with a true SmB of 0.3 might be overestimated to 0.42.

Now if only someone with access to the underlying data would produce a set of research factors for EAFE (and Canada while they're at it), none of this would be necessary! Until then, hope this helps somebody. (It at least helped me understand the model better, going through the process!)
In the above post, what is "TD;DR;"?
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.
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

Lol, TD;DR is the typo version of TL;DR ('too long; didn't read') :D Fixed.

Your other two questions are linked. The t-values in the regression results are a measure of confidence. Specifically, they measure the ratio of the result to its standard error.

So a t-value of 1 (or -1 for a negative factor) means the standard error is equal to the magnitude of the value itself. For example, if we come up with an HmL of 0.3 with a t-value of 1, that means the standard error of that measurement is also 0.3, which means 68% of the time the true value is between 0.3-0.3 and 0.3+0.3, ie between 0 and 0.6. If the HmL result were again 0.3, but the t-value were 3, that would mean the standard error was 0.1, so 68% of the time the true value would be between 0.2 and 0.4. (I'm no statistician, so if someone else is, please feel free to correct me!)

Now, that's if the factor data you are using is accurate. In our case here, there is added uncertainty because the factor data we are using are approximations. (We don't have access to the underlying returns data, so we come as close to Fama-French's specifications for the factors as we can. HmL should be calculated for small and large portfolios and averaged together, but we simply take high BtM minus low BtM across the entire portfolio. For SmB, we're using MSCI index data where small covers the range from 2% to 16% of total market cap, rather than 0 to 10% as specified for the official factors.) These approximations offset the results, which is why I estimate the 'true' results based on the changes I know the approximations are making. How much uncertainty this adds depends on how accurate my assumptions are. I'd say its reasonable to add +-0.1 to our corrected estimate, on top of what the t-value tells us.

Does that make sense? Let me know what's unclear and I'll try and illustrate the process I'm going through better.


Edit: It's also important to keep an eye on R^2 ("R-squared"), which gives a measure of the overall fit of your linear model to the data. R^2=1 means the model perfectly describes the data. The lower R^2, the more unexplained movements there are in your returns data, which means greater uncertainty using your model going forward. As a rough guide, I've found that anything over about 0.97 is good. Over 0.95 is ok. Over 0.94 is useable. Low 90s can still provide some insight, but I would definitely want to consider it a rough estimate only, and consider other things, like the methodology of the index, tracking error, and average P/B, more closely. Anything below 0.90 isn't suitable to draw conclusions from, IMO.

Edit 2: Should note that these R^2 guidelines refer to regressions done over periods up to 10 years or so, and assume no changes in underlying methodology of the fund/index. For longer periods, factors will drift more, so R^2 will naturally be lower.
Last edited by ClosetIndexer on Wed Sep 19, 2012 6:57 am, edited 1 time in total.
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by LadyGeek »

OK, so when I got down to writing a top-level overview, I realized that I didn't understand the big picture. My first question: What is a factor? Engineers think of them as coefficients. Apparently, finance expands them into a whole lot more.

The answer resulted in a new wiki article: Factors (finance) (Anyone is welcome to supply comments.)

Here's what I have so far: Fama-French three-factor model analysis

- Using one paper and Bill Bernstein's Rolling Your Own: Three Factor Analysis, I built a table showing what you need to do the analysis.

- Data quality. See if your data is good enough to work with. I flagged your uncertainty (pun intended) for review, as I don't know the answer, either.

- Show how the Fama-French factors are used to derive an expected return. After a lot of reading, I found a single article that actually used a numeric example. Even then, I had to derive the equation from the text description. :shock:

- Show the equation used for alpha. Here's where Bill Bernstein's article and your R script differ. Excel's output will generate the intercept point, which is alpha. Your R script suppresses the intercept data - alpha is missing. At least that's what I think happens, but I can see it in your posted examples. No alpha.

- To do: The tutorial I thought I was going to write in the first place. :)

=====================

I found some good websites. - take a look at the Software section under "External links." The author of the Three Factor Rolling Regression Viewer is none other than Bogleheads forum member mas. :) I'll PM him for some expert help.
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.
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

My script does calculate the intercept, and I handily multiply it by 12 and pretty-print it as a percentage so you can see the annual alpha as well as the less obviously meaningful monthly intercept.

For the 'what is a factor' question, I'd suggest reading about the CAPM model first, then Fama and French's 1992 paper that started the whole 'multi-factor' method of modelling portfolio returns (and perhaps an easier read, their '1992a' paper, Common risk factors in the returns on stocks and bonds. In short, factors are variables we can use to create a linear model to describe the returns of a portfolio. In the CAPM model, a portfolio's returns can be described reasonably well based on its exposure to one factor: beta. Fama and French expand on this to show that the addition of two additional factors, HmL and SmB, result in a linear model that fits actual portfolio returns more closely. Reading the paper itself should provide more insight.

As far as how the factors are used to derive an expected return, see my recent post here: http://www.bogleheads.org/forum/viewtop ... 2#p1484572

Basically, first you figure out the factor loadings of a fund, using a linear regression. The future returns of that fund, assuming the factors remain constant, should be described by the sum of the returns of the various factors, times the exposures the fund has to each of those factors, and plus the risk free rate and any alpha.

So, if you had a fund with factor exposures like this

Code: Select all

                                    Mkt-Rf    SmB     HmL     Alpha   

        MSCI US SV                   0.87    0.63    0.50     0.05%
           t-values                 22.04   11.82    7.33     0.03

        R^2 = 0.941
And the factor returns for a given year were these

Code: Select all

        Mkt-Rf   SmB   HmL   Rf (%)
2007   2.65   -8.22   -12.04   4.67
then the expected return for that fund in that year would be
(0.87 * 2.65 + 0.63 * (-8.22) + 0.50 * (-12.04) + 4.67 + 0.05)% = -4.17%

Beyond that, give me a few days. Once I've got the final questions in my own portfolio ironed out, I'm planning a thread describing the entire process I went through to build it, including all the analysis, results, and decisions along the way. Should help illustrate how one would actually use this sort of data. Feel free to PM me too if you still have questions after looking at those papers.
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by LadyGeek »

I'll look at your script more closely. My error, I removed the note in the wiki.

As for factors, my employer provides free (they pay for) online access to a book subscription service. I've got full run of several deep-dive finance texts. The content in Factors (finance) was paraphrased from "Equity Valuation and Portfolio Management" by Fabozzi and Markowitz (with proper citations). It was the clearest explanation I could find from an authoritative source - Wikipedia has nothing of significance.

This text gave me enough of a background that, for the first time, I could actually understand what the Fama and French papers were discussing. However, I don't have the experience of working with the data.

I did see reasons why Fama-French factors are used (and why it's better than CAPM), but I put it on low priority - to get to later. Your explanation is simple and clear. I'll incorporate your explanation in Factors (finance). Once that's done, I'll remove the "Please help improve this article" banner at the top. I'll also take a look at putting those comments in the CAPM and Fama-French Three-factor articles as well.

I think I'm getting the hang of this.

Do you have a feel for the t-value confidence levels? I'm looking for similar good, bad, and ugly thresholds that you gave for R^2. For example: is 22.04 good?

It bugged me why the Fama-French examples I used didn't include alpha. But, when used for fund managers, it just showed up (Bill Bernstein used alpha in his article).

I had no idea why this was done, as the original Fama - French papers included some form of alpha (another term in the equation to account for this effect - I think). Since you include alpha, this tells me it's the right way to go. After you get your end-to-end tutorial up and running, I'll replace the Fama article example with yours.
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.
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

Absolutely I feel alpha is an extremely important part of a fund analysis. One of the most important really. Doesn't much matter what factor exposures a fund has if it doesn't capture them efficiently. (Much like it doesn't matter how good an index is if the fund tracking it has a massive tracking error.) Unfortunately, alpha is also the most imprecise output from the regressions, so the output tends to only really be useful when there is a good length of data - at least a few years of monthly data, ideally more like ten. (Like this recent example.)

As for t-values, I can't really give a good/bad/ugly like with R^2, since their meaning and use is different. (And the cutoffs I mentioned for R^2 are just rough estimates from my own personal experience, and are specific to these sorts of investment fund regressions BTW.) For t-values, the best I can say is basically what I described before:
ClosetIndexer wrote:The t-values in the regression results are a measure of confidence. Specifically, they measure the ratio of the result to its standard error.

So a t-value of 1 (or -1 for a negative factor) means the standard error is equal to the magnitude of the value itself. For example, if we come up with an HmL of 0.3 with a t-value of 1, that means the standard error of that measurement is also 0.3, which means 68% of the time the true value is between 0.3-0.3 and 0.3+0.3, ie between 0 and 0.6. If the HmL result were again 0.3, but the t-value were 3, that would mean the standard error was 0.1, so 68% of the time the true value would be between 0.2 and 0.4.
So the higher the t-value the better, but there's no level at which it's too low to be useful at all. Double the t-value, and the standard error is cut in half. Half the t-value, and the standard error is doubled. That link on standard error should help with understanding what that means as far as our confidence in the estimate.

So in the end, the estimated value along with its t-value give you a range that you can be fairly confident the true value fell in, over the samples tested. (Of course, the factors of funds do drift over time, some more than others. That's why I like to run regressions over multiple periods to check for consistency. Shifting loadings aren't necessarily a problem, but they are important to be aware of when you're basing allocation decisions on what you believe to be the best estimate of a fund's factor loadings.)

What I'm planning to write will be a complete run through of my personal portfolio construction process, rather than a generalized tutorial. Still, I hope it will provide a useful real world example of how a person might use this data to make (at least theoretically) rational decisions. I expect I'll learn a few things too, both in the process of writing in it, and hopefully in resulting suggestions from others! (One thing though is that parts of it will be specific to Canadian investors - stuff regarding the Canadian allocation, and certain tax issues. I'm thinking of posting the whole thing on financialwebring.org then linking from bogleheads with excerpts and some added notes.
steve_14
Posts: 1507
Joined: Wed Jun 20, 2012 12:05 am

Re: How to get Fama-French EAFE Factors, with results

Post by steve_14 »

ClosetIndexer wrote:Absolutely I feel alpha is an extremely important part of a fund analysis. One of the most important really. Doesn't much matter what factor exposures a fund has if it doesn't capture them efficiently. (Much like it doesn't matter how good an index is if the fund tracking it has a massive tracking error.)
Kind of makes you wonder how much all those folks who wrote books promoting this method of investing over the past decade really understood it, given that no investing book I'm aware of published before 2012 actually mentions alpha.
User avatar
stevewolfe
Posts: 1672
Joined: Fri Oct 10, 2008 7:07 pm

Re: How to get Fama-French EAFE Factors, with results

Post by stevewolfe »

steve_14 wrote:Kind of makes you wonder how much all those folks who wrote books promoting this method of investing over the past decade really understood it, given that no investing book I'm aware of published before 2012 actually mentions alpha.
"The Intelligent Asset Allocator" by William Bernstein (Copyright 2001) discusses alpha on pages 88-90. Page 98 has a spotlight on performing a three factor regression analysis on an actively managed fund to determine if the manage has produced positive alpha (and, for this part of the discussion in the book, determining manager skill).
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by LadyGeek »

ClosetIndexer wrote:Absolutely I feel alpha is an extremely important part of a fund analysis. One of the most important really. Doesn't much matter what factor exposures a fund has if it doesn't capture them efficiently. (Much like it doesn't matter how good an index is if the fund tracking it has a massive tracking error.) Unfortunately, alpha is also the most imprecise output from the regressions, so the output tends to only really be useful when there is a good length of data - at least a few years of monthly data, ideally more like ten. (Like this recent example.)
Via PM, I received some useful suggestions from mas. The alpha term should be included in all of the equations. If it's not used, set it to 0. A total stock market fund should have beta = 1.0, the SMB and HML coefficients = 0.0, and a slightly negative alpha which accounts for expense ratio and transaction costs. He's done the analysis here: Rolling Regression Results Gallery (VTSMX).
ClosetIndexer wrote:As for t-values, I can't really give a good/bad/ugly like with R^2, since their meaning and use is different. (And the cutoffs I mentioned for R^2 are just rough estimates from my own personal experience, and are specific to these sorts of investment fund regressions BTW.) For t-values, the best I can say is basically what I described before...
I found some background info and added it to the wiki article. The t-statistics depend on the number of data points. With a sufficiently large dataset, the t-distribution approaches a normal distribution. I don't know how many points your dataset contains, but anything over 120 can be considered a normal distribution. Anything less should use the table. (A t-distribution is a normal distribution for a small number of samples.) When you mention 68% of the time, I think you're referring to a large number of data points (1 standard deviation with a normal distribution). I'm not a statistics expert, feel free to correct.

Wiki article link: Fama-French three-factor model analysis
ClosetIndexer wrote:What I'm planning to write will be a complete run through of my personal portfolio construction process, rather than a generalized tutorial. Still, I hope it will provide a useful real world example of how a person might use this data to make (at least theoretically) rational decisions. I expect I'll learn a few things too, both in the process of writing in it, and hopefully in resulting suggestions from others! (One thing though is that parts of it will be specific to Canadian investors - stuff regarding the Canadian allocation, and certain tax issues. I'm thinking of posting the whole thing on financialwebring.org then linking from bogleheads with excerpts and some added notes.
There's no sense doing duplicate work. Go ahead and post the whole thing on financialwebring.org, link back to here. The wiki article will eventually appear in finiki, the Canadian financial Wiki :wink:

I've found that working with a multinational perspective is very enlightening. Differences in market compositions (TSX/composite is not diversified), tax laws, and regulations, force you to see things from a different perspective.
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.
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

All sounds right to me! Note that 120 samples equates to 10 years of data (monthly), which is part of why I prefer to do regressions over at least that period. The thing is though that you're just really looking for a rough range for the values, since this is all educated guesswork anyway, so I don't find myself looking up t-test distributions or anything like that. I just keep in mind that things get less precise as the sample size decreases. (That said, nothing wrong with understanding the basis and specifics behind that!)
steve_14
Posts: 1507
Joined: Wed Jun 20, 2012 12:05 am

Re: How to get Fama-French EAFE Factors, with results

Post by steve_14 »

stevewolfe wrote:
steve_14 wrote:Kind of makes you wonder how much all those folks who wrote books promoting this method of investing over the past decade really understood it, given that no investing book I'm aware of published before 2012 actually mentions alpha.
"The Intelligent Asset Allocator" by William Bernstein (Copyright 2001) discusses alpha on pages 88-90. Page 98 has a spotlight on performing a three factor regression analysis on an actively managed fund to determine if the manage has produced positive alpha (and, for this part of the discussion in the book, determining manager skill).
I meant more with regard to picking index value/growth funds. Of course back then we didn't have much real world data to analyze.
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by LadyGeek »

^^^^ Bill Bernstein has an excellent article on calculating alpha; I use it in the wiki article: Rolling Your Own: Three Factor Analysis. It's from 2001, so I suspect it's linked to the book somehow.

To avoid complications regarding terminology, I redid all the equations to conform with the variable names used in the Fama-French papers. I'm also referencing a textbook by a respected author, Frank Fabozzi (thanks to my employer which pays for an online book subscription service). For me, he clearly explains the concepts and shows detailed examples.

Wiki article link: Fama-French three-factor model analysis

What I got from this textbook is that there is another term ε (epsilon) which is attributed to company-specific risk (unsystematic risk). This term is not specific to Fama-French, but is one used in the general concept of factor models. Thinking about this further, when you do the regressions, the term can't be separated from alpha (active return, related to the tracking error against the benchmark) - they're both constants. I think this is why it's not possible to extract the transaction costs. I could be wrong here...

====================
I found the ClosetIndexer's tutorial in the financialwebringforum.org and posted in his thread: Thorough factor analysis of Can TSM, Value, and Small ETFs (We still need help figuring out how to interpret the statistics metrics, R^2 and t-values.)

It's worthwhile to read, as you can see how the analysis is used to find questionable funds and arrive at a workable compromise.
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.
steve_14
Posts: 1507
Joined: Wed Jun 20, 2012 12:05 am

Re: How to get Fama-French EAFE Factors, with results

Post by steve_14 »

Thanks for the link, LadyGeek. He did indeed!
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by LadyGeek »

LadyGeek wrote:What I got from this textbook is that there is another term ε (epsilon)... I could be wrong here...
I was wrong. :) Thanks to camontgo for providing an excellent explanation (via PM) which is now in the wiki. He also provided rationale for removing the R^2 threshold levels (good, bad, ugly criteria) - see the footnote. I cleaned up the "Data quality" section.

After more reading, I revised alpha to be aligned with the Fabozzi textbook (with help from a Dartmouth paper). I'm going to expand on the details in the Fama and French Three-Factor Model article, as actively managed portfolios are the subject of much discussion.

Wiki article link: Fama-French three-factor model analysis

Comments / questions / corrections are welcome.
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.
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

camontgo (of calculatinginvestor.com!) is definitely far more knowledgeable regarding the math behind this stuff than I am. Glad he caught that issue with guidelines on the R^2s. I actually edited my post above to add a caveat about that earlier, but I think you're right that it's best not to include misleading guidelines at all (beyond 'higher is better'). I suppose you could say something along the lines of, "When comparing several portfolios over the same number of samples, the ones with higher R^2 are explained more completely by the linear model," without giving specific numbers... At least that would give an idea of how the number would be used in this 'real world' application.
Last edited by ClosetIndexer on Wed Sep 19, 2012 10:01 pm, edited 1 time in total.
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by LadyGeek »

I agree, and have incorporated your comments.

Wiki article link: Fama-French three-factor model analysis
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.
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by LadyGeek »

I updated the wiki article Fama and French Three-Factor Model to give a better description of alpha and its use to determine active fund management costs (under "Evaluating fund managers"). Can someone please give it a sanity check?

Also, I revised the equations to be consistent with Fama-French three-factor model analysis.
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.
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by LadyGeek »

I updated Factors (finance) with comments from mas and ClosetIndexer.
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.
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by LadyGeek »

To further complete the picture, I added tracking error to Fama and French Three-Factor Model. Please review.

The content is not specific to Fama-French, so I added it under the "Notes" section. My intention was to describe the two types of tracking error models: What's used for historical fund manager performance shouldn't be used to predict future performance (a lot more complicated).

I stopped short of giving a numerical example, as that would be too much detail for the article. If it's useful, we could put an example in Fama-French three-factor model analysis, but that not be in-scope for the regression analysis (?).
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.
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

Here's an updated version of the factors calculated using this method (csv format):

Code: Select all

,Mkt-Rf,SmB,HmL,RF
199007,1.52,,-0.15,0.68
199008,-10.94,,-1.02,0.66
199009,-14.42,,1.87,0.6
199010,15.63,,-3.77,0.68
199011,-6.62,,1,0.57
199012,0.79,,-2.23,0.6
199101,2.72,,-0.07,0.52
199102,9.48,,5.37,0.48
199103,-5.69,,-0.66,0.44
199104,1.1,,-0.1,0.53
199105,-1,,-0.57,0.47
199106,-7.33,,0.17,0.42
199107,4.45,,1.4,0.49
199108,-2.57,,-0.99,0.46
199109,5.05,,1.13,0.46
199110,1.33,,-3.02,0.42
199111,-4.76,,-2.12,0.39
199112,4.7,,-1.04,0.38
199201,-3.03,,3.95,0.34
199202,-3.61,,0.31,0.28
199203,-6.79,,1.12,0.34
199204,-0.03,,7.25,0.32
199205,6.28,,-0.88,0.28
199206,-4.78,,0.1,0.32
199207,-2.96,,-3.82,0.31
199208,6.05,,-5.59,0.26
199209,-2.69,,0.3,0.26
199210,-5.8,,0.24,0.23
199211,1.11,,-0.45,0.23
199212,0.14,,0.91,0.28
199301,-0.02,1.8693669937,3.38,0.23
199302,2.85,1.4594488238,0.27,0.22
199303,9.57,1.8037349383,5.72,0.25
199304,8.79,2.1853041506,3.61,0.24
199305,2.33,4.5295130074,-1.6,0.22
199306,-3.36,-3.8479055576,-1.27,0.25
199307,4.48,0.8127377487,1.43,0.24
199308,5.03,0.3512133491,0.86,0.25
199309,-3.13,0.019104878,0.5,0.26
199310,2.15,-3.3844254076,-0.23,0.22
199311,-8.65,-1.2410923804,1.62,0.25
199312,6.56,1.4498826468,3.53,0.23
199401,9.9,2.7445923503,3.21,0.25
199402,-1.08,2.2937098951,1.68,0.21
199403,-3.7,2.17044348,1.79,0.27
199404,3.59,0.093933228,0.77,0.27
199405,-1.06,-0.5729502474,1.79,0.32
199406,1.55,0.7571371843,1.03,0.31
199407,0.45,-0.6541441161,-0.51,0.28
199408,1.75,-1.2461712791,-0.19,0.37
199409,-3.51,-0.3264805565,0.28,0.37
199410,3.28,-1.787698085,0.65,0.38
199411,-5.64,-2.0481709969,-0.71,0.37
199412,0.63,0.79595299,-1.06,0.44
199501,-4.44,0.1332569956,-0.04,0.42
199502,-0.78,-0.9064503434,-0.49,0.4
199503,5.2,-4.0332542851,-0.57,0.46
199504,3.8,-1.3971470445,-2.06,0.44
199505,-1.84,-0.9841418727,0.7,0.54
199506,-1.99,-0.3349450572,0.3,0.47
199507,5.36,0.3220806583,-0.46,0.45
199508,-3.55,1.1330390053,-0.08,0.47
199509,0.65,-2.0921553489,-0.9,0.43
199510,-2.13,-1.1813605109,-0.15,0.47
199511,1.83,-1.4242592411,0.28,0.42
199512,3.37,-0.4848581657,0.85,0.49
199601,-0.14,2.7372311396,1.79,0.43
199602,0.17,0.7260658721,0.89,0.39
199603,1.52,-0.0102168598,-0.72,0.39
199604,2.43,2.4672832958,1.01,0.46
199605,-1.48,-0.0642236881,-0.73,0.42
199606,0.16,-0.5585423872,-0.61,0.4
199607,-3.37,-2.4812903261,0.23,0.45
199608,-0.49,0.2503369587,0.22,0.41
199609,2.56,-2.0567675477,-0.17,0.44
199610,-1.75,-0.6277609441,-0.27,0.42
199611,3.63,-2.5659836776,2.57,0.41
199612,-1.46,-2.0349573809,1.37,0.46
199701,-4.22,1.6079504962,0.94,0.45
199702,1.17,0.3624678254,0.13,0.39
199703,0.22,-2.7516887606,-0.8,0.43
199704,0.05,-3.7318727065,-2.62,0.43
199705,6.71,1.2040921255,0.99,0.49
199706,5,-3.5876849894,-1.87,0.37
199707,1.3,-5.0198992423,0.1,0.43
199708,-7.53,1.2018803899,1.77,0.41
199709,5.17,-7.5864871453,-1.78,0.44
199710,-7.61,2.2694448294,3.15,0.42
199711,-1.4,-5.8229283674,-4.04,0.39
199712,-0.07,-6.5734425386,-0.44,0.48
199801,4.47,1.807409329,6.91,0.43
199802,5.97,2.672600701,4.37,0.39
199803,2.66,-0.7466918721,1.64,0.39
199804,0.3,-0.8643238682,-0.5,0.43
199805,-0.15,0.9295421379,-0.24,0.4
199806,-0.16,-4.6167339813,-2.77,0.41
199807,0.56,-2.2618229553,-0.62,0.4
199808,-12.44,-0.7994081923,-3.04,0.43
199809,-2.83,-1.394430087,-1.99,0.46
199810,9.79,-2.2698105152,-0.79,0.32
199811,4.75,-0.2184761536,0.23,0.31
199812,3.27,-3.437582039,-2.44,0.38
199901,-0.6,-0.8457819215,-1.58,0.35
199902,-2.81,1.2399811552,2.51,0.35
199903,3.9,1.3633449892,5.98,0.43
199904,4.12,3.4998726108,7.63,0.37
199905,-4.89,1.6363610051,-0.93,0.34
199906,3.42,1.5347958159,2.99,0.4
199907,2.96,0.2032757573,2.44,0.38
199908,0.87,1.8742453287,2.12,0.39
199909,0.71,-1.2051857247,-0.34,0.39
199910,3.6,-5.0043764806,-3.86,0.39
199911,4.38,-2.8259271768,-7.91,0.36
199912,8.86,-7.9514980107,-5.39,0.44
200001,-6.98,6.5638971595,0.72,0.41
200002,2.44,-0.1375934364,-12.76,0.43
200003,3.17,-1.4090382351,6.85,0.47
200004,-5.89,-2.1581475643,3.82,0.46
200005,-3.66,2.3354366989,8.29,0.5
200006,3.64,4.9935747111,3.6,0.4
200007,-4.55,-2.0020815946,1.55,0.48
200008,0.88,3.9581265828,-0.78,0.5
200009,-5.11,-0.7799318932,3.53,0.51
200010,-3.4,-4.4915301232,3.27,0.56
200011,-3.59,3.5528932657,7.59,0.51
200012,2.55,-3.1197318566,3.2,0.5
200101,-0.18,3.0611497484,3.25,0.54
200102,-7.92,5.4994831063,4.67,0.38
200103,-7.57,-0.9261825866,0.15,0.44
200104,7.02,1.5749712045,-0.06,0.39
200105,-3.71,3.2941338291,-0.03,0.32
200106,-4.28,1.1244839724,0.31,0.28
200107,-2.5,-2.2916913973,1.5,0.3
200108,-2.82,3.6057586366,2.99,0.31
200109,-10.35,-3.2371532242,-4.85,0.28
200110,2.78,2.5598024904,0.84,0.22
200111,3.35,0.167594235,1.46,0.17
200112,0.06,-3.0499916892,-0.68,0.15
200201,-5.04,2.9575254286,-1.71,0.14
200202,0.64,0.9621778569,-1.52,0.13
200203,5.66,1.1954064509,1.4,0.13
200204,0.61,2.9462774778,-0.61,0.15
200205,1.42,3.1281450885,4.09,0.14
200206,-4.21,0.235425777,-2.75,0.13
200207,-9.09,2.153604315,1.62,0.15
200208,-0.53,-1.0725449737,0.19,0.14
200209,-10.7,2.9469089882,-2.89,0.14
200210,5.35,-6.6167653429,-1.07,0.14
200211,4.54,-1.6692524223,3.82,0.12
200212,-3.14,1.4982868927,0.31,0.11
200301,-3.82,2.7547435723,3.61,0.1
200302,-2.37,1.4929289716,-0.15,0.09
200303,-1.96,1.37706204,-1.94,0.1
200304,9.56,-0.4245091749,3.53,0.1
200305,6.34,2.4657093823,4.81,0.09
200306,2.26,2.8936848825,1.61,0.1
200307,2.43,0.6784979796,3.64,0.07
200308,2.41,4.176876539,2.04,0.07
200309,3.45,2.9015733715,3.44,0.08
200310,6.26,1.7493868161,3.9,0.07
200311,2.08,-2.6436922123,-0.09,0.07
200312,7.51,-1.7164056123,0.71,0.08
200401,1.71,3.6634264671,2.07,0.07
200402,2.36,0.5235969124,1.3,0.06
200403,0.52,3.5431496357,0.68,0.09
200404,-2.39,-0.8214089913,0.51,0.08
200405,0.12,-1.9926179631,0.17,0.06
200406,2.01,3.1863554112,1.44,0.08
200407,-3.25,-1.2777149053,0.81,0.1
200408,0.14,0.2651806943,0.87,0.11
200409,2.42,-0.2345125223,1.24,0.11
200410,3.39,0.3457425939,0.64,0.11
200411,6.74,1.1139963949,1.62,0.15
200412,4.46,0.6081787497,0.9,0.16
200501,-1.57,3.8053374118,1.1,0.16
200502,3.88,-0.3159560055,0.45,0.16
200503,-2.76,0.8947292136,0.94,0.21
200504,-2.71,-0.0604988509,-2,0.21
200505,-0.71,-0.1020245909,0.08,0.24
200506,1.57,1.3503615088,-0.42,0.23
200507,2.97,1.4509314365,1.87,0.24
200508,2.73,0.3525562353,0.08,0.3
200509,3.37,-0.3413935297,-0.17,0.29
200510,-2.61,-0.3486860718,1.62,0.27
200511,1.8,0.7567585913,-1.1,0.31
200512,4.82,3.3445971499,-1.09,0.32
200601,5.75,0.6828457689,-0.34,0.35
200602,-0.46,-1.0083203776,3.04,0.34
200603,2.94,1.519744624,-0.42,0.37
200604,4.4,-0.4614814116,-0.2,0.36
200605,-4.85,-1.922220694,-0.1,0.43
200606,-0.33,-2.1979540258,-1.08,0.4
200607,0.71,-3.6967811525,1.26,0.4
200608,2.14,0.0341170896,0.26,0.42
200609,-0.06,0.2458983565,2.06,0.41
200610,3.63,-0.2256303011,1.26,0.41
200611,2.52,1.2828883622,0.65,0.42
200612,2.97,0.1979963733,1.73,0.4
200701,0.34,1.4095992936,1.11,0.44
200702,0.5,0.8603963477,-0.51,0.38
200703,2.42,0.660322116,-0.77,0.43
200704,3.68,-0.7625234345,-0.04,0.44
200705,1.42,-1.1223331931,0.72,0.41
200706,-0.06,-0.338895046,-0.53,0.4
200707,-2.03,1.0825922513,-2.26,0.4
200708,-2.03,-3.6048223248,-1.15,0.42
200709,5.08,-4.2333169497,-1.73,0.32
200710,3.73,2.302389303,0.14,0.32
200711,-3.92,-3.9240073863,-3.91,0.34
200712,-2.59,-1.2128030574,1.07,0.27
200801,-8.76,-0.5278307816,1.5,0.21
200802,1.77,2.7980520916,-2.93,0.13
200803,-1.4,0.7391222511,2.21,0.17
200804,5.08,-3.1398456538,-1.2,0.18
200805,1.22,0.600739637,-3.91,0.18
200806,-7.96,-0.016271979,-3.73,0.17
200807,-3.67,-1.3578938818,2.59,0.15
200808,-4.42,0.1995682824,1.02,0.13
200809,-14.05,-2.6915921093,3.18,0.15
200810,-20.09,-3.5607191294,-2.03,0.08
200811,-5.43,0.9447654423,-1.61,0.03
200812,6.52,0.8226157972,-1.01,0.09
200901,-9.69,3.402454048,-3.18,0
200902,-9.93,0.9811078153,-1.93,0.01
200903,5.82,0.1768823648,5.38,0.02
200904,12.34,2.4774762274,17.03,0.01
200905,12.77,2.2768797688,2.62,0
200906,-0.2,2.5144014269,-0.41,0.01
200907,9.03,-1.5005797172,4.84,0.01
200908,4.77,2.6433615121,6.01,0.01
200909,4.16,1.1578928701,0.11,0.01
200910,-1.42,-0.5202262478,-2.03,0
200911,2.08,-2.0133688876,-1.54,0
200912,1.22,-0.6891584965,-0.12,0.01
201001,-3.66,3.4191129823,0.49,0
201002,-0.96,-0.7112213427,-0.18,0
201003,6.36,1.0448861527,1.37,0.01
201004,-0.94,3.4787817822,1.94,0
201005,-11.22,-1.0066616769,-1.7,0.01
201006,-0.97,0.578914304,-1.41,0.01
201007,9.57,-0.8525246992,2.13,0.01
201008,-3.31,0.2081849533,-1.01,0.01
201009,10.12,1.6021911429,-0.06,0.01
201010,4.06,0.2487135843,0.04,0.01
201011,-4.88,1.2226084665,-1.92,0.01
201012,8.39,3.5474564013,1.89,0.01
201101,2.21,-1.5778468356,4.74,0.01
201102,3.05,-1.0482032677,0.2,0.01
201103,-1.69,2.1334723363,-3.07,0.01
201104,6.16,-0.7612635741,-1.09,0
201105,-2.65,0.3362762552,-2.03,0
201106,-1.2,-0.4083387746,-0.37,0
201107,-1.34,0.535326804,-2.78,0
201108,-8.89,0.8084840366,-3.12,0.01
201109,-9.72,-0.8351980152,0.73,0
201110,8.51,-2.4459848835,-1.05,0
201111,-4.69,-0.5751023912,-1.93,0
201112,-1.25,-0.9668948704,0.3,0
201201,6.06,2.9245538074,4.2,0
201202,5.77,0.2472247456,0.91,0
201203,-0.16,0.5327888251,-2.24,0
201204,-1.65,1.4089399961,-4.05,0
201205,-11.23,-0.3292737807,-3.29,0.01
201206,6.17,-2.9663402196,2.02,0
201207,0.89,-0.7593472695,-2.12,0
201208,2.78,-0.0134337484,1.3,0.01
201209,3.23,1.7218672135,2.35,0.01
201210,0.99,0.01653638,1.49,0.01
201211,2.06,-1.4680715447,-0.68,0.01
201212,3.4,0.9360930045,3.07,0.01
Just keep in mind the caveats here.
pradador
Posts: 156
Joined: Thu Jun 14, 2012 9:20 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by pradador »

Just visited the Fama-French Data Library to update the www.fundfactors.com website and noticed there's a new dataset for the Global Ex-Us Factors. I wonder how long it's been available for and how it compares to these factor returns? EAFE + Canada shouldn't be too different from EAFE.
User avatar
Topic Author
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: How to get Fama-French EAFE Factors, with results

Post by ClosetIndexer »

pradador wrote:Just visited the Fama-French Data Library to update the www.fundfactors.com website and noticed there's a new dataset for the Global Ex-Us Factors. I wonder how long it's been available for and how it compares to these factor returns? EAFE + Canada shouldn't be too different from EAFE.
New to me! Should work alright, especially for evaluating ex-us funds that include Canada obviously, but ya likely for EAFE too. Give it a try!
Post Reply