Small Cap Value heads Rejoice !!!

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Locked
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

Uncorrelated wrote: Tue Nov 05, 2019 2:53 pm
305pelusa wrote: Tue Nov 05, 2019 11:10 am
YRT70 wrote: Tue Nov 05, 2019 11:02 am
Ah sorry, I did not notice that was the case with ZPRV. I just saw it has an MSCI index, which is easy to get data for.

When I get a chance, I'll regress it and compare values. Some things that matter to me are R^2 values and alphas, which I didn't see for ZPRV in your comment. Also, I'm not necessarily trustful of "a Dutch guy online".

Regardless, provided their values are correct, I think you came to a reasonable conclusion.
I suspect I'm the "Dutch guy online". Here are my results with the momentum factor included:

IJS daily data (actual fund performance)

Code: Select all

|                                                    |annualized alpha|beta    |SmB     |HmL     |RmW     |CmA     |MoM      
:--|:--|:--|:--|:--|:--|:--|:--
|IJS, USA small cap value weighted                           |  -0.19%|    1.04|    0.92|    0.21|    0.17|    0.06|   -0.06|
|t stat                                                      |   -0.07|    4.04|   48.06|   10.57|    6.58|    1.94|   -4.86|

expected outperformance: 1.56%
Based on 18.8 years of data, 263 data points per year
R^2 adj = 0.376

ZPRV daily data (index)

Code: Select all

|                                                    |annualized alpha|beta    |SmB     |HmL     |RmW     |CmA     |MoM      
:--|:--|:--|:--|:--|:--|:--|:--
|MSCI USA small cap value weighted                           |  -1.82%|    1.03|    0.68|    0.26|    0.07|    0.16|   -0.12|
|t stat                                                      |   -1.41|    5.50|   76.99|   23.37|    5.36|    9.18|  -17.90|

expected outperformance: 1.25%
Based on 4.9 years of data, 308 data points per year
R^2 adj = 0.889

ZPRV monthly data (index)

Code: Select all

|                                                    |annualized alpha|beta    |SmB     |HmL     |RmW     |CmA     |MoM      
:--|:--|:--|:--|:--|:--|:--|:--
|MSCI USA small cap value weighted                           |   0.30%|    1.05|    0.75|    0.29|    0.30|    0.07|   -0.21|
|t stat                                                      |    0.35|    2.66|   31.25|    9.35|    8.87|    1.54|  -14.33|

expected outperformance: 1.17%
Based on 24.4 years of data, 12 data points per year
R^2 adj = 0.849
Source: own calculations based on MSCI index data and Ken French data library. Expected outperformance assumes that factors with t-stat below 2 are insignificant (equal to zero) and the remaining factors get 50% of the historical factor performance. Expected outperformance is before fees / turnover / tax inefficiencies.

The R^2 on IJS is atrocious. Not sure what's up with that. The factors loadings themselves are almost identical to portfolio visualizer's numbers
Nice work Dutch guy ;)

What could the low R^2 of IJS mean?
User avatar
Steve Reading
Posts: 2959
Joined: Fri Nov 16, 2018 9:20 pm

Re: Small Cap Value heads Rejoice !!!

Post by Steve Reading »

YRT70 wrote: Tue Nov 05, 2019 11:14 am
305pelusa wrote: Tue Nov 05, 2019 11:10 am Ah sorry, I did not notice that was the case with ZPRV. I just saw it has an MSCI index, which is easy to get data for.

When I get a chance, I'll regress it and compare values. Some things that matter to me are R^2 values and alphas, which I didn't see for ZPRV in your comment. Also, I'm not necessarily trustful of "a Dutch guy online".

Regardless, provided their values are correct, I think you came to a reasonable conclusion.
That would be great. And yes I agree, I have no way of knowing that what the Dutch guy calculated is 100% correct.

If true, I'm surprised that the expected outperformance of IJS is that 'much' higher, even though it has a weaker value tilt. I guess, if true, it would mean size is a lot more important.

Making things more interesting for me: IJS probably also has a benefit for me in dividend withholding taxes, giving it another ±0.2% edge.
Aight I promised you some data. I personally like to regress underlying index data and then assume the fund/etf itself will track that index well. This typically gives me more data for regression since the index has more length than the funds themselves. That's what I did here:

IJS (S&P 600 SCV) from May 1997 to Aug 2019 (monthly)
Alpha = 0.05 (p-value of 0.58, not statistically significant)
Beta = 0.96
SmB = 0.63
HmL = 0.6
MOM = -0.06

Had an R^2 of 0.964. All factors were statistically significant.

ZPRV (MSCI USA SC Value Weighted) from May 1997 to Aug 2019 (monthly):
Alpha = 0.19 (p-value of 0.02 so is statistically significant)
Beta = 1
SmB = 0.59
HmL = 0.54
MOM = -0.19

Had an R^2 of 0.974. All factors were statistically significant. The alpha seems to imply exposure to other factors (RmW and CmA) so I'll take it into account below.

Prediction
Assuming the following premiums (arbitrary 50% cuts for all factors except beta, where I just had a 25% reduction)
Beta = 6 (vs 8.3 historical)
SmB = 1.5 (vs 3.3 historical)
HmL = 2.4 (vs 4.8 historical)
MOM = 4.8 (vs 9.6 historical)

IJS expected return = 0.96*6 + 0.63*1.5 + 0.6*2.4 - 0.06*4.8 - 0.25 (ER) = 7.6%
ZPRV expected return = 1*6 + 0.59*1.5 + 0.54*2.4 - 0.19*4.8 + 0.19*12 (monthly alpha) - 0.3 (ER) = 9.2%

Thoughts
You'll see the above results for the S&P 600 SCV index are fairly different than if you just regress IJS itself (say with PV). That's because IJS did not track the index as well during its early years and because the index has gotten less value-y and smaller in recent time (which is exactly when you'd regress IJS). You'll have to come to your own conclusions there.

As for ZPRV, I'm really confused as to why my results are so different than Uncorrolated's. I used monthly index data from MSCI and the 3F (+MOM) from Ken. Perhaps it's just because he used the full 5 factors (6 it looks like? 0_o).

Just thinking out loud but your R^2 values look kinda poor. IJS for sure but even 0.89 on ZPRV implies a poor fit. Could you have a small error on your end?
"... so high a present discounted value of wealth, it is only prudent for him to put more into common stocks compared to his present tangible wealth, borrowing if necessary" - Paul Samuelson
Topic Author
hdas
Posts: 1395
Joined: Thu Jun 11, 2015 8:24 am

[Deleted]

Post by hdas »

[Deleted]
Last edited by hdas on Wed Jan 29, 2020 7:14 pm, edited 1 time in total.
....
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: Small Cap Value heads Rejoice !!!

Post by Uncorrelated »

305pelusa wrote: Thu Nov 07, 2019 6:06 pm Had an R^2 of 0.974. All factors were statistically significant. The alpha seems to imply exposure to other factors (RmW and CmA) so I'll take it into account below.
That sounds fishy. I would ignore the alpha for this fund, the fund doesn't do any special market timing that can explain the alpha.

Regarding the low R^2, I'm probably doing something wrong. The odd part is that my numbers for IJS are much closer to portfoliovisualizer numbers than yours. Comparing my IJS numbers to PV (6 factors daily), the beta, HML and MOM exposure matches, rmw differs by +0.02 and SMB and CMA differ by +0.04 and -0.04 respectively. Comparing your numbers to PV (4 factors monthly), the beta differs -0.01, SMB -.18, HML +.10 and MOM -0.07. PV's MOM is not statistically significant on monthly data.
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

305pelusa wrote: Thu Nov 07, 2019 6:06 pm Aight I promised you some data. I personally like to regress underlying index data and then assume the fund/etf itself will track that index well. This typically gives me more data for regression since the index has more length than the funds themselves. That's what I did here:

IJS (S&P 600 SCV) from May 1997 to Aug 2019 (monthly)
Alpha = 0.05 (p-value of 0.58, not statistically significant)
Beta = 0.96
SmB = 0.63
HmL = 0.6
MOM = -0.06

Had an R^2 of 0.964. All factors were statistically significant.

ZPRV (MSCI USA SC Value Weighted) from May 1997 to Aug 2019 (monthly):
Alpha = 0.19 (p-value of 0.02 so is statistically significant)
Beta = 1
SmB = 0.59
HmL = 0.54
MOM = -0.19

Had an R^2 of 0.974. All factors were statistically significant. The alpha seems to imply exposure to other factors (RmW and CmA) so I'll take it into account below.

Prediction
Assuming the following premiums (arbitrary 50% cuts for all factors except beta, where I just had a 25% reduction)
Beta = 6 (vs 8.3 historical)
SmB = 1.5 (vs 3.3 historical)
HmL = 2.4 (vs 4.8 historical)
MOM = 4.8 (vs 9.6 historical)

IJS expected return = 0.96*6 + 0.63*1.5 + 0.6*2.4 - 0.06*4.8 - 0.25 (ER) = 7.6%
ZPRV expected return = 1*6 + 0.59*1.5 + 0.54*2.4 - 0.19*4.8 + 0.19*12 (monthly alpha) - 0.3 (ER) = 9.2%

Thoughts
You'll see the above results for the S&P 600 SCV index are fairly different than if you just regress IJS itself (say with PV). That's because IJS did not track the index as well during its early years and because the index has gotten less value-y and smaller in recent time (which is exactly when you'd regress IJS). You'll have to come to your own conclusions there.

As for ZPRV, I'm really confused as to why my results are so different than Uncorrolated's. I used monthly index data from MSCI and the 3F (+MOM) from Ken. Perhaps it's just because he used the full 5 factors (6 it looks like? 0_o).

Just thinking out loud but your R^2 values look kinda poor. IJS for sure but even 0.89 on ZPRV implies a poor fit. Could you have a small error on your end?
Thanks for doing that. Appreciated.

First thing that stands out to me is that you found IJS has more loading on value, while the metrics of ZPRV suggest more value loading.

ZPRV (SPDR® MSCI USA Small Cap Value Weighted UCITS ETF USD Acc (EUR))
Avg Market Cap 2.62 Bil
Price/Earnings 13,06
Price/Book 1,24
Price/Sales 0,48
Price/Cash Flow 4,35
Dividend Yield % 2.28

IJS (iShares S&P Small-Cap 600 Value ETF (USD))
Avg market cap 1.49 Bil
Price/Earnings 14.67
Price/Book 1.34
Price/Sales 0.60
Price/Cash Flow 5.44
Dividend Yield % 2.11

Second thing I noticed is that you found a rather small difference in size loading: 0.63 vs 0.59. Is this congruent with the fact that the holding in IJS are a lot smaller?

Comparing the loading of the factors to Ben Felix's analysis there seems to be quite a large difference. His: SmB 0.86 HmL 0.27. Uncorrelated's numbers come quite close to Ben Felix numbers.

About using the indexes: could the tracking error be causing the differences? Tracking error for IJS seems to be "0.06" (not sure if that's 6% or 0.06%?)
https://screener.fidelity.com/ftgw/etf/ ... ymbols=IJS

ZPRV tracking error 0.4%
https://lu.spdrs.com/library-content/pu ... eet_en.pdf
User avatar
Steve Reading
Posts: 2959
Joined: Fri Nov 16, 2018 9:20 pm

Re: Small Cap Value heads Rejoice !!!

Post by Steve Reading »

Uncorrelated wrote: Fri Nov 08, 2019 5:05 am
305pelusa wrote: Thu Nov 07, 2019 6:06 pm Had an R^2 of 0.974. All factors were statistically significant. The alpha seems to imply exposure to other factors (RmW and CmA) so I'll take it into account below.
That sounds fishy. I would ignore the alpha for this fund, the fund doesn't do any special market timing that can explain the alpha.

Regarding the low R^2, I'm probably doing something wrong. The odd part is that my numbers for IJS are much closer to portfoliovisualizer numbers than yours. Comparing my IJS numbers to PV (6 factors daily), the beta, HML and MOM exposure matches, rmw differs by +0.02 and SMB and CMA differ by +0.04 and -0.04 respectively. Comparing your numbers to PV (4 factors monthly), the beta differs -0.01, SMB -.18, HML +.10 and MOM -0.07. PV's MOM is not statistically significant on monthly data.
Why would you ignore the alpha? It just implies it has loadings of RmW and CmA that I would probably catch if I did all factors.

Remember I'm regressing IJS's index, not IJS itself. So I don't expect the numbers to necessarily be similar. In my post, on the first paragraph of my conclusion, I explain why.

When I regress the index from 2001 to today, I get very similar numbers to IJS.

YRT70 wrote: Fri Nov 08, 2019 6:40 am
305pelusa wrote: Thu Nov 07, 2019 6:06 pm Aight I promised you some data. I personally like to regress underlying index data and then assume the fund/etf itself will track that index well. This typically gives me more data for regression since the index has more length than the funds themselves. That's what I did here:

IJS (S&P 600 SCV) from May 1997 to Aug 2019 (monthly)
Alpha = 0.05 (p-value of 0.58, not statistically significant)
Beta = 0.96
SmB = 0.63
HmL = 0.6
MOM = -0.06

Had an R^2 of 0.964. All factors were statistically significant.

ZPRV (MSCI USA SC Value Weighted) from May 1997 to Aug 2019 (monthly):
Alpha = 0.19 (p-value of 0.02 so is statistically significant)
Beta = 1
SmB = 0.59
HmL = 0.54
MOM = -0.19

Had an R^2 of 0.974. All factors were statistically significant. The alpha seems to imply exposure to other factors (RmW and CmA) so I'll take it into account below.

Prediction
Assuming the following premiums (arbitrary 50% cuts for all factors except beta, where I just had a 25% reduction)
Beta = 6 (vs 8.3 historical)
SmB = 1.5 (vs 3.3 historical)
HmL = 2.4 (vs 4.8 historical)
MOM = 4.8 (vs 9.6 historical)

IJS expected return = 0.96*6 + 0.63*1.5 + 0.6*2.4 - 0.06*4.8 - 0.25 (ER) = 7.6%
ZPRV expected return = 1*6 + 0.59*1.5 + 0.54*2.4 - 0.19*4.8 + 0.19*12 (monthly alpha) - 0.3 (ER) = 9.2%

Thoughts
You'll see the above results for the S&P 600 SCV index are fairly different than if you just regress IJS itself (say with PV). That's because IJS did not track the index as well during its early years and because the index has gotten less value-y and smaller in recent time (which is exactly when you'd regress IJS). You'll have to come to your own conclusions there.

As for ZPRV, I'm really confused as to why my results are so different than Uncorrolated's. I used monthly index data from MSCI and the 3F (+MOM) from Ken. Perhaps it's just because he used the full 5 factors (6 it looks like? 0_o).

Just thinking out loud but your R^2 values look kinda poor. IJS for sure but even 0.89 on ZPRV implies a poor fit. Could you have a small error on your end?
Thanks for doing that. Appreciated.

First thing that stands out to me is that you found IJS has more loading on value, while the metrics of ZPRV suggest more value loading.

ZPRV (SPDR® MSCI USA Small Cap Value Weighted UCITS ETF USD Acc (EUR))
Avg Market Cap 2.62 Bil
Price/Earnings 13,06
Price/Book 1,24
Price/Sales 0,48
Price/Cash Flow 4,35
Dividend Yield % 2.28

IJS (iShares S&P Small-Cap 600 Value ETF (USD))
Avg market cap 1.49 Bil
Price/Earnings 14.67
Price/Book 1.34
Price/Sales 0.60
Price/Cash Flow 5.44
Dividend Yield % 2.11

Second thing I noticed is that you found a rather small difference in size loading: 0.63 vs 0.59. Is this congruent with the fact that the holding in IJS are a lot smaller?

Comparing the loading of the factors to Ben Felix's analysis there seems to be quite a large difference. His: SmB 0.86 HmL 0.27. Uncorrelated's numbers come quite close to Ben Felix numbers.

About using the indexes: could the tracking error be causing the differences? Tracking error for IJS seems to be "0.06" (not sure if that's 6% or 0.06%?)
https://screener.fidelity.com/ftgw/etf/ ... ymbols=IJS

ZPRV tracking error 0.4%
https://lu.spdrs.com/library-content/pu ... eet_en.pdf
I refer you also to my first paragraph of the "Conclusion". The index differs from IJS because IJS both did not track the index as much right at the beginning, and because the index was high in value in 1997-2002, when IJS did not exist yet.

Take whatever conclusion you want from that. To me, that means that had IJS existed, and you used PV, you WOULD see my numbers.

As for your metrics, that would represent current loading. When I regress more recently (say 2010 to 2019), I get basically the same values as PV.

So you'll have to ask yourself whether the loadings you see today for IJS are representative of what it will be in the future, or if the long term index data is more representative of that. I lean towards the latter because it includes more business cycles, bubbles, etc. I've seen other posters lean for the former, believing more recent data is more relevant. Take your pick
"... so high a present discounted value of wealth, it is only prudent for him to put more into common stocks compared to his present tangible wealth, borrowing if necessary" - Paul Samuelson
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

305pelusa wrote: Fri Nov 08, 2019 7:27 am I refer you also to my first paragraph of the "Conclusion". The index differs from IJS because IJS both did not track the index as much right at the beginning, and because the index was high in value in 1997-2002, when IJS did not exist yet.

Take whatever conclusion you want from that. To me, that means that had IJS existed, and you used PV, you WOULD see my numbers.

As for your metrics, that would represent current loading. When I regress more recently (say 2010 to 2019), I get basically the same values as PV.

So you'll have to ask yourself whether the loadings you see today for IJS are representative of what it will be in the future, or if the long term index data is more representative of that. I lean towards the latter because it includes more business cycles, bubbles, etc. I've seen other posters lean for the former, believing more recent data is more relevant. Take your pick
I see. Thank you. If it's not too much trouble could you run the regression again only using the last 10 years of data?

I wonder what the difference in expected return will be with IJS being less valuey but smaller. I realise IJS will look like PV but PV has no data for ZPRV.
User avatar
Steve Reading
Posts: 2959
Joined: Fri Nov 16, 2018 9:20 pm

Re: Small Cap Value heads Rejoice !!!

Post by Steve Reading »

YRT70 wrote: Fri Nov 08, 2019 7:46 am
305pelusa wrote: Fri Nov 08, 2019 7:27 am I refer you also to my first paragraph of the "Conclusion". The index differs from IJS because IJS both did not track the index as much right at the beginning, and because the index was high in value in 1997-2002, when IJS did not exist yet.

Take whatever conclusion you want from that. To me, that means that had IJS existed, and you used PV, you WOULD see my numbers.

As for your metrics, that would represent current loading. When I regress more recently (say 2010 to 2019), I get basically the same values as PV.

So you'll have to ask yourself whether the loadings you see today for IJS are representative of what it will be in the future, or if the long term index data is more representative of that. I lean towards the latter because it includes more business cycles, bubbles, etc. I've seen other posters lean for the former, believing more recent data is more relevant. Take your pick
I see. Thank you. If it's not too much trouble could you run the regression again only using the last 10 years of data?

I wonder what the difference in expected return will be with IJS being less valuey but smaller. I realise IJS will look like PV but PV has no data for ZPRV.
Aug 2009 - Aug 2019 (monthly)
S&P 600 SCV
Beta = 0.98
SmB = 0.8
HmL = 0.32
MOM = -0.03 (not statistically significant)

Alpha was positive but not statistically significant. All other factors (except MOM) were statistically significant. R^2 = 0.987.

MSCI USA Small Cap weighted
Beta = 1.04
SmB = 0.68
HmL = 0.33
MOM = -0.1

Alpha was not statistically significant. All other factors were. R^2 = 0.993

IJS Expected Return = 0.98*6 + 0.8*1.5 + 0.32*2.4 - 0.25 (ER)= 7.6%
ZPRV Expected Return = 1.04*6 + 0.68*1.5 + 0.33*2.4 - 0.1*4.8 - 0.3 (ER) = 7.3%

The loading of other factors seems smaller on ZPRV here, you see it from the fact that there's no alpha. The above is so close, it's basically a toss-up. I'd choose IJS purely for its lack of neg. MOM.

The above numbers represent a measly period of 10 years, during which there wasn't even a single depression. So I probably wouldn't take them too seriously. That said, the point is that when one index gets less value-y, the other probably does too. So I just regress the entire index and compare.

Hope that helps a little. And needless to say, these are merely regression parameters not meant as actual predictions. Conclude as you like from them.
"... so high a present discounted value of wealth, it is only prudent for him to put more into common stocks compared to his present tangible wealth, borrowing if necessary" - Paul Samuelson
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

305pelusa wrote: Fri Nov 08, 2019 8:06 am
YRT70 wrote: Fri Nov 08, 2019 7:46 am
305pelusa wrote: Fri Nov 08, 2019 7:27 am I refer you also to my first paragraph of the "Conclusion". The index differs from IJS because IJS both did not track the index as much right at the beginning, and because the index was high in value in 1997-2002, when IJS did not exist yet.

Take whatever conclusion you want from that. To me, that means that had IJS existed, and you used PV, you WOULD see my numbers.

As for your metrics, that would represent current loading. When I regress more recently (say 2010 to 2019), I get basically the same values as PV.

So you'll have to ask yourself whether the loadings you see today for IJS are representative of what it will be in the future, or if the long term index data is more representative of that. I lean towards the latter because it includes more business cycles, bubbles, etc. I've seen other posters lean for the former, believing more recent data is more relevant. Take your pick
I see. Thank you. If it's not too much trouble could you run the regression again only using the last 10 years of data?

I wonder what the difference in expected return will be with IJS being less valuey but smaller. I realise IJS will look like PV but PV has no data for ZPRV.
Aug 2009 - Aug 2019 (monthly)
S&P 600 SCV
Beta = 0.98
SmB = 0.8
HmL = 0.32
MOM = -0.03 (not statistically significant)

Alpha was positive but not statistically significant. All other factors (except MOM) were statistically significant. R^2 = 0.987.

MSCI USA Small Cap weighted
Beta = 1.04
SmB = 0.68
HmL = 0.33
MOM = -0.1

Alpha was not statistically significant. All other factors were. R^2 = 0.993

IJS Expected Return = 0.98*6 + 0.8*1.5 + 0.32*2.4 - 0.25 (ER)= 7.6%
ZPRV Expected Return = 1.04*6 + 0.68*1.5 + 0.33*2.4 - 0.1*4.8 - 0.3 (ER) = 7.3%

The loading of other factors seems smaller on ZPRV here, you see it from the fact that there's no alpha. The above is so close, it's basically a toss-up. I'd choose IJS purely for its lack of neg. MOM.

The above numbers represent a measly period of 10 years, during which there wasn't even a single depression. So I probably wouldn't take them too seriously. That said, the point is that when one index gets less value-y, the other probably does too. So I just regress the entire index and compare.

Hope that helps a little. And needless to say, these are merely regression parameters not meant as actual predictions. Conclude as you like from them.
Thanks a ton. Good to know. For what it's worth, your last expected return calculation is very close to what uncorrelated's data suggested.

And then I also have a slight tax advantage in favour of IJS, probably 0.2%.

You mention no depression in the data. Would 15 year data be a different story?
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: Small Cap Value heads Rejoice !!!

Post by Uncorrelated »

305pelusa wrote: Fri Nov 08, 2019 7:27 am Why would you ignore the alpha? It just implies it has loadings of RmW and CmA that I would probably catch if I did all factors.

Remember I'm regressing IJS's index, not IJS itself. So I don't expect the numbers to necessarily be similar. In my post, on the first paragraph of my conclusion, I explain why.

When I regress the index from 2001 to today, I get very similar numbers to IJS.
If you want to capture WmR and CmA then you should use a 5-factor regression. It sounds like you're discovering a new factor rather than explaining the factor exposure of a portfolio. You also didn't discount the alpha by half, as you did for the historical factor performance.

I missed that you regressed the index instead of live data. I likely made a mistake on my own analysis.
User avatar
Steve Reading
Posts: 2959
Joined: Fri Nov 16, 2018 9:20 pm

Re: Small Cap Value heads Rejoice !!!

Post by Steve Reading »

YRT70 wrote: Fri Nov 08, 2019 8:14 am
305pelusa wrote: Fri Nov 08, 2019 8:06 am
YRT70 wrote: Fri Nov 08, 2019 7:46 am
305pelusa wrote: Fri Nov 08, 2019 7:27 am I refer you also to my first paragraph of the "Conclusion". The index differs from IJS because IJS both did not track the index as much right at the beginning, and because the index was high in value in 1997-2002, when IJS did not exist yet.

Take whatever conclusion you want from that. To me, that means that had IJS existed, and you used PV, you WOULD see my numbers.

As for your metrics, that would represent current loading. When I regress more recently (say 2010 to 2019), I get basically the same values as PV.

So you'll have to ask yourself whether the loadings you see today for IJS are representative of what it will be in the future, or if the long term index data is more representative of that. I lean towards the latter because it includes more business cycles, bubbles, etc. I've seen other posters lean for the former, believing more recent data is more relevant. Take your pick
I see. Thank you. If it's not too much trouble could you run the regression again only using the last 10 years of data?

I wonder what the difference in expected return will be with IJS being less valuey but smaller. I realise IJS will look like PV but PV has no data for ZPRV.
Aug 2009 - Aug 2019 (monthly)
S&P 600 SCV
Beta = 0.98
SmB = 0.8
HmL = 0.32
MOM = -0.03 (not statistically significant)

Alpha was positive but not statistically significant. All other factors (except MOM) were statistically significant. R^2 = 0.987.

MSCI USA Small Cap weighted
Beta = 1.04
SmB = 0.68
HmL = 0.33
MOM = -0.1

Alpha was not statistically significant. All other factors were. R^2 = 0.993

IJS Expected Return = 0.98*6 + 0.8*1.5 + 0.32*2.4 - 0.25 (ER)= 7.6%
ZPRV Expected Return = 1.04*6 + 0.68*1.5 + 0.33*2.4 - 0.1*4.8 - 0.3 (ER) = 7.3%

The loading of other factors seems smaller on ZPRV here, you see it from the fact that there's no alpha. The above is so close, it's basically a toss-up. I'd choose IJS purely for its lack of neg. MOM.

The above numbers represent a measly period of 10 years, during which there wasn't even a single depression. So I probably wouldn't take them too seriously. That said, the point is that when one index gets less value-y, the other probably does too. So I just regress the entire index and compare.

Hope that helps a little. And needless to say, these are merely regression parameters not meant as actual predictions. Conclude as you like from them.
Thanks a ton. Good to know. For what it's worth, your last expected return calculation is very close to what uncorrelated's data suggested.

And then I also have a slight tax advantage in favour of IJS, probably 0.2%.

You mention no depression in the data. Would 15 year data be a different story?
I suppose? But I'm not sure what you want out of this. In my opinion, 15 years is more valid than 10. And 22 years (which includes TWO Bear markets which, BTW, IJS itself did not suffer through), even more valid.

Just my opinion.
Uncorrelated wrote: Fri Nov 08, 2019 8:45 am
If you want to capture WmR and CmA then you should use a 5-factor regression. It sounds like you're discovering a new factor rather than explaining the factor exposure of a portfolio.
No, I'm not "discovering a new factor". I'm ACKNOWLEDGING that there's a source of returns that is statistically significant that is unexplained by my regression. That could exposure to other factors not in the regression, the way the index was reconstituted, how it defines value (it uses metrics other than BtM, and uses an average window of 3 years), or other things I'm unaware of. Some indexes (like RAFI) produce actual alpha unexplained by common factors purely from methodology. Even though these are passive, not actively managed.

So instead of ignoring that (like you suggest), I account for it. I don't discount it by 50% because, again, I don't know how likely it is to continue. The reader can come to their own personal arbitrary conclusions.

BTW, I don't generally use 5F regression because I do want to know MOM. I didn't know it was valid to just tack on MOM like that for a 6F regression. Is that true?
"... so high a present discounted value of wealth, it is only prudent for him to put more into common stocks compared to his present tangible wealth, borrowing if necessary" - Paul Samuelson
Topic Author
hdas
Posts: 1395
Joined: Thu Jun 11, 2015 8:24 am

[Deleted]

Post by hdas »

[Deleted]
Last edited by hdas on Wed Jan 29, 2020 7:14 pm, edited 1 time in total.
....
User avatar
Steve Reading
Posts: 2959
Joined: Fri Nov 16, 2018 9:20 pm

Re: Small Cap Value heads Rejoice !!!

Post by Steve Reading »

hdas wrote: Fri Nov 08, 2019 12:03 pm
305pelusa wrote: Fri Nov 08, 2019 9:22 am In my opinion, 15 years is more valid than 10. And 22 years (which includes TWO Bear markets which, BTW, IJS itself did not suffer through), even more valid.
I don't think this is necessarily true, to some degree you want a fund that sticks to a certain exposure but on the other hand you care about the most current period the most. Cheers :greedy
Lol of course it's not necessarily true. It's just my opinion.

My thinking is that because the methodology is constant and they're not explicitly trying to target some factor exposure, then the factor exposure you see is simply a product of the environment, available stocks, spreads, etc. Because I'm a long term investor, I'd rather look at what the long term average exposure that methodology gave (which "averages out" those environment variables). Otherwise,I'd have to periodically check factor exposure and buy/sell to maintain the desired exposure. Which is valid too but that seems to try to add additional "exactitude" to an already fairly uncertain field. So I prefer to stick with B&H.
"... so high a present discounted value of wealth, it is only prudent for him to put more into common stocks compared to his present tangible wealth, borrowing if necessary" - Paul Samuelson
Fryxell
Posts: 183
Joined: Sat Mar 02, 2013 9:08 pm
Location: Dallas

Re: Small Cap Value heads Rejoice !!!

Post by Fryxell »

Uncorrelated wrote: Tue Nov 05, 2019 2:53 pm
The R^2 on IJS is atrocious. Not sure what's up with that. The factors loadings themselves are almost identical to portfolio visualizer's numbers
That’s strange. I can’t get it to be that bad on Portfolio Visualizer.

It’s hard to tell from what you posted. It looks like you’re looking at the adjusted R-squared. What was the plain R-squared?

What was the time period? Whose factor data did you use to regress against?
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

305pelusa wrote: Fri Nov 08, 2019 9:22 am I suppose? But I'm not sure what you want out of this. In my opinion, 15 years is more valid than 10. And 22 years (which includes TWO Bear markets which, BTW, IJS itself did not suffer through), even more valid.
Only because you said that the factor exposure around 2002 was quite different than it is now. I'm only interested in how IJS would perform with its current factor exposures (smaller, not so 'valuey') because as far as I can tell there's no reason to assume that it's going to change significantly in the future.

That's why I thought the longest period we have of S&P600 SCV with its current factor exposures is maybe 15 years. Although I do not know if that's true.
User avatar
Steve Reading
Posts: 2959
Joined: Fri Nov 16, 2018 9:20 pm

Re: Small Cap Value heads Rejoice !!!

Post by Steve Reading »

YRT70 wrote: Sat Nov 09, 2019 6:47 am
305pelusa wrote: Fri Nov 08, 2019 9:22 am I suppose? But I'm not sure what you want out of this. In my opinion, 15 years is more valid than 10. And 22 years (which includes TWO Bear markets which, BTW, IJS itself did not suffer through), even more valid.
Only because you said that the factor exposure around 2002 was quite different than it is now. I'm only interested in how IJS would perform with its current factor exposures (smaller, not so 'valuey') because as far as I can tell there's no reason to assume that it's going to change significantly in the future.

That's why I thought the longest period we have of S&P600 SCV with its current factor exposures is maybe 15 years. Although I do not know if that's true.
Look:
IJS's size/value was 0.63/0.6 for the last 23 years.
It was 0.8/0.32 for the past 10 years.
It was 0.87/0.54 for the past 3 years (2017 onward). So its "current" value exposure is back up once again. Not quite 2002 levels but much higher than the past decade levels that you'd see on PV.

You simply cannot estimate future exposure based on current exposure.
- Had you done it in 2002, you would've thought it had a value of around 0.8. And you would've been disappointed for the next 15 years.
- Had you done it in 2017, you would've thought it had a value of around 0.3. And you would've been disappointed for the next 3 years (and counting).

Let me recommend an alternative:
Instead of trying to figure out what its current exposure is and then extrapolating it to the future (which just doesn't work), how about you regress over as long of a period of time as you can, and then assume that its future exposure, on average, will be similar to that thanks to the law of large numbers? That requires you buy and hold long term these funds, which is what every factor investor will recommend.

Ironically, its current, most recent loading is very similar to its long term one so if you were really looking at its current loading, we wouldn't be having this conversation about discrepancy. But I'm glad we did because then you can tell precisely the danger of just regressing, say, last 15 years and just tossing out the 8 years before. You run the risk of just regressing over an abnormal period of time in loading.

Just something to think about.
"... so high a present discounted value of wealth, it is only prudent for him to put more into common stocks compared to his present tangible wealth, borrowing if necessary" - Paul Samuelson
Topic Author
hdas
Posts: 1395
Joined: Thu Jun 11, 2015 8:24 am

[Deleted]

Post by hdas »

[Deleted]
Last edited by hdas on Wed Jan 29, 2020 7:14 pm, edited 1 time in total.
....
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: International Small Cap Value

Post by Uncorrelated »

hdas wrote: Tue Nov 12, 2019 8:06 am
MotoTrojan wrote: Mon Nov 04, 2019 6:51 pm For those true believers, what International SCV funds do you hold?
Im not a true believer, I use ISCF. I believe international small cap value has benefited the most in the recent value rebound. Cheers :greedy
ISCF has large negative value exposure according with both FF and ARQ factors (not statistically significant with AA). I don't recommend small cap non-value stocks for any purpose. Small cap historically has been the weakest equity factor.

Small cap non value historically had negative value exposure outside the USA. Such a fund can be seen as bet against value.
Topic Author
hdas
Posts: 1395
Joined: Thu Jun 11, 2015 8:24 am

[Deleted]

Post by hdas »

[Deleted]
Last edited by hdas on Wed Jan 29, 2020 7:14 pm, edited 1 time in total.
....
MotoTrojan
Posts: 11259
Joined: Wed Feb 01, 2017 7:39 pm

Re: Small Cap Value heads Rejoice !!!

Post by MotoTrojan »

Looking to start a position in AVDV and continue to research it against other funds, namely DLS. Will prioritize contributions there to catch up with domestic SCV, which means I’ll still hold most of that in the Vanguard CRSP offering rather than S&P600.
User avatar
Steve Reading
Posts: 2959
Joined: Fri Nov 16, 2018 9:20 pm

Re: International Small Cap Value

Post by Steve Reading »

Uncorrelated wrote: Tue Nov 12, 2019 8:23 am
hdas wrote: Tue Nov 12, 2019 8:06 am
MotoTrojan wrote: Mon Nov 04, 2019 6:51 pm For those true believers, what International SCV funds do you hold?
Im not a true believer, I use ISCF. I believe international small cap value has benefited the most in the recent value rebound. Cheers :greedy
ISCF has large negative value exposure according with both FF and ARQ factors (not statistically significant with AA). I don't recommend small cap non-value stocks for any purpose. Small cap historically has been the weakest equity factor.

Small cap non value historically had negative value exposure outside the USA. Such a fund can be seen as bet against value.
I'm assuming you get that from inception? I would ignore any value loading data for a fund based on just the most recent 4 years (ISCF inception). In that time period, FNDC AND DLS also have negative value exposure.

ISCF's index does target value as one of its factors. It currently has a lower P/E ratio than both FNDC and DLS. I definitely do not think of FNDC or DLS are "non-value".

A regression on ISCF's index since 1996 (a much longer data period) shows a small but positive value loading of 0.18. That's the number I considered when judging inclusion of ISCF in my portfolio.

Just my 2 cents
"... so high a present discounted value of wealth, it is only prudent for him to put more into common stocks compared to his present tangible wealth, borrowing if necessary" - Paul Samuelson
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

MotoTrojan wrote: Tue Nov 12, 2019 8:49 am Looking to start a position in AVDV and continue to research it against other funds, namely DLS. Will prioritize contributions there to catch up with domestic SCV, which means I’ll still hold most of that in the Vanguard CRSP offering rather than S&P600.
Looks like AVDV is deeper value and iirc Larry likes it better. Might want to ask him.
HippoSir
Posts: 269
Joined: Tue Jul 03, 2018 2:56 pm

Re: Small Cap Value heads Rejoice !!!

Post by HippoSir »

I originally put these stats in the thread about the new Avantis funds, but I'll reproduce here (and add DLS):

AVDV vs ISCF vs DLS
Price/Earnings 10.82/12.43/12.03
Price/Book 1.00/1.29/1.10
Price/Sales 0.54/0.72/0.59
Price/Cash Flow 4.34/5.73/5.59
Dividend Yield % 4.29/3.42/4.90
Long-Term Earnings % 9.44/9.87/9.11
Historical Earnings % 1.24/4.15/-3.78
Sales Growth % -55.93/3.66/4.15
Cash-Flow Growth % -28.90/0.69/-3.80
Book-Value Growth % -0.98/6.46/4.78
Avg Market Cap 1.42 Bil/1.40 Bil/1.36 Bil

The numbers effectively match what I see with the etf.com MSCI factor exposure tool. ISCF has a moderate positive value tilt, slightly less than DLS, quite a bit less than AVDV. On the other hand, ISCF has a significantly higher exposure to the quality factor than either DLS or AVDV. Honestly all of these funds seem like fine choices, I'm sticking with ISCF simply because I like the combination of value and quality load.
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: International Small Cap Value

Post by Uncorrelated »

305pelusa wrote: Tue Nov 12, 2019 9:45 am
Uncorrelated wrote: Tue Nov 12, 2019 8:23 am
hdas wrote: Tue Nov 12, 2019 8:06 am
MotoTrojan wrote: Mon Nov 04, 2019 6:51 pm For those true believers, what International SCV funds do you hold?
Im not a true believer, I use ISCF. I believe international small cap value has benefited the most in the recent value rebound. Cheers :greedy
ISCF has large negative value exposure according with both FF and ARQ factors (not statistically significant with AA). I don't recommend small cap non-value stocks for any purpose. Small cap historically has been the weakest equity factor.

Small cap non value historically had negative value exposure outside the USA. Such a fund can be seen as bet against value.
ISCF's index does target value as one of its factors. It currently has a lower P/E ratio than both FNDC and DLS. I definitely do not think of FNDC or DLS are "non-value".
I didn't realize that ISCF targets value and size. My bad.
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: International Small Cap Value

Post by YRT70 »

Uncorrelated wrote: Tue Nov 12, 2019 12:52 pm I didn't realize that ISCF targets value and size. My bad.
Too bad we can't get it in the Netherlands. Afaik the only iShares multifactor fund marketed to Europeans is this one: IE00BZ0PKT83 (there's also a EU and US version).

If possible, could you run a factor regression on it?
MotoTrojan
Posts: 11259
Joined: Wed Feb 01, 2017 7:39 pm

Re: Small Cap Value heads Rejoice !!!

Post by MotoTrojan »

YRT70 wrote: Tue Nov 12, 2019 10:03 am
MotoTrojan wrote: Tue Nov 12, 2019 8:49 am Looking to start a position in AVDV and continue to research it against other funds, namely DLS. Will prioritize contributions there to catch up with domestic SCV, which means I’ll still hold most of that in the Vanguard CRSP offering rather than S&P600.
Looks like AVDV is deeper value and iirc Larry likes it better. Might want to ask him.
I may do just that. There was an interesting comparison on the Avantis thread between the quality of DLS and AVDV (quoted below); would be interested to hear Larry's thoughts, if anything just to inform my knowledge. I did just buy a few shares of AVDV and got the ask price (marketable limit order as well as 2 shares via market order).
HippoSir wrote: Sun Nov 03, 2019 12:51 am
nedsaid wrote: Sat Nov 02, 2019 10:33 pm Larry told me in a personal message that Avantis screens for Profitability which is related to Quality. Avantis uses a combination of cash flow and earnings metrics and uses a modified book value.

Hopefully Larry can comment. I didn't quote from his message as most all of it came directly from Avantis but I can't point you to a particular source. I want other Bogleheads to be able to look things up themselves.
Larry is far more of an expert than me, so you should probably listen to him. I'll explain my thinking anyway though.

Quality is one of the "fuzzier" factors with a lot of different definitions and everybody seems to have their own mix of underlying metrics.

Vanguard's factor funds define quality as:
"For financials the Quality factor is measured by return on equity and share issuance. For non-financials the Quality factor is measured by return on equity, gross profitability, change in net operating assets, and leverage."

MSCI/iShares factor funds define quality as:
"The quality score for each security is currently based on all quality factors, Profitability,
Investment Quality, Earnings Quality, Leverage and Earnings Variability, from the relevant
Barra Equity Model (currently GEMLTL). A sector-relative score is derived from the combined
score by standardizing the latter within each sector and winsorizing at +/- 3.
Quality = (0.2) ∗ Profitability + (0.2) ∗ Investment Quality + (0.2) ∗ Earnings Quality
+ (−1) ∗ (0.2) ∗ Earnings Variability + (−1) ∗ (0.2) ∗ Leverage"

Not all of these are available from morningstar (although perhaps there is a better online source?), but what is available:

DLS/AVDV
Long-Term Earnings % 9.1/9.55
Historical Earnings % -5.67/2.04
Sales Growth % 4.16/-52.95
Cash-Flow Growth % -3.76/-26.39
Book-Value Growth % 4.77/-1.15

So AVDV does have higher earnings % (which would match the quoted methodology) but higher variability and negative growth. It really depends on how you define "quality".
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: International Small Cap Value

Post by Uncorrelated »

YRT70 wrote: Tue Nov 12, 2019 1:19 pm
Uncorrelated wrote: Tue Nov 12, 2019 12:52 pm I didn't realize that ISCF targets value and size. My bad.
Too bad we can't get it in the Netherlands. Afaik the only iShares multifactor fund marketed to Europeans is this one: IE00BZ0PKT83 (there's also a EU and US version).

If possible, could you run a factor regression on it?

Code: Select all

    R^2: 0.86
adj R^2: 0.86
|                                                    |annualized alpha|beta    |SmB     |HmL     |RmW     |CmA     |MoM      
:--|:--|:--|:--|:--|:--|:--|:--
|IFSW, iShares Edge MSCI World Multifactor UCITS ETF         |  -0.93%|    0.96|    0.04|   -0.01|    0.07|   -0.03|    0.03|
|t stat                                                      |   -0.43|   63.73|    1.63|   -0.22|    1.40|   -0.53|    1.65|

expected outperformance based on 50% factor premia: -0.10% + alpha
Based on 3.7 years of data, 261 data points per year (2015-09-03 to 2019-04-30)
Most of the MSCI factor indices don't seem to have any significant factor exposure. The fund above has statistically significantly exposure to none of the FF factors. The poor performance of multifactor funds is one reason why Ben Felix recommends small cap value funds.

Note that MSCI uses factor definitions that differ from the FF factor definitions, which might explain a part of the poor regressions. I regressed a whole lot of other MSCI multi factor indices and they all tell the same story. (I did not test low volatility indices).




For what it is worth, I re-did my analysis on IJS and ZPRV. My R^2 is now in the range of .96-.97. It turns out that I did some pre-processing on the data to regress for beta-1 instead of just beta that messed up the results. Despite the higher R^2 score the results are mostly unchanged.

Code: Select all

    R^2: 0.97
adj R^2: 0.97
|                                                    |annualized alpha|beta    |SmB     |HmL     |RmW     |CmA     |MoM      
:--|:--|:--|:--|:--|:--|:--|:--
|IJS, USA small cap value weighted                           |  -0.75%|    1.05|    0.93|    0.21|    0.15|    0.06|   -0.05|
|t stat                                                      |   -0.86|  309.22|  146.33|   32.69|   18.18|    6.03|  -11.92|

expected outperformance based on 50% factor premia: 1.71% + alpha
Based on 18.8 years of data, 251 data points per year (2000-07-25 to 2019-05-31)


    R^2: 0.96
adj R^2: 0.96
|                                                    |annualized alpha|beta    |SmB     |HmL     |RmW     |CmA     |MoM      
:--|:--|:--|:--|:--|:--|:--|:--
|MSCI_USA_SMALL_CAP_VALUE_WEIGHTED_INDEX                     |   0.29%|    1.05|    0.75|    0.29|    0.30|    0.07|   -0.21|
|t stat                                                      |    0.35|   55.80|   31.25|    9.35|    8.87|    1.54|  -14.33|

expected outperformance based on 50% factor premia: 1.17% + alpha
Based on 24.4 years of data, 12 data points per year (1994-12-30 to 2019-05-31)


    R^2: 0.98
adj R^2: 0.98
|                                                    |annualized alpha|beta    |SmB     |HmL     |RmW     |CmA     |MoM      
:--|:--|:--|:--|:--|:--|:--|:--
|MSCI_USA_SMALL_CAP_VALUE_WEIGHTED_INDEX                     |  -1.37%|    1.03|    0.69|    0.26|    0.07|    0.16|   -0.12|
|t stat                                                      |   -1.24|  184.93|   74.00|   22.87|    5.27|    9.05|  -18.02|

expected outperformance based on 50% factor premia: 1.23% + alpha
Based on 4.5 years of data, 251 data points per year (2014-11-13 to 2019-05-31)
This contradicts 305Pelusa's results. Some differences are expected due to 4 factor vs 6 factor, but I'm not sure of we can explain all differences that way.
305pelusa wrote: Fri Nov 08, 2019 9:22 am BTW, I don't generally use 5F regression because I do want to know MOM. I didn't know it was valid to just tack on MOM like that for a 6F regression. Is that true?
That's a good question. I don't know. Why not?
Last edited by Uncorrelated on Tue Nov 12, 2019 2:23 pm, edited 1 time in total.
User avatar
Forester
Posts: 2400
Joined: Sat Jan 19, 2019 1:50 pm
Location: UK

Re: Small Cap Value heads Rejoice !!!

Post by Forester »

Is an ETF which screens as being deeper value necessarily better? It could also be a collection of laggards. Surely one would want a mixture of cheap stocks + stocks which are moving up.
Amateur Self-Taught Senior Macro Strategist
MotoTrojan
Posts: 11259
Joined: Wed Feb 01, 2017 7:39 pm

Re: Small Cap Value heads Rejoice !!!

Post by MotoTrojan »

Forester wrote: Tue Nov 12, 2019 2:22 pm Is an ETF which screens as being deeper value necessarily better? It could also be a collection of laggards. Surely one would want a mixture of cheap stocks + stocks which are moving up.
I like quality and value. S&P600’s earnings screen helps there.
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: International Small Cap Value

Post by YRT70 »

Uncorrelated wrote: Tue Nov 12, 2019 2:21 pm

Code: Select all

    R^2: 0.86
adj R^2: 0.86
|                                                    |annualized alpha|beta    |SmB     |HmL     |RmW     |CmA     |MoM      
:--|:--|:--|:--|:--|:--|:--|:--
|IFSW, iShares Edge MSCI World Multifactor UCITS ETF         |  -0.93%|    0.96|    0.04|   -0.01|    0.07|   -0.03|    0.03|
|t stat                                                      |   -0.43|   63.73|    1.63|   -0.22|    1.40|   -0.53|    1.65|

expected outperformance based on 50% factor premia: -0.10% + alpha
Based on 3.7 years of data, 261 data points per year (2015-09-03 to 2019-04-30)
Most of the MSCI factor indices don't seem to have any significant factor exposure. The fund above has statistically significantly exposure to none of the FF factors. The poor performance of multifactor funds is one reason why Ben Felix recommends small cap value funds.

Note that MSCI uses factor definitions that differ from the FF factor definitions, which might explain a part of the poor regressions. I regressed a whole lot of other MSCI multi factor indices and they all tell the same story. (I did not test low volatility indices).
Interesting. Thank you.

I was expecting iShares Edge MSCI World Momentum Factor UCITS ETF (Acc) IE00BP3QZ825 to have significant momentum exposure. Does it not? That would surprise me.


For what it is worth, I re-did my analysis on IJS and ZPRV. My R^2 is now in the range of .96-.97. It turns out that I did some pre-processing on the data to regress for beta-1 instead of just beta that messed up the results. Despite the higher R^2 score the results are mostly unchanged.

Code: Select all

    R^2: 0.97
adj R^2: 0.97
|                                                    |annualized alpha|beta    |SmB     |HmL     |RmW     |CmA     |MoM      
:--|:--|:--|:--|:--|:--|:--|:--
|IJS, USA small cap value weighted                           |  -0.75%|    1.05|    0.93|    0.21|    0.15|    0.06|   -0.05|
|t stat                                                      |   -0.86|  309.22|  146.33|   32.69|   18.18|    6.03|  -11.92|

expected outperformance based on 50% factor premia: 1.71% + alpha
Based on 18.8 years of data, 251 data points per year (2000-07-25 to 2019-05-31)


    R^2: 0.96
adj R^2: 0.96
|                                                    |annualized alpha|beta    |SmB     |HmL     |RmW     |CmA     |MoM      
:--|:--|:--|:--|:--|:--|:--|:--
|MSCI_USA_SMALL_CAP_VALUE_WEIGHTED_INDEX                     |   0.29%|    1.05|    0.75|    0.29|    0.30|    0.07|   -0.21|
|t stat                                                      |    0.35|   55.80|   31.25|    9.35|    8.87|    1.54|  -14.33|

expected outperformance based on 50% factor premia: 1.17% + alpha
Based on 24.4 years of data, 12 data points per year (1994-12-30 to 2019-05-31)


    R^2: 0.98
adj R^2: 0.98
|                                                    |annualized alpha|beta    |SmB     |HmL     |RmW     |CmA     |MoM      
:--|:--|:--|:--|:--|:--|:--|:--
|MSCI_USA_SMALL_CAP_VALUE_WEIGHTED_INDEX                     |  -1.37%|    1.03|    0.69|    0.26|    0.07|    0.16|   -0.12|
|t stat                                                      |   -1.24|  184.93|   74.00|   22.87|    5.27|    9.05|  -18.02|

expected outperformance based on 50% factor premia: 1.23% + alpha
Based on 4.5 years of data, 251 data points per year (2014-11-13 to 2019-05-31)
This contradicts 305Pelusa's results. Some differences are expected due to 4 factor vs 6 factor, but I'm not sure of we can explain all differences that way.
Perhaps tracking difference plays a role? going by that Reddit post TD for ZPRV is 0.83% iirc. That seems like quite a lot.

https://www.reddit.com/r/DutchFIRE/comm ... small_cap/

Going by this webpage IJS seems to score well on tracking difference, tracking error and hurst exponent. I have no idea of their methodology though.
https://www.trackinsight.com/fund/US464 ... _perf,flow
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: International Small Cap Value

Post by Uncorrelated »

YRT70 wrote: Wed Nov 13, 2019 7:29 am
I would stay away from momentum funds. Literature claims that although the momentum premium is persistent in theory, there are big issues with it in practice. See this thread from today: viewtopic.php?f=10&t=294890


Tracking error can't be the reason for the disparity between mine and 305Pelusa's result because we both regress the index of ZPRV instead of the actual fund data.
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: International Small Cap Value

Post by YRT70 »

Uncorrelated wrote: Wed Nov 13, 2019 8:05 am I would stay away from momentum funds. Literature claims that although the momentum premium is persistent in theory, there are big issues with it in practice. See this thread from today: viewtopic.php?f=10&t=294890
I've seen that thread. I've seen conflicting claims about iShares MTUM. Robert T has been posting previously on why in his view it is a good choice.
Tracking error can't be the reason for the disparity between mine and 305Pelusa's result because we both regress the index of ZPRV instead of the actual fund data.
Ah, I thought you had analysed IJS & ZPRV, not the index.
User avatar
Forester
Posts: 2400
Joined: Sat Jan 19, 2019 1:50 pm
Location: UK

Re: Small Cap Value heads Rejoice !!!

Post by Forester »

If a factor can't be crowbarred into CAPM then it's claimed to be uninvestable or explained by the politically-correct factors.
Amateur Self-Taught Senior Macro Strategist
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

Forester wrote: Tue Nov 12, 2019 2:22 pm Is an ETF which screens as being deeper value necessarily better?
I don't think so and for what it's worth, I don't think Larry thinks so either. Larry seems to like deeper value as long as it has the appropriate screens (likes Avantis, DFA and Bridgeway have afaik).

Personally I don't mind not so deep value. I quite like Paul Merriman's idea of going half blend, half value. Which is very close to what IJS is imo.
MotoTrojan
Posts: 11259
Joined: Wed Feb 01, 2017 7:39 pm

Re: Small Cap Value heads Rejoice !!!

Post by MotoTrojan »

HippoSir wrote: Tue Nov 12, 2019 12:09 pm I originally put these stats in the thread about the new Avantis funds, but I'll reproduce here (and add DLS):

AVDV vs ISCF vs DLS
Price/Earnings 10.82/12.43/12.03
Price/Book 1.00/1.29/1.10
Price/Sales 0.54/0.72/0.59
Price/Cash Flow 4.34/5.73/5.59
Dividend Yield % 4.29/3.42/4.90
Long-Term Earnings % 9.44/9.87/9.11
Historical Earnings % 1.24/4.15/-3.78
Sales Growth % -55.93/3.66/4.15
Cash-Flow Growth % -28.90/0.69/-3.80
Book-Value Growth % -0.98/6.46/4.78
Avg Market Cap 1.42 Bil/1.40 Bil/1.36 Bil

The numbers effectively match what I see with the etf.com MSCI factor exposure tool. ISCF has a moderate positive value tilt, slightly less than DLS, quite a bit less than AVDV. On the other hand, ISCF has a significantly higher exposure to the quality factor than either DLS or AVDV. Honestly all of these funds seem like fine choices, I'm sticking with ISCF simply because I like the combination of value and quality load.
Thanks, these are very insightful. I am struggling to understand a couple of these though; Sales Growth (-55.93%) and Cash-Flow Growth (-28.90%) are pretty staggering declines, yet the earnings growth seems strong. What does this mean exactly? Cash-flow and sales plummeted for these companies but so did their expenses so earnings was still A-okay?

Would be interested to see something like this for AVUV as well (Avantis US Small Value) to compare to SLYV or similar index offerings, but frankly if I am going to start to invest in AVDV (I have) then I wouldn't go overboard and hold many/any other Avantis offerings, especially not in taxable where I have no qualms about holding say, VIOV.
Longtermgrowth
Posts: 731
Joined: Thu Nov 26, 2015 12:59 pm

Re: Small Cap Value heads Rejoice !!!

Post by Longtermgrowth »

MotoTrojan wrote: Wed Nov 13, 2019 8:21 pm Would be interested to see something like this for AVUV as well (Avantis US Small Value) to compare to SLYV or similar index offerings, but frankly if I am going to start to invest in AVDV (I have) then I wouldn't go overboard and hold many/any other Avantis offerings, especially not in taxable where I have no qualms about holding say, VIOV.
Now I'm going to have to go look that up for those funds plus DES, which is WisdomTree's equivalent small value fund for the US (same fund construction as DLS and DGS).
Maybe someone will post that here before I get around to checking both morningstar and etf.com. Interesting numbers on etf.com though...

DES, AVUV, SLYV

P/E: 31.77, 14.68, -436.91?
P/B: 1.84, 1.19, 1.45

Edited to add: etf.com calculates a different way than morningstar. I can see information for both DES ans SLYV on morningstar without logging in, and for comparison:

DES, SLYV

P/E:15.33, 14.59
P/B:1.63, 1.32
fennewaldaj
Posts: 1097
Joined: Sun Oct 22, 2017 11:30 pm

Re: Small Cap Value heads Rejoice !!!

Post by fennewaldaj »

MotoTrojan wrote: Wed Nov 13, 2019 8:21 pm
HippoSir wrote: Tue Nov 12, 2019 12:09 pm I originally put these stats in the thread about the new Avantis funds, but I'll reproduce here (and add DLS):

AVDV vs ISCF vs DLS
Price/Earnings 10.82/12.43/12.03
Price/Book 1.00/1.29/1.10
Price/Sales 0.54/0.72/0.59
Price/Cash Flow 4.34/5.73/5.59
Dividend Yield % 4.29/3.42/4.90
Long-Term Earnings % 9.44/9.87/9.11
Historical Earnings % 1.24/4.15/-3.78
Sales Growth % -55.93/3.66/4.15
Cash-Flow Growth % -28.90/0.69/-3.80
Book-Value Growth % -0.98/6.46/4.78
Avg Market Cap 1.42 Bil/1.40 Bil/1.36 Bil

The numbers effectively match what I see with the etf.com MSCI factor exposure tool. ISCF has a moderate positive value tilt, slightly less than DLS, quite a bit less than AVDV. On the other hand, ISCF has a significantly higher exposure to the quality factor than either DLS or AVDV. Honestly all of these funds seem like fine choices, I'm sticking with ISCF simply because I like the combination of value and quality load.
Thanks, these are very insightful. I am struggling to understand a couple of these though; Sales Growth (-55.93%) and Cash-Flow Growth (-28.90%) are pretty staggering declines, yet the earnings growth seems strong. What does this mean exactly? Cash-flow and sales plummeted for these companies but so did their expenses so earnings was still A-okay?

Would be interested to see something like this for AVUV as well (Avantis US Small Value) to compare to SLYV or similar index offerings, but frankly if I am going to start to invest in AVDV (I have) then I wouldn't go overboard and hold many/any other Avantis offerings, especially not in taxable where I have no qualms about holding say, VIOV.
I have never been sure what to make of the growth numbers on morningstar. I have noticed that active funds tend to have growth numbers that look better for what ever that is worth. It also makes sense that a dividend focused fund would not include firms that have very negative cash flow growth ect.
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: International Small Cap Value

Post by YRT70 »

Uncorrelated wrote: Wed Nov 13, 2019 8:05 am I would stay away from momentum funds. Literature claims that although the momentum premium is persistent in theory, there are big issues with it in practice. See this thread from today: viewtopic.php?f=10&t=294890
Here's one of the reasons I'm not convinced that momentum investing isn't worth it: https://www.institutionalinvestor.com/a ... hers-Argue

I posted it in the thread above too.
Topic Author
hdas
Posts: 1395
Joined: Thu Jun 11, 2015 8:24 am

[Deleted]

Post by hdas »

[Deleted]
Last edited by hdas on Wed Jan 29, 2020 7:15 pm, edited 1 time in total.
....
User avatar
Forester
Posts: 2400
Joined: Sat Jan 19, 2019 1:50 pm
Location: UK

Re: Small Cap Value heads Rejoice !!!

Post by Forester »

iShares FOVL +11.06%
Amateur Self-Taught Senior Macro Strategist
Topic Author
hdas
Posts: 1395
Joined: Thu Jun 11, 2015 8:24 am

[Deleted]

Post by hdas »

[Deleted]
Last edited by hdas on Wed Jan 29, 2020 7:15 pm, edited 1 time in total.
....
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

Continuing on the topic IJS vs. ZPRV expected return: the other Dutch guy (not Uncorrelated) sent me some new factor regressions and calculations of the tracking difference.

Here's his conclusion:
IJS: 1.82 - 0.18 = 1.64
ZPRV: 2.12 - 0.73 = 1.39

So ZPRV has a rather large tracking difference making the expected outperformance 1.39%. For IJS it's 1.64% (this includes TER and revenues from security lending).

I have not checked if these numbers are correct, I'm just sharing them in case someone finds them interesting.
Topic Author
hdas
Posts: 1395
Joined: Thu Jun 11, 2015 8:24 am

[Deleted]

Post by hdas »

[Deleted]
Last edited by hdas on Wed Jan 29, 2020 7:15 pm, edited 1 time in total.
....
MotoTrojan
Posts: 11259
Joined: Wed Feb 01, 2017 7:39 pm

Re: Small Cap Value heads Rejoice !!!

Post by MotoTrojan »

Big day, especially in the deeper size factor options (S&P600 > CRSP).
wiredaces80
Posts: 15
Joined: Sat Mar 17, 2018 1:02 am

Re: Small Cap Value heads Rejoice !!!

Post by wiredaces80 »

I just posted this in another thread, but it's relevant to this one as well:

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

I think it's important to note that over the last 90-100 years, SCV has not always performed better than everything else, in any conceivable period. The chart from Paul Merriman's website speaks volumes. To those who think SCV is a bad investment because of how it has performed recently....all I have to say is, I'm sure there were some people saying the same thing historically.

It is illogical, in my opinion, to trust the long term returns of the overall stock market, but also not trust the long term returns of a particular asset class over time. Because, the TSM is in fact an asset class in and of itself, with respect to every other possible investment. Since we rely on the long term historical returns of the stock market to estimate our future long term returns overall, I don't see why that same line of reasoning cannot be applied to SCV. Am I wrong? What am I missing?
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

How have you guys decided on how much you want to tilt?

I believe tilting is a good idea but I have hard time deciding on how much tilt I prefer. I saw some people have 90% TSM 10% IJS for their US allocation but that seems like such a small tilt that it's hardly worth it. 100% IJS for US stocks feels to aggressive for me personally even though more diversification across factors makes sense.

Is there anything I can do to get clearer on my desired tilt? Maybe run MC simulations?

I've played around with back testing but since inception of IJS it doesn't seem to matter much how much one tilts, assuming the holding time is long enough.

So how have you guys settled on your amount of tilting?
User avatar
Steve Reading
Posts: 2959
Joined: Fri Nov 16, 2018 9:20 pm

Re: Small Cap Value heads Rejoice !!!

Post by Steve Reading »

YRT70 wrote: Tue Nov 26, 2019 6:56 am How have you guys decided on how much you want to tilt?

I believe tilting is a good idea but I have hard time deciding on how much tilt I prefer. I saw some people have 90% TSM 10% IJS for their US allocation but that seems like such a small tilt that it's hardly worth it. 100% IJS for US stocks feels to aggressive for me personally even though more diversification across factors makes sense.

Is there anything I can do to get clearer on my desired tilt? Maybe run MC simulations?

I've played around with back testing but since inception of IJS it doesn't seem to matter much how much one tilts, assuming the holding time is long enough.

So how have you guys settled on your amount of tilting?
I decided how much factor exposure I wanted, and then chose whatever funds got me that total factor exposure at the cheapest price.

Tilting as a % of assets is very nebulous. 20% of US equity in SCV doesn't tell you anything unless you know the fund. If it's VBR? Maybe won't do much. If it's SLYV? That might be about right. If it's RZV, that might be a lot more than you intended. And so on. When I read Bernstein's book and see he recommends, say 50% of US on SCV, I just shake my head. That honestly tells you nothing.
"... so high a present discounted value of wealth, it is only prudent for him to put more into common stocks compared to his present tangible wealth, borrowing if necessary" - Paul Samuelson
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

305pelusa wrote: Tue Nov 26, 2019 7:17 am I decided how much factor exposure I wanted, and then chose whatever funds got me that total factor exposure at the cheapest price.

Tilting as a % of assets is very nebulous. 20% of US equity in SCV doesn't tell you anything unless you know the fund. If it's VBR? Maybe won't do much. If it's SLYV? That might be about right. If it's RZV, that might be a lot more than you intended. And so on. When I read Bernstein's book and see he recommends, say 50% of US on SCV, I just shake my head. That honestly tells you nothing.
Yeah I noticed your post on that. IIRC you decided for a 0.2 loading on size and value?

So why did you choose 0.2 and not 0.3 for example?
User avatar
Chrono Triggered
Posts: 396
Joined: Wed Feb 17, 2016 11:55 am

Re: Small Cap Value heads Rejoice !!!

Post by Chrono Triggered »

YRT70 wrote: Tue Nov 26, 2019 6:56 am So how have you guys settled on your amount of tilting?
I do 50% SCV for US (mostly S&P 600 Value based funds), and 50% small caps for international. Super simple to re-balance. Research and this thread inspired that decision: viewtopic.php?f=10&t=38374

Hosting service removed the photo, but someone posted it on page 15.
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

Chrono Triggered wrote: Tue Nov 26, 2019 7:37 am
YRT70 wrote: Tue Nov 26, 2019 6:56 am So how have you guys settled on your amount of tilting?
I do 50% SCV for US (mostly S&P 600 Value based funds), and 50% small caps for international. Super simple to re-balance. Research and this thread inspired that decision: viewtopic.php?f=10&t=38374

Hosting service removed the photo, but someone posted it on page 15.
Thanks. I will give that thread a read.
Locked