Comparison of small-cap value ETFs

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
User avatar
Topic Author
grabiner
Advisory Board
Posts: 25411
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Comparison of small-cap value ETFs

Post by grabiner » Sat Sep 15, 2012 5:05 pm

Two years ago, when Vanguard created its small-cap ETFs, I made a comparison between the small-cap value ETFs. Here is an updated comparison (All data from M*; expenses adjusted for Acquired fund fees and expenses from Business Development Companies)

MSCI Small-Cap 1750 Value (VBR, or Admiral shares VSIAX)
Adjusted expense ratio: 0.10%
Average market cap: $1330M, 20% mid-cap, 60% small-cap, 21% micro-cap
P/E: 13.46
P/B: 1.26
\
S&P Small-Cap 600 Value (VIOV)
Adjusted expense ratio: 0.20%
Average market cap: $903M, 4% mid-cap, 58% small-cap, 38% micro-cap
P/E: 15.38
P/B: 1.31

Russell 2000 Value (VTWV)
Adjusted expense ratio: 0.20%
Average market cap: $816M, 1% mid-cap, 58% small-cap, 40% micro-cap
P/E: 13.64
P/B: 1.11

So the Russell 2000 index gives significantly better small-cap value exposure, but you might prefer using more of the MSCI index to get the same benefit because of its lower expenses.

Going beyond Vanguard, you can get what is essentially a micro-cap fund with a very low P/B ratio, and thus much stronger value exposure:

Guggenheim S&P 600 Pure Value (RZV)
Adjusted expense ratio: 0.35%
Average market cap: $453M, 20% small-cap, 79% micro-cap
P/E: 14.93
P/B: 0.88

PowerShares Fundamental Small Value (PXSV)
Adjusted expense ratio: 0.39%
Average market cap: $855M, 5% mid-cap, 54% small-cap, 41% micro-cap
P/E: 13.21
P/B: 1.19
(also weighted by price/dividend and price/cash flow)

Micro-cap ETFs aren't designed to be small-value, but micro-cap stocks have lower P/B anyway, so they can serve a similar function:

Wilshire Microcap (WMCR)
Adjusted expense ratio: 0.50%
Average market cap: $150M, 100% micro-cap
P/E: 12.91
P/B: 1.32

iShares Russell Microcap (IWC)
Adjusted expense ratio: 0.60%
Average market cap: $261M, 1% small-cap, 99% micro-cap
P/E: 14.56
P/B: 1.28

Bridgeway Ultra-Small Company Market (BRSIX, not an ETF)
Adjusted expense ratio: 0.75%
Average market cap: $182M, 1% small-cap, 99% micro-cap
P/E: 12.88
P/B: 1.23

Caution: Tracking error and index composition can have major effects on micro-cap fund performance. BRSIX and IWC track each other almost perfectly, but WMCR has fallen 20% behind them over the last four years. RZV has been much more volatile.

(edited to add PXSV)
Last edited by grabiner on Sun Sep 16, 2012 9:24 am, edited 2 times in total.
Wiki David Grabiner

User avatar
Topic Author
grabiner
Advisory Board
Posts: 25411
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

My own preferences

Post by grabiner » Sat Sep 15, 2012 5:37 pm

My own preferences from this list are Vanguard's MSCI Small-Cap Value Index if you have enough tax-deferred room for your small-cap value exposure, and RZV if you are short on room because it gives the most value exposure per dollar.

I prefer the MSCI index construction, and I don't believe that the extra value and small-cap exposure from the Russell ETF is worth the cost; I would rather have $15,000 in Total Stock Market and $15,000 in Small-Cap Value Index rather than $20,000 in Total Stock Market and $10,000 in the Russell 2000 Value ETF. In addition, I don't have a brokerage account in my Roth IRA, and the extra cost of a brokerage account there isn't worthwhile; I can get the ETF expenses with Admiral shares of Small-Cap Value Index.

If my Roth IRA were much smaller, I would hold RZV there rather than pushing some of my Small-Cap Value Index into my taxable account. I do hold RZV in my Health Savings Account, which is already a brokerage account and thus has no extra cost for opening one, and is also not with Vanguard and thus has less of an incentive for holding Vanguard funds. (RZV might distribute some capital gains, which are taxable in a HSA in NJ, but I usually have enough capital losses to offset them.)
Wiki David Grabiner

User avatar
tetractys
Posts: 4596
Joined: Sat Mar 17, 2007 3:30 pm
Location: Along the Salish Sea

Re: Comparison of small-cap value ETFs

Post by tetractys » Sat Sep 15, 2012 6:11 pm

The MSCI micro cap index is not included in the MSCI small cap index. So it looks like M* defines micro caps quite a bit larger than MSCI does. M* saying that 21% of an MSCI index is micro cap when MSCI says 0% is quite a distortion, and something investors interested in owning small stocks should be aware of.

So for example, if you own an MSCI Small-Cap 1750 Value fund and Bridgeway Ultra-Small Market fund, which is mainly decile 10, there is no overlap. -- Tet

linenfort
Posts: 2241
Joined: Sat Sep 22, 2007 9:22 am
Location: #96151D

Re: Comparison of small-cap value ETFs

Post by linenfort » Sat Sep 15, 2012 6:34 pm

Great comparison!

One question: is it such a big deal to have some small-value in taxable if one's retirement accounts are full of bonds?

staythecourse
Posts: 6993
Joined: Mon Jan 03, 2011 9:40 am

Re: Comparison of small-cap value ETFs

Post by staythecourse » Sat Sep 15, 2012 6:52 pm

Couple of points:

1. VBR ER is 0.21% and not 0.10%
2. If you going to be comparing small value funds then shouldn't we be comparing what is really important and that is how well the different funds capture small and value premiums (HML and SMB)?

Good luck.
"The stock market [fluctuation], therefore, is noise. A giant distraction from the business of investing.” | -Jack Bogle

User avatar
Barry Barnitz
Wiki Admin
Posts: 3160
Joined: Mon Feb 19, 2007 10:42 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by Barry Barnitz » Sat Sep 15, 2012 7:05 pm

staythecourse wrote:Couple of points:

1. VBR ER is 0.21% and not 0.10%
Note that the 0.21% expense ratio includes "acquired fund fee" expenses of 0.11% which come from the fund's investment in Diversified Business Companies.

Acquired fund fees and expenses - Bogleheads
“Acquired Fund Fees and Expenses” are expenses incurred indirectly by the Fund through its ownership of shares in other investment companies, such as business development companies. These expenses are similar to the expenses paid by any operating company held by the Fund. They are not direct costs paid by Fund shareholders and are not used to calculate the Fund’s net asset value. They have no impact on the costs associated with fund operations. Acquired Fund Fees and Expenses are not included in the Fund’s financial statements, which provide a clearer picture of a fund’s actual operating costs."
regards
Additional administrative tasks: Financial Page affiliate blog; finiki the Canadian wiki; The Bogle Center for Financial Literacy site; Wiki Bogleheads® España.

staythecourse
Posts: 6993
Joined: Mon Jan 03, 2011 9:40 am

Re: Comparison of small-cap value ETFs

Post by staythecourse » Sat Sep 15, 2012 7:13 pm

Barry Barnitz wrote:
staythecourse wrote:Couple of points:

1. VBR ER is 0.21% and not 0.10%
Note that the 0.21% expense ratio includes "acquired fund fee" expenses of 0.11% which come from the fund's investment in Diversified Business Companies.

Acquired fund fees and expenses - Bogleheads
“Acquired Fund Fees and Expenses” are expenses incurred indirectly by the Fund through its ownership of shares in other investment companies, such as business development companies. These expenses are similar to the expenses paid by any operating company held by the Fund. They are not direct costs paid by Fund shareholders and are not used to calculate the Fund’s net asset value. They have no impact on the costs associated with fund operations. Acquired Fund Fees and Expenses are not included in the Fund’s financial statements, which provide a clearer picture of a fund’s actual operating costs."
regards
Please correct me if I am wrong, but if this is another fee added into the total fund cost and is the same that is done will all funds then comparing the total costs of holding the fund as an investor would be the 0.21%, no?? It is nice to separate out the different components, but in the end the total cost is what matters. Isn't that the point of them restating the new charges to the ER that are published??

As long as they do the same to all the funds then comparing the stated ER is the best way to compare each fund, no?? Stating the extra cost is not from the fund expenses doesn't matter to the investor. The cost is still applied to the investor's assets in the fund, no??

Thanks.

Good luck.
"The stock market [fluctuation], therefore, is noise. A giant distraction from the business of investing.” | -Jack Bogle

tpm871
Posts: 149
Joined: Sun Sep 12, 2010 10:58 am

Re: Comparison of small-cap value ETFs

Post by tpm871 » Sat Sep 15, 2012 7:40 pm

IJS also tracks S&P 600 Value:

S&P Small-Cap 600 Value (IJS)
Unadjusted expense ratio: 0.29%
Average market cap: $940M, 5% mid-cap, 58% small-cap, 37% micro-cap
P/E: 16.45
P/B: 1.34
Turnover Ratio: 30%
12 month yield: 1.59%

In comparison, VIOV has a 70% turnover and 0.99% yield. My guess is that VIOV trades more aggressively to keep up with changes to its benchmark composition, which is why it is a bit more micro-cap. But the higher turnover may have hidden costs (e.g., less tax efficiency).

VTWV is even more aggressive, with 101% turnover, and tracks even closer to its benchmark as a result.

User avatar
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by ClosetIndexer » Sat Sep 15, 2012 7:48 pm

The problem with the 'pure value' index / fund is that it has a very narrow focus (only 147 stocks compared to 1019 for VBR or 454/455 for VIOV/IJS), which results in a lot of activity in the returns not explained by the 3F model. Here are the regressions for VBR, IJS, and RZV since inception:

Code: Select all

                                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

    IJS                          0.92    0.85    0.40    -1.85%   R^2 = 0.963
       t-values                 38.45   18.27    9.16    -1.46


    RZV                          1.09    1.37    1.32     1.91%   R^2 = 0.849
       t-values                 10.18    5.71    6.61     0.32

Now, since its inception in 2006, this randomness has benefitted RZV, as we see from the large positive alpha and low R-squared. However, if we look over the longer term at the underlying S&P 600 pure value index, as Robert T does here, it paints a different picture.
Robert T wrote:

Code: Select all

                      
July 1995 – December 2011
                       Alpha [t-stat] Market    Size    Value   R^2              
S&P 600 Pure Value    -0.22  [-0.94]   1.13     0.85    1.18   0.82 
S&P 600 Value         -0.06  [-0.60]   0.94     0.66    0.54   0.94
There's no way RZV will continue to outperform its factor loadings by 2% a year. Much more likely is that it will revert to tracking its index minus MER and other expenses, as most good funds tend to do. IMO, VBR or VIOV (same S&P 600 value index as IJS) are safer choices, since you know that the results will much more closely match their underlying factor weightings. With RZV you're adding a lot of unnecessary uncertainty with little long-term benefit (and almost certainly a long-term detriment on a risk-adjusted basis).

User avatar
Topic Author
grabiner
Advisory Board
Posts: 25411
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: Comparison of small-cap value ETFs

Post by grabiner » Sat Sep 15, 2012 8:10 pm

staythecourse wrote:1. VBR ER is 0.21% and not 0.10%
I adjusted all the expense ratios for Business Development Company acquired fund fees and expenses, because that is not a real expense even though the fund must report it as one. If a BDC makes $10 per share of profits before expenses but has $1 per share of expenses, then investors will be willing to pay an appropriate price for a company with $9 per share of profits. If the BDC then reduces its expenses to 80 cents per share, its price will rise because the lower expenses are a guaranteed extra profit for shareholders. Thus it does not make sense to charge the fund for the $1 or the 80 cents, any more than it makes sense to charge a fund for a corporation's advertising expenses.

Acquired fund fees and expenses do need to be included in adjustments for funds which hold open-end funds (such as the Target Retirement funds), because those funds are valued at their net asset value.
2. If you going to be comparing small value funds then shouldn't we be comparing what is really important and that is how well the different funds capture small and value premiums (HML and SMB)?
Yes, in theory. However, I would expect the actual price/book and market cap of a fund to be a better predictor of future HML and SMB returns. If two indexes hold stocks of about the same size (as the Russell and S&P indexes do), but have different construction rules which lead to different returns (for example, the Russell index did much worse in the 2001 Internet bubble), they will have different HML correlations for reasons unrelated to the size of the stocks they hold.
Wiki David Grabiner

User avatar
Topic Author
grabiner
Advisory Board
Posts: 25411
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: Comparison of small-cap value ETFs

Post by grabiner » Sat Sep 15, 2012 8:19 pm

tpm871 wrote:IJS also tracks S&P 600 Value:

S&P Small-Cap 600 Value (IJS)
Unadjusted expense ratio: 0.29%
Average market cap: $940M, 5% mid-cap, 58% small-cap, 37% micro-cap
P/E: 16.45
P/B: 1.34
Turnover Ratio: 30%
12 month yield: 1.59%

In comparison, VIOV has a 70% turnover and 0.99% yield. My guess is that VIOV trades more aggressively to keep up with changes to its benchmark composition, which is why it is a bit more micro-cap. But the higher turnover may have hidden costs (e.g., less tax efficiency).
According to M*, the VIOV portfolio is as of 6/30 and the IJS portfolio is as of 9/13. The actual portfolios should be about the same.

The yield reported by M* is the distribution yield, not the SEC yield, based on the actual amount of dividends; a growing ETF will have lower distributions per share. VIOV reports a 1.44% SEC yield, and IJS reports a 1.32% yield, which is more consistent with the difference in expenses. (The adjusted expense ratio of IJS is 0.25%.)
Wiki David Grabiner

steve_14
Posts: 1507
Joined: Wed Jun 20, 2012 12:05 am

Re: Comparison of small-cap value ETFs

Post by steve_14 » Sat Sep 15, 2012 8:46 pm

grabiner wrote:Yes, in theory. However, I would expect the actual price/book and market cap of a fund to be a better predictor of future HML and SMB returns. If two indexes hold stocks of about the same size (as the Russell and S&P indexes do), but have different construction rules which lead to different returns (for example, the Russell index did much worse in the 2001 Internet bubble), they will have different HML correlations for reasons unrelated to the size of the stocks they hold.
Unfortunately, as we've seen in recent threads, the notion that you can predict performance based on P/B or P/E is a simplistic one. Two funds can have similar loadings, but greatly different alpha. To say nothing of the fact that the more popular value gets, the better growth seems to do. Might want to reconsider whether there is in fact a pot of gold at the end of this particular rainbow.

User avatar
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by ClosetIndexer » Sat Sep 15, 2012 9:45 pm

steve_14 wrote:Unfortunately, as we've seen in recent threads, the notion that you can predict performance based on P/B or P/E is a simplistic one. Two funds can have similar loadings, but greatly different alpha.
Agreed. (Although that doesn't mean it can't be informative to look at these figures as part of a bigger picture.)
steve_14 wrote:To say nothing of the fact that the more popular value gets, the better growth seems to do. Might want to reconsider whether there is in fact a pot of gold at the end of this particular rainbow.
Value investing has been around for a long time. True, it has recently gotten easier to do with low-cost index funds. Perhaps you're right that this will have an effect on the value premium going forward. Perhaps not. We still see plenty of Facebooks too. Personally I believe we will continue to see a value premium over the very long term. However, at the very least, as long as one achieves their tilt while keeping costs and negative alpha low, they should keep pace with the broad market over the long term. I see no reason to believe there will be a persistent growth premium, so the worst case as I see it is a long term value premium of zero. In that case the difference is your added cost for value tilting. (As well as any psychological effect from not tracking the 'market'.) In the US, this added cost works out to a few bps per year.

Yes, there will certainly be periods when growth outperforms value. However the longer these last, the more of the 'bandwagon' value investors you mention will give it up. That's part of why these kind of cycles happen. Also, outside of bogleheads.org, I don't know if value investing is as popular as you think. Looking at the total investment in various split value and growth funds, they look pretty similar.

User avatar
grap0013
Posts: 1884
Joined: Thu Mar 18, 2010 1:24 pm

Re: Comparison of small-cap value ETFs

Post by grap0013 » Sat Sep 15, 2012 11:14 pm

No love for PXSV? Check out these live FF factor loads in this thread: http://www.bogleheads.org/forum/viewtop ... &p=1468207

Also, FF sort by P/B so that will likely underestimate the true valueyness of PXSV as it utilizes an all-four sort. This is a much better way to do it IMO as P/E and P/CF sorts historically have avoided negative alpha more than P/B and P/D. Also, since it utilizes the past 5 years data for weighting its stocks it helps smooth the data and keep turnover low. Finally, it's pretty cheap with an ER of 0.39.
There are no guarantees, only probabilities.

User avatar
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by ClosetIndexer » Sun Sep 16, 2012 12:00 am

grap0013 wrote:No love for PXSV? Check out these live FF factor loads in this thread: http://www.bogleheads.org/forum/viewtop ... &p=1468207

Also, FF sort by P/B so that will likely underestimate the true valueyness of PXSV as it utilizes an all-four sort. This is a much better way to do it IMO as P/E and P/CF sorts historically have avoided negative alpha more than P/B and P/D. Also, since it utilizes the past 5 years data for weighting its stocks it helps smooth the data and keep turnover low. Finally, it's pretty cheap with an ER of 0.39.
Indeed sorting by multiple parameters is likely a good idea for an index. MSCI and S&P use multiple characteristics for their value sorts too. IIRC, Larry Swedroe mentioned that DFA is looking into doing so as well. However, according to the original Fama-French research (Fama and French 1992a), these other factors don't add anything in terms of explaining results, on average. So as long as regressions using the standard BtM-based HmL factor have high R^2s, they should provide useful comparisons.

Unfortunately the regression results in the thread you linked to don't include any measurements of confidence - R^2 or t-values/p-values, so we have no idea how precise they might be. That said, that fund has a MER of 0.8, which seems pretty enormous compared to funds like VBR and VIOV.

User avatar
grap0013
Posts: 1884
Joined: Thu Mar 18, 2010 1:24 pm

Re: Comparison of small-cap value ETFs

Post by grap0013 » Sun Sep 16, 2012 6:57 am

ClosetIndexer wrote:
Indeed sorting by multiple parameters is likely a good idea for an index. MSCI and S&P use multiple characteristics for their value sorts too. IIRC, Larry Swedroe mentioned that DFA is looking into doing so as well. However, according to the original Fama-French research (Fama and French 1992a), these other factors don't add anything in terms of explaining results, on average. So as long as regressions using the standard BtM-based HmL factor have high R^2s, they should provide useful comparisons.

Unfortunately the regression results in the thread you linked to don't include any measurements of confidence - R^2 or t-values/p-values, so we have no idea how precise they might be. That said, that fund has a MER of 0.8, which seems pretty enormous compared to funds like VBR and VIOV.
Care to run the numbers to see if it's statistically significant? Don't go back too far though as this ETF used to be something else. Also, current ER for PXSV is 0.39 according to http://www.powershares.com. VBR is 0.21, but PXSV has much higher loadings and I'm confident they won't change their index strategy to shoot those loadings in the foot. I hold in tax deferred until I know long term plan for ER.
There are no guarantees, only probabilities.

User avatar
Topic Author
grabiner
Advisory Board
Posts: 25411
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: Comparison of small-cap value ETFs

Post by grabiner » Sun Sep 16, 2012 9:08 am

grap0013 wrote:Care to run the numbers to see if it's statistically significant? Don't go back too far though as this ETF used to be something else. Also, current ER for PXSV is 0.39 according to http://www.powershares.com. VBR is 0.21
Adjusted for BDCs, VBR is 0.10, and I think PXSV is still 0.39. The prospectus doesn't say anything about acquired fund fees and expenses, but the 0.39% includes a waiver, and the waiver would not normally apply to acquired fees as that would discourage management from purchasing BDCs. (BRSIX, for example, has expenses limited to 0.75% but reports 0.76% expenses including 0.01% in acquired fees.) I am adding it to the chart.
but PXSV has much higher loadings and I'm confident they won't change their index strategy to shoot those loadings in the foot. I hold in tax deferred until I know long term plan for ER.
I would keep it in tax-deferred anyway; fundamental weighting creates higher turnover and makes capital gains more likely, and price/dividend weighting increases dividend yields. (Edit: the five-year smoothing may reduce both of those effects; PXSV has a higher dividend yield than RZV, but lower than the conventional small-cap value indexes.)
Last edited by grabiner on Sun Sep 16, 2012 9:18 am, edited 1 time in total.
Wiki David Grabiner

User avatar
Topic Author
grabiner
Advisory Board
Posts: 25411
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: Comparison of small-cap value ETFs

Post by grabiner » Sun Sep 16, 2012 9:17 am

grap0013 wrote:No love for PXSV? Check out these live FF factor loads in this thread: http://www.bogleheads.org/forum/viewtop ... &p=1468207

Also, FF sort by P/B so that will likely underestimate the true valueyness of PXSV as it utilizes an all-four sort. This is a much better way to do it IMO as P/E and P/CF sorts historically have avoided negative alpha more than P/B and P/D.
Despite the weighting, it doesn't have a small or value advantage over the Russell 2000 Value (VTWV):

Average market cap: $816M for VTWV, $855M for PXSV
P/E: 13.64 for VTWV, 13.21 for PXSV
P/B: 1.11 for VTWV, 1.19 for PXSV
P/Cash Flow: 5.02 for VTWV, 5.31 for PXSV
Dividend yield: 2.35 for VTWV, 1.98 for PXSV

So VTWV has stocks of the same size as PXSV, and better ratios on three of the four value metrics, for a lower cost. The five-year smoothing that PXSV does to reduce turnover may be the reason that it loses much of its value weighting.

Over the history of PXSV, it seems to have tracked the Russell 2000 Value Index fairly well, based on a chart of PXSV against IWN (iShares Russell 2000 Value Index ETF, which has a longer history than VTWV).
Wiki David Grabiner

Jerry_lee
Posts: 915
Joined: Tue Mar 06, 2012 1:55 pm

Re: Comparison of small-cap value ETFs

Post by Jerry_lee » Sun Sep 16, 2012 10:07 am

ClosetIndexer wrote:
grap0013 wrote:No love for PXSV? Check out these live FF factor loads in this thread: http://www.bogleheads.org/forum/viewtop ... &p=1468207

Also, FF sort by P/B so that will likely underestimate the true valueyness of PXSV as it utilizes an all-four sort. This is a much better way to do it IMO as P/E and P/CF sorts historically have avoided negative alpha more than P/B and P/D. Also, since it utilizes the past 5 years data for weighting its stocks it helps smooth the data and keep turnover low. Finally, it's pretty cheap with an ER of 0.39.
Indeed sorting by multiple parameters is likely a good idea for an index. MSCI and S&P use multiple characteristics for their value sorts too. IIRC, Larry Swedroe mentioned that DFA is looking into doing so as well. However, according to the original Fama-French research (Fama and French 1992a), these other factors don't add anything in terms of explaining results, on average. So as long as regressions using the standard BtM-based HmL factor have high R^2s, they should provide useful comparisons.

Unfortunately the regression results in the thread you linked to don't include any measurements of confidence - R^2 or t-values/p-values, so we have no idea how precise they might be. That said, that fund has a MER of 0.8, which seems pretty enormous compared to funds like VBR and VIOV.

What is more important that single vs. multi factor sort is the absolute level of the price break. Most indexes split the market into halves, and would be better off using 1/3rds or even 1/4rs. Vericimetry, for example, believes a 1/4 sort using multiple factors is the preferred way to go (see here: http://www.vericimetry.com/risk-premiums/). DFA still prefers mostly a single factor sort (see here: http://www.dimensional.com/famafrench/2 ... tocks.html).

Looking at YTD numbers, we see that these two approaches are relatively close, and both are superior to breaking the market in two (50/50 growth & value per Russell):

Total Stock Index = +18.4%
Russell 2000 Value (multi factor "relative value")= +18.3%
DFA US Small Value (single factor "deep value") = +20.7%
Vericimetry US Small Value (multi factor "deep value") = +20.0%

While not able to test the single/multifactor sort on large stocks until Vericimetry releases their LV fund, we can see the difference between bottom 50% and bottom 25% sorts in the large space as well:

Russell 1000 Value (multi factor "relative value") = +18.0%
DFA US Large Value (single factor "deep value") = +20.8%
The most disciplined investor in the world.

psychodoc
Posts: 34
Joined: Tue Jun 10, 2008 11:21 pm

Re: Comparison of small-cap value ETFs

Post by psychodoc » Sun Sep 16, 2012 10:15 am

any thoughts on jkl ?

Park
Posts: 682
Joined: Sat Nov 06, 2010 4:56 pm

Re: Comparison of small-cap value ETFs

Post by Park » Sun Sep 16, 2012 10:19 am

Code: Select all

                     DFLVX   RPV    DFSVX     RZV
        R Square     0.94    0.91   0.97      0.86
        Alpha       -0.31%   0.35% -0.13%    -0.31%
        Mkt          1.06    1.03   1.00      1.00
        SmB         -0.02    0.21   0.81      0.90
        HmL          0.58    0.92   0.62      1.09
        Mom         -0.11   -0.23  -0.07     -0.30
The above is from the following thread:

http://www.bogleheads.org/forum/viewtop ... t=RZV#wrap

I own RZV, because my tax advantaged space is limited. But a problem with it is its comparatively high exposure to negative momentum. If it had a momentum screen, like DFA, that would help. Such a screen might decrease its value exposure, but I would guess that a happy medium could be found.

steve_14
Posts: 1507
Joined: Wed Jun 20, 2012 12:05 am

Re: Comparison of small-cap value ETFs

Post by steve_14 » Sun Sep 16, 2012 10:35 am

ClosetIndexer wrote:Value investing has been around for a long time. True, it has recently gotten easier to do with low-cost index funds. Perhaps you're right that this will have an effect on the value premium going forward. Perhaps not.
My interest is really in actual fund results, and the academic papers (some going back many decades) show no value premium there, at least as defined by P/B. What shows up in the academic data often isn't translated into real world results. A couple of examples from Fama and French: http://www.dimensional.com/famafrench/2 ... .html#more .
ClosetIndexer wrote:We still see plenty of Facebooks too.
That of course is an IPO, which I agree exhibits pricing anomalies before they're allowed to trade freely. There's a reason most indexes wait a while before including them. A better example might be the 930 mini-Facebooks that comprise the Vanguard Small Growth Index. Over the last 10 years, they've outperformed just about everything else. Or look at how the DFA International SV fund stacks up against its foreign small cap blend competitors (with higher P/E, P/B etc) over the last 15 years.
ClosetIndexer wrote:Personally I believe we will continue to see a value premium over the very long term. However, at the very least, as long as one achieves their tilt while keeping costs and negative alpha low, they should keep pace with the broad market over the long term. I see no reason to believe there will be a persistent growth premium, so the worst case as I see it is a long term value premium of zero. In that case the difference is your added cost for value tilting. (As well as any psychological effect from not tracking the 'market'.) In the US, this added cost works out to a few bps per year.
I expect neither a value or a growth premium going forward in real fund results. In terms of costs of value tilting, you have higher ER (minor), tax drag, but most importantly you lose diversification - if growth outperforms, the value/growth investor has plenty of it, while the value only investor loses. But it's really a matter of degree. A modest tilt, as you say, isn't going to harm (or help) much. Your stock/bond and US/foreign split will probably be much more important.
ClosetIndexer wrote:Yes, there will certainly be periods when growth outperforms value. However the longer these last, the more of the 'bandwagon' value investors you mention will give it up. That's part of why these kind of cycles happen. Also, outside of bogleheads.org, I don't know if value investing is as popular as you think. Looking at the total investment in various split value and growth funds, they look pretty similar.
Right, the phenomenon is mainly promoted by a single fund company and its retail affiliates to a group of retail investors. Vanguard certainly doesn't buy it. The hedge funds and other institutional investors, the ones that scoop up any free lunch as soon as it's discovered, don't seem to take it seriously either. But we can't say for sure. The market may well have adjusted prices to account for an value premium with FF "discovered" it. Some companies must have higher P/Bs than others in an efficient market. Based on future, unknowable information, it will turn out that Wall Street either over or under estimated these relative values. Time, and random variation, will tell.

Jerry_lee
Posts: 915
Joined: Tue Mar 06, 2012 1:55 pm

Re: Comparison of small-cap value ETFs

Post by Jerry_lee » Sun Sep 16, 2012 12:15 pm

steve_14 wrote:
ClosetIndexer wrote:Value investing has been around for a long time. True, it has recently gotten easier to do with low-cost index funds. Perhaps you're right that this will have an effect on the value premium going forward. Perhaps not.
My interest is really in actual fund results, and the academic papers (some going back many decades) show no value premium there, at least as defined by P/B. What shows up in the academic data often isn't translated into real world results. A couple of examples from Fama and French: http://www.dimensional.com/famafrench/2 ... .html#more .
ClosetIndexer wrote:We still see plenty of Facebooks too.
That of course is an IPO, which I agree exhibits pricing anomalies before they're allowed to trade freely. There's a reason most indexes wait a while before including them. A better example might be the 930 mini-Facebooks that comprise the Vanguard Small Growth Index. Over the last 10 years, they've outperformed just about everything else. Or look at how the DFA International SV fund stacks up against its foreign small cap blend competitors (with higher P/E, P/B etc) over the last 15 years.
ClosetIndexer wrote:Personally I believe we will continue to see a value premium over the very long term. However, at the very least, as long as one achieves their tilt while keeping costs and negative alpha low, they should keep pace with the broad market over the long term. I see no reason to believe there will be a persistent growth premium, so the worst case as I see it is a long term value premium of zero. In that case the difference is your added cost for value tilting. (As well as any psychological effect from not tracking the 'market'.) In the US, this added cost works out to a few bps per year.
I expect neither a value or a growth premium going forward in real fund results. In terms of costs of value tilting, you have higher ER (minor), tax drag, but most importantly you lose diversification - if growth outperforms, the value/growth investor has plenty of it, while the value only investor loses. But it's really a matter of degree. A modest tilt, as you say, isn't going to harm (or help) much. Your stock/bond and US/foreign split will probably be much more important.
ClosetIndexer wrote:Yes, there will certainly be periods when growth outperforms value. However the longer these last, the more of the 'bandwagon' value investors you mention will give it up. That's part of why these kind of cycles happen. Also, outside of bogleheads.org, I don't know if value investing is as popular as you think. Looking at the total investment in various split value and growth funds, they look pretty similar.
Right, the phenomenon is mainly promoted by a single fund company and its retail affiliates to a group of retail investors. Vanguard certainly doesn't buy it. The hedge funds and other institutional investors, the ones that scoop up any free lunch as soon as it's discovered, don't seem to take it seriously either. But we can't say for sure. The market may well have adjusted prices to account for an value premium with FF "discovered" it. Some companies must have higher P/Bs than others in an efficient market. Based on future, unknowable information, it will turn out that Wall Street either over or under estimated these relative values. Time, and random variation, will tell.
Yobria,

We've conclusively dismissed these falsehoods. Restating them frequently doesn't lessen their inaccuracies. While I know you intended these for Closetindexer, they roll off the keyboard for me having stated them so frequently in the last few months, so here it goes again:

1. The vast majority of academic studies document a value premium. One paper found that the value premium didn't show up in live active funds over a period of time. What active managers are or are not able to deliver is meaningless to the topic at hand. We are passive investors, so your "evidence" is simply a straw man. The same paper you mention did find a value premium in large cap indexes and small cap indexes, which has continued to be the case over the entire period we have to measure including the 6 years post article.

And, your attempt to equate an anomaly with no link to risk (momentum) with risk/return relationships (equity, size, and value) is a clever slight-of-hand. But we all know the difference. Whether any anomaly persists is anyone's guess--we'll leave that to all your hedge fund buddies you revere so much. But as investors, we have reasons to expect risk to be rewarded with a return, so we confidently hold broadly diversified small and value titled portfolios, and can easily appreciate the difference between investing (our approach) and speculating (yours/your hedge fund friends).

2. That Vanguard's Small Growth Index did as well as it did relative to Vanguard Small Value was due to the index family switch Vanguard stuck us with in 2003. Looking at a small growth index that was unchanged over this period (Russell 2000 Growth) compared to a pure Small Value fund that was also unchanged (DFA US Small Value) shows a small but positive value premium. And even if no value premium resulted in any 10 year period (esp. a period that began after the largest 3 year value premium in many decades) proves absolutely nothing. anymore than TSM underperforming TBM since 2000 does not invalidate the existence of an equity premium.

Further, if we compare the DFA Int'l Small Value fund to the DFA Int'l Small Cap fund, we see a 1% per year higher return over the last 15 years. ISV has bested the EAFE by over 4% per year during this period as well, almost exactly what you'd expect based on historical index simulations and multifactor research. So even this is a very weak example on your part, dismissing your point, not proving it.

3. You can choose to "expect" anything you wish. I am sure you expected a large small and value premium in 2007 when you loaded up on RZV only to be disappointed. I am sure you didn't expect a large small and value premium in early 2009 when you bailed on RZV. In each case, you couldn't have been more wrong. So you'll understand if we aren't putting a lot of stock in your expectations.

Also, there is no evidence with ETFs and TM funds that value index funds are any less tax efficient than holding the market (% lost to taxes since inception):

Russell 3000 ETF: -0.3%
Russell 3000 Value ETF: -0.5%
DFA TM Marketwide Value: -0.3%

And, in a multi factor world where returns come from not just your stock/bond decision, but also exposure to value/growth and small/large, by diversifying not just across thousands of securities (which TSM and value investing both do), but across dimensions of return, you reduce the risk that you have concentrated all your risk in the one dimension that fails to produce (like stocks vs. bonds from 2000-now or 1965-1982). Just like an insurance company who has multiple lines of insurance coverage (home, auto, fire, life, etc.) is a more diversified insurer than one who just insures homes, so too is a multifactor portfolio more diversified than a single factor portfolio.

Finally, you are just flat out wrong about the relative importance of stock/bond and US/foreign over size and value. First, SV has beaten TSM by 4% per year for 80 years, MORE than TSM has outperformed 5YR T-Notes. So how much to tilt to small/value vs. large/growth has actually had a larger portfolio return contribution than stock/bond -- but because they are both risk/return dimensions, I'd say they are equally important. As far as US/Foreign, that is not nearly as important as 3F diversification. Any combo of US/EAFE has had almost identical returns with very little difference in risk. Sinquefield summed this up nicely almost 20 years ago: https://www.savantcapital.com/uploadedF ... 950401.pdf

Probably a good paper for you to familiarize yourself with.

4. Value/growth is almost universally accepted as a risk/return dimension in the academic community. Unfortunately, most Wall Street firms are more concerned with Main Street marketing than evidenced based application. As for Vanguard, who knows what they "buy"? They offer hedge fund like strategies, sector funds for tactical management, etc. Not exactly a firm whose across-the-board principles we'd want to adhere to. What we do know is the industry standard is to sort stocks across size and value dimensions (you've seen the M* style boxes, right?). And we also know, despite the active nature of the execution, that there is a huge component of the hedge fund industry devoted to value investing: long/short, distressed debt, relative value, deep value, etc. But more investors taking a risk doesn't mean that risk will magically disappear. Bankers still charge distressed companies a higher interest rate, and markets still price troubled companies with lower prices. In each case, a higher "cost of capital" means a higher "expected return" for those willing to stick around to earn it. As you know, not everyone is able to do so, however.
The most disciplined investor in the world.

Park
Posts: 682
Joined: Sat Nov 06, 2010 4:56 pm

Re: Comparison of small-cap value ETFs

Post by Park » Sun Sep 16, 2012 2:07 pm

Jerry_lee wrote: Also, there is no evidence with ETFs and TM funds that value index funds are any less tax efficient than holding the market (% lost to taxes since inception):

Russell 3000 ETF: -0.3%
Russell 3000 Value ETF: -0.5%
DFA TM Marketwide Value: -0.3%
As market capitalization decreases, the value premium increases. That's why I'm invested in a small cap value fund, not the Russell 3000 Value ETF, which will be dominated by large caps. And with a small cap value fund, there will be increased turnover compared to a total market fund, which means less tax efficiency.

User avatar
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by ClosetIndexer » Sun Sep 16, 2012 3:05 pm

OK, here's what I get for PXSV since the beginning of the new index on 03/31/2011. Note that there are only 13 samples between then and the end of the FF research factor data, so we can't draw strong conclusions. That said, I'm pretty sure the factors listed in the other thread are not accurate. (It lists beta of 0.09 for one thing, which really doesn't seem possible, and is completely different otherwise than these.)

Code: Select all

                                Mkt-Rf    SmB     HmL     Alpha
    PXSV                         1.04    0.89    0.66     6.07%   R^2 = 0.986
       t-values                 12.66    3.87    2.61     1.70

Note that although that alpha is large, over this small number of samples, it is not technically statistically significant. (p-value of 0.12). So we'll have to wait to get a better idea of what the 'true' alpha is. Big positive beats big negative though I suppose! :) The HmL and especially SmB factors should be better estimates though, although again keep in mind the small sample size. I prefer to do regressions over at least 4 or 5 years.

As for the age-old debate, I don't have anything much to add to what Jerry_lee said, and I agree with Steve that in the worse case long-term-scenario what you're risking is any added costs you pay for your tilt, and that diversification is important. So even when tilting, minimizing costs and maximizing diversification remain the most important principles, as always. Beyond that, I'd rather not see this thread go the way of many others 'discussing' the relative merits of TSM vs multi-factor investing, so let's let people do their own research while keeping this one to a discussion of specific SV funds...?

User avatar
baw703916
Posts: 6681
Joined: Sun Apr 01, 2007 1:10 pm
Location: Seattle

Re: Comparison of small-cap value ETFs

Post by baw703916 » Sun Sep 16, 2012 7:15 pm

I use IWC, and this fund:

iShares Morningstar Small Value (JKL)
Adjusted expense ratio: 0.30%
Average market cap: $1584M, 8% mid-cap, 88% small-cap, 4% micro-cap
P/E: 10.44
P/B: 1.10

(edit, fixed typo)

Brad
Last edited by baw703916 on Mon Sep 17, 2012 8:38 am, edited 1 time in total.
Most of my posts assume no behavioral errors.

User avatar
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by ClosetIndexer » Sun Sep 16, 2012 9:35 pm

baw703916 wrote:I use IWC, and this fund:

iShares Morningstar Small Value (IWC)
Adjusted expense ratio: 0.30%
Average market cap: $1584M, 8% mid-cap, 88% small-cap, 4% micro-cap
P/E: 10.44
P/B: 1.10

Brad
FYI, typo there. The Morningstar ETF is JKL. What was your reasoning for choosing those particular funds?

User avatar
baw703916
Posts: 6681
Joined: Sun Apr 01, 2007 1:10 pm
Location: Seattle

Re: Comparison of small-cap value ETFs

Post by baw703916 » Mon Sep 17, 2012 8:40 am

ClosetIndexer wrote:
baw703916 wrote:I use IWC, and this fund:

iShares Morningstar Small Value (IWC)
Adjusted expense ratio: 0.30%
Average market cap: $1584M, 8% mid-cap, 88% small-cap, 4% micro-cap
P/E: 10.44
P/B: 1.10

Brad
FYI, typo there. The Morningstar ETF is JKL. What was your reasoning for choosing those particular funds?
Oops, I'd copied and pasted from the original post to keep the same format, but forgot to change the ticker.. :P

Basically, I chose it because the M* construction seems to gave more of a value tilt.
Most of my posts assume no behavioral errors.

JRA
Posts: 191
Joined: Fri Aug 03, 2007 3:29 pm

Re: Comparison of small-cap value ETFs

Post by JRA » Mon Sep 17, 2012 10:51 am

Park wrote:
As market capitalization decreases, the value premium increases.
While this has been the case for the last 40 years, it has not always been the case and may not be the case going forward: http://papers.ssrn.com/sol3/papers.cfm? ... _id=686880.

Jerry_lee
Posts: 915
Joined: Tue Mar 06, 2012 1:55 pm

Re: Comparison of small-cap value ETFs

Post by Jerry_lee » Mon Sep 17, 2012 11:18 am

Park wrote:
Jerry_lee wrote: Also, there is no evidence with ETFs and TM funds that value index funds are any less tax efficient than holding the market (% lost to taxes since inception):

Russell 3000 ETF: -0.3%
Russell 3000 Value ETF: -0.5%
DFA TM Marketwide Value: -0.3%
As market capitalization decreases, the value premium increases. That's why I'm invested in a small cap value fund, not the Russell 3000 Value ETF, which will be dominated by large caps. And with a small cap value fund, there will be increased turnover compared to a total market fund, which means less tax efficiency.
That the value premium has been larger in small companies is a US phenomenon, and a byproduct of the last few decades. It has been as large in big companies if we look abroad, and ditto for the US prior to the 60s.

As for the added tax inefficiency of small value, we just aren't seeing that (since inception):

Russell 3000 ETF: -0.3%
Russell 2000 Value ETF: -0.5%
DFA TM Targeted Value: -0.5%

So you see less than 0.2% per year difference in tax efficiency when measured against ~3% or more higher expected returns. Further, looking at ETFs, as TILT gains traction, it won't be necessary to hold a separate SV fund to tilt away from the market. This has already proven to be more tax efficient when we look at DFAs Core equity funds (DFTCX has lost -0.25% to taxes since inception).

Further, if you think about distributions in aggregate, some trading is going to be required for most to maintain US/foreign allocations or stock/bond. Using dividends/distributions for this purpose means that you won't actually realize any additional taxes (compared to just liquidating funds and taking capital gains) if you plan to rebalance whether or not you tilt or hold TSM allocations.
The most disciplined investor in the world.

User avatar
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by ClosetIndexer » Sun Sep 23, 2012 1:45 am

I decided to take a closer look into JKL (the M* SV fund) and RZV (the Rydex S&P 600 Pure Value). On the surface, these look very attractive - less of your portfolio required to achieve your desired tilts. Morningstar discusses here how 'pure' their indexes are, and how great that is. And obviously the Pure Value index is... pure. Of course, one person's pure is another's less diversified.

I did regressions over a few periods to compare over several time periods. JKL first:

Code: Select all

M* (iShares JKL):
                                Mkt-Rf    SmB     HmL     Alpha
    2004/08+                     0.96    0.66    0.57    -0.40%   R^2 = 0.945
       t-values                 24.57    8.52    8.31    -0.20

    2004/08-2007/12              1.02    0.68    0.29    -4.82%   R^2 = 0.911
       t-values                 11.08    6.34    2.62    -2.06

    2008/01+                     0.95    0.57    0.68     4.36%   R^2 = 0.960
       t-values                 20.47    5.26    7.88     1.49
If you just look at the entire period, it looks great. Negative alpha estimate is small and not stastically significant. Basically equal to its management expenses of 0.3% as far as we can tell. (Its tracking error over the entire period is also very low and consistent: around 0.25%, so it's actually tracking closer than its expenses, possibly due to securities lending, or just random variation.) Note though that for that alpha, 'not statistically significant' basically means 'completely useless estimate'. The standard error is equal to the estimate divided by the t-value, so 0.4%/0.2 = 2%. So the estimate could easily be off by 2 or 3%.)

The difficulty is seen when you break it up into periods. I chose 2008 as a convenient halfwayish point. Note the large and significant (p-values under 0.05) variations in alpha. ie: the standard error is still about 2%, but the estimate is varying by far more than that; enough that we know the changes are really happening. (Also interesting is that HmL exposure has also fluctuated a fair bit, but that's not uncommon.)

Next RZV:

Code: Select all

Rydex S&P 600 Pure Value (RZV):
                                Mkt-Rf    SmB     HmL     Alpha
    2006/05+                     1.08    1.37    1.33     2.91%   R^2 = 0.848
       t-values                  9.62    5.58    6.49     0.47

    2006/05-2007/12              1.01    1.03    0.92    -6.95%   R^2 = 0.921
       t-values                  9.17    5.75    5.89    -1.93

    2008/01+                     1.08    1.36    1.37     5.70%   R^2 = 0.847
       t-values                  7.73    4.15    5.23     0.64
We don't have quite as much history here, but we see exactly the same thing going on, just to an even greater extent. First though, tracking error for this fund has been even better. Average about -0.1% since inception, and very consistent, so it is tracking the index surprisingly closely given its management expenses of 0.35%. (Does anyone know where to check how much of that is acquired expenses?) So as with JKL, the variations we're seeing are due to the index itself, not tracking error of the fund. We can also see relatively low R^2s for the regressions here, which also shows us there's stuff going on that's not explained by the model. None of this is particularly surprising given the concentrated nature of the index, at about 150 stocks.

As far as we can tell, long term alpha is fine, but again it's not statistically significant. In this case the error is even larger - standard error of 2.91%/0.47 = 6.19%. So the estimate could easily be off by +/- 5 to 10%. I just mention this since without context, that +3% looks good. But along with the t-value, it again means "we really have no idea". Again though, what we do know is that it's varying significantly over time. If possible I'd like to look at the underlying index history to go back further, but it looks like S&P no longer freely provides this. So, next best thing, Robert T did the same thing back in 2008. Note that Robert's alpha numbers are monthly whereas mine are annual, so multiply his by 12 to get standard annual alphas. His estimate shows the pure value index having about 1.4% greater negative annual alpha long term than the MSCI small value (or the S&P 600 value (non-pure)). That said, there is a bit of a difference between his numbers and what I get for the MSCI index over the same period, and his R^2s are lower, so I'm not 100% confident in them. I've PM'ed him to see if we can figure out the difference.

OK, for comparison, here's VBR:

Code: Select all

Vanguard MSCI Small Value (VBR):
                                Mkt-Rf    SmB     HmL     Alpha
    2004/03+                     0.98    0.63    0.43    -0.54%   R^2 = 0.970
       t-values                 36.32   11.92    9.24    -0.41

    2004/03-2007/12              0.98    0.68    0.37    -1.42%   R^2 = 0.945
       t-values                 14.72    8.88    4.83    -0.86

    2008/01+                     0.98    0.57    0.46     0.93%   R^2 = 0.976
       t-values                 28.68    7.12    7.19     0.43

So we still, unsurprisingly, see variations in our estimates of alpha. But in this case the variations are much smaller and aren't statistically significant. So in this case it's still possible that the true underlying alpha is varying, but definitely not to the extent that we know it is for the first two funds. We have higher R^2, showing that the fund's returns are closely matched by the expectations of the 3F model. Also, the factor loadings have been a bit more stable than the other funds, although this is partially just due to the mid-point chosen.

---

So, I'm not saying any of these are bad funds. All three have low, very consistent tracking errors. (In fact, both VBR and RZV have managed to consistently have tracking error significantly lower than expenses. VBR has averaged positive 0.01% tracking error (beat the index!) with expenses of 0.1% and RZV only lagged by about 0.1%, with expenses of 0.35%. JKL's tracking error was -0.25% - still slightly less than its expenses of 0.3%, but not as stellar as the other two.

JKL and especially RZV are much more concentrated in the companies they hold, which allows them to get higher factor exposures, but also adds a great deal of variation. Hopefully this variation is entirely random and will average out over time, as the model would predict. Given the wide variations in alpha though, there is no way to know for sure whether there is in fact a long-term negative bias. (Or a positive one, but I doubt these funds have accidentally discovered a new source of alpha. :) )

Anyway, if I could get the factor exposure I needed with VBR, I would see no reason to look further. Fairly stable loadings, far more diversification, and much less random variation not explained by its factor loadings. If I really needed greater factor exposure, based on this analysis, I would choose RZV since it gives me higher factor exposures than JKL (meaning I need to hold less of it to meet my portfolio target), with essentially the same management expenses, and for some reason, a consistently lower tracking error. It's more concentrated, but not severely so, at 147 stocks compared to 196 for JKL. (Of course, VBR has 1000 or so.) It also has significantly lower yield (0.67% ttm compared to 2.54% for JKL and 2.01% for VBR), which helps with taxes, although it's necessary to look at the effect on the entire portfolio there, as you would hold different amounts of each fund to achieve the same factor goals.

Of course, it's also important to consider one's tax situation, which I haven't touched on here.

Now as it turns out, I actually would like to get a bit more than 0.4 exposure to HmL on the US side, since I don't have any really any good domestic (Canadian) value tilt options. So at the moment I'm still on the fence as to whether I want to use one of the pure value funds (this one and/or the mid-cap probably) or just stick with the best I can get using broader-based value funds. Hence my interest in this thread! :)

Edit: The Rydex S&P 500 Pure Value ETF has averaged tracking error of -0.47%, and the S&P 400 Pure Value has averaged -0.39% since their inceptions, so I'm guessing that excellent tracking error for the small cap fund is an anomaly. (I can't imagine the small-cap fund doing a better job of tracking its index, long term, than the mid and large.)

Edit 2: Thanks grabiner for reminding me to check the prospectus. (It was late. ;) ) RZV's acquired expenses are insignificant, so I still don't have a good explanation for how RZV has managed to track its index so closely given its expenses, unless it gets significantly more revenue from securities lending or some other source than the mid and large pure value funds. Otherwise I would have to chalk it up to randomness that has happened to be slightly positive on average since 2006. One year tracking error was -0.29%, which seems more realistic. I'll dig into old annual reports to see if I can find a clue when I have a chance. (Unless someone else wants to... :) )

Edit 3: Update - just found the S&P 600 Value and Pure value index data! It is on their old site, it's just hidden behind several layers of obfuscation and stupidity. Will do some comparisons of the MSCI SCV (VBR), S&P 600 Value (VIOV/IJS) and S&P 600 PV (RVZ) historical index returns and report back a bit later. (Done below.)

Edit 4: As seen below, RZV's index has had large negative momentum - average of -0.3 since 1997, and ever greater, about -0.55, since the fund's inception. JKL's has been much lower, around -0.15. So if I had to choose between the two, I would definitely pick JKL now, not RZV. Still prefer VBR though, with a negative momentum exposure of only -0.07, and for the reasons above.
Last edited by ClosetIndexer on Wed Sep 26, 2012 8:17 pm, edited 3 times in total.

User avatar
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by ClosetIndexer » Sun Sep 23, 2012 5:46 am

Fun fact: The S&P mid cap and large cap Pure Value ETFs are both actually small too. Just less so:

Code: Select all

Large (RPV):
                                Mkt-Rf    SmB     HmL     Alpha
    2006/05+                     1.11    0.24    0.96     2.83%   R^2 = 0.878
       t-values                 14.40    1.40    6.80     0.66

    2006/05-2007/12              0.84    0.39    0.51    -0.57%   R^2 = 0.877
       t-values                  8.71    2.49    3.71    -0.18

    2008/01+                     1.13    0.15    1.02     5.53%   R^2 = 0.884
       t-values                 11.97    0.67    5.74     0.93


Mid (RFV):
                                Mkt-Rf    SmB     HmL     Alpha
    2006/04+                     1.08    0.56    0.81     2.67%   R^2 = 0.908
       t-values                 16.60    3.90    6.87     0.74

    2006/04-2007/12              0.91    0.72    0.53     1.93%   R^2 = 0.904
       t-values                  9.33    4.51    4.12     0.61

    2008/01+                     1.10    0.48    0.86     4.49%   R^2 = 0.911
       t-values                 13.62    2.57    5.69     0.88


User avatar
baw703916
Posts: 6681
Joined: Sun Apr 01, 2007 1:10 pm
Location: Seattle

Re: Comparison of small-cap value ETFs

Post by baw703916 » Sun Sep 23, 2012 10:01 am

ClosetIndexer,

Thank you for your excellent analysis! It will take me a little while to digest your posts and to respond, but I wanted to give you a quick word of thanks.

Best wishes,
Brad
Most of my posts assume no behavioral errors.

User avatar
Topic Author
grabiner
Advisory Board
Posts: 25411
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: Comparison of small-cap value ETFs

Post by grabiner » Sun Sep 23, 2012 12:05 pm

ClosetIndexer wrote:We don't have quite as much history here, but we see exactly the same thing going on, just to an even greater extent. First though, tracking error for this fund has been even better. Average about -0.1% since inception, and very consistent, so it is tracking the index surprisingly closely given its management expenses of 0.35%. (Does anyone know where to check how much of that is acquired expenses?)
It's in the prospectus, which has a breakdown of expenses. Morningstar has links to prospectuses on the Filings page for all mutual funds it covers; the Summary Prospectus is usually sufficient.

For RZV, the amount is 0.01%, as the S&P index includes very few business development companies; this is why I reported the 0.35% as adjusted expenses from the reported 0.36%.
Wiki David Grabiner

User avatar
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by ClosetIndexer » Mon Sep 24, 2012 12:48 am

Just came across something else interesting here. The reason why the S&P Pure Value indexes have consistently lower turnover than the value indexes is because their constituents are ranked only on their value score. Some other interesting tidbits in there too.
Since Pure Style indices are score-weighted, weights (and, therefore, Modified Index
Shares) of individual stocks are not be affected by corporate actions such as stock splits,
spin-offs and rights offerings. Between rebalancings, the PWF might be adjusted to
ensure there is no change in a stock’s Modified Index Shares after such a corporate
action. This ensures that, in practical terms, most corporate actions do not necessitate
any action on the part of a portfolio manager tracking the index. Because of this feature,
this series has lower number of turnover events in a given year than the Style index
series.
S&P U.S. STYLE INDICES - INDEX METHODOLOGY

And while I'm at it, here's a handy visualization of all the major indexes that I glace at fairly often. Includes some analysis too:
This paper examines how the major index providers construct and
maintain their measures. We also offer insight into the limitations and
challenges of certain indexing methods.
Determining the appropriate benchmark: A review of major market indexes (by Vanguard)

User avatar
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by ClosetIndexer » Tue Sep 25, 2012 4:23 am

OK, so as it turns out, S&P does still post historical returns for their style funds, it's just hidden behind several layers of obfuscation and stupidity. It's also going away for good in a few days, so if you think you might ever want historical S&P index data, better go grab it now!

Anyway, moving on. I found the discrepancy between my data and Robert T's, so it looks like our results agree. I expanded the regressions he did in 2008 to include data from 1997/05 (the earliest month I had returns for the S&P 600 Value fund) until June 2012. Here's what I got. Note that alphas are annual, not monthly.

Code: Select all

1997/05 - 2012/06 monthly data

                                Mkt-Rf   SmB     HmL    Alpha (annual)
    MSCI SMALL VALUE (VBR)       0.96    0.40    0.73    -0.24%     R^2 = 0.932
       Std. Error                0.02    0.03    0.03     1.32%

    S&P 600 VALUE (VIOV/IJS)     0.98    0.60    0.62    -0.89%     R^2 = 0.917
       Std. Error                0.03    0.04    0.04     1.57%

    S&P 600 PURE VALUE (RZV)     1.13    0.86    1.20    -2.05%     R^2 = 0.816
       Std. Error                0.05    0.07    0.08     3.11%
I decided to start showing confidence in terms of Standard Error instead of t-values, since to me it gives a better illustration of the confidence range. Same information though. t-value = Estimate / Std. Error. To get a feel for what these Standard Errors mean, see here. For a large number of samples (which we have), a t-distribution approximates a normal distribution, so we can assume about 68% of the time the true value will fall within 1 standard deviation of the estimate. 95% of the time it would be within two standard deviations. (In our situation, standard deviation is equal to standard error.)

So, we do see a trend in the alphas, with the least concentrated fund looking best, and having the narrowest error band. But the size of the error on all three estimates is large enough that the true alphas for all three could very easily be the same. So why is it that we're able to get such good estimates of the average factor loadings over this period but not of the alphas? Two (related) reasons. First, the alpha estimate depends very heavily on the factor estimates. If one of our factor estimates is off slightly and there was a large (positive or negative) return for that factor, it could cause a large difference in alpha. Second, the factor loadings of funds change over time. By doing one regression over a long period, we are finding a single set of average factors, when in reality they have varied throughout that period. Since alpha depends on our factor estimates, and we're not capturing these variations in factor loadings, it's not possible to estimate alpha very precisely.

An example: say we're looking at a blend fund over a ten year period. For the first 5 years, the fund had a small positive value loading, and for the second 5 years, it had a small negative value loading, just randomly due to the nature of the companies it happens to hold. Then say for the first five years, value stocks happened to do well, and over the second five years, growth stocks happened to do well. If we do two separate regressions over the two five year periods, the results should be pretty accurate. The first period will show the value loading and alpha around zero, and the second period will show the growth loading and alpha around zero. HOWEVER, if we do a single regression over the full period, it will correctly estimate the value loading (average) to be about zero, and will find the alpha was positive! Because even though it had no value loading over the period (on average), it somehow managed to beat the market. (Or in other words, it outperformed the expectation of its average factor loadings over the period, hence alpha.) (Note: this is why we prefer to pick funds whose factor loadings don't vary too much. It removes a source of risk, in that there's less chance your fund's value loading will dip over the periods when value outperforms, and vice-versa. I imagine the best funds for this are the DFA funds, since they're designed to have consistent factor loadings. That said, they still do vary.)

So long time periods are good because they give us lots of samples, allowing for greater precision. But they're also bad, because factor loadings change over time, so although we can get good estimates for the average factor loadings over the period, we still don't know what the true alpha of the fund was, relative to is changing factor loadings over time.

I'm thinking one way to try to get around this would be to do a bunch of rolling regressions over shorter periods and average the results. I'm going to write a little script to do that at some point here, but first, there is one other way to tell how much of an effect the diversification (or lack thereof) of these funds is having.

We can build theoretically identical simple portfolios using several SV funds and compare their returns and standard deviations. To do this, I took the underlying indexes of the three funds that interest me most, VBR, VIOV/IJS, and RZV, and combined them with a combination of the MSCI Large and MSCI Small indexes, such that all three portfolios ended up with the same factor weightings, on average, over the period of May 1997 - June 2012. (Would have included JKL, but I don't believe the M* index data is available, or goes back that far.) Here they are:

Code: Select all

Portfolios       MSCI Small   MSCI Large   MSCI SV   S&P SV   S&P PV    t-bill   |   Mkt-Rf    HmL    SmB
P.MSCIV (VBR):   49.29%       11.72%       36.85%                        2.13%   |   0.95      0.4    0.4
P.SPV (VIOV):    16.32%       24.14%                 59.48%              0.07%   |   0.95      0.4    0.4
P.SPPV (RZV):    43.05%       26.35%                          24.61%     5.99%   |   0.95      0.4    0.4
So you see the three portfolios, the percent weighting they have to each index, and the fact that all the portfolios have identical factor weightings. (A bit of t-bills is used because the betas (Mkt-Rf factors) of the indexes are slightly different.) Now, this isn't perfect, because the MSCI Small and Large cap indexes don't necessarily have exactly zero alpha, so they will affect the results somewhat. Still though, this does give an idea of how these funds might perform as part of a real-world portfolio.

Here are the annualized returns and standard deviations for the three portfolios over the May 1997 - June 2012 period, along with the MSCI total investible market, for comparison.

Code: Select all

                     P.MSCIV   P.SPV    P.SPPV   TSM
Return (annual):      8.81%     8.21%    8.30%    6.09%
SD (monthly):         5.39%     5.45%    5.41%    4.93%
SD (annualized):     18.67%    18.87%   18.73%   17.07%
Sharpe Ratio:         0.33      0.29     0.30     0.20
So, the main thing is, the returns and standard deviations of all three portfolios are very similar... Which is exactly what the 3F model says should happen. All three portfolios beat TSM on a risk-adjusted basis, not surprising given the overall performance of SV over the period. The winner in this test is the MSCI Small Value index. However, any variation in returns over this period is subject to the same issue I described with the alphas above. It is possible that the difference is due to actual systematic inefficiencies in one index vs another, but it's also possible that the MSCI index just happened to fluctuate its value loading at opportune times. There is no guarantee that the higher return will persist.

What's more interesting to me are the Standard Deviation numbers. All three portfolios are nearly identical, meaning the lesser diversification of the pure value index does not result in greater volatility, which surprised me. The logic behind that is that you need much less of it to reach the same overall factor loadings, but I still expected there to be some non-systematic risk from its level of concentration that would show up in the SD of the portfolio. Instead we also see essentially no difference between the standard deviations of all three portfolios.

So one should be able to build a portfolio with the same factor loadings and the same risk using funds tracking any of these indexes. That said, I still prefer VBR, but not as much due to that historical outperformance as for two other reasons:

First, I used the *index* returns, not the fund returns, for this comparison to allow for a longer period to be studied. VBR has had tracking error very near zero, while IJS has consistently been at -0.2%, and so far VIOV looks very close to that at -0.19%. RZV so far has averaged -0.1%, but has varied a bit more, and evidence from the mid and large pure value funds, as well as RZV's management expenses, suggests that somewhere in the -0.2% to -0.3% range would be a more likely result going forward.

Second, the regressions to determine the factor loadings of the funds was done over the the same period that we tested the results over. That guarantees that the factor weights of our test portfolios are the same over the period we're looking. In the real world, we have to estimate factor loads for funds using past data, then get our returns from the future. If the factor weights change in the mean time, it can alter our risk and return. This is bound to happen to some extent, but the more they vary, the more further our results might be from our expectations. As I mentioned (way) above, that's why it's safest to choose funds with relatively stable factor weightings. (In this case that's VBR, but we do have to remember it's the factor weights of the entire portfolio that matter, so the variations of RZV will have less effect since we'll own less of it.) When I'm able to write that script to do rolling regressions we should be able to get a better idea how much the weights for each index have changed over time.

All that said, I am somewhat more convinced now than before that if there were other reasons to choose a different fund - particularly tax efficiency and/or limited portfolio space in which to achieve tilt - RZV (or VIOV) could be a good choice.

WHEW! Hope this example helped someone! I know I found it interesting.

User avatar
Ketawa
Posts: 2192
Joined: Mon Aug 22, 2011 1:11 am
Location: DC

Re: Comparison of small-cap value ETFs

Post by Ketawa » Tue Sep 25, 2012 7:04 am

ClosetIndexer,

Thanks for this analysis, fascinating stuff, as well as for the link to the S&P index data.

Have you considered running the above simulation using momentum factors? VBR and RZV have historically had a negative MOM loading, while it is typically close to zero for VIOV/IJS. Was curious how it would affect the results.

User avatar
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by ClosetIndexer » Tue Sep 25, 2012 8:33 pm

Ketawa wrote:ClosetIndexer,

Thanks for this analysis, fascinating stuff, as well as for the link to the S&P index data.

Have you considered running the above simulation using momentum factors? VBR and RZV have historically had a negative MOM loading, while it is typically close to zero for VIOV/IJS. Was curious how it would affect the results.
Excellent point. I tend to just use 3F myself most of the time since internationally I don't have access to momentum factors. But since we're just looking at US funds, we can do that. And while it's very difficult to pointpoint alpha for the reasons described above, negative momentum is much easier to get a handle on.

Code: Select all

   1997/05 - 2012/06 monthly data
	Without Momentum:
                                Mkt-Rf   SmB     HmL    Alpha (annual)
    MSCI SMALL VALUE (VBR)       0.96    0.40    0.73    -0.24%     R^2 = 0.932
       Std. Error                0.02    0.03    0.03     1.32%

    S&P 600 VALUE (VIOV/IJS)     0.98    0.60    0.62    -0.89%     R^2 = 0.917
       Std. Error                0.03    0.04    0.04     1.57%

    S&P 600 PURE VALUE (RZV)     1.13    0.86    1.20    -2.05%     R^2 = 0.816
       Std. Error                0.05    0.07    0.08     3.11%

Code: Select all

   1997/05 - 2012/06 monthly data
	With Momentum:
                                Mkt-Rf   SmB     HmL     Mom    Alpha (annual)
    MSCI SMALL VALUE (VBR)       0.92    0.42    0.71   -0.08    0.33%     R^2 = 0.939
       Std. Error                0.02    0.03    0.03    0.02    1.24%
                                                         
    S&P 600 VALUE (VIOV/IJS)     0.95    0.62    0.60   -0.07   -0.47%     R^2 = 0.922
       Std. Error                0.03    0.04    0.04    0.02    1.51%
                                                         
    S&P 600 PURE VALUE (RZV)     0.99    0.92    1.11   -0.31   -0.07%     R^2 = 0.864
       Std. Error                0.05    0.07    0.07    0.04    2.66%
VBR and VIOV's indexes are basically unchanged. A bit of negative momentum, which was showing up in the alpha, and isn't surprising. Factor loadings basically the same. In the Pure Value (RZV) case though, it looks like what was showing up as added market exposure (Mkt beta of 1.13) was in fact negative momentum. So in that case we get beta of 1, alpha of essentially zero, but large negative momentum.

Glad you pointed this out, because it strongly suggests that that negative 3F alpha is largely due to momentum, which means it's not a random variation, but likely to persist. (ie the momentum 'premium' has historically been around 9.7% IIRC, so that -0.31 momentum means a drag of 3% per year.) I had hoped that perhaps S&P made methodology changes to improve this since the start of our data, but I did a couple tests over shorter, more recent periods, and -0.3 was actually the best case. Since 2004 it's been closer to -0.6!

Comments or criticisms welcome!

Note that this doesn't really affect the portfolios example above, because what I'm showing there are the actual results of the indexes as part of portfolios with the same factor weightings. The only factor that really changes based on the switch to 4F is the Mkt factor of the Pure Value index, which only slightly changes the portfolio weights, and ends up with essentially the same results over the test period. So the addition of momentum doesn't change our results, it just helps to explain them, and gives us more confidence that the Pure Value index will likely lag the others by a significant amount; IE the larger alpha we see for it in the 3F analysis isn't an anomaly, but is largely due to its larger negative momentum.

User avatar
Topic Author
grabiner
Advisory Board
Posts: 25411
Joined: Tue Feb 20, 2007 11:58 pm
Location: Columbia, MD

Re: Comparison of small-cap value ETFs

Post by grabiner » Wed Sep 26, 2012 10:08 pm

ClosetIndexer wrote:We can build theoretically identical simple portfolios using several SV funds and compare their returns and standard deviations. To do this, I took the underlying indexes of the three funds that interest me most, VBR, VIOV/IJS, and RZV, and combined them with a combination of the MSCI Large and MSCI Small indexes, such that all three portfolios ended up with the same factor weightings, on average, over the period of May 1997 - June 2012. (Would have included JKL, but I don't believe the M* index data is available, or goes back that far.) Here they are:

Code: Select all

Portfolios       MSCI Small   MSCI Large   MSCI SV   S&P SV   S&P PV    t-bill   |   Mkt-Rf    HmL    SmB
P.MSCIV (VBR):   49.29%       11.72%       36.85%                        2.13%   |   0.95      0.4    0.4
P.SPV (VIOV):    16.32%       24.14%                 59.48%              0.07%   |   0.95      0.4    0.4
P.SPPV (RZV):    43.05%       26.35%                          24.61%     5.99%   |   0.95      0.4    0.4
And the next issue is the cost of these portfolios. MSCI small costs 10 basis points, and MSCI large effectively costs 4 basis points (because you pay 5 basis points for VTI, which is about 80% large and 20% small). Thus the VBR-based portfolio costs 9 basis points, the VIOV-based portfolio costs 14 basis points, and the RZV-based potfolio costs 14 basis points. (This assumes no alpha; the alphas are all within experimental error of zero, so that's a reasonable assumption.)

Conclusion: If you can hold enough VBR (MSCI small value) to get your desired small-cap and value tilt, that is the lowest-cost way to get the tilt. If you can't (limited tax-deferred room, want a larger tilt than VBR itself gives), then use RZV, which requires the smallest holding to get the desired tilt.

Do you have three-factor data for the Russell 2000 Value? This appears to have a better value weighting than the S&P 600 Value, and Vanguard offers it at the same cost. (Long-term regressions of the Russell 2000 should have negative alpha, and Vanguard itself proved that by outperforming the index by 1% per year when Small-Cap Index tracked it; I believe the index is now less vulnerable to front-running.)
Wiki David Grabiner

User avatar
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by ClosetIndexer » Wed Sep 26, 2012 11:37 pm

grabiner wrote:
ClosetIndexer wrote:We can build theoretically identical simple portfolios using several SV funds and compare their returns and standard deviations. To do this, I took the underlying indexes of the three funds that interest me most, VBR, VIOV/IJS, and RZV, and combined them with a combination of the MSCI Large and MSCI Small indexes, such that all three portfolios ended up with the same factor weightings, on average, over the period of May 1997 - June 2012. (Would have included JKL, but I don't believe the M* index data is available, or goes back that far.) Here they are:

Code: Select all

Portfolios       MSCI Small   MSCI Large   MSCI SV   S&P SV   S&P PV    t-bill   |   Mkt-Rf    HmL    SmB
P.MSCIV (VBR):   49.29%       11.72%       36.85%                        2.13%   |   0.95      0.4    0.4
P.SPV (VIOV):    16.32%       24.14%                 59.48%              0.07%   |   0.95      0.4    0.4
P.SPPV (RZV):    43.05%       26.35%                          24.61%     5.99%   |   0.95      0.4    0.4
And the next issue is the cost of these portfolios. MSCI small costs 10 basis points, and MSCI large effectively costs 4 basis points (because you pay 5 basis points for VTI, which is about 80% large and 20% small). Thus the VBR-based portfolio costs 9 basis points, the VIOV-based portfolio costs 14 basis points, and the RZV-based potfolio costs 14 basis points. (This assumes no alpha; the alphas are all within experimental error of zero, so that's a reasonable assumption.)

Conclusion: If you can hold enough VBR (MSCI small value) to get your desired small-cap and value tilt, that is the lowest-cost way to get the tilt. If you can't (limited tax-deferred room, want a larger tilt than VBR itself gives), then use RZV, which requires the smallest holding to get the desired tilt.

Do you have three-factor data for the Russell 2000 Value? This appears to have a better value weighting than the S&P 600 Value, and Vanguard offers it at the same cost. (Long-term regressions of the Russell 2000 should have negative alpha, and Vanguard itself proved that by outperforming the index by 1% per year when Small-Cap Index tracked it; I believe the index is now less vulnerable to front-running.)
I largely agree, but would argue that even better than looking at cost is looking at tracking error, as long as you have a few years' sample period, since the tracking error will include the effect of the cost. Also, based on that excellent suggestion to look at momentum, I honestly wouldn't hold RZV in any circumstance; that negative momentum would eat up much of the potential benefit of the small and value weights, reducing your margin of safety considerably. Personally if that was my only choice I would just stick with TSM. It's true that we can't be sure that the alphas in the 3F regressions were statistically insignificant, but the 4F regression suggested by Ketawa shows that most of that negative 3F alpha for RZV is due to negative momentum, which we can estimate fairly precisely.

Here are the 4F regressions for the funds themselves, including IWN (iShares Russel 2000 value):

Code: Select all


                                Mkt-Rf    SmB     HmL     Mom     Alpha
    VBR 2004/03+                 0.96    0.64    0.39    -0.07    -0.40%   R^2 = 0.974
       Std. Error                0.03    0.05    0.04     0.02     1.21%

    IJS 2002/01+                 0.92    0.85    0.40    -0.00    -1.85%   R^2 = 0.963
       Std. Error                0.03    0.05    0.04     0.02     1.28%

    IWN 2000/08+                 0.89    0.75    0.61     0.02    -1.10%   R^2 = 0.961
       Std. Error                0.02    0.04    0.03     0.02     1.20%

    RZV 2006/05+                 0.92    1.39    0.96    -0.54     0.33%   R^2 = 0.920
       Std. Error                0.08    0.18    0.15     0.07     4.34%

    JKL 2004/08+                 0.91    0.69    0.49    -0.15    -0.02%   R^2 = 0.959
       Std. Error                0.03    0.07    0.06     0.03     1.64%

And all since inception of RZV, the newest of the bunch:

                                Mkt-Rf    SmB     HmL     Mom     Alpha
    VBR 2006/05+                 0.95    0.61    0.41    -0.06    -0.42%   R^2 = 0.977
       Std. Error                0.03    0.06    0.05     0.02     1.45%

    IJS 2006/05+                 0.90    0.90    0.34    -0.03    -1.17%   R^2 = 0.973
       Std. Error                0.03    0.06    0.06     0.03     1.58%

    IWN 2006/05+                 0.91    0.78    0.49     0.01    -1.57%   R^2 = 0.975
       Std. Error                0.03    0.06    0.05     0.02     1.50%

    RZV 2006/05+                 0.92    1.39    0.96    -0.54     0.33%   R^2 = 0.920
       Std. Error                0.08    0.18    0.15     0.07     4.34%

    JKL 2006/05+                 0.90    0.65    0.55    -0.14     1.21%   R^2 = 0.962
       Std. Error                0.04    0.08    0.07     0.03     1.96%
As mentioned above, I've decided to start using Standard Error instead of t-values. Both give the same information, but for me at least, standard error makes it easier to visualize the precision of the estimates. Historically, the momentum premium has been almost 10% per year, so that -0.54 of RZV would work out to about a 5% annual drag. Even if you only need half as much of it in your portfolio, that's far too much to deal with for any degree of tilt IMO. JKL is the next-largest value tilt, and unsurprisingly the next-largest negative momentum, although neither are nearly so extreme. However, we only get 0.14 more value exposure than VBR, in exchange for 0.08 more negative momentum and greater uncertainty. Historically the value premium has been about 4% and the momentum premium almost 10, so JKL is unlikely to come out ahead, and is going to have more risk as well, both due to the greater tilt and due to the lesser diversification, which we see in its lower R^2 and more variable alpha.

IWN does have a larger value weighting than IJS or VBR, and zero momentum, compared to small negative for VBR. That said, it does show some negative alpha - not technically statistically significant, but close, and with at least one good explanation behind it - frontrunning the index. Honestly just looking at these numbers, I would say the choice between IJS and IWN (or Vanguard equivalents) depends mostly on whether one is looking for more small or value to complement the remainder of their portfolio. VBR falls somewhere between the two. It has very slightly more negative momentum, enough to account for perhaps -0.7% per year in return, but its 4F alpha has been superior by at least that amount and relatively consistent, and is likely to be higher by at least the difference in their tracking errors, about 0.2%. (My guess is that the S&P and Russel indexes do more than the MSCI to reduce negative momentum, the cost of which is a drag on alpha, but I haven't researched that.) IWN and VBR are more diversified than IJS, but also have higher yield, which would make a slight difference in taxable accounts.

Honestly I expect there's very little separating these three funds and any is as likely as the others to be a good choice. I would steer clear of RZV due to large negative momentum and JKL due to smaller negative momentum, but insufficiently higher factor weightings to justify it.

Thanks again to Ketawa for the suggestion to look at it from a 4F perspective.

Edit: Do you have a reference for the Russel index being less vulnerable to front-running? I did a bit of looking around for that, but couldn't find anything except them adding bands to reduce turnover. That wouldn't only make frontrunning very marginally more challenging though. Given that their methodology is public and the rebalancing date is known, I don't see how frontrunning could be avoided, but that's certainly not to say there isn't a way.

Edit 2: Tried adding the Russel index to my realistic-ish portfolios test above, and it came out with similar results to the S&P-based ones. Sharpe ratio of 0.30, so MSCI was still the winner over that time period. Also tried a couple shorter periods with similar results.

Edit: 1 million: Those Sharpe ratios used the geometric annualized returns and the approximate annual standard deviations (by dividing by sqrt(12)). Better would have been to use arithmetic returns, and just calculate the monthly Sharpe ratio. Anyway, the conclusions remain the same.
Last edited by ClosetIndexer on Fri Oct 05, 2012 7:58 pm, edited 2 times in total.

User avatar
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by ClosetIndexer » Thu Sep 27, 2012 12:01 am

Another thing to keep in mind about those sample portfolios is that while 0 0.4,0.4 SMB,HML tilted portfolio has historically outperformed TSM by about 2.5% annually, that doesn't account for the increase in risk. If we reduce the equities in the portfolio to hold risk (SD) constant, the out-performance is more like 1.6%, before costs. Using the VBR-based portfolio as an example, not even taking into account the costs of the other two funds, just VBR's momentum and alpha, we lose another ~ 37% * 1.1% = 0.4%. Add in the rest of the portfolio and we could estimate around 0.7%.

To hold risk constant we have admittedly reduced our overall equity exposure by about 15%, so that 0.7% is more like 0.6%, but still, it brings the theoretical out-performance of our portfolio to something more like 1.6% - 0.6% = 1%, assuming factor premiums remain the same as they have been historically. Now, that's still worth doing in my opinion, but it's not exactly a massive margin for error. If you then go reaching for additional tilt using a less efficient product like RZV, you can see that remaining expected out-performance disappear very quickly. At least, that's how I look at it.

I believe in multi-factor investing, but I want to give myself the best possible chance of achieving at least market returns over the long term. So if I don't see a strong enough chance that a portfolio will out-perform, long term, on a risk-adjusted basis, I would rather stick with TSM than roll the dice. (Fortunately, we don't have to make that choice! (Except in Canada... :( ))

User avatar
grap0013
Posts: 1884
Joined: Thu Mar 18, 2010 1:24 pm

Re: Comparison of small-cap value ETFs

Post by grap0013 » Thu Sep 27, 2012 10:09 am

ClosetIndexer,

I don't know if you are still taking some requests, but I would love to see the factor analysis on the fundamental pure small value data with loadings, alpha, and momentum included. I would really appreciate it and I think the forum would benefit from it. It goes back to 1979. Here's the monthly returns of the index. http://www.2shared.com/file/DmQvUJC7/RAFI_Data.html

PXSV has done an excellent job of following it's index's returns so I think it's reasonable to see that going forward. http://www.invescopowershares.com/produ ... icker=PXSV
There are no guarantees, only probabilities.

caklim00
Posts: 1943
Joined: Mon May 26, 2008 10:09 am

Re: Comparison of small-cap value ETFs

Post by caklim00 » Thu Sep 27, 2012 11:03 am

Oh where Oh Where has Robert T gone? This kind of discussion is perfect for him.

FWIW, I've use or have used the following US SCV funds:
1) DFFVX - DFA US Targeted Value (old 401k - its all I hold in that one)
2) VBR - MSCI 1750V - (IRA and taxable)
3) VIOV/IJS - S&P 600V (taxable)
4) RZV - S&P 600PV (taxable)

The choices for me are basically VBR, VIOV/IJS, RZV for an ongoing basis.*

Tax deferred
I like VBR in tax deferred since it has lowest cost + high value load, but I'm finding that I need tax deffered now more for international scv options since there is no go tax efficient fund in that space yet.

Taxable
1) I first liked RZV best. It has a super low dividend yeild. But, sometimes bid/ask can cost more than .1% which is something I try to avoid and there is no good TLH partner. I have some gains on what I currently own, so plan on just holding onto my position.
2) I like VIOV/IJS slightly over VBR in taxable. Its more tax efficient, but costs a bit more (ER and/or slighly more on bid/ask spread but still under my .1% threshold). But, if you buy one you probably end up owning both if you tax loss harvest, so right now I think I own about equal amounts of VIOV and VBR in my taxable account.

*PXSV I've considered, but I don't like adding funds with bid/ask typically above .1%

Also, Russell now has introduced bands which help with turnover which prevents some of the frontrunning that happened in the past. But, there still will be frontrunning since its the biggest index where insituational money is tied to.

User avatar
Ketawa
Posts: 2192
Joined: Mon Aug 22, 2011 1:11 am
Location: DC

Re: Comparison of small-cap value ETFs

Post by Ketawa » Thu Sep 27, 2012 12:45 pm

Is there any easy way to get MSCI index data? I haven't been able to track it down and would like to be able to run regressions on the index itself in addition to live funds. I can only go back to March 2004 with live fund data from VBR.

RAFI Fundamental Small Value Index (since inception)

Code: Select all

RAFI Small Value     1976/01 - 2011/09
                Mkt-RF     SmB       HmL      Mom       alpha     yearly alpha
loadings        0.993     0.864     0.667    -0.177     0.245     2.982%
standard error  0.019     0.027     0.028     0.018     0.082
r2              0.921
t-stat          52.835    31.922    23.447   -9.893     2.974
Some more data for comparison to index funds all over the same time period. If you take out fund expenses for VBR it would reduce its negative yearly alpha. The data for S&P 600 Value and Pure Value is of the actual indexes.

Code: Select all

VBR                  2004/03 - 2011/09
                   Mkt-RF     SmB      HmL       Mom       alpha      yearly alpha
loadings           0.947     0.633     0.404    -0.067    -0.074     -0.886%
standard error     0.028     0.052     0.047     0.022     0.108
r2                 0.972
t-stat             33.732    12.288    8.561    -2.976    -0.686


S&P 600 Value        2004/03 - 2011/09
                   Mkt-RF     SmB      HmL       Mom       alpha      yearly alpha
loadings           0.877     0.898     0.375    -0.024    -0.079     -0.945%
standard error     0.027     0.049     0.045     0.021     0.103
r2                 0.975
t-stat             32.723    18.264    8.319    -1.108    -0.767


S&P 600 Pure Value   2004/03 - 2011/09
                   Mkt-RF     SmB      HmL       Mom       alpha      yearly alpha
loadings           0.880     1.391     0.952    -0.567    -0.021     -0.248%
standard error     0.083     0.152     0.139     0.066     0.318
r2                 0.905
t-stat             10.650    9.176     6.850    -8.584    -0.065


RAFI Small Value     2004/03 - 2011/09
                   Mkt-RF    SmB       HmL       Mom       alpha      yearly alpha
loadings           0.946     1.064     0.517    -0.314     0.181      2.196%
standard error     0.035     0.065     0.059     0.028     0.136
r2                 0.972
t-stat             26.814    16.437    8.717    -11.142    1.335
RAFI SV has had higher loadings than the MSCI SV and S&P 600 Value indexes, but at the cost of greater negative momentum, and higher expense ratio. I think it would be cheaper to get a target load using VBR or IJS/VIOV, and you wouldn't have to worry about negative momentum as much.

pinebarrens1
Posts: 127
Joined: Sat Mar 05, 2011 8:08 pm

Re: Comparison of small-cap value ETFs

Post by pinebarrens1 » Thu Sep 27, 2012 12:58 pm

aren't those who use VBR concerned that it holds such a high percentage in Financials/ Reits (38.5%)? 5 of the top 10 holdings are Reits and 1 BDC I think. Also it appears to be very light in Tech and Healthcare compared to other indexes. So which ever way the financial sector goes....has an outsized impact on VBR. In contrast the S&P 600 Small Cap Value holds much less and plain Vanguard Small Cap (VB) holds about 70% less financial/ reit as VBR and much more balanced by sectors.

User avatar
Ketawa
Posts: 2192
Joined: Mon Aug 22, 2011 1:11 am
Location: DC

Re: Comparison of small-cap value ETFs

Post by Ketawa » Thu Sep 27, 2012 1:04 pm

There's many ways to calculate expected returns so I'm not sure how useful this is. I just looked at the factor loadings for that same time period, March 2004 through September 2011, and calculated expected returns based solely on those and the historical equity/size/value/momentum premiums. Very rough calculation. Not even sure if I did it correctly; I took the French-Fama data and tried to figure out the geometric mean of the factors, then extrapolated that out to a yearly expected return for each factor.

Code: Select all

           Equity    Size      Value     Momentum
Premiums   5.8%      2.3%      3.9%      7.0%
Expected Returns
MSCI Small Value 8.1%
S&P 600 Value 8.5%
S&P 600 Pure Value 8.0%
RAFI Small Value 7.8%

Again, this just takes the loadings for the March 2004 to September 2011 and uses them for expected returns in the future. Not entirely sure what to make of all these numbers. Unless you think PXSV will continue with positive alpha, it does not seem worthwhile, nor does RZV. The negative exposure to momentum has a huge effect.

One thing I'm not sure what to make of is that over some periods S&P 600 SV has had a low market loading, over the above period as low as 0.88. MSCI SV (VBR) seems to have had a more stable market loading.

I'm halfway thinking about switching to what caklim00 likes: VBR in tax-deferred, and VIOV in taxable.

User avatar
grap0013
Posts: 1884
Joined: Thu Mar 18, 2010 1:24 pm

Re: Comparison of small-cap value ETFs

Post by grap0013 » Thu Sep 27, 2012 1:34 pm

That's some really interesting data Ketawa. Those are expected real returns correct?

So it looks like the RAFI small value generated statistically signicificant alpha in the 1976-2011 data set and still generated positive alpha in 2004-2011, but it was not statistically significant. If you read "The Fundamental Index" by Arnott, he claims to remove negative alpha with their methodology and their live funds have done just that. I think it has to do with how they define value. P/E and P/CF sorts have historically produced positive alpha while P/B and P/D have produced negative alpha. Could it really be that simple?

So it looks like if you use PXSV with it's higher ER you might underperform VBR by up to 1% if alphas get closer to 0. However, if the past is any indicator it could outperform VBR by a couple of percentage points on an annualized basis.
There are no guarantees, only probabilities.

User avatar
Ketawa
Posts: 2192
Joined: Mon Aug 22, 2011 1:11 am
Location: DC

Re: Comparison of small-cap value ETFs

Post by Ketawa » Thu Sep 27, 2012 2:34 pm

Technically I think those are expected returns over the F-F risk free rate, which is currently 0, so they would be nominal. This leads me to a couple of questions if anyone has any insight.

1. Bill Bernstein's tutorial takes the fund return, then subtracts RF. Would it make more sense to take the Mkt - RF, add the RF rate back to that, then regress the fund return against Mkt? Bernstein's tutorial seems to be approaching the question from the persective of evaluating an active fund manager. Actually, I'm not sure it matters either way, since the RF rate is a constant being added/subtracted to both series the exact same amount, so this would have no effect on determining loadings.

2. Which is more important for determining expected returns, the historical Mkt, or Mkt - RF? The Mkt - RF rate is 5.8%, while Mkt alone is 9.6%. With the RF rate currently zero, does this mean expected returns for something like Vanguard Total Stock Market would be 5.8% nominal/3.3% real, or 9.6% nominal/7.1% real? I'm pretty sure it would be the first, as I don't see many people calling for 7% real returns in the stock market, I would just like a good explanation for it. It seems silly that the real return for stocks would depend so highly on whatever the RF rate happens to be.

EDIT: I think I answered my own question. Future expected returns do depend on the risk-free rate, and it's obvious why. The highest RF rate in the data series is 1.35, or 17% annualized. Nominal expected returns had to be higher than that in a rational market, or nobody would hold equities. The same thing must apply at the other extreme in a zero interest rate environment. The real question is how expected returns interact with real rates, not nominal rates, and this is probably more difficult to answer, depends on valuations, etc.

User avatar
ClosetIndexer
Posts: 288
Joined: Mon Mar 19, 2012 11:00 pm
Contact:

Re: Comparison of small-cap value ETFs

Post by ClosetIndexer » Thu Sep 27, 2012 4:38 pm

Ketawa wrote:Is there any easy way to get MSCI index data? I haven't been able to track it down and would like to be able to run regressions on the index itself in addition to live funds. I can only go back to March 2004 with live fund data from VBR.
You can get MSCI data for the US here. Note that for regressions you want the gross data (including dividends), not the price. Also, it's daily, so to do monthly regressions you'll need to extract the month-end values. If you want to fire me a PM with your email I'll send you the spreadsheet I use to do that.
Ketawa wrote:

Code: Select all

RAFI Small Value     1976/01 - 2011/09
                Mkt-RF     SmB       HmL      Mom       alpha     yearly alpha
loadings        0.993     0.864     0.667    -0.177     0.245     2.982%
standard error  0.019     0.027     0.028     0.018     0.082
r2              0.921
t-stat          52.835    31.922    23.447   -9.893     2.974
Double-checked and confirmed. Agree that it's surprising to see that large positive alpha with relatively small error. However, the negative momentum is indeed large, especially over more recent periods, as you pointed out. Interestingly, PRF (the broad RAFI fund) shows similar results:

Code: Select all

                                Mkt-Rf    SmB     HmL     Mom     Alpha (annual)
    PRF (RAFI)                   0.97   -0.07    0.31    -0.16     0.57%   R^2 = 0.978
       Std. Error                0.02    0.05    0.05     0.02     1.28%
Although less positive alpha, and for a mid-large fund that momentum is even more significant. (Negative momentum is generally worse with small funds.) Presumably if there were a reason to expect persistent positive alphas from the RAFI technique, we would expect to see them here as well, no? (grap, can you share the historical data for that index as well?)
Ketawa wrote:There's many ways to calculate expected returns so I'm not sure how useful this is. I just looked at the factor loadings for that same time period, March 2004 through September 2011, and calculated expected returns based solely on those and the historical equity/size/value/momentum premiums. Very rough calculation. Not even sure if I did it correctly; I took the French-Fama data and tried to figure out the geometric mean of the factors, then extrapolated that out to a yearly expected return for each factor.

Code: Select all

           Equity    Size      Value     Momentum
Premiums   5.8%      2.3%      3.9%      7.0%
Agree with this approach for estimating factor premiums, particularly size and value. There are many approaches to estimate the equity premium (ie: Gordon's equation, for one.) Equity, size, and value appear to line up with the commonly accepted historical numbers. I remember reading that the momentum factor is more like 9% historically, but I double-checked your result and got 7% as well, and I can't find that source. Perhaps the 9 takes into a account global numbers and/or only looks at a more recent period. (For example, just looking since 1960, I get 8%.)

That said, I would hesitate to just extrapolate out the factors to fund returns. For one thing, the various factors are not perfectly correlated, so provide some diversification benefit. More generally, it's important to look at the fund's effect on the overall portfolio, both in terms of return and standard deviation. My very first post here actually contained some graphs to visualize the returns and SDs of various levels of historical tilt. Didn't include momentum though, so you'll have to tweak the estimates for that. Also note that they're theoretical portfolios based on the research returns, so doesn't include fund costs or alphas either.

(Although I agree that just adding up the expected factor premiums times loadings is a reasonable way to get a quick estimate of expected return only.)
grap0013 wrote:So it looks like if you use PXSV with it's higher ER you might underperform VBR by up to 1% if alphas get closer to 0. However, if the past is any indicator it could outperform VBR by a couple of percentage points on an annualized basis.
Remember to take the momentum into account. Depending on what we estimate for the momentum premium, that wipes out most of the difference in alpha. The higher factor loadings do still result in higher expectation, but also higher standard deviation / risk. As part of an entire portfolio with consistent factor loadings, the difference would be much smaller, even if those positive alphas persist. In a minute here I'll add that index to my 0.4/0.4 portfolios example to take a look at what happened over that particular time period. Finally, I think it's good to keep in mind that most of that index data is from before the funds existed. If it is indeed taking advantage of some market inefficiency (creating positive alpha), it's conceivable that it will disappear as the mechanism behind it is better understood.
Ketawa wrote:EDIT: I think I answered my own question. Future expected returns do depend on the risk-free rate, and it's obvious why. The highest RF rate in the data series is 1.35, or 17% annualized. Nominal expected returns had to be higher than that in a rational market, or nobody would hold equities. The same thing must apply at the other extreme in a zero interest rate environment. The real question is how expected returns interact with real rates, not nominal rates, and this is probably more difficult to answer, depends on valuations, etc.
Correct; the regression gives you the exposure to the Mkt-Rf, HmL, SmB, and Mom factors. So the expected future return would be the sum of the exposures to each factor times its expected premium, plus the risk-free rate. (The idea is to remove the effect of RFR when doing the regression, since it simply deals with equity factors. It doesn't have any information to predict interest rate changes. So you definitely don't want to do the regression on Mkt instead of Mkt-Rf.) And yes, all rates are nominal. To get the expected real rate, just take the expected nominal rate and subtract expected inflation. :) Again, the FF model doesn't say anything about what that might be, so you have to estimate by other means.

User avatar
grap0013
Posts: 1884
Joined: Thu Mar 18, 2010 1:24 pm

Re: Comparison of small-cap value ETFs

Post by grap0013 » Thu Sep 27, 2012 5:49 pm

[quote="ClosetIndexer"]
Although less positive alpha, and for a mid-large fund that momentum is even more significant. (Negative momentum is generally worse with small funds.) Presumably if there were a reason to expect persistent positive alphas from the RAFI technique, we would expect to see them here as well, no? (grap, can you share the historical data for that index as well?)
[quote="ClosetIndexer"]

I actually don't have the monthly returns for the other rafi indexes. Rick Ferri got them for the small cap value index. Rick would you mind helping out? It would be great to get him in on this discussion because he is very knowledgeable about using different value strategies.

I've basically just compared a lot of rafi index returns on an annualized basis to other value indexes and they always seem to outperform by a couple of percent. Usually with a little bit higher standard deviation, but I don't think that's the whole story. Live fund returns both domestic and abroad have demonstrated positive alpha as well.

Finally keep in mind since you are running the value factor loads on FF data, anything with low P/B is going to look more valuey than something using other value strategies. Therefore, I think FF techniques always consistently underestimate the valueyness of rafi indexes and funds.
There are no guarantees, only probabilities.

Post Reply