Question for RobertT

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Topic Author
saurabh
Posts: 319
Joined: Thu Aug 20, 2015 11:56 am

Question for RobertT

Post by saurabh »

Robert T,

I have been avid reader of your posts describing your globally diversified value and small tilted equity portfolio, because they bridge the gap between theory and practice in a very accessible way. Many years back you had set up a Morningstar sample portfolio to track a close approximation of your equity portfolio which was trying to capture 40% of the size premium and 20% of the size premium.

Given all the changes occuring in the indexing and ETF world not the least of which is the switch to CRSP indices by Vanguard, several international value and small cap ETFs introduced by iShares and others, what would you preferred funds/ETFs be today?

Also what is your sense of the small cap premium today, do you think it is still intact? Would a portfolio that only tilted to value using large/mid-cap stocks show more tracking error relative to broad equity benchmarks compared to using small stocks?

Thanks!
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

.
Saurabh,

One the benefits of having factor load targets is they provide a great framework to assess implications of changes in indexes and ETFs i.e. would the new offerings achieve my factor load targets at lower cost? In the absence of a clear framework for assessing new ETFs it is easy to get lost in the blizzard of new offerings, and in my view leads to more instability in the investment process with a higher likelihood of frequent changes (driven more by marketing voices and public opinion).
saurabh wrote:Given all the changes occuring in the indexing and ETF world not the least of which is the switch to CRSP indices by Vanguard, several international value and small cap ETFs introduced by iShares and others, what would you preferred funds/ETFs be today?
The Morningstar sample portfolio is still one of the lowest cost ways to achieve my factor load targets.
  • 23% Vanguard MidCap Value
    ..9% iShares S&P600 Value
    ..5% Bridgeway Small Co. Mkt
    23% iShares EAFE Value
    ..5% iShares EAFE Small
    10% Vanguard EM
    25% iShares 3-7yr Treasury
New offerings can potentially provide 3 additional things:

1. Less negative fund alpha (lower cost though less erosion of returns through negative alpha): In this respect I like the Fundamental Index ETFs as they can potentially provide less negative (closer to zero) alpha than traditional value funds (if past performance/simulations are any guide). While many don’t like fundamental indexing some do. Here’s and extract from Asness et al. “Fact, fiction, and value investing.” “Fundamental indexing does several things we like in a value strategy: it uses multiple measures, it uses up-to-date prices when rebalancing, and it does some implicit timing based on the size of valuation differences across stocks that, despite adding only modest benefit historically, we find intuitively appealing. It is a very clear, simple, even cleaver way to explain and implement value investing.” The historical simulated returns from 1950s and actual returns of RAFI linked ETFs over the last 10 years have shown non-negative alphas in the FF 3 factor analysis.

2. Less portfolio level negative momentum loads (which also shows up as negative alpha in FF 3 factor analyses). A value tilt often brings with it exposure to negative momentum, so it’s difficult to match a 0.4 value load and 0.0 momentum load target with only value funds, as the momentum load will likely be negative. Recently launched momentum ETFs potentially provide a way to reduce a portfolio momentum load while still achieving a 0.4 value load (i.e. to get closer to a 0.0 portfolio momentum load). In this respect, I like the iShares USA Momentum (MTUM) ETF. In addition, reducing negative momentum will reduce tracking error of a value tilted portfolio with the market in periods such as 1985-1999 when large growth outperformed. One caveat, is that as momentum strategies have higher turnover they have less capacity than value strategies.

3. More even distribution of value and size tilts across domestic and international markets. The ETF portfolio list above derived most of its size exposure in US markets, and also has a larger value tilt in US than non-US developed. New offerings allow for a more even exposure.

Here is a portfolio that is expected to have less negative (closer to zero alpha, including less negative momentum load) than the portfolio listed above, and has a more even tilt to small and value across US and non-US markets. Both target a 0.2 and 0.4 size and value load respectively.

Allocation (rounded up)
  • ..8% iShares US Momentum
    11% Vanguard MidCap Value
    19% Schwab Fundamental US Small Company
    17% Powershares RAFI Developed Market Non-US
    11% Powershares RAFI Developed Market Non-US small cap
    10% Powershares RAFI Emerging Markets
    25% iShares 3-7 yr Treasury

    [For US = 20:30:50 Momentum:MCV:SV, for non-US Developed = 60:40 LV:SV]
The portfolio above has a 0.05% higher expense ratio than the earlier portfolio listed, but this is offset by lower expected negative alpha. Using the (cheaper) Schwab Fundamental International ETFs would closely match to expense ratio of the first listed portfolio. I don’t have a strong opinion on differences between the FTSE RAFI international series (tracked by Powershares), and the Russell Fundamental International series (tracked by Schwab). The latter includes buybacks in its stock screens, but buybacks are not as prominent in intl as in US markets. In addition it includes lower leverage, so quality loads would likely be somewhat higher in the Russell Series. Either way - the Powershares and Schwab Fundamental ETFs make great tax loss harvesting partners.

The above two portfolios have similar long-term expected returns as they have similar size and value loads. Arguably the second provides ‘purer factor exposure’ given the less negative expected alphas, which may lead to marginally higher returns (or rather less erosion of returns from negative alpha).
Also what is your sense of the small cap premium today, do you think it is still intact?

I think over the long-term there will be a small-cap premium, and I have a long-term tilt to smaller caps (portfolio size load of 0.2). Not planning to change this anytime soon. There will inevitably be long periods of under performance (re the earlier Fama-French paper https://us.dimensional.com/pdf/Volatili ... emiums.pdf ). FWIW I have more confidence in the value premium than the size premium (re: having a value load target double my size load target).
Would a portfolio that only tilted to value using large/mid-cap stocks show more tracking error relative to broad equity benchmarks compared to using small stocks?
It will have lower tracking error, but also lower long-term expected return, given the omission on small caps.

Robert
.
User avatar
in_reality
Posts: 4529
Joined: Fri Jul 12, 2013 6:13 am

Re: Question for RobertT

Post by in_reality »

Robert T wrote:.
Allocation (rounded up)
  • ..8% iShares US Momentum
    11% Vanguard MidCap Value
    19% Schwab Fundamental US Small Company
    17% Powershares RAFI Developed Market Non-US
    11% Powershares RAFI Developed Market Non-US small cap
    10% Powershares RAFI Emerging Markets
    25% iShares 3-7 yr Treasury

    [For US = 20:30:50 Momentum:MCV:SV, for non-US Developed = 60:40 LV:SV]
The portfolio above has a 0.05% higher expense ratio than the earlier portfolio listed, but this is offset by lower expected negative alpha. Using the (cheaper) Schwab Fundamental International ETFs would closely match to expense ratio of the first listed portfolio. I don’t have a strong opinion on differences between the FTSE RAFI international series (tracked by Powershares), and the Russell Fundamental International series (tracked by Schwab). The latter includes buybacks in its stock screens, but buybacks are not as prominent in intl as in US markets. In addition it includes lower leverage, so quality loads would likely be somewhat higher in the Russell Series. Either way - the Powershares and Schwab Fundamental ETFs make great tax loss harvesting partners.
You are targeting 0.2 size, 0.4 value and 0.0 momentum correct?

Please help me understand what I am seeing in portfoliovisualizer's factor analysis tool.

Allocating the 25% of bonds proportionally to stocks to get a percent I can use in portfoliovisualizer brings it to:

11.00% iShares MSCI USA Momentum Factor ETF (MTUM)
15.00% Vanguard Mid-Cap Value ETF (VOE)
25.00% Schwab Fundamental U.S. Small Company Index ETF (FNDA)
22.00% Schwab Fundamental International Large Company Index ETF (FNDF)
14.00% Schwab Fundamental International Small Cap Company Index ETF (FNDC)
13.00% Schwab Fundamental Emerging Markets Large Company Index ETF (FNDE)
------------

Size Exposure (Bsmb) 0.22
Value Exposure (Bhml) 0.11
Momentum Exposure (Bmom) 0.06
Alpha -0.07%
Annual Alpha -0.85%
r2 97.2%

https://www.portfoliovisualizer.com/fac ... x_cdt=true

Are these results due to:
1) use of AQR factor data (it seems the best fit)
2) the Russell series having a shorter history + the value loading varying in RAFI indexes
3) using the "global" selection [is it better to look at global-ex us, and us in two different portfolios?]
4) something else

Knowing what you know about the factors on the portfolio you posted, do you think the setting I have on portfolio visualizer are returning meaningful data?

These settings on my actual portfolio (which for domestic at least are blended with market cap weight) produce:

Size Exposure (Bsmb) 0.11
Value Exposure (Bhml) 0.11
Momentum Exposure (Bmom) 0.08
Alpha -0.18%
Annual Alpha -2.22%
r2 99%
hariom
Posts: 7
Joined: Thu May 17, 2012 1:37 pm

Re: Question for RobertT

Post by hariom »

Robert T:

I am assuming your portfolio is best suited for tax advantaged accounts. Do you have any suggestions/changes for similar factor tilted but tax-efficient portfolio in a taxable account.

Thanks
pauliec84
Posts: 414
Joined: Wed Feb 09, 2011 11:29 am

Re: Question for RobertT

Post by pauliec84 »

You are targeting 0.2 size, 0.4 value and 0.0 momentum correct?

Please help me understand what I am seeing in portfoliovisualizer's factor analysis tool.

Allocating the 25% of bonds proportionally to stocks to get a percent I can use in portfoliovisualizer brings it to:

11.00% iShares MSCI USA Momentum Factor ETF (MTUM)
15.00% Vanguard Mid-Cap Value ETF (VOE)
25.00% Schwab Fundamental U.S. Small Company Index ETF (FNDA)
22.00% Schwab Fundamental International Large Company Index ETF (FNDF)
14.00% Schwab Fundamental International Small Cap Company Index ETF (FNDC)
13.00% Schwab Fundamental Emerging Markets Large Company Index ETF (FNDE)
------------

Size Exposure (Bsmb) 0.22
Value Exposure (Bhml) 0.11
Momentum Exposure (Bmom) 0.06
Alpha -0.07%
Annual Alpha -0.85%
r2 97.2%

https://www.portfoliovisualizer.com/fac ... x_cdt=true

Are these results due to:
1) use of AQR factor data (it seems the best fit)
2) the Russell series having a shorter history + the value loading varying in RAFI indexes
3) using the "global" selection [is it better to look at global-ex us, and us in two different portfolios?]
4) something else
The issue is two fold.

1) Portfolio Visualizer is estimating factor data based on the live return of the ETFs. Several of the ETFs you are putting in have a short return history and thus not very reliable estimate of future factor exposure. I have little doubt that Robert T's estimates are based upon the longer event window made possible by using the indexes which underlie the ETFs.

2) Using the global selection does not work well. General practice is to use US selection for US funds, and then developed ex-us selection for non-US funds. The caveat is that none of the data covers the emerging market data set. My opinion on best practice in estimating emerging market factor loads is to extrapolate factor loads from developed and US market of funds that use similar index construction methodologies. So for example for Fundamental Emerging Markets Large Company Index, I would look to the Fundamental US Large and the Fundamental Developed Large.

Hope this helps.
pauliec84
Posts: 414
Joined: Wed Feb 09, 2011 11:29 am

Re: Question for RobertT

Post by pauliec84 »

3. More even distribution of value and size tilts across domestic and international markets. The ETF portfolio list above derived most of its size exposure in US markets, and also has a larger value tilt in US than non-US developed. New offerings allow for a more even exposure.
Robert T, I agree with you that it is helpful to have an even factor exposure across markets. With that in mind, I am curious what is your logic to load more heavily on the small tilt domestically compared to internationally? This JFQA article (http://sandylai-research.com/uploads/2/ ... 08jfqa.pdf) makes argument to, if anything, concentrate small cap exposure more internationally to maximize diversification benefits.

Furthermore, within the ETF portfolio you laid out it would be VERY easy to modify to achieve this aim while keep quite similar factor exposures (see below). Furthermore, there are reasonable options for getting a more balanced emerging market exposure.

For example:
8% iShares US Momentum
11% Vanguard MidCap Value (Change to 21%)
19% Schwab Fundamental US Small Company (Change to 9%)
17% Powershares RAFI Developed Market Non-US (Change to 7%)
11% Powershares RAFI Developed Market Non-US small cap (Change to 21%)
10% Powershares RAFI Emerging Markets
25% iShares 3-7 yr Treasury
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

.
in_reality,

IMO factor load data over 2 years is too short to draw any meaningful conclusions about long-term characteristics. My view is its best to estimate factor loads over time spans covering different market cycles - as a reflection of long-term characteristics (factor exposure). This means going back to the underlying index data (something you can't do in portfolio visualizer). The assumption is that the the long term characteristics on the underlying index series will be reflective of future characteristics (obviously no guarantees on this). And IMO the longer the series the higher the likelihood of this assumption holding.

The series I used when I did the estimates in 2014 were the following:

For Vanguard Mid Cap Value: CRSP Midcap Value Index back to its start in 2001, then Russell Midcap Value back its start in 01/1986 - the two are not so different, and it gives a longer series for midcap value.
For MSCI USA Momentum: MSCI USA Momentum index back to 01/1986
For Schwab Fundamental Small Cap: Russell Fundamental Small Cap Index back to its start in 8/1996. Ideally its good to use the same time period (over the same market cycles for all the series), so came up a bit short on the Russell Fundamental Index.

US allocation:
  • The resulting factor loads from the FF3 factor analysis had a non-negative alpha with the following mkt,size, and value loads.

    Mkt = 1.01
    Size = 0.26
    Value = 0.46
Developed-ex US allocation
  • For Powershares FTSE RAFI Developed-ex US = FTSE RAFI Developed ex-US 1000 Index back to July 1990 (the start of the FF global-ex US factor series)
    For Powershares FTSE RAFI Developed-ex US Small = FTSE RAFI Developed ex-US Mid-small index back to July 1990.

    The resulting loads from the FF3 factor analysis also had a non-negative alpha with the following mkt, size, and value loads.

    Mkt = 1.00
    Size = 0.19
    Value = 0.40
Emerging market allocation
  • For this we are on our own. Ken French's database, nor Asness has EM factor data. Jason Hsu has some on his website, but I was not comfortable with them. I ended up creating my own EM factor series from MSCI data, but this contrained the series back to end 2000. Not ideal.
    For Powershares FTSE RAFI Emerging Markets = FTSE RAFI Emerging Market Index. The resulting loads from the FF3 factor analysis had non-negative alpha with the following mkt, size, and value loads:

    Mkt = 1.00
    Size = 0.02
    Value = 0.19
While the value load may perhaps be an underestimate (looking at the US and Developed-ex US analysis), it is the result I got.

Portolio
  • Combing these into a portfolio 50:37:13 for US:Developed-ex US:EM gives the following factor loads (with non-negative alpha).

    Mkt = 1.01
    Size = 0.20
    Value = 0.40
---------------

On portfolio visualizer, factor results for longer series are more meaningful than shorter series. I prefer to do US and Developed-ex US seperately, rather than combined with the global factors, to better understand the evenness of factor loads across markets. I also prefer Fama-French factors (where they have them) to AQR factors. On EM were are on out own.

Obviously no guarantees. Just the way I did it.

Robert
Last edited by Robert T on Mon Feb 15, 2016 6:41 am, edited 1 time in total.
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

pauliec84 wrote:Robert T, I agree with you that it is helpful to have an even factor exposure across markets. With that in mind, I am curious what is your logic to load more heavily on the small tilt domestically compared to internationally? This JFQA article (http://sandylai-research.com/uploads/2/ ... 08jfqa.pdf) makes argument to, if anything, concentrate small cap exposure more internationally to maximize diversification benefits.
0.26 vs. 019 size load in US vs. Developed-ex US is not too far off - and am comfortable with this - both relatively close to 0.2. In EM, prefer to keep it simple. If EM was a larger allocation of equities I would consider adding small, but have no plans to change EM allocation share.

Robert
.
hariom
Posts: 7
Joined: Thu May 17, 2012 1:37 pm

Re: Question for RobertT

Post by hariom »

Hey Robert T:

Thanks for your detailed responses. Do you have any suggestions that might improve the tax efficiency of your factor tilted 75:25 portfolio.

Pradeep
User avatar
in_reality
Posts: 4529
Joined: Fri Jul 12, 2013 6:13 am

Re: Question for RobertT

Post by in_reality »

hariom wrote:Hey Robert T:

Thanks for your detailed responses. Do you have any suggestions that might improve the tax efficiency of your factor tilted 75:25 portfolio.

Pradeep
Is it really tax-inefficient though? The Schwab Fundamental (i.e. Russell) Series seems appropriate in taxable to me.

The following are generally considered tax efficient:

VTI (US Total Market)
12-Mo. Yield% 2.10
SEC Yield% 2.12

VXUS (Total International)
12-Mo. Yield% 3.00
SEC Yield% ---

and by comparison:

VBR (US Small Value)
12-Mo. Yield% 2.12
SEC Yield% 2.32


iShares US Momentum (MTUM)
12-Mo. Yield% 1.16
SEC Yield% 1.06

Vanguard MidCap Value (VOE)
12-Mo. Yield% 2.2
SEC Yield% 2.26

Schwab Fundamental US Small Company (FNDA)
12-Mo. Yield% 1.43
SEC Yield% 1.46

Powershares RAFI Developed Market Non-US (PRF)
12-Mo. Yield% 3.29
SEC Yield% --

Powershares RAFI Developed Market Non-US small cap (PDN)
12-Mo. Yield% 2.17
SEC Yield% --

Powershares RAFI Emerging Markets (PHX)
12-Mo. Yield% 3.63
SEC Yield% ---

-------------
alternatives
Schwab Fundamental Developed Market Non-US (FNDF)
12-Mo. Yield% 2.22
SEC Yield% 3.06

Schwab Fundamental Developed Market Small Non-US (FNDC)
12-Mo. Yield% 1.4
SEC Yield% 1.75

Schwab Fundamental Emerging Markets (FNDE)
12-Mo. Yield% 2.13
SEC Yield% 3.08
Topic Author
saurabh
Posts: 319
Joined: Thu Aug 20, 2015 11:56 am

Re: Question for RobertT

Post by saurabh »

Robert T,

Thanks for taking the time to respond and all the thoughtful clarifications.
Topic Author
saurabh
Posts: 319
Joined: Thu Aug 20, 2015 11:56 am

Re: Question for RobertT

Post by saurabh »

Robert T

One question I meant to ask aside from the tax efficiency which I see another poster asked was about how you deal with or see the risks of sector concentration. I haven't looked up the sector weights of the small cap funds you mentioned, but I know the Vanguard Small Cap Value fund is pretty concentrated in Financials (31%) and Industrials (20%), that is half of the fund in just two sectors. Has the issue of industry exposure and controlling for it been studied extensively in examining the value premium, particularly in small caps? Also during the 2008-2009 crisis were there certain value indexes that did a better job of reducing exposure to the parts of the financial sector that turned out to be value traps? One would think if that industry effects were not important than a better way to implement a long-only value strategy would be to retain market weights for the sectors, but focus on value within each sector.

In general my one concern with the implementation of value is that just using classic P/B like FF would favor sectors such as financials. I know that actual index providers do not just use P/B and use a variety of value metrics include trailing EPS, forward EPS, sales, cash flows etc. but then it makes it all a bet on their black box, very much in the vein of active management, i.e. the discretion to the index provider makes it very distinct from traditional market-cap based passive indexing approaches.

Based on the US experience over the past 20 years, which I see as a good out-of-sample test for FF value and size premiums, how large were they and how do they compare to the original FF studies? I am just trying to get a handle on how large they might be in a long-only portfolio. I don't think the point is stressed often enough that in a long-only strategy one would expect to capture only half of the premium.
User avatar
in_reality
Posts: 4529
Joined: Fri Jul 12, 2013 6:13 am

Re: Question for RobertT

Post by in_reality »

saurabh wrote:Robert T

I know the Vanguard Small Cap Value fund is pretty concentrated in Financials (31%) and Industrials (20%), that is half of the fund in just two sectors.
Sorry, not Robert T though I've got this on my spreadsheet now...(imported from Morningstar into google sheets)

In general be aware the RAFI funds (PowerShares and Schwab Russell) do not have the same sector allocations as a traditional value fund. https://www.google.co.jp/url?sa=t&rct=j ... MWAzTvSNxQ

VBR: (Small Value)
Financials 19.42%
Industrials 16.35%
Consumer Cyclical 13.06%
Real Estate 10.8%
Technology 10.08%


FNDA: (Schwab Fundamental Small)
Industrials 19.93%
Consumer Cyclical 16.9%
Technology 12.85%
Financials 12.36%
Real Estate 10.16%


VB (Small) (Market Weighting)
Industrials 15.39
Technology 15.12
Consumer Cyclical 14.49
Financials 13.36
Real Estate 12.64


To me it seems the US small fundamental index is changing sector allocations less than a traditional value fund does. Only the 2nd and 3rd top sectors have switched position.

For the traditional small value index, 4th -->1st and 2nd -->5th

As for a value trap, consider this SCHE (Emerging Markets - FTSE market cap weighting) is at 8% Energy. FNDE (Schwab Fundamental Emerging) is at 24.5%.
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

.
On taxable accounts
  • Personally I think it is okay to hold the equity ETFs in taxable accounts, while the bond portion may be better placed in tax-advantaged accounts (although I would note that Treasuries are exempt from state and local taxes). Long-only ETFs, at least those in the two portfolios, have been about as tax efficient as tax-managed open ended mutual funds. I would just note that tax-managed mutual funds may still hold an advantage as they can add dividend screens if taxes rise dramatically. Holding ETFs in a taxable account is a long-term commitment, unless of course there are tax-loss harvesting opportunities.
On international diversification
  • The previously linked paper reminds me of the earlier paper by Rex Sinquefield on ‘Where are the gains from international diversification?” http://www.cfapubs.org/doi/pdf/10.2469/faj.v52.n1.1961 The paper, using 24 years of data, concludes that international small stocks provide greatest diversification benefit to a domestic holding of the S&P500 (table 6). Intuitively, smaller companies are less likely to be as linked into the global economy, and more responsive to local conditions than large companies, so inclusion of small caps would provide this diversification benefit.

    These types of papers typically don’t:

    (i) use smaller cap value tilted US portfolios as the starting point on which to assess international diversification benefits. For example if we start with the Russell Fundamental US Small Cap series, here are the annual correlation coefficients from 1997 to 2014 (18 years):

    0.72 = FTSE RAFI Developed-ex US
    0.75 = FTSE RAFI Developed-ex US small

    Which provided greater diversification benefit over this period?

    (ii) use the global cap weighted equity market portfolio as the starting point to asses diversification benefits of adding small cap and value tilts. For simplicity, lets take a 50:37:13 US:EAFE:EM market portfolio. Here at the annual correlation coefficients from 1997-2014:

    0.74 = Russell Fundamental US Small
    0.84 = FTSE RAFI Developed-ex US Small

    Which provided greater diversification benefit over this period?

    Just a different way of looking at it. Different initial start points may lead to different conclusions. As indicated earlier, the size loads on the US and Developed-ex US portions of the second listed portfolio above are not dramatically different (0.26 vs. 0.19). And on emerging markets, I think large caps continue to provide some diversification benefit (re the earlier paper by Campbell Harvey http://papers.ssrn.com/sol3/papers.cfm? ... id=2344817 ).
On sector concentration
  • I don’t think the portfolios are excessively concentrated in particular sectors. Here’s a comparison of the top 4 sectors in the above two listed portfolios compared to the global equity market (as reflected by Vanguard Total World Stock [VT]) – according to M*:

    Stock sectors (%)

    P1 / P2 / VT

    22.4 / 17.4 / 17.8 Financial service
    13.4 / 15.2 / 12.1 Consumer cyclicals
    11.9/ 13.5 / 11.4 Industrials
    10.5 / 11.0 / 14.0 Technology

    P1 = first portfolio listed earlier
    P2 = second portfolio listed earlier (with the RAFI linked funds)
    VT = Vanguard Total World Stock

    And I am aware of one value strategy that excluded financials altogether and still had about a 0.4 value load (so value exposure has not equated exclusively to exposure to financials http://news.morningstar.com/articlenet/ ... ?id=674561 ).
Robert
.
User avatar
in_reality
Posts: 4529
Joined: Fri Jul 12, 2013 6:13 am

Re: Question for RobertT

Post by in_reality »

Robert T wrote:.
For Vanguard Mid Cap Value: CRSP Midcap Value Index back to its start in 2001, then Russell Midcap Value back its start in 01/1986 - the two are not so different, and it gives a longer series for midcap value.
For MSCI USA Momentum: MSCI USA Momentum index back to 01/1986
For Schwab Fundamental Small Cap: Russell Fundamental Small Cap Index back to its start in 8/1996. Ideally its good to use the same time period (over the same market cycles for all the series), so came up a bit short on the Russell Fundamental Index.
Thanks. That's helpful.

Would you happen to have the loadings by index:

CRSP Midcap Value Index
MSCI USA Momentum index
Schwab Fundamental Small Cap
Schwab Fundamental Large Cap

I hold these in different proportions to your suggested allocations and am wondering what I really have for loadings.

Also, can you teach me at some point how to calculate these? I'm wondering if I can automate the process (i.e. scrape the web for the data, compute it in the cloud, then update a google sheets page with the results). As a very first step, what data do you need and where did you get it say for example for the Schwab Fundamental Small Cap index?

Thanks again.
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

in_reality wrote:Thanks. That's helpful.

Would you happen to have the loadings by index:

CRSP Midcap Value Index
MSCI USA Momentum index
Schwab Fundamental Small Cap
Schwab Fundamental Large Cap
Here are the component factor loads I have for the second listed portfolio, from when I last calculated them. These may change somewhat over time, not an exact science, but a reasonable approximation IMO.

US allocation
  • P1 = Russell MidCap Value/CRSP Midcap Value
    P2 = MSCI USA Momentum
    P3 = Russell Fundamental US Small Cap
    US Portion = 20% MSCI US Momentum: 30% MidCap Value: 50% Russell Fundamental US Small Cap

    P1 / P2 / P3 / US combined
    1.01 / +0.98 / 1.02 / 1.01 = Mkt
    0.10 / -0.15 / 0.51 / 0.26 = Size
    0.56 / -0.10 / 0.63 / 0.46 = Value

    P1 was the only one with negative alpha.
Developed-ex US allocation
  • P3 = FTSE RAFI Developed ex-US
    P4 = FTSE RAFI Developed ex-US Small Cap

    P3 / P4 / Developed ex-US combined
    +1.01 / 0.99 / 1.00 = Mkt
    -0.07 / 0.57 / 0.19 = Size
    +0.38 / 0.42 / 0.40 = Value
Also, can you teach me at some point how to calculate these? I'm wondering if I can automate the process (i.e. scrape the web for the data, compute it in the cloud, then update a google sheets page with the results). As a very first step, what data do you need and where did you get it say for example for the Schwab Fundamental Small Cap index?
I wouldn't suggest going overboard. Here's the methodology, the Russell Fundamental data, and the FF factor data.

The methodology: http://www.efficientfrontier.com/ef/101/roll101.htm
The Russell fundamental data: http://www.ftse.com/products/russell-in ... 1453600430
Fama-French factor data: http://mba.tuck.dartmouth.edu/pages/fac ... brary.html

Robert
.
Topic Author
saurabh
Posts: 319
Joined: Thu Aug 20, 2015 11:56 am

Re: Question for RobertT

Post by saurabh »

Robert T,

You are right, the FTSE RAFI indices seem less skewed relative to broad market in their industry exposure compared to the Vanguard Value funds, which is great. On the other hand the performance of both the developed ex-US and emerging market funds relative to the traditional broad market Vanguard funds has been just terrible on a trailing 5-year basis, especially in emerging markets. While in developed ex-US FTSE RAFI has done better than the DFA value funds (Vanguard Developed Markets 0.3%, DFA International Value -3.4%, FTSE RAFI Developed -2.4%) in emerging markets it has done significantly worse (Vanguard Emerging Markets: -5.9%, DFA Emerging Markets Value -8.7%, FTSE RAFI Emerging Markets -10.3%).

Are there international and emerging markets benchmark FF HML and SMB portfolios so that one can judge if the issue is with the particular black box implementation of value or if the underlying factors have been out of favor?
Topic Author
saurabh
Posts: 319
Joined: Thu Aug 20, 2015 11:56 am

Re: Question for RobertT

Post by saurabh »

One basic question. Why do we see long-only funds with factor loads of > 0.5 on small or value factors? Should a long-only fund be limited to only capturing half of the premium?
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

saurabh wrote:Are there international and emerging markets benchmark FF HML and SMB portfolios so that one can judge if the issue is with the particular black box implementation of value or if the underlying factors have been out of favor?
Value has been out of favor. Over the past 7-years to end-December 2015, FF Global-ex US HML = annualized return of -1.9% per year. For emerging markets, the past 5-yr annualized returns to date: MSCI EM Growth = -4.0%, MSCI EM Value = -7.9%.

While at the overall portfolio level sector allocations are not dramatically different from the world equity market, there are larger differences across markets. For example Powershares FTSE RAFI EM (PXH) has a 20.7% allocation to energy compared to MSCI EM allocation of 6.8% (and DFA EM Value=11.6%). As we know oil prices have declined by 70% since mid-2014 which likely had a larger negative effect on PXH.

If we look at implementation over the last 5-years to end December 2015: Annualized returns PXH = -9.86%, FTSE RAFI EM Index = -9.31%. Difference = -0.55% which is close to the 0.49% expense ratio difference. Similarly, the annualized returns of Powershares FTSE RAFI Developed-ex US = -0.65%, FTSE RAFI Developed-ex US index = -0.47. Difference = -0.18 which is smaller than the 0.45% expense ratio.

Current funds in these strategies is small relative to the overall market. For example PXH AUM = 250m, Vanguard EM = 29.8 billion, so don’t think this has had such a large impact on ‘overcrowding’ EM value.

Obviously no guarantees. Time will tell.

Robert

PS FWIW here are some Price/Forward Earnings estimates from M*

..8.6% = Powershares FTSE RAFI EM
..8.8% = DFA EM Value
11.3% = DFA EM Small
11.4% = Vanguard EM
13.4% = iShare MSCI EM Min Volatility
.
Topic Author
saurabh
Posts: 319
Joined: Thu Aug 20, 2015 11:56 am

Re: Question for RobertT

Post by saurabh »

Thanks once again Robert T. Definitely a bit concerning that different implementations of EM value between DFA and FTSE RAFI can come to such large differences in sector weights. It is almost as if you are betting on who has the smart value trap for EM. I suppose by extension that is true for US and developed ex-US as well.

Also is there a reason you did not recommend the PowerShares FTSE RAFI US Small Cap fund? Per portfolio visualizer it has a SMB fctor load of 0.84, HML load of 0.32 and annual alpha of 1.27. t-stats for the factor loading have statistical significane although alpha does not.

https://www.portfoliovisualizer.com/fac ... x_cdt=true
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

saurabh wrote:Definitely a bit concerning that different implementations of EM value between DFA and FTSE RAFI can come to such large differences in sector weights. It is almost as if you are betting on who has the smart value trap for EM. I suppose by extension that is true for US and developed ex-US as well.
Personally, I focus more on factor exposure than sector weights. In relatively diversified portfolios and over the long haul "what matters is factor exposure and expense" (to quote Bill Bernstein). Even within specific sectors, factor exposure has explained much of the variation in returns over time ( Fama-French - Industry Costs of Equity ), but industries tend to have more time varying factor loads (as they become bigger/smaller, healthier/more distressed over time).

In the context of the DFA EM Value and FTSE RAFI EM, the DFA fund has a smaller mkt cap (and hence higher size load) than FTSE RAFI EM ($5.2bn vs. 18.6bn), and over the last 5 years EM small caps outperformed large caps (e.g. -2.6% vs. -4.6% 5-yr annualized returns for DFA EM Small and DFA EM respectively). In my view this explains a large share of the performance difference over this time period.
Also is there a reason you did not recommend the PowerShares FTSE RAFI US Small Cap fund? Per portfolio visualizer it has a SMB fctor load of 0.84, HML load of 0.32 and annual alpha of 1.27. t-stats for the factor loading have statistical significane although alpha does not.
The above portfolios are not recommendations, they are simply ways to achieve my personal factor load targets. Others may have different factor load targets (some just want beta), and a different combination of funds may achieve their particular factor load targets at lowest cost - and would also be good choices.

In this respect the Russell Fundamental US Small Cap series was a better fit to help reach a 0.2/0.4 size/value tilt. It had a (statistically significant) larger value load than the FTSE RAFI US Small cap series - based on a comparison of factor loads since the start of the Russell series in 1996 (as below), and a lower size load. For some, the FTSE RAFI US Small Cap may better fit with their factor load targets, which I think would also be fine.
  • Here are the comparison factor loads.

    1996-2014: US Small

    Russell Fundamental / FTSE RAFI
    1.02 / 1.07 = Mkt
    0.51 / 0.70 = Size
    0.63 / 0.54 = Value
Robert
.
Topic Author
saurabh
Posts: 319
Joined: Thu Aug 20, 2015 11:56 am

Re: Question for RobertT

Post by saurabh »

Robert T wrote: In this respect the Russell Fundamental US Small Cap series was a better fit to help reach a 0.2/0.4 size/value tilt. It had a (statistically significant) larger value load than the FTSE RAFI US Small cap series - based on a comparison of factor loads since the start of the Russell series in 1996 (as below), and a lower size load. For some, the FTSE RAFI US Small Cap may better fit with their factor load targets, which I think would also be fine.
  • Here are the comparison factor loads.

    1996-2014: US Small

    Russell Fundamental / FTSE RAFI
    1.02 / 1.07 = Mkt
    0.51 / 0.70 = Size
    0.63 / 0.54 = Value
Hi Robert T,

Can you clear up one confusion for me? Why are factor loads for size and value > 0.5 for a long only fund? Since HML and SMB are long/short portfolios I would have expected both to be below 0.5 if the excess returns are symmetrically distributed for size and value portfolios.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Question for RobertT

Post by larryswedroe »

Saura
Not Robert but, it's hard to load more than 0.5 or so for value because of the definitions, top 30% by whatever metric you use.

But it's much easier to load on size because of definition as top half. So think about micro cap funds, like Bridgeway BRUSX that only invest in the smallest of stocks , it's loading will be very high. With three factor model it loads at 0.7. If the same 30% construction rule was used for size as value it would likely be more like 0.5 or so.

Hope that helps
Larry
Topic Author
saurabh
Posts: 319
Joined: Thu Aug 20, 2015 11:56 am

Re: Question for RobertT

Post by saurabh »

larryswedroe wrote: But it's much easier to load on size because of definition as top half. So think about micro cap funds, like Bridgeway BRUSX that only invest in the smallest of stocks , it's loading will be very high. With three factor model it loads at 0.7. If the same 30% construction rule was used for size as value it would likely be more like 0.5 or so.
Thanks for the explanation Larry. That makes sense, since size break-point is set by median market cap it should be easier to capture more of the small cap premium even in a long-only fund by tilting towards the smallest stocks. It is a little bit ironic that between size and value the one that is considered more debatable is the easier one to capture for retail investors.
User avatar
in_reality
Posts: 4529
Joined: Fri Jul 12, 2013 6:13 am

Re: Question for RobertT

Post by in_reality »

larryswedroe wrote:Saura
Not Robert but, it's hard to load more than 0.5 or so for value because of the definitions, top 30% by whatever metric you use.
Not Saura, but Larry can I ask about the origin of the 30% construction rule for value?

I can't understand it conceptually. Why can size divide in half but not value.

Is that 30% true for non-capweighted indexes such as the Russell Fundamental series. If you look at the 9 box of FNDA (Russell Fundamental Small) there is a sizable allocation to mid-cap growth. Much larger than you find in VBR (Small Value) for instance. Obviously, these mid-cap growth funds are chosen as they are relatively cheap for their economic footprint (as RAFI measures). On an absolute scale, no they are typically considered value in a traditional value index.

So if we include these "relatively value" stocks, then does that 30% construction rule hold?
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

.
saurabh wrote:Can you clear up one confusion for me? Why are factor loads for size and value > 0.5 for a long only fund? Since HML and SMB are long/short portfolios I would have expected both to be below 0.5 if the excess returns are symmetrically distributed for size and value portfolios.
In addition to Larry’s post, my view is there are two parts to the answer (over the long term).
  • (i) There is a difference in the share of the long and the short side contributions to the size (SmB) and value (HmL) premiums (it is not exactly 50% from each). For example, here’s an extract from a paper by Israel and Moskowitz http://www.sciencedirect.com/science/ar ... 401…[i]”We find that long positions make up almost all of size, 60% of value, and half of momentum profits.”[/i] So as the long side has dominated the size premium its common to see a long-only small cap fund with a size load greater than 0.5. And if 60% of the value premium has been from the long-side its possible for the value load on a long-only value fund to be greater than 0.5.

    (ii) A perhaps larger part of the >0.5 small cap value load story IMO, is that a larger share of the HML premium has been from Small Cap HML rather than from Large Cap HML. When combined in the overall HML, large cap value loads often tend to be less than 0.5 - given that a larger share of HML has been from Small cap HML which is not possible to capture with a large cap value fund. Small cap value loads tend to be greater than 0.5 - given that HML includes Large cap HML, in some sense ‘watering down’ HML and resulting in it being ‘easier’ for small cap value funds to achieve a >0.5 value load.
Robert
.
Topic Author
saurabh
Posts: 319
Joined: Thu Aug 20, 2015 11:56 am

Re: Question for RobertT

Post by saurabh »

Robert T wrote:.
In addition to Larry’s post, my view is there are two parts to the answer (over the long term).
  • (i) There is a difference in the share of the long and the short side contributions to the size (SmB) and value (HmL) premiums (it is not exactly 50% from each). For example, here’s an extract from a paper by Israel and Moskowitz
Thanks Robert, that was very helpful.
larryswedroe
Posts: 16022
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Question for RobertT

Post by larryswedroe »

In reality

Sorry don't have good answer for you. And it makes no sense really. And note that the size premium would be larger and more persistent if we measured it the same way we measure value, using 30/40/30. And at least with models factors tend to be that way, not 50/50, with exception of small.
Larry
Theoretical
Posts: 1514
Joined: Tue Aug 19, 2014 10:09 pm

Re: Question for RobertT

Post by Theoretical »

..8% iShares US Momentum - ER .15
11% Vanguard MidCap Value - ER .08
19% Schwab Fundamental US Small Company - ER .25
17% Powershares RAFI Developed Market Non-US - ER .25
11% Powershares RAFI Developed Market Non-US small cap - ER .5
10% Powershares RAFI Emerging Markets - ER .47
25% iShares 3-7 yr Treasury - ER .15

The expense ratio for this portfolio just dropped by 16% from .30% to .25% (with Schwab and Vanguard's fee reductions). That's extremely impressive.
duffer
Posts: 314
Joined: Tue Sep 09, 2014 8:18 pm

Re: Question for RobertT

Post by duffer »

Hi Robert T.

What is your view these days on a portfolio comprising:

30% MTUM
40% VOE
30% PXSV or FNDA

Thanks,
duffer
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

duffer wrote:Hi Robert T.

What is your view these days on a portfolio comprising:

30% MTUM
40% VOE
30% PXSV or FNDA

Thanks,
duffer
Duffer,

I prefer FNDA over PXSV since the latter changed to track the Russell 2000 Pure Value Index.

Factor exposure
  • If you are targeting a long-term 0.15/0.40 size and value load then 30:40:30 MTUM:VOE:FNDA is a reasonable way to try to achieve it (using the factor loads posted earlier in this thread).

    If you are targeting a long-term 0.20/0.40 size and value load then 30:30:40 MUTM:VOE:FNDA is a reasonable way to try to achieve it.
Expense
  • Expense ratios on the above portfolios = 0.2 or less
    Tax cost ratio (3 yr) of the the weighted average of the three funds is about 0.1 less than Vanguard TSM (VTI) according the M*.
    Current expense ratio + weighted average tax cost ratio (past 3 years) of above portfolios is similar to Vanguard TSM (VTI). If you are in the accumulation stage you can rebalance with new savings thereby reducing the need (and trade and tax costs) to sell and buy to rebalance.
"Over the long-haul what matters is factor exposure and expense" - Bill Bernstein.

Obviously no guarantees.

Robert
.
duffer
Posts: 314
Joined: Tue Sep 09, 2014 8:18 pm

Re: Question for RobertT

Post by duffer »

Robert T wrote:
duffer wrote:Hi Robert T.

What is your view these days on a portfolio comprising:

30% MTUM
40% VOE
30% PXSV or FNDA

Thanks,
duffer
Duffer,

I prefer FNDA over PXSV since the latter changed to track the Russell 2000 Pure Value Index.

Factor exposure
  • If you are targeting a long-term 0.15/0.40 size and value load then 30:40:30 MTUM:VOE:FNDA is a reasonable way to try to achieve it (using the factor loads posted earlier in this thread).

    If you are targeting a long-term 0.20/0.40 size and value load then 30:30:40 MUTM:VOE:FNDA is a reasonable way to try to achieve it.
Expense
  • Expense ratios on the above portfolios = 0.2 or less
    Tax cost ratio (3 yr) of the the weighted average of the three funds is about 0.1 less than Vanguard TSM (VTI) according the M*.
    Current expense ratio + weighted average tax cost ratio (past 3 years) of above portfolios is similar to Vanguard TSM (VTI). If you are in the accumulation stage you can rebalance with new savings thereby reducing the need (and trade and tax costs) to sell and buy to rebalance.
"Over the long-haul what matters is factor exposure and expense" - Bill Bernstein.

Obviously no guarantees.

Robert
.
Robert,

Thanks so much for your generosity with your expertise!

Duffer
User avatar
Lieutenant.Columbo
Posts: 1178
Joined: Sat Sep 05, 2015 9:20 pm
Location: Cicely AK

Re: Question for RobertT

Post by Lieutenant.Columbo »

Robert T wrote:
  • ..9% iShares S&P600 Value
    ..5% Bridgeway Small Co. Mkt
Robert T,
What does each of these two holdings give you that the other one doesn't?
In other words: couldn't you as well simpkyfy by having
14% iShares S&P600 Value
or
14% Bridgeway Small Co. Mkt?
Thank you.
Lt. Columbo: Well, what do you know. Here I am talking with some of the smartest people in the world, and I didn't even notice!
DaufuskieNate
Posts: 405
Joined: Wed May 28, 2014 11:53 am

Re: Question for RobertT

Post by DaufuskieNate »

in_reality wrote:
larryswedroe wrote:Saura
Not Robert but, it's hard to load more than 0.5 or so for value because of the definitions, top 30% by whatever metric you use.
Not Saura, but Larry can I ask about the origin of the 30% construction rule for value?

I can't understand it conceptually. Why can size divide in half but not value.
In their 1993 study Common Risk Factors in the Returns on Stocks and Bonds, Fama and French describe how their data sets were constructed. Here is a quote:

“Our decision to sort firms into three groups on BE/ME and only two on ME follows the evidence in Fama and French (1992a) that book-to-market equity has a stronger role in average stock returns than size. The splits are arbitrary, however, and we have not searched over alternatives.”
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

Lieutenant.Columbo wrote:
Robert T wrote:
  • ..9% iShares S&P600 Value
    ..5% Bridgeway Small Co. Mkt
Robert T,
What does each of these two holdings give you that the other one doesn't?
In other words: couldn't you as well simpkyfy by having
14% iShares S&P600 Value
or
14% Bridgeway Small Co. Mkt?
Thank you.
Its important to go back to the framework I use for making portfolio decisions i.e. to target a 0.20/0.40 size and value load within equities. Using 14% iShares S&P600 Value would shift this to 0.18/0.43, and using 14% Bridgeway Small Co. Mkt would shift it to 0.26/0.34. Some may think these are small difference, but then what's the point of having portfolio factor load targets if you are not going to try to achieve these the best way you can.

In addition, arguably an efficient way of getting a small cap tilt is to get it all in CRSP10 (the smallest decile of equities) given apparent non-linearities in the size premium. For example see page 22 of the Bridgeway Annual Report https://s3.amazonaws.com/bridgeway_web/ ... dgeway.pdf

Annualized returns over last 90.5 years

CRSP Deciles
Decile 1 (largest) = 9.26%
Decile 2 = 10.50%
Decile 3 = 10.97%
Decile 4 = 10.78%
Decile 5 = 11.38%
Decile 6 = 11.25%
Decile 7 = 11.41%
Decile 8 = 11.39%
Decile 9 = 11.31%
Decile 10 (smallest) = 13.13%

The Bridgeway fund (BRSIX) tracks the CRSP10 index. Some don't believe its possible to track the CRSP10 index, and indeed annual tracking error of BRSIX has often been huge, but since inception of BRSIX:

7/31/1997 to 12/31/2016: Annualized Returns
  • BRSIX = 10.83%
    CRSP10 = 10.83%
    Not a typo.
The combination of MidCap Value (VOE) and CRSP10 (BRSIX) is along the lines mentioned above i.e. getting value exposure from a fund with a close to zero size load and size exposure from BRSIX (to try to take advantage of the apparent non-linearity in the size premium). The iShare S&P600 Value fund was then added to achieve the 0.20/0.40 portfolio factor load targets.

The downside of BRSIX is that while it used to be tax-managed, now it is not and so is fairly tax-inefficient and would only suggest holding in tax advantaged accounts.

Obviously no guarantees.

Robert
.
Theoretical
Posts: 1514
Joined: Tue Aug 19, 2014 10:09 pm

Re: Question for RobertT

Post by Theoretical »

Robert, that's absolutely shocking long-term tracking, better than a lot of major index funds not named Vanguard or iShares.
User avatar
triceratop
Posts: 5838
Joined: Tue Aug 04, 2015 8:20 pm
Location: la la land

Re: Question for RobertT

Post by triceratop »

Robert T wrote:The combination of MidCap Value (VOE) and CRSP10 (BRSIX) is along the lines mentioned above i.e. getting value exposure from a fund with a close to zero size load and size exposure from BRSIX (to try to take advantage of the apparent non-linearity in the size premium). The iShare S&P600 Value fund was then added to achieve the 0.20/0.40 portfolio factor load targets.
I take it you don't lend too much credence to the thought that the value premium also exhibits nonlinearity with respect to dependence on size? Because then you would want VBR to maximize efficient capture of the premia.
"To play the stock market is to play musical chairs under the chord progression of a bid-ask spread."
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

triceratop wrote:I take it you don't lend too much credence to the thought that the value premium also exhibits nonlinearity with respect to dependence on size? Because then you would want VBR to maximize efficient capture of the premia.
Here is some data on value returns by size.

1927-2014: Annualized returns
The difference in return between Large Value and Small Value = 4.1%. MidCap Value 'captures' 63% of this difference (14.1-11.5)/(15.6-11.5)

1969-2005: Annualized returns
  • Large Value = 10.8%
    MidCap Value = 14.0%
    Small Value = 15.3%
    Source: Ibbotson Yearbook
The difference in return between Large Value and Small Value = 4.5%. MidCap Value 'captures' 71% of this difference (14.0-10.8)/(15.3-10.8)

Re: non-linearities. The largest return difference is not between Large/Mid Value and Small Value, it is between Large and Mid/Small Value - at least historically. Obviously no guarantees.

FWIW - what matters to me is to target specific portfolio factor loads. This includes trying to ensure a portfolio alpha is close to zero (non-negative) and has a high R^2 in the FF 3 factor analysis . Both of these are indicators of "pure" exposure to the factors I want to target i.e. factor exposure that is not eroded by negative alpha, and that exposure to mkt, size and value account for close to all the variability in returns.

Obviously these can only be estimated from historical data and its important to use the longest time frame possible to capture the full market cycle. The assumption is that the historical portfolio factor loads will be indicative of the future factor loads. This says nothing about the size of the factor premiums, just the share of any 'premium' (positive or negative) that is captured by the portfolio. Obviously no guarantees.

Following the above, and getting back to which approach is better to target size and value exposure - For me, no approach is better than another if they both deliver a 0.20/0.40 size/value loads with zero alpha, a high R^2, at the same total cost. IMO this provides a useful framework for making portfolio decisions on the 'best' approach.

I understand this doesn't work for everyone.

Robert
.
Coato
Posts: 185
Joined: Wed Oct 21, 2015 4:34 pm

Re: Question for RobertT

Post by Coato »

Resuscitating this thread with a question....

Broadly speaking could a REIT fund (like the Fidelity index which is concentrated in the top four right side boxes per Morningstar) roughly serve the same purpose of the Momentum ETF? I know they are invested differently, but does it provide the same movement under similar market conditions?
AlohaJoe
Posts: 5539
Joined: Mon Nov 26, 2007 2:00 pm
Location: Saigon, Vietnam

Re: Question for RobertT

Post by AlohaJoe »

Vanguard's REIT fund has negative momentum, so it isn't a substitute. I assume Fidelity's fund is similar.
Theoretical
Posts: 1514
Joined: Tue Aug 19, 2014 10:09 pm

Re: Question for RobertT

Post by Theoretical »

Also, there's nothing indicative about a REIT tilt that gets you to momentum. The closest analogue would be a Midcap growth fund.
Justin W
Posts: 5
Joined: Fri Sep 15, 2017 4:46 pm

Re: Question for RobertT

Post by Justin W »

(Responding to an old thread, but very similar question as the original—)

Robert,

Like the OP, I've been an avid reader of your posts, and they (and especially this thread) have greatly informed my own portfolio decisions. Thank you for sharing your insights freely and regularly.

I'd love to hear if you have any thoughts yet on the new Vanguard factor and multifactor funds. Do you plan to use any of them? Is it possible to approximate some of their factor loads based not on their history (since they're brand new), but on the current composition of the funds?

I'm guessing that you do not plan to use the value factor fund (VFVA) in place of Vanguard Midcap Value (VOE) since more of the value premium is located in mid- and small-caps rather than large-caps. (That, at least, was my understanding of why you prefer VOE over the iShares Value Factor ETF VLUE.) But since VFVA reaches into small caps whereas VLUE doesn't, I'm guessing you would prefer it over VLUE? And maybe it's a tough decision between VOE and VFVA?

What about Vanguard's new momentum fund (VFMO) vs iShares Momentum (MTUM)? Do you have any reasons for or against VFMO replacing MTUM?

And what about the multifactor fund (VFMF/VFMFX)? Do you think that now (or at some point in the future, as we see some results) you might replace portions of MTUM, VOE, and/or FNDA (Schwab Fundamental Small) in your portfolio with VFMF?

I currently have MTUM, VOE, and FNDA for my domestic allocation. I'm considering adding VFMF to the mix, or replacing MTUM and VOE with it.

In regard to VFMF vs a combination of VFVA+VFMO, VFMF seems superior to me. I'm guessing that VFMF will be able to achieve equal (or higher) value and momentum factor loads than a combination of VFVA+VFMO, with a positive quality load and volatility screen to boot. But that's only a guess. Vanguard's paper "Equity Factor-Based Investing: A Practitioner’s Guide" also leads me to believe VFMF would have higher projected returns (potentially with higher tracking error and volatility) than, say, 50% VFVA/50% VFMO.

Justin
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

Justin W wrote: Wed Mar 14, 2018 6:45 pm I'd love to hear if you have any thoughts yet on the new Vanguard factor and multifactor funds. Do you plan to use any of them?
No changes for me. Don't have many thoughts yet on these new funds - but here are some.

When evaluating new products, I simply ask the question: Can it achieve my factor load targets at lower cost?

Factor loads
  • I do not know the factor loads of the new vanguard funds. In the absence of historical underlying index data, we will need to rely on live returns data. This may take a while to accumulate if we want factor load estimates over a period covering different market environments. What we have is current metrics, and historical live data on Vanguard's global factor funds back to the start of 2016. I am not sure if the Vanguard global factor funds use the same construction methodology as the Vanguard US factor funds, but they are managed by the same Quantitative Equity group. Over the very short two-year time period, the factor loads on the Vanguard Global Momentum and Value funds suggest they are small cap (size loads of about 0.6). The Vanguard Global Momentum load was smaller than on MSCI World Momentum index over this period.

    FWIW - from Jan 2016 to Feb 2018.

    Annualized return (%): annualized standard deviation

    19.2/13.9 = MSCI World Momentum Index
    14.0/13.3 = MSCI World Index
    13.9/15.5 = Vanguard Global Momentum fund

    Unfortunately, as far as I can tell, MSCI does not have a smaller cap World Momentum series for a closer comparison. The value factor funds look promising. While current metrics and the two-year implementation period of the Vanguard Global factor funds is interesting, it does not provide enough information for me to make an informed decision (relative to alternatives).
Costs
  • The Vanguard US Momentum factor ETF has an expense ratio of 0.13 compared to 0.15 for the iShares MSCI US Momentum ETF. But this is not the total cost. As the Vanguard fund is smaller cap it will likely have larger market impact/trade costs, and lower capacity. The Vanguard US Value factor ETF also has an expense ratio of 0.13 compared to 0.19 for a 3:5 combination of Vanguard Mid-Cap Value (VOE): Schwab US Fundamental (FNDA). But again, I don’t know the factor loads of the Vanguard US Value factor so have no basis for matching this with my factor load targets.
Personally, I prefer to have a clear framework for making decisions (re: Can it achieve my factor load targets at lower cost?) and to stick to this framework. If this exclude consideration of products without information on factor loads, then so be it. New products are coming out all the time, and the absence of a clear framework can increase instability in the investment processes i.e. tendency for frequent portfolio changes being driven simply by new product releases (and some of the hype around them). In this context, this quote comes to mind – the “enemy of a good plan is the dream for a perfect plan”.

Just my take,
Robert
Elysium
Posts: 3223
Joined: Mon Apr 02, 2007 6:22 pm

Re: Question for RobertT

Post by Elysium »

Do you still believe value premium exists? it hans't outperformed over the past several years, except for a fews years from the tech crash all the way up to the financial crisis. Before and after that in practice small value appears to have done almost same or underperformed small growth / blend.
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

Dieharder wrote: Thu Mar 15, 2018 8:30 am Do you still believe value premium exists?
I have a long-term value load target of 0.4, so believe that averaged over my investment life-span there will be a value premium. Obviously no guarantees that this will be the case https://papers.ssrn.com/sol3/papers.cfm ... id=3081101
.
Justin W
Posts: 5
Joined: Fri Sep 15, 2017 4:46 pm

Re: Question for RobertT

Post by Justin W »

Very helpful. Thanks, Robert.
Jiu Jitsu Fighter
Posts: 180
Joined: Mon Nov 21, 2016 10:22 am

Re: Question for RobertT

Post by Jiu Jitsu Fighter »

Hi Robert,

Thanks for the incredible value (pun intended) that you bring to this group. I would like to get your thoughts on Invesco's pure value indices. More specifically, VOE vs RPV (S&P 500 Pure Value) - any preference? RPV has a higher value load but somewhat offset by more negative momentum as well as a higher expense ratio.

Thanks!
JJF
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

Jiu Jitsu Fighter wrote: Mon Jan 14, 2019 12:32 pm More specifically, VOE vs RPV (S&P 500 Pure Value) - any preference? RPV has a higher value load but somewhat offset by more negative momentum as well as a higher expense ratio.
When considering alternatives I simply ask the question - will inclusion achieve my factor load targets at lower cost? The answer is no if I were to use RPV instead of VOE.

From earlier posts in this thread. The estimated long-term factor load for equity portion = 1.01 / 0.26 / 0.46 for mkt / size / value achieved with MTUM / VOE / FNDA at an expense ratio of 0.18%. If I use RPV instead of VOE then my estimate is that 30% MTUM / 15% RPV / 55% FNDA gives a similar long-term factor load at an expense ratio of 0.24%. So it does not lower the cost of achieving my factor load targets (this does not consider the longer term negative alpha of S&P500 Pure Value relative to CRSP MidCap Value - although this has not shown up recently). This may seem like a small difference (0.18 vs 0.24), but I think it is important to have (and use) a framework for making portfolio decisions.

Robert
.
Random Walker
Posts: 4612
Joined: Fri Feb 23, 2007 8:21 pm

Re: Question for RobertT

Post by Random Walker »

Just to add my 2 cents. I’d just make the point that cost, while very important, isn’t everything. The purpose of factor investing is to diversify away from the market beta factor that dominates most all of our portfolios. Sometimes a factor fund with deeper exposures to the desired factors will cost a bit more, but allow us to achieve our factor targets while taking on less market beta risk. This can be worthwhile and perhaps worth a few basis points. Moreover, with deeper factor exposures, less of this fund is required to achieve the factor targets. This allows the investor to use more of the relatively cheap total market or core fund.

Dave
User avatar
Robert T
Posts: 2740
Joined: Tue Feb 27, 2007 9:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: Question for RobertT

Post by Robert T »

Random Walker wrote: Fri Jan 18, 2019 2:20 pm I’d just make the point that cost, while very important, isn’t everything.
Once you have set your specific factor load targets - then total cost is everything IMO. i.e. achieve them at lowest total cost.

Some investor may have lower or no target loads for value and size, others may have higher value loads for value and size, and lower beta. In either case, once their specific targets are set - then achieving them at lowest total cost is everything.

Robert
.
Post Reply