S&P Indices: Blending Factors In Smart Beta Portfolios

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
Topic Author
Random Walker
Posts: 3961
Joined: Fri Feb 23, 2007 8:21 pm

S&P Indices: Blending Factors In Smart Beta Portfolios

Post by Random Walker » Wed Jul 10, 2019 10:49 pm

https://us.spindices.com/documents/educ ... nload=true

Here’s an article from S&P on factors: low volatility, size, quality, value, momentum. It studies the period 1995-2018 and compares S&P 500 single factor indices to the S&P 500. Over the long run, each factor displays a premium and the correlations between the factors are low. Moreover the correlations remain low during periods of crisis. The different factors have tended to pay off at different times of the market cycle, investment cycle, and investment sentiment regimes.

It is particularly attractive to combine factor exposures in a single equity portfolio. There is clearly a trade off between portfolio tracking error to the benchmark S&P 500 and portfolio volatility. Over this time period, accepting increased tracking error resulted in a more efficient portfolio. Important to appreciate this study is looking at indexes, not actual funds with their associated costs. Exhibit 6 is particularly interesting; the negative correlation between momentum and just about everything else is impressive.

Dave

heyyou
Posts: 3480
Joined: Tue Feb 20, 2007 4:58 pm

Re: S&P Indices: Blending Factors In Smart Beta Portfolios

Post by heyyou » Thu Jul 11, 2019 1:22 am

Seems like using the 500 for factor exposure, is a good middle ground between full slice and dice and a total market portfolio.
accepting increased tracking error resulted in a more efficient portfolio.
There is the risk and reward.

Dave, I was enthusiastic until "This suggests potential for strategic and tactical factor allocation." I'll keep my passive, low maintenance slice and dice of the whole market, and let others do tactics on their slices of the 500 portfolio.

schismal
Posts: 38
Joined: Sat Apr 13, 2019 8:53 pm

Re: S&P Indices: Blending Factors In Smart Beta Portfolios

Post by schismal » Thu Jul 11, 2019 5:17 am

Very interesting article -- thanks for the read.

BlackRock just launched a factor rotation ETF -- DYNF-- that I've been watching with great interest. It operates on similar principles to the S&P model at the end of the paper.

schismal
Posts: 38
Joined: Sat Apr 13, 2019 8:53 pm

Re: S&P Indices: Blending Factors In Smart Beta Portfolios

Post by schismal » Thu Jul 11, 2019 5:22 am

Random Walker wrote:
Wed Jul 10, 2019 10:49 pm
Exhibit 6 is particularly interesting; the negative correlation between momentum and just about everything else is impressive.
Agreed. Although it's notable that the returns for momentum held in perpetuity (fig. 5) are actually not that impressive. That may give credence to -- and I realize this is blasphemy here -- the notion of momentum timing as explored by this NBER paper: LINK

Topic Author
Random Walker
Posts: 3961
Joined: Fri Feb 23, 2007 8:21 pm

Re: S&P Indices: Blending Factors In Smart Beta Portfolios

Post by Random Walker » Thu Jul 11, 2019 8:10 am

heyyou wrote:
Thu Jul 11, 2019 1:22 am
Seems like using the 500 for factor exposure, is a good middle ground between full slice and dice and a total market portfolio.
accepting increased tracking error resulted in a more efficient portfolio.
There is the risk and reward.

Dave, I was enthusiastic until "This suggests potential for strategic and tactical factor allocation." I'll keep my passive, low maintenance slice and dice of the whole market, and let others do tactics on their slices of the 500 portfolio.
I don’t really view tracking error, perhaps better referred to as tracking variance, as a risk. Perhaps it’s better viewed as a behavioral error. What matters is meeting one’s goals, not meeting or outperforming a somewhat random benchmark.

I agree with you on the tactical allocation; my eyes glazed over a bit when I thought they were starting to describe factor timing. I’m a big believer in having a static plan and sticking to it. I believe some extremely smart people have said that factor timing is really really hard!

Dave

User avatar
patrick013
Posts: 2626
Joined: Mon Jul 13, 2015 7:49 pm

Re: S&P Indices: Blending Factors In Smart Beta Portfolios

Post by patrick013 » Thu Jul 11, 2019 12:20 pm

Random Walker wrote:
Wed Jul 10, 2019 10:49 pm

It is particularly attractive to combine factor exposures in a single equity portfolio.

...

Important to appreciate this study is looking at indexes, not actual funds with their associated costs.

Great info. Russell has similar info for their 7 factors. I expected yield factor to do well now that it's established and it was interesting that quality did well in both factor reports. Should be some conservative investors interested in Quality for some equity exposure and happy with the safety a low debt/equity ratio can provide in a possible down market.

I picked 2 of the best and made a low beta AA. Expecting Qual to do well but High Div to do better. It's nice they're releasing info about these current funds with short histories. Got the beta below .7 and SD below 10 so minimum volatility still provides a decent estimate of portfolio return/risk. The metrics are only 3.5 years back but still a good test portfolio example with factor information.

Image

My attempt at risk parity just by limiting portfolio beta, no fancy algorithm. For the adventurists. Tickers are SPHQ, SPYD, AND VFIUX.




...
age in bonds, buy-and-hold, 10 year business cycle

User avatar
nisiprius
Advisory Board
Posts: 38510
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: S&P Indices: Blending Factors In Smart Beta Portfolios

Post by nisiprius » Fri Jul 12, 2019 7:25 am

Random Walker wrote:
Wed Jul 10, 2019 10:49 pm
...Moreover the correlations remain low during periods of crisis...
Let's see how some multifactor funds and ETF did around year-end 2018. That is, did these low correlations pay off in cushioning the decline?

I have not looked at this before and will post whatever I find. I'm going to include VFMF because it's been of interest in the forum. For the others, I'm going to use funds mentioned in a Morningstar report, A Framework for Analyzing Multifactor Funds because Morningstar chose them as credible examples of the genre and seems to have nice things to say about many of them. In Morningstar's list, I omitted a non-US funds, a fund with no ticker symbol, and one with a ticker symbol that doesn't show up in a Morningstar search.
  • VFMF, Vanguard Multifactor Fund
  • QCLEX, AQR Large Cap Multistyle, "well-crafted strategy..."
  • DEUS, Xtrackers Russell 1000 Comprehensive Factor ETF, "solid choice to own a broadly diversified U.S. stock portfolio that tilts toward stocks that score well on five distinct factors..."
  • DFEOX, DFA U.S. Core Equity 1 Portfolio, "great choice to own a broadly diversified U.S. stock portfolio that moderately tilts toward stocks with attractive characteristics..."[/color]
  • GSLC, Goldman Sachs ActiveBeta US Large Cap Equity ETF, "...factor tilts are modest, it has a low expense ratio to match, which gives it a reasonable chance to beat the market over the long term."
  • LRGF, iShares Edge MSCI Multifactor USA ETF, "Its factor tilts are stronger than most of its peers, but its portfolio construction is opaque and complex."
  • JPUS, JPMorgan Diversified Return US Eq ETF, "targets stocks with stronger price momentum, lower valuations, and higher profitability [but] its targeted exposures are overwhelmed by the fund’s risk-weighting approach."
I included the S&P 500 index into the list as the fourth item because Morningstar plots the fourth one in a hard-to-see yellow and I wanted all the actual funds to be easy to see. (As a result the S&P 500 index is almost invisible! However, GSLC and the S&P 500 plotted virtually identically so just look for the blue GSLC curve, second from the top).

Source

Image

The S&P 500 declined -15.4%. Five of the multifactor funds declined more. One, GSLC, basically tied (-15.5%). Only one of them, JPUS, declined less. Interestingly, that's one that Morningstar didn't seem to like. The seven multifactor funds declined by an average of -17.5%.

For other time periods of interest, 2011 and 2008-2009, only one of Morningstar's choices existed: DFEOX, which Morningstar calls a "great choice." It declined lower than the S&P 500 during 2011, -20.2% versus -16.3%, and (microscopically) lower during the global financial crisis, -51.7% versus -50.9%.

None of these results would bother me much if happened to own one of these funds, but it at least raises a question about whether "correlations remaining low in time of crisis" was occurring--and, if so, what benefits did that provide to investors as a result?
Last edited by nisiprius on Fri Jul 12, 2019 8:14 am, edited 4 times in total.
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.

garlandwhizzer
Posts: 2334
Joined: Fri Aug 06, 2010 3:42 pm

Re: S&P Indices: Blending Factors In Smart Beta Portfolios

Post by garlandwhizzer » Fri Jul 12, 2019 1:22 pm

This is a very interesting article but I believe the most important sentence in RandomWalker's post is this:

Important to appreciate this study is looking at indexes, not actual funds with their associated cost
As Nisi points out, results to date for multi-factor funds have been disappointing for investors who hopped on this anticipated gravy train. In a preceding post, JoMoney pointed out in long term comparisons of DFA small value versus small blend funds there was no harvested value premium over the last 26 years of their existence. Returns were essentially identical. Likewise JoMoney pointed out that over 37 years DFSCX, DFA's Small Cap Fund, did not outperform Vanguard's S&P 500 Index Fund, demonstrating a failure to capture the small premium for 37 years. Although presumably factor premiums for small and for value were robust over these time frames, the results from real funds are in my mind more important to investors than the academically defined premiums. Certainly it raises the question how reliable harvesting of these premiums are in real funds.

No one doubts that on cost-free/long short portfolio backtesting using the optimal parameters to define the premiums in the rear view mirror and the optimal cut offs for trading frequency in the rear view mirror, that factor premiums are positive and significant. Unfortunately you can't do that in real funds going forward. In the real world your past mistakes educate you. These academic models are designed to maximally magnify premiums and make them appear more significant without regard to trading costs and trading frictions, not to mention management costs and advisor costs that real investors have to deal with. Publish or perish is the rule in academics and, hence, an incredible amount of signal has been "discovered" in market action which is in fact for practical purposes IMO mostly noise. The above examples demonstrate that how well robust factor research translates into real world funds is a legitimate concern. Results for investors have been spotty, not consistently reliable in terms of risk adjusted return. In contrast as factors are defined in academia, they are sure winners with optimized multi-factor models producing reliable increased risk adjusted returns with lower volatility.

Investors who choose factor dominated portfolios need to realize two things IMO. First, they must have a lot of patience. Second, they should realize that they are making a bet based on their perception of probabilities, not certainties. That bet may or may not pay off for them in their time frame.

Garland Whizzer

dcabler
Posts: 991
Joined: Wed Feb 19, 2014 11:30 am

Re: S&P Indices: Blending Factors In Smart Beta Portfolios

Post by dcabler » Fri Jul 12, 2019 1:26 pm

nisiprius wrote:
Fri Jul 12, 2019 7:25 am
Random Walker wrote:
Wed Jul 10, 2019 10:49 pm
...Moreover the correlations remain low during periods of crisis...
Let's see how some multifactor funds and ETF did around year-end 2018. That is, did these low correlations pay off in cushioning the decline?

I have not looked at this before and will post whatever I find. I'm going to include VFMF because it's been of interest in the forum. For the others, I'm going to use funds mentioned in a Morningstar report, A Framework for Analyzing Multifactor Funds because Morningstar chose them as credible examples of the genre and seems to have nice things to say about many of them. In Morningstar's list, I omitted a non-US funds, a fund with no ticker symbol, and one with a ticker symbol that doesn't show up in a Morningstar search.
  • VFMF, Vanguard Multifactor Fund
  • QCLEX, AQR Large Cap Multistyle, "well-crafted strategy..."
  • DEUS, Xtrackers Russell 1000 Comprehensive Factor ETF, "solid choice to own a broadly diversified U.S. stock portfolio that tilts toward stocks that score well on five distinct factors..."
  • DFEOX, DFA U.S. Core Equity 1 Portfolio, "great choice to own a broadly diversified U.S. stock portfolio that moderately tilts toward stocks with attractive characteristics..."[/color]
  • GSLC, Goldman Sachs ActiveBeta US Large Cap Equity ETF, "...factor tilts are modest, it has a low expense ratio to match, which gives it a reasonable chance to beat the market over the long term."
  • LRGF, iShares Edge MSCI Multifactor USA ETF, "Its factor tilts are stronger than most of its peers, but its portfolio construction is opaque and complex."
  • JPUS, JPMorgan Diversified Return US Eq ETF, "targets stocks with stronger price momentum, lower valuations, and higher profitability [but] its targeted exposures are overwhelmed by the fund’s risk-weighting approach."
I included the S&P 500 index into the list as the fourth item because Morningstar plots the fourth one in a hard-to-see yellow and I wanted all the actual funds to be easy to see. (As a result the S&P 500 index is almost invisible! However, GSLC and the S&P 500 plotted virtually identically so just look for the blue GSLC curve, second from the top).

Source

Image

The S&P 500 declined -15.4%. Five of the multifactor funds declined more. One, GSLC, basically tied (-15.5%). Only one of them, JPUS, declined less. Interestingly, that's one that Morningstar didn't seem to like. The seven multifactor funds declined by an average of -17.5%.

For other time periods of interest, 2011 and 2008-2009, only one of Morningstar's choices existed: DFEOX, which Morningstar calls a "great choice." It declined lower than the S&P 500 during 2011, -20.2% versus -16.3%, and (microscopically) lower during the global financial crisis, -51.7% versus -50.9%.

None of these results would bother me much if happened to own one of these funds, but it at least raises a question about whether "correlations remaining low in time of crisis" was occurring--and, if so, what benefits did that provide to investors as a result?
FYI - there are a number of funds that don't show up in the M* search for some reason, but the info is still there. Simply choose a fund that does show up in the search, then go into the URL and replace that ticker with the one you're looking for... It's an annoying "feature" of M* for some reason.

Forester
Posts: 185
Joined: Sat Jan 19, 2019 2:50 pm
Location: UK

Re: S&P Indices: Blending Factors In Smart Beta Portfolios

Post by Forester » Fri Jul 12, 2019 7:05 pm

27 Sept to 24 Dec 2018

iShares US min vol -12%

S&P 500 -19%

Post Reply