AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
User avatar
Topic Author
matjen
Posts: 2189
Joined: Sun Nov 20, 2011 10:30 pm

AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by matjen »

The latest from AQR’s Cliff Asness. As Cliff says himself:
The Momentum Factor: Its Not Just For Breakfast Anymore

Oh, and it survives t-costs, and studies that say it doesn’t are just poorly done nonsense.


https://www.aqr.com/cliffs-perspective/ ... 20it%20Too
A man is rich in proportion to the number of things he can afford to let alone.
User avatar
White Coat Investor
Posts: 17409
Joined: Fri Mar 02, 2007 8:11 pm
Location: Greatest Snow On Earth

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by White Coat Investor »

Oh? AQR and Cliff Asness think the momentum factor is real? That seems like an unbiased source.
1) Invest you must 2) Time is your friend 3) Impulse is your enemy | 4) Basic arithmetic works 5) Stick to simplicity 6) Stay the course
Random Walker
Posts: 5561
Joined: Fri Feb 23, 2007 7:21 pm

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by Random Walker »

Thanks Matjen for posting the article. I assume this article is discussing CS Momentum as opposed to TS Momentum?

Dave
User avatar
Topic Author
matjen
Posts: 2189
Joined: Sun Nov 20, 2011 10:30 pm

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by matjen »

Random Walker wrote: Wed Dec 06, 2017 5:23 pm Thanks Matjen for posting the article. I assume this article is discussing CS Momentum as opposed to TS Momentum?

Dave
I assume so Dave but I haven't read it thoroughly yet. Just skimmed it but they are discussing MOM so I take that to mean CS Momentum.
A man is rich in proportion to the number of things he can afford to let alone.
garlandwhizzer
Posts: 3565
Joined: Fri Aug 06, 2010 3:42 pm

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by garlandwhizzer »

Mr. Asness's research may well show that MOM can be harvested on an after cost basis and outperform the appropriate comparable index. Research is one thing. Real funds are another. Let's look at the performance of AQR's 2 MOM funds which presumably over a significant time period will reflect this outperformance. Let's compare their returns to funds which operate in the same market arena, growth, and in the same cap weight segment. That seems to be the appropriate comparison, small cap MOM versus small cap growth index, both active in the same market segment but one, growth, which pays no attention to the MOM factor, only to profit growth. Likewise, large cap MOM versus large cap growth index which again pays no attention to the MOM factor. AQR's large cap MOM fund, AMOMX has been around for 9 years, not exactly a lifetime but a significant period of time. Since its inception 9 years ago according to Morningstar it has underperformed VUG, Vanguards low cost large cap growth index fund that pays no attention to MOM. They play in the same cap weight and same spot on the growth value scale, and after nine years MOM trails the plain old growth index that costs a lot less. The very same result occurs when we compare AQR's small cap MOM fund, ASMOX to plain old vanilla Vanguard Small Cap Growth Index fund which has also outperformed ASMOX since its inception 9 years ago. Perhaps 9 years isn't long enough for MOM to show its superiority as true factor adherents believe.

It strikes me as odd that AQR with all its brilliance can't outperform a comparable index in what looks like an era in which growth and MOM ruled. Oddly, it also seems that the SCV premium has also been missing in action for real funds for more than a decade now. Likewise the small factor has withered so much that it has become controversial as to whether it even exists. Those who have 100% equity in factor approaches have been packed and ready, poised for a long time waiting for the rocket ship to take off. It could happen next year or perhaps not for another decade.

Irrespective of a decade or so of underperformance, factor approaches promise eventual outperformance at some indefinite point in the future based on past testing academic research. I am retired and in the withdrawal rather than accumulation phase. I sell equity without fail every year now to provide for living expenses. A decade of underperformance relative to low cost indexes poses a problem for me, selling before the rocket ship takes off, if indeed it ever takes off before I croak. Tracking error is a real, rather than theoretical, problem for me. Hence I use both factors and plain low cost indexes, a smoother ride, selling whatever hasn't suffered to provide for living expenses.

Add to this my basic skepticism about the reliability of academic investment research to accurately project the future now that everyone knows about factors and boatloads of money pour into factor approaches every month. The lion's share of positive factor backtesting research was done prior to widespread knowledge of factors by investors and also before hundreds of funds poured huge amounts of assets into factor index approaches. If the past was significantly different from the present and the future, are its backtesting results to be fully trusted? The case going forward may not be so bright as in the past for factors. I am positioned somewhere between belief in factors on one end and "Nobody knows nothing" on the other.

Garland Whizzer
Random Walker
Posts: 5561
Joined: Fri Feb 23, 2007 7:21 pm

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by Random Walker »

I’m guessing they would find the same result for TS Momentum
User avatar
in_reality
Posts: 4529
Joined: Fri Jul 12, 2013 6:13 am

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by in_reality »

garlandwhizzer wrote: Wed Dec 06, 2017 6:57 pm Perhaps 9 years isn't long enough for MOM to show its superiority as true factor adherents believe.
Funny. I purchased a MOM fund after reading suggestions here that MOM works well with value. Not having access to AQR I went with MTUM.

Over MTUM's life (Jan 2014 - Nov 2017), it's had a 16.20% CAGR vs. 10.20% for AMOMX (AQR large cap MOM).

Factor analysis shows they have about the same MOM exposure (0.23 vs 0.21), but that MTUM has large and significant alpha. AMOMX has large negative alpha. It's not significant in that time frame, but if you extend the time longer - it's negative and significant.

Actually, I just assumed AMOMX would be the better fund because AQR gets pumped here a lot, and that likely MTUM would be too simple, unsophisticated and probably cookie cutter since it's from iShares. I sold in Nov. because I really don't have time to evaluate the methodologies behind the funds.

Anyway, the S&P 500 had 11.77% CAGR in that time, so how can iShares do it but AQR not? I get that MTUM has a short life, but if it is a MOM fund, shouldn't I expect other MOM funds to do well too.

Obviously that is not the case, and it irritated me so much that factor investing is so seemingly dependent on the specific implementation that I am happy to have thrown in the towel on MOM after that nice run by MTUM. Maybe the AQR fund will show good results in another period when MTUM doesn't, but since their MOM loads are about the same and the difference is in alpha ... in the end, the measurement of factor investing is too irritating for me.

https://www.portfoliovisualizer.com/bac ... ion3_3=100

https://www.portfoliovisualizer.com/fac ... sion=false
DaufuskieNate
Posts: 605
Joined: Wed May 28, 2014 11:53 am

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by DaufuskieNate »

garlandwhizzer wrote: Wed Dec 06, 2017 6:57 pm Let's look at the performance of AQR's 2 MOM funds which presumably over a significant time period will reflect this outperformance. Let's compare their returns to funds which operate in the same market arena, growth, and in the same cap weight segment. That seems to be the appropriate comparison, small cap MOM versus small cap growth index, both active in the same market segment but one, growth, which pays no attention to the MOM factor, only to profit growth. Likewise, large cap MOM versus large cap growth index which again pays no attention to the MOM factor.
I would challenge the notion that the funds you are comparing operate in the same arena. Momentum and growth are two entirely different investing concepts. A four-factor analysis (market, size, value, momentum) of VUG shows that it has no exposure to Momentum, as you point out. What it does have are statistically significant negative exposures to size and value. You can think of it as an "anti-SCV" fund. AMOMX has exposure to Momentum and no significant exposures to size and value. Two entirely different animals here. Comparing two funds with entirely different factor exposures over such a short timeframe doesn't really tell you anything.
User avatar
in_reality
Posts: 4529
Joined: Fri Jul 12, 2013 6:13 am

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by in_reality »

DaufuskieNate wrote: Thu Dec 07, 2017 7:05 am
garlandwhizzer wrote: Wed Dec 06, 2017 6:57 pm Let's look at the performance of AQR's 2 MOM funds which presumably over a significant time period will reflect this outperformance. Let's compare their returns to funds which operate in the same market arena, growth, and in the same cap weight segment. That seems to be the appropriate comparison, small cap MOM versus small cap growth index, both active in the same market segment but one, growth, which pays no attention to the MOM factor, only to profit growth. Likewise, large cap MOM versus large cap growth index which again pays no attention to the MOM factor.
I would challenge the notion that the funds you are comparing operate in the same arena. Momentum and growth are two entirely different investing concepts. A four-factor analysis (market, size, value, momentum) of VUG shows that it has no exposure to Momentum, as you point out. What it does have are statistically significant negative exposures to size and value. You can think of it as an "anti-SCV" fund. AMOMX has exposure to Momentum and no significant exposures to size and value. Two entirely different animals here. Comparing two funds with entirely different factor exposures over such a short timeframe doesn't really tell you anything.
By the way, what explains the statistically significant -2.23% Annualized Alpha (α) of AMOMX (AQR Large Cap Momentum)?
Why does MTUM (iShares large cap momentum) have a statistically significant positive Alpha?

They have about the same MOM loadings and very different returns.

My take is that theoretical backtesting doesn't reflect the real life choices that are made for the methodology of an actual fund. It seems from the wide differences in two MOM funds with similar MOM loadings that it makes all the difference in the world.
betablocker
Posts: 522
Joined: Mon Jan 11, 2016 12:26 pm

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by betablocker »

Looks to me like MTUM focuses on companies that are a bit larger and a bit more growth oriented but if you compare the top holdings they are largely the same but have different weights. Is MTUM just more exposed to the part of the market that has grown the most? I don't think anyone here is saying MTUM is bad. Of course many of the other factors (value, small, etc.) will under perform while the market factor is outperforming. That's why they aren't totally correlated and why they get return.
DaufuskieNate
Posts: 605
Joined: Wed May 28, 2014 11:53 am

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by DaufuskieNate »

in_reality -

Very short timeframe for drawing conclusions. But since you asked, let's see what, if anything, we can learn.

AQR has their own formulation of the value factor, known as HmL Devil, which they consider to be superior. Book Value has a time lag as it is an accounting number. Market is basically current information. HmL Devil reduces the timing difference between the two measures.

If we use the AQR factor data with HmL Devil, the negative alpha is reduced to -1.85% and is just under the threshold of statistical significance. Extending the analysis to six factors (Market, Size, Value, Momentum, Quality, and Low Beta) there is a small positive alpha that is indistinguishable from zero. The six factor analysis shows a large and significant load of .33 on momentum. It also has small negative loads on Quality and Low Beta that are significant, but just over the threshold. The loads on size and value are basically zero which is a good thing.

Let's compare AMOMX to MTUM using HmL Devil and a six factor analysis. The timeframe should be changed to a start date of May 2013 to be consistent between the two funds and use all the available data on MTUM. First, MTUM has significant loads on Size (-.29), Value (-.26), and Momentum (.18). The alpha of 3.35% is not statistically significant. AMOMX has significant loads on Value (-.10), Momentum (.31), Quality (-.14) and Low Beta (-.17). The alpha of .61% is not statistically significant. MTUM's R^2 is .88 while AMOMX's is .96. Clearly, there is something else going on with MTUM, but AMOMX is well explained by the factor analysis.

Again this is a short timeframe for drawing any conclusions. However, if I were choosing between the two funds based on these numbers, I would go with AMOMX. How a fund fits into the overall portfolio is, for me, the most important thing. I have a SCV tilt. MTUM's negative size and value loads would not fit my portfolio very well. AMOMX has negative loads on Quality and Low Beta, which bears watching. However, the loads are smaller and not as strong as MTUM's exposures to Size and Value so perhaps they are an artifact of this particular short timeframe. The alphas of both funds are not significant in the six factor analysis and the AMOMX exposure to Momentum is much larger.
User avatar
Topic Author
matjen
Posts: 2189
Joined: Sun Nov 20, 2011 10:30 pm

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by matjen »

in_reality wrote: Thu Dec 07, 2017 7:21 am
By the way, what explains the statistically significant -2.23% Annualized Alpha (α) of AMOMX (AQR Large Cap Momentum)?
Why does MTUM (iShares large cap momentum) have a statistically significant positive Alpha?

They have about the same MOM loadings and very different returns.

My take is that theoretical backtesting doesn't reflect the real life choices that are made for the methodology of an actual fund. It seems from the wide differences in two MOM funds with similar MOM loadings that it makes all the difference in the world.
This does. Congratulations though. :mrgreen:
Trick Question: How is the Momentum Factor Performing YTD?
https://alphaarchitect.com/2017/07/24/m ... -and-size/
But one thing is clear: the bulk of the returns MTUM is capturing are probably unrelated to the academic momentum factor that has historically generated all the “buzz.”
However, the portfolio construction methodology suggests this fund is more of a mega-cap, low-beta-ish, quasi-momentum fund
A man is rich in proportion to the number of things he can afford to let alone.
garlandwhizzer
Posts: 3565
Joined: Fri Aug 06, 2010 3:42 pm

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by garlandwhizzer »

In_reality wrote:
My take is that theoretical backtesting doesn't reflect the real life choices that are made for the methodology of an actual fund. It seems from the wide differences in two MOM funds with similar MOM loadings that it makes all the difference in the world.
1+

This is the whole point. Factor analysis alone does not accurately determine future performance. It's that simple. The idea that you can prospectively select positive factor loads in either single factor or multi-factor funds and they will reliably and definitely outperform in the future is quite simply overstating the case. As I have pointed out before, Vanguard's Primecap Fund, looks terrible on factor analysis, significantly negative on summing up MOM, VAL, QUAL, and SIZE. On retrospect over its lifetime, dating back to 1984, 33 years, plenty of time for factor superiority to show up, the Primecap Fund has shown consistent and massive outperformance of both beta and of factor approaches including the darling of SCV funds, DFSVX, since DFSVX's inception in 1992. On factor analysis alone, Primecap looks like a massive long term loser on either single factor or multi-factor approaches. In reality, it massively and consistently outperforms for more than 3 decades. Such is the Grand Canyon gap between how a real fund did and what its factor analysis predicted.

Warrant Buffett and Charlie Munger have a pretty impressive long term investing record and they believe strongly in value and quality, but they do not believe that indexes capture these rare and difficult to define attributes. They do not do factor analysis as per AQR. Clearly factor index analysis is not the sole determinant of future results as AQR demonstrates from the massive negative alpha of its MOM funds and their underperformance relative to beta and MTUM. After all, AQR is supposed to be the really smart guys. If I were to invest in a MOM fund, which I do not do, I would select MTUM, which uses a less rigid factor approach--larger cap, lower turnover--reducing the high costs which are a major hurdle in employing MOM strategies.

Garland Whizzer
betablocker
Posts: 522
Joined: Mon Jan 11, 2016 12:26 pm

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by betablocker »

I don't think it tells us much to see that a large cap growth fund is outperforming small value at the top of a large cap growth cycle. The question is given a long cycle, what has better odds of outperforming. No one is claiming that factor loads guarantee anything. It's probability. MTUM has outperformed Primecap since inception. I don't take that as defnitive evidence that momentum or MTUM's version of that is superior.
User avatar
Ketawa
Posts: 2521
Joined: Mon Aug 22, 2011 1:11 am
Location: DC

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by Ketawa »

garlandwhizzer wrote: Thu Dec 07, 2017 12:17 pm 1+

This is the whole point. Factor analysis alone does not accurately determine future performance. It's that simple. The idea that you can prospectively select positive factor loads in either single factor or multi-factor funds and they will reliably and definitely outperform in the future is quite simply overstating the case. As I have pointed out before, Vanguard's Primecap Fund, looks terrible on factor analysis, significantly negative on summing up MOM, VAL, QUAL, and SIZE. On retrospect over its lifetime, dating back to 1984, 33 years, plenty of time for factor superiority to show up, the Primecap Fund has shown consistent and massive outperformance of both beta and of factor approaches including the darling of SCV funds, DFSVX, since DFSVX's inception in 1992. On factor analysis alone, Primecap looks like a massive long term loser on either single factor or multi-factor approaches. In reality, it massively and consistently outperforms for more than 3 decades. Such is the Grand Canyon gap between how a real fund did and what its factor analysis predicted.

Warrant Buffett and Charlie Munger have a pretty impressive long term investing record and they believe strongly in value and quality, but they do not believe that indexes capture these rare and difficult to define attributes. They do not do factor analysis as per AQR. Clearly factor index analysis is not the sole determinant of future results as AQR demonstrates from the massive negative alpha of its MOM funds and their underperformance relative to beta and MTUM. After all, AQR is supposed to be the really smart guys. If I were to invest in a MOM fund, which I do not do, I would select MTUM, which uses a less rigid factor approach--larger cap, lower turnover--reducing the high costs which are a major hurdle in employing MOM strategies.

Garland Whizzer
You are simply incorrect on Primecap's factor loadings. It is -0.1 to -0.2 in value loading, +0.1 to 0.2 in size loading, and basically indistinguishable from zero for momentum and quality, depending on the factor model you choose in Portfolio Visualizer. This basically comes out in a wash.

What it does have is a positive alpha with a t-stat of 2.0-3.0, depending on the particular factors you choose. That may be statistically significant and indicate some stock selection skill, it may not. It is economically significant because 2-3% of alpha per year above and beyond the performance explained by factor loadings adds up. With only 130 stocks in large or mid caps, I'm not sure I would be convinced that the managers of the fund are on to something that has not shown up in research, or they simply had a lucky outcome.

Re: AQR, how can you make such definitive statements in light of the performance of QSPIX, which is pure factor investing?
User avatar
in_reality
Posts: 4529
Joined: Fri Jul 12, 2013 6:13 am

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by in_reality »

DaufuskieNate wrote: Thu Dec 07, 2017 9:53 am
Let's compare AMOMX to MTUM using HmL Devil and a six factor analysis.
Ok, if those are your preferred measures, let's do that.
Ketawa wrote: Thu Dec 07, 2017 12:47 pm
garlandwhizzer wrote: Thu Dec 07, 2017 12:17 pm 1+

This is the whole point. Factor analysis alone does not accurately determine future performance. It's that simple.
You are simply incorrect on Primecap's factor loadings. It is -0.1 to -0.2 in value loading, +0.1 to 0.2 in size loading, and basically indistinguishable from zero for momentum and quality, depending on the factor model you choose in Portfolio Visualizer.
Re: AQR, how can you make such definitive statements in light of the performance of QSPIX, which is pure factor investing?
Using HmL Devil and extending the analysis to six factors (Market, Size, Value, Momentum, Quality, and Low Beta) as suggested as the best way to look at AMOMX, what do we see for QSPIX, MTUM and PRIMECAP?

QSPIX has higher loadings on all the juicy "drivers of return" (i.e. factors) except for size, yet over it's life has a 7.65% CAGR vs 15%-16% for the other funds which load on alpha. So if QSPIX is pure factor investing, what is explaining the returns?

OK, value didn't do well and MTUM and PRIMECAP with their negative value loadings benefited from that, but value should be accounted for in HmL-dev. What's the alpha -- random noise? Imprecision? A yet unexplained factor(s)?

https://www.portfoliovisualizer.com/fac ... sion=false

AMOMX vs MTUM and PRIMECAP shows a similar story.

What then do I conclude? MTUM and PRIMECAP don't seem explainable by factor analysis in this time period.

What factor(s) did well over the last 4 years that explains returns? It doesn't seem to be MOM. So growth stocks that don't exhibit MOM?

https://www.portfoliovisualizer.com/fac ... sion=false

Simply claiming that academic research shows that factor investing explains returns but that the stars must align at some future point in time for factor investing to explain real funds doesn't satisfy me. Sure, I am willing to accept that factors don't always pay a premium, but looking at AMOMX, MTUM, PRIMECAP and QSPIX, I can't really identify what factors have had good performance in the past 4 years.
Theoretical
Posts: 1546
Joined: Tue Aug 19, 2014 10:09 pm

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by Theoretical »

QSPIX has higher loadings on all the ju ... e returns?

You're missing the factor loading it does NOT have (much of), market factor/beta. QSPIX isn't set up to be an all-in-one Portfolio with a beta of 1. Its to have huge factor loadings with minimal beta or correllation to the stock market, also across multiple asset classes and not just stocks.

Also, the factors that have done well over the last ten years:

Market
Quality or (Profitability+Investment)
Low Beta/Betting Against Beta (almost to the same extent as market)
Term
Credit

Not so well
Size
Value
Momentum (US)
User avatar
triceratop
Posts: 5838
Joined: Tue Aug 04, 2015 8:20 pm
Location: la la land

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by triceratop »

Theoretical wrote: Thu Dec 07, 2017 11:17 pm QSPIX has higher loadings on all the ju ... e returns?

You're missing the factor loading it does NOT have (much of), market factor/beta. QSPIX isn't set up to be an all-in-one Portfolio with a beta of 1. Its to have huge factor loadings with minimal beta or correllation to the stock market, also across multiple asset classes and not just stocks.

Also, the factors that have done well over the last ten years:

Market
Quality or (Profitability+Investment)
Low Beta/Betting Against Beta (almost to the same extent as market)
Term
Credit

Not so well
Size
Value
Momentum (US)
Also, is it even true that factors are correlated across asset classes? Like, value in both stocks and commodities, or what not?
"To play the stock market is to play musical chairs under the chord progression of a bid-ask spread."
User avatar
Robert T
Posts: 2803
Joined: Tue Feb 27, 2007 8:40 pm
Location: 1, 0.2, 0.4, 0.5
Contact:

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by Robert T »

.
The linked paper in the OP compares the live returns of the AQR Momentum Funds to the AQR Momentum Indexes maintained by S&P.
  • Since inception on July 9, 2009 to December 2016

    AQR US Large Cap Momentum: Fund – Index = +0.9%
    AQR US Small Cap Momentum: Funds – Index = - 1.2%
    AQR International Momentum: Fund – Index = -1.1%
Compared to the indices, the live strategies utilize a number of more sophisticated portfolio construction decisions – many of which are meant to improve expected net returns. In particular, the live strategies utilize an evolved model that includes multiple measures of momentum and tilts towards higher momentum scoring names: weighting the top third of stocks based on a combination of their market capitalization and signal strength (in this case their momentum characteristic). The strategies rebalance more frequently than the indices (monthly rather than quarterly) to effectively capture the gross momentum premium. And, importantly, the live strategies include additional controls to manage liquidity, turnover, transaction costs, as well as other risk management concerns.
These construction decisions seem to have added value to the US Momentum Fund, less so for the US Small Cap Momentum and International Momentum Fund over this period.

In general, agree with the article – that implementation costs of momentum (so far) don’t appear to be so high. For example:
  • Since inception to end November 2016

    iShares MSCI USA Momentum [MTUM] – MSCI USA Momentum Index = -0.23 [Expense ratio = 0.15]
  • According to M* 3-yr Tax-cost ratio

    iShares MSCI USA Momentum [MTUM ]= 0.36
    Vanguard Total Stock Market [VTI] = 0.50
    iShares S&P500 [IVV] = 0.59
The biggest concern for momentum in past studies on costs of various factor tilted portfolio has been market impact (due to high turnover).
  • From linked paper: Estimated market impact (so far)

    AQR US Large Cap Momentum Fund = +0.11%
    AQR US Small Cap Momentum Fund = +0.31%
    AQR International Momentum Fund = +0.16%
I think these are simply hard to estimate, as large market impacts would probably affect the underlying index return on which comparisons are being made.

The paper is focused on implementation costs ...
While seven years of live experience is still too short to evaluate the efficacy of a style portfolio such as momentum (i.e., what is the true gross momentum premium?), it is sufficient to learn something about its implementability in practice.
Robert
.
betablocker
Posts: 522
Joined: Mon Jan 11, 2016 12:26 pm

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by betablocker »

in_reality wrote: Thu Dec 07, 2017 7:59 pm
DaufuskieNate wrote: Thu Dec 07, 2017 9:53 am
Let's compare AMOMX to MTUM using HmL Devil and a six factor analysis.
Ok, if those are your preferred measures, let's do that.
Ketawa wrote: Thu Dec 07, 2017 12:47 pm
garlandwhizzer wrote: Thu Dec 07, 2017 12:17 pm 1+

This is the whole point. Factor analysis alone does not accurately determine future performance. It's that simple.
You are simply incorrect on Primecap's factor loadings. It is -0.1 to -0.2 in value loading, +0.1 to 0.2 in size loading, and basically indistinguishable from zero for momentum and quality, depending on the factor model you choose in Portfolio Visualizer.
Re: AQR, how can you make such definitive statements in light of the performance of QSPIX, which is pure factor investing?
Using HmL Devil and extending the analysis to six factors (Market, Size, Value, Momentum, Quality, and Low Beta) as suggested as the best way to look at AMOMX, what do we see for QSPIX, MTUM and PRIMECAP?

QSPIX has higher loadings on all the juicy "drivers of return" (i.e. factors) except for size, yet over it's life has a 7.65% CAGR vs 15%-16% for the other funds which load on alpha. So if QSPIX is pure factor investing, what is explaining the returns?

OK, value didn't do well and MTUM and PRIMECAP with their negative value loadings benefited from that, but value should be accounted for in HmL-dev. What's the alpha -- random noise? Imprecision? A yet unexplained factor(s)?

https://www.portfoliovisualizer.com/fac ... sion=false

AMOMX vs MTUM and PRIMECAP shows a similar story.

What then do I conclude? MTUM and PRIMECAP don't seem explainable by factor analysis in this time period.

What factor(s) did well over the last 4 years that explains returns? It doesn't seem to be MOM. So growth stocks that don't exhibit MOM?

https://www.portfoliovisualizer.com/fac ... sion=false

Simply claiming that academic research shows that factor investing explains returns but that the stars must align at some future point in time for factor investing to explain real funds doesn't satisfy me. Sure, I am willing to accept that factors don't always pay a premium, but looking at AMOMX, MTUM, PRIMECAP and QSPIX, I can't really identify what factors have had good performance in the past 4 years.
QSPIX trades across multiple asset classes: currency, bonds, etc. with almost no exposure to equities. It shouldn't compete with equities. It's more bond like but even that isn't a great comparison. That makes it an alternative. In Reality, here's a question: if there are factors that are uncorrelated with market beta don't they have to underperform to some extent while market beta is on one of its biggest runs in history? This is recency bias.
garlandwhizzer
Posts: 3565
Joined: Fri Aug 06, 2010 3:42 pm

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by garlandwhizzer »

In response to my post about Primecap Fund, Ketawa wrote:

With only 130 stocks in large or mid caps, I'm not sure I would be convinced that the managers of the fund are on to something that has not shown up in research, or they simply had a lucky outcome
I agree. Primecap's success has been consistent for 33 years so you can rule out luck. Coins don't flip heads consistently for 33 years. The other potential explanation as Ketwawa wrote, "the managers of the fund are onto something that doesn't show up on research." Ketawa is not convinced of that, but I am. Academic research does not come from a place of perfect and complete knowledge. Markets are more complex and unpredictable than current academic models project IMO. The claim that 90% of future returns can be explained by factors is IMO overstating the case. My point is that factors cannot be relied upon to accurately and consistently project future returns. Primecap's very impressive long term outperformance in spite of its totally unimpressive factor loads clearly demonstrates this.

Garland Whizzer
User avatar
Ketawa
Posts: 2521
Joined: Mon Aug 22, 2011 1:11 am
Location: DC

Re: AQR/Asness - You Can Have Your Momentum Factor and Eat it Too

Post by Ketawa »

garlandwhizzer wrote: Fri Dec 08, 2017 11:52 am I agree. Primecap's success has been consistent for 33 years so you can rule out luck. Coins don't flip heads consistently for 33 years. The other potential explanation as Ketwawa wrote, "the managers of the fund are onto something that doesn't show up on research." Ketawa is not convinced of that, but I am. Academic research does not come from a place of perfect and complete knowledge. Markets are more complex and unpredictable than current academic models project IMO. The claim that 90% of future returns can be explained by factors is IMO overstating the case. My point is that factors cannot be relied upon to accurately and consistently project future returns. Primecap's very impressive long term outperformance in spite of its totally unimpressive factor loads clearly demonstrates this.

Garland Whizzer
You can indeed rule out luck for Primecap's performance, depending on your threshold for luck. Among the entire universe of active funds, some would be expected to outperform by luck, especially if you are selecting from the universe of funds that have survived for 33 years.

Primecap's alpha in a Fama-French four-factor regression has a p-value of 0.005, i.e. if you had a large number of funds with this measured alpha, you would estimate 1 in 200 to have it due to chance alone, while 199 of 200 might demonstrate actual skill >0, not necessarily at 3%/year. This is not anywhere near the probability of 33 consecutive heads (on the order of 10^-10), and saying the two is similar is completely disingenuous. It's the difference between me picking a random person on earth, and it happening to be you, Garland Whizzer, vs it being a random person living in California.

Among funds that have survived 33 years, the probability of a positive alpha is much higher due to survivorship bias.

And among funds that are being used to prove a point on the internet, the probability of a positive alpha is much, much higher.

Factor investors would agree with you that factors cannot be relied upon to accurately and consistently project future returns. There are no guarantees about future returns. By investing in a passive or index fund that has had a particular factor loading, factor investors hope that their investments will continue to have those factor loadings in the future. Nobody is making the claim that factors accurately and consistently project future returns. To say otherwise is a red herring.
Post Reply