What is wrong with AQR?

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
User avatar
305pelusa
Posts: 418
Joined: Fri Nov 16, 2018 10:20 pm

Re: What is wrong with AQR?

Post by 305pelusa » Sun Jun 16, 2019 6:28 am

typical.investor wrote:
Sat Jun 15, 2019 11:36 pm
larryswedroe wrote:
Sat Jun 15, 2019 11:11 pm
305 Glad it was helpful.
Adding SV tilts provides two ways to use it. First is adding risk and return (so higher return and higher vol if keep market beta the same), OR my preference (certainly in retirement) is to use the higher expected returns to lower beta exposure (the equities you own have higher expected returns thus don't need to own as much). That allows more bond holdings and gives you more of risk parity portfolio. And leverage not really needed.

But you are right some risk parity portfolios do use moderate amount of leverage to achieve objectives. As long as limit the amount of leverage to small dose and with low vol assets (not like EM bonds) you are fine.

Larry
If overweighting SCV provides better returns at the same risk OR the same returns at lower risk, why doesn't everyone do it? Why doesn't it become the market portfolio?

If an asset offers better returns for the same risk, it would seem like everyone should increase their exposure. Why don't people do that? I mean this can't be a secret to anyone, can it? I mean unless your head is in the sand or you are under a rock or living in a cave.
Huh? SCV doesn't provide better returns at the same risk, no one is saying that. It's more returns with an approximately commensurate increase in risk.

EDIT: Also to be fair to OP, I derailed this thread quite a bit with those questions to Larry. Maybe we should stop discussing factors/risk parity, etc and go back to talking about what's wrong (or what's rght? lol) with AQR.
Last edited by 305pelusa on Sun Jun 16, 2019 6:34 am, edited 1 time in total.

User avatar
coingaroo
Posts: 34
Joined: Fri Apr 26, 2019 11:31 am

Re: What is wrong with AQR?

Post by coingaroo » Sun Jun 16, 2019 6:34 am

Forester wrote:
Thu Jun 13, 2019 7:24 am
Correct me if I'm wrong but these futures markets are the same as betting exchanges which exist for sports and politics. It's an insulated system where all the profits & losses add up to zero, minus the trading costs.
There are market hedgers who pay the cost of hedging, which go to speculators (i.e. AQR).

User avatar
packer16
Posts: 1228
Joined: Sat Jan 04, 2014 2:28 pm

Re: What is wrong with AQR?

Post by packer16 » Sun Jun 16, 2019 8:00 am

I think there are two ideas in the SCV Larry Portfolio. First, is SCV which has its own "show up" risk associated with it which has been discussed extensively. The second is risk parity via more bonds. In the historical examples Larry cites you are getting a more return for a given amount of SD but this result is time dependent depending upon whether the SCV premium return "shows up" in your hold up. You do bring up an interesting question of is the market portfolio a risk parity one? I do not think this is the case the market stock/bond mix is determined by market size not whether it is a risk parity portfolio.

Packer
Buy cheap and something good might happen

larryswedroe
Posts: 15955
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: What is wrong with AQR?

Post by larryswedroe » Sun Jun 16, 2019 9:40 am

Packer, Simple explanation. A typical 60/40 portfolio with TSM as the equity doesn't have 60% of the risk, but closer to 90% because stocks so much more volatile. That's the issue. If you want each asset to contribute similar amount of risk to total portfolio you have to either hold a lot less equity or lever up the safer bonds. The higher EXPECTED returns to SV (which must be true just as it is for market beta, if you believe markets are efficient as they are clearly lot more volatile and expensive to trade and much more volatile earnings and so on) allow you to hold less market beta (so getting closer to risk parity) and adding other unique factors (hard to argue they are not unique when the correlation of value to market beta is close to zero over very long periods) gets you even closer.
Larry

User avatar
Ben Mathew
Posts: 751
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle
Contact:

Re: What is wrong with AQR?

Post by Ben Mathew » Sun Jun 16, 2019 10:36 am

larryswedroe wrote:
Sun Jun 16, 2019 9:40 am
The higher EXPECTED returns to SV (which must be true just as it is for market beta, if you believe markets are efficient as they are clearly lot more volatile and expensive to trade and much more volatile earnings and so on)
The idea that we can expect small value to have a higher expected return after adjusting for market beta has never made sense to me from a theoretical point of view. The fact that small value firms are more volatile, have a higher risk of bankruptcy in bad times, etc. are all characteristics that are captured by market beta. The puzzle is that small value seems to provide a return beyond what we can expect based on their return structure. We are rewarded for investing in small value over large growth with the exact same return structure. But why? Investors should care only about return structure. Why should small value provide an extra return beyond what their return structure justifies, when anyone can hold them?

(Note: I believe there is a small value premium and overweight them in my portfolio. But I think the reason for the premium is mispricing, not risk.)

larryswedroe
Posts: 15955
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: What is wrong with AQR?

Post by larryswedroe » Sun Jun 16, 2019 10:42 am

Ben
CAPM model simply doesn't do a good enough job of explaining return differences, only about 2/3, but 3 factor model explains above 90%. That is because the stocks have different risk characteristics than AVERAGE of market. My factor book provides the various findings demonstrating unique risks, like higher SD of earnings, more leverage, higher SD of divs, and other unique risks like more irreversible capital, which combining them makes them more riskier. Simple, and investors demand risk premium. That's the risk story. Now there are also behavioral explanations and limits to arb that prevent corrections. Read the book (:-))

User avatar
Ben Mathew
Posts: 751
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle
Contact:

Re: What is wrong with AQR?

Post by Ben Mathew » Sun Jun 16, 2019 10:52 am

larryswedroe wrote:
Sun Jun 16, 2019 10:42 am
CAPM model simply doesn't do a good enough job of explaining return differences, only about 2/3, but 3 factor model explains above 90%.
CAPM provides a solid theoretical explanation for why market beta should provide a premium. Factor models are purely an empirical relationship, with no justification for why a particular factor should work.
larryswedroe wrote:
Sun Jun 16, 2019 10:42 am
My factor book provides the various findings demonstrating unique risks, like higher SD of earnings, more leverage, higher SD of divs, and other unique risks like more irreversible capital, which combining them makes them more riskier. Simple, and investors demand risk premium. That's the risk story.
Higher standard deviation of earnings, higher standard deviation of dividends, irreversible capital -- these should all manifest themselves through the return structure. And to the extent that they don't manifest themselves through return structure, investors shouldn't care about them, no?
larryswedroe wrote:
Sun Jun 16, 2019 10:42 am
Now there are also behavioral explanations and limits to arb that prevent corrections.
These stories featuring inefficient markets make more sense, I think.
Last edited by Ben Mathew on Sun Jun 16, 2019 11:07 am, edited 1 time in total.

DaufuskieNate
Posts: 371
Joined: Wed May 28, 2014 11:53 am

Re: What is wrong with AQR?

Post by DaufuskieNate » Sun Jun 16, 2019 11:06 am

larryswedroe wrote:
Sun Jun 16, 2019 10:42 am
Ben
CAPM model simply doesn't do a good enough job of explaining return differences, only about 2/3, but 3 factor model explains above 90%. That is because the stocks have different risk characteristics than AVERAGE of market. My factor book provides the various findings demonstrating unique risks, like higher SD of earnings, more leverage, higher SD of divs, and other unique risks like more irreversible capital, which combining them makes them more riskier. Simple, and investors demand risk premium. That's the risk story. Now there are also behavioral explanations and limits to arb that prevent corrections. Read the book (:-))
There is an interesting parallel in the world of bond ratings. Size is a significant factor in bond ratings. Take two firms with similar quality balance sheets and profitability, one large and one small. The small company will invariably have a lower bond rating and will pay more for debt. The rating agencies have concluded that size is an independent risk factor in bonds because small companies have a higher risk of default.

User avatar
Ben Mathew
Posts: 751
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle
Contact:

Re: What is wrong with AQR?

Post by Ben Mathew » Sun Jun 16, 2019 11:17 am

DaufuskieNate wrote:
Sun Jun 16, 2019 11:06 am
larryswedroe wrote:
Sun Jun 16, 2019 10:42 am
Ben
CAPM model simply doesn't do a good enough job of explaining return differences, only about 2/3, but 3 factor model explains above 90%. That is because the stocks have different risk characteristics than AVERAGE of market. My factor book provides the various findings demonstrating unique risks, like higher SD of earnings, more leverage, higher SD of divs, and other unique risks like more irreversible capital, which combining them makes them more riskier. Simple, and investors demand risk premium. That's the risk story. Now there are also behavioral explanations and limits to arb that prevent corrections. Read the book (:-))
There is an interesting parallel in the world of bond ratings. Size is a significant factor in bond ratings. Take two firms with similar quality balance sheets and profitability, one large and one small. The small company will invariably have a lower bond rating and will pay more for debt. The rating agencies have concluded that size is an independent risk factor in bonds because small companies have a higher risk of default.
Interesting analogy. I can see how small firms are more likely to default than large firms, holding constant balance sheet quality and profitability. But I think the SV premium is more puzzling because these SV firms have been through many market crashes and we know how they perform in bad times. But we still insist on a premium over and above what that performance justifies. It's as if the ratings agencies are saying the risk of default by a small firm justifies a B rating, but we're giving it a C, just because it's small.

larryswedroe
Posts: 15955
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: What is wrong with AQR?

Post by larryswedroe » Sun Jun 16, 2019 11:43 am

Ben, my last comment. The CAPM does provide an explanation but not sufficient. If you read the literature, like in my factor book IMO it presents very clear and simple explanations for why these are unique risks that don't happen to show up in the market beta.
Fama I think said it best. When your theory (CAPM in this case) doesn't fit the data you throw out the theory, not the data.
You obviously can decide for yourself if there are logical risk based explanations for the uniqueness, and the data shows they clearly do not show up in market beta (or the correlations of value to market beta would not be close to zero, and this is true wherever we look). BTW, FF did not discover the two factors, they were in the literature a decade or more before, they just summarized it and put it into a 3 factor model.
Best wishes
Larry

DaufuskieNate
Posts: 371
Joined: Wed May 28, 2014 11:53 am

Re: What is wrong with AQR?

Post by DaufuskieNate » Sun Jun 16, 2019 11:44 am

Ben Mathew wrote:
Sun Jun 16, 2019 11:17 am
DaufuskieNate wrote:
Sun Jun 16, 2019 11:06 am
larryswedroe wrote:
Sun Jun 16, 2019 10:42 am
Ben
CAPM model simply doesn't do a good enough job of explaining return differences, only about 2/3, but 3 factor model explains above 90%. That is because the stocks have different risk characteristics than AVERAGE of market. My factor book provides the various findings demonstrating unique risks, like higher SD of earnings, more leverage, higher SD of divs, and other unique risks like more irreversible capital, which combining them makes them more riskier. Simple, and investors demand risk premium. That's the risk story. Now there are also behavioral explanations and limits to arb that prevent corrections. Read the book (:-))
There is an interesting parallel in the world of bond ratings. Size is a significant factor in bond ratings. Take two firms with similar quality balance sheets and profitability, one large and one small. The small company will invariably have a lower bond rating and will pay more for debt. The rating agencies have concluded that size is an independent risk factor in bonds because small companies have a higher risk of default.
Interesting analogy. I can see how small firms are more likely to default than large firms, holding constant balance sheet quality and profitability. But I think the SV premium is more puzzling because these SV firms have been through many market crashes and we know how they perform in bad times. But we still insist on a premium over and above what that performance justifies. It's as if the ratings agencies are saying the risk of default by a small firm justifies a B rating, but we're giving it a C, just because it's small.
Not puzzling at all. The equity is a security issued by the same firm as the bond. If the small company bond has a higher risk of default, the equity of the small company has a higher risk of being worthless. If the bond has a higher interest rate to compensate for this risk, it only makes sense that the equity holder requires a higher expected return to compensate for the risk.

typical.investor
Posts: 972
Joined: Mon Jun 11, 2018 3:17 am

Re: What is wrong with AQR?

Post by typical.investor » Sun Jun 16, 2019 11:53 am

DaufuskieNate wrote:
Sun Jun 16, 2019 11:44 am
Ben Mathew wrote:
Sun Jun 16, 2019 11:17 am
DaufuskieNate wrote:
Sun Jun 16, 2019 11:06 am
larryswedroe wrote:
Sun Jun 16, 2019 10:42 am
Ben
CAPM model simply doesn't do a good enough job of explaining return differences, only about 2/3, but 3 factor model explains above 90%. That is because the stocks have different risk characteristics than AVERAGE of market. My factor book provides the various findings demonstrating unique risks, like higher SD of earnings, more leverage, higher SD of divs, and other unique risks like more irreversible capital, which combining them makes them more riskier. Simple, and investors demand risk premium. That's the risk story. Now there are also behavioral explanations and limits to arb that prevent corrections. Read the book (:-))
There is an interesting parallel in the world of bond ratings. Size is a significant factor in bond ratings. Take two firms with similar quality balance sheets and profitability, one large and one small. The small company will invariably have a lower bond rating and will pay more for debt. The rating agencies have concluded that size is an independent risk factor in bonds because small companies have a higher risk of default.
Interesting analogy. I can see how small firms are more likely to default than large firms, holding constant balance sheet quality and profitability. But I think the SV premium is more puzzling because these SV firms have been through many market crashes and we know how they perform in bad times. But we still insist on a premium over and above what that performance justifies. It's as if the ratings agencies are saying the risk of default by a small firm justifies a B rating, but we're giving it a C, just because it's small.
Not puzzling at all. The equity is a security issued by the same firm as the bond. If the small company bond has a higher risk of default, the equity of the small company has a higher risk of being worthless. If the bond has a higher interest rate to compensate for this risk, it only makes sense that the equity holder requires a higher expected return to compensate for the risk.
So why does SCV diversify total market, when the bonds (Low credit rating right) don't? At least that's what Larry says - credit risk shows up at the wrong time.

Not saying low credit bonds/weaker companies don't pay a premium, just not sure about the diversification claim.

User avatar
nisiprius
Advisory Board
Posts: 38753
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: What is wrong with AQR?

Post by nisiprius » Sun Jun 16, 2019 12:00 pm

larryswedroe wrote:
Sun Jun 16, 2019 10:42 am
...CAPM model simply doesn't do a good enough job of explaining return differences, only about 2/3, but 3 factor model explains above 90%...
If the three-factor model explains above 90%, then how important can all the rest of them--momentum, quality, profitability, low-vol, investment--really be?
Last edited by nisiprius on Sun Jun 16, 2019 12:09 pm, edited 1 time 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.

User avatar
nedsaid
Posts: 12172
Joined: Fri Nov 23, 2012 12:33 pm

Re: What is wrong with AQR?

Post by nedsaid » Sun Jun 16, 2019 12:06 pm

nisiprius wrote:
Sun Jun 16, 2019 12:00 pm
larryswedroe wrote:
Sun Jun 16, 2019 10:42 am
...CAPM model simply doesn't do a good enough job of explaining return differences, only about 2/3, but 3 factor model explains above 90%...
If the three-factor model explains 90%, then how important can the rest of them--momentum, quality, profitability, low-vol, investment--really be?
Good question. The best answer we can give is that whenever a manager outperforms the market, we can now explain almost all of that performance with factors. Maybe it is a way of saying that individual stock picking is pretty much dead. So the DFA approach of setting screens and buying everything that passes the screen is modern portfolio management and the old Graham-Dodd way of evaluating stocks company by company is obsolete. That is the best answer this back bencher can give.
A fool and his money are good for business.

User avatar
Ben Mathew
Posts: 751
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle
Contact:

Re: What is wrong with AQR?

Post by Ben Mathew » Sun Jun 16, 2019 12:08 pm

DaufuskieNate wrote:
Sun Jun 16, 2019 11:44 am
Ben Mathew wrote:
Sun Jun 16, 2019 11:17 am
DaufuskieNate wrote:
Sun Jun 16, 2019 11:06 am
larryswedroe wrote:
Sun Jun 16, 2019 10:42 am
Ben
CAPM model simply doesn't do a good enough job of explaining return differences, only about 2/3, but 3 factor model explains above 90%. That is because the stocks have different risk characteristics than AVERAGE of market. My factor book provides the various findings demonstrating unique risks, like higher SD of earnings, more leverage, higher SD of divs, and other unique risks like more irreversible capital, which combining them makes them more riskier. Simple, and investors demand risk premium. That's the risk story. Now there are also behavioral explanations and limits to arb that prevent corrections. Read the book (:-))
There is an interesting parallel in the world of bond ratings. Size is a significant factor in bond ratings. Take two firms with similar quality balance sheets and profitability, one large and one small. The small company will invariably have a lower bond rating and will pay more for debt. The rating agencies have concluded that size is an independent risk factor in bonds because small companies have a higher risk of default.
Interesting analogy. I can see how small firms are more likely to default than large firms, holding constant balance sheet quality and profitability. But I think the SV premium is more puzzling because these SV firms have been through many market crashes and we know how they perform in bad times. But we still insist on a premium over and above what that performance justifies. It's as if the ratings agencies are saying the risk of default by a small firm justifies a B rating, but we're giving it a C, just because it's small.
Not puzzling at all. The equity is a security issued by the same firm as the bond. If the small company bond has a higher risk of default, the equity of the small company has a higher risk of being worthless. If the bond has a higher interest rate to compensate for this risk, it only makes sense that the equity holder requires a higher expected return to compensate for the risk.
All of that would be captured in the return structure, which in turn would be captured in the market beta. High default risk of SV firms is not a valid reason for the SV premium, because high default risk should be captured as market beta.

In other words, a high default risk of SV firms should show up as high beta for SV firms, and the premium for SV firms should simply be a premium for their high beta, not a separate premium simply for being SV.

User avatar
vineviz
Posts: 4681
Joined: Tue May 15, 2018 1:55 pm

Re: What is wrong with AQR?

Post by vineviz » Sun Jun 16, 2019 12:23 pm

Ben Mathew wrote:
Sun Jun 16, 2019 12:08 pm
All of that would be captured in the return structure, which in turn would be captured in the market beta. High default risk of SV firms is not a valid reason for the SV premium, because high default risk should be captured as market beta.

In other words, a high default risk of SV firms should show up as high beta for SV firms, and the premium for SV firms should simply be a premium for their high beta, not a separate premium simply for being SV.
This would only be true if default risk was perfectly correlated with market beta. This is obviously not the case.

The notion that risk can be accurately and completely represented as a single unidimensional continuum is a quaint artifact of the work by Sharpe and Markowitz, and not one that any financial economists take to be literally true.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch

DaufuskieNate
Posts: 371
Joined: Wed May 28, 2014 11:53 am

Re: What is wrong with AQR?

Post by DaufuskieNate » Sun Jun 16, 2019 12:44 pm

vineviz wrote:
Sun Jun 16, 2019 12:23 pm
Ben Mathew wrote:
Sun Jun 16, 2019 12:08 pm
All of that would be captured in the return structure, which in turn would be captured in the market beta. High default risk of SV firms is not a valid reason for the SV premium, because high default risk should be captured as market beta.

In other words, a high default risk of SV firms should show up as high beta for SV firms, and the premium for SV firms should simply be a premium for their high beta, not a separate premium simply for being SV.
This would only be true if default risk was perfectly correlated with market beta. This is obviously not the case.

The notion that risk can be accurately and completely represented as a single unidimensional continuum is a quaint artifact of the work by Sharpe and Markowitz, and not one that any financial economists take to be literally true.
Building off vineviz's comment, my intent was not to say that SmB is simply default risk. The rating agencies have decided that size is an independent risk factor for bonds. All the evidence would suggest that the stock market has decided the same thing. Yes, some of the risk of a small stock can be captured by market beta. Just not very well. Run a CAPM model on a SCV fund and the beta is typically higher than one. It also typically explains about 70% of the returns over time. Add SmB and HmL and the beta typically goes down and the explanatory power goes to something like 95%.

garlandwhizzer
Posts: 2387
Joined: Fri Aug 06, 2010 3:42 pm

Re: What is wrong with AQR?

Post by garlandwhizzer » Sun Jun 16, 2019 1:32 pm

I have a question for those who claim that 90% - 95% of market returns are explained by factors on a consistent basis. If that is true how do we account for the aggregate underperformance of multi-factor funds and ETFs since their inception in recent years. These funds are managed by brilliant managers fully versed in factor models and they also get to pick their optimal weightings of different factors and yet to date they have in aggregate underperformed dirt cheap beta. Factor enthusiasts argue that single factors like value can be negative for a long time but when one fact like value is negative, the others--MOM, QUAL, SIZE, and LOW VOL--make up the slack and should continue to outperform plain old beta. In theory you shouldn't have to wait a decade or two for outperformance with multi-facoto approaches. The opposite has happened with real multi-factor funds during their short history, many of which suffer substantial negative alphas. Alpha is not part of the factor zoo and it is often strongly negative in funds where factors are strongly positive. If factor funds were able to capture 90% - 95% or all returns, this would not occur unless we assume that beta is the king of factors and the others in aggregate play close to zero sum game. Primecap Fund from Vanguard, a LCG fund, has lousy factor loads now and throughout its long multi-decade lifetime but it has managed to hugely outperform DFSVX since the SCV fund's inception in the early 1990s. Analysis reveals this is due to Primecap's huge and consistently positive alpha which is entirely unaccounted for in factor models. If 90% - 95% of returns were defined by factor models this strong combination of size and value (SCV) wouldn't underperform its opposite (LCG) for 27 consecutive years. Alpha is the market's way of reminding factor theorists that they come from a place of imperfect and only partial knowledge. Often in real funds the more compellingly positive non-beta factor loads are, the greater the alpha is negative. As Cliff Asness, a brilliant factor enthusiast has written, beta is his favorite factor. Real factor fund results apart from beta have been disappointing for 15 years now while beta results have been robust. Things may change in the future, time will tell. IMO, however, caution is warranted when someone tells you they can prospectively define where 90% - 95% of future returns will come from and, more importantly, can select a portfolio beforehand to capture it.

Garland Whizzer

User avatar
Ben Mathew
Posts: 751
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle
Contact:

Re: What is wrong with AQR?

Post by Ben Mathew » Sun Jun 16, 2019 1:44 pm

vineviz wrote:
Sun Jun 16, 2019 12:23 pm
Ben Mathew wrote:
Sun Jun 16, 2019 12:08 pm
All of that would be captured in the return structure, which in turn would be captured in the market beta. High default risk of SV firms is not a valid reason for the SV premium, because high default risk should be captured as market beta.

In other words, a high default risk of SV firms should show up as high beta for SV firms, and the premium for SV firms should simply be a premium for their high beta, not a separate premium simply for being SV.
This would only be true if default risk was perfectly correlated with market beta. This is obviously not the case.
Not, it does not require that default risk be perfect correlated to market beta. It only requires that it be captured by market beta. Market beta will include other considerations as well--basically how procyclical your business is. So beta need not be perfectly correlated to any one of its determinants.

That aside, the important question here really is: for an investor, does the distribution of returns of an asset capture everything they care about the asset? If so, then the reason for the SV premium should be couched in those terms. i.e. It needs to be explained how SV firms affect the return distribution in a particular way, and why that is not captured by beta. The risk of defaulting during a market downturn is exactly the sort of thing that beta is about, so it's not clear why talking about the default risk of SV firms is supposed to explain the excess premium.
vineviz wrote:
Sun Jun 16, 2019 12:23 pm
The notion that risk can be accurately and completely represented as a single unidimensional continuum is a quaint artifact of the work by Sharpe and Markowitz, and not one that any financial economists take to be literally true.
You are too quick to dismiss the quaint works of some great financial economists. No model in economics is literally true. The question is how useful it is in understanding some aspects of the world. CAPM is useful. It shows how basic risk aversion implies that assets that tend to pay out during bad times (when money is dear) will be more expensive than assets that pay out mostly in good times (when money is cheap). Weakening the strong utility assumptions of CAPM leads to less sharp predictions, but still leaves the basic shape of this result intact. This result helps us understand why stocks yield more than bonds, and why call options on an equity index will yield more than put options. No, it doesn't capture all of the variation in the cross section of stock returns. But, it still explains a bunch of things. And it makes sense from a theoretical point of view--i.e. it is consistent with how we expect humans to act. And deviations from what we expect constitute puzzles. So it's a puzzle that SV firms seem to yield more. It's a puzzle that momentum exists. It's a puzzle that equity pays out so much more than bonds (we expect it to pay more, but not so much more). We can put these down to inefficient markets, or we can make up some other theory. But making up ad hoc risk based stories to explain every puzzle after the fact, doesn't really help us understand anything better, in my opinion.
Last edited by Ben Mathew on Sun Jun 16, 2019 3:53 pm, edited 1 time in total.

User avatar
Ben Mathew
Posts: 751
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle
Contact:

Re: What is wrong with AQR?

Post by Ben Mathew » Sun Jun 16, 2019 1:53 pm

DaufuskieNate wrote:
Sun Jun 16, 2019 12:44 pm
vineviz wrote:
Sun Jun 16, 2019 12:23 pm
Ben Mathew wrote:
Sun Jun 16, 2019 12:08 pm
All of that would be captured in the return structure, which in turn would be captured in the market beta. High default risk of SV firms is not a valid reason for the SV premium, because high default risk should be captured as market beta.

In other words, a high default risk of SV firms should show up as high beta for SV firms, and the premium for SV firms should simply be a premium for their high beta, not a separate premium simply for being SV.
This would only be true if default risk was perfectly correlated with market beta. This is obviously not the case.

The notion that risk can be accurately and completely represented as a single unidimensional continuum is a quaint artifact of the work by Sharpe and Markowitz, and not one that any financial economists take to be literally true.
Building off vineviz's comment, my intent was not to say that SmB is simply default risk. The rating agencies have decided that size is an independent risk factor for bonds. All the evidence would suggest that the stock market has decided the same thing. Yes, some of the risk of a small stock can be captured by market beta. Just not very well. Run a CAPM model on a SCV fund and the beta is typically higher than one. It also typically explains about 70% of the returns over time. Add SmB and HmL and the beta typically goes down and the explanatory power goes to something like 95%.
Empirically, yes, SmB seems to matter. But the question is, why has the stock market decided this? Is it mispricing or fair compensation for risk?

User avatar
305pelusa
Posts: 418
Joined: Fri Nov 16, 2018 10:20 pm

Re: What is wrong with AQR?

Post by 305pelusa » Sun Jun 16, 2019 2:07 pm

garlandwhizzer wrote:
Sun Jun 16, 2019 1:32 pm
I have a question for those who claim that 90% - 95% of market returns are explained by factors on a consistent basis. If that is true how do we account for the aggregate underperformance of multi-factor funds and ETFs since their inception in recent years. These funds are managed by brilliant managers fully versed in factor models and they also get to pick their optimal weightings of different factors and yet to date they have in aggregate underperformed dirt cheap beta. Factor enthusiasts argue that single factors like value can be negative for a long time but when one fact like value is negative, the others--MOM, QUAL, SIZE, and LOW VOL--make up the slack and should continue to outperform plain old beta. In theory you shouldn't have to wait a decade or two for outperformance with multi-facoto approaches. The opposite has happened with real multi-factor funds during their short history, many of which suffer substantial negative alphas. Alpha is not part of the factor zoo and it is often strongly negative in funds where factors are strongly positive. If factor funds were able to capture 90% - 95% or all returns, this would not occur unless we assume that beta is the king of factors and the others in aggregate play close to zero sum game. Primecap Fund from Vanguard, a LCG fund, has lousy factor loads now and throughout its long multi-decade lifetime but it has managed to hugely outperform DFSVX since the SCV fund's inception in the early 1990s. Analysis reveals this is due to Primecap's huge and consistently positive alpha which is entirely unaccounted for in factor models. If 90% - 95% of returns were defined by factor models this strong combination of size and value (SCV) wouldn't underperform its opposite (LCG) for 27 consecutive years. Alpha is the market's way of reminding factor theorists that they come from a place of imperfect and only partial knowledge. Often in real funds the more compellingly positive non-beta factor loads are, the greater the alpha is negative. As Cliff Asness, a brilliant factor enthusiast has written, beta is his favorite factor. Real factor fund results apart from beta have been disappointing for 15 years now while beta results have been robust. Things may change in the future, time will tell. IMO, however, caution is warranted when someone tells you they can prospectively define where 90% - 95% of future returns will come from and, more importantly, can select a portfolio beforehand to capture it.

Garland Whizzer
When you compare DFSVX to true "dirt cheap beta", like VFINX, the former has absolutely crushed it since inception.

If you compare it against an actively-managed fund though, you've introduced the potential for luck/skill (PrimeCap sold low and bought higher consistently perhaps, or was lucky on its concentrated stock picks, etc). You also introduce survirvorship bias. How about all of the actively managed funds that did beat "dirt cheap beta" like VFINX, but did not beat DFSVX? If I find twice as many of those as you find like PrimeCap, then aren't factors truly explaining a lot of the returns above beta?

Some things to think about.

DaufuskieNate
Posts: 371
Joined: Wed May 28, 2014 11:53 am

Re: What is wrong with AQR?

Post by DaufuskieNate » Sun Jun 16, 2019 2:11 pm

Ben Mathew wrote:
Sun Jun 16, 2019 1:53 pm
Empirically, yes, SmB seems to matter. But the question is, why has the stock market decided this? Is it mispricing or fair compensation for risk?
It's could be some of both. Small growth stocks with high investment and low profitability, the lottery stocks, may have returns driven down for behavioral reasons. Small value may be more of a risk story.

User avatar
packer16
Posts: 1228
Joined: Sat Jan 04, 2014 2:28 pm

Re: What is wrong with AQR?

Post by packer16 » Sun Jun 16, 2019 3:10 pm

305pelusa wrote:
Sun Jun 16, 2019 2:07 pm
garlandwhizzer wrote:
Sun Jun 16, 2019 1:32 pm
I have a question for those who claim that 90% - 95% of market returns are explained by factors on a consistent basis. If that is true how do we account for the aggregate underperformance of multi-factor funds and ETFs since their inception in recent years. These funds are managed by brilliant managers fully versed in factor models and they also get to pick their optimal weightings of different factors and yet to date they have in aggregate underperformed dirt cheap beta. Factor enthusiasts argue that single factors like value can be negative for a long time but when one fact like value is negative, the others--MOM, QUAL, SIZE, and LOW VOL--make up the slack and should continue to outperform plain old beta. In theory you shouldn't have to wait a decade or two for outperformance with multi-facoto approaches. The opposite has happened with real multi-factor funds during their short history, many of which suffer substantial negative alphas. Alpha is not part of the factor zoo and it is often strongly negative in funds where factors are strongly positive. If factor funds were able to capture 90% - 95% or all returns, this would not occur unless we assume that beta is the king of factors and the others in aggregate play close to zero sum game. Primecap Fund from Vanguard, a LCG fund, has lousy factor loads now and throughout its long multi-decade lifetime but it has managed to hugely outperform DFSVX since the SCV fund's inception in the early 1990s. Analysis reveals this is due to Primecap's huge and consistently positive alpha which is entirely unaccounted for in factor models. If 90% - 95% of returns were defined by factor models this strong combination of size and value (SCV) wouldn't underperform its opposite (LCG) for 27 consecutive years. Alpha is the market's way of reminding factor theorists that they come from a place of imperfect and only partial knowledge. Often in real funds the more compellingly positive non-beta factor loads are, the greater the alpha is negative. As Cliff Asness, a brilliant factor enthusiast has written, beta is his favorite factor. Real factor fund results apart from beta have been disappointing for 15 years now while beta results have been robust. Things may change in the future, time will tell. IMO, however, caution is warranted when someone tells you they can prospectively define where 90% - 95% of future returns will come from and, more importantly, can select a portfolio beforehand to capture it.

Garland Whizzer
When you compare DFSVX to true "dirt cheap beta", like VFINX, the former has absolutely crushed it since inception.

If you compare it against an actively-managed fund though, you've introduced the potential for luck/skill (PrimeCap sold low and bought higher consistently perhaps, or was lucky on its concentrated stock picks, etc). You also introduce survirvorship bias. How about all of the actively managed funds that did beat "dirt cheap beta" like VFINX, but did not beat DFSVX? If I find twice as many of those as you find like PrimeCap, then aren't factors truly explaining a lot of the returns above beta?

Some things to think about.
I think you also need to consider how much of factor fund performance is based upon luck also. It would be interesting to see how many of the factor funds outperform cheap beta funds vs. active funds. IMO there are alot of similarities between active and factor funds (esp. in that both think they have better weighting of stocks than investor determined market cap weighted funds).

Packer
Buy cheap and something good might happen

larryswedroe
Posts: 15955
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: What is wrong with AQR?

Post by larryswedroe » Sun Jun 16, 2019 5:14 pm

Nisiprius
How important are the rest, not very obviously. And you can gain exposure for example to quality/profitability by using value metrics like p/e and p/cf instead of p/b. And can at least eliminate negative MOM exposure in value funds by use of neg momentum screens and buy and hold ranges and that actually lowers costs of implementation and improves tax efficiency. So IMO you should focus on those and TERM and that is exactly what we do at my firm.
Larry

larryswedroe
Posts: 15955
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: What is wrong with AQR?

Post by larryswedroe » Sun Jun 16, 2019 5:53 pm

Ben, the problem I believe is that volatility (market beta) is only one measure of risk and not the only one people care about. Investors also care about skewness and kurtosis (and overpay for assets that have positive skew and excess kurtosis, partly explaining value premium) and also WHEN risks shows up, not just volatility. So assets that tend to do poorly when labor capital is risky will have higher costs of capital even if have same beta as investors demand a larger premium for asset that do poorly in bad times (which value does because of more irreversible capital for example). The CAPM simply isnt' complex enough. Things need to be as simple as possible, but not simpler (:-)) It's why the CAPM isn't used as the standard asset pricing model in finance--nice and elegant theory that just doesn't work as well as a more complex model. If the data doesn't confirm to your theory, throw out your theory
Hope that addresses your issue
Larry

User avatar
305pelusa
Posts: 418
Joined: Fri Nov 16, 2018 10:20 pm

Re: What is wrong with AQR?

Post by 305pelusa » Sun Jun 16, 2019 6:16 pm

packer16 wrote:
Sun Jun 16, 2019 3:10 pm
305pelusa wrote:
Sun Jun 16, 2019 2:07 pm
garlandwhizzer wrote:
Sun Jun 16, 2019 1:32 pm
I have a question for those who claim that 90% - 95% of market returns are explained by factors on a consistent basis. If that is true how do we account for the aggregate underperformance of multi-factor funds and ETFs since their inception in recent years. These funds are managed by brilliant managers fully versed in factor models and they also get to pick their optimal weightings of different factors and yet to date they have in aggregate underperformed dirt cheap beta. Factor enthusiasts argue that single factors like value can be negative for a long time but when one fact like value is negative, the others--MOM, QUAL, SIZE, and LOW VOL--make up the slack and should continue to outperform plain old beta. In theory you shouldn't have to wait a decade or two for outperformance with multi-facoto approaches. The opposite has happened with real multi-factor funds during their short history, many of which suffer substantial negative alphas. Alpha is not part of the factor zoo and it is often strongly negative in funds where factors are strongly positive. If factor funds were able to capture 90% - 95% or all returns, this would not occur unless we assume that beta is the king of factors and the others in aggregate play close to zero sum game. Primecap Fund from Vanguard, a LCG fund, has lousy factor loads now and throughout its long multi-decade lifetime but it has managed to hugely outperform DFSVX since the SCV fund's inception in the early 1990s. Analysis reveals this is due to Primecap's huge and consistently positive alpha which is entirely unaccounted for in factor models. If 90% - 95% of returns were defined by factor models this strong combination of size and value (SCV) wouldn't underperform its opposite (LCG) for 27 consecutive years. Alpha is the market's way of reminding factor theorists that they come from a place of imperfect and only partial knowledge. Often in real funds the more compellingly positive non-beta factor loads are, the greater the alpha is negative. As Cliff Asness, a brilliant factor enthusiast has written, beta is his favorite factor. Real factor fund results apart from beta have been disappointing for 15 years now while beta results have been robust. Things may change in the future, time will tell. IMO, however, caution is warranted when someone tells you they can prospectively define where 90% - 95% of future returns will come from and, more importantly, can select a portfolio beforehand to capture it.

Garland Whizzer
When you compare DFSVX to true "dirt cheap beta", like VFINX, the former has absolutely crushed it since inception.

If you compare it against an actively-managed fund though, you've introduced the potential for luck/skill (PrimeCap sold low and bought higher consistently perhaps, or was lucky on its concentrated stock picks, etc). You also introduce survirvorship bias. How about all of the actively managed funds that did beat "dirt cheap beta" like VFINX, but did not beat DFSVX? If I find twice as many of those as you find like PrimeCap, then aren't factors truly explaining a lot of the returns above beta?

Some things to think about.
I think you also need to consider how much of factor fund performance is based upon luck also. It would be interesting to see how many of the factor funds outperform cheap beta funds vs. active funds. IMO there are alot of similarities between active and factor funds (esp. in that both think they have better weighting of stocks than investor determined market cap weighted funds).

Packer
I hear what you say but it doesn't apply here specifically. DFSVX had a similar (in fact, slightly lower) Sharpe ratio than VFINX since inception. It crushed the market in returns at the cost of carrying higher risk.
If a fund is beating the market with lower, or similar risk (achieving a higher Sharpe ratio, like Prime Cap), then that's when I start worrying about luck and probability playing a role.

typical.investor
Posts: 972
Joined: Mon Jun 11, 2018 3:17 am

Re: What is wrong with AQR?

Post by typical.investor » Sun Jun 16, 2019 6:34 pm

garlandwhizzer wrote:
Sun Jun 16, 2019 1:32 pm
Factor enthusiasts argue that single factors like value can be negative for a long time but when one fact like value is negative, the others--MOM, QUAL, SIZE, and LOW VOL--make up the slack and should continue to outperform plain old beta
I believe it's been shown that cross-factor correlations are time-varying, with spikes in correlation around periods of factor underperformance.
garlandwhizzer wrote:
Sun Jun 16, 2019 1:32 pm
If 90% - 95% of returns were defined by factor models this strong combination of size and value (SCV) wouldn't underperform its opposite (LCG) for 27 consecutive years.
No -Why?

Saying factors explain 90%-95% of returns does not mean factor returns will be positive. All's it does is explain that during those 27 years you underperformed due to your factor exposure.
garlandwhizzer wrote:
Sun Jun 16, 2019 1:32 pm
Alpha is the market's way of reminding factor theorists that they come from a place of imperfect and only partial knowledge. Often in real funds the more compellingly positive non-beta factor loads are, the greater the alpha is negative. As Cliff Asness, a brilliant factor enthusiast has written, beta is his favorite factor.
Quite true and if you play around with portfolio visualize will see how much alpha changes depending on which combination of factors you select. Also, if factors were truly independent, why do loadings changes so much depending on what model is selected?
garlandwhizzer wrote:
Sun Jun 16, 2019 1:32 pm
Real factor fund results apart from beta have been disappointing for 15 years now while beta results have been robust. Things may change in the future, time will tell.
It's kind of expected to me. If factors weren't difficult to hold and paid a premium, why wouldn't investors generally load up on them?
garlandwhizzer wrote:
Sun Jun 16, 2019 1:32 pm
IMO, however, caution is warranted when someone tells you they can prospectively define where 90% - 95% of future returns will come from and, more importantly, can select a portfolio beforehand to capture it.
I do believe factors will ultimately pay a premium. My question is whether you need to have an institutional investing horizon to be guaranteed to realize them.

bgf
Posts: 969
Joined: Fri Nov 10, 2017 9:35 am

Re: What is wrong with AQR?

Post by bgf » Sun Jun 16, 2019 7:21 pm

vineviz wrote:
Sat Jun 15, 2019 12:03 pm
bgf wrote:
Sat Jun 15, 2019 11:05 am
vineviz wrote:
Sat Jun 15, 2019 10:49 am
bgf wrote:
Sat Jun 15, 2019 9:23 am
are we safely using the assumption of normality or not?
With respect to OLS specifically, normality is not assumed.

There are other places in financial economics where that assumption IS used to simplify theoretical proofs or empirical tests. But not in least squares regression.
modern financial theory makes no assumptions about the distribution of stock price changes?
That is literally the opposite of what I just said.
my ignorance abounds and is seriously limiting. for example, the abstract of this paper sounds really interesting, but will forever be beyond my comprehension.

i dont know if stock price changes can be described as fitting a power law distribution, but if they can, this paper argues least squares won't work. at least, thats what I THINK it says.

"Power-law distributions occur in many situations of scientific interest and have significant consequences for our understanding of natural and man-made phenomena. Unfortunately, the detection and characterization of power laws is complicated by the large fluctuations that occur in the tail of the distribution -- the part of the distribution representing large but rare events -- and by the difficulty of identifying the range over which power-law behavior holds. Commonly used methods for analyzing power-law data, such as least-squares fitting, can produce substantially inaccurate estimates of parameters for power-law distributions, and even in cases where such methods return accurate answers they are still unsatisfactory because they give no indication of whether the data obey a power law at all."

https://arxiv.org/abs/0706.1062
“TE OCCIDERE POSSUNT SED TE EDERE NON POSSUNT NEFAS EST"

User avatar
Ben Mathew
Posts: 751
Joined: Tue Mar 13, 2018 11:41 am
Location: Seattle
Contact:

Re: What is wrong with AQR?

Post by Ben Mathew » Sun Jun 16, 2019 7:54 pm

larryswedroe wrote:
Sun Jun 16, 2019 5:53 pm
Ben, the problem I believe is that volatility (market beta) is only one measure of risk and not the only one people care about. Investors also care about skewness and kurtosis (and overpay for assets that have positive skew and excess kurtosis, partly explaining value premium) and also WHEN risks shows up, not just volatility. So assets that tend to do poorly when labor capital is risky will have higher costs of capital even if have same beta as investors demand a larger premium for asset that do poorly in bad times (which value does because of more irreversible capital for example).
Thanks for putting it in terms of the return distribution--this is a much clearer statement. I think I understand what you meant to say, but I do want to clarify for those who may not know that market beta isn't a measure of volatility, but rather a measure of when the payouts occur--whether the asset pays out when the market does well (high beta) or when the market does badly (low beta). That said, it's possible that the value premium is capturing something about payout timing that isn't being captured by beta. Relaxing the assumptions of mean variance preferences would readily allow for this. The default general model in my mind is something like this (not the CAPM, which requires strong assumptions on preferences). But it's a little hard to see how the value premium would work into this. The labor risk story sounds like value would have to do badly when people lose their jobs but while the market is still doing well. That's one scenario where value could add information over and above beta. And investors will need to know this and price these assets at a discount. I'm skeptical, but open to the possibility. if your book has a summary of the evidence on this, I will try to read it. I'm assuming the book you're referring to is Your Complete Guide to Factor Based Investing?

The alternate hypothesis here is market inefficiency and mispricing. We know that the vast majority of people make terrible financial decisions. They live paycheck to paycheck, save too little, take out payday loans and credit card debt at insanely high interest, pay unforgivably high fees for investment managers to trail the market, chase recent performance, and so on. Most people are not simply not very good at the game of saving and investing. So the possibility that the prices of risky assets are a bit out of whack isn't that much of a stretch to begin with (since there are limits to arbitrage with risky assets). We also know that bubbles have been around for a long time--history gives us dramatic examples stretching back centuries. So it's not unlikely that there would be smaller scale bubbles simmering in our stock markets on an ongoing basis. How would these small scale bubbles manifest themselves in asset prices? Popular assets inside the bubble will have high valuations and low expected return. These would be the growth firms. Unpopular assets outside the bubble will have low valuations and high expected return. These would be the value firms.

There's also a simple story for small vs big: people like to stick with what's familiar. This could be what's behind home bias as well. One explanation for two phenomena--that's something.

It seems to me that these sorts of of inefficiency stories are less complicated and more compelling than the possibility that small, value, quality and momentum are all capturing higher order aspects of the return distribution that are not being captured by beta.
larryswedroe wrote:
Sun Jun 16, 2019 5:53 pm
Ben, The CAPM simply isnt' complex enough. Things need to be as simple as possible, but not simpler (:-)) It's why the CAPM isn't used as the standard asset pricing model in finance--nice and elegant theory that just doesn't work as well as a more complex model. If the data doesn't confirm to your theory, throw out your theory
I'm not saying that we stick with the theory and throw out the data. CAPM is partially right, which is pretty darn good for a model in social science (which, in case anyone forgets, is what finance is). It doesn't get everything exactly right, and we have to address its deficiencies. The question is whether to modify the basic CAPM story by bringing in inefficient markets, or to modify it with other rational risk factors.

One thing to note here is that under the view that the excess returns for factors are fair compensation for risk, people on the right side of the equation--the large growth investors--are also making rational choices in line with their risk preferences. This raises the question--why are factorheads usually on the left side of the equation and not on the right? Most people interested in factors seem to be tilting towards small value and not towards large growth. Shouldn't risk preferences be more evenly distributed among factorheads? To me it looks more like a group of people crowding around the free lunch that is small value, than like a group of people locating themselves evenly along a risk return frontier between small value and large growth. Isn't this odd if factors are about risk and not about mispriced assets?

Post Reply