Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Post Reply
scottj19707
Posts: 219
Joined: Fri Dec 05, 2014 1:25 pm

Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by scottj19707 » Mon Jun 13, 2016 9:35 pm

Q: Do Dimensional Fund Advisors (DFA) funds generate risk-adjusted "alpha?"

[Even further... Do DFA funds generate statistically significant, risk-adjusted "alpha?"]

This question occurred to me when I was recently reading:

"Has Vanguard Added Value as an Active Manager?" (AdvisorPerspectives.com/Larry Swedroe, January 12, 2016): http://www.advisorperspectives.com/arti ... ve-manager

The last table (on p. 8) in Larry's article (above) does not include a row for DFA... and I thought to myself, "why not?"

To answer this question properly, it seems like you should do a multi-factor multiple-regression analysis (e.g., using PortfolioVisualizer.com). The alpha term would represent in-sample in/out-performance on a risk-adjusted basis. Then you'd calculate a t-statistic for the alpha term to determine statistical significance.

[Related question: Is the answer to the main question above (esp. "risk-adjusted") somehow contingent on whether you believe (or assume) a particular factor reflects "risk compensation" vs. a "behavioral mis-pricing?"]

My hypothesis: DFA funds provide only targeted exposure to chosen factors (and the related factor premiums); they do not generate statistically significant, risk-adjusted alpha.

[Admission: I currently invest mainly using VTI/VXUS/BND and do not have a handy list of all DFA funds (and ETFs)... so I've not yet run the numbers myself.]

Any thoughts, comments, opinions, especially the BH.org Factor Posse? Am I generally in the ball park? Insightful, of just plainly obvious?

Thanks and best wishes to all,


scottj19707
Last edited by scottj19707 on Thu Jun 16, 2016 7:49 pm, edited 1 time in total.
“To acquire knowledge, one must study; but to acquire wisdom, one must observe.”-Marilyn vos Savant; “If you can’t explain it simply, you don’t understand it well enough.”-Albert Einstein; VTI/VXUS/BND

Theoretical
Posts: 1076
Joined: Tue Aug 19, 2014 10:09 pm

Re: Factor-Based Investing (F-BI): Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by Theoretical » Mon Jun 13, 2016 11:05 pm

I think the whole point of DFA funds is that they don't have alpha and that they are extremely factor pure. The funds seem to meet this qualification.

User avatar
nedsaid
Posts: 8722
Joined: Fri Nov 23, 2012 12:33 pm

Re: Factor-Based Investing (F-BI): Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by nedsaid » Mon Jun 13, 2016 11:21 pm

Theoretical wrote:I think the whole point of DFA funds is that they don't have alpha and that they are extremely factor pure. The funds seem to meet this qualification.


That is my understanding as well. Alpha has shrunk because academics have identified the causes of a higher and higher percentage of market return. There used to be one factor and that was stock market exposure. Then came size and value. Then momentum and quality. There is also illiquidity. There are probably others that don't come to mind. The known factors now account for 96% of the market returns. There isn't much unknown out there.

So you can increase return by tilting towards certain factors without increasing alpha. Let's say DFA tilts small/value and screens to get momentum at least to neutral. You have four factors now rather than just one and alpha isn't just excess returns over the market but excess returns with market, size, value, and momentum taken into account. Tilting would increase returns over just market exposure but the excess returns could all be accounted for by the factors. DFA Funds should be zero alpha as the fund managers do not pick stocks. Perhaps very patient trading strategies could generate a tiny bit of alpha.
A fool and his money are good for business.

lack_ey
Posts: 5719
Joined: Wed Nov 19, 2014 11:55 pm

Re: Factor-Based Investing (F-BI): Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by lack_ey » Mon Jun 13, 2016 11:40 pm

I think there are some misconceptions upthread. Hopefully I'm not introducing more of them here. I don't own DFA funds and haven't researched them as much as others have.

DFA doesn't religiously adhere to indexes or pure factor exposure (not pure 3-factor exposure). In fact, they do attempt to gain alpha from patient trading. That is, they're not afraid of some tracking error relative to internal benchmarks and ideas of what their funds "should" hold, preferring to transact when they can get a good price. Secondly, their funds include screens based on research to weed out certain stocks that some academics to have found have poor risk-adjusted returns. They also do things like exclude REITs from tilted funds because they expect you to use their separate REIT fund for that exposure.

Now, the question was asked about alpha. No fund that holds a very wide and relatively representative slice of the market is going to demonstrate much alpha, positive or negative, unless its expenses are really high and it has negative alpha, or there were some real oddities in the data and it happened to own the right or wrong stuff. So it's not going to be statistically significant. DFA's tweaks aren't going to somehow overcome the ERs and then a lot more. These are fairly passive funds. I don't think anybody comes to DFA to try to get alpha; though the hope is that active managers generate alpha, that's not what people are gunning for here or in regular index funds.

Here's a look via portfoliovisualizer at four of their funds for about the last 20 years:
https://www.portfoliovisualizer.com/fac ... ssetType=1

larryswedroe
Posts: 15598
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Factor-Based Investing (F-BI): Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by larryswedroe » Tue Jun 14, 2016 2:07 am

lackey
got it right
patient trading and screens help but don't expect alpha

scottj19707
Posts: 219
Joined: Fri Dec 05, 2014 1:25 pm

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by scottj19707 » Thu Jun 16, 2016 7:54 pm

Hi all,

Post by larryswedroe » Tue Jun 14, 2016 12:07 am


by lack_ey » Mon Jun 13, 2016 9:40 pm


by nedsaid » Mon Jun 13, 2016 9:21 pm


by Theoretical » Mon Jun 13, 2016 9:05 pm


Thanks for your comments, and confirmations re: the factor-driven B/beta (not alpha-seeking) focus of DFA funds. I appreciate your responses.


scottj19707
“To acquire knowledge, one must study; but to acquire wisdom, one must observe.”-Marilyn vos Savant; “If you can’t explain it simply, you don’t understand it well enough.”-Albert Einstein; VTI/VXUS/BND

Theoretical
Posts: 1076
Joined: Tue Aug 19, 2014 10:09 pm

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by Theoretical » Thu Jun 16, 2016 8:16 pm

I'd look at it as Alpha vs alpha.

Big Alpha is stock picking or market timing in the macro sense

Small alpha is more along the lines of avoiding either dumb trading decisions (buying/selling in the wrong direction on a high volatility day) or for avoiding grossly overvalued "lottery tickets" as Larry calls them.

It seems like Big Alpha is about making gains that exceed the particular area the fund is investing in whole Small Alpha is more oriented towards avoiding bad decisions in implemtation of a fixed strategy.

FWIW, I'd consider Vanguard's success in zeroing out the expenses of the admiral shares of the S&P 500 and Total Stock Market funds to be small alpha

User avatar
nisiprius
Advisory Board
Posts: 34171
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: Factor-Based Investing (F-BI): Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by nisiprius » Thu Jun 16, 2016 10:55 pm

larryswedroe wrote:lackey
got it right
patient trading and screens help but don't expect alpha
If patient trading results in higher risk-adjusted return than closely tracking an index, how is that not "alpha?"
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.

lack_ey
Posts: 5719
Joined: Wed Nov 19, 2014 11:55 pm

Re: Factor-Based Investing (F-BI): Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by lack_ey » Thu Jun 16, 2016 11:10 pm

nisiprius wrote:
larryswedroe wrote:lackey
got it right
patient trading and screens help but don't expect alpha
If patient trading results in higher risk-adjusted return than closely tracking an index, how is that not "alpha?"

I don't speak for Larry but I think the point is that what they do on the margins essentially attempts to improve 3-factor alpha over not doing it. Here "improve" means something like "make less negative" because the expense ratios are nonzero.

So expect the stuff they do to help but do not expect that to make overall fund alpha positive—certainly not statistically significantly positive.

jalbert
Posts: 2220
Joined: Fri Apr 10, 2015 12:29 am

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by jalbert » Fri Jun 17, 2016 12:52 am

To answer this question properly, it seems like you should do a multi-factor multiple-regression analysis (e.g., using PortfolioVisualizer.com). The alpha term would represent in-sample in/out-performance on a risk-adjusted basis. Then you'd calculate a t-statistic for the alpha term to determine statistical significance.


Such a computed t-statistic would be an invalid measure of statistical significance because it is not generated from a random sample of independent outcomes. It is also required that the tests to be used be decided before the analysis begins.

These are not small issues despite their widespread dismissal. There is a general bias that if you can feed some data to statistical software, and display the result on a computer screen, it must be meaningful.

Ever notice how portfoliovisualizer results vary widely with the time period used? That's because time consecutive finance data is highly biased. Such results are not generalizable or reproducible.

I'm not trying to pick on the OP. This is a pervasive problem in finance.
Last edited by jalbert on Fri Jun 17, 2016 1:46 pm, edited 2 times in total.

User avatar
nisiprius
Advisory Board
Posts: 34171
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: Factor-Based Investing (F-BI): Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by nisiprius » Fri Jun 17, 2016 8:04 am

lack_ey wrote:
nisiprius wrote:
larryswedroe wrote:lackey
got it right
patient trading and screens help but don't expect alpha
If patient trading results in higher risk-adjusted return than closely tracking an index, how is that not "alpha?"

I don't speak for Larry but I think the point is that what they do on the margins essentially attempts to improve 3-factor alpha over not doing it. Here "improve" means something like "make less negative" because the expense ratios are nonzero.

So expect the stuff they do to help but do not expect that to make overall fund alpha positive—certainly not statistically significantly positive.
It won't wash. Do you agree that Dimensional--or at least Dimensional advisors, make an explicit claim that patient trading is better than tracking an index? "Slavishly" tracking would be the rhetoric. And they do not track indexes and that's not a small technicality. They don't call their funds index funds, they don't mention tracking an index as a fund goal, and any plot will show that they do not, in fact, adhere closely to the indexes.

Well, it's one of two things. Either patient trading improves return but does not improve risk-adjusted return,, in which case what good is it? It just yet another meaningless way to distinguish Dimensional from the Vanguard competition.

Or, patient trading does improve risk-adjusted return, compared to the benchmark index, in which case that's alpha.

It is of course quite possible that the amount is negligible and isn't passed along to investors (i.e. is eaten up by Dimensional's very-low-but-higher-than-index-fund fees), and that apart from introducing tracking error it is no better than whatever transactional techniques Vanguard uses to overcome costs.

It's really up to the Dimensional advocates to try to make some kind of quantitative estimate. Does "patient trading" improves risk-adjusted return, and, if so, by how much? How many basis points, please?
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.

lack_ey
Posts: 5719
Joined: Wed Nov 19, 2014 11:55 pm

Re: Factor-Based Investing (F-BI): Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by lack_ey » Fri Jun 17, 2016 10:06 am

nisiprius wrote:
lack_ey wrote:
nisiprius wrote:
larryswedroe wrote:lackey
got it right
patient trading and screens help but don't expect alpha
If patient trading results in higher risk-adjusted return than closely tracking an index, how is that not "alpha?"

I don't speak for Larry but I think the point is that what they do on the margins essentially attempts to improve 3-factor alpha over not doing it. Here "improve" means something like "make less negative" because the expense ratios are nonzero.

So expect the stuff they do to help but do not expect that to make overall fund alpha positive—certainly not statistically significantly positive.
It won't wash. Do you agree that Dimensional--or at least Dimensional advisors, make an explicit claim that patient trading is better than tracking an index? "Slavishly" tracking would be the rhetoric. And they do not track indexes and that's not a small technicality. They don't call their funds index funds, they don't mention tracking an index as a fund goal, and any plot will show that they do not, in fact, adhere closely to the indexes.

Well, it's one of two things. Either patient trading improves return but does not improve risk-adjusted return,, in which case what good is it? It just yet another meaningless way to distinguish Dimensional from the Vanguard competition.

Or, patient trading does improve risk-adjusted return, compared to the benchmark index, in which case that's alpha.

It is of course quite possible that the amount is negligible and isn't passed along to investors (i.e. is eaten up by Dimensional's very-low-but-higher-than-index-fund fees), and that apart from introducing tracking error it is no better than whatever transactional techniques Vanguard uses to overcome costs.

It's really up to the Dimensional advocates to try to make some kind of quantitative estimate. Does "patient trading" improves risk-adjusted return, and, if so, by how much? How many basis points, please?

By mitigating market impact, providing liquidity rather than taking it, not getting front run, you should improve return and risk-adjusted return relative to not doing that. The drawback is not having the AA you want quite when you want it, so tracking error relative to your preferred allocation.

It should be effectively alpha. The actual amount depends on how much trading there is, market conditions, the liquidity and spreads of what's being traded, and so on. But some of the improvement is just relative to most index funds, which willingly eat this typically small drag in order to track the index better. So it's more like they're taking less negative alpha from trading compared to the typical index fund and much less compared to the typical active fund. How many basis points? Probably not much for most asset classes. I'm sure DFA knows and keeps track of this; I'm not an adviser, don't own DFA funds, and don't have the figures.

DFA can get away with more of it compared to most index funds because they're not being compared to some S&P or Russell or whatever index, unlike actual index funds. They can take more tracking error and nobody will notice and complain.

It's just one thing that helps mitigate DFA's higher fees to some degree. Like I said, it's not going to have a huge impact. Just one of the things to consider along with securities lending (which index funds do too), inclusion screens, and so on. I mentioned it to begin with because the original claim being made here was that DFA offers pure factor exposure, which is not what's happening. They throw in these small extras. They're not the draw for DFA funds, though. The people that use them are there for the asset class coverage and higher tilts than some alternatives. That's what does the heavy lifting.

We all know here that typical index fund performance lags the index by a small amount, usually around the ER and sometimes less. This is because securities lending frequently makes up for transaction costs and very small cash drag. (But note that this is just performance relative to an index; an index itself could have negative alpha if others are front running it. Smarter indexes these days mitigate these kinds of things.)

User avatar
nedsaid
Posts: 8722
Joined: Fri Nov 23, 2012 12:33 pm

Re: Factor-Based Investing (F-BI): Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by nedsaid » Fri Jun 17, 2016 10:38 am

nisiprius wrote:
lack_ey wrote:
nisiprius wrote:
larryswedroe wrote:lackey
got it right
patient trading and screens help but don't expect alpha
If patient trading results in higher risk-adjusted return than closely tracking an index, how is that not "alpha?"

I don't speak for Larry but I think the point is that what they do on the margins essentially attempts to improve 3-factor alpha over not doing it. Here "improve" means something like "make less negative" because the expense ratios are nonzero.

So expect the stuff they do to help but do not expect that to make overall fund alpha positive—certainly not statistically significantly positive.
It won't wash. Do you agree that Dimensional--or at least Dimensional advisors, make an explicit claim that patient trading is better than tracking an index? "Slavishly" tracking would be the rhetoric. And they do not track indexes and that's not a small technicality. They don't call their funds index funds, they don't mention tracking an index as a fund goal, and any plot will show that they do not, in fact, adhere closely to the indexes.

Well, it's one of two things. Either patient trading improves return but does not improve risk-adjusted return,, in which case what good is it? It just yet another meaningless way to distinguish Dimensional from the Vanguard competition.

Or, patient trading does improve risk-adjusted return, compared to the benchmark index, in which case that's alpha.

It is of course quite possible that the amount is negligible and isn't passed along to investors (i.e. is eaten up by Dimensional's very-low-but-higher-than-index-fund fees), and that apart from introducing tracking error it is no better than whatever transactional techniques Vanguard uses to overcome costs.

It's really up to the Dimensional advocates to try to make some kind of quantitative estimate. Does "patient trading" improves risk-adjusted return, and, if so, by how much? How many basis points, please?


Nisiprius, my very foggy memory banks recall Gus Sauter making a comment that front-running of the indexes by traders costs 25 basis points a year. Of course, with "Total" indexes like US Total Stock Market or Total International Stock Market you don't have this problem. But even the good old S&P 500 suffers from this issue.

I bought a Micro-Cap Index ETF on the recommendation of the Paul Merriman organization. Morningstar ran an article on this very fund saying that an investor would be better off in a plain old, boring small cap index. One big reason was the front running.

I think this is part of what people refer to when they talk about "dumb" indexes. So if the S&P 500 suffers with a loss of maybe 25 basis points to front running, just imagine what my Micro-Cap Index ETF gets penalized. Micro-Caps are a lot less liquid and I imagine this is a much bigger problem.
A fool and his money are good for business.

lack_ey
Posts: 5719
Joined: Wed Nov 19, 2014 11:55 pm

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by lack_ey » Fri Jun 17, 2016 11:02 am

I doubt it's 25 bp for the S&P 500. The turnover is way too low and the underlying securities too liquid for that. DFA U.S. Large Company Portfolio (DFUSX), which tracks the S&P 500 but explicitly in its prospectus attempts to mitigate the impact of index changes
The U.S. Large Company Portfolio generally invests in the stocks that comprise the S&P 500® Index in approximately the proportions they are represented in the S&P 500® Index. The S&P 500® Index comprises a broad and diverse group of stocks. Generally, these are the U.S. stocks with the largest market capitalizations and, as a group, they represent approximately 75% of the total market capitalization of all publicly traded U.S. stocks. For the U.S. Large Company Portfolio, Dimensional Fund Advisors LP (the “Advisor”) considers the stocks that comprise the S&P 500® Index to be those of large companies. Under normal market conditions, at least 95% of the U.S. Large Company Portfolio’s net assets will be invested in the stocks that comprise the S&P 500® Index. As a non-fundamental policy, under normal circumstances, the U.S. Large Company Portfolio will invest at least 80% of its net assets in securities of large U.S. companies.

Ordinarily, portfolio companies will not be sold except to reflect additions or deletions of the companies that comprise the S&P 500® Index, including as a result of mergers, reorganizations and similar transactions and, to the extent necessary, to provide cash to pay redemptions of the U.S. Large Company Portfolio’s shares. Given the impact on prices of securities affected by the reconstitution of the S&P 500® Index around the time of a reconstitution date, the U.S. Large Company Portfolio may purchase or sell securities that may be impacted by the reconstitution before or after the reconstitution date of the S&P 500® Index. In seeking to approximate the total investment return of the S&P 500® Index, the Advisor may also adjust the representation of securities in the U.S. Large Company Portfolio after considering such securities’ characteristics and other factors the Advisor determines to be appropriate.

doesn't have an edge like that over the S&P 500 itself (DFUSX is DFA's fund with an ER of 0.08% now, while VFIAX is Vanguard's S&P 500 index fund, admiral shares with ER of 0.05%) as evidenced by the performance:
Image

I don't know how much of the above can be attributed to differences in securities lending. Hard to say what the impact is, but it's not consistent with the level of 25 bp.

Also consider the performance relative to other large/mid cap US stock indexes.

larryswedroe
Posts: 15598
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by larryswedroe » Fri Jun 17, 2016 11:16 am

Nisiprius
Lackey has it EXACTLY right. yes patient trading is alpha, but as I said I would not expect the fund to generate alpha because of the fund expenses. DFA and others also benefit from exclusions of assets with poor historical records. In some cases the DFA funds have been able to generate enough alpha to cover expenses, but I would not expect it to occur. It might but should not be expected.

If you want the perfect example of NEGATIVE alpha just look at the historical returns of the R2k vs the very similar CRSP 6-10. Huge difference in returns. Used to be about 2%, but Russell made some changes that helped address the problem and now I think it's down to about 1.5% over the life. Here one could say simply choosing the right index was "alpha" (at least in terms of returns relative to another index). That negative is what caused Vanguard's Sauter to get the board to drop the R2k index as the benchmark. The MSCI indexes in effect "stole" (copied) DFA's buy and hold range strategy to help reduce front running problems.

So bottom line is DFA's strategies add alpha over pure indexing (forced trading and NO exclusions) but should not be counted on to cover all fund expenses.

Hope that clarifies
Larry

FinancialDave
Posts: 1515
Joined: Thu May 26, 2011 9:36 pm

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by FinancialDave » Fri Jun 17, 2016 11:51 am

Here is what Morningstar reports on buy and hold DFA portfolio (no rebalancing) I built slightly over 5 years ago:

http://socialize.morningstar.com/NewSoc ... 48781B645F

It does not really look all that pretty for DFA compared to its category or index.

dave
I love simulated data. It turns the impossible into the possible!

User avatar
nedsaid
Posts: 8722
Joined: Fri Nov 23, 2012 12:33 pm

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by nedsaid » Fri Jun 17, 2016 12:55 pm

I found a thread that addressed this: It was entitled "Index front-running -- any hard numbers?" A few quotes from that.

Wagnerjb said:

Gus Sauter of Vanguard has been quoted as saying that the S&P500 index fund loses 0.25% per year to front running.

The fact that VFINX matches the S&P500 index is irrelevant to assessing the front running damage. You would need to compare the S&P500 index to a similar large cap asset class return.

Best wishes.

Andy


verygoodthings said:

According to Bill Bernstein in The Investor's Manifesto, Vangaurd changed index providers to try to minimize the damage of frontrunning. They now primarily use MSCI (Morgan Stanley), they used to use S&P.


Exeunt said:

There are have been two major studies on this issue. The most recent one, by Antti Petajisto of active share fame, pegs the cost at a modest 20-30 basis points annually for the S&P 500 and a more significant 40-80 bp for the Russell 2000. Sauter was probably quoting this figure. Keep in mind that it's a lower-bound estimate.

The other, earlier study by Chen, Noronha and Singhal pegs the cost at a more modest 3-12 bp.
A fool and his money are good for business.

User avatar
whodidntante
Posts: 2207
Joined: Thu Jan 21, 2016 11:11 pm

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by whodidntante » Fri Jun 17, 2016 1:00 pm

I doubt it. I am not sophisticated enough to prove my answer.

I consider value investing a way of taking on additional, acceptable risk.

jalbert
Posts: 2220
Joined: Fri Apr 10, 2015 12:29 am

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by jalbert » Fri Jun 17, 2016 1:53 pm

Q: Do Dimensional Fund Advisors (DFA) funds generate risk-adjusted "alpha?"


How do you measure risk? Using historical sample volatility, or is there some way of measuring risks that were in fact taken by the investor, but that did not materialize? The latter are what drive the risk premia that were actually realized by the investors, but the former are what can be measured.

larryswedroe
Posts: 15598
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by larryswedroe » Fri Jun 17, 2016 2:45 pm

Eric Nelson has article on this very subject at seekingalpha.com. I'm traveling and not easy for me to copy it but I'm sure someone can find it and paste it

Shows alpha for DFA and similar Vanguard funds back to 1998

Larry

larryswedroe
Posts: 15598
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by larryswedroe » Fri Jun 17, 2016 2:46 pm

jalbert by risk-adjusted alpha people mean AFTER accounting for exposure to common factors, not volatility
Larry

jalbert
Posts: 2220
Joined: Fri Apr 10, 2015 12:29 am

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by jalbert » Fri Jun 17, 2016 3:43 pm

Nonetheless we don't have any way of quantifying the actual risks taken with an investment, only the risks that show up in returns, and even those are hard to quantify.

There are other issues. For instance, if you regress data from different time periods, you get a different fit to the factor model, i.e. different coefficients for factors. So, for different time periods, the alpha term includes different residual components of return not accounted for by the coefficient fit for other factors.

This is analogous to beta drift over time with CAPM.

kolea
Posts: 1234
Joined: Fri Jul 11, 2014 5:30 pm
Location: Maui and Columbia River Gorge

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by kolea » Fri Jun 17, 2016 4:30 pm

larryswedroe wrote:jalbert by risk-adjusted alpha people mean AFTER accounting for exposure to common factors, not volatility
Larry


I thought alpha was exactly defined by CAPM and that it is intimately related to volatility (i.e., risk). I assumed that the OP was being somewhat redundant in using the term "risk adjusted alpha" since the derivation of alpha by CAPM already accounts for risk of the assets being considered.

Is "risk adjusted alpha" different than CAPM's alpha? If so, can someone provide a link to its exact definition so we can see how it is calculated, etc.?
Kolea (pron. ko-lay-uh). Golden plover.

jalbert
Posts: 2220
Joined: Fri Apr 10, 2015 12:29 am

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by jalbert » Fri Jun 17, 2016 5:04 pm

In these models risk is just the variance (or if you prefer, standard deviation) of observed return, which can be separated for the separate terms of the model (market risk and unsystematic risk under CAPM).

But the true variance of return is not known. That would require knowing the probability distribution of future returns, and would consider the probabilities and returns of many outcomes that may never occur.

For instance, we could, in theory, have had serious deflation for the last 7 years with inflation at -3%. This likely would have resulted in strongly negative equity returns. Such an outcome had some probability of occuring in the probability distribution of returns, and thus factors into the calculation of the mean and variance of actual returns, but will not be incorporated in any risk-adjusted return calculations you see on any website or in any published article. Nonetheless, it was a risk one took by holding equities during the period, with some unknown probability of occuring.

larryswedroe
Posts: 15598
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by larryswedroe » Sat Jun 18, 2016 12:42 am

s e Alpha and CAPM alpha

CAPM Alpha is very specific to that single factor model. But we know that the CAPM only explains about 2/3 of differences in returns between diversified portfolios. So we now have better models that explain more than 95%. In other words what was alpha in the CAPM is now BETA, returns explained by other factors, other than market beta. So it's no longer alpha, it's beta (or loading on a common factor).

Suggest you read the incredible shrinking alpha, my last book which explains this process.

Larry

jalbert
Posts: 2220
Joined: Fri Apr 10, 2015 12:29 am

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by jalbert » Sat Jun 18, 2016 1:57 am

I understand the difference between capm and multifactor models, I was just using capm as an example. Risk-adjusted returns calculated using sample variance of some historical period as a risk measure suffer from risk not being accurately quantified in whatever model you prefer.
Last edited by jalbert on Sat Jun 18, 2016 12:47 pm, edited 1 time in total.

afan
Posts: 2943
Joined: Sun Jul 25, 2010 4:01 pm

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by afan » Sat Jun 18, 2016 8:21 am

There is a well developed literature that incorporates higher moments of the return distribution into the definition of risk. One looks at the performance adjusted for these, as well as variance. This does not require any assumptions about what factors are priced. It simply assesses return vs these components of risk. Because there are then multiple dimensions to risk, you do not get a unique measure of risk adjusted performance. How much negative skewness one would accept for a given level of expected return and variance depends on an individual 's utility function. One can derive the market average trade-off of risk and return and ask whether particular slices of the market do better than this average.

Overall, size and value tilting gets you more negatively skewed portfolios for the level of expected return and variance. Some people-those who are less sensitive than average to negative skewness -may be better off with such portfolios. By definition, the average investor has an average aversion to negative skewness and does not improve his/her position by investing in such portfolios. But each individual has to decide whether they are average on this sensitivity to risk.

One avoids the postulated downsides of indexing that are addressed with patient trading by holding TSM portfolios. Since most of us don't know our utility functions, there is not much basis for assuming we are less sensitive than most to the higher moments.

The argument for some of the DFA funds, to the extent it makes sense, revolves around the relatively high and consistent factor weighting they offer. Provided you have good reason for thinking you want that factor exposure, these are reliable places to get it. Because the factors do not appear to improve risk adjusted performance for those whose utility functions are average, most people don't have a reason to want these factor tilts that deviate from the market.

It MIGHT make sense to do your own tilting with investments that are under your control to compensate for factor exposure that is not under your control. Consider an executive in an industry with volatile earnings whose compensation is largely in company stock and retirement depends on a company pension. She has high exposure to the factors implied by a very high exposure to one company in one industry. Ideally the remainder of her portfolio would not be TSM. Rather, it would contain a set of factor tilts that got her overall as close to TSM as she could get. If she is formally forbidden to short company stock, or it would be looked upon with great disfavor although permitted, she needs a highly tilted portfolio to compensate for her undiversified position in her company. Factor tilted funds might make sense for her.
We don't know how to beat the market on a risk-adjusted basis, and we don't know anyone that does know either | --Swedroe | We assume that markets are efficient, that prices are right | --Fama

kolea
Posts: 1234
Joined: Fri Jul 11, 2014 5:30 pm
Location: Maui and Columbia River Gorge

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by kolea » Sat Jun 18, 2016 9:52 am

larryswedroe wrote:s e Alpha and CAPM alpha

CAPM Alpha is very specific to that single factor model. But we know that the CAPM only explains about 2/3 of differences in returns between diversified portfolios. So we now have better models that explain more than 95%. In other words what was alpha in the CAPM is now BETA, returns explained by other factors, other than market beta. So it's no longer alpha, it's beta (or loading on a common factor).

Suggest you read the incredible shrinking alpha, my last book which explains this process.

Larry


Perhaps I do need to read your book. Although at this point I think "staying the course" trumps any shifting of my portfolio to gain factor exposure. It is interesting stuff and I like to read what you guys are doing but the terminology seems to be a little loose. For instance, even in F-F three factor model alpha means the same thing as alpha does in CAPM (the intercept of the market line with the Y axis). It is has not changed meaning at all, but as you point out, as the risk model changes the market line shift in such a way as to shrink the value of alpha. Perhaps when the risk model is perfect, alpha will be zero and the factor model is complete. But I wonder how different that perfectly factored portfolio will be from the TSM portfolio?
Kolea (pron. ko-lay-uh). Golden plover.

larryswedroe
Posts: 15598
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by larryswedroe » Sat Jun 18, 2016 11:01 am

kolea
No the alpha definition has changed as it's not a CAPM alpha when you use a 3 factor model, it's a 3 factor model alpha. And if you add momentum, it's now a 4-factor alpha, add low vol and or low beta and quality/profitability and you have a 5 or 6 factor alpha. That's important to understand as each time you add a factor you increase the explanatory power of the model and show the sources of returns

CAPM alpha only accounts for BETA, meaning specifically MARKET beta, which is related to, but not the same as, volatility. It's the risk of a security or portfolio RELATIVE to the risk of the market portfolio.

Larry

User avatar
Rick Ferri
Posts: 8203
Joined: Mon Feb 26, 2007 11:40 am
Location: Vagabond, USA, Twitter: @Rick_Ferri
Contact:

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by Rick Ferri » Sat Jun 18, 2016 12:28 pm

larryswedroe wrote:kolea
No the alpha definition has changed as it's not a CAPM alpha when you use a 3 factor model, it's a 3 factor model alpha. And if you add momentum, it's now a 4-factor alpha, add low vol and or low beta and quality/profitability and you have a 5 or 6 factor alpha. That's important to understand as each time you add a factor you increase the explanatory power of the model and show the sources of returns

CAPM alpha only accounts for BETA, meaning specifically MARKET beta, which is related to, but not the same as, volatility. It's the risk of a security or portfolio RELATIVE to the risk of the market portfolio.

Larry


To be even more specific, beta is a portfolio's sensitivity to movements in the overall market. It roughly measures that part of a portfolio's return that can be attributed to the market. Since you can buy the market with a low-cost total market fund, there isn't any reason to pay excess fees for a fund that has high sensitivity to the market (which describes most actively managed funds). If you're trying to by other factors, you really need to be looking for funds that have high sensitivity to those other factors (other beta's) and consider the cost per unit of factor exposure.

BTW, the market is always defined as having a beta of 1 regardless of how volatile it is. Thus, Larry's comment that beta is related to, but not the same as volatility.

Rick Ferri
The views expressed by Rick Ferri are strictly his own as a private investor and author and do not reflect the views of any entity or other persons.

afan
Posts: 2943
Joined: Sun Jul 25, 2010 4:01 pm

Re: Factor-Based Investing: Do DFA Funds Generate Risk-Adjusted "Alpha?"

Post by afan » Sat Jun 18, 2016 2:58 pm

Alpha means the same thing in single factor and multifactor models. It is the component of return not explained by the factors.
If there is only one factor- as in CAPM- then the single beta also relates to variance. In a multifactor model, there is a beta for each factor. Alpha is the part of the return left unexplained by the collection of factors.

But factors are not "risk".

Factors are putative drivers of return. One could have high and low risk portfolios with the same factor weights, but varying amounts of risk free assets.

It would be entirely possible for a portfolio to have lots of exposure to the popular factors and have a high (or low) return for the volatility, skewness, kurtosis, etc. If it had high return, then it would have a positive risk adjusted alpha- using risk, rather than factors, in the equation.
We don't know how to beat the market on a risk-adjusted basis, and we don't know anyone that does know either | --Swedroe | We assume that markets are efficient, that prices are right | --Fama

Post Reply