Severe criticism of DFA's "profitability" factor research

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Browser
Posts: 4857
Joined: Wed Sep 05, 2012 4:54 pm

Severe criticism of DFA's "profitability" factor research

Post by Browser » Mon Sep 09, 2013 6:33 pm

I found this article by Michael Edesess to be quite provocative. He raises some serious criticisms about DFA's recent research on "profitability" as a new "factor" explaining equity returns and argues it's flaws demonstrate pseudo-science.
Dimensional Fund Advisors (DFA) is a company with a laudable history, founded on solid principles and a valuable product concept.
But I am afraid the company has succumbed to a dreadful descent into scientism.

The famed Austrian economist Friedrich Hayek defined scientism as “slavish imitation of the method and language of science” when applied to the social sciences such as economics. Scientism takes on the trappings of science without its depth or rigor.

DFA recently advocated for tilting an equity portfolio toward companies with higher profitability. The company’s argument begins with a spurious pseudo-mathematical “derivation” of a reason why companies with higher profitability should have higher expected returns. It then, without further ado, enshrines the results of this deeply flawed analysis into theory, using phrases such as “financial economics shows that…,” as if the principle were an integral part of a long-standing and thoroughly proven theoretical framework. DFA provides historical evidence for the relationship that is poorly presented and looks suspicious. Then, without further discussion or explanation, the company simply extrapolates the results found in historical returns to become “expected returns” – all the while displaying with each exhibit the obligatory statement: “Past performance is no guarantee of future results.”

Pardon my cynicism – if that’s what it is – but something is seriously wrong here. The facade of science that the industry presents – of which DFA is only one example – is nothing more than that, a facade, and an especially transparent one at that.

DFA’s justifications for its portfolio tilts toward small stocks and value and now toward profitability are presented as the results of scientific findings. But they are all extrapolations of past results into the future. At least one of these tilts – toward small stocks – has already proved questionable, as I have previously shown. Past outperformance of an investment strategy has almost invariably been shown to be an unreliable indicator of future performance. Will we ever move beyond lip service to this ubiquitous warning?
We don't know where we are, or where we're going -- but we're making good time.

berntson
Posts: 1366
Joined: Mon Oct 29, 2012 12:10 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by berntson » Mon Sep 09, 2013 7:18 pm

I too am skeptical of the new profitability factor. My old three and four factor models are working just fine, and it will take some convincing for me to take on board yet another factor. I also think that DFA has a tendency to engage in precisely the sort of pseudo-science marketing the author suggests (For example, the picture on their home page is a picture of a "resident scientist" who is really just an economist. I'm a philosopher, and most of economics strikes me as being more like applied empirical decision theory (something we philosophers do) rather than something particle physicists do).

That being said, I get annoyed when people complain about using any past data to project future returns. Like this:
Past outperformance of an investment strategy has almost invariably been shown to be an unreliable indicator of future performance.
We can confidently make all sorts of predictions based on historic data. In the long run, stocks will likely outperform bonds. Corporate bonds will likely outperform government bonds. Long-term government bonds will likely outperform my checking account. And, I think, we can confidently say that value stocks and small stocks will likely outperform large stocks and growth stocks. One of the reason we can be confident in these claims is that they have explanatory power (we can construct models using risk factors to explain past returns) and have shown up in out across various international markets.

But I should track down this guy's "refutation" of the small premium. Maybe there is something to it.

User avatar
matjen
Posts: 1968
Joined: Sun Nov 20, 2011 11:30 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by matjen » Mon Sep 09, 2013 8:06 pm

Seems odd that he goes after DFA but fails to even mention Novy-Marx and his academic work. It is my understanding that all the DFA papers and strategy comes from that original academic source which has been peer reviewed.

"Over time, tilts towards value, momentum and profitability have outperformed the market, and due to the diversification benefits, a combined portfolio of these three has provided much higher reward per unit of risk and a significant reduction in extreme risk or losses. Thus, an approach that combines these three themes using an integrated, straightforward methodology that is designed to endogenously and efficiently capture these themes and take advantage of their natural synergies is ideal"

http://www.forbes.com/sites/phildemuth/ ... ty-factor/
A man is rich in proportion to the number of things he can afford to let alone.

hafius500
Posts: 217
Joined: Sat Sep 10, 2011 4:31 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by hafius500 » Tue Sep 10, 2013 6:54 am

matjen wrote:Seems odd that he goes after DFA but fails to even mention Novy-Marx and his academic work. It is my understanding that all the DFA papers and strategy comes from that original academic source which has been peer reviewed.

"Over time, tilts towards value, momentum and profitability have outperformed the market, and due to the diversification benefits, a combined portfolio of these three has provided much higher reward per unit of risk and a significant reduction in extreme risk or losses. Thus, an approach that combines these three themes using an integrated, straightforward methodology that is designed to endogenously and efficiently capture these themes and take advantage of their natural synergies is ideal"

http://www.forbes.com/sites/phildemuth/ ... ty-factor/
Edesess commented on the Fama's mathematical equation. Did Novy-Marx use this equation?

Comparing the IFA article with the replies of Novy-Marx I note some differences (that I emphasized).

Fama (IFA) seems to argue the higher expected return is a fact that is proved mathematically.
Edesess replies the maths is wrong and I can't see an error.

Novy-Marx simply states the outperformance is an empirical observation. He speculates about explanations and adds he believes the effect can persist.
That's what every active manager can promise.

IFA:
More recent research by Fama2, French, and Robert Novy-Marx3 shows that expected profitability—as measured by the ratio of operating profitability minus interest expense to book value—is another reliable and robust dimension of stock pricing and expected returns. The reason for this is easily inferred from the equation that relates the current price of a stock to the discounted present value of future cash flows...
One of the key findings of Novy-Marx’s research is that current profitability is a good predictor of future profitability, and future profitability is a good predictor of higher returns.
Source: link to IFA article provided by Edesess, click on link to 'prior article')

Interview with Novy-Marx:
To reconcile profitability with efficient markets, there are risk-based stories for the gross profitability premium that posit that profitable firms are more exposed to some type of risk that investors require compensation for bearing. While these risk-based stories are certainly plausible, it is relatively hard to simultaneously tell risk-based stories for both profitability and value. One can also explain profitability from a behavioral standpoint arguing that it is due to investors’ mis-reaction to information. While there is not a single theory to explain the existence of profitability, just as there isn’t for value or momentum, the existence of plausible economic theories coupled with strong empirical evidence gives us confidence the effect is there and likely to persist...
Again, if you believe that there is an economically intuitive explanation for profitability, be it risk-based or behavioral, then you have more confidence in the strong empirical evidence and the likelihood that it will persist as a long term source of return.
And in my eyes the reply of Novy-Marx is inconsistent:
He admits he knows no theory that explains the outperformance. Then he states that the "existence of plausible economic theories coupled with strong empirical evidence gives us confidence the effect is there and likely to persist".
How can the existence of competing theories or explanations prove something?
Every data-miner could have given this response.
But this is very different from a mathematical truth.
prior username: hafis50

David Oransky
Posts: 12
Joined: Wed Aug 03, 2011 11:26 am
Location: St. Louis

Re: Severe criticism of DFA's "profitability" factor researc

Post by David Oransky » Tue Sep 10, 2013 4:51 pm

Like Edesess, when I first heard that higher future earnings would result is higher returns, I though "Duh!" However, my very next thought was this seems too easy. Surely anything this obvious has already been factored into the price -- how did the guy who came up with the Efficient Market Hypothesis miss that? It seems like that is about as far as Edesees got also.

However the more I dig into the research and try to wrap my head around it, the closer I move to understanding why Novy-Marx and DFA think a "profitability" strategy should outperform, much like small and value -- hint: it has nothing to do with alpha and everything to do with risk.

There is essentially universal agreement that the current value (price) of a company is the sum of it's future discounted cash flows. Therefore if Company A and Company B have the same expected future cash flows and the market applies the same discount rate, then the prices will be the same. If we now increase the expected future cash flows of Company A, but leave everything else unchanged, the price of Company A will increase. In the article, Edesess uses this illustration to show that because the price will change with higher cash flows that DFA's logic is flawed. While his statements in that regard are correct his conclusion is wrong. The price of Company A doesn't have to increase if at the same time the discount rate also increases. A higher discount rate implies additional uncertainty (risk) in the expected future cash flows. In order to entice investors, the "extra" discount rate has to not only cover the probability that those cash flows won't be delivered, but also a premium.

The simplest analogy I've been able to come up with, is to look at the world of bonds, where cash flows, prices, and discount rates are much easier to isolate than in the equity world. Let's assume we have a government bond with a $15 coupon and a corporate bond with a $30 coupon. Both bonds were issued at the same time, have the same term, and for simplicity of math are trading at par. Therefore, the government bond's future expected cash flows are $15/yr, while the corporate bond's are $30/yr. The price of both bonds is $1000. How is it that we have two bonds with the same price, but different expected future cash flows? The discount rate, or in bond speak, yield. The expected cash flows from the government bond are much more certain then those of a corporate bond because the later has the potential for default and some corporate bonds will inevitability do so. Because there is expected to be some default, the expected return of the corporate bond is not actually 3%, but something less. The yield spread of 1.5% between the government bond and the corporate bond is made up of two components. First, something to offset the probability that some bonds will default and secondly, a risk premium to entice investors to buy the corporate instead of the government bond.

How much of the 1.5% belongs to each component? We don't know, but expect that the risk premium is greater than zero and can look at historical evidence to get an sense of how much investors have received in the past. Regardless of the magnitude of the premium, the yield spread indicates that some amount of premium should exist. Therefore we can reasonably expect corporate bonds to outperform government bonds in the long-run because of higher perceived risk and the associated premium. The historical data provides further evidence that supports this, but it is not needed to make the case that higher risk bonds should have higher return.

If we apply the principles in the above example to stocks we can see that it is the higher discount rate that delivers the additional return, not the higher expected future cash flows. This shouldn't be surprising as EMH states that markets price risk. The only significant difference between my bond analogy and stocks is the fact that we don't know what expected future cash flows of stocks are. With bonds we know the coupon payment, but with stocks dividends aren't contractual and many companies chose to keep earnings in the company. How then are we supposed to figure out which stocks have higher discount rates and thus higher expected returns? This has been the conundrum for a while. After all everything we've talked about up to this point is purely mathematical. Although Edesess accuses DFA of practicing scientism, in many circles math actually qualifies as a hard science!

The "breakthrough" of the Novy-Marx research and follow-on research by DFA is that they think they've found a proxy for expected future cash flows that they think is reliable enough to build a strategy on. While no proxy is perfect by definition, they seem to have found one that at least historically is highly correlated. Much in the same regard that Book-to-Market is used as a proxy for value and Small-minus-Large is used for small. While neither of these proxies is the actual risk, they seem to capture most of it reliably well. Time will tell if DFA's Direct Profitability metric will be as robust as their other proxies.

I think the mere fact that DFA has defined this factor as Direct Profitability is confusing because it makes people focus on the cash flows, when the important piece is actually the discount rate, but I suppose the same could be said about the other risk factors as well.

User avatar
bertilak
Posts: 6127
Joined: Tue Aug 02, 2011 5:23 pm
Location: East of the Pecos, West of the Mississippi

Re: Severe criticism of DFA's "profitability" factor researc

Post by bertilak » Tue Sep 10, 2013 6:08 pm

Laminar wrote:...[quite a lot]...
Thank you for that lengthy but amazingly clear explanation. Maybe it's wishful thinking, but I think I understand that topic better than I used to.
May neither drought nor rain nor blizzard disturb the joy juice in your gizzard. -- Squire Omar Barker, the Cowboy Poet

berntson
Posts: 1366
Joined: Mon Oct 29, 2012 12:10 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by berntson » Tue Sep 10, 2013 6:27 pm

Thanks Laminar for the great post! Maybe there is something to this profitability stuff after all...

richard
Posts: 7961
Joined: Tue Feb 20, 2007 3:38 pm
Contact:

Re: Severe criticism of DFA's "profitability" factor researc

Post by richard » Tue Sep 10, 2013 6:33 pm

Laminar wrote:The "breakthrough" of the Novy-Marx research and follow-on research by DFA is that they think they've found a proxy for expected future cash flows that they think is reliable enough to build a strategy on. While no proxy is perfect by definition, they seem to have found one that at least historically is highly correlated. Much in the same regard that Book-to-Market is used as a proxy for value and Small-minus-Large is used for small. While neither of these proxies is the actual risk, they seem to capture most of it reliably well. Time will tell if DFA's Direct Profitability metric will be as robust as their other proxies.
We might divide finance research into testing economic hypotheses and data mining. Testing economic hypotheses involves coming up with an economic theory and seeing if it has worked. Data mining involves churning through data to see what has worked. Data mining has a deservedly bad reputation. Although neither method can give any assurance that it produces reliable guides to the future, one has more confidence if there is a reasonable underlying economic theory.

I have a very hard time seeing Novy-Marx's work as other than pure data mining. I've yet to see any reason the factors chosen by Novy-Marx (gross profits-to-assets) should work any better than any similar metric, such as p/e. Defenders claim gross profits is better than earnings, as earnings is subject to more accounting games than gross profits, but assets are very vulnerable to the same charge, especially compared to earnings.

Being historically highly correlated, in itself, is not a great rationale for a strategy.

Small and value at least have the argument of being proxies for risk, although one might wonder why more direct risk measures don't work better than SV. Profitability just seems like a shiny new toy, without an underlying theory.
Laminar wrote:If we apply the principles in the above example to stocks we can see that it is the higher discount rate that delivers the additional return, not the higher expected future cash flows.
It's not clear what the practical difference is between saying something has a higher discount rate and that the something has higher expected returns.

SteveB3005
Posts: 1425
Joined: Mon Feb 19, 2007 9:29 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by SteveB3005 » Tue Sep 10, 2013 7:23 pm

I was having flashbacks of Yobria there! Cool.

larryswedroe
Posts: 15711
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Severe criticism of DFA's "profitability" factor researc

Post by larryswedroe » Tue Sep 10, 2013 7:25 pm

Finally had chance to think about this issue, so here are some initial thoughts after reading the piece.

Re Richard's comments and data mining; You can argue that but the only problem is you have the same results across asset classes (large and small, value and growth)and all over the world, in developed and emerging markets. Plus the data holds up for long periods and is stat significant. Pure coincidence? So that addresses that issue IMO. As to Richard's other comments there is logic to the choice of metric. The higher up the income statement you go the less management can influence the data with things like accruals. That is why GROSS profitability is the choice. Also by using that the persistence of profitability is predictable/reliable, out to about 7 years. That is an important point.
What's interesting is that in the field of financial economist Novy Marx paper is considered probably the most important paper since the FF work in 92-93. Not my opinion but the opinion of pretty much everyone I've spoken to in the field. By the way, another important point IMO is that this factor explains Buffett's alpha as I've pointed out (plus low cost leverage he gets). Another coincidence? Too many IMO to be coincidence.

Re the critique that has drawn attention, as already pointed out the paper IMO is filled with either incorrect statements or just plain shows lack of knowledge by the author, mixing up many issues. What Fama and French said was that GIVEN THE SAME PRICE, companies with higher profitability have higher returns. That means simply that there is some risk investors see and expect compensation for and thus require higher discount rate. That is simple math and nothing wrong with it. The author is just plain wrong. He states that not all else is equal and prices move. But the prices HAVE MOVED to reflect the perceived risks.

As to the risk story, there are simple risk explanations, you can find them satisfactory or not, but they are there, just as they are for value. But value is debated as risk or behavioral. So the same applies here. As I already explained there are two good stories here. The first is that highly profitable companies are growth companies. That means more of the cash flows are further in the future, and longer duration assets are riskier than shorter duration assets and thus require higher discount rates. The second story is also simple---highly profitable companies attract more competition, and thus are at risk of not maintaining the high profitability. And reverse also true. IN fact the research shows clearly that abnormal profits have very strong tendency to revert to mean faster than investors think (which some say explains the value premium). So investors could demand premium for taking the risk of competition entering and reducing profits. Then there is the behavioral story which says that investors just make mistake in pricing these companies. Take your pick for the reason and thus you can decide if it will persist or not (remember even behavioral anomalies persist, like MOM and perhaps at least part of the value premium)

As to the 5%/50bp- story. Remember that factors are LONG-SHORT portfolios--which is why long only mutual funds tend to capture about half the value premium, at best. DFA also isn't building profitability funds, but instead incorporating profitability into existing funds. There is a huge difference. As with MOM, where DFA screens out neg MOM stocks (doesn't buy positive MOM stocks) it will be screening out low profitability stocks. Like with MOM that has a positive impact on returns, call it 50bp worth based on estimates. But there are second/third benefits which should be considered. Because profitability (like MOM) is negatively correlated with value, you get a diversification return, and improve tracking error. And finally you increase the tstat of the above market return, meaning it's more reliable.

I hope this is helpful

Best wishes
Larry

larryswedroe
Posts: 15711
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Severe criticism of DFA's "profitability" factor researc

Post by larryswedroe » Tue Sep 10, 2013 7:31 pm

Thought I would add that there are other papers that have similar take on "profitability"
Here's a piece I wrote a while ago on another multifactor model the authors explain the anomalies the FF 3 factor model cannot, thus is a better model. And it includes a profitability factor
It is now competing with the FF model as to which should be the "standard"

We know that the three-factor model isn’t perfect — it fails to explain a wide range of anomalies including momentum and the negative relationship of average returns with idiosyncratic volatility. The authors of the September 2012 study, “Digesting Anomalies: An Investment Approach,” which covers the period 1972-2011, propose a new multifactor model that goes a long way to explaining many of the anomalies that neither the Fama-French three-factor model nor the Carhart four-factor model (adding momentum as the fourth factor) can explain. In the new model (which they call the q-factor model), the expected return of an asset in excess of the riskless rate is described by the sensitivity of its return to four factors:
• The market excess return (MK T).
• The difference between the return on a portfolio of small-cap stocks and the return on a portfolio of large-cap stocks (rME). The size factor earns an average return of 0.31 percent per month and is statistically significant at the 5 percent level.
• The difference between the return on a portfolio of low-investment stocks and the return on a portfolio of high-investment stocks (R*A/A). The investment factor earns an average return of 0.44 percent per month and is statistically significant. It’s worth noting that the investment factor is highly correlated with the value premium (0.69), suggesting that this factor plays a similar role to that of the value factor.
• The difference between the return on a portfolio of high return on equity (ROE) stocks and the return on a portfolio of low return on equity stocks (rROE). The ROE factor earns an average return of 0.60 percent per month, and is statistically significant. Also of importance is that the rROE factor has very low correlation with the Fama-French factors. Thus, we can conclude that this factor provides important new information missing from the Fama-French model. In addition, it has a high correlation (0.50) with the momentum factor, meaning that rROE would play a similar role as the momentum factor in analyzing performance. They also found that the investment and return on equity factors are almost totally uncorrelated, meaning that they are independent, or unique, factors.

Model Based On Investment Based Asset Pricing Theory
The authors explain that their model is based on investment-based asset pricing theory. They explain that the negative relation between expected returns and investment is intuitive. Firms invest more when their marginal q (the net present value of future cash flows generated from one additional unit of capital) is high. Given expected cash flows, high costs of capital imply low net present values of new projects and in turn low investment, and low costs of capital imply high net present values of new projects and in turn high investment. As such, value firms with high book-to-market should invest less, and earn higher average returns than growth firms with low book-to-market. In general, firms with high valuation ratios have more growth opportunities, invest more, and earn lower expected returns than firms with low valuation ratios.
They go on to explain that “high valuation ratios often result from a stream of positive shocks on fundamentals, and low valuation ratios from a stream of negative shocks on fundamentals. High valuation ratios of growth firms can manifest as high long-term prior returns, and low valuation ratios of value firms as low long-term prior returns. As such, firms with high long-term prior returns should invest more and earn lower average returns than firms with low long-term prior returns.”
Summarizing, the authors believe that the q-factor model’s performance, combined with its clear economic intuition, suggests that the model can be used in practical applications such as evaluating mutual fund performance, measuring abnormal returns in event studies, estimating costs of capital for capital budgeting and stock valuation, and obtaining expected return estimates for optimal portfolio choice. And finally, investment companies can also adjust the structure of the financial products offered to investors, going beyond traditional styles such as size and book-to-market. Stay tuned.
Kewei Hou, Chen Xue , Lu Zhang, “Digesting Anomalies: An Investment Approach,” September 2012
Note that Zhang is one of the leading thinkers in financial theory today.
Hope that helps as well

Larry

medesess
Posts: 5
Joined: Tue Sep 10, 2013 10:23 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by medesess » Wed Sep 11, 2013 1:29 am

Perhaps I should jump in here (someone sent me the link – just registered on bogleheads).
I’ll respond to Larry’s first post:
• “the persistence of profitability is predictable/reliable, out to about 7 years”
No surprise here, but what is the implication? In the DFA papers, this is taken to imply that this year’s profitability can be used as a proxy for the entire future profitability stream. But that implication is wrong. If it were right, then – as I pointed out in my Advisor Perspectives article – the same reasoning would imply that the entire future income stream of a bond is greater than the entire future income stream of a stock, and therefore (a la Fama/French, see next points) bonds have a higher expected return than stocks.
• “the paper IMO is filled with either incorrect statements or just plain shows lack of knowledge by the author”
No surprise here either: IMO it isn’t and it doesn’t.
• “What Fama and French said was that GIVEN THE SAME PRICE, companies with higher profitability have higher returns… That is simple math and nothing wrong with it.”
Sure, given the formula it’s simple math. But it doesn’t imply (or even suggest, logically) that when price is not held fixed, companies with higher profitability have higher returns.
• “That means simply that there is some risk investors see and expect compensation for and thus require higher discount rate.”
Although this sentence is sandwiched between the two sentences quoted in the last point, it has nothing to do with the math. If profitability does correlate with higher expected returns, then we could presume that investors thought higher profitability companies were riskier. But nothing like that comes out of the math.
• “value is debated as risk or behavioral”
If it is behavioral (i.e., presumably, “irrational”) it cannot persist forever. Run-ups in price due only to behavioral factors are by definition bubbles.
• “highly profitable companies are growth companies”
Do you mean that companies that are now highly profitable will have the highest growth in profitability in the future – and if so, why – or are you defining “highly profitable” in terms of their whole future profitability stream?
• “That means more of the cash flows are further in the future”
Ah, so by “highly profitable” you are referring to their whole future profitability stream, and you are assuming it starts relatively small and grows fast. Then how can the first year’s profitability be a good proxy for the whole future profitability stream? One company could have low initial profitability and high future growth, while another has high initial profitability but low future growth. How would you distinguish based on their initial levels of profitability?
• “research shows clearly that abnormal profits have very strong tendency to revert to mean faster than investors think (which some say explains the value premium)”
Hmm… Interesting and bears thought.
• “behavioral anomalies persist, like MOM”
Yes, like other behavioral anomalies, MOM persists until suddenly, and in general at an unpredictable time, it sharply reverses, obliterating its gains.

As to your second post, unfortunately it doesn’t carry much weight with anyone who doesn’t already buy into the meaningfulness of the whole body of finance research beginning at least with Fama/French 1992/93. I’m with richard (if I’m not reading too much into the views he expressed): I think these papers fall into the category of data mining. And so do Fama/French – read the first paragraph in Section I of their SSRN paper, A Four-Factor Model for the Size, Value, and Profitability Patterns in Stock Returns: “asset pricing models work forward” while “Empirical asset pricing models work backward. They take as given the patterns in average returns, and propose models to capture the observed patterns.”

SteveB3005
Posts: 1425
Joined: Mon Feb 19, 2007 9:29 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by SteveB3005 » Wed Sep 11, 2013 11:23 am

^ This you? Enjoyed the book if so.

Image

thx1138
Posts: 784
Joined: Fri Jul 12, 2013 2:14 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by thx1138 » Wed Sep 11, 2013 11:36 am

medesess wrote: • “value is debated as risk or behavioral”
If it is behavioral (i.e., presumably, “irrational”) it cannot persist forever. Run-ups in price due only to behavioral factors are by definition bubbles.
????

Finance is full of persistent behavioral errors. Who said it was only related to run ups and bubbles? What about persistent pricing errors of risk? If what you say is true then there should be no free dinner annuity sales around. No one should be leasing vehicles. There shouldn't be people in CA who have extended warranties on their cars but no earthquake insurance.

Persistent behavioral errors are a fact of humans. Whether return can be harvested from them in the market is another question to be sure but I've never seen any justification for your claim they can't persist nor have I ever seen a bubble defined that way. Bubbles are typically viewed as large, long deviations from fundamental value. Nothing here claims the deviation is large at all.

larryswedroe
Posts: 15711
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Severe criticism of DFA's "profitability" factor researc

Post by larryswedroe » Wed Sep 11, 2013 11:42 am

No desire to get into long debate here so I would just add that the evidence is that current profitability does persist and for fairly long time, meaning it has predictive value--If you look at the evidence it's there, and also the price is what the market decides, so it's simple math. That argument that the price isn't fixed makes no sense whatsover. The price is what it is at the time you form the portfolio. Also it's clearly wrong that behavioral anomalies cannot persist. All you need is presence of limits to arbitrage and aversion to leverage (both are present) to allow anomalies to persist as they do. And finally the fact that there are crashes in MOM doesn't change the fact that the premium exists any more than bear markets means that the ERP doesn't exist.
Best wishes
Larry

User avatar
camontgo
Posts: 324
Joined: Tue Oct 27, 2009 1:46 pm
Location: Chicago, IL
Contact:

Re: Severe criticism of DFA's "profitability" factor researc

Post by camontgo » Wed Sep 11, 2013 12:22 pm

The Novy-Marx profitability effect is based on a factor model. That means that the spread in average returns is related to the degree of loading on the new "profitability" factor. (there is co-variance among the stocks which have higher returns)

This suggests a potential risk explanation, since one reason that the returns of a particular group of stocks might move together is that they are exposed to the same types of risk.

A working factor model doesn't, by any means, guarantee that something risk-based (and, therefore, likely to be persistent) is happening, but it is much stronger evidence than a mere association of a particular characteristic with higher historical returns. It may indeed be "data mining", but it is data mining with a more stringent and carefully thought out filter than critics often are willing to acknowledge. For example, I might discover that stocks whose tickers begin with "C" have outperformed the S&P over the past 50 years....but that is a far cry from having a true factor model which works across markets and in outside of the original sample.

I'm not sold on the "profitability factor" yet because I haven't heard a risk explanation that I find plausible, but I do think it has cleared several important empirical hurdles which many other historical anomalies have not.

Here is a chart from John Cochrane on the steps to a risk-based factor model (from course notes available on his website):

Image

For me, accepting that the profitability factor is likely to be something real and persistent would mean that each of the questions in the chart is answered to my satisfaction. So far that isn't true, but there is enough evidence that it is starting to get interesting.
Last edited by camontgo on Wed Sep 11, 2013 1:43 pm, edited 1 time in total.
"Essentially, all models are wrong, but some are useful." - George E. P Box

David Oransky
Posts: 12
Joined: Wed Aug 03, 2011 11:26 am
Location: St. Louis

Re: Severe criticism of DFA's "profitability" factor researc

Post by David Oransky » Wed Sep 11, 2013 1:42 pm

Nothing in investing is certain - if it was it wouldn't be called investing as there would be no return. Instead, we get compensated for taking risk and hope that the compensation we demand is greater than the risk we realize. If we want to increase our expected return, we should increase our exposure to systematic risk. But how do we find these systematic risks?

We can hypothesize about where risk is (a risk story) and/or we can look at historical data for clues.

When we have a good risk story that is also supported by the historical evidence we can feel pretty good about investing in it. I presume this is how most people here feel about the equity risk premium (that over the long-run stocks should outperform bonds because they are riskier). This "story" makes sense because the future cash flows of equity ownership are not contractual and if the company goes under it goes to the bond holders not the equity holders. The equity risk premium is also supported by tons of historical evidence across many time periods and different markets. The data is also very statistically significant (btw statistically significant means a t-stat of 2+ which basically means 95% confidence or greater, but does not mean 100% certainty).

When we get to the small, value, profitability, momentum, etc. factors we start to get a lot more debate - and we should as the risk stories and evidence are not as convincing as the equity risk premium. When faced with the decision about whether to incorporate the research into my own portfolio I often ask myself:

A) Does it make theoretical sense? (is there a risk story)
B) Is it supported by historical evidence at statistically significant levels, persistent through time, and pervasive through markets? (this helps cut-down on pure data mining)
C) Does the perceived benefit outweigh the cost to implement?
D) What if I'm wrong?

The only premium that is a slam dunk on all four criteria is the equity risk premium - it has a good story, good evidence, cheap to implement, and I hold bonds in my portfolio in case I'm wrong and it doesn't actually exist.

The others factors aren't as easy to draw a conclusion on as they get muddied up with whether or not there is a convincing risk story (some will argue the evidence, but the statistics are pretty robust in my mind). Faced with this dilemma many investors will chose to stick solely with the equity risk premium and there is absolutely nothing wrong with that. In fact, it should be everyone's default. However I think it's at least worth considering what the additional cost of incorporating the additional factors into your portfolio is and asking yourself the question, "what if I'm wrong?" If you don't feel comfortable with the answer, stick with a total market approach!

With small, value, and profitability the costs to implement are relatively low. Yes, they are not as tax-efficient as total market, but if you hold them in tax-advantaged accounts you can get rid of the negative impacts. Momentum requires lots of trading and has extremely high turnover as a standalone strategy and therefore is expensive, but is worthwhile to consider in trading decisions (as DFA has done).

The question of "what happens if we're wrong and these additional premiums don't exist going forward?" is the most interesting. If the alternative hypothesis is that the only premium that exists is the equity risk premium, then all of these other factors add nothing and only take away their incremental cost (maybe 10-20 bps on a portfolio). Assuming we've gained exposure to these factors while maintaining broad diversification, then all we've done is buy slightly more expensive stock funds when we should have just bought the cheapest total market fund. Now some could argue that active investing should be viewed in the same light, but there are important differences. Most active funds lack the diversification and they are much more expensive.

SteveB3005
Posts: 1425
Joined: Mon Feb 19, 2007 9:29 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by SteveB3005 » Wed Sep 11, 2013 2:05 pm

It's handy science, a marketing answer for those buzzword savvy investors who need something that fits the Grantham 'high-quality stocks' mold.

larryswedroe
Posts: 15711
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Severe criticism of DFA's "profitability" factor researc

Post by larryswedroe » Wed Sep 11, 2013 2:19 pm

For the benefit of the board I thought I would add this important thought

Fama-French work appears in the most prestigious of peer-reviewed academic journals, with very high standards (rejection rates of 90%+) and reviews by several top and independent academics to make sure the concepts are sound and the data is accurate, and so on. In other words, the papers are well vetted and cited by independent academics. On the other hand this piece was in an advisor journal, with who knows if anyone ever reviewed it, let alone someone with the knowledge of finance needed to review it.


And one other thing
I tried my best to understand what this statement was about --and could not, and no one I talked to could either, including world class mathematicians. It's just plain jibberish.
“the persistence of profitability is predictable/reliable, out to about 7 years”
No surprise here, but what is the implication? In the DFA papers, this is taken to imply that this year’s profitability can be used as a proxy for the entire future profitability stream. But that implication is wrong. If it were right, then – as I pointed out in my Advisor Perspectives article – the same reasoning would imply that the entire future income stream of a bond is greater than the entire future income stream of a stock, and therefore (a la Fama/French, see next points) bonds have a higher expected return than stocks.
There is nothing that could lead one to that conclusion unless you simply don't understand how stocks are priced or valuation models. And neither Fama and French nor Novy Marx ever said that you can use this year's profits as proxy for next year's, only that there is a tendency for the high profitability firms to persist, and regressions show that out to at least 7 years. (DFA stopped running them beyond that so it might even be longer).


Best wishes
Larry

swaption
Posts: 1190
Joined: Tue Jul 29, 2008 11:48 am

Re: Severe criticism of DFA's "profitability" factor researc

Post by swaption » Wed Sep 11, 2013 2:48 pm

I reference the below quote by Novy-Marx referenced in the earlier post by hafius5000:
Again, if you believe that there is an economically intuitive explanation for profitability, be it risk-based or behavioral, then you have more confidence in the strong empirical evidence and the likelihood that it will persist as a long term source of return.
Note the underlined section, because for me this is the bottom line. For any of these factors to persist, one has to be convinced of either the risk or behavioral story. For a long time, the value and size premiums have generally been viewed as related to risk, but those arguements seem to continually be somewhat speculative in nature, and to some extent a convenience to those that might resist a systematic behavioral explanation for anything. I think the same applies to the new profitability factor, although the risk rationale now seems even more speculative in nature. So maybe it is also behavioral, or maybe size and value were behavioral all along. This of course implies the dreaded free lunch. Does a behavioral rationale imply that it can't persist? I don't know, human nature is human nature, and that is basically the behavior we are talking about. Really not much more complicated than that. What I do find a bit tiresome, after following this now for some time, is all the acrobatics to try to explain everything in risk terms.
Last edited by swaption on Wed Sep 11, 2013 3:41 pm, edited 1 time in total.

Browser
Posts: 4857
Joined: Wed Sep 05, 2012 4:54 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by Browser » Wed Sep 11, 2013 3:33 pm

What I do find a bit tiresome, after following this now for some time, is all the acrobatics to try to explain everything in risk terms.
Me too. I have two problems with the risk story: (1) whichever asset factor shows outperformance over the next 20 years will be ordained ex post to have been risky. (2) the risk--> return model as it is currently enshrined in financial academics is nonfalsifiable -- if it isn't immediately obvious, we just haven't mined up a good explanation of the risk just yet.
We don't know where we are, or where we're going -- but we're making good time.

User avatar
matjen
Posts: 1968
Joined: Sun Nov 20, 2011 11:30 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by matjen » Wed Sep 11, 2013 3:39 pm

Laminar wrote:Nothing in investing is certain - if it was it wouldn't be called investing as there would be no return. Instead, we get compensated for taking risk and hope that the compensation we demand is greater than the risk we realize. If we want to increase our expected return, we should increase our exposure to systematic risk. But how do we find these systematic risks?

We can hypothesize about where risk is (a risk story) and/or we can look at historical data for clues.

When we have a good risk story that is also supported by the historical evidence we can feel pretty good about investing in it. I presume this is how most people here feel about the equity risk premium (that over the long-run stocks should outperform bonds because they are riskier). This "story" makes sense because the future cash flows of equity ownership are not contractual and if the company goes under it goes to the bond holders not the equity holders. The equity risk premium is also supported by tons of historical evidence across many time periods and different markets. The data is also very statistically significant (btw statistically significant means a t-stat of 2+ which basically means 95% confidence or greater, but does not mean 100% certainty).

When we get to the small, value, profitability, momentum, etc. factors we start to get a lot more debate - and we should as the risk stories and evidence are not as convincing as the equity risk premium. When faced with the decision about whether to incorporate the research into my own portfolio I often ask myself:

A) Does it make theoretical sense? (is there a risk story)
B) Is it supported by historical evidence at statistically significant levels, persistent through time, and pervasive through markets? (this helps cut-down on pure data mining)
C) Does the perceived benefit outweigh the cost to implement?
D) What if I'm wrong?

The only premium that is a slam dunk on all four criteria is the equity risk premium - it has a good story, good evidence, cheap to implement, and I hold bonds in my portfolio in case I'm wrong and it doesn't actually exist.

The others factors aren't as easy to draw a conclusion on as they get muddied up with whether or not there is a convincing risk story (some will argue the evidence, but the statistics are pretty robust in my mind). Faced with this dilemma many investors will chose to stick solely with the equity risk premium and there is absolutely nothing wrong with that. In fact, it should be everyone's default. However I think it's at least worth considering what the additional cost of incorporating the additional factors into your portfolio is and asking yourself the question, "what if I'm wrong?" If you don't feel comfortable with the answer, stick with a total market approach!

With small, value, and profitability the costs to implement are relatively low. Yes, they are not as tax-efficient as total market, but if you hold them in tax-advantaged accounts you can get rid of the negative impacts. Momentum requires lots of trading and has extremely high turnover as a standalone strategy and therefore is expensive, but is worthwhile to consider in trading decisions (as DFA has done).

The question of "what happens if we're wrong and these additional premiums don't exist going forward?" is the most interesting. If the alternative hypothesis is that the only premium that exists is the equity risk premium, then all of these other factors add nothing and only take away their incremental cost (maybe 10-20 bps on a portfolio). Assuming we've gained exposure to these factors while maintaining broad diversification, then all we've done is buy slightly more expensive stock funds when we should have just bought the cheapest total market fund. Now some could argue that active investing should be viewed in the same light, but there are important differences. Most active funds lack the diversification and they are much more expensive.
Excellent post Laminar. :sharebeer

The last paragraph, in particular, resonates with me in that I have started to dip my toes in what I call the passive+/factor world of investing. Larry S's and Robert Ts writings, blog posts, and board participation have been particularly influential. Although I have been tilted in the more traditional VTI/IJS - VXUS/VSS manner for a couple of years, I am now adding some of the DFA vector and core funds (US and International and Emerging) where I can. I may well be wrong in this move but I figure at worst I will do about as well as before less the 20 basis points. I have no illusion that I will do 2% better but remain hopeful that 50 basis points+ is in the cards. Only time will tell.
A man is rich in proportion to the number of things he can afford to let alone.

larryswedroe
Posts: 15711
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Severe criticism of DFA's "profitability" factor researc

Post by larryswedroe » Wed Sep 11, 2013 5:40 pm

Let me add this to be clear. I have also been quite skeptical of the risk stories. I presented them so each person can make their own decision and to show that there are risk stories. Of course you are free to disagree with them, just as people disagree about the value premium being a risk story. That's why investing is not like physics, it's a social science, not a physical one.

It could also be a behavioral or mispricing story which certainly could persist (or not). But there is no doubt that the HISTORICAL evidence is very strong, no matter where they looked. Now if you think it's behavioral you should JUMP ON IT, as investors arb it away and you'll get a big one time gain. On other hand, lots of evidence that human behavior doesn't change and that is why for example lots of anomalies persist well after discovery.
Best wishes
Larry

User avatar
Blue
Posts: 1151
Joined: Sat Jul 12, 2008 10:18 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by Blue » Wed Sep 11, 2013 7:13 pm

Laminar wrote: The question of "what happens if we're wrong and these additional premiums don't exist going forward?" is the most interesting. If the alternative hypothesis is that the only premium that exists is the equity risk premium, then all of these other factors add nothing and only take away their incremental cost (maybe 10-20 bps on a portfolio). Assuming we've gained exposure to these factors while maintaining broad diversification, then all we've done is buy slightly more expensive stock funds when we should have just bought the cheapest total market fund. Now some could argue that active investing should be viewed in the same light, but there are important differences. Most active funds lack the diversification and they are much more expensive.
The worst case scenario is that the risks for these factors show up with years of dramatic underperformance, testing tilted investors fortitude. The deficit in this scenario would be many times worse than what you outline.

David Oransky
Posts: 12
Joined: Wed Aug 03, 2011 11:26 am
Location: St. Louis

Re: Severe criticism of DFA's "profitability" factor researc

Post by David Oransky » Wed Sep 11, 2013 7:54 pm

Blue wrote:The worst case scenario is that the risks for these factors show up with years of dramatic underperformance, testing tilted investors fortitude. The deficit in this scenario would be many times worse than what you outline.
I agree that is a significant risk and one that needs to be consider before deciding to "tilt" your portfolio, but that has to do with investor behavior (which I don't discount as it's responsible for a lot of lost wealth). All premiums will have periods of positive and negative performance. The S&P 500 underperformed T-Bills from 1965-1981. Small and value have had periods of underperformance for 20+ years. Anyone who expects a free lunch will be humbled eventually. Nothing in investing is certain, it's all probability and even the probabilities aren't precisely known. This is why you should diversify. Don't own only stocks even if you believe in the equity risk premium. Don't only own small, value, profitability, or momentum even if you believe they exist. This is why people often use the term "tilt." They aren't putting everything behind it, but they are giving it some weight. If you own a tilted portfolio you should expect tracking error in both directions and you should be prepared to stay the course even when things are down.

However, I wonder if by acknowledging that these factors could result in long periods of underperformance relative to a total market portfolio, you must also acknowledge that the factors do indeed exist and that there is more than just market beta explaining returns...

User avatar
Blue
Posts: 1151
Joined: Sat Jul 12, 2008 10:18 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by Blue » Wed Sep 11, 2013 8:13 pm

Laminar wrote:
However, I wonder if by acknowledging that these factors could result in long periods of underperformance relative to a total market portfolio, you must also acknowledge that the factors do indeed exist and that there is more than just market beta explaining returns...
I believe it...... But the question will be after 20 years of relative underperformance how many of those tilted will believe it and remain committed to their investment strategy.

With such a period of underperformance there is likely to be convincing experts, data and articles witnessing it was all beta after all and simply seeing reversion to the mean. Ala the Telltale chart by Bogle.

Because changing course at the wrong time can have such a disastrous impact, one should not casually decide to tilt believing the only potential downside is 20-30 bps of additional costs.

medesess
Posts: 5
Joined: Tue Sep 10, 2013 10:23 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by medesess » Thu Sep 12, 2013 1:22 am

Good discussion. Yes, SteveB3005, that’s my book – another one coming in the spring. Thanks for your kind comment.

On Larry’s earlier attack – the part of it related to content – I’ll leave it be because the quote he cites doesn’t sound like gibberish to me. Would China’s and the U.S.’s “profitabilities” (read GDP) in 1980 have been good proxies for their expected GDPs?

richard
Posts: 7961
Joined: Tue Feb 20, 2007 3:38 pm
Contact:

Re: Severe criticism of DFA's "profitability" factor researc

Post by richard » Thu Sep 12, 2013 5:20 am

larryswedroe wrote:Re Richard's comments and data mining; You can argue that but the only problem is you have the same results across asset classes (large and small, value and growth)and all over the world, in developed and emerging markets. Plus the data holds up for long periods and is stat significant. Pure coincidence? So that addresses that issue IMO. As to Richard's other comments there is logic to the choice of metric. The higher up the income statement you go the less management can influence the data with things like accruals. That is why GROSS profitability is the choice. Also by using that the persistence of profitability is predictable/reliable, out to about 7 years. That is an important point.
Larry,

Data mining - trolling through data, as opposed to testing hypotheses, is data mining, even if the results look good. As you say, investing is not physics. Fama has said in his youth he found lots of historically robust results, but they didn't hold up going forward.

Metric - as I said, gross profitability is less subject to accounting issues than earnings, but assets are much more subject to accounting issues than price (comparing profitability to p/e as a valuation metric).

At least we share skepticism that profitability is a risk story.

None of which necessarily means it won't work going forward, but it does have the flavor of the latest fad.
Last edited by richard on Thu Sep 12, 2013 5:42 am, edited 1 time in total.

richard
Posts: 7961
Joined: Tue Feb 20, 2007 3:38 pm
Contact:

Re: Severe criticism of DFA's "profitability" factor researc

Post by richard » Thu Sep 12, 2013 5:28 am

medesess wrote:• “behavioral anomalies persist, like MOM”
Yes, like other behavioral anomalies, MOM persists until suddenly, and in general at an unpredictable time, it sharply reverses, obliterating its gains.
That's a general problem. If it's not a risk story, it can easily sharply reverse, obliterating gains. If it is a risk story, then the nature of risk is the possibility that it will sharply reverse, obliterating gains (else what does risk mean?).
medesess wrote:As to your second post, unfortunately it doesn’t carry much weight with anyone who doesn’t already buy into the meaningfulness of the whole body of finance research beginning at least with Fama/French 1992/93. I’m with richard (if I’m not reading too much into the views he expressed): I think these papers fall into the category of data mining. And so do Fama/French – read the first paragraph in Section I of their SSRN paper, A Four-Factor Model for the Size, Value, and Profitability Patterns in Stock Returns: “asset pricing models work forward” while “Empirical asset pricing models work backward. They take as given the patterns in average returns, and propose models to capture the observed patterns.”
I don't believe you are reading too much into my post.

User avatar
nisiprius
Advisory Board
Posts: 36886
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: Severe criticism of DFA's "profitability" factor researc

Post by nisiprius » Thu Sep 12, 2013 6:19 am

I wish someone would define clearly what is meant when people say something is a "risk story." I am allergic to the word "story" anywhere in investments and don't understand what it's doing here! However. What I assume--and if I'm wrong then everything that follows will also be wrong--is that a "risk story" means that the characteristics of the asset class are mostly explained by its having higher risk, and, in particular, it doesn't have higher risk-adjusted return than the total market.

Well, it seems to me that if it's just a "risk story," then in fact it is not even a risk story, it is just a "correlation story."

If asset X doesn't have higher risk-adjusted return than the total market, T, then what good is it? Why can't I just tune my risk to taste by adjusting my stock/bond ratio? The only possible answer is that X has low correlation with T, so that mixing it with T--overweighting X if X is already part of T--will produce a modest risk reduction, and a modest risk-adjusted return for the combined portfolio.

But that then brings us to a second set of problems.

1) If it is a correlation story, then correlations are notoriously unstable. Possibly less stable even than return?

2) In theory, X will improve T by some amount, possibly tiny, as long as the correlation is less than 1.00000000. But this is ONLY TRUE if X and T are in fact well matched in risk-adjusted return. If X has lower risk-adjusted return than T, then naïvely one would expect it to drag T down. And it will, unless the correlation is genuinely low. The extreme case, the "magic" case is the zero-return asset that can improve T IF, IF, IF it has, not a low correlation, but a strong, persistent, negative correlation. In that case, even though it adds no return, it cancels out some of the risk in T.

I am (not?) surprised that I have not seen any tables or charts anywhere showing what low correlation is "worth." For example, for a pair of assets with unequal risk-adjusted return, how low must the correlation be for the worse asset to improve the better asset. I am cynical enough to think that the reason such tables and charts are uncommon is that they would show that in real life, it is very rare to find persistently low correlations that are low enough to overcome the drag of a lower return.
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.

larryswedroe
Posts: 15711
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Severe criticism of DFA's "profitability" factor researc

Post by larryswedroe » Thu Sep 12, 2013 8:17 am

Richard
First there has been lots of work on profitability for at least a decade. Fama and French wrote about it in 2004, well before Novy-Marx's paper.

http://papers.ssrn.com/sol3/papers.cfm? ... _id=570343

While you may think they data mined rather than had a theory and then tested it that might not be right. Read the above paper and you can come to your own conclusion

Having said that let's assume it was found/discovered by data mining--that doesn't mean that it isn't correct. You test for these things to see if they persist over long periods, exist across asset classes, exist across markets and have t-stats that are significant, and fit with other data (like in this case explains Buffett's alpha)

As I noted with the citation of the paper by Zhang, many have been working on profitability as factor explaining returns and in fact Zhang shows including it and investment (which FF also looked at in their 2004 paper) do a better job than the three factor model in that they explain most of the anomalies that exist in the three factor world

So whatever you believe as to how it was discovered the results seem quite impressive

As to the comment about China and US growth predicting--making that analogy makes no sense since no one is saying country growth rates are predictive or persist, so it's irrelevant whatever the answer is. We do know that profitability persists and thus can be used to form portfolios. Now what is really amazing is that as I noted both the FF paper and the Novy Marx papers appeared in top journals--so we know they were reviewed by many leading academics who passed judgment on the quality of the work. So you can then decide if you think the article attacking them which has not appeared in any peer reviewed journal (IMO for very good reason) has any merit.

Best wishes
Larry

thx1138
Posts: 784
Joined: Fri Jul 12, 2013 2:14 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by thx1138 » Thu Sep 12, 2013 8:38 am

At the risk of dragging off topic - but there are some knowledgeable folks participating - I have a naive question regarding factors beyond the original FF three. Please keep in mind I am committing the cardinal sin of not reading the original papers and understanding their details but instead going from summaries of FF.

I thought the three factors - beta, SMB and HML - accounted for 90% of the return. An improvement compared to CAPM that accounts for only 70%. That seems to leave little room for MOM or profitability.

Or am I making an assumption there? Do four and five factor models still use SMB and HML in the same way FF3 did or are those factors changed such that MOM or Profit now accounts for some of the return formerly in SMB/HML? That is to say SMB for example might have been an incomplete optimization in FF3 skewed by another hidden factor and when a new factor is added SMB becomes a bit smaller because the new factor has "taken" some of that return along with the previous unaccounted for "hidden" factor.

Apologies if this is a poorly formed question - I'm probably grasping at weak mental analogies to Eigendecomposition with which I'm more familiar.

medesess
Posts: 5
Joined: Tue Sep 10, 2013 10:23 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by medesess » Thu Sep 12, 2013 9:17 am

Larry, surely you're aware that the peer review process is not always a guarantee of soundness. I've seen the same mathematical error propagate through a series of peer-reviewed papers because it becomes established by citation - and not only in finance. A widely-read and discussed source like Advisor Perspectives can provide equally good or even better end results, as can this forum. In the particular case we are discussing I believe the peer-reviewed errors are not just propagating they're compounding. If you want to debate you should do it solely on the merits of the argument not on where or how it was published.

longview
Posts: 362
Joined: Wed Mar 19, 2008 5:26 am

Re: Severe criticism of DFA's "profitability" factor researc

Post by longview » Thu Sep 12, 2013 9:25 am

has provided much higher reward per unit of risk
(emphasis mine).

Most of this breaks down, for me, on the idea of the measurement of risk. I recommend The Black Swan which does a great job speaking to this point.
(To color my comments: my situation is ER trying to make a large portfolio that is 99% taxable last 45 years)

Levett
Posts: 4177
Joined: Fri Feb 23, 2007 2:10 pm
Location: upper Midwest

Re: Severe criticism of DFA's "profitability" factor researc

Post by Levett » Thu Sep 12, 2013 9:27 am

"If you want to debate you should do it solely on the merits of the argument not on where or how it was published."

Speaking as a well-published retired academic (not in finance), let me say that medesses has identified a fundamental principal of scholarly integrity.

The merits of the argument are what matter.

Arguments from authority get brushed aside in Logic 101.

Lev

larryswedroe
Posts: 15711
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Severe criticism of DFA's "profitability" factor researc

Post by larryswedroe » Thu Sep 12, 2013 9:30 am

medesses
Yes it is not a guarantee, having said that the profitability factor has been explored and written about now for about 10 years, and by many people in the field, not just FF of course. And the researchers writing on it are considered among the leaders in the field. Now one should consider has everyone of the peer reviewed publications gotten it wrong and only you know the answer? Besides I believe your analysis if filled with big errors, including what appears to be assumptions on your part as to what DFA is doing and they are wrong. Just the part where you mention a 5% premium and tie to Hebner's piece evidences a big problem.
As to Advisors Perspectives, sure that can happen, but I don't believe anyone seriously reviewed your paper before they published it. Had that happened I don't think it would have been published. Just my opinion of course. And I'll wait patiently to see if you get it to appear in a high quality journal, one that is peer reviewed.
And I have pointed out what I believe are the many errors you are making. I have made the points clear I believe and no wish to get in a mud slinging contest. So let's leave it at that.

THK1138 (btw-thought it was one of worst films ever made)
The problem if you will with the three factor model is that while it explains 90% (not of returns but the variation of returns of well diversified portfolios, a very common error) it also leaves many anomalies unexplained. Like MOM, small growth stocks and so on. The paper I cited by Zhang etal appears to improve the explanatory power of the 4-factor (now the "gold" standard) model, adding MOM and at same time explains almost all the anomalies. Hence it is a competing model which just might become the standard someday.

Best wishes
Larry

YDNAL
Posts: 13774
Joined: Tue Apr 10, 2007 4:04 pm
Location: Biscayne Bay

Re: Severe criticism of DFA's "profitability" factor researc

Post by YDNAL » Thu Sep 12, 2013 9:35 am

larryswedroe wrote:You test for these things to see if they [did] persist over long periods, exist[ed] across asset classes, exist[ed] across markets and have [had] t-stats that are significant, and [may] fit with other data (like in this case explains Buffett's alpha)...
Larry,

Prior to grammatical facts in blue, I take serious issue with statements as in the quote above where you present arguments to deliver your point of view/conclusion. You may find it less effective/impactful (?) to write in past tense - but that doesn't make it correct.
  • That which may have been explain[ed] - whether data-mining or whatever we wish to call it - absent a Fama, French, Novy-Marx, Zheng, Swedroe clear crystal ball, it explains nothing about the future (just the past).
  • The best we can do is hope, when we form portfolios, that what has been observ[ed] manifests itself during our investing timeframe.
    larryswedroe wrote:We do know that profitability persists[ed] and thus can be used to form portfolios.
  • There are no guarantees with regards to the future, or it would be included in asset management agreements with clients in firms as Buckingham Asset Management.
Last edited by YDNAL on Thu Sep 12, 2013 9:46 am, edited 1 time in total.
Landy | Be yourself, everyone else is already taken -- Oscar Wilde

User avatar
matjen
Posts: 1968
Joined: Sun Nov 20, 2011 11:30 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by matjen » Thu Sep 12, 2013 9:45 am

YDNAL wrote:
larryswedroe wrote:You test for these things to see if they [did] persist over long periods, exist[ed] across asset classes, exist[ed] across markets and have [had] t-stats that are significant, and [may] fit with other data (like in this case explains Buffett's alpha)...
Larry,

Prior to grammatical facts in blue, I take serious issue with statements as in the quote above where you present arguments to deliver your point of view/conclusion. You may find it uncomfortable (less effective/impactful ?) to write in past tense - but that doesn't make it correct.
  • That which may have been explain[ed] - whether data-mining or whatever we wish to call it - absent a Fama, French, Novy-Marx, Zheng, Swedroe clear crystal ball, it explains nothing about the future.
  • The best we can do is hope that what has been observ[ed] manifests itself during our investing timeframe, when we form portfolios.
    larryswedroe wrote:We do know that profitability persists[ed] and thus can be used to form portfolios.
  • There are no guarantees with regards to the future, or it would be included in asset management agreements with clients in firms as Buckingham Asset Management.
Fair enough, but it seems to me that the above would be true with regard to virtually everything written on this board. I don't think anyone who suggests that factors/tilting is worthwhile is suggesting that this is a mathematical certainty or that the future is certain and known.
A man is rich in proportion to the number of things he can afford to let alone.

YDNAL
Posts: 13774
Joined: Tue Apr 10, 2007 4:04 pm
Location: Biscayne Bay

Re: Severe criticism of DFA's "profitability" factor researc

Post by YDNAL » Thu Sep 12, 2013 10:06 am

matjen wrote:
YDNAL wrote:
larryswedroe wrote:You test for these things to see if they [did] persist over long periods, exist[ed] across asset classes, exist[ed] across markets and have [had] t-stats that are significant, and [may] fit with other data (like in this case explains Buffett's alpha)...
Larry,

Prior to grammatical facts in blue, I take serious issue with statements as in the quote above where you present arguments to deliver your point of view/conclusion. You may find it uncomfortable (less effective/impactful ?) to write in past tense - but that doesn't make it correct.
  • That which may have been explain[ed] - whether data-mining or whatever we wish to call it - absent a Fama, French, Novy-Marx, Zheng, Swedroe clear crystal ball, it explains nothing about the future.
  • The best we can do is hope that what has been observ[ed] manifests itself during our investing timeframe, when we form portfolios.
    larryswedroe wrote:We do know that profitability persists[ed] and thus can be used to form portfolios.
  • There are no guarantees with regards to the future, or it would be included in asset management agreements with clients in firms as Buckingham Asset Management.
Fair enough, but it seems to me that the above would be true with regard to virtually everything written on this board. I don't think anyone who suggests that factors/tilting is worthwhile is suggesting that this is a mathematical certainty or that the future is certain and known.
We disagree.
  • Explain mathematical certainty of "future" (not past) worth, and then you are free to conclude (or suggest) why factors/tilting are worthwhile in the future.
  • Otherwise, it is conjecture and uncertainty; and claims should minimally be expressed in past tense - the reason for my post.
Landy | Be yourself, everyone else is already taken -- Oscar Wilde

User avatar
matjen
Posts: 1968
Joined: Sun Nov 20, 2011 11:30 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by matjen » Thu Sep 12, 2013 10:24 am

YDNAL wrote:
matjen wrote:
YDNAL wrote:
larryswedroe wrote:You test for these things to see if they [did] persist over long periods, exist[ed] across asset classes, exist[ed] across markets and have [had] t-stats that are significant, and [may] fit with other data (like in this case explains Buffett's alpha)...
Larry,

Prior to grammatical facts in blue, I take serious issue with statements as in the quote above where you present arguments to deliver your point of view/conclusion. You may find it uncomfortable (less effective/impactful ?) to write in past tense - but that doesn't make it correct.
  • That which may have been explain[ed] - whether data-mining or whatever we wish to call it - absent a Fama, French, Novy-Marx, Zheng, Swedroe clear crystal ball, it explains nothing about the future.
  • The best we can do is hope that what has been observ[ed] manifests itself during our investing timeframe, when we form portfolios.
    larryswedroe wrote:We do know that profitability persists[ed] and thus can be used to form portfolios.
  • There are no guarantees with regards to the future, or it would be included in asset management agreements with clients in firms as Buckingham Asset Management.
Fair enough, but it seems to me that the above would be true with regard to virtually everything written on this board. I don't think anyone who suggests that factors/tilting is worthwhile is suggesting that this is a mathematical certainty or that the future is certain and known.
We disagree.
  • Explain mathematical certainty of "future" (not past) worth, and then you are free to conclude (or suggest) why factors/tilting are worthwhile in the future.
  • Otherwise, it is conjecture and uncertainty; and claims should minimally be expressed in past tense - the reason for my post.
Actually I think we agree but for some reason you are only picking on Larry's posts in this thread. Wouldn't your position apply equally to everything written about personal investing pretty much? That was the reason for my response. For instance, no one on this board seriously suggests that having anything other than a diversified portfolio across asset classes (and regions for many but not Bogle I guess) consistent with their need, willingness and ability to take risk makes sense. A portfolio that is usually 40%-50% or greater in equities for almost everyone except the very wealthy and the very old. Shouldn't that be in past tense as well at all times? What would Bogle's portfolio look like if a handful of dirty bombs went off in NYC, LA, Chicago, and SF? The future is uncertain for all whether we tilt or express certainty in equities (generally speaking) in the long haul over gold and diamonds, etc. Why does it apply to only tilting/factors?
A man is rich in proportion to the number of things he can afford to let alone.

richard
Posts: 7961
Joined: Tue Feb 20, 2007 3:38 pm
Contact:

Re: Severe criticism of DFA's "profitability" factor researc

Post by richard » Thu Sep 12, 2013 10:25 am

nisiprius wrote:I wish someone would define clearly what is meant when people say something is a "risk story." I am allergic to the word "story" anywhere in investments and don't understand what it's doing here! However. What I assume--and if I'm wrong then everything that follows will also be wrong--is that a "risk story" means that the characteristics of the asset class are mostly explained by its having higher risk, and, in particular, it doesn't have higher risk-adjusted return than the total market.

Well, it seems to me that if it's just a "risk story," then in fact it is not even a risk story, it is just a "correlation story."
Risk story means the hope for higher returns are explained by higher risk, especially the possibility that an investment will not produce adequate returns when you really need the money, such as in an economic downturn. In other words, an economic story rather than a behavioral story or some such.

Dismissing risk as "just" correlation is somewhat odd. Does badly in bad times can be written as a correlation between performance and economic conditions. Correlations are unstable, but so is just about everything in finance. If there were certainty, there wouldn't be risk.

BTW, in an efficient market, everything has the same risk-adjusted return (assuming diversification). Given the lack of an adequate quantification for risk, this statement is hard to test.

larryswedroe
Posts: 15711
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Severe criticism of DFA's "profitability" factor researc

Post by larryswedroe » Thu Sep 12, 2013 10:28 am

YDNAL
First, IMO you are being extremely picky, nitpicking. Of course it was the past-what else could it have been?
Second, of course there are no clear crystal balls, any more than with stocks and the ERP. No one knows that more than me as I have written about that many, many times.
Third, so the best we can do is to see if the data makes sense, exists over LONG periods, exists across markets, and exists across asset classes. IMO the profitability premium passes every one of those tests. Now you can ignore that if you choose because you don't believe there is a good risk story there nor find a convincing behavioral story, but there is no doubt it has existed, survives transactions costs and even explains Buffett's alpha. Pure coincidence in every case?

As I said, I have difficulty with the risks explanations (theories to explain the premium)--for some reason again people are nitpicking over words like story versus explanation---this is a chat board not an academic paper and I certainly don't have the time to carefully review every post--actually have a full time job. And perhaps it is more mispricing. But if you believe the risk story then should use the information. And if believe the behavioral story then you should jump on it either because it will persist for variety of reasons or before market arbs it away by driving UP the prices of the high profitability stocks and down the prices of the low profitability stocks. IMO this is not likely because of limits to arbitrage, costs and risks of margin and human behavior doesn't change (preference for positive skewness for example). At least that is my view. There are more risks in not incorporating the knowledge than in doing so. Again just my opinion and obviously each can form their own opinion.

Levitt
As to authorities and logic. Of course the logic and evidence is what should be focused on. I believe I addressed SOME of the many problems in the article, pointing out what I believe are very clear errors. And I do think there is a difference between citing an "authority" and citing a whole body of peer reviewed literature from leading journals. These papers have been around for about 10 years now, with not a single opposition view to my knowledge. And we know how academics like to "attack" others. Again, that doesn't win the argument but one should consider that before jumping to conclusions when reading a piece like this one. Just my opinion of course. I would add for the group that DFA has been studying this for over 10 years now. They did not just jump on a new paper and implement its findings as it appears at least some think


Best wishes
Larry

richard
Posts: 7961
Joined: Tue Feb 20, 2007 3:38 pm
Contact:

Re: Severe criticism of DFA's "profitability" factor researc

Post by richard » Thu Sep 12, 2013 10:36 am

Larry,

Theory obviously provides that if you pay less for some future earnings type measure (dividends, cash flow, whatever), you will have higher returns. My issue is choosing gross profitability to assets. There does not appear to be any good theoretical reason why this should work better than any competing metric.

In a world of ideal accounting, p/future earnings (or adjusted dividends) would probably be the right metric. Given accounting problems and our inability to know the future, we need to find a proxy. As we've discussed, profitability is less prone to accounting manipulation than earnings, but assets are much more prone to manipulation to price.

It's choosing a non-intuitive metric that makes this seems like data mining. Of course, nothing is certain, things that have worked may be more likely to continue to work, etc.

best,

larryswedroe
Posts: 15711
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Severe criticism of DFA's "profitability" factor researc

Post by larryswedroe » Thu Sep 12, 2013 10:48 am

Richard
There are STRONGLY disagree, in fact there are very good reasons why gross profitability is better metric and in fact it was that 'finding" that really drove the concept of using it

The further you go down the income statement the more management has the ability to manipulate earnings. That is how for example the "accrual anomaly" was created, discovered and then basically arbed away.
So with that the further down you go the less the persistence of profitability (because it really wasn't there in first place in many cases). The higher up you went the better the persistence, allowing you to form portfolios that would have firms with persistent profitability. I explained this several times.

I would add that AQR uses three screens for profitability--so no one "right" way to address this issue
Total Profits / Assets
Total Profits / Sales
Free Cash Flow / Total Assets

And in their core fund which includes value, MOM and profitability in one fund they rank stocks by the three categories, weight them 40%/40/20 and then implement. o

Best wishes
Larry

User avatar
camontgo
Posts: 324
Joined: Tue Oct 27, 2009 1:46 pm
Location: Chicago, IL
Contact:

Re: Severe criticism of DFA's "profitability" factor researc

Post by camontgo » Thu Sep 12, 2013 10:53 am

medesess wrote:If you want to debate you should do it solely on the merits of the argument not on where or how it was published.
I agree that the merit of the arguments is what really matters. However, in this case, your harshest critique is directed at a paper which I haven't been able to find on the web. I suspect it isn't publicly available (anyone have a link?). It is hard to evaluate the merit of the argument you are making if we can't see the original material you are attacking.

Regardless of whether or not I find a link to the paper, I wish your article had focused more on the Novy-Marx or Fama-French papers on profitability. I think these papers lay out the methodology very clearly, and, if you think that the profitability factor is bogus (I'm still a skeptic myself...but I think the evidence is more impressive than some critics give credit for) then you should attack the best arguments. I think Novy-Marx and Fama-French probably make the case more carefully than any paper rolled out by DFA. You do attack the "holding all else constant" justification that FF use as a starting point in their paper, but I think this is a reasonable starting point for any investigation. Of course, all else is not constant in practice, but the relationship may not be as it first appears unless we dig deeper.

For example, is a higher yield on some types of bonds signaling higher expected return (possibly due to risk), or is the difference in yield completely offset by a difference in the default rate (no expected return difference)? Empirical tests can provide some insight. Do historically high or low PEs for the overall stock market mean that the market is anticipating high or low future earnings growth, or has the risk premium that investors demand changed due to economic conditions? Do high interest rates in foreign markets mean higher returns for U.S. investors, or do they signal future changes in exchange rates? The list goes on.... All these questions are worth investigating in the empirical data...we might learn something!... and they start with looking at the equations and utilizing "all else constant" reasoning.
Last edited by camontgo on Thu Sep 12, 2013 11:09 am, edited 4 times in total.
"Essentially, all models are wrong, but some are useful." - George E. P Box

richard
Posts: 7961
Joined: Tue Feb 20, 2007 3:38 pm
Contact:

Re: Severe criticism of DFA's "profitability" factor researc

Post by richard » Thu Sep 12, 2013 10:59 am

Larry,

You continue to ignore the ability to manipulate assets as shown on a balance sheet. Price has a much stronger theoretical and practical justification. It's much harder to manipulate price.

I've always agreed that profitability has less potential for manipulation than accounting earnings.

Bill Bernstein
Posts: 561
Joined: Sat Jun 23, 2007 12:47 am

Re: Severe criticism of DFA's "profitability" factor researc

Post by Bill Bernstein » Thu Sep 12, 2013 11:03 am

We can wave our hands all we want about risk versus behavioral, science versus pseudoscience, data mining versus real signal, or phases of the moon.

Like the rest of us, I find all of that intellectually fascinating on some level; heck, some of us even make writing careers out of it.

But at the end of the day, DFA is an eminently pragmatic firm, and their process boils down to the following:

1) Look for anomalies.
2) Test the most promising ones they can find in other time periods and in other markets. Small and value passed those tests, and now, so does profitability.
3) Take an ex-ante flier on the ones that test out.

The proof is in the pudding: they now have 8 well diversified small/value-oriented funds with 15 year track records, with the following %ile rankings versus their peers:

DFSCX: 17th %ile
DFSVX: 18th
DFLVX: 6th
DFISX: 22nd
DISVX: 1st
DFIVX: 23rd
DEMSX: 8th
DFEVX: 6th

That's not data mining, behavioral nonsense, mis-specifying of risk, or, shudder, pseudoscience; it's cold, hard cash, and over the past two decades I've been more than happy with the process that produced it.

Bill
Last edited by Bill Bernstein on Thu Sep 12, 2013 11:19 am, edited 1 time in total.

thx1138
Posts: 784
Joined: Fri Jul 12, 2013 2:14 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by thx1138 » Thu Sep 12, 2013 11:04 am

Larry -

Thanks for the reply and correction of my understanding of what the factors account for. I probably need to go do some more reading of the actual source material or I'll just be wasting mine and others time!

Oh, and I think I've only ever seen part of THX 1138 - probably just the chase scene at the end - as a kid. However that experience has been very valuable in catching pop culture references to the movie that appear in other programs. For any financial forums I'm on I try to use unique usernames not easily traceable to my "normal" accounts and for some reason that movie sprang to mind when creating this one.

Browser
Posts: 4857
Joined: Wed Sep 05, 2012 4:54 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by Browser » Thu Sep 12, 2013 11:05 am

I think the point of the debate is that no-one really makes much of an honest effort to actually justify the extrapolation of past findings to the future. It's simply assumed that's the best you can do so there's no point in arguing about it. But that assumption opens the door to any mined-up findings. It's a slim reed to build grandiose models on. Perhaps it's best to resist overbuilding models based how things went in the past, in the hope that they don't come crashing down of their own weight as we go forward. Unlike true science, which can make and precisely test predictions based on theoretical models, economics involves the reading of tea leaves and then looking for patterns in cloud formations that subsequently form in order to discover some sort of confirmation. More distressing is that these models often seem to be non-falsifiable, because it can always be claimed that they might only work some of the time, but not all of the time. Paradoxically, if they worked all of the time then they wouldn't work, because they'd be arbitraged away.
We don't know where we are, or where we're going -- but we're making good time.

thx1138
Posts: 784
Joined: Fri Jul 12, 2013 2:14 pm

Re: Severe criticism of DFA's "profitability" factor researc

Post by thx1138 » Thu Sep 12, 2013 11:38 am

Browser wrote:Unlike true science, which can make and precisely test predictions based on theoretical models, economics involves the reading of tea leaves and then looking for patterns in cloud formations that subsequently form in order to discover some sort of confirmation.
I'll let the astronomers, cosmologists, geologists, evolutionary biologists, climatologists and a whole host of other scientists know they aren't doing "true science" :happy

Actual science is very messy and rarely bears any resemblance to what lay people seem to think its practice in the ideal is. Economics certainly lives in realm that makes it very difficult to study and model, but it shares these problems with many branches of science. Science is at best predictive, not deterministic.

I think Bill's post makes a sound point. If you read the start of the Feynman lectures there is a good exploration of what science is and it tacks closer to what Bill is saying than the idealized version of science and prediction some seem to hold.

All that said - anyone who has spent a little time with noisy systems knows the more parameters you try to tune the more likely you are to starting fitting noise rather than fundamentals. Again, Bill makes the same point in his book when exploring over tuning of an efficient frontier. As Larry points out there is a lot of data and peer review around to suggest that's not what is happening with these factors.

Post Reply