The EMH; reply to my critics

Bryan Caplan asks the following question:

My main complaint: Scott’s still not buying my simple modification that makes the EMH far more plausible.  It’s pretty obvious that investors’ mood swings matter a lot.  Why not just say that their mood swings are yet another important hard-to-predict variable that the best minds in the world struggle to forecast?

I’m not convinced mood swings are as obvious as they might seem.  I’ve argued that the stock market crash of 1929 was a rational response to the sudden awareness that we were rushing headlong into Depression.  I wonder if that stock market crash was one of those examples where Bryan thinks it’s “obvious” there was a mood swing.  Even if Bryan doesn’t believe that, I’d estimate about 99.9% of historians do look at the crash that way.  But they are wrong, as they don’t understand the underlying fundamentals driving AD.

Just to show I’m not being dogmatic, I admit that the 1987 crash is much harder to explain, so perhaps a mood swing did occur.  Does this violate the EMH?  Not as long as the mood swings are unpredictable.  But it is hard to claim that mood is unpredictable.  That would mean that it followed a random walk.  But variables that follow a random walk tend to eventually drift off toward infinity or negative infinity.

I am pretty sure that most people who believe in mood swings (including Bryan) have in mind some sort of mean reversion.  Think of 100 as a normal mood, and let higher numbers represent optimism and/or less risk aversion, while lower numbers represent pessimism and/or high risk aversion.

Now suppose the mood index hit 120 right before the 1987 stock crash, and bottomed out at 85 after the crash.  If people could observe current mood levels, the fact that there was some mean reversion would make future movements somewhat predictable.  And that would violate the EMH.  You might ask “Then why do I believe in the EMH?”  The answer might surprise you.  If events like the 1987 crash happened frequently, I would not believe in the EMH.  I think 1987 is a real problem for the EMH.  Even the 2000 bubble had at least the excuse of the massive uncertainty about the future prospects of the internet companies as a plausible explanation for some of the excesses.

Update:  Oops.  The first commenter below (OneEyedMan) pointed out that this is not correct.  It would not violate the EMH, as the changing expected return in 1987 (as stocks fell) would properly reflect the changing level of risk aversion.  So it looks like Bryan was more correct than I thought.  I think at some point we may enter deep metaphysical discussions of what the term ‘efficient’ means, in a world of investors as moody as 14 year old girls.  But since I’ve already made one silly error, I won’t hazard a guess right now.  End of update.

My other problem with Bryan’s argument is that if mood were truly unpredictable, then the EMH would not have to be modified.  Rorty said “that which has no practical implications, has no philosophical implications.”  If mood was a random walk, it would serve precisely the same role as randomly time-varying risk aversion.   It would be observationally equivalent.  But the EMH already allows for time-varying risk aversion.  (Actually I’ve never taken a finance course, so I don’t know that.  Let me rephrase that; the version of the EMH that I believe in allows for random time-varying risk aversion.)

Part 2.  What about those who did see the crisis coming?

Good for them!  That’s been my answer in several comment sections.  This relates to a post by Adam Ozimek:

I think this is part of the reason Sumner is frustrated by the anti-EMH crowd:  he isn’t distinguishing between those that identify profitability as sufficient proof from those who don’t, so he is faced with some people arguing that “people reliably profit off of bubbles, therefore EMH is false!” and others arguing that “market irrationality prevents reliably profiting off of bubbles, and EMH is still false”. There may be people who hold both of these contradictory points at the same time, and they are wrong. But the existence of two separate arguments against EMH that happen to contradict each other is not evidence against either theory or for EMH.

That’s a good point, and my post was meant sort of half jokingly.  I understand that the two complaints are logically distinct, and somewhat in conflict.  My point was more to show the difficulty with debating the anti-EMH crowd.  No matter what happens in the natural experiment performed by LTCM, it will be viewed as a defeat for the EMH by a sizable contingent of the anti-EMH crowd.

But he also raised another issue that is more serious:

Take Calculated Risk, for instance. He clearly and specifically identified the housing bubble using his extensive knowledge of the entire housing industry- from realtors to securitizers. Are we really going to demand that he identify the next bubble in gold, oil, Beanie Babies, or some other market he knows nothing about until we accept that he identified the housing bubble? I see no reason to expect this to be true. Nor do I see any reason to expect that all bubbles be equally identifiable. Using Sumner’s criteria, I’m not sure how you tell the difference between a world in which 1/5 bubbles are identifiable and a world in which 0/5 bubbles are identifiable.

I have actually made two separate arguments regarding the people who have become famous forecasters (Roubini, Calculated Risk) or famous investors (Soros, Buffett.)  I believe that group includes people who are both lucky and also people who actually did in some sense beat the market, or forecast something that the market as a whole missed.  But this is important, it is almost impossible to distinguish between the lucky and the smart.  As a result, their forecasts don’t have practical implications for ordinary investors, for public policymakers, or for social scientists like me.  They are flukes.  If they did it reliably we could simply observe where they invested, and mimic their stock investments.  Soros has even written accounts of his strategy.  Does anyone think I would do as well as he did if I read his explanation?

I hope I don’t botch this, as I am relying on memory.  But a commenter named 123 recently sent me a recent article by French and Fama that I think perfectly encapsulates what I am trying say.  Years earlier, some studies showed that the excess returns of mutual funds are not serially correlated.   These studies seem to me to be almost the only reliable test of the EMH that I have seen.  Other studies done by people like Shiller just strike me as data mining.  In any case, if markets could be beaten by smart investors, then mutual funds run by above average managers should do better on average.  Not every year, but on average.  This means that excess returns should be serially correlated.  But they aren’t.  The successful investors during one year tend to just do average the next, suggesting it was all luck.

Recently Fama and French took a closer look at the data, and found evidence that if you just looked at the top few percentiles, above 97%, there did seem to be a bit of serial correlation.  But even in that group there are also some funds that were lucky.  Interestingly, I am pretty sure that they found that if you could identify which of the top 3% were smart, and not just lucky, you could slightly beat the market.  But if you invest in all funds in the top 3%, you will get a mixture of skilled managers and lucky managers, and you won’t be able to beat an index fund (which has lower expense ratios.)  Odds are I’ve at least slightly misinterpreted their findings.  If so, someone let me know and I will add an update as soon as possible.

So for me the bottom line is that the EMH is not literally true, indeed no model in the social sciences is precisely true.  There are probably occasional mood swings like the 1987 crash, which set up at least slightly forecastable mean reversion.  And there are probably a few people who do on occasion see things the market misses, but they are hard to identify.  In the end, the EMH is still useful for ordinary investors, regulators and social scientists.  But I hope pundits like Calculated Risk and Nouriel Roubini will keep looking for bubbles on the assumption the EMH is not true, and I hope investors like Buffett and Soros will continue assuming the EMH is not true, so that they can help make it more true (or more useful, if like me you deny objective reality.)

I don’t have the copy at home, but I just finished reading Tyler Cowen’s book on globalization and culture.  I seem to recall that he ended the book arguing that us cosmopolitans can only enjoy sampling from a rich and diverse global environment if many segments of the world’s population refrain from cosmopolitanism, but instead impose constraints on people that generate distinct local cultures, and their associated artistic achievements.  The EMH is like that, perhaps Cowen made the connection himself.

Update 2/7/10:  Commenter 123 just sent me the following:

Fama&French study Scott mentioned is really great, you can see it here:
http://www.dimensional.com/famafrench/2009/11/luck-versus-skill-in-mutual-fund-performance-1.html
But they did not study autocorrelation of excess returns there. Their approach was to have a Monte Carlo simulation of distribution of lifetime fund returns where no funds have alpha, and to compare that simulation with a distribution of actual lifetime mutual fund returns.

People might also want to check out his blog, as he knows far more about the market efficiency literature than I do.  And he often disagrees with me.  So you will get a different perspective.


Tags:

 
 
 

48 Responses to “The EMH; reply to my critics”

  1. Gravatar of OneEyedMan OneEyedMan
    5. February 2010 at 17:22

    “If mood was a random walk, it would serve precisely the same role as randomly time-varying risk aversion….But the EMH already allows for time-varying risk aversion. (Actually I’ve never taken a finance course, so I don’t know that. Let me rephrase that; the version of the EMH that I believe in allows for random time-varying risk aversion.)”

    Time varying risk aversion could be 100% forecast-able and consistent with the EMH. We expect the equity premium to be high when risk aversion is large and therefore equities to on average outperform during these periods. But while your portfolio could outperform in this situation you wouldn’t care because you’d be more risk averse in those periods and so see that higher return as a poor deal than during low risk aversion periods.

    In my finance class we always discussed the EMH with respect to the return under the risk neutral measure which would be preserved but that’s distinct from the idea that share prices or even raw share returns need have no forecast-able structure.

  2. Gravatar of John Paul Lewicke John Paul Lewicke
    5. February 2010 at 17:31

    Scott,

    You might like the book Finding Alpha by Eric Falkenstein. He has some interesting thoughts on how a relative utility function explains a lot of problems with CAPM. If you think that you’ll lose your relative place in society if you don’t overweight the hot new sector, then you may be tempted to overweight stuff you wouldn’t otherwise care about. Think California real estate in good school districts, bubble stocks, Morgan Stanley levering up, etc. Even if you think you have an edge in expected return or lower volatility, people really find it hard to stay the course if it means possibly missing the boat big time.

  3. Gravatar of William William
    5. February 2010 at 18:03

    Lemme see if I understood your views correctly.

    1) In any given bubble, some people may notice it, but…
    2) Even in times when there are no bubbles, you’ll find plenty of people claiming bubbles anyway, so…
    3) The signal to noise ratio is pretty lousy, and identifying the tiny prescient fraction is very unlikely, and beyond that…
    4) Identifying the lucky over the prescient is just as unlikely, thus…
    5) Any effect the prescient may have is negligible on markets, so
    6) The practical effects of EMH still hold up, even though it may not be 100% true

    I would imagine if people were capable of living 250 years old, the bubble-prescient would eventually build credibility over the just-lucky. But how many housing bubbles has Calculated Risk lived through? There isn’t a large enough sample size to determine the prescient from the lucky.

    Hmph, I don’t know how accurate your post has translated into my brain, but in any case you’ve pretty much convinced me.

  4. Gravatar of ssumner ssumner
    5. February 2010 at 18:33

    OneEyedman, Thanks. I just did an update to correct my post.

    John, I am sure I would like the book, I was very impressed by his review of Shleifer’s book on inefficient markets. I think about relative position a lot. That’s why I invest in Asia, I figrue it’s my only shot to jump over other upper middle class baby boomers, and grab a nice positional good like a house in West LA to retire in.

    Willian. Yes, that’s what I was saying. And that reminds me that I am too long-winded.

  5. Gravatar of Lorenzo from Oz Lorenzo from Oz
    5. February 2010 at 20:26

    I thought we had the discussion that bubbles were not an inherent problem for EMH.

    I absolutely think there are bubbles in housing markets. I fail to see how that is a problem for EMH since, if people knew they were going to crash, they would not buy assets in the way they do. But I suspect the word ‘bubble’ gets used in different ways which are not helpful to analytical clarity. I simply mean “asset prices experience a strong rise then a precipitous fall “. I am not investing in some notion of market price versus true value.

    The situation in the US is complicated since you had both over-supply (from the pumping of credit specifically into housing) and constrained-supply (from land use regulations), which is why a “national” analysis of the series of local housing bubbles is so misleading.

    Housing prices in Australia are simply mad, but as long as housing supply remains constrained (or, more precisely, the supply of land for housing is constrained by regulation) I see no reason why they cannot stay up, with minor corrections, almost indefinitely, given our population growth.

    Even a (correct) belief that prices will collapse is, in fact, not knowledge in any market-useful sense if one cannot say when. “Prices will collapse in 20 years” is very different in implications than “prices will collapse next month”.

  6. Gravatar of JP Koning JP Koning
    5. February 2010 at 21:04

    Out of curiosity Scott, how would your GDP futures scheme perform if the EMH doesn’t hold? Alot of exuberant investors herding to take one side of the futures market, forcing the Fed to make a lot of exuberant OMOs?

    “If people could observe current mood levels, the fact that there was some mean reversion would make future movements somewhat predictable. And that would violate the EMH.”

    Specifically, it would violate the EMH because said theory says in its weakest form that historical price data and other historical data of non-fundamental nature cannot be used to predict future prices. Mood is historical data of this type; therefore if mood level information could be used to make money, the EMH is violated.

  7. Gravatar of Mike Sandifer Mike Sandifer
    6. February 2010 at 06:31

    Can’t we just leave this with the observation that markets are generally hard to beat and that both anti-EMH and EMH proponents need each other to a degree for either to be right? It seems there are attempts to split hairs with very dull blades.

  8. Gravatar of Artturi Björk Artturi Björk
    6. February 2010 at 06:52

    I think if you frame EMH in this way it can brush off all sorts of critisism:

    The expected value of the price of market beating information is equal to the value of that market beating information.

    So someone can have market beating information, but if you start from skratch and try to “find” market beating information the cost of finding such information equals the benefit derived from that information.

  9. Gravatar of Lord Lord
    6. February 2010 at 07:35

    There is little conflict between EMH and bubbles. Now if one does pay attention to the past, to fundamentals, they are less likely to be deceived by bubbles, but this is just the truth behind the value anomaly to EMH. It does not mean something that is overvalued will not become more overvalued, and it is still very unpredictable as to when it might pay off.

    Some bubbles are easier to spot than others. There has never been an income bubble and the only rent bubbles are temporary ones in resource boom towns. There have been earnings bubbles due to deceptive accounting and to companies chasing bubbles, such as tech suppliers and finance companies.

  10. Gravatar of Philo Philo
    6. February 2010 at 07:48

    Two questions:

    1. Are “mood swings” just changes in time preference, with pessimism/depression being high time preference and optimism/mania low time preference? Are investors guided more by their own personal time preferences, or by their estimates of the market’s time preference (i.e., of the present and future market interest rates)? (The term ‘mood swing’, with its non-cognitive overtones, is appropriate only for the individual’s personal preference, not for his estimate of future market interest rates.)

    2. Why do you attribute 1987 to a “mood swing,” rejecting the view that it was a rational response to new information? The fact that you can’t point to the relevant information is little reason to believe it doesn’t exist. Or are you confident that (a) *no one* has pointed to such information, and (b) if it existed, *some one* would have done so?

  11. Gravatar of mbk mbk
    6. February 2010 at 08:23

    I have to quote my own old post here, http://www.themoneyillusion.com/?p=2348#comment-7859

    In essence, as of Fama himself, 1970, the EMH does not require any specific distributions, such as a random walk. The EMH does allow for “investors’ tastes” (Fama’s words) to influence prices – even in a regular manner, why not? Expected regularities known ahead of time are also reflected in prices then, and no one can make money out of them. The EMH just states that all information, good, bad, or demented, is reflected in prices – it calls this condition an “efficient market”.

    The EMH is about the distribution of information, not about the quality of information. So, bubbles and investor irrationality are not an argument against the EMH. Price movements not following a random walk are not an argument against the EMH. False information leading to a price collapse are not an argument against the EMH. The 2008 crash being caused by a real shortfall in AD is not an argument for the EMH. The 2008 crash being caused by a panic is not an argument against the EMH. The EMH allows for both above.

    Finally, I disagree with the point about global diversity, culture etc. (though I haven’t read Cowen’s book). I used to think this way – when in a first step of globalization, boundaries are lowered or erased, must not a levelling of previous cultural and economic differences follow?

    Well it may, initially. But the diversity can, and should, reappear in other form. Left to their own devices, both the living world and living economic systems, under pressure of competition, naturally differentiate and diversify. Topographically, climatically uniform rainforests are the most (bio-) diverse living environments on Earth. Life and economics do the same thing, they work towards _producing_ gradients and structure, not towards levelling them. They do not increase entropy. They naturally generate endless order and finer and finer differentiation.

    So what is it then that makes say, the “globalized”, read, integrated US states more similar to each other than the hodgepodge of European nations where most US immigrants originally came from? I believe it is rather legal pressure that makes the US more “flat” than Europe: legal standards enforcing uniformity of offerings, expensive safety requirements on products favoring large corporations with economies of scale, political/constitutional requirements ensuring large degree uniformity of legal-political systems etc.

    Globalization initially removes a certain kind of original differences between cultures, but local idiosyncrasies will naturally remain or re-emerge and strengthen if allowed to do so. Local idiosyncrasies are usually suppressed by law and large scale deviations from the mandated global norm are tolerated only for traditional societies. A lot of things reprehensible to our Western sensibilities would be considered an example of remarkable cultural diversity in some corner of Africa, and as illegal cruelty to gangland initiates in L.A. It is not the removal of trade barriers and international travel that makes countries more uniform, but the emergence of global regulations and legal standards and their enforcement.

  12. Gravatar of ssumner ssumner
    6. February 2010 at 12:48

    Lorenzo. I don’t see how you can say housing prices are mad, and also that they are likely to stay up indefinitely. If they stay up indefinitely, that tells me they reflect fundamentals. Otherwise they would fall. Can builders make huge profits at these prices?

    JP Koning, If the risk premium in NGDP markets was fairly stable, it wouldn’t be a big problem. If it was time-varying, it cause cause problems. There is no evidence that risk premia in commodity futures markets are anywhere near enough to cause major problems for NGDP targeting. NGDP is not all that volatile even when it isn’t being targeting (compared to commodity prices.)
    I won’t try to answer your second question, as I am not clear in my own mind the difference between changes in risk aversion, and mood swings. It all seems very vague to me, so I’d probably get it wrong (as I did in the post, before being corrected.) But I certainly don’t have any objection to what you say.

    Mike, That sounds about right, but my instinct is that people will always fight about this.

    Artturi, That sounds right to me.

    Lord, It depends how you define bubbles. If the bubbles aren’t driven by expectations of fundamentals, then I think they are a problem for the EMH. But the mining boomtowns you mention may be driven by fundamentals.

    Philo, Those are good questions. I think when people talk about mood swings they have irrationality in mind. irrationality is a “I know it when I see it idea.” There is also the implicit assumption that some people are not swept up in the hysteria, and can profit from it. Or can regulate it. I don’t have any good answers.

    Yes, I agree that 1987 could have been caused by factors that were fundamental, but not easily observed. But my hunch is that there aren’t any such fundementals that are larged enough. The market lost 23% in just hours. What changes in the economy could have caused that, especially on a slow news day? It just seems weird, so much bigger than any other change in history. And many of the other huge changes did have identifiable causes. So I still think that if things like that happened often, it would be a problem for the EMH. But they don’t, which I think is very significant. It wasn’t just the only drop of more than 20% in all of American history. It was the only drop of more than 15%. There is something that usually prevents huge crashes within a single trading day, even when very bad things happen in the news. My hunch is that that “something” is efficient markets.

    mbk, Don’t assume that you don’t agree with Cowen until you read him. Much of what you say sounds very similar to what he says. He says there are gains and losses, but on balance he likes globalization.

    I don’t follow your definition of efficient markets. As the term is generally understood, market price movements that are irrational and predictable indicate an inefficent market. That’s how most people on both sides define the term. If people preferred to trade stocks at a 20% higher price on TT, compared to MWF, I think that would massively violat ethe EMH.

  13. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    6. February 2010 at 14:12

    Fama&French study Scott mentioned is really great, you can see it here:
    http://www.dimensional.com/famafrench/2009/11/luck-versus-skill-in-mutual-fund-performance-1.html
    But they did not study autocorrelation of excess returns there. Their approach was to have a Monte Carlo simulation of distribution of lifetime fund returns where no funds have alpha, and to compare that simulation with a distribution of actual lifetime mutual fund returns.

    “No matter what happens in the natural experiment performed by LTCM, it will be viewed as a defeat for the EMH by a sizable contingent of the anti-EMH crowd. ”
    I’m in both sizable contingents. It is the severe anomaly LTCM noticed that is a defeat for EMH. Where is Vernon Smith when we need him? He should perform some experimental economics research and try to replicate Royal Dutch Shell in the lab.

    “Yes, I agree that 1987 could have been caused by factors that were fundamental, but not easily observed. But my hunch is that there aren’t any such fundementals that are larged enough. The market lost 23% in just hours. What changes in the economy could have caused that, especially on a slow news day? It just seems weird, so much bigger than any other change in history”
    The market has learned many lessons from 1987 crash and is more efficient now, and a repeat of 1987 is much less likely. 1987 was caused by the inefficiency in the stock futures market, and models used by participants of that market are now much more sophisticated and probability of such crash on a day with no adverse macro news is now much smaller.

    “If the risk premium in NGDP markets was fairly stable, it wouldn’t be a big problem. If it was time-varying, it cause cause problems. There is no evidence that risk premia in commodity futures markets are anywhere near enough to cause major problems for NGDP targeting.”
    As NGDP futures would have short expiration dates, market in them would be very efficient, as many of the sources of market inefficiency do not apply to such markets.

  14. Gravatar of mbk mbk
    6. February 2010 at 19:23

    Scott,

    as far as I understand it, predictability of price movements violates the EMH, but irrational causation does not. The EMH says that efficient markets reflect all information, including information about investor insanity for instance. In its pure form the EMH is definitional and almost tautological. To make it even empirically testable at all, in the words of Fama 1970, it is usually tested under additional assumptions: fair game, expected returns, and market equilibrium. The very existence of trading implies (to me) that locally, the market is never in equilibrium, so locally, a small marginal profit can always be made from exploiting information differentials that have just arisen. These trades then work towards eliminating this differential, until another arises. The random walk model, again according to Fama 1970 himself, depends on additional assumptions not in the pure EMH, made to make the EMH empirically testable.

    1987: If memory serves the 1987 crash is understood largely as the result of a mechanical chain reaction in selling, stemming from concatenations of mandatory stop loss orders, margin requirements etc., triggered by a random sale. As such it may violate the random walk but not necessarily the EMH. A lot of real life price features quoted by say, Mandelbrot, Taleeb etc., violate the random walk, but not the EMH which does not require it, according to Fama.

    Here the verbatim quotes from Fama 1970, my comments in [brackets]:

    A market in which prices always “fully reflect” available information is called “efficient” . [the efficient market is a definition relating to the distribution of information].

    But we should note right off that, simple as it is, the assumption that the conditions of market equilibrium can be stated in terms of expected returns elevates the purely mathematical concept of expected value to a status not necessarily implied by the general notion of market efficiency. The expected value is just one of many possible summary measures of a distribution of returns, and market efficiency per se (i.e. the notion that prices “fully reflect” available information), does not imbue it with any special importance. [investors’ evaluation of prices may depend on other factors than expected returns.].

    … it is best to regard the random walk model as an extension of the general expected return or “fair game” efficient markets model in the sense of making a more detailed statement about the economic environment. The “fair game” model just says that the conditions of market equilibrium can be stated in terms of expected returns, and thus it says little about the details of the stochastic process generating returns.

    But though transaction costs, information that is not freely available to all investors, and disagreement about investors about the implications of given information are not necessarily sources of market inefficiency, they are potential sources.

  15. Gravatar of Winton Bates Winton Bates
    6. February 2010 at 20:49

    TheMoneyDemand mentioned Vernon Smith. The conclusions of an article Vernon Smith wrote with David Porter seem to me to shed some light on the relationship between mood swings and the informational efficiency of markets:
    “Laboratory stock markets in which shares have a
    well-defined expected fundamental (dividend) value,
    that is common information, exhibit strong price bubbles
    relative to fundamental value. These bubbles diminish
    with experience; trades fluctuate around fundamental
    values when the same group returns for a third
    trading session. Thus, common information is not sufficient
    to induce common rational expectations, but
    eventually through experience in a stationary environment,
    the participants come to have common expectations.
    If we suppose that investors are more “inexperienced”
    the longer it has been since the last stock
    market crash, the laboratory results are corroborated
    by a study showing a 98% correlation between the severity
    of declines in the Standard and Poor index and
    the length of time since the last crash.” (Journal of Behavioral Finance,2003. See: http://www.cs.princeton.edu/courses/archive/spr09/cos444/papers/porter-smith03.pdf

  16. Gravatar of Winton Bates Winton Bates
    6. February 2010 at 21:14

    This result from the Porter/Smith article I just quoted seems particularly relevant even though it is a test of rational expectations rather than EMH:
    “Result 10: The results support the rational expectations
    prediction provided that
    the informed traders are endowed
    with a capacity to sell short and the
    uninformed traders are once experienced.
    When the uninformed traders
    are inexperienced, the bubble forces
    are so strong that the informed traders
    are swamped by the buying wave; by
    period 11 they reach their maximum
    selling capacity, including short sales.”
    There could be something in the old adage that it is a good time to sell when large numbers of inexperienced investors enter the market.

  17. Gravatar of Doc Merlin Doc Merlin
    6. February 2010 at 21:34

    @Lorenzo
    “I thought we had the discussion that bubbles were not an inherent problem for EMH.”
    Yes, you can have EMH theory with bubbles, actually… “bubble like” phenomena will arise any time you have long term investors trading on what they perceive to be fundamentals, and short term investors in the same market trading on “the dips.”

  18. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    7. February 2010 at 01:59

    Winton Bates,
    thanks. Yes, bubbles have been replicated in the lab. But what about Royal Dutch Shell type of situations, where fundamental values are not clear, but there are two markets with the same security?

  19. Gravatar of scott sumner scott sumner
    7. February 2010 at 08:07

    123, Thanks, I added an update at the end of the post and a link to your blog.

    Shleifer says it is a great embarrassment, but others don’t seem to agree. And no one has answered my question of whether there is any conceivable state of the world where the Dutch Shell/Shell agreement breaks down.

    Suppose we still had paper stocks, not electronic. And suppose some Microsoft shares were red and some Microsoft shares were blue. Would their value be equal? I say yes. The fact that the two Shells differ makes me think there is something Shleifer is missing.

    This applies to your comment on 1987, and even more to mbk, I don’t see mechanical trading as providing an explanation for 1987. Weeks after the crash stocks were still nearly 40% below levels of August 1987. How does program trading explain those vastly different LEVELS?

    mbk, Most of what you say I agree with. See my answer above on 1987.

    For me, all these questions of “which EMH” boil down to this question: What’s it for? The purpose of the EMH in my book is to tell ordinary investors to buy index funds, to tell regulators that they shouldn’t regulate on the assumption that market prices are irrational, and to tell academics that for most modeling purposes it’s best to assume RATEX. So whatever form of the EMH yields those outcomes is the one I believe in. At the same time, any EMH should allow a role for individuals to ferret out information, and make markets more efficient. So the extreme versions aren’t literally true. I think that is also what you are saying.

    Winton, I strongly believe that (marginal) market makers are far more sophisticated than the students who do experiments in those labs.

  20. Gravatar of Gene Callahan Gene Callahan
    7. February 2010 at 08:28

    “But while your portfolio could outperform in this situation you wouldn’t care because you’d be more risk averse in those periods and so see that higher return as a poor deal than during low risk aversion periods.”

    Say what?! Because the average risk aversion has risen, yours must have risen as well?

  21. Gravatar of Lord Lord
    7. February 2010 at 09:55

    The mining case is interesting as prices are driven by short term fundamentals that will in the long term knowingly collapse as rents fall to the cost of construction. It is not a bubble if it is recognized as temporary but can be if it is assumed permanent. David Merkel considers these differences in time horizons crucial to bubbles. For those with short horizons, bubbles offer attractive investment opportunities to exploit temporary imbalances, while those with long time horizons can be fooled into thinking these are actually represent long term fundamentals and trends. So when one considers expectations one needs to consider whose and for how long. It is often said short term investors are long term investors that encounter falling prices. While EMH encourages long term investment, that is only useful if you are a long term investor and many deceive themselves into thinking they are when they are not.

  22. Gravatar of Nathanael Nathanael
    7. February 2010 at 19:36

    Here’s where you’re wrong:
    “it is almost impossible to distinguish between the lucky and the smart. ”

    Nope. It’s easy. But *only if you’re smart*…..

    Fundamentally, some *people* have an advantage over others. It’s largely due to asymmetric information, but asymmetric information happens even when the information is in the public domain, because some people have spent much more time and energy than others at finding and analyzing said information.

    The true expert in the housing market will identify the bubble when the person without the information about the housing market won’t.

    The fundamental problem with the efficient markets hypothesis is that asymmetric information is everywhere, and it’s *not* randomly distributed. Under these circumstances, the efficient markets hypothesis is plainly ludicrous.

  23. Gravatar of Nathanael Nathanael
    7. February 2010 at 19:37

    Or to put it another way, *under the correct conditions* markets are efficient, but *those conditions never exist*.

  24. Gravatar of Nathanael Nathanael
    7. February 2010 at 19:41

    Oh, there is a subtle implication in what I said which may have been missed.

    You can’t hire out expertise, because only an expert can accurately evaluate other people’s levels of expertise.

    So if I’m an expert in the housing market, I can tell which other people are experts. But why would I hire them (unless I am very very busy) — I’m already an expert. Meanwhile, you’re not an expert, and therefore *you* can’t tell the difference between an expert and someone who’s just lucky — or even a fraud!

    This is not true in all fields, but in finance it appears to be largely true. Therefore hiring someone to invest for you cannot make sense. Either you know exactly what to do personally, or you can’t beat the market.

  25. Gravatar of Nathanael Nathanael
    7. February 2010 at 19:50

    “Years earlier, some studies showed that the excess returns of mutual funds are not serially correlated. These studies seem to me to be almost the only reliable test of the EMH that I have seen. Other studies done by people like Shiller just strike me as data mining. In any case, if markets could be beaten by smart investors, then mutual funds run by above average managers should do better on average. Not every year, but on average. ”

    Nope nope nope nope nope. See, the trouble is that mutual fund managers are *investors for hire*. I just explained that that never makes sense to do from the investor’s POV. But that’s not why this fails as a test — it fails due to the principal-agent problem.

    The managers, when working for someone *other than themselves*, have strange incentives. Some of them are much better than others — at extracting fees for themselves! Making consistent long-run returns is not actually the best way to do this, as a mutual fund manager.

    You have to compare something like sovereign wealth funds or family owned companies — something where people are investing on their own behalf — to get a real sense of whether anyone can “beat the market”.

  26. Gravatar of Nathanael Nathanael
    7. February 2010 at 19:52

    Winton Bates: That Porter/Smith study is beautiful. Thanks for mentioning it. Beautiful experimental work which matches history to a T.

  27. Gravatar of Nathanael Nathanael
    7. February 2010 at 19:55

    Actually, a correction to what I said above: the fact that only experts can accurately evaluate other experts in the finance world *does* impact on mutual fund performance. A good fund manager will be killed by the outflows (triggered by non-experts who invested in the fund) when he’s “hanging on” contrary to short-term market sentiment, and killed by the inflows when he’s just succeeded at making a big gain — the nonexperts investing with him will be actively frustrating his attempt to beat the market. (This problem does not arise with closed-end funds or private equity funds or ‘locked up’ hedge funds.)

  28. Gravatar of scott sumner scott sumner
    8. February 2010 at 06:45

    Gene, I think he just means that a EM sets prices based on the average investors risk aversion.

    Lord, I’ve never thought the short and long term investor distinction is useful. Even short term investors must care about long term performance. Consider a 2 period model. You invest for only one year, and don’t care about the next. But a year later when you sell, the value will reflect expectations about the following year, the year you supposedly don’t care about. The same is true of 3, 4 or 5 year examples.

    Nathaniel, You said;

    “You have to compare something like sovereign wealth funds or family owned companies “” something where people are investing on their own behalf “” to get a real sense of whether anyone can “beat the market”.”

    People used to cite the Harvard endowment. I notice those people have shut up recently.

    I don’t agree with you about mutual funds. There is a principle agent problem. But people do tend to invest based on past returns, so managers do have some incentive to beat the market, and that is the only assumption you need, that they have some incentive, not perfect. Their failure to do better than predicted by luck is devastating to the EMH.

  29. Gravatar of PieFarmer PieFarmer
    8. February 2010 at 09:28

    If EMH holds for the Great Crash because investors learned of a policy change that would produce a severe contraction, then 1987 can be explained the same way. The 1987 crash was precipitated by Greenspan’s rookie admission that the dolllar was too strong and Sec. Baker’s confession that the US would not support the USD. When the markets opened that Monday, they fell into an elevator shaft.
    When markets get surprised, they react quickly.

  30. Gravatar of Lord Lord
    8. February 2010 at 14:14

    That is just an assertion short term traders shouldn’t or don’t exist. That just isn’t the world we live in. Discount rates vary. Market discount rates vary as the trader/investor mix varies and as risk aversion varies.

  31. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    8. February 2010 at 16:20

    “Shleifer says it is a great embarrassment, but others don’t seem to agree. And no one has answered my question of whether there is any conceivable state of the world where the Dutch Shell/Shell agreement breaks down.”

    This is from Wikipedia but should be true:
    Mispricing in Dual Listed Companies

    The shares of the DLC parents represent claims on exactly the same underlying cash flows. In integrated and efficient financial markets, stock prices of the DLC parents should therefore move in lockstep. In practice, however, large differences from theoretical price parity can arise. For example, in the early 1980s Royal Dutch NV was trading at a discount of approximately 30% relative to Shell Transport and Trading PLC. In the academic finance literature, Rosenthal and Young (1990)[2] and Froot and Dabora (1999)[3] show that significant mispricing in three DLCs (Royal Dutch Shell, Unilever, and Smithkline Beecham) has existed over a long period of time. Both studies conclude that fundamental factors (such as currency risk, governance structures, legal contracts, liquidity, and taxation) are not sufficient to explain the magnitude of the price deviations. Froot and Dabora (1999)[3] show that the relative prices of the twin stocks are correlated with the stock indices of the markets on which each of the twins has its main listing.

    Froot and Dabora say:
    We examine pairs of large, ‘Siamese twin’ companies whose stocks are traded around the world but have different trading and ownership habitats. Twins pool their cash flows, so, with integrated markets, twin stocks should move together.

    We conjecture that: (a) country-specific sentiment shocks might affect share intensity, (b) investors are rational, but markets are segmented by frictions other than international transactions costs, such as agency problems.

    “Suppose we still had paper stocks, not electronic. And suppose some Microsoft shares were red and some Microsoft shares were blue. Would their value be equal? I say yes. The fact that the two Shells differ makes me think there is something Shleifer is missing.”
    If you could not convert one red share to one blue share and vice versa, and if most red share investors are republicans, and blue share investors are democrats, share prices might diverge a bit.

    “This applies to your comment on 1987, and even more to mbk, I don’t see mechanical trading as providing an explanation for 1987. Weeks after the crash stocks were still nearly 40% below levels of August 1987. How does program trading explain those vastly different LEVELS?”
    Mechanical trading is just a tool investors used and is not a problem per se. The problem was that stock index options were severely mispriced before the crash, as investors believed that they could easily replicate those options by using various dynamic trading strategies. Now option traders and stock investors are much smarter and they no longer have this wrong belief.

  32. Gravatar of Doc Merlin Doc Merlin
    8. February 2010 at 19:09

    @123
    “Mechanical trading is just a tool investors used and is not a problem per se. The problem was that stock index options were severely mispriced before the crash, as investors believed that they could easily replicate those options by using various dynamic trading strategies.”

    Indexes are always mispriced. This is due to certain tax incentives and forced retirement plans that require investing in indexes and managed funds. This is why now a fairly common strategy for small investors is to try to see where indexes are moving and to get in before the big index funds have finished getting in.

    This goes back to what Nick Rowe and I have been saying about EMH, the market isn’t 100% efficient if people assume it is efficient and then invest in indexes, and the more people do that, the less efficient it becomes.

  33. Gravatar of Doc Merlin Doc Merlin
    8. February 2010 at 19:17

    @123
    “Yes, bubbles have been replicated in the lab. But what about Royal Dutch Shell type of situations, where fundamental values are not clear, but there are two markets with the same security?”

    The whole point is that they aren’t the same security. They are fundamentally different just because they are listed on different exchanges with different laws and different funds buying them. For a retail investor, for example, the transaction costs of buying the London security is different than the transaction costs for buying the other security.

    They have replicated similar effects between a stock and an option for that stock, btw.

    Also, I agree with you, 123, that the only way to make them be the same routinely is for someone to impose some form of convertibility between the two, and even this won’t be perfect. This is why the only way that NGDP future targeting would work in Scott’s or Bill’s regime is if there was convertibility between the the futures and future money.

  34. Gravatar of scott sumner scott sumner
    9. February 2010 at 06:43

    Piefarmer, I can’t argue with your logic, but I just can’t see how those factors were important enough to produce such a big crash. The government has made far worse mistakes, with only a tiny impact on thr stock market.

    Lord, I never asserted that short term traders shouldn’t exist, I said they should care about the long run fundamentals when making short term trades. That is because changes in expected long run fundamentals will affect short run prices.

    123, That information is interesting, but it doesn’t answer my question. It says that the differences that do exist don’t seem big enough to explain the price diversion. But that doens’t mean that those differences could not explain the price diversion, just that the authors don’t think they are large enough. That discussion basically assumes my question away.

    You said;

    “If you could not convert one red share to one blue share and vice versa, and if most red share investors are republicans, and blue share investors are democrats, share prices might diverge a bit.”

    Bingo. Everyone complains my hypothesis is irrefutable. There’s the test. I disagree, and believe the marginal investors would be numerous enough to arbitrage away any differences. Most investors would buy the cheaper one, and that would keep their prices almost identical.

    If stocks were way too high before the crash (Dow at 2700) why did they rise far above 2700 in the 1990s and the 2000s? Wasn’t 2700 a good buying opportunity for long term investors?

    Doc, That’s a good point about transactions costs, but my hunch is that other factors are even more important–such as uncertainty over the future structure of Shell.

  35. Gravatar of Eric Eric
    9. February 2010 at 13:42

    Regarding bubbles being consistent with EMH, I’d like to mention the following book “A world of chance” by Reuven Brenner. I’m reminded here of the tulip bubble story where although it appeared irrational, when faced with plague or invasion, trying to make a few bucks flipping bulbs seems quite rational when faced with those alternatives. Risk of bankruptcy doesn’t seem to be so bad when compared to death.

    Anyhow, great book worth mentioning. Thought the perspective was interesting.

  36. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    10. February 2010 at 06:44

    “But that doens’t mean that those differences could not explain the price diversion, just that the authors don’t think they are large enough. That discussion basically assumes my question away.”
    Everyone who looked at Shell thinks that these differences are very small.

    “If stocks were way too high before the crash (Dow at 2700) why did they rise far above 2700 in the 1990s and the 2000s? Wasn’t 2700 a good buying opportunity for long term investors?”
    There was even better opportunity for long term investors to invest in T-Bills and wait for better entry point.

  37. Gravatar of ssumner ssumner
    10. February 2010 at 07:55

    Eric, Yes, and there is also another book that reached that conclusion.

    123, You said;

    “Everyone who looked at Shell thinks that these differences are very small.”

    Does “everyone” include the 1000s of investors (many of them sophisticated institutional investors) who bought the more expensive version of the stock? What was their motivation?

    123, You said;

    “There was even better opportunity for long term investors to invest in T-Bills and wait for better entry point.”

    That doesn’t prove the price was irrational unless you think people are capable of market timing. Do you really think the average investor can do that? And if so, why don’t they?

  38. Gravatar of Doc Merlin Doc Merlin
    13. February 2010 at 05:46

    @Scott
    ‘Does “everyone” include the 1000s of investors (many of them sophisticated institutional investors) who bought the more expensive version of the stock? What was their motivation?’

    Because of the nature of indices stocks on an index respond to the index as well as the company’s finances. Its like what we were talking about earlier about market inefficiency due to people thinking the market is efficient. Also, because of how indexed stocks tend work, the risk/reward of an index is stock is somewhat spread out to the other members of the index.

  39. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    14. February 2010 at 14:40

    “Does “everyone” include the 1000s of investors (many of them sophisticated institutional investors) who bought the more expensive version of the stock? What was their motivation?”
    Some of them were index funds, some of them were sophisticated domestic funds that don’t invest in foreign exchanges.

    “123, You said;

    “There was even better opportunity for long term investors to invest in T-Bills and wait for better entry point.”

    That doesn’t prove the price was irrational unless you think people are capable of market timing. Do you really think the average investor can do that? And if so, why don’t they?”

    Average investor is not rational and is not capable of stock market timing and housing market timing. You should never rent and you should never invest in T-Bills are two very popular misconceptions.

  40. Gravatar of ssumner ssumner
    15. February 2010 at 09:20

    123 and Doc, What is the barrier to investing in foreign exchanges? And if there is a barrier, as you both imply, then doesn’t this explain the anomaly? The EMH applies only when there are no barriers.

    123, Even Buffett says he can’t do market timing, doesn’t that suggest it is pretty hard? Buffett lost billion in 2008 because he couldn’t do market timing.

  41. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    17. February 2010 at 10:18

    There is no legal barrier. The only barrier is that too many investors prefer to invest in domestic-only funds.

    Buffett does not do market timing on a daily or a monthly scale, but if you look at the annual data of his investments, it is quite consistent with the Shiller’s model.

  42. Gravatar of Some very clever ideas, mostly from Americans « Freethinking Economist Some very clever ideas, mostly from Americans « Freethinking Economist
    18. February 2010 at 01:54

    […] efficient market hypothesis: I see this post about the Invincible Markets Hypothesis as dealing with Sumner’s EMH views, which are all about how difficult it is to beat the market.  Sure. It’s also difficult to […]

  43. Gravatar of scott sumner scott sumner
    18. February 2010 at 10:47

    123, If he likes Shiller’s model, why was he so heavily invested in 2007, when the market was supposed to be grossly overvalued according to Shiller’s model? He lost a fortune in 2008.

  44. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    20. February 2010 at 07:35

    Buffet almost never sells because of the taxation of capital gains, but he attempts to time his stock investments. In 2007 he was mostly accumulating cash.

  45. Gravatar of ssumner ssumner
    21. February 2010 at 07:59

    123, Is he allowed to buy puts?

  46. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    21. February 2010 at 13:21

    Yes he is, but he doesn’t do that. He has a very interesting risk management philosophy, and he prefers to spend his risk budget on a different kind of profitable risks.

  47. Gravatar of buffet investing buffet investing
    30. July 2011 at 13:15

    buffett strategy…

    TheMoneyIllusion » The EMH; reply to my critics…

  48. Gravatar of Is a Better World Possible? Neoliberals and Knowing Your Enemy | Last Men and OverMen Is a Better World Possible? Neoliberals and Knowing Your Enemy | Last Men and OverMen
    17. March 2017 at 05:49

    […]     http://www.themoneyillusion.com/?p=4140 […]

Leave a Reply