Heads I win, tails you lose

The commenter 123 sent me a link to an article by Andrei Shleifer, which argues that markets aren’t efficient.  Before quoting Shleifer, a bit of background on the EMH.

The efficient market hypothesis implies that it should be very difficult to beat the markets, as asset prices should already reflect all publicly available information.  Many people found this hard to accept; surely really smart people are better investors than the average schmuck!  Not surprisingly, the very smartest people of all, including not one but two Nobel Prize-winning finance professors, gave in to temptation and joined a hedge fund that was set up to find market anomalies and to make investments that took advantage of the market’s inefficiencies.  That hedge fund was called “Long Term Capital Management.”  Of course anyone who has read ancient Greek tragedies knows what happened next.  Things didn’t work out quite as well as planned.  LTCM went bankrupt, and had to be rescued by a cartel of banks assembled by the Fed. 

You might think this would lead the anti-EMH forces to give up their fruitless quest for the finance equivalent of the Holy Grail, the fountain of youth, or Prester John’s Kingdom.  The search for the hidden “intrinsic value,” that diverges from the vulgar market value.  You would be wrong.  The anti-EMH forces are stronger than ever.  But here is something I never would have expected.  The failure of LTCM, a firm founded and run on the premise that the EMH was wrong, actually shows that .  .  . the EMH is wrong!

A shrewd investor who noticed, for example, that in the summer of 1997, Royal Dutch traded at an 8% to 10% premium relative to Shell, would have sold short the expensive Royal Dutch shares and hedged his position with the cheaper Shell shares. Sadly for this investor, the deviation from the 60-40 parity only widened in 1998, reaching nearly 20% in the autumn crisis.  This bet against market inefficiency lost money, and a lot of money if leveraged.

In this case, it is said that when Long Term Capital Management collapsed during the Russian crisis, it unwound a large position in the Royal Dutch and Shell trade. Smart investors can lose a lot of money at the times when an inefficient market becomes even less efficient. In fact, as the
LTCM experience illustrates, their businesses might not survive long enough to see markets return to efficiency.

The inefficiency in the pricing of Royal Dutch and Shell is a fantastic embarrassment for the efficient markets hypothesis because the setting is the best case for that theory. The same cash flows should sell for the same price in different markets. It shows that deviations from efficiency can be large and persistent, especially with no catalysts to bring markets back to efficiency. It also shows that market forces need not be strong enough to get prices in line even when many risks can be hedged, and that rational and sophisticated investors can lose money along the way, as mispricing deepens.

You have to be impressed by the resourcefulness of the anti-EMH, crowd.  If LTCM and its merry band of Nobel-Prize winning economists had actually beat the market, if they had used market anomalies to get rich, well then it would have been the death knell of the EMH.  Every time Fama said “if you’re so smart how come you’re not rich,” people would have responded that Scholes and Merton did get rich by spotting market inefficiencies.  Instead they failed miserably, and this shows . . . it show that markets are inefficient because the market can stay irrational longer than you can stay solvent

The more I study the psychology of the anti-EMH crowd, the more surprised I am that anyone still believes in the EMH.  It seems like anything that happens undercuts the EMH.  It reminds me of people who see monopoly everywhere.  High prices?  Clearly monopolistic exploitation.  Low prices?  Ah, that’s predatory pricing.  The same price as your competitor?  Obviously price fixing.

OK everyone, tell me where I’m wrong.  What outcome of LTCM’s bet on Shell Oil would have supported the EMH?

PS.  I don’t know enough about Dutch corporate law to comment on the specifics.  Is there any possible state of the universe that might cause the agreement between Royal Dutch and Shell to break down at a future date?  If so, it would seem to me that Shleifer’s example is invalid.

PPS.  After I wrote this I came across this excellent book review by Eric Falkenstein.  I should just stop discussing the EMH and reference him from now on.


Tags:

 
 
 

58 Responses to “Heads I win, tails you lose”

  1. Gravatar of David David
    3. February 2010 at 19:42

    Scott,

    I’ve noticed that contradiction as well, especially when reading Lowenstein’s “When Genius Failed”. How can anyone believe Merton and Scholes were EMH idealogues when LTCM’s whole strategy was to profit from market inefficiencies?!

    I’m glad you’ve referenced Falkenstein. His blog is excellent, and has forced me to revise my opinion of EMH.

  2. Gravatar of Jon Jon
    3. February 2010 at 20:42

    This seems like another example of logic that stems from that notion that “efficient” means right. “Efficient” does not mean right; it means only that the market knows at least as much as a technocrat would…

    But I don’t think that’s entirely on point.

    Royal Dutch Shell is a DLC. LTCM lost money because they liquidated their arbitrage position BEFORE they intended to. i.e., they were forced to liquidate at a moment when the model would have told them to buy.

    In this way, the juxtaposition of the LTCM collapse and the EMH is entirely wrong. The two were not connected.

    But what about the arbitrage opportunity itself? The frequently cited reason is that Royal Dutch was included in the S&P 500 index… you can chew on whether you believe that.

  3. Gravatar of Chad Hime Chad Hime
    3. February 2010 at 20:59

    The only problem with bashing LTCM’s strategies is that, in a ertain sense, they did work exactly as advertised. The syndicate of banks that purchased and unwound that portfolio ended up making a tidy profit on LTCM’s positions. The issue was liquidity, not necessarily the paired arbitrage positions held by LTCM. LTCM’s issue was really liquidity (isn’t it always though?). They held too many illiquid positions, many of which could not be moved at any price in the market environment following Russia’s sovereign default.

    The other issue was that many of the other ‘smart’ players in the markets (inlcuding many of LTCM’s brokers) were mimicking LTCM’s positions, which led to their positions being even less liquid. In effect, LTCM was operating on such a large scale that they actually moved the markets that they wre attempting to exploit the inneficiencies of.

    EMH cannot be relied on in every circumstance, but it is certainly an effective lens to view market behavior through in a macro sense. You can likely (but not always) rely on EMH to generally describe market behavior in the same sense that you can expect data sets to revert to their long term mean. But don’t forget to use your head – no bhavioral theory operates in a vacuum!

    An imperfect analogy might be found with string theory and relativity. One is consistent on a macro level, the other predicts results on a micro level. Both are useful in their own right. While that analogy may break down at some point(depending on whether you believe there are immutable scientific laws), maybe it an help to show that there are shads of grey, especially in a social science like economics. After all, we’re trying to describe the behavior of people!

  4. Gravatar of Mike Sandifer Mike Sandifer
    4. February 2010 at 03:36

    Scott,

    The section you quoted is all based on a logical fallacy. I hope there are better anti-EMH arguments than that. Unfortunately, other than a few limiting arguments, I can’t think of any.

    It’s not whether EMH is true to me. It’s the degree. It’s clearly true to a degree. And it’s whether it’s the best we can do. I doubt it is.

    Personally, I’ve just started considering the Adaptive Market Hypothesis. You can find the first paper on it here:

    http://web.mit.edu/alo/www/Papers/JPM2004.pdf

    And the wiki: http://en.wikipedia.org/wiki/Adaptive_market_hypothesis

    This offers that markets are hard to beat, but that successful niche strategies are found, but quickly disappear. It claims that stronger EMH theories are sometimes hyperbolic approximations. It has to do with a lot of ecological psychology.

    Will this approach, if it has merit, make for better investors? I don’t know, but if not maybe some versions of the EMH are good enough.

  5. Gravatar of Master of None Master of None
    4. February 2010 at 04:50

    As someone who is paid by investors to spot market inefficiencies, I think about this issue a lot.

    My view is 1) that EMH is mostly useful as a framework for thinking about securities prices, and 2) EMH is only “true” sometimes, depending on your defined time period and/or scale (i.e. housing is an “efficient” market only if you consider it over a long-term and from high-level macro perspective).

    When betting that markets will become more efficient is it CRITICAL to identify the catalyst. Otherwise, you’re just playing roulette.

  6. Gravatar of scott sumner scott sumner
    4. February 2010 at 05:33

    Thanks David.

    Jon, Yes, I agree. So in the future when people say “markets aren’t efficient because Soros or Buffett did so well,” I will remind them of LTCM.

    Chad, I like the analogy to physics. To me the EMH is like Newton’s laws; not precisely true, but close enough to give useful predictions.

    I didn’t suggest that the LTCM strategy was a bad one. Indeed it probably would have been good if they were less leveraged. My point is that it is silly to argue “if genius X succeeds, it will disprove the EMH, and if he fails, it will also disprove the EMH.”

    Mark, You said;

    “The section you quoted is all based on a logical fallacy. I hope there are better anti-EMH arguments than that. Unfortunately, other than a few limiting arguments, I can’t think of any.”

    I am not going to argue with you, but I’ll just point out that Shleifer is one of the world’s leading economists, and a leading proponent of the anti-EMH view.

    You said;

    “Will this [behavioral] approach, if it has merit, make for better investors? I don’t know, but if not maybe some versions of the EMH are good enough.”

    If investors adopt this model, all the gains from it will disappear. It must be kept secret to work.

    Master of none, I have similar views, but would approach it slightly differently. I don’t think scientific theories in general are “true,” only useful or not useful. I think the EMH is useful for government regulators, as even Shleifer says the government isn’t smarter than the markets. And it’s useful for ordinary investors, who are better off in index funds.
    It is not as useful for Wall Street pros, because someone has to dig up the information that makes markets efficient in the first place. Rather, Wall Street pros should be aware of the theory, so that they have a sense that beating the market isn’t easy, and they should be leery of too good to be true get rich quick schemes. But they also need confidence in their own ability to perhaps add some value to the market information set.

  7. Gravatar of Eli Eli
    4. February 2010 at 06:47

    LTCM was writing naked puts on volatility. It turns out the market had priced volatility correctly, and LTCM lost a lot of money. This is completely consistent with the EMH.

  8. Gravatar of azmyth azmyth
    4. February 2010 at 07:14

    Chad Hime: “LTCM was operating on such a large scale that they actually moved the markets that they wre attempting to exploit the inneficiencies of.”
    That’s exactly how the EMH works. As sophisticated investors trade, they move the market price toward a more efficient value. Markets aren’t efficient as decreed by God, they become efficient by aggregating information of the people trading.

  9. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    4. February 2010 at 07:53

    My answer is here:
    http://themoneydemand.blogspot.com/2010/02/answer-to-scott-sumner.html

  10. Gravatar of Dan Dan
    4. February 2010 at 08:04

    My problem with the EMH lies more in its formulation and its implication. The EMH remains a hypothesis and never graduated to a theory precisely because it is not falsifiable. Which is why the alternative, the “anti-EMH” is also non-falsifiable. Every piece of empirical evidence can be interpreted either as supporting the theory or refuting the theory. Thus, Scott’s statement:

    “The more I study the psychology of the anti-EMH crowd, the more surprised I am that anyone still believes in the EMH. It seems like anything that happens undercuts the EMH,”

    could be restated as such:

    “The more I study the psychology of the EMH crowd, the more surprised I am that anyone still does not believe in the EMH. It seems like anything that happens supports the EMH.”

    For instance: the notion that professional investors are incapable of earning excess profits over time beyond what would be predicted by the probability distribution related to market volatility. While this claim can be and is hotly debated (by itself, it is also non-falsifiable, because the probability distribution could be nothing more than the distribution of investor skill), for purposes of this illustration, I’ll take it at face value.

    Pro-EMH: See! if markets were not efficient, then we would see persistent outperformance by more skilled investors. Since Nobel-prize winners are obviously more skilled than others (the skills required for winning Nobels surely are transferable to the investing world), their failure proves the EMH! Now only if Krugman would start his own fund …

    Anti-EMH: The inability of professional investors to “outperform the market” is explained by the fact that markets are highly volatile and unpredictable … and inefficient. If we started with the hypothesis that market prices do not reflect available information or fundamentals, then it would be impossible to predict the direction of asset prices and therefore make consistent excess profits. If markets were efficient, then the volatility implicit in the probability distribution would decline and investor returns would be reasonably predictable. Otherwise, investors could profit off of significant price swings. (Indeed, you referred to Falkenstein, who shows that investors are not rewarded for taking on extra volatility. There is also Mandelbrot, who shows that return distributions are not normal, as is usually predicted by the EMH, but instead follow a power law).

    The other problem with the EMH is the fact that its reasoning is circular.

    The problem with index funds and index fund investors: they are free riders (something you alluded to indirectly in your post), and very likely exaggerate market volatility by reducing the number of professional investors actively engaged in the market.

    Regarding Merton and Scholes, most of the anti-EMH crowd points to the fact that they relied on false assumptions about the distribution of asset prices, assumptions also used by some of their Nobel prize winning theories. Very similar kinds of assumptions were used in the models that predicted the future course of prices in the residential housing markets as well as default rates, at least until 2007 (VAR-type models originated in academia). We all now know how that turned out.

    I think we would all be better off if the academics would follow the evidence, use basic logic, and not create grand theories of everything that cannot be proven. And no, I do not think the EMH is a particularly useful policy insight, and may even do more harm than good.

    Btw, I don’t mean to only post critical comments, because I do like your blog. It is one of the three or four blogs I read regularly, even though the posts are long and sometimes time consuming to read. It is certainly much better than the blogs written by Nobel-prize winning economists that I have read.

  11. Gravatar of Jon Jon
    4. February 2010 at 08:33

    Azmyth writes:
    “That’s exactly how the EMH works. As sophisticated investors trade, they move the market price toward a more efficient value. Markets aren’t efficient as decreed by God, they become efficient by aggregating information of the people trading.”

    Or something like that. There are reasons for the process to breakdown. Studies have shown that arbitrage is not exclusively riskless. e.g., DLC stocks are not convertible. Therefore this is not a classic arbitrage opportunity. LTCM was working something known as pair-trades–two things prices that should move together. They look for divergences and bet on the mean-reversion. (The premise of HFT is that these divergences arise from time-lagged information between markets. Time lags are real. So if you can crunch the “truth” faster, you win. LTCM was betting against longer-run divergences).

    These kinds of arbitrage trades mean that the arbitrager may fail IF the market does not move. Studies have shown that the risk of such strategies arises from too few arbitragers collectively forcing the same result. Now arbitrage ought to be fully self-financing. You should be able to take an unlimited position in your arbitrage. Therefore there should be no game-theory considerations here. Except there is Regulation T which requires collateral to be posted for these trades.

    Regulation T makes the arbitrage trade risky. Consequently it raises the costs of arbitrage, and in so doing increases the range of price discrepancy for which there is no profitable arbitrage.

    You should think about the irony of this. Regulations introduce pricing irregularities. Academics document these irregularities and question whether the market is efficient. These studies are then used to layer on… more regulation.

  12. Gravatar of Eric Eric
    4. February 2010 at 09:28

    I liked one of the comments to Alex Tabarrok’s post a few days back on MR (Daniel Gross, Me, and the Efficient Market Hypothesis) regarding “money on a sidewalk”:

    “For the markets to be efficient, someone has to make them efficient. You should be the one to pick it up, if everyone expects others to be ‘picking it up’ (or making the market efficient) the markets wouldn’t be efficient.

    People providing information is what makes the market efficient.”

  13. Gravatar of Lord Lord
    4. February 2010 at 09:39

    I think the primary problem is people misinterpreting and overinterpreting EMH, often as not its proponents as its opponents. It actually says very little. Like Yglesias, if it were called the unpredictable market hypothesis, it would likely both be more widely accepted and less controversial. The problem is efficiency is equated to rationality but if markets were rational prices would not be subject to tectonic changes, nor would they vary widely from fundamentals. Often people interpret it as markets may not be efficient but they tend towards efficiency (rationality), leading to the type of investing LTCM did. They earned 40% returns their first several years, not a sign of great rationality of markets before they started. In part it was markets catching on to their style that directed them towards increased leverage on a smaller capital base that increased the risk of their strategy that when correlations moved to one ruined them. Yes, markets can stay and increase their irrationality longer than you can stay solvent, and when everyone is being irrational, there is efficiency in being irrational.

  14. Gravatar of Doc Merlin Doc Merlin
    4. February 2010 at 09:48

    Thats an excellent point Jon.

  15. Gravatar of StatsGuy StatsGuy
    4. February 2010 at 10:15

    EMH… Do you realize that entire dollar value of all assets in the world just dropped by 2% today? Based on an employment report that came in 6,000 under expectation, when that is totally within the standard of error? Or, alternatively, the dollar increased by that amount. The correlation in assets is something like 0.92.

    Good grief, BONDS and STOCKS are correlated at approaching 1 today! That is not supposed to happen. What does it mean? It means that international capital flows (USD carry trade unwind, Greece/Italy/Spain/Portugal/UK fears, and now China bubble fears) are totally swamping domestic policy.

    Dow 10,067.33 -203.22 -1.98%
    Nasdaq 2,141.61 -49.30 -2.25%
    S&P 500 1,072.10 -25.18 -2.29%
    10 Yr Bond(%) 3.6150% -0.8800
    Oil 73.18 -3.80 -4.94%
    Gold 1,067.00 -44.40 -3.99%
    FTSE 100 5,139.31 -113.84 -2.17%
    DAX 5,533.24 -138.85 -2.45%
    CAC 40 3,689.25 -104.22 -2.75%

    The Fed will sit on its thumbs for another month till the next meeting. To protect the dollar. It’s all about the dollar. The “integrity” of the currency trumps all, even though the silly thing is that nothing is ensuring the long term risk to the currency more than massive deficits caused by economic deterioration. Up is down. The short term value of the dollar itself is more important than the economy it is supposed to support.

  16. Gravatar of Patrick R. Sullivan Patrick R. Sullivan
    4. February 2010 at 10:20

    Eric Falkenstein’s blog is here:

    http://falkenblog.blogspot.com/

    Which might be of interest to Scott simply because he’s recently found a way to cut down on his blogging, to get his (professional) life back.

    He has a book:

    http://www.amazon.com/dp/0470445904?tag=defrprob-20&camp=14573&creative=327641&linkCode=as1&creativeASIN=0470445904&adid=1SENDHAJRK5MB35QHME3&

    Videos based on his book here:

    http://www.efalken.com/video/index.html

    Interestingly, he seems to have to found anomalies to exploit himself!

  17. Gravatar of Master of None Master of None
    4. February 2010 at 10:49

    Scott said: “And it’s useful for ordinary investors, who are better off in index funds.”

    This is generally accepted as true, although I rarely see much empirical support for this (much the same way that “stocks always perform over the long run” and “home prices always go up in the long run” were generally accepted as true).

    If everyone invested only in ETFs, what would our markets look like?

    One implication is that the firms that choose index constituents would hold more and more power. Do we trust them not to manipulate markets or take advantage of that power?

    As a single-name equity analyst that focuses on fundamental analysis, this might create more opportunities (for me) to beat the market. But it would also create more risk relating to index changes.

    In any case, I’m not convinced.

  18. Gravatar of Master of None Master of None
    4. February 2010 at 11:48

    …Scott Sumner on efficient (aka unpredictable) markets: ‘The failure of LTCM, a firm founded and run on the premise that the EMH was wrong, actually shows that . . . the EMH is wrong!’…

  19. Gravatar of D. Watson D. Watson
    4. February 2010 at 13:18

    I’m reminded of what you wrote (or copied) earlier this week: It’s not that hundred dollar bills never exist, but that if you look at any particular stretch of sidewalk you won’t find any. Any one example is never going to disprove anything any more than it will prove it.

  20. Gravatar of david glasner david glasner
    4. February 2010 at 13:45

    Scott, I am afraid that I just can’t get past that annoying 3 percent. We may have discussed this already, but didn’t Bob Shiller refute EMH in his papers on excess volatility? Maybe you will say that EMH is only about the unpredictability not the volatility of stock prices. If that is your position, then I am wondering what part of Keynes’s beauty contest theory of stock market prices in chapter 12 of the General Theory is inconsistent with your understanding of EMH?

    I agree that stock prices and asset prices are hard to predict, but the financial industry is devoting billions of dollars in research activities to try to predict them. Is that rational or irrational behavior on their part? Obviously there is no model of price formation that allows one to predict the future course of asset prices. In what sense does the absence of such a model imply anything about the “efficiency” of asset markets?

    EMH is a loaded term, because it is meant to suggest a normative evaluation of how markets work that carries a lot more weight than mere unpredictability. Arguments about EMH are therefore more about how people feel about markets than about deep theory or evidence. But it’s Fama who is mainly responsible for the direction that the debate has taken. I happen to like a lot of what he has written, but he is a bit of a nut.

    In my view, for what it is worth, the conceptual problem with EMH is that there are no fundamentals. Thus, there is never any “true” or “correct” valuation of an asset. There are only expectations. The value of any asset at any moment depends on the distribution of expectations about its future price and the current price is the median of all expectations at that moment. There is excess volatility because expectations tend to be positively correlated (the Keynes effect, but it is also implied by network effects). It is only in futures markets for certain commodities where futures price must at some point coincide with spot price that fundamentals can ever determine pricing. There is no possible justification for the price of gold being $1100 an ounce. But the value of gold consists almost entirely in expectations of its future value and the future will never arrive. So when will fundamentals govern the price of gold?

    By the way, I just love it when commercials on television touting gold investment announce that gold has appreciated at a 40 percent annual rate for the last two years, surpassing any other asset. What is the rational investor supposed to conclude from that? That there is a gold bubble? What other asset in the whole world has a value less related to fundamentals than gold?

  21. Gravatar of TGGP TGGP
    4. February 2010 at 15:39

    Mike Sandifer, I’ve also become interested in adaptive expectations recently. I believed that was actually the background assumption Milton Friedman worked under. Adaptive expectations would also seem to make it hard (though not impossible) to beat the market. Does anyone know on what basis ratex was able to displace it?

    Dan: it sounds like you are saying for the anti-EMH position that fundamentals have nothing to do with prices, which is why they can’t predict future prices. So then why bother with talk about “fundamentals” at all? Also, does the EMH really predict normal distributions rather than “fat tails”? Fama, who more than anyone is associated with EMH, agrees with Mandelbrot on the issue of tails. He thinks they actually form stable distributions, of which normal distributions are a subset.

  22. Gravatar of TGGP TGGP
    4. February 2010 at 15:54

    I commented before checking out Sandifer’s wiki article. Adaptive Market Hypothesis is apparently a new thing from Andrew Lo, I was thinking of the old adaptive expectations theory:
    http://en.wikipedia.org/wiki/Adaptive_expectations
    Apparently its predictions do not hold up well (see “Cobweb model”), people behave along the lines of rational expectations.

  23. Gravatar of ssumner ssumner
    4. February 2010 at 19:33

    Eli, That sounds right.

    AZmyth, I agree.

    123, You missed my point. I agree that logically it doesn’t refute the EMH at all. I am saying that it would be used by the anti-EMH crowd to refute the EMH, regardless of the outcome. You might argue that if a smart anti-EMH person keeps getting rich by taking advantage of market anomalies, that this fact it doesn’t support the anti-EMH position. But believe me, there are plenty of people who will argue otherwise, many have on this very blog.

    Dan, Few social science theories are falsifiable. But I find it extremely useful in my research. For instance, when I look at the impact of news on equity prices, the EMH allows me to focus on the immediate effect, as news should be immediately incorporated into equity prices. That makes it 100 times easier to isolate the market response.

    Jon, Good point about regulation. It is often that way. Think health care.

    Eric, I liked that comment too.

    Lord. 40% returns is nothing for a leveraged company in an up market. I could have easily gotten 40% in the late 1990s from simple index funds; just get 2/1 leverage. The market was rising 20% a year.

    Statsguy, You said;

    “EMH… Do you realize that entire dollar value of all assets in the world just dropped by 2% today? Based on an employment report that came in 6,000 under expectation, when that is totally within the standard of error?”

    How do you know why stocks dropped? I just did a post on this, and have a different view.

    But I certainly agree with you on the Fed.

    Patrick, Thanks for the info, I’ll check it out.

    Master on None, That is the fallacy of composition. As individuals, it may be in our interest to go the index fund route, but as a society, we might be better off if people did research and picked stocks. My hunch is some of each. I should buy an index fund, and you should pick stocks.

    D. Watson. Thanks.

    David, Neither Shiller nor anyone else has refuted the EMH. You can’t refute the EMH with data mining. I found Keynes’s Chapter 12 to be very weak. He simply asserts that the stock market is irrational, without providing any evidence. Or I should say he mentioned the ice market, but a commenter said that was checked out and was wrong.

    You said;

    “I agree that stock prices and asset prices are hard to predict, but the financial industry is devoting billions of dollars in research activities to try to predict them. Is that rational or irrational behavior on their part? Obviously there is no model of price formation that allows one to predict the future course of asset prices. In what sense does the absence of such a model imply anything about the “efficiency” of asset markets?
    EMH is a loaded term, because it is meant to suggest a normative evaluation of how markets work that carries a lot more weight than mere unpredictability. Arguments about EMH are therefore more about how people feel about markets than about deep theory or evidence.”

    There is no way to use simple logic to figure out whether the EMH is useful. By the way, I don’t like to talk about whether it is “true,” as no social science theory is literally true, in the sense of works perfectly.

    I don’t know why so much is invested in stock reserach, but the fact that it is makes the market more efficient. The question of free markets is totally unrelated to regulation. The gasoline market may be completely efficient in an EMH sense, but may still need regulation because of externalities from polution. On the other hand even Shleifer doesn’t think markets are so inefficient that government can improve them by trying to guess when they diverge from “intrinsic values.” So I see the EMH as un-related to left/right arguments about regulation. Nobody thinks Soros and Buffett are going to work as federal regulators for $80,000 a year.

    I don’t agree about fundamentals. The price of a stock is the discounted value of the expected future dividends, adjusted for risk. That seems pretty fundamental to me. Gold is valued according to an expected flow of services provided (monetary or industrial/cosmetic.)

    Maybe we are more than 3% apart!

    TGGP. Rational expectations are far superior to adaptive when you have information that is not yet reflected in past data. I.e. , when some sort of sudden shock hits the economy.

    I agree, the EMH is perfectly consistent with fat tails.

  24. Gravatar of David Smith David Smith
    4. February 2010 at 19:37

    You saw, didn’t you, that elsewhere in the theoretical universe, random walk professor Burton Malkiel is launching a China hedge fund.

    Someone up a few comments back seemed exercised that correlation approached one on a big down day as the intial claims were off by 6,000. That’s not really all it was, of course. It was that, + PIIGS, + China, +Volcker, + Budget, + Jon & Kate and whatever else. But that issue of increasing correlation within and across markets in downturns over time really is an interesting one. Has it been studied to death yet?

  25. Gravatar of Winton Bates Winton Bates
    4. February 2010 at 23:13

    The EMH seems to me to have the same status as the law of one price. If there are enough other people out there shopping around looking for a bargain, then we can assume prices will be the same in all shops and there is no point shopping around. But we all know that prices are not the same in all shops – even economists can sometimes find bargains if they look hard enough for them.
    The best excuse for not shopping around is high transactions costs.

  26. Gravatar of Quote Of The Day « Thinking Things Through Quote Of The Day « Thinking Things Through
    5. February 2010 at 04:33

    [...] via TheMoneyIllusion » Heads I win, tails you lose. [...]

  27. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    5. February 2010 at 04:48

    Scott said:
    “Chad, I like the analogy to physics. To me the EMH is like Newton’s laws; not precisely true, but close enough to give useful predictions.”
    Just like you should apply Einstein’s laws when doing astrophysics, you should also keep in mind the impact of noise traders when designing macro policies.

    “My point is that it is silly to argue “if genius X succeeds, it will disprove the EMH, and if he fails, it will also disprove the EMH.””
    Shleifer’s point was a little bit different. He said that noise traders have caused the mispricing of Royal Dutch Shell, and this is what proves EMH is wrong. And there is no wonder that some informed investors win despite noise trader risk, and some lose because of it. A direct consequence of noise trader risk hypothesis is a balance of successful and unsuccessful informed investors. If all informed investors were unsuccessful, then we would get the stone age, if all informed investors were successful, we would get EMH.

    “I think the EMH is useful for government regulators, as even Shleifer says the government isn’t smarter than the markets. And it’s useful for ordinary investors, who are better off in index funds.”
    Yes, ordinary investors are better of in index funds, except for those richer individuals for whom it is better to replicate market portfolio themselves by using some randomization technique and save index fund costs.
    No, the EMH is harmful for governemt regulators. They should be aware of the harm of noise trader risk, and take care that they themselves do not become a source of noise. EMH trivialises the harm of Fannie and Freddie.

    “123, You missed my point. I agree that logically it doesn’t refute the EMH at all. I am saying that it would be used by the anti-EMH crowd to refute the EMH, regardless of the outcome. You might argue that if a smart anti-EMH person keeps getting rich by taking advantage of market anomalies, that this fact it doesn’t support the anti-EMH position. But believe me, there are plenty of people who will argue otherwise, many have on this very blog.”
    Anti-EMH position predicts that there is a probability distribution, according to which some smart anti-EMH persons get rich by taking advantage of market anomalies, and some get poor.

    Eli said:
    “LTCM was writing naked puts on volatility. It turns out the market had priced volatility correctly, and LTCM lost a lot of money. This is completely consistent with the EMH.”
    Good point, but I have a little bit different interpretation. LTCM priced volatility as if EMH is true. Because EMH is false (Shiller proved that volatility of the market is larger than volatility of fundamentals), volatility spiked and LTCM crashed.

  28. Gravatar of What Is A Bubble? «  Modeled Behavior What Is A Bubble? «  Modeled Behavior
    5. February 2010 at 06:08

    [...] efficient market hypothesis seems to stem from a disagreement about what constitutes identifiable. Scott Sumner, and a lot of other EMH proponents, define bubbles as identifiable only if the method of [...]

  29. Gravatar of Mr. E Mr. E
    5. February 2010 at 06:48

    Scott,

    They blew up because they had too much leverage. They didn’t know about Kelly betting because they were mostly fixed income guys.

    With proper leverage, LTCM would still be here today. Yes, their returns would be only 8-12% a year, but they wouldn’t have blown out in the crisis.

    You’re making a huge logical mistake. Huge. You are mistaking returns with trading strategy success. This is one of the biggest fallacies of non-expert traders – that the returns of a strategy is only dependent on the quality of the trading idea behind it, and has nothing to do with risk management. In fact, it is the other way around. The risk management is far more important than the strategy.

    In fact, John Henrys strategy is simply risk management, with some simple rules.

    I can prove this to you very easily. Here is a gambiling game we can play. Heads I win 2 dollars, tails I lose 1 dollar. I can’t lose at this game right? Wrong.

    If I bet more than 50% of my entire account, I will inevitably go broke. I will go broke with 100% odds, it is just a matter of how long it takes to go broke. This part is not perfectly known, but can only be estimated.

    LTCM had 250 to 1 leverage. It was just a matter of time before they went broke – no matter how good the trading strategy they had.

    I suggest that Noble prizes in economics are almost worthless when looking at risk management, and probably counter productive. Especially when one of them got a prize for using a normal distribution as a proxy for asset return distribution.

    Because this crew of people came out of the Banks and traded fixed income and equities, they didn’t know really anything about risk management, either position, portfolio, or account risk management. There is a very good reason why so many billionaire traders started in commodities and futures – I will leave that to you to figure out.

  30. Gravatar of Lord Lord
    5. February 2010 at 06:59

    Just my point. If you look at retrospective studies of what stocks should have been worth given future earnings and dividends and their actual prices, you see how irrational a 20% swing is, yet how extremely common it is. I do think prices efficiently represent our future beliefs, but those beliefs are not rational but swayed heavily by emotions.

  31. Gravatar of Dan Dan
    5. February 2010 at 07:12

    Lord says, “Like Yglesias, if it were called the unpredictable market hypothesis, it would likely both be more widely accepted and less controversial. The problem is efficiency is equated to rationality but if markets were rational prices would not be subject to tectonic changes, nor would they vary widely from fundamentals. ”

    I agree, an UMH would be better. However, the definitions of the EMH (strong and semi-strong) preclude that: that information, public and/or private, is incorporated instantaneously into prices. The implication strongly supports the notion of rationality – a definition that includes incorporating information into prices in an irrational or erratic manner would be, well, counter-intuitive.

    Scott says, “Dan, Few social science theories are falsifiable. But I find it extremely useful in my research. For instance, when I look at the impact of news on equity prices, the EMH allows me to focus on the immediate effect, as news should be immediately incorporated into equity prices. That makes it 100 times easier to isolate the market response.”

    While I am not intimately familiar with the expansive body of social science theory, I will go out on a limb. Most social science theories qualify as theories because they can be tested and potentially contradicted by the evidence. The structure of the EMH formulation precludes contradiction by the evidence.

    I should clarify: I am not arguing that information flow does not have an impact on prices; it clearly does. However, whether that impact is asymetric or of a magnitude that is not rational is where the debate lies. What does a company missing earnings estimates by a few pennies really say about the long-term “intrisic value” of the company? (or do we believe that companies have no intrisic value, only socially constructed value, in which case stocks are nothing more than poker chips?). I do believe that for large companies, stocks are mostly efficient with respect to information flow in the short run – it is not usually possible to consistently outsmart the market with respect to making predictions, without the expenditure of huge amounts of resources.

    TGGP says, “Dan: it sounds like you are saying for the anti-EMH position that fundamentals have nothing to do with prices, which is why they can’t predict future prices. So then why bother with talk about “fundamentals” at all? Also, does the EMH really predict normal distributions rather than “fat tails”? Fama, who more than anyone is associated with EMH, agrees with Mandelbrot on the issue of tails. He thinks they actually form stable distributions, of which normal distributions are a subset.”

    I should clarify that as well: the point of the illustration was to highlight the fact that there are alternative explanations for the observation I mentioned. I used an extreme one to make my point. I believe that stocks converge to fundamentals over time, but in the short run are moved by exogenous liquidity flows in the aggregate and a handful of large players on a single issue level (either directly or by proxy). For example, a stock went to 1/2 free cash flow last year because a single owner was forced to sell.

    Regarding “fat tails,” I agree that EMH does not force the use of a normal distribution, just that it is normally associated with it. “Fat tails” are difficult (but not impossible) to reconcile with a notion of rationality implied by the EMH.

  32. Gravatar of Mr. E Mr. E
    5. February 2010 at 07:28

    Then the EMH:

    It isn’t true. Support:

    1. Robert Haugen work.
    2. Robert Shiller’s 10 year P/E ratio work. This is probably the work the above people were thinking about. For God’s sake – it is right on the EMH wikipedia entry.

    Then there are the “anomalies”. The word anomaly is a mealy mouthed way of saying – “my theory is wrong but I don’t want to admit it.” I have a theory the sun comes up in the west. When the sun comes up in the east, should I call it an anomaly?

    There list of published anomalies is long. But, not many have been published in the last decade. EMH guys assume thats because there isn’t any. This is the wrong assumption.

    A much more likely – and rational – explanation for the lack of published anomalies is: “anomalies” are no longer published as it has become very clear you can make huge amounts of money by exploiting them. Why pass up the chance to make several million dollars? This would not be rational behavior at all. Especially when your proof the theory is wrong will be ignored.

  33. Gravatar of Mr. E Mr. E
    5. February 2010 at 07:48

    Scott,

    I guess what I am trying to say about the EMH is that you can either have the EMH – which looks very unlikely with the published data, or you can have rational agents.

    In today’s world, rational agents would not publish their anomalies.

    You really cannot have both, and you probably get neither. Given the number of “anomalies” that were published at a pretty good clip in the past, and the seeming desert of “anomalies” we face today, it is almost assuredly the case that:

    1. We don’t have rational agents – because rational agents would not publish a found anomaly. Since we have many published anomalies, people do not always act in utility maximizing fashion. Rational agents are gone.

    2. Some People that are utility maximizing are finding anomalies and not publishing them. EMH is gone

  34. Gravatar of Lord Lord
    5. February 2010 at 10:23

    If one interprets information in an objective sense, then it would imply rationality, but if one interprets it as future expectations, and especially as expectations of others expectations, then there can be no objectivity nor rationality because we are not considering the past but the future, all information concerns the past, and we are free to imagine any possible future. As future becomes past these expectations can swing wildly and imaginations can trample the hardiest of fundamentals. Imaginations are free to soar and plummet even as fundamentals keep up tethered to the realities of the past.

  35. Gravatar of rob rob
    5. February 2010 at 12:02

    Does the EMH purport to concern simply information (public or private) that is *known to some person/trader*, or all (knowable?) information regardless of whether it is known to anyone? Or something in between?

  36. Gravatar of jb jb
    5. February 2010 at 13:00

    Mr. E.

    “The inefficiency in the pricing of Royal Dutch and Shell is a fantastic embarrassment for the efficient markets hypothesis because the setting is the best case for that theory.”

    No, it is not an embarassment to EMH. The Royal Dutch / Shell is an anomaly, that is, an instance of something that conflicts with accepted theory. To suggest that we throw out EMH because an instance has emerged that appears to refute it is to reject the scientif method. I always thought that the challenge to the economist is to try to explain the anomaly within the context of what you think you know, in this case EMH. Many interesting explanations for the Royal/Shell anomaly have been proposed that are consistent with EMH. But even if none of these are satisfactory EMH remains the only rational model for explaining how capital is priced UNTIL A BETTER THEORY REPLACES IT. Nothing even remotely as strong as EMH has emerged as far as I know.

    I love Steve Lansburg’s explanation: “Imagine a physicist, well versed in the laws of gravity, which he believes to be excellent approximations to the ultimate truth. One day he encounters his first helium-filled balloon, a blatant challenge to the laws he knows so well. Two courses are open to him: He can say, “Well, the laws of gravity are usually true, but not always; here is one of the exceptions”. Or he can say, “Let me see if there is any way to explain this strange phenomenon without abandoning the most basic principles of my science”. If he takes the latter course, and if he is sufficiently clever, he will eventually discover the properties of objects that are lighter than the air and recognize that their behavior is in perfect harmony with existing theories of gravity. In the process, he will not only learn about helium-filled balloons; he will also come to a deeper understanding of how gravity works.

    Now it might very well be that there are real exceptions to the laws of gravity, and that our physicist will one day encounter one. If he insists on looking for a good explanation without abandoning his theories, he will fail. If there are enough such failures, new theories will eventually arise to supplant the existing ones. Nevertheless, the wise course of action, at least initially, is to see whether surprising facts can be reconciled with existing theories. The attempt itself is good mental exercise for the scientist, and there are sometimes surprising successes. Moreover, if we are too quick to abandon our most successful theories, we will soon be left with nothing at all.”
    The Armchair Economist Steven Landsburg, p. 12 1993

    What is your alternative to EMH? Without one, it stands.

  37. Gravatar of david glasner david glasner
    5. February 2010 at 13:24

    Scott, I’m not an expert on Shiller’s work on excess volatility, but I don’t think it’s fair to say that what he did was data mining. He showed that stock market prices fluctuate much more than the dividends that they are supposed to reflect. Why is that data mining?

    I used to think that Keynes’s Chapter was 12 was not just weak, but an embarrassment. I don’t necessarily agree with it now, but it is not totally daft. But you cleverly avoided answering my question by bringing up his idiotic point about the value of ice companies in the summer time. My question was aren’t Keynes’s theory of stock prices, which you disdain, and EMH observationally equivalent in that they both imply that stock prices will move in an unpredictable fashion.

    You said:

    “I don’t agree about fundamentals. The price of a stock is the discounted value of the expected future dividends, adjusted for risk. That seems pretty fundamental to me.” Do you mean that the price of a stock is equivalent to the discounted value of the expected future dividends. I hope not, then you would be arguing by definition and speaking in tautologies. You have a theory of stock market prices in which people are trying to predict future dividends and discount them back into the present. Those expectations/predictions are not the same as the realized values and the very expectations may affect the actual outcomes. So no, it is not fundamental at all, it is as ethereal and subjective as could be.

    You said:

    “Gold is valued according to an expected flow of services provided (monetary or industrial/cosmetic.)”

    Please tell me what has happened to the expected service flow generated by gold over the past two years to cause a 40 percent annual rate of appreciation.

    You said:

    “Maybe we are more than 3% apart!”

    We are obviously way more than 3 percent apart on this one, but you have to adjust for the frequency with which we disagree. So far we only disagree about Chinese sterilization of FX inflows and EMH.

  38. Gravatar of JTapp JTapp
    5. February 2010 at 19:13

    Even guys like Shiller have to believe in SOME form of market efficiency, otherwise they wouldn’t be designing housing price futures markets and other derivative markets with predictive value for asset prices.

    I think the general teaching in finance right now is that the EMH mostly holds but large hedge funds make the market less efficient because they’re moving markets with their large positions and are very opaque in design.

  39. Gravatar of Mike Sandifer Mike Sandifer
    5. February 2010 at 20:09

    Dan,

    You said:

    “The EMH remains a hypothesis and never graduated to a theory precisely because it is not falsifiable. Which is why the alternative, the “anti-EMH” is also non-falsifiable. Every piece of empirical evidence can be interpreted either as supporting the theory or refuting the theory.”

    I see EMH as being the null hypothesis, as one must demonstrate non-randomness and serial dependence in the first place.

  40. Gravatar of JP Koning JP Koning
    5. February 2010 at 21:42

    Bah, Andrew Lo and adaptive markets. Guys like Hayek and Brian Arthur at the Santa Fe Institute said all that ages ago. Lo just slapped a logo on it and gave it a name.

  41. Gravatar of ssumner ssumner
    6. February 2010 at 07:43

    David, Yes, it was a lot of things. Glad to hear about the China fund.

    Winton, That’s a good analogy.

    123, How do noise traders affect macro policy? Does it affect the optimal NGDP target rate?

    The harm of F&F was that they existed, I’ve opposed them from the beginning, but that has nothing to do with inefficient markets, they aren’t a free market institution. Of course if they exist they should be regulated.

    The bottom line is that LTCM was founded and operated on the premise that the EMH was false. If people thought it was true, it would never have been created. It was a natural experiment that tested whether the EMH was false. And they lost. It may have been due to the fact that two Nobel prize winning economists are stupid, and a smarter group of people could have done better, but I doubt it.

    Mr. E, You say LTCM was stupid? See my previous answer to 123. Hope springs eternal—if only they had done it right . . . But these are the smartest finance guys on the planet.

    Lord, Obviously that is possible. But until humans are replaced by Vulcans, all we can say is humans do the best that can, and that is efficient in terms of what humans are capable of.

    Dan, I agree with the first part of your comment, but not this:

    “While I am not intimately familiar with the expansive body of social science theory, I will go out on a limb. Most social science theories qualify as theories because they can be tested and potentially contradicted by the evidence. The structure of the EMH formulation precludes contradiction by the evidence.”

    Social science theories are not easy to refute. This is because no one expects them to work perfectly. How bad is bad? There is one natural test of the EMH. If it is false, then excess returns at mutual funds should be serially correlated. But they don’t seem to be, or perhaps only to a tiny extent. So maybe its a tiny bit false.

    I don’t see any problem at all reconciling fat tails with the EMH. We have fat tails because real world shocks have fat tails.

    Mr. E. As you may know most anomaly studies are worthless. They are a misuse of statistics. There are millions of anomalies in Vegas. Does that mean the roulette wheels are fixed?

    People publish anomalies to get tenure. They don’t think they are giving away valuable information, because deep down they don’t think the anomalies are real.

    Lord#2, Tht’s pretty philosophical. The bottom line is that I find the EMH useful. Why that is I don’t know, but it seems to explain a lot to me.

    rob, There are different versions of the EMH, some include private info, some don’t.

    jb, I like that analogy. Steven Landsburg is very smart.

    David, I think the theories are only observationally equivalent if people believe fundamentals play no role at all. But almost all anti-EMH people (like Shiller) think asset prices tend to move back toward fundamentals over time, and thus that excess returns are partly predictable.

    Shiller’s studies are searches for statitical “anomalies” but even if the EMH was true, for every 20 million computer studies you do, you expect to find 1 million anomalies, signficant at the 95% level. He is constructing models to explain the past! That is incredbly easy to do. I could create statistically signficant models showing how to beat a given roulette wheel, if you gave me the data produced by the last 100 spins of the wheel.

    My point was that stock prices are based on forecasts of fundamentals. I don’t see the problem with that.

    You asked:

    “Please tell me what has happened to the expected service flow generated by gold over the past two years to cause a 40 percent annual rate of appreciation.”

    All you need to do is read Hayek to realize that question is unanswerable. But I’ll speculate. There are recent articles about a sudden slowdown in new gold discoveries in the last couple years, and great pessimism about future supplies. This is combined with huge increases in gold demand in booming China and India, as well as prospective increases as those countries get richer over time. If supply is inleastic, the price of an infinitely lived zero depreciation asset can be very unstable if demand is suddenly expected to grow slightly more rapidly.

    JTapp, Good point.

  42. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    6. February 2010 at 14:59

    “123, How do noise traders affect macro policy? Does it affect the optimal NGDP target rate?”
    For small and medium sized countries currency noise traders are very important for macro stability. For US noise traders are important for supervision of banking system and for proposed macroprudential regulation.

    “The harm of F&F was that they existed, I’ve opposed them from the beginning, but that has nothing to do with inefficient markets, they aren’t a free market institution. Of course if they exist they should be regulated.”
    Conventional analysis calculates the harm of F&F by estimating tax & subsidy related deadweight losses. This underestimates the total harm of F&F as F&F are huge noise traders and they severely diminish the efficiency of credit markets. This diminished efficiency of credit market means that other private non-F&F sector credit transactions happen at less efficient prices.

    “The bottom line is that LTCM was founded and operated on the premise that the EMH was false. If people thought it was true, it would never have been created. It was a natural experiment that tested whether the EMH was false. And they lost. It may have been due to the fact that two Nobel prize winning economists are stupid, and a smarter group of people could have done better, but I doubt it.”
    The real bottom line is that LTCM was founded on premise that noise trader risk is not important. Failure of LTCM supports Shleifer’s noise trader risk thesis, success of LTCM would have supported theories that emphasise other sources of market inefficiency..

    “Mr. E, You say LTCM was stupid? See my previous answer to 123. Hope springs eternal—if only they had done it right . . . But these are the smartest finance guys on the planet.”
    It is widely agreed that LTCM has severely underpriced risk, and many smart people have called them stupid for that. LTCM were smart in other areas, but not in the pricing of risk.

  43. Gravatar of david glasner david glasner
    7. February 2010 at 09:02

    Scott

    You wrote:

    “I think the theories are only observationally equivalent if people believe fundamentals play no role at all. But almost all anti-EMH people (like Shiller) think asset prices tend to move back toward fundamentals over time, and thus that excess returns are partly predictable.” That’s a good point. However, since I deny that there is any clear distinction between fundamentals and non-fundamentals, I am not so sure that that answer really works, but it is worth some more thought. Just to elaborate on fundamentals, what I am saying is that people have all kind of different expectations about what will happen in the future and that these expectations relate to future demand and cost conditions. But future demand and cost conditions also depend on current expectations. So when you talk about fundamentals it sounds as if there is some objective underlying reality out there that already exists and people are just groping to try to find it and they gradually gravitate toward it. That is what I reject. Another way to think about it is whether there is a unique general equilibrium or there are multiple general equilibria. If there are multiple general equilibria then changes in expectations can change the equilibrium towards which the economy is moving. Why is one of those potential equilibria any more “fundamental” than any of the others. If no single equilibrium is more special than any other, what meaning can be attached to the term “fundamentals” in the context of EMH?

    You wrote:

    “Shiller’s studies are searches for statitical “anomalies” but even if the EMH was true, for every 20 million computer studies you do, you expect to find 1 million anomalies, signficant at the 95% level. He is constructing models to explain the past! That is incredbly easy to do. I could create statistically signficant models showing how to beat a given roulette wheel, if you gave me the data produced by the last 100 spins of the wheel.”

    Shiller’s 1981 paper “Do Stock Prices Move Too Much to be Justified by Subsequent Changes in Dividends?” AER 1981, derives a theoretical bound on the volatility of observed stock prices in terms of the volatility of dividends. Theoretically, the volatility of dividends should be greater than the volatility of stock prices, but the volatility of stock prices was 31 times greater than the volatility of the dividends. Is that data mining? Also see Shiller’s article on Excess Volatility in the 1997 volume that I edited Business Cycles and Depressions: An Encyclopedia.

    You wrote:

    “All you need to do is read Hayek to realize that question [what has happened to the expected service flow generated by gold over the past two years to cause a 40 percent annual rate of appreciation] is unanswerable.”

    Hey, who’s this Hayek dude anyway? Didn’t you just post some rap video in which he makes a walk-on appearance?

    “But I’ll speculate. There are recent articles about a sudden slowdown in new gold discoveries in the last couple years, and great pessimism about future supplies. This is combined with huge increases in gold demand in booming China and India, as well as prospective increases as those countries get richer over time. If supply is inleastic, the price of an infinitely lived zero depreciation asset can be very unstable if demand is suddenly expected to grow slightly more rapidly.”

    Interesting speculation, but what are the real services that gold is providing and have those services become more valuable? My point is that the forces governing movements in the price of gold are so far removed from anything that corresponds to “fundamentals” (a point that in your own way you seem to concede by citing that Hayek dude) that it just doesn’t make sense to me to talk about it in those terms. Just to reiterate, I think that people are guessing about what the future price will be based on all kinds of factors that have nothing to do with the underlying service flows generated by gold, which to me account for only an insignificant part of the total stock of gold which is being held entirely in the expectation of future appreciation. And by the time the world comes to an end, it will all be worthless, so somebody is going to be taking a huge capital loss on all that gold.

  44. Gravatar of ssumner ssumner
    7. February 2010 at 12:04

    123, You said;

    “For small and medium sized countries currency noise traders are very important for macro stability. For US noise traders are important for supervision of banking system and for proposed macroprudential regulation.”

    I thought you meant the existence of noise traders had to be offset by policy. That they were a rogue element that was out of control.

    We are just going around in circles with LTCM. I do understand how others can say they did it the wrong way. But the whole anti-EMH argument seems to be that smart people like Soros can beat the markets. Maybe these guys weren’t smart enough in the end. I can accept that. But it tells me the market is pretty damn efficient if even they couldn’t beat it.

    David: There is a lot of debate in macro about multiple equilibria, but the models are all over my head, so I won’t comment.

    You said;

    “Theoretically, the volatility of dividends should be greater than the volatility of stock prices, but the volatility of stock prices was 31 times greater than the volatility of the dividends. Is that data mining? Also see Shiller’s article on Excess Volatility in the 1997 volume that I edited Business Cycles and Depressions: An Encyclopedia.”

    My response is that his model is probably wrong. He may have assumed a constant rate of discount.

    It is possible (indeed I’d say likely), that there will be zero speculative demand for gold long before the world comes to an end. So I don’t anticipate any big capital losses. Ironically, Nick Rowe just did a post asking why the value of land isn’t infinite. He claimed it should be if the real growth rate equals the real interest rate, and the demand for land is unit elastic. The same argument applies to gold. Now you are asking why gold is so valuable, as the service flow seems small. I think the truth is somewhere in between. The price of gold isn’t infinite, but it’s near zero depreciation gives it an extremely high value even if the service flow is modest.

    I believe 100% of the value of gold is for its expected future service flow. Sure sume people invest in it, but only because it is assumed that it can later be used for jewelry, etc. If it had no real world uses, no one would use it as a hedge against infaltion. It would be like pre-USSR Russian currency, totally worthless.

  45. Gravatar of david glasner david glasner
    8. February 2010 at 10:43

    Scott,

    You said:

    “There is a lot of debate in macro about multiple equilibria, but the models are all over my head, so I won’t comment.”

    Another clever evasion on your part, Scott. If I can talk about the multiple equilibria without knowing the literature, why can’t you?

    “His [Shiller's] model is probably wrong. He may have assumed a constant rate of discount.”

    Shiller did use a constant rate of discount to get the result that stock prices were 31 times too volatile. When he adjusted for varying rates of discount, he found that lo and behold they were a mere 17 times too volatile. Again, I am not up on the literature on excess volatility either, but I don’t think that anyone has called his basic result into question. Like you, people come up with various rationalizations to explain away his result, but as far as I can tell the result still stands.

    “It is possible (indeed I’d say likely), that there will be zero speculative demand for gold long before the world comes to an end. So I don’t anticipate any big capital losses.”

    In other words, you think it is likely that people in the future will discover sufficiently valuable new uses for gold to cause it to appreciate so much in the future that holders of gold today will be compensated for holding the gold until those new uses are discovered. At that point, they will sell their gold to be put to those new uses. Now be honest, do you really believe that?

    “Ironically, Nick Rowe just did a post asking why the value of land isn’t infinite. He claimed it should be if the real growth rate equals the real interest rate, and the demand for land is unit elastic. The same argument applies to gold. ”

    The same argument doesn’t apply to gold, because gold is homogeneous, and land is not. Some land is generating very high locational rents, and some land is generating no rents at all. So the land generating no rents is now pretty much worthless except in those cases where undeveloped land is being eyed for future development. Because there is a surplus of land, it is not clear why the locational rents accruing to already developed land should continue increase at a rate equal to the overall rate of growth. So I don’t see why any land should have infinite value.

    “Now you are asking why gold is so valuable, as the service flow seems small. I think the truth is somewhere in between. The price of gold isn’t infinite, but it’s near zero depreciation gives it an extremely high value even if the service flow is modest.”

    Zero depreciation can’t give a worthless asset any value.

    “I believe 100% of the value of gold is for its expected future service flow.”

    Really? Is that why most of the gold in the world is sitting in the vaults of central banks?

    “Sure sume people invest in it, but only because it is assumed that it can later be used for jewelry, etc.”

    Central bankers are just waiting to sell their gold to be made into trinkets?

    “If it had no real world uses, no one would use it as a hedge against infaltion. It would be like pre-USSR Russian currency, totally worthless.”

    Well, not totally worthless, but close. I think that these gold bugs think that the world is going to hell in a handbasket, that all the world’s currencies are going to collapse, and that gold will again become the world’s only or dominant currency. They even do some crazy calculations converting all outstanding dollar currency into the available stock of gold, and derive the implied dollar price per ounce. It come to something over (maybe way over) $6000 an ounce. That’s the rational expectation that governs gold investment these days. My goodness!

  46. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    8. February 2010 at 16:55

    “I thought you meant the existence of noise traders had to be offset by policy. That they were a rogue element that was out of control.”
    It is not possible to offset noise traders by policy. But institutions matter, if short selling is prohibited, how can the information held by short seller can be reflected in the price?

    “We are just going around in circles with LTCM. I do understand how others can say they did it the wrong way. But the whole anti-EMH argument seems to be that smart people like Soros can beat the markets. Maybe these guys weren’t smart enough in the end. I can accept that. But it tells me the market is pretty damn efficient if even they couldn’t beat it.”
    Current anti-EMH argument is that both smart people and noise traders can beat the markets. When noise traders win, prices are wrong, market becomes a lottery and no Nobel will help you win the powerball. But I have seen no Efficient Powerball Hypothesis.

  47. Gravatar of scott sumner scott sumner
    9. February 2010 at 07:23

    David, You said;

    “Again, I am not up on the literature on excess volatility either, but I don’t think that anyone has called his basic result into question. Like you, people come up with various rationalizations to explain away his result, but as far as I can tell the result still stands.”

    People don’t question the accuracy of his statistics, they question his interpretation. Again, if you went to Vegas and just wrote down the first 100 numbers from a roulette wheel I could easily use “statistics” to prove they are not random. Whenever you are examining a past data set, it will never look random. There are almost an infinite number of models you can try to fit to the data, so I’m not surprised that Shiller was able to find a model inconsistent with the EMH that fit the data pretty well.

    People need to ask themselves some basic questions. For instance, the US is itself an outlier, with a much more stable economy than almost any other country in the world. 3% growth for over 150 years. If people didn’t know, ex ante, that the US was going to be a “winner” then you would expect excess volatility in the US even if the EMH was true. I am certainly not claiming that explains Shiller’s results, but it is part of the story. And there may be lots of other factors. Does he account for changing real tax rates on capital, for instance? What if his estimates of risk aversion are wrong? If volatilty is 17 fold too high, why don’t mutual funds take advantage of all those free dollars lying on the table? Why weren’t the anti-EMH guys screaming BUY last March, when the market severely overreacted on the downside? (Roubini said don’t buy last spring.)

    You said;

    “In other words, you think it is likely that people in the future will discover sufficiently valuable new uses for gold to cause it to appreciate so much in the future that holders of gold today will be compensated for holding the gold until those new uses are discovered. At that point, they will sell their gold to be put to those new uses. Now be honest, do you really believe that?”

    You shouldn’t jump to conclusions. I never said what you claim here. It may well be that no new uses are found for gold. But even so, I could easily imagine a 5 fold increase in the number of people who are able to afford gold jewlery over the next 100 years. Nor did I claim that gold prices would never decline and that people would never suffer capital losses. I think gold price declines are almost certain at some point in the future, as are further increases. I said I didn’t think the value would fall to zero. I would never claim that the current price of gold will seem correct, ex post. What I do claim is that it isn’t ovious whether it is currently too high or too low.

    Nick was referring to oceanfront land, not all land. He assumed it had an income elasticity of 1. I.e.,that it was a positional good.

    I said;

    “Now you are asking why gold is so valuable, as the service flow seems small. I think the truth is somewhere in between. The price of gold isn’t infinite, but it’s near zero depreciation gives it an extremely high value even if the service flow is modest.”

    You responded;

    Zero depreciation can’t give a worthless asset any value.

    Nor did I make that claim, in fact I made exactly the opposite claim, if you will reread what I wrote. The claim was that gold was not worthless, but had a positive service flow.

    I said;

    “I believe 100% of the value of gold is for its expected future service flow.”

    You responded:

    Really? Is that why most of the gold in the world is sitting in the vaults of central banks?

    There is no conflict there at all. Why would central banks hold an commodity that didn’t have intrinsic value. Surely you are not saying central banks are indiferent between holding gold bars or mud cakes? As you know, the causation ran this way—first gold became valuable for jewelry, etc, then central banks decided to hold it as reserves. But they only hold it because it has intrinsic value. In the long run Keynes’ barbarous relic comment will be right, and gold will be completely demonetized. So why don’t central banks sell it off now, while the going is good? Because they think future demand for jewelry will be high enough that even when to world goes all electronic, and gold has no monetary value, then they will be able to sell it off for private uses at that point in the future.

    Regarding your last point, gold bugs aren’t a big enough part of the market to explain high prices of gold. The same applies to diamonds by the way. My model can explain the high price of diamonds. The gold bug model cannot, as no one expects we will return to a diamond standard.

    123, You said;

    Current anti-EMH argument is that both smart people and noise traders can beat the markets.

    There’s the problem right there. “Current” arguments. The argument has to be revised each time. The anti-EMH argument is a backward looking argument, and hence useless. I am not in the slightest impressed by the fact that some investors are luckier than others, or that people are able to fit patterns to past data. When new data comes along these anti-EMH models often can’t explain it, so they have to revise their models. But they are still backward-looking.

    Here is the problem:
    1. Anti-EMH types think all they need to do is say “Look! I found a pattern”
    2. pro-EMH types like me think the question is whether mutual fund excess returns are serially correlated. If smarter investers do better on average, then they should be. But the effect is so trivial that it is hard to take advantage of, hence the EMH is roughly true. Not precisely true, BTW, otherwise people would have no incentive to gather information.

  48. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    10. February 2010 at 06:55

    The finding that mutual fund industry adds no significant excess return value is very old. But this finding only tells us that it is very hard to pick a good mutual fund, but not that market prices are efficient.
    The only think Shleifer did was to take very old bits from Keynes and build a formal model from them.
    Respectable anti-EMH position is that in practice deviations from the assumptions of EMH (unlimited leverage, shorting, complete markets etc.) have significant effects. Prices are frequently wrong, but it is harder to exploit that than it looks because of the noise trader risk.

  49. Gravatar of 123 – TheMoneyDemand 123 - TheMoneyDemand
    10. February 2010 at 07:18

    Noise trader risk thesis is also supported by the fact that Buffet minimizes noise trader risk by having AAA balance sheet and zero risk of investor redemptions, and Soros who has no such luxures has adapted his trading strategy to noise traders (i.e. he often likes to ride the bubbles up).

  50. Gravatar of David Glasner David Glasner
    10. February 2010 at 10:04

    Scott,

    On Shiller and excess volatility

    I am puzzled by your criticisms of Shiller. On the one hand you accuse him of data mining and suggest that it is easy to find a data set that ex post has what seem to be pricing anomalies in it. Thus, you said:

    “If you went to Vegas and just wrote down the first 100 numbers from a roulette wheel I could easily use “statistics” to prove they are not random. Whenever you are examining a past data set, it will never look random. There are almost an infinite number of models you can try to fit to the data, so I’m not surprised that Shiller was able to find a model inconsistent with the EMH that fit the data pretty well.”

    But that is not at all what Shiller did in his excess volatility paper (he may have in his other work, but that’s not what I’m concerned with). He theoretically derived an implication about the relationship between stock prices and dividends if EMH is true. He then looked at the data and saw that the data are massively, overwhelmingly, inconsistent with the relationship between stock prices and dividends implied by EMH. Then you said:

    “People need to ask themselves some basic questions. For instance, the US is itself an outlier, with a much more stable economy than almost any other country in the world. 3% growth for over 150 years. If people didn’t know, ex ante, that the US was going to be a “winner” then you would expect excess volatility in the US even if the EMH was true.”

    Sorry, I can’t figure out what you are trying to say.

    “I am certainly not claiming that explains Shiller’s results, but it is part of the story. And there may be lots of other factors. Does he account for changing real tax rates on capital, for instance? What if his estimates of risk aversion are wrong?”

    Look, Shiller published his excess volatility paper in 1981. EMH people have thus had almost 30 years to show that his result depends on some special assumption about any of the factors that you mentioned. Don’t you think that if his results are as fragile as you suggest, some eager young Chicago grad student would have by now shown that a correct treatment of risk aversion or capital taxation makes Shiller’s result go away? If someone has written such a paper, it seems t o be a pretty well kept secret. That’s why I say the result still stands. Just because you can think up alternative ad hoc explanations for his result doesn’t mean that you can pretend the result doesn’t exist. Your rationalizations for the result are also testable. And if no one in the past 30 years has been able to show that his result is conditional on a particular assumption about discount rates, or risk aversion, or tax rates, that tells me that his result is pretty darn robust.

    “If volatilty is 17 fold too high, why don’t mutual funds take advantage of all those free dollars lying on the table? Why weren’t the anti-EMH guys screaming BUY last March, when the market severely overreacted on the downside? (Roubini said don’t buy last spring.)”

    Does EMH mean anything besides prices in period t+1 not being predictable based on prices in t-1, t-2, t-3 . . .? If not, it is a very weak but not useless concept that can’t even rule out Keynes’s beauty contest theory of asset pricing. If it does mean something more than unpredictability, your question about money left on the table does not address what is problematic about EMH?

    Scott, you said;

    “It may well be that no new uses are found for gold. But even so, I could easily imagine a 5 fold increase in the number of people who are able to afford gold jewlery over the next 100 years. Nor did I claim that gold prices would never decline and that people would never suffer capital losses. I think gold price declines are almost certain at some point in the future, as are further increases. I said I didn’t think the value would fall to zero. I would never claim that the current price of gold will seem correct, ex post. What I do claim is that it isn’t ovious whether it is currently too high or too low.”

    Okay, fine. Just tell exactly what are the fundamentals that govern the value of gold. I don’t think that there are any, I think the value of gold is based on total, utter, speculation about unknowable and unpredictable future demands
    for gold. My own hunch is that there will never be enough future non-monetary demands for gold to allow current holders of gold to earn a positive (much less competitive) return on their investment. Is there any argument based on “fundamentals” that disproves or even casts serious doubt on my hunch. I maintain that most holders of gold to the extent that they are governed by any implicit or explicit theory of the future are betting on an increased monetary role for gold in the future, not on new or increased real demand for gold.

    You said:

    “Why would central banks hold an commodity that didn’t have intrinsic value. Surely you are not saying central banks are indiferent between holding gold bars or mud cakes? As you know, the causation ran this way—first gold became valuable for jewelry, etc, then central banks decided to hold it as reserves. But they only hold it because it has intrinsic value.”

    Now I am going to put on my Austrian hat and repeat 5 times, “value is not intrinsic, it is subjective.” This is the correct and valid core of Austrian economics to which the rest of the profession pays (partial) lip service. There is nothing intrinsic about value, value comes from the subjective valuations (including expectations) that people assign to commodities.

    What you meant to say is that central banks hold valuable assets like gold not worthless hunks of matter. That’s a truism. The whole point of our discussion is whether increasing gold prices and increased holdings of gold by central banks over the last five years are a reflection of anticipations of an increased real service flow from the existing stock of gold or of a bubble like expectation of future appreciation of gold that is unrelated to any anticipation of an increased future real service flow.

    You said:

    “In the long run Keynes’ barbarous relic comment will be right, and gold will be completely demonetized. So why don’t central banks sell it off now, while the going is good? Because they think future demand for jewelry will be high enough that even when to world goes all electronic, and gold has no monetary value, then they will be able to sell it off for private uses at that point in the future.”

    This confident assessment of what central bankers are thinking is based on what?

    You said:

    “Regarding your last point, gold bugs aren’t a big enough part of the market to explain high prices of gold. The same applies to diamonds by the way. My model can explain the high price of diamonds. The gold bug model cannot, as no one expects we will return to a diamond standard.”

    And this confident assessment of how much current demand for gold derives from gold bugs is based on what? A quick look at historical trends shows that Since 2007, diamond prices have risen by about 40 percent and gold prices by 80 percent.

  51. Gravatar of scott sumner scott sumner
    11. February 2010 at 07:03

    123, Prices are often wrong? Ex ante or ex post? Everyone agrees ex post, but ex ante is much trickier. How do you define wrong?

    David; You said;

    “But that is not at all what Shiller did in his excess volatility paper (he may have in his other work, but that’s not what I’m concerned with). He theoretically derived an implication about the relationship between stock prices and dividends if EMH is true. He then looked at the data and saw that the data are massively, overwhelmingly, inconsistent with the relationship between stock prices and dividends implied by EMH.”

    We have no way of knowing that this is what he did. He might have noticed that the data looked too volatile, and decided to construct his model after noticing that fact. That’s what almost all anomaly studies are done. Try submitting a paper to a journal that has a study unable to find an anomaly. You have zer chance of getting it published, unless it disproves some previous anomaly paper. And that is precisely the problem, people actually seek out anomalies. Not surprisngly, they find them, as there are million out there. There real problem is that they then assume the existence of anomalies conflicts with the EMH

    You said;

    Sorry, I can’t figure out what you are trying to say.

    I was discussing a well known problem with anti-EMH studies. They mostly pick a very unrepresentative country. Suppose you picked Warren Buffett to see whether investors could beat the market more than 50% of the time. It would look like they can, but only because you picked someone who was very successful.

    Much of the excess volatility comes from the fact the US kept bouncing back from adversity. Russia never bounced back from 1918. Argentina never bounced back from 1929. Japan never bounced back from 1991. Part of the reason our stock market looks too volatile is that people have selected the one country that ex post proved very resiliant. But in the 1930s lots of people thought we were going toward socialism. Is it any wonder stocks were so low? (Too low in retrospect)

    You said;

    “Look, Shiller published his excess volatility paper in 1981. EMH people have thus had almost 30 years to show that his result depends on some special assumption about any of the factors that you mentioned. Don’t you think that if his results are as fragile as you suggest, some eager young Chicago grad student would have by now shown that a correct treatment of risk aversion or capital taxation makes Shiller’s result go away?”

    I think you misunderstand my point. I don’t question the actual statistics of the study, I question the interpretation. That’s not how the EMH should be tested. The right way to test it is to check whether one class of people consistently pick stocks better, or market time better, than another group of people. All Shiller has done is create his own model, which isn’t my model, and find it doesn’t fit the data. But that proves nothing to me. It simply suggests his model is wrong. Why is it wrong? How would I know, I’ve already mentioned that maybe he underestimates the extent of volatility in risk aversion. Or he ignores taxes. Or there could be a million other problems.
    If you were right that the study was beyond reproach, then why do so many finance types still believe in the EMH? They must have read Shiller’s study.

    You said;

    “Does EMH mean anything besides prices in period t+1 not being predictable based on prices in t-1, t-2, t-3 . . .? If not, it is a very weak but not useless concept that can’t even rule out Keynes’s beauty contest theory of asset pricing. If it does mean something more than unpredictability, your question about money left on the table does not address what is problematic about EMH?”

    Shiller doesn’t agree with you here. He claims (rightly in my view) that anti-EMH models imply you can beat the market, by following advice. That’s where I disagree with Shiller. I don’t think people can reliably beat the market, hence I think his anti-EMH model is wrong.

    You said:

    “Okay, fine. Just tell exactly what are the fundamentals that govern the value of gold. I don’t think that there are any, I think the value of gold is based on total, utter, speculation about unknowable and unpredictable future demands
    for gold.”

    On the supply side the most important fundamental is the cost of production. On the demand side, it is the demand for jewelry, and in the short term the demand for inflation hedges and central bank assets. But as I said, I expect those demands to fade over time, leaving mostly jewelry.

    I don’t think central bank demand for gold is rising because they’ve changed their views, rather I think it simply reflects a few newly emerging nations like China and India getting in on the action. As far as I know the developed country central banks have reduced their demand for gold.

    BTW, when I said “intrinsic value” I should have said “with non-monetary value” Then it is not a tautology, as central banks also hold assets with zero non-monetary value, like cash.

    You said;

    “This confident assessment of what central bankers are thinking is based on what?”

    Logic. Why would banks hold assets expected to have zero value in the future?

    BTW, I’ve always agreed there is some demand for gold from gold bugs, I simply argued that that fact alone can’t explain why gold is highly valuable. Diamonds show another asset that is very valuable, without the monetary backdrop. You seem to agree. If so perhaps we are getting back close to 3%. All we seem to disagree about is the percentage of gold demand that comes from gold bugs. There must be data on private gold hoards (excluding jewelry) as a share of the total world gold stock. Perhaps it is higher than I thought, but that would leave open the question of whether gold bugs held gold because of its expected future use as jewelry, or because they expect us to return to the gold standard.

  52. Gravatar of David Glasner David Glasner
    11. February 2010 at 18:11

    Scott,

    More on Shiller.

    I defended Shiller against your accusation of data mining, by summarizing the approach he took in his 1981 excess volatility paper. You responded:

    “We have no way of knowing that this is what he did. He might have noticed that the data looked too volatile, and decided to construct his model after noticing that fact.”

    Again he did not construct any model. He simply derived an implication of EMH about the relative volatility of asset prices and dividends. The implication is either validly deduced on invalidly deduced. Since you are not claiming that the implication is invalid, I am sorry to be so blunt, but everything else that you say is, ahem, beside the point.

    “That’s what almost all anomaly studies are done.”

    We’re not talking about almost all anomaly studies. We are talking about Shiller’s paper on whether the observed patterns of stock prices and dividends conforms to the deep structure implied by the basic logic of EMH.

    “Try submitting a paper to a journal that has a study unable to find an anomaly. You have zer chance of getting it published, unless it disproves some previous anomaly paper.”

    I view excess volatility as way more than an anomaly. Has anyone tried to check to see if stock prices since 1981 have exhibited less than volatility than before? I wonder what that would show. If stock prices in the US and in every other major stock exchange consistently exhibit excess volatility, what would you make of that?

    You said;

    “I was discussing a well known problem with anti-EMH studies. They mostly pick a very unrepresentative country. Suppose you picked Warren Buffett to see whether investors could beat the market more than 50% of the time. It would look like they can, but only because you picked someone who was very successful.

    Much of the excess volatility comes from the fact the US kept bouncing back from adversity. Russia never bounced back from 1918. Argentina never bounced back from 1929. Japan never bounced back from 1991. Part of the reason our stock market looks too volatile is that people have selected the one country that ex post proved very resiliant. But in the 1930s lots of people thought we were going toward socialism. Is it any wonder stocks were so low? (Too low in retrospect)”

    The question is not volatility. It is excess volatility. Everything that you mentioned should, under EMH, be reflected in both asset prices and dividends. The problem is the excess volatility of asset prices relative to the volatility in dividends. None of what you just said addresses that issue. That’s the deep structure that I referred to earlier.

    You said;

    “Look, Shiller published his excess volatility paper in 1981. EMH people have thus had almost 30 years to show that his result depends on some special assumption about any of the factors that you mentioned. Don’t you think that if his results are as fragile as you suggest, some eager young Chicago grad student would have by now shown that a correct treatment of risk aversion or capital taxation makes Shiller’s result go away?”

    “I think you misunderstand my point. I don’t question the actual statistics of the study, I question the interpretation.”

    I think you’ve done both. Otherwise, why did you mention varying discount rates, and tax rates, and measurement of risk aversion?

    “That’s not how the EMH should be tested.”

    I think that you basically view EMH as so fundamental, simply an extension of the rationality postulate, that since any empirical test is always a test of more than one hypothesis, you will always reject the auxiliary hypotheses that are also being assumed rather than question EMH.

    “All Shiller has done is create his own model, which isn’t my model, and find it doesn’t fit the data. But that proves nothing to me. It simply suggests his model is wrong.”

    You keep saying that and you are misrepresenting what he has done. He derived an implication — a deep implication — of EMH and tested it against the data. You have every right to dispute his derivation, but you chosen not to. If you aren’t willing to dispute his logic, don’t say that he has just created his own model. He is simply deducing logical implications of EMH and testing them against the data.

    “Why is it wrong? How would I know, I’ve already mentioned that maybe he underestimates the extent of volatility in risk aversion. Or he ignores taxes. Or there could be a million other problems.”

    If so, his result is not robust, and it could be reversed by adjusting his assumptions and replicating his test under the revised assumptions. Nobody, as far as I know, has done that. If nobody has done that, you can’t just dismiss his result with a wave of your hand.

    “If you were right that the study was beyond reproach, then why do so many finance types still believe in the EMH? They must have read Shiller’s study.”

    Well, maybe I’m not right, but I’m still waiting to hear a powerful criticism of his paper. I am assuming that your criticisms are about the best that are out there, but if you have a better one, please share it with me. In other words, I think that finance types, like you, do not regard EMH as a refutable hypothesis. As you have made clear, correct me if I’m wrong, you don’t regard most economic theories as refutable. I am not totally unsympathetic to that view, as I am less of a Popperian than I used to be.

    I asked;

    “Does EMH mean anything besides prices in period t+1 not being predictable based on prices in t-1, t-2, t-3 . . .? If not, it is a very weak but not useless concept that can’t even rule out Keynes’s beauty contest theory of asset pricing. If it does mean something more than unpredictability, your question about money left on the table does not address what is problematic about EMH?”

    You responded:

    “Shiller doesn’t agree with you here.”

    Believe it or not, I do not agree with Shiller on everything, so I am not shocked by this statement.

    “He claims (rightly in my view) that anti-EMH models imply you can beat the market, by following advice. That’s where I disagree with Shiller. I don’t think people can reliably beat the market, hence I think his anti-EMH model is wrong.”

    People do beat the market all the time. Sometimes, it is pure arbitrage and sometimes it is just a hunch. But according to Knight profit is the return to entrepreneurial speculation on uncertainty, which is by definition unquantifiable. Do you believe that the Knightian return to entrepreneurial activity is totally random or do you believe that some people have insights into market anomalies or the mispricing of assets that allow them to earn profits that in excess of the returns generated by holding the market portfolio? Do those insights emerge from the study of historical trends in market prices? No. Do those insights reflect a correct understanding that the current structure of market prices is in principle not sustainable (as it would be if market prices were efficient forecasts of future prices)? Yes.

    You wrote:

    “On the supply side the most important [gold] fundamental is the cost of production. On the demand side, it is the demand for jewelry, and in the short term the demand for inflation hedges and central bank assets. But as I said, I expect those demands to fade over time, leaving mostly jewelry.”

    First of all, the stock of gold is enormous relative to the current rate of output. So at any moment, only the expected future cost of production, not the current cost, is relevant to the current price. I think that the current gold price and the future time path of gold prices is determined not by fundamentals, but only by expectations of a future that is suffused in Knightian uncertainty. The notion that the information exists with which to fully list the fundamentals governing the future price path of gold, let alone insert that information into a model that would allow that path to be specified boggles the mind. I think that is true for most assets but especially for gold. I reject EMH because the notion that there are fundamentals which could be known that, if they were known, would determine an equilibrium value for assets does not reflect the extent of our ignorance about the future states of the world on which the current value of assets depends. Nor does it reflect the fact that our expectations of the future also can affect what the future will be, so that the idea that fundamentals have any objective existence is simply wrong.

    Scott, I’m sorry if this has turned into a harangue, but I am trying to articulate for myself just why I find EMH so unsatisfactory. I think that Shiller’s work has made me sensitive so some of these issues, but I think a careful working out of the implications of Hayek’s 1937 “Economics and Knowledge” paper must ultimately lead to a rejection of EMH on purely conceptual grounds.

    PS The main difference between diamonds and gold is the overhang of gold in the hands of the central banks for which there is no analog in the market for diamonds. And I don’t think that central banks historically have done a very good job of managing their asset portfolios, so I don’t think that your assumption that if they didn’t have a good reason for expecting the assets that they acquire or hold to appreciate they wouldn’t hold them has little basis in the historical record.

  53. Gravatar of ssumner ssumner
    12. February 2010 at 07:46

    David; You said;

    “Again he did not construct any model. He simply derived an implication of EMH about the relative volatility of asset prices and dividends. The implication is either validly deduced on invalidly deduced. Since you are not claiming that the implication is invalid, I am sorry to be so blunt, but everything else that you say is, ahem, beside the point.”

    You are still not getting my point. The EMH has 20 million implications. Statistical theory suggests that tests of the EMH should refute 1 million of those 20 million implications at the 95% confidence level. Shiller selected one of those 20 million implications, and found it was refuted. (And even that assumes the discount rates assumptions he made are correct, not a sure thing.) I’m not as impressed as others seem to be.

    You said;

    “I view excess volatility as way more than an anomaly. Has anyone tried to check to see if stock prices since 1981 have exhibited less than volatility than before? I wonder what that would show. If stock prices in the US and in every other major stock exchange consistently exhibit excess volatility, what would you make of that?”

    I would assume Shiller’s model is probably wrong. It isn’t obvious that the assumptions that he made are correct. I seem to recall that Fama or someone else once said that Shiller’s findings are observationally equivilent to an EMH model where the discount rate varies over time.

    You said;

    “The question is not volatility. It is excess volatility. Everything that you mentioned should, under EMH, be reflected in both asset prices and dividends. The problem is the excess volatility of asset prices relative to the volatility in dividends. None of what you just said addresses that issue. That’s the deep structure that I referred to earlier.”

    If you are right, you should try to publish this. The idea I put forward is not my own, but rather is a very well know problem that is widely discussed in the literature. The problem CANNOT be fixed by looking at ex post dividends. In a way it is similar to the peso problem. Suppose that after 1990 Mexico had gotten its act together and never once devalued the peso (of course in the real world they did, but suppose they didn’t) Ex ante, people would have still feared devaluations, based on Mexico’s history. Ex post, the interest rate on Mexican bank accounts (compared to the US)would look “too high” for 240 consecutive months. That’s statistically significant at an extremely high level, far beyond 95%. But we know that statistical signficance of that anomaly would mean nothing about the efficiency of the North American banking market.

    You responded to me:

    ““I think you misunderstand my point. I don’t question the actual statistics of the study, I question the interpretation.”

    I think you’ve done both. Otherwise, why did you mention varying discount rates, and tax rates, and measurement of risk aversion?”

    Those are economic issues, not statistical. My point is that his math may be accurate, but he may be testing the wrong model. Suppose he left out important variables. The EMH doesn not say that only future dividends affect stock prices, there are also tax rates, risk, discount rates, etc.

    You said;

    “If so, his result is not robust, and it could be reversed by adjusting his assumptions and replicating his test under the revised assumptions. Nobody, as far as I know, has done that. If nobody has done that, you can’t just dismiss his result with a wave of your hand.”

    I think people have shown that if you assume a fluctuating discount rate, you can explain his results with the EMH. Shiller would reply the extent of the discount rate fluctuations required are implausible large. And I actually would agree with Shiller. My argument is fundamentally about data mining. I think the entire profession is in deep denial about data mining. Over one half of scientific paper’s cannot be replicated. It isn’t due to fraud, it’s due to data mining.

    You responded to me:

    “That’s not how the EMH should be tested.”

    I think that you basically view EMH as so fundamental, simply an extension of the rationality postulate, that since any empirical test is always a test of more than one hypothesis, you will always reject the auxiliary hypotheses that are also being assumed rather than question EMH.”

    Not at all, I’ve said many times that the proper way to test the EMH is not to look for anomalies, but to test whether excess returns in mutual funds are serially correlated. Otherwise the anti-EMH theory is totally vacuous. Otherwise it is merely people saying “the price seems wrong to me, but there are no useful implications of my view.”

    You said;

    “People do beat the market all the time. Sometimes, it is pure arbitrage and sometimes it is just a hunch. But according to Knight profit is the return to entrepreneurial speculation on uncertainty, which is by definition unquantifiable. Do you believe that the Knightian return to entrepreneurial activity is totally random or do you believe that some people have insights into market anomalies or the mispricing of assets that allow them to earn profits that in excess of the returns generated by holding the market portfolio? Do those insights emerge from the study of historical trends in market prices? No. Do those insights reflect a correct understanding that the current structure of market prices is in principle not sustainable (as it would be if market prices were efficient forecasts of future prices)? Yes.”

    Finally something we agree on. I do think people beat the market on occasion, due to their entrepreneurial activities. In the case of fiancial speculation, I don’t know who is smart and who is lucky. But like Fama I agree that there are a few smart people out there. My point is a point about models. Models like Shiller’s are already incorporated into market prices, and hence are not useful for investment purposes. My argument has never been that the EMH is precisely “true” whatever that means. Rather
    my argument is that the anti-EMH view is not useful to academaics, ordinary investors or government regulators. It may well be useful to Soros and Buffett.

    I think your final comment on gold helps explain why our views diverge so much. You rightly argue that there is enormous uncertainty about the future fundamentals for gold. I agree, indeed I think there is massive uncertainty about the fundamentals for many assets. But I see the same factor as a point in favor of the EMH. It doesn’t surprise me at all that asset prices are so volatile. I have no idea what stocks or gold should be worth. And I assume most investors are just as in the dark as I am. But we do the best we can in a world of great uncertainty. To me, that is the EMH, investors doing the best they can to figure out what’s going on at a fundamental level, and what will happen decades out in the future, and then investing based on their best guess.

    We know fundamentals matter a lot, because we frequently see fundemantals dramtically impact asset prices. The question is this:

    Are the asset movements that we can’t easily connect to fundamentals actually cased by shifting expectations over future fundamentals (as I argue) or are they due to mood swings, beauty contests, etc, as the anti-EMH people argue.

    We can’t know for sure, but Occam’s Razor pushes me toward the EMH view, until the anti-EMH people can give me a useful implication of their theory for how I should do research, or how I should vote, or how I should invest.

  54. Gravatar of David Glasner David Glasner
    14. February 2010 at 22:00

    Scott,

    You said:

    “You are still not getting my point. The EMH has 20 million implications. Statistical theory suggests that tests of the EMH should refute 1 million of those 20 million implications at the 95% confidence level. Shiller selected one of those 20 million implications, and found it was refuted. (And even that assumes the discount rates assumptions he made are correct, not a sure thing.) I’m not as impressed as others seem to be.”

    There may be 20 million implications of EMH, but some are more important than others. I think that excess volatility is a more important implication than say whether small cap stocks earned higher than average returns over some time period. Moreover, if you are right that Shiller was just lucky in finding excess volatility, then his result should not hold up as one alters the time period in which the volatility of stock prices relative to dividends is measured. Do you have reason to believe that Shiller’s result is sensitive to the time period over which volatility is measured? If you are right, smart Chicago grad students should be churning out articles showing that excess volatility was a fluke dependant on the time period cleverly selected by Shiller to run his test.

    You said;

    “I would assume Shiller’s model is probably wrong.”

    Scott, Shiller’s model is EMH. He is deriving an implication from EMH about the relative volatility of stock prices and dividends and showing that the data are inconsistent with the implication of EMH. What other model are you talking about?

    You said:

    “It isn’t obvious that the assumptions that he made are correct. I seem to recall that Fama or someone else once said that Shiller’s findings are observationally equivilent to an EMH model where the discount rate varies over time.”

    Scott, talk is cheap. Why hasn’t a smart Chicago graduate student written a paper showing that a time varying discount rate can account for the excess volatility of asset prices relative to the volatility of dividends. Has it occurred to you that the variation in discount rates necessary to account for the observed excess volatility is implausibly large?

    I said;

    “The question is not volatility. It is excess volatility. Everything that you mentioned should, under EMH, be reflected in both asset prices and dividends. The problem is the excess volatility of asset prices relative to the volatility in dividends. None of what you just said addresses that issue. That’s the deep structure that I referred to earlier.”

    And you responded:

    “If you are right, you should try to publish this.”

    What is “this?”

    “The idea I put forward is not my own, but rather is a very well know problem that is widely discussed in the literature. The problem CANNOT be fixed by looking at ex post dividends. In a way it is similar to the peso problem. Suppose that after 1990 Mexico had gotten its act together and never once devalued the peso (of course in the real world they did, but suppose they didn’t) Ex ante, people would have still feared devaluations, based on Mexico’s history. Ex post, the interest rate on Mexican bank accounts (compared to the US)would look “too high” for 240 consecutive months. That’s statistically significant at an extremely high level, far beyond 95%. But we know that statistical signficance of that anomaly would mean nothing about the efficiency of the North American banking market.”

    I think that you are making a good point here which I want to think about. My initial response is that although the argument works in a single case like the peso, I am not sure that it works over a large number of cases. If there is a significant ex ante probability that dividends will fluctuate, we should observe ex post fluctuations in a corresponding number of stocks even though in any particular case, there may be no observed fluctuation. But I will have to keep thinking about this.

    You said:

    “My point is that his math may be accurate, but he may be testing the wrong model. Suppose he left out important variables. The EMH doesn not say that only future dividends affect stock prices, there are also tax rates, risk, discount rates, etc.”

    We are repeating ourselves. There are enough smart graduate students at Chicago to have written papers showing that including any or all of these variables would account for the excess volatility in Shiller’s test of EMH. Can you cite any such article?

    You said;

    “I think people have shown that if you assume a fluctuating discount rate, you can explain his results with the EMH. Shiller would reply the extent of the discount rate fluctuations required are implausible large. And I actually would agree with Shiller.”

    Well, that’s progress.

    “My argument is fundamentally about data mining.”

    As I said above, I think that you have a good point about the difference between ex ante and ex post volatility in dividends. And I agree that data mining is a serious problem,, but subject to the above qualification, which is not really about data mining, but about how to understand the meaning of volatility, I see no data mining issue at all in Shiller’s work.

    You said:

    “I’ve said many times that the proper way to test the EMH is not to look for anomalies, but to test whether excess returns in mutual funds are serially correlated. Otherwise the anti-EMH theory is totally vacuous. Otherwise it is merely people saying “the price seems wrong to me, but there are no useful implications of my view.””

    But by interpreting EMH in that fashion, you drain all the meaning out of EMH, because Keynes’s beauty contest theory of asset prices would also imply that excess returns in mutual funds are not serially correlated. Like it or not, EMH is saying more than you are willing to test for empirically.

    You said;

    “I do think people beat the market on occasion, due to their entrepreneurial activities. In the case of fiancial speculation, I don’t know who is smart and who is lucky. But like Fama I agree that there are a few smart people out there. My point is a point about models. Models like Shiller’s are already incorporated into market prices, and hence are not useful for investment purposes.”

    The Black-Scholes model was useful for investment purposes. How many other models in the history of economics have been useful for investment purposes? I don’t understand why that criterion matters for this discussion. And I repeat as far as I can tell, Shiller’s work on excess volatility does not involve any model other than EMH, so I have no idea what you mean when you say “models like Shiller’s.”

    “My argument has never been that the EMH is precisely “true” whatever that means. Rather my argument is that the anti-EMH view is not useful to academaics, ordinary investors or government regulators. It may well be useful to Soros and Buffett.”

    I’m sorry but I can’t take your statement that the anti-EMH view is not “useful to academics, ordinary investors or government regulators” seriously. I guess that may be due to the residual Popperianism in my philosophical outlook which makes me cringe whenever I hear that sort of pragmatic claptrap.

    You said:

    “You rightly argue that there is enormous uncertainty about the future fundamentals for gold. I agree, indeed I think there is massive uncertainty about the fundamentals for many assets. But I see the same factor as a point in favor of the EMH. It doesn’t surprise me at all that asset prices are so volatile. I have no idea what stocks or gold should be worth. And I assume most investors are just as in the dark as I am. But we do the best we can in a world of great uncertainty. To me, that is the EMH, investors doing the best they can to figure out what’s going on at a fundamental level, and what will happen decades out in the future, and then investing based on their best guess.”

    But how do you know that shifting expectations don’t themselves react on the fundamentals? You assume that there is no interaction between expectations and fundamentals? What is the basis for that assumption?” Do you think that changing expectations of deflation during the Great Depression had no effect on the course of the price level?

    “We know fundamentals matter a lot, because we frequently see fundemantals dramtically impact asset prices. The question is this:

    Are the asset movements that we can’t easily connect to fundamentals actually cased by shifting expectations over future fundamentals (as I argue) or are they due to mood swings, beauty contests, etc, as the anti-EMH people argue.” No the question is whether shifting expectations can alter the fundamentals. EMH assumes that fundamentals are invariant with respect to expectations. That assumption is not based on logic or evidence.

    “We can’t know for sure, but Occam’s Razor pushes me toward the EMH view, until the anti-EMH people can give me a useful implication of their theory for how I should do research, or how I should vote, or how I should invest.”

    Sorry, but I just don’t know how to respond to this. I think I’ve already said enough. I think I need to get back to the good old 97 percent now and start reading your Great Depression stuff.

  55. Gravatar of ssumner ssumner
    15. February 2010 at 10:23

    David, Shiller’s model isn’t the EMH, it is the EMH with added assumptions added in. It is impossible to directly test the EMH because many of the parameters are unobservable.

    You said;

    “Scott, talk is cheap. Why hasn’t a smart Chicago graduate student written a paper showing that a time varying discount rate can account for the excess volatility of asset prices relative to the volatility of dividends. Has it occurred to you that the variation in discount rates necessary to account for the observed excess volatility is implausibly large?”

    Not only has it occurred to me, I made that piont to you in the last email. I don’t claim that any one factor explains Shiller’s findings. I think it is lots of factors, a number of which I have selected. Country selection bias, fluctuating discount rates, luck, etc. No single factor. The bottom line is that I think that if one invests on the basis of Shiller’s formula (market timing) you will do no better than a buy and hold index fund strategy. Shiller disagrees. He says he can beat the market. Good for him, but that is no basis for developing an academic theory. Do we want to write textbooks explaining to undergraduates that if they follow Shiller’s advice they will probably do better than buy and hold with an index fund? Do we want to encourage them to market time? I don’t think this is a good strategy, even if it seemed to work with past data.

    You asked;

    We are repeating ourselves. There are enough smart graduate students at Chicago to have written papers showing that including any or all of these variables would account for the excess volatility in Shiller’s test of EMH. Can you cite any such article?”

    I’m the wrong person to ask. I’ve never taken a finance course and don’t follow the literature. I don’t doubt that Fama could cite 10 reasons not to believe Shiller. You might want to check out his defense of the EMH, he is certainly aware of Shiller’s work. I also linked at the bottom of this post to a book review by Eric Falkenstein (of a Shleifer book) that I think nicely discusses the pros and cons of the anti-EMH position.

    You said;

    “But by interpreting EMH in that fashion, you drain all the meaning out of EMH, because Keynes’s beauty contest theory of asset prices would also imply that excess returns in mutual funds are not serially correlated. Like it or not, EMH is saying more than you are willing to test for empirically.”

    I strongly disagree, and I think Shiller would as well. The anti-EMH position doesn’t say fundamentals don’t matter at all. So if you can observe markets being above or below fundamental value, it should give you a 60-40 chance of beating the market, as (beauty contest or not) values tend to drift back toward fundamentals in the long run.

    I don’t think Black-Scholes allowed one to predict asset prices. If it did, Black and Scholes should have kept their mouth shut.

    David;

    You said;

    “I’m sorry but I can’t take your statement that the anti-EMH view is not “useful to academics, ordinary investors or government regulators” seriously. I guess that may be due to the residual Popperianism in my philosophical outlook which makes me cringe whenever I hear that sort of pragmatic claptrap.”

    Which social science models are true, and how do you define “true?” Is Newton’s laws of motion true? Or did the theory of relativity make it false? I don’t think these sorts of simplistic statmetns about reality are useful. I say Newton’s theory is useful, but not precisely true. Social science is a very complex area, and even the laws of S&D aren’t true except in the non-existent limiting case. But S&D is a very useful model, sometimes even useful for industries that are actually monopolistically competitive.

    I agree with your comment that expectations, even false expectations, can impact fundamentals like the rate of deflation.

    You said;

    “Sorry, but I just don’t know how to respond to this. I think I’ve already said enough. I think I need to get back to the good old 97 percent now and start reading your Great Depression stuff.”

    If I am going to get the last word, let me tell you where you are right. I have argued that 1987 crash doesn’t seem consistent with the EMH. Shiller’s finding is really just a generalization of that comment. So I agree his result seems troubling for the EMH. My bottom line (which you don’t accept) is that the EMH remains a very useful idea, despite the seeming flaws, whereas the anti-EMH position isn’t very valuable, despite the seeming plausibility of models that allow one to beat the market, to market time. Some other time we’ll have to debate pragmatism vs. realism.

  56. Gravatar of The Height of Irony « It Don't Mean Much, These Seats are Cheap. The Height of Irony « It Don't Mean Much, These Seats are Cheap.
    12. March 2010 at 23:21

    [...] Anti-trust is a veritable black box of confusion, and the psychology of people who are “strongly pro-anti-trust” is filled with endlessly backward logic such that anything that doesn’t fit into their normative intuition of the way the world should work is, de facto, a monopoly. Because of their political and philosophical dispositions, this roles is almost always played by leftist-liberals. According to this worldview, monopoly literally exists everywhere. [...]

  57. Gravatar of Patrick McCann Patrick McCann
    1. November 2011 at 11:07

    Wow, the anti-EMH folks have certainly gotten a hold of the wikipedia page, many of the citations are things like this: http://www.nytimes.com/2009/06/06/business/06nocera.html

    Someone needs to fix this.

  58. Gravatar of Scott Sumner Scott Sumner
    5. November 2011 at 13:08

    Patrick, Yeah, that’s a terrible article. I doubt there is any reporter who actually understands the EMH.

Leave a Reply