Robert Shiller agrees to disagree with Fama

28 October, 2013 at 19:23 | Posted in Economics | 33 Comments

Professor Fama is the father of the modern efficient-markets theory, which says financial prices efficiently incorporate all available information and are in that sense perfect. In contrast, I have argued that the theory makes little sense, except in fairly trivial ways. Of course, prices reflect available information. But they are far from perfect. Along with like-minded colleagues and former students, I emphasize the enormous role played in markets by human error, as documented in a now-established literature called behavioral finance …


Actually, I do not completely oppose the efficient-markets theory. I have been calling it a half-truth. If the theory said nothing more than that it is unlikely that the average amateur investor can get rich quickly by trading in the markets based on publicly available information, the theory would be spot on. I personally believe this, and in my own investing I have avoided trading too much, and have a high level of skepticism about investing tips.

But the theory is commonly thought, at least by enthusiasts, to imply much more. Notably, it has been argued that regular movements in the markets reflect a wisdom that transcends the best understanding of even the top professionals, and that it is hopeless for an ordinary mortal, even with a lifetime of work and preparation, to question pricing. Market prices are esteemed as if they were oracles.

This view grew to dominate much professional thinking in economics, and its implications are dangerous. It is a substantial reason for the economic crisis we have been stuck in for the past five years, for it led authorities in the United States and elsewhere to be complacent about asset mispricing, about growing leverage in financial markets and about the instability of the global system. In fact, markets are not perfect, and really need regulation, much more than Professor Fama’s theories would allow …

We disagree on a number of important points, but there is nothing wrong with our sharing the prize. In fact, I am happy to share it with my co-recipients, even if we sometimes seem to come from different planets.

 Robert Shiller



  1. But the EMH doesn’t say that the average investor trading on public information cannot beat the market. It says that THE investor ON AVERAGE trading on public information cannot beat the market. The idea being that if you take, say, a Soros or a Buffett and track their performance over a long period of time, they will merely hit the market average. I’m afraid Shiller is giving the theory far too much credit. Even the “humble wisdom” that it purports to give to investors is obvious garbage.

    • The flaw of averages takes many forms …


      • Indeed. I’ve come to think that the EMH is actually just a tautology based on a flawed use of averages.

        Think about it: the investor, on average, cannot beat the market.

        Okay, well what is “the market”? Well, it’s a price set in line with the net effect of investment in assets from investors. Put another way: the price is set in line with bidding by the average of investors.

        So, what the EMH really is saying is: the investor, on average, cannot beat the average of actions undertaken by investors.


        • “what the EMH really is saying is: the investor, on average, cannot beat the average of actions undertaken by investors.”

          With the important qualification: And there is no other investment strategy based on public information that would allow an investor, on average, to beat the market.

          But I am not sure how this is based on a “flawed use of averages”. The average is just there to filter out luck.

          And I certainly do not think that investors like Buffet or Soros have been purely trading on public information.

          • I don’t think you understand what I’m getting at. “The Market” is just the average outcome of investor interaction. Saying that the investor cannot beat The Market is thus saying that the investor cannot beat the average of investor interaction. Thus all the EMH is saying is that “all investors are really just average”.

            Seen in this way it is a silly little moral message wrapped in a tautology. “Don’t get ahead of yourself, humble mortal,” says the EMH adherent. Of course, it seems to me that this is, like all such messages, just based on Nietzschean ressentiment: academic economists tend to make rubbish investors because they are crap at forecasting, so they placate their egos with EMH garbage.

            Also, I don’t think you know what you’re talking about regarding, for example, Soros. When he broke the sterling in 1993 he was obviously trading on public information and he made a huge amount of money. If you look back to his strategies surrounding energy companies in the 1970s it is also clear that he was trading on public information.

            I find it amusing that when people prove EMH adherents wrong they resort to literal libel. You might want to be careful where you go around making such accusations, Pontus from Cambridge University… you might find yourself in trouble one day.

            Oh, and you might want to also accuse Keynes of insider-trading. At least he’s not around to sue you for libel!

        • Very interesting point.
          Don’t we need another element for this “the investor, on average, cannot beat the average of actions undertaken by investors.” to be a tautology? Namely that the individual average (for every investor) and global average are equal. And for this to be true, don’t we have to consider that all investors form their expectations the same way, or that they are identical?
          As you pointed out below “all investors are really just average”.
          So EMH is an application of rational expectations to financial market. They cannot beat the market because the market is the average of their on-average-right expectations.

          • Bingo! You nailed it! It’s based on a single agent. And in the statement “The investor on average cannot beat the market”, the terms ‘investor’ and ‘market’ are simply this single agent. They are synonymous and thus the theory is a tautology.

  2. What leaves me scratching my head with the all of this Capital Asset Pricing Model stuff is that it doesn’t seem to incorporate the way that volatility can actually provide a source of returns when asset holdings are rebalanced. If an asset randomly doubles in price and halves in price then rebalancing against cash (or against other assets) will create a stream of returns (as would a covered call option strategy or whatever). That seems to all get missed when these models focus on holding an asset and just taking the dividends. It seems to me the beneficial aspects of price volatility (that require trading for them to be captured) actually are crucial for how assets are actually priced. It doesn’t require “knowing the future”; such volatility capture would work if prices were bobbing up and down as directed by a coin being flipped.

    • I’m not sure that is a criticism. A CAPM enthusiast would say, “Well, you just have to count the price moves (volatility) in the expected returns…”. I think this is a cop-out but then you’ll find that most of these models are either wrong or too vague to be useful. I wrote up a separate criticism here:

      • I guess what I was trying to get at was that if someone just held such a security over time as the price randomly bobbed up and down, they would gain nothing from that volatility. Rebalancing (or writing call options or whatever) is essential if those potential returns are not to simply pass by. Activist shareholders want management to cause returns to be passed on by way of price volatility so that the returns pass by the “dumb money” (such as pension funds etc) and pass on through to be captured by investment banks etc. That induces companies into taking on lots of debt, conducting buybacks rather than paying dividends, paying management by way of options etc etc -anything to crank up the volatility.
        Am I mistaken in thinking that that entire realm of the real world is ignored by CAPM?
        To be honest this subject is a bit alien to me. Have you seen Ole Peter’s stuff?

        • What you’re saying then is that big banks and funds that use modelling that insulates them from volatility have an edge over the rest of the market when there is volatility? Certainly they have an edge over the individual investor who buys shares, but probably not over pension funds etc. who probably have people controlling for volatility too.

          As to the EMH, I think the stance is vague here. Generally, they say “Buy index funds”. But given the above they might say “Actually we meant buy index funds that control for volatility”. The whole thing is impossibly vague to be honest. It soon becomes clear that what exactly their “market tracking portfolio” is is quite vague. The EMH is just a lot of hot air, to be honest.

          • Sorry I haven’t managed to get across what I’m trying to say. I’m not simply talking about diversifying to protect against overall portfolio volatility. I’m talking about rebalancing between assets so as to capture the “rebalancing bonus”. That rebalancing bonus comes about because if something is bobbing up and down randomly then every 50% loss is matched in the end with a 100% gain. But 50% of £100 is £50 whilst 100% of £100 is £100. If you are not convinced then just try a reconstruction where you have a backdated portfolio with something very volatile at say 25% and cash at 75% and see what happens over say 10 years if you do rebalance or if you don’t rebalance. Then think about how leverage compresses time and think about having a variety of divergent assets (eg 30year US treasuries, stocks or whatever).

          • I think I know what you’re talking about now. You’re saying that if you see an asset swing -50%, +100%, -50%, +100%… and so on, that you try to capture this dynamic to make money? So far as I can see there are two related problems: (a) you cannot be sure that this trend will carry into the future because financial markets are non-ergodic and (b) others in the market will soon spot the pattern, do the same as what you are doing and the pattern will disappear.

    • Also, it does require “knowing the future”. If the volatility is mimicking a coin toss then it is, by definition, ergodic.

      • Phil, I’m not saying it is a “patten” of alternate -50%, +100% etc. I’m simply saying that if the price of ANYTHING is, over the long run, trending around some sort of “typical” value, then implicitly that equates to every 50% fall having a corresponding 100% gain at some point, every 75% fall having a corresponding 300% gain, every 20% fall having a 25% gain, every 33% fall having a 50% gain etc etc etc. That holds up with anything that more or less holds its value over time eg a stock index, a commodity, 30year US treasury bonds, a currency pair or derivatives based on those underlying prices or whatever.
        If you are saying that people would eliminate that phenomenon by exploiting it then basically you are saying that prices shouldn’t be volatile. But they are and so long as they are, that phenomenon will hold true. Leverage and information asymmetries build volatility. Like I said, managers are induced to use capital restructuring to induce stock price volatility.
        It is a fundamental deep seated glitch in the nature of finance. I think a large part of the “financial services industry” is simply an artifact of that glitch. It bleeds the global economy by a lot.

        • “….if the price of ANYTHING is, over the long run, trending around some sort of “typical” value…”

          That’s an ergodic assumption. It assumes a long-run trend equilibrium. It will not typically occur in the real world. But when it does, traders will exploit it until it disappears.

          Why does volatility continue to exist? Because in series that do not have an equilibrium trend it is much more difficult to exploit volatility.

          • Phil the example I gave below with the five “coin flip” assets, clearly shows that it works even without a trend equilibrium. Such coin flip assets obviously do not have a trend equilibrium, they can wander freely.

            Anyway real assets do actually have a trend equilibrium. Commodities typically trend to the production cost. Stock indices trend around some sort of ratio to GDP. They won’t wander off by many orders of magnitude.

            You seem to make out that I’m talking about a potential way to play things as though I’m not talking about what is actually done.

          • I can give you a probability of the whether heads or tails will arise out of the coin tosses prior to the tosses. It is therefore an ergodic phenomenon.

            Commodities trending to production cost means little. When speculation drove gold to $1800 an oz, less profitable mines opened and the production cost rose. The production cost can, and often is, the dependent variable.

            If you believe you can find trends in the market that are exploitable I would encourage you to exploit them. I’m 99% sure that they are not as easy to find as you think.

          • Gold price and production cost:

  3. Phil, try this experiment: Have a pretend portfolio starting with £1000 worth of five “assets” A, B, C, D and E. Toss five coins, one for A, one for B one for C etc. If you get heads double the value of the corresponding “asset”. If tails, then halve it. Try that for say twenty rounds with and without rebalancing between your assets each round.

    • Phil the “coin flip portfolio” is NOT ergodic. It would only be ergodic if it had an infinite number of of “coin flip assets”. It only has five. Try it out yourself like I suggested. Also check out

      I really think you are closing your eyes to a very fundamental core part of how the real world works.

      • A coin toss is ergodic. If I toss a balanced coin once there is a priori a 50/50 chance I will get heads. Non-ergodic processes are ones in which probabilities cannot be assigned.

        • Phil:

          “the term describes a random process for which the time average of one sequence of events is the same as the ensemble average.”

          The example that Ole Peters uses to illustrate non-ergodicity is where a coin flip decides whether the player gains 50% or loses 40%. Clearly winning gains more than losing loses BUT a subsequent win does not make up for a preceding loss and a subsequent loss loses more than was gained by a preceding win (1.5×0.6=0.9<1); that is the “magic” of compounding. This simple set up means that a large enough population of players will, in aggregate, steadily gain from playing the game but all of the winnings will randomly accrue to an ever smaller minority of players whilst almost everyone loses almost everything. Only if wins and losses are pooled and rebalanced across many players does the game become universally attractive. The time average for an isolated player is NOT the same as the ensemble average.

          I really recommend checking out his video, he explains it very clearly

          • Coin tosses are ergodic. They have a probability distribution that can be known before the event. I am not willing to argue this further. And you’re misrepresenting the Peters example completely… It is not saying what you think that it is saying.

            However, if you are correct on the volatility argument then why not prove me wrong? Why not use the technique and earn lots of money? You can document it on my blog if you’d like… But you have to bet real money in a real market.

  4. It’s been pointed out many times that coin flips are nothing like asset prices. Each coin flip is a new event. It’s interesting though that if you use a coin flip to decide if a value increase or decreases by 1 every time unit, the resulting graph represents a system that has no mean. You can, of course, calculate a mean if you want but that gives you no insight into the system because it has no trends. The only thing you can know about such a system is that it will at some time before infinity return to zero. Otherwise it could rise or fall however. And that’s only 1 dimensional systems, for higher dimensional systems there is not even a guarantee that it’ll return to zero.

    On the EMH: If it is the case that the EMH says that any specific speculator’s gains, over their lifetime, will average at 0 (that is not just that investors overall will average at 0). Then the only mathematical model I can think of that gives rise to that property is one in which the market, whatever it is, is purely and literally random in a non-deterministic. And not just random as in ‘conforms to a gaussian distribution’ but random in ‘that’s just the way it is’ way. If it were simply conforming to a gaussian distribution that wouldn’t rule out it not actually being random. But EMH people don’t seem to be arguing that the market is random, but that it’s ‘fair’. Now unless ‘fair’ is just some economics technical term for ‘random’ surely it makes no sense. How can prices be random and fair? If something is really random, all you can say about it is that it’s random.

    It’s also a little funny that in order to conclude that humans are too ignorant to exploit the market economists start by assuming that ‘agents’ in the economy are perfect predictors and have unlimited cost benefit calculation resources.

  5. I still don’t think flipping a coin is relevant because flipping a coin is, each time, a new instance of the system. An asset price at a given time is a sample of a continuous function, not an event. The value of that function is contingent on its history and other asset prices.

    • This is exactly right.

      Coin tosses = ergodic as they are based on known probability distributions.

      Asset prices = non-ergodic because they are not.

  6. Phil, ergodicity is where the ensemble average equals the time average -that is the definition of the word. A cumulative series of doublings and halvings based on coin flips such as I described is NOT ergodic. Ole Peters actually uses it as the key example in that video, he shows the plots demonstrating how different the ensemble average is from the time average.

    • I didn’t say that the cumulative series is ergodic. I said that the coin tosses are. They are. End of story.

      You keep equating the two and that is why you are confusing yourself.

      • Phil, I really don’t see what your problem is. Are you saying that it is a bad idea to rebalance assets because we don’t know that say a stock index won’t fall forever unlike a coin flip sequence for which we do know that a string of heads is vanishingly unlikely? My take is that it is a fair enough approximation to accept that a fall could be followed by a subsequent additional fall but that it might be followed by a subsequent rise just as if a coin flip were deciding. Don’t you accept that in the past that has served well? I mean stock indices have kept more or less the same relative to the economy over the long run. Same thing with say gold prices or whatever. For such prices to recover, they have necessarily provided the potential for a “rebalancing bonus”.

        • “Don’t you accept that in the past that has served well?”

          There you go. Ergodic argument, like I said.

          If you believe this cash in on it. It shouldn’t be hard.

Sorry, the comment form is closed at this time.

Blog at
Entries and comments feeds.