OK you like EMH so much that you think 9 students from one professor all outperforming for decades is cherry picking and data mining. I think it is finding a small group of people wh oclaim to be learning from someone who has empirically verified methods, and who, when they apply these methods, get the predicted results consistently for decades. I think characterizing this as cherry picking and data mining is at more likely to be a bad explanation for what is being seen than is mine, which is that they are doing what ehy say they are doing, and it is working.
Even a broad index fund is “managed.” The conditions for being listed are quite stringent, and involve “survival bias” filters, if stocks fall below a certain value they are delisted. I actually don’t think that the difficulty of beating the SP500 is much of a proof of EMH as much as it is a proof that very straigtforward standards applied on a slow timescale capture almost all of the value available from managing a portfolio. I think people investing more broadly than SP500, people investing with people who come in to their living rooms seeking “angel” investors do a lot worse. If the market was efficient in principle, then one wouldn’t need the SP500 or even the NASDAQ seal of approval to wind up with results that were at the market mean. If usnig your brain is required to pick SP500 over living room pitch man, then in principle, using your brain is required to get reasonable results.
I think if a proposition of efficiency is to be proved true, ti si not by looking at the average performance of every tom dick and harry and noticing that with mathematical necessity they tend to have the same mean as the market which of course they comprise. I think a proper proof of efficinecy requires showing in detail that there are no consistent outliers of high performance. That funds with decades long records of outperformance occur at the proper rate to be consistent with pure luck. Indeed, to show that while it appears that some people predictably outperform, that for all these actors past performance is no predictor of future performance, and that the hangers on that joined Buffett in the 60s or 70s or 80s or 90s after seeing his record THOUGHT their outperformance was due to their identifying a winner, but that it was consistent with just pure dumb continuous luck.
I think their is a gigantic difference between “we cannot prove that their is alpha” and “the most likely explanation of what we see is that their is no alpha.”
As to identifying bubbles that were not bubbles, the only bubbles I have identified are tech,and real estate. I identified a “bubble” in a small company stock (Conductus) where a company with no real products generated excitement by talking about how they were getting in to the cellular industry, driving their stock price from 3 to about 80 before they crashed back down to 3. I shorted them at about 70, took my returns a few weeks later at 60 or so, they proceeded to rise to 80 and then within a year drop back to 2.5. I identified another mispricing in NHCS where numerically they were spinning out a company which was being completely undervalued in their current stock price. I asked others “can this really be true,” they said only in general yeah stuff like that happens. I bought a few thousand dollars worth, made the 20% or so return it seemed I was seeing laying on the table a few months later.
The main sense in which the market seems efficient is that the prices are predominantly set using sensible analyses, presumably because those who do not follow a proven technique of picking sensible prices do not survive, so the main component of market efficiency is that the processes for beating the market are broadly exercised and dominating the market. So it is hard to do better than free-riding on that. But does it turn out that some people do better at that process than others? I think the best explanation for what we see is that yes, some do, and that they are a smallish minority is not because they are just the tail of a random distribution, but because of mathematical necessity beating the average significantly can only be done by a minority.
Anyway, thanks for sticking with it and explaining your position to me.
OK you like EMH so much that you think 9 students from one professor all outperforming for decades is cherry picking and data mining.
To expand even further on my critique: you are placing a huge amount of weight on 9 students, of unknown veracity, out of an unknown number of students (itself out of an unknown number of millions of people who have tried to beat the market over the past century), who have not released audited records much less ones comparing them to indexing, who started half a century ago (which is the investing dark ages compared to what goes on now, in 2013), and at least one of whose successes seem to be partially explained by non-efficiency-related factors?
This is roughly as convincing as Acts of the Apostles documenting the 12 apostles’ successes in beating the (religious) market and earning converts.
I think people investing more broadly than SP500, people investing with people who come in to their living rooms seeking “angel” investors do a lot worse. If the market was efficient in principle, then one wouldn’t need the SP500 or even the NASDAQ seal of approval to wind up with results that were at the market mean.
Those angel investors are forfeiting diversification and so can easily earn below-average returns. EMH doesn’t mean that you cannot deliberately contrive to lose money.
I think their is a gigantic difference between “we cannot prove that their is alpha” and “the most likely explanation of what we see is that their is no alpha.”
I think in an adversarial environment where everyone claims to be able to beat the market and you should give them their money, and there are compelling theoretical reasons that any beating of the market would wipe out whatever advantage was posssed, there is not such a gigantic difference.
I bought a few thousand dollars worth, made the 20% or so return it seemed I was seeing laying on the table a few months later.
Congratulations on your day-trading success. You know what happens to most of them, right?
EMH doesn’t mean that you cannot deliberately contrive to lose money.
Under EMH is pretty hard to deliberately and consistently lose money. It’s very easy to get additional risk (e.g. by not diversifying), but I don’t think EMH envisions assets with negative expected return.
Mm, the way I remembered was that by not diversifying, you were taking on additional uncompensated risk; not diversifying wasn’t completely neutral, expected-value wise. (Also, there’s obvious ways to guarantee losing money: trade a lot. The fees will kill you.)
… who have not released audited records much less ones comparing them to indexing, …,
Actually the records ARE audited, they ARE compared to indexing, and those records and comparisons are reported by the original article I mentioned, which I finally link to here.
you are placing a huge amount of weight on 9 students
If a professor’s students dominate some part of engineering or biology or chemistry, it is generally taken as evidence that the professor was teaching something real. I suppose if we had an Efficient Knowledge Theory we would understand that going to Caltech or MIT was as wasteful as picking up $20 bills on the sidewalk (which don’t exist in a classic EMH joke).
Should we be questioning whether a good education in philosophy or math or physics or engineering or biology or… is just a mismatch between the power of random chance and the human bias towards seeing patterns? Or is there something special about learning how to value companies that puts it in a category of analysis that is different from all other observations of the effects of knowledge?
In any case, the article linked discusses the randomness hypothesis extensively pointing out among other things that the various investors reported upon had exceedingly small amounts of overlap in what they actually invested in.
Gwern, these comments are not so much aimed at you, you have obviously been down these roads and decided which way you would turn. They are aimed at anybody reading this who is still not sure about EMH. The article linked is excellent and written by a guy who walks the walk better than anybody else in human history (so far).
Actually the records ARE audited, they ARE compared to indexing, and those records and comparisons are reported by the original article I mentioned
I don’t see any mention of how they were audited (Buffett merely says that they ‘were audited’, no mention of by whom, when, what the audits said, whether he saw the results, etc, and offers as reassurance that checks were paid for the appropriate amounts, which is not my problem here), and if you really want to nitpick, then I would bring up that Buffet does not talk about ‘9’ students, he actually talks about 4 people who worked for Graham, tells us that ‘it’s possible to trace the record of three’ (well, there’s some selection bias right there...) and does not explain how the 3 partners did (more selection bias), and some of his other examples are questionable at best—including his very good friend Munger, including two funds he ‘influenced’ (while disclaiming that he might have influenced any other funds and this isn’t cherrypicking, which I don’t understand how he can honestly say how he knows for sure he has not similarly influenced any others), reporting different metrics for different examples (why is Munger compared against the Dow while others are compared against the S&P?), not comparing against an index (table 8), and some do not beat the comparison index at all (Table 9, Becker, underperforms S&P by 3%)
If a professor’s students dominate some part of engineering or biology or chemistry, it is generally taken as evidence that the professor was teaching something real.
Buffett doesn’t dominate the markets, and the proper comparion is to ideas, not students—if a single professor’s students dominated, I’d be more inclined to suspect corruption or logrolling or the professor being a genius at academic infighting and bureaucracy...
I suppose if we had an Efficient Knowledge Theory we would understand that going to Caltech or MIT was as wasteful as picking up $20 bills on the sidewalk...Should we be questioning whether a good education in philosophy or math or physics or engineering or biology or… is just a mismatch between the power of random chance and the human bias towards seeing patterns?
Markets are very different from electronic circuits or particle physics or philosophy or engineering. Circuits don’t care if you found a more efficient way to design them. The properties of steel will not change when you discover it lets you build profitable bridges.
Or is there something special about learning how to value companies that puts it in a category of analysis that is different from all other observations of the effects of knowledge?
Er, yes, there is. That’s kind of the point of the efficient markets concept! Markets are unusual and special in that the attempt to find predictable regularities leads to the exploitation of the regularities and their disappearance. (Eliezer describes this as “markets are anti-inductive”, which is not wrong, but I’m convinced there must be some more intuitively understandable phrase than that.)
Gwern, these comments are not so much aimed at you, you have obviously been down these roads and decided which way you would turn. They are aimed at anybody reading this who is still not sure about EMH.
Is that one article really the best, solidest, most convincing criticism of EMH you can come up with, which you think will persuade people reading this conversation that EMH is to a meaningful degree false and markets are often beatable—some cherrypicked questionable examples from the dawn of time?
Is that one article really the best, solidest, most convincing criticism of EMH you can come up with, which you think will persuade people reading this conversation that EMH is to a meaningful degree false and markets are often beatable—some cherrypicked questionable examples from the dawn of time?
In its way, yes it is. You get a guy who has impeccable credentials, a massive public record who thinks he has been investing intelligently for decades, who if he IS performing randomly is a few sigma out on the positive side of the random distribution. You get to see what he has to say about what he thought he was doing, how it fits with what a whole bunch of other people were doing, a cogent description for why it might work, and a bunch of numbers about how it does indeed seem to work. Buffett understands the idea that he could just be lucky and he addresses it.
If you think the best explanation of Buffett’s life and results are that he has been fooled by randomness, then you are a very different judge of character and information than me or millions of others like me.
If the EMH was “the markets are really really efficient, it is hard to produce alpha (outperformance), hard to know when you have alpha, and easy to fool yourself because of human biases” then who would argue with that? Not me. But that step from “really hard” to “impossible” is unreasonable. It is not impossible to be a great baseball player. It is not impossible to consistently beat other players at poker, even though everybody playing has the same information, on average across all the hands. It is not impossible to understand 10 languages, even though to most of us most of them sound like noise.
If EMH was right, wouldn’t the smartest, most quantitative participants in the market have figured that out? Wouldn’t Renaissance Technologies have 1) failed, and 2) figured out that their failure was consistent with randomness where they thought there was order?
EMH is the hypothesis that because bunches of smart people all work to figure out what the best investment is, there can be no excess returns available to the smart people who all work hard to figure out what the best investment is. Well if there are not excess returns available to them, why do they do it?
Isn’t EMH the hypothesis that, for EVERYBODY in the market, it would be more efficient to free ride and use your intelligence on something where you can actually produce a return?
Isn’t EMH ultimately a big floppy tent held up by a tent pole which the EMH’ers deny exists?
In its way, yes it is. You get a guy who has impeccable credentials, a massive public record who thinks he has been investing intelligently for decades, who if he IS performing randomly is a few sigma out on the positive side of the random distribution. You get to see what he has to say about what he thought he was doing, how it fits with what a whole bunch of other people were doing, a cogent description for why it might work, and a bunch of numbers about how it does indeed seem to work. Buffett understands the idea that he could just be lucky and he addresses it. If you think the best explanation of Buffett’s life and results are that he has been fooled by randomness, then you are a very different judge of character and information than me or millions of others like me.
First, let me point out that I put a fair amount of work into pointing out all those flaws and holes in your last best citation, and I’m a little annoyed that you completely ignored all of them in favor of saying “but Buffett is so high-status and I like him so much”. Yes, and George W. Bush famously said of Vladimir Putin, “I looked the man in the eye. I found him to be very straight forward and trustworthy and we had a very good dialogue. I was able to get a sense of his soul. He’s a man deeply committed to his country and the best interests of his country and I appreciate very much the frank dialogue and that’s the beginning of a very constructive relationship.” We all know how that turned out.
What you think about Buffett’s “character” is irrelevant to me, and for me, further emphasizes your extremely poor reasoning in this area—that when pushed back, you resort to one man and your beliefs about his “character”.
If EMH was right, wouldn’t the smartest, most quantitative participants in the market have figured that out? Wouldn’t Renaissance Technologies have 1) failed, and 2) figured out that their failure was consistent with randomness where they thought there was order?
I don’t know why RenTech performs as well as they seem to. Presumably it’s not the same reason that Madoff was able to beat the market for so many years in contravention of EMH. Perhaps it was the same reason SAC did well (insider trading) and they simply haven’t been caught yet. Or maybe there were some inefficiencies back when they started which they erased and have since been coasting on their reputation. Given that it’s a very private hedge fund, we’ll probably never know.
EMH is the hypothesis that because bunches of smart people all work to figure out what the best investment is, there can be no excess returns available to the smart people who all work hard to figure out what the best investment is. Well if there are not excess returns available to them, why do they do it?
Because there is demand for investment services, considerable cognitive biases at play along with wishful thinking (‘I will be the next Buffett!’), and normal profits available. After all, if no one was there taking even the normal profits, there would immediately be excess returns attracting people to the enterprise...
Isn’t EMH the hypothesis that, for EVERYBODY in the market, it would be more efficient to free ride and use your intelligence on something where you can actually produce a return?
No.
Isn’t EMH ultimately a big floppy tent held up by a tent pole which the EMH’ers deny exists?
Because there is demand for investment services, considerable cognitive biases at play along with wishful thinking (‘I will be the next Buffett!’), and normal profits available. After all, if no one was there taking even the normal profits, there would immediately be excess returns attracting people to the enterprise...
So your hypothesis is that some process ties all the people there needing to provide skull sweat to get normal returns (and create a market for everybody else that is efficient) works equally across all such players? It sure doesn’t work that way in any other human enterprise I can think of. Intel and AMD produce different quality chips for laptops. At the other end of tilt, Intel and Qualcomm produce very different quality of chips for mobile. The physics department at Caltech produces a very different product, research wise and teaching wise, than the physics department at USC. Writer Stephen King produces a very different quality of novel than do a thousand or more other authors populating the increasingly virtual shelves of bookstores. Even here on lesswrong, some of us write wonderful stuff which is read by many and admired, while others of us struggle to get our karma up to 1000 and then hang on by our fingernails stopping from saying what we really want to say to keep it there.
So why on FSM’s tomato-colored earth would you expect these financial creators of efficiency to all get the same results from their efforts?
And when shown the spread in effectiveness in results, to deny the evidence of your own eyes and declare it all to be the distant tail of millions of coin flippers?
It doesn’t seem like a stretch to you? It doesn’t seem that the evidence is strong that the market is VERY efficient, but that the evidence is not there that it is COMPLETELY efficient?
First, let me point out that I put a fair amount of work into pointing out all those flaws and holes in your last best citation, and I’m a little annoyed that you completely ignored all of them in favor of saying “but Buffett is so high-status and I like him so much”.
If you want I’ll go through them point by point.
I don’t see any mention of how they were audited (Buffett merely says that they ‘were audited’, no mention of by whom, when, what the audits said, whether he saw the results, etc, and offers as reassurance that checks were paid for the appropriate amounts, which is not my problem here)
Presumably you can see the difference between your stating these are NOT audited, and then when pointed out that they are, you back off to this.
The results of the audit are the results in this article. That is, these are results reported which survived the audits.
In many of the cases, the audits are “typical” of the investment advisory business, but I do not know what that means exactly. But it is a level playing field against all other investment advisers.
Also for a few, not all, of the investors cited here, they run/ran for decades public investment businesses. Isn’t the preponderance of your Bayesian a posts that if at least these members of this widely read, cited, and discussed “superinvestors” article was just wrong, that this would at least have lead to traceable reports of the discrepancies on the internet, finable with google search?
To the extent your objections amount to “Buffett could be an idiot and a fraud, either not knowing or not caring what it means to make these claims” my answers are going to be we have 5 decades of impeccable record, if you think Buffett is that unreliable then generally there is no arguing with you as anybody who says something with which you disagree you will question as an idiot or a fraud. If you cannot tell that Buffett is not an idiot or a fraud, or have not followed him well enough to be sure one way or the other, then I would suggest you have no business weighing in on the subtle subject of whether the market is so inefficient that the best investors in the world are just coin-flippers.
What you think about Buffett’s “character” is irrelevant to me, and for me, further emphasizes your extremely poor reasoning in this area—that when pushed back, you resort to one man and your beliefs about his “character”.
I suggest relying upon Buffett because you and everyone else out there who can read has infinitely more reason to rely upon Buffett than to rely upon me. And further, what is needed int he discussion of EMH vs non-EMH is not some brilliant new insight that I can provide that you haven’t seen somewhere else already. EMH vs non-EMH is a subtle question, is the market so efficient that Buffett can’t consistently beat it without committing a crime, either insider trading or some other information-twisting fraud, or is it just a little less efficient than that? The “insight” I have is that what pushes it towards efficiency is competing analyses on opposite sides of each trade. The “insight” I have is that every bit of evidence suggests that in business some people have superior skill or algorithms or SOMETHING and are more successful than others. And they can do it serially, command high prices in very competitive markets, blah blah blah, and show EVERY BIT as much evidence of being “real” as are great pitchers or tennis players or tenors or talk show hosts or porn stars. And your case is that no, with investing it is different, the people who do the work are so smart that they get it right in an unbeatable way, but so stupid that they don’t even realize they would be better off free riding.
What is needed is not any great insight from one or the other of us, I don’t think, but evidence that is hard to deny that yes, the market can be beat. I think that evidence would consist of market beaters coming from a narrowly defined group of people who set out to beat the market by studying it and allowing evidence to drive their future hypotheses and efforts. And what do we find in the market? Exactly that, market beaters are smart and talk in terms of causality, of what makes a business great, of where the momentum traders and the chartists missed the boat.
But my causal chain of how the market could be merely VERY efficient has been, I hope, presented by now. Let me know if it hasn’t.
Markets are very different from electronic circuits or particle physics or philosophy or engineering. Circuits don’t care if you found a more efficient way to design them. The properties of steel will not change when you discover it lets you build profitable bridges.
As much as you might hypothesize that we will not see securities markets make the same mistakes they have made in the past, does the evidence support that? And in any case, the idea that markets do learn or have learned SOMETHING supports only the VEMH, the very efficient market hypothesis, which is not controversial. By this I mean the hypothesis that it is hard to beat the market, because all the easy stuff has been figured out and is properly accounted for by the bulk of the traded money in the market.
I tracked Chiplotle stock on and off from around 2000 forward. There were two classes of shares, A, and B, with the B’s trading at a very consistent 10% discount to the A’s. I would check once or twice a year to see if this difference persisted, and it did. The thing that was surprising was that the documentation of the company explained that these shares had equal value, represented identical fractions of the total company. Why they traded at a 10% difference I never saw an explanation, and I always questioned whether there was some detail I was missing. Here, in late 2007 is documentary evidence that the difference persisted. Here, two years later, is Chipotle’s report that they were eliminating the two classes in favor of one class, and that the exchange rate would be 1:1 just as I had always believed.
In my case, I am an electrical engineer/physicist, trying to concentrate on building new cell phone algorithms for at least a few hours a day. Instead of organizing the financing to exploit this weird inefficiency at low cost, I just checked in on it every year or two. Wanting to see if I was right. Had I been a professional trader, I would have looked more at creating an arbitrage on the A and B shares and capturing the collapse of the arbitrary pricing difference. As an amateur I didn’t know if it would ever collapse, and the brokers are neither smart enough nor dumb enough to let me buy the As and short the Bs without a lot of capital in my account to anchor what they see as two uncorrelated risky bets.
My point here is this is just ONE of MANY possible stories of moderate sized inefficiencies I have seen with my own eyes. Others I have traded. Yes, every one of them is an anecdote. The plural of anecdote is not data. But a bunch of anecdotes like that creates, it would seem, market beating performance for many traders trading different stocks.
Maybe markets COULD be different than circuits and so on, and maybe as computers and AI takes over more and more, they will get more and more efficient. But even then, the most powerful AIs will be beating the market, even as they essentially set th prices at levels that make it incredibly hard for anybody else to beat the market. The thing that drives market makers is not their stupidity, but their intelligence and rationality. Seems to me.
Er, yes, there is. That’s kind of the point of the efficient markets concept! Markets are unusual and special in that the attempt to find predictable regularities leads to the exploitation of the regularities and their disappearance.
THIS is a hypothesis. And the only word in that hypothesis I will argue with is the last one: disappearance. The predictable regularities don’t disappear from the time-stream of prices, if there is a mispricing at 2:31 PM on Thursday it is frozen there in the permanent record. What changes is how long it takes for the record to close those various gaps. Maybe before computers a broad class of inefficient prices were never traded away. Maybe in the 1980s a broad class of inefficiencies were capitalized upon by people with computers over the course of a two week period. Maybe by the 2000s those same inefficiencies were traded away within hours or minutes.
But my points are: 1) we are not arguing efficiency vs inefficiency, we are arguing too efficient to beat vs nearly too efficient to beat and 2) without the inefficiencies, no one would be there to pay the actors making the market more efficient by trading the inefficiencies, and that no, it is not their stupidity that keeps them working for free.
I hope this is what you wanted when you suggested I was ignoring your point and merely arguing pro hominem, citing people who I thought should be much more believable than I am. If I missed anything that still seems critical, flag it to me and I’ll answer it.
OK you like EMH so much that you think 9 students from one professor all outperforming for decades is cherry picking and data mining. I think it is finding a small group of people wh oclaim to be learning from someone who has empirically verified methods, and who, when they apply these methods, get the predicted results consistently for decades. I think characterizing this as cherry picking and data mining is at more likely to be a bad explanation for what is being seen than is mine, which is that they are doing what ehy say they are doing, and it is working.
Even a broad index fund is “managed.” The conditions for being listed are quite stringent, and involve “survival bias” filters, if stocks fall below a certain value they are delisted. I actually don’t think that the difficulty of beating the SP500 is much of a proof of EMH as much as it is a proof that very straigtforward standards applied on a slow timescale capture almost all of the value available from managing a portfolio. I think people investing more broadly than SP500, people investing with people who come in to their living rooms seeking “angel” investors do a lot worse. If the market was efficient in principle, then one wouldn’t need the SP500 or even the NASDAQ seal of approval to wind up with results that were at the market mean. If usnig your brain is required to pick SP500 over living room pitch man, then in principle, using your brain is required to get reasonable results.
I think if a proposition of efficiency is to be proved true, ti si not by looking at the average performance of every tom dick and harry and noticing that with mathematical necessity they tend to have the same mean as the market which of course they comprise. I think a proper proof of efficinecy requires showing in detail that there are no consistent outliers of high performance. That funds with decades long records of outperformance occur at the proper rate to be consistent with pure luck. Indeed, to show that while it appears that some people predictably outperform, that for all these actors past performance is no predictor of future performance, and that the hangers on that joined Buffett in the 60s or 70s or 80s or 90s after seeing his record THOUGHT their outperformance was due to their identifying a winner, but that it was consistent with just pure dumb continuous luck.
I think their is a gigantic difference between “we cannot prove that their is alpha” and “the most likely explanation of what we see is that their is no alpha.”
As to identifying bubbles that were not bubbles, the only bubbles I have identified are tech,and real estate. I identified a “bubble” in a small company stock (Conductus) where a company with no real products generated excitement by talking about how they were getting in to the cellular industry, driving their stock price from 3 to about 80 before they crashed back down to 3. I shorted them at about 70, took my returns a few weeks later at 60 or so, they proceeded to rise to 80 and then within a year drop back to 2.5. I identified another mispricing in NHCS where numerically they were spinning out a company which was being completely undervalued in their current stock price. I asked others “can this really be true,” they said only in general yeah stuff like that happens. I bought a few thousand dollars worth, made the 20% or so return it seemed I was seeing laying on the table a few months later.
The main sense in which the market seems efficient is that the prices are predominantly set using sensible analyses, presumably because those who do not follow a proven technique of picking sensible prices do not survive, so the main component of market efficiency is that the processes for beating the market are broadly exercised and dominating the market. So it is hard to do better than free-riding on that. But does it turn out that some people do better at that process than others? I think the best explanation for what we see is that yes, some do, and that they are a smallish minority is not because they are just the tail of a random distribution, but because of mathematical necessity beating the average significantly can only be done by a minority.
Anyway, thanks for sticking with it and explaining your position to me.
To expand even further on my critique: you are placing a huge amount of weight on 9 students, of unknown veracity, out of an unknown number of students (itself out of an unknown number of millions of people who have tried to beat the market over the past century), who have not released audited records much less ones comparing them to indexing, who started half a century ago (which is the investing dark ages compared to what goes on now, in 2013), and at least one of whose successes seem to be partially explained by non-efficiency-related factors?
This is roughly as convincing as Acts of the Apostles documenting the 12 apostles’ successes in beating the (religious) market and earning converts.
Those angel investors are forfeiting diversification and so can easily earn below-average returns. EMH doesn’t mean that you cannot deliberately contrive to lose money.
I think in an adversarial environment where everyone claims to be able to beat the market and you should give them their money, and there are compelling theoretical reasons that any beating of the market would wipe out whatever advantage was posssed, there is not such a gigantic difference.
Congratulations on your day-trading success. You know what happens to most of them, right?
Under EMH is pretty hard to deliberately and consistently lose money. It’s very easy to get additional risk (e.g. by not diversifying), but I don’t think EMH envisions assets with negative expected return.
Mm, the way I remembered was that by not diversifying, you were taking on additional uncompensated risk; not diversifying wasn’t completely neutral, expected-value wise. (Also, there’s obvious ways to guarantee losing money: trade a lot. The fees will kill you.)
Yep, that’s what I said—that you can easily get additional risk by not diversifying.
And the trading fees are outside of EMH—there are certainly plenty of ways to reliably lose money in the real world, but not in the EMH world.
I said ‘uncompensated’ risk.
EMH doesn’t say anything about uncompensated risks.
To get to risk premium you need something like CAPM or APT which are a different kettle of fish.
Actually the records ARE audited, they ARE compared to indexing, and those records and comparisons are reported by the original article I mentioned, which I finally link to here.
If a professor’s students dominate some part of engineering or biology or chemistry, it is generally taken as evidence that the professor was teaching something real. I suppose if we had an Efficient Knowledge Theory we would understand that going to Caltech or MIT was as wasteful as picking up $20 bills on the sidewalk (which don’t exist in a classic EMH joke).
Should we be questioning whether a good education in philosophy or math or physics or engineering or biology or… is just a mismatch between the power of random chance and the human bias towards seeing patterns? Or is there something special about learning how to value companies that puts it in a category of analysis that is different from all other observations of the effects of knowledge?
In any case, the article linked discusses the randomness hypothesis extensively pointing out among other things that the various investors reported upon had exceedingly small amounts of overlap in what they actually invested in.
Gwern, these comments are not so much aimed at you, you have obviously been down these roads and decided which way you would turn. They are aimed at anybody reading this who is still not sure about EMH. The article linked is excellent and written by a guy who walks the walk better than anybody else in human history (so far).
I don’t see any mention of how they were audited (Buffett merely says that they ‘were audited’, no mention of by whom, when, what the audits said, whether he saw the results, etc, and offers as reassurance that checks were paid for the appropriate amounts, which is not my problem here), and if you really want to nitpick, then I would bring up that Buffet does not talk about ‘9’ students, he actually talks about 4 people who worked for Graham, tells us that ‘it’s possible to trace the record of three’ (well, there’s some selection bias right there...) and does not explain how the 3 partners did (more selection bias), and some of his other examples are questionable at best—including his very good friend Munger, including two funds he ‘influenced’ (while disclaiming that he might have influenced any other funds and this isn’t cherrypicking, which I don’t understand how he can honestly say how he knows for sure he has not similarly influenced any others), reporting different metrics for different examples (why is Munger compared against the Dow while others are compared against the S&P?), not comparing against an index (table 8), and some do not beat the comparison index at all (Table 9, Becker, underperforms S&P by 3%)
Buffett doesn’t dominate the markets, and the proper comparion is to ideas, not students—if a single professor’s students dominated, I’d be more inclined to suspect corruption or logrolling or the professor being a genius at academic infighting and bureaucracy...
Markets are very different from electronic circuits or particle physics or philosophy or engineering. Circuits don’t care if you found a more efficient way to design them. The properties of steel will not change when you discover it lets you build profitable bridges.
Er, yes, there is. That’s kind of the point of the efficient markets concept! Markets are unusual and special in that the attempt to find predictable regularities leads to the exploitation of the regularities and their disappearance. (Eliezer describes this as “markets are anti-inductive”, which is not wrong, but I’m convinced there must be some more intuitively understandable phrase than that.)
Is that one article really the best, solidest, most convincing criticism of EMH you can come up with, which you think will persuade people reading this conversation that EMH is to a meaningful degree false and markets are often beatable—some cherrypicked questionable examples from the dawn of time?
In its way, yes it is. You get a guy who has impeccable credentials, a massive public record who thinks he has been investing intelligently for decades, who if he IS performing randomly is a few sigma out on the positive side of the random distribution. You get to see what he has to say about what he thought he was doing, how it fits with what a whole bunch of other people were doing, a cogent description for why it might work, and a bunch of numbers about how it does indeed seem to work. Buffett understands the idea that he could just be lucky and he addresses it.
If you think the best explanation of Buffett’s life and results are that he has been fooled by randomness, then you are a very different judge of character and information than me or millions of others like me.
If the EMH was “the markets are really really efficient, it is hard to produce alpha (outperformance), hard to know when you have alpha, and easy to fool yourself because of human biases” then who would argue with that? Not me. But that step from “really hard” to “impossible” is unreasonable. It is not impossible to be a great baseball player. It is not impossible to consistently beat other players at poker, even though everybody playing has the same information, on average across all the hands. It is not impossible to understand 10 languages, even though to most of us most of them sound like noise.
If EMH was right, wouldn’t the smartest, most quantitative participants in the market have figured that out? Wouldn’t Renaissance Technologies have 1) failed, and 2) figured out that their failure was consistent with randomness where they thought there was order?
EMH is the hypothesis that because bunches of smart people all work to figure out what the best investment is, there can be no excess returns available to the smart people who all work hard to figure out what the best investment is. Well if there are not excess returns available to them, why do they do it?
Isn’t EMH the hypothesis that, for EVERYBODY in the market, it would be more efficient to free ride and use your intelligence on something where you can actually produce a return?
Isn’t EMH ultimately a big floppy tent held up by a tent pole which the EMH’ers deny exists?
First, let me point out that I put a fair amount of work into pointing out all those flaws and holes in your last best citation, and I’m a little annoyed that you completely ignored all of them in favor of saying “but Buffett is so high-status and I like him so much”. Yes, and George W. Bush famously said of Vladimir Putin, “I looked the man in the eye. I found him to be very straight forward and trustworthy and we had a very good dialogue. I was able to get a sense of his soul. He’s a man deeply committed to his country and the best interests of his country and I appreciate very much the frank dialogue and that’s the beginning of a very constructive relationship.” We all know how that turned out.
What you think about Buffett’s “character” is irrelevant to me, and for me, further emphasizes your extremely poor reasoning in this area—that when pushed back, you resort to one man and your beliefs about his “character”.
I don’t know why RenTech performs as well as they seem to. Presumably it’s not the same reason that Madoff was able to beat the market for so many years in contravention of EMH. Perhaps it was the same reason SAC did well (insider trading) and they simply haven’t been caught yet. Or maybe there were some inefficiencies back when they started which they erased and have since been coasting on their reputation. Given that it’s a very private hedge fund, we’ll probably never know.
Because there is demand for investment services, considerable cognitive biases at play along with wishful thinking (‘I will be the next Buffett!’), and normal profits available. After all, if no one was there taking even the normal profits, there would immediately be excess returns attracting people to the enterprise...
No.
No.
So your hypothesis is that some process ties all the people there needing to provide skull sweat to get normal returns (and create a market for everybody else that is efficient) works equally across all such players? It sure doesn’t work that way in any other human enterprise I can think of. Intel and AMD produce different quality chips for laptops. At the other end of tilt, Intel and Qualcomm produce very different quality of chips for mobile. The physics department at Caltech produces a very different product, research wise and teaching wise, than the physics department at USC. Writer Stephen King produces a very different quality of novel than do a thousand or more other authors populating the increasingly virtual shelves of bookstores. Even here on lesswrong, some of us write wonderful stuff which is read by many and admired, while others of us struggle to get our karma up to 1000 and then hang on by our fingernails stopping from saying what we really want to say to keep it there.
So why on FSM’s tomato-colored earth would you expect these financial creators of efficiency to all get the same results from their efforts?
And when shown the spread in effectiveness in results, to deny the evidence of your own eyes and declare it all to be the distant tail of millions of coin flippers?
It doesn’t seem like a stretch to you? It doesn’t seem that the evidence is strong that the market is VERY efficient, but that the evidence is not there that it is COMPLETELY efficient?
If you want I’ll go through them point by point.
Presumably you can see the difference between your stating these are NOT audited, and then when pointed out that they are, you back off to this.
The results of the audit are the results in this article. That is, these are results reported which survived the audits.
In many of the cases, the audits are “typical” of the investment advisory business, but I do not know what that means exactly. But it is a level playing field against all other investment advisers.
Also for a few, not all, of the investors cited here, they run/ran for decades public investment businesses. Isn’t the preponderance of your Bayesian a posts that if at least these members of this widely read, cited, and discussed “superinvestors” article was just wrong, that this would at least have lead to traceable reports of the discrepancies on the internet, finable with google search?
To the extent your objections amount to “Buffett could be an idiot and a fraud, either not knowing or not caring what it means to make these claims” my answers are going to be we have 5 decades of impeccable record, if you think Buffett is that unreliable then generally there is no arguing with you as anybody who says something with which you disagree you will question as an idiot or a fraud. If you cannot tell that Buffett is not an idiot or a fraud, or have not followed him well enough to be sure one way or the other, then I would suggest you have no business weighing in on the subtle subject of whether the market is so inefficient that the best investors in the world are just coin-flippers.
I suggest relying upon Buffett because you and everyone else out there who can read has infinitely more reason to rely upon Buffett than to rely upon me. And further, what is needed int he discussion of EMH vs non-EMH is not some brilliant new insight that I can provide that you haven’t seen somewhere else already. EMH vs non-EMH is a subtle question, is the market so efficient that Buffett can’t consistently beat it without committing a crime, either insider trading or some other information-twisting fraud, or is it just a little less efficient than that? The “insight” I have is that what pushes it towards efficiency is competing analyses on opposite sides of each trade. The “insight” I have is that every bit of evidence suggests that in business some people have superior skill or algorithms or SOMETHING and are more successful than others. And they can do it serially, command high prices in very competitive markets, blah blah blah, and show EVERY BIT as much evidence of being “real” as are great pitchers or tennis players or tenors or talk show hosts or porn stars. And your case is that no, with investing it is different, the people who do the work are so smart that they get it right in an unbeatable way, but so stupid that they don’t even realize they would be better off free riding.
What is needed is not any great insight from one or the other of us, I don’t think, but evidence that is hard to deny that yes, the market can be beat. I think that evidence would consist of market beaters coming from a narrowly defined group of people who set out to beat the market by studying it and allowing evidence to drive their future hypotheses and efforts. And what do we find in the market? Exactly that, market beaters are smart and talk in terms of causality, of what makes a business great, of where the momentum traders and the chartists missed the boat.
But my causal chain of how the market could be merely VERY efficient has been, I hope, presented by now. Let me know if it hasn’t.
As much as you might hypothesize that we will not see securities markets make the same mistakes they have made in the past, does the evidence support that? And in any case, the idea that markets do learn or have learned SOMETHING supports only the VEMH, the very efficient market hypothesis, which is not controversial. By this I mean the hypothesis that it is hard to beat the market, because all the easy stuff has been figured out and is properly accounted for by the bulk of the traded money in the market.
I tracked Chiplotle stock on and off from around 2000 forward. There were two classes of shares, A, and B, with the B’s trading at a very consistent 10% discount to the A’s. I would check once or twice a year to see if this difference persisted, and it did. The thing that was surprising was that the documentation of the company explained that these shares had equal value, represented identical fractions of the total company. Why they traded at a 10% difference I never saw an explanation, and I always questioned whether there was some detail I was missing. Here, in late 2007 is documentary evidence that the difference persisted. Here, two years later, is Chipotle’s report that they were eliminating the two classes in favor of one class, and that the exchange rate would be 1:1 just as I had always believed.
In my case, I am an electrical engineer/physicist, trying to concentrate on building new cell phone algorithms for at least a few hours a day. Instead of organizing the financing to exploit this weird inefficiency at low cost, I just checked in on it every year or two. Wanting to see if I was right. Had I been a professional trader, I would have looked more at creating an arbitrage on the A and B shares and capturing the collapse of the arbitrary pricing difference. As an amateur I didn’t know if it would ever collapse, and the brokers are neither smart enough nor dumb enough to let me buy the As and short the Bs without a lot of capital in my account to anchor what they see as two uncorrelated risky bets.
My point here is this is just ONE of MANY possible stories of moderate sized inefficiencies I have seen with my own eyes. Others I have traded. Yes, every one of them is an anecdote. The plural of anecdote is not data. But a bunch of anecdotes like that creates, it would seem, market beating performance for many traders trading different stocks.
Maybe markets COULD be different than circuits and so on, and maybe as computers and AI takes over more and more, they will get more and more efficient. But even then, the most powerful AIs will be beating the market, even as they essentially set th prices at levels that make it incredibly hard for anybody else to beat the market. The thing that drives market makers is not their stupidity, but their intelligence and rationality. Seems to me.
THIS is a hypothesis. And the only word in that hypothesis I will argue with is the last one: disappearance. The predictable regularities don’t disappear from the time-stream of prices, if there is a mispricing at 2:31 PM on Thursday it is frozen there in the permanent record. What changes is how long it takes for the record to close those various gaps. Maybe before computers a broad class of inefficient prices were never traded away. Maybe in the 1980s a broad class of inefficiencies were capitalized upon by people with computers over the course of a two week period. Maybe by the 2000s those same inefficiencies were traded away within hours or minutes.
But my points are: 1) we are not arguing efficiency vs inefficiency, we are arguing too efficient to beat vs nearly too efficient to beat and 2) without the inefficiencies, no one would be there to pay the actors making the market more efficient by trading the inefficiencies, and that no, it is not their stupidity that keeps them working for free.
I hope this is what you wanted when you suggested I was ignoring your point and merely arguing pro hominem, citing people who I thought should be much more believable than I am. If I missed anything that still seems critical, flag it to me and I’ll answer it.