Efficient Market Hypothesis - EMH

[yielder]

Since we don't have a thread on the subject and I didn't want to take the RRSP 30% foreign content limit GONE!!! thread off on a tangent in response to martingale's comments on EMH .................

It's a terrible mistake to throw the EMH out on the grounds that it has a few problems. The people who do this throw up all kinds of anecdotes to support their alternative theories and trading strategies, but the reality is that those other hypotheses have even more problems than the EMH, and are even less well supported by the data.

Care to explain why DFA are active not passive funds? Most strange that many of the EMH academics are associated with an investment style that is decidedly not EMH. They are closet behaviourists.

The few attempts of the EMHers to deal with behavioural finance are weak at best. Malkiel's explanantion of the Internet Bubble is telling:

Another stock market event often cited by behavioralists as clear evidence of the irrationality of markets is the Internet “bubble” of the late 1990s. Surely, the remarkable market values assigned to internet and related high-tech companies seem inconsistent with rational valuation. I have some sympathy with behavioralists in this instance, and in reviewing Robert Shiller’s (2000) Irrational Exuberance I agreed that it was in the high-tech sector of the market that his thesis could be supported. But even here, when we know after the fact that major errors were made, there were certainly no arbitrage opportunities available to rational investors before the bubble popped.

No opportunities? It's called short selling!!!!!!!!!!

He goes on to say "We know now, with the benefit of hindsight" and says even the pros had it wrong. Most but not all. There were some who recognized that dot.com multiples were folly, that companies with no revenue, not to mention no earnings or cash flow, were mostly illusion.

Direct tests of the actual performance of professionals, who often are compensated with strong incentives to outperform the market, should represent the most compelling evidence of market efficiency.

It's arguable that compensation which is paid on a short term basis will generate short term performance chasing behaviour, ie, style drift.

One can count on the fingers of one hand the number of professional portfolio managers who have managed to beat the market by any significant amount.

Exactly. But you won't find out why in the CRSP database.

[NormR]

Seems like a good place for What Cost “Noise”?

Benjamin Graham was fond of saying that in
the short run the market is a voting machine but in
the long run it is a weighing machine. The volatility
tied to the noisy short-term departures from the
rational quest for true fair value dissipates in the
very long run and thus has little impact on annualized
returns relative to the constant search for true
fair value. But if merely one-eighth of the idiosyncratic
volatility is noise, rather than sensible reaction
to the changing fundamentals that set the
ultimate, unknowable true fair value, then we will
see 14 percent noise in pricing.

This result would cut the return on any portfolio
that tracks this noise—as if it were reality—
by 2 percent a year. Cap-weighted “market portfolios”
track prices, so they may be pulled down by
2 percent a year as a consequence of simple noise
in market pricing. Even if the noise is in the opposite
direction, there is no way for this noise to boost
returns for the cap-weighted market portfolio.
Underpricing is just as damaging to the return on
a cap-weighted portfolio as overpricing. Mispricing
relative to an unknowable true fair value is still
mispricing. This noise is pure loss.

[martingale]

Behavioral finance has changed the face of academic investment theory, yes. But, contrary to the tone you've set, behavioral finance is less of a refutation of the EMH than a superceding of it. The data that supported the EMH didn't go away; it turned out there were corner cases that were better explained by the addition of some behavioral factors.

To conclude from this that you should abandon a core indexing strategy is nonsense. The correct conclusion is a core indexing strategy with deviations to capture what behavioral financial theory has uncovered. I think you'll find that is what DFA is doing, and in the other thread, that is also what I recommended.

Also it's worth noting that the provably exploitable trading strategies all involve a lot more money and access to the markets than either you or I have. Individual stock picking has not been shown to be a viable strategy by behavioral finance either. What has been shown is that massive hedged investments sometimes outperform the market over periods of less than a year, but neither you nor I have access to the trading methods that exploit those. To make those anomolies robust to trading costs you have to get trading costs down very low; and you have to be able to use the money from a short position to go long. Neither of us can do either of those.

Putting forward the idea that you should go out and pick individual stocks because there are difficulties with some corner cases in the EMH is folly. The EMH will be superceded by some better theory that includes the findings of behavioral finance--but it will agree 80% of the time with the EMH.

[Brix]

Care to explain why DFA are active not passive funds?

Probably shouldn't be jumping in here, but among the reasons would be that active management, whether as a basic principle or an additive/enhancement to a passive approach, is more easily 'productized', branded and marketed. Passive indexing tends toward commoditization at the product level.

Even advisory firms in which noted finance academics participate promote elaborate, proprietary, portfolio-construction methodology.

Unless you're some sort of purely co-operative, profit-sharing entity it's likely you're likely going to want to construct a 'moat' of distinctiveness around your offerings. Index funds, no. IndexPlus(tm), yes. IndexPro+(tm) -- we've got yours right here!

[like_to_retire]

Passive indexing tends toward commoditization at the product level.

Indeed, and if some spin isn't attached to it, doesn't it comes down to 'who's got the lowest costs'. Well, no money to be made there, I better come up with my take on 'enhanced super' indexing. I can beat it...I know I can, I know I can

ltr

[martingale]

I'm blown away that this "what cost noise" article was allowed to run in something calling itself a journal. It smacks of innumeracy. I don't know whether the innumeracy is an honest failure of the author to understand the mathematics, or whater it is meant to be intentionally misleading.

The article would lead you to believe that any statistic with a standard deviation (like "market return", for example) has a built in drag. It does not. The author got the mathematics wrong: It is not "+50% -50% = -25%" it is rather "1.5 * 1/1.5 = 1".

The article also has some nonsense about how an indexing strategy sometimes over or undervalues a particular security by 40% or so. Well, only on paper. An indexer does not actually execute any trades based on this information--they hold their position, and the shift in value automatically adjusts the cap weight. Also, an indexer does not hold just one security, but very many. Only if there is systemic bias would the indexing strategy go wrong--random noise cancels out, with equally many securities over and under valued. Now it's true there are people who claim there is systemic bias in the market--but that is not the point this author is trying to make. There is no systemic error introduce by random noise.

As the article points out nobody, no matter committed to the concept of an efficient market, believes that there is no error in security pricing. Rather, the question is whether there are any deviations that are robust to trading costs. Trading costs MUST include the cost of research, the cost of equipment, and labour, as well as trading fees. Few anomolies have been discovered that are even robust to trading costs.

Here is where lay people, those who have inadequate education in economic theory, typically misunderstand the EMH and other economic theories about efficiency:

In ordinary economic theory there is a difference betwen "normal profit" and "economic profit". Normal profit is the return that could have been obtained from the best alternative forgone. In other words, if you had put your time and energy into the next best thing you could have been doing, the return on that is your normal profit. Economic profit is a return in excess of the next best alternative--money that a business returns beyond the other opportunities available to the entrepreneur. (The technical term here is "opportunity cost", and the point of market equilibrium occurs when the return equals normal profit equals opportunity cost).

The same concept is embedded in the EMH. Sure, you can earn an excess return by spending time and energy researching stocks. The problem is that the excess return is at most normal profit, or, more often, less than normal profit, in which case you are making an economic loss. Despite beating the market, your time and energy would have been better invested in some other activity. This is why I make the joke that you would have been better flipping burgers at McDonalds (and investing your salary) than researching stocks: Flipping burgers in exchange for minimum wage is an opportunity forgone. Most of you here could earn in excess of minimum wage and so your opportunity forgone (your required normal profit) is higher still.

In real terms, what does that mean to an investor? It means this: The EMH is comfortable with the idea that an active manager can beat the market, so long as the active manager's salary, trading costs, and other expenses roughly equal the excess return. In such a situation, the market is still efficient as the investor has obtained nor more than normal profit. Someone with the abilities and resources of that active manager had other opportunities with a higher return than active management.

In reality study after study has shown that, from a retail investor perspective, active management mutual funds return an economic loss. The fund itself is probably earning a normal profit, but the fund managers extract enough money from the return (the "MER", plus trading costs) that for the investor it's an economic loss.

This is typically measured by the extent to which the return on the fund trails the index.

[jiHymas]

The few attempts of the EMHers to deal with behavioural finance are weak at best. Malkiel's explanantion of the Internet Bubble is telling:

Another stock market event often cited by behavioralists as clear evidence of the irrationality of markets is the Internet “bubble” of the late 1990s. Surely, the remarkable market values assigned to internet and related high-tech companies seem inconsistent with rational valuation. I have some sympathy with behavioralists in this instance, and in reviewing Robert Shiller’s (2000) Irrational Exuberance I agreed that it was in the high-tech sector of the market that his thesis could be supported. But even here, when we know after the fact that major errors were made, there were certainly no arbitrage opportunities available to rational investors before the bubble popped.

No opportunities? It's called short selling!!!!!!!!!!

Another opportunity is called "issuing". But be fair to Mr. Malkiel - he does discuss the problematical mechanics of short-selling in connection with the Palm Pilot / 3-Com arbitrage: I can't remember ever having seen an EMH paper concede that the market is not, in fact, infinitely liquid.

My greatest complaint about the article is that it is assumed that transaction costs (which do not include the compounding effects of taxes on capital gains) are large and negative.

"Mechanical" trading costs (commissions, settlement costs, etc.) are, for practical purposes, zero. (OK, OK, "insignificant"). The other component of trading costs is "market impact". If transaction costs were always negative, dealers would be best described as bankrupt philanthropists, but they continue to do rather well from what I see in the papers. Somebody's market impact COST is somebody else's market impact GAIN. I harp on the issue, I know, but selling liquidity is an extremely valuable and well-rewarded endeavor. But the dealers siphon off a huge chunk of this excess return and, according to those academics who limit their returns universe to a classically defined "buy-side", it just disappears ... which helps lower the average return of the examined universe.

A lot of the problem here is that there are many consultants who are averse to turnover. They're afraid of churning. It doesn't fit the buy-and-hold philosophy - which is also pitched by the managers themselves to clients. But churning, in and of itself, is only a problem for (i) taxable investors trading in securities which could reasonable double in time (i.e., equities, not bonds) and (ii) investors who have mechanical trading costs too high relative to expected incremental returns from trading.

Churn, baby, churn! A bond desk might turn over its net inventory a dozen times a day and be very pleased with their efforts.

Malkiel also attempts to explain away excess returns on low Price/Earnings stocks with a claim that it is confounded with a low Price/Book effect that he then re-labels as a risk-factor.

Additionally, I have found that most research in this area is performed using regression analysis, which is totally unsuitable for excess returns analysis. If I buy a stock at $25 in January because I think it will go up to $30, and it is priced at $30 from July-November, returning to $25 in December, I call that a win. Regression calls it a loss. It's hard enough to make the call that an equity is undervalued - predicting the precise time that it will take for the market to reprice it is something that I would not care to attempt. Classical regression with common end-points is simply not a suitable tool for securities analysis.

[adrian2]

I'm blown away that this "what cost noise" article was allowed to run in something calling itself a journal. It smacks of innumeracy. I don't know whether the innumeracy is an honest failure of the author to understand the mathematics, or whater it is meant to be intentionally misleading.

The article would lead you to believe that any statistic with a standard deviation (like "market return", for example) has a built in drag. It does not. The author got the mathematics wrong: It is not "+50% -50% = -25%" it is rather "1.5 * 1/1.5 = 1".

The article also has some nonsense about how an indexing strategy sometimes over or undervalues a particular security by 40% or so.

Robert Arnott is too much of an academic to be caught up pants down on this one: read his footnote #2

2. This conclusion assumes that the market’s pricing errors are
normally distributed.

which we know it's an approximation useful only for small values of volatility: a stock can be up 150% but cannot be down 150% - that's why usually a log normal distribution is used for stock prices.

[adrian2]

No opportunities? It's called short selling!!!!!!!!!!

In addition to James' quote on 3-Com / Palm impossibility of executing an in-your-face arbitrage due to the lack of stock to short, I'll mention again my hands-on experience with an Air Canada "guaranteed profit" trade involving options + short sale, from which I was bought back twice.

And the obvious "replace a short with buying a put" will not work in extreme cases like this because the whole market in the specific security (stock or options) can become, and stay, distorted.

[jiHymas]

No opportunities? It's called short selling!!!!!!!!!!

In addition to James' quote on 3-Com / Palm impossibility of executing an in-your-face arbitrage due to the lack of stock to short, I'll mention again my hands-on experience with an Air Canada "guaranteed profit" trade involving options + short sale, from which I was bought back twice.

And the obvious "replace a short with buying a put" will not work in extreme cases like this because the whole market in the specific security (stock or options) can become, and stay, distorted.

On the other hand, it has occurred to me that it is precisely such rudeness on the part of the real world in refusing to allow efficient arbitrage that embarrasses the EMH.

[Shakespeare]

It is the leap from "mostly efficient most of the time" to dogma that is my fundamental difficulty with many EMH apologists.

Over many years of market-watching, I have seen too many obvious mispricings to accept that EMH - particularly strong EMH - can be accepted unquestioningly.

I am quite willing to trade my time for the opportunity to pick up the free money that doesn't exist. However, for most people, indexing remains the best approach, since the required contrarian mindset is very difficult (and usually expensive) to acquire. Nonetheless, once you have paid the tuition, the opportunities can be taken advantage of when recognized. But the tuition can be expensive, and has a high present value because it is generally paid early in one's investment career.

[martingale]

It's worth noting that the market can be efficient even if the EMH is inadequate. The EMH claims that the market is an efficient application of a specific set of factors ("three factor model", etc.). Behavioralists have shown that these factors are inadequate--that real world investors judge risk differently than the pure rational investor envisioned by the EMH.

Adjust the EMH to account for the correct definition of risk (ie: the one actual investors care about, and act on) and suddenly efficiency is restored. So you wind up with an invalid/inadequate EMH, but nevertheless, an efficient market.

I firmly believe that the market is efficient. On the other hand, I also believe that the EMH is likely to be replaced by a better model which more accurately captures the factors that apply to real investors. The model in the original EMH is too simplistic.

This is what I mean when I say that the behavioralist theories should ultimately lead to a supercession of the EMH, rather than a rejection of it, just like Newton's theory of gravity was superceded by Einstein's. Rather than saying Newton got it wrong, and Einstiein right, it's better to say Einstein got it less wrong.

That is the situation with the EMH and behavioral finance. The EMH is wrong. Behavioral Finance should lead to a less wrong version. People who disregard all of this and go off to concentrate on a few "winning stocks" get it wrongest of all.

[martingale]

Another opportunity is called "issuing"..

The current or past issue of the Journal of Finance included a study that looked into whether corporate managers are capable of timing the market when they issue shares. It turns out they cannot. Over the past few decades corporate managers have not done any better than random in their ability to pick the best time to issue shares.

Corporate managers DO issue more shares when the prices are at historic highs (ex post, looking back) but an analysis of their issuing indicates that they had no power to predict future prices (ex ante). Prices were just as likely to go higher or lower regardless of the activity of corporate issuers. Were issuers able to time the market then there should be evidence that the market went down more often than up after a successful issue. That is simply not the case.

So, this is yet more evidence that the core of the EMH hangs together, and in fact it is evidence for some of the stronger forms of the EMH. There really are a few situations in which it appears not to, but nobody should be under the illusion that those situations are commonplace. They are not.

[yielder]

Aw, you guys are no fun. Here I was hoping to get roughed up and most of you agree in varying degrees.

behavioral finance is less of a refutation of the EMH than a superceding of it

su·per·sede

1. To take the place of; replace.
2. To cause to be set aside, especially to displace as inferior or antiquated.

My emphasis.

To conclude from this that you should abandon a core indexing strategy is nonsense.

Martin, I never concluded that you should abandon indexing. In fact, given the general & high-priced incompetence of the great majority of fund managers and the general inability of individuals to pick stocks consistently successfully over the long run, indexing is the only way to go unless there is no product available.

DFA may be be using behavioural theory but Fama still sings the old tune publicly and does so poorly - "Fama finds that apparent underreaction to information is about as common as overreaction, and post-event continuation of abnormal returns is as frequent as post-event reversals." It's not that there is underreaction or overreaction but that there is incorrect reaction. Jean Coutu announces the acquistion of Eckerdt and the market reacts as if higher earnings were in the bank. Jean Coutu announces a poor first post-acquisition quarter and the market reacts as if the acquisition has destroyed Coutu.

Individual stock picking has not been shown to be a viable strategy by behavioral finance either.

And it never will be as long as researchers persist in regressing the hell out of the CRSP database. Malkiel in his Appendix 9 has pointed the direction in which to look.

To make those anomolies robust to trading costs

I don't dispute Malkiel's observations about trading. From a behavioural perspective, the more frequently you trade, the more you are likely to fail.

Putting forward the idea that you should go out and pick individual stocks because there are difficulties with some corner cases in the EMH is folly.

For most people, yes. But anyone can learn how to pick stocks effectively. Controlling behaviour is something else that is far from easy.

Despite beating the market, your time and energy would have been better invested in some other activity. This is why I make the joke that you would have been better flipping burgers at McDonalds

You and I have debated this one before so I won't go over plowed ground.

[jiHymas]

Another opportunity is called "issuing"..

The current or past issue of the Journal of Finance included a study that looked into whether corporate managers are capable of timing the market when they issue shares. It turns out they cannot. Over the past few decades corporate managers have not done any better than random in their ability to pick the best time to issue shares.

Corporate managers DO issue more shares when the prices are at historic highs (ex post, looking back) but an analysis of their issuing indicates that they had no power to predict future prices (ex ante). Prices were just as likely to go higher or lower regardless of the activity of corporate issuers. Were issuers able to time the market then there should be evidence that the market went down more often than up after a successful issue. That is simply not the case.

I was referring to the ability of tech companies (in the period, say, 1997-99) to raise huge sums of money to finance on-line pet-food stores. These issuers were extremely successful in their efforts to time the market.

In the bond market there is certainly an intent (by more 'normal' issuers than those to whom I was referring) to time the market and evidence that it is indeed attempted (The maturity of debt issues and predictable variation in bond returns) but at least some indication that this is not successful ( Interest Rates and The Timing of Public Issues of Corporate Debt), a result that, quite frankly, surprises me. I will read that last paper more carefully.

Adjust the EMH to account for the correct definition of risk (ie: the one actual investors care about, and act on) and suddenly efficiency is restored. So you wind up with an invalid/inadequate EMH, but nevertheless, an efficient market.

This is getting awfully close to playing with words - you can simply claim that every incidence of excess return was accompanied by excess risk of some nature and therefore the strong EMH holds.

It's not clear to me where you believe market impact costs disappear to under the EMH.

[martingale]

I was referring to the ability of tech companies (in the period, say, 1997-99) to raise huge sums of money to finance on-line pet-food stores. These issuers were extremely successful in their efforts to time the market.

So, the article in the JoF examined that data and concluded that they weren't actually successful in timing the market. Rather the issuing of shares can be fully explained by ex post share prices. Ex ante prices had no impact whatsoever on the amount of shares issues. In plain English, when share prices are at a historic high (looking back) managers issue more shares. When prices are at a historic low (looking back) they issue fewer. Yet being at a historic high/low is no indication of whether future prices will be higher or lower, and similarly, the rate of issues (which depends on historic prices) contains no information about the future either. In yet other words, there was a correlation between past share prices and the rate of issue, but no correlation between the rate of issue and future prices. So, no evidence that corporate managers have any skill at all in timing the market.

In the bond market there is certainly an intent (by more 'normal' issuers than those to whom I was referring) to time the market

There are a lot of people who intend to time the market. There are people who get lucky and make the right trades and then say that was their intent and describe their plan. As Malkiel points out in RW if you held a coin flipping contest with a few thousand contestants, eliminating everyone who guessed heads/tails wrong, then after a few iterations you'd be down to a few "experts". Some of those experts would claim to have a skill in predicting coin flips, and would tell you their method, and so on. In reality, sometimes you get lucky, sometimes you don't.

Pointing out that there were a few people who got lucky with the timing in 2000 is a nice anecdote. Yet over the long run corporate issuers behavior is completely predicted by historic share prices, and contains no predictive power.

This is getting awfully close to playing with words - you can simply claim that every incidence of excess return was accompanied by excess risk of some nature and therefore the strong EMH holds.

Sorry, I am misunderstood. I am pointing out that saying "the EMH is wrong" is different than saying "the market is inefficient." The market very probably is efficient--just with a different utility function for investors than the one that the EMH assumes. This is where the behavioral finance research is going: towards the idea that real investors are very much not like the rational investor in the EMH. They have a different utility function, and therefore the market will reach a different equilibrium, with different prices, than the equilibrium and prices predicted by the EMH.

The theory of efficiency in competitive markets has been around a good while longer than French, Fama, Malkiel, or the EMH.

It's not clear to me where you believe market impact costs disappear to under the EMH.

They don't disappear. It's not clear to me why you think the EMH is incompatible with impact costs? The EMH does not assert that securities will trade at their equilbrium prices. It asserts that they will trend close enough to their equilibrium price that you cannot profit from the differences, given impact costs, trading fees, research costs, salaries, resources, and the other money you have to spend.

[jiHymas]

It's not clear to me where you believe market impact costs disappear to under the EMH.

They don't disappear. It's not clear to me why you think the EMH is incompatible with impact costs?

The market price of the security has been changed by a non-fundamental circumstance. That is to say, the counter-party to the trade has agreed to execute in exchange for a price concession. If the price of the security has gone down, for instance, solely because there was a highly motivated seller, it may be expected to rise as the temporary imbalance gets absorbed.

Another example is the necessity for a distinction between "on-the-run" and "off-the-run" bonds. The on-the-run issues are benchmarks, the most recently (re-)issued bonds of their tenor. They trade at a premium to off-the-run issues, until a new issue occurs.

The point is that the market pays up on a regular basis for liquidity. Selling that liquidity is a very profitable business; the existence of that premium is not compatible with any strong-form EMH I have yet seen.

The EMH does not assert that securities will trade at their equilbrium prices. It asserts that they will trend close enough to their equilibrium price that you cannot profit from the differences, given impact costs, trading fees, research costs, salaries, resources, and the other money you have to spend.

So is it your claim that the brokers' institutional trading desks all lose money? There have been recent reports (Flashy Bay St. trader faces probe) that Scotia's Preferred Desk made $45-60 million annually. Is this a mere anecdote?

[yielder]

I firmly believe that the market is efficient.

The market is not efficient; it is irrational. I have the sense that you cling to EMH because to reject it suggests a rejection of indexing, your preferred investment process. I believe that you can make a stronger argument for indexing based on irrationality than on efficiency. There are enough exceptions to conclude that EMH is flawed to the point of being worthless. Irrationality (Behavioural Market Hypothesis???) will have most managers and individual stock pickers screwing up, perhaps even consistently, over time. On that basis alone, indexing is the only viable alternative. A disciplined investor using a well-founded selection process can beat the market over time. Using an irrational rather than an efficient argument means that indexing can co-exist along side active stock selection.

EMH and random walk are now defunct although, along the way, they made substantial contributions to our understanding of the market. The problem is that behavioural finance is still so undeveloped that there is a vacuum. It's likely that BF will be fairly slow to develop and always be full of inconsistencies because regressing CRSP will have to be replaced with interview-based information gathering. Surveys are likely, at least initially, to reflect the biases of the researcher. Even as subsequent surveys address this problem, there are likely to be many different explanations for the same market characteristic simply because human behaviour can't be neatly summarized in some statistical test.

I guess it's time for the academics to shut down their databases and come down from their ivory towers to talk to people.

[like_to_retire]

I believe that you can make a stronger argument for indexing based on irrationality than on efficiency

I believe that you can make a stronger argument for indexing based on diversification over the long term than on efficiency?

conclude that EMH is flawed to the point of being worthless

Would active management not consistantly beat the market? I don't believe the evidence supports this....

ltr

[Shakespeare]

The primary argument for indexing is low costs, not EMH.

[jiHymas]

I believe that you can make a stronger argument for indexing based on irrationality than on efficiency

I believe that you can make a stronger argument for indexing based on diversification over the long term than on efficiency?

The primary argument for indexing is low costs, not EMH.

A very strong case can be made for indexing based on diversification and low cost.

There is inefficiency in the market, yes. This does not imply that every active manager has the ability to exploit it. It does not imply that all those who are able to exploit it will be able to do it consistently. It does not imply that the frictional effects of taxation can be ignored*. And it most particularly does not imply that those who can exploit it reasonably consistently will be able to do so by enough to cover their fees - particularly if by "fees" we mean "2.5% on an equity mutual fund".

* I suggest that the most effective investment plan a taxable investor with a time horizon of more than ten years can make is to buy rock-solid blue chips, 100 shares at a time whenever enough cash is accumulated, and sit on them. This does not apply to the entire portfolio, of course, but to the taxable equity core.

[martingale]

The market price of the security has been changed by a non-fundamental circumstance. That is to say, the counter-party to the trade has agreed to execute in exchange for a price concession. If the price of the security has gone down, for instance, solely because there was a highly motivated seller, it may be expected to rise as the temporary imbalance gets absorbed.

In economic literature liquidity costs are routinely used to demonstrate that the market is efficient. This is sort of a straw-man argument against the EMH. You've simplified the EMH so that it doesn't include liquidity costs, and then said that liquidity costs refute your simplified version.

When economists look at the market and judge its efficiency liquidity is not only taken into account, but used to prove efficiency by demonstrating that the liquidity costs prevents participants from earning economic profit.

The EMH does not assert that securities will trade at their equilbrium prices. It asserts that they will trend close enough to their equilibrium price that you cannot profit from the differences, given impact costs, trading fees, research costs, salaries, resources, and the other money you have to spend.

So is it your claim that the brokers' institutional trading desks all lose money?

No, I'm talking about economic profit, which is a technical term. Unfortunately if we want to debate the EMH we will have to introduce some core concepts from economic theory. The EMH is a theory of economics, and assuming that it uses the ordinary definition of "profit" leads to straw-men.

A "normal profit" is the amount of money an entrepreneur could have earned doing something different. The people at the Scotia desk are smart, skilled, highly educated people, with access to a lot of capital. Those resources could have been put to some other money-making task, and "normal profit" is defined to be the return from the best opportunity forgone. "Economic profit" is any profit in excess of normal profit.

The EMH states that if anyone is earning an economic profit then others will notice, and copycat them, until there's no more economic profit to be made. If anyone is making an economic loss, then eventually they will quit trading and do something that earns normal profit. The market equilibrium occurs at the point where all market participants are earning exactly normal profit. No-one expects that to occur--but the EMH states that the market should continually trend towads this equilibrium. The degree of efficiency of the market determines how close it gets.

There's an important difference between "earning exactly normal profit" and "earning no profit", which is what you get if you try and use the lay definition of "profit" when you read the EMH. Earning no profit would be an economic loss: no-one would trade in the market if they were able to earn more money flipping burgers at McDonalds, or working as an accountant, or whatever their best opportunity forgone is.

We also know that the market is not perfect. There is not perfect information, nor is there infinite liquidity, nor are trading costs infinitely low. That's all fine with the EMH, so long as no-one is able to exploit any of those features to consistently earn greater than normal profit.

The word "consistently" is important here. According to the EMH, people WILL sometimes earn economic profit (or suffer economic loss) and it is precisely that event which causes people to enter or exit the market, driving it towards efficiency.

Also note that brokerage houses earn money on things OTHER than trading. They earn money charging fees to their clients, for exampe. They may also be able to consistently over-charge their clients for securities. There's no violation of the EMH in any of that, because that all happens off-market.

[martingale]

Yielder: When I said I believe the market is efficienct, I wasn't simply referring to the EMH. I believe the EMH uses an incorrect utility function for investors. It simplistically assumes that investors seek the maximum return for the minimum risk. They do not. They use a "human" definition of risk, and they demonstrably choose lower profits in some scenarios

Nevertheless, given an appropriate utility function for participants I believe the market to be efficient. We could call that BMH.

[Shakespeare]

given an appropriate utility function for participants

That gets close to changing the parameters to hide the evidence of failure.

There is a story of a mathematician, a physicist, and an engineer asked to test the hypothesis that "all odd integers are prime".

The mathematician said, "1 is prime, 3 is prime, 5 is prime, 7 is prime, 9 is not prime, the hypothesis is false."

The physicist said, "1 is prime, 3 is prime, 5 is prime, 7 is prime, 9 is not prime, 11 is prime, 13 is prime. 9 must be experimental error, the hypothesis is true."

The engineer said, "1 is prime, 3 is prime, 5 is prime, 7 is prime, 9 is prime...."

[adrian2]

I believe that you can make a stronger argument for indexing based on irrationality than on efficiency

I believe that you can make a stronger argument for indexing based on diversification over the long term than on efficiency?

I've probably mentioned this before, but it's a good (re-)read:
The Inefficient Markets Argument for Passive Investing