Traders, let's have a discussion.
- Thread starter DrOlmifon
- Start date
Like all theories, one counterexample disproves the whole theory. I saw a counterexample, took 10 grand and traded it. The inefficiency was as tradeable as I had expected by observing it, and for about 5 years I was doing 30-40 trades per day, having about 90% profitable days, no losing months, and achieving extremely high percentage returns with no peak-valley drawdown above 10%.
Those results are statistically *very* significant, for several reasons:
i) high frequency of trades, approx 5-10,000 per annum.
ii) overheads of 20-30,000 per annum, plus my own salary.
iii) high win rate
iv) winning trade profit in excess of losing trade profit, both average and maximum
v) significant transactions costs, due to the high frequency of trades
vi) bulk of my profits came during a 50% bear market in the S&P 500, where the majority of my trades were long index futures. The index was lower when I stopped trading that inefficiency than when I started in 1998.
This inefficiency eventually went away, I had found several other markets where it was replicable and traded again, with similar win rates albeit lower return on capital due to liquidity constraints. Still way ahead of portfolio investment returns though, and with far lower risk.
I've also observed various arbitrages available in the markets since I started. My first job was at a bond futures arbitrage shop which also had a very high trading frequency, win rate, and annual return relative to drawdown. The existence of arbitrage opportunitites by itself disproves strong market efficiency. I saw numerous isolated arb opportunities.
So basically, the reason I don't agree with strong market efficiency is that I've seen on several occasions, track records based on tens of thousands of trades which are statistically significant way beyond the confidence level that any scientific testing demands. The fact that I've seen this from several firms/traders, trading totally different strategies and instruments, and the fact that my own "experiment" was forward looking as opposed to data-mining, makes it highly improbable that it's the result of random chance. Therefore the logical conclusion is that it was almost certainly down to inefficiency and the skilled exploitation thereof.
If in future markets become competitive to the extent that no one can find efficiencies that last long enough to outperform consistently, then I would come around to your point of view and invest in a diversified investment portfolio. But so far I see further examples of inefficient market behaviour, so I'm continuing to trade as long as it continues to offer such good opportunities.
As for academic evidence for efficiency, that's impossible since you can't prove a negative. All you can say is that there currently is not evidence supporting inefficiency. This then raises a problem - are academics aware of all the evidence? There are logical reasons why they would not be:
i) lack of ability. Since pay in the markets is so much higher than in academia, most people who are truly skilled at finding inefficiencies will be in the private sector rather than at universities. Those left behind will be of poorer quality and therefore less likely to make breakthrough discoveries.
ii) economic incentives. If an academic does discover a true market inefficiency, they will earn far more money by trading it or selling the idea, than by publishing. Therefore all good insights are likely to remain secret rather than be disclosed.
iii) lack of market experience. Most academics do not trade full time, which puts them at a disadvantage in discovering and trading inefficient markets, compared to professional traders.
Summary: the existence of numerous trading track records that are statistically incredibly unlikely to have occured by chance (due to high frequency, high returns, low risk, and high transactions costs) demonstrates strong evidence that market inefficiencies still exist. There are also very strong economic reasons why practicioners are far more likely to trade these inefficiencies secretly than academics are to discover and then publish them. Thus all the main points supporting your assertion are contradicted by the evidence. There is far more evidence for inefficiency than can be explained by sample size and survivorship bias, and there is no evidence from academia that markets are efficient (there is only evidence so far that most people cannot beat the market - but that is obvious and accepted by all).
If you are aware of any evidence or findings that contradict what I've said, now is your chance to put them forward.
Those results are statistically *very* significant, for several reasons:
i) high frequency of trades, approx 5-10,000 per annum.
ii) overheads of 20-30,000 per annum, plus my own salary.
iii) high win rate
iv) winning trade profit in excess of losing trade profit, both average and maximum
v) significant transactions costs, due to the high frequency of trades
vi) bulk of my profits came during a 50% bear market in the S&P 500, where the majority of my trades were long index futures. The index was lower when I stopped trading that inefficiency than when I started in 1998.
This inefficiency eventually went away, I had found several other markets where it was replicable and traded again, with similar win rates albeit lower return on capital due to liquidity constraints. Still way ahead of portfolio investment returns though, and with far lower risk.
I've also observed various arbitrages available in the markets since I started. My first job was at a bond futures arbitrage shop which also had a very high trading frequency, win rate, and annual return relative to drawdown. The existence of arbitrage opportunitites by itself disproves strong market efficiency. I saw numerous isolated arb opportunities.
So basically, the reason I don't agree with strong market efficiency is that I've seen on several occasions, track records based on tens of thousands of trades which are statistically significant way beyond the confidence level that any scientific testing demands. The fact that I've seen this from several firms/traders, trading totally different strategies and instruments, and the fact that my own "experiment" was forward looking as opposed to data-mining, makes it highly improbable that it's the result of random chance. Therefore the logical conclusion is that it was almost certainly down to inefficiency and the skilled exploitation thereof.
If in future markets become competitive to the extent that no one can find efficiencies that last long enough to outperform consistently, then I would come around to your point of view and invest in a diversified investment portfolio. But so far I see further examples of inefficient market behaviour, so I'm continuing to trade as long as it continues to offer such good opportunities.
As for academic evidence for efficiency, that's impossible since you can't prove a negative. All you can say is that there currently is not evidence supporting inefficiency. This then raises a problem - are academics aware of all the evidence? There are logical reasons why they would not be:
i) lack of ability. Since pay in the markets is so much higher than in academia, most people who are truly skilled at finding inefficiencies will be in the private sector rather than at universities. Those left behind will be of poorer quality and therefore less likely to make breakthrough discoveries.
ii) economic incentives. If an academic does discover a true market inefficiency, they will earn far more money by trading it or selling the idea, than by publishing. Therefore all good insights are likely to remain secret rather than be disclosed.
iii) lack of market experience. Most academics do not trade full time, which puts them at a disadvantage in discovering and trading inefficient markets, compared to professional traders.
Summary: the existence of numerous trading track records that are statistically incredibly unlikely to have occured by chance (due to high frequency, high returns, low risk, and high transactions costs) demonstrates strong evidence that market inefficiencies still exist. There are also very strong economic reasons why practicioners are far more likely to trade these inefficiencies secretly than academics are to discover and then publish them. Thus all the main points supporting your assertion are contradicted by the evidence. There is far more evidence for inefficiency than can be explained by sample size and survivorship bias, and there is no evidence from academia that markets are efficient (there is only evidence so far that most people cannot beat the market - but that is obvious and accepted by all).
If you are aware of any evidence or findings that contradict what I've said, now is your chance to put them forward.
EMH is a theory and like all theories means nothing when you are in front of a trading screen 
.
.Oh I should just add, to be scientific, the hypothesis that outperformance is due to survivorship bias needs to be falsifiable. I assume this would be done by requiring the results to be statistically significant relative to what could be achieved by chance.
Now if you assume high frequency of trades, thus very high annual transactions costs as a % of portfolio, then the average trader or random trade algorithm with no edge will lose most if not all of their capital within a short period of time. Because the negative edge is so high (due to the high costs), even a very large sample size will see most if not all of the samples go to zero within a few years. It's no different to taking 1 million roulette players, making them bet 50 times a day, and seeing how much they have left in 5 years - they will all be broke, because the negative edge is far higher than they can overcome by luck.
So, if we find just 1 trader who has not only not gone bust over 5 years, not only beaten the S&P, but made say 100% per annum with only single digit drawdowns, trading with such a large inherent negative edge, then the odds are overwhelming that it is the result of market ineffiency. The odds of this happening by chance are way way higher than even the most demanding of statistical confidence intervals used in other scientific experiments.
Thus to support your claim it is the result of survivorship bias/luck, you need to explain i) whether your thesis is falsifiable ii) what evidence you would accept as falsifying it iii) why you are selecting a far more rigorous set of criteria than scientists use in their testing of any other theory.
Now if you assume high frequency of trades, thus very high annual transactions costs as a % of portfolio, then the average trader or random trade algorithm with no edge will lose most if not all of their capital within a short period of time. Because the negative edge is so high (due to the high costs), even a very large sample size will see most if not all of the samples go to zero within a few years. It's no different to taking 1 million roulette players, making them bet 50 times a day, and seeing how much they have left in 5 years - they will all be broke, because the negative edge is far higher than they can overcome by luck.
So, if we find just 1 trader who has not only not gone bust over 5 years, not only beaten the S&P, but made say 100% per annum with only single digit drawdowns, trading with such a large inherent negative edge, then the odds are overwhelming that it is the result of market ineffiency. The odds of this happening by chance are way way higher than even the most demanding of statistical confidence intervals used in other scientific experiments.
Thus to support your claim it is the result of survivorship bias/luck, you need to explain i) whether your thesis is falsifiable ii) what evidence you would accept as falsifying it iii) why you are selecting a far more rigorous set of criteria than scientists use in their testing of any other theory.
I'm not sure how these claims can be supported. All they demonstrate is that the studies & academics in question *did not find any inefficiency*. That does not prove that no inefficiency exists.
Typical sloppy Wikipedia entry. It also gives no statistical significance testing of the results of Buffett et al. In fact, I don't understand why no academic has gone through his performance and run a significance test on it.
If you are basing the claim for long-term drift on the performance of US stocks, isn't this a case of survivorship bias? What if you had invested 100 years ago in stocks on the Moscow or South Vietnamese stock exchanges? What if you had invested in Zimbabwe in 1970? Or French 30 year bonds in 1789?
I would like to see your evidence that stocks rise in the long-run.
Stocks do not rise per se in the long run, indexes where losers are dropped and new companies are added do rise in the long-term because they are always fitted with survivor stocks.
Where would the Dow be if its value was based on the original 30 stocks without substitutions?
Where would the Dow be if its value was based on the original 30 stocks without substitutions?
EMH is a nice way to capture the idea of feedback from all participants being factored in. It provides a framework or something that can be visualized. If there are assumptions that all the participants are informed or even interested in the actual value of a financial instrument, at least in the short term, they are way off base. Most of the transactions are done by money managers with their own agendas which can include short term bonuses and the like... not that when those same managers are making unselfish decisons they are informed or able to form correct information from the data at hand all of the time either.. a good example is the current subprime mortage disaster. It is a repeat of previous similar situations, it could have been expected actually, well it is easy to say that with hindsight. People hung on to the idea that real estate was going up in value long after it was clear that it was not, and was not going to for quite awhile, it was easy enough to foresee a huge slowdown in the loan industry and really, it would not be too hard to look at all the subprime loans and make a guess about defaults... maybe.. but lots of people failed to do that and bankrupted their funds in rather short order...
Efficiency may exist if you look at a very long term view only and I've never looked for it personally so I don't know. LTCM attempted to correct for all the short term inefficiency with derivatives and were able to do so for awhile until they got blindsided. Maybe they liked EMH so much they were determined to make it happen?
The main reason a trader would not focus on the longer term to capture the EMH is that potential gains are huge, unbelievably huge, compared to buy and hold. Few realize those gains apparently but those that do are way ahead of any EMH based trading system...
Efficiency may exist if you look at a very long term view only and I've never looked for it personally so I don't know. LTCM attempted to correct for all the short term inefficiency with derivatives and were able to do so for awhile until they got blindsided. Maybe they liked EMH so much they were determined to make it happen?
The main reason a trader would not focus on the longer term to capture the EMH is that potential gains are huge, unbelievably huge, compared to buy and hold. Few realize those gains apparently but those that do are way ahead of any EMH based trading system...
I have a Ph.D. in History, not in Finance, but I would say that my degree is more relevant to this topic than one might think.
I don't have the time to get into this in any depth, for I have to help put the kids to bed, but for the most part, the markets ARE efficient and are getting more so every day. Chalk this up to the increasing availability of information, SEC laws, and round the clock news, to name a few. Fama, Bachalier, and Malkiel were for the most part spot on.
But the history of financial markets suggests that although speculators would like to THINK they are intelligent, governed by reason, they often are not. Greed and fear are the two primary market forces. When one adds leverage to the equation, you have emotion on steroids. The fundamental value of Japanese Yen means little when the margin clerk is breathing down your neck, threatening to liquidate your account.
I was trained in logic as an undergraduate and even teach the subject on occasion in my ancient history course. It has great application to my own personal life, but often has little relevance in the study of manic personalities such as the market.
In short, the EMT begs a significant question, one that cannot be answered by traditional finance, but perhaps by psychology, or behavioral finance. I am still awaiting proof from these fields that investors are reasonable, rational individuals.
I'd like to thank everyone for replying!
In regards to your first question, I'll try and provide what I see as very strong evidence for market efficiency without citing a bunch of academic journal articles (which often seems to be the case for some).
Take the monthly returns from a mutual fund. If you use the Fama-French three factor model, you can usually describe well over 95% of returns. The three factors take into account basically two things, the market return and the return of being exposed to two particular kinds of risk. You do get a significant constant in the regression, and a very small error term, but are they independent? I don't know, but I'd be willing to say they are.
Now, for the specifics on "portfolio strategies that are based around somewhat efficient markets". Things that take on exposure to situations where efficiency in the particular market is lacking such as a recently bankrupted public companies where government-forced selling is imposed onto large shareholders (like mutual funds) is present, and other such similar situations.
In regards to your first question, I'll try and provide what I see as very strong evidence for market efficiency without citing a bunch of academic journal articles (which often seems to be the case for some).
Take the monthly returns from a mutual fund. If you use the Fama-French three factor model, you can usually describe well over 95% of returns. The three factors take into account basically two things, the market return and the return of being exposed to two particular kinds of risk. You do get a significant constant in the regression, and a very small error term, but are they independent? I don't know, but I'd be willing to say they are.
Now, for the specifics on "portfolio strategies that are based around somewhat efficient markets". Things that take on exposure to situations where efficiency in the particular market is lacking such as a recently bankrupted public companies where government-forced selling is imposed onto large shareholders (like mutual funds) is present, and other such similar situations.