The absence of linear autocorrelation in asset price movement has been empirically observed countless times. It is usually accompanied by an explanation that goes something like this:

If there was linear autocorrelation in the price movement of a particular asset, it would be exploited by traders, causing the autocorrelation to vanish. Therefore it doesn’t exist.

I don't see why these observations wouldn't support the opposite conclusion. The logic is as follows:

Let's assume for sake of contradiction that the above statement is true. Now suppose we currently observe no linear autocorrelation in the market. Any strategy that attempts to exploit autocorrelation if it isn't present would not be profitable. Therefore they aren't being used at the moment. But if they aren't being used at the moment, and we are assuming they are the reason autocorrelation doesn’t exist, we should see a reversion to an autocorrelated price. Since we don’t observe this reversion, these strategies cannot be the reason for an absence of autocorrelation in prices.

Since this is a rather obvious idea I'm assuming its wrong but I'm not exactly sure why. I was wondering if someone could help direct me to an explanation for why the first conclusion is in fact more likely.

  • $\begingroup$ Hi. No offense intended but I don't see your argument ? Why should we see a reversion to the autocorrelated prices ? If there's a reversion to the autocorrelated price, then that's exactly the time where profits could be made so any significant autocorrelation should go away instantaneously. Remember that, when the term autocorrelation is used, it doesn't mean that autocorrelations don't exist. It just means that they aren't strong enough consistently to be consistently profitable from. So, autocorrelation is a tricky thing because it can exist without allowing for profits. $\endgroup$
    – mark leeds
    Apr 20, 2022 at 1:46
  • 1
    $\begingroup$ @markleeds Thank you for the comment! I should note that my assumptions aren't consistent with your statement that, "any significant auto correlation should go away instantly." I'm assuming that a pattern needs to be observer before it can be exploited. This process wouldn't be as instantaneous as you suggest. Also, I might be wrong, but when researchers say that there is an absence of linear autocorr', I think they mean this in a strictly statistical sense. Not that they aren't strong enf' to profitably trade on them, but that they aren't strong enf' to even be statistically significant. $\endgroup$
    – stam_a
    Apr 20, 2022 at 2:39
  • $\begingroup$ I'm not certain but I always took it to mean "strong enough to profitably trade off of". The reason I say this is because statistical significance often doesn't translate to profitability because of effect size. For example, one could have autocorrelation but the exploitation of it may not be able to overcome transaction cost. As far as the instantaneousness question, that's also controversial because people do take advantage ( or try to ) of autocorrelation in say "trend following". I bet there are trend followers out there who will stand by there claim that trend following works. $\endgroup$
    – mark leeds
    Apr 20, 2022 at 11:55
  • $\begingroup$ @markleeds I understand what you are saying about statistical significance and profitability. Also I agree that there are examples of academic papers and practitioners supporting strategies built on trivial linear autocor. That being said, my post is not as concerned with whether or not we actually observe profitable autocor but about the proposed explanation which follows when we assume prices are not linearly autocor. I'm asking if the logic I provide is (or, more likely, why it isn't) sufficient to show that a complete lack of autocor is actually inconsistent with the weak EMH hypothesis $\endgroup$
    – stam_a
    Apr 20, 2022 at 15:36
  • 1
    $\begingroup$ Here's my take. When the ME experts say say: "If there was autocorrelation , then it would be exploited and this exploitation would cause it to vanish". ( I guess price would get too high and supply would be greater than demand and cause price to revert ), I don't think they mean to imply that autcorrelation naturally exists in prices. They are saying that, if it did exist, it would be washed away. Your argument assumes that they are assuming that there is a natural tendency for autocorrelation to exist in prices which I don't think is what they are assuming. $\endgroup$
    – mark leeds
    Apr 20, 2022 at 17:44


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.