Suppose I define order flow imbalance as volume(aggress buy)/volume(aggress sell), or some variant of that.

Is this variable more, or less, correlated with price movements when I sample less frequently: 1 minute, 1 hour, 1 day...?

  • 1
    $\begingroup$ Intuitively I would think it would have a bigger effect in the short term... but without empirical tests I don't know. $\endgroup$
    – Alex C
    Feb 6, 2016 at 20:11
  • $\begingroup$ Alex: can you explain your intuition a little? $\endgroup$ Feb 8, 2016 at 4:20
  • $\begingroup$ I dunno. Philosophically I believe in the longer run (hours, days, weeks, months) prices are set by "information" in the Efficient Market theory sense, i.e. announcements, earnings, fundamentals, etc. Only in the very short run (hours, minutes, seconds) does the sheer weight of buying and selling have a direct mechanical effect on price. For ex. when there is a "fat finger error" the price gets clobbered but it soon returns to its normal level because the fundamentals haven't changed. Given enough time the arbitrageurs go to work and restore the equilibrium. $\endgroup$
    – Alex C
    Feb 9, 2016 at 2:44
  • $\begingroup$ @AlexC: I have the same intuition. I haven't confirmed that from a causal look at the data yet, but there are unfortunate issues with trade direction classification: they rely on price changes, the very thing I want to explain. Do you have ideas how to get around that? $\endgroup$ Feb 9, 2016 at 4:02


Your Answer

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