Dynamic autocorrelation of intraday stock returns

We discover three significant periodicities in the autocorrelation of intraday stock returns. We demonstrate that (i) the autocorrelation is 64% more negative during afternoons than during mornings, (ii) the autocorrelation is more negative Tuesdays through Fridays than on Mondays, (iii) overall serial correlation becomes less negative when salient information events arrive, i.e., earnings months, but measures less negative during mornings and on Mondays. Our results support the hypothesis that informational demand is more critical following daily and weekly market closures when information accumulated cannot easily be traded on, while liquidity demand intensifies closer to the no-trading periods