Buy-sell imbalances, market returns and macro-announcement days

Using a unique dataset from the Brazilian equity market, we document that retail buy-sell imbalances better explain and predict market returns than institutional imbalances. This behavior is attributed to order-splitting strategies of sophisticated investors, who aim to gradually profit from private information, rendering their trades less informative. We support this conjecture with evidence suggesting that institutional imbalances exhibit strong autocorrelation, whereas retail imbalances do not. We also show that the predictive capacity of all imbalance types disappears on the days of macroeconomic announcements. More interestingly, on these days, the imbalances’ autocorrelation ceases, suggesting that institutional investors rapidly incorporate the newly available information into prices.