We develop a theoretical model that replicates three observed phenomena in securities markets: serial correlation in trades; serial correlation in squared price changes (conditional heteroskedasticity); and more persistent serial correlation in trades than in squared price changes. In the model exogenous news is captured by signals that informed agents receive. Agents trade anonymously through a market specialist, who does not receive a signal. We show that entry and exit of informed traders following the arrival of news produces serial correlation in the number of trades and serial correlation in squared price changes. Because the bid-ask spread of the market specialist tends to shrink as individuals trade and reveal their information, the serial correlation in trades is more persistent than the serial correlation in squared price changes.
