Description:
In this paper, an anomaly in the daily cumulative abnormal returns (CARs) of corporate site visits (CSVs) is documented. In this study, the anomaly observation period is divided into three time segments to show that the anomaly results in the CAR trending upward (downward) one month before the event day, temporarily declining (increasing) on the event day, and then trending upward (downward) one year after the event day. We refer to this anomaly as the CSV manipulative effect. Based on a unique data set of site visits to listed firms in China, we obtain the following main findings. First, the CSV manipulative effect is not caused by other company announcements disclosed around the CSV event date, and the probability of manipulative effect brought by site visits is significantly greater than that brought by other company announcements. Second, factors related to market manipulation significantly affect the CSV manipulative effect. Third, we construct a machine learning model to detect the CSV manipulative effect and find that it is capable of detecting this effect. Fourth, the portfolio constructed by the CSV manipulative effect performs better than the portfolio constructed by the momentum or reversal effect