Imagine a situation where points are removed for which the X values are judged to be outliers.
Note that this restricts the range of X values. Without looking at the data, can you predict whether Pearson's correlation will increase or decrease after these points are removed?