摘要：Political identification is the basis of enduring conflict, suggesting that political attitudes are difficult to change. Here we show that in the 2016 U.S. Presidential Election, political identities underwent modification in response to salient political events. We investigate these dynamics in detail by collecting data at periodic intervals from mid-June 2016 through the general election (N = 3,958). We operationalize identification using prosocial giving in Dictator Games played between supporters of competing primary candidates recruited from Amazon Mechanical Turk. The observed dynamics differed across political parties. In-group bias among Democrats remained high until the Democratic National Convention, disappeared shortly thereafter, and then returned during the final stage of the election. Bias among Republicans was generally high until the final days of the election. The late resurgence of bias among Democrats was not reflected in voting intentions, but may have presaged the Democratic election loss.
