By counting the number of articles published in major US newspapers containing carefully selected keywords, we construct a time varying measure of ethnic tension. Then, we empirically test the predictions of a theoretical model by using the aforementioned measure, and investigate how ethnic tension affects presidential approval ratings by different ethnic groups. Our results show that while ethnic tension decreases the approval by white voters, the opposite is true for the approval by African American voters. Further scrutiny reveals that this may be explained by the fact that government transfers to African Americans increase as a result of higher ethnic tension.