Testing the political replacement effect: A Panel data analysis