Many contend that President Bush's reelection and increased vote share in 2004 prove that the Iraq War was either electorally irrelevant or aided him. We present contrary evidence. Focusing on the change in Bush's 2004 showing compared to 2000, we discover that Iraq casualties from a state significantly depressed the President's vote share there. We infer that were it not for the approximately 10,000 U.S. dead and wounded by Election Day, Bush would have won nearly 2% more of the national popular vote, carrying several additional states and winning decisively. Such a result would have been close to forecasts based on models that did not include war impacts. Casualty effects are largest in "blue" states. In contrast, National Guard/Reservist call-ups had no impact beyond the main casualty effect. We discuss implications for both the election modeling enterprise and the debate over the "casualty sensitivity" of the U.S. public.