Responsible party government theory requires that voters hold parties electorally accountable for their performance in control of government. Existing literature suggests that voters do this only to a limited extent—holding the presidential party's candidates responsible for government performance on Election Day. While this method of voting may hold the executive accountable for his performance, it is not really an effective way to hold the party in control of Congress accountable for its performance. The method falls short particularly when Congress is controlled by a different party than the president, but also whenever a Congress controlled by the same party pursues policies different from the president's. Using surveys of voters leaving the polls in the 1990, 1994, and 1998 midterm congressional elections, this study tests whether voters' evaluations of Congress's job performance also affect their support for majority party candidates in House and Senate elections, during both unified and divided government.