Ballot positions often affect the electoral performance of parties. Existing theories for this effect focus on voter-specific behavioral explanations. We present evidence on an additional unexplored mechanism: parties adjust their behavior to account for their position on the ballot. This adjustment matters for how we interpret existing results. First, we use a constituency level lottery of ballot positions across two elections and 1971 races in Colombia. We find evidence for a ballot position effect of 6.7 percent on vote share, and 8.8 percent on seat share.