The reproducibility of published academic work is increasingly important across a wide array of fields from neuroscience to sociology to political science. If we as academics wish to create impactful research, we need to create faith in the larger research community that our work is valid, and a series of reproducibility scandals over the past decade has degraded that faith. A commonly cited solution is for all data-driven research to include complete replication files, including all code and data necessary to reproduce all key statistics and figures. In this paper, we examine some of the norms and best practices surrounding this notion across various fields. We test the hypothesis that providing replication files increases a paper's citation count, finding that there is a notable negative effect. We conclude by offering policy prescriptions for academic journals in light of this finding.