Agency theory foments the expectation that corporate criminal prosecutions mitigate crime by inducing firms to self-regulate. Whether this bears out in reality remains a topic of contentious debate. Although the U.S. government began prosecuting firms over a century ago, insufficient empirical evidence exists to determine how the costs of prosecutions actually affect firms. Moreover, the limited empirical record tells a surprising and somewhat confusing story. Scholars of management and related disciplines have consistently found that criminal convictions have negligible impacts on shareholder wealth despite theoretical expectations to the contrary. To explain these surprising findings and better understand how firms experience the costs of prosecutions, I apply agency theory to the criminal prosecution process and propose that each legally defined action in that process communicates unique information regarding potential and actual agency costs (i.e., sanctions). I also propose that formal convictions appear to elicit no reaction from principals because firm responses to early events in the prosecution process, what I call "conviction harbingers," foretell unfavorable verdicts and sanctions well before courts make them official. Findings derived from a sample of 177 cases largely confirm my expectations by showing that prosecutions cause firm value to decline nearly 11%. In addition to exceeding previous estimates by a factor of three, this amount exceeds direct fines by a factor of five (i.e., shareholders lost more than $5 of wealth for each dollar of sanctions). I conclude by outlining implications of these findings for agency theory, corporate governance, policy, and practice.