Why do lower courts treat Supreme Court precedents favorably or unfavorably? To address this question, we formulate a theoretical framework based on current principal-agent models of the judiciary. We use the framework to structure an empirical analysis of a random sample of 500 Supreme Court cases, yielding over 10,000 subsequent treatments in the U.S. Courts of Appeals. When the contemporary Supreme Court is ideologically estranged from the enacting Supreme Court, lower courts treat precedent much more harshly. Controlling for the ideological distance between the enacting and contemporary Supreme Courts, the preferences of the contemporary lower court itself are unrelated to its behavior. Hence, hierarchical control appears strong and effective. At the same time, however, a lower court's previous treatments of precedent strongly influence its later treatments. The results have important implications for understanding legal change and suggest new directions for judicial principal-agency theory.