When an administrative agency changes a policy, a reviewing court requires a reasoned explanation. The demand is sometimes described as a rule against inconsistency, as though an agency were bound to its first answer until facts themselves change. That description is misleading. Agencies are institutions created to make policy within delegated bounds; elections, accumulated experience, and altered priorities may legitimately produce a new judgment from the same record. Requiring a new administration to prove that its predecessor's view was erroneous would entrench policy without legislative authorization.

But permission to change direction is not permission to ignore what the old policy set in motion. Regulated parties may have purchased equipment, organized businesses, or entered long contracts in reliance on the existing rule. Those interests do not create a permanent veto. They do, however, become facts that a rational decision maker must notice. An agency may conclude that the new policy's benefits outweigh transition costs, or it may phase the policy in; what it may not do is speak as though the world at the time of change were the world before the old rule existed.

This distinction clarifies the court's role. A court should not compare the substantive wisdom of two policies and select its favorite. It should ask whether the agency recognized the change, addressed serious reliance interests and important alternatives, and connected the evidence to the choice made. The intensity of that inquiry should reflect the agency's own claims. A minor clarification requires less explanation than a reversal that disclaims the factual premises on which a durable regime was built.

Judicial review so understood protects neither stasis nor a judicial policy agenda. It protects deliberative continuity: the obligation of government to account for the consequences of its prior acts even when it remains free to choose a different future. That obligation makes change more candid and often more costly, but those are features of responsible administration, not constitutional defects.

Experience with phased compliance illustrates the distinction. When an agency identifies investments induced by its prior rule, a transition period may preserve much of their value while still advancing the new policy. But transition is not mechanically required: delay may itself impose urgent public costs. What matters is that the agency compare those costs openly. A court reviewing that comparison should demand evidence proportionate to the agency's departure, not a demonstration that only one policy could be rational. This keeps review searching without turning reliance analysis into disguised substantive veto.