Tax authorities must often draw a line between transactions that are entitled to preferential tax treatment and those that are not. To do so they employ four different policy-making strategies: rule-making, adjudication, advance ruling and licensing. These are illustrated, respectively, by the policy-making strategies applied to controlled foreign corporations, the economic substance doctrine, corporate spin-offs and nonprofits. How should tax authorities choose among the possible policy-making strategies? I characterize each policy-making strategy along two dimensions, timing and breadth, and lay out the problem that tax authorities face when choosing among them, namely their lack of information on firms’ circumstances and their reaction to policies. While rule-making allows tax authorities to commit to a policy, it does not allow the policy to be narrowly tailored to the circumstances of specific firms. Adjudication allows tax authorities to harness information that firms have on their unique circumstances to narrowly tailor the policy to those circumstances, but since it takes place after firms have already undertaken their investments, tax authorities’ choice of policy may be affected by this fact. This latter problem is mitigated when adjudication is supplemented with advance ruling. Licensing eliminates this problem, but only by requiring licenses even when the problem does not arise. Given the benefit and cost of each policy-making strategy, I show that tax authorities’ choice of policy-making strategy should depends on how informed they are about firms’ circumstances and their reactions to possible policies. When tax authorities are relatively well informed they should choose rule making, but when they are not they should choose either licensing or adjudication supplemented with advance ruling. Finally, I consider the effect of additional factors outside the basic framework of analysis, and note that the analysis can be applied to other areas of law, where a similar choice among policy-making strategies is made.