This simple and alluring idea has farreaching implications. One of the most salient concerns political control of the bureaucracy. Delegating Powers is one of the first transaction-cost analyses to suggest that in delegation, Congress cannot have its cake and eat it, too. If Congress loads down a bureaucrat with a) statutory specification and b) administrative procedures, then both a) and b) must be specified by a committee. That is, the restraint of the executive itself compels delegation with its concomitant agency losses. When (plausibly) the preferences of the specifying committee diverge from the floor's, procedural control is much less compelling. By showing that delegation is genuine, and that procedural control carries its own costs, the authors masterfully refute legislative dominance scholarship.

Another kudo deserved by Epstein and O’Halloran concerns empirical measurement. They create a measure of delegation that aggregates, for each of 257 important legislative measures passed in the post–World War II period, the number of planks that delegate, divides this sum by the number of planks, and then subtracts from the resulting “delegation ratio” a weighted aggregate of the number of constraints placed upon delegation. Their measure is simple and intuitive, and it has the advantage of comparing delegation across policy domains. From tests using this measure, they generate some nontrivial results supporting their theory: 1) legislative delegation is less common under divided government (pp. 129–40); 2) legislators are less likely to vote for delegation when the president is of the opposite party (pp. 142–50); 3) delegation to the executive induces Congress to structure oversight committees as “contrary outliers” that compensate for the divergence of the executive from the floor by skewing the other way (pp. 167, 171–77); and 4) these outliers emerge more frequently under divided than unified government (pp. 177–82).

Despite these triumphs, noteworthy liabilities accrue to Epstein and O’Halloran's analysis. First, the static foundations of their model remove the most important transaction-cost considerations from view, namely, the asset value of agencies and Williamsonian “firstmover” advantages. A rational legislator will condition present delegation decisions upon the results of past choices, and will anticipate the necessity for future delegation to any agency he or she creates. Suppose a new policy area— where circumstances change repeatedly— emerges, requiring repeated delegation. If Congress delegates authority to a nuclear power commission, for example, it will create new capacities in that agency that will undoubtedly condition future delegation. When nuclear regulation issues emerge again, Congress will be less likely to delegate new nuclear regulatory authority to a newborn agency or a legislative committee. Agencies to which authority is delegated the “first time” will, by the logic of transaction costs, be more likely to receive it again when new policy in the area is created.

This point is neither academic nor a mere modeling extension. As Epstein and O’Halloran's data show, the vast majority of delegations since 1947—fully 759 of 961 (p. 158)—go to existing agencies, not to agencies created de novo. About four in five acts of delegation, then, give authority to agencies whose personnel, structure, and behavioral history are at least somewhat known by Congress. The authors grant passing mention to this crucial fact. The recipients of delegation in their model are characterized only by preferences, not at all by “historical” characteristics, such as their observed ability, their observed propensity to defect from the floor median, or their experience with the policy.

Empirically, it would be worthwhile to replicate Epstein and O’Halloran's statistical analyses with controls for whether the policy in question has been delegated to the executive before. Then the authors’ regressions would ask not “did Congress delegate Policy X to the executive?” but, rather, “Did Congress delegate Policy X at time t to the executive, given that it also did so (or did not) at some time t-k?” Attention to the dynamics of delegation would also generate different explanations for the authors’ qualitative inferences. In a seemingly counterintuitive finding, they observe higher delegation ratios for agriculture policy, long assumed the cozy home of distributive politics and congressional dominance. They explain this pattern by reference to the need for presidential assent to enhance an otherwise unstable logrolling coalition on the floor (pp. 220–22). A far better explanation would point to the long history of congressional delegation to the U.S. Department of Agriculture before World War II, including the Meat Inspection Act of 1885 and the Pure Food and Drugs Act of 1906. When it came time to delegate highly complex policymaking to the USDA in the New Deal, as Kenneth Finegold and Theda Skocpol have noted, Franklin Roosevelt and congressional Democrats saw the USDA as the efficient home of commodity price supports.

The second problem is that Epstein and O’Halloran's model is not truly strategic. Neither presidents nor agencies wait idly by as Congress decides whether or not they shall receive new authority. In some cases—witness the Food and Drug Administration and tobacco, or President Clinton and fast-track trade negotiating authority—the executive branch actively seeks to gain new powers that Congress might otherwise be unwilling to grant it. In other cases, as James Q. Wilson has observed, agencies wish to avoid new tasks. In both the agriculture and fast-track negotiation cases, the authors point to presidential lobbying as instrumental to the final outcome. But neither presidential bargaining nor bureaucratic lobbying— both of which are empirical regularities—plays a role in their theory.

Third, Epstein and O’Halloran ignore those situations in which discretion is taken away from the executive (a transaction-cost perspective would not). This artificial unidirectionality of delegation generates some faulty inferences. The most notable comes when they note that the delegation ratio has been declining over time (Figure 5.9), and infer that claims of a runaway bureaucracy are “mythical”: “the amount of discretionary authority given to the executive has declined since 1947” (p. 117). Yet the authors have shown only that the rate of increase of executive discretion has slowed. Since they do not observe delegation taken away, and since authority delegated in one Congress is not returned at the end of the session, they cannot claim that the total amount of bureaucratic discretion has subsided. Put mathematically, the authors make inferences about the “cumulative distribution function” of delegation (which is strictly increasing if removed delegation is unobserved) when they observe only its “density.”

There are other pleasures and other faults, but Delegating Powers clearly hews its way into a complex question and makes valuable progress. Further advances will occur when scholars, inspired by transaction-cost considerations or by other theories, look at how the history of delegation conditions its present.