In their book, John Huber and Charles Shipan propose and test a theory identifying the conditions under which bureaucracies are imbued with independent policymaking authority. The core theoretical development derives from the authors' efforts to bridge the gap between existing theorizing in the American politics literature on the sources of bureaucratic discretion and in the European parliamentary context about policymaking in minority and coalition governments. Huber and Shipan argue that rather than addressing these topics separately, progress in understanding the implications of institutional arrangements for executive governance is best made by simultaneously comparing policymaking within and across these institutional forms. To judge by their robust and interesting results, this is indeed a fruitful enterprise.

The core theoretical arguments, with technical details relegated to an appendix, are presented in Chapter 4. The theory focuses attention on the relative costs and benefits of producing detailed legislation. High levels of political conflict between legislators and bureaucrats, who are presumed to implement the wishes of the chief executive or relevant cabinet minister, lead legislators to desire restraints on bureaucratic discretion. Writing long and detailed legislation, however, is costly. It takes legislators, particularly those without high levels of expertise, away from other tasks. Furthermore, the very demand for this type of legislation can increase the difficulty of securing it because new laws are subject to review by a hostile executive and a potentially hostile second chamber in bicameral legislatures. These effects are mitigated, however, when legislators have other tools for controlling bureaucratic behavior ex post.

The empirical analysis is divided into two sections, the first examining the design of Medicaid laws in the American states during 1995 and 1996 and the second examining European labor legislation from 1986 to 1998. These undertakings build on an in-depth and comparative look at the nature of statutory controls in Chapter 3 and the case study of Michigan in Chapter 5. Chapter 3 deserves attention from a wide audience for demonstrating that 1) longer laws place greater constraints on bureaucratic behavior, and 2) the use of procedural controls to limit bureaucratic discretion is more common in separation-of-powers systems than in parliamentary systems.

Huber and Shipan's statistical analysis of the length of state Medicaid legislation largely confirms their theoretical predictions, showing that policy conflict (divided government) leads to longer legislation, but that this effect is mitigated by 1) a lack of legislative capacity, 2) conflict among the chambers of the legislature, and 3) the existence of a legislative veto that otherwise checks agency discretion. In the parliamentary con-text, legislation is longer during periods of coalition government and longest during periods of minority government, but these effects are diminished by rapid government turnover early in a government's tenure. These chapters are also note-worthy for their close link to the posited theory and for the careful attention given to evaluating alternative explanations.

Despite the overall strength of this work, several limitations remain. First, the statistical models are constructed around imperfect measures of autonomy. The authors make a strong case for the number of words governing agency behavior as a proxy for constraints on bureaucratic discretion, but then use the total number of new words in state Medicaid legislation (p. 141) and pages per legislative enactment for European labor laws (p. 177) as their dependant variables. The problem with both measures is that the “words in legislation” approach does not account for the cumulative degree of constraints on bureaucratic discretion. In the Medicaid case, that is because new laws rarely create policy anew or without reference to existing statutes. In Huber and Shipan's coding, however, a new law that adds 100 words to an existing statute is treated the same way as a law that deletes 200 words of text and replaces it with only 100. Similarly, states that already have long (or short) laws on the books may have no need to replace them with short (or long) laws absent a shift from (to) divided government. In the analysis of European legislation, using passed laws as the unit of observation compounds this limitation. Absent amendment, constraints on discretion persist. It would seem more appropriate to use the cumulative total of words of labor law in force in a given year (perhaps broken down by issue) to measure the contemporaneous degree of bureaucratic discretion. By comparison, it would be inappropriate to evaluate the degree of judicial autonomy by examining the sentencing discretion given to judges in statutes enacted in a given year. Existing laws and sentencing guidelines might otherwise constrain judges. Looking in isolation at the number of words in individual pieces of legislation is different from examining active constraints on discretion.

Second, there is scant evidence in this work that statutory constraints on bureaucratic discretion substantially alter policy outcomes. This is ironic, given that Huber and Shipan frame their work in Chapter 2 around a forceful criticism of the existing empirical work on the political control of the bureaucracy (pp. 23–26). Perhaps legislative specificity is simply a substitute for other forms of agency control, or perhaps it is purely symbolic and has no effect at all on agency behavior. If one accepts Huber and Shipan's assertion that bureaucrats are largely under the control of the executive, it would be straightforward to demonstrate (perhaps with a survey of state legislators) that oppositional legislators are less satisfied with bureaucratic performance when another legislative chamber sympathetic to the governor inhibits the passage of highly specific legislation.

That this hypothetical exercise would constitute a test of Huber and Shipan's theory points to perhaps a more fundamental limitation of their packaging of the book as identifying the “institutional foundations of bureaucratic autonomy.” Bureaucrats in this model are not autonomous when imbued with bureaucratic discretion, but instead are perfect agents of some other actor. It is a thin autonomy, however, that is nothing more than the slavish implementation of a governor's or cabinet minister's ideal policy. In separation-of-powers systems, it is critical to determine which actor—the executive, the legislature, or the bureaucracy itself—“wins” in the presence of agency discretion. This is, of course, the goal of the “political control of the bureaucracy” scholarship that the authors seek to move beyond. Unfortunately, a pure focus on statutes cannot resolve this question.

These limitations notwithstanding, Deliberate Discretion is a substantial step forward in the institutionally based study of bureaucracy and policymaking. Huber and Shipan's work is an impressive theoretical and empirical undertaking of great relevance to students of American and comparative politics alike.