TABLE OF CONTENTS INTRODUCTION I. SENTENCING IN THE UNITED STATES: THE DIALECTIC OF UNIFORMITY AND DISCRETION II. DELEGATION AND FEEDBACK LOOPS A. The Actors Determining Sentencing Policy 1. Legislatures 2. Judges 3. Sentencing Commissions B. Combining Individual and Systemic Justice Through Dialogue C. Dialogue in Presumptive Systems D. Dialogue in Advisory Systems E. The Legislature's Role in Sentencing Feedback Loops III. INSTITUTIONAL DESIGN QUESTIONS A. Aggregating Past Practice or Imposing a New Philosophy B. Appellate Review C. Reason-Giving by Judges and Commissions D. Sources of Stickiness E. The Sentencing Commission's Branch and Membership CONCLUSION INTRODUCTION
The current federal sentencing system is incoherent. Before sentencing a defendant, a judge must first calculate the sentence range provided by the Federal Sentencing Guidelines. (1) But the judge is then free to ignore this range, and impose a sentence outside of the Guidelines. This is because the Supreme Court's decision in United States v. Booker rendered the Guidelines "advisory," so judges can no longer be reversed merely for failing to adhere to them. (2) But the Guidelines were not written to "advise" judges in any meaningful sense. They were written to dictate judges' sentencing decisions. The Guidelines manual contains no explanatory discussion that might persuade a judge. Instead, it provides an opaque formula that receives facts about the crime and the defendant as inputs and spits out a range of months. Thus the Guidelines calculation is now little more than a mandatory homework assignment for judges. (3) Booker has created a basic contradiction--we have advisory guidelines that provide no advice. Federal judges are given two conflicting messages: the Sentencing Commission tells them to follow the Guidelines, while the Supreme Court tells them to follow their consciences. (4) The Supreme Court has permitted appeals courts to consider the degree of variance from the Guidelines when deciding whether a sentence is substantively reasonable (and thus valid on appeal).5 But the Court has also held that deviation from the Guidelines cannot make a sentence presumptively unreasonable, (6) and has left it to circuit courts to decide whether within-Guidelines sentences are presumptively reasonable. (7)
These deferential instructions have created a divide between sentencing judges over how often to follow the Guidelines, as well as significant regional variation in Guidelines compliance rates. (8) They have also led judges to issue a large number of below-Guidelines sentences for certain crimes, such as child pornography possession, where the Guidelines are especially harsh. (9) Further, they have created a divide among different appellate courts, some of which presume that within-Guidelines sentences are substantively reasonable, and some of which give the Guidelines no such presumption. (10) The same uncertainty about the normative status of sentencing guidelines exists in a number of state systems. Arkansas, Delaware, Maryland, and other states have adopted "advisory" guidelines that have much the same structure as the federal system--a black box formula that provides a recommended sentence with no real explanation. (11) Judges are understandably unsure about why they should follow such recommendations, especially in cases where the recommended sentence seems unjust or inappropriate.
There have been a variety of proposals to reform the post-Booker sentencing system so as to clarify the normative status of the Guidelines. These Booker "fixes" come in two basic categories. Some argue that the system should be hardened in order to remove or sharply curtail judges' sentencing discretion. (12) Others argue that the Guidelines should be abolished, significantly scaled back, or simply kept as advisory so as to preserve judges' sentencing authority. (13) Each of these proposals carries the assumption that sentencing policy is a zero-sum game. We must choose either (1) a system where judges control sentencing decisions, or (2) a system where judges are constrained by guidelines.
This Article transcends that dichotomy by proposing a different architectural logic for sentencing guidelines systems: one of dialogue between judges, legislatures, and sentencing commissions. In the conventional architecture, the legislature delegates primary control over sentencing to one actor--either the judiciary or the commission--and that actor dictates sentencing policy while the other is merely an adjunct. In a dialogic architecture, by contrast, the legislature delegates sentencing authority to both actors in concert, making judges and the commission coauthors of the sentencing system. Judges and the commission then establish sentencing policy through iterative feedback loops, wherein judges systematically influence the decisions of sentencing commissions, which in turn systematically influence the decisions of judges. Meanwhile, the legislature can direct the development of this sentencing system by enacting framework statutes (establishing certain values or criteria to guide sentencing decisions) or enacting specific sentencing dictates (mandating particular changes to the guidelines formula). Through such policy feedback loops, the sentencing system can realize the twin goals of individualizing sentences and reducing unwarranted disparity. Empowering only judges or only the commission forces the legislature to choose between those goals. But a dialogic architecture allows judges' departure decisions to be incorporated into the guidelines themselves, so that the sentencing system adapts over time to accommodate more variables without creating unwarranted disparities. Dialogue thus allows the branches to cooperate in building a common law of sentencing.
There are two models for building such feedback loops into sentencing institutions. In the first model ("presumptive" feedback loops), sentencing guidelines are made presumptively binding on judges, but there is a general departure power letting judges give above- or below-guideline sentences if merited by the facts of a case. These departures are then subject to appellate review to determine whether they were in fact justified, and over time the appellate courts' departure case law can be incorporated into the guidelines themselves. In the second model ("advisory" feedback loops), the guidelines are not binding on judges, but are designed to reduce disparity by persuading judges to follow them. They do so in two ways. First, the commission can design the guidelines to reflect judicial practice, such that they inform judges of existing sentencing norms. Second, the commission can issue its own recommendations for changing sentencing practices, and can use its policy expertise and facility with empirical social science to provide arguments for such changes. Judges then respond to the commission's guidelines by following or departing from them, and the guidelines are updated to reflect judges' sentencing decisions. In both models, the sentencing system uses judges' departures as an engine to update the guidelines. These models thus each carry two benefits. First, they permit judges to individualize sentences while limiting the incidence of interjudge disparity. Second, they permit judges to systematically correct the guidelines' oversights and irrational recommendations as these are revealed in particular cases.
A dialogic model is compatible with many existing sentencing institutions. The states of Kansas, Minnesota, North Carolina, and Washington have adopted the basic tools of presumptive feedback loops-presumptively mandatory guidelines, a general departure power for unusual cases, and appellate review of departures. (14) The draft sentencing provisions of the Model Penal Code adopt the same structure, and also instruct the sentencing commission to include commentary explaining the reasoning behind each guideline provision. (15) Virginia has an advisory guidelines system that is designed to reflect judicial practice. Its sentencing commission compiles extensive data about judges' sentencing decisions, provides empirically grounded policy guidance about different sentencing factors, and issues annual recommendations for updating the guidelines to better reflect current sentences. (16) Virginia has thus basically embraced the model of advisory feedback loops, using the guidelines system both to advise judges of existing sentencing norms and to provide policy arguments for adopting new norms.
Even the framework of the federal sentencing system is compatible with the logic of dialogue. The federal guidelines are now advisory, the federal Sentencing Commission collects exhaustive data about federal judges' sentencing practices, (17) and the Sentencing Reform Act instructs the Commission to consult with judges and other actors when updating the guidelines. (18) Indeed, several of the authors of the federal guidelines originally envisioned that the system would incorporate elements of interbranch dialogue, with the Commission studying and learning from the justifications that judges articulate for their sentences. (19) However, even this modest dialogic practice was not implemented. Post-Booker, the federal Sentencing Commission could easily follow Virginia's lead and adopt dialogue as its governing logic. Embracing the concept of feedback loops would return coherence to the federal sentencing system by clarifying the guidelines' force to judges. Indeed, in some ways, federal judges and the Commission have (albeit very recently) begun to engage in such dialogue. The Commission has taken serious note of judicial critiques of (and departures from) certain guidelines, and has responded with statistical reports and modest policy changes. (20) Hopefully this nascent trend will continue.
The thesis of this Article is thus both theoretical and practical. Policy feedback loops between judges and sentencing...