Studying Discretion in the Processes that Generate Criminal Justice Sanctions

Abstract

This paper presents a typology of discretion that differs from the standard typology of discretion in the legal literature: Type A discretion, which is the generally recognized discretion of individual actors to make decisions within a set of laws and rules, and Type B discretion, which is the crafting of laws and setting of rules in the first place. An analysis using aggregate data from 1992 to 2002 points to the prosecution stage as the primary source of Type A discretion that contributed to the massive increase in incarceration during this period. To understand how this could happen, this paper argues that actions need to be linked to outcomes. We identify three key outcomes of the sanctioning process—crime control, justice (guilt/innocence), racial disparity (fairness)—that deserve more attention from social scientists.

Keywords:

• sentencing
• discretion
• criminal justice system
• prosecutors
• crime reduction

Acknowledgements

We are grateful to Gail Hsiao and Brendan Lantz for research assistance, and two anonymous reviewers for helpful suggestions. This paper was originally prepared for the Conference on The Past and Future of Empirical Sentencing Research, 23-24 September 2010, Albany, supported by a grant from the National Science Foundation (SES-0939099). We are also grateful for the three discussants, Kim Hunt, Nancy King, and Max Schanzenbach, at the Symposium who provided useful comments.

Notes

A prior version of the paper was titled “Studying Discretion in Sentencing.” The authors subscribe to the notion, first described in National Academy of Sciences (1983), that sentencing is not just the stage at which judges assign punishment, but a product of all the steps along the way that lead to the ultimate punishment experienced by the offender, including indictment and parole decisions. Indeterminate sentencing, for example, includes the parole board by definition. However, this broad description of sentencing has not been adopted in the criminal justice literature, which has largely studied sentencing using conviction data-sets.

1. The idea of rule of law precedes John Adams by more than two millennia. Plato wrote, in around 350BC, that “if law is the master … and the government is its slave, then the situation is full of promise and men enjoy all the blessings that the gods shower on a state.”

2. As we discuss below, we find it useful to distinguish the flow variable, proportion of convicted persons incarcerated, from the stock variable, the percentage of the population in prison or jail. It might make better sense to regard the former as the “incarceration rate” and the latter as the “proportion of the population incarcerated.”

3. These two important elements can be analyzed separately and are likely to be influenced by criminal justice agents; we can simplify the analysis for current purposes by regarding those elements as exogenous to the problem of discretion by criminal justice agents and agencies.

4. Consider the identity: Inmates/Population = (Victimizations/Population) × (Reported Crimes/Victimizations) × (Arrests/ReportedCrimes) × (Convictions/Arrests) × (Incarcerations/Convictions) × (Average term/Incarcerations). In a steady state, the number of inmates will equal the number of incarcerations annually times the average term per incarceration.

5. For the period 1992-2002, the bias associated with declining crime was substantial, while that for punitiveness was, on the whole, negligible. This is evident in comparisons of the actual change in prisoners per capita with that estimated by the components of the steady-state model over the ten year period. For murder, the gross incarceration rate increased by 53% from 1992 to 2002, while the steady-state estimate of the rate declined by 20%. For aggravated assault the rate more than doubled, while the steady-state estimate of the rate increased by just 20%.

6. The results reported here are from the Bureau of Justice Statistics National Judicial Reporting Program, reported in its Felony Sentences in State Courts bulletins and the BJS State Court Processing Statistics, reported in its Felony Defendants in Large Urban Counties series.

7. The prison population was 850,566 in 1992 and 1,367,547 in 2002; the combined prison-jail population was 1,295,150 in 1992 and 2,033,022 in 2002, which yield absolute and per capita increases in the gross incarceration rate of 57 and 38%, respectively. Source: Glaze, Minton, and West (2009). The US population was 255 million in 1992 and 290 million in 2002.

8. This finding shows up in both the BJS State Court Processing Statistics series for 75 large urban counties, which accounted for about one third of the US population in 2002 and over half of the crime, and the BJS National Judicial Reporting Program series for 300 counties, which accounted for even more of both population and crime.

9. In the most recent American Society of Criminology meetings in Washington DC, Richard Rosenfeld spoke on a Presidential Panel Discussion entitled “Innovations in Criminal Justice Policy and Practice” about the need to evaluate policies based on what they were trying to accomplish, rather than some external objective goal unrelated to the goals of the policy. One example of this type of disconnect includes assertions that the federal sentencing guidelines fail to reduce crime. Given that the policy was not created to reduce crime, this seems like an odd criticism. We endorse this linkage of evaluation to the goals of the policy.

10. More recent examples where scholars create an explicit counterfactual for theory and empirical testing include the application of outcome analysis to the question of racial discrimination in police searches and bail setting (Ayres & Waldfogel, 1994; Knowles, Persico, & Todd, 2001).

11. We applaud those who hold up a high standard of proof for causal research on the impact of sentencing policy on crime. However, we caution researchers to use the same standards when evaluating all causal work on sentencing policies, including the impact of incarceration on future life outcomes for the individual and his/her family. Individuals who are opposed to the idea of using sentencing policies to prevent crime on philosophical grounds should state this opposition up front—empirical work on the causal impact of the policies on crime is clearly irrelevant for these individuals.

12. This dataset could be improved even further if the arrest data could be linked with dispositions, so that analyses could be conducted on a sample of arrested individuals, rather than simply those who are convicted, in order to deal with the selection problem discussed above.

13. An anonymous reviewer raised reasonable questions about our approach to generating these estimates. We are unaware of other estimates of false negatives, and we encourage an active discussion about the best way to estimate false negatives as a productive line of future social science research.

14. See also Schanzenbach and Tiller (2008) for a similar call.

15. An anonymous reviewer raised the issue that researchers are beholden to the system for access to the data. This makes them understandably unwilling to criticize the very actors who granted access to the data. We agree that this is a real issue, but hope that over time situations are created where access to the data is not limited to those who must necessarily curry favor with the actors who are to be analyzed in the data.

16. For a more general treatment of this problem, with a compelling and interesting example, see a paper on peer effects by Carrell, Fullerton, and West (2009).

17. See Note 1, supra.