Questioning Blackmun's Thesis: Does Uniformity in Sentencing Entail Unfairness?

• DOI: http://doi.org/10.1111/lasr.12141
• Published date: 01 June 2015
• Date: 01 June 2015

Questioning Blackmun’s Thesis: Does Uniformity in

Sentencing Entail Unfairness?

Ben Grunwald

It is commonly believed among criminal justice scholars that sentencing

guidelines increase uniformity in sentencing at the cost of fairness. They rea-

son that guideline systems rarely take all relevant case characteristics into con-

sideration, and as a result, impose sentences in particular cases that are biased

relative to the ideal or best sentence. This bias effect is one of the primary theo-

retical and practical challenges faced by courts and sentencing commissions in

the last 30 years, and provides one of the strongest arguments against manda-

tory sentencing guidelines. This article identifies a second effect of guidelines

on fairness, which has not been sufficiently acknowledged by the scholarly lit-

erature: the variance effect increases the fairness of sentences directly by

increasing uniformity. This article uses statistical simulation to examine the

relationship between the variance effect and the bias effect. The results pro-

vide substantial evidence that the variance effect is comparatively large, and

that it may often outweigh the negative effects of bias. Under these conditions,

sentencing guidelines will both increase uniformity and increase fairness.

Until the 1970s, judges in the United States enjoyed nearly

unlimited discretion in assigning sentences to criminal offenders.

Judges were free to adopt their own theory of punishment, to

determine how that theory applied to the facts of a case, and to

select the most appropriate punishment scheme on that basis.

Typically, only very wide statutory ranges constrained judges’

power to individualize sentences (Stith and Koh 1993). Early

empirical research from the 1960s to the early 1980s revealed

that the existing system produced large sentencing disparities,

finding that similar defendants with similar convictions often

received different punishments. Scholars theorized that these dis-

parities arose from variations in judges’ ideological background

(Forst and Wellford 1981; Partridge and Eldridge 1974), from

discrimination based on legally irrelevant characteristics such as

race and ethnicity (Baldus, Pulaski, and Woodworth 1983), and

from differences in local context (see Dixon 1995). Contemporary

Please direct all correspondence to Ben Grunwald, Department of Criminology, Uni-

versity of Pennsylvania; e-mail: Ben.Grunwald@gmail.com.

Law & Society Review, Volume 49, Number 2 (2015)

V

C2015 Law and Society Association. All rights reserved.

499

research confirms that these phenomena continue to impact sen-

tencing today (Abrams, Bertrand, and Mullainathan 2012; Albo-

netti 1997; Bushway and Piehl 2001; Kaut 2002; Ulmer and

Johnson 2004).

In response to evidence of disparity, federal and state legisla-

tures established sentencing guideline systems to constrain judi-

cial discretion. These systems typically used a limited number of

variables to determine a sentence range for each defendant.

Judges were then encouraged and in some systems required to

impose a sentence within that range (Frase 1995). The level of

discretion left up to the judge varied widely by jurisdiction, but

no system was more restrictive than the Federal Sentencing

Guidelines, which gave nearly all weight to crime severity and

criminal history (Tonry 1993).

Almost immediately, the Federal Guidelines were criticized

for promoting uniformity at the cost of fairness in individual

cases (e.g., Ogletree 1987). Critics widely argued that the federal

guidelines decreased the fairness of sentences by constraining

judges’ ability to take relevant case characteristics into considera-

tion. I call this the bias effect of sentencing guidelines: guidelines

can bias sentences away from the fairest or most appropriate sen-

tence by limiting judges’ ability to take all relevant case character-

istics into consideration and to fully individualize punishment.

While discussing disparities in the administration of the capital

punishment, Justice Harry Blackmun articulated a well-known

statement of the bias effect: “Experience has shown that ... consis-

tency and rationality ... are inversely related to [fairness]. A step

toward consistency is a step away from fairness” (Callins v. Collins

1994). Although Blackmun articulated this thesis in the context

of the death penalty, I take his critique as a paradigmatic formu-

lation of a broad and popular criticism of mandatory sentencing

guidelines echoed by judges (Schwarzer 1991) and scholars

(Alschuler 1991; Freed 1992; Ogletree 1987; Tonry 1993; Ulmer

1996), which persists until today (Kim 2004; O’Hear 2006; Osler

2003).

This article advances the sentencing literature in two ways.

First, it develops a novel framework for conceptualizing the rela-

tionship between uniformity and fairness. Second, it tests Black-

mun’s Thesis by identifying and examining a second effect of

sentencing guidelines that has not been acknowledged by the aca-

demic literature: increasing uniformity through guidelines has a

second effect—a variance effect—of directly increasing the fairness

of sentences on average.

This article uses simulation modeling to examine the relation-

ship between sentencing guidelines, uniformity and fairness. I

begin by defining an “ideal” sentence for a set of equivalent

500 Does Uniformity in Sentencing Entail Unfairness?

criminal cases based on assumptions that are favorable to Black-

mun’s Thesis. I then randomly generate a preguideline distribu-

tion of sentences centered around the “ideal” sentence. Based on

existing estimates from the literature, I introduce a sentencing

guideline system, and posit a series of plausible bias and variance

effects on the sentence distribution. The average distance of sen-

tences in the preguideline and postguideline distributions from

the “ideal” sentence is then compared to determine whether,

under these assumptions, Blackmun’s Thesis would hold. This

approach provides an important methodological benefit. Unlike

many other analytic methods in the sentencing literature, this

article avoids the need to assume a thick normative theory about

the purposes of punishment and the case characteristics that are

relevant in sentencing.

The results of the analysis show that, under plausible condi-

tions, the variance effect of sentencing guidelines is comparatively

large and may often outweigh the negative effects of bias. When

the bias effect is outweighed, Blackmun’s Thesis does not hold,

and the guidelines both increase uniformity and increase or

maintain the existing level of fairness, thereby defusing one of

the most formidable arguments against restrictive sentencing

guideline systems.

The remainder of the article proceeds as follows. I begin with

a brief history of sentencing guidelines, and a review of the

empirical literature on sentencing disparity. Next, I develop a

conceptual framework to clarify and explore the contours of

Blackmun’s Thesis, and then describe the basic design of the

study. I conclude by reporting the results of the analysis, and by

discussing the implications for the academic and policy debate on

sentencing guidelines.

Sentencing Guidelines

From the late 1970s to 1990s, Congress and a number of

state legislatures established sentencing commissions to design

guidelines that would increase uniformity in sentencing. Con-

gress, for example, established the United States Sentencing

Commission through the Sentencing Reform Act of 1984. The

commission was authorized to develop general rules to regulate

the form (e.g., fine, probation, or imprisonment) and intensity

of sentences on the basis of a list of variables including serious-

ness of offense, criminal history, age, education, vocational skills,

mental and emotional condition, family responsibilities, commu-

nity ties, and extent of participation in the offense. In 1987, the

commission passed a guideline system based primarily on two of

Grunwald 501