Sentencing and Conventional Number Preferences: A Research Note

• DOI: 10.3818/JRP.8.1.2006.67
• Published date: 01 June 2006
• Date: 01 June 2006
• Subject Matter: Article

Sentencing and Conventional Number Preferences • 67

* Sentencing and Conventional Number

Preferences: A Research Note

Andrew Wiseman

Daniel Fischer

Michael Connelly

Wisconsin Sentencing Commission

* Abstract

Prior research on sentencing has shown that judges rely heavily on “conventional

number preferences” (CNPs), often defaulting to certain “preferred” sentences, e.g.,

three years, four years, ve years, etc., that come easily to mind. Data from Wiscon-

sin conrm that its judges, virtually unbound in their discretion, are likewise heavily

inuenced by CNPs, with 10 prominent sentences explaining almost all non-proba-

tion felony sentences. Wisconsin currently spends nearly $2,300 per prisoner, per

month. Strong preferences for three-, four-, and ve-year sentences, when shorter

terms would sufce, cost the state millions of dollars each year. Additionally, CNPs

may result in disproportionate sentences for offenders who commit the same crime.

This paper was funded in part by a grant from the National Institute of Justice (NIJ).

The authors thank their anonymous reviewers at Justice Research and Policy for their

careful reading and thoughtful recommendations. The authors also thank the Honorable

Thomas H. Barland, Eau Claire County Circuit Court (emeritus), who provided valuable

insight into the ndings reported here.

JUSTICE RESEARCH AND POLICY, Vol. 8, No. 1

© 2006 Justice Research and Statistics Association

68 • Justice Research and Policy

Sentencing and Conventional Number Preferences • 69

* Introduction

Since the 19th century, researchers have documented strong preferences among

judges for particular criminal sentences—terms that correspond with whole,

often round, numbers. In 1895, Francis Galton observed,

The terms of imprisonment that are most frequently awarded fall into

rhythmic series. Beginning with sentences reckoned in months, we see

that their maxima of frequency are at 3, 6, 9, 12, 15 and 18 months,

which are separated from one another by the uniform interval of 3

months . . . a round gure that must commend itself to the judge by its

simplicity (p. 175).

A century later, Ostrom and Ostrom (2002) observed the same phenomenon:

Actual sentence length may not be consistent with a continuous “scale

of severity.” Consider the distribution of prison sentences among 9,586

offenders convicted in the State of Michigan in 1995. Prison sentences

range from one month to 480 months. Michigan judges are free to as-

sign any term of days, months or years they wish. However, it is clear

that a small number of sentences predominate: 12, 18, 20, 24, 30, 36,

48, 60, 72, 84, 96 and 120 months. These 10 [sic] terms account for

over 78 percent of sentences issued in 1995 (p. 277).

As the Ostroms contend, profound policy consequences follow from observed

judicial preferences for particular sentences: public safety demands that offenders

receive their just deserts, but prisoners suffer unduly, and we squander scarce, ex-

pensive prison space if our afnity for particular numbers leads to sentences that

exceed just penal deserts (pp. 279–280). Essentially, the Ostroms’ ndings con-

rm what Fitzmaurice and Pease (1986) had stated years before—that judges increase

sentence length apace with what they perceive inexactly as greater culpability.

The Ostroms maintain that the policy implications of conventional num-

ber preferences (CNPs) are three. First, strong preferences for certain numbers

may result in disproportionate sentences—12 years vs. 15, or 15 vs. 20—for

similar offenders who commit the same crimes. Second, this inconsistency can

easily become racial disparity if judges consistently impose longer sentences on

minorities. And third, coupled with rising pressure to increase sentence length,

conventional number preferences put heavy pressure on available prison space.

These effects can have profound consequences anywhere. This research note

attempts to replicate the Ostroms’ ndings in Wisconsin, and raises additional

implications that follow from our analysis.

* Literature Review

As noted above, scholars have found that actual sentence length does not co-

incide with a continuous scale of severity. Ostrom and Ostrom (2002) offer

68 • Justice Research and Policy

Sentencing and Conventional Number Preferences • 69

three explanations for the appearance of CNPs among sentencing judges. First,

a restricted choice set reects the human tendency to simplify complex deci-

sions—judges proceed toward the “right” sentence from several options that

are “good enough.” This accords with Farrell and Holmes (1991), and Guthrie,

Rachlinsky and Wistrich (2001), who maintain that judges use mental shortcuts,

including stereotyping, framing, hindsight bias, and anchoring (making nal de-

terminations based on arbitrary starting points) to solve complex problems. Ac-

cording to Farrell and Holmes (1991), while these “tools” may facilitate sound

decision-making, they often produce cognitive illusions that systematically yield

erroneous sentences:

Alleged perpetrators of violent and property crimes, for example, are

likely dealt with more severely if their alternative statuses conform to

lower class and minority group stereotypes popularly associated with

such offenses. More severe treatment may be deemed necessary to in-

capacitate offenders whose behavior is thought deeply rooted in their

social backgrounds, [and] thus likely to reoccur.

. . .

Wheeler and his associates (1982) nd a strong relationship between

white-collar offenders’ occupational status and probability of impris-

onment. Contradicting popular and social scientic opinion, their data

show that offenders of higher occupational status are more likely to

re

ceive prison sentences than lower status, “run of the mill” offenders. . . .

Held more responsible, defendants of higher social status are perceived

[to be] exceptional and their deviation more serious, warranting more

severe penalties (pp. 534–535).

Second, when reducing their choice set, judges default to sentences that are

easily recalled. Ostrom and Ostrom (2002) suggest that this process may illus-

trate what Kahneman, Slovic and Tversky (1982) have termed the “availability”

heuristic. The availability heuristic leads to several decision-making biases, chief

among them being that judges in discretionary environments will most frequent-

ly impose those few “prominent” sentences that come easily to mind. Dhami

(2005) and Dhami and Ayton (2001) have similarly reported that the decision-

making process in bail decisions in the United Kingdom is less complicated than

one might imagine:

[P]sychological research has examined the cognitive processes involved in

making bail decisions (Dhami 2003; Dhami & Ayton 2001). It has been

found that the decisions of both judges responding to simulated cases

and judges responding to real cases in the courtroom are better predicted

by a simple heuristic called the matching heuristic than by more complex

strategies. According to the heuristic, judges search through only a small

subset of the available case information, and then base their decision on

the level of one factor alone (Dhami, 2005, p. 371).