         
       
    

   

\

if"
7,,

l‘
‘1

;—#*
.—
r if

7 7 777

WW1

W
‘H‘ *'

 

155% W

 

i
to ’l

 

UBRARY
Michigan State
University

 

 

 

This is to certify that the
" thesis entitled

THE EFFECT OF HIGHER EDUCATION ON
POLICE BEHAVIOR

presented by

JASON RYDBERG

has been accepted towards fulﬁllment
of the requirements for the

MS. degree in CRIMINAL JUSTICE

 

 

(it/W4 W

 

Major Professor’s Signature

7/17/07?

Date

MSU is an Afﬁrmative Action/Equal Opportunity Employer

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K.IProj/Acc&Pres/ClRC/DateDue indd

 

THE EFFECT OF HIGHER EDUCATION ON POLICE BEHAVIOR
By

Jason M. Rydberg

A THESIS
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
MASTER OF SCIENCE

Criminal Justice

2009

ABSTRACT
THE EFFECT OF HIGHER EDUCATION ON POLICE BEHAVIOR
By

Jason M. Rydberg

In the past, police scholars have examined the impact of higher education on
different measures of ofﬁcer behavior. These previous studies were conducted in such a
way that comparing ﬁndings was difﬁcult, such as focusing on a single behavior per
study. The research presented in this manuscript examines the effect of college education
on three dimensions of police behavior (e.g., arrests, searches, and the use of force), so as
to improve comparability. The results of the analysis indicate that higher education
carries no inﬂuence over the probability of an arrest or search occurring in a police
suspect encounter. College education does, however, signiﬁcantly reduce the likelihood
of force occurring in an encounter. More speciﬁcally, some exposure to college reduces
the likelihood of verbal force, but a four-year degree is necessary to reduce the likelihood
of physical force. Results may be due to the amount of discretion ofﬁcer’s exercise in
pursuing these behaviors. Limitations in the previous literature to develop and test
hypotheses regarding why higher education should be expected to inﬂuence police
behavior are highlighted. Finally, recommendations for future inquiries revolving around
theory development and incorporation of research from the ﬁeld of education are

presented along with implications for policy.

AKNOWLEDGEMENTS

First and foremost, I would like to thank the members of my thesis committee
(Dr. Timothy Bynum, Dr. Christina DeJong, and Dr. William Terrill) for their guidance
in preparing this manuscript. I would like to thank Dr. Terrill, in particular, for the
numerous thoughtful revisions and taking the time to discuss the ideas in this thesis with
me at length in order to draw out the best ways to present them. His direction during my
masters studies as a whole has been immensely important to my time at Michigan State
University. I would also like to thank Rebecca Stone for her constructive criticism and
Eric Grommon for his incessant nagging, both of which undoubtedly improved the
quality of this thesis well beyond what I thought myself capable of writing on my own.

In a broader sense, I would like to thank my parents for their commitment to my
education and for their ever-present support. And I would also like to thank Dr. Sean
Varano, for without his advice and ﬁ'iendship, I would probably still be working in an

intellectually unfulﬁlling job, wondering whether to continue my education.

iii

TABLE OF CONTENTS

LIST OF TABLES .......................................................................................................... v
CHAPTER I: INTRODUCTION .................................................................................... 1
CHAPTER H: HISTORICAL BACKGROUND ........................................................... 4
The Status of the American Police Force in the Early Twentieth Century ......... 4
National Commissions from the 1930s to the 19703 .......................................... 7
Higher Education for Police from the 19805 to Present ................................... 11
CHAPTER HI: LITERATURE REVIEW .................................................................... 15
The Effect of Ofﬁcer Education on Attitudes ................................................... 16
The Effect of Ofﬁcer Education on Arrests ...................................................... 18
The Effect of Ofﬁcer Education on Searches ................................................... 21
The Effect of Ofﬁcer Education on the Use of Force ....................................... 22
Previously Employed Theoretical Frameworks ................................................ 26
Proposed Theoretical Framework ..................................................................... 29
The Gap in the Literature .................................................................................. 31
The Current Inquiry .......................................................................................... 31
CHAPTER IV: METHODOLOGY .............................................................................. 33
Study Sites ........................................................................................................ 33
Data ................................................................................................................... 35
Variable Descriptions ........................................................................................ 37
Data Analysis .................................................................................................... 4O
Validity, Reliability, and Generalizability ........................................................ 43
CHAPTER V: RESULTS ............................................................................................. 45
Descriptive Statistics ......................................................................................... 45
Bivariatc Analysis ............................................................................................. 49
Multivariate Analysis ........................................................................................ 52
CHAPTER VI: DISCUSSION ..................................................................................... 72
Overview of Findings ....................................................................................... 72
Further Questions for Consideration ................................................................. 74
Limitations of the Current Inquiry .................................................................... 79
Directions for Future Research ......................................................................... 80
Implications for Policy ...................................................................................... 83
APPENDICIES ............................................................................................................. 85
REFERENCES ............................................................................................................. 91

iv

LIST OF TABLES

Table 4.1: Variable Descriptions and Hypothesized Effects ........................................ 42
Table 5.1: Descriptive Statistics for Dependent Variables ........................................... 46
Table 5.2: Descriptive Statistics for Ofﬁcer Level Independent Variables .................. 47
Table 5.3: Descriptive Statistics for Suspect Level Independent Variables ................. 48
Table 5.4: Descriptive Statistics for Encounter Level Independent Variables ............. 48
Table 5.5: Bivariate Distribution of Arrests by Ofﬁcer Education ............................... 50
Table 5.6: Bivariate Distribution of Searches by Ofﬁcer Education ............................ 50
Table 5.7: Bivariate Distribution of Force by Ofﬁcer Education ................................. 51
Table 5.8: Bivariate Distribution of Verbal/Physical Force by Ofﬁcer Education ....... 52

Table 5.9: Binary Logistic Regression of Arrests on Ofﬁcer, Suspect, and Encounter
Characteristics ............................................................................................................... 59

Table 5.10: Binary Logistic Regression of Searches on Ofﬁcer, Suspect, and Encounter
Characteristics ............................................................................................................... 62

Table 5.11: Binary Logistic Regression of Force on Ofﬁcer, Suspect, and Encounter
Characteristics ............................................................................................................... 64

Table 5.12: Multinomial Logistic Regression of Force on Ofﬁcer Characteristics ...... 68

Table 5.13: Multinomial Logistic Regression of Force on Ofﬁcer and Suspect
Characteristics ............................................................................................................... 69

Table 5.14: Multinomial Logistic Regression of Force on Ofﬁcer, Suspect, and
Encounter Characteristics ............................................................................................. 70

Table A. 1: Samples, Data, and Direction of Findings in Ofﬁcer Education
and Attitudes ................................................................................................................. 85

Table A2: Samples, Data, and Direction of Findings in Ofﬁcer Education
and Arrests .................................................................................................................... 87

Table A.3: Samples, Data, and Direction of Findings in Ofﬁcer Education and Use
of Force ......................................................................................................................... 88

Table B. 1: Distribution of Ofﬁcer Characteristics by Education Level ....................... 90

vi

CHAPTER I: INTRODUCTION

Police scholars and practitioners have long called for the adoption of a college
education requirement for police ofﬁcers as a condition of employment (Carter & Sapp,
1990). Beginning with the professional movement in the early 19003, the importance of
education was seen as a means to a better style of policing. As the century progressed
several high proﬁle national commissions (e.g., Wickersham Commission and President’s
Commission on Law Enforcement and Administration of Justice) also heralded the
beneﬁts of police education. More recently, ofﬁcer education has been linked to
community policing, which is thought to require a greater aptitude for innovation and
creativity in problem solving among line-level ofﬁcers (Carter & Sapp, 1992).

Despite repeated calls for a college education requirement, few agencies have
instituted such a policy. In fact, only one percent of local police departments in the
United States require a four-year college degree (Hickman & Reaves, 2006). The
discrepancy between the recommendations of national leadership and actual departmental
requirements may be attributable to a paucity of evidence suggesting that education has a
desirable impact on police behavior. While it seemed sensible to police reformers that a
college education would result in a better police ofﬁcer, the rationale for believing Such a
relationship existed was grounded primarily in rhetoric (Carter & Sapp, 1990). Indeed,
numerous opponents of college education requirements question exactly which aspects of
higher education enhance an ofﬁcer’s performance on the job (Bayley & Bittner, 1997;
Bittner, 1970; Miller & Fry, 1976a; Shemock, 1992).

Since the 19703 a body of research has developed that attempts to empirically

examine the impact of college education on the police. The majority of this research has
focused on the effect of education on ofﬁcer attitudes, ﬁnding that college-educated
ofﬁcers hold beliefs that are less authoritarian (Dalley, 1975), rigid (Roberg, 1978), and
punitive (Carlan & Byxbe, 2000; Guller, 1972) than their non-college educated
colleagues. To a lesser extent, some research has focused on speciﬁc aspects of police
behavior, examining the effect of education on an ofﬁcer’s propensity to arrest (Bozza,
1973; Fickenauer, 1975) and to use force (Paoline & Terrill, 2007; Worden, 1996).

Unfortunately, many of these previous inquiries have been driven by weak methodologies

and atheoretical frameworks.1 The National Academics Panel on Police Policy and
Performance, in particular, cited concern over poor methodologies such as inadequate
samples, as well as the inability to control for potentially theoretically relevant variables
(Skogan & Frydl, 2004). Moreover, while there is some evidence suggesting college
educated ofﬁcers behave differently than non-college educated ofﬁcers, ﬁndings
regarding the direction of education’s impact on police behavior have been largely

inconsistent. Prior work in this area has also been limited to studies focusing on but one

 

1Many of the works in this body of literature devote little effort to developing hypotheses regarding why or
how education impacts police attitudes or behaviors. For example, Cunningham (2006) states that data
show education to have a signiﬁcant impact on an ofﬁcer’s disciplinary infractions because of “some
unknown reason” (p. 22), but does not offer evidence for what that reason may be. Yet more studies cite the
ﬁndings of previous literature as a reason for believing education should have an impact on ofﬁcer
behavior, when those earlier studies suffer from their own shortcomings. For example, Smith, Locke, and
Fenester ( 1970) are commonly cited as evidence that ofﬁcers with a college degree are less authoritarian
than their non-college educated colleagues. Their study was actually the third in a series of examinations on
the impact of education on the New York City Police Department. In the ﬁrst of the three studies (Smith,
Locke, & Walker, 1967), the authors determined that there was existing differences in authoritarianism
between their college and non-college samples well before the ofﬁcers had completed their college degrees.
This signiﬁes a time-order problem that was never remedied but continues to be cited as evidence of
educations’ impact on the police.

behavioral outcome at a time (e.g., arrest or force), thereby making it difﬁcult to compare
outcomes across studies (e.g., the role of education on arrest versus force).

The current inquiry will attempt to overcome some of the limitations associated
with previous studies, particularly the latter one. Using observational data ﬁ'om two
medium-sized cities, this research will examine the impact of ofﬁcer education on three
speciﬁc measures of behavior (i.e., arrests, searches, and the use of force). The analysis
will allow for a comparison of education’s impact on these behaviors, which has not been
available in the literature to date. If ofﬁcer education has a signiﬁcant effect on more than
one of these behaviors, these effects may be compared to determine where education’s
impact on police behavior is strongest.

This manuscript will ﬁrst review the historical background of higher education in
policing, from the ﬁrst reform period in the early twentieth century until present day.
Following this section, a literature review will cover previous empirical examinations of
education’s impact on police attitudes, and propensity to arrest, search, and use force.
The thesis will continue with a deduced theoretical framework, research questions, and a
discussion of the methodology. Univariate statistics, bivariate analyses, and multivariate
analyses will then examine the impact of ofﬁcer education on arrests, searches, and the
use of force. Finally, the results of the analyses will be interpreted and discussed, along

with directions for future research and implications for policy.

CHAPTER II: HISTORICAL BACKGROUND
The movement advocating a college-educated police force arose from efforts in
the early 19003 to professionalize the American police. Though there have been
competing deﬁnitions of professionalism between scholars and practitioners (Skolnick &
Fyfe, 1993; Baro & Burlingarne, 1999), reformers sought to attain it primarily through

redeﬁnition of the police mission, bureaucratization of police departments, and the

.r.-_‘

improvement of personnel quality (F ogelson, 1977). From the origin of the
professionalism movement, increasing the education level of ofﬁcers was thought to be .
one such avenue of improvement (Sherman, 1978). By improving the education level of

the police ofﬁcer, reformers believed that not only would the ofﬁcer become more

effective in his/her day-to-day tasks and more proﬁcient with technology, but the public

image of the police as a whole would be heightened to that of a professional occupation

(LeDoux, Tully, Chronister, & Gansneder, 1984).2
The status of the American police force in the early twentieth century

The commonly held image of the American police in the early 19003 was that of
ineffectiveness and corruption, both criminal and political (Walker, 1977). The image
was largely accurate for the time. Fogelson (1977) writes of media interviews in the
19303 where former ofﬁcers depicted urban police as irresponsible drunks, dabbling in

corrupt activities such as gambling and prostitution, and purposefully employing

 

2 Bittner posits that the driving force behind the attempts of reformers to link higher education to policing

“is to abolish permanently the idea that is all too prevalent in our society that if one does not want to take
the trouble of becoming something worthwhile, he can always become a cop” (1970: 83).

4

excessive force to gain compliance from suspects.3 Political interference from city ward
bosses was partially to blame for the poor state of police character, as they held political
inﬂuence over almost all ranks in the police hierarchy. Reformers of the period
recognized that the removal of partisan politics from policing was necessary to improve
the public image of the police (Walker, 1977), though they also recognized that reform
could not stop at the moral character of ofﬁcers.

Improvement of the police required attention towards the quality of personnel.
Fogelson (1977) writes that police in the early part of the twentieth century were not only
morally corrupt but also somewhat incompetent. Several sources mention an inquiry at
the time of the First World War which indicated that 75 percent of police personnel could

not pass an Army intelligence test (LeDoux et al., 1984; National Commission on Law

Observance and Enforcement, 1931; Sherman, 1978).4 Increasing the education of police
recruits, among other options, was viewed as a vehicle to propel the police ofﬁcer to a
similar status as other professional occupations (Bittner, 1970; Fogelson, 1977). The ﬁrst
such organized attempts to tie the policing profession to academic achievement were lead
by August Vollrner, the police chief in Berkley, California from 1905 to 1932 (Carte,

1973)

 

3 The use of excessive force was frequently referred to as the “third degree” during this time (Fogelson,
1977)

4 The army intelligence test being referred to was the Army Alpha intelligence test created by Yerkes
(1921). Gould (1996) offers a scathing critique of Yerkes methods, questioning the test’s reliability, and
providing evidence that it does not actually measure intelligence, but rather familiarity with American
culture. As such, this ﬁgure of “75 percent” should be approached with caution. It is unfortunate then, that
the Wickersham Commission would later advocate using the Army Alpha to determine whether an
applicant would be worthy of employment as a patrol officer (National Commission on Law Observance
and Enforcement, 1931).

Volhner is best-known for his effort to single-handedly professionalize the police
force in Berkley (Walker, 1977). His contributions to the department were largely in the
form of implementing the use of new technologies to aid the police in their work. Of
particular importance to the current inquiry, Vollmer encouraged recruits to attend classes
in police administration at the nearby University of California (Carte, 1973). In viewing
the new advances in technology available to the policing practice, Volhner posited that
recruits would require skills that they could not learn while on the job. Advocating a
college education was an effort then to provide would-be ofﬁcers the background
necessary to be a part of the prestigious, more technologically inclined police profession
(Carte 1973).

The implementation of Vollmer’s reforms created a publicly popular department
referred to as Berkley’s “college cops.” In reality the Berkley police department was
never made up of a majority of college graduates (Carte, 1973). This was not surprising
given that the average American citizen at the time did not ﬁnish high school (Roberg &
Bonn, 2004). Elsewhere in the country, survey data indicated that only “two out of three
[ofﬁcers] ﬁnished grade school” and “only one out of ten graduated from high school”
(Fogelson, 1977: 102). These data indicate that police recruits in the early 19003 were
representative of the communities they served, in terms of education at least. In light of
this, police reformers seeking to elevate the profession were willing to require more of
recruits than the average citizen could offer. In the next decade these wishes would serve
as the ﬁrst calls for a college education requirement for the police.

Vollmer’s efforts to link higher education and policing showed little impact

outside of Berkley before the 19303. As he gained national recognition he was invited to
aid other departments as a consultant, and spent a year in Los Angeles in the 19203. He
found that it was far more difﬁcult to implement reform in a department that was many
times larger than that of Berkley (Walker, 1977).

National commissions from the 19303 to the I970s

Wickersham commission. Despite lack of success elsewhere, Volhner’s
achievements in Berkley and his tenure as the president of the International Association
of Chiefs of Police (IACP) in 1922 gained him enough national recognition to be selected
as a key contributor to the 1929 National Commission on Law Observance and
Enforcement (1931; Walker, 1977) to report on the state of the police. The commission
would be better known as the Wickersham Commission, named after its chairman, then
Attorney General George W. Wickersham (F o gelson, 1977). The ﬁnal report of the
Commission discussed many crime-related topics relevant to the time - prohibition and
corruption in law enforcement receiving the most attention (Walker, 1977).

Most relevant to the current research, the Wickersham Commission believed that
higher education played a potential role in improving the quality of police personnel.
Coauthoring the report on the police, David Monroe with the University of Chicago
optimistically wrote that he wished to see a college education become available to all
police recruits. He believed that only under that circumstance would the police “ever
hope to successfully cope with the crime situation” (National Commission on Law
Observance and Enforcement, 1931: 85). The recommendations of the Commission

would receive little follow-through from police agencies. Police administrators were

skeptical of the beneﬁts of a college education for recruits (Decker & Huckabee, 2002),
especially when it was likely that they never attended college themselves (Bittner, 1970;
Fogelson, 1977). At the time of the Wickersham ﬁnal report, the majority of police
agencies in this period did not require even a high school education of their recruits
(Roberg & Bonn, 2004). This most likely reﬂected the educational attainment of the
average citizen at the time, and as the American population became more educated,
ultimately the police force would as well.

The issue of college education for police recruits received very little attention
between the 19303 and the 19603. However, the lack of congruence between commission
recommendations and the educational requirements of police agencies did not stop
academic institutions from providing law enforcement-based education. During that time
some of the ﬁrst full-time college curriculums designed for police ofﬁcers were being

created. In 1935, Michigan State University became the ﬁrst institution to offer a full-

time ﬁve-year pre-service curriculum (LeDoux et al., 1984).5 Besides a few select
programs at some well-known universities, the majority of curriculums designed for
police recruits existed in community colleges as two-year associate degree programs in
police science. By 1957, survey data indicated that 77 degree granting programs in
police-related science existed in the United States (Germann, 1957; Stephens, 1976).
Commissions of the 1960s and 19703. The 19603 would see increased relevance

for a college education for police recruits and the issue again received attention from

 

5 The Michigan State University curriculum for police administration in 1935 included three years and a
term of physical education courses such as wrestling and boxing, along with a heavy load of mathematics
and science courses such as bacteriology, mammalian anatomy, and general chemistry. In the fourth year of
the program the curriculum required that students begin an 18 month training period at the Michigan State
Police barracks, where they received the same training as police recruits (Michigan State College, 1936).

8

national commissions. Following the relatively low crime rates in the 19503, the 19603
brought with it a national wave of crime and civil unrest. Public attention on the growing
crime rate called into question the competence and ability of the police to effectively
cope (Sherman, 1978). Walker (1977), reﬂecting on that period, concluded that the
previous era of police reform in the 19203 and 19303 was the cause of the more recent
lack of adaptability. During that period, while educational reforms failed to gain traction,
organizational reform from a service oriented approach to a legalistic, paramilitary
approach had become widely adopted. This shift was posited to be the cause of the police
becoming further separated from, and antagonistic towards, the communities in which
they served. Future commissions would recognize that the problem was not the rising
crime rates, but that the paramilitary shift in the police threatened democratic values
(Jacobs & Magdovitz, 1977). .

Beginning with the 1967 President’s Commission on Law Enforcement and the
Administration of Justice, numerous reports would be ﬁled recommending that the police
would need to reform their role in order to adapt to the changing American society.
Among the numerous commissions a common perception was that in order for the police

to effectively adapt to complex tasks in a complex society, educational standards must be
raised (Carter & Sapp, 1990; Hawley, 1998).6 Many of the commissions went as far as
recommending that all police ofﬁcers obtain a four-year college degree within the next

decade (National Advisory Committee on Criminal Justice Standards and Goals, 1973).

 

6 The other commissions being the 1969 National Commission on Causes and Prevention of Violence, the
1971 President’s Commission on Campus Unrest, the 1973 American Bar Association Project on Standards
for Criminal Justice, and the 1973 National Advisory Commission on Criminal Justice Standards and Goals
(Sherman, 1978).

Unlike the recommendation of the Wickersham Commission some 30 years earlier, the
educational recommendation made in the 19603 and 19703 would receive some follow-
through by police agencies.

In order to allocate ﬁmding towards higher education for the police (and other
criminal justice personnel), the 1968 Omnibus Crime Control Act created the Law
Enforcement Education Program (LEEP). The LEEP provided federal funding to colleges
and universities to create curriculums for police and funding to police departments to
provide incentives for ofﬁcers to return to school to obtain a degree (Carter & Sapp,
1990). Between 1968 and 1978 LEEP provided approximately $300 million towards the
education of criminal justice ofﬁcials, many of which were police (Sherman, 1978). The
result was something of an explosion in the number of degree-granting programs
designed for law enforcement personnel. To illustrate, in 1966 there were 184 colleges
and universities offering such programs. By 1976 this ﬁgure had increased to 1,070
institutions, though the majority were being offered in community colleges, with classes
being taught by fellow police ofﬁcers (Sherman, 1978).

As a result of the availability of federal subsidies for their education, American
police ofﬁcers took advantage of the available funds and their average education level
increased. In 1960, only three percent of ofﬁcers held a four-year degree. By 1974 this
ﬁgure had increased to nine percent of ofﬁcers. The proportion of ofﬁcers that had taken
at least one college course increased substantially over the same period (National
Planning Association, 1978). Despite increases in the education level of ofﬁcers, police

agencies had not increased their requirements by similar lengths. By 1975, only six

10

percent of departments required any college education of their recruits, and less than one
percent required a four-year degree (National Planning Association, 1978). It would
appear that police agencies were unwilling to increase their educational requirements to
the extent recommended by national commissions in the 19603 and 19703, similar to the
reaction of agencies following the Wickersham report some 30 plus years earlier.
Higher education for police from the 1980s to Present

Following the criticism of the 19603 which accused the police of becoming too
militaristic and disconnected ﬁ'om the communities they served, the 19803 would see
reform that attempted to mend the gap between ofﬁcers and citizens. Community
policing, or problem-oriented policing, was a product of this era of police reform. The
community policing approach attempted to change the focus of police effectiveness ﬁ'om
quantities such as the number of arrests or stops made, to a focus on the ability of police
to solve problems. This approach was considered as a strategy for the police to target the
underlying issues causing crime in communities (Goldstein, 1979; Troj anowicz, 1994;
Xu, Fiedler, & Flaming, 2005). Adapting such an approach was also thought to broaden
the role of the police in the context of community issues. Ofﬁcers would be required to
establish rapport with citizens in order to more effectively identify sources of disorder
and locate problems as identiﬁed by residents (Paoline, Myers, & Worden, 2000; Wilson
& Kelling, 1982).

In 1988 the Police Executive Research Forum (PERF) conducted a national
survey examining the state of police education in America, which included a review of

the literature to date and the surveying of a large number of police agencies. The research

11

population for the PERF survey included all local police departments serving populations
of 50,000 residents or more, all state police/highway patrol agencies, all countywide
police departments with 100 or more employees, and all county sheriff's departments
with 100 or more employees, for a total of 699 agencies surveyed (Carter & Sapp, 1990).
The results of the PERF study indicated that the adoption of a community policing model
increased the responsibilities of patrol ofﬁcers which required them to be better decision
makers, more innovative, and more tolerant. The authors of the PERF report believed that
these ﬁndings made college education for police ofﬁcers more relevant than in the past
(Carter & Sapp, 1990, 1992).

The PERF survey also reported changes in the education level of police ofﬁcers.
The most dramatic change from the past was seen in ofﬁcers holding a four-year degree.
Survey ﬁndings indicated that 23 percent of the 250,000 ofﬁcers surveyed held four-year
degrees, up from nine percent in 1974. The dramatic increase in the percent of ofﬁcers
with college degrees appeared to have happened despite little grth in the number of
police agencies requiring a college degree. By 1988 only 14 percent of police
departments surveyed required any college of their recruits, up from six percent in 1975.
Of those 14 percent, less than one percent required a four year degree, while nine percent
required a two-year degree. Police agencies surveyed did indicate, however, that recruits
with college credits would hold a competitive advantage in selection (Carter & Sapp,
1 990, 1992).

More recent data from the Law Enforcement Management and Administrative

Statistics Survey (LEMAS) indicates that while police ofﬁcers became more educated,

12

police departments lag behind in minimum education requirements. The 2000 LEMAS
indicated that 83 percent of all departments required only a high school diploma of their
recruits, while eight percent required a two-year degree and one percent required a four-
year degree. For departments serving cities with one million or more residents, a third
required some college of their recruits, while none required a degree of any kind. By
2003, the percentage of departments requiring degrees of their recruits remained stable,
with nine percent requiring a two-year degree and one percent requiring a four-year
degree (Hickman & Reaves, 2003, 2006).

These are the most recent data available to describe the state of higher education
in policing. Since the Wickersham Commission ﬁrst called for a college education
requirement for the police in 1931, departments have been reluctant to follow this
suggestion. In attempting to explain the disconnect, Bell (1979) and Hawley (1998)
suggest that police administrators may be concerned about factors such as keeping
ofﬁcers representative of the communities they serve, or that instituting a minimum
college requirement may be discriminatory towards women and minorities. Carter, Sapp,
and Stephens (1988) posit that police agencies seeking to hire college educated recruits
would have to make serious considerations of the available applicant pool. These
considerations include how many college educated recruits would be available, how
much additional salary would the department have to offer to remain competitive with
private industry, and how much additional background investigation of recruits would
have to be conducted, given that college students tend to be a more transient population.

Another potential reason for the small proportion of departments requiring a college

13

 

degree may be the lack of evidence national commissions have presented to make the
case that a college education requirement will produce a desirable outcome (Skogan &

Frydl, 2004).

14

 

CHAPTER HI: LITERATURE REVIEW

As illustrated in the preceding section, numerous national commissions
recommended that police agencies require a four-year degree, however, national statistics
indicate that few departments have followed suit (Hickman & Reaves, 2006). Lack of
evidence for a desirable effect of education on police behavior has been cited as a major
reason for the discrepancy between national recommendations and actual department
requirements. As Carter and Sapp (1990) point out, the recommendations of the
commissions were largely based on intuition and rhetoric, rather than empiricism. Higher

education, they posit, was viewed as a solution to the problems of policing because it was

logical that a young profession would require more rigorous academic preparation.7
Reﬂecting on the lack of empirical basis for the arguments of previous commissions, the
1973 National Advisory Commission on Criminal Justice Standards and Goals concluded
that they had not provided sufﬁcient justiﬁcation for a minimum education requirement
for police recruits, and failed to link higher education to desirable objectives.

As noted, the previous endorsement for a college education requirement was
grounded in perceived commonsense and emotion. These are, however, superﬁcial
methods for understanding complex relationships when compared to scientiﬁc inquiry.
Since the early 19703, a body of literature has formed which attempts to empirically
examine education’s effect on police ofﬁcers. Research on this relationship has measured

police performance across several different dimensions. Some studies focus only on

 

7 Here Carter and Sapp ( 1990) were comparing the relatively young policing profession to older, more
established professions such as law or medicine.

15

education’s effect on the attitudes of ofﬁcers.8 Other studies have focused on
perforrnance-based measures. Given that the performance of police ofﬁcers is multi-
faceted, previous authors have employed a multitude of measures in their research. These
include measuring performance through perceptions of quality or satisfaction (e. g. citizen
complaints) (Brandl, Stroshine, & Frank, 2001), supervisor ratings (Smith & Aamodt,
1997), injuries on the job (Cohen & Chaiken, 1973), and personal job satisfaction
(Dantzker, 1993). The current inquiry is most related to previous studies that focus on
speciﬁc ofﬁcer-suspect encounter outcomes such as conducting a search, pursuing an
arrest, or using force.

The evidence available does not provide a clear picture of the nature of the
relationship between education and police performance. It is possible to recognize
common themes in the results (Carter & Sapp, 1990), yet these are undermined by poor
methodologies such as small samples or inadequate deﬁnitions of performance (Sherman,
1980; Skogan & Frydl, 2004). The following section will review the available literature
within the areas of ofﬁcer attitudes, arrests, searches, and the use of force.

The Eﬂect of Oﬁ‘icer Education on Attitudes

Previous studies have assessed education’s impact on the attitudes of ofﬁcers.9
These works posit that the attitudes ofﬁcers hold in regards to their work shape their
behavior on the job, though behavior sometimes is not measured directly after the fact

(Worden, 1989). While studies focusing on ofﬁcer attitudes are intuitively appealing

 

8 . . . . .
Because the current mqurry rs focused on education’s effect on actual ofﬁcer performance, attitude-
focused literature will only be mentioned for the sake of completeness.

9 See Table 2 in the Appendix for additional information on each study’s samples, data, and direction of
ﬁndings.

16

there is actually little evidence for a strong link between attitudes and behavior (Muir,
1977; Sherman, 1978; Worden, 1989). Nevertheless, researchers have still spent
considerable effort examining the effect of education on ofﬁcer attitudes.

Research hypotheses in the literature examining this relationship tend to posit that
police ofﬁcers with more education will hold attitudes that deviate from those typically
associated with traditional police work, which primarily center on authority and rigidity
(Muir, 1977; Skolnick, 1964). Using scales measuring authoritarianism, conservatism,
and propensity to arrest, Dalley (1 97 5) found that university educated senior ofﬁcers

were signiﬁcantly less authoritarian, more liberal, and more ﬂexible than senior ofﬁcers

without such an education.10 He did not ﬁnd a difference between recruits with and
without a college education, leading him to conclude that a college education may hinder
the formation of stereotypical authoritarian and conservative attitudes as they gain
experience. Researchers performing analyses similar to Dalley (1975) have found that
obtaining a college education may ease ofﬁcer’s authoritarian attitudes (Smith, Locke, &
Fenester, 1970), diminish punativeness (Carlan & Byxbe, 2000), correlate with liberal or
more open belief systems (Guller, 1972; Roberg, 1978), and foster greater emphasis
placed on ethical conduct (Shemock, 1992).

These generally desirable ﬁndings are conﬂicted by contradictory ﬁndings in
other inquiries. Using data from the Project on Policing Neighborhoods, Paoline and
colleages (2000) found that college educated ofﬁcers were signiﬁcantly more likely to

distrust citizens than non-college educated ofﬁcers. Dantzker (1993) found that after

 

IO Dalley (1975) created a “Role Interpretation Scale” in which the ofﬁcer was asked to rank various
criminal offenses in order of severity and discuss whether he/she would pursue formal arrest action.

17

several years of service, college educated patrol ofﬁcers became more dissatisﬁed with
their work than patrol ofﬁcers with a high school diploma. Several previous studies found
that education has either very weak or no impact on ofﬁcer attitudes (Buckley, McGinnis,
& Petrunik, 1992; Miller & Fry, 1976a, 1978; Weiner, 1976). Further, Worden (1990)
found education to have a weak to modest effect on a variety of perceptions ofﬁcers hold
towards their role, however, he warned against overemphasizing education’s effect on
attitudes by selectively underscoring a small number of statistically signiﬁcant ﬁndings.
Reviewing the literature indicates that the relationship between ofﬁcer education
and attitudes remains unresolved. Nearly all ﬁndings regarding the relationship between
education and attitudes face the criticism that these studies assume a strong relationship
between attitudes and subsequent performance when there is little evidence supporting
such an assumption (Frank & Brand], 1991; Worden, 1989). Until a strong case can be
made that changing ofﬁcer attitudes through education is an effective means of
improving performance, research efforts should focus on measuring education’s impacts
on ofﬁcer performance directly. The remainder of the literature review will focus on
empirical works which have examined the direct impact of education on ofﬁcer
performance, as measured by several outcomes of police-suspect encounters (e.g., arrests,
searches, and force).
The Eﬁ'ect of Oﬂicer Education on Arrests

Previous examinations of the relationship between ofﬁcer education and the

propensity to arrest have produced inconclusive results.11 Researchers have posited that

 

II See Table 3 in the Appendix for additional information on each study’s samples, data, and direction of
ﬁndings.

18

 

education affects arrests in different ways. For example, Glasgow, Green, and Knowles
(1973) hypothesize that ofﬁcers with lower levels of education will make more arrests
because of an assumed theoretical tie between higher education and dissatisfaction with
police work. That is, a lesser educated ofﬁcer will be more satisﬁed and thus produce
arrests at a higher level than an ofﬁcer with more education, and thus, an inherent
dissatisfaction with his/her work. Contrary to their hypothesis, their sample of 24 ofﬁcers
from the Costa Mesa Police Department indicated that higher levels of education were
associated with higher rates of arrest.

Bozza (1973), using the same sample of 24 ofﬁcers as Glasgow and colleagues
(1973), chose to alter the hypothesis to posit that young ofﬁcers with high levels of
education would be eager to prove themselves and thus arrest at higher levels than older
ofﬁcers with lower levels of education. Bozza found support for his hypothesis in that the

young ofﬁcers with more education made more arrests than the older ofﬁcers with less

education.12

The only studies mentioned thus far have found that higher levels of ofﬁcer
education were associated with higher rates of arrest. Fickenauer (1975) found the
opposite. In his sample of 98 police ofﬁcers he noted that ofﬁcers with college education
were signiﬁcantly less likely to invoke the arrest process than ofﬁcers without a college
education. Smith and Klein (1983) constituted the only other study that was found which

indicated that better educated ofﬁcers made fewer arrests. In this study the authors note

 

I2 Bozza (1973) did not attenrpt to isolate the effect of education, instead comparing groups of young,
higher educated ofﬁcers to older, lesser educated ofﬁcers. He concedes that the difference between the
groups could be attributed to more experienced ofﬁcers learning methods of solving problems without
deferring to the arrest process.

19

that this was only the case when considering the education level of the department as an
aggregate measure. Using data from the Police Services Study (PSS), the authors found

that ofﬁcers in departments with a higher average level of education made fewer arrests

than ofﬁcers in departments with a lower average level of education.13 Smith and Klein
(1983) believed that the ﬁnding may indicate an impact of education on the informal
police culture at the department level. Their hypothesis followed that departments with
higher average levels of education would foster informal rules viewing arrest as a less
desirable means of achieving an end. Ofﬁcers that frequently used arrest would then not
ﬁnd much support for their behavior among their fellow ofﬁcers.

Smith and Klein (1983) also examined models which used a measure for the
individual ofﬁcer’s education level, but found that it did not produce an effect on the

ofﬁcer’s propensity to arrest. The effect of education appeared to be completely

controlled for by the victim’s request to arrest or not arrest and the suspect’s demeanor.l4

Similarly, Worden (1989), Brandl and colleagues (2001), and Smith and Aamodt (1997)
found no direct relationship between individual ofﬁcer education and the propensity to
arrest. Interestingly, Worden (1989) found that no ofﬁcer background characteristics ——

age, race, gender, or experience —- explained his/her propensity to arrest.

 

I3 Examining education’s effect on the aggregate level is fairly rare in the literature, as most of the
previous studies are at the individual level. Fyfe (198 8) points out that ﬁndings at the aggregate level may
be conrplicated by the fact that non-patrol ofﬁcers are included in the calculation of the department’s
education level. Thus personnel in technical capacities (i.e. roles that more explicitly demand a college
degree) my unduly inﬂuence the results of aggregate analyses.

I4 In this case, “no relationship” meant that Smith and Klein’s (1983) regression analysis found that the
partial regression coefﬁcient for individual ofﬁcer education level on arrests was literally 0.000 in all

models.

20

In sum, the literature pertaining to the association between ofﬁcer education and
arrests does little to provide clarity and the relationship remains unresolved. The studies
in this area are characterized by poor methods (Hudzik, 1978; Skogan & Frydl, 2004),
competing theoretical frameworks, and contradictory ﬁndings. The small number of
studies is surprising given the important part that arrest plays in the central role of the
police (Bittner, 1970). Studies focusing on the effect of education on arrests have been
outnumbered by studies examining education’s effect on ofﬁcer attitudes and other
general measures of performance such as injuries (Cascio, 1977), supervisor evaluations
(Finnegan, 1976; Roberg, 1978; Truxillo, Bennet, & Collins, 1998), and citizen
complaints (Cascio, 1977; Cohen & Chaiken, 1972; Kappeler, Sapp, & Carter, 1992;.
Lersch & Kunzman, 2001; Wilson, 1999). The difference could be due in part to
ambiguity regarding what is a desirable outcome regarding arrest. While few would argue
that lowering citizen complaints is a desirable outcome, the same consensus may not be
reached regarding lower rates of arrest, depending on the goals of the department. While
some may see higher education for recruits as a path to increased productivity in arrests,
others may see it as a path to improved community relations. In an era where police
agencies may be attempting to improve rapport between ofﬁcers and their communities,
increasing rates of arrest may not be a desirable mechanism.

The Eﬂ'ect of Oﬁicer Education on Searches

The relationship between ofﬁcer education and his/her propensity to search

suspects has received no scholarly attention. Of the two major reviews of the causes of

police behavior (Riksheim & Chermak, 1993; Sherman, 1980) that review the effect of

21

education on a variety of ofﬁcer behaviors, neither examines the effect of education on
searches. Despite the dearth of scholarly attention there has been no lack of research
examining factors associated with police searches, especially in trafﬁc stop encounters.
There have been a number of studies concemed with the effect of citizen or suspect
characteristics on ofﬁcer decision-making in this area (see Schafer, Carter, & Katz-
Bannister, 2004 for a review). However, only three works are identiﬁed which include
characteristics of ofﬁcers in their analyses ODecker & Rojeck, 2002; Paoline & Terrill,
2005; Smith & Petrocelli, 2001), none of which include ofﬁcer education as an
independent variable.
The Effect of Oﬁicer Education on the Use of Force

Though the relationship between ofﬁcer education and the use of force has

received more attention than arrests and searches, it would be an overstatement to call

this body of research large.15 These studies incorporate different measures of force, some

focusing on particular types of force such as discharging weapons, while others examine

any show of force. I6 As with previous studies examining arrests, the literature on the

relationship between ofﬁcer education and use of force has produced mixed results.
One beneﬁt with the force literature is that there is general uniformity in the

hypothesized effects. In a seminal piece, Muir (197 7) posited that the mark of a good

police ofﬁcer was his/her ability to use coercive force skillfully (i.e., avoiding

 

I5 See Table 4 in the Appendix for additional information on each study’s samples, data, and direction of
frndings.

I6 This body of literature exempliﬁes the wide range of deﬁnitions which are considered to be incidences

of force. These deﬁnitions include differentiating between levels of force such as verbal commands,
physical contact (Paoline & Terrill, 2007) and deadly force (McElvain & Kposowa, 2008), excessive and
reasonable force, and excessive and unnecessary force (Worden, 1996).

22

unnecessary or unreasonable force while still being comfortable using reasonable force to
obtain compliance). He hypothesized that good communication skills played a key role in
the skillful use of force, and a college education made an ofﬁcer a superior interpersonal

communicator. Much research in this area follows this line of thought and this notion has

received strong empirical support.17 The result is that most of the literature on ofﬁcer
education and force employ a hypothesis which states that college educated ofﬁcers will
use force less ﬁ'equently than non-college educated ofﬁcers.

Theoretically, research hypotheses regarding ofﬁcer education and force do not
have to be in complete agreement. As noted, most authors posit that college educated
ofﬁcers will be less likely to use force as their exposure to a college environment instilled
in them self control and a greater ability to reﬂect on the consequences of their actions
(Sherman & Blumberg, 1981). Yet, it is plausible to posit that college-educated ofﬁcers

are more likely to use force because college creates a greater drive to exceed expectations

while following the mission of the department (Bozza, 1973).18 There is also evidence to
support a third hypothesis, that college-educated ofﬁcers will not use force differently
from non-college educated ofﬁcers. Sherman (1978) draws comparisons between the
training recruits receive in academics and the practitioner-oriented classes many ofﬁcers
take at community colleges. He states that because the ofﬁcers typically attend colleges

that are so similar in curriculum to what is received in a training academy, there will be

 

17 In a meta-analysis of research on the impact of college on students from the 19703 and 19803, Pascarella
& Terenzini (2005) found strong freshman-to-senior gains in verbal skills and speaking skills. The
strongest gains, however, were in critical thinking, reﬂective judgment, and conceptual complexity.

18 Though Bozza (1973) is primarily concerned with arrests, it is not difﬁcult to apply the same concept to
police use of force.

23

no appreciable difference in the use of force between college-educated and non-college-
educated ofﬁcers.

The majority of studies in this area have found support for the hypothesis that
college-educated ofﬁcers use force less often than ofﬁcers with a high school diploma.
Several studies in this area have directly observed ofﬁcers using force. Using
observational data from the Project on Policing Neighborhoods (POPN), Terrill and
Mastrofski (2002) found that ofﬁcers with more education were signiﬁcantly less likely
to be involved in suspect encounters where force was used. Paoline and Terrill (2007),
while examining the effect of education on escalating types of force, learned that simple
exposure to college signiﬁcantly reduced incidences of verbal force, but not physical
force. They found that in order to see a signiﬁcant reduction in physical force, ofﬁcers
needed to have obtained a four-year degree. Also using POPN data, Paoline and Terrill
(2004) examined gender differences in the use of force in police-suspect encounters.
They found that college educated male ofﬁcers were less likely to use verbal force, but
not physical force. Education did not appear to have an appreciable impact on the use of
verbal or physical force by female police ofﬁcers.

Yet more studies have examined education’s impact on an ofﬁcer’s propensity to
discharge their weapon. Using data ﬁ‘om 186 ofﬁcer-involved shootings in Southern
California, McElvain and Kposowa (2008), found that ofﬁcers with a college degree were
more than 41 percent less likely to discharge their ﬁrearms than ofﬁcers with a high
school diploma or some college but no degree. A similar ﬁnding was produced by Binder

and colleagues (1982) in a report to the National Institute of Justice two decades earlier.

24

In a meta-analysis of studies examining the effect of education on ofﬁcer performance,
Aamodt (2004) found that better-educated ofﬁcers use force less often, though he does
not include a discussion of how force was measured.

Several studies in this group found that ofﬁcers with college backgrounds were
subject to fewer complaints of force. Cohen and Chaiken (1972) discovered that ofﬁcers
entering the New York City Police Department with some college experience received
fewer citizen complaints of force than ofﬁcers without such experience. With a sample of
940 ofﬁcers in Dade County, Florida, Cascio (1977) found that white, college-educated

ofﬁcers submitted fewer use of force reports and were subject to fewer allegations of

physical force.19

A small number of studies found that college-educated ofﬁcers were more likely
to use force as opposed to non-college educated ofﬁcers. A report submitted by Milton
and colleagues (1977) to the Police Foundation stated that ofﬁcers involved in violent
incidents typically held more years of education than the average of the department in
which they served. Using PSS data, Worden ( 1996) found that college-educated ofﬁcers
were signiﬁcantly more likely to use reasonable force in suspect encounters, however, the
same was not found for incidents involving excessive force.

The ﬁnal group of studies presented found that college-educated ofﬁcers and non-
college educated ofﬁcers do not behave differently in regards to their use of force.
Analyzing seven years of shooting data from the Kansas City, Missouri Police
Department, Sherman and Blumberg (1981) found that ofﬁcer education appeared to

have no signiﬁcant effect on discharging ﬁrearms when controlling for assignment, age,

 

I9 Cascio (1977) concedes that he did not control for age or length of service in reporting his ﬁndings.

25

and length of service.20 Four years earlier, Inn and Wheeler (1977) produced similar
results, ﬁnding that college education did not cause signiﬁcant differences in shooting
incidents among ofﬁcers. Hayden (1981) found that individual ofﬁcer characteristics,
including education, did not predict the decision to use deadly force.

Sherman and Blumberg (1981) point out, “depending on where and how police
use of force is measured. . .more educated police ofﬁcers appear to use force less often,
more often, or just as often as less educated ofﬁcers” (p. 318). Since the early 19803,
ﬁndings have gravitated towards college-educated ofﬁcers using force less often than
their less educated counterparts. Nonetheless, the relationship between ofﬁcer education
and the use of force remains unresolved. As F yfe (1988) remarks, results in the 19703 and
19803 may be suspect because ofﬁcers with college educations were typically found
operating computers or technical equipment within the police department, where
substantially fewer incidences of force occur. Future research could determine if ﬁndings
are indeed trending towards a negative relationship between education and force.
Previously Employed Theoretical Frameworks

As noted, much of the literature regarding the impact of education on ofﬁcer
behavior does little to provide an explanation as to why education should have any
impact on ofﬁcer behavior at all. Most studies do not question the mechanism by which
college education results in differences in behavior between college educated and non-
college educated ofﬁcers. Instead, authors simply note correlations between education

and particular attitudes or behaviors without entertaining a discussion of why education

 

20 Sherman and Blumberg (1981) concede that their ﬁndings may be the result of a lack of older, more
experienced, college-educated ofﬁcers in their sample.

26

should theoretically be expected to have any impact on those attitudes or behaviors.21
Exceptions to this norm have brieﬂy discussed possible theories for education’s effect on

ofﬁcer behavior (See Carter, Sapp, & Stephens, 1988), but few have followed through

and tested those hypotheses.22

Roberg (1978) posits that college education impacts ofﬁcer performance by
shaping the ofﬁcer’s belief system. He believes that college educated ofﬁcers will hold
more open belief systems and thus have a greater appreciation for psychological and
sociological intricacies in their communities. Roberg’s results support this hypothesis.23
Wilson (1975) entertains several hypotheses. Wilson muses that college educated ofﬁcers
would be expected to be less likely to use their ﬁrearms because college instills a more
liberal, reﬂective outlook. However, similar to Bozza (1973), Wilson (1975) also
considers that college may indoctrinate ofﬁcers with a stronger sense of duty, resulting in
more vigorous application of authority. As noted earlier, some authors (Sherman, 1978;
Sherman and McLeod, 1979) have posited that college education should not be expected
to have any affect on performance because the majority of ofﬁcers attend courses which

are quite similar to academy training. Miller and Fry (1976a) state that even if the ofﬁcers

 

21 For exanrple, Cascio (1977) examines correlations between ofﬁcer education and a variety of outcomes.
In an attempt the address the criticism that the correlations may not be due to college education itself, but
the possibility that college selects ofﬁcers which are more motivated and intelligent, he performs a path
analysis to control for the effects of motivation and intelligence. Upon ﬁnding that college education has a
unique direct effect on ofﬁcer performance, Cascio concludes that education should be considered as a
selection criteria for police ofﬁcers, but does not offer a discussion of how education could produce such an
effect independent of intelligence or motivation.

22 Worden (1990) provides an excellent attempt to examine a variety of hypothesized relationships
between ofﬁcer education and attitudes and performance, but produced inconclusive results.

23 Roberg (1978) measured performance through supervisor ratings.

27

partake in a liberal arts education, the content has so little to do with actual police work,
that one should still not expect college educated ofﬁcers to behave differently than
ofﬁcers without such an education. Bittner (1970) comes to a similar conclusion:

In particular, making the college degree a requirement for admission to police
work should not be misunderstood: four years of a liberal arts education of any
kind will not prepare a young man for police work. And it would be absolutely
pernicious to encourage the belief, either in the minds of the new recruits or of
existing personnel, that a BA. in sociology or psychology equips a person to do
peace keeping or crime control (p. 86).

Nevertheless, Bittner (1970) goes on to recommend requiring a college degree of new
recruits in order to help the status of the police occupation as a whole.
Of the studies examining the impact of higher education on the police, virtually

none draw on the literature ﬁ'om the ﬁeld of education concerning the impact of college

on students.24 Indeed, it would seem that the previous literature does little to make a
compelling case why one would expect college-educated police ofﬁcers behave
differently than their non-college educated colleagues.

Previous research in the ﬁeld of education has done much to examine the impact
that education has on status attainment for students, using outcomes such as income and
occupational prestige (Knox, Lindsay, & Kolb, 1993; Meyer, 1977). More relevant to the
current inquiry, within the education literature there is evidence that higher education
affects more than a student’s prospects for status attainment (Pascarella & Terenzini,

1991, 2005). Unfortunately for behavioral researchers, the focus of empirical research in

 

24 The exceptions being Jacobs and Magdovitz (1977), Sherman and Blumberg (1981), and Weiner (1976).

28

education has centered on higher education’s impact on the attitudes of students.
Supportive of ﬁndings in the area of policing by Smith, Locke, and Fenester (1970),
Dalley (1975), and Roberg (1978), educational research indicates that college seniors (as
opposed to college freshmen) hold beliefs that are less authoritarian (Altemeyer, 1988),
dogmatic, and more liberal (Feldman & Newcomb, 197 3; Knox, Lindsay, & Kolb, 1993).

In Feldman and Newcomb’s (1973) The Impact of College on Students, the
authors mention little regarding education’s impact on student behavior, especially
behavior related to policing. Their analyses indicate that over the course of four-years
living within a university setting, college seniors express more individualism and
readiness to express impulses. Potentially related to ofﬁcer interactions with citizens,
college upperclassmen were more conﬁdent, assertive, and autonomous. These ﬁndings
were repeatedly replicated in research between 1970 and 2000 (Pascarella & Terenzini,
2005)
Proposed Theoretical Framework

Despite the dearth of ﬁndings regarding education’s impact on behaviors directly
linked to policing, previous studies in this area are not completely without use. In
reviewing literature ﬁ'om education and policing, a common theme emerges regarding
education’s impact on an individual’s democratic values. Indeed, previous research has
shown education to have a positive effect on values regarding humanitarianism, civil
rights, social conscience, and political interest (Knox, Lindsay, & Kolb; Muir, 1977;
Pascarella & Terenzini, 1991, 2005; Wilson, 1999). Historically, education has been

viewed as playing a key role in the average American’s understanding and participation

29

in democratic society. Thomas Jefferson was quoted as saying:

I know of no safer depository of the ultimate powers of the society but the people

themselves. And if we think them not enlightened enough to exercise their

control with a wholesome discretion, the remedy is not to take it from them, but

to inform their discretion by education (Padover, 1939: 89-90).

The notion that higher education has an impact on the democratic values of police
ofﬁcers was visited in a classic piece by Muir (1977). Muir posited that in order for one
to become a good police ofﬁcer, he or she had to grasp the nature of human suffering -
forming what he termed to be the tragic perspective of humanity. By this, he meant that

an ofﬁcer had to be able to empathize with the individuals which he/she encountered on

the street.25 Muir hints that a college education may be a pathway to developing this
perspective on humanity.

Relating this notion to the current inquiry, a police ofﬁcer in modern American
society holds great power to overcome human obstacles, and exercises much discretion in
wielding that power (Bittner, 1970). Education may be hypothesized then to instill in
police ofﬁcers a certain appreciation for the discretion they hold through enhanced
democratic values and the development of Muir’s (1977) tragic perspective. College-
educated ofﬁcers would be hypothesized to hold a greater understanding of not only the
democratic rights of the individuals they encounter, but also their own ability to
circumvent or remove those rights through formal action (i.e., arresting, searching, or

using force against the suspect). Through the experience of higher-education, college-

 

25 Muir (1977) contrasts the tragic perspective with the cynical perspective, through which the ofﬁcer sees
themselves and the citizens they encounter as belonging to two entirely different types of people. Muir
found that cynical ofﬁcers use varying forms of an “us versus them” dichotomy to separate themselves
from citizens.

30

educated ofﬁcers would be expected to pursue these formal actions more cautiously, and
thus would be less likely to use them than their non-college educated counterparts.
The Gap in the Literature
Upon examining the available literature, the National Acaderrrics Panel on Police

Policy and Performance stated that there was insufﬁcient evidence to conclude that
higher education has a desirable effect on police performance (Skogan & F rydl, 2004).
The panel commented that the literature is characterized by inconsistent ﬁndings and
generally poor methodologies (Hudzik, 1978; Sherman, 1980). The Panel did, however,
suggest “rigorous research on the effects of higher education on [ofﬁcer] job
performance” (Skogan & F rydl, 2004: 141). In this area of the literature there appears to
be a need for ﬁuther studies employing valid, reliable methodologies. Another important
area where the current state of knowledge is lacking was best highlighted by Worden
(1990). Worden points out that the majority of the previous literature examines a single
outcome, and thus the results cannot be interpreted cumulatively. Though Worden was
speciﬁcally referring to studies of the effect of education on ofﬁcer attitudes, the same
could be said of studies regarding speciﬁc police behaviors.
The Current Inquiry

The current research will attempt to bridge the gap in the literature by examining the
effect of education on three speciﬁc measures of ofﬁcer behavior (i.e. arrest, search, and
force). In addition, if there is a signiﬁcant effect present in more than one of the outcomes
a comparison will be made to determine where the effect of education is the strongest.

More directly, using data ﬁ'om the Project on Policing Neighborhoods, the present

31

analysis will proceed with two research questions:
1. What is the effect of an ofﬁcer’s education level on his/her propensity to arrest,
search, and use force?
2. If there are signiﬁcant effects in more than one outcome, where is the inﬂuence of
education the strongest?26
The current analysis is open to the possibility that in regards to the ﬁrst research
question, no signiﬁcant relationship, or a single signiﬁcant relationship will be detected.

If that is the case then the second research question cannot be pursued.

 

6 As the analysis Will use brnary and multmorrnal logrstrc regressron, odds wrll be compared to make this
determination.

32

CHAPTER IV: METHODOLOGY

The data used for the present inquiry will come from the Project on Policing
Neighborhoods (POPN). On the whole, the POPN data were collected to allow
researchers to describe how policing occurs on a day-to-day basis. The data were
collected in the cities of Indianapolis, Indiana in the summer of 1996, and St. Petersburg,
Florida in the summer of 1997 (Paoline et al., 2000). Of particular interest to the current
inquiry were the data depicting the systematic observation of ofﬁcers in encounters with
citizens and interviews with those same ofﬁcers to gather background information (e. g.,
education level).
Study Sites

Indianapolis and St. Petersburg were selected for POPN based on several criteria.
Both cities exhibited diverse social, economic, and demographic characteristics, were
receptive to a hosting large-scale research projects for approximately one year each, and
had implemented community policing initiatives (Paoline et al., 2000). Each city showed
certain similarities and differences which were relevant to studying police behavior.
While both cities experienced similar crime rates, St. Petersburg held a smaller
population and employed fewer ﬁrll-time sworn patrol ofﬁcers, resulting in a higher
workload than the ofﬁcers in Indianapolis. Indianapolis did, however, show greater signs
of economic distress whereas it outpaced St. Petersburg in relation to percentage of the
population that was minority, percent unemployed, percent of residents living in high
poverty areas, and percent of female-headed households (Terrill, 2001).

The police forces of each city were different in regards to several measures. As

33

noted, the police force in Indianapolis was larger than St. Petersburg, with 416 and 246
sworn full-time patrol ofﬁcers, respectively. While both cities had implemented
community policing initiatives, they differed in size and strategy. St. Petersburg had
begun their initiative two years prior to Indianapolis, and committed 60 ofﬁcers to
community policing (23% of the police force) as opposed to 25 for community policing
in Indianapolis (6%). St. Petersburg also chose to deploy their community policing
ofﬁcers geographically, instilling each ofﬁcer with a sense of responsibility for the
problems residents experience in the geographic area which they served (Terrill, 2001).
Indianapolis used a “get tough” strategy for community policing emphasizing directed
and aggressive proactive patrols, but also some of the elements more commonly
associated with community policing such as personal interaction (Terrill and Mastrofski,
2002)

Of particular interest to this study, the departments in each city were not equal in
traditional measures of professionalism (Shemock, 1992). Indianapolis required more
hours of training for recruits (1,392) than St. Petersburg (1,280). More than one-third
(36%) of Indianapolis patrol ofﬁcers held a four-year college degree (Terrill, 2001). This
1996 ﬁgure indicates that the patrol ofﬁcers in Indianapolis were more educated than the
general populace of the city, of which 26 percent held a four-year college degree in 2000
(U .8. Census Bureau). In St. Petersburg, a quarter (26%) of patrol ofﬁcers held a four-
year degree (Tenill, 2001). Comparatively, 23 percent of St. Petersburg residents held a

four-year degree in 2000 (U .8. Census Bureau).

34

Data

As noted, the current inquiry utilizes two aspects of the POPN data: systematic
observation of ofﬁcers in the ﬁeld, and in-person interviews with those ofﬁcers. The
observational data account for the most important part of the current study, detailing a
direct and disinterested account of what occurs on ofﬁcer patrols (Mastrofski et al.,
1995). The interviews provide data on ofﬁcer backgron characteristics of interest (e.g.,
education level). ’

Observational data. The observation data were collected through a technique
called systematic social observation (SSO) (Mastrofski et al.,1998) in the manner of
observer-participant. As an observer-participant, the researcher makes no effort to keep a
distance from the subject they are observing (in this case, police on patrol), but also make
no effort to participate in the interactions between the observed ofﬁcers and citizens
(Babbie, 1995). Prior to beginning observations in the ﬁeld the observers (students from
Michigan State University and the State University of New York at Albany) took a
semester-long course in $80 and participated in training rides with ofﬁcers at local
police departments (Terrill & Mastrofski, 2002).

Regarding the physical process of data collection, observers accompanied ofﬁcers
on patrol throughout a matched sample of shifts in sampled beats for each city. The
observers were instructed to take notes on activities and encounters the ofﬁcer made with
citizens, detailing the persons involved and how the encounter transpired. An encounter
was deﬁned as “a face-to-face communication that took place between ofﬁcers and

citizens that took over one rrrinute, involved more than three verbal exchanges between

35

ofﬁcer and citizen, or involved signiﬁcant physical contact between the ofﬁcer and
citizen” (Terrill, 2001: 50). Following each day of ﬁeld observation, the observers would
transcribe their notes according to POPN protocol. Ofﬁcers received assurances of
conﬁdentiality and were allowed to read the notes of the researcher for which they were
assigned, but not the notes of researchers assigned to other ofﬁcers (Parks et al., 1999).

A sample of beats was selected for each city for the observations to take place. In
Indianapolis, 12 of the 50 total beats were selected, and similarly 12 of the 48 beats in St.
Petersburg were selected. Beats were selected based on the expectation of higher
frequencies of ofﬁcer and citizen encounters as POPN research directors wanted the
sample data to include a large number of encounters. Selection criteria for beats were
based on an index of socioeconomic conditions in neighborhoods. These conditions
included: percent of families with children that are headed by a female parent, percent of
adults employed, and percent of the population living below 50 percent of the poverty
level (Parks et al., 1999). Being the more distressed of the two cities, the socioeconomic
index scores for beats in Indianapolis ranged ﬁom 4 to 76, with the median being 36. In
St. Petersburg the index values ranged from 4 to 103, with a median of 15. Beats were
then selected from the 2nd, 3rd, and 4th quartiles of index scores in Indianapolis. St.
Petersburg beats were then selected to match those of Indianapolis (Terrill, 2001).

The ride sampling procedure called for the sample to include observations over
every shift in each beat, include both general duty and community policing ofﬁcers, and
include shifts on both slow and busy days. Given that the project directors wanted a large

sample of encounters, busy days (Thursday through Saturday) were oversampled.

36

After all ﬁeld observations had taken place, observers in Indianapolis had ridden-
along with 194 different patrol ofﬁcers and 48 supervisors over the course of 2,800 hours
of observation and witnessed 6,485 encounters with citizens. Observers in St. Petersburg
had ridden-along with 128 different patrol ofﬁcers as well as 37 supervisors over the
course of 2,900 hours, witnessing 5,500 encounters with citizens (Paoline et al., 2000).

Interview data. In—person interviews were conducted with the ofﬁcers with whom
the observers were riding. These interviews lasted approximately 25 rrrinutes each and
were conducted by personnel hired and trained for the interview data collection only.
Questions on the surveys pertained to ofﬁcer background characteristics, most
importantly in the case of the present study - education, and the ofﬁcer’s perceptions and
experiences of the beats on which they patrolled. The response rates were 95 percent and
97 percent in Indianapolis and St. Petersburg, respectively (Terrill, 2001).

Variable Descriptions

Dependent variables. The present analysis employs three dependent outcome
measures (see Table 4.1). All three will be measured dichotomously, but an additional
force variable will be a trichotomous measure in order to differentiate the effect of ofﬁcer
education on verbal force versus physical force. Because the outcomes are measured in
such a way, the analysis will rely on stepwise logistic regression, and multinomial
logistic regression for the verbal/physical force measure.

The goal of the analysis will be to determine how the primary independent and
control variables explain variation in the following outcomes. The ﬁrst, arrest, is deﬁned

as taking a person into custody for the purpose of charging him or her with a criminal

37

offense. The second, search, is deﬁned as a search of the suspect, suspect’s vehicle, or

the area immediately surrounding the suspect which goes beyond plain view. The third,

force, is deﬁned as acts that threaten, or actually inﬂict physical harm on citizens.27 An
additional dependent variable for the use of force will distinguish between verbal and
physical force.

Independent variable. The primary independent variable in the analysis is the
education level of the ofﬁcer. The ofﬁcer’s education will be captured using two
dichotomous measures, one for some college exposure but no baccalaureate degree, and
one for a four-year degree. The reference category will consist of encounters involving
ofﬁcers with a high school diploma or less. As Worden (1990) points out, measuring
education in this way (as opposed to by years of formal education or credits accumulated)
is desirable in that it has the potential to capture the nature of the curriculum that the
ofﬁcer was exposed to (e. g., a four-year institution versus some college experience versus
no college experience). As a downside, the data do not allow researchers to capture
indications of the quality of the education the ofﬁcer received. Higher levels of education
are hypothesized to reduce incidences of the three outcome measures.

Control variables. To maintain an appropriately speciﬁed model the analysis will
also include variables which have exhibited theoretical relevance and/or have been found
to be signiﬁcant in previous research (Paoline & Terrill, 2007; Terrill & Mastrofski,

2002). As noted, there are three groupings of control variables. The ﬁrst group captures

 

27 There are certain limitations to be aware of when measuring these outcomes as dichotomous variables.
In regards to searches, variation in discretionary and non-discretionary searches will be lost. Measuring the
use of force in such a way ignores the severity of force (e.g. painful restraint techniques versus irrrpact
methods) (Terrill & Mastrofski, 2002). However, if the outcome variables are not all measured in such a
way, their comparability would subsequently be reduced in the analysis.

38

characteristics of the ofﬁcer. In addition to the ofﬁcer’s education, control variables in
this group include a measure of the ofﬁcer’s experience. Higher levels of ofﬁcer
experience are hypothesized to reduce the likelihood of arrests, searches, and force. The
other control variables at the ofﬁcer level are age in years, and dichotomous measures for
gender, and race. Older ofﬁcers are hypothesized to be less likely to use searches, arrests,
and force, while male ofﬁcers are hypothesized to be more likely to engage in those
behaviors. Previous research has shown inconsistent ﬁndings regarding the effect of
ofﬁcer race in police-suspect encounters (see Terrill, 2001).

The second grouping of control variables describes characteristics of the suspect
in the encounters. Previous reviews of research (Riksheim & Chermak, 1993; Sherman,
1980) have shown situational factors, such as suspect characteristics, to be inﬂuential on
police behavior. These variables include demographic measures such as age (categorical),
race, gender, and socioeconomic status. It is important to note that as opposed to ofﬁcer
characteristics, which were available from interview data, the suspect characteristics were
based on the researcher’s observations. This grouping also includes information about the
suspect’s behavior in the encounter, including whether they showed resistance to the
ofﬁcer, if the suspect was in a conﬂict with another citizen during the encounter, if the

suspect was carrying a weapon, his/her demeanor, if the suspect was using alcohol, and

an index of evidence of a violation of the law.28 The hypothesized effect for each of these

variables is displayed in Table 1. With the exception of suspect socioeconomic status,

 

28 This evidence scale is a summative index of evidence implicating the suspect as a violator of the law. It
is weighted to appropriately capture the relative importance of some pieces of evidence over others. The
value increased by (3) if the ofﬁcer observed the suspect engage in illegal behavior, (2) if the ofﬁcer heard
the suspect confess, (1) for physical evidence on the scene, (1) for circumstantial evidence, and (l) for
hearsay evidence from citizens on the scene, for a maximum possible score of 8.

39

which is hypothesized to decrease the likelihood of the outcome measures at higher
levels, each variable in this group is hypothesized to increase the likelihood of arrests,
searches, and force.

The third grouping of variables describes characteristics of the setting of the
encounter. The group includes measures of the number of ofﬁcers present on the scene,
the number of citizens present, whether the ofﬁcer initiated the encounter with the
suspect, and whether the encounter took place in Indianapolis or St. Petersburg. While
both the number of ofﬁcers and the number of citizens on the scene may have an impact
on the observed ofﬁcer’s behavior, the direction of the impact may be contingent on a
number of factors such as whether the ofﬁcer feels in control of the situation, or if the
additional citizens are non-participating bystanders or are attempting to support the
suspect in the encounter. Because of these reasons, the hypothesized effect of both these
variables is unclear. As for the remaining variables, ofﬁcers that are proactive in
initiating the encounter are hypothesized to be more likely to arrest, search, and use force.
Because of the less-aggressive, problem-solving approach of the St. Petersburg police
department, their ofﬁcers are hypothesized to be less likely to engage in the measured
outcomes.

Data Analysis

As the dependent variables (e. g., arrest, search, and force) in the current analysis
are measured categorically, the analysis will utilize two forms of logistic regression.
Logistic regression is preferable when dependent variables are measured in such a way

(Liao, 1994). Once full models have been run for all three dependent variables, the odds

4O

ratios for ofﬁcer education on the dependent measures can be compared to determine
where education’s effect is the strongest. This aspect of the analysis is contingent on a

signiﬁcant effect for education in more than one of the outcome measures.

41

Table 4.1. Variable Descriptions and Hypothesized Effects

 

Variable Name
Dependent variables
Arrest

Search

Force
Force (verbal/physical
distinction)

Independent/Controls
Oﬂicer Characteristics
Some college, no degree
Four-year degree or higher

Experience
Male
Non-white

Suspect Characteristics
Age

Non-white
Male
Socioeconomic Status

Arrest“

Resistance

Conﬂict with other citizen
Weapon

Evidence
Demeanor
Drug or Alcohol

Encounter

Number of Officers
Number of Citizens
Proactive Encounter
Observation site

Hypothesized Effect

+

+/-

+/-
+
+

Variable Coding

1 = Suspect is arrested, 0 = Suspect is not arrested

1 = Suspect, vehicle, or area is searched, 0 = Not
searched

1 = Ofﬁcer used force, 0 = Ofﬁcer did not use force
1 = Ofﬁcer used verbal force against suspect, 2 =
Ofﬁcer used physical force against suspect, 0 = No
force was used

1 = Some college, no degree, 0 = All other
1 = Four-year degree or higher, 0 = All Other

Years of experience
1 = Male, 0 = Female
1 = Non-white, 0 = white

1=0t05years,2=6to12years,3=l3to17years,
4= 18 toZOyears,5=21 to29years,6=30to44
years, 7 = 45 to 59 years, 8 = 60 or more years

1 = Non-white, 0 = White

1 = Male, 0 = Female

1 = chronic poverty, 2 = low, 3 = middle, 4 = above
middle

1 = Suspect was arrested prior to use of force, 0 =
Suspect not arrested

1 = none, 2 = passive, 3 = verbal, 4 = defensive, 5 =
active

1 = none, 2 = calm verbal, 3 = agitated verbal, 4 =
threatened assault, 5 = assault

1 = Suspect has weapon, 0 = Suspect not visibly
armed

Summative scale (0 to 8)

1 = Suspect disrespectful to police, 0 = All other

1 = Suspects behavior indicates drug or alcohol
effects, 0 = All other

Number of ofﬁcers on the scene

Number of citizens observing the scene

1 = Ofﬁcer initiates encounter, 0 = All other
1 = Indianapolis, 0 = St. Petersburg

*This variable is only used in the models in which force is the dependent variable.

42

Validity, Reliability, and Generalizability

As noted, the most important pieces of these data are comprised of systematic
ﬁeld observations. Compared with other methods, ﬁeld research typically produces valid
data. It allows researchers to capture events or phenomena that prove more difﬁcult or
elusive to survey research or ofﬁcial records. Field research does introduce a threat to
validity with the presence of the observer. A commonly cited concem in observational

research is reactivity on behalf of the observed (Mastrofski & Parks, 1990). In these

instances, ofﬁcers would alter their behavior due to the presence of the observer.29 The
POPN protocol provided two built-in measures of reactivity. With the ﬁrst, ride level
reactivity, the observer made an assessment of whether the ofﬁcer reacted to their
presence over the course of an entire shift. The second, activities and encounters with
reactivity, was an indication of whether the observer felt the ofﬁcer altered their behavior
because of the observer’s presence in the context of speciﬁc events or encounters during
the ride (Spano, 2003). Observers noted that only 0.5 percent of all observations showed
any indication that the ofﬁcer altered their behavior due to the presence of the observer
(Terrill, 2001). More in-depth analysis of reactivity by Spano (2003) indicates that
beyond these built-in measures, signs of reactivity appear in ofﬁcer’s concerns for the
observer’s safety, especially if the observer was female. Project directors took several

steps in order to minimize reactivity of the patrol ofﬁcers to their observers. As noted

 

29 In a seminal piece taking place some 40 years prior to the POPN observations, Westley (1970) wrote of
reactivity concerns. He mentioned that ofﬁcers initially responded to his presence negatively, “this proved
to be both personally painful, in the sense thereafter he had to push himself on men who he felt disliked and
were afraid of him, and practically disastrous, since if the men refused to talk to him the research would
stop” (p. vii). The solution, he found was “to stay around for such long and continuous periods that it was
not possible for the men to keep up the pretense” (p. viii) and the ofﬁcers would eventually return to their
daily routines.

43

earlier, ofﬁcers were promised conﬁdentiality and allowed to read the notes taken by
their observers (Mastrofski et al., 1998).

The generalizability of the POPN observational data is somewhat limited due to
the beat selection process. Because beats were selected as clusters of economically
distressed areas, it may be difﬁcult to generalize ﬁndings to ofﬁcers working in the more
economically prosperous areas of Indianapolis and St. Petersburg. When attempting to
generalize ﬁndings to patrol ofﬁcers in other departments around the nation, the POPN
observational data has a particular advantage over previous studies of the same kind
(Paoline et al., 2000). Previous observational studies have focused on large cities with
substantially large patrol forces and bureaucratic structures, thus making them difﬁcult to
generalize to ofﬁcers in medium-sized cities or rural areas. Indianapolis and St.
Petersburg have the advantage of being more similar to more areas of the nation than the

larger cities that were subjects of previous inquiry (Terrill, 2001).

CHAPTER V: RESULTS

The following section contains the statistical analyses and ﬁndings for the current
inquiry. First,'descriptive statistics present the distribution of the dependent variables
(e.g., arrest, search, and force) and the three groupings of independent variables (e.g.,
ofﬁcer, suspect, and encounter). Bivariate distributions of the dependent measures by
ofﬁcer education level are then shown. The ﬁnal portion of this section presents and
describes the multivariate analyses, consisting of binary logistic regression for the
dichotomous dependent measures and multinomial logistic regression for the dependent
variable distinguishing between verbal and physical force.
Descriptive Statistics

Table 5.1 displays descriptive statistics for the three dependent variables in the
present analysis. As noted, the dependent variables are measured dichotomously, while
there is also an additional force variable measured in three categories. For the
dichotomous measures, their means may be interpreted as the proportion of cases which
fall in the measured category. As such, 15 percent of encounters involve the suspect
being arrested, 23 percent of encounters involve the suspect being searched by the
primary observed ofﬁcer, and 5 8 percent of the encounters involve the ofﬁcer using force
against a suspect. Distinguishing between verbal and physical force, verbal force is the

highest level of force used in 37 percent of the encounters, and physical force is the

highest level in the remaining 21 percent of encounters.30

 

30 The variable that distinguishes between verbal and physical force captures the highest level of force used

in the encounter. For example, in the 1,246 encounters coded as verbal force, the ofﬁcer did not use any
force higher than verbal commands or threats. In the 699 encounters coded as physical force, the ofﬁcer

45

Table 5.1: Descriptive Statistics for Dependent Variables (N=3,356)

 

Variable Range Mean Standard Deviation
Arrest“ 0 to l 0.15 0.36
Searchi 0 to 1 0.23 0.42
Force 0 to l 0.58 0.49
Force (verbal/physical distinction) 0 to 2 0.79 0.76

(0) No force, n = 1,411 (42%)
(l) Verbal force, n = 1,246 (37%)
(2) Physical Force, II = 699 (21%)

 

*Two missing cases (11 = 3,354)
IThree missing cases (11 = 3,353)

The next three tables show the univariate statistics for the ofﬁcer, suspect, and
encounter level variable groupings. Beginning with Table 5.2 (i.e., ofﬁcer level),
encounters involving ofﬁcers with a high school diploma or less comprise a minority of

the police-suspect encounters. Roughly 44 percent of the encounters involve ofﬁcers with

some exposure to college, but have not obtained a baccalaureate degree.“ The remaining

42 percent of encounters involve ofﬁcers having obtained a four-year degree or higher.32
Regarding the other ofﬁcer level variables, the average encounter involves an

ofﬁcer with approximately seven years experience. With a standard deviation of 5.9

years, however, ofﬁcer experience levels show substantial variation. The majority of

 

may also have used verbal force, but as physical force is considered a higher level of force than verbal
threats or commands, these cases are recorded as being physical force encounters.

3 This category also Includes encounters Involvrng ofﬁcers wrth 2-year assocrate’s degrees. Overall, this
subset comprised only 5 percent of encounters. As such, it was combined with the “some college, but no
degree” category.

32 The majority of encounters in this category involved ofﬁcers with a baccalaureate degree (n = 1,224,
87%). One hundred and seventy-eight (13%) encounters in this category involved ofﬁcers with some
exposure to graduate school, and only 8 (0.06%) encounters were with an ofﬁcer holding a graduate degree.

46

encounters involve male ofﬁcers (84.7%) and just one-ﬁfth (20.8%) of the encounters

. . 33
Involve a non-white ofﬁcer.

Table 5.2: Descriptive Statistics for Ofﬁcer Level Independent Variables (N = 3,356)

 

Variable Range Mean Standard Deviation
Education
Some college, no BA/BS 0 to l 0.44 0.50
Four-year degree or higher 0 to 1 0.42 0.49
Experience 1 to 32 7.73 5.97
Male 0 to 1 0.85 0.36
Non-white 0 to 1 0.21 0.41

 

Below, Table 5.3 displays descriptive statistics for the suspect level variables.
Regarding individual characteristics, the average suspect in the sample is between the
ages of 21 and 29 years old, non-white, male, and lower class. Regarding race in
particular, 63 percent of encounters involve a non-white suspect. As a point of contrast,

Census 2000 data indicated that the non-white populations of Indianapolis and St.

Petersburg were 31 percent and 29 percent, respectively (U .S. Census Bureau).34 The
decision to arrest, prior to using force against the suspect, occurred in 11 percent of the
encounters. Regarding suspect behavior on average, suspects show no resistance towards
ofﬁcers, are not in conﬂict with other citizens on the scene, are unarmed, and are not
disrespectful. Approximately one-ﬁfth (21%) of encounters involve suspects showing
signs of intoxication. With respect to the evidence measure, 50 percent of encounters

provide at least some evidence of law violation.

 

33 Appendix B contains a table showing the joint distribution between ofﬁcer education level and ofﬁcer

sex and race.

34 As a cautionary note, though these ﬁgures show that a disproportionate number of non-white suspects
were represented in police-citizen encounters, beats were selected on the basis of socioeconomic factors.
More speciﬁcally, areas in high socioeconomic distress were over-sampled in order to ensure a high
ﬁ'equency of encounters.

47

Table 5.3: Descriptive Statistics for Suspect Level Independent Variables (N = 3,356)

 

Variable Range Mean Standard Deviation
Age 1 to 8 5.24 1.35
Non-white 0 to 1 0.63 0.48
Male 0 to 1 0.72 0.49
Socioeconomic Status 1 to 4 2.36 0.56
Arrest“ 0 to 1 0.11 0.31
Resistance 1 to 5 1.21 0.66
Conﬂict with other Citizen 1 to 5 1.13 0.57
Weapon 0 to l 0.02 0.12
Evidence 0 to 7 1.32 1.70
Evidence prior to ArrestI O to 7 1'27 1'53
Demeanor 0 to 1 0.10 0.30
Drug or Alcohol 0 to l 0.21 0.41

 

*This variable is not used in the models in which arrest is the dependent variable.
IThis variable was specially coded to measure the amount of evidence available prior to making an arrest.
It is only used when arrest is the dependent variable.

Table 5.4 (i.e., encounter level) provides the descriptive statistics for the ﬁnal set

of control variables. The average encounter involves approximately two ofﬁcers and four
citizens, though there are several outlier encounters which included much larger numbers
of ofﬁcers and bystanders.35 Slightly less than half (45%) of the encounters are
proactively initiated by ofﬁcers, and just more than half (56%) of the encounters take
place in Indianapolis, with the remaining 44 percent occurring in St. Petersburg.

Table 5.4: Descriptive Statistics for Encounter Level Independent Variables (N = 3,356)

 

Variable Range Mean Standard Deviation
Number of Ofﬁcers 1 to 26 2.21 1.61
Number of Citizens 1 to 100 4.20 5.63
Proactive Encounter 0 to 1 0.45 0.50
Observation site 0 to l 0.56 0.50

 

 

35 These outliers represent a very small number of encounters in which there was a very large grouping of
either citizens or ofﬁcers. In the citizen’s case, there was an encounter at a party where the estimated
number of guests was around 100. The removal of the highest values from both these variables changes
their mean values very little (a change of 0.01 and 0.06, for ofﬁcers and citizens, respectively), and with no
theoretical justiﬁcation for removing these encounters from the sample they were left as is.

48

Bivariate Analysis

The bivariate portion of the analysis presents joint distribution tables and utilizes
the chi-square test of independence to examine the relationships between ofﬁcer
education and the three dependent variables (e.g., arrest, search, and force). The chi-
square test of independence tests the null hypothesis that ofﬁcer education and the
dependent measures are independently distributed (Bachman & Paternoster, 2004). If the
test yields a signiﬁcant value, the null hypothesis may be rejected and one could then
state that the two variables are in some way related. Included with each contingency table I
is the gamma (7) statistic, which is the preferred measure of association when using
ordinally measured variables. As a proportional reduction in error measure, gamma is an
indication of the degree to which error in predicting the dependent variable is reduced by
knowledge of the independent variable (Bachman & Paternoster, 2004).

Table 5.5 presents the joint distribution of ofﬁcer education level and arrests. The
table shows that these two variables are independently distributed (1 2 = 1.66, p = .437),

indicating that they are not related, and the gamma statistic demonstrates that ofﬁcer
education level is not a useful variable for reducing error in predicting arrests (y = .04).
Across all levels of education, the proportion of encounters involving arrest varies very
little, from 15 percent for some college exposure to 16 percent for a four-year degree. At
the simple bivariate level, these results support previous ﬁndings of a null relationship
between ofﬁcer education and arrests (Brandl et al., 2001; Smith & Aamodt, 1997; Smith
& Klein, 1983; Worden, 1989). Later, multivariate analysis will examine if the presence

of additional independent variables reveals a relationship between education and arrests.

49

Table 5 .5: Bivariate Distribution of Arrests by Ofﬁcer Education (N = 3,354)*

 

High school or less Some college, no BA/BS Four-year degree or higher

11 (%) n (%) n (%)
Suspect not arrested 412 (85%) 1,248 (86%) 1,180 (84%)
Suspect arrested 74 (15%) 211 (15%) 228 (16%)
Total 487 1,459 1,408

12: l.66,p= .437;y= .04

 

*Two missing cases
The joint distribution of ofﬁcer education and searches is displayed in Table 5.6.

As with ofﬁcer education and arrests, the results indicate that education and searches are

independently distributed (2’ 2 = 0.69, p = .710). Furthermore, ofﬁcer education is not a

useful measure for reducing error in predicting whether the suspect will be searched (y =
.02). As the present analysis is the ﬁrst to examine the distribution of ofﬁcer searches by
education level, it cannot be stated whether this result is in support of previous literature,
there simply is no prior point of comparison. At this level of analysis ofﬁcer education
and searches appear to be unrelated. In the next section of this inquiry, multivariate
analysis will indicate if education becomes a better indicator of searches when additional
variables are included in the analysis.

Table 5.6: Bivariate Distribution of Searches by Ofﬁcer Education (N = 3,353)*

 

High school or less Some college, no BA/BS F our-year degree or higher

It (%) n (%) n (%)
Suspect not searched 374 (77%) 1,127 (77%) 1,070 (76%)
Suspect searched 113 (23%) 331 (23%) 338 (24%)
Total 487 1,458 1,408

12= 0.69,p = .710; y = .02

 

*Three missing cases

The next two tables examine the joint distribution of ofﬁcer education level with
the use of force. First, Table 5.7 uses a dichotomous measure for force. Opposed to the

previous dependent variables (e.g., arrest and search), ofﬁcer education level and the use

50

of force are not independently distributed (1 2 = 22.71, p < .001). However, the gamma

statistic shows the relationship between the variables to be a weak one (7 = -.09), with the
knowledge of ofﬁcer education level reducing the error in predicting force by only nine
percent. The distribution of encounters in the table indicates that ofﬁcers with only a high
school background use more force than ofﬁcers with either some college exposure or a
four-year degree. Interestingly, the proportion of ofﬁcers who use force is slightly larger

for the group with completed degrees (57%) compared to those with only some exposure

 

to college (56%).
Table 5.7: Bivariate Distribution of Force by Ofﬁcer Education (N = 3,356)
High school or less Some college, no degree Four-year degree or higher
11 (%) n (%) n (%)
No force used 157 (32%) 644 (44%) 610 (43%)
Force used 330 (68%) 815 (56%) 800 (57%)
Total 487 1,459 1,410

12= 22.7l,p < .001; y = -.09

 

The ﬁnal bivariate table (Table 5.8) displays the joint distribution of ofﬁcer
education level and force, but distinguishes between verbal and physical force. Unlike

arrests and searches, ofﬁcer education level and the levels of force do not appear to be
independently distributed (2’ 2 = 27.44, p < .001). The largest difference across the

categories lies with verbal force, as 47 percent of verbal force incidents are in encounters
involving ofﬁcers with a high school diploma or less, compared to 36 percent and 35
percent for some college exposure and four-year degree, respectively. As a point worth
mentioning, the proportion of ofﬁcers using physical force is nearly identical across all
education categories. Encounters involving ofﬁcers in the collegiate categories more

often involve no use of force compared to ofﬁcers with a high school background. Like

51

arrests and searches, however, ofﬁcer education level does not produce a strong measure
of association with the use of force (y = -.05). The reason for this may be that while
ofﬁcer education does predict some of the variation in the use of force, it is by no means
the only valid predictor. The next section of the paper will examine impact of ofﬁcer
education on the use of verbal and physical force when addition variables are controlled
for by using a multinomial logistic regression.

Table 5.8: Bivariate Distribution of Verbal/Physical Force by Ofﬁcer Education
(N = 3,356)

 

High school or less Some college, no degree Four-year degree or higher
11 (%) n (%) n (%)
No force used 157 (32%) 644 (44%) 610 (43%)
Verbal force 227 (47%) 520 (36%) 499 (35%)
Physical force 103 (21%) 295 (20%) 301 (21%)
Total 487 1,459 1,410

12: 27.44,p < .001; y = -05

 

Multivariate Analysis

The present inquiry utilizes binary logistic regression and multinomial logistic
regression for the purpose of multivariate analysis. Because the dependent variablesare
measured dichotomously and categorically, certain implications carry to the speciﬁcation
of models, particularly underscoring the preference for logistic regression as opposed to
linear regression (Hanushek & Jackson, 1977; McKelvey & Zavoina, 1975) As noted, the
dependent variables in the present analysis capture whether a particular event occurred or
did not occur in a police-suspect encounter (e. g., whether a suspect was arrested, or not
arrested). The mean of these dependent variables not only represent the proportion of
cases which fall in the measured categories (e.g., the proportion of encounters involving
arrests), but may be interpreted as the probability that an encounter will result in one of

these outcomes (DeMaris, 1995). The purpose of the analysis presented here is to model

52

the probability of arrests, searches, and force as a function of ofﬁcer education and a
number of control variables.

Much has been written regarding the inability of linear regression to produce
meaningful results when using a dichotomous dependent variable (Hanushek & Jackson,
1977; Long, 1997; McKelvey & Zavoina, 1975). The difﬁculties that face linear
regression in such a scenario are due to the fact that it is ill-suited to model using a
variable with a ﬂoor of zero and a ceiling of one (Pampel, 2000). When using a
continuous dependent variable the regression line may extend beyond zero and one to
inﬁnity, depending on the strength of the slope. Because of this, linear regression with a
dichotomous dependent variable may produce nonsensical probabilities both less than
zero and greater than one (DeMaris, 1995). Linear regression also faces difﬁculty in that
measuring the dependent variable as a dichotomy violates two of its assumptions,
normalin and homoskedasticity. The assumption of normality becomes violated when
only two residuals are possible for any independent variable, given that there are only
two possible values for the dependent variable (Pampel, 2000). The assumption of
homoskedasticity, or constant error variance, is violated by a dichotomous dependent
variable as the error term will vary depending on the value of the independent variable.
At very low or very high values for the independent variable, the error term will be small,
given that the dependent variable may only have values of zero or one. At middle values,
however, the error term will be much larger. Thus, dichotomous dependent variables are
innately heteroskedastic (Berk, 2004; DeMaris, 1995; Parnpel, 2000).

Logistic regression models correct for the difﬁculties that linear regression

53

encounters with dichotomous dependent variables in several ways. First, due to the innate
characteristics of dependent variables measured in such a way, logistic regression relaxes
the linear assumptions of normalin and homoskedasticity (Grommon, 2005; Winship &
Mare, 1984). Second, logistic regression performs a transformation on the dependent
variable to linearize the non-linear relationships between the dichotomous dependent
variable and the independent predictors (DeMaris, 1995). This transformation involves
converting probabilities into odds and then taking the natural logarithm of the odds to
obtain logged odds. Logged odds eliminate the difﬁculties of regression with
dichotomous dependent variables by removing the ﬂoor of zero and the ceiling of one
(Pampel, 2000).

As opposed to linear regression, which expresses coefﬁcients as representing the
amount of change in the dependent variable given a one unit change in the independent
variable (Allen, 1997), logistic regression coefﬁcients carry a similarly structured yet
distinct interpretation. Logistic regression coefﬁcients represent the change in the logged
odds of an event based on a one unit change in the independent variable (Pampel, 2000).
In the tables below these coefﬁcients are displayed as ‘b’ along with their standard errors
(SE).

Another helpful interpretation of logistic regression coefﬁcients involves taking
the antilogarithrn of the logged odds to display the impact of the independent variable

directly on the odds of an event occurring. In the tables below these are shown under the

columns for exponentiated coefﬁcients (Exp(b)).36 For these values, odds equal to one

 

36 These values are also known as odds ratios. The “ratio” title may be more appropriate because the
values appearing in the tables below are not speciﬁcally the odds of a college degree-holding ofﬁcer

54

indicate that the independent variable has no impact on the odds of the event in the
dependent variable occurring. Odds lower than one mean that a one unit change in the
independent variable decreases the odds of an event, and odds greater than one represent
an increase (Liao, 1994). These odds also offer a simple interpretation. For example, an
exponentiated coefﬁcient of 1.34 for suspect sex in the arrest model indicates that a
suspect’s being a male increases the odds of being arrested. More speciﬁcally, 134 males
are arrested for every 100 females. Another way of interpreting the same ﬁgure is to say
that the odds of a male being arrested is 1.34 times higher than the odds of a female
suspect being arrested (Liao, 1994).

In order for partial logistic regression coefﬁcients to be interpreted in any
meaningﬁrl fashion, the model as a whole must show that it can explain the dependent
variable signiﬁcantly better than a null model, which only contains the intercept (Liao,
1994). In logistic regression this is done using log likelihood values, similar to linear
regression using the F statistic. The model signiﬁcance is determined by comparing the
log likelihood value of the null model to the log likelihood of the model being evaluated.
The greater the difference between the null value and the tested value the less likely that
all coefﬁcients are equal to zero in the population (DeMaris, 1995; Liao, 1994). A test of
signiﬁcance using a chi-square distribution indicates if the difference between the null
and tested log likelihoods is due to chance (Pampel, 2000). This value is presented in the
tables below as the model chi-square.

Each model in the tables below also includes a pseudo-variance explained

measure, or pseudo-R2. The Cox and Snell (1989) R2, and other measures like it, is called

 

arresting a suspect. Rather, the value that appears is the ratio of the odds of a degree holding ofﬁcer
arresting a suspect to the odds of a high school educated ofﬁcer arresting a suspect.

55

a pseudo-variance explained measure because it is calculated via log likelihood values,
which do not capture variance in the same way that the sum of squared deviations does in
linear regression (Pampel, 2000). These pseudo-measures are nonetheless capable of
providing an indication of explained variance based upon model speciﬁcation.

The analysis is presented in a stepwise manner, beginning with ofﬁcer
characteristics and subsequently adding the suspect and encounter level variables.
Presenting the analysis in this way allows one to examine how each group of variables
contributes to the model, culminating in the Full Model (represented as Model 3 in Table
5.9). This method of presentation is preferred for the present analysis because unlike an
exploratory study seeking to ﬁnd what variables, if any, are related to arrests, the present
analysis has a primary independent variable of interest — ofﬁcer education. By
sequentially adding in groups of variables, a researcher is capable of monitoring how the
impact of education changes as more variables are considered. The incomplete models
are thus presented as a matter of interest. The relationship between ofﬁcer education and
the dependent variables in the full models represent the ultimate ﬁndings of the present
analysis.

Arrest modeling. Table 5.9 displays the logistic regression of arrests on
characteristics of ofﬁcers, suspects, and encounters. The model containing only ofﬁcer
characteristics is presented in Model 1 in Table 5.9. This modeling indicates that, relative
to encounters involving high school educated ofﬁcers, encounters involving ofﬁcers with
some college or a four-year degree (b = -0.10, p =.537 and b = 0.01, p = .927,

respectively) are not any more or less likely to result in an arrest. In fact, the inﬂuence of

56

this group of variables is so weak that they do not explain arrests signiﬁcantly better than
a null model containing only the intercept (Model 2' 2 = 2.84, p = .725).

When suspect characteristics are included (Model 2 in Table 5.9), multiple
variables are found to signiﬁcantly affect the odds of an arrest taking place.
Characteristics of suspects themselves — age, sex, race and socioeconomic status — all
inﬂuence the odds of arrest in model 2. More speciﬁcally, suspects who are younger,
male, non-white, and of low socioeconomic status are more likely to be arrested in an
encounter with an ofﬁcer. Also, offering higher levels of resistance, being intoxicated,
and greater amounts of evidence present all increase the odds of an arrest taking place.

Model 3 represents the Full Model, when variables from all three groups are
considered. Here, higher education, whether it is just some college (b = —O. 14, p = .431)
or a four-year degree (b = -0.02, p = .919), is not a signiﬁcant predictor of arrests relative
to ofﬁcers with no college education. All other ofﬁcer characteristics also show no
signiﬁcant inﬂuence on the odds of an arrest taking place. This ﬁnding replicates that of
Worden (1989) who, using PSS data, found that no ofﬁcer characteristics inﬂuence arrest
behavior. Concerning suspect characteristics, some interesting differences between
Model 2 and the Full Model are apparent. Introducing the encounter level variables
causes some physical characteristics of suspects — age, sex, and race - to no longer
signiﬁcantly predict the likelihood of an arrest. Also, suspect demeanor (e.g., whether the
suspect was disrespectful to the ofﬁcer) now signiﬁcantly increases the odds of an arrest
taking place. As with Model 2, resistance on the part of the suspect, being intoxicated,

and the amount of evidence present prior to the arrest all signiﬁcantly increase the odds

57

of an arrest, while belonging to a higher socioeconomic status signiﬁcantly lowers the
likelihood of being arrested. All four encounter level variables appear to signiﬁcantly
inﬂuence the odds of an arrest. More speciﬁcally, encounters involving higher numbers
of ofﬁcers on scene and being in Indianapolis are more likely to result in a suspect being
arrested, relative to encounters with fewer ofﬁcers, and those that take place in St.
Petersburg. Finally, encounters with more citizens present and were initiated by the
ofﬁcer are less likely to result in an arrest, relative to encounters involving fewer citizens

and those encounters that the ofﬁcer responds to reactively.

58

Table 5.9: Binary Logistic Regression of Arrests on Ofﬁcer, Suspect, and Encounter
Characteristics (N = 3,354)

 

Variable
Intercept

Oﬂicer Characteristics
Education
Some college
Four-year degree
Experience
Male
Non-white

Suspect Characteristics
Age

Non-white

Male

Socioeconomic Status
Resistance

Conﬂict

Weapon

Evidence

Demeanor

Drug or Alcohol

Encounter Setting
Number of Ofﬁcers
Number of Citizens
Proactive Encounter
Observation site

N
-2 Log Likelihood

Model 1 2
df

Cox & Snell R 2

Model 1. Officer

Model 2. Ofﬁcer & Suspect

Model 3. Full Model

 

b (SE)

-1.65 (.20)

-010 (.15)
0.10 (.15)
-001 (.01)
0.06 (.14)
-0.06 (.12)

3,354

Exp(b)

0.19

0.91
1.01
0.99
1.06
0.95

2,870.23

2.84

5
.001

b (SE) Exp(b)
-210 (.42) 0.12
-011 (.17) 0.90
-001 (.17) 0.99
0.00 (.01) 1.00
-010 (.16) 0.93
-0.06 (.14) 0.95
-0.11 (.04) 0.89"
0.31 (.12) 136*
0.29 (.13) 134*
-050 (.10) 061*"
0.48 (.07) 1.62***
-0.06 (.09) 0.95
-0.03 (.42) 1.05
0.58 (.03) 1.78***
0.32 (.17) 1.38
0.72 (.13) 2.05***
3,354
2,266.78
606.29***
15
.165

b (SE) Exp(b)
-353 (.50) 0.03
014 (.18) 0.87
-0.08 (.18) 0.98
0.01 (.01) 1.01
-013 (.16) 0.88
-010 (.15) 0.91
-004 (.05) 0.96
0.21 (.13) 1.23
0.21 (.14) 1.24
-0.46 (.11) 063*"
0.35 (.08) 142*"
-0.08 (.10) 0.92
0.03 (.44) 0.97
0.65 (.04) 191*"
0.49 (.18) 1.63“
0.59 (.13) 1.80***
0.46 (.04) 1.59***
-003 (.01) 0.97"
-039 (.13) 0.68M
0.58 (.13) 1.80""
3,354
2,047.84
825.23***

.218

 

*p < .05, "p < .01, ***p < .001

Search modeling. The binary logistic regression models for searches are presented

in Table 5.10. Unlike the models for arrests, all three models were signiﬁcantly better at

explaining searches than a model containing only the intercept. Model 1 displays the

logistic regression of ofﬁcer characteristics alone on searches. As with arrests, ofﬁcer

education does not appear to inﬂuence the odds of a search taking place in an encounter.

59

However, unlike the models for arrests, some ofﬁcer characteristics appear to have a
signiﬁcant effect on searches. Speciﬁcally, ofﬁcers who are more experienced (b = -0.02,
p < .01) and non-white (b = -O.34, p < .01) are less likely to search suspects than less
experienced ofﬁcers and white ofﬁcers, respectively.

When suspect characteristics are included in Model 2, the effects of ofﬁcer
experience and race retain their signiﬁcance, and multiple suspect characteristics prove to
signiﬁcantly inﬂuence the odds of a search taking place. As with arrests, suspects who
are younger, male, non-white, and of low socioeconomic status are more likely to be
searched. The strongest predictor of being searched in this model is being arrested (b =
1.78, p < .001). Other factors found to increase the likelihood of being searched are
carrying a visible weapon and being intoxicated. Surprisingly, suspects involved in
higher states of conﬂict with other citizens on the scene are less likely to be searched, and
the amount of evidence available does not seem to be a signiﬁcant predictor as well.

When the encounter level variables are included in the Full Model, only a small

amount of pseudo—explained variance is gained, however, the difference between the
models is signiﬁcant (12 = 33.98, df= 4, p < .01).” When all other variables are

considered, some exposure to college (b = 0.06, p = .652) and holding a four-year degree
(b = -0.04, p = .777) do not signiﬁcantly inﬂuence the odds of an encounter involving a
search relative to encounters with non-college educated ofﬁcers. Regarding the other
ofﬁcer characteristics, introducing encounter level variables mediates the relationship

between ofﬁcer experience and searching, but ofﬁcer race maintains its inﬂuence,

 

37 This chi-square value is calculated by subtracting the model chi-square and degrees of freedom of the
smaller model from the full model (DeMaris, 1995).

60

indicating that non-white ofﬁcers are less likely to search suspects. Considering suspect
characteristics, the encounter level variables control the relationships between suspect
race and citizen conﬂict with searches. All other signiﬁcant relationships found between
suspect variables and searches in Model 2 retain their signiﬁcance in the Full Model. For
the encounter level variables themselves, only the encounter site did not signiﬁcantly
impact the odds of a search taking place. More citizens being present decreases the odds
of a search taking place, encounters with more ofﬁcers present and proactive encounters

are more likely to result in a search of the suspect.

61

Table 5.10: Binary Logistic Regression of Searches on Ofﬁcer, Suspect, and Encounter
Characteristics (N = 3,353)

 

Variable

Intercept

Oﬂicer Characteristics

Education

Some college
Four-year degree

Experience
Male
Non-white

Suspect Characteristics

Age
Non-white
Male

Socioeconomic Status

Resistance
Arrested
Conﬂict
Weapon
Evidence
Demeanor

Drug or Alcohol

Encounter Setting
Number of Ofﬁcers
Number of Citizens
Proactive Encounter
Observation site

N

-2 Log Likelihood

Model 1 2
df

Cox & Snell R 2

Model 1. Ofﬁcer

Model 2. Ofﬁcer & Suspect

Model 3. Full Model

 

b(SE)
-1.10(.17)

-007 (.13)
-009 (.13)
-002 (.01)
0.23 (.12)
-034 (.11)

Exv(b)
0.34

0.94
0.92
0.98“
1.25
0.71 **

3,353
3,816.52
2380*"

5
.007

b (3E) Exp(b)
-074 (.37) 0.48
0.03 (.14) 1.03
-004 (.14) 0.96
-002 (.01) 098*
0.17 (.13) 1.19
-034 (. 12) 0.72**
-0.13 (.03) 0.88***
0.24 (.10) 1.28*
0.75 (.11) 2.12***
-031 (.08) 0.73***
0.01 (.07) 1.01
1.78 (.13) 5.95***
-0.19 (.09) 083*
0.98 (.31) 2.65**
0.04 (.03) 1.05
-025 (.16) 0.78
0.61 (.11) 1.83***
3,353
3,193.77
448.54***
16
.125

b (SE) Exp(b)
-1.23 (.39) 0.29
0.06 (.14) 1.07
0.04 (.14) 0.96
-001 (.01) 0.99
0.13(.13) 1.14
-033 (.12) 0.72**
-014 (.04) 0.87***
0.17 (.10) 1.19
0.71 (.12) 204*"
-031 (.08) 0.74***
0.04 (.07) 0.96
1.71 (.13) 5.52***
-011 (.09) 0.90
0.97 (.31) 2.64**
0.03 (.03) 1.03
-O.16 (.17) 0.86
0.64 (.11) 1.90***
0.23 (.03) 1.26***
-005 (.01) 0.95***
0.38 (.10) 1.46***
0.08 (.10) 1.08
3,353
3,119.79
522.52***
20
.144

 

*p < .05, "p < .01, ***p < .001

Force modeling. The results of the binary logistic regression for the use of force

are shown in Table 5.11. Consistent with the bivariate analysis, ofﬁcer education level

appears to be signiﬁcantly related to the probability of an ofﬁcer using force in an

encounter. Model 1 displays the regression of ofﬁcer characteristics on the use of force.

Here it appears that ofﬁcers with some college exposure and completed degrees are

62

signiﬁcantly less likely to use force in an encounter (b = -0.61, p < .001, and b = -0.66, p
< .001, respectively). This is a stark contrast to the previous models for arrests and
searches as until this point ofﬁcer education did not appear to be related to the current
inquiry’s dimensions of police behavior. Of the other ofﬁcer characteristics, more
experienced ofﬁcers are less likely to use force, and non-white ofﬁcers are signiﬁcantly
more likely to use force in an encounter.

Model 2, presented in Table 5.11, controls for the effects of suspect
characteristics. Introducing these variables does not mediate the inﬂuence of ofﬁcer
education level on the probability of force occurring. Interestingly, essentially all suspect
level variables are signiﬁcantly related to the use of force, with the exception of suspect
demeanor. Regarding the direction of those effects, suspects who are younger, non-white,
male and of a low socioeconomic status are more likely to be subject to force by the .
ofﬁcer in each encounter. Suspects who are resisting, in conﬂict with bystanders, have a
weapon within reach, intoxicated, presenting much evidence, and arrested are also more
likely to be recipients force.

The introduction of encounter level variables in Model 3 changes very few of the
relationships from the previous models. When all other variables are held constant,
ofﬁcer education level is still signiﬁcantly related to the use of force. Speciﬁcally,
ofﬁcers with some college exposure and four-year degrees are signiﬁcantly less likely to
use force in an encounter (b = -0.49, p < .001 and b = -0.68, p < .001, respectively)
relative to non-college educated ofﬁcers. While more experienced ofﬁcers are also less

likely to use force, the introduction of encounter level variables controlled for the

63

 

relationship between ofﬁcer race and the use of force. All significant relationships

between suspect characteristics and the use of force retained their signiﬁcance in the full

model. Regarding the encounter level variables themselves, encounters initiated by the

ofﬁcer are more likely to involve force, as are encounters which take place in

Indianapolis, compared to encounters in St. Petersburg.

Table 5.11: Binary Logistic Regression of Force on Ofﬁcer, Suspect, and Encounter
Characteristics (N = 3,356)

 

Variable
Intercept

Oﬂicer Characteristics
Education

Some college

F our-year degree
Experience
Male
Non-white

Suspect Characteristics
Age

Non-white

Male

Socioeconomic Status
Resistance

Arrested

Conﬂict

Weapon

Evidence

Demeanor

Drug or Alcohol

Encounter Setting
Number of Ofﬁcers
Number of Citizens
Proactive Encounter
Observation site

N
-2 Log Likelihood

Model 1 2
df

Cox & Snell R 2

Model 1. Ofﬁcer

Model 2. Ofﬁcer & Suspect

Model 3. Full Model

 

b (SE)
0.97 (.15)

-0.61 (.11)
-O.66 (.12)
-003 (.01)
0.13 (.10)
0.20 (.09)

3,356

Exp(b)
2.64

055*"
052*“
0.97***
1.13
1.22*

4,509.37
57.70***

5
.017

b (3E) Exp(b)
0.11 (.32) 1.12
-0.61 (.12) 0.54***
-0.67 (.13) 0.51***
-0.03 (.01) 0.97***
0.11(.11) 1.11
0.21 (.10) 123*
-015 (.03) 0.86***
0.27 (.08) 1.32***
0.35 (.08) 1.41***
-023 (.07) 0.80***
0.77 (.11) 2.16***
1.05 (.15) 2.90***
0.41 (.09) 1.51***
1.38 (.42) 3.98***
0.16 (.02) 1.18***
-0.11 (.15) 0.90
0.51 (.10) 1.66***
3,356
4,060.38
506.69***

b (SE) Exp(b)
-30 (.34) 0.74
-049 (.12) 0.61***
-O.68 (.13) 0.51***
-003 (.01) 0.97***
0.11(.11) 1.11
0.14 (.10) 1.15
-014 (.03) 0.87***
0.22 (.09) 1.25**
0.33 (.09) 1.39***
-025 (.07) 0.78***
0.77 (.11) 2.16***
1.08 (.16) 2.95***
0.46 (.09) 1.59***
1.36 (.43) 3.91***
0.16 (.02) 1.17***
-003 (.15) 0.97
0.57 (.10) . 1.76***
-0.01 (.03) 1.00
-000 (.01) 1.00
0.36 (.08) 1.43***
0.43 (.08) 1.53***
3,356
4,008.07
559.01***

.153

 

*p < .05, Mp < .01, ***p < .001

64

 

Verbal versus physical force modeling. The force modeling presented in Table
5.11 utilized a dichotomous dependent measure. Measuring force in such a way
potentially ignores differences between verbal force and physical force. In the bivariate
analysis section, a contingency table (Table 5.8) demonstrated that verbal force is
distributed differently across education categories when compared to physical force.
Examining how this difference manifests in a multivariate analysis is certainly warranted.

The following tables (Tables 5.12-14) use a dependent variable which
distinguishes between verbal and physical force. Unlike the multivariate analyses for the
previous variables which utilized binary logistic regression, the analysis for this use of
force variable requires a different form of logistic regression. As the force categories are
discrete and cannot be naturally ordered, multinomial logistic regression is the preferred
approach (Liao, 1994; Long, 1997). Multinomial logistic regression, or polytomous
logistic regression, is a natural extension of the binary logistic regression model
(DeMaris, 1992). While a binary logistic regression of the use of force would compare
the likelihood of force with that of no force, a multinomial logistic regression using the
three force categories from the current analysis compares the likelihood of verbal force to
no force, and physical force to no force (Liao, 1994).38 As such, each table presents the
addition a group of variables — ofﬁcer, suspect, and encounter — and includes two sets of
coefﬁcients and odds; one for comparing verbal force to no force, and another for

comparing physical force to no force.

 

3
8 A multinomial logistic regression model will always make k-l comparisons, where k is equal to the
number of categories in the dependent variable (Liao, 1994).

65

 

The interpretation of odds ratios may become somewhat confusing in a
multinomial logistic regression model (DeMaris, 1992). The models may be interpreted
as comparing the probability of verbal force and physical force occurring in an encounter
relative to no force. For example, the exponentiated beta coefﬁcient for suspect race and
physical force in the Full Model below is 1.50. This means that the odds of a non-white
suspect being the recipient of physical force instead of no force is 1.5 times the odds of a
white suspect being on the receiving end of physical force instead of no force (Liao,
1994)

Table 5.12 below displays the ﬁrst step of the multinomial logistic regression with
just ofﬁcer characteristics. Unlike the models for arrests and searches, ofﬁcer
characteristics appear to inﬂuence the odds of both verbal and physical force. Encounters
involving ofﬁcers with some college exposure (b = -0.67, p < .001) and four-year degrees
(b = -0.72, p < .001) are signiﬁcantly less likely to involve verbal force relative to non-
college educated ofﬁcers. This model also indicates that some college exposure (b = -
0.48, p < .001) and holding a degree (b = -0.54, p < .001) signiﬁcantly reduce the
likelihood of physical force in an encounter, compared to high school educated ofﬁcers.
Higher levels of ofﬁcer experience appear to reduce the odds of both verbal and physical
force. While non-white ofﬁcers are more likely to use verbal force instead of no force in
an encounter, the same ﬁnding is not apparent for physical force.

Table 5.13 presents the second model of the force analysis, with suspect
characteristics included. The introduction of suspect characteristics does not fully control

for the relationships between ofﬁcer characteristics and both verbal force and physical

66

force. Among the suspect variables, however, there are some interesting distinctions
between predictors of verbal and physical force. Suspect sex does not inﬂuence the odds
of verbal force in an encounter, but male suspects are signiﬁcantly more likely to receive
physical force from ofﬁcers. Encounters where suspects are arrested are signiﬁcantly less
likely to involve verbal force, but signiﬁcantly more likely to involve physical force.
Also, suspects in conﬂict with other citizens on the scene are signiﬁcantly more likely to
receive verbal force from ofﬁcers, but not more or less likely to receive physical force.
Otherwise, there are many similarities at this level between predictors of verbal and ,
physical force. Suspects that are younger, non-white, holding a low socioeconomic status,
resisting of the ofﬁcers requests, intoxicated, visibly armed, and have much evidence
present are signiﬁcantly more likely to receive both verbal and physical force.

The Full Model for the use of force is presented in Table 5.14. With all other
ofﬁcer, suspect, and encounter variables being controlled for, both college education
variables appear to signiﬁcantly reduce the odds of verbal force in an encounter (b = -
0.56, < .001 and b = -0.73, < .001, for some college exposure and four-year degree,
respectively) relative to non-college educated ofﬁcers. Regarding physical force, simple
exposure to college does not result in a signiﬁcant effect, however, encounters involving
ofﬁcers with a four-year degree are signiﬁcantly less likely to involve physical force (b =
-0.53, p < .01). The introduction of encounter level variables changed little else among
ofﬁcer and suspect variables. As opposed to the second force model, suspects in conﬂict
with other citizens are signiﬁcantly more likely to be subjected to physical force instead

of no force. Among the encounter level variables, encounters with higher numbers of

67

police present are more likely to involve physical force, and less likely to involve verbal

force. Encounters initiated by the ofﬁcer and those that take place in Indianapolis are

more likely to involve both verbal and physical force instead of no force at all.

Table 5.12: Multinomial Logistic Regression of Force on Ofﬁcer Characteristics

 

 

(N = 3,356)
Model 1. Officer Characteristics
Variable Verbal Force v. No Force Physical Force v. No Force
b (SE) Exp(b) b (SE) Exp(b)

Intercept 0.56 (.16) -- -0.14 (.20) --
Oﬁ‘icer
Education

Some college -0.67 (.12) 051*“ -0.48 (.15) 062*“

4-yr Degree -0.72 (.13) 050*" -0.54 (.15) 059*”
Experience -0.03 (.01) 0.97*** -0.05 (.01) 096*"
Male 0.06 (.11) 1.07 0.25 (. 14) 1.28
Non-white 0.27 (.10) 1.31" 0.04 (.12) 1.05
N 3,356
-2 Log Likelihood 1,043.53
Model 2' 2 72.21***
(If 10
Cox & Snell R 2 '021

 

*p < .05, **p < .01, ***p <

001

68

 

Table 5.13: Multinomial Logistic Regression of Force on Ofﬁcer and Suspect
Characteristics (N = 3,356)

 

Model 2. Ofﬁcer and Suspect

 

 

Variable Verbal Force v. No Force Physical Force v. No Force
b (SE) Exp(b) b (SE) Exp(b)
Intercept -0.35 (.34) -- -0.87 (.47) --
Oﬂicer
Education
Some college -0.68 (.13) 051*“ -0.39 (.18) 068*
4-yr Degree -0.72 (.13) 049*" -0.54 (.18) 0.59"
Experience -0.02 (.01) 098*" -0.05 (.01) 096*"
Male 0.07 (.11) 1.07 0.22 (.16) 1.25
Non-white 0.24 (.10) 1.27* 0.10 (.14) 1.10
Suspect
Age -0. 10 (.03) 0.91" -0.29 (.04) 075*”
Non-white 0.20 (.09) 1.22* 0.48 (.12) 162*"
Male 0.16 (.09) 1.18 0.97 (.14) 264*"
Socioeconomic Status -0.18 (.07) 0.84”“ -0.37 (.10) 069*"
Resistance 0.76 (.11) 222*" 0.82 (.12) 2.26"”MI
Arrested -1.60 (.30) 0.20*** 2.52 (.17) 1248*”
Conﬂict 0.46 (.09) 159*" 0.23 (.12) 1.25
Weapon 0.97 (.46) 2.65* 2.10 (.47) 8.00""
Evidence 0.17 (.03) 1.19*** 0.12 (.03) 1.13“”
Demeanor -0.03 (.16) 0.97 -0.29 (.21) 0.75
Drug/Alcohol 0.34 (.12) 1.40" 0.94 (.14) 255*"
N 3,356
-2 Log Likelihood 5,454 .69
Model I 2 1,214.66***
df 32
Cox & Snell R 2 3’03

 

*p < .05, "p < .01, "*p < .001

69

 

Table 5.14: Multinomial Logistic Regression of Force on Ofﬁcer, Suspect, and Encounter
Characteristics (N = 3,356)

 

Model 3. Full Model

 

Variable Verbal Force v. No Force Physical Force v. No Force
b (SE) Exp(b) b (SE) Exp(b)
Intercept -0.35 (.37) -- -1.82 (.50) --
Ofﬁcer
Education
Some college 056 (.13) 0.57*** -0.31 (.18) 0.73
4-yr Degree -0.73 (.13) 048*" -0.53 (.18) 0.59“
Experience -0.02 (.01) 0.98" -0.04 (.01) 096*"
Male 0.07 (.12) 1.07 0.19 (.16) 1.22
Non-white 0.16 (.10) 1.19 0.05 (.14) 1.05
Suspect
Age -0.10 (.13) 090*" -O.27 (.04) 076*"
Non-white 0.17 (.09) 1.19* 0.41 (.13) 150*"
Male 0.16 (.09) 1.17 0.91 (.15) 249*"
Socioeconomic Status -0.22 (.08) 0.80" -0.37 (.10) 0.69"”""I
Resistance 0.77 (.11) 2.15*** 0.77 (.12) 2.17“"
Arrested -1.47 (.30) 023*" 2.47 (.17) 1184*"
Conﬂict 0.51 (.09) 1.67“” 0.30 (.12) 1.35“
Weapon 0.97 (.46) 2.63* 2.11 (.47) 821*"
Evidence 0.17 (.03) 1.19*** 0.12 (.03) 1.12“"
Demeanor 0.05 (.16) 1.06 -0.19 (.22) 0.83
Drug/Alcohol 0.42 (.11) 1.52*** 0.98 (.14) 2.66"”
Encounter
Number of Ofﬁcers -0. 16 (.04) 085*" 0.20 (.04) 122*"
Number of Citizens -0.00 (.01) 0.10 -0.01 (.01) 0.99
Proactive Encounter 0.28 (.09) 1.14” 0.46 (.12) 158*"
Observation Site 0.44 (.09) 1.55*** 0.38 (.12) 1.46"
N 3,356
-2 Log Likelihood 5,692.77
Model 12 1,347.90***
(If 40 .
Cox & Snell R 2 '331

 

*p <05, **p < .01, ***p < .001

Summary of multivariate modeling. The preceding analysis examined the impact
of college education on three measures of ofﬁcer behavior (e.g., arrest, search, and the
use of force) while controlling for a number of variables that have been found to be
theoretically relevant or signiﬁcant in past research. The results indicate that college

education has no signiﬁcant inﬂuence on the odds of an arrest or a search taking place in

70

 

a police-suspect encounter. While encounters involving a college educated ofﬁcer were
signiﬁcantly less likely to involve verbal force, in order to see a similar result for physical
force, ofﬁcers needed to have obtained at least a four-year degree.

Multivariate Diagnostics. Collinearity statistics were assessed for all multivariate
models to search for potentially confounding interrelationships among the independent
variables. No problematic variance inﬂation factors (V IF) were present in any model, as
most ranged between 1.0 and 1.3. The highest VIFs in any models were for the education
variables, which were approximately 2.4 and 2.5 for some college and four-year degree,
respectively. This is not unexpected given that these are two dummy-coded variables
measuring the same concept. The tolerance values for these variables were approximately
.400, and no tolerance value outside of the education variables fell below .700 in any
model. From these ﬁgures it was concluded that multicollinearity did not undermine any

of the ﬁndings from the analysis above.

71

 

CHAPTER VI: DISCUSSION

The purpose of the research presented here was to examine the impact of higher
education on particular outcomes in police-suspect encounters. Previous examinations of
this nature sought to analyze the effect of higher education on but a single outcome per
study, impeding the comparability of education’s potentially differential impact on a
variety of ofﬁcer behavioral outcomes. For the current inquiry the dependent measures -
arrests, searches, and force - were modeled as functions of ofﬁcer education level and
other controlling factors. This analysis allowed for a comparison of higher education’s
impact on these measures to be made across more than 3,000 police-suspect encounters in
two cities over two summers. The results both conﬂicted and supported individual
ﬁndings from previous examinations, but also provided new ﬁndings in relationships that
had not been analyzed before.
Overview of Findings

In general, the impact of higher education on ofﬁcer arrests, searches, and force
ranged from negligible to modest. Ofﬁcer education level yielded no inﬂuence over the
probability of an arrest taking place in an encounter. This was true When the
characteristics of suspects and encounters were held constant, but also when individual
ofﬁcer characteristics were considered alone. This ﬁnding supports previous indications
of a null relationship between ofﬁcer education and arrests (Brandl et al., 2001; Smith &
Aamodt, 1997; Smith & Klein, 1983; Worden, 1989). Indeed, only a minority of studies
ﬁom the mid-19703 indicated a discemable relationship between ofﬁcer education and

arrest behavior. Two of these studies used a very small sample of ofﬁcers (Bozza, 1973;

72

 

Glasgow et al., 1973) and one did not actually measure arrests per se, but rather ofﬁcer
attitudes towards the use of arrests (F ickenauer, 1975).

Suspect and encounter characteristics appeared to provide the best predictors of
arrests. These factors included the amount of evidence present, the suspect’s
socioeconomic status, whether the suspect resisted the ofﬁcer, was disrespectful, or
intoxicated, as well as the number of ofﬁcers and citizens present at the scene. It is
interesting to note that while certain suspect characteristics (e.g., age, race, sex) were
found to be signiﬁcantly related to arrests in the model containing only ofﬁcer and
suspect characteristics, the introduction of encounter characteristics controlled for these
relationships. Also, suspect demeanor (e. g., disrespect) was found to be signiﬁcantly
related to the probability of arrest in the full model, contradicting ﬁndings from Klinger
(1996).39

Schafer and colleagues (2004) hypothesized that ofﬁcer education could be
related to search decision making, but were unable to include such a variable in their
analysis. No previous inquiry had examined the effect of ofﬁcer education level on
his/her propensity to search suspects in police-suspect encounters. The results of the
preceding analysis indicate that at both the bivariate and multivariate levels, higher
education does not affect whether a search will take place in a given police-suspect

encounter. Suspect and encounter characteristics again proved to be the best indicators of

 

39 The demeanor variable was measured as a dichotomy. This ﬁnding contradicts that of Klinger (1996),
who, using a sample of ofﬁcers from Dade County, Florida, found that measuring suspect demeanor as a
dichotomy was not significantly related to arrest when controlling for the type of crime and conﬂict with
other citizens on the scene. Instead, Klinger had to measure demeanor as the incidence of extreme hostility
to produce a signiﬁcant ﬁnding. While the present analysis did not include the type of crime supposedly
committed by the suspect as a controlling factor, the summative evidence scale may approximate such an
effect.

73

 

search behavior. These included the suspect’s age, sex, and socioeconomic status, and
whether they were visibly intoxicated. The results from the present inquiry do not
indicate that ofﬁcer education is a signiﬁcant factor concerning the probability of a
search taking place; however, non-white ofﬁcers were signiﬁcantly less likely to search a
suspect.

The use of force was the only dependent measure on which ofﬁcer education
showed a signiﬁcant impact. It was found that some exposure to college was sufﬁcient to
reduce the likelihood of verbal force being used in an encounter. However, for a similar
effect to be produced for physical force, the ofﬁcer in the encounter needed to be holding
at least a four-year degree.40 In addition to holding a four-year degree, ofﬁcer experience
was also found to reduce the likelihood of verbal and physical force taking place in an
encounter. Suspect characteristics appeared to be the strongest predictors of both verbal
and physical force. Interestingly, the suspect being arrested signiﬁcantly reduced the
likelihood of the ofﬁcer using verbal force, but was the strongest predictor of the ofﬁcer
resorting to physical force in an encounter.41
Further Questions for Consideration

The results of the analysis performed here indicate that higher education does not

impact all facets of police behavior equally. The education level of ofﬁcers did not

 

40 These ﬁndings are not surprising given that the analysis utilized the same data and a similar set of
variables as an analysis by Paoline and Terrill (2007). However, the point of including ofﬁcer use of force
in the analysis was to conrpare it to arrests and searches in terms of how higher education may or may not
irrrpact these decisions.

41 This ﬁnding may be due to the fact that the verbal/physical force variable captured the highest level of
force used. Arresting a suspect is much more likely to involve physical restraint, and thus it is not unlikely
that arrests become a signiﬁcant predictor of physical force during an encounter. The correlation coefﬁcient
between arrest prior to use of force and a dummy variable for physical force cases was 0.505.

74

generate any meaningful impact on their propensity to arrest or search a suspect, but did
produce an effect on an ofﬁcers’ likelihood of using force in an encounter. Had higher
education been found to signiﬁcantly impact the odds of either arrest or search, it would
be here that some determination could be made as to where education bears its strongest
effect on these dimensions of police behavior. However, given the results of the analysis,
higher education shows its strongest and only effect on an ofﬁcer’s propensity to use
force.

The ﬁndings as a whole beg the question of why higher education has a
differential impact on these dimensions of police behavior. The analysis itself is not
capable of providing a satisfactory answer to this question. This is partially due to the
lack of empirically tested hypotheses regarding higher education and policing which
could be used to interpret the results. But also, because of the constraints of foresight and

practicality, the analysis above simply did not ask why higher education may or may not

inﬂuence ofﬁcer behavior.42 Some speculative interpretation of the analysis may be
useful, and this interpretation will hopefully be subjected to empirical analysis in future
research.

The nature or distinctive roles that these three behaviors play within the police
suspect-encounter may be one reason why higher education offers a differential impact.
Arrests are an end to the encounter, representing the point at which the police have
brought an individual into the formal criminal justice system. Force, on the other hand,

represents a means for ofﬁcers to achieve an end. Searches occupy a middle ground

 

42 It was only after research had begun that the lack of knowledge in the area of how or why higher
education affects police became apparent. The available data contained very little information which could
be used to answer this question.

75

 

 

between arrests and force in that they can represent an end to the encounter if no evidence
is uncovered, or may be used as a means to obtain evidence towards an end (e. g, the
arrest of the suspect). This distinction carries implications towards the exercise of
discretion in resorting to these behaviors. An example focusing on the decision to arrest
as an end may be useful. Previous research has indicated that ofﬁcers exercise discretion
in pursuing arrests (e. g., Terrill & Paoline, 2007). It may be possible that, relative to
force, arrests present fewer situations in which the ofﬁcer has a large amount of
discretion to wield. In certain encounters, it follows that an ofﬁcer may be presented with
a situation where even if his/her college education was somewhat responsible for a
decision to ‘give a guy a break’ and not pursue an arrest, that discretion may be mediated

by the practicality of such a decision (e.g., the suspect was involved in a serious

offense).43 In these encounters, the ofﬁcer’s discretion to arrest would be constrained.

What is left is a smaller number of encounters in which the ofﬁcer maintains full

. . 44
drscretron to arrest.

 

43 While this exarrrple posited that college educated ofﬁcers may be less likely to pursue an arrest in an
encounter with a suspect, an equally meaningful example could be drawn from positing that suCh ofﬁcers
are actually more likely to pursue an arrest. Several scholars (Bozza, 1973, Sherman & Blumberg, 1981,
Wilson, 1975) have hypothesized that higher education may inoculate ofﬁcers with a greater motivation to
exceed expectations. As such, ofﬁcers may be presented with fewer opportunities, relative to force, to
zealously pursue an arrest because of practical reasons (e.g., lack of evidence).

44 Though this example is focused speciﬁcally on arrests, the same point on limited discretion could be
applied to searches. Engel (2008) demonstrated that between 60 and 88 percent of searches are non-
discretionary (e.g., inventory searches, searches incident to arrest, searches based on warrants, or consent
searches). The current inquiry did not differentiate between the discretionary and non-discretionary
searches, which may be one possible explanation for the null relationship between ofﬁcer education and
searches.

76

 

Unlike arrests, the use of force is not an end to the encounter, but primarily a very
different phenomenon. Force may be used throughout encounters as a means to
controlling the behavior of suspects (Muir, 1977). In any given encounter, police ofﬁcers
are permitted to use force if they feel it would be appropriate (Bittner, 1970). The police
are thus afforded great discretion in the application of force to suspects as it is not legally
required or prohibited in any encounter. On this point, Klockars (1996) states, “the police
need not invoke ‘the law’ to use force, though they may decide to use force to invoke ‘the

3”

law (p. 12). This carries implications with respect to the potential role higher education
may play in relation to police use of force behavior.

In one sense, as opposed to the decision to arrest or search, there is more room for
ofﬁcer education to have an impact on discretion with respect to force. This means that
ofﬁcer education can have an impact on the use of force, because of the amount of
discretion ofﬁcers wield in using it. Though ofﬁcers are afforded more discretion in using
force than for pursuing arrests or searches, to imply that there are not certain factors
which make force more or less desirable in an encounter would be untrue. If an encounter
involves a suspect who is highly resisting of the ofﬁcer’s requests, or is intoxicated, or
becoming violent with bystanders, the ofﬁcer’s discretion to use force becomes
constrained by practicality. Such encounters may require some kind of force on the part
of the ofﬁcer in order to gain compliance from the suspect. The question which follows is
after these factors are considered, does higher education have an impact on the use of

force. The current inquiry controlled for these inﬂuences and still found that higher

education inﬂuenced the likelihood of force occurring.

77

 

 

The relevant question that remains unresolved is how higher education inﬂuences
the ofﬁcer’s propensity to use force. This question not only concerns the direction of the
impact (a factor the current inquiry was able to examine), but also the means by which
college education affects this propensity (a factor the current inquiry was unable to
examine). In the past, inquiries have approached explanations for why or how higher
education inﬂuences ofﬁcer behavior in three different ways.

With the ﬁrst approach, several studies have proceeded atheoretically (e. g.,
Cascio, 1977; Cunningham, 2006; Fickenauer, 1975). These studies have not directly
discussed why higher education would impact ofﬁcer decision making but instead only
focus on the question of does higher education inﬂuence ofﬁcer behavior. Studies

following the second approach identify some underlying hypotheses, but fail to test them

directly (e. g., Paoline & Terrill, 2007; Sherman & Blumberg, 1981).45 It cannot be said
then that these studies contribute to the empirical status of any hypotheses they present.
The ﬁnal group of previous inquiries includes those that have tested various hypotheses.
Unfortunately, this literature has produced inconclusive results (Dantzker, 1993; Paoline
et al., 2000; Shemock, 1992; Worden, 1990). These studies have operated under a broad
theoretical framework positing that higher education inﬂuences ofﬁcer behavior by
shaping attitudes and values (Worden, 1990). Given the tenuous relationship between
attitudes and behavior in the previous literature (Frank & Brandl, 1991; Muir, 1977;
Worden, 1989), it is not difﬁcult to see why research on higher education and ofﬁcer

behavior is often lacking a theoretical explanation.

 

45 Carter and colleagues (1988: 16-18) compiled a list of these hypotheses, most of which revolve around
differences in attitudes.

78

 

 

Limitations of the Current Inquiry

Several limitations in the present analysis are worth mentioning. First, the manner
in which this analysis attempted to capture police behavior - by dichotomizing particular
outcomes — has the potential to ignore a wide variety of intermediate decisions which
may have been relevant. Because the analysis was concerned only with whether the
behaviors of interest (e.g., arrests or searches) were present during the encounter, the
results may signiﬁcantly underestimate the inﬂuence of higher education on ofﬁcer
behavior. Future researchers may ﬁnd it beneﬁcial to expand the quantiﬁcation of police
behavior beyond simply marking the presence or absence of an act.

Second, similar to previous studies of education and ofﬁcer behavior, the present
inquiry was not capable of testing speciﬁc hypotheses regarding the mechanism by which
higher education inﬂuences ofﬁcer behavior. Because the focus of POPN was not

speciﬁcally on higher education, ofﬁcer education level was captured using a single

eight-category variable.46 Future research endeavors will undoubtedly have to collect
education parameters beyond a single categorical variable. For instance, research on
education should include a multitude of measures to better capture ofﬁcer exposure and
involvement in higher education. Potential measures may include whether the degree was
earned prior to, or during police employment, as well as whether college major,
extracurricular activities, and grade point average matters in some way (See Hudzok,

1978 for more examples and a discussion).

 

46 Undoubtedly this is a limitation which many inquiries face, not to be limited to education. In writing a
literature review concerning the impact of higher education on police behavior, the author is typically
searching for a single regression coefﬁcient within a study that did not focus on ofﬁcer education, but only
included a single measure of education as a control variable. The same could be applied to studies of use of
force, for exanrple.

79

 

 

Third, while a qualitative analysis would provide richer detail as to how higher
education may inﬂuence ofﬁcer decision making, such an approach was ultimately
dismissed for two reasons: lack of an explicit theoretical ﬁamework within the prior
literature and insufﬁcient detail found within the narrative accounts of the POPN data.
More directly, the paucity of literature on possible explanations fails to offer a sufﬁcient
guide as to what to look for within narrative accounts (e.g., certain cues or patterns that
may lead one to better understand the role of higher education). Further, the level of
detail included in POPN observer debrieﬁng sessions was generally sparse, if provided at
all. Hence, with little theoretical guidance and limited narrative descriptions, the ability to
conduct any type of meaningful qualitative analysis was essentially eliminated.
Directions for-Future Research

The potential for research in the area of higher education and policing is both
abundant and exciting. The research to date has mostly centered on whether higher
education has an impact on a variety of ofﬁcer behaviors, and if so, the direction of the
impact. As was demonstrated, the relationship between higher education and ofﬁcer
behavior remains unresolved. The development of hypotheses regarding how and why
higher education inﬂuences ofﬁcer behavior may be helpful in guiding future analyses.

It is important then for future research not to ignore the need for hypotheses
regarding how and why higher education inﬂuences ofﬁcer behavior. Analyses which
produce-ﬁndings regarding the direction of education’s impact on ofﬁcer behavior will be
useful in deciphering unresolved relationships, and those ﬁndings will certainly beneﬁt

from empirically derived, stable explanations. In the past, hypotheses have been

80

 

presented as a means for interpreting results, but have not been tested themselves. Given
this fact, ﬁndings regarding higher education’s impact on ofﬁcer behavior have been
interpreted through any number of hypotheses, each with little empirical support. Rather
than interpreting results through the best explanation, researchers have simply picked
explanations compatible with their ﬁndings. Research on higher education and police
behavior may gain new relevance if previously proposed hypotheses are revisited and
rigorously tested, discarding unsupported hypotheses and moving forward with stronger
explanations.

In order to demonstrate an approach to examining hypotheses presented in the
previous literature, a hypothesis from Carter and colleagues (1988) list of assumed
educational beneﬁts may be used as an example. According to these researchers, one
advantage to hiring college educated ofﬁcers as opposed to non-college educated ofﬁcers
is that higher education “[permits] the individual to learn more about the history of the
country, the democratic process and appreciation for constitutional rights, values and the
democratic form of government” (1988: 16). This particular hypothesis may be reduced
to posit that higher education positively inﬂuences an ofﬁcer’s appreciation of and
commitment to democratic values. It would be possible to test this hypothesis by creating
a scale which measures such appreciation and commitment. One could then test whether
college and non-college educated ofﬁcers differ signiﬁcantly in their attitudes; and if so,
whether such attitudes help explain differences in behavior. If it is found that it does, this
ﬁnding would begin to chip away at the questions of how and why higher education

inﬂuences ofﬁcer behavior as it would provide evidence that higher education impacts

81

 

ofﬁcer behavior by positively inﬂuencing democratic values.

Aside ﬁom testing hypotheses which have been forwarded by police scholars,
research in the ﬁeld of education has produced a large amount of empirical studies
regarding education’s impact on students in general (e. g., Feldman & Newcomb, 1973;
Pascarella & Terenzini, 1991, 2005). Very little has been done to relate this research to
criminal justice actors. What could be gained through a careful examination of education
research are new insights on research methodology in order to isolate the effects of
education on behaviors, and a better understanding of the salient effects of higher
education. As an example, previous research in education illustrates that some salient
effects of college on students are improvements in critical thinking, reﬂective judgment,
and communication skills (Pascarella & Terenzini, 1991, 2005). Police scholars may
consider examining the extent to which college educated ofﬁcers differ from non-college
educated ofﬁcers in these respects. If it is found that they do differ signiﬁcantly,
researchers may examine whether differences in these skills explain any variation in i
ofﬁcer behavior or decision-making. As with the example above, ﬁndings from such an
inquiry would begin to build a better explanation for how and why higher education
inﬂuences ofﬁcer behavior. An examination of higher education and ofﬁcer
communication skills would be especially relevant to the literature on police use of force,
as Muir (1977) posited that communication skills play an important role in the skillful use
of force. This is a falsiﬁable hypothesis that may be tested and such an examination could
shed light on how and why higher education inﬂuences the use of force. Also, armed with

information approximating the areas where one may expect to see higher education have

82

an inﬂuence on ofﬁcer behavior, researchers may wish to examine the extent to which
police training or culture degrades, mediates, or ampliﬁes those effects. The potential for
research in this area is bountiful.

Implications for Policy

In any examination of the impact of higher education on police behavior, the most
prudent policy implication regards whether police departments should include a college
degree as a prerequisite for employment. The ﬁndings of the present analysis alone do
not warrant a reversal of the statement of the National Acaderrrics Panel on Police Policy
and Performance (Skogan & Frydl, 2004) when they found that there was insufﬁcient
evidence to recommend a college education requirement for employment as a police
ofﬁcer. The results of the analysis here do little to change the state of the available
evidence. This is not particularly good news for proponents of higher education (although
it does not represent bad news either). There is simply not enough quality evidence to
determine whether higher education has a desirable effect on police performance.

As noted, this is an area of policy which can beneﬁt immensely from future lines
of research. It may be possible in the future to address whether the effect of college
diminishes over time, as such studies would attest to the potential value of providing
incentives for ofﬁcers to retru'n to school later in their careers. Research may also identify
characteristics of higher education which produce outcomes most beneﬁcial to the police,
such as college major, the impact of the quality of the institution, or enhanced critical
thinking, reﬂective judgment, and communication skills. Police administrators would

beneﬁt from a better picture of what their department could expect to receive should they

83

 

decide to require a college degree of their recruits. While there is much potential for
evidence of a beneﬁcial effect of higher education to be found, scholars and practitioners
also need to remain open to the notion that higher education may be unrelated to some
dimensions of police performance.

In closing, research in the area of higher education and policing has the potential
to produce valuable knowledge regarding the nature of education, but also the
improvement of police performance. It should not be expected that a college education
requirement will provide amelioration of all the intricacies of police discretion. The
enterprise of using scientiﬁc inquiry to improve the practice of policing, however small

an improvement it may be, should not be discounted.

84

Appendix A: Samples, Data, and Direction of Findings in the Available Literature

Table A.l: Samples, Data, and Direction of Findings in Ofﬁcer Education and Attitudes

 

 

Authors Sample Data Type Direction of Finding*_
Srrrith, Locke, & 226 ofﬁcers in the New Rokeach (1960) and +
Walker (1967) York City Police Piven (1961) Scales
Department (NYPD), 104 of
which were enrolled in a
nearby college
Smith, Locke, & 78 officers in the NYPD, 39 Rokeach (1960) and +
Fenester (1970) of which had completed Piven (1961) Scales
baccalaureate degrees
Guller (1972) 63 ofﬁcers with the NYPD, Rokeach Scale (1960) +
32 of which were college
seniors and 31 of which
were college freshmen at the
John Jay College of
Criminal Justice
Dalley (1975) 139 ofﬁcers from the Royal Adorno Authoritarian +
Canadian Mountain Police Scale (1950),
Kerlinger Social
Attitude Scale (1970),
and Role
Interpretation Scale
Miller and Fry (1976a, 136 sworn ofﬁcers from a Miller and Fry null“
1978) California county Professionalism Scale
(1976b)
Weiner (1976) 115 police science students Multi-scale attitudinal null
who were also police survey
ofﬁcers and 396 ofﬁcers in a
nearby department
Roberg (1978) 118 nonsupervisory patrol Rokeach Scale (1960), +
ofﬁvers in Lincoln, supervisor ratings
Nebraska
Worden (1990) Surveys of ofﬁcers and Attitudinal Likert null
citizens in Rochester, St. scales

Buckley, McGinnis, and

Louis, and Tampa-St.
Petersburg

156 Canadian regional

Career attitude survey

Petrunik (1992) police constables

Shemock (1992) 177 patrol ofﬁcers from 11 Attitudinal -
municipal departments in questionnaire
New England and New York

 

85

 

Table A.l Continued: Samples, Data, and Direction of Findings in Ofﬁcer Education and

 

Attitudes
Dantzker (1993) 535 patrol ofﬁcers from Job satisfaction survey -
municipal departments in
Illinois, Texas, and
California
Carlan and Byxbe 235 college students who Vignettes to measure +
(2000) were aspiring police officers punativeness

Paoline, Myers, and
Worden (2000)

and 428 non-criminal justice
majors

Interviews with 398 ofﬁcers
in Indianapolis, Indiana
(Summer 1996) and 246
officers in St. Petersburg,
Florida (Summer 1997)

Ofﬁcer’s conceptions -
of the police role

 

*Note: “+” = college educated ofﬁcers possess more desirable attitudes, “-” = college educated ofﬁcers

ssess less desirable attitudes, “null” = college education has no signiﬁcant effect on ofﬁcer attitudes.
An important point to make is that this direction represents an overall picture of the study ﬁndings. Miller
and Fry (1976) measured many different types of attitudes and their results “ranged from slightly positive,
to nonexistent, to slightly negative” (p. 192).

86

 

 

Table A2: Samples, Data, and Direction of Findings in Ofﬁcer Education and Arrests

 

 

Authors Sample Data Type Direction of Finding*_
Glasgow, Green, & 24 ofﬁcers with the Costa Arrest records +
Knowles (1973) Mesa, California Police
Department
Bozza (1973) 24 oﬂicers with the Costa Arrest records +
Mesa, California Police
Department
Fickenauer (1975) 98 police recruits from Vignettes measuring -
training academics in New discretion in invoking
Jersey formal action
Smith & Klein ( 1983) 24 municipal departments in Independent _ / null¥
Rochester, St. Louis, and observations of
Tampa-St. Petersburg and ofﬁcer encounters
950 observed encounters with suspects
between police and suspects
Worden (1989) Observation of 900 patrol Independent null
shifts by ofﬁcers in 60 observations of
neighborhoods in Rochester, ofﬁcer encounters
St. Louis, and Tampa-St. with suspects
Petersburg during the
summer of 1977
Srrrith & Aamodt (1997) 299 municipal ofﬁcers ﬁ'om Arrest records null
12 departments in Virginia
Brandl, Stroshine, & 800 ofﬁcers ﬁ'om a large Arrest records null
Frank (2001) mid-western municipal
police department

 

*Note: “+” = college educated ofﬁcers more likely to arrest,

arrest, “null” = college education has no signiﬁcant effect on propensity to arrest.

¥Smith & Klein (1983) produced a negative relationship between education and arrests when education was
measured as the entire department’s average education. They produced a null relationship when education

was measured as an individual ofﬁcer’s years of education.

87

“-” = college educated ofﬁcers less likely to

 

Table A.3: Samples, Data, and Direction of Findings in Ofﬁcer Education and Use of

 

 

(2004)

police and suspects
Indianapolis (Summer 1996)
and St. Petersburg (Summer
1997)

observations of ofﬁcer
encounters with
suspects

Force
Authors Sample Data Type Direction of F inding*_
Cohen and Chaiken 1,608 officers in the NYPD Citizen complaints of -
(1972) who were appointed in 1957 force
and still active in 1968
Cascio (1977) 940 ofﬁcers in the Dade Citizen complains of -
County Public Safety excessive force
Department
Inn and Wheeler (1977) 347 ofﬁcers in a large Letters of explanation null
metropolitan police for shooting incidents
department
Milton, Halleck, Seven urban police Ofﬁcer involvement +
Lardner, and Abrecht departments selected by in violent incidents
(1977) population, administration
style, and rates of ofﬁcer-
involved shootings
Hayden (1981) 50 ofﬁcers from a large Vignettes measuring null
northeastern city discretion in deadly
force
Sherman and Blumberg 473 ofﬁcers in the Kansas Investigation reports null
(1981) City Police Department on discharging
ﬁrearms
Binder, Scharf, and Report submitted to the Police shooting -
Galvin (1982) National Institute of Justice records
Worden (1996) 1,528 encounters between Independent +
ofﬁcers and suspects in 60 observations of ofﬁcer
neighborhoods in Rochester encounters with
and Tampa-St. Petersburg suspects
Terrill and Mastrofski 3,116 encounters between Independent -
(2002) police and suspects in observations of ofﬁcer
Indianapolis (Summer 1996) encounters with
and St. Petersburg (Summer suspects
1997)
Aamodt (2004) Meta-analysis of 10 studies Unclear¥ -
‘ on ofﬁcer education and use
of force
Paoline and Terrill 3,356 encounters between Independent - / null€

 

88

 

Table A3 Continued: Samples, Data, and Direction of Findings in Ofﬁcer Education and
Use of Force

Paoline and Tenill 3,356 encounters between Independent -

(2007) police and suspects observations of ofﬁcer

Indianapolis (Summer 1996) encounters with
and St. Petersburg (Summer suspects

1997)
McElvain and Kposowa 186 ofﬁcer-involved Police shooting -
(2008) shootings in the Riverside records

County Sheriffs

Department, California

 

*Note: “+” = college educated ofﬁcers more likely to use force, “-” = college educated ofﬁcers less likely
to use force, “null” = college education has no signiﬁcant effect on propensity to use force.

¥Studies were included in the meta-analysis as long as the effect of education on use of force could be
converted into an effect size, but Aamodt (2004) did not divulge what types of force were examined in
these studies.

€Paoline and Terrill (2004) found that college educated male ofﬁcers used less verbal force, but not
physical force. Education did not have a signiﬁcant impact on use of force by female ofﬁcers.

89

 

Appendix B: Additional Bivarite Tables

Table B. 1: Distribution of Ofﬁcer Characteristics by Education Level (N = 3,356)

 

High school or less Some college, no degree F our-year degree or higher

 

m% m% m%
Ofﬁcer Sex
Female 82 (17%) 288 (20%) 143 (10%)
Male 405 (83%) 1,171 (80%) 1,267 (90%)
Ofﬁcer Race
White 383 (79%) 1,116 (77%) 1,160 (82%)
Non-white 1MQ%) 343 (24%) 250 (18%)

 

90

REFERENCES

Aamodt, MG. (2004). Research in Law Enforcement Selection. Boca Raton, FL:
BrownWalker Press.

Adomo, T.W. (1950). The Authoritarian Personality. New York: Harper and Row.
Allen, MP. (1997). Understanding Regression Analysis. New York: Plenum.

Altemeyer, B. (1988). Enemies of Freedom: Understanding Right— Wing
Authoritarianism. San Francisco, CA: Jossey—Bass Publishers.

Babbie, E. (1995). The Practice of Social Research, Seventh Edition. New York, NY:
Wadswoth.

Bachman, R., & Paternoster, R. (2004). Statistical Methods for Criminology and
Criminal Justice. New York: McGraw Hill.

Baro, A. L., & Burlingame, D. (1999). Law enforcement and higher education: Is there
an impasse? Journal of Criminal Justice Education, 10, 57-73.

Bayley, D. H., & Bittner, E. (1997). Learning the skills of policing. In R. G. Dunharn &
G. Alpert (Eds), Critical issues in policing: Contemporary readings (3rd ed., pp.
114-137). Prospect Heights, IL: Waveland.

Bell, D. J. (1979). The police role and higher education. Journal of Police Science and
Administration 7(4), 467-475.

Berk, RA. (2004). Regression Analysis: A Constructive Critique. Thousand Oaks, CA:
Sage.

Binder, A., Scharf, P., & Galvin, R. (1982). Use of deadly force by police ofﬁcers. Final
Report Submitted to the National Institute of Justice, Grant 79-NI-AX-0134.

Bittner, E. (1970). The Functions of Police in Modern Society. Washington, DC: US.
Government Printing Ofﬁce.

Bozza, C. M. (1 97 3). Motivations guiding policemen in the arrest process. Journal of
Police Science and Administration, 1, 468-475.

Brandl, S.G., Stroshine, M.S., & Frank, J. (2001). Who are the complaint prone ofﬁcers?
An examination of the relationship between police ofﬁcers’ attributes, arrest

activity, assignment, and citizens’ complaints of excessive force. Journal of
Criminal Justice 29, 521-529.

91

 

Breci, M. G., & Simons, R. L. (1987). An examination of organizational and individual
factors that inﬂuence police response to domestic disturbances. Journal of Police
Science and Administration, 15, 93-104.

Brown, LP. (1974). The police and higher education: The challenge of the time.
Criminology 12, 114-124.

Buckley, L.B., McGinnis, J .H., & Petrunik, MG. (1992). Police perceptions of education
as an entitlement to promotion: An equity theory perspective. American Journal
of Police 12(2), 77-100.

Carlan, RE. (2007). The criminal justice degree and policing: conceptual development or
occupational primer? Policing: An International Journal of Police Strategies &
Management 30(4), 608-619.

Carlan, P. E., & Byxbe, F. R. (2000). The promise of humanistic policing: 13 higher
education living up to societal expectation? American Journal of Criminal
Justice, 24, 235-245.

Carte, G. E. (1973). August Vollmer and the origins of police professionalism. Journal of
Police Science and Administration, 1, 274-281.

Carter, D.L., & Sapp, AD. (1990). The evolution of higher education in law
enforcement: Preliminary ﬁndings from a national study. Journal of Criminal
Justice Education 1(1), 59-85.

Carter, D. L., & Sapp, A. D. (1992). College education and policing: Coming of Age. FBI
Law Enforcement Bulletin, January, 8-14.

Carter, D.L., Sapp, A.D., & Stephens, D.W. (1988). Higher education as a bona ﬁde
occupational qualiﬁcation (BF OQ) for police: A blueprint. American Journal of
Police 7, 1-28.

Cascio,W. F. (1977). Formal education and police ofﬁcer performance. Journal of Police
Science and Administration, 5, 89-96.

Cohen, B., & Chaiken, J. M. (1973). Police Background Characteristics and
performance. Lexington, MA: Lexington Books.

Cox, D.R. & Snell, E]. (1989). Analysis ofBinary Data (2” ed. ). London: Chapman and
Hall.

Cunningham, S. (August, 2006). The Florida research Police Chief 73, 20-22.

Dalley, A. F. (1975). University and non-university graduated policemen: A study of

92

police attitudes. Journal of Police Science and Administration, 3, 45 8-468.

Dantzker, M. L. (1993). An issue for policing — educational level and job satisfaction: A
research note. American Journal of Police, 12, 101-1 18.

Decker, L. K., & Huckabee, R. G. (2002). Raising the age and education requirements for
police ofﬁcers: Will too many women and minority candidates be excluded?
Policing: An International Journal of Police Strategies & Management, 25, 789-

802.

Decker, S. H., & Rojek, J. (2002). Saint Louis Metropolitan Police Department trafﬁc
stop patterns. St. Louis: University of Missouri- St. Louis.

DeMaris, A. (1992). Logit Modeling: Practical Applications. Thousand Oaks, CA: Sage.

DeMaris, A. (1995). A tutorial in logistic regression. Journal of Marriage and the Family
5 7, 956-968.

Engel, RS. (2008). A critique of the outcome test in racial proﬁling research. Justice
Quarterly 25, 1-36.

Feldman, K.A., & Newcomb, T.M. (1973). The Impact of College on Students, Volume I:
An Analysis of F our Decades of Research. San Francisco, CA: Jossey-Bass
Publishers.

Fickenauer, J .O. (1975). Higher education and police discretion. Journal of Police
Science and Administration 3(4), 450-457.

Frank, J ., & Brandl, S. (1991). The police attitude-behavior relationship: Methodological
and conceptual considerations. American Journal of Police, I 0, 83-103.

Fogelson, RM. (1977). Big City Police. Cambridge, MA: Harvard University Press.

F yfe, J.J. (1988), “Police use of deadly force: research and reform”, Justice Quarterly,
5(2), 165-205.

Germann, AC. (1957). Law enforcement education and training in the United States.
Police Chief 24(10), 22-28.

Glasgow, G., Green, R., & Knowles, L. (1973). Arrest performance among patrolrnen in
relation to job satisfaction and personal variables. Police Chief 28-34.

Goldstein, H. (1979). Improving policing: A problem-oriented approach. Crime &
Delinquency, 25, 236-258.

93

Gould, SJ. (1996). The Mismeasure of Man. New York: Norton.

Grommon, EL. (2005). The Eﬂect of Socialization and Onset Behavior: An Examination
into Youthful Firearm Oﬂending Patterns in Michigan. Masters Thesis: Michigan
State University.

Guller, LB. (1972), “Higher education and policemen: attitudinal differences between
freshman and senior police college students”, Journal of Criminal Law,
Criminology, and Police Science, Vol. 63, pp. 396-401.

Hanushek, E.A. & Jackson, J.E. (1977). Statistical Methods for Social Scientists. New
York: Academic Press.

Hawley, T. J. (1998). The collegiate shield: Was the movement purely academic? Police
Quarterly 1(3), 35-59.

Hayden, G. A. (1981). Police discretion in the use of deadly force: An empirical study of
the information usage in deadly force decisionmaking. Journal of Police Science
and Administration, 9, 102-107.

Hickman, M. J ., & Reaves, B. A. (2003a). Local police departments 2000. Washington,
DC: Bureau of Justice Statistics.

Hickman, M. J ., & Reaves, B. A. (2006). Local police departments 2003. Washington,
DC: Bureau of Justice Statistics.

Hudzik, J. K. (1978). College education for police: Problems in measuring component
and extraneous variables. Journal of Criminal Justice, 6, 69-81.

Inn, A. & Wheeler, AC. (1977) Individual differences, situational constraints, and police
shooting incidents. Journal of Applied Social Psychology 7(1), 19-26.

Jacobs, J .B., & Magdovitz, SB. (1977). At LEEP’s end?: A review of the Law
Enforcement Education Program. Journal of Police Science and Administration
5(1), 1-18.

Kappeler, V. E., Sapp, A. D., & Carter, D. L. (1992). Police ofﬁcer higher education,
citizen complaints and departmental rule violations. American Journal of Police,

11, 37-54.

Kerlinger, F.W. (1970). A social attitude scale: Evidence on reliability and validity.
Psychological Reports 26, 379-383.

Klinger, DA. (1996). More on demeanor and arrest in Dade County. Criminology 34(1),
61-82.

94

Klockars, CB. (1996). A theory of excessive force and its control. In W. A. Geller & H.
Toch (Eds), Police violence: Understanding and controlling police abuse of force
(pp. 11-29). New Haven, CT: Yale University Press.

Knox, W.E., Lindsay, P., & Kolb, MN. (1993). Does College Make a Diﬂerence?: Long-
term Changes in Activities and Attitudes. Westport, CT: Greenwood Press.

LeDoux, J ., Tully, E., Chronister, J .L, & Gansneder, BM. (1984). Higher education for
law enforcement: Half a century of growth. The Police Chief 51 , 22-24.

Lersch, K. M. & Kunzman, L. L. (2001). Misconduct allegations and higher education in
a southern sheriffs’ department. American Journal of Criminal Justice, 25, 161-
1 72.

Liao, T. F. (1994). Interpreting Probability Models: Logit, Probit, and other Generalized
Linear Models. Thousand Oaks, CA: Sage.

Long, J .S. (1997). Regression Models for Categorical and Limited Dependent Variables:
Analysis and Interpretation. Thousand Oaks, CA: Sage.

Mastrofski, S.A., Parks, R.B., Reiss, A.J. Jr., & Worden, RE. (1995). Community
policing at the street level: A study of the police and the community. Proposal
submitted to the National Institute of Justice. Washington, DC.

Mastrofski, S.D., Parks, R.B., Reiss, A.J., Worden, R.E., DeJong, C., Snipes, J.B., Terrill,
W. (1998). Systematic observation of public police: Applying ﬁeld research
methods to policy issues. Washington, DC: National Institute of Justice. NCJ
172859.

McKelvey, R.D., & Zavoina, W. (1975). A statistical model for the analysis of ordinal
level dependent variables. Journal of Mathematical Sociology 4, 103-120.

McElvain, J .P., & Kposowa, A.J. (2008). Police ofﬁcer characteristics and the likelihood
of using deadly force. Criminal Justice and Behavior 35(4), 505-521.

Meyer, J.W. (1977). The effects of education as an institution. American Journal of
Sociology 83, 55-77.

Michigan State College. (1936). Bulletin of Michigan State College of agriculture and
applied science, Catalog number for 1935-1936, Announcements for 1 936—3 7,
30(9).

Miller, J ., & Fry, L. (1976a). Reexamining assumptions about education and
professionalism in law enforcement. Journal of Police Science and

95

Administration, 4, 187-196.

Miller, J., & Fry, L. (197 6b). Measuring professionalism in law enforcement.
Criminology 14, 401-412.

Miller, J., & Fry, L. (1978). Some evidence on the impact of higher education for law
enforcement personnel. The Police Chief 45, 30-33.

Milton, C.M., Halleck, J .W., Lardner, J ., & Abrecht, G.L. (1977). Police Use of Deadly
Force. Washington, DC: Police Foundation.

Muir, W. K., Jr. (1977). Police: Streetcomer Politicians. Chicago: University of Chicago
Press.

National Advisory Committee on Criminal Justice Standards and Goals (1973). A
National Strategy to Reduce Crime. Washington, DC: US. Government Printing
Ofﬁce.

National Commission on Law Observance and Enforcement (1931). Report on Police.
Washington, DC: US. Government Printing Ofﬁce.

National Planning Association. (1978). A nationwide survey of law enforcement criminal
justice personnel needs and resources. Washington, DC: United States
Government Printing Ofﬁce.

Padover, SK. (1939). Thomas Jeﬂerson on Democracy. New York, NY: New American
Library.

Parnpel, RC. (2000). Logistic Regression: A Primer. Thousand Oaks, CA: Sage.
Paoline, E. A., III, Myers, S. M., & Worden, R. E. (2000). Police culture, individualism,
and community policing: Evidence from two police departments. Justice

Quarterly, 1 7, 575-605.

Paoline, EA, 111, & Terrill, W. (2004). Women police ofﬁcers and the use of coercion.
Women & Criminal Justice 15(3/4), 97-117.

Paoline, E.A., HI, & Terrill, W. (2005). The impact of police culture on trafﬁc stop
searches. Policing, An International Journal of Police Strategies & Management

28(3), 455-472.

Paoline, E. A., HI, & Terrill, W. (2007). Police education, experience, and the use of
force. Criminal Justice and Behavior, 34, 179-196.

Parks, R.B., Mastrofski, S.D., DeJong, C., Gray, MK. (1999). How ofﬁcers spend their

96

time with the community. Justice Quarterly 16(3), 483-518.

Pascarella, E.T., & Terenzini, PT. (1991). How College Aﬂects Students: Insights from
Twenty Years of Research. San Francisco: J ossey-Bass.

Pascarella, E.T., & Terenzini, PT. (2005). How College Aﬂects Students (Vol. II): A
Third Decade of Research. San Francisco: Jossey-Bass.

Piven, (H. (1961). Professionalism and Organizational Structure. Unpublished doctoral
dissertation, Columbia University School of Social Work.

Riksheirn, E. C., & Chermak, S. M. (1993). Causes of police behavior revisited. Journal
of Criminal Justice, 21, 353-382.

Roberg, R. R. (1978). An analysis of the relationships among higher education, belief
systems, and job performance of patrol ofﬁcers. Journal of Police Science and
Administration, 6, 336-344.

Roberg, R., & Bonn, S. (2004). Higher education and policing: Where are we now?
Policing: An International Journal of Police Strategies & Management, 27, 469-
486.

Rokeach, M. (1960). The Open and Closed Mind: Investigations into the Nature of Belief
Systems and Personality Systems. New York: Basic Books.

Schaefer, J .A., Carter, D.L., & Katz-Bannister, A. (2004). Studying trafﬁc stop
encounters. Journal of Criminal Justice 32, 159-170.

Sherman, L.W. (197 8). College education for police. The reform that failed?
International Journal of Police Strategies and Management I, 32-38.

Sherman, L. W. (1980). Causes of police behavior: The current state of quantitative
research. Journal of Research in Crime and Delinquency, I 7, 69-100.

Sherman, L. W., & Blumberg, M. (1981). Higher education and police use of deadly
force. Journal of Criminal Justice, 9, 317-331. '

Sherman, L.W., & McLeod, M. (1979). Faculty characteristics and course content in
college programs for police ofﬁcers. Journal of Criminal Justice 7(3), 249-267.

Shemock, SK. (1992). The effects of college education on professional attitudes among
police. Journal of Criminal Justice Education 3(1), 71-92.

Skogan, W. & Frydl K. (2004). Fairness and eﬂectiveness in policing: the evidence,
National Research Council of the National Academies. Washington, DC: The

97

National Academies Press.

Skohrick, J .H. (1964). A sketch of the policeman’s working personality. In T. Newbum
(Ed.) Policing: Key Readings, Uffclume, U.K.: Willan Publishing.

Skolnick, J.H., & Fyfe, J .J . (1993). Above the Law: Police and the Excessive Use of
F orce. New York, NY: The Free Press.

Smith, A., Locke, B., & Fenster, A. (1970). Authoritarianism in policemen who are
college graduates and non-college police. Journal of Criminal Law, Criminology,
and Police Science, 61, 313-315.

Smith, A., Locke, B., & Walker, W.F. (1967). Authoritarianism in college and non-
college oriented police. Journal of Criminal Law, Criminology, and Police
Science, 58, 128-132.

Smith, D. A., & Klein, J. R. (1983). Police agency characteristics and arrest decisions. In
G. P. Whitaker and C. D. Phillips (Eds.) Evaluating performance of criminal
justice agencies, 63-97.

Smith, M. R., & Petrocelli, M. (2001). Racial proﬁling? A multivariate analysis of police
trafﬁc stop data. Police Quarterly, 4(1), 4 —27.

Smith, S. M., & Aamodt, M. G. (1997). The relationship between education, experience,
and police performance. Journal of Police and Criminal Psychology, 12, 7-14.

Spano, R. (2003). Concerns about safety, observer sex, and the decision to arrest:
Evidence of reactivity in a large-scale observational study of police. Criminology
41(3), 909-932.

Stephens, G. (1976). Criminal justice education: Past, present, and ﬁlture. Criminal
Justice Review 1, 91-120.

Terrill, W. (2001). Police Coercion: Application of the Force Continuum. New York:
LFB Publishing.

Terrill, W., & Mastrofski, S. D. (2002). Situational and ofﬁcer based determinants of
police coercion. Justice Quarterly, 19, 215-248.

Terrill, W. & Paoline, EA. (2007). Nonarrest decision making in police-citizen
encounters. Police Quarterly 10(3), 308-331.

Trojanowicz, R. (1994). The future of community polcing. In D. Rosenbaum (Ed.) The

Challenge of Community Policing: Testing the Promises. Thousand Oaks, CA:
Sage.

98

Truxillo, D. M., Bennet, S. R., & Collins, M. L. (1998). College education and police job
performance: A ten-year study. Public Personnel Management, 27, 269-280.

Walker, S. (1977). A Critical History of Police Reform: The Emergence of
Professionalism. Lexington, MA: Lexington Books.

Weiner, N.L. (1976). The educated policeman. Journal of Police Science and
Administration 4(4), 450-457.

Westley, WA. (1970). Violence and the Police: A Sociological Study of Law, Custom,
and Morality. Cambridge: MIT Press.

Wilson, H. (1999). Post-secondary education of the police ofﬁcer and its effects on the
ﬁequency of citizen complaints. Journal of California Law Enforcement, 33, 3-
1 0.

Wilson, J .Q. (1975). Thinking About Crime. New York: Basic.

Wilson, J .Q., & Kelling, G.L. (1982). Broken windows. The Atlantic Monthly 249(3), 29-
38.

Winship, C. & Mare, RD. (1984). Regression models with ordinal variables. American
Sociological Review 49(4), 512-525.

Worden, R. E. (1989). Situational and attitudinal explanations of police behavior: A
theoretical reappraisal and empirical assessment. Law & Society Review 23(4),
667-71 1.

Worden, R. E. (1990). A badge and a baccalaureate: Policies, hypotheses, and further
evidence. Justice Quarterly, 7, 565-592.

Worden, R. E. (1996). The causes of police brutality: Theory and evidence on police use
of force. In W. A. Geller & H. Toch (Eds), Police violence: Understanding and
controlling police abuse of force (pp. 23-51). New Haven, CT: Yale University
Press.

Xu, Y., F eidler, M.L., & Flaming, K.H. (2005). Discovering the impact of community
policing: The broken windows’ thesis, collective efﬁcacy, and citizens’ judgment.
Journal of Research in Crime and Delinquency 42(2), 147-186.

Yerkes, RM. (1921). Psychological examining in the United States Army. Memoirs of
the National Academy of Sciences 15, 1-890.

99

   

IllllllllllIllllllllIlllllllllllllllllllllllllllllllllllllll