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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 c:/5lRC/DateDue.indd-p.15

 

THE IMPACT OF NEIGHBORHOODS ON A VARIETY OF POLICE ACTIONS

By

Michael Kevin Gray

A DISSERTATION
Submitted to
Michigan State University

in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Criminal Justice

2004

ABSTRACT
THE IMPACT OF NEIGHBORHOODS ON A VARIETY OF POLICE ACTIONS
By

Michael Kevin Gray

Using conﬂict theory and the benign neglect hypothesis, a variety of police services are
examined to determine if police activities differ by neighborhood. While previous
research has examined contextual level inﬂuences of police behavior, prior examinations
have been limited in terms of the activities examined (typically limited to law
enforcement activities) and the contextual characteristics included. The current research
examines ofﬁcer initiated order maintenance, law enforcement, and service actions of the
police across neighborhoods. Neighborhood characteristics, derived from a conﬂict
theoretical approach, include concentrated disadvantage and percent homeownership.
Results indicate that ofﬁcer level variables are more consistently related to the three types
of ofﬁcer initiated activities than are neighborhood characteristics. Differences were
found for how much time officers spend inside assigned beats based on neighborhood
characteristics as well as how much order maintenance is conducted. Findings are

discussed and placed within the larger context of existing research of police behavior.

Dedicated to my father, John Gray.

iii

ACKNOWLEDGMENTS

I would like to take this opportunity to thank several people without whom I
would not have ﬁnished this dissertation. First, I would like to thank the members of my
committee: Chris Maxwell, Kevin Ford, John McCluskey, and Ed McGarrell. I
appreciate your time and input into this paper; your suggestions and insights have made
this research much stronger.

I would like to extend particular thanks to the chair of my dissertation, Chris
Maxwell. You have helped me throughout my time at Michigan State - from my ﬁrst
years working for you (looked in your basement with no heat), through job searches,
conferences, until now when I am ﬁnally ﬁnished with the dissertation. I sincerely
appreciate all the advice, assistance, and support you’ve provided.

I would also like to thank my family and friends — for their support,
encouragement, and their patience. I dedicated this dissertation to my dad because, from
an early age, he always told me that I could accomplish anything I put my mind to. Of
course, some things take longer than others, but I’m ﬁnally done! I appreciate the
encouragement and patience of my mother especially — even after hearing “I don’t want to
talk about it!” time and time again when asking how the dissertation was going. I have
been blessed with such a wonderful and supportive family.

I have also been fortunate to have had so many good friends who have encouraged
me along the way. My MSU friends are among the best people I have ever met: Meghan
Stroshine, Stephen Jefferson, Hao Lu, Amanda Robinson, Roni Mayzer, Barb Koons,

Heather Hansel, Gwen Bramlet - to name a few. My friends in Terre Haute have had to

iv

listen to me rant and rave and fret more than anyone should have to, thank you to Jenny
Krozier, Kelly and Joe Spitznagel, Jennifer Boothby, and Jon Schaefer.

I would also like to thank my colleagues at Indiana State University.

Table of Contents

List of Tables ......................................................... viii
List of Figures .......................................................... x
Chapter 1 .............................................................. 1
Introduction ...................................................... 1

The Importance of Place ............................................ 9

Place and the Police ............................................... 13

A Conﬂict Theoretical Approach to Police Behavior ..................... 15
Benign Neglect and the Underclass Hypothesis ......................... 21
Summary ....................................................... 24
Chapter 2 ............................................................. 26
Introduction ..................................................... 26
Individual Level Inﬂuences of Police Behavior .......................... 27
Contextual Level Inﬂuences of Police Behavior ......................... 29
Qualitative Accounts of Differential Police Behavior by Neighborhood ...... 31
Quantitative Examinations of Neighborhood Inﬂuences on Ofﬁcer Behavior . . 35

Levels of Analysis ................................................ 44
Contextual Variables .............................................. 45
Summary ....................................................... 46
Chapter 3 ............................................................. 49
Data ........................................................... 49

Study Sites ...................................................... 54
Indianapolis, Indiana ........................................ 54

St. Petersburg, Florida ....................................... 55

Unit of analysis .................................................. 55
Measurement of the Dependent Variables .............................. 57
Measurement of the Independent Variables ............................. 63
Ofﬁcer Level Variables ...................................... 63
Neighborhood Variables ..................................... 67

Analytic Techniques .............................................. 71
Hypotheses ...................................................... 76
Officer Level Variables ...................................... 76
Neighborhood Level Variables ................................ 79

vi

Chapter 4 ............................................................. 81

Analyses ........................................................ 81
Bivariate Correlations Among Predictor Variables ....................... 81
Preliminary Analyses — Free Time and Time in Beat ..................... 84

Free time ................................................. 85

Time in Beat ............................................... 88

Dependent Variables - Distributions .................................. 94
Bivariate Correlations - Predictor and Outcome Variables ................ 100
Analysis of Variances ............................................ 102
Multivariate Analyses ............................................ 110
Ordinary Least Squares ..................................... 110

Hierarchical Linear Modeling ................................ 112

Chapter 5 ............................................................ 120
Discussion and Conclusions ....................................... 120
Interpretation and Discussion of the Findings .......................... 122

Order Maintenance ......................................... 130

Legalistic ................................................ 131

Service ...................................... i ............ 132

Bivariate and Multivariate Analyses ................................. 135

HLM .......................................................... 141
Limitations of the Current Research ................................. 148

Future Research ................................................. 152
Appendix A. Contextual Variables - Previous Research ........................ 157
Appendix B. Categorization and Distribution of the Three Outcome Measures ..... 164

APPENDIX C. Contextual Information From Citizen Surveys - Original Analyses . . 169

REFERENCES ....................................................... l 83

vii

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Table 7

Table 8.

Table 9.

Table 10

Table 11

Table 12

Table 13

Table 14

Table 15

Table 16

Table 17

Table 18

Table 19

Table 20

Table 21

LIST OF TABLES

Event Type by Study Site ....................................... 51
Outcome Variables - and Variables Used in Calculations .............. 62
Ofﬁcer Characteristics ......................................... 65
Neighborhood Characteristics ................................... 69
Disadvantage Factor ............................................ 70

Zero-Order Correlation Coefﬁcients and Tolerance Statistics: Ofﬁcer

Level Variables ............................................ 83

Zero-Order Correlation Coefﬁcients of Neighborhood Variables ....... 84
Bivariate Correlations - Free Time ............................... 86
One-way ANOVA - Free Time .................................. 87
OLS Regression Results - Free Time .............................. 88
Bivariate Correlations - Time in Beat ............................. 90
One-way ANOVA - Time in Beat ................................ 91
OLS Regression Results - Percent of Time in Beat .................... 93
Inter-correlations Among Dependent Variables ..................... 100
Bivariate Correlations: Ofﬁcer Level and Dependent Variables ........ .101

Bivariate Correlations: Neighborhood Level and Dependent Variables . . 102

One-way AN OVA by Officer Characteristics - Order Maintenance . . . . . 103

One-way AN OVA by Officer Characteristics - Legalist .............. 105
One-way ANOVA by Ofﬁcer Characteristics - Service ............... 107
One-way AN OVA - Site and Outcome Measures ................... 1 10
OLS Regression Results ....................................... 111

viii

Table 22

Table 23

Table 24

Table 25

Table A

Table B]

Table B.2

Table 8.3

Table C]

Table C.2

Table C.3

HLM - Order Maintenance .................................... 1 15
HLM - Legalist ............................................. 116
HLM - Service .............................................. 117
Signiﬁcant Bivariate and Multivariate Findings .................... 136
Contextual Variables - Previous Research. ........................ 157
Order Maintenance Events ......................... 164
Legalist Events ........................................... 165
Service Events ........................................... 167
Original Neighborhood characteristics .......................... 171
Zero-order Correlation Coefﬁcients of Neighborhood Variables ..... 178
Correlation Coefficients, Neighborhood Level Variables ........ 181

ix

Figure 1
Figure 2
Figure 3
Figure 4

Figure 5

LIST OF FIGURES

Histogram — Time on order maintenance events ..................... 96
Histogram -- Time on legalist events .............................. 96
Histogram -— Time on service events .............................. 97
Log transformed time on legalist events ........................... 98
Log transformed time on service events ........................... 99

Chapter 1
Introduction

The importance of considering context in explaining behavior is evidenced by the
number of disciplines that include context as a key construct of interest. Considerable
research addressing innumerable topics has been conducted in ﬁelds such as psychology,
sociology, and criminology examining contextual level inﬂuences. Both social and
physical conditions of neighborhoods have been shown to be associated with emotional
and behavioral outcomes of residents (Austin, Furr, and Spine, 2002). Brodsky,
O’Campo, and Aronson (1999) note the important role that community level resources
play in the development, treatment, and prevention of various mental and physical public
health concerns. Brodsky et al. (1999) identify the improvement in mental health and
well-being in relation to environmental design. According to Oxford (1992) person-in-
context is the hallmark of the community psychological approach; what are termed
“healthy communities” are seen to play a pivotal role in the physical and psychological
well-being of its residents and increase the ability to resist physiological and
psychological problems (Chavis and Newbrough, 1986).

Place and place characteristics have also been examined in efforts to identify
determinants of police actions. Explaining and identifying the inﬂuences of police
behavior have been goals of police researchers for decades. Approaches have followed
several dimensions — individual, situational, organizational, and contextual level
inﬂuences. Using conﬂict theory and the benign neglect hypothesis, differential police
initiated activities are examined to determine if neighborhood contextual level inﬂuences

of police behavior can be identiﬁed.

Conﬂict theory suggests that the police, as an agent of the state, behave in a more
punitive manner towards members of groups who pose a threat to the established order
and in places where other forms of social control are weak or nonexistent. As a result of
this, differential patterns of police behavior are expected — based on neighborhood
characteristics such as income levels.

A second theoretical approach that explains differential police behavior is the
benign neglect hypothesis. This approach suggests that the police operate towards certain
areas with a type of “benign neglect” (Liska and Chamlin, 1984) — ignoring behaviors that
may warrant attention if seen in other areas and providing fewer and lower quality
services to some groups than would be provided to others. Benign neglect suggests that
poor or minority areas would receive fewer and lower quality of public services — such as
police services.

Both of these theoretical approaches underscore the differential behavior of the
police towards different groups in society. Both of these approaches recognize the
theoretical importance of place and place characteristics in explaining police behavior.

This research endeavor examines whether context is a relevant factor when
considering the delivery of a variety of police services. Speciﬁcally, police provision of
services and the types of services provided to different neighborhoods are examined.
While no clear relationship has been shown in terms of the provision of traditional
reactive policing activities for different neighborhoods (based on racial and economic
characteristics) (Mladlenka and Hill, 1978; Ostrom, 1983), other forms of policing,
namely, order maintenance and service activities have yet to be examined for potential

differential distribution by neighborhood. The ﬁ'equency with which police engage in

such non-crime services involving order maintenance and service activities has long been
recognized (Mastrofski, 1983) and is substantial (Bayley and Mendelsohn, 1969; Walker
and Katz, 2002; Wilson, 1968). This line of research is important for several reasons,
including the signiﬁcant amount of time that police are engaged in non-law enforcement
activities. This line of research is also important considering the potential for inequities
in the delivery of police services. Finally, this line of research is important due to the
prevalence of community policing reforms being adopted across the country and
principles inherent in these reforms, such as increased discretion and neighborhood
speciﬁc approaches.

During the past 15 years, more and more police departments across the country
have adopted principles of community policing, especially since the 1994 passage of the
President’s Crime Bill which included provisions for hiring 100,000 more community
police ofﬁcers. The adoption of community policing constitutes a major reorientation of
departmental patrol efforts and operational strategies. Community policing as a reform
movement has several principles that differentiate it from previous police practices, four
of which are particularly relevant for the current study. These differences include (1)
calls for a different relationship between the police and the public, (2) expansion and/or
recognition of non-law enforcement functions, (3) place-speciﬁc approaches, and (4)
increases in ofﬁcer discretion (Cordner, 1995- cited in Novak, Frank, Smith and Engel,
2002).

The relationship and partnership between the police and the public is one
signiﬁcant difference between traditional police strategies and community policing. The

professionally remote relationship with citizens that departments pursued during the

professional era of policing, is seen under a community policing philosophy as a
hindrance to effective policing. Community policing attempts to address and change the
relationship between the police and public - calling for working
relationships/partnerships between the two. The professional remoteness which was
sought by previous police reforms ultimately led to relationships characterized by
isolation, alienation, and mistrust. Isolation led to resentment and misunderstanding and
a lack of support by the community of the police (Trojanowicz and Carter, 1988). For
community policing, this support and cooperation is seen as a necessity. Community
policing reforms recognize the role of the public in maintaining order in their
neighborhoods and speciﬁes changes in the role of the police with respect to duties,
activities, and general operating policies. As Sparrow, Moore, and Kennedy (1990) note,
“police are rediscovering that ordinary people and communities are the ﬁrst line of
defense in controlling crime and fear” (p. 46).

In addition to a change in the relationship between the police and the public,
community policing efforts recognize and embrace a broader police mandate.
Throughout the majority of the 20th century, police agencies emphasized the law
enforcement/crime control aspects of policing. With community policing, the goals and
focus of police action is expanded to include such considerations as fear of crime,
disorder, and quality of life issues; service behaviors and order maintenance functions are
recognized as essential police functions. While traditional police models have
emphasized the crime ﬁghting/law enforcement role of police, community policing

efforts accommodate the breadth of functions that the police routinely handle.

Another substantive change from traditional policing, is that community policing
calls for a place-speciﬁc approach, recognizing that different areas have different
problems and needs. More tailored or neighborhood-speciﬁc approaches in terms of
police service delivery replace the traditional “one size ﬁts all” approach. With this
tailored approach, one potential problem is that differential and unequal service delivery
by neighborhoods could manifest; the quality of services which police provide to
different areas could be different. The impact of community policing reorientation
regarding the potential for differential service delivery is not known and is examined by
this research.

A ﬁnal and signiﬁcant change involves ofﬁcer discretion. While most police
reform during the 20th century aimed (explicitly or implicitly) to reign in ofﬁcer
discretion, as a reform, community policing is unique in its call for increased ofﬁcer
discretion (Novak, et al., 2002). Speciﬁcally, ofﬁcers under a community policing
philosophy are given more autonomy to determine how their time is spent, with whom
they interact, and the targets of their actions. The outcome of this increased discretion is
only partially coming to light and indicates that ofﬁcers freed from traditional 911-
response driven contacts with citizens, come into contact with signiﬁcantly different
groups (Parks, Mastrofski, DeJong, and Gray, 1999).

Police discretion has been a focus of research for some time, involving the
examination of police behaviors in terms of organizational, situational, individual ofﬁcer,
and neighborhood or contextual level inﬂuences. The ﬁndings from previous research
regarding these inﬂuences on police discretionary behaviors have yielded some consistent

ﬁndings, such as the inﬂuence of suspect demeanor, and some inconsistent ﬁndings, such

as the race of the offender (see for example, Riksheim and Chermak, 1993, and Sherman,
1980). One possibility is that inconsistencies of prior research can be explained by
considering the different contexts in which decisions, such as arrest, occur (Smith, 1984).
This notion is highlighted by Chambliss and Seidman (1971) who note that the total
social milieu in which people operate must be examined to determine the way in which
rules actually operate.

While some research has been conducted on the inﬂuences of neighborhood
characteristics and police discretionary actions, the primary focus of these studies has
been limited to police decisions to arrest and police use of force (See for example, Brown
and Warner, 1992; Crank, 1990; Liska, Chamlin, and Reed, 1985). While important,
these two areas of police action leave much left to be examined. A further consideration
is that different activities involve varying levels of discretion and that arrest decisions and
use of force practices arguably involve the least amount of discretion. Bayley and
Mendelsohn (1969) note that some events entail clear cut enforcement decisions while
others are more ambiguous and more discretionary. Wilson (1968) identiﬁes the type of
activity (law enforcement or order maintenance) as a key factor affecting the amount of
discretion that an ofﬁcer has and notes that order maintenance functions involve
exercising substantial discretion.

As Brooks (1997) notes, the fewer the rules about handling situations, the more
discretion ofﬁcers have. There are numerous rules concerning situations when arrests
should be made, as well as when the use of force is appropriate.

Walker and Katz (2002) also acknowledge the considerable discretion that

ofﬁcers possess when handling non-crime incidents and Kelling (1997) refers to order

maintenance activities as highly discretionary. The question remains, how might
neighborhood context inﬂuence situations in which discretion is higher?

Traditional police practices have served to provide a seemingly egalitarian
distribution of police services (Parks et al., 1999). Speciﬁcally, equity in services has
been examined in relation to the allocation of resources and response time (Ostrom, 1983)
and the research conducted in these areas suggest no relationship between these outcome
measures and racial or economic indicators (Maxﬁeld, Lewis, and Szoc, 1980; Mladlenka
and Hill, 1978; Ostrom, 1983). Mastrofski (1983) notes the general lack of support for
the underclass hypothesis (the idea that members of the lower class receive fewer public
services) in regard to the delivery of these police services.

Community policing may change the egalitarian nature of police service delivery.
Principles inherent in community policing drive this potential, namely an appreciation of
non-law enforcement type activities, place-speciﬁc approaches, and increased discretion
in terms of how ofﬁcers spend their time. This paper examines the degree to which
discretionary police behaviors are inﬂuenced by neighborhood context. As noted by
Smith (1987): “Police patrol both people and places, hence characteristics of both people
and places may inﬂuence police behavior” (p. 780).

Explaining and identifying the factors which inﬂuence police behavior has been a
target of police researchers for some time; various approaches have been used in such
efforts. The current study attempts to add to the current knowledge of inﬂuences of
police behavior by examining a variety of police activities and by examining contextual
level variables. While not negating the potential inﬂuences of ofﬁcer-level, situational,

or organizational factors, environmental or contextual factors have often gone

unexarnined in explaining differential police practices. This is especially true when one
considers the examination (or lack thereof) of activities outside the realm of traditional
reactive policing — considering the broad provision of services in which police are
engaged, namely order maintenance and service activities.

For the remainder of Chapter One, a theoretical framework is provided for the
current study. Speciﬁcally, a conﬂict theoretical approach is provided as a foundation for
explaining and understanding the potential differences in the provision of police services
(law enforcement, order maintenance, and service activities) to different neighborhoods.

In Chapter Two, previous research examining neighborhood and contextual level
inﬂuences of police behavior is examined. The state of what we know about
neighborhood level inﬂuences is presented as well as how this information was gained.
The body of prior research is examined along a several dimensions, including the speciﬁc
contextual variables included and the level of aggregation (e.g., neighborhood, police
beat, city). The current study is placed within a broader context of existing scholarship
conceming the inﬂuences of police behavior. The current study broadens the knowledge
base by examining neighborhood level inﬂuences on police practices especially non-law
enforcement activities.

Chapter Three describes the methodology for the current study. Data collected as
part of the Project on Policing Neighborhoods (POPN) are used for the current analyses.
Chapter Three begins with a discussion of the original study and how the data were
gathered. Next, the neighborhood level variables used for the current study are outlined.

In addition, speciﬁc hypotheses about the relationship between these neighborhood level

variables and police behavior are provided. Finally, the data analytic techniques to be
employed are outlined in Chapter Three.

In Chapter Four, the results from the analyses outlined in Chapter Three are
presented. Signiﬁcant ﬁndings are highlighted.

In Chapter Five, a summary of the results is provided and then discussed.
Implications are also discussed, as are the limitations of the current study and avenues for

future research.

The Importance of Place '

Numerous disciplines within the social and behavioral sciences have addressed
the importance of place in affecting human behavior. Sociological examinations have
included studies identifying neighborhood level effects on such topics as sense of
community (Brodsky et al., 1999; Kingston, Mitchell, Florin, and Stevenson, 1999;
Prezza, Amici, Roberti, Tedeschi, 1999) and feelings of attachment (Buttel, Martinson,
and Wilkening, 1979; Kasarda and Janowitz 1974; Oropesa, 1989; Sampson, 1998). One
interesting example of this line of research was conducted by Kuo, Sullivan, Coley, and
Brunson (1998) who discovered that the level of vegetation in common spaces predicted
the use of these common spaces and the social ties that subsequently developed.

Neighborhood satisfaction has also been shown to be inﬂuenced by the physical
and social characteristics of neighborhoods (Fried, 1982; Herting and Guest, 1985; St.

John ct al., 1986; White, 1985). Several researches have determined that the physical

 

' Place is used to denote environmental or social characteristics of a geographical location. References in
the extant literature utilize inconsistent terminology in this regard. Context could be used to indicate
several dimensions but is used for the current study to denote characteristics of geographic locations.

9

characteristics of a community have signiﬁcant impacts on residents’ satisfaction (Fried,
1982; Herting and Guest, 1985; St. John et al., 1986; White, 1985). Although community
attractiveness is necessarily subjective (White, 1985), residents’ perception of physical
characteristics have been shown to explain satisfactiOn. These include such things as
general physical appearance, beauty, housing quality, access to nature, neatness, and
home ownership. In these studies, the physical characteristics of the communities were
found to be among the most powerful predictors of community satisfaction. Mobility is
identiﬁed by Baldassare (1986) as one important ramiﬁcation of neighborhood
satisfaction; dissatisfaction among residents results in more geographic mobility.

In other studies, the role of these positive community attributes (e.g., satisfaction)
are examined to determine their effect on various outcomes. For example, examinations
have been conducted to determine the effects Of positive community attributes on citizen
participation in organizations, such as community groups (Arcury, Austin, Quandt,
Saavedra, 1999; Chavis and Wandersman, 1990; Howell et al., 1999; Julian, Reischl,
Carrick, and Katrenich 1997) and other collective action efforts (Haeberle, 1989; Hyman
and Wright, 1971; McPherson, 1981; Perkins et al., 1990; Skogan and Lurigio,1992).
Community characteristics have been shown to function as a catalyst for civic
participation in change — improving an area’s ability to combat and resist problems
(Chavis and Pretty, 1999; Prezza et al., 1999).

Research in the ﬁeld of Criminal Justice has also long recognized the importance
of settings and the inﬂuence of contextual level characteristics. Criminological research
on crime (Sampson and Groves, 1989; Sampson and Raudenbush, 2001; Shaw and

McKay 1942/1969), victimization (Cohen and Felson, 1979; Sampson, 1986), fear of

10

crime (Austin et al., 2002; Greenburg, 1986; Ross and J ang, 2000), attitudes towards the
police (Cao, Frank, and Cullen, 1996; Reisig and Parks, 2000; Sampson and Jeglum-
Bartusch, 1998; Weitzer, 1999) and recidivism (Gottfredson and Taylor, 1986) have all
illustrated the inﬂuence of place and place characteristics.

The interest of criminologists in place can be traced to the works of Social
Ecologists such as Park, Burgess, and McKenzie. These early social ecologists stressed
an ecological focus on the process of spatial development. McKenzie (1926) describes a
community as “an ecological distribution of people and services in which the spatial
location of each unit is determined by its relation to all other units” (p. 312). This
approach adapts a model from the physical sciences, speciﬁcally plant ecology, and
applies it to the evolution of the city. It is the spatial distribution of the city which
determines the interactions and relations among people and units comprising the city
(McKenzie, 1926). According to McKenzie (1926), social organization accommodates
itself to spatial dimensions; as areas expand, social disorganization (e.g., crime, disorder,
disease) increases and relationships are increasingly strained.

In one of the most inﬂuential examinations of ecology and crime, Shaw and
McKay (1942/1969) explained higher delinquency rates in certain parts of Chicago —
speciﬁcally, identifying what were termed transition zones. Areas characterized as
socially disorganized were found to have higher rates of delinquency. These areas
maintained high delinquency rates despite signiﬁcant shifts in residential populations
(Shaw and McKay 1942/1969).

In a cursory review of neighborhood-level research, Byme and Sampson (1986)

identify several community characteristics that have been shown to have a positive

11

relationship with crime and/or delinquency. These characteristics include percent
nonwhite, proportion of younger males, housing density, and mobility. At the city-level,
research has also linked crime and delinquency to physical characteristics such as density,
crowding, population, and population change, as well as aggregate residential
characteristics such as ethnicity, age composition, economic factors, and family
composition (Byme and Sampson, 1986).

Wilson and Kelling (1982) argue that neighborhood disorder and crime are
inextricably linked. In their seminal piece, “Broken Windows” Wilson and Kelling
(1982) suggest that neighborhood disorder leads to fear of crime which in turn leads to
isolation among citizens and eventual withdrawal from the community. This isolation
and withdrawal reduces the informal social controls that exist, and begets more disorder,
more crime, and more serious crime (Wilson and Kelling, 1982). The popularity of this
line of reasoning is evidenced by the intense focus which community policing programs
have received and the adoption of community policing principles by police departments
across the country.

Examinations have also been conducted examining contextual characteristics and
victimization. In describing a routine activities approach to explaining crime and
victimization, Cohen and Felson (1979) note that physical characteristics of dwellings, as
well as architectural and environmental designs impact victimization — speciﬁcally,
through target suitability.

Non-physical contextual characteristics have also been examined. Sampson
(1986) illustrated how the nature of family structures (i.e., divorced/separated and female

headed households) affected victimization rates. Areas with more dissolved families had

12

higher victimization rates, according to Sampson (1986) because of lower levels of social
controls exhibited in these areas. The inﬂuence of structural characteristics of
neighborhoods as well as levels of informal social control exerted was later examined and
found to signiﬁcantly affect both levels of crime and levels of disorder in a neighborhood
(Sampson, Raudenbush, and Earls, 1997; Sampson and Raudenbush, 2001).

McGahey (1986) discusses economic factors and their inﬂuence on the stability of
an area which impacts the informal social controls that develop and resulting
susceptibility of a neighborhood to crime. With regard to recidivism, Gottfredson and
Taylor (1986) conclude that the interaction effects between environmental factors and
offender characteristics are important areas to be considered.

As described, the inﬂuence of place and contextual characteristics has been found
across multiple disciplines and used to explain a variety of outcomes and behaviors.
Characteristics and inﬂuences of place have received special attention by researchers and
practitioners in the area of crime and justice as a broader understanding of crime and
delinquency has been sought. Program development and implementation to address
problems associated with crime, at least implicitly, recognize the importance of context.
The inﬂuence of place and contextual characteristics on the day-to-day operations of

police is the topic of the current research.

Place and the Police
In addition to numerous other examinations, the inﬂuence of context on behavior
has also been a focus of examinations conducted involving police discretionary

behaviors. The idea that police behave differently and provide different services in

13

different neighborhoods is neither a novel nor unexamined notion. Sherman (1986) states
that there is signiﬁcant variation in policing across cities and within cities across
neighborhoods. Previous research (discussed in detail in Chapter 2) illustrates the
variation of police behaviors for activities such as arrest and use of force by considering
neighborhood characteristics.

There are two general approaches found when discussing the operating styles of
police and variations of behavior by type of neighborhood, both of which can be
supported with reference to the extant policing literature. One approach is founded upon
the idea that police act more formally in areas characterized as poorer or higher
percentage minority. More formal actions including more frequently handling situations
with arrest as opposed to warnings and engaging in more crime control and law
enforcement activities and fewer order-maintenance and service activities. This approach
can be traced to conﬂict theory generally and from the threat hypothesis contained within
conﬂict theory speciﬁcally. Works such as those done by Turk (1972), Blalock (1967),
and Quinney (1970) exemplify this approach.

The second approach suggests that police tolerate higher levels of criminal
behavior in areas characterized as poor, minority, and unstable than is tolerated in other
areas. This latter approach would explain why there appears to be a higher threshold for
action in some areas, why police overlook some violations and disturbances, are less
likely to take formal actions such as an arrest, and tolerate certain public behaviors in
speciﬁc areas which would warrant formal action in others. With this second approach,
the police operate towards certain areas with a type of “benign neglect” (Liska and

Chamlin, 1984); Warner (1997) describes this as the underclass hypothesis. This

14

approach applied to explaining police behaviors is described by Klinger (1997) and Stark

(1987)

A Conflict Theoretical Approach to Police Behavior

The ﬁrst approach considered in explaining police behaviors and linking these to
community characteristics stems from conﬂict theory. Conﬂict theory posits that law
enforcement agencies, as agents of the state, process individuals with less (economic and
political) power (V old and Bernard, 1986). Conﬂict theory attempts to explain crime
control as a means by which a dominant group in society controls other groups whose
actions may threaten the interests of that dominant group (Liska and Chamlin, 1984).
Instead of viewing the law as impartially imposed in the general interests of the majority
(i.e., the consensus view), conﬂict theorists view crime control and the imposition of law
as in the interests of those with power to maintain the status quo (Jacobs, 1979).
Accordingly, even minor incidents when committed by members of these threat groups
are handled formally (Warner, 1997). Proponents of conﬂict theory cite manifestations
such as police force size (Jacobs, 1979; Liska, Lawrence, and Benson, 1981), higher
arrest rates (Jacobs and Britt, 1979), and brutality of the poor by the police (Chambliss
and Seidman, 1971) as evidence supporting the legitimacy of a conﬂict theoretical
approach.

Turk (1972) details the conﬂict perspective with regard to who is initially deﬁned
as criminal and as a result, who are the targets of police actions and also the ofﬁcial
responses which these groups receive. With this perspective, the criminal justice system

in general and the police speciﬁcally are enforcers of legal and social norms; they are

15

authorized coercive agents (Turk, 1972). Considerable variation in both law formulation
and law enforcement is understood as efforts to control culturally dissimilar subordinate
groups (Liska et al., 1981; Turk, 1972). Turk (1972) recognizes that police can routinely
enforce or ignore violations, and when responding, take formal or informal actions. As
an example, Turk cites police routinely ignoring weekly suburban poker games but being
on the lookout for cards and craps games in the “slum”. Chambliss and Seidman (1971)
echo these sentiments noting that law enforcement ofﬁcials will process a
disproportionately high number of the politically weak, at the same time ignoring the
violations of those with power. According to Chambliss and Seidman (1971) police
operate more in a crime control manner with poor populations and engage in more non-
enforcement practices with politically and economically powerful groups.

In terms of police behavior, the power and threat hypotheses contained in the
conﬂict perspective indicate that police act more formally and more vigorously in areas
characterized as less powerful (i.e., poorer, minority, heterogenous, unstable areas) to
compensate for the lack of other social controls. Sampson et a1. (1997) deﬁne social
control a the capacity of a group to regulate its members. According to this view, poorer
and more minority neighborhoods present a threat to the established order and elicit
higher levels of formal social controls by the police (Warner, 1997).

Characteristics of communities play an integral role in the development of
informal social controls. Bursik (1986) discusses communities’ ability to engage in self
regulation and notes the importance of the development of formal and informal networks.
These networks are difﬁcult to maintain during periods of compositional change (Bursik,

1986). Sampson et a1. (1997) identify speciﬁc examples of the types of informal social

16

controls in consideration, including the monitoring of play groups and teenagers and

conclude that,

the capacity of residents to control group level processes and visible signs of
disorder is thus a key mechanism inﬂuencing opportunities for interpersonal crime

in a neighborhood (p. 918).

When these controls are absent, the threat hypothesis contends, formal mechanisms for
control, such as the police, are necessary.

As Jacobs (1979) notes, the police are the major institution responsible for the
maintenance of order and stability and a strong police force is the most direct way of
maintaining the current order. Consistent with a conﬂict approach, higher levels of social
control are expected by the police in certain locations and towards certain groups. As
Black (1976) details, law and the imposition of law are variable and quantiﬁable — being
more heavily or formally applied in some places, to some people, and at certain times.

Following a conﬂict perspective, Jacobs and Britt (1979) use the power
hypothesis to explain police behavior, speciﬁcally, identifying differential police behavior
based on the social class of the target citizen. Jacobs and Britt (1979) state that the police
have fewer restraints when interacting with members of the lower class because these
groups have less power. As Sherman (1986) notes, differential police behavior by
neighborhood is a product of what police feel that they can get away with based on the

political power of the neighborhood. The threat hypothesis of conﬂict theory is one

17

explanation of discretionary police behavior and why law is more heavily or formally
applied in some areas more than in others.

Quinney (1970) addresses police discretion and community context and discusses
differential police behavior associated with community heterogeneity. Quinney (1970)
states that “the most signiﬁcant characteristic that accounts for community differences in
law enforcement is the extent to which the community is homogenous . . .” (p. 115). The
degree to which police act more or less formally is claimed to be a function of the degree
of heterogeneity/homogeneity that a community exhibits. Basically, if the police believe
that order can be maintained in an area without the invocation of formal law enforcement
efforts, then informal efforts will be applied. Areas which are characterized as more
heterogenous receive more formal police actions while more homogenous areas receive
more informal measures. Law enforcement is used as a means of maintaining order and
used more formally in heterogenous areas to compensate for the lack of other
mechanisms of control. Even for the same behaviors, law violations in homogenous
areas are handled informally while handled formally in heterogenous areas (Quinney,
1970)

Black (1976) also provides an explanation of how and when law is imposed and
how the implementation of law varies from one time to another and from one setting to
another. The foundation upon which Black’s (1976) propositions rest is that law is a
quantiﬁable variable. A central tenet of Black (1976) is that law varies according to
social context and that law varies inversely with other social controls (i.e., where there are
more social controls there is less law). According to Black, where social controls are

weak or nonexistent, law is imposed in a more vigorous manner. The police are but one

18

type of social control and they can impose law more or less. Black (1970) notes that
when no other type of social control is present, people are more likely to rely on the
police.

The relationship between the police and other forms of social control has also
been addressed by Banton (1964). Banton (1964) states that when other controls are
present, there is less reliance by the public on the police. Additionally, when other social
controls are weaker, police act more formally, hence, imposing more law. Goldman
(1963) describes how police ofﬁcers’ impressions of a family’s ability to control their
children impacts the ofﬁcer’s decision to refer a juvenile to court. Speciﬁcally, if an
oﬂicer believes that a family is able to control the juvenile, then the police act less
formally than if they believe such familial, informal social controls are weaker or
nonexistent.

Black (1976) explains the behavior of law and how law is imposed in certain
settings, not how individual police ofﬁcers act. In addition to the amount of law, the style
of law also varies from setting to setting: law can be applied in a punitive, compensatory,
therapeutic, or conciliatory manner (Black, 1976). Black claims that the conciliatory style
of policing is common in areas where people are considered equal, intimate, and
homogenous (Black, 1980).

Among several inﬂuences of the quantity of law, Black (1976) identiﬁes
stratiﬁcation (vertical relationship), morphology (horizontal relationship), organization,
culture, and social control. Goldman’s (1963) ﬁndings concerning juvenile court referrals
by the police illustrates the horizontal relationship as described by Black: where ofﬁcers

and citizens had impersonal or rigid relationships, formal actions are more likely taken,

19

whereas ofﬁcer behavior was more informal in communities where more personal
relationships existed.

Place and changes in place play a key role in the implementation of law. The
development of cities and population growth and density are related to the differing levels
of law. Black (1976) notes that in the development of law, it is adopted ﬁrst and most
fully in the cities and “last and least to the bush where tradition is still strong” (p. 15). As
populations increase and as people crowd together, relational distance grows. Black
identiﬁes two important developments which result from growths in population and
increased density: the development of laws and the adoption of full time police ofﬁcials.
As groups grow, communal relationships are weakened or broken and the need for laws
and for law enforcement increases. Black (1976) states that, “in any setting where people
closely watch each other’s conduct and readily criticize and punish deviants in their own
way, law is less important” (p. 109). Chambliss and Seidman (1971) similarly state that
as societies move from primitive to more complex, less consensus and more conﬂict is
present; as societies became more complex, police forces were developed.

Black’s (1976) Behavior of Law is particularly useful in explaining the behavior
of the police, in particular, why police behavior is different in different settings.
Characteristics of neighborhoods, such as the degree of organization present (such as
neighborhood groups), indications of levels of social controls, and socioeconomic status
(as an indicant of stratiﬁcation), determine the quantity of law that is imposed. Black
(1980) notes that while in theory the law is available to all, the imposition of law tends to
be reserved for those at the bottom and this serves to perpetuate systems of social

stratiﬁcation.

20

Consistent with conﬂict theory, as stratiﬁcation increases, the frequency or
intensity of legal-coercive control increases (Williams and Drake, 1980). More formal
police actions (e.g., arrest versus warning) are manifestations of the varying quantities of
law and this control. Differential police ofﬁcer actions, as exhibiting more or less vigor

in different settings can be understood in light of the behavior of law.

Benign Neglect and the Underclass Hypothesis

The alternate and contrasting approach to the conﬂict theory in explaining
differential police behavior by neighborhoods centers around what Liska and Charnlin
(1984) term “benign neglect” and Klinger (1997) and others refer to as the underclass
hypothesis. This approach posits that the police have a general disregard for areas
characterized as poorer and more unstable and act less formally, less vigorously, and
provide fewer services in general than in other areas (Warner, 1997). In terms of speciﬁc
officer behavior, the benign neglect hypothesis contrasts with the threat and power
hypotheses of conﬂict theory, suggesting that police will behave less formally in poor and
minority neighborhoods (e. g., fewer arrests, fewer reports taken) (Liska and Chamlin,
1984; Warner, 1997).

Two important tenets of the benign neglect hypothesis are identiﬁed by Liska and
Chamlin (1984) that have particular relevance for the study of differential police
behavior. One is that as the percentage of nonwhites and segregation increases arrests of
nonwhites decreases because the ratio of intra-racial crimes increases; and secondly, that

nonwhites are arrested less often for intra-racial crimes (p. 392).

21

Klinger (1997) provides a theoretical foundation consistent with principles found
in the benign neglect hypothesis. Klinger (1997) notes the relatively common belief that
ofﬁcer discretionary behavior is inﬂuenced by contextual characteristics of a
neighborhood or community and sets out to provide a theory to explain this inﬂuence.
Klinger’s (1997) approach to explaining differential police behavior by place, while
centered around the same principal as found in the benign neglect hypothesis, goes further
to explain why such neglect exists. Speciﬁcally, Klinger provides a theoretical
ﬁ'amework explaining why officers behave more vigorously in some area than in others,
accounting for differences in police behaviors that have been previously noted (e.g.,
tolerating citizen actions in one area that would receive a formal response from police in
another area).

Several principles are presented by Klinger (1997) in explaining differential police
activity by location. Klinger notes the importance of boundaries in policing and the
recognition, identiﬁcation, and association of locations by ofﬁcers with certain
characteristics. Klinger also notes the inﬂuence that ofﬁcers have on one another, so that
perceptions among ofﬁcers are shared regardless of personal experiences in a speciﬁc
area. This is similar to the near unanimity of beliefs held by ofﬁcers that was found by
Bayley and Mendelsohn (1969) regarding areas of town and problems associated with
these areas. It is claimed that ofﬁcers stereotype the population of individual districts,
one stereotype involving ofﬁcers’ evaluation of victim worthiness. Other police
researchers have addressed the issue of victim worthiness as well.

Bittner (1970) states that the differential treatment of citizens by the police stems

from the distribution of esteem, credit, and desserts in society at large. Victim worthiness

22

inﬂuences how ofﬁcers interact with citizens and the services they provide to all people
in certain areas. What Klinger (1997) terms “territorial stereotyping” is comparable to the
concept Werthman and Piliavin (1967) call “ecological contamination” and similar to
what Wilson (1980) refers to as “collective liability”.

Stark (1987) also discusses principles consistent with a benign neglect approach.
In deﬁning a theory of place and the ecology of crime, Stark (1987) discusses the impact
of neighborhood characteristics on law enforcement. From a set of propositions, Stark
delineates how such factors as neighborhood density, poverty, mobility, level of
dilapidation, and residential/commercial use inﬂuence the rates of deviancy. Of
particular relevance and value in understanding diverse police behaviors is how poorer,
more densely populated neighborhoods are more mixed-use (i.e., containing both
residential and commercial establishments) and how these areas tend to be more transient
or mobile. Neighborhoods characterized as more mobile have weakened extra-familial
bonds and fewer voluntary organizations -— both of which directly reduce formal and
informal sources of social control. These areas are also characterized as more
dilapidated, with higher rates of deviance; these in turn lead to stigma being attached to
the area. Stark (1987) presents that residents in these stigmatized neighborhoods will
receive more lenient law enforcement, for example, allowing things to pass in these
neighborhoods that ofﬁcers would respond to in non-stigmatized neighborhoods.

Klinger (1997) also states that in addition to victim worthiness, differential ofﬁcer
behavior is inﬂuenced by workload levels. Ofﬁcers who work in high workload areas
devote less attention to and even ignore minor crimes than officers patrolling in low

workload areas. Stark (1987) states that police frequently share the general community’s

23

view of stigmatized places as ﬁlled with morally disreputable individuals who deserve
what they get. Kress (1980) illustrates this line of reasoning noting that while there may
be a public perception of an “overarrest” of minorities, that within the “ghetto” the
opposite is often true. Kress (1980) uses a stabbing as an example: while in a middle-
class suburban area a stabbing is viewed as a serious matter, in poorer, predominantly
minority areas, criminal justice ofﬁcials including the police, view the same offense as
somewhat trivial.

Examination of Turk (1972) reveals an ideological foundation for the benign
neglect hypothesis within conﬂict theory. Turk notes that enforcers (i.e., police) are
indifferent towards opportunities to arrest what he terms the helpless. According to Turk,
as long as these helpless people only bother each other, they are not worth the trouble. A
curvilinear relationship is suggested in that those on one end (i.e., the powerful) and those

at the opposite end (i.e., the very weak) are essentially ignored.

Summary

Both the conﬂict perspective and the benign neglect hypothesis are useful in
explaining differential behavior by the police (with opposing outcomes) and account for
potential variations in the delivery of police services across neighborhoods. The conﬂict
perspective indicates more stringent, strict enforcement of the laws when focused on
poorer or minority populations, so that even minor infractions warrant formal responses.
On the other hand, following a benign neglect approach, minor inﬁactions within these

same populations would be ignored. Additionally, a benign neglect approach suggests a

24

generally lower level and quality of service provision results when the focus involves
poorer or minority areas.

In sum, both of the approaches discussed can be used to explain differential police
behavior by neighborhoods. Following these perspectives and recognizing the
importance and the inﬂuence of place and contextual level characteristics on police
behavior, the variation of police behavior across neighborhoods will be examined. The
current study will use these theoretical approaches as a foundation to test whether police
provide different services to different neighborhoods. Following a conﬂict perspective, it
is believed that police will engage in more criminal and law enforcement activities in
areas characterized as poorer, unstable, and racially or economically heterogeneous and
also engage in less order maintenance and service activities. From a benign neglect
position, it is believed that poorer, more unstable areas will receive fewer services from
the police. While police practices have previously been seen as somewhat equitable
(based on limited measures, such as response time and allocation of resources),
examination of a broader array of police activities could illuminate inequities that may

exist and which may be exaggerated with current police practices.

25

Chapter 2
Introduction

A signiﬁcant amount of research has been conducted in an attempt to identify the
determinants of police actions. Speciﬁc activities of ofﬁcers such as arrest and use of
force have been the primary focus of such research efforts, examinations being conducted
to identify factors at various levels which may inﬂuence ofﬁcer behavior. During the
19703 and early 1980s research was conducted to assess the equity of police services
across racially and economically disparate areas. These examinations, however, were
limited to measures capturing traditional and limited police practices. Speciﬁcally,
research involving the equity of the provision of police services has examined the
allocation of police services in terms of personnel levels and also response times (Ostrom,
1983). The research conducted in these areas suggests no evident relationship between
these speciﬁc outcome measures and racial or economic indicators (Maxﬁeld et al., 1980;
Mladlenka and Hill, 1978; Ostrom, 1983). These measures however, leave unexamined
signiﬁcant activities in which the police are engaged and for which the potential for
differential distribution exists. These studies also fail to capture the wide range of
environmental factors which may potentially inﬂuence ofﬁcer discretion.

While environmental factors are important, they are by no means the only
inﬂuence of officer discretion. Areas that have traditionally been examined include
individual ofﬁcer level characteristics, situational and organizational characteristics, as
well as community level characteristics. At the individual level, demographic,
background, and attitudinal variables have commonly been examined. Probably the most

researched area of inﬂuences of police behavior involve situational factors, such as

26

suspect demeanor, victim/suspect relationship, victim preference, and the seriousness of
the crime (Riksheim and Chermak, 1993; Sherman, 1980) . Organizational factors which
have been examined include managerial styles and mandates, bureaucratization, and
levels of professionalism. Research surrounding these different spheres of inﬂuence have
provided voluminous results. Some consistent and inconsistent ﬁndings have been
produced and some areas of examination have yet to be fully developed (Riksheim and

Chermak, 1993, and Sherman, 1980).

Individual Level Inﬂuences of Police Behavior

Research has indicated that differences exist between male and female police
ofﬁcers. There is evidence that female ofﬁcers perform their duties and interact with
citizens differently from male ofﬁcers. Research conducted in the 19705 (e.g., Bloch and
Anderson, 1974; Sherman, 1975) found that female police ofﬁcers initiated fewer citizen
encounters and also made fewer arrests. Later research such as Worden (1989) found no
difference between male and female ofﬁcers in terms of initiating encounters nor arrest
patterns.

Other research indicates that female officers are less likely to have misconduct
complaints ﬁled against them (Brandl, Stroshine, and Frank, 2001; Hickman, Piquero,
and Greene, 2000; Lersch, 1998; Lersch and Mieczkowski, 1996; Wagner, 1980). Lersch
(1998) explains such ﬁndings, indicating that female ofﬁcers were less authoritarian,
better communicators and possessed greater pacifying qualities. Others have noted that

female officers are more adept at avoiding violence, de-escalating potentially violent

27

situations, and providing sympathy and support to citizens (Brandl et al., 2001 , Toch,
1996)

Mixed result have been provided for ofﬁcer race and potential behavioral
differences. Reiss (1972) found differences in the use of unjustiﬁed force — with black
ofﬁcers exhibiting higher levels; Fyfe (1978) found similar results though these were
attributable to neighborhood differences. Sherman (1980) described research that found
that black officers patrolled more aggressively, initiated more citizen contacts, and ﬁled
more crime reports than white ofﬁcers. Worden (1989) however, found no racial
differences between officers in arrest decisions and Friedrich (1980) found no racial
differences between white and black officers with concern towards force — excessive or
justiﬁed.

As with race, ofﬁcer education has been found to affect police behaviors by some
researchers and not by others. Breci and Simmons (1987) found that more educated
ofﬁcers engaged in more criminal detection activities. Bozza (1973) found that college
educated ofﬁcers made more arrests (though the sample was small). Other researchers
have found no difference between college educated ofﬁcers and those ofﬁcers without a
college education — Smith and Klein (1983) and Worden (1989) examining arrest, and
Sherman and Blumberg (1981) examining force.

These are just a few of the numerous ﬁndings indicating that differences may exist
between ofﬁcers when considering individual level variables. While not negating the
importance of individual level inﬂuences, other levels such as contextual levels have also

been shown to impact the actions of the police.

28

Contextual Level Influences of Police Behavior

While examination of individual level inﬂuences on police behavior is substantial,
one area of research which is less developed is the examination of community level
inﬂuences on police behavior. From a summary and review of quantitative analyses of
the causes police behavior, Riksheim and Cherrnak (1993) conclude that neighborhood
characteristics (such as stability and racial homogeneity/heterogeneity) affect police
behavior. This same summary of research indicates that the inﬂuence of community level
variables on police behaviors has received little quantitative attention, especially with
regard to police service behaviors. Smith (1984) states that consideration of the contexts
in which discretion is exercised is essential for understanding decision-making by the
police. While considerable research on police discretionary actions has been conducted,
the primary focus of such research has been on situational and case characteristics leaving
unexamined the possibility that discretionary decisions are inﬂuenced by the contexts in
which these decisions occur (Smith, 1984).

Previous research examining contextual inﬂuences of police discretion differs
along several dimensions - the data used, the methods employed, theoretical frameworks
(or lack thereof), outcome measures (i.e., police behaviors), the level of “place,” key
contextual variables, and ﬁndings. Research into police discretionary behavior and the
inclusion of context is noticeable beginning in the 19605; earlier examinations were more
qualitative and ethnographic in nature (e.g., Banton, 1964, Bittner, 1967). In the 19703,
the Police Services Study provided a rich source of data that has been examined along
many lines, including examinations of contextual inﬂuences (e.g., Smith, 1984, 1986,

1987).

29

The outcome measures of previous research into contextual level inﬂuences of
police behavior have been limited primarily to arrest practices and the use of force (See
Appendix A, column 6 “outcome variables”). Various levels of aggregation have also
been used in previous research, such as neighborhood, census tract, police beat, police
precinct, city, and state (Smith, 1984; Weiner and Willie, 1971; Maxﬁeld et al., 1980;
Hepburn, 1978; Swanson, 1978; Jacobs and Britt, 1979 respectively). A signiﬁcant
amount of research has been conducted examining aggregate arrest rates by examining
state and city characteristics such as income levels and population demographics (e.g.,
Brown and Warner, 1992; Crank 1990; Jacobs and Britt, 1979; Liska et al., 1985). Lower
levels of aggregation such as neighborhoods have also been used (e.g., Smith 1984, 1986,
1987, Warner, 1997).

A variety of community characteristics have been examined in an attempt to
identify environmental inﬂuences of police behavior. Most commonly examined
characteristics include racial and socioeconomic indicators (e.g., Crank, 1990; Hepburn,
1978; Liska and Chamlin, 1984). While most research examining the inﬂuence of place
on police behavior has found some inﬂuence, a few studies ﬁnd little impact of
contextual level variables (e.g., Novak et al., 2002, Slovak, 1986, 1987).

What follows is a review of the existing research into environmental and
contextual level inﬂuences of police behavior along the dimensions mentioned above.
The limitations of previous research are addressed and the current study is placed in the

broader context of extant research.

30

Qualitative Accounts of Differential Police Behavior by Neighborhood

The idea that police ofﬁcers engage in different behaviors and treat citizens in
different areas differently has been present in the police literature for some time. As
Warner (1997) notes, there is a growing body of literature which suggests that police
respond differently to similar behavior in different communities. Several early
ethnographic works conducted in the 1960s identiﬁed differential police practices in
various neighborhoods. In an early work illustrating police behavior and community
level inﬂuences, Banton (1964) describes how ofﬁcers in different neighborhoods provide
different services; Banton attributes this difference to the social distance that officers
have with the residents whom they police. Black (1980) states that the greater relational
distance between disputants, the more likely law is used to settle disputes; this reasoning
could easily be extended to consider the social distance between the police and citizens
and the resulting differential imposition of the law.

Banton (1964) discusses the varying levels of social control that are present in
different communities and the notion that residents in areas characterized by more social
control come to rely on the police less. Communities are more or less stable, more or less
homogenous, and more or less able to exert social control. Police officers function to
enforce the standards accepted by the community (Banton, 1964). These community
standards differ and therefore police ofﬁcer behavior differs by community. As Banton
describes, at different times and in different districts, the police enforce the law more

strictly. It is pointed out that in:

31

“rougher neighbourhoods [the police] will disperse groups from the street corners
to prevent the conditions arising in which ﬁghts and disturbances most usually
start. A larger group on the pavements at the end of a church service or in a

middle-class neighborhood will be left undisturbed’ (p. 131).

The police turn a “blind eye” upon “drunks and revellers” in some neighborhoods and
arrest them in other neighborhoods; police are seen to provide services in some areas that
are not requested nor required in others (Banton, 1964).

Bittner (1967) also identiﬁes differential police behavior by neighborhood.
Bittner describes police behavior as adaptable to different settings, noting that ofﬁcers
adopt a style of policing that ﬁts both the inhabitants and the conditions of an area.
Bittner recounts ofﬁcer behavior on ‘Skid Row,’ noting that “police have a particular
conception of the social order of skid-row life that determines the procedures of control
they employ” (p. 699). Skid row is seen as completely different from other areas and
other districts.2 According to Bittner (1967) ofﬁcer behavior towards residents of skid
row are markedly different from behavior towards residents in other areas. The
invocation of law, as a formal response to deviance, is not the same in skid row and
towards skid row inhabitants as in other parts of the city.

Bittner (1967) also points out that ofﬁcers had a fundamental agreement about the
structure of skid-row life and residents. Residents are presumed to be incompetent and

disinclined to be “normal.” In terms of perceptions of skid row inhabitants, residents are

 

2 According to Bittner (1967), skid row is identiﬁed as an area with a heavy concentration of persons
classiﬁed as not living “normal” lives.

32

painted with broad brush strokes — identifying all who dwell in the area as alike and
coloring how they are treated by the police. This process is similar to what Werthman
and Piliavin (1967) call ecological contamination, wherein all people in an area are
treated differently (from people in other areas) because of a few. Werthman and Piliavin
(1967) identify differential behavior of police as motivated by a set of expectations
regarding appropriate conduct in a given neighborhood - for example, gang boys were
allowed privileged use of certain parts of the streets, while similar behavior in other areas
would be grounds for arrest.

Among the determinants of ofﬁcer behavior identiﬁed by Werthman and Piliavin
(1967), such attributes as the attitude of juvenile and their “moral character,” location is
said to play a pivotal role. Who is (and who is not) deemed suspicious is a function of
both personal characteristics and place characteristics. In the natural course of police
business, people and places are divided into categories and initial assumptions are
associated with these categories. Accordingly, police treat persons and places in unique
ways. By their residency in certain “bad” neighborhoods, most of the people in these
neighborhoods are regarded with suspicion (Werthman and Piliavin, 1967).

In another early work, Bayley and Mendelsohn (1969) examined police in Denver,
Colorado and discuss police relationships with minorities and also ofﬁcer attitudes and
expectations towards different racial and socioeconomic groups. The expectations that
officers have concerning the parts of a city where these groups are concentrated are also
discussed. Bayley and Mendelsohn note that ofﬁcers consistently and uniformly profess
that different socioeconomic groups request different services as well as present different

challenges for police. When Denver police ofﬁcers were presented with a map of the

33

city, ofﬁcers identiﬁed speciﬁc duties that they expected would be required in different
areas of the city. Speciﬁcally, in areas characterized as having higher socioeconomic
status, officers expected more assistance calls — those not involving criminal violations.
In lower income areas, ofﬁcers consistently stated that more ambiguous situations (for
example, family disputes and general disturbances such as noise complaints or crowds)
were common and assistance calls were less common. In other words, situations in which
ofﬁcers exercised the greatest amount of discretion were associated with lower income
and minority areas. Denver police associated their work in disadvantaged areas with a
high amount of “discretionary intervention”. Bayley and Mendelsohn (1969) concluded

that:

Police ofﬁcers not only carry very detailed maps of the city in their minds but they
evidently color these maps almost identically with respect to the kinds of behavior

to expect in different locations (p. 74).

The perceptions of different neighborhoods and expectations identiﬁed by Bayley and
Mendelsohn (1969) were consistent among ofﬁcers, despite their personal experiences in
speciﬁc neighborhoods. This indicates sharing of information among ofﬁcers and the
development of a “climate of opinion” about a neighborhood (Bayley and Mendelsohn,
1969). Ofﬁcers carry these predispositions (i.e., cognitive maps) into their contacts with
individuals in these neighborhoods and different levels of enforcement result. As noted

by Banton (1964) and Bittner (1967), ﬁom this, differential police responses are

34

presented (e.g., a brawl that is tolerated in one area producing an arrest in another)
(Bayley and Mendelsohn, 1969).

These works conducted in the 19608 provide some of the earliest attempts to
explain police behavior and include geographic location and characteristics of place as
central inﬂuences. The assertions, depictions, and accounts provided in these earlier
accounts of police behavior have since been used as a foundation for empirical attempts
to quantify and specify the inﬂuences of police ofﬁcer behavior. In addition to other
levels of inﬂuence, place and characteristics of place have been included in examinations
of police discretionary behaviors and have generally been shown to inﬂuence the actions

of officers.

Quantitative Examinations of Neighborhood Influences on Officer Behavior

While examination of neighborhood level inﬂuences is less common than other
examined levels of inﬂuence of police behavior, several researchers have looked at the
role of neighborhood and other environmental inﬂuences. The majority of ﬁndings
(twenty-ﬁve out of twenty-eight studies examined) support the signiﬁcance of context as
an inﬂuence on police ofﬁcer behavior.

Wilson (1968) and Liska and Chamlin (1984) have identiﬁed the effects of
community characteristics on styles of policing. In his seminal work, Varieties of Polig

Behavior Wilson (1968) recognizes that police behavior differs from city to city and

 

identiﬁes three distinct types of police organizational styles (watchman, legalistic, and
service) which vary according to city (or community) characteristics. Wilson notes that

police ofﬁcers’ conception of the police role varies with the character of the community.

35

For example, the service style of policing identiﬁed by Wilson (1968) is characterized by
frequent but informal police interventions and is said to be most often found in
homogenous, middle class communities comprised of citizens who have a generally high
level of agreement concerning the deﬁnition of public order. Crank (1992) credits
Wilson’s Varieties of Police Behavior with sparking an interest in jurisdictional
variations in police practices based on organizational and environmental factors; Kania
and Mackey (1977) describe Wilson as the foremost advocate of explanations of the
variability of police behaviors based on community characteristics.

One relatively common type of police decision-making which has been examined
in light of community characteristics is arrest. Arrest patterns and decisions have been
examined in regard to a variety of community contextual characteristics including racial
composition, city government types, urban-rural differences, and socioeconomic factors.

Of the research examining community characteristics and police behavior,
examinations of the racial compositions of areas in which certain police activities, such as
arrest, are among the most common. The research that has been conducted indicates that
police are more likely to make arrests in areas containing larger minority populations.
Swanson (1978) found that environmental variables (as well as organizational variables)
affected arrest rates — the percentage of the population as non-white being the most
important predictor. Others such as Hepburn (1978), Liska and Chamlin (1984), Liska et
a1. (1985), and Williams and Drake (1980) also found the racial composition of an area to
be a predictor of arrest —- speciﬁcally, as the percent non-white population increases, the
likelihood of arrest increases. Liska and Chamlin (1984) found that arrest rates varied

according to the racial composition of cities, independent of police size and reported

36

crime rates. Hepburn (1978) found that the racial composition of the area in which an
arrest was made had a signiﬁcant impact on whether or not the prosecutor eventually
refused a warrant, indicating to Hepburn that in these areas, people are more likely to be
arrested on insufﬁcient evidence.

Crank (1990, 1992) also examined and found signiﬁcant relationships between
arrest rates and the racial composition of areas. Crank (1990) measured racial
heterogeneity in 3 ways (percent Black, percent Hispanic, and percent foreign-language
speaking) and found that increases in percent Black and percent Hispanic were related to
increases in arrest rates.3 Crank (1992) continued with this line of research and found that
police were more likely to make arrests of minority suspects when the percentage of
minorities in the community was small. In other words, black and Hispanic suspects
were arrested at higher rates when they were in areas where their percentage of the
population was small (Crank 1992). Although Crank (1990) did not ﬁnd the percentage
of foreign speaking residents to be a predictor of arrest, Brown and Warner (1992) found
percent foreign-born to be a signiﬁcant predictor of arrests for drunkenness (B = 3.04) in
an examination of arrest rates in the early 20th century, though this effect was mediated by
cities’ political culture and structural characteristics.

Characteristics pertaining to the economic conditions of certain locales have also
been shown to inﬂuence arrest practices. Smith (1984) found that as the percentage of
households below the poverty level increased, the probability of arrest increased. Others

such as Crank (1990) and Smith (1986, 1987) have similarly shown that arrest rates are

 

3 ln urban communities, percent Hispanic was associated with higher arrest rates for trespassing (B=.29);
percent Black was a consistent predictor of arrest in both urban and rural settings for trespassing (B=.34 and
.15), disorderly conduct (B=.36 and .19) and in rural settings, percent Black was predictive of arrest for
motor vehicle offenses (B=.35) and cannabis control (B=.18) (Crank, 1990).

37

inﬂuenced by the economic conditions of an area. Williams and Drake (1980) found that
income inequality was positively related to arrest rates; as income inequality increased,
arrest rates increased.

The type of city government has also been examined as an environmental or
community characteristic affecting arrest decisions. Speciﬁcally, Wilson’s (1968)
typology of police departments (i.e., legalistic, watchman, and service) has been tested
empirically and found some support. Departments have been measured according to their
degree of professionalization and bureaucratization, classiﬁed according to Wilson’s
typology, and examined in relation to the type of city government present in the area (city
council, city managers, mayors, and combinations thereof). Results indicate that the type
of city government inﬂuences the levels of arrests for different types of crimes (Crank,
1990; Crank 1992; Langworthy, 1985). Slovak (1986) found that policing styles were
more legalistic (i.e., more arrests and more formal actions) in cities that were governed by
city managers and where political decision-making was more decentralized. Brown and
Warner (1992) found that the type of political culture (including the strength of local
political machines) mediated the effect found in regard to the percent of immigrant
residents.

A second type of police behavior that has been examined in light of community
level inﬂuences is police use of force. Several community characteristics have been
shown to inﬂuence police use of force. In particular, police use of force (including police
use of deadly force) has been shown to be affected by the crime rate, socioeconomic
status, and racial composition of an area. Kania and Mackey (1977) examined inﬂuences

of police use of deadly force across states (including what they termed quality of life,

38

safety, mobility, and recreation) and found the strongest associations to be crime rates.
Waegel (1984) also found the level of crime in the environment to be related to police use
of lethal force. Findings by Geller and Karales (1981) point to differential residential
patterns in police shootings in Chicago - larger percent minority and higher crime rates
were associated with more shootings by and of the police.

While the validity of the ﬁndings drawn by Kania and Mackey (1977) have been
questioned (see for example, Fyfe, 1988), several of the measures used captured
indicators of socioeconomic status (e.g., owning an automobile, owning a television, hot
water in the home) and these indicators were found to be related to the rates of police use
of deadly force at the state level.4 Jacobs and Britt (1979) also examined instances of
police use of deadly force and found that economic inequality (measured as the
differences in economic resources) explained police use of deadly force better than
absolute poverty.

In another examination of police use of force and community characteristics, Fyfe
(1980) examined police shootings in New York City from 1971-1975, speciﬁcally
examining arrest and homicide rates of police zones and found that as these rates
increased, so too did the prevalence of police shootings.’ Both Kania and Mackey (1977)
and Fyfe (1980) concluded that police respond to the characteristics of a community in

developing operating styles; speciﬁcally, in more violent areas, police act more violently.

 

4 These measures were initially used by Kania and Mackey (1977) to indicate such things as “mobility,”
“quality of life” and “recreation.”

5 This ﬁnding was the highest correlation found (r = .72) in the analysis —- and was between arrest rates and
on-duty uniformed ofﬁcers’ shooting rates (Fyfe, 1980).

39

The impact of community characteristics on other discretionary police activities,
such as police contact with juveniles and decisions to formally ﬁle a crime report have
also been examined, though to a lesser extent. Weiner and Willie (1971) found that in
Washington, D.C., racial and socioeconomic characteristics of census tracts were related
to police contacts with juveniles. No relationship was found between these
characteristics and subsequent court referrals, similar to Hepbum’s (1978) ﬁnding
regarding the effects of race on arrest and the subsequent decision not to issue warrants
(cited above).

In an examination of police decisions to take a report or not, Maxﬁeld et a1.
(1980) and Warner (1997) both found community level inﬂuences. Examining police
decisions to ﬁle an offense report, Maxﬁeld et a1. (1980) found that as demands for
service increased in different areas, the likelihood that ofﬁcers would ﬁle an ofﬁcial
report increased. Warner (1997) also examined police decisions to record a crime and
found neighborhood inﬂuences. Police were less likely to record burglary complaints in
poor neighborhoods, more likely to record in more mobile neighborhoods, and more
likely to record in neighborhoods with higher percentages of foreign-born residents.6 For
calls that involved assaults, only percentage foreign-born residents was a signiﬁcant
predictor of non-recording (B = -.222); police were more likely to make an official record
of assault calls in areas with higher percentage of foreign-born residents. In
neighborhoods with a lower percentage of black residents, reported burglaries were more
likely to be recorded as less serious and in neighborhoods with higher percentages of

black residents, it was more likely that no crime was recorded (Warner, 1997).

 

6 Percent Black was not a signiﬁcant predictor of non-recording of burglaries (Warner, 1997)

40

One of the most thorough examinations of the effects of neighborhood context
was conducted by Smith (1986) using the Police Services Study from the late 19705.
Smith (1986) examined ‘neighborhood context of police behavior’ looking at various
police activities and numerous neighborhood characteristics. Types of police ofﬁcer
behavior that were examined included proactive assistance and investigation, the use of
coercive authority (e.g., taking a person into custody or arrest), and report ﬁling.
Characteristics of neighborhoods were obtained using a random survey of approximately
200 residents of each of the study neighborhoods. Variables created that were examined
included: a crime scale, socioeconomic scale, residential stability (i.e., percentage of
residents residing in the neighborhood less than ﬁve years), interaction of residents (i.e.,
how often neighbors socialized), household composition, and racial heterogeneity of the
neighborhood (Smith, 1986).

Several ﬁndings signiﬁed the importance of the inﬂuence of neighborhoods on
police ofﬁcer behavior. Smith (1986) found that ofﬁcers were more active in racially
mixed neighborhoods, arrested suspects more in lower status neighborhoods, were more
likely to exert coercive authority in minority or racially mixed neighborhoods, and were
less likely to ﬁle reports in higher crime neighborhoods. Smith (1986) discovered that the
likelihood of arrest declines with increasing community status in terms of SES (B = -
1.13). Coercive authority was more likely to be used or threatened to be used in black or
racially mixed neighborhoods (B = .19) and less coercive authority exhibited in more
transient areas (B = -.24). Socioeconomic status, as hypothesized, had an inﬂuence on
several police activities. Police actions were considered more legalistic (i.e., more arrests

/ more formal actions taken) as the socioeconomic status of a neighborhood decreased

41

(Smith, 1986). Additionally, police used more coercive authority in non—white and
racially mixed areas, even taking into consideration suspect race, sex, demeanor and the
type of problem involved (Smith, 1986).

In an examination of police responses to interpersonal violence, Smith (1987)
again included individual, situational, and contextual variables in an attempt to explain
ofﬁcers’ responses (i.e., separate, mediate, arrest) towards interpersonal conﬂicts. Two
measures of the context in which the police work were developed by aggregating survey
responses from study neighborhood residents (i.e., violent crime rate and economic
status). Smith (1987) found that the poverty level of the neighborhood where the
encounter occurred had a signiﬁcant inﬂuence on how police ofﬁcers responded to calls
involving interpersonal violence. Speciﬁcally, encounters involving interpersonal
violence in neighborhoods with lower economic status were signiﬁcantly less likely to be
mediated, independent of encounter-level characteristics such as severity of the incident
or victim/offender characteristics.7 In sum, the economic status of the neighborhood
“independently affects how police handle violence between citizens” (Smith, 1987, p.
779).8

Smith (1986) described these ﬁndings as “disturbing” because, he said, they
“undermine our ability to develop general theories of police discretion or decision
making” (p. 339). Smith states that any complete understanding of police decision

making must include the contexts in which such decisions are made. Theories attempting

 

7 Results are presented as probabilities of mediation, separation, and arrest. As the poverty level of a
neighborhood increases, the probability of mediation decreases (-.2l9) and the probability of an arrest
increases (.335) (Smith, 1987).

8 Smith (1987) found no signiﬁcant effect of the neighborhood crime rate on police decision-making.

42

to explain police behavior and police discretionary decisions “must explicitly recognize
the contextual variability of police decision making” (Smith, 1986, p. 339).

While the above research highlights evidence indicating the inﬂuence of
contextual factors in police activity, other research has failed to ﬁnd such effects. In an
examination of types of police activity across neighborhoods, Slovak (1987) found little
variation between the types of activities, which were divided into 3 categories (service,
legalistic, and watchman-type behaviors — per Wilson, 1968) and examined across 8
neighborhoods. Neighborhood characteristics that were examined included the age and
income of residents and also how safe residents’ felt. Slovak (1987) concluded that
variations in police activity were but nuances on a general city-wide pattern of policing
with little variation from one neighborhood to the next. In an examination of police use
of deadly force, Lester (1982) found patterns of civilians killed by the police in areas
characterized with higher violent crime rates and those with a higher percentage of black
residents, but the associations rarely reached a statistically signiﬁcant level.

Novak et a1. (2002) included situational and community level variables to
examine the impact of community policing and found that arrest decisions did not vary
due to community level characteristics. Novak et a1. (2002) examined characteristics
such as crime rate, mobility and a community factor capturing economic and racial
characteristics. Slovak (1986) found that organizational characteristics impacted styles of
policing and concluded that the style of police work was generally consistent across
neighborhoods within a given city — the one exception being differences found in police

styles between downtown business districts and residential neighborhoods.

43

Despite these lastly mentioned studies, there is reason to believe that contextual
variables inﬂuence a variety of police behaviors. While most of the existing research has
been limited to traditional measures of police activity, the current study examines a wider
range of police practices as well as contextual indicators which may play a role in

inﬂuencing the types of services police provide to neighborhoods.

Levels of Analysis

One consideration in attempting to better understand persons and systems such as
the police, is the recognition that context is multi-leveled (Kingry-Westergaard and Kelly,
1990). The studies cited above have examined inﬂuences at various levels of aggregation
including states, cities, police districts and beats, census tracts, and neighborhoods
(Jacobs and Britt, 1979; Liska and Chamlin, 1984; Maxﬁeld et al., 1980, Smith, 1984).
With the embedded nature of systems within systems, the consideration of different levels
of aggregation are important in providing a complete context.

As Klinger (1997) notes, “because ecological research examines ‘communities’. .
. the ﬁrst order of business is any ecological work is to specify precisely the areal units
that constitute the ‘community of interest’” (p. 280). For Klinger (1997), identifying the
ecological unit of interest is a somewhat less complicated task when discussing the police
since departments operate within police districts with clearly deﬁned boundaries. These
police-deﬁned units have been utilized in previous research: police beats (Maxﬁeld et al.,
1980), police zones (Fyfe, 1980), police districts (Geller and Karales, 1981) and police
precincts (Hepburn, 1978). Others, such as Smith (1986), described the locus of

inﬂuence in terms of smaller units —- the neighborhood, but also recognized potential

44

inﬂuences of larger units. Smith (1986) and Klinger (1997) come to similar conclusions,
that any attempt to understand police behavior in neighborhoods, must also take into
account the nature of the patrol district in which they are situated. Neighborhoods have
commonly been the unit of analysis in examinations of contextual inﬂuences of police
behavior (e.g., Novak et al., 2002, Smith, 1984; Smith 1986; Warner, 1997).

Kelly (1988) notes that the inclusion of multiple levels of community necessitates
the consideration of a number of different variables such as economic, social structural,
and physical. As seen, examinations of police behavior have been conducted at these
multiple levels but have been limited in the range of contextual variables which have

been included in analyses.

Contextual Variables

For the studies described above, several community contextual variables have
been examined to determine if an effect on police behavior is discemable. One of the
most commonly included contextual variables examined involves the racial composition
and/or racial heterogeneity of an area (e.g., Crank, 1990, Crank 1992; Geller and Karales,
1981; Hepburn, 1978). Income characteristics are also commonly included (e.g., Crank
1990; Hepburn, 1978; Jacobs and Britt, 1979; Kania and Mackey, 1977, Lester, 1982;
Liska and Chamlin, 1984). A third important and frequently included contextual variable
of interest in terms of potential effects on police practices are crime rates (e. g., Jacobs and
Britt, 1979; Karria and Mackey, 1977; Lester, 1982; Slovak 1986).

Other variables capturing community context have been included but less

frequently. The inclusion of these characteristics are particularly relevant in terms of the

45

potential inﬂuences of police behavior based on levels of other forms of social control.
Speciﬁcally, variables indicating levels of social control, such as residential stability and

mobility, have been included (Novak et al., 2002, Smith, 1986, Warner, 1997).

Summary

These previous studies involving the inﬂuence of place on police behaviors
provide the foundation for the current research. The previous research conducted can be
placed along several dimensions, such as the type of data examined, key contextual
variables included in the analysis, and the level of aggregation. The current research
attempts to add to the current state of knowledge regarding contextual level inﬂuences of
police behavior by examining the patterns of policing by neighborhood context.
Speciﬁcally, while previous research has been limited primarily to examinations of arrest
decisions and decisions to use force, the current study takes a broader approach in
examining a variety of police behaviors, including order maintenance and service-type
activities. Previous examinations have also been limited in the scope of variables
available - largely relying on racial and economic indicators. The current study includes
neighborhood level social and economic characteristics, as well as stability indicators,
which may prove useful in determining the inﬂuences of police behavior.

While researchers have examined the role of contextual level inﬂuences of police
behavior, the current study attempts to broaden our understanding of determinants of
police behavior. Previous studies have been limited in numerous respects, primarily with
regard to the types of behavior examined. With the exception of Slovak (1986 / 1987)

and Smith (1986), examinations of contextual level inﬂuences have been limited to

46

traditional law enforcement practices, speciﬁcally, crime control aspects such as arrest
(e.g., Brown and Warner, 1992; Crank, 1990, 1992; Langworthy, 1985; Liska and
Chamlin, 1984; Smith 1984), police use of force (e.g., Fyfe, 1980; Geller and Karales,
1981; Lester, 1982; Waegel, 1984), and decisions to ﬁle a police report (Warner, 1997;
Maxﬁeld et al., 1980).9 These studies, as previously mentioned have generally supported
the notion that contextual level factors inﬂuence police behaviors. While providing
important insight into discretionary police practices, the focus of previous studies has
been narrow and ignores a considerable, some would argue, the majority of police
activities in which the police engage and activities involving the most discretion.

Cognizant of the amount of time that ofﬁcers spend engaged in non law
enforcement type activities, the current study examines a variety of police behaviors.
Speciﬁcally, the current study examines three broad areas of police activities: order
maintenance, law enforcement, and service activities. While previous research has
illustrated differential law enforcement/crime control behavior by neighborhood, other
activities in which the police engage have been all but ignored.

To be fair, a couple of previous studies have examined a broader sphere of police
behaviors, most notably, Smith (1986) and Slovak (1987). The study conducted by
Slovak (1987) is of particular utility for methodological considerations for the current
undertaking. Slovak (1987) categorized all police activities into one of three style
categories, following Wilson’s organizational styles: legalist, watchman, and service.

These styles correspond to the three generally accepted categories of police behavior

 

9 Categories of crimes for Warner (1997) included only burglaries and assaults; Maxﬁeld et a1.
(1980)examined UCR Part I and II offenses.

47

mentioned above (crime control/law enforcement, order maintenance, and service).‘0
Taken together, Smith (1986) and Slovak (1987) are unique for their breadth and also
their conﬂicting ﬁndings: Smith (1986) ﬁnding support for the contention that
neighborhood level inﬂuences affect police behavior while Slovak (1987) found that
differences across neighborhoods were insigniﬁcant.

Previous studies have also been limited in terms of a coherent theoretical
underpinning. Most commonly examined is a conﬂict perspective (e. g., Brown and
Warner, 1992; Jacobs and Britt, 1979; Liska and Chamlin, 1985; Williams and Drake,
1980) and while these studies supported conﬂict theory, Crank (1990) warns that a
conﬂict theoretical approach may not be appropriate to all areas, namely rural areas.
Alternative approaches have centered around the underclass hypothesis (e.g., Maxﬁeld et
al., 1980) and organizational approaches such as Wilson’s styles of policing (e.g.,
Langworthy, 1985; Smith, 1984).

While previous studies have broadened the understanding of the determinants of
police behavior, signiﬁcant shortcomings prohibit a complete picture. The current study
draws from and builds on existing knowledge and also attempts to ﬁll in existing gaps.
Speciﬁcally, extant research indicates differential police behavior by neighborhood
characteristics when examining certain (and limited) police practices. How do these
characteristics affect the delivery of other services? How do these characteristics affect
the amount of order maintenance activities in which officers engage? The present study
considers a range of police activities that captures the diversity of the job and potentially

illuminate differentials and even discrepancies in police practices.

 

'0 Chapter Three details the speciﬁc police activities and categories into which each was included.

48

Chapter 3
Data

The current study involves secondary data analyses of data from the Project on
Policing Neighborhoods (POPN). The original study was funded by the National Institute
of Justice and involved multiple data collection methods including systematic social
observation (SSO), structured and unstructured interviews with police officers, and
citizen surveys. The study sites were Indianapolis, Indiana in 1996 and St. Petersburg,
Florida in 1997.

Systematic Social Observation was the primary method of data collection and will
be the main data source analyzed. The methods involved in SSO date back to the 19305
and were initially applied to the study of policing by Reiss in the 19605 (Mastrofski et al.,
1998). In a research report prepared for the National Institute of Justice, Mastrofski et a1.
(1998) describe the methods involved in the systematic social observation of police, “In
880 of patrol work, trained observers accompany police ofﬁcers in their cars, on foot, or
on bicycle to observe everything [police ofﬁcers] do during a typical tour of duty” (p. vii).
Data collected for POPN involved project observers taking brief notes during police
observation sessions and later recounting what they observed in semi-structured narrative
accounts of events from the patrol observation period. Next, highly structured
computerized questionnaires were completed about events that occurred during the
observation session. Information was coded concerning the ride, the officer, citizens
encountered, and other details of events that were observed.

The observation instruments used for the initial project included multiple forms

(i.e., ride, activity, encounter, and citizen), providing for different levels of analysis. The

49

ride form provided general information about an observation period (e.g., date, shifi, time
of day, assigned beat, officer’s assigned unit); one ride form was completed per
observation session.

When ofﬁcers were engaged in some activity in the absence of citizens, this was
deﬁned as an “activity” and an activity form was completed detailing such an event.
Activity form questions included how the activity began and ended, the type and length of
the activity, as well as the focus of the activity. Information was also obtained as to how
much of the activity (0 - 100 percent) occurred inside the study beat. Each activity was
assigned an activity code.” In addition, an activity could be directed towards a speciﬁc
problem and therefore also given a problem code. Problem codes were from an
exhaustive list of potential problems and situations in which the police could become
involved. These were broadly categorized as: problems with persons, problems with
property, trafﬁc problems, service problems, information problems, legal procedures,
administrative problems, and miscellaneous problems. To illustrate, an ofﬁcer engaged
in an activity involving administrative report writing (activity code 701) could also
involve the speciﬁc problem of drug violations (problem code 040).12

When citizens were present and ofﬁcers were interacting or otherwise engaged
with these citizens, these events were considered “encounters” and a different form (an

encounter form) was used to code information about these events. Encounter forms

 

11 For example, the activity code for general patrol was 1 10; for directing traffic, 404.
'2 Activities did not always involve a problem code. For example, general patrol which was focused on no

speciﬁc problem would not be assigned a problem code, whereas general patrol focused on a speciﬁc
problem (e.g., drugs) or area (e.g., a park) could have a problem code attached.

50

captured a variety of information including the initiation source and location for the
encounter (using geocodes of

cross-streets), the length of the encounter, and the problem(s) addressed during the
encounter, to name only a few. All encounters were assigned problem codes.

The ﬁnal form was the citizen form. Information about citizens with whom
ofﬁcers interacted was coded using the citizen form. The citizen form captured
information about general citizen characteristics such as age and race, speciﬁc officer and
citizen interactions, requests made by citizens, and responses of the officer to such
requests, as well as situational characteristics such as demeanor.

Observers collected and coded data for a total of 729 observation sessions
(commonly referred to as ride-alongs), totaling more than 5,700 hours of ﬁeld observation
(Parks et al., 1999). A total of 35,433 events were observed; a breakdown of these can be

found in Table 1.

Table 1. Event Type by Study Site (n = 35,433)

 

Activities Encounters
Indianapolis, IN 13,418 3,967
St. Petersburg, FL 14,572 3,476
Total 27,990 7,443

 

The POPN project had twenty-four study beats that were selected for observation
(twelve in each site). The observation sampling plan was created to capture all days,

work shifts, units working the study beats (both community police ofﬁcers and general

51

patrol unit ofﬁcers), and both busy and slow times (Parks et al., 1999).13 Observers
conducted approximately 240 hours of observation in each study beat (DeJong et al.,
2001)

A number of trained observers were involved in these 5,700 hours of systematic
social observations over several months, during two summers. As such, two
considerations generally considered important when using methods such as 880 —
reliability and reactivity — need to be addressed. These issues however do not necessarily
become problems methodologically, as POPN recognized and attempted to address such
concerns. According to Mastrofski et al. (1998) reliability issues were addressed-in
several ways: ﬁrst, reliability was enhanced by the structured nature of the observations
and the reconstruction of these observations. Second, reliability was enhanced by
extensive training and supervision of observers. Third, the data entry system contained
quality control programs which tested logical arrangements of coded data.

Officer reactivity to observers was a second important methodological
consideration with ﬁeld observations, especially in observations of the police. The
potential exists that observed ofﬁcers may change their behavior when being observed,
resulting in researchers collecting data about behaviors that are different from the true
nature of police work. The most important measure to address potential subject reactivity
was strict conﬁdentiality agreements and restrictions regarding information sharing.

Observers were not allowed to discuss or disclose any aspect of an observation session

 

'3 Busier times (Thursday through Saturday) were over sampled to ensure sufﬁcient numbers of
observations of encounters with the public (Parks et al., 1999)

52

with anyone except the observed ofﬁcer. Observers were also trained to look for and note
potential signs of reactivity (Mastrofski et al., 1998)

In addition to information collected through the systematic observation of officers,
further information about ofﬁcers was gathered through structured interviews. These
interviews were conducted outside of any observation sessions and by separate research
personnel (DeJong et al., 2001). General demographic characteristics, education,
training, and attitudinal variables were obtained through these interviews. Attitudinal
questions pertained to ofﬁcers’ beliefs about their role, attitudes towards supervisors, and
attitudes about citizens (DeJong et al., 2001). The data gathered from these interviews
supplemented the observational data.

One signiﬁcant strength of S80 is that it allows for the observation of social
phenomenon in its natural setting following explicit, duplicatable procedures (Mastrofski
et al., 1998). Since the focus of the current study is the context of police behavior, SSO
provides an excellent opportunity to consider important contextual characteristics — in
this case, neighborhood characteristics — and examine differential police practices.

For the original project, neighborhoods were deﬁned by the police departments
using the boundaries of primary police assignment areas. Assignment areas (known as
patrol beats in Indianapolis and Community Policing Areas in St. Petersburg) “were
drawn to match existing neighborhood boundaries as far as possible, given available
patrol resources” (Reisig and Parks, 2000, p. 613).

Information about neighborhoods was obtained from 1990 Census data. For the
current project, census data was aggregated to the study beat boundaries. While the level

of aggregation for the current study is the neighborhood, since there are multiple levels to

53

be considered, a description of the study sites is provided to place the police behavior in a

larger context.

Study Sites '4
Indianapolis, Indiana

At the time of the study, the Indianapolis Police Department (IPD) had an
estimated service population of 377,723. Of this population, 39 percent was minority,
eight percent was unemployed and nine percent was 50 percent below the poverty level.
The Uniform Crime Report Index Crime rate was 100 per 1,000 residents. IPD employed
1,013 full-time sworn ofﬁcers (approximately 2.7 ofﬁcers per 1,000 residents), of which
48.6 percent, or 492 ofﬁcers were assigned to the patrol division. Patrol ofﬁcers were
the focus of the original study; observation sessions and interviews were conducted with
these ofﬁcers. At the time of the study, 83 percent of IPD patrol ofﬁcers was male and 21
percent was minority. Thirty-six percent of IPD ofﬁcers had a four-year college degree.

The Indianapolis Police Department service area is divided into 4 districts (North,
South, East, and West) and within these districts, there are 50 police beats. Twelve of
these beats were included in the original project as study beats. The average residential

population of a beat in Indianapolis was approximately 7,500 (Mastrofski et al., 2002).

 

M Information concerning the study sites was obtained ﬁ'om published articles using POPN data, primarily,
Parks et a1. (1999). 1990 census data (and projections) were originally used as was the most recent (1996)
UCR data.

54

St. Petersburg, Florida

At the time of the original study, the St. Petersburg Police Department (SPPD)
had a service area of approximately 240,318 residents and employed 505 full-time sworn
ofﬁcers (approximately 2.1 per 1,000 residents). Of the residential population, 24 percent
was minority, ﬁve percent was unemployed, and six percent was 50 percent below the
poverty level. The UCR Index Crime Rate per 1,000 residents was 99. Of the 505 full-
time sworn officers, 283 of these were in the patrol division (56 percent). Eighty-seven
percent of SPPD patrol officers was male, 22 percent was minority, and 26 percent had
four-year college degree.

The St. Petersburg Police Department jurisdiction was divided into three districts
(North, South, and West) containing a total of 48 beats. Twelve of these beats were
included in the original project as study beats and were selected to approximate the
socioeconomic conditions of the study beats selected in Indianapolis. The average
residential population of a beat in St. Petersburg was approximately 5,000 (Mastrofski et

al., 2002).

Unit of analysis

Observers were assigned to observe and record events of officers in one of twenty-
four study beats in each study site. Approximately 240 hours of observation were
conducted in each site’s 24 study beats (40 rides per beat). A total of 361 observation
sessions were conducted in Indianapolis and 368 observation sessions were conducted in

St. Peterburg. Observation sessions were designed to average approximately eight hours

55

but this time varied as some shifts ended early (for various reasons) and some were
extended (DeJong et al., 2001).

Observation of an assigned study beat was not always possible primarily due to
departments not always assigning an officer primary patrol duties to a speciﬁc study beat.
In such events, observers were assigned to ride with the ofﬁcer who would most likely be
responsible for covering calls from the study beat (typically an ofﬁcer from an adjacent
beat). These rides were excluded from the current analyses.IS

The unit of analysis for the current research is the ofﬁcer, with 224 ofﬁcers being
observed — 173 ofﬁcers observed in Indianapolis and 70 ofﬁcers observed in St.
Petersburg. It should be noted that several ofﬁcers were observed more than once and
though rare, some ofﬁcers were observed in more than one study beat. For ofﬁcers
observed multiple times in the same study beat, these times are summed. If an officer
was observed in multiple study beats, times were calculated and summed separately for
each study beat. The 224 observed ofﬁcers therefore resulted in 243 unique ofﬁcer-in-
beat observations. A total of 529 rides were included in the analyses (281 observation
sessions in Indianapolis and 248 observation sessions in St. Petersburg); the mean
number of observation sessions per ofﬁcer was 2.18 (SD. = 1.8) and ranged from 1 to

10."

 

‘5 A total of 166 observation sessions were excluded for this reason. No pattern was observed based on
beats and department assigned officers or lack of assigned officers.

'6 Signiﬁcantly more repeat observations of the same officer occurred in St. Petersburg. The mean number

of observation sessions in Indianapolis was 1.6 (SD. = 0.85, range 1 to 4) while in St. Petersburg, the mean
number of observations sessions was 3.5 (SD. = 2.6, range 1 to 10).

56

Measurement of the Dependent Variables:

Several variables were available and used to calculate the dependent measures:
time variables, initiation source indicators, event location, and event type descriptions.
For each event, observers coded the time the event began and the time the event ended —
in minutes. For the current analyses, the time spent on each event during each
observation session was summed to give a total observed time for each ofﬁcer in minutes.

The initiation source for each event was also available - whether ofﬁcer initiated,
dispatch, on-scene citizen initiated, or other ofﬁcer initiated. This variable was used to
provide a total amount of “free” time that an ofﬁcer had per ride. Free time was deﬁned
as time that did not involve being dispatched, engaged with a citizen by citizen request, or
directed by a supervisor or other ofﬁcer to perform some duty or function. In other
words, the calculated free time was the summed time of all ofﬁcer initiated events per
observation session. For each ofﬁcer, a value was calculated indicating the amount of
free time that an ofﬁcer had divided by the total observed time. While this variable was
used to create subsequent dependent measures, it is useful in and of itself in examining
the possibility that ofﬁcers in some areas had more free time than officers in other areas.
This will be examined at the outset to determine if some areas placed higher demands on
ofﬁcers and in which areas officers had more self-directed (or free) time.

Indicators were also available that identiﬁed where events occurred. Only events
that occurred within the geographical boundaries of the assigned study beat were included
in the current analyses. For activities, information was available indicating the percent of
the activity that occurred within the study beat (0 - 100 percent); for encounters, geocodes

indicated the location of the encounter (by using cross-streets) and beat designations were

57

later tied to these geocodes. Again, while this variable is used in the subsequent creation
of the dependent measures, it should be useful to examine this variable independently and
before the primary analyses. Examination of the location and time variables can
determine if officers on some rides were more involved in (and/or restricted to) an
assigned beat while more time on others was spent outside of the assigned area.

The amount of free time that ofﬁcers had and the amount of time that ofﬁcers
spent inside their beat will be examined ﬁrst to determine what inﬂuences are able to be
identiﬁed for these. These preliminary analyses have implications in terms of selection
effects - that the dependent measures and primary analyses described below may be
affected by the amount of free time that officers had and the amount of time ofﬁcers spent
inside the assigned beat. Ofﬁcers may not be equal in the amount of free time or time
inside the beat that they had available to engage in the behaviors of interest, described
below.

Finally, for the creation of the dependent measures, indications were available
identifying the type of event observed. Events were placed into three categories: order
maintenance, legalistic, and service - inspired by the typology introduced by Wilson
(1968). Activity codes (and problem codes if available) were used to classify activities
and problem codes were used to classify encounters into one of the three categories.'7

Order maintenance activities involve situations where the police are functioning in
a peacekeeping capacity. Order maintenance type activities and encounters involve

situations in which a crime may not have been committed, but the potential for such a

 

‘7 Complete lists are provided in Appendix B (Tables 8.1, 8.2, 8.3) for the various police activities and
encounters, how they were categorized, as well as their distributions.

58

commission is present and possible. Order maintenance events also include events that
by their mere presence, police function to instill some order — such as general patrol.
Examples of other order maintenance behaviors include the relatively common practice of
settling disputes.

Legalist events revolve around law enforcement and crime control behaviors.
These events involve incidents of law violations (or suspected violations). As discussed
previously, the majority of previous research has focused on legalist activities or
orientations of the police (e.g., arrests). The events included in this third category of
behaviors are the most varied - due to the speciﬁcity of problem codes indicating the
precise violation or issue addressed.

Service behaviors involve and indicate behaviors that anyone could perform; the
full weight of the state and authority given to the police are not necessary to attend to
these functions. As Wilson (1968) notes, “it is only a matter of historical accident and
community convenience ,
that [these services] are provided by the police” (p. 5). Examples of service behaviors
include helping someone locked out of a car or house, transporting a citizen, or giving
directions.

It is important to note when considering speciﬁed activities, that certain activity
categories were not mutually exclusive. Some activities are not considered in and of
themselves to be one of the three types of behavior. Activities such as meetings with
non-police service providers and checking or ﬁxing property were exclusively
categorized as service, while other activities were placed into a speciﬁc category based

upon the focus of the problem (i.e., problem codes). For example, backing up other

59

ofﬁcers, information gathering, or “checking out the situation” activities were classiﬁed
as ‘service’ only if the activity was tied to a service problem code considered of a service
nature.18 Problem codes provided more detail and speciﬁcs than did activity codes,
therefore, they were more amenable to classiﬁcation; the categorizations of problem
codes were mutually exclusive.

Values were calculated that indicated the percentage of observed, officer initiated
time spent on each variety of behavior in a study beat, out of the total observed free time.
Free time was used as the denominator in an effort to determine how ofﬁcers were
choosing to spend the time they had that was not otherwise obliged.

Ofﬁcer initiated service, order maintenance, and legalist events are the focus of
the current analyses. The use of oﬂicer initiated events is useful in uncovering the
discretionary activities engaged in by ofﬁcers in different areas — as opposed to
examining all activities which would include what citizens or supervisors request.

Table 2 provides descriptive information about the variables used to create the
three dependent measures. Descriptive statistics are provided for the total amount of
observed time for each ofﬁcer, time inside and outside the assigned beat, ofﬁcer and
other-initiated time, total time on each of the three behaviors comprising the dependent
measures, and ﬁnally, the dependent measures: percent of free time on each type of
behavior inside the beat.

As mentioned previously, in addition to the three types of behaviors that are the

focus of the current paper, the total amounts of time spent within and outside of an

 

'8 The same is true for order maintenance and legalist events. For ambiguous activities such as those listed,
placement of the event into one of the three categories was dependent upon a problem code consistent with
the respective category.

60

assigned beat were available for examination. Also, summed times engaged in citizen or
other ofﬁcer requested action — as well as ofﬁcer initiated time for each beat were
available. These measures will also be examined in an effort to provide a fuller picture of

how ofﬁcers’ time is spent and the policing patterns for different types of neighborhoods.

61

Table 2. Outcome Variables - and Variables Used in Calculations (n = 243 Officers)

Total observed time (in minutes)

Percent of observed time in assigned beat
(Percent spent outside of beat
Percent of observed time on Ofﬁcer initiated events

(Percent on other than Officer initiated events) _
Percent of observed time on ofﬁcer initiated events in beat

(Percent on officer initiated out of beat

Total time on Order Maintenance events (in minutes)
Total time on Legal events (in minutes)

Total time on Service events (in minutes)

Percent of free time on OM events in beat
Percent of free time on legal events in beat

Percent of free time on service events in beat

62

Mean SQ, Range

1,010.6 846.3 15to 4,698

46.1 26.7 0 to 90.8
53.9 23.7 9.2 to 100)
70.5 13.0 3.6 to 99.4

29.5 13.0 0.64 to 96.4
47.3 26.2 0 to 100

52.7 26.2 0 to 100)

293.7 229.2 Oto 1,569
185.1 178.9 Oto 1,202

28.2 87.2 0 to 1,009

24.3 17.3 0 to 80
5.5 6.3 O to 34.7

0.5 1.5 Oto 10.1

Measurement of the Independent Variables
Oﬁ‘icer Level Variables

A number of officer level variables believed to impact the delivery of police
services were available, speciﬁcally, ofﬁcer sex, race, age, unit, education, and an
attitudinal variable capturing “crime ﬁghting” orientation. These variables were available
from ofﬁcer surveys. As mentioned in Chapter Two, research has shown that police
practices vary by individual ofﬁcer characteristics as well as contextual factors. As such,
ofﬁcer characteristics were included as one level of analysis for the current research.

Several demographic variables were available from the ofﬁcer survey and
included in the current analyses. Table 3 contains descriptions, coding, and distributions
of these variables. In addition to the ofﬁcer demographic variables included, Table 3
describes the distribution of officer unit. For the current study, police ofﬁcers were
identiﬁed as either 911 officers or Community Police Ofﬁcers (CPOs). Differences are
expected between the two types of ofﬁcers.

Speaking generally, CPOs were freed (for the most part) from answering dispatch
calls for service. As Parks et al. (1999) found using these same data, CPOs may spend
less time with citizens than 911-responders. In addition, the persons with whom CPOs
have contact have been described as a “better class of people” (e. g.,“non-criminals,”
community leaders, neighborhood organizers, and business people) (Parks et al., 1999).

One attitudinal variable considered important for inclusion was also available —
officer’s crime ﬁghting orientation. Ofﬁcers were asked to identify on a likert scale how
strongly they agreed with the following statement: “Enforcing the law is by far a patrol

ofﬁcer’s most important responsibility.” This variable was included in an attempt to

63

capture differences in ofﬁcers’ attitudes and identify priorities that may predispose an
ofﬁcer towards engaging in one type of behavior over another — i.e., law enforcement
over order maintenance activities or service activities.

Ofﬁcers working different days of the week or different times of the day are
confronted with different situations and problems. Some days of the week and some
shifts are busier than others, late week and weekend days and night shifts particularly.
Ofﬁcers working busier days of the week or working busier shifts would have less free
‘ time and therefore may be less likely to engage in activities due to the workload (expected
or real). In other words, ofﬁcers working less busy days or shifts may not be concerned
that they will be swamped with calls and therefore may engage in different activities than
ofﬁcers scheduled on busier days and shifts. To control for these possibilities, shift times
and the percent of free time that an observed ofﬁcer had were included as control
variables at the ofﬁcer level.

Shift times were calculated using the start time for ofﬁcers’ shifts. Since start
times varied across departments, values for shift were categorized into day shift, evening
shift, and night shift. Day shifts included start times ranging from 5:30 am to 11:10 am;
evening shifts included start times ranging from 1:30 pm to 4:09 pm; night shifts included
start times from 7:00 pm to 12:55.” For the multivariate analyses, dummy variables were

created for evening and night shifts.

 

'9 Thirteen ofﬁcers bad start times that varied across shifts (5.3% of all observed ofﬁcers). For these
ofﬁcers, the modal start time category was used.

64

Percent of free time was calculated as all officer initiated events of an observed
officer divided by the total amount of observed time. Table 3 contains descriptive
statistics for this and the other officer level variables described above.

As can be seen in Table 3, most ofﬁcers were male (84 percent) and white (77
percent). The mean age for officers was approximately 35 years. The majority of the
observed ofﬁcers were general response, or 911 ofﬁcers (70 percent). Around one-third
of the observed ofﬁcers had a college degree; over 80 percent had some college or higher.
Regarding the attitudinal variable, 81 percent agreed (somewhat or strongly) with the

statement that crime ﬁghting was the most important responsibility of an ofﬁcer.

Table 3. Ofﬁcer Characteristics (n = 224)

 

 

Variable Coding n %

Female 0 - male 187 83.5
1 - female 37 16.5

Minority O - white 173 77.2
1 - minority 51 22.8

Age Mean: 34.6 SD: 7.0 Range: 24 to 65

CPO 0 - 911 Ofﬁcer 155 69.2

1 - Community Police Officer 69 30.8

65

Table 3 (cont’d).

 

Variable Coding n %
Education 1 - Less than High School 2 0.9
2 - High School or GED 34 15.2
3 - Some college (no degree) 50 22.3
4 - Associates degree 15 6.7
5 - More than 2 years of college 40 17.9
6 - Bachelors degree 73 32.6
7 - Some graduate school 10 4.5
Crime ﬁghting orientation
1 - disagree strongly 12 5.4
2 - disagree somewhat 30 13.4
3 - agree somewhat 115 51.3
4 - agree strongly 67 29.9
Shift“
0 - Day 88 36.2
1 - Evening 66 27.2
2 - Night 89 36.6
% free time" Mean: 70.5 SD: 13.0 Range: 3.6 to 99.4

 

"' includes all unique ofﬁcer-in-beat combinations (n = 243)

66

Neighborhood Variables 2°

Due to the limited number of neighborhood units available for observation (n =
24) the number of variables at the neighborhood level is restricted. Several contextual
variables believed to impact how police deliver services were available for the current
study. Some of these neighborhood level variables have been included in previous
research efforts examining differences in police behavior, others have rarely been
included. The neighborhood variables described below were drawn from 1990 census
data.21 These data were aggregated to the beat level.22

Buckner (1988) discusses the aggregation of individual level data and states that if
individual level scores on variables are aggregated and averaged, the resulting mean score
is a measure of the collective of community residents. The aggregated variable is said to
describe a setting-level attribute (Buckner, 1988). A5 M005 and Lemke (1983) note, the
“character” of an environment depends in part on the typical characteristics of its
members. Brodsky et al. (1999) echo this statement stating that the combined responses

of individuals in a setting contribute to that community’s overall character. According to

 

2° An important issue faced when examining community level variables is delineating appropriate and
meaningful boundaries. The notion of neighborhood possesses the same vague qualities as the notion of
community (e.g., what exactly constitutes a neighborhood '2). The deﬁnitions of both are necessarily
subjective. For the current project, the delineation of neighborhoods was established by the police
departments and as K1inger(1997) notes, this makes identifying the ecological unit of interest less
problematic than in other inquiries as police districts and beats contain clearly deﬁned boundaries.
Mastrofski et a1. (2002) address this issue, recognizing the complexities involved in deﬁning a
neighborhood and stating that “in both departments efforts had been made to draw beat boundaries
consistent with the community’s sense of where neighborhoods began and ended” (p. 4).

2‘ Initially, variables drawn from residential surveys, gathered as part of the POPN project were to be used.
Speciﬁcally, levels of mobility/stability, satisfaction, involvement, and heterogeneity were obtained for each
study beat. These data presented problems and the inclusion of these initial variables was reconsidered. In
place of these, census data was examined and chosen for inclusion; census data proved to be more reliable.
Appendix C outlines the initial variables under consideration and the problems that arose from them.

22 Beat is used to refer to the smallest geographical police service area. The SPPD used the term
Community Policing Area (CPA) to denote this.

67

Shinn (1990), the aggregating of individual units assumes that individual biases cancel
each other out and that a more reliable, valid picture of the community as a whole
emerges.

Individual level census information from neighborhood residents was aggregated
to the neighborhood level in order to provide information and differentiate areas on key
contextual variables. The measures selected for the current study are believed to be
indicative of different types of neighborhoods in terms of economic and social factors.

The neighborhood variables included are theoretically rooted in conﬂict theory.
Speciﬁcally, the variables tap into economic and racial indicators that are relevant when
examining how police respond to different neighborhoods from a conﬂict theoretical
perspective. These variables are similar to contextual level variables used by previous
researchers testing macrosociological inﬂuences of police activity. Williams and Drake
(1980) describe the importance of dominant/subordinate relationships inherent in conﬂict
theory and historically relevant measures such as percent black and percent unemployed.
With the premise of conﬂict theory being power differentials between groups, racial and
economic indicators are particularly salient. Differences in poverty rates, minority status,
and economic equality have commonly been included when testing conﬂict theory (see
for example, Hepburn, 1978; Jacobs and Britt, 1979; Liska and Chamlin, 1984; Warner,
1997). The variables described below were designed to capture these differences. Table

4 describes the neighborhood level variables included in the current analyses.

68

Table 4. Neighborhood Characteristics (n = 24)

 

Variable
n °/_o

Site

1 = Indianapolis 12 50

2 = St. Petersburg 12 50

M4111. & R_an_ge

Concentrated disadvantage“ 1.9 0.87 0.63 to 3.5
Percent homeownership 39.66 15.30 2.0 to 66.1

 

"' Factor score: % poor, % unemployed, % female headed family, % African-American

Site. Site was included primarily as a control variable due to differences that
might manifest between study sites. While it could be argued that site is an
organizational level, it is included at the neighborhood level for two reasons: 1) it places
neighborhoods within a larger context — neighborhoods have the quality of being in one
of the two speciﬁc jurisdictions, and 2) increasing the number of levels in the models
increases the complexity of the models and potentially confuses the primary focus of
identifying contextual level inﬂuences of police behavior.

Concentrated Disadvantage. Concentrated disadvantage is a factor score from
four census items: percentage poverty, percentage unemployed, percentage female headed
families, and percentage African-American. Concentrated disadvantage has been used by
previous researchers to distinguish between neighborhoods. Sampson et a1. (1997)
examined neighborhood level inﬂuences and included a concentrated disadvantage factor

similar to that outlined above. Sampson et a1. discussed the importance of resource

69

distribution and economic stratiﬁcation that included income, racial, and familial
indicators (e.g., female headed households). Sampson et al. (1997) noted the cumulative
effect of these forms of disadvantage.” These items loaded onto the factor included with
an eigenvalue of 3.13 (See Table 5). The factor had a skewed distribution and, as had
been done with previous use of this factor, a constant (1.5) was added to the term to
eliminate negative values and adjust the distribution (Morenoff, Sampson, and
Raudenbush 2001; Terrill and Reisig, 2003). The communalities and factor loadings for

this variable are presented in Table 5.

Table 5. Disadvantage Factor

 

Factor

Item Communalit_v_ Ladig
Percent poor 0.84 0.92
Percent unemployed 0.76 0.87
Percent female headed families 0.83 0.91
Percent African-American 0.71 0.84
Eigenvalue 3.13

Percent of Variance ermlained 78.33

 

 

23 The inclusion of percent African-American in the factor variable described as concentrated disadvantage
is not intended to describe any inherent meaning in terms of “disadvantage”. Percent African-American
was included for three reasons: 1) the factor loading and communality indicated a good ﬁt with this factor;
2) racial indicators have been included with this factor by previous researchers; and 3) as described above,
for conﬂict theory, racial indicators are theoretically relevant and in conjunction with the other items for
this variable, indicate a cumulative form of disadvantage.

70

The inclusion of concentrated disadvantage as a neighborhood level variable is
designed to identify areas that are socially and economically challenged. Previous
research outlined in Chapter Two indicates that these areas are subjected to different
types of policing. Additionally, previous research using this speciﬁc measurement has
indicated an effect on homicide rates (Morenoff et al., 2001) and also on police use of
force (Terrill and Reisig, 2003).

Percent Homeownership. In addition to tapping negative social indicators,
indications of positive social control were also sought. As noted by Morenoff et al.
(2001), a reliance on indications of concentrated disadvantage may obscure protective
effects of more stable affluent neighborhoods. One indication of higher levels of stability
captured and used for the current study was the percentage of homeowners out of all
housing units. Previous research has identiﬁed home-ownership as a key inﬂuence of
attachment and stability (Fried, 1982; Riger and Lavrakas, 1981; Taylor et al., 1985).
Neighborhoods that have a higher percentage of homeowners are judged as exhibiting

higher levels of attachment and resulting higher levels of social control.

Analytic Techniques

To consider the possible contextual effects of neighborhood level characteristics
on police ofﬁcer discretionary behavior, several analytic procedures will be used.
Chapter Four begins by examining the distributions of the independent variables and
determining if potential problems of multicollinearity exist. Multicollinearity would
cause logical and statistical problems; logically, redundant variables would be useless to

include. Statistically, multicollinearity creates problems with unstable estimations of

71

error terms (Tabachnick and Fidell, 2001). Zero-order correlation matrices will be
provided and tolerance statistics calculated for the independent variables and the three
dependent variables.

Next, an examination of the dependent variables will be conducted. In addition to
analyzing the three outcome measures described above, preliminary analyses will be
conducted to determine if inﬂuences of free time and time inside the beat can be
identiﬁed. Using the same independent variables, these two measures will be designated
as outcome measures to determine if ofﬁcer demographics, assignment, attitude, or
neighborhood characteristics inﬂuence the amount of free time that ofﬁcers had or the
amount of time that ofﬁcers spent within the boundaries of their assigned beat.

After examining amount of free time and time spent inside the beat, attention is
turned to the three dependent measures. It is important to examine the distribution of the
dependent variables and determine if these distributions approximate normality or if
problems of overdispersion exist. The dependent variables will be examined both
statistically and graphically - the former by examining the skewness and kurtosis; the
latter by examining histograms of the variables with a normal distribution overlay
(Tabachnick and Fidell, 2001). If there are indications of non-normality, data
manipulations to normalize the distribution will be conducted (e.g., using natural logs).

After these analyses, bivariate correlations between the independent variables and
the outcome measures will be examined. Mean group comparisons and differences
between the categorical independent variables and the three dependent variables will be
examined using One-way ANOVAs; signiﬁcant differences identiﬁed using F-ratios

(Agresti and Finlay, 1986).

72

Next, the data will be analyzed using traditional linear techniques, speciﬁcally,
Ordinary Least Squares regression models.“ Predictor variables from both levels will be
entered into these models.

Since the data under examination involve variables at multiple levels - ofﬁcer
level information (e.g., officer race, sex, unit, etc.) which are nested within the
community level (in this case, police beat — e.g., neighborhood level variables),
hierarchical linear modeling (HLM) will be used. HLM allows analyses of multilevel
models. In the present case, there are two levels of data under consideration: (1) officer
level variables —- Level 1, and (2) neighborhood level data - Level 2. Before this more
advanced analytic technique can be performed, the data ﬁrst need to be examined to
determine the suitability for these procedures and diagnostics need to be performed.

The unit of analysis is the ofﬁcer, and these ofﬁcers were observed in twenty-four
study beats repeatedly. This being the case, neighborhood level variables and values
would have to be duplicated in each case. These repeated observations at the community
level make the use of traditional linear analytic models questionable. With multilevel
investigations, individuals nested within groups are exposed to the same stimuli and are
therefore expected to be more similar to one another than individuals in other groups
(Hofrnann and Gavin, 1998). Traditional linear regression techniques, such as OLS,
assume that the random errors are independent, normally distributed, and have constant
variance (I-Iofmann and Gavin, 1998). Traditional linear models therefore underestimate
the size of the standard error (Bryk and Raudenbush, 1992). With HLM, each level in the

structure is formally represented by its own submodel which expresses relationships

 

2‘ The choice of analytic techniques will be dependent upon the normality of the distribution of the
dependent variables and the appropriateness or inappropriateness of traditional linear techniques.

73

among variables at that level and also speciﬁes how variables at one level inﬂuence
relationships at another level (Bryk and Raudenbush, 1992).

With HLM, parameter estimates from the ﬁrst level are subsequently used as
outcome variables in the second level. For the second level, the outcome measures are
modeled as a ﬁmction of group level variables (Hoﬁnann and Gavin, 1998).

Outlined below are the equations used for HLM. Both Level 1 and Level 2

models are presented, followed by a combined model.

Level 1 model (ofﬁcer level):
Y; = b0 + bIXi + r;
where:

b0 = the intercept (expected value of Y if X = 0)

bl = the slope (expected change in Y with 1 unit increase in X

r = the error term

HLM allows consideration of multiple neighborhoods (Level 2). Below is a Level 2
equation, for j neighborhoods; the equation describes the relationship within any

neighborhood j as follows:

Yij=b01+blj (Xv-.1) +171

74

where:

subscript j : = each neighborhood unique intercept and slope.

The strength of HLM is that it allows the consideration of both Level 1 and Level
2 variables in a combined model for hierarchical data.

Combined model:

Yij = 70 o + 'Yo le + 710(xij '_j) + Y11Wj(Xij '_-j) + qu +

”nix ij '_J) + rij
where:

y = Level 2 coefﬁcients

70 0 / 71 0 = level 2 intercept terms

70 1 / ’Y] 1 = level 2 slopes

W = Level 2 predictor

uoj and 111 j = the Level 2 random effects

rij = the Level 1 random effects

Following the description of Bryk and Raudenbush (1992) the steps of the HLM
analyses will ﬁrst involve a determination of the appropriateness of HLM for the current
data. A One-way ANOVA will be performed involving Level 1 data; this will provide
descriptive information as to the sample means and also indicate the reliability of the

sample mean as a representation of the true mean. This step will determine the

75

applicability of the data to hierarchical linear modeling. If it is determined that HLM is
not the most appropriate technique for the available data, other regression techniques or

adjustments will be used to evaluate the data and judge the hypotheses outlined below.

Hypotheses
Oﬂicer Level Variables

It is expected, based on previous research ﬁndings, that different ofﬁcers will
behave differently based on individual ofﬁcer characteristics. Considerable research has
been conducted examining this premise (for a review, see Riksheim and Chermak 1993;
Sherman, 1980). The present research adds to this long line of research by including
individual ofﬁcer level variables and examining the relationship of these variables to the
levels of ofﬁcer initiated order maintenance, legalistic, and service behaviors. The
following hypotheses are presented regarding ofﬁcer level variables and the three

dependent measures.25

Hypothesis 1 - Oﬂicer sex: Female ofﬁcers engage in less order maintenance

behaviors, less legalistic behaviors, and more service behaviors.

While previous research has been mixed regarding ofﬁcer sex and behaviors,
there is some evidence to support the idea that male and female ofﬁcers behave

differently. Bloch and Anderson (1974) found that female ofﬁcers made fewer felony and

 

25 While speciﬁc hypotheses are not posited for all ofﬁcer level variables, these variables are included in the
analyses. Variables such as ofﬁcer education levels and ofﬁcer race are traditionally included in such
research, were available, and therefore included in the current analyses.

76

misdemeanor arrests and this ﬁnding was supported by the ﬁndings of Sherman (1975).
Sherman (1980) cites research indicating that female ofﬁcers are less likely to initiate
encounters. On the other hand, Homant and Kennedy (1985) found that females were

more likely to get involved in disturbances.

Hypothesis 2 — Oﬂicer age: Older officers engage in less legalistic behavior;

stated conversely, younger ofﬁcers will engage in more legalistic behavior.

Age and length of service are difﬁcult to differentiate from each other as they are
typically (and with the current data) highly correlated — older ofﬁcers have more
experience. Several researchers have found differences between younger less
experienced ofﬁcers and older ofﬁcers with more experience. Sherman (1980) described
ﬁndings indicating that officers with less experience engage in more law breaking
detection activities (legalistic) and also engaged in more preventive patrolling (order
maintenance). F orst, Lucianovic, and Cox (1977) found that less experienced ofﬁcers
made more arrests. Worden (1989) found that more experienced ofﬁcers made fewer

trafﬁc stops and also fewer stops of suspicious persons.

Hypothesis 3 — Ofﬁcer unit: Community Police Ofﬁcers will engage in more order

maintenance and service oriented policing and less legalistic policing.

Community policing, in general, recognizes a broader mandate for police — more

than a simple focus on law enforcement activities. As such, it is expected that

77

community police ofﬁcers will spend less time engaged in legalistic behaviors and more
time on order maintenance and service policing.

CPOs, generally freed from dispatched calls, have more discretion in how they
spend their time. These ofﬁcers should have more free time (which can be tested) and
therefore may be more likely to devote that time to service or order maintenance policing
— instead of limiting what they become engaged in due to expectations of being
interrupted by being dispatched. This premise can be further examined by controlling for

the amount of free time that the ofﬁcers in different units have.

Hypothesis 4 — Attitudes: A positive relationship is expected between law
enforcement attitudes (enforcing the law as the top priority of ofﬁcers) and

legalistic behaviors.

While previous research has provided weak support for the link between attitudes
and behaviors, an ofﬁcer level attitudinal variable was included in the current research in
an attempt to control for preferences of ofﬁcers that might be present. In other words,
ofﬁcers who viewed their primary role as enforcing laws will choose to spend more of
their free time engaged in more legalistic activities. Altemately, ofﬁcers who expressed
that enforcing the law was not the highest priority of police will engage in more diverse

behaviors, including order maintenance and service activities.

Hypothesis 5 — Free time: Ofﬁcers who have more free time will engage in more

order maintenance, legalistic and service policing.

78

It is expected that ofﬁcers with more free time will use this free time to engage in
a variety of activities — including the three used for the present research as outcome
measures. Logically, officers with less ﬁee time will have less time to devote to any
activity. Also, ofﬁcers with less free time could be less inclined to become involved in

self-initiated events as these may be interrupted by dispatched calls.

Neighborhood Level Variables

The main premise of the current research is that police function differently in
different areas. While previous studies have examined a set of speciﬁc police behaviors
and a limited number of key contextual characteristics, much has been left unexamined.
The current study attempts to include key contextual variables believed to impact the
delivery of a variety of police behaviors.

In short, it is believed that police behave differently in different neighborhoods,
acting more legalistic in some than others and providing more services in some than
others. As described above, it is expected that in areas characterized as having more
concentrated disadvantage, police will provide fewer services and engage in more order
maintenance and legalistic behavior. In areas with higher percentage of homeowners
(versus renters) — indicating more attachment and commensurate higher levels of informal
social control - it is expected that the police will provide more service functions and
engage in less order maintenance and legalistic behavior. The expected relationships
revolve around the idea that the police compensate for the lack of other controls. As

such, the following hypotheses are presented:

79

Hypothesis 6 — Concentrated disadvantage: A positive relationship is
hypothesized between concentrated disadvantage and both legalistic and order
maintenance policing. A negative relationship is hypothesized between

concentrated disadvantage and service policing.

Hypothesis 7 — Percent homeowner: A negative relationship is hypothesized
between percent homeowner and both legalistic and order maintenance policing.
A positive relationship is hypothesized between percent homeowner and service

oriented policing.

In the next chapter, these hypotheses will be tested and results of the analyses
(i.e., bivariate correlations, AN OVAs, traditional linear regression analyses, and HLM)
presented. In Chapter Five, these results will be interpreted, implications drawn, and

directions for further research discussed.

80

Chapter 4
Analyses

As described in Chapter Three, the ﬁrst steps of the data analyses involved
determining the suitability of the data for the proposed analytic techniques. To this end,
several diagnostic procedures were performed. Speciﬁcally, collinearity checks were
performed on the Level 1 data to ensure that correlation problems did not exist among the
independent variables chosen for the current analyses. Collinearity diagnostics were also
performed for the Level 2 data to determine if the items selected for inclusion were inter-
correlated. For the current analyses, the number of Level 2 variables is limited since there

were only 24 units of observation at Level 2 (i.e., 24 neighborhoods).

Bivariate Correlations Among Predictor Variables

First, correlational analyses for Level 1 data (i.e., officer level) were examined. A
correlation matrix was produced that included all of the ofﬁcer level variables (i.e., sex,
race, age, unit, education, crime ﬁghting orientation, shift, and percent free time). These
results can be seen in Table 6. While there were correlations that were statistically
signiﬁcant, none reached or exceeded the generally accepted cut-off point of .80. Table 6
does indicate that ofﬁcer sex and race were signiﬁcantly correlated, as were ofﬁcer age
and education. The shift indicators were signiﬁcantly correlated to several other ofﬁcer
level variables. Day shift was positively correlated with female ofﬁcers, minority
ofﬁcers, age, CPOs, and amount of free time. Day shift was negatively correlated with
ofﬁcer education. Evening shift was negatively correlated with ofﬁcer age and free time.

Night shift was negatively correlated with female ofﬁcers, minority ofﬁcers, age, and

81

CPOs. Percent free time was positively correlated with day shift and negatively
correlated with evening shift. None of these bivariate correlations was large enough to
indicate concerns of collinearity.

Additional test for indications of multicollinearity involved tolerance statistics
(See column 12, Table 6). Tolerance statistics indicate the percentage of variance of
speciﬁc predictor variables that is not explained by the other predictor variables.
Tolerance statistics were calculated for all of the independent variables for each of the
three dependent variables. Tolerance statistics range from 0 to 1 and lower values
indicate problems (< .20) (Memard, 1995). Tolerance levels for these Level 1 variables

also indicated no problems with multicollinearity.

82

 

 

Eva: 3v?
3.3 2: 2.? :2? :83 3.? 23 8.3 3.3 23 8.3 2:: noes 2
8.3 2: :3? :3? 8.? 8.3 :2? t 3.? .2? t a? 25.322 3
3.3 2: :8? :3 3.3 8.3 :2? 3.? 23 2.355% a
83.35 2: 3.? .2? .23 :53 :23 .23 2528 a
qozﬁcoto
3.3 2: 2.? 3.? 2.? 8.3 8.? magma 2:5 o
23 2: 8.3 :3? 3.3 3.? 828:8 m
3.3 2: 3.3 3.? 3.3 0,6 4
23 2: 3.? 3.? ow< m
3.3 2: .23 2352 m
23 2: 235a _
335m 2 a a a e m 4 m N _
3:828. oE~ta>

 

mo_aa_._a> 364 .5050 ”8:235 35.5.8. can 356580 noun—8.80 ect,—OracoN .w Bank

83

Next, Level 2 (i.e., neighborhood) variables were examined. Bivariate correlation
coefﬁcients are provided in Table 7. As with the ofﬁcer level variables, these Level 2
variables did not present problems due to high inter-correlations. In fact, none of the
correlations was signiﬁcant at the p = .05 level. Tolerance statistics were also calculated
using aggregated / summed time variables at the neighborhood level; these also indicated

no problems of multicollinearity.

Table 7. Zero-Order Correlation Coefﬁcients of Neighborhood Variables (n = 24)

 

Tolerance

Variable 1 2 3 Statistics
1 St. Petersburg 1.00 0.82
2 Disadvantage factor 0.00 1.00 0.85
3 Percent homeowner -0.40 "‘ -0.35 1.00 0.72

 

*p=.052

Preliminary Analyses - Free Time and Time in Beat

To determine if potential selection effects were inﬂuencing the sample of ofﬁcers
observed, preliminary analyses of the amount of free time that officers had and the
amount of time spent inside the study beat were examined. It was necessary to determine
the inﬂuences of these in order to have an accurate and complete picture of how ofﬁcers

were spending their time.

84

Free time

As mentioned in Chapter Three, while the primary focus of the current research
involves the three types of behavior engaged in by the police, also of interest is the
amount of free time that officers had as well as the time that ofﬁcers spent within their
assigned study beat. Bivariate and multivariate analyses were performed using these
calculations as dependent measures.

First, the amount of free time was examined to determine if any of the included
variables predicted this. As mentioned, the mean percent of observed free time was 70.5
(SD. 13.0) and ranged from 3.6 percent to 99.4 percent. For amount of free time, only
ofﬁcer unit, shift, and the neighborhood level variables were included - as there was no
reason to believe that an ofﬁcer’s demographic information would determine how often
dispatch operators or supervisors directed the ofﬁcer to engage in some activity?" The
results of the bivariate correlations are presented in Table 8. As Table 8 indicates,
bivariate correlations indicate that ofﬁcers on day shifts had more undirected (i.e., free

time) and ofﬁcers on the evening shifts had less free time.

 

26 Just to be certain, a separate model was run that included all of these variables and none reached
signiﬁcance.

85

Table 8. Bivariate Correlations - Free time

 

% Free Time

 

Variable r (Sig)
Oﬂicer variables
CPO 0.03 0.640
Day shift 0.26 0.000
Evening shift -0.17 0.007
Night shift -0.10 0.107
Neighborhood variables
St. Petersburg 0.06 0.362
Concentrated 0.01 0.85 1
disadvantage
Percent homeowner -0.11 0.086

 

A series of One-way ANOVAs were also calculated to examine the categorical
variables (i.e., ofﬁcer unit, shift, and site) and potential signiﬁcant differences for officer
amounts of free time. As Table 9 indicates, the ANOVAs indicated that shift was
signiﬁcant — officers with day shift assignments had signiﬁcantly higher mean free times

than other shifts; ofﬁcers on evening shift had the least amount of free time.

86

Table 9. One-way ANOVA - Free time

 

 

Variable Mean SD.

Unit
911 70.21 12.21
CPO 71.05 14.64
F -ratio 0.22

Shift
Day 75.02 14.22
Evening 66.79 12.50
Night 68.70 10.71
F -ratio 9.48 ***

Site
Indianapolis 69.99 13 .01
St. Petersburg 71 .67 12.95
F-ratio 0.83

*** p < .001

87

Table 10 provides the results of an OLS regression model with the percentage of
free time as the outcome measure. As with the bivariate correlations and the AN OVAs,
shift was the only variable that was signiﬁcantly related to the amount of free time that
ofﬁcers had. Officers who were observed on evening shifts and night shifts had
signiﬁcantly less free time. This ﬁnding should be kept in mind when examining the
speciﬁed outcome measures — time on order maintenance, legalist, and service events.

Ofﬁcers on different shifts did _n_q_t have the same amount of free time to potentially

devote to any or all of these three types of behaviors. Ofﬁcers assigned to different units

and in different types of neighborhoods did have similar amounts of free time.

Table 10. OLS Regression Results - Free time

 

 

Variable B S.E.
Constant 80.52 *** 4.79
Oﬁ‘icer variables
CPO -1.98 2.29
Evening shift -8.61 *** 2.07
Night shift -6.44 *** 1.93
Neighborhood variables
St. Petersburg 1.79 2.48
Concentrated Disadv. -0.29 1.09
Percent own 011 0.07
R squared 0.09

 

‘p<.05; **p<.01;***p,<,001

Time in Beat

Next, the amount of time that officers spent within and outside their assigned
study beats was examined. All of the variables used were included in these analyses.
First, bivariate correlations were calculated, followed by a series of one-way AN OVAS,
and lastly an OLS regression model using percent of time in beat as the dependent
variable. Recall, that the mean percent of time spent inside an assigned study beat was
46.1 (SD. = 26.7) and ranged from 0 to 90.8 percent.

88

Table 11 provides the bivariate correlations of both ofﬁcer level and
neighborhood level variables and the amount of time that ofﬁcers spent within the
assigned study beat. Officer age, unit, shift, and site had signiﬁcant bivariate correlations.
The bivariate correlations revealed that older officers, CPOs, ofﬁcers on day shifts, and
ofﬁcers in St. Petersburg spent less time within the assigned study beat.

Next, a series of one-way AN OVAs were produced examining group differences
for the amount of time ofﬁcers spent within their beat. As with the bivariate correlations,
ofﬁcer unit was signiﬁcant - 911 ofﬁcers spent a signiﬁcantly larger portion of the
observed time inside the study beat than CPOs. Officer shift was also signiﬁcant —
ofﬁcers on day shift spent less time in the assigned beat than officers on evening or night
shift. Also, ofﬁcers in Indianapolis spent more time inside the assigned beat. Age, which
was signiﬁcant at the bivariate level failed to reach signiﬁcance though it was close (p. =

.078) — with younger ofﬁcers spending more time in the assigned beat than older ofﬁcers.

89

Table 11. Bivariate Correlations - Time in beat

 

 

Variable r (Sig)
Ojﬁcer variables
Female -0.08 0.224
Minority 0.01 0.882
Age -0.16 0.015
CPO -0.36 0.000
Education 0.06 0.333
Attitude -0.01 0.897
Day shift -0.18 0.005
Evening shift 0.10 0.107
Night shift 0.08 0.188
Percent free time 0.00 0.959
Neighborhood variables
St. Petersburg -0.31 0.000
Concentrated 0.06 0.314
disadvantage
Percent homeowner -0.01 0.833

 

90

Table 12. One-way AN OVA - Time in beat

 

Variable Mean SD.
Oﬂicer level variables

Sex
Male 46.88 23.56
Female 41.94 24.24
F-ratio 1.48

Race
White 45.93 23.87
Minority 46.49 23.26
F-ratio 0.02

Age
30 and younger 50.63 21.07
31-40 44.53 24.08
41 and older 41.74 26.20
F -ratio 2.57

Unit
911 Ofﬁcer 51.79 22.56
CPO 33.45 21.24
F-ratio 35.76 ***

Education
Less than HS 59.64 1.19
HS / GED 43.17 24.53
Some college 44.02 22.94
Associates degree 50.35 22.10
2 + years college 46.96 25.92
Bachelors degree 45.51 23.40
Some graduate school 57.73 22.01
F -ratio 0.82

91

Table 12 (cont’d).

 

Variable Mean SD.
Law Enforcement Orientation
Strongly disagree 42.53 29.63
Disagree somewhat 44.69 25.70
Agree somewhat 48.32 22.94
Agree strongly 43.58 22.92
F -ratio 0.78
Shift
Day shift 40.38 25.82
Evening shift 50.06 21.17
Night shift 48.69 22.42
F -ratio 4.12 *
Neighborhood level variables
Site
Indianapolis 50.76 24.36
St. Petersburg 34.41 17.25
F-ratio 26.18 ***

 

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

Lastly for time in beat, an OLS model was produced using both ofﬁcer level and
neighborhood level characteristics as predictor variables and percent of time inside the

study beat as the outcome variable. Results are presented in Table 13.

92

Table 13. OLS Regression Results - % Time in Beat

 

 

Variable B S.E.
Constant 69.11 "* 18.41

Oﬂicer variables
Female -3.93 3.89
Minority -0.23 3.56
Age -0.34 0.23
CPO -12.59** 4.03
Education 0. 12 0.91
L.E. orientation -0.79 1.78
Evening shift 7.79 * 3.95
Night shift 3.09 3.76
Percent free time 0.08 0.11

Neighborhood variables
St. Petersburg -11.83 ** 4.37
Concentrated 0.67 1 .94

Percent homeownership -0.27 * 0.13

 

R squared 0.21

 

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

93

Several signiﬁcant predictor variables were found for time inside the beat. Ofﬁcer
unit was signiﬁcant; Community Police Officers spent signiﬁcantly less of their free time
in the assigned beat. Officers on evening shifts spent more time in the study beat and
ofﬁcers in St. Petersburg spent less time in the study beats. Ofﬁcers in neighborhoods
that had a higher percentage of homeowners also spent less time in the assigned beat.

These ﬁndings should be kept in mind when examining the subsequent analyses.
Only ofﬁcer shift was signiﬁcant in predicting how much free time that ofﬁcers had.
With the exception of ofﬁcers observed on day shifts, observed officers had similar
amounts of free time to which they could devote to engaging in the three sets of outcome
behavior measures. Observed officers generally had equal amounts of free time to engage
in the activities targeted for examination.

Ofﬁcers were n_o_t_ equal in terms of the amount of time that they spent inside their
study beat. Ofﬁcer age, assignment, shift, city, and neighborhood inﬂuenced the amount
of time spent inside the assigned beat and therefore result in different amounts of time
and opportunity to engage in the identiﬁed behaviors that are the primary focus of the

current research.

Dependent Variables - Distributions

Attention was next turned to the examination of the dependent variables.
Considering the possibility of non-normal distributions of the dependent variables, a
cursory look at the distribution of the dependent variables (see Table 2) indicates that
overdispersion could be a problem with the current data (for two of the three dependent

variables). This can be determined by examining the means and standard deviations. The

94

standard deviation for time spent on legalistic behaviors is greater than the mean (mean =
5.5, SD. = 6.3) and the same is true for time spent on service behaviors (mean = 0.6, SD.
= 2.5) indicating overdispersion. To examine this possibility further, calculations of
skewness and Kurtosis were conducted.

Skewness indicates the symmetry of a distribution while Kurtosis indicates the
peakedness of a distribution. When a distribution is normal, the values of each are zero
(T abachnick and Fidell, 2001). Additionally, histograms with a normal curve overlay are
provided (See Figures 1-3).

When the distribution of the three dependent variables were examined and
. statistics for normality calculated, indications of non-normal distributions were evident.
Skewness and Kurtosis values indicated problems with legalist and service behaviors -
two of the three dependent variables: order maintenance (skewness = 0.84; Kurtosis =
0.58), legalist (skewness = 1.62; Kurtosis = 3.35), and service (skewness = 7.73; Kurtosis
= 68.91). The high Kurtosis value for service was due to the considerable number of zero

values and the peak that this caused in the distribution.

95

Figure 1. Histogram — Time on Order Maintenance Events

 

30

 

 

       
   

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

20. m-
E
O E.__"r .
.1“! ,
, ,. .;.. -.
a 3 .2": is". .3;
. f], 2'. .u
o '
l git-J .0.
‘I-I «c: ':‘.-“
o " ‘ '-
.1» ~ 2‘;
“‘1‘“ . ‘4.v
‘. . -- ..'.'..
41: :1 .10" 1 . ..
0.5 -- : _. ..
I0 "t1 .1: <. . tr ;'='-
f; q 7. f; -. v__.j
‘4‘! ' ’ .1 '
(. ..- ' . ‘ ' . .- v
m.“ E : ' - .. ‘ 5‘
:“f: b. »' ' 1;:
5E“; ;,' .E .1. f 3".
" s”: I I" ‘ i E“:
3', . l j .
,n: w.
ctr: ,0.
n3; : '
i. -:'«- . -.'*.‘
5‘5; 1' ' E‘
13’5““ . ' '
7115:»?- ’- l '1 L . “
«r it 6’"! '1'.'. ..'.'_. - - .17.: ':i f]. xi. 2 _ “. ‘, -‘ -
O Liar n.3- 1'4: {1...}. tic-«ma; p " '~. . -.
t:- ->m--‘)t.:.~.!c r. f: 7.21 manta-3m atria-EJ131111 30:12:: 2.35.51. .‘ a :‘t‘. a. .1.” at...

4:"; “5‘!
g $53; $3!“ 1
.1‘"0

oqxxeeee 0220
'0 0 0-0‘r00-0‘r00-0‘E0700Y" 0 -0 ‘?00-0 "‘00-0 00-0

% free time on Order Maintenance events

Figure 2. Histogram - Time on Legalist events
80

 

#1m'. .4- ._. .- . .:
.‘3‘ '1 .1?! ‘ . new
i 'I_;“',.ur
. ._. .r .v‘

. . . . a _ ., ..
.rh. - . - -'-.. .._,

-1721???

nh

' .. .-"~..'
:1.
-.:.. .rt. 1.: . .
._. ,_ , J ..
:..-..

# of Ofﬁcers
‘3
are
twat.

 

. 12.0 16.0 E 24.0 ‘7 28.0
10.0 14.0 . 22.0 26.0 30.0 34.0

  

% free time on Legalist events

96

Figure 3. Histogram - Time on Service events

 

 

 

 

 

200
m
E , l
D
U l:
g 100.”
“-1
o .
t .
:2
0‘ m — ~-
o.00 ' 2.00 T 4.00 6.00 8.00 10.00
1.00 3.00 5.00 7.00 9.00

% ﬁ'ee time on Service events

The histograms produced (with a normal curve overlay) also indicated distribution
problems with the same two dependent variables (legalist and service). A number of
responses to address this problem were available. First, an attempt was made to
normalize the distribution of legalist behavior. To better approximate a normal
distribution, a natural log transformation was performed, after adding a constant (1.0) in
order to transform zero values. After the transformation, the mean value for legalist
events was 1.57, SD. = 0.90. As Figure 4 indicates, the distribution appeared much more
normal. Skewness and Kurtosis were also improved (0.1 and - 1.2 respectively; higher

Kurtosis due to large number of zero values — ofﬁcers not engaging in legalist behaviors).

97

Figure 4. Log Transformed Time on Legalist Events

 

# of Ofﬁcers

 

 

 

 

 

.63 1.13 1.63 2.13 2.63 3.13 3.63

Log transformed % free time on Legalist events

Time on service events also has a skewed distribution. As with the legalist
dependent variable, a constant was added and a log transformation was conducted as to
better approximate a normal distribution. Even after these had been conducted, time on
service events, though improved, still presented a skewed distribution (See Figure 5).
After transforming this variable, free time on service had a mean value of 0.85 (SD. =
0.34). The skewness was 2.99 and Kurtosis was 9.4 — again, due to the high number of
zero values (i.e., ofﬁcers not performing service activities). Considering the persisting
distribution problem, the hierarchical models run will be adjusted to take this into
consideration. Transformed values of legalistic and non-transformed values of service

will be used and adj ustrnents will be made to the multivariate statistical procedures used.

98

Figure 5. Log Transformed Time on Service Events

IQ
t:
1::

 

 

# of Ofﬁcers
2'

 

 

 

 

 

:75 1.00. 1.25 " fso 1.75 2.00 2.25 2.50
.88 1.13 1.38 1.63 1.88 2.13 2.38

Log transformed % of free time on Service events

The three dependent variables were next examined to determine if signiﬁcant

inter-correlations were present between these three types of behaviors. Table 14 provides

the correlations among the three dependent measures. As Table 14 illustrates, there were

no signiﬁcant correlations between the outcome measures.

99

Table 14. Inter-correlations Among Dependent Variables

 

 

Variable 1 2 3
1 O-M 1.00

2 Legalistic 0.06 1.00

3 Service -0.04 0.07 1.00

 

Bivariate Correlations - Predictor and Outcome Variables

Table 15 provides the correlations between ofﬁcer level (Level 1) variables and
the three dependent variables. The signiﬁcant bivariate correlations are in bold. For free
time on order maintenance events, ofﬁcer sex, ofﬁcer unit, shift, and percent of free time
had signiﬁcant bivariate correlations. Female ofﬁcers, community police ofﬁcers,
ofﬁcers on day shift, and ofﬁcers with lower percentages of free time spent less of their
free time engaged in order maintenance activities. Night shift and time on order
maintenance were positively correlated, indicating that officers on night shifts spent more
time engaged in order maintenance policing.

For time on legalist events, age and shift were the only variables that were
signiﬁcantly correlated with the percent of free time that officers spent on legalist events.
Younger officers spent more time engaged in legalist events. Day shift was negatively
correlated with amount of free time on legalist events and night shift was positively
correlated.

For service, ofﬁcer sex, shift, and percentage of free time were signiﬁcantly

correlated with time on service events. According to the bivariate correlations, female

100

ofﬁcers engaged in more service behavior than male officers, ofﬁcers on day shifts
engaged in more, ofﬁcers on night shift engaged in less, and ofﬁcers with more free time

engaged in more service events (See Table 15).

Table 15. Bivariate Correlations: Ofﬁcer Level and Dependent Variables

 

 

 

 

 

OM Legal Service
Variable r (Sig.) r (Sig.) r (Sig.)

Female -0.13 .040 -0.03 .688 0.15 .017
Minority -0.06 .370 -0.06 .359 -0.08 .247
Age -0. 12 .070 -0.25 .000 0.04 .522
CPO -0.32 .000 -0.05 .469 0.06 .372
Education 0.05 .413 0.12 .069 0.02 .764
Attitude -0.05 .436 0.07 .250 -0.1 1 .084
Day shift -0.24 .000 -0.23 .000 0.18 .006
Evening shift 0.10 .1 12 0.01 .830 0.01 .910
Night shift 0.15 .022 0.22 .001 -0.18 .004
Percent free time -0.13 .044 0.03 .671 0.21 .001

 

Next, neighborhood level variables were examined for bivariate correlations with
the dependent variables. Table 16 presents the bivariate correlations between the
neighborhood level variables and the three dependent variables. The signiﬁcant
correlations are in bold type. As can be seen in Table 16, St. Petersburg police engaged

101

in less ofﬁcer initiated order maintenance events and more service events.
Neighborhoods exhibiting higher levels of concentrated disadvantage were signiﬁcantly

correlated with more legalistic police behavior.

Table 16. Bivariate Correlations: Neighborhood and Dependent Variables

 

 

 

 

OM Legal Service

Variable r (Sig.) r (Sig) r (Sig.)
St. Petersburg -0.73 .000 -0.1 5 .489 0.49 .016
Conc. Disadv. 0.06 .796 0.50 .012 -0.1 1 .612
% homeowner 0.19 .379 -0. 19 .366 -0.11 .599

 

Analysis of Variances
Next, a series of One-way AN OVAs were calculated for the independent variables
at the ofﬁcer level and the three dependent variables. Tables 17 - 19 provide group mean

values and f-ratios for these variables.

102

Table 17. One-way AN OVA by Officer Characteristics - Order Maintenance

 

Variable Mean SD. n

Sex
Male 25.31 17.38 202
Female 19.22 16.35 41
F -ratio 4.26 *

Race
White 24.80 18.06 192
Minority 22.34 14.26 51
F -ratio 0.81

Age
30 and younger 27.30 16.60 82
31-40 22.41 17.04 114
41 and older 23.56 18.86 47
F-ratio 1.97

Unit
911 Ofﬁcer 27.98 17.02 167
CPO 16.15 15.17 76
F—ratio 26.98 ***

Education
Less than HS 29.54 1.91 2
HS / GED 24.57 19.46 38
Some college 21.92 15.61 54
Associates degree 29.30 14.95 16
2 + years college 21.65 14.86 42
Bachelors degree 24.41 18.01 80
Some graduate school 35.71 22.84 11
F -ratio 1.40

103

Table 17 (cont’d).

 

Variable Mean SD. n

Law enforcement orientation
Strongly disagree 20.26 15.93 14
Disagree somewhat 27.19 20.86 33
Agree somewhat 25.51 16.94 122
Agree strongly 21.72 16.37 74
F -ratio 1.31

Shift
Day shift 18.73 15.10 88
Evening shift 27.17 17.44 66
Night shift 27.63 18.09 89
F -ratio 7.47 ***

 

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

104

Table 18. One-way AN OVA by Officer Characteristics - Legalist

 

Variable Mean SD. n

Sex
Male 5.61 6.33 202
Female 5.18 6.13 41
F-ratio 0.16

Race
White 5.73 6.48 192
Minority 4.82 5.52 51
F-ratio 0.84

Age
30 and younger 6.93 6.98 82
31-40 5.67 6.18 114
41 and older 2.79 4.15 47
F -ratio 6.82 **

Unit
911 Ofﬁcer 5.73 6.28 167
CPO 5.10 6.33 76
F-ratio 0.53

Education
Less than HS 14.61 0.75 2
HS / GED 4.24 5.84 38
Some college 4.14 5.19 54
Associates degree 4.39 6.04 16
2 + years college 7.12 6.75 42
Bachelors degree 6.43 6.88 80
Some graduate school 5.54 4.31 11
F -ratio 2-36 *

105

Table 18 (cont’d).

Variable Mean SD. n

 

Law enforcement orientation

Strongly disagree 5.02 6.06 14
Disagree somewhat 3.28 6.01 33
Agree somewhat 6.18 6.60 122
Agree strongly 5.58 5.79 A 74
F-ratio 1.90
Shift

Day shift 3.59 4.96 88
Evening shift 5.68 5.57 66
Night shift 7.36 7.37 89
F -ratio 8.47 ***

 

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

106

Table 19. One-way AN OVA by Ofﬁcer Characteristics - Service

 

Variable Mean SD. n

Sex
Male 0.42 1.15 202
Female 1.03 2.59 41
F-ratio 5.73 *

Race
White 0.58 1.60 192
Minority 0.31 1 .00 5 1
F-ratio 1.35

Unit
911 Ofﬁcer 0.47 1.38 167
CPO 0.65 1.73 76
F—ratio 0.80

Age
30 and younger 0.47 1.11 82
31-40 0.47 1.50 114
41 and older 0.76 2.03 47
F-ratio 0.69

Education
Less than HS 0.87 1.23 2
HS / GED 0.21 0.46 38
Some college 0.71 1.93 54
Associates degree 0.13 0.39 16
2 + years college 0.74 1.90 42
Bachelors degree 0.57 1.49 80
Some graduate school 0.04 0.12 11
F-ratio 0.97

107

Table 19 (cont’g.

 

Variable Mean SD. n

Law enforcement orientation
Strongly disagree 1.09 2.93 14
Disagree somewhat 0.76 2.11 33
Agree somewhat 0.47 1.21 122
Agree strongly 0.40 1.20 74
F -ratio 1.18

Shift
Day shift 0.88 2.10 88
Evening shift 0.54 1.38 66
Night shift 0.16 0.39 89
F-ratia 5.17"

 

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

The One-way AN OVAs for the ofﬁcer characteristics revealed several signiﬁcant
relationships. For order maintenance behaviors, the AN OVAs revealed mean differences
for ofﬁcer sex, unit, and shift. The mean order maintenance time for male ofﬁcers was
signiﬁcantly higher than for females (25 percent versus 19 percent respectively); the mean
for 911 ofﬁcers was signiﬁcantly higher than for CPOs (approximately 28 percent versus

16 percent). The mean order maintenance time for day shift ofﬁcers was signiﬁcantly

lower than evening and night shift officers (18.7 percent versus 27.2 and 27.6 percent

respectively).

For legalist behaviors, officers in the younger age categories engaged in more

legalist behavior. The One-way AN OVA for ofﬁcer education indicated signiﬁcantly

different mean values between groups, though no clear pattern was evident. The

108

AN OVA for shift indicated that ofﬁcers on day shift spent less time engaged in legalistic
events (3.5 percent), compared to evening (5.7 percent) and night shift ofﬁcers (7.4
percent).

The AN OVAs revealed that ofﬁcer sex and shift had signiﬁcant mean differences
for time on service events. Female ofﬁcers had signiﬁcantly higher mean scores than did
male ofﬁcers (1.03 percent for females versus 0.42 percent for males). Ofﬁcers on day
shift had a mean percentage of time on service of 0.88 percent, whereas evening shift
ofﬁcers had 0.54 percent and night shift ofﬁcers spent 0.16 percent of their observed free
time on ofﬁcer initiated service events.

A series of One-way AN OVAs including study site and the three dependent
variables was also conducted (See Table 20).27 As indicated in Table 20, ofﬁcers in
Indianapolis engaged in signiﬁcantly higher percentages of ofﬁcer initiated order
maintenance policing. Ofﬁcers in Indianapolis had a mean score of 27.8 for percent of
free time on ofﬁcer initiated order maintenance compared to 15.6 percent for St.
Petersburg ofﬁcers. Ofﬁcers in St. Petersburg engaged in more service policing than

ofﬁcers in Indianapolis (1.05 percent versus 0.31 percent, respectively).

 

27 Study site was the only categorical variable at Level 2.

109

Table 20. One-way ANOVA - Site and Outcome

 

 

 

Measures
O-M Legalist Service
Mean SD. Mean SD. Mean SD.
Site
Indianapolis 27.80 18.20 1.59 0.92 0.31 1.08
St. Petersburg 15.60 10.80 1.51 0.84 1.05 2.14
F-ratio 27.52 *** 0.38 12.65 ***
*" p < .001

While several signiﬁcant relationships were uncovered with these bivariate

calculations, these relationships may not hold up once other variables are entered into the

analyses. The next sections begins with traditional linear analyses to examine the

multivariate relationships between ofﬁcer and neighborhood characteristics and police

behavior, followed by multi-level analyses containing the full models which differentiate

between Level 1 and Level 2 variables.

Multivariate Analyses

Ordinary Least Squares

First, traditional linear models were produced that included both ofﬁcer level and

neighborhood level variables. Ordinary Least Squares (OLS) regression coefficients were

produced and are presented in Table 21.

110

3. v 3.

~8.V.Q**a£o.vaaammc.vaa

 

 

 

2.3 2.3 :3 Base... m
86 86- 8.0 cod mod m _ .o- :30 258$
Q .o 86 mod _ mo 94 we; .33me .280
mg .1... no: u _ .o m _ .o- m _ .m 2.... 5.3- 352802 am
83623 meefeammmez
Ed .. Nod cod 83 83 o _ .o- 2.5 8a 888.:
vmd a... med- 2.0 I. 93 EN _.. an... «Em 2*me
mad 5.6. m mo 36 mad .. mad «Em meagm
:6 Ed- 36 :6 cm.“ mm. 7 5:858 m...—
eod cod vod mod Need 23. conga—um
omd 2.. :6- Ed 83 Sad _. $6- 0.5
8.3 3.? 3.3 :8? 83 S- 0?...
mad _.. 3.? Ed _ mo- mmmN mm. 7 .3852
mg _. 02 m : .o 3.3 mmwd ”me. 228":
833.8.» aeo§
m _ ._ end and _N._ ~12 1. 3.3. E5380
.m.m m .m.m m .m.m m 0523/
ooEum 5:de ooggﬂmaz hop—O

 

3282 828% who .8 some

111

As Table 21 indicates, for percentage of free time spent on ofﬁcer initiated order
maintenance events, the OLS results indicate that ofﬁcer’s unit, shift variables, and site
were signiﬁcantly related to time on order maintenance. CPOs and officers in St.
Petersburg engaged in less order maintenance behavior. The OLS regression coefﬁcients
indicate that ofﬁcers on evening and night shiﬁs spent more time on order maintenance
policing.

For time on legalist events, the OLS results indicate that older ofﬁcers spent less
of their free time engaged in activities deﬁned as legalistic. The results also indicated
that ofﬁcers on night shifts spent more of their free time on legalist events.

For service behavior, signiﬁcant ofﬁcer level variables included officer sex, race,
unit, shift, the percent of free time, and site. Female ofﬁcers, ofﬁcers with more free
time, and ofﬁcers in St. Petersburg engaged in more service behavior. Minority ofﬁcers,

CPOs, and officers on night shifts engaged in less service behavior.

Hierarchical Linear Modeling

For the multilevel models using HLM, steps suggested by Bryk and Raudenbush
(1992) were followed. Speciﬁcally, intraclass correlation coefﬁcients (ICC) and
reliability coefﬁcients were calculated, followed by multilevel models.

First for the HLM analyses, one-way AN OVAs were calculated for each of the
three dependent variables. By computing these, several important descriptive indicators
are provided. First, the reliability of the three dependent variables at the neighborhood
level can be assessed. HLM calculations for the AN OVAs provide reliability estimates

(1) for each of the three dependent variables. For order maintenance, A = 0.64, indicating

112

that this measure was a reliable indicator of the true value at the neighborhood level. For
legalist and service behaviors, the reliability coefﬁcient was lower, indicating
questionable suitability of HLM in analyzing these behaviors. For legalist behaviors, the
ANOVA produced a l= 0.36; for service behaviors A = 0.17.

Next, intraclass correlation coefﬁcients were calculated (p). ICCs indicate the
amount of variance in the outcome variable that was between neighborhood units (Reisig
and Cancino, 2004). For order maintenance events, p = 16.3 7, meaning that about 17
percent of the variance in ofﬁcer initiated order maintenance policing was between
neighborhoods. Altemately, approximately 83 percent of variance was within
neighborhoods. The ICCs for legalist and service behaviors were lower: for legalist, p =
6.3 and for service p = 10.4.

For order maintenance activities, the reliability and intraclass correlation
coefﬁcients indicate that HLM is appropriate; the analysis follows below. The low
reliability coefﬁcients and ICCs for legalist and service behaviors indicate that HLM may
not be the most appropriate technique for these two dependent variables. Taken together,
these calculations indicate that legalist and service behaviors as measured may not vary
across neighborhoods in a consistent manner. Despite these indications, HLM analyses
will be run and results provided in order to compare ﬁndings across the three dependent
measures. Signiﬁcant ﬁndings however should be interpreted with these considerations
in mind.

Tables 22 - 24 present the results of the HLM using percent of free time on ofﬁcer
initiated order maintenance (Table 22), legalist (Table 23), and service events (Table 24)

as the dependent variables. Both officer level and neighborhood level characteristics are

113

entered into the model. The results are from a “ﬁxed effect” hierarchical model. Using a
ﬁxed effect model constrains the slopes of the ofﬁcer level measures, not allowing them
to vary as a function of neighborhood level characteristics. Using ﬁxed effects ensures a
test of the primary objective of the current research — testing the potential effects of cross-

level interactions (Reisig and Cancino, 2004; Terrill and Reisig, 2003). The constant, B0,,

was allowed to vary across neighborhoods and was modeled as a function of the
neighborhood variables: site, concentrated disadvantage, and percent homeowner. Both
the ofﬁcer and neighborhood level variables were centered around the grand mean.

As mentioned, there were some concerns about the distribution of the dependent
variables, speciﬁcally legalist and service behaviors. The problem identiﬁed with legalist
behaviors was the large number of zero values - indicating that numerous officers were
not observed engaging in such behaviors. As such, for the HLM for legalist behaviors, a
Poisson distribution was used. The Poisson distribution, which is also called the
distribution for rare events compensates for a non-normal distribution. The distribution
was more normal after log transforming the variable so this problem was not an issue for
this dependent variable.

For service behaviors, the distribution described above indicated the rarity of
these events and also the non-normal distribution. As such, a Poisson distribution was
used, as well as modiﬁcations for an over-dispersed distribution. This was done because
even after transforming the service dependent variable, the distribution was still heavily
skewed. Since order maintenance did not have these problems, a “norm ” HLM model

was run for order maintenance. The results are presented below.

114

Table 22. HLM - Order Maintenance

 

 

 

 

Variable _l_)__ S.E.
Intercept 23.07 ** * 1 .06

Oﬁicer variables
Female -5 .72 a 2.79
Minority -2.28 2.68
Age 0.21 0.15
CPO -7.55 * 3.08
Education -0. 13 0.77
L.E. Orientation -0.75 0.88
Evening shiﬁ 8.06 *** 1.82
Night shift 4.87 2.60
Percent free time -0.06 0.08

Neighborhood variables
St. Petersburg -9.58 ** 2.80
Conc. Disadv. 0.49 1.12
Percent own -0. l9 ** 0.04

* p < .05; ** p < .01; *** p < .001

a p = .052

115

Table 23. HLM - Legalist

 

Variable __b__
Intercept 0.41 ***

Oﬂieer variables
Female 0.05
Minority -0.06
Age -0.02 ***
CPO 0.02
Education 0.02
L.E. Orientation 0.07
Evening shift 0.18 *
Night shift 0.30 **

Percent free time 0.01 *

Neighborhood variables
St. Petersburg -0. 10
Cone. Disadv. 0.07
Percent own 0.00

SE.

 

0.04

0.11

0.10

0.00

0.10

0.02

0.04

0.08

0.10

0.00

0.10

0.04

0.00

 

* p< .05; ** p< .01; *** p < .001

116

Table 24. HLM - Service

 

 

Variable __tl_
Intercept -022

Oﬂicer variables
Female 0.74a
Minority 0.26
Age -0.01
CPO -0.70
Education 0.08
L.E. orientation -O.44 **
Evening shift 0.44
Night shift -062
Percent free time -0.02

Neighborhood variables

St. Petersburg 2.77 * * *
Conc. Disadv. 7 0.71 *
Percent own 000

SE.

 

0.23

0.36

0.50

0.02

0.71

0.07

0.11

0.35

0.37

0.02

0.45

0.27

0.02

 

*p<.05;**p<.01;***p <.oo1
a p= .052

117

As seen in Table 22, for time on order maintenance events, the HLM results
indicate several signiﬁcant ﬁndings. Community police ofﬁcers spent less of their free
time engaged in order maintenance and ofﬁcers on the evening shifts spent more. For the
Level 2 variables, the HLM coefficients indicated that ofﬁcers in St. Petersburg spent less
time on order maintenance policing as did ofﬁcers in areas characterized by more
homeowners. For homeownership, as the percent of home owning increased, the amount
of ofﬁcer initiated order maintenance decreased. Female ofﬁcers were also seen engaged
in less order maintenance policing than male ofﬁcers and this ﬁnding was close to
signiﬁcant (p = 0.052).

For Legalist events, Table 23 provides the ﬁndings for the HLM for this
dependent variable (using a Poisson distribution in the HLM model). As can be seen,
age, shift, and free time were signiﬁcantly related to ofﬁcer initiated time on legalist
events. Age was negatively related to ofﬁcer initiated legalist activity while evening
shift, night shift, and percent of free time were positively related to time on legalist
events. Older ofﬁcers engaged in less ofﬁcer initiated legalistic policing; ofﬁcers
observed on evening shifts, night shiﬁs, and officers with more free time engaged in more
legalistic policing.

Table 24 provides the ﬁndings for the HLM for service behaviors. For ofﬁcer
level variables, ofﬁcer attitude was signiﬁcant and ofﬁcer sex was close to signiﬁcant (p
= 0.052). The attitudinal variable — law enforcement orientation — indicated that ofﬁcers
who identiﬁed law enforcement as the top priority engaged in less service policing.
Ofﬁcer sex was close to signﬁcant, indicating (as did the AN OVAs and OLS) that female

officers engaged in more service policing. For the service HLM, site and concentrated

118

disadvantage were signiﬁcant Level 2 predictors. Ofﬁcers St. Petersburg engaged in
more service policing. Ofﬁcers in more disadvantaged areas are also seen to engage in
more service policing.

In the next chapter, these ﬁndings are discussed in detail. Consistent and

inconsistent ﬁndings across methods are addressed and conclusions are drawn.

119

Chapter 5
Discussion and Conclusions

The present research attempted to answer key questions concerning the contextual
determinants that inﬂuence police practices and impact the delivery of police services.
As with much research, some questions were answered, some were not, and new
questions arose.

The current research was designed to examine potential differential police
behavior across neighborhoods and how neighborhood contextual characteristics
inﬂuence the behaviors in which police engage. The importance of this line of research
was addressed and included the potential for variations for the provision of police
services — especially in a time of community policing. As found in the previous chapter,
CPOs and general response ofﬁcers did differ in terms of the types of activities engaged
in by type of ofﬁcer. For example, community police ofﬁcers engaged in less order
maintenance. CPOs and general response ofﬁcers did not vary in the amount of
discretionary time that they had available but did differ from regular patrol ofﬁcers in the
amount of time that they spent inside their assigned beat, surprising ﬁndings which are
discussed in more detail below.

While previous research has, in a limited way, examined the impact of contextual
level inﬂuences on speciﬁc police behaviors, the current research examined a variety of
police behaviors across neighborhoods. Previous inquiries were limited in terms of both
the behaviors examined (limited primarily to legalistic practices such as arrest) and the
contextual characteristics included. The question was posed as to whether, in addition to

the previously studied legalistic activities, other police behaviors, ones that by their nature

120

are more discretionary, would be inﬂuenced by neighborhood characteristics. Limited
signiﬁcant ﬁndings were uncovered for the contextual variables included in these
analyses. In short, most of the behaviors included in the present analyses and examined
using the contextual variables did not vary across neighborhoods. The one exception is
that as homeownership increased, ofﬁcers’ order maintenance behaviors decreased. This
and other ﬁndings, as well as the lack of ﬁndings for other behaviors, are discussed
below.

These analyses were approached from two theoretical perspectives. Conﬂict .
theory suggests that police act in a more legalistic manner in areas characterized by higher
socioeconomic distress. The benign neglect hypothesis suggests that police provide fewer
services in these same areas. The unsupported hypotheses concerning the contextual
variables lend support to neither of these theoretical approaches. The ﬁndings from the
preliminary analyses concerning how much time officers spent inside study beats provide
support for a workload hypothesis. In areas characterized as more stable, ofﬁcers spent
less of their free time inside the beat. One interpretation is that ofﬁcers in these beats
leave the area in search of activities and work that need their attention.

In this ﬁnal chapter, the ﬁndings presented in Chapter Four are interpreted and
discussed, the current research is placed in the larger context of current policing literature,
limitations of the data are addressed, and areas for future research are presented. First,
ﬁndings from the data are presented, interpreted and discussed. Several interesting
ﬁndings resulted from examination of the current data. These will be discussed in the

order that they were uncovered.

121

Interpretation and Discussion of the Findings

When considering the totality of the project from which the data used was drawn,
a tremendous amount of information was available. For the original study, over thirty-
ﬁve thousand individual events from 729 observation sessions were observed, coded, and
included in the data sets. The current study used this as a beginning point. As indicated
previously, though the original research design assigned project observers for speciﬁc
beats and shifts, occasionally there was no department assigned officer for that beat and
shift. In such events, observers rode with ofﬁcers who would most likely respond to calls
coming ﬁ'om inside that study beat. These observation sessions were excluded from the
analyses. As a result, a total of 529 rides (i.e., observation sessions) were used; 281 in
Indianapolis and 248 in St. Petersburg. A total of 224 different ofﬁcers were included in
the current analyses.

The data was reduced further (and signiﬁcantly so) by examining only the
identiﬁed three types of behavior — order maintenance, legalistic, and service. These
three measures do not capture the entire range of what police do during the normal course
of business. Excluded was unobserved time, personal time, time spent with other ofﬁcers
discussing non-police business, time to check in and out of service (e.g., at the beginning
and end of a shift), certain administrative activities, and others. This is an important
aspect of the examination conducted. The behaviors measured did not total 100 percent
of ofﬁcer activity, where an increase in one type would have necessarily resulted in a
decrease in another type. The lack of inter-correlations among these three outcome
measures (See Table 14) indicates that the measures were distinct from one another and

not capturing an underlying construct such as ofﬁcer productivity.

122

In addition, only ofﬁcer initiated events were included as the target behaviors for
examination. Finally, only those events occurring inside the study beat were used to
calculate the outcome measures. After considering all of this, the original data was
reduced signiﬁcantly, resulting in 4,532 classiﬁable events for the targeted measures of
police behavior out of the initial 35,433 events.

The total observed time and also the time for all of these target events were
summed for each unique ofﬁcer-in-beat combination. In other words, if Ofﬁcer 1 was
observed three times in beat 321, the total time observed on these events was calculated.
IfOﬂicer 1 was observed three times in beat 321 and once in beat 344, two summed
times were calculated — one for beat 321 and one for beat 344. This occurrence was
relatively rare however — happening for only 19 observed ofﬁcers (8.5 percent).

Times for observation sessions varied, as did the number of times ofﬁcers were
observed. There were more repeated observation sessions of officers in St. Petersburg
than in Indianapolis (mean = 1.6 versus 3.5 respectively). The average time for observed
ofﬁcers was 1,011 minutes - close to 17 hours. The shortest observation session was 15
minutes and the longest observed ofﬁcer totaled over 78 hours of observation.

Several computations were involved in calculating the three dependent variables.
These are interesting in their own right and are worthy of discussion. The mean observed
time that ofﬁcers spent in their assigned beat was 46.1 (SD. 26.7). In other words,
ofﬁcers spent over half of the time they were observed outside their assigned beat (close
to 54 percent). While it is not unexpected for a certain portion of a shift to be spent
outside of an assigned beat, the ﬁnding that more than half is spent outside the beat is

somewhat surprising. In the normal course of work, ofﬁcers leave their beat for multiple

123

reasons — to back up another ofﬁcer in a nearby beat, to take a suspect to the station for
processing, meetings (including roll call / read-off at the beginning of a shift), etc. Fifty-
four percent of time outside the beat does seem high however, and a closer examination
was conducted.

Further examination of this ﬁnding involved a review of the events in which
ofﬁcers engaged outside of their beats. The most frequently engaged in activities outside
of the assigned beat were moving en route to a location other than a dispatched call. The
possibilities of this are numerous: driving to or ﬁ'om the station, driving to or from a
restaurant, driving back to the assigned beat after being dispatched to a call outside the
beat, just to name a few.

The second most common activity outside an assigned beat was general motorized
patrol. One cautionary point is in need of mentioning regarding this: activities that were
not 100 percent inside a study beat were not included in the analyses, therefore, an ofﬁcer
could spend ten minutes conducting general motorized patrol, nine minutes within the
beat and one minute outside, and this would not be included. This was due to the
impossibility of classifying the geographic location of the remainder of the patrol and to
distinguish patrol that was purposely targeted to the study beat. Most activities (even
patrol) were either totally inside or totally outside the assigned beat (approximately 80
percent). Despite this, what is interesting is that it appears routine for ofﬁcers to leave
their assigned beats while conducting motorized patrol.

Other common activities outside an ofﬁcer’s beat included meeting with other
police (official business), personal time (lunch breaks, personal errands, etc.),

administrative activities (such as report writing), backing up other police ofﬁcers, and

124

meeting with other police (non-police business). These identiﬁed activities make logical
sense when considering the types of activities that would draw an ofﬁcer outside of his or
her assigned beat.

Also included in the calculation of the three dependent variables was the amount
of observed time that was ofﬁcer initiated (as opposed to dispatched events, citizen
initiated events, or other ofﬁcer, including supervisor, initiated events). The mean
amount of observed time that was ofﬁcer initiated was 70.5 percent (SD. = 13) and
ranged from around 4 percent to almost 100 percent. In other words, the majority of an
ofﬁcer’s shift involved undirected time.

Time spent in the beat (as mentioned above, 46 percent of observed time) was
dissected further into officer initiated versus other person initiated events. Of the time
spent within the boundaries of the beat, officer initiated time was reduced from 70.5
percent to 47 percent. While within the boundaries of their beat, the majority of ofﬁcer
time was directed either by dispatch, other ofﬁcers or supervisors, and citizens. Most of
the time spent within the geographic boundaries of an assigned beat was other than the
observed ofﬁcer initiated.

An examination of time outside the beat (54 percent of observed time) indicated
that 70.9 percent of this time was ofﬁcer initiated. This is an interesting ﬁnding. When
ofﬁcers are free, they tend to leave their beat. It appears that if not directed by others

(police or citizens) ofﬁcers would spend even more time outside their beat.28

 

28 This ﬁnding is supported further by indications that day shift ofﬁcers have the most free time and also
spend the least time inside their assigned beat.

125

The “preliminary analyses” section of the previous chapter which examined free
time and time in beat calculations (variables that were subsequently used to create the
dependent variables) provided some interesting ﬁndings. Analyses were performed to
determine if any of the other variables included in the analyses were predictive of the
amount of free time that ofﬁcers had or the amount of time that ofﬁcers spent in their
primary beat.

The examination of free time and potential inﬂuences of this revealed that only
ofﬁcers’ shift was signiﬁcant. Ofﬁcers working day shifts had the most free time and
ofﬁcers working the night shift had the least. Day shift ofﬁcers had a mean percent of
ﬁee time of approximately 75 percent, evening shift ofﬁcers was 67 percent, and night
shift officers was 69 percent. This ﬁnding is expected as different shifts have previously
been shown to involve varying workloads and demands on ofﬁcers’ time. None of the
other variables was signiﬁcant in predicting the amount of free time that ofﬁcers had.29
Observed ofﬁcers, across study sites, across beats, and across units had similar amounts
of free time. Other than ofﬁcers on different shifts, ofﬁcers are being engaged in different
beats and cities at similar rates. The results discussed below should be viewed with this
in mind.

Signiﬁcant ﬁndings by shift in terms of the rates of engagement in the targeted
behaviors further highlight the differing workloads and activities that ofﬁcers choose to

engage and the signiﬁcant differences in ofﬁcer behavior across shifts. The other ofﬁcer

 

29 Note that only ofﬁcer unit, shift, study site, concentrated disadvantage, and percent homeowner were
included in these analyses.

126

characteristics included to predict free time indicate that ofﬁcers are left with similar
amounts of ﬂee time to do with what they will.

One surprising ﬁnding from the examination of free time was that community
police ofﬁcers and general response ofﬁcers did not differ in the amount of free time that
they had. This is surprising since one would expect general response ofﬁcers to have
more directed time due to responding to dispatch calls. Perhaps community police
officers are engaged by citizens in a different way from general response ofﬁcers —-
speciﬁcally, in person as opposed to telephonically. It could also be that community
police ofﬁcers are being directed by superiors as to the activities in which to become
engaged. The lack of differences found between the two types of ofﬁcers could counter
arguments that community police ofﬁcers who are freed (in most situations) ﬁ'om
responding to dispatched calls for service, left to their own devices, could have
misconduct problems. Even though CPOs are not primarily responsible for handling calls
for service, they are still being engaged by citizens or directed by superiors at the same
rate as general response ofﬁcers.

Unlike the analyses for ofﬁcer’s free time, there were several signiﬁcant ﬁndings
revealed from the examination of the amount of time that ofﬁcers spent in their primary
beat. Age was consistently and signiﬁcantly negatively related to the amount of time that
ofﬁcers spent within their beat. As ofﬁcer’s age increased, they spent less time within the
beat. One explanation is that younger officers are also newer ofﬁcers and these ofﬁcers
may be less likely to deviate from proscribed beat boundaries. Younger ofﬁcers may be
hesitant to leave their beats in the event of an important call for service being ordered.

Older ofﬁcers having been there a while may feel freer to wander and have more

127

conﬁdence in their ability to respond to calls, or recognize that the frequency of these
high priority calls is relatively low.

Another signiﬁcant ﬁnding for amount of time spent inside the beat was that
community police ofﬁcers spent less time in their beat than did general 911 ofﬁcers. This
ﬁnding is surprising and counterintuitive. One would expect that community police
ofﬁcers, being relatively permanently assigned to their respective beats and urged to take
ownership of these beats would outspend general patrol ofﬁcers in the amount of time
spent inside these beats. It would also be expected that in the normal course of business,
ofﬁcers are routinely dispatched outside of their beat; this occurrence would seem more
rare for community police ofﬁcers.

Explaining the ﬁnding that observed CPOs spent less time in their beats than did
911 officers is difficult. One possibility is that CPOs, who have projects, programs, and
reports that traditional ofﬁcers do not, spend more time in the department (headquarters
or beat ofﬁce) working on these. It could also be that observed CPOs are helping other
CPOs from different beats with programs or events and this takes them out of their own
beat. It could be that CPOs are meeting outside of their beat with other agencies or
ofﬁcials for projects. It could also be that community police ofﬁcers have more
discretion and choose to spend their time outside of the assigned beat for personal
matters. Any of these are possibilities and further research would need to be conducted to
determine the answer.

Ofﬁcers on different shifts were also signiﬁcantly different from one another in
terms of time spent in the assigned beat. Officers observed on day shifts spent the least

amount of time in the assigned beat (approximately 40 percent), followed by night shift

128

officers (48 percent), and evening shift ofﬁcers (approximately 50 percent). The ﬁndings
for ﬂee time and time in beat exhibit an interesting pattern: officers with more ﬂee time
spend less time inside the assigned beat. Day shift ofﬁcers had the most ﬂee time and
spent the least time in the assigned beat; ofﬁcers on evening shift had the least amount of
ﬂee time and spent the most time inside the assigned beat. This supports the earlier
statement that when ofﬁcers have ﬂee time they tend to leave their beat.

Also signiﬁcant for the amount of time spent inside the assigned beat was the
neighborhood variable, percent homeownership. The results ﬂom the OLS analyses
indicated that as the percentage of homeowners in an area increased, ofﬁcers spent less of
their time in the beat. As described for the current research, percent homeowner indicates
more stable areas. This result makes intuitive sense. If these areas are indeed more
stable, then the reliance on the police should be less and ofﬁcer activity in these areas
would be lower. As such, ofﬁcers would be more likely to leave their beat to back up
other ofﬁcers or simply to look for things to do. This ﬁnding lends support to the premise
of differential policing; that more stable areas do not receive the same services, even
general patrol, that more troubled places receive. This is common sense and empirically
backed up. This ﬁnding is consistent with the workload hypothesis that Klinger (1997)
offered. Klinger (1997) states that differential neighborhood policing is due to the
variations in workloads that are present in different neighborhoods.

From the present ﬁndings, if neighborhoods are more stable and the workload is
lower, ofﬁcers may leave their primary beat to assist others or locate activities in need of
their attention. As indicated by these results, ofﬁcers in these areas spent less of their ﬂee

time in these areas — these areas which have been characterized in the present case as

129

more stable areas. So even the most commonly observed activity of general patrol, which
was classiﬁed as order maintenance, was less in these areas and helps explain the
signiﬁcant ﬁnding of percent homeowner and less time spent inside the beat.

This ﬁnding supports partially the proposition that ofﬁcers treat different areas
differently. It makes intuitive sense, and now shown empirically, that in areas that can be
characterized as more stable, ofﬁcers spend less of their unobligated time. If these areas
are not in need of assistance from the police in maintaining order, then why should the
police linger about. It is likely the case that general patrol of these areas is rare and that
police presence is more common on an as-needed basis. After the examination of free
time and time inside the assigned beat, attention was turned to the three outcome
variables, the primary focus of the current research. Times values were smnmed for each
officer: total observed time, time inside the assigned study beat, ofﬁcer initiated time, and
ofﬁcer initiated time inside the beat. The values outlined and discussed above were used
to calculate, as the denominator, the percentage of time spent on the three speciﬁc types
of police behaviors — order maintenance, legalistic, and service events. These behaviors
are described next — the ﬂequency of their occurrence, time spent on each, and the
distribution of these dependent variables across officers.

Order Maintenance. Of the three types of police behavior examined, order
maintenance was by far the most common. The total number of order maintenance events
was 2,928. The large number of order maintenance events was due to the inclusion of
general patrol in this category. General patrol was the most common type of order
maintenance activity engaged in by observed officers. It should be noted however, that

for inclusion as one of the three types of behaviors used for this dependent measure, 100

130

percent of the activity had to occur within the boundaries of the study beat. These
restricted general patrol activities were observed 2,431 times.

Other order maintenance events that were more commonly observed included
checking out suspicious circumstances (n = 93), encounters with suspicious persons (11 =
52), residential and business security checks (n = 35), loiterers (n = 33), persons identiﬁed
or suspected of being drunk (n = 26), public nuisances (n = 19), and domestic arguments
(n = 15). The ﬁrll list of events identiﬁed as order maintenance and the distribution of
these events can be found in Appendix B. As mentioned previously, some activities were
classiﬁed as one of the three types of measured activities based on the focus of that
activity — not the activity itself. For example, backing up another ofﬁcer did not receive a
classiﬁcation as one of the three outcome measures unless the problem on which that
activity was focused was able to be identiﬁed and categorized as either order
maintenance, legalist, or service oriented policing.

Legalistic. Legalistic behaviors initiated and engaged in by ofﬁcers was the next
most commonly seen. A total of 1,332 legalist events were observed. The most common
legalist activities involved attempts by the observed officer to locate suspects or witnesses
(n = 133).30 Trafﬁc focused activities were also common — mobile and stationary trafﬁc
activities occurred 108 times. Law enforcement focused surveillance of particular
persons or particular locations was also commonly observed (n = 75).

The most common officer initiated legalist encounters involved automobile

violations, speciﬁcally, moving violations (n = 154) and improper equipment or plates (11

 

30 Here, activity is speciﬁcally used to denote activities as originally coded — differentiated ﬂom events by
the presence of a citizen interaction. Encounters are discussed subsequently.

131

= 111). The most commonly observed speciﬁc criminal violations (or suspected) that
ofﬁcers engaged in involved drugs. Eighty-two events (57 encounters and 25 activities)
involved ofﬁcer initiated events involving suspicion of illegal drug activity. Table B.2
located in Appendix B provides all observed legalist events. The list of legalist events is
the most lengthy due to the speciﬁcity of problems or crimes on which police were
observed to engage and the comprehensive nature of the list.

Service. Of the three types of police behaviors under examination, service
behaviors was by far the rarest. Only 272 observed events were classiﬁed as ofﬁcer
initiated within beat service events. Service was deﬁned as behavior for which the
authority and enforcement powers of the police were not necessary in order to complete.
In other words, these activities were not necessarily considered police responsibilities —
anyone could conduct these. The ﬁndings indicated that these observed service behaviors
included activities commonly associated with community policing programs.

The most common types of service activities involved meetings that the observed
police ofﬁcer attended. These were not necessarily formal meetings but could also
include brief face-to-face interactions with certain people. Classiﬁed as service were
meetings with other government ofﬁcials (n = 26), meetings with the public — such as
neighborhood group meetings (n = 18), and meetings with non-police service providers.
Common service encounters involved what was termed “ofﬁcer friendly” or community
relations encounters (n = 45). Another commonly observed service event involved
ofﬁcers giving citizens information (11 = 40). Assisting motorists and providing
directions occurred 31 times. A complete list of these service events can be found in

Appendix B, Table B3.

132

The speciﬁc activities and encounters described above and detailed in tables in
Appendix B detail the types of behaviors that were included in the classiﬁcation of the
three dependent variables for the current study. While common perceptions exist as to
how ofﬁcers spend their time and what types of activities they engage in, the detail
provided above is to move beyond vague notions of what is considered order
maintenance, legalist, or service behaviors. Most of what has been described is not
surprising - police spend a large part of their time engaged in general patrol and a large
part conducting trafﬁc enforcement. What is surprising is the relative inﬂequency of
service events. While other police researchers have identiﬁed larger percentages of time
on service (e.g., Walker and Katz, 2002); the current research found that this type of
behavior is quite rare.

Simple counts of events however do not necessarily capture the degree to which
police engage in these behaviors of interest. For this reason, time spent on such
behaviors, as a percentage of all observed time was used. Again, the mean observed time
was 1,010.6 minutes (SD. = 846.6) and ranged ﬂom 15 minutes to 4,698 minutes.
Approximately 70 percent of this time was ﬂee time — not dispatched or otherwise
engaged; 46 percent of this time was spent in the assigned beat. Using these calculations,
the percent of ﬂee time spent on each of the three dependent variables inside the assigned
beat was produced for the 243 unique ofﬁcer-in-beat combinations.

The mean percent of ﬂee time spent on order maintenance inside a beat was 24.3
(SD. = 17.3) and ranged ﬂom zero to 80 percent. The mean percent of ﬂee time spent on
legalist events inside the beat was 5.5 percent (SD. = 6.3) and ranged ﬂom zero to 34.7

percent. The mean percent of ﬂee time spent on service events was considerably less, 0.5

133

percent (SD. = 1.5) and ranged ﬂom zero to 10.1 percent. In addition to the simple
counts described above, the percentages of time ofﬁcers spend reiterate the relative
common practice of spending time on ofﬁcer initiated order maintenance events and the
small amount of time spent on service behaviors. Not only are the number of service
events offered small, the time spent on them is also limited.

Further examination of the distribution of the three outcome measures also reveals
some interesting ﬁndings. While the distribution of order maintenance policing is
somewhat normal (See Chapter 4, Figure 1), the distribution for the other two outcome
measures are negatively skewed, and in the case of service behaviors, seriously so. An
exarrrination of the histograms provided in Chapter 4 (Figures 2 and 3), indicate that a
large number of ofﬁcers engaged in no ofﬁcer initiated, inside the beat legalist oriented
policing and an even larger number did not engage in any service. For ofﬁcer initiated
legalist behaviors inside the study beat, 54 ofﬁcers had values of zero; for ofﬁcer initiated
service behaviors inside the study beat, 166 ofﬁcers had values of zero (several more
ofﬁcers for each outcome measures had very little, less than two percent of their ﬂee time
spent on these activities). These ﬁndings indicate that a signiﬁcant number of ofﬁcers
choose to spend no time trying to invoke the law or offer services inside their beat of
primary responsibility.

For the three dependent measures, an examination was done to determine if these
were correlated with each other (See Table 14). It would be useful to know if ofﬁcers
who had high levels of one type of officer initiated activity also had high levels of other
behaviors. In other words, there is the potential for the designed measures to be tapping

into an underlying indicator of productivity. This however did not appear to be the case.

134

Correlations among the three dependent variables were low (none above 0.07) and none
was signiﬁcant. This ﬁnding indicates that the outcome measures were not simply
identifying officers that were more productive, or engaged in more ofﬁcer initiated
activities - that there is some pattern of activities in which police engage and that ofﬁcers
choose to spend their time differently ﬂom one another. The next set of analyses
discussed attempted to identify ofﬁcer and neighborhood characteristics that helped

predict these patterns.

Bivariate and Multivariate Analyses
Several signiﬁcant ﬁndings were uncovered by examining both the bivariate and
multivariate relationships between the ofﬁcer, neighborhood, and behavior variables.

Table 25 contains the signiﬁcant relationships that were found.

135

Table 25. Signiﬁcant Bivariate and Multivariate Findigs

 

 

Bivariate
Variable Correlations ANOVA OLS HLM
Oﬂicer variables
Female - (OM) - (OM) - (OM)*
+ (Service) + (Service) + (Service) + (Service)*
Minority - (Service)
Age - (Legal) - (Legal) - (Legal) - (Legal)
CPO - (0M) - (0M) - (0M) - (0M)
- (Service)
Education +/- (Legal)
Attitude - (Service)
Day shift - (OM) - (OM) (excluded) (excluded)
- (Legal) - (Iﬁgal)
+ (Service) + (Service)
Evening shift " + (OM) + (OM)
+ (Legal)
Night shift + (OM) + (OM) + (OM)
+ (Lagal) + (Legal) + (Lesal) + (Legal)
- (Service) - (Service) - (Service)
% ﬂee time - (OM) N/A + (Legal)
+ (Service)
Neighborhood variables
St. Petersburg - (OM) - (OM) - (OM) - (OM)
+ (Service) + (Service) + (Service) + (Service)
Concent Disadv. + (Legal) N/A + (Service)
% own N/A 40M)
‘ p = .06

" Evening shift had median value for all 3 measures

Both bivariate and multivariate analyses indicate that female ofﬁcers spent less
time engaged in order maintenance policing and more time engaged in service policing.
These ﬁndings partially support Hypothesis I offered in Chapter Three (that female
ofﬁcers engage in less order maintenance, less legalistic, and more service).

Previous research has indicated that male and female officers perform their
functions differently. Previous studies have indicated that female officers made fewer
arrests (Bloch and Anderson, 1974; Sherman, 1975). The ﬁndings here lend support to
these previous ﬁndings regarding the types of activities in which female ofﬁcers choose
to engage. While the AN OVA for legalistic behaviors indicated that female ofﬁcers had
lower mean values for legalistic behaviors, the differences failed to reach a level of
signiﬁcance.

Analyses also indicated that as the age of ofﬁcers increased, the percentage of ﬂee
time spent on officer initiated legalistic behaviors decreased; this was true for both
bivariate and multivariate analyses. Stated conversely, younger officers engaged in more
legalistic behavior. This ﬁnding is consistent with previous research that has indicated
that younger ofﬁcers (generally speaking, ofﬁcers with less experience) engage in more
law enforcement oriented behaviors such as arrests and ﬁeld interrogations. Forst et al.
(1977) and Worden (1989) found similar results concenring legalistic, or law
enforcement, types of behaviors and age — that younger/newer ofﬁcers engage in more
activities identiﬁed as law enforcement oriented. This ﬁnding supports Hypothesis 2 —
that older ofﬁcers engage in less legalistic behavior.

As illustrated in Table 25, there were also signiﬁcant ﬁndings concerning

differences between general response ofﬁcers and community police ofﬁcers. It was

137

found that community police ofﬁcers spent less time on order maintenance policing.
This ﬁnding is opposite of what was expected and expressed in Hypothesis 3 which
proposed that community police ofﬁcers would engage in more order maintenance
policing, less legalistic, and more service. The regression analyses do not support this
hypothesis — conﬁrming that community police ofﬁcers actually engaged in not only less
order maintenance but also less service.

These ﬁndings are somewhat surprising when considering the role of community
police ofﬁcers and the broadened mandate, that by deﬁnition, these ofﬁcers have. It is
also surprising since it would be expected that community police ofﬁcers would have
more “ownership” of their beat and exert more effort in providing services. When one
considers the importance of building community relationships and partnerships, the fact
that community police officers spent less of their ﬂee time on service activities is
interesting to discover. It should be kept in mind however that as deﬁned, service events
were rare and only accounted for a minuscule amount of observed officer time. It could
be that the way that service was deﬁned here was limiting and therefore did not capture
the full range of potential services that police, especially community police ofﬁcers
provided.

As outlined in Chapter One, community policing as a reform differs ﬂom
“traditional” policing in many respects. The expanded role of police — in terms of
recognizing and embracing the varied nature of the role, not just enforcing the law — is
not supported by the ﬁndings. It was thought that the broader mandate for community
police ofﬁcers that includes quality of life issues would be evidenced in higher levels of

order maintenance activities, but this was not the case.

138

The ﬁnding of differences between general patrol and community police officers
and the types of behaviors in which they engage could be capturing underlying
neighborhood needs not identiﬁed by the current research. The neighborhood speciﬁc
approaches called for by many community police advocates might be operating here and
evidenced by the different patterns of ofﬁcer initiated behaviors.

Differences were also uncovered for ofﬁcer behavior by shift. In addition to the
amounts of ﬂee time and the time spent inside the assigned beat (discussed above)
ofﬁcers on various shifts (i.e., day, evening, and night) engaged in different levels of
order maintenance, legalistic, and service policing. Day shift ofﬁcers were seen to engage
in less order maintenance policing. The larger amounts of ﬂee time discovered for day
shift officers helps to explain this ﬁnding and the indication that ofﬁcers with more ﬂee
time leave their beats. General patrol outside of the assigned beat would not be included
in the current measure of order maintenance, and may be responsible for the lack of
signiﬁcant ﬁndings for day shift ofﬁcers’ provision of order maintenance to their assigned
beat. There were indications at the bivariate and ﬂom the AN OVAs that day shift
ofﬁcers engaged in more service policing. Further research is needed to detect how
speciﬁcally day shift ofﬁcers are spending their time.

Day shift ofﬁcers were also seen to engage in less legalistic behavior. This may
be due to the opportunity for such enforcement. Day shifts are known to be slower for
ofﬁcers in terms of workloads and the fact that most people work or are in school during

the day shift may lirrrit the opportunity to engage in legalistic activities.

139

The bivariate analyses and AN OVAs for shift indicated that day shift ofﬁcers
engaged in more service policing however this ﬁnding failed to reach signiﬁcance for the
more advanced regression models.

Evening and night shift ofﬁcers were identiﬁed as engaging in more order
maintenance and more legalistic policing. These ﬁndings may be due to differential
opportunities for activities and enforcement that present themselves during different times
of the day. Night shift officers were also seen to engage in less service policing.
Considering the time of the evening and after midnight hours that these ofﬁcers work,
this is not surprising — people may not be around and situations may not arise during
these hours for service opportunities.

There were also differences between study sites. Ofﬁcers in St. Petersburg
engaged in less order maintenance and more service policing. Differences in
departmental styles and organizational mandates may help explain these ﬁndings.
Describing the departments used for the original study, Parks et a1. (1999) indicate that
the Indianapolis Police Department at the time, was directed by a chief who urged officers
to engage in aggressive order maintenance policing and traditional law enforcement
activities such as suspicious stops. The results here consistently showed that ofﬁcers in
Indianapolis spent more of their ﬂee time engaged in order maintenance policing. With
such support ﬂom management for aggressive order maintenance and traditional law
enforcement practices, it is not surprising that time on service behaviors in Indianapolis
was consistently and signiﬁcantly lower than in St. Petersburg.

These ﬁndings lend some support to Wilson (1968) and the differences between

departments and organizational inﬂuences of police behavior. While the focus of the

140

current research did not attempt to capture organizational inﬂuences, the differences
between departmental priorities and mandates point to areas that warrant further
examination. Clear differences were present in terms of the priorities and styles of the
departments included in the original study and these differences manifested in the
ﬁndings of signiﬁcant differences across sites. Further analyses, measuring speciﬁc
organizational characteristics could more solidly identify potential organizational effects
and lend support to the inﬂuence of departmental styles such as those described by

Wilson (1968) — i.e., watchman, legalistic, and service styles.

HLM

For the Hierarchical Linear Models, similar results to those found in the preceding
discussion were found though some slight differences also appeared. For order
maintenance, the multilevel model indicated that female ofﬁcers engaged in less order
maintenance policing, community police officers engaged in less, and ofﬁcers in St.
Petersburg engaged in less. The HLM indicated that ofﬁcers observed during evening
shifts engaged in more order maintenance. Two differences between the HLM and
previous analyses were found: (1) percent of ﬂee time which was signiﬁcant in the earlier
analyses was not signiﬁcant for the HLM, and (2) percent homeownership was signiﬁcant
for the HLM and had not been for the previous analyses.

The Level 2 variable, percent homeownership being signiﬁcant indicated that in
areas that had more homeowners, ofﬁcers engaged in less order maintenance policing.
This supports in part, Hypothesis 7 and was one of only two signiﬁcant ﬁndings

concerning contextual characteristics and variations in police behavior.

141

There are two possible interpretations for this ﬁnding. One interpretation is that
these areas did not have problems that were categorized as order maintenance and
therefore the observed ofﬁcers did not need to engage in these activities. In other words,
perhaps these areas were relatively stable and not in need of the control mechanisms
offered by the police.

The alternate explanation becomes apparent by looking at the “preliminary
analyses” section of the previous chapter. One area that was examined was the amount of
time that ofﬁcers spent in their assigned beat. Table 13 provides the OLS regression
results examining the independent variables and the outcome of percent of time spent
within a beat. As indicated by these results, ofﬁcers in areas with more homeowners
spent less of their ﬂee time in these areas - these areas which have been characterized in
the present case as more stable areas. So even the most common observed activity of
general patrol which was classiﬁed as order maintenance was less in these areas and this
helps explain the signiﬁcant ﬁnding of percent homeowner and less order maintenance
policing.

The HLM for legalist behavior also provided results that were similar to the
bivariate and multivariate ﬁndings. Speciﬁcally, only age and ofﬁcer shift variables were
signiﬁcant. The HLM analyses supported the ﬁndings previously discussed that younger
ofﬁcers spent more of their ﬂee time on ofﬁcer initiated events categorized as legalistic in
nature. For ofﬁcers’ shifts, the HLM conﬁrmed earlier ﬁndings that ofﬁcers observed
during evening and night shifts both engaged in more legalistic policing. None of the
Level 2 variables was signiﬁcant — indicating that the variables used at the second level

did not capture variations across neighborhoods that would account for different legalistic

142

behaviors. It is worth noting that only a small percentage of the variance was able to be
attributed across neighborhoods, speciﬁcally, 6.3 percent. Tire measures used in the
current analyses failed to capture or identify these differences.

One explanation for the ﬁndings of no differences between neighborhoods for
legalistic behaviors and the relatively small amount of variation attributable to between
neighborhoods involves the nature of police work in general. Previous research has
indicated that ofﬁcers behave somewhat similarly in terms of triggers for initiating events
involving actual or suspected criminal activities. More speciﬁcally, when engaging in
crime ﬁghting activities, police might need stronger evidence to motivate their
involvement. While order maintenance policing is relatively easily engaged in (e.g.,
general patrol or asking a crowd‘to move on), a higher threshold exists for police to
implant themselves in a situation which could result in someone being ticketed or
arrested. It is possible that the areas observed for the current project provided similar
opportunities for such engagement in these crime ﬁghting opportunities.

The ﬁnal HLM to be discussed involves service policing. The ﬁndings ﬂom the
HLM for service policing provided some interesting and unique ﬁndings -— ﬁndings
different ﬂom the less sophisticated bivariate and OLS analyses. It is worth repeating that
there were problems with this dependent variable: a non-normal distribution, a high
number of ofﬁcers who engaged in no service policing, and the low reliability coefﬁcient
(0.174). Accommodations for the model were made before it was rrm, namely, the use of
a Poisson distribution and indicating that the distribution was over dispersed. These

considerations make this HLM not directly comparable to the other multilevel models.

143

Four variables in the HLM were found to be signiﬁcant predictors of the amount
of ﬂee time that ofﬁcers spent on ofﬁcer initiated service policing: officer sex, law
enforcement orientation (the attitudinal variable), study site, and concentrated
disadvantage. Female ofﬁcers were seen to engage in more service policing. The
variable capturing the law enforcement orientation of ofﬁcers was also signiﬁcant and in
the expected direction. Ofﬁcers who identiﬁed law enforcement as “by far” the most
important duty of a police ofﬁcer engaged in less service policing. This ﬁnding is
consistent with Hypothesis 4, however the problems with the reliability of the model
prevent full support for this hypothesis. Ofﬁcers in St. Petersburg engaged in more
service policing. The signiﬁcance of concentrated disadvantage would suggest that
ofﬁcers in more disadvantage areas engaged in more service policing, opposite of what
was expected and proposed following both conﬂict theory and the benign neglect
hypothesis.

What is to be made of these conﬂicting ﬁndings? Due to the problems with the
distribution of this variable and the low reliability, HLM is probably not an appropriate
method for analysis for this variable. The inconsistencies provided by the HLM further
indicate the questionable applicability of this method for this variable. The rarity of
service policing also created a problem for meaningful analyses. As such, the bivariate
and OLS results seem more appropriate: that female officers engaged in more, ofﬁcers on
day shift engaged in more, and that ofﬁcers in St. Petersburg engaged in more. The
“more” described above however is minuscule.

In addition to the signiﬁcant ﬁndings discussed above, several variables that were

not signiﬁcant are worth mentioning — namely, officer race, education, and attitudes.

144

There were no signiﬁcant differences uncovered between minority and white ofﬁcers in
terms of observed time spent maintaining order or time engaged in legalistic activities.
Additionally, no differences were found between officers who had different levels of
education and the three outcome variables. While the AN OVA revealed signiﬁcant mean
differences, these differences were not in any discemable pattern.

No consistent ﬁndings were uncovered between officers who held different
attitudes concerning law enforcement as the top priority of ofﬁcers. The lack of ﬁnding
for the attitudinal variable is not altogether surprising since previous research has
struggled to link attitudes and behavior. Hypothesis 4 which posited a relationship
between law enforcement attitudes and priorities and behaviors was therefore not
supported.

Concerning the neighborhood variables, the hypotheses concerning the expected
effects of the contextual level inﬂuences on police behavior are generally unsupported.
Concentrated disadvantage was signiﬁcant for the bivariate correlations with legal
activity - positively correlated — and for the HLM, positively related to service. The
correlation with legal activity is in the proposed direction, but the more sophisticated
analyses fail to ﬁnd an effect making this bivariate relationship questionable. The
problems with the service model HLM also make the impact of concentrated
disadvantage questionable.

These ﬁndings fail to support the contention proposed concerning police officer
compensation for the lack of other social controls. In addition, Hypothesis 6, which

expressed that areas characterized by higher levels of concentrated disadvantage would be

145

related to higher levels of legal and order maintenance policing (and less service) was not
supported by the analyses.

The positive indicator - percent homeownership — was found to be signiﬁcant in
the multilevel model and in the direction proposed by Hypothesis 7. This lends support
to the idea that areas are indeed policed differently. In addition, differences were
uncovered in terms of how much time ofﬁcers spend in communities that are
characterized as more stable.

The analyses described above help shed light on how police ofﬁcers are choosing
to spend their time and advance our understanding of police behavior. While ultimately,
conﬂict theory and benign neglect fail to explain most of the differences in police
behavior, the signiﬁcant ﬁndings uncovered do help create a fuller understanding of
police practices. While consistent signiﬁcant patterns of variations of neighborhood
policing are absent, there are indications of neighborhood effect that are signiﬁcant and
help explain police conduct.

Our understanding of police behavior has focused primarily on ofﬁcer and
situational level explanations leaving contextual level characteristics under-examined.
Evidence supporting multiple levels of inﬂuence are present ﬂom the current research —
ofﬁcer level, organizational level, and contextual level. Effects of ofﬁcer level
characteristics were identiﬁed ﬂom the current research indicating that male and female
ofﬁcers behaved differently and that ofﬁcers of different ages behaved differently. There
is also evidence that organizational inﬂuences are acting to shape ofﬁcer behavior — site
differences presumably the result of different organizational mandates ﬂom departments

-- though further analysis is needed to conﬁrm this.

146

The ﬁndings ﬂom the current study also support the idea that police behave
differently in different neighborhoods - the original premise of the current study.
Although ofﬁcers in different neighborhoods had similar amounts of ﬂee time, the
amount of time that they devoted to these areas and what they did in these areas was
different. In areas characterized as more stable, ofﬁcers spent less time inside the beat
and spent less time focusing on order maintenance problems.

As shown, where and how ofﬁcers are spending their time is affected by
characteristics of place. As described here and by previous police research, to understand
fully the inﬂuences of police behavior, contextual level characteristics need to be
included. The current research is an attempt to ﬁll the void in terms of macro level
examinations of police behavior, especially approaching these ﬂom a theoretical
perspective. It may be that different outcome measures and alternate community
characteristics can be shown to have a more powerful effect and can ultimately conﬁrm
and advance police theory. These and other areas for future research endeavors are
discussed in more detail below.

Though several signiﬁcant ﬁndings and interpretations have been discussed and
though all of the expected ﬁndings were not revealed (especially at the neighborhood
level), certain methodological and measurement considerations might help explain these.
There are certain examinations that could and were done and some that were not possible.
Attention is next turned to discussing the limitations of the data used for the current

project.

147

Limitations of the Current Research

There are important limitations faced by the current research, namely, the choice
of neighborhoods, the limited number of neighborhoods available for study, the exclusion
of activities due to the limited ability to capture the locations of these activities, and the
limitations posed by the use of secondary data. While these limitations are present and
ones that face many researchers, they do not prohibit meaningful analyses and
interpretations.

The ﬁrst and most important limitation of the current research involves the
neighborhoods that were chosen for study by the original research project. The
neighborhoods chosen for the Project on Policing Neighborhoods were deliberately
chosen to represent areas where more police-citizen encounters would occur. In other
words, these areas were not necessarily representative of all the neighborhoods for which
the Indianapolis and St. Petersburg Police Departments were responsible. The
neighborhoods that were chosen represented the higher end of the continuum of
socioeconomic distress. This being the case, an accurate picture of how police act in the
most stable neighborhoods is difﬁcult to glean. What is left to analyze are areas of high
economic distress and areas of medium economic distress. The limited number and type
of neighborhoods raises questions of the generalizability of the results uncovered.

A related problem is the number of neighborhoods selected. The use of only
twenty-four neighborhoods was limiting. First, it was limiting in terms of the variation
between all types of neighborhoods, ﬂorn the best to the worst. Second, it was limiting in
terms of the number of variables that could be used at the second level. Certainly other

indicators of “good” or “bad” neighborhoods could have been chosen, however with such

148

a small number of level two units, the number of independent variables at level two was
restricted. The goal was to choose indicators that had been used before, were
theoretically relevant, and also indicators not only pointing to negative traits (i.e.,
disadvantage) but also positive assets (i.e., percent homeowner).

A problem that is directly related, or better, underlying the issue of chosen
neighborhoods is the use of secondary data. Obviously secondary data is limiting - in
terms of how data was collected and what speciﬁc data was collected. The current
project is no exception. While enough and a variety of information was gathered in the
original project, the data was not originally designed for the purposes of the current
project. There are speciﬁc issues for the current research that result ﬂom using this
particular secondary data.

The use of secondary data precluded the cmrent analyses ﬂom using all observed
events. The geographic location of activities was not known other than whether or not it
was within or outside the study beat. As such, only events that were 100 percent inside
the study beat were able to be used. The reason for this was to be able to identify
activities geographically. Events that occurred outside of the designated study beat could
have been inside another study beat, thus muddling the picture. This being the case, the
decision was made to eliminate events that did not occur inside the study beat for its
entirety." The problem here is one of precision and not necessarily validity.

Another problem for the present research presented by the use of secondary data

involves the categories compiled. The three dependent variables that were used for most

 

3 ' While this may seem an extreme limitation, it should be noted that the vast majority of events (upwards
of 80 percent) either occurred 100 percent (all) or zero percent (none) inside the study beat.

149

of the analyses were created by collapsing problem and activity codes into one of the
three categories. While great effort was expended to appropriately categorize these
variables, the fact remains that these were a reclassiﬁcation of originally coded data. The
three variables created and used for the current analyses attempted to capture differences
in types of police activity using admittedly broad categories. The problem with the
categorization of these behaviors in ways not initially designed for is that there could be
some overlap in exactly what type of behavior is being exhibited by the police. For
instance, while “checking out suspicious circumstances” was categorized as order
maintenance, it could be argued that it is also legalistic.

As discussed previously, clear logical lines were drawn to differentiate the three
types of behavior. The order maintenance label was applied if no violation of the law was
evident or if the event involved nuisance or quality of life issues; legalistic events
involved crime ﬁghting or law violations, and service events were ones that could be
done by anyone. There are certainly speciﬁc event type classiﬁcations that could be
debated, however, justifying each of these as one type of behavior or another would be
cumbersome and distracting. While not initially coded as one or other type of behavior,
careful attention was given to appropriately identifying and classifying each type of
behavior.

While these broad categories of behavior have been used in previous research and
are in wide use when classifying types of policing or police work, they may gloss over
types of order maintenance or services provided. For instance, if an ofﬁcer is involved
with a group of people suspected of drug activity (maybe after dark in a park), the ofﬁcer

could also be ﬁmctioning to maintain order, or providing services such as suggesting

150

alternate recreational activities for young people. The broad categories of behavior may
not adequately capture what is precisely going on at the encounter level.

While the problems and limitations are serious issues that merit attention and ones
that should be kept in mind when reading and interpreting these results, they are not fatal
ﬂaws. There is still variation between the neighborhoods that was able to be identiﬁed
and used; the secondary data while limiting, is still a widely accepted way to analyze data.
The data set while ﬂawed, is still the best police research data that we currently have.
This data set is substantial and enough information was available, useful, and captured to
give this research and a variety of other researchers ﬂuitful avenues to carry out their
work.

It is also worth noting that other researchers have analyzed data collected ﬂom the
same project, including examining contextual inﬂuences. Terrill and Reisig (2003) used
data ﬂom the same project and hierarchical linear modeling to examine police use of
force by neighborhood. They found that ofﬁcers used more force in neighborhoods
characterized as disadvantaged. This examination involved encounter-level measures and
was able to geographically locate all of these encounter events, resulting in 98 Level 2 .
cases or neighborhoods. Reisig and Parks (2000) examined and found that residents ﬂom
more disadvantaged neighborhoods expressed less satisfaction with the police. The focus
of this research involved citizen surveys and was able to geographically place the citizens
in 58 neighborhoods between the two study sites. The availability of more (and a wider
range of) neighborhoods may explain why these other multi-level examinations utilizing

the same data have been able to identify signiﬁcant neighborhood effects.

151

Future Research

There are several avenues for further research that arise ﬂom the current research
and alternate avenues for future research. The amount of contextual level research is still
dwarfed by other police research such as individual and situational level examinations. In
this section, these areas will be offered.

The importance of considering context in explaining behavior is highlighted in the
psychological maxim that behavior is a function of the person, his or her environment,
and the interaction between the two (Levvin, 1935). The relationship between person and
setting is reciprocal - people inﬂuence their settings and settings inﬂuence people (Allen,
1990; Oxford, 1992). Future research of contextual level inﬂuences of policing should
examine the potential interaction of place and personal characteristics.” For example,
community police officers may not only provide different services ﬂom general response
ofﬁcers, but may also behave differently in different areas. Likewise, demographic
characteristics of ofﬁcers (e.g., race and sex) should be exarrrined to determine if
interactions exist between these characteristics and place characteristics such as the racial
composition of neighborhoods. For instance, does the race of the officer interact with the
race of the place resulting in differential policing? Future research could help answer
these questions and identify potential interaction effects of persons and places.

As described above, the current research has several limitations that future
research could address. One of these areas involves the number and type of police beats

sampled. Future research should strive to ensure that all types of neighborhoods are

 

32 Variables capturing potential interaction effects were not included in the models examined due to the
limited Level 2 cases and resulting number of Level 2 variables that could be included.

152

represented. It is important to know how police act in all types of neighborhoods, how
they spend their time and the services that they provide. More neighborhoods available
for examination would be ideal, and if possible, an entire jurisdiction. Not only would
different types of neighborhoods be represented, this would allow more level two
variables to be included in the analyses.

It would also be useful to be able to pinpoint the geographical location of
activities. While this presents serious methodological concerns, especially considering
the movement of ofﬁcers while conducting general patrol, effort should be made to
capture this information. At a minimum, the speciﬁcation of which neighborhood or beat
an activity may have been carried into would be beneﬁcial.

Future research should also consider different measures, both independent and
dependent that could impact the delivery of police services. At the neighborhood level
there is an abundance of possibilities. One of the most obvious neighborhood variables
that was not available for the current study, which could impact how police behave in
certain areas are crime and victimization rates. This would be an objective measure and
one that police would be familiar with and potentially respond to in their choice of
actions.

Other contextual level inﬂuences that should be considered include measures that
previous neighborhood research have included. One measure in particular, collective
efﬁcacy has ﬂequently been shown to have important implications for community level
research. Sampson and Raudenbush (2001), Sampson et al. (1997), and Sampson and
Bartusch (1998) deﬁne collective efﬁcacy as cohesion among neighbors and the

willingness and/or ability of residents to intervene on behalf of the common good.

153

Collective efﬁcacy has been shown to impact neighborhoods, for instance, in terms of
levels of violence. The neighborhood variables included for the current analyses have
been shown to be related to collective efﬁcacy efforts: concentrated disadvantage
decreasing collective efficacy efforts and stability increasing collective efﬁcacy efforts
(Sampson and Raudenbush, 2001). This same research indicated that the effects of
concentrated disadvantage and residential stability was mediated by levels of collective
efﬁcacy. It may be that collective efﬁcacy is a better, more direct measure of informal
social control than those included in the current study. Indicators for this variable were
not available for the current study. The neighborhood variables that were available and
included (i.e., concentrated disadvantage and percent homeownership) may be too distal
ﬂom inﬂuential neighborhood forces that would inﬂuence police behavior — such as
collective efﬁcacy or other measures of informal social control.

In addition to the measures described above, more subjective contextual measures
should also be examined. One important subjective measure would be officers’
perception of an area - as stable or unstable, as helpful or unhelpﬁrl, as having crime
problems or being relatively crime ﬂee. Previous research has indicated that officers
carry images and perceptions of areas, not only of areas where they work, but by
communicating with other ofﬁcers, of all areas in a city (Bayley and Mendelsohn, 1969).
These cognitive maps, as they have been termed, could impact they types of activities in
which ofﬁcers engage in these different areas and should be studied.

Another variant of the current research might consider, instead of the broad
categories of outcome measures used, more speciﬁc encounter level variables. It could be

that the failure to ﬁnd signiﬁcant variation between neighborhoods is a product of the

154

choice of outcome measures. Encounter level variables may reveal differences in the
types of services offered, the helpfulness or unhelpfulness of ofﬁcers, or a variety of other
behaviors dealing with how police interact with citizens in different areas. It may be that
the categories used here were too broad to capture the nuances of police behavior.

The current research contains some interesting ﬁndings that also point to potential
areas for ﬁrrther inquiry. For instance, results revealed that community police ofﬁcers
and general patrol officers had the same amount of ﬂee time; this despite the fact that
CPOs do not routinely take calls for service. This ﬁnding should be examined and efforts
exerted to duplicate or refute this ﬁnding and determine how and why this is the case.

Efforts should also be made to determine why CPOs are spending signiﬁcantly
more time outside of their assigned beat than general response officers. While potential
explanations have been offered, the true nature of what CPOs are doing is not known.

These are only a few of possible areas for future research. Some of these
suggestions arise ﬂom unique and surprising results uncovered ﬂom the current research,
while others extend the important research that has and should be conducted in the area of
contextual level inﬂuences of police behavior.

Though the current study is limited in its ability to identify strong inﬂuences of
police behavior at the contextual level, ﬁndings point to the differential delivery of police
services across neighborhoods. The two key ﬁndings endorsing this point are that
officers in more stable neighborhoods left those neighborhoods more and that ofﬁcers in
these same neighborhoods engaged in less ofﬁcer initiated order maintenance policing.
Future research should continue along these lines of research in an effort to identify

neighborhood level inﬂuences on these and a variety of other police behaviors.

155

APPENDIX A

156

.985 85:00

099:

2.28 8::8 :88

£050 :8 .85:8

:88 .88: 8:08:88
8098.080: 8 .88.: 8:88

.085 85:8

MW.58.: :5 8:88

0:0:
88.55 .5830 :8 .80
838 8:0: .853
85:8 :8 .88:
.8888 :88»: 80:88

:8: 80.8 8850
80882 850

a. 8850 .888 8x8
80£8>

.8882 .8088.
.3850

88 8850
my: .8800

S8 8850

S8 888
8.88: :89

888: :89

8:82 9%

8800 .9098 B850
8800 do:

88 3850

8.8: 88 8838: 5:5 858 on 5 88.. «8.8

8:8 .8 8.8 _m a 8:8 E 88 as a. 88 8:8
:5 a 8.8: 8.6.8: 8:8 9888: 8:8 «2

23.85 8838: .8: 2: a 8:8. 8:8
838: 3% :08 8: 88 88:8: 888 8:8 .8:
$2.88 8: a 885 88 :8 .88 828322

88".: 3838: d: 888 3 8o 8.88.: :<
88".: 5-8: 88:8 >2 8 8:88 8:88:
8m.m 88.8: 83.8 9— 5 :9: 3:88.: 5 88.. 808
88 .8 8085.: 888838 :8 80:88:30
coma 88.8: 8:68 A: 5 M9: 8:888 E 88: 80.8

8:: 8.38 8 88...:

«gm: 8:8:

8:: 8.2.8.3

:8: 8:8: as 88:
a8: :8 :8 88$.

:8: £88:

8:: 888: a. 8:8
s8: .8:

8:: «:20

38: «55

 

D85 85:8 :88 .88 8850 8:— :_ 8:8 .m.D 808— on :_ 80.8 «N8 Q .88.: :0 58$
08: 028:8: 30.. 2% :0 8088: 8:0: .8 888: 02:80:50 9:: t 88.5
88 - 88m
80: 8:0: - 88m @888 :8 8858 88: «88:09 8:0: .8809 «8.: 5:88:83 we 8:99
.m.D 05 E 388:8 m .8 8:88
.0080 380: 038:8: 888m :_ 8058880 $303.88 .88..» 800: «8:: 889
0.8.889: .85885 .89 8:8 :0 8:82 3853‘

888: 8:2: - 8.885 .8380 .< 8.88

.3

.m—

.m—

.:

.0—

800'

vi

157

0:...38: o\o

 

88: :88 8:5 80.88800 5:888 0800:. .08: 080 .80: o\.. «8.: Q 5.8.5 u: 3:5

8:0: .3 8...: 0:0: 08.8 ::0.0.>
88.8 \ 808.0 .3 8...: 8:0: 8.:.U ..0>0. 808: 30.0: 8:88.. ..0 o\.. 60.88.80... .8: «N8: 8:8:
08: :88 8.:.U 0%: ::08::0>0m .80 «mm: 1 8.80383
820.80: .0088 00:0: 08:: 08: 08.8 ::0.0.> 80:85.88. 0.8088088 «Rot 8:00: a. 0.8::
mw:....0. 8:0: 08:m 08: 080 ::0.0.> 18.9 ..\o 5.8.8:. 0800:. «:8: 2.2% :8 800::
.088. 0:83 8 :88 8800:: ..0: a. 0.0.35.0 80...: ..0 :0.:.8:800 mmm u... 8.8: «:8: :88:
88:00:: 00:0: 8.8.: 00:0: a. 0.0.3-80 :88 8.0:: 8.8.8:: :88 .8885 8.888: «30: Q 8.93% a: 8:06

88:80:: 8.8.80:

08:00:: 00:0: Wm. 8:8 8:0: 08: 020.80.. .8: 88: :88 «3.: t 0:0:
:88 .08 2888. 8:808 8088:: 8:088 «N8 C «8.5

.8 \ 8 .038. :wE b8
:88 b8 .80.:..::00 0.80800 5.08888: 8.08.880 «:8: «8.6
:88 .08 8.8.8.» 80800 .00.::0: .Eo: 8.08.. .8 «N8 : .853: a: 58$
8. 38 8:. 8. 8:: 85...:
05:88.0 x..::::.:: 800508.80: 88588.0: ..0 m.0>0. 0800:. .08 800.08 8.0:: «:9: 5:080:83. a... 8.89
.2800 ..0 8.80:8. 800502.80: 0.08. 0800:. .8005858: 28:80:: «8:: 285%
0.8.8.» 080080 .08. 80:80 830...: 3:80:80 N0: Any—05.5

158

0:8

088 8:88: :8 08m 8:8 .8 80:80:08: 8:8 :88 080:8 £28888 8:8 .8 8:80:88 080800 :8 :28:
8:80:80 .80: :8 b8 :3 .088 0: :80: :05: 88:8 00:0: ”08:8 8083 .8 8:8 8:8:

08 08:: 08:3 880 8 88:8 00:8 .5: 8::— 82:>8 .8 88:88 8:8:

89:88

38:82: 8.: 8:8 :88 8:8: :.>0w .08 88:88: .59 :8 88:2 8.: 8:8 :88 8808:: 28880>0w 8888 .08
2:088:38 0:: 0: :08: 08 :0::3 ..8:::28m8a..

08 88:8 8:288:88 88888 0: 08098: a 8 88.03 .8 08 00:2: 8298 8:880: 00:0: :8 0088?
0:8: :0 208— ::3 9:80:20: :8808 888 cm 0:: .8 888888888 0: 8:20: 8:880: :088 8:0:

00:8 >: 00.8.: 850— .8

0m: 8808 838 8 80808: 080800 8 88:88:: 8.: 8:888 8:3 .885: 00:0: .8 88808 .80 8:8 00:88;
808

0::38: o\.. w: _: 0::38: 8.: 88:0: 383: :00:.: 0:: .8 8:80:88 :28: :8 8::0.:.:0 .8 008 8038: 88808:

88:8 :02: 8.: 8:8 8:805. 8:9: 8298 0880:8882: 8298 8:08: 9:80: .808 08:8 :8:
88:00:»: 0: 8:20.: .888 8:8 0:888: 8 38:820.: 8088 .8 8:8 - : 80...: 0: 8.888 0:: 0: .88.: 80:8
8:880 :8 8:2: 8.: 80:95: - 88:8: :88 8 8082.8: 88:88:88 :8 80:02.8: 008

:28 30820.: 08 .888 0:: _: 888 08:: .8 o\.. 8:3 88.888 9:: :8 :02m :88 0: :00:— 088 8:0:
808 :88 0: 038:8: 0: :0: 88 .085 8:8 0:: :8 E005 8:88 8.: 80:98

:88 80828: 088:: .m8 8 8:888:80 :8:

s8 2:: :3:an 8888 .888 :82 .8 582:. 2:: 0828: 238 .888 :o 28

088088: 8.: $88 .8 8880.8 .wﬁ 88:888.: .8 - .0005 8:88 8.: 80:98

0800: 8:: 8:8: 80:88 0888 0: 8808

:8 5:82:50 8:: 8:8: :2: 8 :08: 8080: :8: :0808 89:80:20.: 9:83:80 x: 08230:: .8 80: :2: a

«8 a 5:88 :. 33

«No.2: 8:8:

«8: @5283

at 3%: u. 38:
at at: :8 38%
88: 88:

:8: 838: a. 8:8
e8: «8:

88: «5.8
88: «55

a?» 8:89 my 88.5

 

0:88 8850 808880 8:88 .8 88:80.8 888.88: 8:: 0:: 38-208 .8 8:8 .208 .8 88888 8:08: 8:0: Rout 88.5
0.80:8: .8 088:: :8 008 8 :08: 808 808.8: 8 80:8: :8: 88:28.: :8 8:8: 8.: 8098 O :88 8:88 «at 5:88:83: 8 83$

80:8 :: 30:: :0: :803 8:: 8:: :00:.: 08 :: 828:0: 30:: 00:0: .2888
:88: 80:0 38:808.: 8888 .0880 088 8880: 8:8 888:8:8 .8 82:08 8 ”88088 888.80 88:8 «8:: 88%
1.8.0:: «895::

159

85:00
500580800 30mm2000
0:0:

8030: :wEB

0:0 85:00

A5005 0000 0 005
85:00 0:00:08

3000—08003 .0005 0800000

.50— 00 E005 0.0.005

0:0:
00000 :0003

0:0:

85:00 3 00:08.00: 0800

000305
85:00 0:00:05 0:0:

500 8000, 3:03 ”A0000
:8: 00.00 000v 0:0:00
0300,0005

0:0 800030000
000000.. 0003000

58> 80 .800
0:0:00

0% 0000005009
00:000.. .0009

do:
000200

.08 0:0 :000>._0000

000200
.08 0:0 000080000

_0:000>._000 0.00800
080,000 050020000

do: 03:00

03:00 0:0 :000>._0000

88: 800088 +0880 55 0.020 00

0.80808 wgw 0:0 00000— ..m.m 0:0 000200
0000: >2 .8800 a 0288 209% 8:8 _03

A000: 0.0 8205083 a 85:8 0:028 8:8 80.0
800.280 88: .00 8.80 .8050 80 0:00

$0.10 00-2.2 08.20220 5 00008.0 8:8

08.80.. 9 08.80 .08 a? 800 0.: 8 s0 5% 005
800% S 85.22008 05225 23888 2: Ea: 000

0003000 00000 .00 @005, 0
8 0688 08.08000: Ea: 000

0000:0800 000:0 000080 003 93: 000

005 800000.000:
5 b8 0:0 5 b0>000 000600 .00 0500000

85.03: 00000 .m.D N0

8.01% 0850500 0.2888 8:8 2888 .5308 :0

s00:

335 0.3 30.5.8
000:

0.5.0.: 0 0020200

:8:
0.5.00 0 00:30

0 8: 0530
000: 0030
08: 085020

a 00: .350.

000: .330
000: 5.20.
0 00: .3600
«000 0 0820

«0800 .00 a .362

 

00000093 000—0000:D S00 dd .00000 03— 0—0700.. 008000 00.0 0:00 .. 88000009 00000 0$005 «000: .00 00 30008:
E005 86:00 500 000.00 30000 000.02 00>0 :0003000 500 0008 E. E 000000 «0.0.0 Q .00 8 30.3
._.—03050.: 08000005 30G .0050 ._0 9:502 3005.5.

.wm

KN

.0m

.3

.vm

.mN

.NN

._N

.om

.0—

.m—

.2

.0—

.m—

160

monau amok—d

0.:

0.0000000 00000
9.00000 0:0 0008000.
08.00 00 0:00.05 w5000000

0800000 00000

00000 000000

900000 60000000 000.008. 0:0.0.>
.0:000000005 00 00000000

000000 00.00080 0>80000

000000 600000.000 0>000000
:000w000>5 0>000000

80.0800 000.000

.ms. 60005050800000 .00.>000

0.50 00... 00800003

0% 00055000030 00000
000000

00000000 008.00

000000 .00 550000
0.00.00.» 080080

0008

008.008.0000
0.0>0. n 00 000080030
08000 000000 mm.

008.008.00.00 s
0.0 0

30.00 8
088.888 8
88.08.0002. 8

008.008.00.00 8
008.008.0002. 0

80.0 N0.

008.008.0002. N...
00000 b08888

E. 8 0000w00ww0 00000

000.0
.200. 88:00

0000 000058000, .0000 08.00
.0850 .050 00.00.0000 00.00.0800: ..\0 500.0 .x. 5:00.005 0.80:00m.

00000000030: .0008 0030.. 0000 0:0 ..000:

50.000800 .0800 0000.: .A0w5..030 0:000 ..\.. .080080

.0800 .00800 .00 050> .0.0>0. :0000000 60.0 80800—0800 0000 .00 mmm
000.>000

00.. 0:0800 .0000 b.0008 500.0 o\.. .800 08800.. ..\0 50050 5 .x.

00000 08.00
A ...... 8080000 0058 00 :000 $800.. .0580 00.03 .0... 008.00 800000
a0w0 60.0300: ..\0 5000005 08005. 8085580 08.00

00000 50058.88 0000.: 0000.: .00 08000 0.80:000

we 00>0 .x. 6:20 w5>.. .x. 000.0000

000...: ..000 08005 5.0:0m00080 .0800 £000.08? 080000 0.m50

0000050: 00.3 :00000005 50.008 80:00.000 .0800 mmm

0000 80058.80, 500.60 800000

w50000 0080.0000 ..0>0. 08005 .0w0 :0008000

:00000. E00800

00.0008 .x. 850.00.: :0803 ..\o .0000 08.00.0000 .0000 08.00 80.08

08800.. .x. 100.0 .x. .08005 0»... .x. .0:.0> 0800 05.00000 00.008

00.000000 00:30 .0. 8000.08 o\.. 58:00 0.0000000 .000w0:000 ..\o
:08. b00>00 0:0 00.000000 00800 00.000002. 0.80.. 0.w50

000800800 .0800 00000.0. 008008800 0 0:0 b.0080 8.800.000

000.>000 00.. 000800 .08005 0.80.. 000.008 .080...— ..\o

0000 X. 5000005 08005 80009800000
00.00.09, .00-0000000 00v—

800

0.00.0 0.3 2.0.58
«0 0.0

5.0.0.0 0 005.00.»...
3.. 0.00.0. 0 .0000...
«080 .0500.
0.080 0008.00
.0080 008030

0 00: .320.

«000 0. 00.20

000: .0020
008. 096.0

008. .3600
080. .00 0... .0962
s00 00 .00 .0 0.0009:

0.0000 .00 0.. S03
@0203.

161

0... 0000 00.0

000000 .00 0000 000 000 0.00.00 ..\0 000 03 00000 00 ..03 00 0.00000 0000>00ww0 00.0 0000 000000 000 0:0 00.000005 0.800000 0003000 0.00:0.00.00
00005800000 8080.000 00>: 00 00.03 5 0000.: 000 I 000.0 00 00.0 0000.0

0. 00.03 00008.0 000 .00 000000000 .0008 000 000.05 00.000 00000 m55800000 000000.. “0000005 0000808 w00w 000 00000 300 00580000
0000.. 0000.00 00 - 000000.0m 0. 0.0005002 00 .0000000 8000.000 00000 00000000

0>00 0.00.0 000 0000 5 000000 00030. 00..00>0.a 000 00000 000 0000000 .00 00:00.05 000 00 0000.00 000 00000 .00 mum 000 0000 - .0d 5
00.00000 .00 w5000000 w500.0000 5 w.0 0003 00.00.00> 0000.: 000 .00 30.0.0000? 000800000

000 000.w0: 0w.:00 00000000 - 00.03000 .00 w5000000-000 00000 00 0000.00.w.0 0003 480.0 o\o 0000000 00.00.00> 0000.: 000 .00 ..0

00000005 0w500000 00000 .0000 00000 08.00 00 I 0008000080 000 5 08.00 .00 .0>0. 000 000 8500000 00000 0003000 500050080
0.00.00> 008.0000 000000000. 0008 000 w500 00.0300: ..\0 00.3 00000 000000 00 0000.00 0030.000 0. 808000.80 08.00 0008 .00 000000.05
80000.: 0000000030. 0. 000000 00 .000... 0008.

000008 0003000 00:0.0.> 0.0000 00000 300 0000.00 0800000005 00000000 000000000 000 00.03 5 0000.: 00.0 .00 000000 0.800000 000
00000.: 08.00 000w.0 5 000000000085 0.0 00 .000... 000. .000.>.05

.00 0000 00: 0000.: .00 005000800 .0800 00 000.5. 0.000000 0>80000 .00 000 .0000.: 000000 0030. 5 0008 00000000 0800000 08.00 0w.0 5
000000.0000 0000 0.00... 000. ”00000.0.000 0.0800 5 00000.0.> 0000 d 000000000003 80800 0008 80.0.5 0% 00:00.000 00 000000.000 0008
000000 :0 0000 00.000.00.000, 0000.0 .00 000.00.».0 00

00000005 00.0 000000000 00>.» >00 5 000000 .00 50000000 000 .00000005 05. >000>00 000 30.00 00.0000000 .00 0w00000000 000 00
000000: 000 000 000000000880 0003000 0000000....0

b0>000 00.>000-00..00 .000. ..0 0000000 000 w500000000 5 00:0 .00008000.>00 0803000 00.. 0000.00 080.000.0090

005800000.

00000 000 u 000050080 .0w000 0 n 0008000080 - .0000 5 000008.08 800.000.0080 00 030000.000 0.0008 00.0.0 0003000 00000000000
00000.00000 0. 0.003 00000 000 03000030 0008 000 .0000 00 .00 mmm 000 0030. 000

000: 3080
0.... 305...:

000: 00.0.0.0
% 080000.000:
03:

000...: 0 02.000
0 a: 00.2.00

0.00: 000%.:
.03: .8820.

0 000 £050

000: 5.....0
000: 5.....0

Q 00: «0.000%

000: 0.300

00.00.000000000 .0>0. 00.008800 000 00 000 000000.080 b0> 00: 00 000.0800 000000 «30% .00 00 00.82

000.8000 w5000000 5 0000.09 00 000000050 00.. 0000000 00: 00 00.00.800.208 08005 000 .0800 00000000
000000 .00 00.000 0030. 0>00 000.>000 00000 00.. 0000800 000w.0 00.3 00000.: ”008.00 w5000000 00 0000.00 003 000.>000 00.0 000800 .000

8000
.00 .0 008.5

00000 .00 050000 000 000.00 00008 - 00.00.00> 008000000 000.0000 000 00.. 000000 0300008000 $.00 C .00 00 30......

 

00.00:0

@0203.

162

APPENDIX B

163

Appendix B. Categorization and Distribution of the Three Outcome Measures.

Table B.l. Order maintenance events (n = 2,928)

 

Actm ﬁes _n Problem codes 1; 3
General motorized patrol 2,443 Suspicious person 52 (4)
Check out susp. 93 Loitering 33 (4)
Back up other police 35 * Drunk 26 (5)
Res. security check 19 Public nuisance 19 (2)
Com. security check 16 Domestic argument 15 (13)
Parade/crowd control 13 Susp. circumstances l4 (3)
Foot patrol 12 Juvenile problem 13
Direct trafﬁc/parade 8 Noise disturbance 12
Information gathering 7 * Trafﬁc accident 12 (3)
Problem focused activity 6 * Argument 10 (3)
Check out situation 4 * Vagrancy 8
Conduct research 1 * Adult subject of 8
police concern
Juvenile subject of 6 (1)
police concern
Peddling/begging 5
Neighbor trouble 5 (2)
Mental disorder 4 (l)
Landlord/tenant 4 (2)
dispute
Abandoned vehicle 4 (2)
Trafﬁc ﬂow 4
Litter/trash 4
Suspicious motor 3 (1)
vehicle
Disorderly conduct 2 (2)
Family trouble 2
Suicide or attempt 2
Alarm 2 (2)
Inter-group conﬂict 1
Crowd control 1
Police protection (3)
Total 2,657 271 (53)

 

*53 activities involved an "order maintenance" problem

164

Table B.2. Legalist events (n = 1,332 )

Agility 0
Attempt to locate suspect/witness 133
Back up other police 81 *
Surveillance - address 75
Trafﬁc mobile 65
Trafﬁc - stationary 43
Surveillance - person 33
Information gathering 27 *
Search property 23
Problem focused activity 20 *
Pursue ﬂeeing suspect 18
Traffic enforcement 13
Parking 13
Search crime scene 12
Building code 11
Guard crime scene 7
Process evidence/property 7 *
Conduct research 7 *
Warrant/ subpoena service 3
Health/sanitation/trash violation 3
Check out situation 3 *
Ordinance enforcement 1
Total 598

 

*142 activities involved a "legalist" problem code

165

Table B.2 (eont’d).

 

Problem codes

Moving violation
Equipment/tags violation
Drugs

0 wants crime information
Warrant to be served
Trespass

Non domestic assault
Vehicle violation
Parking violation
Domestic ﬁght

Motor vehicle theft
Suspected violator
Burglary

Prostitution
Curfew/truancy
Vandalism

Obscene activity
Alcohol

Problem with money

Fight

Theft from residence
Routine check
Harassment/stalking

Missing person
Burglary - commercial
Civil code problem

Theft - unspeciﬁed

Theﬁ from motor vehicle

134
111

(12)
(5)
(25)
(9)

(3)
(3)
(1)
(4)
(3)

(3)
(5)
(1)
(1)
(1)

(1)
(3)

(4)
(1)
(3)

(1)

(1)

(2)

166

Child abuse/neglect
Robbery - citizen
Missing/stolen property
Juvenile runaway

Flight from police

Robbery - commercial
Rape

Theft (attempted) - comm.
Arson

Hit and run

Leaving the scene

Gunshot

Weapons violation
Arrest/booking

Domestic assault

Homicide (or attempted)
Buy/possess stolen property
Damaged property
Nonpayment of child support
Physical injury (or threat) by
person

Child molestation

Break in - residential
Intentional damage of
property

Citizen case related
information

Citizen crime tip

Transport person in custody
Officer needs aid - weapons
involved

Interference with police

Total

0311—-
vv

-NNN0www0hhhhhhhmtﬂO~
AAA A
NNN N
vvv v

AA
ﬂ
V

._.—._.-
p—A
V

1 (2)

(2)
(1)

734

Table B.3. Service events (11 = 272)

 

Activi 2 0101000100210: 11. 1
codes)

Meet gov’t agency ofﬁcials 26 Ofﬁcer friendly/ 45 (2)
comm. relations

Meet with public 18 Give information 40 (7)

Meet non police service provider 18 Assist motorist 24

Back up other police 9 * Directions 7

Meet neighborhood group 8 General request for service 6 (2)

Meet business individuals 7 Animal problem 6

Service 5 Compliment/complaint 5

Check lﬁx road conditions 5 Crime prevention info 4

Check/ﬁx property 5 Medical assist/person down 3 (1)

Problem focused activity 4 * Help disable person 3

Check out situation 3 * F ire/ﬁre prevention 6 (1)

Escort 2 Funeral/parade/escort 3

Information gathering 4 * Lost property 2 (1)

Medical/health service 1 Emergency transportation 1

Transport prisoner/witness/evidence 1 Road block 1
Dead body (2)
Environmental hazard (1)
Escort (1)
Transport (2)

Total 116 Total 156 (20)

 

* 20 activities involved a "service" problem/problem code

167

APPENDIX C

168

APPENDIX C. Contextual Information From Citizen Surveys - Original Analyses

Described below are the initial contextual level variables identiﬁed for inclusion
for in the current research. Due to data limitations however, the use of these variables
was not feasible. Speciﬁcally, the reliability of the citizen survey data was low making
the use of the variables questionable. Census data was used instead.

Initial community-level variables were obtained from aggregating citizen survey
data. This information about neighborhoods and citizens was obtained through telephone
interviews with citizens conducted in conjunction with (although by different research
personnel) the observational study. The target sample was 100 interviews with residents
18 and older in each study beat (neighborhood). The sample was stratified by
neighborhoods; households within the study beats were randomly selected from telephone
directories and telephone surveys were conducted. Interviews were conducted by Indiana
University’s Center for Survey Research (Reisig and Parks, 2000). The citizen survey
used for the project collected a variety of citizen information including (but not limited
to) length of residence, neighborhood satisfaction, identiﬁcation and levels of
neighborhood problems, perceptions of crime and fear, satisfaction with the police, and
demographic characteristics. A total of 2,343 citizen surveys in the 24 study beats were

available (average number for each study beat = 96.6).

169

Neighborhood variables

A number of contextual variables believed to impact how police deliver services
are available for the current study. Some of these neighborhood level variables have been
included in previous research efforts examining differences in police behavior, others
have rarely been included. Table C.1 details these variables, provides a description and
coding, and the distribution of these variables.

As mentioned in the previous chapter, race and income are commonly included;
other less commonly examined variables involve social aspects of the neighborhood —
indicators which tap other forms of control. These are especially relevant to the current
study due to the proposition that police compensate for the lack of other controls by

engaging in more control oriented types of behavior.

170

 

 

33 e 2 3% ~53 been? .x. been? .x.
bmmtmwexﬁmt
couﬁﬁawco
mom 9 m3. :3 3.8 83.8%? .8 33808 as Ba .x. Suﬁswaa
AKA—S. W I .HO>OG. 0v
mad 8 mm; cad EN 03% ﬁonnwmoa 53> coﬁomou How aoﬂo 3o: 3030835
0:.“ Hmcadd. w I .06—Ga. 0v
NVN 8 9V.— wmd mod 03% no.8 E moZﬁ—otmucowa go 898:: x838:
.Eugwégﬁé
@me b0>. M I .0.“de b0>. 0v
SN 3 2: m3 3: one... 5% a 8% magma co we as
92 3 35 m3 a: $2.85. m - .88. 8 Seesaw»: co muse consumes 82:22.35:
totoﬁntum
m. a 9 m: 8.: Sam 8832 35.95 8V maize .x. .56 8a? as
a? e 3 :3 2.3 was» m as use Seesaw? a was: as $38. sunﬁsh
Sages bees:
owqam dd :32 @538 :oumwtomoﬂ £38835 vooﬁoﬁwmoz

 

 

g n 5 moumuaoaﬁo @8533 ._.o 0:5

171

The neighborhood variables described below are drawn from the citizen survey of
residents of study neighborhoods, aggregated to the beat level. A total of 2,343
respondents from the study beats were used to create aggregate neighborhood level
indicators. The mean number of respondents per beat used to create the aggregated
neighborhood characteristics was 97.6 (SD. = 14.4).

The key variables aggregated and presented for each neighborhood involve
indicators of mobility and stability, satisfaction, involvement, and heterogeneity. It is
believed that these variables can be used to distinguish neighborhoods from one another
and that these characteristics potentially inﬂuence the types of activities in which the
police engage.

The measures selected for the current study point to indicators of informal social
control present or absent in the study neighborhoods. As previously mentioned and as
previous studies and scholars have indicated, varying levels of informal social control
have implications for social control mechanisms (i.e., the police). The indicators chosen
point to geographical areas that exhibit higher levels of cohesion and social control and as
such, the treatment by formal social control agents may differ.

Mobility. The item on the citizen survey used to gauge mobility asked for the
length of residence in the neighborhood; this item is used as an indicator of residential
mobility or transience. Original responses ranged from 0 (less than 1 year) to 51 years.
The length or residency variable was recoded providing a dichotomous variable
indicating those who had lived in the neighborhood less than ﬁve years and those living
there 5 years or more. These categories indicated longer term and shorter term residents

and have commonly been used in other studies. Responses were aggregated to indicate

172

areas that had more or fewer long-term residents. Areas that scored higher on this
measure indicated more mobile areas and alternately, considered less stable.

Length of residence and residential mobility have been examined previously and
have been shown to have an impact in the development of social networks, sentiments of
attachment, and sense of community (Kasarda and J anowitz, 1974; Sampson, 1988; St.
John et al., 1986). High community mobility is said to reduce local friendship ties by
limiting and constraining friendship ties and hence reducing attachment to the community
(Sampson, 1988). These are all important considerations for the levels of social controls
exhibited by areas.

One indication of higher levels of stability used for the current study is the
percentage of homeowners (or buyers). Previous research has identiﬁed home-ownership
as a key inﬂuence of attachment and stability (Fried, 1982; Riger and Lavrakas, 1981;
Taylor et al. 1985). Neighborhoods that have a higher percentage of homeowners will be
judged as exhibiting higher levels of attachment and resulting higher levels of social
control.

Satisfaction. Community satisfaction refers to an individual resident’s subjective
evaluation of his or her residential environment - how people evaluate the place in which
they live (Baldassare, 1986; Hummon, 1992). Residents were queried as to their level of
overall satisfaction with where they lived; speciﬁcally, residents were asked to rate the
neighborhood on a scale (excellent, good, fair, poor). Shurnaker and Hankin (1984)
enumerate three consistent assumptions in the areas of community attachment and

satisfaction: that individuals can become attached to their physical and social

173

environments, that these attachments vary across individuals, groups, and locations, and
that attachment has important implications for both the individual and the community.

An additional item considered important in terms of community satisfaction is the
level of fear that residents have. The speciﬁc item included in the citizen survey asked
residents “how safe would you feel walking alone in your neighborhood aﬁer dark ?”
with possible responses ranging from very safe to very unsafe. Feelings of safety in their
own neighborhood provides an alternate indication of citizen satisfaction as well as
indications of levels of controls that citizens feel are present or absent in their
neighborhood.

Involvement. Involvement in the community refers to an individual’s networks
and interactions with other residents. For the current study, three indicators of
involvement were gathered and aggregated to the beat level: a measure of the number of
ﬁiends and relatives living in the area, a measure of how often residents interact with
neighbors socially, and the percentage of residents who are members of a neighborhood
organization.

One key consideration concerning involvement is the interpersonal or social
relations and interactions. A number of studies have demonstrated the importance of
these social ties, sentiments and feelings of attachment (Goudy, 1982; Herting and Guest,
1985; Kasarda and Janowitz, 1974; Sampson, 1988; St. John et al., 1986; White, 1985).
Social ties refer to factors such as the number of friends living in the area, involvement in
social networks, and participation in local activities or organizations. Kasarda and
Janowitz (1974) note that the number of friends in an area is the most important type of

social bond inﬂuencing community sentiments. Sampson (1988) states that sparse

174

ﬁiendship ties impede an individual’s integration into the community. St. John et al.
(1986) note that the number of friends one has is an important determinant of attachment
to the community. Collective attachment refers to a macro level measurement of the
percentage of people who exhibit indicators of attachment (Sampson, 1988).

Two indicators of these type of social ties were available and used for the current
study. Residents were asked how many of their ﬁiends lived in their immediate
neighborhood and also how many of their relatives lived in their neighborhood.
Responses ranged from 0 (none) to 4 (almost all). These 2 questions were combined to
give a measure of the number of ﬁiends and relatives in an area —- indicating social
networks, an important determinant indicating the involvement of residents in the
community.

Residents were also asked how oﬁen they interacted with their neighbors.
Possible responses ranged ﬁ'om 0 (never) to 5 (daily). Higher mean scores on this
variable indicated stronger social networks and were taken as an indicating areas with
more interactions and therefore higher levels of control.

Participation in local organizations is the ﬁnal indicator of involvement.
Responses from the citizen survey were aggregated to indicate areas that had a larger
number of respondents who were members of local neighborhood organizations. The
speciﬁc item on the survey asked residents if they were member of an organization or
group in their neighborhood that addresses neighborhood problems.33 Membership in

neighborhood organizations is especially relevant to the current research effort as an

 

33 The speciﬁc item referred to was a contingency question — respondents ﬁrst being asked whether or not
there was an organization or group in their neighborhood that dealt with problems. Respondents who were
unaware of such groups could logically not be members and were scored on the contingency membership
question as “no”.

175

indicator of informal mechanisms of social control. It is believed the areas exhibiting
higher levels of membership display more collective action efﬁcacy (or at least efforts)
and resulting less reliance on police. It is worth mentioning that citizen participation in
such organizations is a cornerstone of many community policing efforts. While
membership in organizations does not necessarily indicate active participation, it does
indicate a certain level of personal investment.

Heterogeneity. The potential effects of residential heterogeneity have been
studied in terms of racial or cultural heterogeneity and also income heterogeneity; for the
current project, heterogeneity is measured in terms of the racial composition of
neighborhoods. Previous community studies have provided mixed results in terms of
residential heterogeneity. Wirth (1938) posits that increased heterogeneity breaks down
rigid social structures and lead to increased mobility, instability, and insecurity. Other
community researchers have found that despite signiﬁcant changes in population, social
ties, networks and community sentiments still existed in what were considered
heterogeneous areas (Christenson, 1979; Goudy, 1982; Kasarda and Janowitz, 1974;
Rodgers, 1980). The heterogeneous indicator for the current study tapped the percent
minority population of the study beat.

The neighborhood characteristics outlined above capture differences between
areas and indicate levels of stability, satisfaction, involvement, and heterogeneity. The
individual level responses of residents are aggregated to provide a picture of
neighborhood level characteristics. Generally, these measures provide indications of the

levels of informal social control exhibited in different areas. It is hypothesized that the

176

differences in these characteristics will be related to differences in the types of activities
in which police operating in these areas engage.

Level 2 variables were examined carefully. Since many of the variables proposed
for inclusion in the Level 2 model bore some theoretical relevance to each other, these
individual items were included in an exploratory factor analysis. First, however, bivariate
correlations were calculated for the Level 2 data. Table C.2 presents the zero-order
correlation matrix for the neighborhood-level variables drawn ﬁ'om the aggregated citizen

data.

177

 

a. V9:

 

 

mQVQ.
2: :85- mod 36 :85- .. :3- So .23.. 85 beoazxa
2: 86. so mad and 1mg- :de one e236: emanates- w
2: ”3 So- Ed- N8 taco- 43- 83 59305805- 5
2: one :35 3o :6- .53 . 8808 .9038
2: 1&8 33 26 Ed ae €82 - m
2: 8d .53 moo Seance so: - a.
2: r. «2. 8o- museum ax. - m
8; ”we can» w v maﬁam .x. - N
2: 32m -_
a w a e m a. m N _ osaﬁ>

 

an n 5 83%; BoeBmwez .8 986880 8388 52903 .8 03.;

178

Several signiﬁcant correlations were revealed ﬁom the bivariate correlation
matrix which is a good indication that the items may reveal some underlying factor. In an
effort to reduce the number of variables and considering that the neighborhood levels
could be tapping into some underlying concept, the neighborhood levels were factor
analyzed. Since the data had been aggregated to the neighborhood level from individual
respondents residing in each neighborhood, the non-aggregated, citizen-level data was
examined for potential factors. Twenty-four neighborhood cases would not provide the
number of cases necessary to perform factor analysis — as it is generally accepted that 300
cases are necessary for such functions (Tabachnick and F idell, 2001).

The original neighborhood respondent data set, from which the aggregated
neighborhood level variables had been created, was examined to identify potential
factors. An initial correlation matrix was calculated ﬁom the neighborhood residents
database (See Table C.3 ). The zero-order correlation matrix revealed that there were
several signiﬁcant correlations (ranging from r = .-.20 to r = .46). As Tabachnick and
F idell (2001) note, if an initial correlation matrix does not reveal numerous variables
correlated at high levels, the use of factor analysis should be reconsidered; this would
indicate that it is unlikely that they share common factors (Norusis, 1990). As seen in
Table C.2, only 2 correlations are greater than the recommended level of .30
(renting/length of residence, r = .31 and fear/satisfaction r = .46). Since the bivariate
correlations among the items examined are small, the use and appropriateness of factor
analysis is called into question. Although low, there were signiﬁcant correlations and

therefore the items were factor analyzed.

179

Factor analysis attempted with these variables and the items within the
dimensions as described above. Scores for these factors were weak and reliability
coefﬁcients were extremely low (< .10). I therefore elected to seek more conclusive

contextual level variables for inclusion — choosing census data.

180

 

_o.Vm..:._

 

 

8.1:...
2: :NS- 8.2. 8.2 :86- 1.2.2 N22 :23 3% been: N
2: :oNd ::.o :23 :N; 8.2 Nod manages, 8:83: N
2: :85 :2; :85 2:. :25 0303332385 0
2: :85 ::.o :35 .25 sﬁqmouagaouawawho m
2: :23 2; :22 82.2530”: v
2: *2; :23- 82838 @892 N
2: :23 aeaaggo N
2: .205sz No 523 _
N N o w v N N _

 

22.3 n 5 8235, 3.3 85.32202 £56580 839:8 .8 22¢

18]

REFERENCES

182

REFERENCES

Agresti, A. and Finlay, B. (1986). Statistical methods for the social sciences 2"d Edition.
San Francisco: Dellen Publishing Company.

 

Allen, L. (1990). A Developmental perspective on multiple levels of analysis in
community research. In P. Tolan, C. Keys, F. Chertok, and L. Jason (Eds.) Researching
comb: psychology (pp. 131-134). Washington, D.C.: American Psychological
Association.

Arcury. T.A., Austin, C.K., Quandt, S.A., and Saavedra, R. (1999). Enhancing
community participation in intervention research: F armworkers and agricultural
chemicals in North Carolina. Health Education and Behavior 26 (pp. 563-578).

 

Austin, D.M., Furr, LA, and Spine, M. (2002). The Effects of neighborhood conditions
on perceptions of safety. Jom of Criminal Justice, 30 (pp. 417-427).

Baldassare, M. (1986). Residential satisfaction and the community question. Sociology
go Sociﬂ Research, 70 (pp. 139-142).

Banton, M. (1964). The Policeman in the communig. London: Tavistock Publications.

Bayley, DH. and Mendelsohn, H. (1969). Minorities and the police: Confrontation in
Amorica. New York: The Free Press.

Bittner, E. (1967). The Police on skid-row: A Study of peace keeping. American
Sociological Review, 32 (pp. 699-715).

Bittner, E. (1972). Ibo Functiops of the police in modern socieg. Washington, D.C.:
National Institute of Mental Health.

Black, D. (1976). The Behavior of law. New York: Academic Press.
Black, D. (1980). The Manners and customs of the police. New York: Academic Press.

Blalock, H.M. (1967). Toward a theog of minority-gloup relatiops. New York: John
Wiley and Sons, Inc.

Bloch, P. and Anderson, D. (1974). Policewomen on patrol: Final report. Washington,
D.C.: Police Foundation.

Bozza, CM. (1973). Motivations guiding policemen in the arrest process. Jour__nal of
Police Science and Administration, 1 (pp. 468-476).

183

Brandl, S.G., Stroshine, MS, and Frank, J. (2001). Who are the complaint-prone
ofﬁcers? An Examination of the relationship between police officer attributes, arrest
activity, assignment, and citizens’ complaints about excessive force. Journal of Criminal

Justice, 29 (pp. 521-529).

Brecci, M.G. and Simmons, KL. (1987). An Examination of organizational and _
individual factors that influence police response to domestic disturbances. Journal of

Police Science Administratiop, 15 (pp. 93-104).

Brodsky, A.E., O’Campo, P.J., and Aronson, RE. (1999). PSOC in community context:
multi-level correlates of a measure of psychological sense of community in low-income,

urban neighborhoods. Journal of Communig Pachology, 27 (pp. 659-679).

Brooks, L.W. (1997). Police discretionary behavior: A Study of style. In R.G. Dunham

and GP. Alpert (Eds) Critical Issues in Policing: Contempom Readings (3" Edition)
(pp. 149-166). Prospect Heights, IL: Waveland Press, Inc.

Brown, M.C. and Warner, ED. (1992). Immigrants, urban politics, and policing in
1900. American Sociological Review, 57 (pp. 293-305).

Bryk, AS. and Raudenbush, SW. (1992). Hierarchical linear models: Applications and
dag Mysis mothods. Newbury Park, CA: Sage Publications.

Buckner, J .C. (1988). The Development of an instrument to measure neighborhood
cohesion. Amgican Journal of Communigg Psycholog, 16 (pp. 771-791).

Bursik, R.J., Jr. (1986). Ecological stability and the dynamics of delinquency. In A.J.
Reiss and M. Tonry (Eds) Communities and crime (pp. 35-66). Chicago: The University
of Chicago Press.

Buttel, F.H., Martinson, OB, and Wilkening, EA. (1979). Size of place and community
attachment: A Reconsideration. Social Inoicatog Research, o (pp. 475-485).

Byme, J .M. and Sampson, R]. (1986). Key issues in the social ecology of crime. In
J.M. Byme and RJ. Sampson (Eds) The Social ecology of crime (pp. 1-22). New York:
Springer-Verlag New York, Inc.

Cao, L., Frank, J .T., and Cullen, F .T. (1996). Race, community context and conﬁdence
in police. American Journal of Police, 15 (pp. 3-22).

Chambliss, W.J. and Seidman, RB. (1971). Law, order, and mwer. Reading, MA:
Addison-Wesley Publishing Company.

Chavis, D.M. and Newbrough, J .R. (1986). The Meaning of “community” in community
psychology. loom of Communig Eychology, 14 (pp. 335-340).

184

Chavis, D.M. and Pretty, G.M.H. (1999). Sense of community: Advances in
measurement and application. Journal of Communig Pachology, 27 (pp. 63 5-642).

Chavis, D.M. and Wandersman, A. (1990). Sense of community in the urban
environment: A Catalyst for participation and community development. American

JogﬂioJ of Comqu Psychology, 18 (pp. 55-81).

Cohen, L.E. and Felson, M. (1979). Social change and crime rate trends: A Routine
activity approach. Amop’cg Sociological Review. 44 (pp. 588-605).

Cordner, G.W. (1995). Community policing: Elements and effects. Police Forum, 5 (pp.
1-8).

 

Crank, J .P. (1992). Police style and legally serious crime: A Contextual analysis of
municipal police departments. low! of Criminal Justice, 20 (pp. 401-412).

Crank, J .P. (1990). The Inﬂuence of environmental and organizational factors on police

style in urban and rural environments. Journal of Research in Crime and Delinquency, 27
(pp. 166-189).

DeJong, C., Mastrofski, SD, and Parks, RB. (2001). Patrol officers and problem
solving: An Application of expectancy theory. Justice Merly, 18 (pp. 31-61).

Forst, B., Lucianovic, J ., and Cox, SJ. (1977). What happens after arrest? A Court
perspective of police operations in the District of Columbia. Washington, D.C.: Institute
for Law and Social Research.

Fried, M. (1982). Residential attachment: Sources of residential and community
satisfaction. Journal of Social Issues, 3 (pp. 107-119).

Friedrich, R.J. (1980). Police use of force: Individuals, situations, and organizations.
m Amorican Agdemy Politioal Social Sciences, 452 (pp. 82-97).

Fyfe, J .J . (1980). Geographic correlates of police shooting: A Microanalysis. Journal of
Remoh in Crime and Dolinguenoy, 17 (pp. 101-113).

Fyfe, J .J . (1988). Police shooting: Environment and license. In J .E. Scott and T. Hirschi
(Eds). Controversial issues in crime and justice. Newbury Park, CA: Sage Publications.

Geller, W.A. and Karales, K.J. (1981). Shootings of and by Chicago police: Uncommon
crises Part I: Shootings by Chicago police. The Joon_1_al of Comm—al' Law Ed
mum; (pp- 1813-1866)-

Geller, W.A. and Karales, K.J. (1981). Shootings of and by Chicago police: Uncommon
crises Part II: Shootings of police, shooting correlates and control strategies. The Journal

of Qriminal Law and Qriminology, 73 (pp. 331-378).

185

Goldman, N. (1963). The Differential selection of juvenile offendors for court
W New York: National Council on Crime and Delinquency.

Gottfredson, SD. and Taylor, RB. (1986). Person-environment interactions in the

prediction of recidivism. In J .M. Byme and R.J. Sampson (Eds) The Social Ecology of
Crimo (pp. 133-155). New York: Springer-Verlag New York, Inc.

Goudy, W.J. (1982). Further consideration of indicators of community attachment.
Social Indicators Research, 11 (pp. 181-191).

Greenburg, SW. (1986). Fear and its relationship to crime, neighborhood deterioration,
and informal social control. In J .M. Byme and R.J. Sampson (Eds) The Social Ecology
of Crime (pp. 47-62). New York: Springer-Verlag New York, Inc.

Haeberle, SH. (1989). Planting the wsroots: Structuring citizen participation. New
York: Praeger.

Hepburn, J .R. (1978). Race and the decision to arrest: An Analysis of warrants issued.
Jogn_al of Research in Crime and Delinquency, 15 (pp. 54-73).

Herting, J .R. and Guest, A.M. (1985). Components of satisfaction with local areas in the
metropolis. The Sociological Merly, 26 (pp. 99-115).

Hickman, M.J., Piquero, A.R., and Greene, J .R. (2000). Does community policing
generate greater numbers and different types of citizen complaints than traditional

policing. Police Merly, 3 (pp. 70-84).

Hofmann, DA. and Gavin, MB. (1998). Centering decisions in hierarchical linear
models: Implications for research organizations. Journal of Mmement, 24 (pp. 623-
641).

Homant, R.J. and Kennedy, DB. (1985). Police perceptions of spouse abuse: A
Comparison of male and female ofﬁcers. Jomal of Qriminal Justice, 13 (9-47).

Hummon, D.M. (1992). Community attachment: Local sentiment and sense of place. In
I. Altman and S.M. Low (Eds) Place attachment (pp. 253-278). New York: Plenum
Press.

Hyman, H.H. and Wright, GR. (1971). Trends in voluntary association memberships of
American adults: Replication based on secondary analysis of national sample surveys.

' 3 i1 i Rviw36(pp. 191-206).

Jacobs, D. (1979). Inequality and police strength: Conﬂict theory and coercive control in
metropolitan areas. American Sociological Review. 44 (pp. 913-925).

186

Jacobs, D. and Britt, D. (1979). Inequality and police use of deadly force: An Empirical
assessment of a conﬂict hypothesis. Sooial Problems, 26 (pp. 403-412).

Julian, D.A., Reischl, T.M., Carrick, R.V., and Katrenich, C. (1997). Citizen
participation - lessons from a local United Way planning process. Journal of the

American Planning Association, 63 (pp. 345-355).

Kania, RE. and Mackey, W.C. (1977). Police violence as a function of community
characteristics. Criminology, 15 (pp. 27-48).

Kasarda, J .D. and Janowitz, M. (1974). Community attachment in mass society.
American Sociological Review, 39 (pp. 328-339).

Kelling, G.L. (1997). Crime control, the police, and culture wars: Broken windows and
cultural pluralism. In National Institute of Justice, Perspectives on Crime and Justice:
1997-1998 Lecture Series (pp. 1-28). Washington, D.C.: US. Department of Justice.

Kelly, J .G. (1988). A Guide to conducting prevention research in the community. New
York: The Haworth Press.

Kingry-Westergaard, C. and Kelly, J .G. (1990). A Contextualist epistemology for
ecological research. In P. Tolan, C. Keys, F. Chertok, and L. Jason (Eds) Researching
community psychology (pp. 23-31). Washington, D.C.: American Psychological
Association.

Kingston, 8., Mitchell, R., Florin, P., and Stevenson, J. (1999). Sense of community in

neighborhood as a multilevel construct. Journal of Qommunity Psychology, 27 (pp. 681-
694).

Klinger, DA. (1997). Negotiating order in patrol work: An Ecological theory of police
response to deviance. Criminology, 35 (pp. 277-306).

Kress, J .M. (1980). The Spatial ecology of criminal law. In D.E. Georges-Abeyie and
K.D. Harris (Eds) Crime: A Spatial perspgctive (pp. 58-71). New York: Columbia
University Press.

Kuo, F.E., Sullivan, W.C., Coley, R.L., and Brunson, L. (1998). Fertile ground for
community: inner-city neighborhood common spaces. American Journal of Community

Pachology, 26 (pp.823-851).

Langworthy, RH. (1985). Wilson’s theory of police behavior: A Replication of the
constraint theory. Jyjice Merly, 2 (pp. 89-98).

Lersch, KM. (1998). Predicting citizen race in allegations of misconduct against the
police. Jom of Criminal Justice, 26 (pp. 87-97).

187

Lersch, KM. and Mieczkowski, T. (1996). Who are the problem prone officers? An
Analysis of citizen complaints. American Journal of Polico, 15 (pp. 23-44).

Lester, D. (1982). Civilians who kill police ofﬁcers and officers who kill civilians: A
Comparison of American Cities. Journal of Police Science and Administration, 10 (pp.
384-387).

Lewin, K. (1935). Principles of tomlogjcal psychology. New York: McGraw-Hill.

Liska, A.E., and Chamlin, MB. (1984). Social structure and crime control among
macrosocial units. American Journal of Sociology, 90 (pp. 383-395).

Liska, A.E., Chamlin, M.B., and Reed, MD. (1985). Testing the economic production
and conﬂict model of crime control. Social Forces 64 (pp. 119-139).

 

Liska, A.E., Lawrence, J .J ., and Benson, M. (1981). Perspectives on the legal order: The
Capacity for social control. American Journal of Sociology, 87 (pp. 413-426).

Mastrofski, SD. (1983). The Police and noncrime services. In G.P. Whitaker and CD.

Phillips (Eds) Evaluating mrformang of criminal jm'ce ngenoies (pp. 33-61). Beverly
Hills, CA: Sage Publications.

Mastfrofski, S.D., Parks, R.B., Reiss, A.J. Jr., Worden, R.E., DeJong, C., Snipes, J .B.,
and Tenill, W. (1998). Systematic observation of public police: Applying ﬁeld research
methods to policy issues. National Institute of Justice, US. Department of Justice,
Washington, DC.

Mastrofski, S.D., Parks, R.B., Worden, R.E., and Reiss, A.J., Jr. (2002). Project on
policing neighborhoods in Indianapolis, Indiana, and St. Petersburg, Florida, 1996-1997,
Part 18 (Citizen survey data). US. Department of Justice, National Institute of Justice,
Data Resources Program.

Maxﬁeld, M.G., Lewis, D.A., and Szoc, R. (1980). Producing ofﬁcial crimes: Veriﬁed
crime reports as measures of police output. SocinJ Sciengp Qquterly, 61 (pp. 221-236).

McGahey, R. M. (1986). Economic conditions, neighborhood organization, and urban
crime. In A. J. Reiss and M. Tonry (Eds. ) Communities and crime (pp. 231-270).
Chicago: The University of Chicago Press.

McKenzie, RD. (1926). The Scope of human ecology. In E.W. Burgess (Ed) Ih_e
Urban communion Selected pamrs from the proceedings of the American Sociology

Society, Chicago: University of Chicago Press.
McPherson, J .M. (1981). A Dynamic model of voluntary afﬁliation. Sooial Forces, 59
(pp. 705-728).

188

Menard, SW. (1995). Applied logistic regmssion analysis. Sage series:
Quantitative applications in the social sciences, vol. 106. Thousand Oaks,

CA: Sage.

Mladenka, K.R. and Hill, K.Q. (1978). The Distribution of urban police services.
Journal of Politics, 49 (pp. 112-133).

Moos, RH. and Lemke, S. (1983). Assessing and improving social-ecological settings.
In E. Seidman (Ed) Handbook of Social Intervention (pp. 143-162). Beverly Hills, CA:
Sage Publications.

Morenoff, J.D., Sampson, R.J., and Raudenbush, SW. (2001). Neighborhood inequality,
collective efﬁcacy, and the spatial dynamics of urban violence. Criminology, 39 (pp.
517-553).

Norusis, M.J. (1990). SPSS / PC + Statistics 4.0 for the IBM PC/XT/AT and PS/2.
Chicago, IL: SPSS Inc.

Novak, K.J., Frank, J ., Smith, B.W., and Engel, RS. (2002). Revisiting the decision to
arrest: Comparing beat and community ofﬁcers. Crime and Delinquency, 48 (pp. 70-98).

Oropesa, RS. (1989). The Social and political foundations of effective neighborhood
improvement associations. Social Science Quarterly, 70 (pp. 723-743).

Ostrom, E. (1983). Equity in police services. In G.P. Whitaker and CD. Phillips (Eds)

Evaluating p_erformance of criminal justice agencies (pp. 99-125). Beverly Hills, CA:
Sage Publications.

Oxford, J. (1992). Commqu psychology: Theory and practice. New York: Wiley.

Parks, R.B., Mastrofski, S.D., DeJong, C., Gray, MK. (1999). How ofﬁcers spend their
time with the community. Justice Qngerly, 16 (pp. 483-518).

Perkins, D.D., Florin, P., Rich, R.C., Wandersman, A., and Chavis, D.M. (1990).
Participation and the social and physical environment of residential blocks: Crime and

commtmity context. American Journal of Community Psychology, 18 (pp. 83-115).

Prezza, M., Amici, M., Roberti, T., and Tedeschi, G. (1999). Sense of community
referred to the whole town: Its relations with neighboring, loneliness, life satisfaction, and

area of residence. Journal of Community Psychology, 29 (pp. 29-52).

Quinney, R. (1970). The Social reality of crime. Boston: Little, Brown and Company.
Reisig, MD. and Cancino, J.M. (2004). Incivilities in nonmetropolitan communities:

The Effects of structural constraints, social conditions, and crime. Journal of Criminal
W (PP- 1549)-

189

Reisig, MD. and Parks, RB. (2000). Experience, quality of life, and neighborhood
context: A Hierarchical analysis of satisfaction with police. Justice Quarterly, 17 (pp.
607-630).

Reiss, A.J., Jr. (1972). Police brutality. In L. Radzinowicz and ME. Wolfgang (Eds)
The Criminal in tho arms of the law, Volume 2. New York: Basic Books.

Riger, S. and Lavrakas, P.J. (1981). Community ties: Patterns of attachment and social

interaction in urban neighborhoods. American Journal of Community Psychology, 9 (pp.
55-66).

Riksheim, EC. and Chermak, S.M. (1993). Causes of police behavior revisited. Journal
of Criminal Justice 21 (pp. 353-382).

 

 

Rodgers, W. (1980). Residential satisfaction in relationship to size of place. Socig
Psychologioal Qu_art_erly, 43 (pp. 436-441).

Ross, CE. and Jang, SJ. (2000). Neighborhood disorder, fear, and mistrust: The
Buffering role of social ties with neighbors. Amerioan Joumnl of Communin;

Psychology, 28 (pp. 401—420).

Sampson, R.J. (1986). Neighborhood family structure and the risk of personal
victimization. In J .M. Byme and R.J. Sampson (Eds) Tho Social Ecology of Crime (pp.
25-46). New York: Springer-Verlag New York, Inc.

Sampson, R.J. (1988). Local ﬁiendship ties and community attachment in mass society:
A Multilevel systemic model. American Sociological Review, 53 (pp. 766-779).

Sampson, R.J. and Groves, W.B. (1989). Community structure and crime: Testing social
disorganization theory. American Journal of Sociology, 94 (pp. 774-802).

Sampson, R.J. and Jeglum-Bartusch, D. (1998). Legal cynicism and (subcultural?)
tolerance of deviance: the neighborhood context of racial differences. Law and Society
Review, 32 (pp. 777-804).

Sampson, R.J., Raudenbush, S.W., and Earls, F. (1997). Neighborhoods and violent
crime: A Multilevel study of collective efﬁcacy. Science 277 (pp. 918-924).

 

Sampson, R.J. and Raudenbush, S.W. (2001). Disorder in urban neighborhoods - does it
lead to crime? National Institute of Justice, Research in Brief. Washington, D.C.: US.
Department of Justice.

Shaw, CR. and McKay, H.D. (1942/1969). Juvenile dolinquency and urban areas; A

Stony of rates of delinquoncy in relation to differential characteristics of local
communities tn' American cities. Chicago: The University of Chicago Press.

190

Sherman, L.J. (1975). An Evaluation of policewomen on patrol in a suburban police
department. Journal of Police Science and Administration, 3 (pp. 434-438).

Sherman, L.W. (1980). Causes of police behavior: The Current state of quantitative
research. Journal of Research in Crime and Delinquency, 17 (pp. 69-100).

Sherman, L.W. (1986). Policing communities: What works? In A.J. Reiss and M. Tonry
(Eds) Comnnnn'ties an! Crime (pp. 343-3 87). Chicago: The University of Chicago

Press.

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

Shinn, M. (1990). Mixing and matching: levels of conceptualization, measurement, and
statistical analysis in community research. In P. Tolan, C. Keys, F. Chertok, and L. Jason

(Eds) Researching Commqu Psychology (pp. 111-126). Washington, D.C.: American
Psychological Association.

Shumaker, SA. and Hankin, J. (1984). The Bonds between people and their residential
environments: Theory and research. Population and Environmenn 7 (pp. 59-60).

Skogan, W.G. and Lurigio, A.J. (1992). The Correlates of community antidrug activism.
Crime and Delinquency, 38 (pp. 510-521).

Slovak, J .S. (1987). Police organization and policing environment: Case study of a
disjuncture. Sociological Focus, 20 (pp. 77-94).

Slovak, J .S. (1986). Sgles of tnban mlicing: Organiz_ation, environment, and mlice
$195 in selected American cities. New York: New York University Press.

Smith, D.A. (1987). Police response to interpersonal violence: Deﬁning parameters of
legal control. Social Forces, 65 (pp. 767-782).

Smith, D.A. (1986). The Neighborhood context of police behavior. In A.J. Reiss and M.

Tonry (Eds) Communities and crime (pp. 313-341). Chicago: The University of Chicago
Press.

Smith, D.A. (1984). The Organizational context of legal control. Criminology, 22
(pp.19-38).

Smith, D.A. and Klein, J .R. (1983). Police agency characteristics and arrest decisions.

In G.P. Whitaker and CD. Phillips (Eds) Evalnnting mrformanoo of criminﬂ jnstjce
ngenoieo (pp. 63-97). Beverly Hills, CA: Sage Publications.

Sparrow, M.K., Moore, M.H., and Kennedy, D.M. (1990). Beyond 911: A New era for
mlioing. New York: Basic Books.

191

St. John, C. Austin, D.M, and Baba, Y. (1986). The Question of community attachment
revisited. Smiological Smcm, 6 (pp. 411-431).

Stark, R. (1987). Deviant places: A Theory of the ecology of crime. Criminology, 25
(pp. 893-909).

Swanson, C. (1978). The Inﬂuence of organization and environment on arrest policies in

major US. cities. Policy Studies Journal, 7 (pp. 390-398).

Tabachnick, B.G. and Fidell, LS. (2001). Using multivariate statistics (4th Edition).
Boston: Allyn and Bacon.

Taylor, R.B, Gottﬁedson, SD, and Brower, S. (1985). Attachment to place:
Discriminant validity, and impacts of disorder and diversity. American Journal of

Community Psyohology, 13 (pp. 525-524).

Terrill, W. and Reisig, MD. (2003). Neighborhood context and police use of force.
Re e h in rim d Delin enc 40 (pp. 291-321).

Toch, H. (1996). The “Violence prone” officer. In W. Geller and H. Toch (Eds) A_no
Jnm'g fot; 111: Cnoognnding go Qntrolling mlice abm of fotoo. Washington, D.C.:

Police Executive Research Forum.

Trojanowicz, R. and Carter, D. (1988). The Philosophy and role of community policing.
Community Policing Series, No. 13. National Foot Patrol Center, School of Criminal
Justice, Michigan State University.

Turk, AT (1972). mum—(mien, Chicago: Rand McNally & Company.

Vold, GB. and Bernard, T.J. (1986). Theoretical criminology (3" Edition). New York:
Oxford University Press.

Waegel, W.B. (1984). The Use of lethal force by police: The Effect of statutory change.
Crime Ed Delinquency, 30 (pp. 121-140).

Wagner, A.E. (1980). Citizen complaints against the police: The Accused officer.
Jonrh_al of Police Scieng go Aginistration, 8 (pp. 373-377).

Walker, S. and Katz, C.M. The Police in America (4* edition). New York: McGraw-
Hill.

Warner, B.D. (1997). Community characteristics and the recording of crime: Police
recording of citizens’ complaints of burglary and assault. Justice Merly, 14 (pp. 631-
649).

192

Weiner, N.L. and Willie CV. (1971). Decisions by juvenile officers. American Journal
of Sociology, 77 (pp. 199-210).

Weitzer, R. (1999). Citizens’ perception of police misconduct: Race and neighborhood
context. Justice Quarterly, 16 (pp. 819-846).

Werthman, C. and Piliavin, I. (1967). Gang members and the police. In DJ. Bordua
(Ed) The Police: Six Sociological Essays (pp. 56-98). New York: John Wiley and Sons,
Inc.

White, M.J. (1985). Determinants of community satisfaction in Middletown. American
Journal of Communig Psychology, 13 (pp. 583-597).

Williams, K.R. and Drake, S. (1980). Social structure, crime and criminalization: An

Empirical examination of the conﬂict perspective. The Sociological Merly, 21 (pp.
563-575).

Wilson, J. Q. (1968). Varieties of police behavior. The Manngement of law and order 1
eigm Qmm mmunities. Cambridge, MA: Harvard University Press.

Wilson, J .Q. and Kelling, G.L. (1982). Broken windows: The Police and neighborhood
safety. Atlantic Monthly (March 1982) (pp.29-3 8).

Wirth, L. (1938). Urbanism as a way of life. The American Journal of Sociology, 44
(pp. 1-24).

Worden, RE. (1989). Situational and attitudinal explanations of police behavior: A

Theoretical reappraisal and empirical assessment. Law and Sociegt Review, 23 (pp. 667-
707).

193

\ﬂl'llllililllﬂlﬂllllll

02736 3401

      

1)")le