. 5%..er
A . . . . .w: ‘ O?» 3‘ ...
. > o I ..P V’Jh .
... .212... .
ﬁﬁvﬁxaaan. ﬂan:

Hi . ._

it»:

1a....

.
....

3 ‘F- 0,.
£51.... 1. ..

A
4.53 hiavﬂr .
..-..1. 2‘. a
. aﬁﬂqh .w
"ll.l

.31. Sips...
4:35.13 1. p u r
Inh.v.’~...x 1251 I it .

f.

V :
1.03.»?
..l.‘ I!”

 

 

 

 

95%.. §I§ #3.. .. .,

.53.; .o .

 

 

 

This is to certify that the
dissertation entitled

A MULTILEVEL TEST OF FEAR OF CRIME:

THE EFFECT OF SOCIAL CONDITIONS, PERCEIVED
COMMUNITY POLICING ACTIVITIES, AND PERCEIVED
RISKS OF VICTIMIZATION IN A MEGALOPOLIS

presented by

Eui-Gab Hwang

has been accepted towards fulﬁllment
of the requirements for the

PhD. degree in Criminal Justice

 

 

////%L./7mﬂ

Major Professor’ 5 Signature

ﬂ“: ;/ /64#.,2L7C%
/
Date

MSU is an Afﬁrmative Action/Equal Opportunity Institution

 

LIBRARY
Michigan State
L «ersity

 

 

 

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

 

 

—'v'

 

C7-

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/ClRC/DaleDue.indd-p.1

A MULTILEVEL TEST OF FEAR OF CRIME:
THE EFFECT OF SOCIAL CONDITIONS, PERCEIVED COMMUNITY POLICING
ACTIVITIES, AND PERCEIVED RISKS OF VICTIMIZATION IN A MEGALOPOLIS

By

Eui-Gab Hwang

A DISSERTATION

Submitted to
Michigan State University
in partial ﬁilﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
School of Criminal Justice

2006

ABSTRACT
A MULTILEVEL TEST OF FEAR OF CRIME:

THE EFFECT OF SOCIAL CONDITIONS, PERCEIVED COMMUNITY POLICING
ACTIVITIES, AND PERCEIVED RISKS OF VICTIMIZAT ION IN A MEGALOPOLIS
By
Eui-Gab Hwang

This study extends research on communities, crime, and quality of life beyond
Western contexts in the course of developing a causal model of fear of crime in a
megalopolis, Seoul, the ﬁﬁh largest city in the world. It investigated four different
conceptual models—the community context model, the community concern/control
model, the victimization model, and the disorder model—for directional accuracy and
ability to explain fear of crime. These conceptual factor models were tested in a causal
frame. Building upon F erraro’s risk interpretation frame, perceived risk was considered
as the major mediator between the conceptual predictors and fear, and behavioral
adaptations were hypothesized as the response to fear as well as perceived risk. The
community-context approach was extended to consider district conditions as macro—level
predictors. The research focus of the community concern/control approach was given to
the effect of citizen perceptions of community-policing activities on fear of crime.

To answer the research questions, this study combined census and crime statistics
with survey data of 654 citizens nested in 25 districts in Seoul, the capital city of South
Korea. Overall, results indicate that the community context model, the disorder model,
and the victimization model are applicable in the Seoul context, while the community

concern/control model, especially citizen perceptions of the formal social control

dimensions do not play an important role in determining fear of crime among Seoulites.

Several sets of Hierarchical Linear Models revealed that district-level economic
disadvantage, mobility, and crime were signiﬁcantly positive predictors of fear.
Unexpectedly, however, district-level population density was negatively associated with
fear, which could be based on the distinct process of city development that Seoul has
experienced. The results also revealed that citizen perceptions of formal social control
dimensions such as perceived community policing activities and satisfaction with police
were not inﬂuential. Interestingly, females expressed higher fear, more constrained
actions, and more cautious behaviors, but age and SES were not signiﬁcantly associated
with fear and constrained actions. Structurally, perceived risks played a substantial role
as a mediator between victimization and fear. In addition, behavioral adaptations were
well explained as the response to fear. Policy implications for city development and
community oriented policing are suggested, and the results are discussed in comparison

to previous research in Asia and the United States and other Western countries.

Copyright by
Eui-Gab Hwang

2006

ACKNOWLEDGEMENTS

This dissertation is the result of support from many individuals. Particular
gratitude is given to Dr. Edmund F. McGarrell, my advisor and chair of my guidance and
dissertation committee, whose conﬁdence, leadership, and thoughtful direction
inﬂuenced me to reach greater heights. He always knew in which stage I was and in what
direction I needed to go, and his crime and communities class provided me with the main
ideas of my dissertation. I am profoundly grateful to my committee. To Dr. Bruce L.
Benson, for his years of friendship and support. My ability to teach policing courses is a
debt to my experience as a TA for his outstanding class for three years. To Dr. Vincent J.
Hoffman, for his support and encouragement. I could cover issues on Korean and Asian
societies owing to his continuous encouragement. To Dr. Christopher D. Maxwell for his
support for my dissertation, job application, and crime analysis techniques. He was
willing to sacriﬁce his schedule to prioritize my requests for letters of reference and
dissertation defense meetings. To Dr. William G. J acoby, for his support and statistical
advice. His public opinion class and methodological insights were very helpful
throughout my dissertation.

I would like to thank other scholars who provided me with insight and advice. To
Dr. Kimberly S. Maier for allowing me to sit in her HLM class and for being available
for many questions. To Dr. Robyn R. Mace for providing me with great feedback for the
ﬁnal draft of my dissertation. Special acknowledgement goes to Dr. Min-Sik Lee and Dr.
Seong-Un Kim. Their pioneering studies in the United States and in Korea enabled me to

conduct this dissertation research.

My appreciation is expressed to my friends and colleagues. All along the way,
editorial assistance and advice were provided by Jason Ingram, Carol Zimmerrnann,
Susan Gade, and Mark VerNooy. Jason, Carol, Susan, and Mark always understood my
situation as an international student and were willing to assist me in many ways. To my
supportive Korean friends, J inseong Cheong, Dae-Hoon Kwak, Suyeon Park, Mirang
Park, and Seok-J in Jeong for their close ﬁiendship and support. Special thanks to Mirang
for her assistance for data collection. To my colleagues in the Korea National Police
Agency for assisting me to collect census and crime statistics. I could design the
multilevel study with their assistance.

Most importantly, my sincere thanks go to my precious family. To my mother
and parents in law for their positive encouragement and support. To Yun, Jeeseon, and
J oonyeon for their faithful love, encouragement, and support. You are truly God’s gift to

me.

vi

TABLE OF CONTENTS
LIST OF TABLES -------------------------------------------------
LIST OF FIGURES ------------------------------------------------

CHAPTER 1

INTRODUCTION -------------------------------------------------
1. Statement of the Problem
2. Importance of the Study beyond U.S. Context
3. Organization of Dissertation

CHAPTER 2

SOCIAL CONTROL AND POLICING AMONG KORE ANS ..............
1. PhiIOSOphy 0f Social Control ------------------------------------
2. Modern Aspects ----------------------------------------------
3. Policing in Korea .............................................
4. Crime and Fear of Crime .......................................
5. Applicability of Models of Fear of Crime --------------------------

CHAPTER 3

COMMUNITY POLICING AND FEAR OF CRIME .....................
1. Reform Models of Policing .....................................
2. Community Policing ------------------------------------------
3. Community Policing and Fear of Crime ...........................
4. Impact of Community Policing on Fear of Crime ....................

CHAPTER 4
REVIEW OF THEORETICAL MODELS OF FEAR OF CRIME -----------
1. Concepts of Fear of Crime ......................................
(1) Concepts -------------------------------------------------
(2) Standard Measures -----------------------------------------
2. Models of Fear of Crime .......................................
(1) Personal Factors: The Victimization Model ......................
Victimization -------------------------------------------
Vulnerability ............................................
Indirect Experience of Crime -------------------------------
(2) Community Based Factors of Fear of Crime .....................
Theory in Brief ------------------------------------------
The Disorder Model ......................................
The Community Concem/Control Model ----------------------
The Subcultural Diversity Model ............................
The Community Context Model -----------------------------
(3) Implications for Research ------------------------------------

vii

xi

Hui—bud

12
12
16
18
25
29

36
36
4O
44
49

52
53
53
59
61
62 .
63
65
73
77
77
78
82
87
89
94

CHAPTER 5

METHODOLOGY ------------------------------------------------- 102

1. Hypotheses -------------------------------------------------- 103

(1) Social Conditions and the Community Context Model ............. 103

(2) Community Policing and the Community Concern/Control Model- - - - 104

(3) The Victimization Model ------------------------------------ 104

(4) The Disorder Model ---------------------------------------- 105

(5) Perceived Risk, Fear, and Behavioral Adaptations ---------------- 106

2. Data -------------------------------------------------------- 107

(1) Survey Data --------------------------------------------- 107

(2) Census and Crime Statistics ---------------------------------- 111

3. Measurement in the Study --------------------------------------- 112

( 1) Endogenous Variables --------------------------------------- 112

(2) Exogenous Variables ---------------------------------------- 117

(3) Control Variables ------------------------------------------- 12 5

4. Analytic Strategy ---------------------------------------------- 127

5. Hierarchical Linear Modeling (HLM) Method ----------------------- 128

(1) HLM and Traditional Approach ------------------------------- 128

(2) Data Structure for HLM ------------------------------------- 129

(3) Traditional Approach --------------------------------------- 130

(4) Advantages of Multilevel Analysis ----------------------------- 1 3 3
CHAPTER 6

FINDINGS ------------------------------------------------------- 137

1. Preliminary Statistics ........................................... 137

(1) General Characteristics -------------------------------------- 137

(2) Bivariate Correlation ---------------------------------------- 140

Citizen Level Correlation ---------------------------------- 140

District Level Correlation ---------------------------------- 144

2. Hierarehieal Linear Modeling Analyses ---------------------------- 145

(l) MOdel Building Procedure ----------------------------------- 145

Unconditional (one-way AN OVA) Model --------------------- 145

Random Coefﬁcient Regression Model ----------------------- 147

Fixed Effect Model (Full Model) ---------------------------- 150

(2) ReSUItS 0f Analyses ----------------------------------------- 15 3

Fear Of Crime ----------------------------------------- 153

Perceived Risk and Fear of Crime --------------------------- 157

Perceived RlSk, Fear, and Behavioral Adaptations -------------- 161

(3) Summary Of Findings --------------------------------------- 165
CHAPTER 7

DISCUSSION AND CONCLUSION

1. Discussion --------------------------------------------------- 169

(1) District Level Conditions and the Community Context Model -------- 169

(2) Community Policing and the Community Concern/Control Model - - - - 172

(3) The Disorder Model ---------------------------------------- 176

viii

(4) The Victimization Model .................................... 177

(5) Perceived Risk, Fear, and Behavioral Adaptations ----------------- 181

(6) Cross Level Interaction -------------------------------------- 182

2. Conclusion and Policy Implications ------------------------------- 184

3. Limitations and Future Research ---------------------------------- 188
ENDNOTES ------------------------------------------------------ 194
BIBLIOGRAPHY -------------------------------------------------- 198

LIST OF TABLES

Table 1: Administrative Districts, Population, and Sample Size in Seoul ----- 109
Table 2: Description of Variables ------------------------------------- 114
Table 3: Rotated Factor Matrix of Behavioral Adaptations ---------------- 116
Table 4: Factor Pattern for Social Condition Variables (N = 25) ------------- 118
Table 5: Individual Level Descriptive Statistics (N=654 in 25 Districts) ------- 138
Table 6: District Level Descriptive Statistics (25 Districts) ----------------- 140
Table 7: Bivariate Correlations between Fear and Citizen Level Variables ----- 141
Table 8: Bivariate Correlations between Fear and District Level Variables - - - ' 144
Table 9: Decomposition of Variance and District Level Reliabilities --------- 147
Table 10: Hierarchical Linear Models for Fear of Crime ------------------ 154
Table 11: Hierarchical Linear Models for Perceived Risk and Fear of Crime- - - 159

Table 12: Hierarchical Linear Models for Behavioral Adaptations °°°°°°°°°°° 162

LIST OF FIGURES

Figure 1: Risk Interpretation Model (Ferraro, 1995) ......................

Figure 2: A Revised Causal Model -----------------------------------

Figure 3: POpulation Density in Seoul --------------------------------

Figure 4: Population living in Poverty in Seoul ..........................

Figure 5: A Path Model of Fear of Crime (signiﬁcant effects only) ----------

98

100

108

108

166

CHAPTER ONE
INTRODUCTION

1. Statement of the Problem

Quality of life can be measured both subjectively and objectively. Fear of crime
ﬁts into the broad framework of subjective experiences, compared to objective indicators
(e. g., economic levels). Since the President’s Crime Commission Report in 1967 which
suggested that a signiﬁcant proportion of the population experiences crime-related fears
on a regular basis (see Hale, 1996), fear of crime has become a serious issue for scholars
and practitioners in the United States and in Eastern Asian countries such as Korea (Lee,
1997), China (Curran & Cook, 1993), and Japan (Ito, 1993). Despite the decreasing
crime trend in recent decades, fear of crime in major US. cities remains consistently high
(Lamon, 2000; Skogan, 1990) and crime and fear of crime shows consistently increasing
trends in Eastern Asian countries (Ito, 1993; J00, 2003; Lee, 1997). Taylor and Hale
(1986) argued that increasing crime trends raised fear, yet fear did not necessarily
decrease when crime declined. This suggests that the negative impact of crime endures
once people sense crime prevalence. According to two public opinion polls conducted by
the National Law Journal, only a third of the American public was “truly desperate”
about crime in 1989 but more than sixty percent was in 1994 (Haghighi & Sorensen,
1996; Sherman, 1994), an indication that concern for safety and fear of crime has
increased dramatically in recent decades. This situation is more so in Eastern Asian
countries due to rapid urbanization that generates large social disorganization in urban
communities. Cross-national comparison of ofﬁcial statistics showed that major crimes

were substantially increasing in Korea and Japan (J 00, 2003), contrary to their decreasing

trends in the United States, which raised social concern for increased fear of crime among
Koreans.

Following a Durkheimian notion of the role of fear for social reinforcement, there
is a certain positive aspect of fear when it motivates citizens to shoulder some of the
burdens of crime control and this leads to social solidarity by reinforcing the normative
order (Garofalo, 1981; Moore & Trojanowicz, 1988). Yet, as Conklin (1975) notes, a
general impact of fear is that it produces a series of undesirable social outcomes such as
mutual distrust, decreased social interaction, and withdrawal of support of public and
private activities such as crime control. Quality of life would not be high when people
are wonied that a stranger might hurt them on the street or at home. People in such
situations may suffer from psychological discomfort and withdraw from community
activities (Skogan & Maxﬁeld, 1981). The National Crime Survey (NCS) data in the
United States indicated that a majority of people limited their activities because of crime
concerns (Garofalo, 1979), suggesting that restriction of behavior in relation to fear is a
social problem. People may avoid certain routes, places, or transportation when they are
afraid of crime (Box, Hale, & Andrews, 1988). This situation is also applicable in the
Eastern Asian context due to the increasing trend of crime and fear of crime in recent
decades (J00, 2003).

More importantly, the impact of fear of crime on community life was so
signiﬁcant among people who perceived a high personal vulnerability (i.e., females or the
elderly) that they did not walk, but instead drove, took an escort, avoided certain places,
and carried something to protect themselves (Skogan & Maxﬁeld, 1981). Such a passive

life pattern among community members further deteriorated the community by

decreasing liveliness and integration (Kelling & Coles, 1996; Moore & Trojanowicz,
1988). Many scholars have singled out the negative aspect of fear as an interrupter of the
quality of community life and informal social control, which is eventually responsible for
community decline (Markowitz, Bellair, Liska, & Liu, 2001; Moore & Trojanowicz,
1988; Skogan & Maxﬁeld, 1981; Taylor & Hale, 1986). In response, studies of fear of
crime have become one of the main concerns of criminal justice academics and policy
makers in recent years.

Extant literature on fear of crime has identiﬁed distinct frameworks or models.
These models include the victimization model, the disorder model, the community
concern/control model, the subculture diversity model, and the community context model
(DuBow, McCabe, & Kaplan, 1979; Hale, 1996; Lane & Meeker, 2000; Taylor & Hale,
1986), which will be addressed in depth in Chapter 4. These models recognize direct and
indirect victimization, individual characteristics, perceptions of disorder, informal and
formal social control dimensions in the neighborhood, and community environments as
factors of fear of crime. While the victimization model and the disorder model have
often been tested, relatively less attention has been given to the community context
approach and the community concern/control model. In addition, even though these
conceptual models have been tested independently, little research exists to investigate the
effect of these models simultaneously in a comprehensive causal frame. Even more rare
is the literature regarding the effect of these conceptual models on fear of crime in non-
U.S. settings.

The social contextual approach or the community context model addresses the

role of neighborhood conditions on fear of crime (Cancino, 2002; Ross, Mirowsky, &

Pribesh, 2001). This social contextual approach emphasizes the importance of conditions
of social units, in addition to the individual characteristics and perceptions of
neighborhood quality of life, in explaining social disorganization that fosters crime and
delinquency (Rountree & Land, 1996; Sampson & Groves, 1989), and that reduces
citizen perceptions of security in communities (Skogan, 1990). Researchers have given
attention to small units defending their choices on the basis of within-group homogeneity
in street blocks, block groups, or census tracts that are similar on various census
indicators (Wooldredge, 2002). Yet, the aspects of the social contextual approach
deserve further testing in larger units. Since district or city-level environments are also
an important part of the social context, there should be a certain amount of variation in
behavior and perceptions that is explained by district or city-level conditions, in addition
to neighborhood-level environments. As community environment is important so would
be city or district level environment in explaining fear of crime. It is predicted that, in
general, citizens living in a relatively disadvantaged and high crime city or district, would
be more fearful of crime than those in a well organized and afﬂuent city or district,
controlling for community conditions, individual perceptions, and demographics.
Formulating models to identify city or district-level predictors as well as neighborhood or
community level conditions would help develop theories and policies for social control
and safety enhancement in districts or cities.

In addition, in the community concem/control model, community concern for
public social control as well as parochial and private level social control deserves special
attention especially in relation to the police-initiated endeavors for community safety.

The popular debate over the role of community policing in fear reduction could be

addressed in this regard. In addition to informal social control dimensions such as
community cohesion and collective efﬁcacy, citizen perceptions of public social control
endeavors and conﬁdence in public social control agents are among the predictors of fear
of crime in this model (Box et al., 1988; McGarrell et al., 1997; Velez, 2001).
Community-policing strategies and studies emphasize that residents can address
problems of crime and fear of crime by securing ties to public ofﬁcials and the police
(Skogan & Hartnett, 1997; Velez, 2001), proposing a substantial role of police activities
in communities to reduce fear of crime. The effect of community policing on fear of
crime showed mixed results in the United States (Zhao, Schider, & Thurman, 2002). The
western ideology of community policing was introduced as one of the most important
reform programs in Korea, while its impact on crime or fear of crime has not been
studied empirically in this cultural context. Considering the high priority of police work
on community safety and quality of life, it would be helpful to empirically identify the
role of community - policing activities on fear of crime especially in those countries that
recently introduced Western forms of community policing.

Scholars have attempted to identify and test a causal frame to investigate the
effects of multiple predictors on fear of crime. Early research explained the causal link
between a position in social space (i.e., different life style, vocation, leisure), fear, and
consequences of fear (Garoralo, 1981). Others modeled multiple predictors (e. g.,
personal and household vulnerability, knowledge of victims) that are structurally related
to fear and in turn, responsive behaviors (Skogan & Maxﬁeld, 1981). Criticizing these
early causal frames due to their lack of testability, Taylor and Hale (1986) designed three

separate causal models; the victimization, the disorder, and the community concern

models. In an attempt to link these models to test them simultaneously, Ferraro (1995)
proposed one of the most comprehensive models in a testable format, but this model did
not consider all predictors in theoretical frames. In addition, based on the reciprocal
relationship between fear and constrained behavior, this model considered behavioral
adaptations (i.e., constrained actions) as a predictor of, rather than a response to, fear. It
is interesting to investigate how fear inﬂuences different types of behavioral adaptations
(i.e., constrained actions, cautious actions, active defense), and to assess the ability of the
causal frame in explaining fear after relocating all predictors based on conceptual or
theoretical perspectives.

Finally, it is unclear if these fear of crime models are applicable in culturally
different contexts due to the lack of empirical research beyond Western countries (some
exceptions in Eastern Asian context include Curran & Cook, 1993; Ito, 1993; Lee, 1997;
Lee, 1998). A study in a comparative context is beneﬁcial to the ﬁeld by extending the
range of variation in crucial variables, testing the universality or robustness of initial
ﬁndings, and exploring the boundaries of generalizability (Kohn, 1987). Even with these
beneﬁts, there exists little research for testing these theoretical models and the effect of
social conditions and the effectiveness of police practice on fear of crime in Eastern
Asian contexts. This study intends to ﬁll this void by using a sample of citizens in Seoul,
the ﬁfth largest city in the world. A cross-national study is a challenge for scholars due
to the difﬁculties of establishing both linguistic and conceptual equivalence in different
countries (Shin, Putkowski, & Park, 2003). A test of Western theories and conceptual
models in a culturally different context may help overcome some of these methodological

difﬁculties and provide reliable results for comparative understanding of predictors and

levels of fear of crime. The value of conducting such a study is discussed in the next

section.

2. Importance of the Study beyond U.S. Context

A study in a comparative context provides a useful method for generating, testing,
and further developing theory as well as observing and explaining curious phenomena
(Kohn, 1987), through comparative insight into historical, cultural, political, or economic
idiosyncrasies. Such a study allows for exploring diversity, interpreting cultural or
historical signiﬁcance, and advancing theory (Ragin, 1994). Researchers look at how
social structures affect people’s lives by examining which of the many historical, cultural,
political, or economic differences in different countries are relevant for explaining some
unique patterns of theoretical issues, while they also extend the scope of knowledge by
ﬁnding general patterns of criminological phenomena (Kohn, 1987). Fear of crime and
the risk of criminal victimization are two types of phenomena that vary from country to
country. Nieuwbeerta (2002) explained this disparity based upon the differences in
individual life patterns, social and geographical constructs, and the practice of law
enforcement policies. That is, life style and routine activities, social or cultural
conditions, and the styles of policing in various nations affect crime rates and citizens’
perceptions of safety. International comparative studies suggest that different countries
share some aspects of fear and victimization, yet unique features exist. For example,
people are less likely to avoid speciﬁc locations due to fear of crime in Asian countries
than are citizens in Western countries (Vanderveen, 2000). Research on diverse ethnic
groups in the United States illustrates distinct aspects of victimization, where Asian-
Americans have consistently been ranked amongst the lowest in victimization rates

compared to other ethnic groups (Poole & Pogrebin, 1990), due to their emphasis on

group solidarity and informal social control. Moreover, unlike other countries that have
experienced rapid modernization and industrialization, in general, Asian-Paciﬁc nations
had relatively moderate increases in delinquency and crime rates (Curran & Cook, 1993).

Perceptions of law enforcement agencies as well as organizational features also
differ according to cultural settings and environments. Compared to other immigrants,
“Korean American immigrants distrusted and avoided the police because of their
negative conceptions of and experiences with the police in their home country” (Poole &
Pogrebin, 1990, p.90). Research on victimization, fear of crime, and conﬁdence in police
suggest that multiple cultural dimensions that inﬂuence the aforementioned conceptions
may exist. Cultural differences in the presence of social controls, routine activities, or
other phenomena may offer some explanatory value. Therefore, the issues surrounding
fear of crime, features of social control, and other predictors are worthy of further
discussion in the context of culture. Such a comparative approach would contribute to
the existing knowledge base regarding fear of crime and victimization from a global
perspective.

Societal differences in regard to the presence of social controls, cultural, or
philosophical dynamics in South Korea may offer some comparative value. Compared to
the United States, South Korea is homogenous—composed of one race, one language,
and represented by one set of collective norms (Lee, 2003). This country was under the
inﬂuence of Confucianism and Buddhism for several centuries, sharing this cultural
tradition with China and Japan. Unlike the Western emphasis on values of autonomy,
individualism, freedom from authority, and the presence of the strong public social

control system (Bellah, Madsen, Sullivan, Swidler, & Tipton, 1985; Cao & Hou, 2001),

collectivism, respect of authority, the informal social controls of mutual sanctioning and
shaming among residents characterize this cultural context (Asakura, 1998; Cao, Stack, &
Sun, 1998; Kim, 1998). That is, social controls in Korean society were largely based on
the constructs of moral socialization with origins of Confucianism. Conformity was
produced through moral socialization (Yang, 2003), not via the deterrent effect of law.
Confucian tradition still enjoys a dominant status among Koreans due to its lasting
ideological tradition (Kim, 1998), despite the strong inﬂuence of the Christian right,
Western ideologies and lifestyles, and perspectives or value systems of individualism.

In addition, South Korea is one of the most rapidly urbanizing countries,
recognized as one of the four dragon states 1 in Asia in addition to Hong Kong, Singapore
and Taiwan. Having emerged as an industrial giant in the second half of the twentieth
century, South Korea is often referred to as the economic “Miracle of the Han-River”
(Korea National Statistical Ofﬁce, 2004). South Korea serves as an important research
base to study the effect of social conditions and police practice on fear of crime, due to
the rapid urbanization, increasing indicators of social disorganization, and consequently
soaring crime in urban communities (J 00, 2003).

Moreover, the historical image of the South Korean police as frontline
representatives of government authority raises noteworthy comparative issues regarding
the effect of police practice (i.e., community policing) on fear of crime. Whereas the
majority of citizens in western democracies may View the police from a functional
perspective as the agency responsible for crime control, order maintenance, and citizen
safety (Lab & Das, 2003; Sims, Hooper, & Peterson, 2002), this may be less accurate in

developing nations. The distinct historical, political, and cultural traditions of many

developing countries, particularly those that experienced colonization and coup d'état,
may portray police as controlled by the central government, serving the interests of the
political elite (Lab & Das, 2003). South Korea seems to ﬁt this pattern. The perception
of the South Korean police force was largely shaped by its oppressive role at the time of
Japanese colonization and by military coup (Pyo, 2001). The Korean War and
subsequent regional division in the nation, as well as constant threats from the communist
North, created a preoccupation with issues of national security (Kim, 2003; Lee, 2004).
Accordingly, citizens often view the highly centralized South Korean police force as an
agent of government bureaucracy. Due to the historical understanding of the role of
police as a government arm, citizens may perceive it difﬁcult to link police activities for
community safety such as community policing to their perceptions of the quality of life.

In sum, the dimensions of effective facilitators of the quality of life vary from
society to society due to the uniqueness in the process of developing philosophies and
theories of social control. For example, informal social controls in Asian countries are
based on moral socialization that emphasizes positive values and ethos. This is in sharp
contrast to the emphasis on individualism, independence, and reliance on formal social
controls in Western countries. Globalization, however, inﬂuences the norms and values
of different cultures. Recent interest in reintegrative shaming, restorative justice, strong
families, and community organization and collective efﬁcacy in contemporary Western
societies, for example, are equivalent to the Asian traditions of community based positive
socialization and informal social controls. Numerous reform endeavors for criminal
justice agents in Asian societies, on the other hand, popularized the programs of

community oriented policing in these countries, which represents their efforts to learn

10

from the Western emphasis on the effective utilization of formal social control agents. It
is also the case, however, that unique traditions still enjoy a dominant status in each
country due to their lasting ideology, despite the strong inﬂuence of different cultures and
lifestyles. It is believed that this empirical research focusing on the effects of social
conditions and community-policing activities on fear of crime in the Seoul context will
enable investigations into the cultural aspects inherent to Eastern Asian countries and into
the situations of the rapidly developing countries.
3. Organization of Dissertation

The remainder of the current dissertation is organized into six chapters. Chapter
two focuses on social control, both informal and formal, in the context of Korea.
Applicability of theoretical perspectives developed and tested in the United States to
Koreans is also discussed. Chapter three then addresses the connection between fear of
crime and community policing. Chapter four reviews the models of fear of crime
developed primarily in the United States. This chapter also considers various deﬁnitions
and measures of fear of crime. Chapter ﬁve speciﬁes the hypotheses to be tested, the
methodological design, and identiﬁes and defends the measurements of the independent
and dependent variables. Hierarchical Linear Modeling (HLM) is also discussed in this
chapter as the major analytic strategy. The major research ﬁndings are reported in
chapter six, and the implications of the results are discussed in both theoretical and

practical terms in the ﬁnal chapter.

11

CHAPTER TWO
SOCIAL CONTROL AND POLICING AMONG KOREANS

1. Philosophy of Social Control

Every society induces or forces people to comply with social norms through
formal and informal measures to maintain social order. In ancient East Asian society,
moralist and legalist philosophers laid the foundation for societal controls with emphasis
on the moralist approach of informal social controls based upon Confucian philosophy
(Chen, 2002; Westermann & Burfeind, 1991; Young, 1942; Yusuf, 2003). Their
emphasis on informal sanctioning, shame, and community level social control was
different from legalists’ emphasis on the evil nature of human being that should be
controlled by strict rules of law, equivalent to modern public social control via the
criminal justice system. Confucius’s ideology for social control was represented in his
famous teaching for governors (J iao, 2004). He explained that if the rulers lead people
with government measures and regulate them by law and punishment, they may avoid
wrongdoings, but they may have no sense of honor and shame, and such a method of
regulation takes people away from rulers. On the other hand, if the rulers lead them with
virtue and regulate them by the rules of propriety of established morality, citizens obtain
a sense of shame and set themselves right and come closer to the rulers.

This moralist tradition based on Confucian principles has been accepted as
ofﬁcial philosophy in most Eastern Asian countries, including Korea, Japan, and China,
and has dominated Eastern Asian culture for centuries. Also stemming from
Confucianism tradition, law was perceived either as the second measure to be considered
only when a person could no longer be cured by moral education or as the last line of

defense for maintaining social order when a person does extremely immoral or serious

harm to the society (J iao, 2004). Informal resolution of conﬂict and reliance on group or
communal pressures in the service of conformity and compromise were valued over
ofﬁcial regulation, since law was viewed as secondary to the collective moral principles
of Eastern Asian countries under the inﬂuence of Confucianism (J iao, 2004; Li, 2003;
Yang, 1961). Confucius believed that the society could be maintained by establishing a
moral conduct of respectability toward others and social orders (Wong, 2001; Yang,
1961). Therefore, traditional Confucian texts consist of colorful descriptions of how to
be a dutiful child, an ideal wife, a loyal ofﬁcial, and even a benevolent ruler (Chen, 2002).
It also emphasized reinforcement of the overall hierarchy in society by teaching people to
recognize their position and role in family, community, and society. Ceremonial respect
for ancestors was extended to parents, husbands, teachers, elderly, and public ofﬁcials.
People were encouraged to moderate their personal emotions, evade self-interests, and
develop a sense of obligation to the family or group (J iao, 2004).

The long historical process of Confucianization shaped the thinking of Koreans
during the ﬁve centuries of the Choson dynasty (A.D.1392—1910), and it was
reproduced in Korea’s social and cultural transformations. This philosophy served as the
basis for governing the nation and integrating society (Suenari, 1998). Unlike modem
music ringing with dominant themes of personal emotions, ancient songs and poems were
about the pleasure of serving the king, parents, and husbands. Until the end of the
Chosun dynasty, songs and poems that did not hold these themes were considered lower
quality works, representing a dominant cultural focus of society. Yoo (1998) recognized
the key to the Korean Confucian system was the family. The elders of centric

Confucianism educated their youth to conform to authority, elderly, and rules. The

family and group base of social support and control is still a dominant aspect of society in
South Korea. Today, in everyday practice, Confucianism takes the form of etiquette and
courtesy in interpersonal relations, unlike the emphasis of ceremonial worshiping from
the past (Asakura, 1998). The cultural values of self-control and shame avoidance were
linked to Confucian philosophy since shame or success was linked to the entire family
(Suenari, 1998). Therefore, social control and social support are the result of the family,
and community as the extension of it, even in the present day, leaving government
sanction as a secondary measure.

Buddhism 2 provides another tradition of social control. According to Buddhist
philosophy, one can control the mind and its contents by practicing sound mental habits,
such as abandoning corrupt thoughts or impulse, recognizing them, and letting them
dissolve naturally (Kim, 2003; Li, 2003; Yang, 1961). Buddha recommends meditation,
or the emptying of the mind, as a way not only to cleanse unhealthy thoughts such as
envy, violence, greed, or sexual impulse but also mental unease resulting from difﬁcult
situations or low social class. Furthermore, the belief of predetermined fate was
foremost.

“Everything is interconnected, and not only does the development of the child

inﬂuence the formation of the adult within one lifetime, but what happens to one

within this lifetime is due to one’s behaviors and actions in one’s past life; and
similarly one’s behaviors and actions in this life dictate what will happen to one in

the next” (Kalupahana, 1976; Li, 2003, p.20).

Buddhism purports that one’s current status is the result of previous conduct from a past
life, and one’s present behaviors and actions have profound consequences in the future.

This ideology functions as a social control, since individuals who follow these teachings

behave well in pursuit of the next life. The philosophy also contains profound moral

14

dimensions, such as not killing insects, because everything is interconnected and one will
pay for one’s misdeeds.

This religion played an important role in resolution of conﬂict through emphasis
on conformance to status through persuasion under the guise that the actions and
practices of today will inﬂuence tomorrow. Like Confucian ethics, Buddhism’s inﬂuence
has been a dominant philosophy among Koreans. Buddhists’ determinism is still
widespread in Korean society. Although younger generations have grown accustomed to
Western materialism and individualism, they generally follow previous practices, such as
visiting a fortuneteller to determine if their partner is the right person for marriage. It is
also custom for Koreans to visit their ancestor’s tombs for remembrance and good
fortune. In addition, on Lunar New Year’s Day, many Koreans visit websites or
fortunetellers to seek their destiny of the upcoming new year. Many of these customs
resulted from the philosophies of Buddhism and Confucianism. Some may argue that
Koreans could be considered potential Buddhists as well as latent Confucians due to the
inﬂuences of these philosophies on every day life, despite the strong presence of
Christianity and Western individualism.

Kim (2003) noted that today’s Confucianism survives largely as a set of social
ethics emphasizing the importance of the family, education, loyalty, hierarchy, and
propriety. Yang (2003) argues in this regard that Koreans cannot expect an autonomous
or equal relationship between members and social relationships due to the strict
hierarchical understanding of the social structure. In general, Korean youth, students and
employees respect their elders, teachers, and employers regardless of their situation.

Above all, the family has remained the dominating force of social controls while the

15

community structure, as the extension of the family, also provides social support or
sanctioning of citizens. In contrast to the western ideology of freedom from authority,
cultural tradition in Korea has long emphasized the value of conforming to authority and
order, and avoiding conﬂict under the inﬂuence of Confucian moral codes (J 00, 2003).
2. Modern Aspects

The disruption of basic norms and structures of Confucianism and religious
beliefs based on Buddhism occurred during Japanese colonization (1910-1945) and the
Korean War (1950-1953). Industrialization following social upheavals invited Western
norms and philosophies into Korean society. Korea began its journey from extraordinary
emphasis on economic development; during the 19603, the country achieved substantial
economic growth, noted as the miracle of the Han River; during the 19705, the country
increased production and development (Asakura, 1998). Industrialization was
accelerated by the community movement, known as “Saemaul Undong,” which regulated
shamans and other rituals while encouraging modern lifestyles and customs (Hidemura,
1998). The New Community Movement “Saemaul Undong” in Korea has been
considered a successful example of a community development campaign in a developing
country (Ito, 1998). The inﬂux of people into cities from different regions has
contributed to social disorganization, since people were severed from the intimate
familial relationships and communal networks. South Korean economic development has
displayed some modern features including rapid industrialization involving the
progression from an agriculturally based economy to a diversiﬁed industrial base
including heavy and chemical industry, electronics, and engineering and high technology

products (Moran, 1998). This process of industrialization increased the number of

16

nuclear families in urban areas, since many citizens moved away from the intimate social
ties with parents and community relatives (Hidemura, 1998). The socioeconomic gap in
urban societies generated enmity amongst Koreans that was seldom experienced in
traditional societies.

In addition, the generation gap widened in recent Korean society as the older
generations who subscribed to Confucian norms, could not tolerate women or children
smoking in public, students and youngsters antagonizing teachers or elders, aspects
which developed as a result of the complexity and anonymity of urban cities coupled with
the inﬂux of the Western individualism and freedom from authorities (Inglehart, 1997).
The effect of Christianity on Koreans’ social lives was also important. Hidemura (1998)
noticed that Christian churches offered substitute contexts to build new social networks
especially in an urban community after Korean urbanites lost strong ties upon abandoning
their hometown. Christianity-based Western values offered more equality and a sense of
importance that Koreans could not ﬁnd in the Confucian system (Lee, 2003). In addition,
expansion of higher education was tremendous. In just eleven years (1971-1982), the
number of students enrolled in four year universities more than quadrupled in Korea
(Asakura, 1998). Based on extreme emphasis on educational success in Korea and other
Eastern Asian societies, it is not surprising that students left behind become a major
source of delinquency (Curran & Cook, 1993; Kim & Lee, 2003; Takahashi & Chang,
1985). The strain associated with failure is profound since only some of the students who
take the admissions test are granted entrance to prestigious colleges in these countries
(Curran & Cook, 1993; Kim & Lee, 2003). Chinese scholars realized that “students in

such a situation lose care for good deeds as well as trust of educators, some of them turn

17

from single poomess in studies turn into double poomess of both in studies and in moral
character, and furthered into law violators” (Curran & Cook, 1993, p.304; Zhou & Xia,
1988, p.4). This explanation is duly applicable in South Korea since the extreme strain
for higher education is one of the biggest social issues in the country.
3. Policing in Korea

In contrast to the western principles of individualism and dependence on formal
social controls, social control has traditionally relied upon community-based informal
social control mechanisms in South Korea. Due to the long cultural tradition of moral
governance and a relatively short period of development of the nation’s modern legal
system, Korea has established a criminal justice system that coexists with the informal
social control system. The modern legal system recognizes the importance of rule of law,
due process, protection of individual rights, and judicial independence and all police
ofﬁcers are required to possess a complete understanding of laws and procedures
pertinent to police work through education and training (Yoon, 1998). However, whereas
the majority of citizens in Western democracies may view the police from a functional
perspective as the agency responsible for crime control, order maintenance, and citizen
safety (Lab & Das, 2003; Sims et al., 2002), this may be less true in developing nations
like South Korea. The distinct historical, political, and cultural traditions of many
developing countries, particularly those that have experienced colonization and military
coup, may lead to images of the police as controlled by the central government and
intended to serve the interests of the political elite (Lab & Das, 2003; Hwang, McGarrell,

& Benson, 2005).

18

A national modern police system was established in South Korea in 1894, when
Japan and Western nations compelled Korea to open her borders. Prior to this, there was
no national police organization and no clear distinction between the police and the
military under the feudal system. Japan maintained and utilized extremely centralized,
nationalistic, and a pyramid style police system to control Koreans and help enforce
colonization during its occupation from 1910 tol945 (Park, 1988). Police substations
throughout every corner of the country were the mainstream for the surveillance of
citizens and for the deterrence of anti-Japanese movement (Park, 1988). Koreans
perceived the police at this time as the frontline enforcer of colonization. After World
War II and Korea’s independence from Japan in 1945, the US. military was in charge of
the establishment of new government and the police system. The national police system
was still maintained for peacekeeping and order maintenance purposes. Historians
understood that the US. military decided to maintain the national bureaucratic police
system to ensure social order against leftists in the peculiar situation of the division of the
country between Capitalist South and Communist North (Cummings, 1997). Due to this
political situation, the focus of the national police was national security controlling
leftists and political opponents. The regional situation also afforded power to the military,
and South Korea was under military government until a civilian president and democratic
government was implemented in 1993. The democracy, thus, has very short history in
South Korea. The role of the police throughout the colonial and military regime was
mostly to support the illegitimate ruling authorities (Hoffman, 1982; Hwang et al., 2005;
Park, 1988). To overcome the oppressive image of the past and revive as an agency for

the people, the police were compelled, encouraged, and motivated to shift their role from

19

government protector to citizen protector and service provider (Korea National Police
Agency, 2003). As most countries in the developing world struggle to move from
authoritarian policing to democratic policing (Davis, Henderson, & Merrick, 2003),
South Korea has recently been through tremendous reform endeavors to establish
community oriented police practices.

One of the most signiﬁcant reform endeavors was the “Police Grand Reform” in
2000 (Pyo, 2002), which included both internal and external issues. Internal reforms
included improved payment and work environment as well as encouraging bottom-up
communication (Korea National Police Agency, 2002). After months of internal efforts
to change the organizational culture from authoritative to democratic, external reform
measures followed, which were directed toward improving the relationship between the
police and citizens. External reform endeavors tried to eradicate the previous image of
police as being unkind, unfair, and unfaithful (Korea National Police Agency, 2002; Pyo,
2002). Operational strategies to strengthen external relationships with the public
included citizen police academies, community based patrol in partnership with volunteer
residents working closely with the local police, and emphasis of a police role in domestic
violence disputes and crimes against women/children (Pyo, 2002). Even though the
reform initiatives were not strictly based on philosophies and models of community
policing, they were considered as a general trend toward community policing, given the
emphasis on enhancement of the public image of the police, community crime prevention
through collaboration with voluntary citizens, and attention to traditionally ignored

problems (i.e., domestic violence).

20

 

For continued reform, the police created the Police Reform Committee in 2003,
which was composed of 18 external ﬁgures who were selected among experts in
government administration and reform expertise (Korea National Police Agency, 2004).
Reform programs included citizen boards for monitoring civil rights violations in the
police organization, the recording and videotaping of the investigation of the child
victims of sexual abuse to ensure protection, and the renovation of the police substation
system. One of the most important products of the committee was the reform of a police
substation system that had been utilized since the birth of the modern police without
signiﬁcant modiﬁcations.

South Korea has long maintained a police substation system as the bottom level
main structure of the hierarchical system. These police substations are observed at every
comer of the country and have been a major tool of social control since its introduction
with the establishment of the modern police system during the 18908 (Hoffman, 1982).
After the dark age of colonization and military dictatorship, this police substation system
became a main post for community crime control. There were 2,944 substations
throughout the country and about 44 percent of ofﬁcers (39,988) were assigned to mini-
stations in 2002 (Korea National Police Agency, 2003). This system is also observed in
other Asian countries. Singaporean police posts and Japanese Koban are a couple of
examples (Skolnick & Bayley, 1988). Substations in Korea, Japan, and Singapore were
responsible for all aspects of policing except criminal investigation, including receiving
complaints, responding to calls for service, providing information and advice, patrolling
on foot or bicycle, organizing community crime prevention, and developing personal

contacts (Korea National Police Agency, 2003; Miyazawa, 1992; Skolnick & Bayley,

21

1988). Foot patrols were the mainstay of patrol coverage in Korea, Japan, and Singapore
up until the late 19805 (Korea National Police Agency, 2004; Miyazawa, 1992;
Mizumachi, 1982), which was largely transformed to motor patrols since then. This
multi-function of substations in Asian countries was different from those of the United
States. Detroit’s mini-stations, for example, did not conduct general police work but
were responsible only for community crime prevention tasks (Skolnick & Bayley, 1988).
In addition, the substation system in Korea illustrates some unique features due to its
national identity. Although substations are spread across the country, ofﬁcers working in
the substations do not have much discretion. Often, several substations are commanded
and controlled by a main-station and several police main-stations are directed by a
regional police agency that is subordinate to Central Headquarters.

Recently, Korean police ofﬁcials recognized some negative aspects of the
historical substation and initiated changes in the system (Korea National Police Agency,
2004). Although it was commendable for substation ofﬁcers to familiarize themselves
with the people and problems in the neighborhood, multiple tasks in substations from bar
regulation to trafﬁc enforcement were responsible for police corruption and poor public
perceptions of ofﬁcers. The spread of ofﬁcers throughout so many mini-stations often
made it difﬁcult to handle speedy, violent, and multi-jurisdictional urban problems. In
many rural areas, ofﬁcers had no choice but to stay in substations for administrative work
without patrolling the neighborhood, due to the lack of ofﬁcers. In addition, residents
perceived police substations as a frontline of prior illegitimate ruling authority despite the
nation’s democratization. Recognizing this, the Korean police implemented reforms into

the police substation system in 2003 in order to centralize the patrol system and to allow

22

substations to focus on community service functions, subsequently referred to as the
“community policing system” (Korea National Police Agency, 2004).

In this system, every three to ﬁve substations were grouped together and most
ofﬁcers were assigned in a central station. The central stations, then, provided motor
patrol while the substations provided ofﬁce based community services (i.e., consultation,
crime reporting, information exchanges). The substation system of the past was
maintained only in limited rural areas and some urban areas. Although this system was
equipped to handle some violence by dispatching several patrol units at the same time,
the police recognized some negative effects of this system (Korea National Police
Agency, 2004). A public opinion poll showed that response time became longer due to
the broadened jurisdictions, while substations were vulnerable to law violators due to the
minimized number of ofﬁcers, and citizen perceptions of the centralized patrol system
remained negative (Yonhap News, 2005). The Korean police have attempted to
implement changes into the system to model Western inﬂuences, but difﬁculties lie in its
programmatic approach due to some misunderstandings of community policing and
cultural differences in police organizations. This parallels community policing in the
United States and other Western countries that struggle with the problem of internal
resistance and misunderstandings of the concept (Leighton, 1991; Manning, 1984).

In sum, despite the origins of the Korean police in the primary functions of
community duty and protection of the village (Pyo, 2001), the role of police was
complicated by its national history. Korea’s colonization by Japan (1910-1945), coupled
with military governments for much of the twentieth century, resulted in the police force

having to play an oppressive role to serve the ruling powers that wanted to maintain their

23

authority by suppressing opposing citizens. College students and educated citizens
accepted the need to resist unjust governments and historical records illustrate that many
citizens considered it shameful not to resist such authority. Even after Korea’s
independence and democratization, due to the government priority of national security
based on the country’s regional division between the North and South, ofﬁcers in the
national police system possessing power and authority have dominated the public they
are supposed to serve (Hoffman, 1982). Due to the constant threats from the communist
North, the nation has been preoccupied with issues of national security (Kim, 2003).
Thus, Korean citizens have perceived the police as a frontline representative of
government authority rather than as a public protector of the people. This is important
since an assumption of community policing requires that the public be willing to trust and
work in conjunction with police (Manning, 1984), and share some of the responsibilities
that come with crime prevention (Sadd & Grinc, 1994). In South Korea, one study noted
that citizens still doubted the trustworthiness and fairness of police ofﬁcers (Choi & Gy,
1997), and police ofﬁcers also deﬁned their relationship with citizens as being negative
or hostile (Lee, 2002; Moon, 2004). Although South Korea has intensive community
cohesion and collective efﬁcacy, the historically negative image of police may launch a
unique outcome regarding the impact of community policing on fear of crime and public
satisfaction with police. For successful implementation of community support, it may be
a prerequisite for South Korean police and many other Asian countries to shift the
historical image of the police as a government arm toward a democratic role as the public

partner or protector. Recognizing the necessity, political and police leaders have

24

implemented signiﬁcant reforms based on a community policing model intended to
change the image of the police to one of service to the community.

The impact of some Western programs of community policing in Korea (i.e.,
citizen police academy) on public satisfaction with police or fear of crime is not clear due
to the lack of empirical research. Studies of community policing in the United States
showed that the negative image of police among residents was an obstacle of community
policing in some communities with historically poor relationship with police (Sadd &
Grinc, 1994). This unique relationship between the police and the residents in South
Korea provides a comparative context for the studies of the impact of community
policing on public satisfaction with police and fear of crime. There are some positive
aspects given that the government and citizens are acknowledging the role of Korean
police as crime prevention professionals and the police are extensively being involved in
designing new cities. As an example, Korean police have been introducing Crime
Prevention Through Environmental Design (CPTED) methods in collaboration with other
govermnent agencies. The police launched a task force and are designing a crime
prevention model in cooperation with the Ministry of Construction and Transportation to
apply the CPTED methods to a newly constructed Pankyo City (Korea National Police
Agency, 2005).3 These extensive roles of police for community safety suggest a positive
progress of policing toward the community in this country.

4. Crime and Fear of Crime

Using ofﬁcial crime data from Korea, the United States, and Japan, J00 (2003)

observed that the overall crime rate of Criminal Code Offenses (CCO) in Korea has been

relatively low over the past three decades. In 1990, “the overall crime rate per 100,000

25

people was the lowest in Korea (910) compared to the US. (5,820) and Japan (1,324) in
major crimes” (J 00, 2003, p.250). He also indicated that during 1990-98, the crime rate
of the CC0 in Korea has risen steeply from 910 to 2,330, while that of Japan has risen
moderately from 1,324 to 1,608 and that of the United States decreased from 5,820 to
4,616. Rapid industrialization and urbanization, from the 1960’s until the 1980’s, appear
to be responsible for social disorganization and crime prone circumstances in society. In
addition to the rapid urbanization, the Korean economic crisis in 1997, and the overall
tough policies on crime (e. g., war on crime) in the 1990’s, were responsible for dramatic
increases in crimes reported to ofﬁcial law enforcement agencies (J ang, 2003; Yoon,
1998). Contrary to developed countries (e. g., the United States and Japan) that had
already experienced industrial development accompanied by social changes, Korean
society is in the process of changes in social demographics and circumstances through
rapid industrialization (Hong, 2003), which appears responsible for the increasing trend
of crime.

In addition to the ofﬁcial crime indicators, the victimization survey conducted by
the Korean Institute of Criminology claims that “the victimization rate in Seoul is one of
the highest among most metropolitan cities in the world” (Lee, 1997, p.318). As J 00
(2003) argued, Confucian moral codes and collectivism in traditional agricultural society
historically resulted in effective social control among Koreans. Conﬁrcianism based
ethics, however, were very much disrupted during the social uprisings such as Japanese
colonization (1910-1945) and the Korean War (1950-1953). Under this situation of norm
disruption, rapid industrialization may again accelerate social disorganization in the

urbanized Korean society, which may be responsible for the dramatic increase in crime.

26

Cho’s (2002) descriptive study of 500 police ofﬁcers and 304 citizens in a
metropolitan city of South Korea suggested differences in the perceptions of police and
citizens. Speciﬁcally, whereas the police expressed conﬁdence in citizen’s safety,
citizens perceived that crime was widespread in their neighborhoods. Regarding
victimization and fear of crime, Lee’s (1997) review of the limited literature in Korea
maintained that women were less likely to be victimized because of their decreased
exposure to crime, since routine activities under the cultural tradition of Confucianism do
not usually include going out at night. Korean women, however, were more fearful of
crime than men according to the Social Indicators in Korea by the Korean National
Statistical Ofﬁce. Utilizing a sample of 528 Korean women residing in Seoul, Lee (1997)
reported that 76 percent of the respondents felt somewhat or very unsafe while they were
walking alone in their neighborhood at night. The majority (90 percent) reported that the
probability of being a victim of street robbery at night to be somewhat or very high. The
study included both global fear, “how safe do you feel while walking along in your
neighborhood?,” and speciﬁc fear “how likely do you feel that you might be robbed
while walking alone in your neighborhood at night?” The results showed that speciﬁc
fear was most inﬂuenced by indirect experience of victimization through acquaintances
and the media, followed by satisfaction with community quality of life, and perceived
seriousness of crime in general. On the other hand, global fear was explained by their
perceived seriousness of crime problem, followed by satisfaction with community quality
of life. Interestingly, neighborhood cohesion and perceptions of police ability and
visibility were not significantly associated with global fear or speciﬁc fear of robbery.

This is contrary to the general ﬁndings in Western societies where community cohesion

27

and conﬁdence in the police related to the reduced fear of crime (Box et al., 1988; Velez,
2001). Historically and culturally constructed oppressive and authoritative aspects of
public social control, along with the long tradition of informal social controls and
corresponding insigniﬁcant role of other community organizations could possibly be
responsible for the differences of fear of crime between Western and Korean societies.

In a similar cultural setting, research conducted by the Research Institute of
Public Security in China, out of 12,652 citizens in 15 provinces, autonomous regions and
direct jurisdiction cities, Chang (1990) found that almost 50 percent of respondents
reported that they were afraid of going out alone at night. The fear of property crime was
highest, followed by violent crime and assault. Interestingly, only 19 percent of citizens
responded that they would go to police ofﬁcers for help when they perceived their
personal safety (or the safety of relatives or property) was at risk. However, the majority
(70 percent) stated they would report victimization to the police. Respondents reported
that enhanced patrol, enforcement of justice, and helping and educating youth and
juvenile delinquents to be the preferred approaches to enhancing safety.

Utilizing survey data from 15,000 citizens throughout China, conducted by the
Chinese Ministry of Public Security, Curran and Cook (1993) reported that juveniles in
China have a greater fear of crime than their elders. Unlike the Western cultures, age was
negatively associated with fear of crime. Comparing ages from 29 to 45, 46 to 60, and 61
years and above, those from the 16 to 28 age group had the highest fear of murder, rape,
assault, robbery, and hooliganism. Also, these juveniles’ general sense of safety was
lower than the average of the nation. The authors attributed this to the fact that

contemporary Chinese youth are mostly residing in an urban society where crime is high.

28

Statistics also reveal that homicides, rapes, and assaults increased dramatically among
juveniles, whereas overall crime rates stayed low. Another possible explanation,
however, could be related to the Confucian-based long history of respect of the elderly.
That is, Chinese and Korean elderly may perceive less vulnerability than their Western
counterparts due to cultural taboos.

In another similar cultural context, Ito’s (1993) literature review showed that fear
levels were considerably higher than actual crime rates in Japan. Contrary to the general
belief that Japan’s crime rates were lower and more stable than those in other
industrialized countries, perceived fear was substantially high. Among predictors of fear
of crime, women were more fearful of crime than men but the effect of age was mixed.
Interestingly, fear level for several speciﬁc crimes were generally lower in megalopolises
and small cities and higher in large cities, middle scale cities, and farming districts. It was
interpreted that in farming districts, larger populations per police ofﬁcer and lower levels
of patrol activities by police were responsible for higher levels of fear, while in
megalopolises, higher levels of police forces and activities helped reduce fear of crime,
despite many other factors seen as increasing fear. As Ito (1993) admitted, however, lack
of empirical research in Eastern Asian contexts makes it difﬁcult to conclude the
directional accuracy and the magnitude of predictors of fear of crime in this cultural
context.

5. Applicability of Models of Fear of Crime

A number of the models of fear of crime in major Western countries appear to be

applicable in the South Korean context. Traditional emphasis on informal social controls

and collectivism, coupled with historically negative perceptions of the role of the police

29

may result in different patterns of fear of crime in Korean context, however, compared to
the Western context that has emphasized the role of public social control, individualism,
and equal rights in social life.

The victimization model considers prior experience with victimization, awareness
of victimization of close people, and vulnerability as factors of fear of crime (Hale, 1996).
In general, variables of physical vulnerability include gender and age, while those of
social vulnerability include low socioeconomic status and racial minority status. Since
South Korea maintains racial homogeneity, social vulnerability as a racial minority is not
applicable in this context. Yet, gender, age, and socioeconomic status account for some
distinction in this country. Under the Confucian tradition, women were largely perceived
to be subordinate to men, while the elderly were highly recognized in society. It is not
only the physical vulnerability of women or the elderly that is generally perceived among
Western scholars, the low social status of women and the higher status of the elderly in
the Korean context suggest that these variables would have distinct effects from that of
Western societies. Women may perceive fear due to their social as well as physical
vulnerability. For example, Korean newspapers distributed the story of two women who
were hit by a street vendor because they were smoking on the street (Yonhapnews, 2005).
The street vendor argued about women smoking in public while he was arrested by the
police. This is a current theme in modern Korean society, which signiﬁes the social
vulnerability of women due to the traditional Confucianism-based male domination in
society, where quiet and obedient women have been valued. Confucian culture is
characterized by patriarchal family structure where females are socialized to be passive or

docile, which makes them socially as well as physically vulnerable (Lee, 1998).

30

On the other hand, the Korean elderly enjoy their social status compared to their
counterparts in Western society. Korean culture teaches young people to respect the
elderly, and to treat elders with politeness and honor (Lee, 1998). Even under the strong
inﬂuence of individualism and equal treatment of social status of Western values,
Korean’s believe the elderly should at least be prioritized for general social activities.
For example, it is often noticed that the young usually yield their seats to the elderly in
public transportation. This high status of the elderly may provide them with a sense of
safety despite their general physical vulnerability. The previous ﬁnding of the negative
association between age and fear of crime in China (Curran & Cook, 1993), which is
contrary to general ﬁndings in Western countries, may be due to the high social status
that the elderly enjoy in a Confucian culture.

Regarding social vulnerability, factors in Korean context may include education
and socioeconomic status. Ofﬁcial statistics generally suggest that Korean society was
less successful in reducing the unequal distribution of income during the rapid economic
grth period and the level of poverty has risen due to the dramatic increases in
unemployment (Lee, 2003). The heavy concentration of wealth in a small proportion of
people is considered a major obstacle to the ﬁuther enhancement of the quality of life
since lives of the poor are at risk due to the minimal social security system in Korea (Lee,
2003). Also, as previously discussed, reliance on social achievement through education
is a phenomena that Korea and other Eastern Asian countries hold (Kim & Lee, 2003). It
is important to note in this regard that Korean society is very hierarchical not only in
gender and age-related relationships but also in other social aspects that include

educational and income status. Challenges to the socioeconomic hierarchy are not

31

culturally tolerated. This hierarchical social relationship structure may render those of
low educational background and income more socially vulnerable than their counterparts
in Western societies.

Indirect experience of victimization through conversation with friends, relatives,
and other acquaintances is also considered a general factor of fear of crime in Western
literature (Hale, 1996). The strong dependence on informal networks due to the family
centric group life in Korean society may allow the spread of victimization news faster
and more intensely than in Western societies. That is, this integration through familial
and communal networks may act as a basis of the spread of crime news. Also, previous
research indicates that the Korean society has rapidly mobilized through cell phones, the
Internet, and other high technology measures (Kim, 2003). Thus, family and community
based every day interactions, along with the modernized aspect of high accessibility to
news media in Korea, suggest that the effect of indirect victimization through family and
community networks as well as the media would have substantial impact on their fear of
crime.

The community control/concem model posits that perceptions of informal social
controls and public social control dimensions in communities are linked to people’s fear
of crime. Since the traditional social control practice has relied heavily on informal
sanctioning and social shame that were linked to people’s endeavor to maintain a good
family name, perceptions of such traditional norms may be signiﬁcantly associated with
people’s fear of crime. The role of Christianity for social networks in urban Korean
societies needs attention in this regard. The church consists of more than believers and is

a place where people interact closely as in previous traditional societies (Kim, 2003).

32

Their integral relationship is considered a substitute of the traditional familial community.
Such a network would also function as an inhibitor of fear just as community integration
does in Western societies (Gibson et al., 2002; McGarrell et al., 1997), suggesting that
the role of family-like Christian networks in Korean cities could be studied in relation to
the community control/concern model. However, different from the Western notion of
public social control, Koreans have perceived the public social control system as an arm
of government authority to supervise citizens (Hoffman, 1982; Pyo, 2001). In general,
conﬁdence in the police and perceptions of police activities were helpful in the reduction
of fear in Western societies (Moore & Trojanowicz, 1988; Velez, 2001), but this may not
be true in the Korean context where distrust of the police has been largely distributed
among citizens due to the oppressive role of the police in colonization and military
regime (Hoffman, 1982; Lee, 1990). The innovation to change the police as a function
for the people and the community has been in process. The success of this endeavor
could at least boost the image of police as guardians and protectors of the citizenry,
though further investigation is needed.

The disorder model and the community context model appear to be applicable in
the Korean context in light of the country’s rapid urbanization. As social disorganization
theorists explain, rapid industrialization in Korean society has generated problems of
disorder, economic gap, and instability in urban societies even though racial
heterogeneity was not an issue. As a result of urbanization and westemization, disruption
of informal norms of traditional shaming and sanctioning capacity has become a social
problem in Confucianism’s conservative society. Since people have been under the

inﬂuence of the conservative norms, the impact of juveniles’ resistance against parents,

33

elderly, teachers, and government ofﬁcials has been enormous for the general public,
when people started to experience a symptom of social disorganization throughout the
urban development process. Therefore, people’s perceptions of disorder as well as
community indicators of concentrated disadvantage (e. g., poverty, unemployment, etc.)
and crime rates may be responsible for the increased fear of crime in Korea. The impact
may be stronger in this cultural context due to the traditionally conservative social
atmosphere.

Most Western models of fear of crime appear generally applicable in Korean
society. However, social control processes and modernization affecting Korea, that differ
from general social control processes in Western societies, suggest that the effect of some
predictors of fear of crime may have different directions or magnitudes. Consequently,
comparative study would be beneﬁcial to both theoretical and policy understanding of
fear of crime. In addressing crime and fear of crime, it appears that progress in the west
derives from both public social controls and development of community mobilization. At
the same time, Eastern Asia lost much of her informal sanctioning capacity through the
process of establishing a formal social control system. Maintaining informal social
control networks would be valuable in Korea as it pursues globalization or
Westemization. Finally, research in a comparative context could present a model
containing multiple predictors of effective social controls as well as the reduction of fear
of crime.

In sum, this chapter indicated that social controls in Korean society were largely
informal in nature due to the origins of Confucianism and Buddhism, compared to the

emphasis on public social controls in the United States. It was realized that this society

34

was still maintaining such a tradition even under the processes of industrialization and
globalization. In addition, citizens perceived the police as representatives of government
authority in Korea, compared to the view of the police as crime controllers in the Western
society. Even though it was difﬁcult to discern the directions and magnitudes of factors
of fear of crime in the Korean context, a number of the models of fear of crime developed
in major Western countries appeared to be applicable in this cultural context. Some
unique aspects were discussed (i.e., age, public satisfaction with police) regarding fear of
crime among Koreans. The next chapter will address the connection between fear of

crime and community policing.

35

 

 

CHAPTER THREE
COMMUNITY POLICING AND FEAR OF CRIME

1. Reform Models of Policing

Research on the effectiveness of traditional policing, increasing crime and fear of
crime, and hostility between the police and the community led to a new style of policing.
New styles emerged as a reaction to historical demands. These reform models included
team policing, zero tolerance policing, problem oriented policing, and community
oriented policing.

Team policing was one of the earliest strategies to bring police closer to the public.
The concept of team policing, though adopted by only a small number of agencies,
became the buzzword of the 19705 (Riechers & Roberg, 1990, p. 106). A report from the
Presidents Commission on Law Enforcement and Administration of Justice in 1967
proposed the implementation of team policing (Palmiotto & Donahue, 1995). The
premise behind team policing was that units of police ofﬁcers, consisting of ﬁve to ﬁfty
ofﬁcers in a group assigned to a geographic area, working toward commonly deﬁned
goals in a cooperative effort can provide more comprehensive police service. Teaming
was an innovation that enabled police personnel from various divisions to participate as
full partners in the development of a superior police service. The concept of team
policing included increasing familiarity with neighborhoods which would provide police
ofﬁcers a sense of personal responsibility, and, in turn, the community’s familiarity with
the ofﬁcers would breed trust and cooperation, as well as the community’s sense of
responsibility for crime and social problems (Riechers & Roberg, 1990, p.106). The
major elements to this strategy included “stable geographic assignment, decentralized

authority, emphasis on community relations, emphasis on crime prevention, improvement

36

 

 

of internal communications, and reduction of reliance on department specialists such as
centralized detective units” (Wasson, 1977, p.23). Although some studies on team
policing experiments illustrated a reduction in crime and improvement in community
relations (Bloch & Specht, 1973; Bristow, 1969), the results were different depending on
management strategies and quality of personnel (Wasson, 1977). This strategy was
typically supported by chief administrators, but middle management resisted due to a
substantial reduction in their authority as a result of decentralization (Wasson, 1977).
Team policing failed because it required restructuring of the organization on a massive
scale in a dominant bureaucratic environment, which resulted in internal resistance and
the decline of this strategy in late 1970s (Rosenbaum & Lurigio, 1994)" Despite the
failure of the team-policing concept in some organizations, the idea of a community
context of policing remained strong (Allen, 2002). It is worth noting, therefore, that the
elements of team policing have been encompassed in the later reform trend of community
policing (Riechers & Roberg, 1990).

Wilson and Kelling (1982) developed the ‘broken windows’ theory, also known
as, ‘order maintenance’ or ‘quality of life’ policing, which emphasized the importance of
disorder rather than serious crime. Left uncorrected, signs of physical decay (i.e., broken
windows, dilapidated buildings, grafﬁti, etc.) and social disorder (i.e., public drunkenness,
street beggars, delinquent teenagers, etc.) invites widespread disorder, sending a message
of tolerance, which breeds more serious types of crime (Skogan & Roth, 2004). In turn,
frightened law abiding citizens avoid the area or move out, buildings deteriorate and
become vacant or are occupied by the disorderly and the criminal element (Skogan &

Roth, 2004). Disorder is directly linked to the fear of crime. Broken windows

37

symbolized physical and social disorders that foster fear of crime as well as criminality,
lead residents to withdraw from the public life of the neighborhood, resulting in the
reduction of the efﬁcacy of informal social controls (Greene, 1999; Herbert, 2001;
Walker & Katz, 2005). According to the notion of the feedback loop, a decrease of
informal social controls will eventually attract more serious offenders (Herbert, 2001). In
other words, broken windows theory holds that if not ﬁrmly suppressed, disorderly
behavior in public will frighten citizens and attract predatory criminals, thus increasing
fear and leading to more serious crime problems (Greene, 1999). Broken windows
theory resulted in several models of policing including zero tolerance policing and
problem oriented policing. Zero tolerance policing was implemented in 1993 in New
York City. The key notion was that serious crime problems may be quelled by mounting
a large scale attack on petty crimes and disorderly conduct through a zero-tolerance
strategy (Greene, 1999). Broken windows redirected the attention of police from a
historical emphasis on serious types of crime toward disorder or misdemeanor violations
in an effort to reinstitute mechanisms of informal social controls. Zero tolerance differs
from community policing since it turned more exclusively to the police to solve disorder
problems by, for example in New York City, arresting individuals such as farebeaters
(avoid paying to ride the subway) and squeegee men (wash cars without being asked and
force to pay) (Skogan & Roth, 2004). New York subsequently experienced a dramatic
decrease in crime after implementing this strategy, yet police organizations experienced a
dramatic rise in complaints by encouraging ofﬁcers to be overly aggressive (Greene,
1999). This, in turn, often negatively affected police community relations (Walker &

Katz, 2005).

38

Herbert (2001) explained that broken windows policing dominated due to three
main reasons “police culture and organization, wider cultural understanding of crime and
deviance, and political dynamics” (p.446). The police preferred the broken windows
approach to community policing because of the emphasis in the police subculture on
masculinity to enforce the law bravely and morality of good guys to punish bad guys; the
cultural understanding of crime as a result of too lenient or faulty mechanisms in the
criminal justice system; the breakdown of informal social controls; and ﬁnally, political
preference to allow state actors to justify being tough on crime (Herbert, 2001 ). Green
(2004, p.48) also argued that “many police agencies have shifted toward zero tolerance
policing and it is not clear whether the police identify them with community facilitation
and partnerships or with the suppression of social disorder, or bot .”

Broken windows theory also supported the notion of problem oriented policing
that had been advocated by Herman Goldstein (1979). Goldstein argued that the police
should take categories of crime and disorder and break them down into discrete problems
and then develop speciﬁc responses to each problem, since the root causes and
dimensions would vary. This concept emerged due to the necessity to identify and
analyze problems, develop and implement strategies, and evaluate their effectiveness
(J iao, 1998). This was a reaction to the dominant means over ends syndrome, where
police focused on organizational changes and operating methods rather than the
substantive outcome of their work (Goldstein, 1979). Goldstein (1979, p.238) thought
“the failure of team policing was due to a focus on secondary considerations” such as
organizational change with a management dominated concept of police reform without

concentrating on problems to be solved. To develop a more systematic process for

39

examining and addressing problems in communities, problem oriented policing was
typically implemented through a four-stage process known as scanning, analysis,
response, and assessment (Eek & Spellman, 1987). They initiated systematic research
and evaluations, focused on crime prone locations, and designed crime speciﬁc
intervention strategies to handle repeat victims and recurring problems. Unlike
community policing, the goal of problem oriented policing does not necessarily include
building a close relationship with citizens but serves to reduce problems of concern to the
public where the community partnership is merely an element of that process (Eck &
Spellman, 1987; J iao, 1998; Walker & Katz, 2005). Problem oriented policing is often
confused with the process of problem solving in the course of community policing. The
initial concept of problem solving in community policing tended to be small in nature
because it was a neighborhood level strategy but was broadened to include all elements of
problem oriented policing, which became an important part of community policing
(Cordner, 1998; Walker & Katz, 2005). Emphasis on community policing regarding the
permanent assignment of ofﬁcers to a particular neighborhood to instill geographic
ownership and responsibility, is often considered a signiﬁcant difference between the two
policing strategies (Benson, 2004; Trojanowicz, Benson, & Trojanowicz, 1988). In
practice, problem oriented policing can be implemented alone or as part of community
policing (Walker & Katz, 2005).
2. Community Policing

The concept of community policing in the United States began with the

community relations programs of the 19505 and 19605 that were developed to increase

interactions between minorities and the police, and continued through the 19705 with the

40

concept of team policing (Greene, 1987). Community policing contains the basic
strategy of Peel’s philosophy and practices adopted by police in the United States in the
nineteenth century (Eck & Rosenbaum, 1994; Leighton, 1991; Peak et al., 1992). That is,
contrary to the general notion of community policing as a new approach, it was
considered a reemergence or renewal of former approaches in the London Metropolitan
Police Department (Leighton, 1991; Thompson, 1991). A comprehensive theory of
community policing has yet to emerge and is theoretically an undeveloped set of policing
principles and practices (Leighton, 1991).

Troj anowicz (1994), however, has stated that normative sponsorship theory and
critical social theory are the bases of community policing. “Normative sponsorship
theory emphasizes problem solving through shared values and a community of interest
and critical social theory requires enlightened and empowered citizens and police ofﬁcers
so that emancipation results” (Trojanowicz, 1994, p.258). The ideology maintains that
the relationship between the police and the community has an effect on crime, and police
access to the community can serve as a vital link in restoring a traditional sense of
community (Goldstein, 1987; Trojanowicz et al., 1988).

In addition to community partnerships, both problem oriented policing and broken
windows policing provide the root philosophy of community policing. Problem oriented
policing became an important part of community policing for its attention toward the root
causes of crime as well as chronic or repeated problems (Goldstein, 1987; Walker & Katz,
2005). Building upon the theories and research of Goldstein, Wilson and Kelling called
for the police to focus upon quality of life issues and to increase order maintenance

activities rather than diverting all the attention to felonious crimes (Perez & Shtull, 2002).

41

Adams, Rohe, and Arcury (2005) considered social disorganization as another basis of
community policing. The social disorganization theory posits that neighborhoods with
high residential mobility, heterogeneity, and a large socioeconomic gap formulate social
disorganization in the community, such as weak institutions, a low level of social
cohesion and efﬁcacy, which make it difﬁcult to enforce both formal and informal social
control over residents and strangers (Adams et al., 2005; Sampson & Groves, 1989; Shaw
& McKay, 1942). Increasing community collective efﬁcacy, therefore, leads to a
decrease in incivilities and results in low level of fear of crime and a high level of
attachment to the local community, and eventually provides a community with more
effective social controls.

There is confusion in the use of the term “community policing.” It has often been
utilized to describe programs for police-community relations, team policing, foot patrol,
or unit beat policing (Manning, 1984). Trojanowicz and Bucqueroux (1990) emphasized,
however, that unlike earlier reform efforts, community policing is value based rather than
solely programmatic.

It is deﬁned as:

a new philosophy of policing, based on the concept that police ofﬁcers and

private citizens can work together in creative ways to solve contemporary

community problems related to crime, fear of crime, social and physical disorder
and neighborhood decay. The philosophy is predicated on the belief that
achieving these goals requires that police departments develop a new relationship
with the law-abiding people in the community, allowing them a greater voice in
setting local priorities, and involving them in efforts to improve the quality of life

in their neighborhoods (Trojanowicz & Bucqueroux, 1990, p.5).

Synthetically, “. .. is both a philosophy and an organizational strategy that allows the

police and community residents to work closely together in new ways to solve the

problems of crime, reduce fear of crime, and improve neighborhood conditions”

42

(Trojanowicz et al., 2002, p.37). In short, these deﬁnitions outline the idea that the police
initiate problem solving approaches in partnership with the community to improve
quality of life and prevent crime.

Among the three primary and interrelated functions including crime control, order
maintenance, and service provision, the professional policing model focused primarily on
crime control while the community policing model broadened the scope and placed an
equal emphasis on all three activities (Scheider, Rowell, & Bezdikian, 2003; Zhao,
Scheider, & Thurman, 2002). The goals of community policing often include reducing
fear, enhancing public satisfaction with the police, and deterring crime through problem
solving (Goldstein, 1987; Trojanowicz & Bucqueroux, 1990). That is, while the police
have traditionally deﬁned their primary mission in terms of crime control, community
policing expands the role of law enforcement to include such issues as “fear of crime,
order maintenance, conﬂict resolution, neighborhood decay, and social and physical
disorder as basic functions” (Walker & Katz, 2005, p.313).

Community policing rearranges priorities among functions and adds new ones
(Eck & Rosenbaum, 1994). Fear of crime was well recognized as one of the major goals
of community policing where changes are being implemented that are designed to be
more responsive not only to crime, “but to the fear of crime, more broadly, to a wide
range of problems that affect the quality of life in our urban areas” (Goldstein, 1987, p.7).
In addition, social and physical disorder are important foci of community policing as
some of the root causes of serious crimes (Wilson & Kelling, 1982) as well as the
facilitators of fear of crime (Skogan, 1990). The objectives of community policing

include reducing fear of crime and increasing community cohesion in part through

43

decreasing physical and social disorder (Lanier & Davidson H, 1994), which is consistent
with broken windows theory that community informal social control is enhanced by
handling disorderly matters.

Together, these goals are intended to achieve an often unstated but important goal
of community policing, which is crime prevention (Riechers & Roberg, 1990). In sum,
the goals of fear reduction and public satisfaction with police are achieved through
enhancing informal social control (i.e., facilitating community institutions, building
collective efﬁcacy, and increasing social cohesion) and order maintenance (i.e., handling
social and physical disorder). Although prevention was an explicit community policing
goal, though rarely the ﬁrst priority (Roth et al., 2004), this process eventually is intended
to have a long term effect on crime (Riechers & Roberg, 1990).

3. Community Policing and Fear of Crime

In the mid 19605, as the crime rate began to increase sharply so did public fear of
crime in the United States (Walker & Katz, 2005). As noted earlier, the President’s
Commission on Law Enforcement and Administration of Justice in 1967 regarded fear of
crime as a major social problem. Scholars recognized that fear of crime created
community deterioration where citizens were unable to exert informal social control,
which eventually increased crime. Skogan’s (1986) explanation of the feedback
mechanisms of community deterioration as a result of fear included—physical and
psychological withdrawal from community life, a weakening of the informal social
control processes that foster crime and disorder, a decline in community organization and
mobilization capacity of the neighborhood, deterioration of business conditions,

production and invitation of delinquency and deviance, and further dramatic changes in

44

the composition of the population. Fear of crime, in a chain of factors, is linked to the
abandonment and deterioration of neighborhoods (Wilson & Kelling, 1982), which leads
to neighborhood decline that may in turn, lead to a higher crime rate (Skogan, 1986,
1990). Moore and Trojanowicz (1988) note that if peOple retreat and hide in their homes,
they make their homes safer but make the streets more dangerous. Theoretically,
therefore, fear reduction programs were viewed as another form of crime prevention, a
core function of the police. In addition, levels in fear of crime were linked to citizen
participation in community social control. During the various crime prevention efforts
throughout the 19705 and 19805 (Moore & Trojanowicz, 1988), it was recognized that
fear of crime was one of several barriers to citizen participation in community crime
prevention and problem solving efforts (Grinc, 1994).

Due to the traditional focus of police on crime control, however, reducing fear
was typically not viewed as a police objective. Some scholars argued that police may not
reduce fear since it is correlated with racial heterogeneity, socioeconomic gap, and
mobility, which the police could not control (see Eck & Rosenbaum, 1994). The police
also considered fear of crime as either an insigniﬁcant issue or assumed that they were
already taking care of it indirectly by their focus on crime (Scheider, Rowell, &
Bezdikian, 2003).5 Research, however, revealed that victimization did not always
determine the levels of fear of crime, suggesting a complex mechanism of fear.
Speciﬁcally, fear patterns did not follow crime patterns (i.e., fear remained stable even
when crime declines), the least victimized (i.e., old women) were most fearful, the impact
of indirect victimization due to a sensational media was signiﬁcant (Hale, 1996).

Recognizing that reducing victimization might not necessarily reduce fear, fear was

45

considered as a separate problem. Instead of traditional approaches for crime control, an
approach directly targeted to fear reduction thus emerged (Skogan, 1994).

Research about policing also provided insight on the effects of police strategies on
fear as well as crime. Traditionally, motorized patrol and rapid response to calls for
service were considered effective tools to reduce crime and possibly the fear of crime
(Bristow, 1969). Contrary to the dominant assumption that fear would decrease by
reducing victimization through such traditional strategies, research showed that these
strategies were not helpful in reducing either crime or fear of crime (Kelling, Pate,
Dieckman, & Brown, 1974). The Kansas City experiment showed that citizens were
unaware of the level of patrol in their neighborhood and their levels of fear of crime were
not inﬂuenced by reduced patrol (Kelling et al., 1974), even though the reliability of
these ﬁndings has been controversial due to some limitations related to the research
design. As well, rapid response to calls for service was not a signiﬁcant factor for
arresting criminals and enhancing citizen satisfaction with police (National Institute of
Justice, 1978). On the other hand, the implications of both the foot patrol and fear
reduction experiments was that close interaction between the police and residents did
reduce fear (Moore & Trojanowicz, 1988). In addition to fear reduction through foot
patrols and disorder management, it was realized that as police work with the community
to solve community identiﬁed problems, citizens’ fear of crime decreased (Troj anowicz
& Carter, 1988). In previous research, the Flint experiment suggested that foot patrol and
policing disorder in partnership with residents had enormous effect on the feelings of
safety of residents (Trojanowicz, 1986). The foot patrol experiment in Newark, New

Jersey also suggested that while crime and victimization rates were not likely to decrease

46

in response to foot patrol, public perceptions of social disorder and fear of crime were
affected (Police Foundation, 1981). Practitioners and researchers, therefore, identiﬁed
visible foot patrols and policing disorder as promising strategies to reduce fear of crime.
Reducing the fear of crime became a ﬁmdamental issue during the 19805 and
19905 as a police problem (Skogan & Roth, 2004; Trojanowicz, 1986). Underlying
theories of community policing explain that reduced fear would enhance community
cohesion, collective efﬁcacy, community quality of life, and eventually increased social
control against crime and delinquency (Trojanowicz, 1994). Community policing from
its inception, therefore, brought fear of crime to the attention of law enforcement as a
major goal and encouraged direct efforts to reduce it (Trojanowicz, Kappeler, & Gaines,
2002). Community policing was believed to reduce fear of crime for three theoretical
reasons: increased police presence, attention to disorder, and community partnership.
First, one of the most important assumptions of community policing is that “the
presence of the police through increased visibility reduces the public’s fear of crime”
(Manning, 1984, p.212; Riechers & Roberg, 1990, p.107). Community policing
programs (i.e., foot patrol, substations, and youth programs) encourage ofﬁcers to
increase their positive visibility in the community. Such a presence has an effect on fear
of crime directly and indirectly (Scheider et al., 2003). Frequent presence of police in the
community deters crime and decreases victimization by curbing criminality of potential
law violators, which will indirectly reduce fear of crime by reducing crime. Police
presence also fosters a sense of security and makes law-abiding citizens feel safer, which

will directly reduce fear of crime.

47

Second, community policing shifts the focus of police away from reactive
responses to crime issues and toward proactive problem solving approaches that focus on
the root causes of crime, such as disorder (Trojanowicz et al., 2002). The disorder model
posits that people’s sense of decay heightens their feeling of insecurity (Taylor & Hale,
1986). This focus of community policing on disorder, therefore, directly reduces fear of
crime by minimizing one of the most consistent predictors of fear of crime, disorder.
Police attention to disorder by ﬁxing broken windows thus signals that someone cares for
the community, which decreases fear.

Finally, community policing is a proactive problem solving strategy in partnership
with the community (Skogan & Hartnett, 1997; Trojanowicz et al., 2002). Police
endeavors for community partnership through consultation and mobilization are believed
to increase community cohesion and security. The community concern/control model
posits that community cohesion and social bonds are direct predictors of the reduction of
fear of crime (Taylor & Hale, 1986). Community based police activities such as youth
programs, community crime prevention meetings, and neighborhood watches build
strong communities by increasing collective efﬁcacy and strengthening social bonds
(Trojanowicz et al., 2002), which enhance security and reduce fear. Reduction of fear of
crime has been associated with community policing programs from the beginning and
was the core element in the deﬁnition, goal, key assumptions, and theories of community
policing. As well, practitioners and researchers identiﬁed visible foot patrols and
policing disorder as promising strategies to reduce fear of crime. The implication of both
the foot patrol and fear reduction experiments was that close interaction between the

police and residents reduces fear (Moore & Trojanowicz, 1988). In addition to fear

48

reduction through foot patrols and disorder management, it was realized that as police
work with the community to solve community identiﬁed problems, citizens’ fear of crime
decreased (Trojanowicz & Carter, 1988).

4. Impact of Community Policing on Fear of crime

In reality, the impact of community policing on fear of crimes has not been
consistent in previous studies. In their review of the quasi-experimental literature on the
relationship between community policing and fear reduction, Zhao et a1. (2002) realized
that out of 50 experimental studies, signiﬁcant reduction in fear of crime was reported in
31 studies, no changes in 18 studies, and increase in fear in one study. For example,
Newark coordinated a community policing program (Williams & Pate, 1987) which was
a comprehensive community policing model that included foot patrol, newsletter,
substation, community organization, problem solving, and residential surveys, where
decreases in fear of personal victimization and decreases in fear of property crime were
reported. The Hartford comprehensive community oriented policing program
implemented aggressive arrest, foot patrol, vehicle safety checks during the “weed” phase
and fostered better social integration during “seed” phase, where a general decline in
concerns about crime was reported (Tien & Rich, 1994). Also, within the Reclaiming
Our Area Residences project in public housing facilities in an economically
disadvantaged neighborhood of Spokane, Washington, McGarrell et al (1999; 1997)
found that comprehensive community policing and crime prevention programs such as
police substation, public education, social events, and target hardening was associated
with a reduction in fear across four waves of face-to-face interviews with representative
samples of housing residents.

On the other hand, no signiﬁcant change in perception of fear of crime was

49

reported in community policing project in public housing developments in Philadelphia
(Piquero, Green, F yfe, Kane, & Collins, 1998) and violence continued to be perceived as
a serious problem after implementing a problem oriented policing project (i.e., proactive
arrests, citations for code violations, removal of disorder) in conjunction with community
policing programs (i.e., better information sharing between ofﬁcers and residents) to
reduce disorder, street robberies, and gang problems in San Diego (Capowich & Roehl,
1994). In addition, a study in twelve US. cities reported that citizen perceptions of
community policing had no effect of their levels of fear of crime (Scheider et al., 2003).
The impact of citizen perceptions of community policing on fear of crime has also
been examined. Data from ﬁve small to medium size cities in North Carolina showed
that awareness of the police department’s community-oriented policing was strongly
related to fear of crime and citizens felt more at ease and less fearful of being a crime
victim (Adams, Rohe, & Arcury, 2005). In Chicago, Skogan and Hartnett (1997)
reported lower fear of crime levels and that citizens avoided fewer places due to worries
about being victimized in community policing areas than in traditional policing areas. It
was also reported that residents who knew about community policing and participated in
some problem solving efforts were the least fearful and avoided the fewest places.
Closely related to perceptions of community policing and based on the
assumption that increasing government responsiveness to crime and neighborhood
problems would reduce the fear of crime, McGarrell, et al. (.1997) tested and found a
signiﬁcant association between citizen perceptions of government and neighborhood
responsiveness and fear of crime. On the other hand, as an extended dimension of citizen

perceptions of community policing, in their comparison of fear of crime between police

50

volunteers and general citizens, Zhao and associates (2002) reported that volunteers
participating in community-policing activities feared both violent crime and property
crime victimization substantially more than ordinary citizens. The authors suggested that
participation in crime prevention activities might build fear of crime and heighten the
suspicion of others. Another possible interpretation was that their fear could be a
motivator for these citizens to volunteer for community policing activities.

Overall, tests of the impact of community policing on fear of crime in Western
contexts suggest mixed results. Many developing countries have accepted the
philosophies and programs of community policing. The extent to which community
policing works in these countries is not clear due to the lack of empirical research.
Testing the impact of community policing on fear of crime in Seoul may enhance our
knowledge on community policing in a comparative context.

In sum, this chapter reviewed several police reform models that emerged in
reaction to the limits of the professional police model, the community policing model
speciﬁcally, and the relevance of community policing in explaining fear of crime.
Following the review of literature indicating the mixed impact of community policing on
fear, it was argued that a test of the effect of community policing on fear of crime in
Seoul would help enhance our knowledge of this popular police strategy. The next
theoretical chapter reviews the models of fear of crime developed in the United States.

The chapter also considers various deﬁnitions and measures of fear of crime.

51

CHAPTER FOUR
REVIEW OF THEORETICAL MODELS OF FEAR OF CRIME

Over the past three decades, fear of crime has been an important theme in
scholarly work and policy debate and the body of literature has substantially grown (Hale,
1996). Scholars generally acknowledge that fear of crime is particularly an urban
problem and its impact on urban unease is enormous (Garofalo & Laub, 1978; Hale,
1996). The majority of urban residents are afraid of walking somewhere nearby at night
(Skogan & Maxﬁeld, 1981). In the course of investigating the causes and results of fear
of crime, scholars have observed some unexpected outcomes. Taylor and Hale (1986)
noted three inconsistencies: young males are the most victimized but the least fearful of
crime, elderly women are victimized the least but are the most fearful, and crime patterns
do not spatially match the fear patterns showing that residents in higher crime
communities do not necessarily report a higher fear than those in lower crime areas. In
addition to these paradoxes, Skogan and Maxﬁeld (1981) added some more unexpected
patterns: exposure to excessive crime and violence in the media did not have any
substantial impact on fear of crime, major protective measures against crime were taken
the most by people who needed it least due to their low probability of victimization (i.e.,
elderly women), and levels of fear were increasing while governments were expanding
funding to improve neighborhood quality and to encourage people to take protective
measures. Scholarly efforts to build theoretical models were partly motivated by desire
to answer these unexpected patterns of fear as well as the general recognition of fear as a
social problem. Some have attempted to answer these puzzles based on the inconsistency
of the deﬁnition and measure of fear, while others have focused on developing-theory

based models. Prior to designing a research model, it is necessary to synthesize what

52

scholars have done to deﬁne the concept of fear and to build models for factors of fear of
crime.
1. Concepts of Fear of Crime

(1) Concepts

Fear of crime refers to a wide variety of subjective and emotional assessments and
behavioral reports (DuBow et al., 1979, p.1) and lack of consistency of deﬁnition and
measurement is in part responsible for conﬂicting ﬁndings in the studies concerning fear
of crime (Hale, 1996). The phrase “fear of crime” has acquired so many divergent
meanings that it may not be reasonable to directly compare the explanatory magnitude
and directional consistency of predictors of fear of crime across various studies. Skogan
and Maxﬁeld (1981) explained fear of crime as a physiological state and expressed
attitude and described the physical manifestations of fear as “a rapid heartbeat, high
blood pressure, and an increased ﬂow of blood to the large muscles, etc.” (p.49). More
speciﬁcally, Ferraro and LaGrange (1987) deﬁned it as “a negative emotional reaction to
crime or the symbols associated with crime” (p.72). Considering most elements
suggested by scholars, Garofalo (1981) comprehensively deﬁned fear of crime as
“emotional reaction characterized by a sense of danger and anxiety produced by the
threat of physical harm elicited by perceived cues in the environment that relate to some
aspect of crime” (p.840). Consistent with these deﬁnitions, Maxﬁeld (1984) noted that
since fear is “an emotional and physical response to a threat,” people experience
physiological changes when confronted with a dangerous situation (p.3). Many scholars
have deﬁned fear with psychological and physiological reasoning, yet existing research
has used multiple measures of fear of crime due to the lack of consistency in the

deﬁnition.

53

A major problem in conceptualizing fear of crime is its confusion with risk
assessment. Fear of crime is different from general concern for safety, that is, risk
assessment. Fear of victimization is conceptualized as being afraid, worried, and
concerned with being victimized, compared to a cognitive perception of perceived risk
(Ross & Jang, 2000). Ferraro and LaGrange (1987) compared both the cognitive
dimension of perceptions of the risk of victimization and the emotional dimension of
psychological or physiological reactions to the threat of victimization and argued that
measures which did not differentiate emotional reactions from judgments of general
safety are invalid measures. Risk assessment occurs when the measure of fear relies on
judgment without considering any emotional component. An example of how risk
assessment is measured is the item “how likely it is that a person walking around here
might be held up or attacked?” This is related to the safety assessment of the community
in general (Ferraro & LaGrange, 1987, p.283). Such a measure has often been used to
assess fear of crime. Therefore, Garofalo and Laub (1978) argued that what has been
measured in research as the fear of crime is “simply not fear of crime,” but rather a
concern for community and quality of life. In addition to fear of crime, therefore, what
we call “fear” has perhaps been a measure of some other concepts which might be
characterized as “insecurity with modern living, quality of life, perception of disorder, or
urban unease” (Hale, 1996, p.84). Similarly, DuBow, McCabe, and Kaplan (1979) found
that fear, risk, concern, worry, or anxiety have often been interchangeably used in much
research, suggesting a serious lack of both consistency and speciﬁcity. Calling for a
distinction between risk assessment and fear, Ferraro and LaGrange (1987) explained that

risk to others or self, concern about crime to others or self, and fear for others’

54

victimization were distinct from fear of self-victimization, indicating that emotions of
self-victimization should be differentiated from judgments and values of others or their
own risk. Also, it is important to note that the measure becomes a risk assessment when
it does not include an imminent threat even when it is about the self (Hale, 1996). To
avoid the general judgment of risk, the measure should include some stimulus of fear,
such as walking alone at night (Skogan & Maxﬁeld, 1981).

Closely linked to the risk assessment dimension, fear has often been used without
considering a personal perspective. One of the most important components of fear is that
it is a response to a threat to oneself. In that way, it was differentiated from perceptions
of crime as a general problem or concern for others. Maxﬁeld (1984) believed that fear
involves a personalized threat rather than abstract concern or beliefs about crime in
general as a problem or concern for others’ victimization. Warr (1984) stated that fear
should actually be fear of victimization deﬁned as fear of criminal acts committed against
one’s own person or property. Likewise, F urstenberg (1971) separated fear and general
concern by explaining that “fear of crime is usually measured by a person’s perception of
people’s own chances of victimization and concern by their estimation of the seriousness
of the crime” (p.603). In his Baltimore study, he realized that the majority of the
respondents believed crime had risen, yet such an opinion of crime did not correspond to
their fear of crime. Therefore, individual’s attitudes about crime in general as well as an
individual’s assessment of others’ risk are distinct from fear of personal victimization.
Fear involves a personalized threat rather than abstract beliefs about crime since, for
example, “one might be easily concerned about the rising trend of a certain crime in

society but not at all afraid of being personally attacked” (Maxﬁeld, 1984, p.3).

55

Consistently, Ferraro and LaGrange (1987) argued that fear becomes a general concern
when it does not include personal threat even though an emotional dimension is
considered.

Fear is an emotional and physical response to a threat (Maxﬁeld, 1984). It is “the
emotional dimension of people’s response to crime that most appropriately includes
measures of fear” (DuBow et al., 1979, p.5). One might be concerned with a murder case
reported in the newspaper but not be emotionally afraid of being attacked. Hale (1996)
also realized that fear was appropriate when it was perceived as “the negative emotional
reaction generated by crime or associated symbols” (p.92). Emotion necessarily
accompanies irrationality. F attah (1993) noted that emotional fear by deﬁnition is not
based on rational objective assessments of the chances of becoming a victim, rather it is
irrational. Therefore, the concept of fear contains a situation-speciﬁc stimulus
component that includes speciﬁc kinds of risks and potential consequences to stimulate
fear irrationally and emotionally (Skogan & Maxﬁeld, 1981). An emotional threat was
illustrated as “a possible danger to oneself in the course of walking home from a late
evening visit to a bar” (Maxﬁeld, 1984, p.3). The major attention to street crime as a
stimulus of fear might be related to this emotional dimension because of its association
with signs of imminent danger in a helpless situation.

Fear of physical harm was also differentiated from worry about property crime.
Maxﬁeld (1984) differentiated fear from worry even though he suggested that the concept
of fear included both themes. Garofalo (1981) argued for the necessity of differentiating
fear of physical harm from worry about property 1055 indicating that “our losing of some

property is related to calculating, while physical threat especially at night on the street is

56

emotional and irrational because of the possibility of the physical harm” (p.854). Skogan
and Maxﬁeld (1981) noted that the main deﬁnition of fear in terms of street crime in the
neighborhood at night is persuasive in this regard since people are concerned with their
physical harm rather than property loss when they meet a stranger in such a situation.

Anticipated fear was also differentiated from actual fear. Fear at the moment of
interview may not be considered actual but anticipated fear. It is reasonable to think that
“the person walking alone in a high crime area at night is experiencing something quite
different than the subordinate who is telling an interviewer that he or she would be fearful
in such an area at night” (Garofalo, 1981, p.841). It was also recognized that anticipated
fear is more inﬂuenced by distorted information about crime, while actual fear is more
inﬂuenced by the objective threat of crime (Garofalo, 1981; Skogan & Maxﬁeld, 1981).
However, researchers cannot interview people in the fear-provoking situation. Therefore,
Garofalo (1981) suggested measuring actual fear by asking how often people ﬁnd
themselves in speciﬁc situations and how strongly they have emotionally reacted to such
situations in the past, arguing that if the question is hypothetical in nature, it may be more
likely to have tapped into anticipated fear but not actual fear.

Speciﬁc fear was emphasized over global fear. One general shortcoming in early
research on fear of crime is the failure to differentiate fear for various crime types, which
is the result of over-reliance on global measures of fear (Warr, 1984). Global fear is fear
that is measured without referring to any crime type. Even though scholars attempted to
separate worry of property crime from fear of personal attack (Garofalo, 1981; Skogan &
Maxﬁeld, 1981), others suggested that fear needed to be further speciﬁed (F attah, 1993;

Warr, 1984). Fattah (1993) argued that women might be afraid of rape and the elderly

57

might fear burglary especially when they are at home. Fear of crime research has failed

to consider the obvious fact that fear varies from crime to crime as its seriousness and
consequences are divergent, by measuring fear globally or by mostly focusing on street
crimes (F attah, 1993; Warr, 1984). It may be necessary to measure fear of crime for each
crime type (Warr, 1984), since one single crime type (i.e., rape, burglary) may capture
overall attitudes about crime and safety only for some people (i.e., women, elderly). Also,
the global approach may not capture explanatory factors of fear that vary by different
types of victimization.

Even though scholars argue that fear should contain emotional reaction to
imminent personal danger of speciﬁc crime (Garofalo, 1981; Skogan & Maxﬁeld, 1981;
Taylor & Hale, 1986; Warr, 1984), fear has often been studied as a perception or an
attitude toward overall safety issues of global crime due to the difﬁculty of interviewing
people in a fear-provoking speciﬁc situation (Garofalo & Laub, 1978; Hale, 1996). Most
of all, it is necessary to differentiate speciﬁc fear from both general concern of safety and
global fear. That is, it is necessary to distinguish speciﬁc fear, global fear, and risk
assessment. Maxﬁeld (1984) argued that while expressing doubts about one’s safety on
the streets at night reﬂects “a rather narrow set of anxieties that may or may not be linked
to other threats faced daily by urban residents, broader concern about crime and how it
affects the lives of people who may not be unsafe, but rather anxious, concerned, or
worried is more accurate and more important as a policy issue” (p.4). Haghighi and
Sorensen (1996) indicated that researchers have neglected the more common non-violent
victimizations by limiting their focus to fear of violent victimization. Rountree (1998)

found that crime experiences have differential effects on fear of violence in comparison

58

to fear of burglary. In her Seattle neighborhood study, violent victimization increased
both fear of violence and fear of burglary, but burglary victimization increased only fear
of burglary. Most of all, it is necessary to theoretically and empirically distinguish risk
assessment and emotional fear as well as global fear and speciﬁc fear (Rountree, 1998).
(2) Standard Measures

As stated, the conceptualization behind the composition of fear of crime has been
problematic due to the lack of consistent and standardized measure. The measure of fear
has also been divergent in its deﬁnition. In general, Hale (1996) reviewed research on
fear of crime and addressed some elements that have been associated with the measure of
fear: it was measured quantitatively, it has been focused on ordinary street crime, and it
was considered to be an emotional or psychological property of certain people. One of
the most popular measures of fear of crime is in the National Crime Survey, which is also
referred to as the National Crime Victimization Survey, based on the question “how safe
do you feel or would you fear being out alone in your neighborhood at night?” (Garofalo
& Laub, 1978; McGarrell et al., 1997; Skogan & Maxﬁeld, 1981; Taylor & Hale, 1986).
By being out alone at night in the neighborhood, people may fear physical violence
through personal confrontation with a stranger since it can result in injury or death
(Skogan & Maxﬁeld, 1981). Also, this situation may arouse emotional fear because of
the anxiety generated from the unpredictability of a stranger. Therefore this measure has
been considered to include most elements required for fear.

Although popular, this measure is controversial. More than anything else, it
limits types of threats. Even though this measure has been advantageous in deﬁning fear

as “a dangerous street crime in the neighborhood at night,” the narrow deﬁnition misses

59

the frequent crime “burglary” which may cause physical damage as well as property loss
(Skogan & Maxﬁeld, 1981, p.50). Rountree and Land (1996) suggest a more speciﬁc
measure of affective, personal, or emotional reactions to the possibility of being
victimized by a speciﬁc type of crime by using the question “how afraid are you of
becoming a victim of rape, robbery, murder, etc?” (p.1354). Their suggestion was
consistent with Warr (1984) and Fattah’s (1993) argument about global versus speciﬁc
fear. Other scholars criticized the NCS measure as just reﬂecting general risk assessment
rather than emotional fear. “Most people do not routinely walk alone in the
neighborhood at night and so this question reﬂects a general safety concern” by forcing
people to imagine and assess their expected fear (Haghighi & Sorensen, 1996, p.17).
Also, Skogan and Maxﬁeld (1981) argued that, without stating the term “crime,” urban
dwellers might include their fear of meeting an African American on the street at night as
fear of crime, stating that fear of crime among whites often represents fear of blacks.
Their arguments suggest, at least, that the NCS measure of fear is not independent of
other concerns. Garofalo (1981) also cautioned that there were other potential physical
harms that are not generally classiﬁed as crimes such as car accidents or natural disasters.
Another criticism is related to the quantitative assessment. Maxﬁeld (1984)
indicated that this measure concealed the true extent of fear among men due to the
tendency of men’s reluctance to show their weakness when they respond to the
quantitative scale. Another general measure of fear is used in the General Social Survey
(GSS) by asking “is there any place right around here, that is, within a mile, where you
would be afraid to walk alone at night?” (Haghighi & Sorensen, 1996). Flaws similar to

the NCS measure were raised in the GSS measure. The reference to crime is implied

60

rather than explicit, single item indicators are more error prone than multiple-item
indicators, and the variation in fear of different types of crime is not to be captured in this
way (LaGrange & Ferraro, 1989). Even with the controversy, most assume that “the
standard National Crime Survey measure captures the fear people anticipate” due to the
possibility of violent crime at night on the street (Hale, 1996; Taylor & Hale, 1986,
p.153). However, considering the criticism of the NCS item, adopting multiple items,
including more crime types, and clearly stating crime in the question, would seem to
increase the validity of the fear of crime measure.

2. Models of Fear of Crime

In the course of revealing predictors of fear, scholars have developed several
theoretical models. These models include the victimization model, the disorder model,
the community control model, the subculture diversity model, and the community context
model (DuBow et al., 1979; Hale, 1996; Lane & Meeker, 2000; Taylor & Hale, 1986).
The most traditional victimization model was related to the link between direct or
vicarious experience with crime and fear. People are aﬁaid of crime because they know
the pain of crime victimization based on their own, or their friends’ experience, or
through the media as well as their awareness that they are socially or physically
vulnerable to crime. The disorder model, based on the social disorganization and broken
window thesis, proposes that perceived disorder stimulates fear. The community control
or concern model explains that the perceptions or concerns of the deterioration of social
control in the community are sources of fear. On the other hand, the subcultural diversity
model focuses on racial and ethnic diversity as a factor predicting fear of crime, based on
the notion that people are afraid of those who are racially and ethnically different from

themselves due to the difﬁculty of understanding others’ culture.

61

More recently, methodological advancement in multilevel analysis allows
simultaneous assessment of various community conditions in the community-context
model. Different from demographics and personal experiences in the victimization
model, as well as perceptions or concerns of community issues in the disorder and the
community control/concem model, this model investigates the impact of actual indicators
of community conditions. Also, this community context model goes beyond the
subcultural diversity model and the disorder model by considering a broader range of
indicators such as crime rates, instability, concentrated disadvantage, etc. Dimensions of
direct victimization, vulnerability, and indirect victimization are personal factors in fear
of crime. In contrast, aspects of disorder, community concern or control, subcultural
diversity, and community context models are social contextual factors.

(1) Personal Factors: The Victimization Model

The victimization model is one of the oldest among the well-reported models of
fear of crime. As a common sense approach, the initial idea was a direct linear
relationship between victimization and fear. Later, the vulnerability thesis was developed
to explain the fear among the physically and socially weak. It was also recognized that
people experience crime not only by directly encountering it as a victim, but also by
learning about it indirectly from a family member, friend, neighbor, or the mass media.
Therefore, this victimization model attempts to address the effect of personal experience
of victimization, perceived social or physical vulnerability, and vicarious experiences

with victimization through stories of people they know or the media (Hale, 1996).

62

Victimization

In the direct victimization approach, researchers tended to assume that those who
experienced criminal victimization would be more fearﬁil of crime in their daily lives
because of the painful memory of physical and/or psychological harm (DuBow et al.,
1979). Being a victim may make people more cautious, but it remains unanswered
whether the experience makes people more fearful (DuBow et al., 1979; M. Lee, 1998).
It is often indicated as one of the most obvious paradoxes regarding fear of crime that
those who were victimized the most (i.e., young males), had the lowest level of fear of
crime, while those who were victimized the least (i.e., the elderly and women) had the
highest fear (Sacco, 1990). In addition to the paradox, Agnew’s (1985) explanation, that
victims may experience less fear because they tend to neutralize their experience by
learning effective defense tactics and other ways to avoid further victimization, counters
the direct victimization thesis.

Empirical evidence supporting a relationship between direct personal
victimization and fear of crime is not consistent. Prior victimization, as an individual
level factor, was important in explaining fear of crime in a number of studies (Bursik &
Grasmick, 1993; Katz et al., 2003; Skogan, 1987; Skogan & Maxﬁeld, 1981), but in other
research, victimization experience did not stimulate the feeling of fear of crime or the
relationship was very weak (Garofalo, 1979; Liska et al., 1988; McGarrell et al., 1997 ;
Rifai, 1982). Skogan (1987, p.135) concluded that “victimization affects both fear-
related attitudes and behavior in a clear and consistent manner,” while Rifai (1982)
reported that victimization had little effect on fear and daily lives of victims through a

number of case studies of burglaries and thefts. The indirect victimization model, based

63

on vulnerability or indirect experience with victimization, was developed as a potential
answer for this inconsistency.

A methodological issue regarding the measure of fear in relation to vulnerability
also deserves attention. For example, regardless of their victimization, women and the
elderly may inﬂate fear of crime in the case that they are asked about a general safety
issue rather than a speciﬁc or emotional fear especially when speciﬁc crime type is not
mentioned in the question (Smith & Hill, 1991), due to their generally high concern for
community safety. Types of threat were important. Miethe and Lee (1984) found that
prior victimization was associated with fear of violent crime but not fear of property
crime. Rountree (1998) reported that violent victimization increased both fear of
violence and fear of burglary, but burglary victimization increased only fear of burglary.

Some unexpected ﬁndings have been reported. Interestingly, Smith and Hill
(1991) found property victimization rather than personal victimization as a signiﬁcant
predictor of fear of crime. Skogan and Maxﬁeld (1981) found that while victims were
more fearful than those who have not been victimized, many of those fearful were not
actually victimized, suggesting that personal experience with crime may not explain
much. Therefore, when direct victimization is included as a predictor of fear, it would be
necessary for researchers to design a fully speciﬁed model especially controlling for
social and physical vulnerability between countering groups (i.e., age, gender, race, and
groups with different levels of socioeconomic status) and indirect experience with the

victimization (i.e., victim friend, violence in the media).

Vulnerability

 

The indirect victimization model includes the idea that social and physical
vulnerability, based on personal characteristics such as gender, age, race, and
socioeconomic status, accelerates fear of crime. Vulnerability is related to the ability to
defend, avoid, and solve the harmﬁil event when the attack occurs (Pantazis, 2000).
Physical vulnerability was generally measured based on demographic characteristics,
while social vulnerability was measured by “the actual risks faced by population groups
and by their resources for dealing with consequences of crime” (Skogan & Maxﬁeld,
1981, p.73). Three key factors of fear proposed by Killias (1990) in relation to
vulnerability included exposure to non-negligible risk, lack of effective defense, and
anticipation of serous consequences. According to this explanation, women are exposed
to sexual attacks, both women and the elderly are expected to have serious and lasting
crime consequences, and they are less able to defend themselves. Also, social
vulnerability posits that those who have risky jobs (i.e., taxi drivers, jobs with late closing
hours, prostitutes, etc.) are more likely to be exposed to risk, those without a network of
social support and adequate resources would expect serious crime consequences, and
these people tend to see difﬁculty in defending themselves due to the high levels of risk.
Similarly, it has been noted that women and the elderly are physically vulnerable because
of their difﬁculty in resisting attacks, and that racial minorities and those in a lower social
class are socially vulnerable because of their frequent exposure to crime and insufﬁcient
resources to handle the consequences (Box et al., 1988; Killias & Clerici, 2000; Taylor &
Hale, 1986). Those with less physical or social ability would be more fearful of crime

because of their expectation of the consequences when they confront an attack.

65

The bitterness of the gender speciﬁc crime, especially rape, would partly be
responsible for the high levels of fear among women since fear of sexual assault in
everyday life may shadow women’s fear of other crimes. As evidence, Killias (1990)
noted women’s higher levels of fear with higher exposure to risk in relation to sexual
attacks and Ferraro (1996) revealed that fear of sexual assault substantially increased the
explained variance in fear, especially personal crime. Others explain vulnerability in
terms of gender speciﬁc socialization. Gender role socialization creates stereotypical
female gender characteristics such as passivity, dependency, and timidity, which force
women to avoid risk-taking or toughness and eventually promote the feeling of
vulnerability (Garofalo, 1979; M. Lee, 1998; Sacco, 1990). Differential sensitivity as
well as frailty is the element to explain vulnerability among the elderly. The elderly are
more sensitive to offenses and risks than younger people (Warr, 1984), which intensiﬁes
their feeling of vulnerability. The common paradox, that despite their lower objective
risk of criminal victimization older persons show signiﬁcantly higher fear levels than
younger ones, may be due to the fact that the elderly are more sensitive to offenses and
behave in an appropriately cautious manner, thus yielding less victimization (Hale, 1996;
Pantazis, 2000). Also, the lower levels of victimization among the elderly and women
was explained in relation to routine activities since they tended to be more cautious and
reluctant to go risky places (Fattah & Sacco, 1989).

On the other hand, social vulnerability was expressed as “daily exposure to the
threat of victimization and limited means for coping with the medical and economic
consequences of the victimization” (Skogan & Maxﬁeld, 1981, p. 69). Those who live in

a disadvantaged area would be vulnerable because of their everyday exposure to crime

66

(Pantazis, 2000; Skogan & Maxﬁeld, 1981), and those individuals who lack ﬁnancial
resources to protect themselves or recover from injuries or recover properties tend to be
more fearful of crime (Hale, 1996). Minority status as well as low income and education
were often used as indicators of social vulnerability (Pantazis, 2000; Skogan & Maxﬁeld,
1981; Taylor & Hale, 1986). Minorities and those of lower socioeconomic status were
more likely to live in a disadvantaged area where disorder and crime rates were high
(Covington & Taylor, 1991). People in such a situation are more fearful of crime
because of their fear-provoking surroundings. The integration and collective efﬁcacy to
enhance community quality is usually low in these areas because of their lack of
sufﬁcient resources, thus further provoking their feelings of vulnerability (Sampson et al.,
1997a). As well, Hale (1996) indicated racism as an issue related to fear of crime among
racial minorities.

In previous studies in general, racial minorities, seniors, and females were more
likely to be fearful of crime because of their social or physical vulnerability, even though
they are, in many cases, less likely to be victimized than their counterparts such as the
racial majority, young people, and males (Kury et al., 2000; Nieuwbeerta, 2002; Skogan
& Maxﬁeld, 1981; Taylor & Hale, 1986). In one of the ﬁrst tests of the vulnerability
thesis, Skogan and Maxﬁeld (1981) found that throughout all cities in the study, women,
the elderly, blacks, and those with lower education and income tended to have higher fear
of crime than their counterparts. In this study, physical vulnerability explained more
variance in fear than social vulnerability. They also demonstrated that physical
vulnerability was strongly related to the frequency of adopting protective tactics. Their

study, therefore, provides a partial answer for the paradox. The least victimized (i.e., the

67

elderly and women) were more likely to be the most fearful and took protective actions
even though they were less likely to use those measures, all due to their vulnerability
(Skogan & Maxﬁeld, 1981). Supporting the vulnerability thesis, Killias (1990)
recognized dimensions of vulnerability as exposure to risk, seriousness of consequences,
and loss of control in relation to physical and social vulnerability. Garofalo (1979) found,
in an eight American city study, that the fear of crime was not simply a function of the
risk of and actual experiences with victimization, rather, social role expectations
represented as gender and age in particular are important factors regardless of their
objective risk of victimization, thus indicating that gender or age speciﬁc role
socialization stimulates fear. Overall, the evidence of the importance of the measures of
vulnerability in predicting fear of crime is relatively consistent and extensive (Hale,
1996), but some controversial ﬁndings are addressed in the following section.
Regarding gender, most research supported the position of a higher level of fear
among women in relation to their physical vulnerability (Covington & Taylor, 1991;
Garofalo, 1979; Haghighi & Sorensen, 1996; Maxﬁeld, 1984; McGarrell et al., 1997;
Taylor & Hale, 1986; Warr, 1984). Some others explained women’s greater fear in
relation to their lack of power in society (Stanko, 1995), or the impact of gendered
socialization process (Goodey, 1997). These studies in general have supported women’s
vulnerability thesis, just in ways mostly focusing on physical vulnerability and giving
slight attention to women’s social position. Several controversial issues have been
addressed. Goodey (1997) argued that the stereotypical image of women’s fear and
men’s fearlessness was just the socialized expression of femininity and masculinity

regardless of their actual levels of fear, suggesting that general survey methods failed to

68

capture the full nature and extent of women’s fear. Sacco (1990) also explained the
gender gap in fear in relation to the gender speciﬁc socialization that discouraged risk-
taking among women and encouraged it among men. Stanco (1995) argued that lack of
power in society generated hidden violence against women such as domestic violence and
sexual harassment, which was responsible for women’s higher levels of fear. Also, fear
of crime was predominantly fear of rape among women (Warr, 1984).

Recently, Haynie (1998), using standard opinion survey data of the General
Social Survey (1973-94) and the US. Uniform Crime Reports, revealed that although
women were much more likely than men to fear crime, the gender gap has narrowed as
men's reported fear of crime has slowly increased over time while women's has remained
stable. Also, even though gender itself was the overriding determinant of women's fear
of violent crime, social and economic factors compounded it (Pain, 1997). Through
qualitative interview data, Gilchrist, Bannister, and Ditton (1998) revealed striking
similarities between men and women regarding their fear of crime, the steps taken to
avoid crime, and the impact of crime in their everyday lives. They argued that the
stereotypical responses might be attributable to the survey methodologies employed in
previous studies. In addition to the general vulnerability thesis of gender differences in
fear of crime, researchers may have to give attention to the methodology to capture true
expression of fear as well as the gender speciﬁc socialization process.

Regarding age, Hale (1996) noted the general lifestyle of the elderly in the USA.
that “they isolate themselves from the outside world, live a life of self-imposed
conﬁnement and are captives in their own homes” (p.100), partly due to fear of crime.

The effect of age on fear of crime appeared to be less consistent or weaker than that of

69

gender. Contrary to the literature that suggested great fear among the elderly (Box et al.,
1988; Hale, 1996; Skogan & Maxﬁeld, 1981; Taylor & Hale, 1986), some scholars
reported no association between age and fear (Chadee & Ditton, 2003; Kury, Obergfell,
& Ferdinand, 2001; LaGrange & Ferraro, 1989). Yin (1982) found that crime was less
important than health and money as a source of personal concern among the elderly,
suggesting that fear of crime might not be the reason for their staying at home. Also,
throwing into question our belief that fear of crime is primarily a problem of the aged,
young women had the greatest fear of crime in Germany (Kury et al., 2001). In addition,
in the study using the General Social Survey conducted in Canada, most indicators of fear
did not suggest that people steadily grow more fearful as they age, thus suggesting that
broad generalizations about the greater fearfulness of the elderly might be exaggerated
(Sacco & Nakhaie, 2001).

It was suggested that the mixed results of age effect on fear might be related to
compounding effects of environmental conditions, routine activities, and measurement of
fear. Review of prior research on the elderly’s fear of crime by Fattah and Sacco (1989)
and Hale (1996) suggested that fear was higher in disorganized areas, fear in the elderly
might not drop after considering their lower exposure to crime, and the elderly tended to
inﬂate their fear on questions about general safety issues. Older adults did not have
higher levels of fear on both personal and property crime when the alternative speciﬁc
measures of fear instead of the standard NCS measure were used (LaGrange & Ferraro,
1989)

Through a critical assessment of the research on fear of crime among the elderly,

LaGrange & Ferraro (1987) argued that the amount of fear experienced in the everyday

70

lives of most older persons has been overstated due to the poor operationalization of fear
of crime. The elderly tended to show greater fear when they were asked about a general
safety issue rather than emotional, actual, and speciﬁc fear (LaGrange & Ferraro, 1989).
In other studies, it was suggested that the effect of age on fear might be conditional to
social status or environment. That is, old age combined with low income produced
greater fear (Baldassare, 1986). Consistent with such a result, the interaction effect
between age and urban residence was greater than the individual effects of the two
variables (Baumer, 1985; J effords, 1983). In this regard, Maxﬁeld (1984) argued that the
elderly were more fearful of crime in deteriorated areas since they were sensitive to the
presence of social and physical incivilities compared to younger people. The effect of
age deserves further tests controlling for measurement error and community conditions.
Regarding social vulnerability, much research has supported the notion that
minorities and those of lower socioeconomic status were considered to be more socially
vulnerable because of their disadvantaged contexts and limited resources either to
improve their environments or to handle consequences of crime (Box et al., 1988; F ishkin,
Rohrbach, & Johnson, 1997; Garofalo, 1981; Gomme, 1986; Hale, 1996; Skogan, 1986;
Taylor & Hale, 1986). The poor, the less educated, and blacks were more likely to fear
crime in their neighborhood than their counterparts (Erskine, 1974). All else being equal,
people with higher levels of education and income tended to be less anxious for their
safety (Grabosky, 1995). In an early study, Biderman, Johnson, McIntyre, and Weir
(1967) found that those with highest income status showed the lowest anxiety of crime
and they interpreted that people with sufﬁcient ﬁnancial resources were better able to

protect themselves from harm and to afford to live in a safe neighborhood. Also,

71

Moeller’s (1989) review of the studies of the effect of social vulnerability on fear of
crime showed that, overall, blacks expressed more concerns about crime than whites. It
was interpreted based on Brooks’s (1981) argument that blacks were more likely to be
involved in crime with other Blacks than were Whites, and they were more likely to
become victims of crime and to be fearful of crime. Consistently, Skogan and Maxﬁeld
(1981) argued that higher fear of crime among blacks and other minorities would be
explained by environmental and contextual factors of incivilities and criminogenic
conditions.

Along the same lines, it was reported that living in lower socioeconomic-status
neighborhoods increased fear (Covington & Taylor, 1991; Hale, Pack, & Salked, 1994)
probably due to residents’ exposure to disorder and crime. Burby and Rohe (1989)
emphasize the importance of environmental conditions in their survey of a random
sample of 127 residents with very similar socioeconomic status residing in 8 public
housing units, half of them located in black neighborhoods in the inner city, and the other
half outside the inner city. Compared with inner city residents, those living in less
concentrated developments were more satisﬁed with their living arrangements and less
fearful of crime. Their study was not focused on the effect of social vulnerability but it
suggested that people who were forced to live in disadvantaged areas due to their lack of
material or social resources would be more concerned with their safety. Findings
contrary to the social vulnerability thesis were also reported. In opposition to general
support to the social vulnerability thesis, some reported that race was not a signiﬁcant
factor (Gibson et al., 2002), regardless of the levels of disorder in the neighborhood

(McGarrell et al., 1997), or race was somewhat less important than has generally been

72

supposed and the socioeconomic variables, such as income and education also had minor
effects (Clementer & Kleiman, 1977). These mixed results suggest that in certain
situations, factors in social vulnerability might interact with each other or with
community conditions, reducing the utility as an independent predictor of fear.

Indirect Experience of Crime

A small number of people are actually victimized, but many experience
victimization indirectly by learning about a crime through friends, relatives, or the media.
The indirect victimization perspective explains the incongruity between fear and crime in
which fear of crime is much more widespread than actual crime, related to vicarious
experiences of victimization (Covington & Taylor, 1991). That is, those who hear from a
victim, observe a crime, or watch crime news become fearful of crime. Taylor and Hale
(1986, p.157) noted this as a “crime multiplier” in that people who hear about a crime
through conversations with victims or observations in neighborhoods become indirect
victims. Hearing of the victimization from a relative, neighbor, or ﬁiend, allows “one’s
imagination full scope” (Hale, 1996, p.105). People get impressions about the nature and
magnitude of crime in such a way. Both media and informal social networks were
indicated as sources of vicarious information about crime (Skogan & Maxﬁeld, 1981;
Taylor & Hale, 1986).

A body of research focusing on the media effect on fear has emerged asking
whether exposure to violent or crime related programs in the media stimulates people’s
fear of crime. Empirical evidence on this issue is quite mixed. Some studies reported a
positive association between exposure to violent programs or news in televisions or

newspapers and fear (Garofalo, 1979; Haghighi & Sorensen, 1996; Tyler, 1980). Using

73

two different samples, those in suburban areas and in three major US. cities, Tyler
(1980) found that crimes learned from the media signiﬁcantly increased fear of crime in
both samples. Other studies reported a different effect of television and newspapers. The
frequency of watching television news and listening to news on the radio was
signiﬁcantly related to fear of crime, yet reading newspapers and newsmagazines were
unrelated to fear of crime (Chiricos, Eschholz, & Gertz, 1997). In a study of White and
Latino residents in California, Lane and Meeker (2003) realized that for Whites, relying
on the newspaper had a negative indirect impact on fear through perceived risk, but
television had no effect. On the other hand, among Latinos, newspapers had no effect on
fear while television had both a positive direct effect and an indirect effect on fear
through perceived risk. This variation of effect based on the type of media may be
explained by the possible education disparity between Whites and Latinos since in
general those more educated tend to read newspapers more. Still others showed that
regular watching of crime dramas was weakly related to fear of crime (Dowler, 2003).
Doob and MacDonald (1979) found that signiﬁcant association between media
exposure and fear of crime disappeared when neighborhood crime was controlled,
suggesting that media messages might have an impact only in high crime areas due to the
high risk of victimization in relation to their crime information. On the other hand, Sacco
(1982) found that exposure to both television and newspapers did not inﬂuence the levels
of fear of crime. Skogan and Maxﬁeld (1981) also reported television inﬂuences to be
unrelated to fear of crime and they argued that media effect would not be signiﬁcant
since in most cases it is about crime events remote ﬁom the neighborhood, and does not

contain useful or speciﬁc information for readers to assess their own risks. In his later

74

review of literature, Tyler (1984) also observed that citizens failed to use the mass media
as a source of information about personal crime risk since they did not ﬁnd media reports
informative. In the studies of media effect on fear, the confounding role of community
conditions and education, especially when newspapers are used as an indicator, need to
be considered further.

In addition to the media, people learn about crime by talking with friends,
neighbors, and other acquaintances. People’s beliefs about the nature of crime were
greatly inﬂuenced by others who share their experiences (Maxﬁeld, 1984). Unlike the
media, getting information through local social networks is more speciﬁc, contains useful
details to assess people’s own risk of victimization, and is usually about a crime event
close to people’s residential location (Skogan & Maxﬁeld, 1981). People would be more
fearful when they know a serious crime has occurred in the neighborhood than when they
learn such a crime in a far away place and their magnitude of fear would be greater when
they know details about the crime. The thesis that geographical proximity played an
important role and learning about crime in their neighborhood was related to people’s
assessments of their personal safety is persuasive. Perceived seriousness of the offense
and perceived likelihood of being attacked were proximal causes of fear (Warr &
Stafford, 1983). Along the same lines, Skogan and Maxﬁeld (1981, p.181) found that
“urban residents learned about crime from each other concentrating on stories on crimes
of violence and knowing crime victims was related to higher levels of fear especially
when such an event occurred close to home.” Therefore, knowing a crime victim was
related to higher levels of fear, especially when those cases occurred close to home (Hale,

1996).

75

Arnold (1991) provided much support for the indirect victimization thesis by
reporting that vicarious victimization but not direct victimization was a signiﬁcant
predictor of fear of crime in all three different samples from the US. and Germany.
Similarly, Box et al. (1988) found that indirect victimization but not direct victimization
was signiﬁcantly associated with fear. Local social contacts would serve to amplify the
fear-stimulating impact of local crime. Taylor and Hale (1986) included local social
networks as positive predictors of fear of crime, arguing that “those with more local ties
will be more fearful because victimization information is transferred through the
networ ” (p.161). In this study, however, community networks signiﬁcantly increased
worry of property crime but not fear of physical harm.

Mixed results were also reported. Tyler (1980) found that crimes learned about
from others were signiﬁcant predictors of ratings of the crime problem in the city sample
but not in the suburban sample. This may be interpreted based on the contextual
difference between cities and suburban areas. People may link their crime information to
their own risk in high-crime urban areas, which may not be the case in relatively safe
suburban areas. Irnportantly, Hale’s (1996) review of literature provided a strong
argument about the link between crime rate and fear of crime in relation to indirect
victimization. People might have more crime information in high-crime areas in relation
to their risk assessment but it has often been reported that total levels of fear and crime
did not match (Haghighi & Sorensen, 1996; Hale, 1996; Skogan & Maxﬁeld, 1981;
Taylor, 2001; Taylor & Hale, 1986). It appears that the crime rate could also be
discussed along with other conditions in the community context model since not only

crime rate but also other community conditions would inﬂuence residents’ assessment of

76

their safety. As stated, however, community conditions of crime should be considered in
the investigation of indirect victimization as a predictor of fear. It may be responsible for
some mixed results in prior studies.
(2) Community Based Factors of Fear of Crime
Theory in Brief

There are many factors that tie one’s personal life to one’s local community.
Irrespective of their personal experiences or level of vulnerability, people will be affected
by neighborhood conditions (Skogan & Maxﬁeld, 1981) in as much as their lives are tied
to the community. Shaw and McKay (1942) theorized that economic hardship, instability,
and ethnic heterogeneity in neighborhoods caused a state of social disorganization
characterized by a lack of informal social control mechanisms. They understood that this
disorganization was linked to increased delinquency, thus suggesting that social patterns
of the urban environment produce social disorganization, which stimulates crime and
deviance and possibly fear of crime. Social organization involves an integration of
customs, teamwork, high morale, and bonding, which formulates a harmonious
community. Sampson and Groves (1989) measured community social organization as
local ﬁiendship networks, control of street-comer teenage peer groups, and prevalence of
organizational participation. Further developing social disorganization theory, Bursik
and Grasmick’s (1993) systemic approach to social control posits that neighborhood-level
social control is inﬂuenced through the effects of social disorganization on the private
(i.e., relationships among family), parochial (i.e., informal networks of friends and
neighbors), and public (i.e., neighborhood networks with formal agencies) dimensions.

They realized that crimes were most likely to develop in communities with limited ability

77

to regulate behaviors at the parochial (community) and public levels (Bursik & Grasmick,
1993) by decreasing community efforts to keep strangers out and to promote informal
social control. This systemic approach, therefore, embraces both informal and formal
social control dimensions and the partnership with formal authorities as well as factors
affecting crime and fear reduction.

In relation to explanatory models of fear, both perceived disorder and concern
with deterioration of social control are positively linked to dimensions of social
disorganization. Also, consistent with initial explanations of social disorganization thesis,
factors for subcultural diversity (i.e., racial composition) and community context models
(i.e., instability, crime rate, concentrated disadvantage) are causes of social
disorganization accompanying problems of disorder and concern with community
deterioration. Therefore, all these community-based models, that is, the disorder model,
the community concern model, the subcultural diversity model, and the community
context model, build on the broader social disorganization framework.

The Disorder Model

As a quality of life model, the disorder model considers perceptions of poor
quality of life and neighborhood incivilities as predictors of perceptions of increased fear
of crime. It is based on the notion that perceived social and physical disorders are
associated with fear of crime (Cole & Kelly, 1992). This perspective considers both
social and physical disorders. “Social disorder is a matter of behavior that people can see
it (public drinking, prostitution), experience it (catcalling or sexual harassment), or notice

direct evidence of it (grafﬁti, or vandalism), while physical disorder involves visual signs

78

of negligence and unchecked decay such as abandoned or ill-kept buildings, broken
streetlights, trash-ﬁlled lots, and alleys strewn with garbage” (Skogan, 1990, p.4).

According to Ross and J ang (2000) social disorder was related to situations such
as people hanging out on the streets, public drinking, drugs, and panhandling, while
physical disorder was the appearance of a community such as noise, dirt, run-down
buildings, vandalism, and grafﬁti. In general, as Skogan (1990) noted, social disorders
appear to be a series of events, while physical disorders tend to be ongoing conditions.
Also, extreme disorderly acts become crimes. Therefore, even though some researchers
tend to separate disorderly reports from other major crimes (Skogan, 1990), others were
reluctant to separate crime from disorder (Bursik & Grasmick, 1993).

In this disorder perspective, the research focus was on perceptions rather than
objective incivilities (Taylor & Hale, 1986). Perceptions of disorder are related to
people’s concern for their safety since disorder signals “physical-lack of concern about
the neighborhood, social-lack of adherence to norms of public behavior” (Taylor & Hale,
1986, p.154). Disorder in this case also represents “weakened local social control and the
attenuation of traditional norms” (McGarrell et al., 1997, p.481). Wilson and Kelling’s
(1982) broken window thesis signiﬁes the signiﬁcant role of neighborhood disorder in
community deterioration. According to the thesis, community disorder persisting in the
neighborhood is likely to further develop into serious crime. Disorder demonstrates that
people in the community are indifferent or less willing to intervene to maintain public
order such as supervising groups of rowdy teenagers or addressing physical signs of
deterioration. Recognizing this, teenagers and other possible offenders become bolder

and intensify their vandalism or even crime, which makes residents retreat due to fear.

79

The rowdy situations are developed further and attract offenders from outside the
community since they feel that the community has become a vulnerable or less risky
place for illegal activities (Bursik & Grasmick, 1993; Wilson & Kelling, 1982). PeOple
in the community perceive even higher levels of fear and begin to withdraw from
community activities or decide to move out, which eventually further deteriorates the
community.

Similarly, Skogan’s (1990) disorder and decline thesis also linked disorder, crime
and fear, and community decline. Rooted in social disorganization theory, his disorder
and decline thesis posits that neighborhood conditions such as racial heterogeneity, low
socioeconomic status, and instability cause signs of disorder, which is eventually
connected to neighborhood changes in victimization, fear, desire to leave, and ﬁtrther
declines of community structure (Skogan, 1990). More recently, modeling the link
among cohesion, disorder and crime, and fear, Markowitz, Bellair, Liska, and Liu (2001)
suggested a feedback loop in which decreases in neighborhood cohesion increase crime
and disorder, raising the level of fear, which, in turn, further decreases cohesion. All
these theses take into account that people’s perceptions of disorder in the neighborhood
stimulate not only generalized anxiety but a speciﬁc fear of crime by signaling that they
are surrounded by symbolic or actual threats, and that their neighborhood is declining
(Hale, 1996).

The thesis that perceived disorders in the neighborhood increase fear of crime has
been widely supported (Box et al., 1988; Garofalo, 1981; Gibson et al., 2002; Hale, 1996;
Markowitz et al., 2001; Maxﬁeld, 1984; McGarrell et al., 1997; Skogan, 1990; Skogan &

Maxﬁeld, 1981; Taylor & Hale, 1986). For example, Skogan (1990), in his study of 40

80

urban neighborhoods, observed the evidence of Wilson and Kelling’s proposition that
“disorder spawns more serious crime” reporting that social and physical disorder were
linked to both actual victimization and fear of crime (p.74). More speciﬁcally, the effect
of social disorder was signiﬁcant in sparking residents’ fear since “disorderly people
were unpredictable and were potentially violent” and physical decay was also associated
with fear, recognized as a“sign of crime” (Skogan, 1990, p.48). Some studies reported
different levels of effect of social and physical disorder on fear. Rohe and Burby (1988)
realized that social incivilities were stronger predictors than physical incivilities due to
their link to direct danger of being attacked by disorderly people. LaGrange, Ferraro, and
Supancic (1992) found both showing signiﬁcance as predictors of fear but they reported
social disorder as a stronger predictor of perceived risk. Irnportantly, Taylor and Hale
(1986) cautioned against the spurious relationship between perceived disorder and fear
due to social class, suggesting that researchers should consider the fact that disorder is
more commonly perceived in a lower-class neighborhood. Also, Ross and Jang (2000)
observed a buffering role of social ties in that perceived neighborhood disorder and social
ties signiﬁcantly interacted and informal social ties with neighbors reduced fear even
when people perceived disorder. That is, the effect of perceived neighborhood disorder
on fear was mediated by social ties.

Although most studies have focused on people’s perceptions of disorder at the
individual level, several scholars have attempted to investigate the effect of objective
measures of community disorder usually through observation. That is, objective
conditions of community disorder as well as perceptions of disorder were used in some

limited studies. Covington and Taylor (1991) in their study of 66 Baltimore

81

neighborhoods realized that both perceived and objective measures of disorder were
signiﬁcant predictors of fear but the effect of perceived disorder was stronger than the
objective measure of disorder. Taylor (2001) also realized that the connection between
disorder and fear was mostly in individual levels of perception instead of neighborhood
levels of objective indicators. Recently, Robinson, Lawton, and Taylor (2003) conducted
a panel study of urban residents on 50 Baltimore street blocks, interviewing two times in
one year. They found that changes in perceived incivilities accompanied changes in
resident satisfaction and fear at individual levels. At the street-block level, however,
objective disorder failed to demonstrate expected lagged impacts on fear of crime.
Interestingly, however, changing objective disorder was accompanied by changing
community satisfaction and perceptions of relative risk, but not fear.

Overall, perceptions of both physical and social disorder were signiﬁcantly
associated with fear of crime, social disorder showing stronger effects in some cases.
Also, the effect of perceived disorder was more consistent than actual indicators of
disorder in the neighborhood, implying that there would be contextual or other factors
that affect symbolic interpretation of objective disorder even in the same neighborhood.
The Community Concem/Control Model

The community concern/control perspective proposes that fear rises as concerns
of the neighborhood increase. Following the social disorganization tradition,
neighborhoods high in social disorganization are less able to exert social control over
residents, and provide fewer opportunities for residents to obtain needed social support
(Sampson et al., 1997), which would increase concern with community problems and

reduce feelings of safety. Crime instigates fear when residents’ perceptions of

82

community dynamics developed into concem for the community (Taylor & Hale, 1986).
Box et al. (1988) conceptualized neighborhood concern as a perception that the
neighborhood is declining economically, and eventually environmentally. Garofalo and
Laub (1978) explained that the sum of anxieties of public deviance is the basis for the
concern with community, and since fear reﬂects urban unease (Garofalo & Laub, 1978),
those who are concerned with neighborhood deterioration are more likely to fear crime.

This model also explained fear by way of social control dimensions. Fear is not
just a consequence of a subjective response to victimization, it is a consequence of the
erosion of social control (Ross & J ang, 2000). Neighborhood social control is “the ability
of neighborhoods to control themselves and their environment through formal and
informal relational networks so that the risk of crime is minimized” (Bursik & Grasmick,
1993, p.12). Rooted in systemic neighborhood control theory (Bursik & Grasmick, 1993),
this perspective proposes that formal, informal, and private level social control dynamics
are important in addressing crime and fear related issues. That is, reduced collective
efﬁcacy and increased social disorganization progress in both internal dynamics and
external ties (i.e., public control), which increase concerns for safety (Taylor, 2001).
This perspective is closely related to the disorder perspective since disorder represents the
deterioration of local social control. Along this line, Taylor and Hale (1986) proposed a
general path model in which perceptions of disorder as well as individual and
environmental characteristics were linked to community concern, altogether predicting
fear.

In prior studies, this perspective took into account concerns of neighborhood

deterioration (Taylor & Hale, 1986), erosion of local social control (Lewis & Salem,

83

1986), and social instability and moral decline (Garofalo & Laub, 1978), as facilitators of
fear. In a reverse direction, social ties or social integration (Gibson et al., 2002; Lewis &
Salem, 1986), neighborhood collective efﬁcacy (Gibson et al., 2002), and ties to local
power structure, government responsiveness, or conﬁdence in the local police (Box et al.,
1988; Lewis & Salem, 1986; McGarrell et al., 1997) were considered as inhibitors of fear
of crime. That is, community quality was maintained either by local customs, norms and
informal sanctions, or by ofﬁcial rules and regulations usually through government
institutions. Perceptions of deterioration of these informal and formal social control
mechanisms were considered serious community concerns that increase fear of crime.

Much attention was given to informal social control dimensions in the belief that
perceptions of neighborhood integration conceptualized as the capacity of a community
to exert social control over its members and even strangers would enhance feelings of
safety (Bursik & Grasmick, 1993). It was suggested that strong relational ties tended to
formulate higher feelings of safety and less fear (Baumer & Hunter, 1979; LeBailly &
Gordon, 1981). Also, collective efﬁcacy, conceptualized as “social cohesion among
neighbors combined with their willingness to intervene on behalf of the common good”
(Sampson et al., 1997, p.918), was considered as an agent of informal social control
(Gibson et al., 2002).

Skogan and Maxﬁeld (1981, p.120) tested two different measures of social
integration, “cognitive and sentimental.” Cognitive integration indicated people’s
awareness of their communities and knowledge of their prominent features, while
sentimental integration represented emotional attachment, identiﬁcation, positive

evaluations, and other affective components. Their thesis was that fear would be low

84

when people knew about the areas of danger and safety and when community members
expressed their willingness in their collective efforts to solve community problems based
on their knowledge about local community conditions. As expected, both residential
commitment and integration into local community systems were signiﬁcant predictors of
reduced fear. McGarrell, et a1. (1997) realized that perceptions of informal social control,
social support, and integration were related to lower levels of fear. Consistent with
theoretical positions, they found that the more one perceived one’s neighbors responding
to rowdy youths, and the more one felt rooted in the community, the less one felt fear.
Along the same lines, Box et a1. (1988) found that people who perceived that their
neighbors helped one another experienced lower fear. Similarly, research has reported
that the effect of social integration on fear was mediated by collective efﬁcacy, deﬁned as
the trustworthiness of neighbors and their willingness to intervene as informal social
control agents (Gibson et al., 2002).

Some contrary results were also reported for the proposed association between
community concern and fear of crime. LeBailly and Gordon (1981) investigated the
effects of four forms of community involvement: neighborhood bonds measured as
feeling of attachment to the locality, residential ties, social interaction with neighbors,
and use of local facilities. Social interaction with neighbors and use of local facilities
were not signiﬁcant predictors of fear of crime even though neighborhood bonds and
residential ties to the community were signiﬁcantly related to the reduction of fear of
crime. Katz et al (2003) attributed these mixed results to the wide range of variation in

the conceptualization and the lack of simultaneous investigation of the various

85

dimensions of community concern on fear. Overall, much support for the community
concern in informal social control dimensions and fear was evidenced.

The community concern perspective also takes into account perceptions of public
social control in the community. Public control refers to the “ability of the community to
secure public goods and services that are allocated by agencies located outside the
neighborhood” (Bursik & Grasmick, 1993, p. 1 7). The agencies with the most obvious
relevance to crime control are law enforcement agencies. Conﬁdence in the police and
perceptions of government responsiveness have been recognized as two such dimensions
(Box et al., 1988; McGarrell et al., 1997; Velez, 2001). In addition, the recent popularity
of community policing is related to this theoretical development. One of the important
goals of community policing strategies includes the reduction of fear of crime through
diagnosing and managing problems in the community, fostering closer relationships with
residents to facilitate mutual support, and building self-defense capabilities in the
community (Trojanowicz, Kappeler, & Gaines, 1998). That is, police activities are
expected to control local problems either directly by working closely with the
neighborhood or indirectly via helping build informal social control abilities among
residents (Trojanowicz, 1985).

Conceptualizing public social control as community ties to the local government
and the police, Velez (2001) reported that public social control reduced household and
personal victimization risk. Similarly, McGarrell et a1 (1997) found that perceptions of
government responsiveness as well as informal social control in the neighborhood were
important predictors of fear. It was also realized that people were less likely to fear crime

if they believed the police were effective in crime prevention and clearance as well as

86

handling calls for services (Baker, Nienstedt, Everett, & McClery, 1983). Box et a1
(1988) also found that those who were conﬁdent in the effectiveness of the police for
prevention and apprehension showed less fear. Also, people were less fearful of crime
when they believed their neighbors would call the police if they noticed grafﬁti painting
in progress (Covington & Taylor, 1991). The effect was greater in disadvantaged
neighborhoods. “Public social control yielded greater beneﬁts in neighborhoods with
prevalent structural disadvantage than in neighborhoods with more afﬂuence” (Velez,
2001, p.858).

Similarly, strong support for these assumptions was reported in neighborhood foot
patrol program (Troj anowicz, 1985), and in the comparison between community policing
neighborhoods and neighborhoods without elements of community policing (Cole &
Kelly, 1992). In contrast some studies have found no signiﬁcant effect in the assessment
of the community contact program (Bennett, 1991) and in citizen perceptions of
community policing activities (Dietz, 1997; Scheider et al., 2003).

The Subcultural Diversity Model

The subcultural diversity perspective has recently gained more attention in
explaining fear of crime. Community conditions as stimulators of social disorganization
include racial heterogeneity along with low socioeconomic status and residential mobility,
which are proposed to be linked to crime and fear via social disorganization (Sampson &
Groves, 1989; Shaw & McKay, 1942). Rooted in this key point of the social
disorganization theory, the subcultural diversity perspective posits that the difference
between one’s ethnicity and that of surrounding neighbors stimulates fear because of the

uncertainty and distrust coming from the difﬁculty of understanding values, attitudes, and

87

norms of people with different cultural backgrounds (Katz et al., 2003; Lane & Meeker,
2000). That is, the manners and behaviors of “others” are difﬁcult to interpret, which
lead to unease and fear (Katz et al., 2003, p.105). Since fear results from living near
persons who have different cultural backgrounds, the extent of cultural or racial diversity
in the neighborhood is related to levels of fear (Merry, 1981). Skogan and Maxﬁeld
(1981) noted that expression of concern about crime by many Caucasians was assumed to
be rooted to a signiﬁcant degree in their fear of African-American people. Merry (1981)
also observed that Chinese residents were afraid of African-Americans because they
could not understand the loud and unruly life style of African Americans in comparison to
the quiet and Confucian culture of China. Therefore, racial heterogeneity was a major
step toward explaining fear according to this perspective focusing on the variation of
subculture in different racial and ethnic groups.

Research based in the subcultural diversity perspective generally have reported a
signiﬁcant link between racial dissimilarity or heterogeneity and fear of crime (Bennett &
F lavin, 1994; Covington & Taylor, 1991; Katz et al., 2003; Lane & Meeker, 2000),
suggesting the conﬂict of various backgrounds and cultures as a facilitator of fear of
crime. Katz et al. (2003) found that concern about diversity had a signiﬁcant impact on
both general fear of crime and fear of gangs, suggesting that policy-makers need to
reduce the public’s fear of persons who were viewed as different from them to reduce fear
of crime. Liska, Lawrence, and Sanchirico (1982) suggested that whites were less fearful
where blacks and whites lived separately, and for nonwhites, fear of crime was inﬂuenced
by the presence of other nonwhites. The authors concluded that cultural beliefs and

stereotypes linking nonwhites to crime increased fear not only for whites but for

88

nonwhites as well. Covington and Taylor (1991) also reported that subculture differences
among people living in the same neighborhood increased fear. Also, they suggested that
the link between diversity and fear would be processed in the racially changing
neighborhoods but not in the integrated ones, suggesting the importance of historical
background of communities rather than simply exploring ethnic diversity alone. Racial
prejudice was also considered important. St. John and Heald-Moore (1996) found that
prejudiced Whites were more fearful than nonprejudiced Whites when they encountered a
Black stranger.

However, Katz et al. (2003) cautioned that studies of race based in the subcultural
diversity perspective could be misspeciﬁed because of its focus on the interpretation and
perception of the behavior of “others” instead of the image or symbolism that “others”
represent. They argue, based on conﬂict theory, that the majority group members may be
threatened by the symbolism of the presence of the minority populations, but not the
personal perceptions of different culture. Their arguments are consistent with Blumer’s
(1958) group position perspective of race prejudice in that a symbolic sense of group
position based on structural power relationships is more important than feelings of
individual members of different racial groups in explaining race-based prejudice and fear.

The Community Context Model

 

Community based predictors of fear of crime include perceptions of disorder,
concerns of community social control, and different subcultures among racial or ethnic
groups. Left out are other community conditions. These community conditions include
the crime rate, concentrated disadvantage, and community instability, which are

understood in the community context model. As social disorganization theory suggests,

89

community characteristics such as the economic gap, instability, and ethnic heterogeneity
in neighborhoods are negative outcomes of urbanization (Sampson & Groves, 1989;
Shaw & McKay, 1942). These negative conditions in an urban society lead to
disorganized social values and norms, therefore providing crime-prone environments and
stimulating offenders, all of which are responsible for crime and fear in the community.
Pioneered by Park, Burgess, Shaw, and McKay, the Chicago school inﬂuenced the
early work on the effect of social conditions (i.e., economic disadvantage, population
turnover) on social disruption (Vold, Bernard, & Snipes, 1998). Dividing the city of
Chicago into ﬁve concentric zones, Park and Burgess (1925) explained that crime and
delinquency burgeoned in the most disorganized area, Zone H, characterized as the area
immediately around zone I, the central business district. Supporting this evidence, in
their analysis of Chicago neighborhood characteristics and ofﬁcial records of juvenile
delinquency, Shaw and McKay (1969) found that neighborhoods in transition showed the
highest levels of delinquency. They explained the result based on the reduced capacity
for social control and the weakened community social organization in urban communities.
They recognized that neighborhoods characterized by economic disadvantage, residential
mobility, and population heterogeneity were less likely to produce social organization and
informal social control due to the lack of resources, time, and cultural commitment.
Unlike the overemphasis on dimensions of disorganization of early social disorganization
perspectives, the systemic model emphasizes social control based on social networks
such as ﬁiends, family, and institutions (Bursik & Grasmick, 1993), as a mechanism to
increase residents’ capacity for social control. Simply put, social ties among family

members, neighbors, and local institutions enhance residents’ ability to control juveniles,

90

recognize strangers, and engage in guardianship behavior. Developing inquiries of Shaw
and McKay (1942; 1969), Sampson and Groves (1989) reported that neighborhood
cohesion mediates the relationship between social disorganization and crime. They
explained the chain in that economic disadvantage, residential mobility, ethnic
heterogeneity, and family disruption generated social disorganization (i.e., sparse local
ﬁiendship networks, unsupervised teenage groups, and low organizational participation)
which in turn, increased crime and delinquency in communities. More recently, Sampson
et a1 (1997) found that collective efficacy, the combination of social cohesion and
informal social control, had the potential to reduce and or reverse the effect of social
disorganization on crime and fear of crime.

In addition, new urban sociology proposes that socioeconomic status and racial
composition are important in situations for social control in the community. Rooted in
conﬂict theory, this perspective posits that communities low in economic status and high
in racial minorities would have fewer resources and less political power for community
development (Sampson & Wilson, 1995; Taylor, 2001; Wilson, 1996). Fewer public
service investments will be made in these communities compared to the afﬂuent
neighborhoods of a dominant race, thus hindering positive development and social
organization in the community. Related to this perspective, Bursik and Grasmick (1993)
emphasize the importance of the social control processes through the connection with the
external coalitions and public agencies, suggesting that some communities are better
connected than others and more effective in their social control on a parochial level.
These theories suggest that community conditions are important in relation to community

development and reduced violence and fear. Therefore, community-based models of fear

91

should include these indicators of community conditions, without restricting their scope
to a few factors (i.e., disorder, racial composition).

In general, this community context approach differs from other community-based
models (i.e. the disorder model, the community concern model, subcultural diversity
model) in that the unit of analysis is not the individual (micro—level) but the neighborhood
(macro-level) and in that the scope is beyond disorder and racial heterogeneity. The
interests of the prior community based models are how neighborhood conditions are
interpreted by various individuals instead of how such conditions directly inﬂuence
people’s fear of crime (Taylor, 2001). Closely related to the disorder and the subcultural
diversity model, this community context perspective holds that community conditions
such as crime rates, concentrated disadvantage, and residential stability would stimulate
people’s fear of crime (Cancino, 2002; Rountree & Land, 1996; Sampson et al., 1997a;
Taylor, 2001). Ever since the broken window thesis (Wilson & Kelling, 1982), disorder
has been considered to be a main accelerator of community decline. However, Taylor
(2001) demonstrated that the magnitude of the impact of disorder on community
deterioration was no more so than other community conditions such as high poverty rates
and high initial crime rates. Sampson et a1 (1997) also demonstrated a signiﬁcant
association between variations in violence and three community conditions: concentrated
disadvantage, immigrant concentration, and residential stability. Concentrated
disadvantage was measured as the proportion of residential income below the poverty
line, the proportion of households on public assistance, the proportion of female-headed
families, the proportion of persons over 16 who are unemployed, the proportion of the

population that is African-American, and persons under 18 per square mile (Sampson et

92

al., 1997, p.920). This study clearly demonstrated that models of fear of crime as well as
crime itself should be broadened to include a wide range of indicators of community
conditions.

Methodologically, research in this community context approach has combined
citizen level surveys, census data for community characteristics, and police statistics for
crime rates. Techniques for multilevel data analysis inevitably consider how community
conditions are linked to fear of crime directly and indirectly through individual level
perceptions (i.e., perceptions of disorder, community concern), providing the sound
structure of this community context perspective. Such a design is consistent with the
social disorganization theory proposing the association between neighborhood conditions
and crime, and possibly fear, which is designed to be indirect via social disorganization in
the neighborhood (Bursik & Grasmick, 1993; Shaw & McKay, 1942).

Multilevel research on fear of crime, based on this perspective, is relatively new
and limited compared to investigations of the effect of community conditions on crime
and delinquency. Relying on block to block ﬁeld studies conducted in Baltimore in 1980
and 1994 and analysis of data collected in some ofﬁcial statistics, Taylor (2001) found
that high poverty rates, high initial crime rates, racial composition, neighborhood
exchange value, and home ownership were strong predictors of fear. He also observed
that these structural factors were stronger than disorder in predicting fear, thus suggesting
the position of theories and policies on neighborhood development would need to be
furthered from the dominant focus on disorder to consider other factors such as economic

well being as a long term strategy to reduce fear and crime.

93

By using the data from a victimization survey and annual police statistics in
Seattle, Rountree & Land (1996) demonstrated the importance of social context as well as
individual level factors in understanding increased levels of fear of crime. In this study,
crime rates, measured by tract-level burglary rates, had important direct positive effects
on perceived crime risk. In a non-metropolitan setting, using survey data from 1,125
citizens nested in 31 residential units located in Michigan, Cancino (2002) demonstrated
that citizens from residential units with higher concentrations of property crime report
higher levels of perceived fear. Also, using multilevel statistical analysis for the
community, crime, and health data in Illinois, Ross, Mirowsky, and Pribesh (2001) found
that residents of disadvantaged high crime neighborhoods have a higher sense of
powerlessness to avoid the threat. Based on the perspective of this model, therefore,
where one lives affects one’s overall perceptions of fear of crime.

(3) Implications for Research

Predictive models designed to ﬁnd factors of fear of crime have been addressed
and research evidence has been reviewed. As stated, inconsistency in the measure of fear
of crime was partly responsible for the mixed results of some models of fear. It is
necessary to distinguish cognitive fear (or risk assessment) and emotional fear as well as
global fear and crime-speciﬁc fear. Taylor and Hale (1986) indicated that outcome
measures have been limited focusing on the popular NCS format, which raised
methodological weaknesses discussed above. As an effort to overcome these
shortcomings, Ferrero (1995) proposed a multi-item, speciﬁc, and emotional measure of
fear of crime. That is, offense-speciﬁc measures were suggested to measure speciﬁc

types of crime (Ferraro, 1995; Ferraro & LaGrange, 1987; Warr & Stafford, 1983).

94

Among several models of fear of crime, the community concern/control model
appears to be less developed than others. Community concern based on the public,
parochial, and private levels of social control needs attention, especially in relation to
police-initiated endeavors for community safety. The popular debate over the efﬁciency
of community policing in fear reduction could be addressed in this regard. The mixed
results of community policing efforts on fear encourage further research. Box et al.
(1988) reported that conﬁdence in the police was an important predictor of fear of crime.
In addition, the community policing thesis emphasizes that residents can address
problems of crime and safety by securing ties with the police (Kelling & Coles, 1996;
Troj anowicz, 1985), suggesting that the police could become an important ally. Hence,
police-initiated endeavors to improve neighborhood quality and safety may be studied
simultaneously with public satisfaction with police to clarify which of those dimensions
are linked to reduced levels of fear of crime. In addition, the western ideology of
community policing was introduced as a form of reform programs in many developing
countries, but the effectiveness of such programs has not been tested empirically in those
countries. It will build on our current knowledge to empirically identify the role of
community - policing activities on fear of crime especially in the contexts that recently
introduced Western forms of community policing.

In addition, the community context model was constructed to include community
conditions that were not considered in prior community based models (i.e., the
community concern model, the disorder model, the subcultural diversity model). These
community conditions included crime rates, concentrated disadvantage, and community

instability. Taylor (2001) reported that these community conditions were stronger

95

predictors of fear than the popular predictor, disorder. In this regard, he suggested that
resident groups working in close collaboration with police and other city agencies and
targeting needed resources to the community would be more important than strictly
regulating order maintenance. To further imagine the magnitude and direction of the
effect of the overall community conditions on fear, these community characteristics need
to be simultaneously modeled along with disorder and individual level factors.
Additionally, larger social contexts need to be taken into account for the explanation of
fear. Based on the theory’s deﬁnition of community, a multilevel approach on these
community issues has given attention to small jurisdictions, narrowing the focus to
micro-level communities or neighborhoods (Lauritson, 2001). Emphasis on micro units
of community is important since community is a place where residents share basic social
life and norms (Bursik & Grasmick, 1993; Wooldredge, 2002). Researchers, therefore,
have given attention to small units defending their choices on the basis of within-group
homogeneity in street block groups, tract groups, or census tracts that are similar on
various census indicators (Wooldredge, 2002). Yet, the aspects of social contextual
approach deserve further testing in bigger units. In a cross-national study of criminal
victimization in 19 countries, for example, researchers realized that the chance to become
the victim of crime was not only determined by the opportunities and conditions
associated with the neighborhood’s capacity to exercise social control, but also by aspects
of larger social structures such as districts, cities, or countries (Van Wilsem, De Graaf, &
Wittebrood, 2000).

All of the predictive factor models reviewed so far could be designed in a

testable causal model. Some relatively early research attempted to explain the causal link

96

connecting predictors and fear of crime. Garofalo’s (1981) general model proposed both
causes and consequences of fear of crime in a causal path from the position in social
space to social outcome via fear of crime. According to the model, socioeconomic
structure decides a position in social space (i.e., different life style for daily activities,
vocation, leisure, etc.) The position in social space inﬂuences the perceptions of the
amount and the nature of crime to which the person is exposed, which inﬂuences their
risk of victimization assessment based on the information obtained about the prevalence
of crime, likelihood, physical vulnerability, and consequences of possible injuries. This
risk assessment eventually shapes actual and anticipated fear. Also, the level of fear
generates responses such as avoidance and protective behavior, etc. Again, these
responsive behaviors have some feedback effects, for example, on risk assessment by
decreasing the risk through avoidance or protection. Eventually, this process, from the
position in social space to responsive behavior, is linked to both positive and negative
outcomes in society. Skogan and Maxﬁeld (1981, p.17) proposed another causal but very
complex model suggesting multiple predictors such as “personal and household
vulnerability, city of residence, neighborhood conditions, victimization experience, media
exposure, personal communication networks, knowledge of events, knowledge of
victims,” which are all structurally related to fear of crime that eventually generates
responsive behavior (e. g., role constraints).

Garofalo’s (1981) and Skogan and Maxﬁeld’s (1981) models were designed with
the intention to capture the causal link among predictors, fear of crime, and personal and
social consequences. However, they do not fulﬁll an important element of theory, which

is testability (see Lee, 1998; Taylor and Hale, 1986). These two general models

97

exacerbate the difﬁculties of empirical tests by not connecting together different
propositions and not providing a testable range of design. To overcome this limitation,
Taylor and Hale (1986) designed three separate causal models in a testable range: the
victimization, the disorder, and the community concern models. Among the three causal
models, the community concern causal model is more comprehensive than the others
because it includes community conditions and perceived disorder and community
concern in one causal dimension. Taylor and Hale’s (1986) three separate models,
however, beg the question of how to link these models to test them simultaneously.
Ferraro (1995) proposed one of the most comprehensive models in a testable format (see
Figure 1). His model considered both macro and micro conditions as well as perceived

risk and behavioral adaptation as causal predictors of fear.

 

Ecological (Macro)\
Crime prevalence
Community traits

Neighborhood Traits _> Behavioral Adaptations
/' lncivility Constrained action
Cohesion Defensive action

\ Perceived Risk————> Fear

Status characteristics/

Victimization
Residential traits

 

 

Figure 1. Risk Interpretation Model (Ferraro, 1995)
Ferraro (1995) explained how macro and micro level factors shape individual
perceptions, risks of victimization, behavioral changes, and eventually fear of crime.

Adding to the three models by Taylor and Hale, Ferraro’s model included risk assessment

98

and behavioral adaptations as major factors. F erraro’s (1995) model is more inclusive
than Taylor and Hale’s (1986) three separate models by framing all predictors together.
Lee (1998) extended F erraro’s works by furthering the predictors at both macro and
micro levels and successfully investigated factors of fear of crime among Korean
Americans residing in metropolitan Chicago, but the basic structure of Lee’s (1998)
model was the same as Ferraro’s risk interpretation model. While Ferraro’s model
signiﬁcantly enhanced our knowledge on the causal factors of fear of crime, there appear
to be some ways to improve this model. First, it is possible to explain all predictors in
theoretical or conceptual perspectives. As described in this chapter, scholars have
recognized the community context model, community concern/control model, disorder
model, and victimization model in explaining fear of crime. The causal frame is more
theoretically parsimonious when all of the variables in the causal process are considered
in conceptual or theoretical models that have been developed to explain theoretical links
between these factors and fear. Second, different from previous causal models, Ferraro
(1995) included behavioral adaptations as one of the factors of fear instead of considering
it as the result of fear. In the risk interpretation model, therefore, behavioral adaptation
plays the role as reaction to perceived risk and as a predictor of fear. This design was
based on a reciprocal relationship between fear and constrained behavior such as an
escalating loop exists (Liska, Sanchirico, & Reed, 1998). While such a design helps
explain how constrained behavior further enhances fear, the impact of fear on varying
behavioral adaptations cannot be investigated even though this has been one of the major
interests in studies of fear of crime (see Garofalo, 1981; Skogan & Maxﬁeld, 1981).

Earlier causal models considered behavioral adaptations as the negative result of fear

99

This trend is in contrast to the analysis of urban life in major US. cities where poverty is
clustered right next to the business district. Park and Burgess (1925), for example,
identiﬁed ﬁve concentric zones radiating out from the center of Chicago. Unlike in Seoul,
the zone in transition characterized by poor residents, population turnover, and

abandoned buildings was located right next to the central business district in Chicago.
This difference in the process of city development in Seoul, compared to major US.

cities, may generate unique patterns in the effect of district-level predictors on fear of
crime in Seoul.

Table 1. Administrative Districts, Population, and Sample Size in Seoul

 

 

 

Gu Dong Tong Population Sample
(District) (Community) (Sub-Community)
1. Gangnam-Gu 26 988 536,031 40
2. Gangdong-Gu 21 766 479,270 32
3. Gangbuk-Gu 17 368 362,094 25
4. Gangseo-Gu 22 543 539,673 28
5. Gwanak-Gu 27 723 526,971 25
6. Gwangjin—Gu 16 530 388,659 13
7. Guro-Gu 19 576 419,438 24
8. Geumcheon—Gu 12 383 267,355 26
9. Nowon-Gu 24 945 633,934 45
10. Dobong-Gu 15 512 378,166 12
1 1 . Dongdaemun-Gu 26 608 386,814 24
12. Dongj ak-Gu 20 615 413,204 26
13. Mapo-Gu 24 666 383,629 27
14. Seodaemun-Gu 21 552 361,754 14
15. Seocho-Gu 1 8 732 400,220 27
16. Seongdong-Gu 20 489 343,929 26
17. Seongbuk-Gu 30 565 456,535 25
18. Songpa-Gu 28 979 623,267 40
19. Yangcheon-Gu 20 489 489,257 42
20. Yeongdeungpo-Gu 22 695 410,952 19
21. Yongsan-Gu 20 473 240,723 13
22. Eunpyeong-Gu 20 669 476,843 41
23. Jongno-Gu 19 402 181,441 10
24. Jung-Gu 15 337 138,798 10
25. Jungnang-Gu 20 662 438,01 1 40
Total 522 15,267 10,276,968 654

 

Note: The source of the Census data is the Korea National Statistical Ofﬁce (2005).

109

 

As described in Table 1, there are 25 autonomous “Gu” and these were under the
control of the Seoul Metropolitan government, but have been functioning as self-
goveming administrative units since July 1995. Under the control of 25 Gu (district),
there are total 522 sub-units of “Dong (community)” that provide close, ﬁrst-hand
services for the residents. As sub—units of “Dong,” there are several thousand “Tong
(sub-community)”

At the request of the KIC, a professional private research company in Seoul
conducted door to door interviews with a sample of randomly selected citizens residing in
the city. The survey strategy here is described based on Kim’s (2003) explanation. A
stratiﬁed sampling method was utilized to develop a representative sample of Seoul
citizens. There are 25 Gus, 522 Dongs, and 15,267 Tongs in the city. To begin, 50
Dongs were selected. The number of selected Dongs in each Gu varied from one to three
depending on the population size of the Gu. As the next step, two Tongs were randomly
selected from each Dong. The total number of selected Tongs in 50 Dongs therefore
total 100. The research administrators then chose six or seven households in each of the
100 Tongs and one family member of each household was interviewed. The ﬁrst criteria
for the selection of interview subjects was citizens between the ages of 20 and 70 years
old. The survey administrator then decided one citizen in each household in a way to
consider the even distribution of gender (50 percent of male and 50 percent of female)
and age (20 percent for each of twentieth, thirtieth, fortieth, and ﬁftieth and over). After
selecting the sample based on the telephone book and census data, the trained company
employees contacted selected citizens in all 25 Gus in Seoul. The survey was based on

door-to-door interviews. Respondents answered the survey by themselves in the presence

110

2. Data

To assess the effects of social conditions and individual level attitudinal
dimensions on fear of crime, data were collected from three independent sources in Seoul,
South Korea. These data sources include: 1) community survey data, 2) census data, and
3) ofﬁcial police crime records. The collection procedures are described in the following
subsections.

(1) Survey Data

In Seoul, the Korean Institute of Criminology (KIC) directed the design and
collection of the survey data. The participants in the study were citizens over 19 years
old in Seoul, the capital city of South Korea. In order to provide insight into the
geographical composition of Seoul, a brief description is provided based on information
from the city’s ofﬁcial publication (Seoul City Government, 2005). The ﬁfth largest city
in the world, the Seoul megalopolis has a population of about 10.3 million, which is
about a fourth of the total population of South Korea. The total area of Seoul is 605.52
square kilometers, or 0.6 percent of the entire country. The Han River divides the city
into two parts: northern (Gangbuk) and southern (Gangnam).

Figure 3 presents population density and Figure 4 presents population living in
poverty in Seoul. The comparison between patterns of density and poverty indicates that
the process of city development in this city is different from that of major US. cities.
The density is intense except in the business district in the center of the city and the areas
close to the city border (see Figure 3), while the poverty is clustered around the city
border (see Figure 4). It indicates that wealthy citizens are residing in inner city areas in

SeouL

107

explain not only fear, but also perceived risk and behavioral adaptations. Structurally,
perceived risk was considered as a major mediator between conceptual models and fear.
Unlike Ferraro’s risk interpretation frame, behavioral adaptation was considered as the
reaction to fear as well as perceived risk. Every variable is understood in conceptual or
theoretical perspectives and the hypothesized direction between fear and behavioral
adaptation is straightforward in this causal frame. Interrelationships among exogenous
variables were also speciﬁed in this revised model, but they may not be tested in a small
sample situation, a possible limitation of this study.

Finally, it is not clear if models of fear of crime are applicable in culturally
distinct settings due to the lack of sufﬁcient empirical research to test these models
beyond US. or Western contexts (Some exceptions in Eastern Asian context include
Curran & Cook, 1993; Ito, 1993; Lee, 1997). Limited research implied that cultural
differences might generate some unique patterns in Eastern Asian cultural context
especially regarding age (Curran and Cook, 1993), and perceptions of informal and
formal social controls (Lee, 1997). Tests of these models in a comparative context may
help extend our understanding of the ability of these models to explain fear of crime.

This chapter described controversies regarding the concept of fear, reviewed
conceptual factor models, and clariﬁed research directions based on the knowledge drawn
from the review of literature. To further investigate these research inquiries, the next
chapter speciﬁes the hypotheses to be tested, the methodological design, and identiﬁes
and defends the measurements of the independent and dependent variables. Hierarchical
Linear Modeling (HLM) is also discussed in the next chapter as the major analytic

strategy.

101

(Garofalo, 1981; Skogan & Maxﬁeld, 1981), indicating that behavioral adaptations

occurred as responses to fear as well as perceived risk. People constrain their actions or

ad0pt defensive actions to deal with not only perceived risk but also fear (Skogan &

Maxﬁeld, 1981), and a passive life pattern among community members further

deteriorated the community by decreasing liveliness and integration (Kelling & Coles,

1996; Moore & Trojanowicz, 1988). It will build on our knowledge to investigate the

impact of fear on different types of life patterns by considering behavioral adaptations as

the response to fear as well as perceived risk.

 

Community Context
Mobility
Economic disadvantage
Crime prevalence

Perceived Risk

 

 

 

 

. it

 

Community
Concern/Control

Community policing

Perceived cohesion

Neighborhood watch
Disorder

Perceived incivility

 

Satisfaction with police

 

 

Fear

 

 

 

 

 

 

 

t it

 

 

 

 

Victimization
Direct
Indirect
(vicarious, media)
Vulnerability

 

  

v v

Behavioral Adaptation
Constrained actions
Cautious actions
Active defense

 

 

 

 

 

 

Figure 2. A Revised Causal Model

Figure 2 presents a revised causal frame for this study. All predictors of fear are

included in four conceptual models in this causal frame. These four conceptual models

100

CHAPTER FIVE
METHODOLOGY

Following up on theoretical frames of fear of crime and the revised version of the
risk interpretation model, this chapter speciﬁes a series of research hypotheses to be
tested. A detailed outline of the methodological procedures including a description of the
sample and operationalization of variables is also provided in this chapter. In addition,
this chapter provides an overview and justiﬁcation of the analytic strategies that will be
employed. Prior studies have recognized four different conceptual models applicable to
fear of crime in Korean context: the victimization model, the disorder model, the
community context model, and the community control model. This study tests these
models in a causal frame in which risk assessment, fear, and behavioral adaptations are
all predicted by these conceptual factors. In addition, perceived risk is considered as a
mediator between these conceptual models and fear, and behavioral adaptations are
designed as the response to fear as well as perceived risk. It is expected that, in addition
to their direct effects on fear, conceptual factors increase perceived risk, and this
eventually enhances fear. Then, both perceived risk and fear would force citizens to
adopt behavioral adaptations. In addition, two of the conceptual models, the community
context model and the community concern/control model, are major foci of this study,
thus yielding two principal hypotheses. First, based on the extension of the community
context model, those who live in a district with higher economic disadvantage, higher
mobility, higher crime, and higher population density are likely to perceive a higher level
of fear of crime than those who do not, holding constant the effect of demographics and
other predictors of fear of crime. Second, based on the community concern/control

perspective, it is expected that citizen perceptions of community-policing activities will

102

be related to fear of crime. Also, the inclusion of some control variables is essential to
control for spurious effects in assessing fear of crime. This study therefore includes other
factors of fear of crime such as length of residency, volunteer activities for crime
prevention, and cross level interactions.

1. Hypotheses

(1) Social Conditions and the Community Context Model

As discussed, economic disadvantage and residential instability facilitates social
disorganization while difficulties of informal social control in communities contribute to
an increase in crime and violence and a decrease in the sense of quality of life (Sampson
& Groves, 1989; Shaw & McKay, 1942). The mechanism of community deterioration is
that social disorganization and the reduced ability of a community for informal social
control increases social and physical disorder and stimulates fear of crime which in turn
further deteriorates the community (Skogan, 1990). Poor neighborhoods lack social,
political, and economic resources to build an organized community. Mobility, as an
indicator of population turnover, reduces the ability to build mechanisms of social
organization, which leads to crime and fear. Individuals in high crime areas are likely to
perceive their vulnerability and tend to express fear due to crime in their neighborhood
(Taylor, 2001). Finally, population density is recognized as a direct indicator of crime
(Osgood and Chamber, 2000) or indirect factor via inhibiting community collective
efficacy (Sampson and Raudenbush, 1999). It is expected that citizens in areas of higher
population density are likely to higher levels of fear, considering the positive effect of
crime on fear. Scholars have recognized, therefore, that aggregate levels of economic
disadvantage, mobility, crime, and population density serve as predictors of fear of crime.

The following relationships are hypothesized:

103

Hla: Citizens in districts of higher mobility are likely to express higher levels of
fear of crime.

Hlb: Citizens in districts of higher economic disadvantage are likely to express
higher levels of fear of crime.

ch: Citizens in districts with a higher crime rate are likely to express higher
levels of fear of crime.

Hld: Citizens in districts with a higher population density are likely to express
higher levels of fear of crime.

(2) Community Policing and the Community Concern/Control Model

Community concern/control dimensions include both informal and formal social
control dimensions. Perceptions that public social control agents are reliable and
community based crime prevention efforts are on the right track enhance the sense of
safety and reduce fear of crime (Moore & Trojanowicz, 1988). In addition, the notion
that neighbors are willing to support each other reduces fear of crime (Gibson et al.,
2002). Irnportantly, community-policing practice was recognized under this model and it
is expected that citizen perceptions of community-policing activities will be related to
fear of crime. The following relationships are hypothesized:

H2a: Citizens who perceive community policing activities in their neighborhood
are likely to express lower levels of fear of crime.

H2b: Citizens who are more satisﬁed with police are likely to express lower levels
of fear of crime.

H2c: Citizens who perceive higher levels of community cohesion are likely to
report lower levels of fear of crime.

H2d: Citizens who have a neighborhood watch team in their community are likely
to express lower levels of fear of crime.

(3) The Victimization Model

104

The victimization model of fear of crime posits that direct and indirect

victimizations predict fear of crime. In addition, the vulnerability thesis is also

understood as a dimension of indirect victimization. Those who are less able to defend

themselves and those who have less resources to handle the consequences of crime tend

to be more afraid of crime (Smith & Hill, 1991). Thus, the following relationships are

hypothesized:

Direct Victimization:

H3a: Citizens who have experienced victimization in the past are likely to express
higher levels of fear of crime.

Indirect Victimization:

H3b: Citizens whose family members, relatives, or ﬁiends have been victimized
in the past are likely to express higher levels of fear of crime.

H3c: Citizens who feel the media cover crime issues over other issues are likely to
report higher levels of fear of crime.

Vulnerability:

H3d: Citizens with lower socioeconomic status are likely to express higher levels
of fear of crime.

H3c: Female citizens are likely to express higher levels of fear of crime than male
citizens.

H3f: Older citizens are likely to express higher levels of fear of crime than
younger citizens.

(4) The Disorder Model

Disorder is a result of social disorganization and a symbol of the deterioration of

community social control. Addressing disorder, as suggested by the “ﬁxing broken

windows” thesis has been considered an important measure to reduce both crime and fear

of crime. Studies have recognized both social and physical disorders as predictors of fear

105

of crime (Kelling & Coles, 1996; Skogan, 1990). The following relationship is
hypothesized:

H4: Citizens who perceive higher levels of physical and social disorder in their
neighborhoods are likely to express higher levels of fear of crime.

(5) Perceived Risk, Fear, and Behavioral Adaptations

Perceived risk of victimization produces a variety of outcomes including
behavioral adaptations and fear of crime (Ferraro, 1995). Fear of crime eventually forces
citizens to adopt constrained and cautious actions (Skogan and Maxﬁeld, 1981). Based
on the revised causal frame in Figure 2, the question of how perceived risk affects
behavioral adaptation and how fear inﬂuences behavioral adaptation is tested here. The
following relationships are hypothesized:

H5a: Citizens who perceive higher risks of victimization are likely to express
higher levels of fear of crime.

H5b: Citizens who perceive higher risks of victimization are more likely to adopt
constrained actions, cautious actions, and active defense.

H5c: Citizens who perceive higher levels of fear are more likely to adopt
constrained actions, cautious actions, and active defense.

H5d: Citizens who perceive higher risks of victimization are likely to express

higher levels of fear of crime which is likely to increase their constrained actions,

cautious actions, and active defense.

A series of research hypotheses were speciﬁed above focusing on the predictors
and the results of fear of crime. Even though all hypotheses are designed regarding fear
since the focus of this study is investigating factors and results of fear, the analyses will

include tests of the associations between the conceptual factors of fear and perceived risk

as well as behavioral adaptations.

106

 

Figure 3. Population Density in Seoul

 

Figure 4. Population living in Poverty in Seoul

108

Table 2. Description of Variables
Wariable Name
ENDOGENOUS VARIABLE
Fear of crime (a)
Perceived risk (a)
Constrained Actions (b)
Cautious Actions (b)
Active Defense (b)
EXOGENOUS VARIABLE
Community Context Model
Economic disadvantage (b)
Mobility (b)
Crime (b)
Population Density (0)
Community Concern/Control Model
Perceptions of community policing (a)
Satisfaction with police (a)
Community cohesion (a)
Neighborhood Watch
Victimization Model
Direct victimization
Indirect victimization
Indirect experience
Crime on media
Vulnerability
Socioeconomic status (b)
Female
Age
Disorder Model
Perceived incivility (a)
CONTROL VARIABLE
Length of residency
Volunteer activities
Cross level interactions

Description

8 survey items
8 survey items
5 survey items
5 survey items
2 survey items

Percent living in poverty and rented house
Percent moved out and percent moved in
(crime/population)*100,000 for 9 types
Log (Population/kmz)

6 survey items
7 survey items
5 survey items
Dummy coded (Yes=1)

5 survey items: Dummy coded (Yes=1)

8 survey items: Dummy coded (Yes=l)
Dummy coded (Over covered=l)

Education, income, housing
Dummy variable (l=female)
Age in years

7 survey items
In years and months

Dummy coded (Yes=1)
see description

 

Note: (a) Factor based additive scale.
(b) Weighted factor score.
(c) Natural log.

Perceived Risk

 

As a measure of risk assessment, a perceived risk was a factor-based scale

corresponding to the respondents’ ratings of the general risk victimization for each of

eight types of crime:

Street robbery

114

computers in the Korean National Police Headquarters. Nine types of crime data for the
entire year in 2004 were obtained.

3. Measurement in the Study
(1) Endogenous Variables

Fear of Crime

 

The major outcome variable is fear of crime. As discussed, the popular measures
of fear of crime research, especially those used for the National Crime Victimization
Survey and General Social Survey, were criticized due to the vague nature. Speciﬁcally,
they do not consider types of crime and emotional stimulus. Recognizing this
shortcoming, Ferraro (2005) introduced a crime speciﬁc measure of fear in which
respondents are asked to rate how fearful they would be being a victim of twelve
different types of crime largely classiﬁed into personal and property crimes. This
measure considers the emotional state of fear by including the phrase “how fearful” or
“how afraid.” This measure is also advantageous because it asks how individuals feel in
everyday situations instead of hypothetical situations of walking alone at night. The
measure used in this survey, therefore, follows F erraro’s (1995) speciﬁc measure of fear.
Respondents were asked to rate how afraid they would be in their everyday life regarding
these eight speciﬁc crimes:

Having somebody break into their home and steal money or valuables while they are
away

Having somebody break into their home and take money or valuables by threatening
or assaulting them and their family

Being pick-pocketed or purse-snatched on the street

Being robbed or mugged on the street

Being assaulted by a gangster or a stranger (except sexual assault)

Being sexually assaulted by a stranger

Having somebody kidnap or abduct them or their family
Having somebody break into their home and hurt their family

112

The response categories were provided based on the four point scale “not afraid at all, not
afraid, afraid, very afraid.” Exploratory factor analysis with Varimax rotation method
conﬁrmed that these items were associated with a single latent construct (Eigenvalue =
4.93, percent of variance = 61.63, factor loadings > .73). The magnitude of internal
consistency (alpha=.91) indicated a high level of reliability among items. Fear of crime,
therefore, was operationalized as both a factor based scale and a weighted factor
regression score. A factor based additive scale was considered over a weighted factor
regression score to allow for a meaningful interpretation of various results. A weighted
factor regression score was also used in case standardized score is necessary for outcome
variables for direct comparison purposes. It was checked, however, that the analyses
using the weighted factor regression score did not show any substantial difference from

the results of the analyses employing this factor based additive scale.

113

of an interviewer or they were assisted by the interviewer in case it was necessary due to
age or other reasons. For example, the survey was read and ﬁlled in by the interviewer
based on the answers from the respondents when the subjects preferred such a method.
The survey was conducted for 20 days during the summer of 2003 and a total of 654
responses were collected by the use of this method.
(2) Census and Crime Statistics

In addition to the survey data, census statistics and crime data were obtained to
measure district level social conditions. Since the citizen survey was conducted late in
2003, district-level statistics for the entire year of 2003 were considered to represent the
district-level environments at the time of the survey response. Crime, however, was
measured based on police statistics in 2004. In 2004, the Korean police introduced a new
place-based crime recording system in which crimes were recorded based on the place it
occurred, contrary to the previous system focusing on police stations that handled those
crimes regardless of where they occurred. The place-based crime statistics were
considered more accurate than those in the previous system. Furthermore, crime statistics
were classiﬁed based on government jurisdictions in addition to police jurisdictions from
the year 2004. For this study, it was necessary to classify both census statistics and crime
data based on government jurisdictions because the survey was conducted in all
government districts in Seoul.

In this study, social context is deﬁned as districts (Gu) in Seoul. Census statistics
to measure district level economic disadvantage, mobility, and population density for the
entire year in 2003 were gathered from the Korean Statistical Ofﬁce data archive. Police

statistics regarding crime rates were obtained from the ofﬁcial statistics recorded in

111

Sexual assault by a stranger
Gang assault
Theft

Burglary

Kidnapping, abduction

Pick-pocketing, purse-snatching

Auto theft (including car accessories and other valuables)
Respondents were asked to indicate their assessment of the risk of victimization for each
type of crime based on the scale “1: (very low), 2: (low), 3: (medium), 4: (high), 5: (very
high)” Factor analysis and reliability test were employed to check the latent construct
(Eigenvalue = 4.83, percent of variance = 60.44, factor loadings > .72) and the internal
consistency (alpha = .91).
Behavigad Adaptations

Some of the important hypotheses for this study were that citizens adopt
behavioral adaptations as their response to perceptions of risk and fear. To test the
hypotheses, three different types of defensive actions were operationalized: constrained
actions, cautious actions, and active defense.

These three variables were measured based on weighted factor regression scores.
There were several survey items for various types of defensive actions. It was necessary
to combine these items into scales to address the issue of multicollinearity and to build
reliable variables. Factor analysis with oblique rotation method was used to guide this
process. Each item incorporated a four-point scale ranging from never to always. Table
3 illustrates that these survey items are associated with three latent constructs, thus
yielding a clear three factor solution (Eigenvalue = 3.41, 1.70, 1.53). Five survey items

under factor 1 were associated with constrained actions for self-safety. Mean of items

(1.97) indicated that citizens sometimes adopted constrained actions for self-safety such

115

as avoiding dangerous places, avoiding going out at night, not possessing cash or
valuables, etc. Five survey items under factor 2 were associated with cautious actions.
Mean of these items (2.11) indicated that citizens sometimes adopted cautious actions
such as looking windows, leaving the light on while they are away, asking neighbors to
watch their homes, etc. Two survey items under factor 3 were associated with active
defense for self-safety. Mean of these items (1.29) indicated that citizens rarely adopted
active actions for self-safety such as carrying a self-defense weapon, learning a self-
defense tactic, etc.

Table 3. Rotated Factor Matrix of Behavioral Adaptations.

I Variables Factorl FactorII FactorIII ]
PL / PS (a) PL / PS (a) PL / PSfa)

 

 

Constrained Action (Mean = 1.97*)

I avoid dangerous places even when it is time consuming .79/ .76 .06/ .24 -.21/—.03

I avoid going out after 10 pm so not to be victimized .60/ .67 .06/ .29 .26/ .40

I ask one of my family members to pick me up .64/ .71 -.01/ .26 .31/ .45
when I come home late

I ask someone to accompany me when I go out .74/ .77 -.05/ .21 .19/ .35
late at night

I do not carry much cash or valuables .76/ .66 -.08/ .07 -.33/-.18

Cautious Action (Mean = 2.11)

Ialways lock windows -.06/ .10 .58/ .55 -.06/ .06

I keep my cash or valuables in a safe -.08/ .16 .64/ .67 .22/ .34

I leave the light on at home when I go out .30/ .11 .70/ .72 -.O6/ .12

I ask my neighbor to watch my home when I’m away -.07/ .15 .71/ .70 .06/ .21
for one or two days

I arrange newspapers and milk not to be delivered .09/ .27 .73/ .72 -. l 6/ .02

to my home when I’m away for several days

Active Defense (Mean = 1.29)

I carry a self-defense weapon such as gas gun .04/ .22 -.04/ .16 .85/ .85
or electronic shocker
Ileam self-defense tactics -.07/ .10 -.01/ .15 .80/ .78
Eigenvalue 3 .41 1.70 l .53
% of Variance 28.35 14.19 12.72

 

Note: Principal Component Analysis of extraction and Promax with Kaiser Normalization rotation method.
(a) Pattern Loading / Pattern Structure.
"' 1:Never, 2: Sometimes, 3:Often, 4: Always

116

Weighted factor regression scores were generated for these three factors. Item means
indicated that citizens were most likely to adopt cautious actions, followed by constrained
actions, and active defense.
(2) Exogenous Variables
Social Conditions

In addition to citizen-level measures, the current study also considered the impact
of social conditions that are suggested to be correlated to fear of crime. In this study,
social context is deﬁned as districts (Gus) in Seoul. Social conditions are operationalized
using four variables: economic disadvantage, mobility, crime, and population density.
Previous research has demonstrated that residents of disadvantaged, high-mobility, and
high-crime areas expressed higher levels of both crime and fear of crime than those in
relatively afﬂuent, high-stability, and low-crime areas (Lewis & Maxﬁeld, 1980;
Sampson, Raudenbush, & Earls, 1997; Skogan, 1986). Community-level population
densities were also recognized as a predictor of both crime (Osgood & Chambers, 2000;
Rasko, 1979), and subsequently fear of crime. It was expected that the situation would be
the same with a context larger than areas or communities, in this case, districts. Census
data included percent of individuals living in poverty, percent rented house, percent
moved in, percent moved out, population, and size of each district. Crime statistics
included murder, robbery, rape, larceny, assault, arson, drug offense (i.e., buying, selling,
using), abduction, and gambling. Table 4 presents the loading pattern from the factor
analysis among district-level variables. Three factors with eigenvalues greater than 1.0
were extracted. Given the loading pattern for three factors, three variables were

constructed: economic disadvantage, mobility, and crime.

117

Table 4. Factor Pattern for Social Condition Variables (N = 25)

 

 

Wariable Factor I Factor H Factor III
ECONOMIC DISADVANTAGE
Percent Rented House .297 .170 £41
Percent living in Poverty .017 -.404 w
MOBILITY
Percent Moved In -.154 Q -.1 19
Percent Moved Out .069 3% .158
CRIME
Murder (b) ._5_2_0 -.227 -.329
Robbery (b) L7 .277 .027
Rape (b) :81 .128 .246
Larceny (b) ._8_7_2 -.013 .230
Assault (b) ._lo -.092 .229
Arson (b) E -.249 .081
Drug offense (b) ._8__1_ -.121 .230
Abduction (b) ,6_7_ -.037 -.236
Gambling (b) ,912 -.028 .046
Eigenvalue 6.351 2.029 1 .400
% of Variance 48.855 15.605 10.773

 

Note: Factor loadings greater than .50 are underlined.
(b) Number of crimes per 100,000 persons.

Economic disadvantage: This was a weighted factor regression score that includes
percent below poverty level and percent renter-occupied housing units. These two census
statistics for the year 2003 were obtained from the Korean Statistical Ofﬁce. Items were
selected referring to those designed by guiding theorists in the ﬁeld (see Sampson et al.,
1997). Some Census statistics available in the US. are not necessarily available in Korea.
For example, census items such as percent black, percent female-headed household are
available in the US. but not in Korea, and so these census items could not be included.
Considering each item’s differential contribution to concentrated disadvantage, a factor
regression score was calculated to weight each variable by its factor loading.

Mobility was measured as a weighted factor regression score that included the

following 2003 census items: percent moved in and percent moved out. It was based on

118

the percent of people moved in and moved out during the year 2003. Percent moved in
and percent moved out were calculated based on the number of citizens moved in or
moved out divided by population and multiplied by 100.

Crime: This is an important characteristic to gauge the district-level social
condition. This variable was a weighted factor regression score based on nine-types of
crimes. In Korea, crimes were generally classiﬁed as ﬁve-index crimes such as murder,
rape, robbery, larceny, and assault. More recently, crimes are often classiﬁed as nine
types adding arson, drug offense (i.e., buying, selling, using), abduction, and gambling.
These crimes are slightly different from index crimes in the United States. The
difference may come from the variation in societal recognition of crime as serious and
frequent as well as the discrepancy in crime deﬁnition between the two countries. In the
US, there are eight types classiﬁed as index crime: murder, rape, robbery, aggravated
assault, burglary, larceny, motor vehicle theft, and arson. The same deﬁnition is used for
murder and rape in both countries, although assault in Korea is deﬁned as both
aggravated and non-aggravated assault. In addition, burglary is classiﬁed into either
larceny or robbery in Korea. When someone enters a building with the intent to commit
a felony or to steal valuable property, the terms “entering and larceny” or “entering and
robbery” are used instead of “burglary” for legal process and punishment based on
Korean criminal law. These crimes, however, are classiﬁed into either robbery or larceny
in the police crime statistics, showing a discrepancy of the deﬁnition between statistical
classiﬁcation and legal procedure. For example, if someone enters a house and steals
valuable property, it is classiﬁed into “entering and larceny” for legal procedure but

“larceny” in police data archive. On the other hand, if she or he enters a house and

119

threatens people to take valuable property, it is classiﬁed into “entering and robbery” for
legal procedure but “robbery” for statistical purposes. In each category (i.e., robbery,
larceny), however, the speciﬁc method of crime such as entering and robbery, or street
robbery, are explained in police statistics. Statistics of nine index crimes for the year
2004 were obtained from the Korea National Police Agency. Exploratory factor analysis
conﬁrmed that these nine types of crime are associated with a single latent construct
(Eigenvalue = 6.12, percent of variance = 67.99). Factor loadings ranged from .52 to .93.
Considering each item’s differential contribution, a factor regression score was calculated
to weight each crime by its factor loading.

Population density: Another indicator of social condition, population density was
measured based on population divided by square kilometer for each district. This
variable, therefore, measures number of residents per square kilometer. Due to the
varying distribution of population from district to district, a natural log transformation
was employed for this variable.

CommunityConcern/Control

Four variables were considered under this model: perceived community-policing
activities in the neighborhood, satisfaction with police, community cohesion, and
neighborhood watch.

Perceived community-policing activities: This variable was a factor-based scale.
Citizen perceptions of community-policing activities in the neighborhood are measured
based on citizens’ perceptions of police activities related to community oriented policing
strategies. Considering the deﬁnition of community policing, operational dimensions

were classiﬁed into three categories: community partnership, problem solving, and

120

organizational change (Dietz, 1997; Trojanowicz et al., 1998). Among these three
dimensions, police efforts for community partnership and problem solving were directly
visible activities in the community. Six items were designed to reﬂect such activities:
The police in my community let residents participate in such activities as planning
or evaluating crime prevention activities
The police in my community often promote a friendly meeting with residents
The police and residents work together in a high crime district
The police in my community visit residents’ houses or business places
to provide helpful advice for crime prevention and to promote public relations
The police in my community ask residents for active cooperation
The police in my community provide helpful information for crime prevention
through a letter or neighborhood meeting
Respondents were asked to rate the police practices described above based on the four
point scale “strongly disagree, disagree, agree, strongly agree.” Factor analysis and
reliability test were employed to check the latent construct (Eigenvalue = 3.85, percent of
variance = 64.10, factor loadings > .74) and the internal consistency (alpha = .89).

Public satisfaction with police: This variable was a factor based-scale. Various
aspects of police practice were considered regarding public satisfaction with police in
seven items:

I trust the ability of the police in crime prevention

1 can rely on the police for the safety of my family

The police are the best experts regarding crime issues

The police do their best for crime prevention

The police are sincere and kind

The police are prompt and proper in handling a report by a citizen

The police are fair in handling a report by a citizen
Respondents were asked to rate police practices described above based on the ﬁve point
scale “strongly disagree, disagree, neutral, agree, strongly agree.” Factor analysis and

reliability tests were employed to check the latent construct (Eigenvalue = 4.06, percent

of variance = 58.01, factor loadings > .70) and the internal consistency (alpha = .88).

121

Community cohesion: Perception of cohesiveness among neighbors was a factor
based scale measuring ﬁve items:

Residents in my neighborhood help each other with difﬁculties

Residents in my neighborhood know about each other

Residents in my neighborhood talk about community issues

Residents in my neighborhood go to community events and meeting

Residents in my neighborhood share what they have
Respondents were asked to rate residents’ cohesiveness based on the four point scale
“strongly disagree, disagree, agree, strongly agree.” Factor analysis and reliability test
were employed to check the latent construct (Eigenvalue = 3.67, percent of variance =
73.36, factor loadings > .83) and the internal consistency (alpha = .91).

Neighborhood watch: A neighborhood watch team organized among residents
was considered an important component for informal social control in communities. This
was a dummy variable (yes=1) measured based on the question “does your community
have a neighborhood watch team organized among residents for crime prevention?”
Victimization

The victimization model is considered the most traditional model to explain fear
of crime. In addition to direct victimization, indirect experience and vulnerability are
included under this model (Hale, 1996; Taylor & Hale, 1986). Three dimensions are
considered under this model: direct victimization, indirect victimization (indirect
experience and media effect), and vulnerability (SES, age, gender).

Direct victimization: This was a dummy variable measure based on whether the
respondents have ever been victimized in each of the ﬁve types of crime described below:

Street Robbery

Sexual Assault by a Stranger
Assault by Gang

122

Burglary
Kidnapping, Abduction

All these crimes inﬂict physical harm that would instill fear. Respondents were asked to
indicate their experience based on the scale “1: (never), 2: (1-2 times), 3: (3-4 times), 4:
(5-6 times), 5: (7 times or more)” Factor analysis and reliability test were employed to
check the latent construct (Eigenvalue = 2.54, percent of variance = 50.75, factor
loadings > .64) and the internal consistency (alpha = .73). Scholars of victimization
caution aggregating different types of crimes, yet exploratory factor analysis showed that
these crimes could be aggregated as a single measure. The factor based scale was
extremely positively skewed with the majority having not experienced victimization, and
so a dummy variable was considered over an additive scale or weighted factor regression
score to help avoid the skewing effect.

Indirect victimization: This dimension includes variables of indirect experience
of victimization and media effect. Indirect experience was a dummy variable measure
based on the question of whether the respondents have family members or relatives who
have ever been victimized for the eight types of crimes (street robbery, sexual assault by
a stranger, assault by gang, theﬁ at home, burglary, kidnapping, abduction, pickpocketing,
snatching, and auto theft including car accessories and other valuables). In addition to
crimes related to physical harm, those types related to property were also included.
Respondents were given ﬁve scales to choose from “1: (never), 2: (1-2 times), 3: (3-4
times), 4: (5-6 times), 5: (7 times or more)” As in the direct victimization measure,
factor analysis and reliability tests were employed to check the latent construct

(Eigenvalue = 3.85, percent of variance = 48.07, factor loadings > .66) and the internal

123

consistency (alpha = .82). The factor based-scale was dummy coded to help avoid the
skewing effect in the multivariate analysis.

As a second component of indirect victimization, the media effect was measured
based on if the respondents think newspapers or broadcasting stations cover crime issues
over other issues such as politics, economy, or culture. It was measured based on the ﬁve
point Likert type scale ranging from strongly agree to strongly disagree. For the
multivariate analysis this scale was dummy coded (strongly agree and agree = 1,
otherwise = 0). This variable, therefore, differentiated those who agreed or strongly
agreed to the question about the over-coverage of crime on media from those who
thought otherwise.

Vulnerability: this dimension includes variables for physical vulnerability
(gender, age) and social vulnerability (socioeconomic status). Gender was measured
based on respondents’ self-classiﬁcation of their gender as male or female. Age was
measured by respondents’ description of their age in years. Socioeconomic status was a
weighted factor regression score based on three survey items: education, monthly income,
and housing situation. Education was measured in six categories such as elementary
school or under, middle school, high school, 2-year college, 4-year college, and graduate
school. Monthly income was measured based on the following ten categories: less than
$1,000; $1,000-$1,499; $1,500-$1,999; $ 2,000-$2,499; $2,500—$2,999; $3,000-$3,499;
$3,500-$ 3,999; $ 4,000—$ 4,499; $4,500-$ 4,999; and more than $5,000. Housing
situation was measured based on respondents’ self report about their homeownership
such as monthly rented room, long-term rented room, long-term rented house, and owned

house. Factor analysis showed that these three survey items (education, monthly income,

124

and housing situation) were associated with a single latent construct (Eigenvalue = 1.53,
percent of variance = 51.12, factor loadings > .60).
Disorder

Perceived incivility: As a measure of citizen perceptions of neighborhood quality
of life, the variable “perceived incivility” was a factor based scale. To reﬂect both
physical and social disorder in the neighborhood, ﬁve related survey items were
considered:

There are many dark and ignored places in my neighborhood

There are some places where delinquent juveniles get together

My neighborhood is dirty with rubbish

I oﬁen see groups of juveniles wandering around

I oﬁen see drunken people wandering around at night

Each item was measured based on the four point scale “strongly disagree, disagree,

agree, strongly agree.” Factor analysis and reliability test showed that these survey items
were associated with one latent construct (Eigenvalue = 2.94, percent of variance = 58.77,
factor loadings > .68) and with a high internal consistency (alpha = .82).
(3) Control Variables

Volunteer activities: Additionally, control variables were included to help control
for potential spurious associations. First, scholars suggest that volunteer activities for
crime prevention were associated with levels of fear of crime. People may volunteer for
crime prevention activities to deal with perceived risk, but the nature of the relationship
between volunteer activities and fear of crime is controversial. One study reported that
volunteers participating in community-policing activities expressed fear of crime

substantially more than ordinary citizens (Zhao et a1, 2002). The authors interpreted that

participation in crime prevention activities might heighten the suspicion of others or fear

125

could be a motivator for citizens to volunteer for crime prevention activities. In another
study, however, patrol participation was related to a slightly lower fear of crime (Ronald
& Dean, 1985). To control for this effect, therefore, the variable “volunteer activities”
was measured based on whether the respondents have done any of the crime prevention
activities described below:

Member in a neighborhood watch team organized among residents

for crime prevention

Member in a community team for proper guidance of delinquent juveniles

Member in a community advisory team for crime prevention

Donated money to support crime prevention activities

Participated in a campaign to improve a community situation for public safety
Respondents were asked to answer based on the scale “no, no but I would like to do it,
yes.” Factor analysis and reliability test were employed to check the latent construct
(Eigenvalue = 2.50, percent of variance = 50.04, factor loadings > .59) and the internal
consistency (alpha = .75). Since factor based scale was extremely skewed toward
positive direction, this scale was dummy coded (yes = 1, otherwise = 0).

Length of Residence: Length of residence was recognized as another factor of fear
of crime. Whereas the stability of the neighborhood is a measure of community
environment, length of residency is a measure of personal situation for familiarity and
intimacy in the neighborhood, which may act as an inhibitor of fear of crime. Prior
studies showed that those who lived longer in the neighborhood expressed lower levels of
fear of crime due to their familiarity in the community and intimacy with residents (Paul,
Dennis, & Frank, 1983). To test this hypothesis, the variable “length of residence” was
measured based on the question “how long have you lived in your present

neighborhood?” This variable was measured based on the self—expression of respondents

of their length of residence in years.

126

Cross-level interactions: Finally, not many researchers have given attention to the
effect of interaction among predictors of fear. Box et a1. (1988) found that previous
victimization was a negative predictor of fear of crime, but it turned positive when it
interacted with incivility. The authors interpreted those who had been victimized might
take more precautions and be less fearful, but this did not work in areas of high incivility
since victims might have to do more to take precautions in their environment of danger
and threats. Also, in other studies, the interaction effect between age and urban residence
was greater than the individual effects of the two variables (Baumer, 1985; J effords,
1983). Others considered the interaction effect through subgroup analysis. McGarrell et
al. (1997) conducted a model assessment in each of the low-, medium-, and high-disorder
neighborhoods to examine whether the role of victimization, perceived disorder, and
community concern vary depending on levels of disorder in a neighborhood. They did
ﬁnd different effects of factors across those neighborhoods. The positive effect of
victimization was signiﬁcant in medium- and hi gh-disorder neighborhoods, but not in
low-disorder neighborhoods, suggesting an interaction between prior victimization and
neighborhood disorder. The review of literature revealed that these important interaction
dimensions were not considered in a majority of studies. To control for the effect of
district-level conditions and individual-level perceptions the following interactions were
considered: mobility and perceived incivility, economic disadvantage and volunteer
activities, and crime and perceived risk. All of these interactions are assumed to have
positive effects on fear of crime, considering their individually positive inﬂuences on fear.

4. Analytic Strategy

To begin, a correlation matrix for citizen-level variables will be estimated to

check bivariate relationships as well as to provide a preliminary assessment of

127

multicollinearity. Next, multivariate analyses will proceed in two stages. First,
Hierarchical Linear Modeling (HLM) analysis for fear of crime is conducted to
investigate the effect of social conditions on fear of crime while holding constant the
effect of individual level factors such as demographics and citizen perceptions of
neighborhood quality and police service. HLM provides full-information maximum-
likelihood estimates and allows for the simultaneous testing of the district-level and
citizen-level predictors. Different from the conventional multiple regression analysis,
this method provides statistical indexes of within- and between-group variation.

Second, turning to the structural approach to examine the mediating effect of
perceived risk on the relationship between conceptual models and fear as well as the
effects of fear on behavioral adaptations, additional HLM analyses are conducted for
perceived risk, fear, and, behavioral adaptation measures. The comparison of HLM
models between perceived risk and fear reveals the structural association between them.
Finally, the comparison of HLM models between fear and behavioral adaptation
measures (i.e., constrained actions, cautious actions, active defense) reveals the effects of
fear and perceived risk on each of the behavioral adaptation measures.

5. Hierarchical Linear Modeling (HLM) Method
(1) HLM and Traditional Approach
Hierarchical Linear Model (Raudenbush & Bryk, 2002), Multilevel Model

(Goldstein & Rasbash, 1996), or Random Coefﬁcient Model was chosen as method of
analysis for this study based on theoretical and statistical grounds. HLM is a particular
regression technique that is designed to take into account the hierarchical structure of

data. Illustrating a multilevel approach for survey data of citizens nested in communities,

128

this section brieﬂy overviews the strength and weakness of the HLM analysis, compared
to the conventional methods.
(2) Data Structure for HLM

Hierarchical Linear Modeling (HLM) is an estimation technique speciﬁcally
designed for the analysis of multilevel data where individuals are nested or grouped
(Raudenbush & Bryk, 2002). Researchers oﬁen encounter a multilevel structure in social
science. Citizens are nested in communities or cities, students are nested in schools, and
public ofﬁcials are nested in agencies. In such a structure, researchers are interested in
assessing the effect of level-2 predictors on level-1 issues, controlling for level-l
predictors. Community environment, school circumstance, and organizational culture or
policies are possible level-2 factors predicting level-1 measures such as fear of crime,
juvenile delinquency, and job satisfaction among public ofﬁcials. Researchers expect a
certain amount of between level variation controlling for individual level variation that is
explained by demographics and personal perceptions of and preferences to certain issues.
In the multilevel studies of fear of crime, we usually have survey data for citizens. We
also have data about different climate characteristics of the communities or cities in
which these citizens reside. We are interested in determining simultaneously the effect of
community characteristics as well as various citizen demographics and attitudinal
dimensions on citizen fear of crime. Prior to the multi-level estimation techniques, there
were some ways to handle multilevel data structure. In general, conventional regression
methods were used after aggregation or disaggregation of the multilevel data, handling
the data as if they were single level data (Goldstein & Rasbash, 1996; Raudenbush &

Bryk, 2002; Snijders & Bosker, 1994).

129

(3) Traditional Approach

First, as the most frequently used method, data are disaggregated to lower level
(i.e., citizen level). Various community level variables such as crime rate, divorce rate,
and mobility rate were assigned to each citizen. That is, all observations are pooled
(Raudenbush & Bryk, 2002). Citizens’ community membership or between community
variation is ignored. In such a case, for example, citizens nested in the same community
will share exactly the same community characteristics on the contextual variables. All
citizens in a given community would have the same level of crime, mobility, etc. Then,
citizens are used as the unit of analysis for conventional regression analyses. In this case,
units for analysis increase but data from fewer units at the higher level is falsely handled
as if it were data for the individual level. For example, mobility rate in a given
community is falsely considered as individual mobility tendency for each citizen, which
causes ecological fallacy by generalizing group-level observations to individual citizens
(Schwartz, 1994). Furthermore, traditional OLS regression analysis would assume that
all citizens, even those coming from the same community, are considered to be unrelated
to each other, which is the independence of observations. In a hierarchy of social
contextual design, as citizens were grouped within communities, however, their
perceptions would be related according to their speciﬁc communities. Thus, it is
considered that the community characteristics are completely homogenous for all citizens
in a speciﬁc community when we use such pooled data. The complete homogeneity
assumption is very unrealistic in a real situation. Citizens would share common
experiences to some degree given structural environments but their experience would

vary due to the speciﬁcs of each individual. In other words, this disaggregated or pooled

130

data approach violates the assumption of independence among observations within
groups and heterogeneity of variance assumption, resulting in biased estimation of
variances in the contextual variables and incorrect standard errors (Green, 2000).
Assigning group level variables down to the individual level results in statistical tests that
are based on the number of individuals instead of the number of groups, thus the standard
errors associated with the tests of the group level variables may be underestimated. More
speciﬁcally, the size of standard errors are underestimated in a statistical test, if we
disregard heterogeneous structure, which would provide a false sign of statistical
signiﬁcance, generating a risk for a Type I error by accepting false hypotheses (Green,
2000). Scholars also caution that even wrongly signed coefﬁcients will be obtained in
some situations (Raudenbush & Bryk, 2002).

Second, data are aggregated to a higher level (i.e., community level). For
example, citizen fears of crime are averaged to the community level and community
jurisdictions are used for the unit of analysis (Goldstein & Rasbash, 1996; Raudenbush &
Bryk, 2002; Snijders & Bosker, 1994). Coefﬁcients are estimated as if the aggregations
are the primary object of interest (i.e., aggregated fear of crime measures for each
community are modeled), therefore ignoring within group variation, yet in fact, citizens
nested in a community would have different levels of fear of crime. In this case, fewer
units at the community level suppress more units at the citizen level generating fewer
units for analysis, which results in weakness of statistical power for analysis and loss of
information. This approach also carries a risk for ecological fallacy when a researcher
makes generalizations about an individual based on the results from aggregate data

(Schwartz, 1994).

131

Third, one of the least frequently used methods was that communities were
separated and individual level regression or other analyses were conducted within
different communities, grouping cases according to some criteria (Eugen, 1996). These
separate analyses within communities can produce unbiased estimates of relationships
within communities, but comparisons between communities are difﬁcult due to the
likelihood of unequal parameter variances, and also the difﬁculty of comparing the
results across many communities.

Mostly relying on disaggregated or pooled data approach, researchers still use
conventional OLS regression method for the reason of insufﬁcient number of cases
required especially in the bottom level (citizen-level) to perform a reliable HLM analysis.
For example, research tested the effect of neighborhood disorder on citizen perceptions of
powerlessness by using survey data for more than 2,482 citizens in Chicago with linked
data on respondents’ census tracts (Geis & Ross, 1998). Level 2 data were disaggregated
to citizens and they were used as the unit of analysis. In this study, approximately two-
thirds of the census tracts (766 out of 1,169) contain only one respondent. Under such
situations, HLM analysis could not generate reliable estimates of the variance within
census tracts and so OLS regression analysis was used. Due to the data structure, this
method was chosen even though this approach violates the assumption of independence
among observations within groups. Similarly, the number of respondents living in each
census tract or block ranged from 1 to 72 with a mean of 8.6 people and so the OLS
regression method was utilized in examining the effect of community level
socioeconomic status on individual level adult health using a national sample of adults in

the US (Robert, 1998). The researcher in this study made a correction to ﬁx the

132

clustering effect. More speciﬁcally, he realized that clustering of observations within
communities produces serial correlation and so he corrected the serial correlation by
adjusting standard errors. He used a replication based survey sampling error program
which adjusts for the fact that respondents clustered within the same communities are
likely to share characteristics compared to respondents chosen randomly from the
population. Regardless, conventional regression analysis methods are still used when the
survey sampled only a few respondents within each census community (Duncan, Connell,
& Klebanov, 1997).
(4) Advantages of Multilevel Analysis

A multilevel approach in criminal justice is theoretically concerned with the effect
of community macro characteristics on individual-level perceptions of crime and the
justice system. Thus, the research design involves both community-level and citizen-
level variables. As stated, prior to the multilevel analysis approach, traditional regression
methods were used even with the nested data. One of the most important assumptions of
the traditional regression approach is independence of observations, where the
observations of any respondent are not related to those of others, which is violated
especially when the multilevel data are disaggregated or aggregated, where individuals
are nested in groups such as organization, community, or city. Violation of the
independence of the observations yields biased estimates of the relationships among
variables (Williams, 1999) by violating the assumption of heterogeneity of variance. The
HLM analysis method avoids the problems of aggregation bias and misestimated
variances that affected the validity of multilevel analysis in conventional methods

(Raudenbush & Bryk, 2002). The assumption of the independence of observations is

133

maintained since the HLM analysis allows cross-level hypotheses to be addressed in one
analysis without having to aggregate citizen perceptions to community means or to
disaggregate community ecological characteristics to individual citizens. The HLM
analysis is based on multilevel theory to specify direct effects of variables on each other
within any one level, and to specify cross-level interaction effects between variables
located at different levels (Goldstein & Rasbash, 1996; Raudenbush & Bryk, 2002).
Compared to complete pooling or unpooling methods, the distinction can be made
between causal effects of, for example, citizen-level variables and constraining effect of
community-level variables. Therefore, one of the obvious advantages is the ability to
explicitly model and test cross-level relationships by considering interactions between
group level and individual level characteristics with correct estimates of standard errors
(Koenig & Lissitz, 2001; Raudenbush & Bryk, 2002; Willms, 1999). Based on theory
and logic, the researcher should design mediating mechanisms that cause variables at one
level to inﬂuence variables at another level. For example, community-level concentrated
disadvantage and crime may produce individual-level fear of crime through increasing
individual level perceptions of neighborhood incivility. In a traditional method, these
three variables would be incorrectly modeled at the same level for data analysis.
Furthermore, if we use pooled data, OLS regression analysis would assume that
the relationships between variables are to be identical across all groups (Reise & Duan,
2003). In traditional OLS approach, all of the regression parameters are ﬁxed and level-2
variance components are not separable from the individual level residual (Hoffman &
Gavin, 1998). Unlike traditional statistical approaches, HLM accounts for intraclass

correlation in the data, and therefore provides more efﬁcient estimates of the effects

134

associated with different individuals and groups than is provided by ordinary least
squares regression methods (Beyers, Bates, Pettit, & Dodge, 2002). HLM generates
correct decomposition of variances in parameter estimates between and within group
components (Goldstein & Rasbash, 1996). HLM estimates random coefﬁcient models in
that the level-1 parameters are allowed to vary across groups (Raudenbush & Bryk,
2002). In HLM, the 510pes of predictors are ﬁxed within communities or allowed to be
random across communities by considering the variance component of each variable.
That is, if the effect of a community-level variable on a dependent variable plays out
differently across communities, multilevel models can explain this causal heterogeneity
by specifying the slope of the variable as random across communities, or ﬁxed within
communities if the variance of the variable is not signiﬁcant across communities. Also,
HLM allows the intercept in a model to be random across communities and assess how
the effects of community-level variables are signiﬁcantly different across communities.

Most of all, Ordinary Least Squares (OLS) regression in aggregated or
disaggregated data would ignore the aggregate clustering of individuals, thereby
suppressing the independence of observations and heterogeneity of variance. HLM
analysis has an advantage over traditional methods by testing multilevel theories and
simultaneously modeling variables at different levels, which decreases the probability of
a Type I error in traditional methods. HLM resolves many of the problems that
conventional methods encounter, yet it is still based on traditional assumptions of
independence of observations, normality, and heterogeneity of variance. Despite its
effect on the studies of multilevel theoretical models, HLM does not allow the

examination of covariance structure models, such as factor analysis, path analysis, and

135

structural equation models, which will be discussed in the later section. The next
chapters will report the major research ﬁndings and will discuss the implications of the

results in both theoretical and practical terms.

136

CHAPTER SIX
FINDINGS

In this chapter, the major research ﬁndings are presented. To begin, several
diagnostic procedures are introduced to help understand the sample characteristics and
district level situations. For this, descriptive statistics in both individual and district
levels are presented ﬁrst followed by bivariate correlations. The analysis proceeds with
the estimation of a series of hierarchical linear models. For the multilevel analyses, a
three-stage modeling procedure is used for each outcome: the one-way AN OVA to obtain
descriptive statistics such as intraclass correlation and reliability estimates; the random
coefﬁcients model to look at the effects of citizen-level predictors, and the ﬁxed effect
full model to determine the effects of both district-level and citizen-level predictors.

1. Preliminary Statistics

(1) General Characteristics

To begin with, descriptive statistics were obtained to report sample characteristics.
The subjects of this study are 654 male and female adults over 19 years old. Table 5
includes the general characteristics of the sample. Among the total respondents, 53
percent were females and 47 percent were males. Respondents were on average 38 years
old with the age distribution between 20 and 69. On average, respondents completed
education of about 2-year college (4 = 2-year college). The average monthly household
income was distributed between $2,500 and $3,490. On average, the respondents were
living in either owned or long-term rented house. The respondents were on average
living 8.91 years in the neighborhood. About 20 percent of the respondents reported that
they had a neighborhood watch team in the community. About 17 percents of the

respondents were victims of crime, while a majority of the respondents (59 percent) had

137

friends, relatives, or family members who were victims of crime in the past. About 43
percent of the citizens reported that the media covered crime issues over politics,
economics, and other issues. A minority of the respondents (9 percent) has volunteered
for crime prevention activities (i.e., neighborhood watch).

Table 5. Individual Level Descriptive Statistics (N=654 in 25 Districts)

 

 

Variable Mean SD Min Max 7
ENDOGENOUS VARIABLES
Fear of crime 17.44 4.93 8 32
Perceived Risk 24.95 5.81 8 40
Constrained Actions (a) 0 1 -1.65 3.09
Cautious Actions (a) 0 l -1.73 2.86
Active Defense (a) 0 1 -1.54 4.61
EXOGENOUS VARIABLES
Community Concem/Control Model
Satisfaction with Police 20.36 4.33 7 34
Perceived Community Policing 12.53 2.98 6 23
Perceived Cohesion 11.39 3.28 5 20
Neighborhood Watch (Yes=1) .20 .40 0 1
Victimization Model
Direct Experience (Yes=l) .17 .38 0 1
Indirect Experience (Yes=1) .59 .49 0 1
Crime on Media (Yes=1) .43 .50 0 1
Vulnerability
Gender (female=1) .53 .50 0 1
Age 38.42 11.45 20 69
SES (a) 0 1 -3.16 1.91
Education 4. 19 l .52 1 7
Income 5.69 2.46 l 10
Housing Situation 3.57 .77 1 4
Disorder Model
Perceived Incivility 11.67 2.75 5 20
CONTROL VARIABLES
Length of Residence 8.91 7.91 .l 40
Volunteer Activities (yes=1) .09 .29 0 l

 

Note: (a) Weighted factor regression score

The average level of fear among the respondents was 2.18 points (mean/number

of items) in the four-point scale (l=not afraid at all, 2=not afraid, 3= afraid, 4=very

afraid) per item for all 8 items. The average satisfaction with police was 2.91 points in

138

the ﬁve-point scale (l=strongly disagree, 2=disagree, 3=neutral, 4=agree, 5=strongly
agree). On average, the respondents perceived community-policing activities at 2.09
points in the four-point distribution (l=strongly disagree, 2=disagree, 3=agree,
4=strongly agree). Community cohesion was perceived on average at 2.28, and the
average perception of incivility was 2.33 points in the four-point distribution (l=strongly
disagree, 2=disagree, 3=agree, 4=strongly agree). These statistics indicated that, on
average, fear of crime, satisfaction with police, perceived cohesion, and perceived
incivility were not very high. On the other hand, the average perceived risk was 3.12
points in the ﬁve-point distribution (l=very low, 2=low, 3=medium, 4=high, 5=very
high), which indicated perceived risk was above median.

Table 6 presents district-level descriptive statistics. About 22 percent of citizens
moved in and about the same percent of citizens moved out during 2003. About 33
percent of citizens were living in a rented house. Average percent of citizens living in
poverty indicated that on average 1.52 percent of citizens had received government
support in 2003. The population per square kilometer in the 25 districts ranged from
7,588 to 28,102 with the average of 18,063. The most frequent type of crime per 100,000
persons in 2004 was assault (828.72), followed by larceny (372.47), rape (17.51), and

robbery (15.91).

139

Table 6. District Level Descriptive Statistics (25 Districts)

 

 

Variable Mean SD Min Max
MOBILITY (a) 0 1 -l .21 1.92
Percent Moved In 21.65 1.79 18.95 24.52
Percent Moved Out 22.33 1.81 19.94 27.56
ECONOMIC DISADVANTAGE (a) 0 1 -1.52 2.51
Percent Rented House 33.48 5.92 24.30 46.52
Percent in Poverty 1.52 .64 .62 3.18
CRIME (a) 0 1 -1.03 3.93
Murder (b) 2.35 1.28 .95 6
Robbery (b) 15.91 7.67 3.15 33.14
Rape (b) 17.51 7.01 9.78 41.07
Larceny (b) 372.47 210.12 180.08 1245.70
Assault (b) 828.72 389.51 526.49 2373.95
Arson (b) 3.34 2.09 1.04 10.81
Drug Offense (b) 6.31 6.34 1.23 31.70
Abduction (b) .38 .38 .00 1.44
Gambling (b) 11.87 8.34 3 38.18
DENSITY (c) 9.76 .32 8.93 10.24
Population Density 18063.71 5203.73 7588.50 28102.07

 

Note: (a) Weighted factor regression score
(b) Number of crimes per 100,000 persons
(c) Natural log transformation

1 (2) Bivariate Correlation

Citizen Level Correlation

 

A series of bivariate correlations were ﬁrst estimated. Table 7 presents bivariate
correlations between fear of crime and citizen-level predictors. The results indicated that
multi-collinearity was not a problem considering the moderate Pearson correlation values.
The highest correlation was observed between fear and perceived risk (.3 8). OLS
regression diagnostics (not presented here) provided additional support that multi-
collinearity did not show, demonstrating the tolerance statistics for each independent

variable equal to or greater than .82.

140

.882 3.5-8 8. V a ... mmEﬂoEooU coca—2.80 .588; 08 mogcm ”202

 

 

 

 

 

LP 8.- ...8. .2. .87 .8. 88>8<52§8>8
8 8.- ...8. 8.- 8. 8528.88 883.:

I ...m- 8.- 8.- ...:. 886888288

8.- 8.- 8.- mum-.8

I 8.- 8. 09...:

| 8. 0388.2

I as: 8 2:5 .2

8 S 8 2 E 8 8 _

...8. ...2. am. ...8. ...S. 8.- 8. 8.- 8.- 8. ...8. 8838<§§8>8
8. 8. *8. 8m. 8. 8. 8.- 8.- 8.- *8- 8.- 852350 £83.:
...2. 8. ...8. ...S. 8.- ...:.- 8. 8. 8. ...8. ...2. 8565838588
8. 8. ...8- :.- 8. 8. 8. ...2. 8. 8.- 8. 88.2
8.- 8.- 8. ...2. ...8. ...2. 8.- ...S. 8. 8.- 8. 03.3
8.- 8.- 8.- 8. 8. 8. 8. ...2. 8m. 8. *8. 2258.2
m9. 8. 8. 8. 8.- 8.- 8.- 8. 8.- 8. 8.-. a8»: 8 225.2
...m 8.- 8.- 8. 8.- ...8. E. ...8. *8. ...8. 8888553285.:

WP. 88.. 8.- :.- ...:. 8. *2. *8. ...2. 8888852

8.8:. *2. ...8. 8. 8.- 8.- ...8- .88.- 533 8888852 .8

.8. *8. 8. ...8. 8. 8.- 8. 87.28 8385 .8

I LM *8. *2. 8. 8.- 8. 838 85888 82838 .8

*8- 8. 8. 8. 8. 8:3 .23 88888 .8

I ...-W *8. 8.- 8. 3.585 3:3. .m

*8. ...2. ...S. 885886 .4

l ...E. ...8. :23. 85888 .m

I ...8. 88 82838 .N

I as“. .8

: 2 8 w a 8 m 4 m N 8 _

 

Sumo H Zv 3388> $54 58:0 use Son 5953 macaw—8.50 0853mm .5 033-

141

As expected, perceived risks, constrained actions, cautious defense, victimization,
indirect victimization, female, and perceived incivility were positively associated with
fear, while neighborhood watch was negatively associated with fear. These were all
consistent with the hypothesized direction (H2d, H3a, H3b, H3e, H4, H5a, and H5c).
Volunteer activity was positively associated with fear, suggesting that those volunteering
for crime prevention activities perceived a higher fear of crime. On the other hand, active
defense, satisfaction with police, perceived community-policing activities, perceived
cohesion, crime on media, age, SES, and length of residence were not signiﬁcantly
associated with fear. The results, therefore, did not support the hypotheses regarding the
link between these predictors and fear (H2a, H2b, H20, H3c, H3d, H3f, and H5c).
Regarding the association between fear and behavioral adaptations, the results indicated
that fear was positively associated with constrained actions and cautious actions, while it
was not signiﬁcantly associated with active defense. The hypothesis between fear and
behavioral adaptations, therefore, was partially supported (H5c).

For the outcome “perceived risk,” fear, constrained action, cautious action,
victimization, indirect victimization, and perceived incivility were positively associated
with perceived risk, while neighborhood watch and length of residency were negatively
associated with it. These associations were consistent with the expectation. On the other
hand, active defense, satisfaction with police, perceived community policing, perceived
cohesion, crime on media, female, age, SES, and volunteer activities were not
signiﬁcantly associated with perceived risk. The results indicated that the hypothesized

relationship between perceived risk and behavioral adaptations was partially supported

142

(H5b) since perceived risk was positively associated with constrained actions and
cautious actions, while it was not signiﬁcantly associated with active defense.

For the behavioral adaptation measures, the associations among behavioral
adaptations themselves will not be reported here since the relationships among them were
not intended to be tested. The signiﬁcant associations between predictors and behavioral
adaptations, therefore, were reported here without giving attention to the association
between behavioral adaptation measures. For the outcome “constrained actions,” fear,
perceived risk, victimization, indirect victimization, and female were positively
associated with constrained actions. The association between these predictors and
constrained actions were consistent with the expected direction. For the outcome,
“cautious actions,” fear, perceived risk, active defense, perceived community policing,
perceived cohesion, indirect victimization, female, age, and SES were positively
associated with cautious actions. It was interesting to observe the positive associations
between perceived community policing activities and cautious behaviors as well as
perceived cohesion and cautious behaviors. For the outcome “active defense,” perceived
community-policing activities, victimization, and indirect victimization were positively
associated with active defense, while satisfaction with police was negatively associated.
The negative relationship between satisfaction with police and active defense was
interesting, compared to the positive relationship between community-policing activities
and active defense. Bivariate relationships reported here, however, should be closely
examined in the multivariate analyses to see their relative relationships after controlling

for the effects of other predictors.

143

District Level Correlation

Table 8 presents bivariate correlation between fear and district-level variables.
Fear was measured as an individual-level perception, but was aggregated into district-
level for this correlation test. Bivariate results from the independent variables showed no
signs of multi-collinearity. In addition, no evidence of multi-collinearity was detected
when fear was regressed on district-level variables (not shown), indicating tolerance
statistics greater than .79.

Table 8. Bivariate Correlations between Fear and District Level Variables (N = 25)

 

 

1 2 3 4 5
1. Fear _
2. Mobility .15 _
3. Economic Disadvantage .43* -.13 __
4. Crime .49* -.08 .29 __
5. Population Density -.45* -.03 -.34 -.37 _

 

Note: Entries are Pearson Correlation Coefﬁcients;
* Correlation is signiﬁcant at the 0.05 level (2-tailed)

As expected, economic disadvantage and crime were positively associated with fear.
Districts with higher economic disadvantage and higher crime were likely to be exposed
to higher levels of fear of crime, which was consistent with the hypothesized
relationships (Hlb, ch). Interestingly, the association between mobility and fear was
positive, but was not statistically signiﬁcant (Hla). Contrary to the hypothesis (Hld),
population density was negatively associated with fear of crime. Districts with a higher
population per square kilometer were likely to be exposed to lower levels of fear.
Correlations among district-level predictors also showed interesting patterns. For
example, even though it was statistically negligible, population density was negatively
associated with mobility, economic disadvantage, and crime. In addition, mobility was
negatively associated with economic disadvantage, crime, and population density, even

though those associations were not statistically signiﬁcant. Overall, the results suggested

144

that, unlike major US. cities, population density and mobility in Seoul might not be
indicators of social disorganization. The next section examines the extent to which the
relationship reported in the correlation analyses still holds while controlling for the effect
of other variables.
2. Hierarchical Linear Modeling Analyses

( 1) Model Building Procedure

The models for fear of crime were estimated using HLM Version 6, following a
three-step modeling strategy: the unconditional (one-way AN OVA) model, random
coefﬁcient model including only citizen-level predictors, and the ﬁxed effects full model
including both citizen-level and district-level variables simultaneously. Here, the
dependent variable “fear of crime” was treated as an interval-level measure.
Unconditional (one-way AN OVA) Model

Model building begins with a simple unconditional model, which partitions the
variance of dependent variables into between group variance and within group variance
(Raudenbush, Bryk, & Cheong, 2000). This one-way ANOVA with random effects,
therefore, provides preliminary information about how much variation in fear lies within
and between districts and if the reliability of each district’s sample mean is reliable as an
estimate of its true population mean. That is, the unconditional or the one-way AN OVA
models determine the amount of variation in the outcomes within and between districts,
and provide reliability estimates for the outcome variable at the aggregate level. In this
one-way AN OVA model, reliability is a function of sample size in each of the districts
and intraclass correlation is the proportion of the total variance that is between districts

relative to the amount that is within districts (Raudenbush & Bryk, 2002). The decision

145

if the data are reasonable for multilevel analyses is made in this model. The models are
expressed as:

Level-l model: Yij = Bo]: + In
Level-2 model: BOj = 700 + U01

In this one way ANOVA model, the level-1 model (Yij = BOJ' + rij ) represents
variation in citizens’ fear of crime within each district: where Yij is the fear of crime of
citizen i in district j, the intercept [30" represents the average fear of crime of district j, and
the random effect rij is assumed to be normally distributed with a mean of zero and a
variance of oz. The level-2 model (130,- = 700 + qu) accounts for variation in fear of crime
between districts: where the intercept We is the grand mean of citizens’ fear of crime
across all districts. The random effect “01' is assumed to be normally distributed around
mean of zero and variance of too. In this fully unconditional model, the intraclass

correlation coefﬁcient ( ,0 = too /02+ too =0.060) shows a portion of between-group

variance in the total variance (Duncan & Raudenbush, 1999). Approximately 6 percent
of the variation in fear of crime was between districts and the remainder was attributed to
citizen-level variation and random error. This between-group variation is reasonably
acceptable considering other studies that encountered small variance between macro-
level units ranging from around 5 to 10 percent (Reisig & Parks, 2000; Sampson &
Bartusch, 1998; Sampson, Morenoff, & Earls, 1999). The district-level reliability (.57)
indicates that the sample mean was a reliable measure of the true district mean for fear of
crime. “The reliability will be close to 1 when the group means, Boj, vary substantially
across level-2 units holding constant the sample size per group” (Raudenbush & Bryk,

2002. p. 257). This reliability indicates that district-level differences can be modeled

146

with a reasonable degree of precision, another encouraging result for multilevel analyses
along with the intraclass correlation.

Table 9. Decomposition of Variance and District-Level Reliabilities

 

 

 

Variance Fear of Perceived Constrained Cautious Active
Components Crime Risk Actions Actions Defense
Within-District
Variance (oz) .943 .917 .880 .938 .913
Between-District

Variance (too) .059 .077 .086 .037 .096

lntraclass

Correlation .060 .077 .089 .038 .095

District-Level
Reliability .568 .632 .647 .444 .663

 

 

Note: N = 654 citizens nested in 25 district units.
Outcome measures are standardized (mean=0, standard deviation=1).

The intraclass correlation coefﬁcient for the outcome measures of perceived risk,
constrained actions, cautious defense, and active defense was 0.077, 0.089, 0.038, and
0.095 respectively. The district-level reliability for perceived risk, constrained actions,

cautious actions, and active defense was .632, .647, .444, and .663 respectively.

R_andom Coefﬁcient Regression Model

Prior to modeling district-level effects, a random coefﬁcient regression model was
estimated including only citizen-level variables, X, that were selected based on their
theoretical relationship to fear of crime. This model examines the multivariate
association between citizen-level variables and fear of crime and shows whether any of
the citizen-level slopes vary signiﬁcantly across districts. Unlike ordinary linear
regression analyses, this multilevel analysis requires not only proper speciﬁcation of the
individual—level regression equations, but also careful speciﬁcation of the variance

components to be estimated. Random coefﬁcient regression model helps determine

147

whether a slope is to be ﬁxed within districts or should be speciﬁed as random across
districts depending on the signiﬁcance of variance across districts. If a district-level
slope varies across districts, the slope can be estimated using district-level predictors
(Rountree, Land, & Miethe, 1994). In this model, all of the interval or ratio-level
variables in the individual level were centered around the group means since this allows
interpretation of parameter estimates as person-level effects within each group. Dummy
coded variables were remained uncentered. The intercept term takes on a different
meaning according to the type of centering. Three different options are considered
(Hoffman & Gavin, 1998). First, researchers choose raw metric scaling with no centering
and the intercept is the expected value of Yij when Xij is zero. The second option is grand
mean centering where the grand mean of the level-1 predictor is subtracted from each
level-1 case and the intercept is the expected value of Yij when ij is the average across
all individuals in the sample. The third option is group mean centering where the
relevant group mean of the level-1 predictor is subtracted from each case and the
intercept is the expected value of Yij when Xij is equal to the group’s mean. Researchers
often use centering for meaningful interpretation of the intercept. Since 301' becomes the
dependent variable in the level-2 models, its meaning must be clear so as to understand
what is being predicted. For example, the level of fear of crime when SES is zero does
not provide sensible information. Therefore, SES is often centered around the group or
grand mean. In the case that SES is grand mean centered, for instance, the intercept is
interpreted as the expected fear of crime when SES is the average of all citizens. There is
no statistically correct choice among centering options and the choice should be driven

by theory and by the intent of the research considering the conceptual paradi gm and

148

research question under investigation (Hoffman & Gavin, 1998; Koenig & Lissitz, 2001).
The random coefﬁcient regression or the within-district model was expressed as:

Citizen Level: Fear ij = BOj + 2(q=1-k) Bqujj + ril-
This citizen-level model can be conceived of in the same way as a multiple regression
model. Here, i refers to the individual (citizen) and j to the group (district). Yij is the
outcome variable (fear of crime) measured for citizens. [30,- is a intercept for a given
district. Bq represents the effect of a certain independent variable on fear of crime. Xq-j is
the value of predictor q for citizen i in district j. The unique effect associated with the
individual is r-j. In contrast to the ordinary least squares (OLS) regression model, the
random coefﬁcient model allows the intercept to take different values in each of the
districts. The results indicated that neighborhood watch, perceived incivility, perceived
risk, length of residency, and volunteer activities had effects which varied across districts
and showed a high reliability of variance, thus the slopes of these variables remained
random across districts (Raudenbush & Bryk, 2002). Slopes for all other variables were
speciﬁed as ﬁxed within-districts due to the lack of enough variation across districts.
These decisions relied on not only statistical results but also theoretical considerations.
In this study, statistical results were maintained since existing theories did not strongly
suggest that some of these variables inﬂuenced fear of crime in different ways across
districts. Following the same procedure, the slopes of socioeconomic status, crime on
media, neighborhood watch, and length of residence remained random across districts for
the outcome measure of perceived risk; the slopes of fear, perceived community policing
activities, indirect victimization, and volunteer activities for crime prevention remained

random across districts for the outcome measure of constrained actions; the slopes of fear,

149

perceived incivility, and perceived risk remained random across districts for the outcome
measure of cautious actions; the slopes of fear, perceived community policing activities,
and indirect victimization remained random across districts for the outcome measure of
active defense.
Fixed Effect Model (Full Model)

After modeling the random coefﬁcient model, the level-2 models are formulated.
In the level-2 model, the coefﬁcients from the level-l model become the dependent
variables. This model allows the study of the effects of city-level variables on the
variance among the values of the coefﬁcients. The models are expressed as:

District Level: 1301- = 700 + 2(s=1--)ysWSj +qu

I311: 710 + 71 W, +qu

OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

The intercept ([30]) is hypothesized to be a function of the overall mean of fear of crime
(700), district characteristics (W51), and a unique (or random) effect associated with each
district (uoj). The slope (BU) is hypothesized to be a function of the mean of the slopes
across districts, the effect of some district characteristics (711), and a unique (or random)
effect associated with each district (ulj). Here, the slope was considered random across
districts since u”- is included in the model. If the effect does not signiﬁcantly vary across
districts, it is ﬁxed by excluding the error term (Ulj) in the model. Several similar models
for slopes are generated based on the number of variables in citizen level.

Through this model building procedure, ﬁnally, a series of combined hierarchical
models are estimated to investigate simultaneously the effects of both citizen and district
level variables. District-level variables were grand mean centered in order not to discard

the compositional effects with group-mean centering. Then, these district-level

150

predictors were entered in the model. Two separated models were estimated considering
the sample size (N=25) at the district level. The ﬁrst model only included three district-
level variables such as economic disadvantage, mobility, and crime. Then, population
density was added in the next model. The estimation of two separate models would help
avoid statistical bias due to the sample size at the district-level. The recommended rule is
that one needs at least 10 observations for each predictor at the aggregate-level (Bryk &
Raudenbush, 1992). The number of level-2 variables would also depend on the structure
of the data such as intraclass correlation and reliability. Next, cross-level interaction was
considered. One of the obvious advantages of multilevel analysis is the ability to model
and test cross-level relationships by considering interactions between group level and
individual level characteristics. The preliminary analysis showed that the district-level
variable “economic disadvantage” had a signiﬁcant effect not only directly on the
intercept of the model but through the slope parameter of a level-1 variable “volunteer
activities.” Therefore, the model intercept and the slope of the variable “volunteer
activities” were modeled as a function of economic disadvantage. Likewise, the model
intercept and the slope of the variable “perceived incivility” were modeled as the ﬁrnction
of mobility, while the model intercept and the slope of the variable “perceived risk” were
modeled as a function of crime. On the other hand, the district-level variable “population
density” was signiﬁcant in predicting the intercept of the model but it did not play a
signiﬁcant role in modeling the slope parameters of level-l variables. Thus, only the
intercept was modeled as a function of population density. On the basis of theoretical
consideration and statistical results, the ﬁnal model was deﬁned as (fear of crime as the

outcome measure):

151

Level-1 (citizen-level):

Yij = [301+ [31 j(satisfaction) + [Sh-(perceived community policing) + [SM-(perceived cohesion)
+ B4j(neighborhood watch) + [kl-(victimization) + [lg-(indirect victimization) + B7j(crime
on media) + ng(female)+ B9j(age) + Ble(SES) + I311j(perceived incivility) + I312j(perceived
risk)+ [Sm-(length of residency)+ B14j(volunteer activities)+ rij

Level-2 (district-level):

BOj = 700 + 701 (economic disadvantage) + 702(mobility)+ 703(crime)
+ “ﬂy-(population density)+ U0]-

I311 = 710

52;“ = Yzo

st = 730

B4j = 740 + U4j
[351' = 750

Boj = 760

[371' = 770

st = 780

I39j = “1’90

I310j = Y 100

[3111' = 7110+ 7111 (mobility) + U111

[3123'= 7120+ 7121(crime)+ Ule

[3131: Y13o+ Ul3j

[3141' = 7140+ 7141 (economic disadvantage)+ U14j

These equations could also be combined through substitutions of terms and expressed as:

Yij = 700 + 701 (economic disadvantage)+ 702(mobility)+ 703(crime)+ 704(population
density)+ 710(satisfaction) + 720(perceived community policing) + ”OJ-(perceived cohesion)
+ 740(neighborhood watch) + 750(victimization)+ 760(indirect victimization)+ 770(crime
on media)+ 730(female)+ 790(age)+ 7100(SES) + 7110(perceived incivility)+
7111(mobility)x(perceived incivility)+ 7120(perceived risk)+ 7121(crime)x(perceived risk) +
7130(length of residency)+ 7140(volunteer activities)+ 7141(econ0mic
disadvantage)x(volunteer activities) + [UOj + U4j+U| 11' + UIZj + U13j +U|4j + r-j]

The model is now run under full maximum likelihood in which variance-covariance

parameters and ﬁxed level-2 coefﬁcients are estimated by maximizing their joint

likelihood (Raudenbush et al., 2000).

152

(2) Results of Analyses
Fear of Crime

Table 10 presents three models for fear of crime. Fear of crime was measured as
a factor based scale in this analysis. Model I presents the random coefﬁcient model. The
model examines the association between citizen-level variables and the outcomes in a
multivariate context. The model also helps determine which of the citizen-level slopes
vary signiﬁcantly across districts. All citizen-level variables, except dummy coded
variables, were group mean centered. Dummy coded variables remained uncentered.
Based on the statistics for variance components, neighborhood watch, perceived incivility,
perceived risk, length of residency, and volunteer activities had effects which varied
across districts and showed a high reliability of variance, thus the s10pes of these
variables remained random across districts. Models 11 and [[1 present ﬁxed effect
hierarchical models, which are fully combined models. All district-level variables were
grand mean centered. Then, these district level predictors were entered in the model one
by one. As previously mentioned, two separated models were estimated to help avoid the
statistical bias due to the sample size at the district level. Only three district level
variables (i.e., economic disadvantage, mobility, crime) were included in model H, and
population density was added in model III. The models converged well up to three
district level variables. The comparison between model H and model 111 indicates that
the size of t-ratio for crime and economic disadvantage decreases when one more district-
level variable (population density) is added in model HI, which may possibly be due to
the sample size effect as well as the control effect of this variable. This result suggested

that up to three district-level variables were recommendable for this data structure.

153

Finally, as mentioned before based on the theoretical and statistical reasons, cross—level
interactions were considered for district-level mobility and citizen-level perceived
incivility, district-level economic disadvantage and citizen-level voluntary activities, and
district-level crime and citizen-level perceived risk.

Table 10. Hierarchical Linear Models for Fear of Crime (N=654 in 25 Districts)

 

 

Volunteer Activities

CROSS LEVEL INTERACTION

Mobility x Perceived Incivility

Economic Disadvantage x Volunteer

 

1973* (.798)

 

2.030M (.695)

4207** (063)
-L568** (432)

2.076" (.675)

4210** (066)
:L811** (445)

I Variables Model I Model H Model III I
Constant 15.505***(.407) 15.534***(.413) 15.594*** (.391)
DISTRICT LEVEL (N=25)
Mobility .425* (.175) .352* (.170)
Economic Disadvantage .647***(.151) .443" (.118)
Crime .717* (.307) .473 (.239)
Population Density -2.078** (.599)
CITIZEN LEVEL (N=654)
Satisfaction with Police .026 (.040) .019 (.039) .020 (.03 8)
Perceived Community Policing .046 (.091) .063 (.094) .058 (.092)
Perceived Cohesion -.006 (.056) -.001 (.055) -.001 (.056)
Neighborhood Watch -.677 (.562) -.735 (.548) -.677 (.539)
Victimization .629 (.419) .681 (.394) .642 (.389)
Indirect Victimization 2.171***(.440) 2.177***(.451) 2.142*** (.462)
Crime on Media .062 (.307) .024 (.324) —.024 (.321)
Female .729** (.253) .837“ (.253) .844" (.261)
Age .036 (.019) .036 (.020) .038 (.020)
SES .054 (.198) .03 8 (.206) .073 (.209)
Perceived Incivility .324" (.093) .322" (.078) .336" (.081)
Perceived Risk .267***(.050) .290***(.047) .291 ***(.048)
Length of Residence -.001 (.039) -.001 (.039) -.003 (.03 8)

Crime x Perceived Risk .124"‘ (.045) .127* (.045)
VARIANCE EXPLAINED

Within-District .340 .346 .351

Between-Districts . 108 .565 .755

 

Note: Entries are unstandardized coefﬁcients and numbers in parenthesis are standard errors;
* p < .05, ** p < .01, *** p < .001

At the district-level, the results from Model H showed that citizens residing in
districts characterized by higher levels of mobility, higher levels of economic
disadvantage, and higher levels of crime reported signiﬁcantly higher levels of fear of

crime. The ﬁndings from Model II, therefore, indicated that the hypothesized

154

relationships between mobility (Hla), economic disadvantage (Hlb), crime (ch) and
fear of crime were all supported. On the other hand, the results from Model III indicated
that citizens residing in districts characterized by higher population density reported
signiﬁcantly lower levels of fear of crime. This ﬁnding indicated that the hypothesized
relationship between population density and fear of crime was not supported (Hld).
Instead, the direction of the relationship was the opposite of the hypothesis. The district
level variables accounted for 56 percent and 75 percent of the between district unit
variance in Model H and Model III.

At the citizen level (see Models I, H, and HI), indirect victimization, gender,
perceived incivility, perceived risk, and volunteer activities had signiﬁcant effects on fear
of crime. As expected, citizens who had friends, relatives, or family members who were
victims of crime in the past expressed signiﬁcantly higher levels of fear of crime.
Consistent with the hypothesis, females reported signiﬁcantly higher levels of fear of
crime. Citizens perceiving higher levels of social and physical incivility in the
neighborhood reported signiﬁcantly higher levels of fear of crime. Citizens perceiving
higher risks of victimization expressed higher levels of fear of crime. The frndings,
therefore, indicated that the hypothesized relationships between indirect victimization
(H3b), gender (H3e), perceived incivility (H4), perceived risk (H5a) and fear of crime
were supported. The control variable “volunteer activities” were positively associated
with fear. That is, citizens volunteering for crime prevention activities reported higher
levels of fear of crime. Interestingly, the hypothesized relationships between satisfaction
with police (H2b), perceived community policing activities (H2a), perceived cohesion

(H2c), neighborhood watch (H2d), victimization (H3a), crime on media (H3c), age (H3 1),

155

 

SES (H3d), and length of residence and fear of crime were not supported. The effects of
these predictors on fear of crime were not statistically signiﬁcant. Even though
statistically negligible, the relationships between perceived cohesion, neighborhood
watch, victimization, crime on media, age, length of residency and fear were consistent
with the hypothesized directions, while the relationships between satisfaction with police,
perceived community policing activities, SES and fear showed unexpected signs.

One of the advantages of multilevel analysis is the ability to explicitly model and
test cross-level relationships by considering interactions between group level and
individual level characteristics. In this study, cross-level interactions were considered
between district-level mobility and citizen-level perceived incivility, district-level
economic disadvantage and citizen-level volunteer activities, and district-level crime and
citizen-level perceived risk (see Model H and 111). Based on theory and logic, mediating
mechanisms that cause variables at one level to inﬂuence variables at another level were
designed. It was hypothesized that district-level mobility, economic disadvantage, and
crime would produce citizen-level fear of crime through increasing individual level
perceived incivility, decreasing volunteer activities, or increasing perceived risk. The
results indicated that citizens perceiving higher levels of incivility were likely to express
lower levels of fear of crime when they reside in districts characterized by higher levels
of mobility. This was an unexpected result, considering that the effects of district-level
mobility and citizen-level perceived risk on fear were positive when they were estimated
independently. It was necessary to check the association between mobility and perceived
incivility by aggregating perceived incivility into district-level (not shown). Interestingly,

the bivariate correlation between mobility and aggregated perceptions of incivility was

156

negative (-.16), even though the association was not statistically signiﬁcant. This result
will be explained further in the discussion section. The interaction between economic
disadvantage and volunteer activities was negative. Citizens volunteering for crime
prevention activities expressed lower levels of fear of crime when they reside in districts
characterized by higher levels of economic disadvantage. This indicated that the effect of
volunteer activities on fear depended on economic situations in the district, since the
effect of volunteer activities on fear was positive when it was estimated independently.
Consistent with the expectation, the interaction between crime and perceived risk was
positive. This indicated that citizens perceiving higher levels of risk of victimization
expressed higher levels of fear of crime when they reside in districts characterized by
higher levels of crime. Individual-level variables altogether accounted for 34 percent of
the within district unit variance (see Model I).
Perceived Risk and Fear of Crime

Next, a hierarchical linear model for perceived risk was estimated and included in
the table (see Table l 1) to assess the effect of citizen-level factors on perceived risk, and
to investigate whether perceived risk mediates the effect of any predictors on fear of
crime. In the one-way ANOVA model, the intraclass correlation coefﬁcient (.07 7)
suggested that about 7.7 percent of the variation in perceived risk was between districts
(see Table 9). The district-level reliability estimate (.63) indicated that the sample means
tend to be quite reliable as indicators of the true district means, and so encouraging
further multilevel analysis. Next, random coefﬁcient regression models were estimated
including all of the citizen-level variables. All citizen-level variables, excepted dummy

coded variables, were centered around the group mean. The slopes of socioeconomic

157

status, crime on media, neighborhood watch, and length of residency remained constant
across districts along with the intercept since these citizen-level variables had effects that
varied across districts, indicating some multilevel interactions. All other citizen-level
slopes were ﬁxed within districts. In the combined ﬁnal model, district-level variables
such as mobility, economic disadvantage, and crime were grand mean centered. To
consider cross-level interaction, based on theoretical and statistical suggestions, mobility
was modeled to predict the intercept and the slope of length of residence, while economic
disadvantage was modeled to predict the intercept and the slope of neighborhood watch.
Crime, however, was modeled to predict only the intercept since it was not signiﬁcant in

predicting any slopes of the level-1 variables.

158

Table 11. Hierarchical Linear Models for Perceived Risk and Fear of Crime

(N=654 in 25 Districts)

 

 

 

 

 

 

Variables Perceived Risk Fear of Crime
Constant -.213 (.117) -.387***(.084)
DISTRICT LEVEL (N=25)
Mobility -.063 (.035) .085* (.036)
Economic Disadvantage .075 (.059) .133***(.031)
Crime .002 (.063) .146* (.062)
CITIZEN LEVEL (N=654)
Satisfaction with Police .021 * (.010) .004 (.008)
Perceived Community Policing -.001 (.014) .012 (.019)
Perceived Cohesion -.025* (.011) -.001 (.011)
Neighborhood Watch -.236 (.160) -.146 (.111)
Victimization .288***(.069) .138 (.080)
Indirect Victimization .283***(.071) .443***(.09l)
Crime on Media .030 (.102) .006 (.065)
Female .046 (.093) .167" (.051)
Age -.003 (.003) .007 (.004)
SES -.010 (.056) .007 (.042)
Perceived Incivility .016 (.021) .066***(.016)
Length of Residence -.004 (.007) -.001 (.008)
Volunteer Activities .069 (.147) .409" (.141)
Perceived Risk .059***(.009)
CROSS LEVEL INTERACTION
Mobility x Length of Residency .015M (.005)
Econo Disadvantage x Neighborhood Watch -.205* (.085)
Mobility x Perceived Incivility -.042** (.013)
Economic Disadvantage x Volunteer -.325** (.088)
Crime x Perceived Risk .025“ (.009)
VARIANCE EXPLAINED
Within-District .198 .348
Between-Districts nt (a) .547

 

Note: Entries are unstandardized coefﬁcients and numbers in parenthesis are standard errors;

All outcome variables were standardized (mean=0, standard deviation=1)

(a) Variance explained between districts was negative

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

Even though the hypotheses were not explicitly stated, the hypothesized

relationships between predictors and perceived risks were considered the same as those

between predictors and fear. At the district level, the results suggested that none of the

district-level variables had signiﬁcant effects on perceived risk. At the individual level,

satisfaction with police, perceived cohesion, victimization, and indirect victimization

were signiﬁcantly associated with perceived risk. Consistent with the expectation,

159

citizens perceiving higher levels of cohesion expressed signiﬁcantly lower levels of risks
of victimization, while those who were victims in the past, and those who had relatives,
friends, or family members who were victims of crime reported signiﬁcantly higher
levels of risks of victimization. Contrary to the expectation, citizens perceiving higher
satisfaction with police expressed higher levels of risks of victimization. The cross level
interaction between district-level mobility and citizen-level length of residence had
signiﬁcantly positive effects on perceived risk, while the interaction between district-
level economic disadvantage and citizen-level neighborhood watch had signiﬁcantly
negative effects on perceived risk. These results suggest that citizens living in the
neighborhood longer express higher levels of risks of victimization when they reside in
districts characterized by higher levels of mobility. As well, citizens having
neighborhood watch teams in the community expressed lower levels of risks of
victimization when they reside in districts characterized by higher levels of economic
disadvantage. Finally, these two models together showed that, consistent with the
hypothesis, perceived risks of victimization had a positive effect on fear of crime (H5a).
That is, citizens perceiving higher levels of risks of victimization reported higher levels
of fear of crime. Interestingly, the effects of perceived cohesion and victimization were
signiﬁcant on perceived risks, but not on fear of crime, which indicated that perceived
risks mediated the effects of perceived cohesion and victimization on fear of crime. That
is, even though perceived cohesion and victimization did not directly increase the levels
of fear of crime, they played an important role in reducing or enhancing fear of crime via
decreasing or increasing perceived risks of victimization. The citizen-level variables

altogether accounted for 19.8 percent of the within district unit variance. Surprisingly,

160

 

the explained variance between district-units in this model was negative, which might be
interpreted based on Snijders and Bosker’s ( 1993) arguments against calculating
explained variance between groups based on the proportional reductions in the estimated
variance components by comparing the size of variance between one-way ANOVA
model and the ﬁxed effect full model, especially in an unbalanced design. Scholars in
multilevel analysis cautioned for meaningfully interpreting the explained variance
between groups in some unbalanced design situations. 6
Perceived Risk, Fear, and Behavioﬁnal Adaptations

Finally, three hierarchical linear models for constrained actions, cautious actions,
and active defense were estimated to investigate the structural associations between
perceived risk, fear and behavioral adaptations. These models also served to assess the
effects of both district-level and citizen-level predictors on behavioral adaptations. In
one-way ANOVA models, the intraclass correlation coefﬁcients (.089, .038, and .095)
suggested that about 8.9 percent of the variation in constrained actions, about 3.8 percent
of the variation in cautious actions, and about 9.5 percent of variation in active defense
was between districts (see Table 9). The district-level reliability estimate (.65, .44,
and .66) indicated that the sample means tend to be reliable as indicators of the true
district means, hence encouraging further multilevel analysis for these outcome measures.
Next, random coefﬁcient regression models were estimated including all of the citizen-
level variables. All citizen-level variables, except dummy coded variables, were centered
around the group means. As previously mentioned in the model building procedure,
some of the citizen-level variables had effects that varied across districts, indicating some

multilevel interactions. As a result, the slopes of these variables remained constant

161

across districts along with the intercept. For the citizen-level variables that did not have
effects varied across districts, the slopes were ﬁxed within districts. In the combined
ﬁnal model, district-level variables such as mobility, economic disadvantage, and crime
were grand mean centered. They were modeled to predict only the intercept since these
variables were not signiﬁcant in predicting the slopes of level-1 variables.

Table 12. Hierarchical Linear Models for Behavioral Adaptations (N=654 in 25 Districts)

 

 

 

Variables Constrained Cautious Active
Actions Actions Defense
Constant -.543***(.098) —.219* (.086) -.l42 (.105)
DISTRICT LEVEL (N=25)
Mobility .084 (.054) -.010 (.046) .052 (.058)
Economic Disadvantage .015 (.058) -.026 (.062) .001 (.056)
Crime .1 18 (.070) -.1 15 (.062) .016 (.042)
CITIZEN LEVEL (N=654)
Satisfaction with Police -.01 l (.009) -.011 (.015) -.024 (.013)
Perceived Community Policing .020 (.023) .050***(.012) .079***(.019)
Perceived Cohesion .020 (.012) .025* (.012) .009 (.018)
Neighborhood Watch .043 (.099) -.219* (.099) .052 (.112)
Victimization .227* (.093) .130 (.128) .258"‘ (.114)
Indirect Victimization .300* (.107) .088 (.105) .125 (.121)
Crime on Media -.101 (.071) .041 (.106) -.142 (.093)
Female .671 ***(.068) .246" (.068) .040 (.087)
Age .004 (.003) .010" (.003) -.003 (.004)
SES -.034 (.036) .130" (.044) .032 (.040)
Perceived Incivility .026 (.015) .012 (.017) .019 (.016)
Length of Residence -.007 (.005) -.003 (.005) -.004 (.007)
Volunteer Activities .012 (.164) -.1 17 (.181) -.005 (.114)
Perceived Risk -.002 (.007) .003 (.010) -.006 (.008)
Fear of Crime .041** (.010) .021* (.010) .004 (.011)
VARIANCE EXPLAINED
Within-District .377 .192 .177
Between-Districts .182 .217 .45 8

 

Note: Entries are unstandardized coefﬁcients and numbers in parenthesis are standard errors;
All outcome variables were standardized (mean=0, standard deviation=1)
* p < .05, ** p < .01, *** p < .001

As in the perceived risk model, the effects of district and individual level variables
on behavioral adaptations were assumed to be the same as those on fear of crime. At the
district level, the results suggested that none of the district-level variables had signiﬁcant

effects on constrained actions, cautious actions, or active defense. At the individual-level,

162

 

victimization, indirect victimization, female, and fear were positively associated with
constrained actions, which was consistent with the expectation. Victims, citizens who
had family members, relatives, or friends as victims of crime, females, and citizens
perceiving higher levels of fear of crime were more likely to adopt constrained actions.
The hypothesized relationship between fear and constrained actions was supported (H5c),
while that between perceived risk and fear was not supported (H5b). That is, citizens
perceiving higher levels of fear of crime were more likely to take constrained actions,
while the effect of perceived risk on constrained actions was not statistically signiﬁcant.
The citizen-level variables altogether accounted for about 37 percent of the within-district
unit variance in constrained actions.

Perceived community-policing activities, perceived cohesion, female, age, SES, and
fear were positively associated with cautious actions, while neighborhood watch was
negatively associated with cautious actions. Citizens perceiving higher levels of
community-policing activities, those perceiving higher levels of community cohesion,
females, the elderly, those in higher levels of socioeconomic status, and those perceiving
higher fear of crime were more likely to adopt cautious actions. On the other hand, those
who had a neighborhood watch team in the community were less likely to adopt cautious
actions. As in constrained actions, the hypothesized relationship between fear and
constrained actions was supported (H5c), while that between perceived risk and fear was
not supported (H5b). That is, citizens perceiving higher levels of fear of crime were
more likely to take cautious actions, while the effect of perceived risk on cautious actions
was not statistically signiﬁcant. The citizen-level variables accounted for about 19

percent of the within-district unit variance in cautious actions.

163

Only perceived community-policing activities and victimization were signiﬁcantly
associated with active defense. The effects of these predictors on active defense were
positive. That is, citizens perceiving higher levels of community-policing activities in the
neighborhood were more likely to adopt active defense tactics, and crime victims were
more likely to adopt active defense tactics. Unlike the previous two behavioral
adaptations (i.e., constrained actions and cautious actions), the hypothesized relationship
between fear and active defense as well as between perceived risk and active defense was
not supported (H5b, H5c). That is, the effects of fear and perceived risk on active
defense were not statistically signiﬁcant. The citizen-level variables accounted for about
18 percent of the within-district unit variance in active defense.

As stated, one of the objectives of estimating these models of behavioral adaptations
was to investigate associations between perceived risk and behavioral adaptations as well
as between fear and behavioral adaptations. Contrary to the hypothesis, perceived risk
was associated with none of the behavioral adaptation measures (H5b). That is,
perceived risks of victimization had no statistically signiﬁcant effects on constrained
actions, cautious actions, or active defense. On the other hand, fear of crime had
signiﬁcantly positive effects on constrained actions and cautious actions, while the effect
on active defense was not signiﬁcant. Citizens perceiving higher levels of fear of crime
were more likely to take constrained actions and cautious actions. The hypothesized
relationship between fear and behavioral adaptations, therefore, was partly supported
(H5c). It was interesting to see the results that fear of crime was associated with
constrained actions and cautious actions, while perceived risk was not linked to any of

the behavioral adaptations measures.

164

(3) Summary of Findings

Figure 5 presents a path diagram to summarize and synthesize the ﬁndings from the
multilevel models for fear, risk, and behavioral adaptations. Overall, the ﬁxed effects
hierarchical models revealed some general patterns. In the community context model,
mobility, economic disadvantage, and crime were positively, and population density was
negatively associated with fear of crime. The hypothesized relationships between
mobility, economic disadvantage, crime and fear, therefore, were supported (Hla, Hlb,
H 1 c). On the other hand, the hypothesized relationship between population density and
fear was not supported (Hld). Rather, the relationship between population density and
fear was the opposite of the hypothesis. Interestingly, unlike their signiﬁcant inﬂuence
on fear of crime, none of the predictors in the community context model was signiﬁcant
in estimating perceived risk or behavioral adaptation measures.

In the community concern/control model, none of the predictors in this model had
signiﬁcant effects on fear of crime. The hypothesized relationships, therefore, were not
supported between perceived community-policing activities, satisfaction with police,
perceived cohesion, neighborhood watch and fear (H2a, H2b, H2c, H2d). On the other
hand, these predictors had some signiﬁcant effects on perceived risk or behavioral
adaptation measures. More speciﬁcally, perceived community-policing activities had
positive effects on cautious activities and active defense. Satisfaction with police had
positive effects on perceived risk, which was an unexpected result. Perceived cohesion
had negative effects on perceived risk, while it had positive effects on cautious actions.

Neighborhood watch had negative effects on cautious actions. Finally, the structural path

165

dimensions indicated that among the predictors in this model, only perceived cohesion
was indirectly and negatively linked to fear via reducing perceived risk.

Community Context

  
 

Density
Mob

  
  
    
 

Disadvantage

 
 
 
 

Crime

Community Concern/Control
COP

Satisfaction

  
   
 

Cohesion

Watch

  
   
  

Disorder

   
 
 

 

++

 

Incivility

 

Victimization Constrained

   
   
  

  

+

 

Victim

 

  

Indirect
Media
SES

    
 

Cautious

   

  
  

  

Female

 

Active

     

Age

Figure 5. A Path Model of Fear of Crime (signiﬁcant effects only).

The sole predictor in the disorder model, perceived incivility, had positive effects on
fear, hence the hypothesized relationship between perceived incivility and fear was
supported (H4). Interestingly, however, this perceived incivility did not have any
signiﬁcant effects on perceived risk or behavioral adaptation measures.

In the victimization model, victimization had no signiﬁcant effects on fear, and the
hypothesized relationship was not supported (H3a). On the other hand, the effects of

victimization were signiﬁcantly positive on perceived risk, constrained actions, and

166

active defense. Even though, there was no direct effect of victimization on fear,
victimization was indirectly associated with fear via increasing perceived risk,

considering the signiﬁcant association between perceived risk and fear. Indirect
victimization had signiﬁcantly positive effects on fear of crime. The hypothesized
relationship between indirect victimization and fear, therefore, was supported (H3b).
Indirect victimization also had signiﬁcantly positive effects on perceived risk and
constrained actions. Surprisingly, crime on media had signiﬁcant effects on neither fear
nor any of the other outcome measures. The hypothesis between crime on media and fear,
therefore, was not supported (H3c).

The victimization model also includes vulnerability measures such as gender, age,
and SES. Female had signiﬁcantly positive effects on fear, which supported the
hypothesized relationship between the two (H3c). The inﬂuence of being a female on
other outcome measures was also signiﬁcant. That is, being a female had positive effects
on constrained actions and cautious actions. On the other hand, age and SES had no
signiﬁcant effects on fear. The hypothesized relationships between age, SES and fear,
therefore, were not supported (H3d, H3 f). In addition, the effects of these predictors on
perceived risk, constrained actions, and active defense were not signiﬁcant. The only
outcome statistically inﬂuenced by age and SES was cautious actions. Both age and SES
had positive effects on cautious actions.

It is also worth noting some structural dimensions among outcome measures.
Perceived risk had signiﬁcantly positive effects on fear, but it did not have any signiﬁcant
effects on all of the behavioral adaptation measures. The hypothesized relationship

between perceived risk and fear was supported (HSa), while the relationships between

167

perceived risk and behavioral adaptation measures (i.e., constrained actions, cautious
actions, active defense) were not supported (H5b). On the other hand, fear of crime had
positive effects on constrained and cautious actions, which partly supported the
hypothesized relationships between fear and behavioral adaptations (H5c). Perceived
risk, therefore, was indirectly associated with constrained and cautious actions via
increasing fear of crime. This result indicated that the hypothesis for the structural
association among perceived risk, fear, and behavioral adaptations was partly supported
(H5d).

Finally, perceived risk acted well as a mediator between some conceptual predictors
and fear. More speciﬁcally, perceived risk mediated the effects of perceived cohesion,
victimization, and indirect victimization on fear of crime. That is, regardless of their
direct effects on fear, these predictors were connected indirectly to the levels of fear via
reducing or enhancing perceived risk. Overall, the results supported the revised causal
frame over the risk interpretation model by Ferraro (1995). It was indicated that the
structural associations were well estimated when perceived risk was considered as a
major mediator between conceptual factors and fear, and when behavioral adaptations
were considered as responses to fear as well as perceived risk. This chapter reported the
major research ﬁndings. The next chapter will discuss the implications of the results in

both theoretical and practical terms.

168

CHAPTER SEVEN
DISCUSSION AND CONCLUSION

1. Discussion

The purpose of this study was to test four conceptual models of fear of crime—the
community context model, the community concern/control model, the disorder model,
and the victimization model—for directional accuracy and the ability to explain fear of
crime in Seoul, the ﬁfth largest city in the world. This research also sought to discover
the effect of social conditions and perceived community policing activities on fear of
crime. More speciﬁcally, this study was aimed at understanding how district
environments and programs of community policing inﬂuence citizens’ fear of crime in
this city. The objective of this research included evaluating the causal frame revised
from Ferraro’s risk interpretation model in the Seoul context. Different from Ferraro’s
model, all predictors were considered in conceptual or theoretical models, perceived risk
was considered as the major mediator between conceptual factors and fear, and
behavioral adaptations were considered as the response to fear as well as perceived risk.
In this chapter, these research purposes are discussed based on the ﬁndings and in
comparison to previous research in Asia and the United States and other Western
countries.
(1) District Level Conditions and the Community Context Model

District-level social conditions such as mobility, economic disadvantage, and
crime were positively associated with fear of crime. Consistent with the hypotheses,
citizens residing in districts characterized by higher mobility, higher economic
disadvantage, and higher crime reported higher levels of fear of crime. These ﬁndings

were best interpreted in light of theoretical insights drawn from Shaw and McKay (1942),

169

Sampson and Groves (1989), Bursik and Grasmick (1993), and their followers, indicating
that social disorganization theory offered powerful constructs for explaining fear of crime
in this city. The ﬁfth largest city in the world, Seoul has been rapidly industrialized as a
main center of economic and political activities. As in most developing cities, rapid
industrialization accompanied by economically disadvantaged groups of people,
residential mobility, and other negative urban characteristics might result in social
disorganization and the deterioration of informal social control mechanisms in Seoul,
despite the long Confusion tradition of private and parochial level social controls.
Accordingly, therefore, social disorganization exogenous variables (i.e., mobility,
economic disadvantage) and negative social conditions (i.e., crime) were responsible for
higher levels of fear of crime in this city.

While these social conditions offer solid constructs for explaining fear of crime,
the results revealed that they were not directly responsible for perceived risks of
victimization or behavioral adaptation measures. The path model in ﬁgure 5 indicated,
however, that these district-level social conditions were responsible for fear, and fear was
structurally linked to both perceived risk and behavioral adaptations. It is important to
note, therefore, that these social conditions speciﬁcally inﬂuence behavioral adaptations
(i.e., constrained actions, cautious actions) indirectly, if not directly, via increasing fear.
District-level social conditions, therefore, were important in explaining the quality of life
among Seoulites, considering the signiﬁcant roles of fear and constrained behaviors in
further deteriorating social control in communities (Markowitz et al., 2001; Taylor, 2001).
Social disorganization theory has not been examined in its entirety in this study, but the

tests with these major social disorganization predictors in Seoul revealed that theoretical

170

mechanisms of rapid industrialization, negative social conditions, social disorganization
or deterioration of neighborhood cohesion, crime, and fear developed in Western settings
(see Sampson et al., 1997; Shaw & McKay, 1942; Taylor, 2001) generalized well to the
Seoul context.

Contrary to the hypothesis, however, population density was negatively
associated with fear. Citizens residing in districts characterized by higher population
density expressed lower levels of fear of crime, which was contrary to the hypothesized
expectations. This ﬁnding was not consistent with the recognition of population density
as one of the indicators of crime in Western research (see Osgood & Chamber, 2000;
Park & Burgess, 1925). Park & Burgess (1925) realized that Zone H, as the zone of
transition, was responsible for much of the crime and delinquency in the ﬁve concentric
zones in Chicago. Located right next to the central business district, this zone was the
most disorganized area, characterized by poor residents, population density and turnover,
and deteriorated buildings. Perhaps the process of city development in Seoul, compared
to the situation in Western cities, may provide an answer for the negative association
between population density and fear. The distribution of population in the ﬁve concentric
zones (Park & Burgess, 1925) showed that poor residents were residing in Zone H, blue-
collar residents were in Zone ID, and middle-income residents with yards and garages
were in Zone IV. Ecological analysis of urban life in the US, therefore, indicated that
wealthy citizens tended to reside toward city borders and the poorest residents were right
next to the central business districts. On the other hand, only wealthy citizens appear
able to afford residing toward the center of the city in Seoul. As the comparison between

p0pulation density (see Figure 3) and population living in poverty (Figure 4) illustrates,

171

contrary to the situation in US. cities, poor citizens resided toward the city border areas
of lower population density in Seoul, which suggested urban life in this city was different
from that of major US. cities. Due to the availability of quality education and positive
social networks, wealthy citizens may tend to reside in inner city areas of higher
population density in Seoul. Population density, therefore, does not appear to be an
indicator of social disorganization in this city. In such a situation, perspectives from the
routine activities theory (Cohen & Felson, 1979) may supplement the social
disorganization perspectives in explaining the negative association between population
density and fear of crime. Surrounded by many dependable residents, citizens residing in
districts characterized by higher population density in Seoul may perceive lower levels of
fear of crime due to both physical and psychological guardianship, despite the
possibilities of more opportunities of crime, more motivated offenders, and higher
complexity and anonymity.7 The results in the community context model in the Seoul
context, therefore, suggest a differential effect of social conditions due to the unique
social structural characteristics of Seoul.
(2) Community Policing and the Community Concern/Control Model

One suggestion is that fear rises as concerns about the level of formal and
informal social control in the neighborhood increases (Taylor & Hale, 1986). The results
indicated that formal social control dimensions such as perceived community policing
activities and satisfaction with police had no signiﬁcant effects on fear of crime, which
was not consistent with the hypotheses and the ﬁndings in previous research. Prior
studies in Western contexts reported mixed results regarding the effect of community

policing activities on fear (Zhao et al., 2002). In some research, community policing was

172

not a signiﬁcant factor, but in other studies, citizens perceiving higher levels of
community policing activities expressed lower levels of fear of crime. On the other hand,
public satisfaction with or conﬁdence in the police were consistently associated with fear
(Box et al., 1988; McGarrell et al., 1997), indicating that Western citizens satisﬁed with
or conﬁdent in the police expressed lower levels of fear of crime. In the Seoul context,
however, the results of this study suggested that neither community policing nor
satisfaction with police played any role in explaining citizens’ fear of crime. Perhaps,
traditional perceptions of the police as authority ﬁgures based on the Confucian ideology
of respect for authority (J 00, 2003), as well as the historically carved image of the police
as the arm of the state throughout colonization and military regimes (Pyo, 2002), might
be responsible for the insigniﬁcant role of community policing activities and satisfaction
with police in explaining citizens’ fear of crime. This stands in contrast to the results in
the Western context where citizens perceive the police as a public partner or protector
(Trojanowicz & Dixon, 1974) or a functional understanding of the police as an agency
responsible for crime control (Lab & Das, 2003).

The historical image of the police as an agency for the government might make it
difﬁcult for Seoul citizens to accept the role of police as a legitimate function for a public
partnership or protection. Scholars have recognized that respect for law enforcement and
acceptance of the law as being legitimate as some of the prerequisites of the successful
implementation of community policing (Davis et al., 2003; Grarnckow, 1995; Manning,
1984). In addition, deﬁance theory and procedural justice emphasize legitimacy as well
as fairness (Sherman, 1993; Tyler, 2003) for law enforcement to be successful. Although

community policing may ultimately increase the legitimacy and perceived fairness of the

173

police in Korea, this study does not ﬁnd that such perceptions inﬂuence fear of crime at
this point in time.

In addition, predictors of informal social control dimensions such as perceived
neighborhood cohesion and neighborhood watch were not signiﬁcantly associated with
fear of crime, which was not consistent with the hypothesized expectations. This was a
surprising result considering the long tradition of informal social controls in this cultural
context as well as the consistently negative effects of these predictors on fear in Western
contexts (Bursik & Grasmick, 1993; Gibson et al., 2002; Sampson et al., 1997). The
limited variation in neighborhood cohesion in urban areas in Korea due to cultural and
ethnic homogeneity might be responsible for this result, compared to the high variation of
neighborhood cohesion in Western countries due to the unequal distribution of race and
wealth as well as cultural diversity and conﬂicts. Furthermore, based on the inﬂuence of
Confucianism in Asian countries, the emphasis on informal social controls has largely
characterized the culture in Korea (Cao & Cullen, 2001; Wong, 2001). Neighborhood
cohesion and neighborhood watch as factors of fear of crime, therefore, might be more
important in individualistic Western societies, and would possibly be less important
estimating fear of crime in the Seoul context due to traditionally stable informal social
networks. It is worth noting, however, that even though neighborhood cohesion had no
direct effects on fear of crime, it contributed to the reduction of the levels of fear of crime
via decreasing perceived risks of victimization in Seoul (see Figure 5), which suggested
that neighborhood cohesion was playing an indirect role for the reduction of fear in the

Seoul context.

I74

Unlike their insigniﬁcant effects on fear of crime, some of the predictors in the
community concern/control model were signiﬁcantly associated with perceived risk or
behavioral adaptation measures. Perceptions of community policing activities had
positive effects on both cautious actions and active defense. Citizens perceiving more
community-policing activities were likely to adopt more cautious and active defense
tactics. This is explained convincingly based on the emphasis of community policing
activities in this city, which included sending out ofﬁcers to educate citizens in strategic
tactics such as cautious actions and active defense (i.e., carrying a self-defense weapon).
As in many Western police agencies, the implementation of community policing in
Korean police agencies is programmatic. Citizen police academies, education of strategic
defense tactics, and encouragement of active defense are some of the examples. Police
ofﬁcers in Seoul are often encouraged to attend monthly meetings with residents and
provide information and education for strategic and active tactics for defense. This
ﬁnding indicated that the police strategies for these community-policing programs
worked in the way the police intended, regardless of the effectiveness of these tactics on
crime prevention or fear reduction.

Interestingly, satisfaction with police was positively associated with perceived
risk, which was the opposite of the hypothesized relationship. That is, citizens with
higher satisfaction with police were likely to perceive higher risks of victimization. This
ﬁnding was difﬁcult to understand. Perhaps, citizens perceiving higher risks of
victimization would like to rely on the police. Such a psychological aspect of the reliance
on the police from citizens perceiving higher risks may be responsible for their higher

satisfaction with police. This inverse of speculation may help interpret the positive

175

association between satisfaction with police and perceived risk, but clearly it requires
further investigation.

Perceived cohesion had a negative effect on perceived risk, while it had a positive
effect on cautious actions. Citizens noticing greater cohesion perceived lower risks of
victimization, which was consistent with the hypothesis. This ﬁnding was also consistent
with the scholarly notion that neighborhood cohesion enhances a sense of safety (Baumer
& Hunter, 1979; LeBailly & Gordon, 1981). Irnportantly, due to this negative effect of
perceived cohesion on perceived risk, perception of cohesion was linked to decreased
levels of fear via reducing perceived risks of victimization. On the other hand, citizens
perceiving greater cohesion were likely to adopt more cautious actions. The measure of
cautious action in this study was related to strategic tactics for crime prevention,
sometimes collaborating with neighbors. Accordingly, citizens might adopt more
strategic tactics when they resided in a community with stronger relational ties with
neighbors. As stated, even though neighborhood cohesion was not signiﬁcantly
associated with fear of crime, the results indicated that neighborhood cohesion is
structurally connected with fear and is important in explaining aspects of quality of life
among Seoulites due to its negative inﬂuence on perceived risk and positive inﬂuence on
cautious actions. Interestingly, however, citizens residing in the community with
neighborhood watch teams were less likely to adopt cautious actions, perhaps due to their
reliance on the role of neighborhood watch teams for crime prevention activities.

(3) The Disorder Model
Disorder signals physical lack of concern about the neighborhood and social lack

of adherence to norms of public behavior (Taylor & Hale, 1986). The results supported

176

this disorder thesis, suggesting that citizens perceiving physical incivilities (i.e., neglected
places, rubbish) and social incivilities (i.e., delinquent juveniles, drunken people) in the
neighborhood expressed higher levels of fear of crime in Seoul. This is consistent with
research ﬁndings in Western contexts (see Ross & Jang, 2000; Skogan, 1990; Taylor &
Hale, 1982), broken windows thesis (Wilson & Kelling, 1982), and disorder and decline
thesis (Skogan, 1990). As scholars have noted, rapid urbanization envelops communities
with high social and physical incivilities which stimulate not only general anxiety but
also fear by signaling that residents are surrounded by symbolic and actual threats of
victimization. As the capital of a small country that has grown into the ﬁfth largest city
in the world, Seoul is the symbol of the economic miracle of South Korea. The
signiﬁcantly positive association between perceived incivilities and fear of crime
supported the thesis linking urbanization, disorder, crime, fear, and community decline
(Skogan, 1990; Wilson & Kelling, 1982) in the Seoul context.
(4) The Victimization Model

There was no direct effect of victimization on fear of crime, but the effect of
victimization on fear was indirect via increasing perceptions of risks of victimization.
This ﬁnding, even though it was contrary to the hypothesis, was not surprising
considering the inconsistent ﬁndings in prior research and the debate that being a victim
might make people more cautious, but it might not necessarily make people more fearful
(DuBow et al., 1979; Lee, 1998). Rather, this result was consistent with some of the
research ﬁndings in the Western context (Garofalo, 1979; Liska et al., 1988; McGarrell et
al, 1997), the paradox of the discrepancy between victimization pattern and fear pattern

(i.e., young males versus elderly women), and Agnew’s (1985) notion that victims tended

177

to experience less fear because of their neutralization by enhancing their ability for
defense. Victimization in Seoul was only indirectly associated with fear via enhancing
perceived risks of victimization.

On the other hand, indirect victimization had a signiﬁcantly positive effect on fear
of crime as well as on perceived risks. Indirect victimization, therefore, was associated
with fear directly and indirectly via increasing perceived risks. Indirect victimization in
this study was measured based on the knowledge of victimization through social contacts.
These ﬁndings suggested that indirect victimization was more signiﬁcant in explaining
fear of crime among Seoulites, which is consistent with the notion of a crime multiplier
(Taylor & Hale, 1986) and the notion that hearing of the victimization ﬁom a relative,
neighbor, or ﬁiend stimulates one’s full scope of imagination (Hale, 1996; Skogan &
Maxﬁeld, 1981). The web-like network among neighbors and residents in Korea may
explain this signiﬁcant role of indirect victimization in explaining fear. Due to the
traditionally intense informal social networks among Koreans (J 00, 2003), gossip and
crime related news spreads easily, which might act as a signiﬁcant source of the vicarious
experience of victimization.

Regarding the association between victimization and behavioral adaptations, it
was interesting to see that victims were more likely to adopt active defense tactics such as
carrying self-defense weapons, while indirect victims were more likely to adopt
constrained actions. Again, considering the negative role of constrained actions in
community deterioration, the inﬂuence of indirect victimization (i.e., having friends,
relatives, and family members who were victimized in the past) was more inﬂuential than

victimization itself to quality of life in the Seoul context.

178

In the case of indirect victimization, scholars have noted that crime on media
serves as another source of vicarious experience of victimization (Skogan & Maxﬁeld,
1981), in addition to the informal social network. Citizen perception of crime in the
media, however, was not signiﬁcantly associated with fear of crime in this study. In
addition, crime on media was not signiﬁcantly associated with perceived risk or
behavioral adaptation measures. One possible explanation for this is that Seoulites share
very similar sources of media (i.e., newspapers and broadcasting services). Such
homogeneity in media services may be responsible for the insigniﬁcant role of crime on
media in explaining fear, perceived risk, and behavioral adaptation measures. Another
possibility was related to the weakness in the method that crime on media was measured
based on a single survey item in this study. Utilizing multiple items to include the time
exposed to crime related news and the sources (i.e., television versus newspaper) that the
respondents rely on might generate different results.

The victimization model also includes the concept of vulnerability. The
vulnerability thesis notes that some people, regardless of their actual victimization,
express higher levels of fear due to their social and physical vulnerabilities (Killias, 1990;
Skogan & Maxﬁeld, 1981). Based on this thesis, it was hypothesized that citizens of
lower socioeconomic status would likely express higher levels of fear of crime because
they have fewer resources to handle the consequences of victimization, and women and
the elderly express higher levels of victimization because they have les physical ability to
defend themselves. Among the predictors of gender, age, and SES, however, only gender
was signiﬁcantly associated with fear of crime in this Seoul study. Females expressed

signiﬁcantly higher levels of fear of crime. In addition, females were more likely to

179

adopt constrained and cautious actions. On the other hand, both age and SES were not
signiﬁcant predictors of fear or constrained actions. Rather, they had positive effects
only on cautious actions. The elderly and the rich were likely to adopt cautious or
strategic actions for crime prevention. The signiﬁcant role of gender in fear, constrained
actions, and cautious actions, compared with the limited role of age and SES, suggested
that gender was much more important than any other vulnerability measure in explaining
quality of life aspects in Seoul. This result is consistent with the traditional Confucian
ideology emphasizing passive and obedient roles for women (J 00, 2003; Lee, 2003; Yoon,
1998), as well as traditions of educating citizens to respect the authority of the elderly.
Females might perceive higher levels of fear and be more likely to adopt constrained and
cautious actions due to their socially low status as well as their physical vulnerability. On
the other hand, the elderly might enjoy the traditional respect for age, and so might not
perceive higher levels of fear of crime and not necessarily adopt constrained action,
despite their physical weakness. Surprisingly, even a negative association between age
and fear of crime was reported in China (Curran & Cook, 1993) where the elderly were
less fearful than younger citizens probably because of their high social status based on
Confucian tradition. This result is in contrast to the ﬁndings in the Western context
where the elderly as well as those in lower social classes expressed higher levels of fear
and adopted more constrained actions (Kury et al., 2001; Skogan & Maxﬁeld, 1981;
Taylor & Hale, 1986). In addition, compared to their culturally and racially diverse
counterparts in Western context, citizens of lower socioeconomic status in Seoul may
perceive less social vulnerability due to the homogeneity of race and culture. This

situation may explain the insigniﬁcant association between SES and fear as well as SES

180

and constrained actions in this city. In contrast to the ﬁndings on fear, the results in this
study indicated that the elderly and the citizens of higher socioeconomic status might
tend to adopt cautious actions probably due to their maturity and their willingness to
protect their wealth in the Seoul context.
(5) Perceived Risk, Fear, and Behavioral Adaptations

Unlike Ferraro’s (1995) risk interpretation model in which both perceived risk
and behavioral adaptation were considered as major mediators between conceptual
predictors and fear, the revised model for this study considered perceived risk as the
major mediator between conceptual predictors and fear and relocated behavioral
adaptations as the response to fear as well as perceived risk. The results indicated that
some conceptual predictors (i.e., perceived cohesion, victimization, indirect
victimization) had signiﬁcant effects on perceived risk, and perceived risk had
signiﬁcantly positive effects on fear of crime. These ﬁndings supported the interpretation
that perceived risk mediated the effects of perceived cohesion, victimization, and indirect
victimization on fear of crime. That is, perceived cohesion, victimization, and indirect
victimization were linked to levels of fear of crime, via decreasing and increasing
perceptions of risk of victimization. Perceived risk, therefore, acted as the mediator
between these conceptual predictors and fear of crime. This result was consistent with
Ferraro’s (1995) model in which perceived risk performed well as a mediator between
predictors and fear.

Based on the reciprocal relationship between fear and constrained behavior (Liska,
et al., 1988; Taylor, et al., 1986), Ferraro (1995) focused on the effect of constrained

behavior on fear, and he reported the positive effect of constrained behavior on fear. As

181

stated, the revised model in this study, however, relocated behavioral adaptations (i.e.,
constrained actions, cautious actions, active defense) as the response to fear as well as
perceived risk to investigate the effect of fear as well as perceived risk on constrained,
cautious, and active actions. The results indicated that behavioral adaptation measures
were well explained when they were considered as the response to fear of crime. Fear of
crime had positive effects on both constrained and cautious actions, which was consistent
with the understanding of fear as a facilitator of community deterioration (Garofalo,
1981; Skogan & Maxﬁeld, 1981). Those who perceived higher levels of fear of crime
were more likely to adopt constrained and cautious actions, perhaps as their responses to
fear. On the other hand, none of the behavioral adaptations were directly inﬂuenced by
perceived risk. Perceived risk was indirectly linked with behavioral adaptations via
increasing fear. In the case that behavioral adaptation was located between perceived risk
and fear as in Ferraro’s (1995) model, therefore, perceived risk would not be linked to
behavioral adaptation measures, considering the lack of signiﬁcant association between
perceived risks and behavioral adaptation measures in this study. Rather, the results of
this study indicated that perceived risk was indirectly linked to behavioral adaptation
measures via increasing fear of crime, which suggested that behavioral adaptations were
well estimated when they were considered as the response to fear as well as perceived
risk. The results, therefore, supported this revised model over F erraro’s (1995) risk
interpretation model in the Seoul context.
(6) Cross Level Interaction

Finally, this study recognized the importance of the interaction between social

contexts and individual-level perceptions in investigating factors of fear of crime. For

182

example, the positive effect of volunteer activities on fear of crime did not have the same
effect in economically disadvantaged districts. Speciﬁcally, the interaction between
district-level economic disadvantage and citizen-level volunteer activities suggested that
citizens volunteering for crime prevention activities in economically disadvantaged
districts perceived lower levels of fear of crime. In contrast, crime prevention volunteers
in economically advantaged districts experienced increased fear. A possible
interpretation is that crime prevention activities might increase the suspicion of others
through these activities (Zhao et al., 2002) since citizens might recognize problems they
have not previously perceived due to the high quality of life in wealthy districts. Citizens
in disadvantaged districts might address their fear through such volunteer activities. On
the other hand, the positive effect of the interaction between crime and perceived risk
suggested that the positive effect of perceived risk on fear was heightened in districts
characterized by higher crime, probably due to the existence of both perceived and actual
risks of victimization. That is, perceived risk of victimization may accelerate fear in
environments of higher crime.

Unlike the positive effect of both mobility and perceived incivility on fear of
crime when they were assessed individually, citizens perceiving higher incivilities
expressed lower levels of fear in districts characterized by higher mobility. The negative
bivariate correlation between mobility and aggregated perceptions of incivility indicated
that the patterns of mobility and perceived incivility were different. The negative effect
of their interaction suggested that mobility, like population density, might be somewhat
different from that of Western cities. Based on this result, it was reasonable to speculate

that mobility might include some dimensions other than social disorganization in the

183

Seoul context. As wealthy citizens reside in districts characterized by higher population
density (see Figure 3 and Figure 4), high mobility may also be associated with the
popularity of the districts. A study indicated that the population of Seoul doubled
between 1970 and 1988, but 80 percent of the population growth was concentrated in
several wealthy districts in the part south of Han-River (Kang, 1989). Seoulites tended to
move into these popular districts due to quality education for their children and move out
from there after some period due to the high costs of living. Even though mobility and
perceived incivility were associated with fear, citizens perceiving incivilities in higher
mobility districts expressed less fear than those perceiving incivilities in lower mobility
districts, probably due to the high quality of living in some high mobility districts despite
their perceptions of incivilities. Overall, the cross-level interactions indicated that the
effects of individual—level predictors are not to be readily generalized without considering
social contexts.
2. Conclusion and Policy Implications

The study of fear of crime goes beyond learning about the perception of safety in
societies. The primary concern is how community environments, individual traits, and
formal and informal social control agents contribute to the quality of life in communities.
Drawing on the conceptual models of fear of crime and the revised causal frame of the
risk interpretation model, this study investigated conceptual factors of fear of crime in the
Seoul context. The results indicated that the community context model, the disorder
model, and the victimization model were well applicable in Seoul. For example, district-
level social conditions, perceived incivility, direct/indirect victimization, and gender were

important in explaining fear of crime among Koreans residing in the megalopolis. On the

184

 

 

other hand, the community concern/control model, especially citizen perception of the
formal social control dimensions (i.e., community policing activities, satisfaction with
police) did not play an important role in determining fear of crime among Seoulites.
While Western scholars recognized social disorganization indicators (i.e.,
mobility, economic disadvantage, population density) as predictors of fear of crime, the
results presented here suggested some of these indicators are not readily generalized in
the Seoul context. Speciﬁcally, population density and the interaction between mobility
and perceived incivility inﬂuenced citizens’ fear of crime in an inverse way in this city.
Contrary to the hypothesized expectation, citizens residing in districts characterized by
higher population density expressed higher levels of fear, and citizens perceiving higher
incivilities in higher mobility districts reported lower levels of fear. This was not readily
understandable without recognizing the process of city development in this country. The
process of city development progressed in a way to attract wealthy citizens to the inner
city districts in Seoul. The Korean government intentionally supported policies of
hosting luxury hotels and department stores as well as good school systems in some inner
districts, especially the part south of the Han River (Kang, 1989). Unlike many major
Western inner-cities, therefore, wealthy citizens tend to reside in high density inner city
districts, which resulted in the unexpected ﬁnding of the signiﬁcantly negative inﬂuence
of population density on levels of fear among Seoulites. In addition, the interaction
between district-level mobility and individual-level perceived incivility suggested that the
patterns of mobility and incivility did not match in Seoul, which was responsible for the
lower levels of fear among citizens perceiving incivilities in higher mobility districts.

These were unexpected ﬁndings because population density and mobility are often

185

. (5:. \ rm I?

 

 

understood as major indicators of social disorganization in Western cities (Osgood &
Chamber, 2000; Park & Burgess, 1925). These ﬁndings in the Seoul context, in
comparison to those in the Western context, suggested a practical policy for urban
planners in Western countries as well as in South Korea. Speciﬁcally, planned city
development might result in a better quality of life in inner city areas.

In the community concern/control model, the insigniﬁcant inﬂuence of
satisfaction with police and community policing activities on fear of crime was
interpreted based on the historically carved image of the public social control agents as
the arm of the state as well as the traditional emphasis on informal social control.
Consistently, studies of community policing in the United States showed that the negative
image of police among residents was an obstacle to community policing in some
communities with historically poor relationship with police (Sadd & Grinc, 1994). For
law enforcement to be effective, scholars suggested that procedural justice (e. g.,
legitimacy, fairness) was essential (Sherman, 1993), and the perception of legitimacy was
critical to citizens’ adherence to rules (Tyler, 2003). In addition, respect for law
enforcement was recognized as a prerequisite of successful implementation of
community policing (Manning, 1984). To obtain legitimacy and to make connections
between policing and citizen perceptions of quality of life (i.e., fear), it appears essential
for the police in South Korea to alter the image of the police to one of service to or
partner with the community. This is one argument for community policing strategies,
which seek to strengthen the connections between the police and the community and to

enhance the sense of security among citizens.

186

..mﬁ

 

 

Even though citizen perceptions of community policing activities did not
inﬂuence their fear of crime, such perceptions did play a signiﬁcant role for their cautious
actions and active measures for crime prevention. As stated, the positive association
between perceived community policing and cautious/active behavioral adaptations might
be due to police programs to educate citizens for such tactics in this city. The results
indicated that the police emphasis on crime prevention tactics was inﬂuential for citizen’s
behavioral adaptations. The insigniﬁcance between perceived community-policing
activities and fear, however, suggested that the police might need to plan their
community policing activities beyond these self-defense tactics. Merely adopting
programmatic approaches is not enough to reduce fear of crime. Scholars in the United
States realize that organizational struggles continue because of misunderstanding and
skepticism of community policing (Greene, 2004; Roth, 2000). This is also true in South
Korea where a programmatic approach is dominant. In a nation in transition from an
authoritative to democratic form of policing, a broader perspective, beyond mere
programmatic strategies, may be critical. Recently, political and police leaders in South
Korea implemented signiﬁcant reforms based on an expanded community policing model
emphasizing the need to build partnerships with citizens and to enhance the community
oriented attitudes of police ofﬁcers. In addition to building an equal partnership, core
elements of community policing (e. g., geographic ownership) should be fully considered
to avoid the simple programmatic approaches and to obtain the effectiveness of
community policing in reducing fear among citizens. Lack of ofﬁcer discretion and
structural constraints due to the bureaucratic aspect of the police organization may also

be some limitations in implementing community policing in Korea. A multidimensional

187

 

 

approach including philosophical issues and internal situations as well as programmatic
dimensions and external relationships is, therefore, suggested.

Finally, gender was very signiﬁcant in explaining fear, constrained actions, and
cautious defense, while age and SES were not signiﬁcantly associated with either fear or
constrained actions. Korea’s cultural emphasis on traditional gender roles and on respect
for authority appears to explain the differential inﬂuence of gender and age on fear.
Similarly, cultural and ethnic homogeneity appears to reduce social vulnerability for
lower SES citizens. The results, therefore, indicated that conceptual predictors of fear
should not be generalized universally without considering culturally unique situations.
Irnportantly, quality of life was much lower for females than males. Females were
suffering from higher fear, more constrained actions, and more caution in Seoul. The
inﬂuence of gender on fear and behavioral adaptations was well contrasted to the
insigniﬁcant role of age in these outcomes. In turn, this result suggested that gender was
associated beyond physical vulnerability, and that gender speciﬁc socialization and the
traditional culture of male domination might still be inﬂuential in South Korea. In
addition to physical vulnerability, scholars have noted gendered socialization process
(Goodey, 1997) and lack of power in society (Stanko, 1995), as factors affecting fear
among females. Political leaders in this country may need to take steps to enhance the
quality of life for females, probably by reducing gender inequality in society along with
the encouragement of attitudes toward equal relationships early in the socialization
process of citizens.

3. Limitations and Future Research

188

3'. " .mk'm. “IM.]".4~

 

 

This dissertation has extended research on communities (i.e., social
disorganization, disorder, formal/informal social control), crime, and quality of life (i.e.,
fear, perceived risk, behavioral adaptations) beyond Western contexts in the course of
developing a causal model of fear of crime in a megalopolis, Seoul. The research is
limited because the data did not provide enough evidence to outline the distinct process
of city development in comparison to Western cities; even though the evidence indicated
that some social disorganization indicators (i.e., population density) are not readily
applicable in this city in explaining fear of crime. Social context was deﬁned as districts
(Gu) with the population of several hundred thousand, and this study was justiﬁed based
on the reasoning that there would be a certain amount of variation in behavior and
perceptions that is explained by district-level environment. As scholars noted, however,
emphasis on micro units of community would be more consistent with the concept of
community as a place where residents share basic social life and norms (Bursik &
Grasmick, 1993; Wooldredge, 2002). Research, therefore, should closely examine the
effect of these social conditions in smaller units of within-group homogeneity in Dongs
(communities) or Tongs (sub-communities) in the Seoul context, as scholars in the
Western context focus on blocks, block groups, or census tracts that are similar on
various census indicators (Wooldredge, 2002). Further investigation of the patterns of
social disorganization indicators (i.e., mobility, density, economic disadvantage,
incivility) in smaller units of this city would provide clearer answers of differential
effects of some of these indicators on fear of crime.

In addition, this study reinforced the need for comparative research. Although

there were similarities in the ﬁndings of this study and previous Western research, there

189

were also important differences. Particularly surprising was the ﬁnding that population
density was negatively associated with fear, and citizen perception of formal social
control dimensions as well as age and SES were not signiﬁcant predictors of fear in the
Seoul context. A distinct process of city development along with history and culture
were discussed in this regard. A comparative perspective of this study relied on prior
research ﬁndings due to the lack of comparative data. A cross-national study at the
global level, therefore, is encouraged in the future to investigate whether theoretical
perspectives and ﬁndings transcend borders.

Methodological limitations are also noted here. This study designed two separate
full models of fear of crime (see Table 10) to help avoid any bias generated from the
district-level sample size (N=25). In hierarchically nested data, two sample sizes exist,
one for group size and the other for number of groups. Scholars argue that a large
number of groups is important as the estimation of the ﬁxed parameters and their
standard errors in the group level variance components are underestimated in small
groups (Snijders & Bosker, 1994). Researchers suggested that, for the same total sample
size of individuals, it is better to spread them across as many groups as possible to reduce
the design effect due to clustering (Reise & Duan, 2003), as it is better to use larger
cluster sizes to place as many individuals as possible in the same group. To perform a
reliable multilevel analysis, some researchers recommend 10 cases within 10 aggregates
(Mok & Flynn, 1998), while others were more strict by suggesting at least 10
observations for each predictor at the aggregate-level (Bryk & Raudenbush, 1992). Even
though intraclass correlation and reliability in the one-way ANOVA model of this study

encouraged multilevel analyses, the sample size at level-2 was a restrictive element in the

190

design for further study including more macro-level variables. Even though evidence
indicated that the data structure of this study allowed up to three or four district-level
predictors, it would be limited especially when the study is designed to include some
more district-level predictors simultaneously.8 Small group-level sample size in relation
to a large number of variables included in the model is linked to insufﬁcient power to
detect signiﬁcant effects due to the low between-group variation, which was not the case
in current study but would possibly be when several more macro-level variables are
included and modeled simultaneously.9

For future research, inclusion of more group-level sample size as well as citizen-
level sample size is highly recommended. If a researcher still perceives a problem due to
the sample size, the Bayesian approach is recommended. Raudenbush and Bryk (2002)
present a Bayesian approach in the case when the number of units is too small to support
a plausible asymptotic normal distributional assumption. The pioneers of multilevel
methods suggest using a Bayesian formulation of such methods when a researcher starts
to feel uncomfortable with the number of level-2 units in the dataset. This approach is
persuasive since Bayesian methods estimate model parameters without relying on normal
approximations. ‘0

Finally, this study modeled a series of hierarchical linear models to assess the
direct and indirect effect of conceptual predictors on fear, perceived risk, and behavioral
adaptation measures. Even though the results revealed a substantial indirect effect of
some conceptual predictors on fear via perceived risks, it could not gauge the actual

magnitude of the effect. Although HLM covers many limitations of conventional

regression analysis to accurately estimate multilevel models, like other conventional

191

techniques, it cannot measure indirect effects, it cannot incorporate latent variables into
the representation of theoretical variables, and it cannot model the interrelationship
among independent variables (Maier, 2001; Raudenbush & Sampson, 1999). Structural
Equation Modeling (SEM) analysis, which is a path analysis with latent variables,
overcomes such limitations, yet it is not designed for multilevel study. Statistical
packages for structural equation modeling analyses such as LISREL and AMOS allow for
the simultaneous testing of the effects of predictor (exogenous) variables and outcome
(endogenous) variables in the speciﬁed structural equation model providing statistical
indexes of the overall ﬁt of the model to the data as well as informative results such as
direct, indirect, and total effects. However, generally, the method is limited to a
continuous outcome, does not allow for missing data, and cannot handle nested nature of
data. Recognizing this limitation, Multilevel Covariance Analysis (MCA), using M-plus
package, seeks to incorporate SEM’s ability to capture the complexity of multivariate
relationships with HLM’s ability to capture the nested nature of those relationships and
HLM’s ﬂexibility in allowing the data to be binary, ordinal, or multinomial (Farmer,
2000). One limitation of MCA is that it is a large sample technique and the group size
should be from 50 to 100. MLwiN incorporated both HLM and SEM’s advantages in the
Bayesian approach in which the weakness of the small sample size is relatively overcome
through computer simulation technique (Maier, 2001). Unlike the conventional
frequentist method relying on large-sample approximations to produce estimates of
model parameters by postulating that the true values of the parameters are ﬁxed and the
data are random, “empirical Bayesian methods make use of both paradigms where a

subset of parameters is estimated and treated as ﬁxed and known values in a subsequent

192

 

Bayesian data analysis technique to estimate the remaining unknown parameters” (Maier,
2001, p. 309). For a future study, multilevel path analysis based on Bayesian approach
could support statistically robust results even with the limited sample size in district-level
and could measure the magnitudes of both direct and indirect effects along with

interrelationships among predictors.

193

 

ENDNOTES

. South Korea, Hong Kong, Singapore, and Taiwan were often described as four
dragon states to be acknowledged as Eastern Asian economic miracles, which was
based on the idea that their rapid industrialization was similar to the image of an
emerging dragon (see Li, 2002).

. Buddhism was accepted in Korea when the peninsula was divided into three
countries: Koguryo in AD. 372, Paekche in AD. 384, and Shilla in AD. 572. It
was the ofﬁcial religion in the united Koryo dynasty (918-1392) and Choson
dynasty (A.D.1392—1910).

. The task force is currently experimenting with the effectiveness of these tactics in
three communities in Gyeonggi province for the past three years. Tactics
included the arrangement of locations of windows and balconies, street lights,
playgrounds, fences, CCTVs, entrance regulations using security code or guard,
emergency alarms, etc. (Korea National Police Agency, 2005).

. The reduction on the high status of investigation expertise due to emphasis on
generalists resulted in resistance by detectives (Rosenbaum & Lurigio, 1994;
Wasson, 1977). Moreover, its demise was largely at the hands of middle level
managers within police organizations, who were interested in maintaining a more
centralized form of command and control. The concept was well designed but the
lack of planning and resistance by mid-level managers in special units made the
success of this strategy difﬁcult.

. It was a general belief that the traditional focus on motorized patrol, rapid

response to calls for service, and retrospective investigation would reduce crime,

194

 

which would eventually decrease levels of fear of crime (Moore & Trojanowicz,
1988)

. “Adding a predictor that models a part of the within-group variability must
decrease the estimate of 02; if this predictor does not model part of between-group
variability then, because unexplained between-group variability remains the same,
the decrease of (I2 must be balanced by an increase of the estimate of Too (Snijders
& Bosker, 1993, p. 347),” therefore resulting in the negative explained variance
between groups.

. Lee (1998) reported a very similar result among Korean Americans in the
Chicago neighborhood, but the ﬁnding could not be generalized to the general
public due to the situations of Korean Americans who tended to reside in selected
communities.

. As an example, in a doctoral dissertation on hierarchical analysis of race, class,
and delinquency in Seattle schools using 634 students nested in 10 schools (Engen,
1996a), the researcher realized that including more than one variable in school
level resulted in highly unstable parameter estimates, and so a series of ﬁnal
models were estimated including one aggregate level variable each time.

. The sample size in higher level may limit the design of a model but it depends on
data structure including intraclass correlation and group—level reliability to
understand whether the model is applicable to the data (Maier, 2001). In addition
to the sample size, therefore, intraclass correlation also affects the accuracy of the
estimates (Willms, 1999). The method of parameter estimation in HLM

(maximum likelihood estimation) provides more power than ordinary least

195

 

10.

squares regression when there is substantial intraclass correlation in the data. The
intraclass correlation coefﬁcients in current study indicated a reasonably
acceptable between group variance to estimate district-level models, considering
other studies reported a large effect associated with differences between groups
with small variance between macro-level units ranging from around 5 to 10
percent (Reisig & Parks, 2000; Sampson ‘& Bartusch, 1998; Sampson et al., 1999).
In addition, the magnitude of group-level reliability shows whether group-level
differences can be modeled with a reliable degree of precision (Raudenbush &
Bryk, 2002). The city-level high reliability (.57) in current study showed that city
means vary substantially across cities, thus encouraging multilevel analyses.

The use of conventional estimation methods, including maximum likelihood,
introduces constraints on the minimum allowable sample size or the degree to
which the hierarchical structure can be sparsely populated since they depend on
normal distribution theory (Maier, 2001; Seltzer, Wong, & Bryk, 1996), which
results in biased parameter estimates in analyses with small sample size.
Bayesian methods allow a solution strategy that produces unbiased estimates and
eliminates the need for directly computing complex integrations without relying
on normal approximations. Bayesian inference supplements the likelihood
equation with prior beliefs the analysts may have about the distributions of the
parameters, via prior distributions (Maier, 2001). This approach speciﬁes a prior
probability distribution for the variance components and then integrates over the
variance components as well as other unknowns in the HLM to obtain a marginal

posterior distribution of interest (Seltzer et al., 1996). Estimation and inference

196

 

via Bayesian simulation methods such as Markov Chain Monte Carlo (MCMC)
are advantageous in various aspects (J ackman, 2000). Large number of simulated
random draws from the joint posterior distribution of all the parameters can be
made instead of ﬁnding simple point estimates for the parameters (Maier, 2001).
These random draws can be used to form a summary of the underlying
distribution to generate more reliable estimates. MLwiN statistical package was

developed for this purpose (Maier, 2001).

197

 

Bibliography

Adams, R. E., Rohe, W. M., & Arcury, T. A. (2005). Awareness of community-oriented
policing and neighborhood perceptions in ﬁve small to midsize cities. Journal of
Criminal Justice, 33, 43-54.

Agnew, R. S. (1985). Neutralising the impact of crime. Criminal Justice and Behavior,
12, 221-239.

Alba, R. D., & Nee, V. (1997). Rethinking assimilation theory for a new era of
immigration. International Migration Review, 31(4), 826-874.

Arnold, H. (1991). Fear of crime and its relationship to directly and indirectly
experienced victimization: A binational comparison of models. In K. Sessar & H.
J. Kemer (Eds.), Developments in crime and crime control research. New York:
Springer-Verlag.

Asakura, T. (1998). Aspects of Yangbanization. Osaka: National Museum of Ethnology.

Baker, M. H., Nienstedt, B. C., Everett, R. S., & McClery, R. (1983). The impact of a
crime wave: perceptions, fear and conﬁdence in the police. Law and Society
Review, 1 7, 319-335.

Baldassare, M. (1986). The elderly and fear of crime. Sociology and Social Research,
70(3), 218-221.

Baumer, T. L. (1985). Testing a general model of fear of crime: Data from a national
sample. Journal of Research in Crime and Delinquency, 22, 239-255.

Baumer, T. L., & Hunter, A. (1979). Street traﬂ‘ic, social integration, and fear of crime.
Evanston, 111: Center for Urban Affairs, Northwestern University.

Bazemore, G., & Grifﬁths, C. (2003). Police reform, restorative justice and restorative
policing. Police Practice and Research, 4(4), 335-346.

Bellah, R., Madsen, R., Sullivan, W. M., Swidler, A., & Tipton, S. M. (1985). Habits of
the heart. New York: Harper & Row.

Bennett, R. R., & F lavin, J. M. (1994). Determinants of fear of crime: The effect of
cultural setting. Justice Quarterly, 11, 357-381.

Bennett, T. (1991). The effectiveness of a police initiated fear reduction strategy. The
British Journal of Criminology, 31(1), 1- 14.

Benson, P. R. (1981). Political alienation and public satisfaction with police services.
Pacific Sociological Review, 24, 45-64.

198

 

Beyers, J. M., Bates, J. E., Pettit, G. S., & Dodge, K. A. (2002). Neighborhood Structure,
Parenting Processes, and the Development of Youths' Extemalizing Behaviors: A
Multilevel Analysis. American Journal of Community Psychology, 31(1/2), 35-53.

Biderrnan, A. D., Johnson, L., McIntyre, J ., & Weir, A. (1967). Report on a pilot study in
the District of Columbia on victimization and attitudes toward law enforcement.
Washington, DC: Government Printing Ofﬁce.

Blumer, H. (1958). Race prejudice as a sense of group position. The Paciﬁc Sociological
Review, 1(1), 3-7.

Box, 8., Hale, C., & Andrews, G. (1988). Explaining fear of crime. British Journal of
Criminology, 28(3), 340-356.

Bristow, A. P. (1969). Effective police manpower utilization. Springﬁeld, Ill: Charles C.
Thomas.

Brooks, J. (1981). The fear of crime in the United States. In B. Gallaway & J. Hudson
(Eds.), Perspectives on crime victims. St. Louis, MO: C. V. Mosby Co.

Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical Linear Models: Applications and
Data Analysis Methods. Newbury Park: Sage Publications.

Burby, R. J., & Rohe, W. M. (1989). Deconcentration of public housing: effects on
residents' satisfaction with their living environments and their fear of crime.
Urban Affairs Quarterly, 25(1), 117-141.

Bureau of Labor Statistics. (1998). Unemployment Rates by Cities and States. Retrieved
20 September, 2004

Bursik, R. J ., & Grasmick, H. G. (1993). Neighborhoods and crime: The dimensions of
effective community control. New York: Lexington Books.

Cancino, J. M. (2002). Nonmetropolitan social disorganization: A multilevel analysis of
quality of life. Unpublished doctoral dissertation, Michigan State University, East
Lansing, MI.

Cao, L., & Cullen, F. T. (2001). Thinking about crime and control: A comparative study
of Chinese and American ideology. International Criminal Justice Reivew, 11,
58-81.

Cao, L., & Hou, C. (2001). A comparison of conﬁdence in the police in China and in the
United States. Journal of Criminal Justice, 29, 87-99.

Cao, L., Stack, S., & Sun, Y. (1998). Public conﬁdence in the police: A comparative
study between Japan and America. Journal of Criminal Justice, 26(4), 279-289.

199

 

Capowich, G. E., & Roehl, J. A. (1994). Problem-oriented policing: Actions and
effectiveness. In D. P. Rosenbaum (Ed.), The challenge of community policing:
Testing the promises (pp. 127-146). Thousand Oaks, CA: Sage.

Chadee, D., & Ditton, J. (2003). Are older people most afraid of crime? Revisiting
Ferraro and LaGrange in Trinidad. British Journal of Criminology, 43(2), 417-
433.

Chang, Y. (1990). Research on fear of crime in China. Police Studies, 13(3), 125-127.

Chen, X. (2002). Social control in China: Applications of the labeling theory and the
reintegrative shaming theory. International Journal of Oﬂender Therapy and
Comparative Criminology, 46(1), 45-63.

Chiricos, T., Eschholz, S., & Gertz, M. (1997). Crime, news and fear of crime: toward an
identiﬁcation of audience effects. Social Problems, 44(3), 342-357.

Cho, S. (2002). A study on community based crime control strategies: An example of
Daegu metropolitan city, South Korea. Dongguk University, Seoul, Korea.

Choi, I., & Gy, K. (1997). A study on relationship between social class and crime. Seoul,
Korea: Korean Institute of Criminology [in Korean].

Chung, J. S. (1997 ). Korean American enterpreneurship is an indication of economic
exploitation. In D. L. Bender (Ed.), Asian Americans: Opposing vieWpoints (pp.
184-191). San Diego, Califomia: Greenhaven Press, Inc.

Clementer, F ., & Kleiman, M. B. (1977). Fear of crime in the United States: a
multivariate analysis. Social Forces, 56(2), 519-531.

Cohen, L., & F elson, M. (1979). Social Change and Crime: A Routine Activity Approach.
In J. Jacoby, Classics of Criminology, 2nd ed. (pp. 66-74). Prospect Heights:
Waveland Press, Inc.

Cole, A. W., & Kelly, D. (1992). Fear and disorder index: Measuring what really matters.
Law and Order, October, 97-100.

Conklin, J. E. (1975). The impact of crime. New York: Macmillan.
Covington, J., & Taylor, R. B. (1991). Fear of crime in urban residential neighborhoods:
Implications of between- and within-neighborhood sources for current models.

The Sociological Quarterly, 32, 231-249.

Cummings, B. (1997). Korea's place in the sun: A modern history. New York: W.W.
Norton & Company.

200

 

Curran, D. J ., & Cook, S. (1993). Growing fears, rising crime: Juveniles and China's
justice system. Crime and Delinquency, 39(3), 296-315.

Davis, J. R. (1990). A comparison of attitudes toward the New York City police. Journal
of Police Science and Administration, 1 7, 233-243.

Davis, R. C., Henderson, N. J ., & Merrick, C. (2003). Community policing: Variations on
the western model in the developing world. Police Practice and Research, 4(3),
285-300.

Dietz, A. S. (1997). Evaluating community policing: quality police service and fear of
crime. Policing An International Journal of Police Strategy and Management,
20(1), 83-100.

Ditton, J ., Bannister, J ., & Gilchrist, E. e. a. (1999). Afraid or angry? Recalibrating the
‘fear' of crime. International Review of Victimology, 6(2), 83-99.

Doob, A. N., & MacDonald, G. E. (1979). Television viewing and fear of victimization:
Is the relationship causal? Journal of Personality and Social Psychology, 3 7, 170-
179.

Dowler, K. (2003). Media consumption and public attitudes toward crime and justice:
The relationship between fear of crime, punitive attitudes, and perceived police
effectiveness. Journal of Criminal Justice and Popular Culture, 10(2), 109-126.

DuBow, F ., McCabe, E., & Kaplan, G. (1979). Reactions to crime: A critical review of
the literature. Washington, DC: National Institute of Law Enforcement and
Criminal Justice, US. Government Printing Ofﬁce.

Dukes, R. L., Hughes, R. H., & Link, W. (1997). Crime survey of citizens in El Paso
County: victimization, fear, and perceptions of police. Colorado Springs, CO:
Center for Justice Studies Colorado University.

Duncan, G. J ., Connell, J. P., & Klebanov, P. K. (1997). Conceptual and Metholological
Issues in Estimating Causal Effects of Neighborhoods and Family Conditions on
Individual Development. In J. Broks-Gunn, G. J. Duncan & J. L. Aber (Eds.),
Neighborhood Poversty: Context and Consequences for Children. New York:
Russell Sage.

Eck, J. E., & Rosenbaum, D. P. (1994). Effectiveness, equity, and efﬁciency in
community policing. In D. Rosenbaum (Ed.), The challenge of community
policing: Testing the promises. California: Sage.

Engen, R. L. (1996). Linking Social Structure and Social Context: A Hierarchical

Analysis of Race, Class, and Delinquency in Seattle Schools. Unpublished
Doctoral Dissertation, University of Washington, Washington, DC.

201

 

Erskine, H. (1974). The polls: Fear of crime and violence. Public Opinion Quarterly, 38,
131-145.

Fattah, E. A. (1993). Research on fear of crime: Some common conceptual and
measurement problems. In W. Bilsky, C. Pfeiffer & P. Wetzels (Eds.), Fear of
crime and criminal victimization. Hannover, Germany: Enke.

F attah, E. A., & Sacco, V. (1989). Crime and victimization among the elderly. New York:
Springer-Verlag.

Ferraro, K. F. (1995). Fear of crime: Interpreting victimization risk. New York: SUNY
Press.

Ferraro, K. F., & LaGrange, R. L. (1987). The measurement of fear of crime.
Sociological Inquiry, 57, 70-101.

Finiﬁer, B. M. (1977). The robustness of cross-cultural ﬁndings. Annals of New York
Academy of Sciences, 285, 151-184.

Fishkin, S. A., Rohrbach, L. A., & Johnson, C. A. (1997). Correlates of youths’ fears of
victimization. Journal of Applied Social Psychology, 27(18), 1601-1616.

Furstenberg, F. F ., Jr. (1971). Public reaction to crime in the streets. The American
Scholar, 40(601-610).

Garofalo, J. (1979). Victimization and the fear of crime. Journal of Research in Crime
and Delinquency, January, 80-97.

Garofalo, J. (1981). The fear of crime: Causes and consequences. Journal of Criminal
Law and Criminology, 72(2), 839-857.

Garofalo, J ., & Laub, J. (1978). The fear of crime: Broadening our perspective.
Victimology: An International Journal, 3(3-4), 242-253.

Geis, K. J ., & Ross, C. E. (1998). A New Look at Urban Alienation: the Effect of
Neighborhood Disorder on Perceived Powerlessness. Social Psychology
Quarterly, 61, 232-246.

George, D., & Yancey, G. (2004). Taking stock of America's attutides on cultural
diversity: An analysis of public deliberation on multiculturalism, assimilation and
intermarriage. Journal of Comparative Family Studies, 35(1), 1-19.

Gibson, C. L., Zhao, J ., Lovrich, N. P., & Gafﬁrey, M. J. (2002). Social integration,

individual preceptions of collective efﬁcacy, and fear of crime in three cities.
Justice Quarterly, 19(3), 537-564.

202

 

Goldstein, H. (1987). Toward community-oriented policing: Potential, basic requirements,
and threshold questions. Crime and Delinquency, 33(1), 6-30.

Goldstein, H., & Rasbash, J. (1996). Improved Approximation for Multilevel Models
with Binary Responses. Journal of Royal Statistical Society, 159(3), 505-513.

Gomme, I. M. (1986). Fear of crime among Canadians: a multi-variate analysis. Journal
of Criminal Justice, 14(3), 249-258.

Goodey, J. (1997). Boys don't cry: Masculinities, fear of crime and fearlessness. British
Journal of Criminology, 37(3), 401-418.

Gordon, A. F., & Newﬁeld, C. (1996). Mapping multiculturalism. Minneapolis, MN:
University of Minnesota Press.

Gordon, M. M. (1964). Assimilation in American life. Oxford: Oxford University Press.

Grabosky, P. N. (1995). Fear of crime, and fear reduction strategies. Current Issues in
Criminal Justice, 7(1), 7-19.

Graham J r., O. L. (2004). Unguarded gates: A history of America's immigration crisis.
New York: Rowan and Littleﬁeld Publishers, Inc.

Green, W. H. (2000). Econometric Analysis (4th edition ed.). New Jersey: Prentice-Hall,
Inc.

Grinc, R. (1994). Angles in marble: Problems in stimulating community involvement in
community policing. Crime and Delinquency, 40, 437-468.

Haghighi, B., & Sorensen, J. (1996). America's Fear of Crime. In T. J. Flanagan & D. R.
Longrnire (Eds.), Americans view crime and justice: A national public opinion
survey. Thousand Oaks, London: SAGE Publications.

Hale, C. (1996). Fear of crime: A review of the literature. International Review of
Victimology, 4, 79-150.

Hale, C., Pack, P., & Salked, J. (1994). The structural determinants of fear of crime: an
analysis using census and crime survey data from England and Wales.
International Review of Victimology, 3(3), 211-233.

Hamilton, C. (1996). Multiculturalism as political strategy. In A. F. Gordon & C.

Newﬁeld (Eds.), Mapping multiculturalism. Minneapolis, MN: University of
Minnesota Press.

203

 

Hidemura, K. (1998). Social change and Christianity in modern South Korean society. In
M. Shima & R. L. J anelli (Eds.), The Anthropology of Korea: East Asian
perspectives. Osaka: National Museum of Ethnology.

Hoffman, D. A., & Gavin, M. B. (1998). Centering Decisions in Hierarchical Linear
Models: Implications for Research in Organizations. Journal of Management,
24(5), 623-641.

Hoffman, V. J. (1982). The development of modern police agencies in the republic of
Korea and Japan: A paradox. Police Studies: The International Review of Police
Development, 5(3), 3-16.

Hong, D. (2003). Social change and stratiﬁcation. Social Indicators Research, 62(1), 39-
50.

Hwang, E., McGarrell, E.F., & Benson, BL. (2005). Public satisfaction with the South
Korean Police: The effect of residential location in a rapidly industrializing nation.
Journal of Criminal Justice, 33, 585-599.

Inglehart, R. (1997). Modernization and postmodernization: Cultural, economic and
political change in 43 societies. Princeton, NJ: Princeton University Press.

Ito, A. (1998). Coordination and Brokerage: Leadership in community development in
rural Korea. Osaka: National Museum of Ethnology.

Ito, K. (1993). Research on the fear of crime: perceptions and realities of crime in Japan.
Crime and Delinquency, 39(3), 385-392.

J ang, J. (2003). Economic crisis and its consequences. Social Indicators Research, 62(1),
5 1-70.

Jeffords, C. R. (1983). The situational relationship between age and the fear of crime.
International Journal of Aging and Human Development, 1 7, 103-1 11.

Jiao, A. Y. (2004). Integration of formal and informal social control in
China.Unpublished manuscript, Rowan University, New Jersey.

J 00, H. (2003). Crime and crime control. Social Indicators Research, 62(1), 239-263.

Kalupahana, D. J. (1976). Buddhist philosophy: A historical analysis. Honolulu:
University of Hawaii Press.

Kang, B. (1989). Divided Seoul: Special districts in the South of Han River and general
districts in the North of Han River. Retrieved March16th, 2006 from

www.spins.co.kr

 

204

 

Katz, C. M., Webb, V. J ., & Armstrong, T. A. (2003). Fear of gangs: A test of alternative
theoretical models. Justice Quarterly, 20(1), 95-130.

Kelling, G. L. (1987). Acquiring a taste for order. Crime and Delinquency, 33, 90-102.

Kelling, G. L., & Coles, C. M. (1996). Fixing broken windows: Restoring order and
reducing crime in our communities. New York, NY: Touchstone.

Kelling, G. L., Pate, T., Dieckman, D., & Brown, C. E. (1974). The Kansas City
preventive patrol experiment: A summary report. Washington, DC: Police
Foundation.

Killias, M. (1990). Vulnerability: Towards a better understanding of a key variable in the
genesis of fear of crime. Violence and Victim, 5, 97-108.

Killias, M., & Clerici, C. (2000). Different measures of vulnerability in their relation to
different dimensions of fear of crime. British Journal of Criminology, 40, 437-450.

Kim, A. (2003). Religious inﬂuences on personal and societal well-being. Social
Indicators Research, 62(1), 149-170.

Kim, H. (2003). Media, the public, and freedom of the press. Social Indicators Research,
62(1), 345-364.

Kim, J., & Lee, J. (2003). Educational attainment and policy. Social Indicators Research,
62(1), 187-195.

Kim, K. (1998). The Confucian construction of a cultural community in contemporary
South Korea. In M. Shima & R. L. Janelli (Eds.), The Anthropology of Korea:
East Asian perspectives. Osaka: National Museum of Ethnology.

Kim, M. (2003). Social welfare system. Social Indicators Research, 62(1), 265-289.

Kim, S. (2003). Rearranging Korean police functions: Community policing and fear of
crime [In Korean]. Seoul, Korea: Korean Institute of Criminology.

Kim, R. (1999). My trek. In P.-G. Min & R. Kim (Eds.), Struggel for ethnic identity:
Narratives by Asian American professionals (pp. 49-58). Walnut Creek, CA: A
Division of Sage Publications, Inc.

Koenig, J. A., & Lissitz, R. L. (2001). The Eﬂects of Centering Method on Parameter
Estimates and Variance Explained in Multilevel Models. Paper presented at the
Annual Meeting of the American Educational Research Association, Seattle.

Kohn, M. L. (1987). Cross-national research as an analytic strategy. American
Sociological Review, 52(December), 713-731.

205

 

Korea National Police Agency. (2002). Korea police annual report [In Korean]. Seoul,
Korea: Korea National Police Agency.

Korea National Police Agency. (2003). Korea police annual report [In Korean]. Seoul,
Korea: Korea National Police Agency Press.

Korea National Police Agency. (2004). Korea police annual report [In Korean]. Seoul,
Korea: Korea National Police Agency Press.

Korea National Statistical Ofﬁce. (2004). Approved statistics in Korea [In Korean].
Retrieved March 16, 2005, from www.nsogo.kr/eng/handbook

Korea National Statistical Ofﬁce. (2005). Statistical facts [In Korean]. Retrieved October
15, 2005, from www.nso.go.kr

 

Kury, H., Obergfell, F. J ., & Ferdinand, T. (2001). Aging and the fear of crime: Recent
results from East and West Germany. International Review of Victimology, 8(1),
75-112.

Kury, H., Obergfell-Fuchs, J., Woessner, G., & Wuerger, M. (2000). Fear of crime and
victimization. In P. Nieuwbeerta (Ed.), Crime victimization in comparative
perspective: Results from the international crime victims survey, 1989-2000. Den
Haag, Netherlands: Boom J uridische uitgevers.

Lab, S. P. (2004). Crime prevention: Approaches, practices and evaluations (5th ed. ).
Cincinnati, OH: Anderson.

Lab, S. P., & Das, D. K. (2003). International perspectives on community policing and
crime prevention. Upper Saddle River, NJ: Prentice Hall.

LaGrange, R. L., & Ferraro, K. F. (1987). The elderly's fear of crime: a critical
examination of the research. Research on Aging, 9(3), 372-391.

LaGrange, R. L., & Ferraro, K. F. (1989). Assessing age and gender: Differences in
perceived risk and fear of crime. Criminology, 2 7, 697-719.

LaGrange, R. L., Ferraro, K. F ., & Supancic, M. (1992). Perceived risk and fear of crime:
Role of social and physical incivilities. Journal of Research in Crime and
Delinquency, 29, 311-334.

Lamon, P. (2000). Crime trends in thirteen industrialized countries. In P. Nieuwbeerta
(Ed.), Crime victimization in comparative perspective: Results from the
international crime victims survey, 1989-2000. Den Haag, Netherlands: Boom
J uridische uitgevers.

206

Lane, J., & Meeker, J. W. (2000). Subcultural diversity and the fear of crime and gangs.
Crime and Delinquency, 46(4), 497-521.

Lauritson, J. L. (2001). The social ecology of violent victimization: Individual and
contextual effects in the NCVS. Journal of Quantitative Criminology, 1 7(1), 3-32.

Lavrakas, P.J., Rosenbaum, D.P., & Kaminski, F. (1983). Transmitting information about
crime and crime prevention to citizens: Evanston newsletter quasi-experiment.
Journal of Police Science and Administration, 11 (4), 463-473.

LeBailly, R. K., & Gordon, M. T. (1981). Community ties and urbanites' fear of crime:
An ecological investigation. American Journal of Community Psychology, 9(6),
653-665.

Lee, A. (2003). Stability and change in Korean values. Social Indicators Research, 62(1),
93-117.

Lee, J. (2003). Income, consumption, and poverty. Social Indicators Research, 62(1),
197-209.

Lee, M. (1998). Fear of crime among Korean Americans in the Chicago area: A
multilevel analysis. Unpublished Doctoral Dissertation, Purdue University.

Lee, M. (2004). Accounting for rapid grth of private policing in South Korea. Journal
of Criminal Justice, 32, 113-122.

Lee, M., & Ulmer, J. T. (2000). Fear of crime among Korean Americans in Chicago
communities. Criminolog, 38(4), 1173-1206.

Lee, S.-K., Sobal, J., & Frongillo, E. A. (2003). Comparison of models of acculturation:
The case of Korean Americans. Journal of Cross-Cultural Psychology, 34(3),
282-296.

Lee, S. Y. (1990). Morning calm, rising sun: National character and policing in South
Korea and in Japan. Police Studies, 13(3), 91-110.

Lee, Y. (1997). Explaining Korean women's fear of crime. International Journal of
Comparative and Applied Criminal Justice, 21(2), 317-329.

Lee, Y. (2002). A cross-national comparative study of police: Criminal investigative
policies and practices in the US. and South Korea. Unpublished Doctoral
Dissertation, Michigan State University, East Lansing, MI.

Leighton, B. N. (1991). Visions of community policing: Rhetoric and reality in Canada.
Canadian Journal of Criminology, July-October, 485-522.

207

 

Lewis, D. A., & Maxﬁeld, M. G. (1980). Fear in the Neighborhoods: An Investigation of
the Impact of Crime. Journal of Research in Crime and Delinquency, 1 7(2), 160-
189.

Lewis, D. A., & Salem, G. (1986). Fear of crime: Incivility and the production of a social
problem. New Brunswick, NJ: Transaction Books.

Li, K. (2002). Capitalist development and economism in East Asia: The rise of Hong
Kong, Singapore, Taiwan, and South Korea. Asia-Paciﬁc Development Journal,
9(1), 109-120.

Li, W. (2003). Confucius on the couch: T ime-Limited Dynamic Psychotherapy ( T LDP)

and the Chinese American population. Unpublished Doctoral Dissertation, Wright
Institute Graduate School of Psychology.

Liska, A. E., Lawrence, J. J ., & Sanchirico, A. (1982). Fear of crime as a social fact.
Social Forces, 60(3), 760-770.

Liska, A. E., Sanchirico, A., & Reed, M. D. (1988). Fear of crime and constrained
behavior: Specifying and estimating a reciprocal effects model. Social Forces, 66,
827-837.

Lynes, D. A. (1996). Cultural diversity and social order: Rethinking the role of
community policing. Journal of Criminal Justice, 24(6), 491-502.

Manning, P. K. (1984). Community policing. American Journal of Police, 3(2), 205-227.

Markowitz, F. E., Bellair, P. E., Liska, A. E., & Liu, J. (2001). Extending social

disorganization theory: Modeling the relationships between cohesion, disorder,
and fear. Criminology, 39(2), 293-320.

Maxﬁeld, M. G. (1984). Fear of crime in England and Wales. London: Her Maj esty's
Stationery Ofﬁce.

McGarrell, E. F., Giacomazzi, A. L., & Thurman, Q. C. (1997). Neighborhood disorder,
integration, and the fear of crime. Justice Quarterly, 14(3), 479-500.

Mehren, E. (1997). Korean American enterpreneurship is an indication of economic

success. In D. L. Bender (Ed.), Asian Americans: Opposing vieWpoints (pp. 184-
191). San Diego, California: Greenhaven Press, Inc.

Merry, S. E. (1981). Urban danger: Life in a neighborhood of strangers. Philadelphia:
Temple University Press.

Miethe, T., & Lee, G. R. (1984). Fear of crime among older people: A reassessment of
the predictive power of crime related factors. Sociological Quarterly, 25, 397-415

208

Miyazawa, S. (1992). Policing in Japan: a study on making crime. Albany, NY: State
University of New York Press.

Mizumachi, O. (1982). Patrol police in Japan. International Criminal Police Review,
June-July(3 82), 150-162.

Moeller, G. L. (1989). Fear of criminal victimization: The effect of neighborhood racial
composition. Sociological Inquiry, 59(2), 97-109.

Moon, B. (2004). The inﬂuence of organizational socialization on police oﬂicers'
acceptance of community policing: The example of Korean police. Unpublished
Doctoral Dissertation, Michigan State University, East Lansing, MI.

Moore, M. H., & Trojanowicz, R. C. (1988). Policing and the fear of crime. Washington,
DC: National Institute of Justice, US. Department of Justice.

Moran, J. (1998). Corruption and NIC development: A case study of South Korea. Crime,
Law, and Social Change, 29, 161-177.

Murty, K. S., Roebuck, J. B., & Smith, J. D. (1990). The image of the police in Black
Atlanta communities. Journal of Police Science and Administration, 1 7, 250-257.

National Institute of Justice. (1978). Response time analysis: Kansas City, Missouri,
Police Department. Washington, DC: US. Department of Justice.

Nieuwbeerta, P. (2002). Crime victimization in comparative perspective: Results from the

international crime victims survey, 1989-2000. Den Haag, Netherlands: Boom
J uridische uitgevers.

Nunez, A.-M. (2004). Using segmented assimilation theory to enhance conceptualization
of college participation. InterActions: UCLA Journal of Education and
Inofrmation Studies, 1(1), Article 4.

Osgood, D.W., & Chambers, J .M. (2000). Social disorganization outside the metropolis:
An analysis of rural youth violence. Criminology, 38, 81-116.

Pain, R. H. (1997). Social geographies of women's fear of crime. Transactions of the
Institute of British Geographers, 22(2), 231-244.

Pantazis, C. (2000). Fear of crime: Vulnerability and poverty. British Journal of
Criminology, 40, 414-436.

Park, B. R. (1988). The history of Korean policing [in Korean]. Seoul, Korea: National
Police College.

Park, R. E., & Burgess, E. W. (1925). The city. Chicago, IL: University of Chicago Press.

209

Piquero, A., Green, J. R., Fyfe, J ., Kane, R. J., & Collins, P. (1998). Implementing
community policing in public housing developments in Philadelphia: Some early

results. In G. P. Alpert & A. Piquero (Eds.), Community policing: Contemporary
readings (pp. 95-122). Prospect Heights, H3: Waveland Press Inc.

Police Foundation. (1981). The Newark foot patrol experiment. Washington, DC: Police
Foundation.

Poole, E. D., & Pogrebin, M. R. (1990). Crime and law enforcement policy in the Korean
American community. Police Studies, 13(3), 57-66.

Portes, A., & Zhou, M. (1993). The new second generation: Segmented assimilation and
its variants. Annals, 530, 74-96.

Pyo, C. (2001). Policing: The past. Crime and Justice International, 24, 5-6.

Pyo, C. (2002). An empirical research on the public perception of the police and various
police reform initiatives in Korea: 1999-2001. Paper presented at the Asian
Association of Police Studies, Hong Kong.

Ragin, C. C. (1994). Using comparative methods to study diversity. In C. C. Ragin (Ed.),
Constructing social research: The unity and diversity of method. Thousand Oaks,
CA: Pine Forge Press.

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical Linear Models: Applications and
Data Analysis Methods (Vol. 2nd edition). Thousand Oaks: Sage Publications.

Raudenbush, S. W., Bryk, A. S., & Cheong, Y. F. (2000). HLM5: Hierarchical Linear
and Nonlinear Modeling. Chicago: Scientiﬁc Software International.

Reise, S. P., & Duan, N. (2003). Design Issues in Multilevel Studies. In S. P. Reise & N.
Duan (Eds.), Multilevel Modeling: Methodological Advances, Issues and
Applications. Hillsdale, NJ: Lawrence Erlbaum.

Reisi g, M. D., & Giacomazzi, A. L. (1998). Citizen perceptions of community policing:
Are attitudes toward police important? Policing: An International Journal of
Police Strategies and Management, 21, 547-561.

Riechers, L. M., & Roberg, R. R. (1990). Community policing: A critical review of
underlying assumptions. Journal of Police Science and Administration, 1 7(2),
105-114.

Rifai, M. Y. (1982). Methods of measuring the impact of criminal victimization through
victimization surveys. In H. J. Schneider (Ed.), The victim in international
perspective. Berlin and New York: de Gruyter.

210

Robert, S. A. (1998). Community-Level Socioeconomic Status Effects on Adult Health.
Journal of Health and Social Behavior, 39, 18-37.

Rohe, W. M., & Burby, R. J. (1988). Fear of crime in public housing. Environment and
Behavior, 20, 700-720.

Ross, C. E., & Jang, S. (2000). Neighborhood disorder, fear, and mistrust: The buffering
role of social ties with neighbors. American Journal of Community Psychology,
28(4), 401 -420.

Ross, C. E., Mirowsky, J., & Pribesh, S. (2001). Powerlessness and the ampliﬁcation of
threat: Neighborhood disadvantage, disorder, and mistrust. American Sociological
Review, 66(4), 568-591.

Rountree, P. W. (1998). A reexamination of the crime-fear linkage. Journal of Research
in Crime and Delinquency, 35(3), 341-372.

Rountree, P. W., & Land, K. C. (1996). Perceived risk versus fear of crime: Empirican
evidence of conceptually distinct reactions in survey data. Social Forces, 74(4),
1353-1376.

Rumbaut, R. G. (1997). Assimilation and its discontents: Between rhetoric and reality.
The International Migration Review, 31(4), 923-960.

Ryu, T. (1988). A history of Christ methodist church, 1903-1988 (Hawaii ui Hanin kwa
Kyohoe). Hawaii: Christ Methodist Church.

Sacco, V. F. (1982). The effects of mass media on perceptions of crime. Pacific
Sociological Review, 25, 475-493.

Sacco, V. F. (1990). Gender, fear, and victimization: A preliminary application of power-
control theory. Sociological Spectrum, 10, 485-506.

Sacco, V. F., & Nakhaie, M. R. (2001). Coping with crime: An examination of elderly
and nonelderly adaptations. International Journal of Law and Psychiatry, 24(2-3),
305-323.

Sadd, S., & Grinc, R. (1994). Innovative neighborhood oriented policing: An evaluation
of community policing programs in eight cities. In D. P. Rosenbaum (Ed.), The
challenges facing community policing: Testing the promises. Thousand Oaks, CA:
Sage.

Sampson, R. J ., & Groves, W. B. (1989). Community structure and crime: Testing social-
disorganization theory. The American Journal of Sociology, 94(4), 774-802.

211

Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent
crime: A multilevel study of collective efﬁcacy. Science, 277, 918-924.

Sampson, R. J., & Wilson, W. J. (1995). Toward a theory of race, crime, and urban
inequality. In J. Hagan & R. Peterson (Eds.), Crime and inequality. Stanford:
Stanford University Press.

Scheider, M., Rowell, T., & Bezdikian, V. (2003). The impact of citizen perceptions of
community policing on fear of crime: Findings from twelve cities. Police
Quarterly, 6(4), 363-386.

Schwartz, S. (1994). The fallacy of the ecological fallacy. American Journal of Public
Health, 84(5), 819-824.

Seoul City Government. (2005). About Seoul. Retrieved October 15, 2005, from
www.seoul.go.kr

Shaw, C., & McKay, H. (1942). Juvenile delinquency and urban areas. Chicago:
University of Chicago Press.

Shaw, C. R., & McKay, H. (1969). Juvenile delinquency and urban areas: A study of
rates of delinquency in relation to differential characteristics of local
communities in American cities. Chicago, H3: University of Chicago Press.

Sherman, L.W. (1993). Deﬁance, deterrence, and irrelevance: A theory of the criminal
sanction. Journal of Research in Crime and Delinquency, 30(4), 445-473.

Sherman, R. (1994). Crime's toll on the US: Fear, despair and guns. National Law
Journal, pp. AI, A19-A20.

Shin, D., Putkowski, C. P., & Park, C. (2003). The quality of life in Korea: Comparative
and dynamic perspectives. Social Indicators Research, 62(1), 3-16.

Sims, B., Hooper, M., & Peterson, S. A. (2002). Determinants of citizen's attitudes
toward police: Results of the Harrisburg citizen survey-1999. Policing: An
International Journal of Police Strategies & Management, 25(3), 457-471.

Skogan, W. G. (1978). Citizen satisfaction with police services: Individual and contextual
effects. Police Studies Journal(469-479).

Skogan, W. G. (1986). Fear of crime and neighborhood change. In A. J. Reiss & M.
Tonry (Eds.), Communities and crime (pp. 203-229). Chicago: University of
Chicago Press.

Skogan, W. G. (1987). The impact of victimization on fear. Crime and Delinquency,
33(1), 135-154.

212

 

Skogan, W. G. (1990). Disorder and decline: Crime and the spiral of decay in American
neighborhoods. Berkeley and Los Angeles: University of California Press.

Skogan, W. G. (1994). The impact of community policing on neighborhood residents: A
cross-site analysis. In D. P. Rosenbaum (Ed.), The challenge of community
policing: Testing the promises. Thousand Oaks, CA: Sage.

Skogan, W. G., & Hartnett, S. M. (1997). Community Policing, Chicago Style. New
York: Oxford University Press.

Skogan, W. G., & Maxﬁeld, M. G. (1981). Coping with crime: Individual and
neighborhood reactions. Beverly Hills: Sage.

Skogan, W. G., & Roth, J. A. (2004). Introduction. In W. G. Skogan (Ed.), Community
policing: Can it work? Belmont, CA: Wadsworth/Thomson Learning.

Skolnick, J. H., & Bayley, D. H. (1988). Theme and variation in community policing. In
M. Tonry & N. Morris (Eds.), Crime and justice (Vol. 10, pp. 1-38). Chicago, H3:
University of Chicago Press.

Smith, L., & Hill, G. D. (1991). Victimization and fear of crime. Criminal Justice and
Behavior, 18, 217-239.

Snijders, T. A. B., & Bosker, R. (1994). Modeled Variance in Two-level Models. Journal
of Educational Statistics, 18, 259-273.

St. John, C., & Heald-Moore, T. (1996). Racial prejudice and fear of criminal
victimization by strangers in public settings. Sociological Inquiry, 66, 267-284.

Stanko, E. (1995). Women, crime and fear. The Annals of the American Academy of
Politics and Social Science, 539, 46-59.

Suenari, M. (1998). Unilateral kindred: A reconsideration in East Asian societies. Osaka:
National Museum of Ethnology.

Takahashi, S., & Chang, D. H. (1985). Crime and delinquency prevention in Japan.
Comparative Criminal Policy, 221-234.

Taylor, R. B. (2001). Breaking away from broken windows: Baltimore neighborhoods
and the nationwide fight against crime, grime, fear, and decline. Boulder,

Colorado: Westview Press.

Taylor, R. B., & Hale, M. (1986). Testing alternative models of fear of crime.
Criminology, 77(1), 151-189.

213

Taylor, D. G., Taub, R. P., & Peterson, B. (1986). Crime, community organization, and
causes of neighborhood decline. In R. Figleo, S. Hakim, & G. Rengert,
Metropolitan crime patterns (pp. 161-77). New York: Willow Tree Press.

Thomas, C. W., & Hyman, J. M. (1977). Perceptions of crime, fear of victimization, and
public perceptions of police performance. Journal of Police Science and
Administration, 5, 305-317.

Tien, J. M., & Rich, T. F. (1994). The Hartford COMPASS program: Experiences with a
weed and seed-related program. In D. Rosenbaum (Ed.), The challenge of
community policing: Testing the promises (pp. 192-206). Thousand Oaks, CA:
Sage.

Trojanowicz, R., Kappeler, V. E., & Gaines, L. K. (1998). Community policing: a
contemporary perspective (2nd ed. ). Cincinnati, OH: Anderson.

Troj anowicz, R. C. (1985). An evaluation of the neighborhood foot patrol program in
Flint, Michigan. In. East Lansing, MI: Michigan State University.

Troj anowicz, R. C. (1986). Evaluating a neighborhood foot patrol program: The Flint,
Michigan, Project. In D. P. Rosenbaum (Ed.), Community crime prevention.
Beverly Hills, California: SAGE Publications, Inc.

Trojanowicz, R. C. (1994). The future of community policing. In D. P. Rosenbaum (Ed.),
The challenge of community policing: Testing the promises. Thousand Oaks, CA:
Sage.

Trojanowicz, R. C., Benson, B. L., & Trojanowicz, S. (1988). Community policing:
University input into campus police policy-making. Community Policing Series
14. National Neighborhood Foot Patrol Center, East Lansing: Michigan State
University.

Trojanowicz, R. C., & Carter, D. L. (1988). The philosophy and role of community
policing. East Lansing, MI: The National Neighborhood Foot Patrol Center,
School of Criminal Justice, Michigan State University.

Trojanowicz, R. C., Kappeler, V. E., & Gaines, L. K. (2002). Community policing: A
contemporary perspective (3rd ed.). Cincinnati, OH: Anderson Publishing Co.

Troyer, R.J., & Wright, RD. (1985). Community response to crime: two middle-class
anti-crime patrols. Journal of Criminal Justice, 13 (3), 227-241.

Tyler, TM. (2003). Procedural justice, legitimacy, and the effective rule of law. In M.

Tonry, Crime and justice: A Review of Research (pp. 431-505). Chicago:
University of Chicago Press.

214

 

Tyler, T. R. (1980). Impact of directly and indirectly experienced events: The origin of
crime-related judgements and behaviors. Journal of Personality and Social
Psychology, 39, 13-28.

Tyler, T. R. (1984). Assessing the risk of crime victimization: the integration of personal
victimization experience and socially transmitted information. Journal of Social

Issues, 40(1), 27-38.

Van Wilsem, J., De Graaf, N. D., & Wittebrood, K. (2000). Variations in cross-national
victimization: The impact of composition and context. In P. Nieuwbeerta (Ed.),
Crime victimization in comparative perspective: Results from the international
crime victims survey, 1989-2000. Den Haag, Netherlands: Boom J uridische
uitgevers.

Vander, V. T. M. (1998). Fear of victimization and the interactional construction of
harassment in a Latino neighborhood. Journal of Contemporary Ethnography,
27(3), 374-398.

Vanderveen, G. (2000). Proposing a novel approach to scout differences between men
and women. In P. Nieuwbeerta (Ed.), Crime victimization in comparative
perspective: Results from the international crime victims survey, 1989-2000. Den
Haag, Netherlands: Boom J uridische uitgevers.

Velez, M. B. (2001). The role of public social control in urban neighborhoods: A
multilevel analysis of victimization risk. Criminology, 39, 837-864.

Vold, G. 3., Bernard, T. J ., & Snipes, J. B. (1998). Theoretical criminology. New York:
Oxford University Press.

Walker, 8., & Katz, C. M. (2005). The police in America: An introduction (5th ed.). New
York, NY: McGraw-Hill.

Warr, M. (1984). Fear of victimization: Why are women and the elderly more aﬁ'aid?
Social Science Quarterly, 65, 681-702.

Warr, M., & Stafford, M. (1983). Fear of victimization: A look at the proximate causes.
Social Forces, 61 , 1033-1043.

Westermann, T. D., & Burfeind, J. W. (1991). Crime and justice in two societies: Japan
and the United States. Paciﬁc Grove, California: Brooks/Cole Publishing
Company.

Wildsmith, E. (2004). Race/ethnic differences in female headship: Exploring the
assumptions of assimilation theory. Social Science Quarterly, 85(1), 89-106.

215

 

Williams, H., & Pate, A. M. (1987). Returning to ﬁrst principles: Reducing the fear of
crime in Newark. Crime and Delinquency, 33, 53-70.

Willis, J. D. (1999). Basic Concepts in Hierarchical Linear Modeling with Applications
for Policy Analysis. In G. Cizek (Ed.), Handbook of Educational Policy. San
Diego, CA: Academic Press.

Wilson, J .Q., & Kelling, G. L. (1982). The police and neighborhood safety: Broken
Windows. The Atlantic Monthly, 249(March), 29-38.

Wilson, W. J. (1996). When work disappears: The world of the new urban poor. New
York: Knopf.

Wong, K. C. (2001). Community policing in China: Philosophy, law and practice.
International Journal of the Sociology of Law, 29, 127-147.

Wooldredge, J. (2002). Examining The (Ir)Relevance of Aggregation Bias for Multilevel
Studies of Neighborhoods and Crime with an Example Comparing Census Tracts
to Ofﬁcial Neighborhoods in Cincinnati. Criminology, 40(3), 681-709.

Wycoff, M. A. (1988). The beneﬁts of community policing: Evidence and conjecture. In
J. R. Green & S. D. Mastrofski (Eds.), Community policing: Rhetoric or reality
(pp. 103-120). Westport, CT: Praeger.

Yang, C. K. (1961). Religion in Chinese society. Berkeley: University of California Press.

Yang, O. (2003). Family structure and relations. Social Indicators Research, 62(1), 121-
148.

Yin, P. (1982). Fear of crime as a problem for the elderly. Social Problems, 30, 240-245.

Yonhap News. (2005). Police considers changes in plice Jigoodae system [In Korean].
JoongAng Daily (http://newsﬂrins. com/society) retrieved on August 3.

Yonhap News. (2005). Two women hit by a man for smoking in public [In Korean].
Yonhap News, p. A1 (http://www.vonhapnews.co.kr/) retrieved on February 7.

Yoo, M. (1998). Community and kinship in a Southern Korean village. In M. Shima & R.
L. J anelli (Eds.), The Anthropology of Korea: East Asian perspectives. Osaka:
National Museum of Ethnology.

Yoon, O. (1998). The economy, political repression and crime control: A time-series
analysis of Korean data, 1964-1993. Unpublished Dissertation, University of
Georgia, Athens, Georgia.

Young, K. (1942). Sociology. Cincinnati, OH: American Book.

216

Yusuf, S. (2003). Innovative East Asia: The future of growth. Washington, DC: World
Bank and Oxford University Press.

Zhao, J., Gibson, C., & Lovrich, N. (2002). Participation in community crime prevention:
Are volunteers more or less fearful of crime than other citizens? Journal of Crime
and Justice, 25(1), 41 -61.

Zhao, J., Scheider, M., & Thurman, Q. C. (2002). The effect of police presence on public
fear reduction and satisfaction: A review of the literature. The Justice
Professional, 15(3), 273-299.

Zhou, C., & Xia, Z. (1988). The psychological characteristics as well as the prevention
and cure for the formation of youngsters' unlawful and criminal cliques. Paper

presented at the International Symposium on Education of Juvenile Delinquents,
Shanghai, China.

217

   

IIIIIIIIIIIIIIIII-III