a randomized experimental study of sharing crime data with

A randomized experimental study of sharing crime data with

citizens: Do maps produce more fear?

ELIZABETH R. GROFF*, BROOK KEARLEY, HEATHER FOGG,

PENNY BEATTY and HEATHER COUTUREUniversity of Maryland, College Park, Maryland, USA

* corresponding author: E-mail: [email protected]

JULIE WARTELLSan Diego County District Attorney’s Office, San Diego, California, USA

Abstract. The larger roles of the community in crime prevention and improvements in technology

have increased policeYcitizen communication and the distribution of information from police

departments to private citizens. Combined, these changes have led to the current movement among

law enforcement agencies toward sharing both summary reports and maps of crime with community

groups. Although the dissemination of crime information is intended to benefit community members,

there is a lack of empirical evidence demonstrating the effects of crime mapping on citizen perceptions

and fear of crime. This experiment compared three formats for disseminating crime data; two popular

types of crime maps (i.e., graduated symbol and density) and the traditional tabular format of crime

statistics. A randomized experimental design was used to measure residents’ fear of crime and their

perception of the safety of different areas of Redlands, CA. Overall, residents who viewed either type

of map reported less fear than those who viewed tabular statistics. Respondents who viewed graduated

symbol maps consistently reported less fear than either density maps or tabular statistics. However,

there were differences depending on the type of map. While graduated symbol maps were associated

with the lowest levels of fear of robbery, theft or assault; density maps produced different reactions

depending on the area of the city. Finally, the maps did not stigmatize high crime areas of Redlands.

Where statistically significant differences existed, respondents who were given maps were more likely

to recommend someone move into an area than those who were given tabular statistics.

Key words: community policing, crime data, fear of crime, geographic information systems, GIS,

information sharing

Introduction

Recent shifts in police practice emphasizing the role of the community in crime

prevention and the widespread adoption of technology have been accompanied by

increased policeYcitizen communication (Sherman 1986), and more routine

distribution of crime pattern and trend information from police departments to

the public (Buslik and Maltz 1998; Maltz et al. 1989; McEwen and Taxman 1995;

Rich 2001; Weisburd and Lum 2001). These efforts have been supported by

Journal of Experimental Criminology (2005) 1: 87–115 # Springer 2005

millions of dollars in federal, state and local investments to promote a more

involved community and improved technology.1 In particular, crime mapping has

been gaining in popularity as a communication tool. However, debate continues

concerning the impact of releasing crime data to the public. Opponents argue that

neighborhoods with higher crime rates will be unnecessarily stigmatized and

redlined, while proponents assert that public dissemination of crime information

will empower community members to get involved in their communities (Harries

1999; Lavrakas et al. 1983; Pate et al. 1986a; Ratcliffe 2002; Wartell and McEwen

2000). Presently, there is little research guiding policing agencies on this important

issue. In fact, researchers have yet to test the effects of presenting crime

information via maps, as opposed to tabular data, on citizen fear of crime.

Despite the uncertainty surrounding the effects, many departments nationwide

have continued to make crime information more easily accessible to the public.

One example is the Redlands, CA Police Department, which has a long tradition

of publicizing crime data and recently began providing public access to crime

maps at the end of 2002 with the implementation of the Community Mapping for

Safety Strategies (COMPASS) program.2 The Redlands Police Department, in an

effort to better understand the consequences of making crime maps publicly

available, invited a team of researchers from the University of Maryland to test the

impact of maps versus tabular information.

This paper reports the results of a randomized experiment that tested (1)

whether the format (i.e., crime statistics3 vs. crime maps) for communicating

crime information to citizens differentially impacted their fear of crime and

moving recommendations, and (2) whether or not different types of crime maps

(i.e., graduated symbol maps vs. density maps) differentially impacted citizen fear

of crime and moving recommendations. Three different experimental conditions

corresponded to three different representations of crime information (1) a table of

crime statistics and an orientation map; (2) a graduated symbol map and (3) a

density map. Each of these treatment conditions depicted crime information by

identical police beat areas. The subjects were volunteer citizens of Redlands who

were from a variety of community settings. Participants were blocked on sex and

randomly assigned to receive one of the treatments. All of the participants

responded to a brief survey concerning their fear of crime in particular areas of

Redlands, their willingness to encourage acquaintances to move into particular

areas, and the clarity of the crime information presented to them.

The results of this experiment have implications for both policing agencies and

policy makers in deciding how best to communicate crime information to local

citizens without significantly increasing their fear of crime. The findings show that,

in general, residents who viewed either type of map reported less fear than those

who viewed tabular statistics. In fact, respondents who viewed graduated symbol

maps consistently reported less fear than either density maps or tabular statistics.

Finally, the maps did not appear to stigmatize high crime areas of Redlands. Where

statistically significant differences existed, respondents who were given graduated

symbol and density maps were more likely to recommend someone move into an

area than those who were given tabular statistics.

ELIZABETH R. GROFF ET AL.88

The next section of this paper provides background information on why the

format of crime data dissemination is an important question in the current policing

climate. It is followed by a discussion of the specific methodology and procedures

followed in conducting the experiment, and the experimental analyses and

findings. The final section of the paper discusses the implications of the study.

Roots of the question

PoliceYcitizen communication and crime mapping

Since the 1970s, the increased role of the community in crime prevention and

improvements in information technology have influenced policeYcitizen

communication. The McGruff BTake a Bite Out of Crime^ campaign is one

example of a nationwide program sponsored by the United States federal

government to increase citizen participation in crime prevention (Lavrakas et

al. 1983). The movement from professional policing to community policing

espoused a new and expanded role for the community as Bco-producers^ of public

safety (Goldstein 1987; Lavrakas et al. 1983; Maltz et al. 1989; Moore 1992) and

stimulated testing of innovative strategies for successfully mobilizing communities

(Sherman 1986). Improved policeYcitizen communication became an integral part

of these efforts (Rich 1999; Sherman 1986). One aspect of improved communi-

cation involved the release of crime data to the public. However, these efforts were

undertaken without a clear understanding of the potential effects on fear of crime

that would result from the release of crime data (Lavrakas et al. 1983).

As policing agencies have adopted new technology and begun to use it,

information has become far more accessible and the technical barriers to producing

and disseminating information about crime events have been significantly reduced.

The widespread adoption of computerized information systems, which began in the

1970s (Manning 1992), has enabled easy access to crime statistics and changed

both the scope and resolution of the crime information available to police and

citizens. Initial policeYcitizen information-sharing efforts consisted of summary

crime reports for neighborhoods or police beat areas, but as police information

systems have evolved, so have their capabilities to collect, analyze and disseminate

information. These changes have impacted both the level of aggregation (i.e., city,

police beat, address) and the type of product (i.e., report or list) that can be

routinely provided. By the late 1980s and 1990s, some departments had begun to

release crime data at the street block level.

During the same time period, advances in computer hardware and software

made geographic information systems (GIS) available to a wide variety of users in

many industries, including policing (Harries 1999; La Vigne and Groff 2001;

Weisburd and McEwen 1997). GIS, also commonly known as computer mapping,

allows existing crime data to be spatially analyzed and visually displayed and has

been rapidly adopted by policing agencies. A 1997 survey by the National Institute

of Justice indicated that approximately 13% of all law enforcement agencies were

SHARING CRIME DATA WITH CITIZENS: MAPS 89

using computer mapping (Mamalian et al. 1999). However, the adoption rate

varied by the size of the agency; among larger agencies with 100 or more officers,

approximately one third reported using mapping. Two years later, a 1999 survey by

the Police Foundation found a marked increase in adoption with over half of the

agencies surveyed using computer mapping for crime analysis (Weisburd et al.

2001a).4 This unusually swift implementation of crime mapping in law

enforcement agencies was examined by Weisburd and Lum (2001); they found

that the adoption of crime mapping diffused more rapidly as compared to other

policing innovations. The rapid adoption of computer mapping technology over the

last 10 years has facilitated data sharing between law enforcement and the

community (Greene 2000; La Vigne and Groff 2001; McEwen and Taxman 1995;

Rich 1995).

Sharing information with the public via maps increased quickly in the

early to mid-1990s (McEwen and Taxman 1995; Rich 1995). Since then, the

use of maps to communicate with the public has continued to grow (La Vigne and

Groff 2001). One reason for this trend is the power of maps as a communication

tool (Harries 1999; MacEachren 1994; Monmonier 1991; Tufte 1983). In his book

discussing the visual display of data, Tufte states, BNo other method for the display

of statistical information is more powerful^ (p. 26). This characteristic of maps

makes them a natural choice for disseminating crime data to the public.

As mentioned before, historically, different policing agencies have shared data

at various resolutions and in different formats. While some agencies released

monthly paper reports by police district, others provided community groups with

direct access to the mapping system and data so that individual crimes as well as

neighborhood-level aggregations could be displayed (Buslik and Maltz 1998; Rich

2001). Still others posted maps on Internet websites (Harries 1999; Wartell and

McEwen 2000). Regardless of the specific method, it was clear that citizens

were rapidly gaining access to crime data. According to the most recent Bureau

of Justice Statistics (BJS) survey, this trend is continuing. In 1999, BJS found

that 92% of citizens had Broutine access to crime statistics or crime maps,^ up

from 70% in 1997 (Hickman and Reaves 2001: 1). Today, many citizens have

access to crime data over the web 7 days a week, 24 hours a day. They can query

and produce maps and reports in their own homes. Law enforcement agencies with

computer mapping capabilities can choose among a myriad of products (e.g., maps,

summary reports, lists of events by address, etc.) and media (e.g., bulletins,

Internet, newspaper, etc.) to share information with the public (Harries 1999;

Wartell and McEwen 2000).

Data sharing concerns

Despite improvements in the technological ability to share crime data with the

public, core issues impacting data sharing remain. Although many policing

agencies are making crime information available to the public, there is significant

debate over the relative costs and benefits of sharing such information, and many


agencies have voiced strong concerns about sharing crime information with the

community (Harries 1999; Ratcliffe 2002; Wartell and McEwen 2000). McEwen

and Wartell (2000) and Harries (1999) offer good overviews of general issues

related to privacy and confidentiality.5 They identify three main issues that

characterize the debate over information sharing with the public, including: (1)

achieving a balance between the victim’s right to privacy and other citizens right to

know; (2) publicizing crime problems without stigmatizing neighborhoods; and (3)

ensuring only facts of record are made available, not intelligence reports based on

unverified information (Harries 1999; Wartell and McEwen 2000).

In addition to the issues identified in recent reviews, researchers in the 1980s

speculated about further reasons that police might be reticent to share crime

information (Lavrakas et al. 1983). They noted that many police and local officials

felt that releasing crime data would only increase fear of crime and that crime

prevention was best done by the professionals, without community involvement.

They also suggested that police may be acting in their own interest by limiting

information that could be used to evaluate their performance.

Most agencies are now willing to share particular forms of crime data with

the public, but there is still a continuum of comfort with the type and specificity

of information shared. For example, while almost all departments are com-

fortable with sharing aggregated crime rates (e.g., a map of crime shaded by

beat), a point map of those incidents would be unacceptable to many due to

concerns over violating a victim’s right to privacy. All agencies who share crime

data with the public face important concerns regarding privacy and confidenti-

ality regardless of the method and medium of communication (e.g., statistics,

charts, maps, hard copy, or electronic) (Harries 1999; Ratcliffe 2002; Wartell and

McEwen 2000). The distribution of crime data in the form of maps has proven to

be an especially controversial method of communicating crime data to the public.

These competing issues deserve careful consideration every time a decision is

made regarding the release of crime data. However, little guidance exists to

assist agencies with the practical issue of identifying which format, tabular or

mapped, best communicates this information while minimizing the impact on

fear of crime.

Empirical research on impact of sharing crime data

The impact of sharing crime data on residents’ perceptions of crime has received

little attention in the literature. The authors could find only four studies on the

subject, the first three of which were part of a group that looked at whether crime

statistics included in crime prevention newsletters increased fear of crime. Another

study investigated the change in the perception of neighborhood safety when

community groups were given their own mapping capability. The first three

studies were part of a set of research that included one quasi-experimental design

and two randomized experiments. These studies tested whether crime prevention


newsletters by themselves produced more fear than the same newsletter with crime

statistics included. The studies were conducted in Evanston, IL (Lavrakas and Herz

1982), Houston, TX (Lavrakas et al. 1983), and Newark, NJ (Pate et al. 1985).6

The results for the quasi-experimental study in Evanston indicated no difference

in fear levels between those who received crime statistics and those who did not.

The Houston and Newark studies were randomized experiments and used the same

methodology in each city. Specifically, two experimental conditions, newsletter and

newsletter with crime information, and a control group were used. The newsletter

with crime information showed the boundaries of the neighborhood and reported

detailed crime information for the previous month. They also used two study

designs, a panel design (with pre- and post-test) and post-test only. In both Houston

and Newark, there was no significant difference between groups with respect to fear

of personal victimization. However, among the panel design participants in

Houston, the pretest scores indicated they were slightly but significantly more

fearful of property crime victimization. Overall, the results of these three studies

provide preliminary evidence that sharing tabular crime data produces little or no

additional fear among residents. In addition, they found that neighborhood residents

reacted positively to receiving the detailed crime information. However, the authors

called for additional research on the relationship between sharing crime data and

fear levels (Pate et al. 1985).

While the first set of investigations tested the effect of sharing crime statistics on

fear of crime, the final study examined the consequences of giving residents their

own information system populated with crime data on perceptions of neighborhood

safety (Rich 2001). The Neighborhood Problem Solving (NPS) system was

developed as part of the Comprehensive Communities Program in Hartford, CT.7

The NPS system was installed in 14 different sites to be used by community groups

for analysis of calls for service, crimes and arrests. Training and on-going support

was provided by the researcher. At the end of the 10-month study period, over half

the respondents felt that working with crime data had not changed their perception

of neighborhood safety.

These findings highlight the limits of extant knowledge regarding the effects

of communicating crime data to the public in general. The studies using stronger,

quasi-experimental and experimental designs did not include maps, and the one

study that included maps (Rich 2001) did so in the context of an information

system rather than a product and had a less rigorous design. To date, the literature

provides no evidence regarding the impact of crime maps as compared to crime

statistics on the way information is perceived. Using a randomized experimental

design, this research examines the relative impact of maps versus statistics on both

fear of crime and housing recommendations.

Perception of mapped data

Recent research has explored the cognitive aspects of map use, looking to

individual differences associated with how various people perceive the same map.


Research findings demonstrate that these individual differences stem largely from

individual differences in Bdomain-specific prior knowledge^ (McGuinness 1994:

186). Thus, people can be divided into categories of experts or novices based on

their familiarity with the purpose of the map. Research into the interaction between

map and user found that experts tend to have better pattern recognition than do

novices (McGuinness 1994). Wood also found that the manner in which map users

interact with maps depends on a multitude of factors such that, BSign, symbol,

code, culture and myth mingle, separate and regroup according to each reader’s

peculiar structures of knowledge, education, experience, beliefs and motivation^(Wood 1994: 2). In other words, those who read the map interpret the information

on the map differently depending on their prior experience with the area and with

the type of data described.

Assuming that the general public lacks extensive experience with crime data, it

may be too much to expect that citizens use map products to conduct their own

crime trend analysis or develop crime prevention strategies without assistance from

the police. However, because the purpose of communicating crime statistics is to

inform the public about crime activity in their area, it is highly likely that citizens

will benefit from the ability to visualize the general pattern of crime. In sum, the

perception of mapped data is dependent on a variety of individual characteristics

that would be difficult to include in a nonrandomized design. Consequently, this

study uses a randomized experimental design because it allows for random

distribution of potentially confounding variables among the treatment groups,

ensuring that these confounds will not be systematically allocated into one group

over another (Boruch 1997; Weisburd et al. 2001b; Shadish et al. 2002).

Materials and methods

Participants

Participants in the study were 314 residents of Redlands, CA, at least 18 years of

age, recruited from various community activity venues.8 Participants were

recruited from BMarket Night,^ a weekly outdoor evening event held on a main

street in Redlands where local merchants sell their wares; the University of

Redlands campus; the Redlands community center; and local senior citizen

centers. Originally, the sole data collection site was going to be Market Night

because it attracted a cross-section of Redlands residents. However, after conducting

a pilot survey, we discovered there were not enough attendees to get the required

number of participants. After extensive conversations with Redlands Police

personnel, we identified the University of Redlands, two Senior Centers and the

Community Center as other venues that together would provide a cross-section of

Redlands residents. The use of these specific venues means that the results of the

study may be biased toward the subset of people who attend outdoor festivals and

Universities and/or use community and senior centers. In general, these people

may differ from a random sample of Redlands residents. They may also view


crime levels differently than Redlands residents who do not patronize these types

of venues.

Materials

The decision about the type of maps to produce for this study was made after

lengthy discussions with Redlands Police, COMPASS staff, and the COMPASS

research partner, as well as an extensive review of the existing cartography

literature. The data to be mapped included the Uniform Crime Report (UCR) Part

I crimes of robbery and aggravated assault over the 3-month period of July 1, 2002

to September 30, 2002. Three months of crime data were shown to residents

because the time period depicted enough data to reveal patterns but not so much

that it was difficult to distinguish the individual symbols on the graduated symbol

map. The specific time period was chosen because it was the most recent data

available at the time the experiment began and thus would provide timely crime

information to residents who participated in the study.

Three types of maps were used: (1) graduated symbol (2) density and (3)

orientation. However, the orientation map (Appendix A) did not depict any crime

data and functioned exclusively as a reference companion to the crime statistics

(see table in Appendix A). Its purpose was to enable respondents to see the

geographic extent of the areas for which statistics were provided (i.e., Area 1,

Area 2, etc.).

The graduated symbol map (Appendix B) was chosen for a number of reasons.

First, graduated symbol maps were regularly used by the Redlands Police

Department and were one of the more popular types of maps used by policing

agencies in general, thereby making the results of this study relevant and

informative to many police agencies. Second, the number of crimes over the 3-

month period used in the study was relatively small, so the graduated symbols were

easily distinguished from one another. Third, using a graduated symbol map gives

the map reader more information. Readers are able to identify which type of crime

occurred along a specific street, in this case aggravated assault or robbery. They

can also see the number of crimes at any location by noting the size of the symbol.

Finally geometric symbols were used to represent the two crime types, rather than

more realistic symbols, because they are easier to distinguish (e.g., a triangle

versus a handgun icon to represent gun-related crimes) (MacEachren 1994).

The second type of map chosen to represent crime data was the density map

(Appendix C), which has been gaining popularity as these types of maps have

become easier to create. In addition, the Redlands Police Department was

interested in testing these maps to see if they produced different fear levels than

the more traditionally used graduated symbol maps. Density maps provide an easy

to understand visualization of the concentration of crime. The shading of the map

indicates the amount of crime, with lighter colored areas depicting less crime than

darker ones.9 However, the crime typesVin this case, robbery and aggravated

assaultVare aggregated making it impossible to distinguish the type of crime being


shown on the map. Another unique feature of density maps is that they do not show

the locations of crimes but rather depict the average density of crime at any point

on the map. Thus, a reader cannot tell exactly how many crimes occurred at a

specific location.

The final condition used consisted of an orientation map and crime statistics for

the same time period (Appendix A). Redlands was divided into five areas, based on

prior police beats no longer in use.10 The summary table consisted of a list of

crimes by type and street block for each area (see table in Appendix A). The map

produced for this treatment condition simply showed the area boundaries and the

roads (see map in Appendix A). This map was not meant to convey crime

information on its own, but rather to provide a context for identifying the

boundaries of the areas (i.e., Area 1, Area 2, Area 3, etc.) described by the

accompanying table of crime statistics.

The maps were created using commonly accepted cartographic principles (Dent

1990; Harries 1999; MacEachren 1994). An identical base map was used for each

treatment so that they would be consistent except for the symbolization method

chosen. Blues were used on both crime maps because it is often associated by map

users with calm and security (Dent 1990: 387). A consistent color scheme was used

for both the graduated symbol and the density maps so that any variation in the

reactions of respondents would be due to the type of map and not the colors.11 The

symbol size for the graduated symbol map was the largest size that could be used

without significant overlap among symbols.

Experimental procedure and measures

The study used a statistical block on sex based on prior research findings that, on

average, women are more likely than men to report higher levels of fear of crime

(Clemente and Kleiman 1977). Therefore, the subjects were first blocked into two

groups (i.e., men and women). Subjects in each block were then randomly assigned

to the three treatment groups in order to rule out the potential confounding impact

of sex on fear of crime. The specific blocking procedure involved creating a table

of randomly generated conditions for each sex (male and female) within each

venue (Market Night, University of Redlands, community center and senior

center). In an effort to maintain balance (i.e., equal numbers) between the

treatment conditions, randomization tables were created for each sex recruited in

each venue by randomly ordering the conditions in a series of triplets to ensure that

one in every three respondents was allocated to each of the three conditions. By

using a block, the results indicate both significant differences between the

treatment groups, as well as significant differences between males and females.

In venues of open recruitment, the researchers made every effort to approach

individuals who appeared to be at least 18 years of age, who were not talking on

cell phones, and who were of the opposite sex from the last respondent recruited.

All potential participants approached were first asked whether or not they were

residents of Redlands. Those who responded Byes^ were then told that, BWe’re


conducting a study today in conjunction with the Redlands Police Department. The

study concerns citizens’ perception of crime in the community. It entails a three-

page survey that takes about 5Y10 minutes to fill out. At the end of the survey, we

invite you to enjoy free candy. Would you be willing to participate in the study?^For those individuals who refused to participate in the study, the researchers

collected general information including the approximate age category, sex and

reason given for not participating. Each Redlands resident, who agreed to

participate in the study, received a manila envelope containing an informed

consent form, the appropriate treatment materials (i.e., summary statistics table,

graduated symbol or density map) and the survey instrument.

The survey protocol employed in this study draws from existing surveys used in

prior evaluations of fear of crime, including the Police Foundation’s evaluation on

reducing fear of crime in Houston and Newark, and the Bureau of Justice

Statistics’ National Crime Victimization Survey (Bureau of Justice Statistics 2001;

Pate et al. 1986b). The survey instrument incorporates measures of fear of crime

and mobility as well as demographic characteristics and information regarding

crime data sources.12 A subset of the questions from the survey instrument and

their coding are listed in Appendix D.

Results

As previously mentioned, the specific objectives of this study were to determine:

(1) whether the method of communicating crime information (i.e., crime statistics

vs. crime maps) to citizens differentially impacts their fear of crime, and (2)

whether or not different types of maps (i.e., graduated symbol map vs. density

maps) differentially impact citizen fear of crime. The key research findings are

presented here in graphic and text form. Data are also presented on the

characteristics of the total sample, and clarity of the information presented.

Characteristics of the sample

The survey respondents were all adults (at least 18 years of age) with an average age

of 41. Fifty-seven percent (57%) were female, and sixty-five percent (65%) were

white. Forty-five percent (45%) of the sample had at least some college education,

and the most common response to yearly household income was B$20,000 to

$49,999.^ The average length of time that the respondents had lived in Redlands

was 14 years, and the majority of those surveyed were renting their homes (54.3%).

Most of the respondents lived with one or two adults (55%), and the majority

did not have children living in their home (60%).

Related to sources of official crime information and victimization, the following

picture emerged. Sixty-seven percent (67%) of respondents said that they had not

seen crime statistics or maps from the Redlands Police Department in the last year.


When asked about their primary source of crime information, fifty-eight percent of

the subjects (58%) reported that they read the newspaper. BWord of mouth^ was

the second most frequently reported source of crime information. With regard to

crime and victimization, nine percent (9%) of the sample said that they were the

victim of a violent crime during the past year and seventeen percent (17%) were

the victim of a property or Bother^ crime. These rates were mirrored when

respondents were asked about the victimization of a household member, with nine

percent (9%) of household members reported as the victim of a violent crime and

fifteen percent (15%) reported as the victim of a property or Bother^ crime.

When asked about their perception of crime in Redlands, the majority of res-

pondents (68%) felt safe in all areas of Redlands during the day, but at night, sixty-

four percent (64%) of the sample felt unsafe in some areas of the city. When asked

if they thought violent crime was a problem in Redlands, the majority of respondents

answered that it was Bsomewhat of a problem^ (59%), and fifty-eight percent (58%)

of the sample stated that the violent crime rate greatly affected their housing

decisions.

Data on non-respondents

During the recruitment period, the interviewers collected descriptive data on

those who declined to participate in the study. The data in this area are brief due

to the limited interaction period, but they provide some insight into who was

most likely to refuse participation and why. A little over half of the people

approached to participate in the study declined to do so (N = 335 for non-

respondents vs. N = 321 for respondents).13 Men and women were almost equally

likely to refuse participation (51% male vs. 49% female), and those between 18

and 30 years of age represented the largest age category of those who refused

(49.3%). The most common reason for refusal cited was that Bthey were in a

hurry^ (49.3%), with the second most common reason being that they were not a

Redlands resident (12.8%). Other, less common, reasons for non-participation

included lack of interest, a language barrier, and an inability to read the survey

instrument without corrective glasses.

Data analysis

Results presented in this report are based primarily on comparisons of means for

the three conditions (Table 1) and statistical tests indicating the probability of

obtaining a difference between the three groups as large as that observed if, in fact,

no true difference existed. Because there was a slight imbalance in case numbers

among some of the group categories, General Linear Model (GLM) analyses were

conducted. GLM represents a more conservative analytic approach than ANOVA

because it does not assume equal cell sizes among all groups (Littell et al. 2002).

When statistically significant differences among the three conditions were found,


post-hoc Tukey tests were conducted to determine exactly which of the treatment

conditions differed significantly from the others (Table 2). For example, examining

the question BFear someone will rob/steal from you in Area 1,^ the post-hoc Tukey

test shows that the significant difference between the three conditions is between

the graduated symbol map and the density map, with neither mapping condition

significantly differing from the statistics condition. A thorough examination of the

findings outlined in Tables 1 and 2 are reported in the sections below.

Interaction effects were also considered. An interaction effect occurs when the

effect of one independent variable varies across levels of another independent

variable. For this report, we looked to see whether the effects of the treatment group

categories on fear of crime differed depending on whether an individual was male or

female. We found no interaction effects for any of the variables of interest.

Fear of crime

Respondents were asked to rate how worried they would be that someone might try

to attack them or beat them up while in Areas 1 and 4 (Figures 1 and 2). Significant

differences among the groups were found in both areas. In Area 1, the graduated

symbol map group reported a significantly lower amount of worry than either the

density map or statistics group (P G 0.01). The density map group had the highest

amount of worry. This pattern remained in Area 4 with the graduated symbol map

Table 1. Fear of crime and residential recommendation outcomes by randomization condition.

Statistics

(N = 109)

Density map

(N = 108)

Graduated symbol

map (N = 97)

Feel safe alone in Area 1 2.844 2.611 2.711

Fear someone will rob/steal from you in Area 1* 1.789 1.851 1.628

Fear someone will attack you in Area 1** 1.814 1.813 1.567

Feel safe alone in Area 4* 2.584 2.849 2.906

Fear someone will rob/steal from you in Area 4* 1.813 1.710 1.568

Fear someone will attack you in Area 4** 1.850 1.648 1.525

How would crime information presented affect

your moving recommendations to Area 1 2.915 2.710 2.793






your moving recommendations to Area 4+ 2.932 3.173 3.197


your moving recommendations to Area 5** 3.238 3.990 3.789

+ Differences among one or more of the groups is significant at P G 0.10.

*Differences among one or more of the groups is significant at P G 0.05.

**Differences among one or more of the groups is significant at P G 0.01.


reporting the lowest amount of worry (P G 0.01). However, in Area 4 the density

map respondents reported fear levels comparable to that of the graduated symbol

map group, and both mapping conditions had significantly lower levels of worry

than the statistics group (see Table 2).

A similar set of questions were posed to respondents regarding how worried

they would be that someone might rob or steal something from them while in

Areas 1 and 4 (Figures 1 and 2). Again, significant differences among the groups

were found in both areas, and the patterns were fairly consistent with those for

assaults. In Area 1, the graduated symbol map group reported significantly lower

levels of fear than the density map group (P G 0.05) but the differences between

map groups and the statistics group were not significant (Table 2). In Area 4, the

graduated symbol map group reported significantly lower amounts of worry than

the statistics group (P G 0.05) (Table 2).

Table 2. Post-hoc tests of mean differences.

(I) Randomization

condition

(J) Randomization

condition

Mean difference

(IjJ)

Standard error Significance

Fear someone will attack you in Area 1

Statistics Density map 0.0017 0.096 1.00

Graduated symbola 0.2478 0.098 0.031

Density map Graduated symbola 0.2461 0.099 0.033

Fear someone will attack you in Area 4

Statistics Density mapa 0.2023 0.092 0.072


Density map Graduated symbol 0.1224 0.095 0.399

Fear someone will rob/steal from you in Area 1

Statistics

Density map j0.0628 0.091 0.768

Graduated symbol 0.1601 0.093 0.200

Density map Graduated symbola 0.2230 0.094 0.045

Fear someone will rob/steal from you in Area 4

Statistics Density map 0.1028 0.089 0.483


Density map Graduated symbols 0.1419 0.092 0.272

Feel safe alone in Area 4

Statistics Density mapaj0.2642 0.127 0.094

Graduated symbola j0.3213 0.130 0.036

Density map Graduated symbol j0.0571 0.130 0.899

Crime information and moving recommendations to Area 4

Statistics Density map j0.2410 0.125 0.133


Density map Statistics 0.2410 0.125 0.133

Graduated symbol j0.0248 0.128 0.979

Crime information and moving recommendations to Area 5

Statistics Density mapaj0.7523 0.141 0.000


Density map Graduated symbol 0.2009 0.144 0.345

aSignificant differences were found between randomization conditions I and J.


Figure 3 presents data on how safe the respondents in each group would feel being

alone in Areas 1 and 4. Within Area 1, no significant differences by randomization

condition were found. However, in Area 4, the graduated symbol map and density

map groups felt significantly safer than the statistics group (P G 0.05) (Table 2).

Figure 2. Impact of presented crime information on fear of violent crime in Area 4.

Figure 1. Impact of presented crime information on fear of violent crime in Area 1.


Overall, the graduated symbol map appeared to produce the lowest levels of

fear of crime among the respondents. The density map group and the statistics

group had fairly similar results, though the density map group had lower overall

levels of fear in Area 4 while statistics were lower in Area 1.

Moving recommendation

To address whether or not the method of communicating crime information

differentially impacts residents’ housing decisions, respondents were asked if the

crime information presented to them would affect their housing recommendations

to others (Figure 4). In Areas 1, 2 and 3, there were no significant differences in

housing recommendations by randomization condition. In Area 4, however,

significant differences were found among the treatment conditions (P G 0.05),

with the density map and graduated symbol map groups having higher levels of

encouragement to move into the area than the statistics group. Between the groups,

graduated symbol maps respondents were significantly more likely to recommend a

move to Area 4 than statistics respondents (P G 0.10). Similarly, within Area 5, the

density map and graduated symbol map groups had significantly higher levels of

encouragement than the statistics group (P G 0.01).

Clarity

Although not the primary focus of the current research, one survey question

addressed whether or not the three conditionsVstatistics, graduated symbol

Figure 3. Feel safe being alone in the area.


maps, and density mapsVdiffered in terms of clarity. Respondents from each of

the randomization conditions were asked to rank how clearly they felt the

information on assaults and robberies was presented. Figure 5 presents the results

of that analysis. No significant differences were found between the randomization

Figure 4. Impact of presented crime information on moving recommendation.

Figure 5. Clarity of crime information.


conditions with regard to the clarity of the information.14 This finding suggests that

significant differences among the three conditions cannot be attributed to disparate

clarity levels. In addition, this finding is opposite the commonly held assumption that

maps communicate information more clearly than statistics.

While the results are somewhat mixed, a number of interesting patterns emerge.

With regard to fear of crime, when significant differences among the groups were

found, the group viewing crime information via graduated symbol maps evidenced

the lowest fear of crime related to those areas. Further, with the exception of Area 1,

the group viewing tabular crime data had higher levels of fear than the ones viewing

one or both of the maps, regardless of the question. Area 1 was the only area in which

the group viewing density maps evidenced the highest fear levels with regard to

violent crime. Examining the treatment condition’s impact on housing recommendations,

the findings again favored the mapping conditions, with the density map and graduated

symbol map groups producing the highest levels of comfort with recommending a move

to Areas 4 and 5.

Discussion

Advances in technology and changes in policing have led to an increased emphasis

on sharing crime information with the public (Greene 2000; McEwen and Taxman

1995; Rich 1999; Sherman 1986; Weisburd and Lum 2001), and police depart-

ments have responded by purchasing new technology, including geographic

information systems (GIS). Recent Federal government programs, such as the

Strategic Approaches to Community Safety Initiative (SACSI) and Community

Mapping and Analysis for Safety Strategies (COMPASS), have invested heavily in

information technology and GIS to facilitate data sharing. All of these changes

have been made despite a lack of empirical evidence regarding the impact of the

mode of communicating crime information on citizen fear of crime.

In this experiment, we tested the effect of three different formats for com-

municating crime information to the public: traditional tabular crime statistics,

graduated symbol maps, and density maps. Two major findings emerged. First,

using maps to report crime data did not consistently cause Redlands residents to be

more fearful than they would be viewing the same information reported in the form

of statistics. In fact, graduated symbol maps were associated with the lowest levels

of fear of robbery or assault. Second, the maps did not stigmatize high crime areas

of Redlands. On the contrary, for the areas in which the differences between the

treatments were significant, citizens who viewed maps tended to be more likely to

encourage someone they knew to move into specific areas than those who were

given a table of statistics.

Given the initial evidence provided here that maps, by themselves, do not increase

fear of crime in comparison to tables of crime statistics, the next logical question

concerns whether the type of map matters. In other words, do graduated symbol

maps produce less fear than density maps? In this experiment, respondents who were


given graduated symbol maps consistently reported the lowest fear ratings. This

finding held for both fear of assault as well as fear of robbery. However, the findings

for density maps were less clear and seemed to be dependent upon the specific area

of Redlands. Density map viewers felt more worried about crime in Area 1 (the

Central Business District) than density map viewers in Area 4 (a more residential

area). One explanation for this effect lies in the relative sizes of Area 1 and Area 4,

which affects the density of crime depicted. Even though Area 1 had two fewer

crimes than Area 4, it had a higher density of crime. Consequently, the whole of

Area 1 is shaded in a darker color suggesting high crime (in this case dark blue,

see Appendix C). In contrast, the overall pattern in Area 4 is one of a few dark

blue areas with large sections of light blue in between. Thus density maps seem

to produce more fear in areas where crime density is uniform and high and less

fear in areas where there are localized hot spots. This initial evidence points

toward the use of graduated symbol maps as the overall preferred method of crime

information transmission to citizens without significantly increasing fear of crime.

The experiment also examined another element of fear of crime that is

frequently raised about providing mapped data; specifically, that crime mapping

will encourage stigmatization of areas through the visualization of crime patterns.

To address this issue, the survey asked respondents about the impact of crime

information on their recommendation regarding moving to each of the five areas in

Redlands. The differences among the three formats of crime information

communication were significant in only two of the five areas, Area 4 and Area

5. For those two areas, groups that received maps were more likely to encourage a

move than those respondents who received crime statistics. Regarding Areas 1, 2

and 3, the mode of crime information communication did not have a significant

effect on moving recommendations. Interestingly, Area 1 and Area 3 had the

highest crime rates over the time period depicted in the crime information. Thus,

communication of crime information via maps has either an equal or a

significant more positive effect on moving recommendations when compared with

statistics.

Another commonly raised issue concerns the complexity of crime maps is

that they are too complicated for citizens to correctly interpret. Although this study

did not directly address the question of whether the respondents correctly

interpreted the information presented, one survey question asked about the clarity

of the information presented. No significant differences in the degree of clarity were

found between statistics, density maps and graduated symbol maps. This result was

somewhat unexpected give the common perception that maps communicate

information better than tables (Harries 1999; MacEachren 1994; Monmonier 1991;

Tufte 1983). However, the finding does suggest that the differences in perceived

crime levels were due to the treatment condition and not to the clarity of the

information reviewed. If the respondents were confused by the crime data provided,

it extended to all three modes of communication.

Within this study, there are several limitations that must be addressed. The

choice of a 3-month period in a specific year, rather than a longer period of time,

means the crime data are subject to seasonal variations. It is possible that other


crime patterns may have emerged during a different time period than the period

chosen. Taking this further, the reactions to the crime data may have been different if

different levels and patterns of crime were reported. Another possible methodological

issue is the use of areas to divide Redlands in order to ask questions about specific

parts of the city. The specific boundaries for the areas may have affected the

perceptions of the residents by manipulating the number of incidents in each area. In

order to minimize the potential bias, Redlands was divided into five large areas, a

downtown and then four areas that corresponded to north side, south side etc.

However, the findings may have changed if alternate areas were defined. Another

issue relates to the use of residents rather than both residents and nonresidents. The

perceptions of residents may have been biased by personal knowledge of areas rather

than simply the format of crime information. However, the varying extents of personal

knowledge for each participant should have been randomly distributed across the

groups. In addition, a main purpose of the study is to provide guidance to police on

how to provide crime statistics to residents in the most productive manner and not

to test the perception of mapped information in general. Thus the use of residents met

the core objective of the study.

Two final limitations of the study deserve discussion: (1) the generalizability

of the results and (2) empirical questions about the perception of crime data.

First, the generalizability of the current study to other populations is limited.

Although the participants in the study were randomly allocated to one of the three

treatment conditions, the participants themselves were not randomly selected.

Thus, caution must be exercised when applying the findings to different populations

both inside and outside of Redlands. Replication is required to strengthen the

validity of the findings in different cities using dissimilar populations. Second,

there are empirical questions left unanswered by this initial study of the effect

of visualization on the perception of crime. As explained by Harries (1999), an

individual’s perceptions are affected by both the design of the map and the

characteristics of the map user. Therefore, the conclusions drawn from this

experiment are only valid for maps that are designed following commonly

accepted cartographic guidelines as done here (Dent 1990; MacEachren 1994). It

is expected that other maps using those guidelines will have a good probability of

producing similar results. However, different map designs that do not follow those

guidelines may produce conflicting results. It is important to stress that maps can

easily be manipulated to convey specific messages by varying symbol size, scale

and other design elements (MacEachren 1994).15 For example, larger symbols on a

citywide map convey a general crime wave, while smaller symbols allow the

differential patterns of crime across the city to be perceived. Our use of a blue color

scheme, rather than a red one, to indicate concentrations of crime may have affected

the perceptions of residents. The effect of color on the perception of crime data de-

serves additional study.

While a randomized experiment provides a good design to examine the specific

question of whether crime maps produce more fear than crime statistics, the design

is unable to assist with more nuanced questions. As just mentioned, this study

leaves questions about how the individual characteristics of the map user affect the


perception of mapped information (Harries 1999). There are race, social class, and

gender differences in mobility that affect an individual’s geographic knowledge

and consequently his or her ability to comprehend maps. An individual with greater

mobility has more knowledge of his or her geographic area and consequently a better

ability to interpret maps. In addition to demographic characteristics and mobility,

there are also psychological factors influencing one’s interpretation of maps. For

instance, the concept of Bselective perception^ states that people Bmay fail to

perceive content that does not accord to his [or her] own spatial knowledge^ (Gold

1980: 59), while simultaneously interpreting information in a way that reinforces

their personal knowledge.

In sum, the findings of this experiment provide initial support for the use of

maps, especially graduated symbol maps, to communicate crime information.

The results suggest that both density and graduated symbol maps often

communicate crime information without significantly increasing fear of crime,

although the three treatments sometimes showed no significant differences. Of

the three types of media tested, graduated symbol maps most often produced less

fear than either density maps or statistics tables. In addition, graduated symbol

maps were found to produce consistent perceptions of crime regardless of the

crime pattern in a particular area. Density maps produced more fear related to

areas of uniformly high crime density and should be avoided when communi-

cating crime data in high crime areas. If stigmatization of particular areas is a

concern, maps should be used, as map users reported more positive moving

recommendations than tabular statistics. Finally, the experiment’s findings

provide the first scientific support of law enforcement’s investment in crime

mapping as a useful communication tool for sharing crime information with the

public.

Acknowledgements

The authors gratefully acknowledge funding for this project from the City of

Redlands, through the City of Redlands East Valley COMPASS Initiative

(2001-MU-MU-K012), and the Police Foundation under a grant from the

National Institute of Justice entitled East Valley COMPASS: Research Partner

(2002-MU-CX-K013). In addition, we are indebted to David Weisburd for his

guidance throughout the project and Chief Jim Bueermann for allowing us to

work with his department. Rachel Boba provided thought-provoking comments

during the design phase of the research. Our anonymous reviewers supplied

insightful comments. Finally, this project could not have been completed

without the extraordinary efforts of Raquel Perez, Vanessa Ruvalcaba and

Philip Mielke of the Redlands Police Department. Points of view expressed in

this article are those of the authors and do not necessarily reflect the views of

the Redlands Police Department, the Police Foundation or the Department of

Justice.


Appendix A: Treatment one: Statistics and orientation map

REDLANDS POLICE DEPARTMENT

REPORT OF INCIDENTS: ASSAULTS AND ROBBERIES

JULY 1YSEPTEMBER 30, 2002

*Note: A map identifying the locations of the areas is attached for your reference

Area 1

Hundred block/intersection/area Street name Number and type(s) of crime

1,400 ARLENE ST 1 ASSAULT

600 CHURCH ST 1 ROBBERY

600 E REDLANDS BL 1 ROBBERY

10 E STATE ST 1 ASSAULT

300 E STATE ST 3 ROBBERIES

200 ELEVENTH ST 1 ASSAULT

300 N FIFTH ST 2 ASSAULTS

300 N SIXTH ST 1 ROBBERY

300 ORANGE ST 1 ASSAULT, 1 ROBBERY

Intersection ORANGE ST/STATE ST 1 ASSAULT

100 REDLANDS MALL 1 ASSAULT

200 REDLANDS MALL 1 ASSAULT

70 SAN GORGONIO DR 2 ASSAULTS

500 THE TERRACE 1 ASSAULT

100 W VINE ST 1 ASSAULT


Area 2


Intersection COLTON AV/REDLANDS BL 1 ASSAULT

1,600 INDUSTRIAL PARK AV 1 ROBBERY

1,700 PLUM LN 1 ASSAULT

500 TEXAS ST 1 ASSAULT

2,000 W REDLANDS BL 3 ROBBERIES

Area 3


Intersection ALTA ST/SAN 1 ASSAULT

BERNARDINO AV

700 BALDWIN AV 2 ASSAULTS

600 CHURCH ST 1 ASSAULT

900 COLUMBIA ST 1 ASSAULT

1,000 COLUMBIA ST 1 ASSAULT

Park COMMUNITY PARK 1 ASSAULT

700 E BROCKTON AV 1 ASSAULT

Intersection GLOVER ST/SAN 1 ASSAULT

BERNARDINO AV

100 MULVIHILL AV 1 ASSAULT

1,200 N GROVE ST 1 ASSAULT

1,300 ORANGE ST 1 ASSAULT

Intersection ORANGE ST/PIONEER AV 1 ROBBERY

Intersection ORANGE ST/SAN 1 ASSAULT

BERNARDINO AV

900 OXFORD DR 1 ASSAULT

1,200 POST ST 1 ROBBERY

800 SINCLAIR CT 1 ASSAULT

30 SUN AV 1 ASSAULT

700 W COLTON AV 1 ASSAULT

100 W PENNSYLVANIA AV 1 ASSAULT

100 W PIONEER AV 3 ASSAULTS

1,000 WEBSTER ST 4 ASSAULTS

1,500 WEBSTER ST 1 ASSAULT

Area 4


1,100 CENTRAL AV 1 ASSAULT

1,200 CENTRAL AV 1 ASSAULT

1,100 CERO CT 1 ASSAULT

Park COMMUNITY PARK 1 ASSAULT

1,100 CORNELL AV 1 ASSAULT

Intersection CORNELL AV/EDWARDS ST 1 ASSAULT

1,100 E CITRUS AV 1 ASSAULT

1,200 E CITRUS AV 1 ASSAULT

1,300 E CITRUS AV 1 ROBBERY

1,200 E COLTON AV 1 ASSAULT

1,100 E LUGONIA AV 3 ASSAULTS

1,200 E LUGONIA AV 1 ASSAULT

1,300 E PENNSYLVANIA AV 1 ASSAULT


Appendix B: Treatment two: Graduated symbol map

1,200 EDWARDS ST 1 ASSAULT

800 JONI LN 1 ASSAULT

400 JUDSON ST 1 ASSAULT

1,000 PARKFORD DR 1 ROBBERY

1,500 POWELL LN 3 ASSAULTS

Area 5


1,600 CALVARY CL 1 ASSAULT

500 CITRUS AV 1 ASSAULT

1,600 COUNTRY CLUB DR 2 ASSAULTS

10 E CLARK ST 1 ASSAULT

31,000 FLORIDA ST 3 ASSAULTS

600 HIBISCUS DR 1 ASSAULT

26,000 KEISSEL RD 2 ASSAULTS

10 N SAN MATEO ST 1 ASSAULT

900 PINE AV 2 ASSAULTS

800 PINE AV, #D 1 ROBBERY

40 PRICE ST 1 ASSAULT

300 SONORA ST 1 ASSAULT

200 TERRACINA BL 2 ASSAULTS

100 W FERN AV 1 ASSAULT

1,000 W STATE ST 1 ASSAULT


Appendix C: Treatment three: Density map

Appendix D: City of Redlands Survey Questions

1. How safe would you feel being alone in Area 1?1

2. How worried would you be that someone would try to rob you or steal something from

you while in Area 1?2

3. How worried would you be that someone would try to attack you or beat you up while

in Area 1?2

4. How safe would you feel being alone in Area 4?1

5. How worried would you be that someone would try to rob you or steal something from

you while in Area 4?2

6. How worried would you be that someone would try to attack you or beat you up while

in Area 4?2

7. If someone you knew was considering moving to Area 1, would the crime information

presented here lead you to:3


presented here lead you to: 3








12. Is the crime information provided here on assaults and robberies in Redlands presented

clearly?4

Answer choices and coding:

1 1YVery unsafe, 2YSomewhat unsafe, 3YSomewhat safe, 4YVery safe2 3YVery worried, 2YSomewhat worried, 1YNot worried at all3 1YStrongly discourage them, 2YDiscourage them, 3YHave no effect, 4YEncourage them,

5YStrongly encourage them4 1YVery unclearly, 2YSomewhat unclearly, 3YSomewhat clearly, 4YVery clearly

Notes

1 For specific information on amounts, see the website of the Office of Justice Programs

at www.ojp.usdoj.gov and the Community Oriented Policing Services Office at

www.cops.usdoj.gov.2 The COMPASS program, created and funded by the National Institute of Justice,

involved multiple city agencies, social science researchers and community members in a

collaborative effort to develop new crime prevention strategies, create an interagency

database for the distribution of community information via the Internet and evaluate the

impact of recent public safety initiatives. The Redlands Police Department serves as lead

agency for the East Valley COMPASS initiative. The Police Foundation is the research

partner.3 For the purpose of this paper, the term crime statistics is used to refer to the

dissemination of crime data in the form of a list of crime incidents by hundred blocks

and sorted by district.4 The primary purpose of the Police Foundation’s survey was to uncover how widespread

the adoption of Compstat had become in policing. However, it included a question about

whether agencies had mapping software available for crime analysis. They surveyed all

agencies with 100 or more sworn officers and a sample of smaller agencies. They found

that over half of the non-Compstat agencies (52.9%) and 85.1% of the Compstat

agencies were using computer mapping technology.5 The following sources provide in-depth analysis of the issues related to data sharing,

privacy and confidentiality and sources of error in official law enforcement data. For a

good review of potential sources of error specific to crime mapping, see Ratcliffe (2002).

For a comprehensive review of data sharing and confidentiality issues please see Harries

(1999) and Wartell and McEwen (2000). Gove et al. (1985) provides an excellent

examination of the validity of official crime statistics.6 These studies focused solely on crime statistics rather than maps because they were

completed prior to mapping software’s general use in law enforcement.7 Hartford was one of 12 sites participating in the Comprehensive Communities Program

(CCP), established by the Bureau of Justice Assistance in 1994. The program

emphasized the active role that citizens need to play in identifying and solving

community problems.8 A power analysis indicated that a total of 105 participants per group were needed to

detect a moderate effect size in the current experiment and this minimum threshold was

achieved. The power analysis was done using the following parameters and the tables


available in Cohen (1977: 28Y39): (1) a two-tailed direction; (2) an effect size of 0.50;

(3) an alpha of 0.05, and (4) a power level of 0.95. A two-tailed test was used to enable

detection of an effect in either direction. The alpha of 0.05 is customary and the beta of

0.05 is stringent. This test is able to detect a moderate effect size.9 The density map was created using a kernel density with a cell size of 144 feet and a

bandwidth of approximately a quarter mile. The map was symbolized using natural

breaks (Jenks) in the data and with five categories. Natural breaks classification

emphasizes the differences between classes and minimizes differences within classes.10 Initially, the authors intended to use neighborhood boundaries or other naturally bounded

areas with which citizens could identify. However, upon review it was determined that

these conditions did not exist for Redlands, thus, old police beats were used. These had

the advantage of clearly delineating the central business district and dividing the

remainder of the city into four areas roughly corresponding to South, North Central,

Northeast and Northwest. Any type of area delineation exposes the study to the

modifiable area unit problem in which the aggregate totals for subareas are a product of

where the boundaries are drawn (Openshaw, 1983). However, this potential hazard was

outweighed by the necessity of having a way of talking about different sections of

Redlands.11 Psychological research indicates people react differently to different colors (Dent 1990).

Blues tend to connote calmness and security and were chosen to minimize reaction to the

map simply due to a more incendiary color such as the more commonly used reds. In this

way any increased fear reaction among respondents viewing maps versus tabular

information was more likely to be from the content and its presentation (graduated

symbol vs. density map) and not simply the colors used in the map.12 In order to minimize the survey administration time, only Areas 1 and 4Vchosen to

represent Blow^ and Bhigh^ crime regions within RedlandsVwere selected for questions

related to fear of crime. Please see the survey questions in Appendix D for the full scale.13 Seven surveys had to be discarded because of incomplete information.14 A score of 2 corresponds to the response FSomewhat unclearly_ and a score of 3

represents FSomewhat clearly_.15 Recall the earlier example where two maps with the same number of incidents can

portray either a large-scale crime problem using larger symbols or localized hot spots of

crime in particular areas with smaller symbols.

References

Boruch, R. F. (1997). Randomized experiments for planning and evaluation. Thousand

Oaks, CA: Sage Publications.

Bureau of Justice Statistics. (2001). National crime victimization survey. Washington, DC:

Office of Justice Programs, Bureau of Justice Statistics.

Buslik, M. & Maltz, M. D. (1998). Power to the people: Mapping and information sharing in

the Chicago Police Department. In D. L. Weisburd & T. McEwen (Eds.), Crime mapping

and crime prevention (pp. 113Y130). Monsey, NY: Criminal Justice Press.

Clemente, F. & Kleiman, M. (1977). Fear of crime in the United States: A multivariate

analysis. Social Forces 56, 519Y531.

Dent, B. D. (1990). Cartography: Thematic map design. Dubuque, IA: William C. Brown

Publishers.


Gold, J. R. (1980). An introduction to behavioural geography. New York: Oxford University

Press.

Goldstein, H. (1987). Toward community-oriented policing: Potential, basic requirements,

and threshold questions. Crime and Deliquency 33(1), 6Y30.

Greene, R. W. (2000). GIS in public policy: Using geographic information for more effective

government. Redlands, CA: ESRI.

Harries, K. D. (1999). Mapping crime: Principle and practice. Washington, DC: U.S.

National Institute of Justice.

Hickman, M. J. & Reaves, B. A. (2001). Community policing in local police departments,

1997 and 1999. Washington DC: Bureau of Justice Statistics, US Department of Justice.

La Vigne, N. G. & Groff, E. R. (2001). The evolution of crime mapping in the United States:

From the descriptive to the analytic. In A. Hirschfield & K. Bowers (Eds.), Mapping and

analysing crime data: Lessons from research and practice (pp. 203Y222). London: Taylor

& Francis.

Lavrakas, P. J. & Herz, E. J. (1982). Citizen participation in neighborhood crime prevention.

Criminology 20, 479Y498.

Lavrakas, P. J., Rosenbaum, D. P. & Kaminski, F. (1983). Transmitting information about

crime and crime prevention to citizens: The Evanston newsletter quasi-experiment.

Journal Police Science and Administration 11, 463Y473.

Littell, R. C. Stroup, W. W. & Freund R. J. (2002). SAS for Linear Models (4th Edition):

SAS Publishing/Wiley-Interscience.

MacEachren, A. M. (1994). Some truth with maps: A primer on symbolization and design.

Washington, DC: Association of American Geographers.

Maltz, M. D., Gordon, A. C. & Friedman, W. (1989). Mapping crime in its community

setting: Event geography analysis . New York: Springer-Verlag.

Mamalian, C. A. & LaVigne, N. G. Crime Mapping Research Center Staff. (1999). The use

of computerized crime mapping by law enforcement: Survey results. Washington, DC:

National Institute of Justice.

Manning, P. K. (1992). Information technologies and the police. In M. Tonry & N. Morris

(Eds.), Modern policing (Vol. 15, pp. 349Y398). Chicago: University of Chicago.McEwen, T. & Taxman, F. S. (1995). Applications of computer mapping to police

operations. In J. E. Eck & D. L. Weisburd (Eds.), Crime and place. Monsey, NY: Willow

Tree Press.

McGuinness, C. (1994). Expert/novice use of visualization tools. In A. M. MacEachren, D.

R. Taylor & Fraser (Eds.), Visualization in modern cartography. Tarrytown, NY: Elsevier

Science.

Monmonier, M. (1991). How to lie with maps. Chicago, IL: University of Chicago.

Moore, M. H. (1992). Problem-solving and community policing. In M. Tonry & N. Morris

(Eds.), Modern policing (pp. 99Y158). Chicago: University of Chicago.

Pate, A. M., Lavrakas, P. J., Skogan, W. G., Wycoff, M. A. & Sherman, L. W. (1985).

Neighborhood police newsletters: Experiments in Newark and Houston (Executive

Summary). Washington, DC: Police Foundation.

Pate, A. M., Skogan, W. G., Wycoff, M. A. & Sherman, L. M. (1986). Reducing fear of

crime in Houston and Newark: A summary report. Washington, DC: Police Foundation.

Pate, A. M., Lavrakas, P. J., Skogan, W. G., Wycoff, M. A. & Sherman, L. M. (1986).

Reducing fear of crime in Houston and Newark: A summary report. Washington, DC:

Police Foundation.

Ratcliffe, J. H. (2002). Damned if you don’t, damned if you do: Crime mapping and its

implications in the real world. Policing and Society 12(3), 211Y225.


Rich, T. (1995). The use of computerized mapping in crime control and prevention

programs. Washington, DC: National Institute of Justice, US Department of Justice.

Rich, T. (1999). Mapping the path to problem solving. NIJ Journal 241, 2Y9.

Rich, T. F. (2001). Crime mapping and analysis by community organizations in Hartford,

Connecticut. Washington, DC: National Institute of Justice, US Department of Justice.

Shadish, W. R., Cook, T.D. & Campbell, D.T. (2002). Experimental and quasi-experimental

designs for generalized causal inference. Boston: Houghton-Mifflin Company.

Sherman, L. W. (1986). Policing communities: What works? In A. J. Reiss, Jr. & M. Tonry

(Eds.), Communities and Crime (pp. 343Y386). Chicago: University of Chicago Press.

Tufte, E. R. (1983). The Visual Display of Quantitative Information. Cheshire, CT: Graphics

Press.

Wartell, J. & McEwen, T. (2000). Privacy in the Information Age: A Guide for Sharing

Crime Maps and Spatial Data. Washington, DC: U.S. Department of Justice, National

Institute of Justice.

Weisburd, D. L., & Lum, C. (2001). Translating Research Into Practice: Reflections on the

Diffusion of Crime Mapping Innovation. Dallas, TX: Fifth Annual Crime Mapping

Research Conference.

Weisburd, D. L. & McEwen, T. (1997). Introduction: Crime mapping and crime prevention.

In L. Weisburd David & T. McEwen (Eds.), Crime Mapping and Crime Prevention:

Crime Prevention Studies (Vol. 8 ). Monsey, NY: Criminal Justice Press.Weisburd, D., Mastrofski, S. D., McNally, A. M. & Greenspan, R. (2001). Compstat and

organizational change: Findings from a national survey. Washington, DC: National

Institute of Justice.

Weisburd, D., Lum, C. & Petrosino, A. (2001). Does research design affect study outcomes?

Annals of the American Academy of Political and Social Science 578, 50Y70.

Wood, M. (1994). Visualization in Historical Context. In A. M. MacEachren, D. R. Taylor

& Fraser (Eds.), Visualization in Modern Cartography (pp. 13Y26). Tarrytown, NY:

Elsevier Science.

About the authors

Elizabeth Groff is a Senior Research Associate at the Institute for Law and Justice. She

initiated the use of GIS in the Research and Planning Bureau of the Charlotte-Mecklenburg

Police Department during the mid-1990s and gained national level experience while

working at the National Institute of Justice’s Crime Mapping Research Center. Elizabeth is

also a doctoral candidate at the University of Maryland in the Geography Department.

Brook Kearly is a graduate student in the Department of Criminology and Criminal Justice

at the University of Maryland. She is also a data analyst with the Baltimore City State’s

Attorney’s Office, Project Safe Neighborhoods initiative.


Heather Fogg is a graduate student in the Department of Criminology and Criminal Justice at

the University of Maryland, where she serves as the project manager of the Global Terrorism

Database project. She is also a program analyst with the Court Services and Offender Supervision

Agency in Washington, DC.

Penny Beatty, M.S. is a graduate teaching and research assistant at the University of

Maryland at College Park. She completed her Masters degree in 2000 at Shippensburg

University. Penny is currently working on several policing and public safety initiatives

at the Institute of Governmental Service, and is also a teaching assistant for the Psychology

Department at the University of Maryland.

Heather Couture is a doctoral student at the University of Maryland, College Park. She is

also an instructor for the undergraduate program.

Julie Wartell is the Crime Analyst Coordinator for the San Diego District Attorney’s Office.

Prior to this position, Julie was the Project Director of the East Valley COMPASS Initiative,

worked as a crime analyst for the San Diego Police Department, and completed a Fellowship at

the National Institute of Justice Crime Mapping Research Center. Julie has a Masters in Public

Administration with an emphasis in Criminal Justice Administration.


a randomized experimental study of sharing crime data with

Documents