Content uploaded by Lallen T. Johnson
Author content
All content in this area was uploaded by Lallen T. Johnson on Jul 01, 2014
Content may be subject to copyright.
Original Article
A partial test of the impact of a casino on
neighborhood crime
Lallen T. Johnson
a,
*and Jerry H. Ratcliffe
b
a
Program in Criminology and Justice Studies, Drexel University, Philadelphia, PA 19104, USA.
E-mail: ljohnson@drexel.edu
b
Department of Criminal Justice, Temple University, Gladfelter Hall, 5th Floor, 1115 Polett Walk,
Philadelphia, PA 19122, USA.
E-mail: jhr@temple.edu
*Corresponding author.
Abstract Ninety-six months of crime incident data were examined to determine the extent to
which crime counts changed within the Philadelphia neighborhood of Fishtown after the opening
of a new casino. Count modeling regression results indicate that the operation of the casino had no
significant effect on violent street felonies, vehicle crime, drug crime or residential burglary in the
surrounding community. Weighted displacement quotient analyses suggest that the operation of
the casino may be associated with an increase in vehicle crime in the area surrounding the casino
neighborhood, indicative of crime displacement. Drug and residential burglary offenses in the
area surrounding the casino neighborhood decreased after the casino opened, suggestive of a dif-
fusion of benefits possibly tied to a change in local police patrols. Net of unexamined police patrol
changes and casino opening simultaneity effects, the current study is unable to identify a neigh-
borhood level effect of the casino on crime. Additional research is necessary to examine localized
effects of casinos on various offenses.
Security Journal advance online publication, 30 June 2014; doi:10.1057/sj.2014.28
Keywords: casino; Philadelphia; time series; count modeling; neighborhood
Introduction
Before the last decade the presence of casinos had been limited to a few well-known sites in
the United States. Atlantic City (NJ), Reno and Las Vegas (NV), and numerous Indian
reservations have been able to capitalize on this limited presence by becoming central hubs
for gaming and associated industries. The American recession of the early 2000s as well as
continued national (and international) economic woes have, however, caused many law-
makers to reconsider whether the added tax revenue of the gaming industry outweighs its
perceived costs. For example, the Pennsylvania Gaming Control Board (2011b) was charged
with the oversight of the casino industry and was the first such state organization created in
40 years. Since then, Pennsylvania has authorized the development of 10 gaming establish-
ments within its borders. This move received substantial (though not unequivocal) support
© 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
www.palgrave-journals.com/sj/
from local politicians and it is clear why; according to the Pennsylvania Gaming Control
Board (2011a), slot machines were responsible for US$2.2 billion in revenue for the state
during the 2009–2010 fiscal year. Thus the desirability of casinos in cash-strapped states
should come as no surprise.
Notwithstanding expected monetary benefits, anti-casino interest groups have articulated
potential costs that are expected to be associated with casino development. Among these
include beliefs that gambling will precipitate a host of social problems such as alcoholism,
gambling addiction and organized crime, and that the gambling industry will exploit the poor
and elderly (Casino-Free Philadelphia, 2012). Social and economic cost concerns have also
been expressed by state-level governmental agencies (Pennsylvania Intergovernmental
Cooperation Authority, 2007). In Maryland, Attorney General Curran (1995) released a
damning opinion of casinos, arguing that they would lead to substantial increases in violent,
property, domestic, white collar and organized crime, as well as child abuse.
The federal government also has weighed in on the effects of gambling on American
society. In 1996 Congress authorized creation of the National Gambling Impact Study
Commission (1999). The Commission was charged to ‘…conduct a comprehensive legal
and factual study of the social and economic impacts of gambling on (A) Federal, state,
local, and Native American tribal governments; and (B) communities and social institu-
tions generally, including individuals, families, and businesses within such communities
and institutions’(National Gambling Impact Study Commission, 1999, pp. IV–2).
Although it was hesitant to form conclusions on the casino/crime relationship, citing
issues with current research, two points are noteworthy. First, the Commission’sreviewof
research suggested that studies were disproportionately based on pathological gamblers –
individuals that may engage in crime to fund their gambling habit. As such, crimes
attributed to pathological gambling may be substantially different from those attributed to
non-pathological gambling. Second, in order to accurately associate crime with the casino
industry, one must distinguish the gaming industry from the larger tourism industry (see
also Miller and Schwartz, 1998).
Unfortunately, research to address the perceived disorder and crime-producing effects of
casinos has been limited to the municipality and county levels, leaving our understanding of
more micro-level neighborhood impacts limited. When the SugarHouse Casino opened its
doors in September 2010, just outside of downtown Philadelphia and in the largest city in the
United States to host a representative of the gaming industry (Associated Press, 2010), we
had an opportunity to examine the criminogenic impact of casinos within a more localized
environmental context.
The contributions of the current study are threefold. First, we examine the relationship
between the location of an urban casino and its immediate surrounding neighborhood, rather
than focusing on a larger areal unit such as the city or county. Second, we provide a new
perspective on the casino/crime link by considering it within the context of Philadelphia.
Urban casino research to date has overwhelmingly focused on cities historically known for
their expansive gambling/entertainment districts. Philadelphia has no such district with legal
gambling, which allows this research to avoid the need to theoretically or empirically
disentangle the effect of tourism from any crime-generating effects of the casino. Third, the
availability of geolocated crime data allows us to examine any changes in crime volume in
the immediate neighborhood environment of the casino alongside potential displacement or
diffusion of benefits effects (Clarke and Weisburd, 1994).
Johnson and Ratcliffe
2 © 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
The remainder of the article is structured as follows. First, we review prior literature on
casinos and crime, organized by spatial unit of analysis. Second, the study site is described.
Third, a methodology for examining the relationship between casinos and neighborhood
crime is presented. After results are outlined, we conclude with a discussion of the
implications of neighborhood level investigations of the casinos and crime link.
Literature Review
City-level research
The most common research on casinos and crime has focused on the effects of casino
development on cities and towns in the Atlantic City (NJ) region. Albanese (1985) was
among the first to investigate the role of casinos on urban crime, using index crime data from
1978–1982 for murder, forcible rape, robbery, aggravated assault, burglary, larceny and motor
vehicle theft. Results indicated that there was a positive correlation between casinos and index
crimes, but after controlling for the population at risk the correlation was negligible. This
suggested that crime increases were likely due to increases in Atlantic City’s population over
time. Police manpower and state crime rates also had negligible effects.
Research suggests that towns more accessible via road networks tend to have higher
rates of crime than those less accessible (Groff et al, 2014). Furthermore, property values
in accessible communities appear to be negatively affected by violence, burglary, robbery
and vehicle theft rates to a greater extent than in non-accessible communities (Buck et al,
1991). This may explain why towns accessible to the casino resort of Atlantic City via
major roads tended to have higher crime rates than non-accessible towns
1
(Friedman et al,
1989). Studies of casinos in Biloxi (MI), however, provided inconsistent findings on the
casinos and crime linkage. While one study found no effect of casino introduction on city
crime rates (Chang, 1986), a subsequent study, however, did identify statistically
significant increases in larceny theft and motor vehicle theft after the introduction of
riverboat gambling (Giacopassi and Stitt, 1993).
Matched quasi-experimental designs have been used to compare monthly crime rates in
municipalities with and without casinos, pre- and post-casino introduction (Stitt et al, 2003).
2
Non-parametric tests revealed that the addition of casinos was associated with marginally
significant increases in larceny, liquor violations, homicide and prostitution; however, when
controlling for the population at risk these marginal differences disappeared. Time series
analyses of Part I crime in two Indiana towns with riverboat gambling have yielded divergent
results (Wilson, 2001): In Hammond, the introduction of riverboat gambling had no
discernible effect on crime, but in the case of Rising Sun the post-intervention period
demonstrated significant increases in assaults and thefts.
More recent scholarship has employed spatial and temporal analysis to describe hot spots
of disorder offenses possibly associated with casinos. Casino hot spots in Reno (NV) were
more likely to be associated with public drunkenness, drugs and trespassing as well as
reports of suspiciousness, than non-casino hot spots (statistical tests lacking) (Barthe and
Stitt, 2009a).
3
Related research looking at the distribution of disorder crimes by hour in
casino versus non-casino hot spots indicated little temporal disagreement between hot spot
types (Barthe and Stitt, 2009b).
A partial test of the impact of a casino on neighborhood crime
3© 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
County-level research
At the county level, casino presence appears to be related to significant increases in violent
and property crime in the state of Wisconsin (Gazel et al, 2001). Furthermore, spatial lag
modeling suggested non-casino counties adjacent to casino counties should have higher
crime rates as a result of their proximity. Such findings may be related to the statistical
technique employed. For example, while a panel-design study found significant adjacency
effects when using an ordinary least squares regression model, the use of the more robust
generalized estimation model indicated no real effects (Koo et al, 2007). Research
considering temporal lag effects found that the effect of casino presence on county crime
shortly after opening was minimal (Grinols and Mustard, 2006); however, as time increased
from the intervention date the presence of the casino was associated with significant
increases in crime.
In addition to index crimes, the association of gambling with social deviance and socially
undesirable conditions has been a focus of anti-casino interest groups (Casino-Free
Philadelphia, 2012; CBS News Miami, 2012). While some concerns are difficult to
substantiate, others may have merit. In particular, the serving of alcohol at most casinos
may have implications for the prevalence of driving under the influence of alcohol. Counties
with casinos have tended to have more alcohol-related fatal accidents than non-casino
counties (Cotti and Walker, 2010). And, counties adjacent to casino counties also experience
higher levels of alcohol-related fatal accidents. The relationship between alcohol-related
fatalities and gambling may not be surprising considering that pathological gambling and
alcohol abuse tend to be co-occurring disorders (Grant et al, 2002). On the other hand,
research appears inconsistent in validating assumed linkages between certain negative social
conditions and locations of the gaming industry. A study matching casino counties to non-
casino counties found no relationship between the presence of the gambling industry and
suicide rates, and differences in divorce rates between casino and non-casino counties were
statistically non-significant in half of the matched pairs (Nichols et al, 2004). Other research
found no relationship between casinos and unemployment and bankruptcy (Koo et al, 2007);
yet, a matched design methodology of eight counties revealed significant increases in casino
county bankruptcies post-intervention (Nichols et al, 2000).
In summary, the extant literature has concentrated on spatial units that are larger than the
neighborhood level, examining crime rates at the city and county scale, often with conflicting
findings. The dearth of literature on spatial effects of crime at the immediate neighborhood or
sub-neighborhood level suggests that the impact of a casino on local conditions is difficult to
predict. With the development of the SugarHouse Casino in a Philadelphia neighborhood,
we have an opportunity to examine the introduction of the casino using pre–post intervention
neighborhood crime data.
Theoretical Perspective at the Neighborhood Level
The more socially active members of local communities often view the introduction of
facilities such as casinos as likely to have a negative effect on crime rates in the immediate
neighborhood, with expectations of increased street and property crime. Other concerns
include worries about traffic upheaval, public intoxication, more raucous activity late at night
Johnson and Ratcliffe
4 © 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
and a general perception of increased disorder. With these concerns being paramount in the
minds of local residents and the police, an environmental criminology/crime science
perspective that examines the introduction of a casino as a potential crime generator or
crime attractor is appropriate. According to Brantingham and Brantingham (1995) crime
generators are places to which a large volume of people are drawn for non-criminal reasons
(such as coming to a casino to gamble), yet their collective presence provides increased
opportunities for offenders to interact with potential victims or to exploit criminal
opportunities. Crime attractors are places where people are specifically drawn for the
opportunity to commit crime or engage in deviant behavior (bar and entertainment areas with
known lax enforcement standards can be an example).
It is also important to consider the role of routine activities in casino-related crime. In this
sense crime is a function of the extent to which casinos provide opportunities for motivated
offenders and potential victims to interact in an environment absent capable guardians (Cohen
and Felson, 1979). Criminal opportunities abound and potential offenders are likely to be aware
of them. First, casinos serve alcoholic beverages, which may lower the vigilance of potential
victims and serve to boost the confidence of potential offenders. Second, they are locations that
exchange large amounts of money with patrons, and patrons on the receiving end of cash
winnings may be seen as potential targets. Third, the nature of casino gaming requires
individuals to make economic decisions about how much money should be invested in playing.
Individuals with low self-control may be less judicious in their playing decisions and as a result
more likely to find themselves in compromised economic situations. In turn, casinos may be
attractive places to those with low self-control because they encourage risk-taking behaviors, a
characteristic found to be associated with criminal behavior (Gottfredson and Hirschi, 1990).
Such individuals may engage in criminal activity to compensate for their losses.
In response, casinos employ a number of strategies to prevent criminal behavior. For
example, the use of place managers (Eck, 1995) such as table attendants, supervisors and
bartenders directly or indirectly set behavioral norms for the casino premises. Furthermore
security guards and law enforcement may provide a sense of formal guardianship over
potential victims, deterring motivated offenders. The presence of place managers may also
displace criminal activity away from the casino grounds. For example, crime prevention at and
around the immediate casino location may cause motivated offenders to rob pedestrians of
their winnings after they have left the casino. If this is the case, localized crime prevention at a
casino may result in an increase in neighborhood violence and crime. Conventional thinking
suggests that focused crime intervention and prevention strategies as well as highly monitored
environments may merely push crime to surrounding areas; however, some modest displace-
ment may indicate that crime prevention tactics are working –if matched with significant crime
reduction at the target site. In other words opportunity reduction (or the removal of the most
optimal offense locations) may cause some crime to be displaced to surrounding, less
advantageous areas, yet still result in a net decrease in overall crime (Reppetto, 1976).
Alternatively, it may be that the removal of opportunities at one site spreads a crime
reduction side-effect to nearby locations. Crime reduction evaluations of Camden, New
Jersey (Ratcliffe and Breen, 2011) and Philadelphia, Pennsylvania (Lawton et al, 2005) hot
spots policing strategies not only found decreases in target area crime but also a diffusion of
benefits, whereby surrounding areas not the target of any additional prevention also
experienced decreases in crime. Thus a crime science theoretical perspective suggests that
place management and guardianship at a casino might displace offender behavior or diffuse
A partial test of the impact of a casino on neighborhood crime
5© 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
reduction benefits to surrounding streets to a lesser or greater degree. While violent offenders
may seek out potentially wealthy casino patrons on the way to or from the casino in nearby
streets, given that identifying casino patrons (from non-casino-attending citizens) becomes
increasingly difficult as distance from the casino increases, this effect is likely to be
measured as a local neighborhood disturbance rather than a citywide impact.
In the next sections, we describe the study site and the security arrangements at the casino,
and then explain the methodology employed to assess the local crime impact of the opening
of the casino.
Study Site and Evaluation Design
SugarHouse Casino is located on the Philadelphia waterfront in the 26th police district, just
north of Philadelphia’s downtown area. The neighborhood is known to locals as Fishtown
for its historical prominence in the fishing industry (see Figure 1). The area north and west of
Interstate-95 is composed of row homes common to the city, while the area south and east of
the interstate is less residential with a number of old factory buildings.
In May 2011, we had an opportunity to tour the facilities, as well as conduct semi-
structured interviews with the Director of Security and the Director of Communications
regarding crime and security inside the casino and on its grounds. These interviews
Figure 1: SugarHouse Casino area map.
Johnson and Ratcliffe
6 © 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
(summarized below) were predominantly used to gain insight into the security arrangements
in the area, and revealed a collaborative relationship with the local and state police
departments, transportation arrangements and environmental characteristics that may
collectively contribute to increased safety.
Transportation
Respondents reported that 2 million visitors enter the casino in a year, with peak time being
during the weekend. About 50 visitors per day arrive by taxi Sunday through Thursday and
about 75–100 on Friday and Saturday. The local public transportation agency, the South-
eastern Pennsylvania Transportation Authority (SEPTA), has a bus route that provides
service to the main road at the casino. About 150 people per day arrive, and 205 depart, via
SEPTA bus. Valet service parks about 13 cars per day in a lot with approximately 1800
spaces. In addition to the public bus routes that service the casino, the casino operates its own
shuttle with pick-up and drop-off locations throughout the city. Its two bus routes operate
from Sunday to Thursday, 14:00−22:30, and weekends 23:00−12:00 (SugarHouse Casino,
2011). Interviewees estimated that 55–100 patrons arrive by walking
4
and about 1–2 private
hire bus trips arrive per week.
Local, state and private policing
Policing on the property of SugarHouse Casino is divided between the Pennsylvania
State Police and the Philadelphia Police Department. The Pennsylvania State Police has
exclusive policing jurisdiction over the gaming floor of the casino. The casino has its own
State Police office with a local commanding officer, and a regular detail of plain-clothes
officers has been assigned to police the gaming floor.Inadditiontopossessingtypical
police powers, the State Police is also responsible for gaming violations that occur inside
the casino. Both respondents reported that disorderly conduct is the most common
offense, albeit rare.
As the municipal law enforcement agency, the Philadelphia Police Department has
jurisdiction over the casino neighborhood outside of the casino building. The 26th Police
District has been provided with a radio that is connected to the casino radio system, making it
easier to provide assistance as necessary. Respondents agreed that the local police can be
seen daily patrolling through the parking lot. In spite of their official jurisdiction being
limited to the outdoor areas of the casino, officers can sometimes be seen on the casino floor
and can be summoned when a uniformed presence is desired. The Philadelphia Police
provide backup assistance to the Pennsylvania State Police as necessary, most commonly
with the coordination of protection for armored truck money deliveries and exchanges.
The Pennsylvania Gaming Control Board mandates a minimum of three security officers
at the front entrance and two in the parking lot, 24 hours a day. After dark, as many as seven
security personnel may be stationed in the parking lot to provide an additional presence.
Additionally, two to three security officers patrol the outdoor grounds via bike patrol.
The Director of Security reported that security resources are shifted as necessary to assist
employees during shift changes. The security detail also maintains two golf carts and a
small SUV to escort employees and guests to their vehicles upon request. An extensive
A partial test of the impact of a casino on neighborhood crime
7© 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
closed-circuit television (CCTV) camera system is used to monitor the outdoor decks, gaming
floor and parking lots. Emergency call boxes are also stationed throughout the parking lots.
Patrons are able to receive free drinks as long as they are gambling, but are limited to two
drinks per hour. Staff members receive Responsible Alcohol Management Program
training.
5
The Director of Communications noted that inebriated persons are not allowed to
play or leave the premises driving a vehicle; instead, they are offered coffee and water and a
staff member will call a cab or relative to transport them home. Disruptive patrons (that are
not inebriated) are asked to leave the establishment in lieu of prosecution. Those who return
are charged with trespassing.
Methodology
When the casino opened in September 2010, the 26th Police District created a special casino
patrol area. This area of slightly less than half a square mile (shown in Figure 1) is patrolled
by one sergeant and 13 officers who provide coverage 24 hours a day, 7 days a week. An area
adjacent to the patrol area was selected to examine potential displacement and diffusion of
benefits effects. This area is roughly two city blocks wide and is comparable in size to
displacement areas used in prior urban hot spots research (Weisburd and Green, 1995b). It is
also similar in size to the casino patrol area (0.49 square miles), but not so large that it
incorporates areas with crimes that are unlikely to occur due to displacement (Ratcliffe and
Breen, 2011). A control area of 0.77 square miles was selected within the 26th Police District
that is 2000 feet wide, and set 1200 feet away from the displacement area. This was done to
reduce the possibility of contamination effects (Weisburd and Green, 1995a). In other words,
the spatial separation of the control and displacement areas was designed to create a control
area that retained much of the neighborhood demographic and structural characteristics, but
would be free from influence of processes related to the casino.
Crime data were sourced from the Philadelphia Police Department’s Incident Transmittal
System. Crime incidents that occurred between January 2004 and December 2011 were
considered in the analyses and aggregated by month. For the time series analysis, this
permitted 80 pre-casino data points and 16 post-casino opening measurements. Analysis was
limited to four specific offense categories theoretically relevant to casino operation as
described in the literature review. ‘Violent street felonies’was a category that included
homicides, aggravated assaults and street robberies. Given the proximity of residential housing
in the neighborhood as well as the concerns of residents, we also include a ‘residential
burglary’category. Monthly vehicle theft reports and monthly theft from vehicle reports were
examined. These two data sets were analyzed together as a ‘vehicle crime’category. Finally,
we included a ‘drug crime’category to represent all detected illicit drug activity in the area
(buying and selling) in case the opening of the casino was associated with an increase in illicit
drug crimes in the immediate vicinity of the facility.
Analytical strategies
We chose two analytical strategies: weighted displacement quotients (WDQs) and time
series analysis. First, we adopted the WDQ of Bowers and Johnson (2003) in order to
Johnson and Ratcliffe
8 © 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
address two of the perennial questions that are raised with regard to potentially crimino-
genic facilities: Has the casino’s presence led to increased crime in the immediate area
(Ratcliffe, 2012), and if not, has crime been simply displaced to nearby locations (Cornish
and Clarke, 1987)?
The WDQ is a ratio measure designed to compare the change in crime before and after the
introduction of a crime prevention initiative relative to both the change in crime in a potential
displacement area and a control area (Bowers and Johnson, 2003). First, a success measure
compares the change in crime after the introduction of a crime prevention initiative in
relation to a control area. If there has been an improvement in the target intervention area
relative to any change in the control area, then it can be claimed that the intervention was a
success. In this case, the researcher can continue to examine the relationship between the
displacement area and the control area. This will indicate whether any of the crime reduction
in the target area was potentially the result of a shift of criminal activity to the displacement
area. This second calculation is termed the buffer displacement measure (also termed the
displacement measure in Bowers et al, 2011). Finally, the WDQ is calculated using the
success measure as the denominator and the buffer displacement measure as the numerator.
The WDQ indicates whether any displacement has been greater or less than the amount of
crime reduced in the target area. The full equation is as follows, using the terminology from
Bowers et al (2011):
WDQ ¼
Da
Ca
-Db
Cb
Ra
Ca
-Rb
Cb
(1)
where the crime count in the intervention (target) area is represented before the intervention
(R
b
) and after (R
a
), the crime count in the control area is measured before (C
b
) and after (C
a
)
the intervention, and the crime count in the buffer (catchment) area is measured before (D
b
)
and after (D
a
) the intervention. In our case, we replaced the area targeted by a theoretical
crime prevention initiative in the above discussion and equation with the timing and
neighborhood of the area likely to be affected by the introduction of the casino. With regard
to the success measure (the denominator in equation 1), we would expect this value to be
positive if crime had increased in the area after the opening of the casino. For a time period,
we chose 1 year before and 1 year after the casino opening, a time period described by
Bowers and Johnson (2003)as a ‘reliable sampling frame’(p. 284).
While the WDQ allows us to determine crime change and estimate any displacement (or
diffusion of benefits) it is less clear on whether any changes in crime are statistically
significant. To reinforce this analysis we included a second approach, a time series analysis
employing a non-linear regression using time-varying covariates to model temporal trends.
Time-varying covariates were employed to model particular dimensions of temporal trend
and seasonality, as well as the opening of the casino in the study area. A linear variable
represents the sequential position of the particular month in the data set, starting in January
2004. This centered variable captured any long-term temporal trend in the data. If crime were
generally increasing over time then this variable would be positive, and if the general trend
were downward it would be negative. Similarly a centered quadratic variable was included
to capture non-linear changes to any long-term trend. Some types of crime display a
significant seasonal component, especially robbery (Sorg and Taylor, 2011). As a result, it is
important to model seasonal effects. One option is to include 11 dummy variables
A partial test of the impact of a casino on neighborhood crime
9© 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
representing each month except a reference month (Greenberg and Roush, 2009), or to
include a measure of temperature in the analysis to represent the seasonality displayed in
some street crime data (Ratcliffe et al, 2009). In the current study we address both
perspectives by modeling a linear month sequence variable, as well as an average monthly
temperature variable. Data for the latter indicator were obtained from the historical archives
available at www.wunderground.com. Daily weather observations for the zip code 19102
were downloaded and the monthly mean temperature was calculated. A dummy casino
variable was coded 0 before the opening of the casino and 1 from September 2010 onwards.
Data for the casino area were modeled using a quasi-experimental time series design. The
advantage of a time series design is that it can compare multiple observations of a pre-
intervention period to a post-intervention period, and account for temporal autocorrelation
(Shadish et al, 2002). Count data, however, are inherently problematic for time series
analysis. Real-valued time series models, such as autoregressive integrated moving average
(ARIMA) models (Box and Jenkins, 1976; MacCleary and Hay, 1980; Chatfield, 1989),
have been applied to crime data for many years (Krimmel and Mele, 1998; Novak et al,
1999; Degenhardt et al, 2005; Chamlin and Myer, 2009). Unfortunately, when attempting to
model non-negative integer-valued data such a low volume crime counts, ARIMA processes
may be inappropriate given a key assumption of ARIMA time series modeling is of
normality in the random shocks of the underlying error structures (Quddus, 2008; Greenberg
and Roush, 2009). Simply put, the inability of recorded crime data to exhibit negative values
truncates low volume crime counts such that the data more often exhibit Poisson or negative
binomial distributional qualities.
While crime count data tend to be non-negative and integer, and often follow a
generalized Poisson distribution, they are usually over-dispersed where the variance is not
found to equal the mean as in a true Poisson distribution. As such, a negative binomial
regression model can also be appropriate. Examples of negative binomial modeling of low
count crime data can be found in the crime analysis literature (Greenberg and Roush, 2009).
Results
Table 1 shows the monthly descriptive statistics of the dependent variables for the casino
area over the time period 2004−2011 inclusive. As shown in Table 1 the casino-area crime
count variances are greater than the respective means, suggesting over-dispersion of the data.
Graphing the data series (not included in the article but available from the authors on request)
showed that most of the series remained stable during the study period, with the exception of
vehicle crime, a series that appeared to have a long-term trend increase over the period of
Table 1: Descriptive statistics of dependent variables (casino patrol area, 96 months)
Series Minimum Maximum Mean Standard deviation Variance
Violent street felonies 0 8 2.92 1.92 3.70
Vehicle crime 1 30 12.89 5.82 33.87
Drug crime 0 6 1.26 1.24 1.54
Residential burglary 0 6 1.21 1.20 1.43
Johnson and Ratcliffe
10 © 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
2004−2011. This potential increase over time is dynamically modeled by the linear time-
varying covariate.
WDQ results
Table 2 shows the pre/post crime counts for the 12 months before and after the opening of the
casino, the success measure, buffer measure and the WDQ value. In the immediate casino
area, violent street felonies increased by seven in the year following the casino opening, and
vehicle crime also increased. Detected drug crime was down, as was residential burglary.
The buffer area saw a decrease in drug offenses, but increases in all other crimes examined.
The control area, representing the overall trend of the region, showed an increase across all
crime types except drug offenses.
The success measure for the four crime types showed that the casino area performed worse
than the control area in violent street felonies (indicative of a relative crime increase) but better
than the control area in residential burglary, vehicle crime and drug crime. As Bowers and
Johnson (2003, p. 285) note in regard to the assessment of a crime prevention initiative, ‘If the
denominator is positive then this means that the scheme has been unsuccessful, and it is
difficult to relate any change in the buffer zone to the scheme. Moreover, it would be
theoretically inappropriate to look for displacement/diffusion of benefit, and therefore, in this
case, the weighted displacement quotient should not be used’. While we are not examining a
crime prevention initiative in this article, we would argue that this line of thinking is equally
applicable, though in a different way. If crime has increased in the target area then there is no
reason to expect displacement to the buffer area, given the apparent increase in criminal
opportunities in the intervention neighborhood. As such, we concur with Bowers and Johnson
that reporting the buffer measure of WDQ results is inappropriate. The buffer measure for the
remaining three crime types shows a negative value for both residential burglary and drug
crime, indicative of a diffusion of crime prevention benefits to the surrounding buffer area,
though some displacement for vehicle crime.
The WDQ is the last value reported in Table 2. This shows values greater than one for
residential burglary and drug crime, but less than −1 for vehicle crime. For a summary
of the overall results we can cross-reference Table 1 of Bowers and Johnson (2003) and
summarize the WDQ analysis as follows. Violent street felonies increased in the target area
compared with the control area. Vehicle crime decreased in the target area relative to the
Table 2: Success measure, buffer measure and WDQ results
Violent street felonies Residential burglary Vehicle crime Drug crime
Target before (R
b
) 35 19 180 15
Target after (R
a
) 42 16 207 13
Buffer before (D
b
) 37 45 286 26
Buffer after (D
a
) 64 51 353 18
Control before (C
b
) 109 81 345 110
Control after (C
a
) 121 114 405 97
Success measure 0.03 −0.09 −0.01 −0.002
Buffer measure —−0.11 0.04 −0.05
WDQ —1.15 −4.01 21.68
A partial test of the impact of a casino on neighborhood crime
11© 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
control area; however, there was substantial displacement indicating that the introduction of
the casino made the vehicle crime problem in the combined treatment/buffer area worse than
before the casino was opened. Residential burglary and drug crime decreased in the target
area relative to the control area; furthermore, there was substantial diffusion of benefits to the
surrounding areas indicating that the introduction of the casino made the residential burglary
and drug crime problem in the combined treatment/buffer area better than before the casino
was opened with an overall positive net effect.
The issue of increasing violent street felonies would clearly be of concern to casino
management, local residents and the police; however, the question remains as to whether this
increase is statistically significant or could be attributed to the natural fluctuation that is
partially associated with most crime patterns. We are unable to address this issue with the
ratio measure; therefore, we turn to the regression component of the analysis.
Casino area regression results
The issue of data over-dispersion was examined statistically. For the most part, Poisson
goodness of fitχ
2
values were statistically significant, indicating over-dispersion (violent street
felonies χ
2
=115.182, P=0.04; drug crime χ
2
=120.598, P=0.02; vehicle crime
χ
2
=148.849, P<0.01). The exception is residential burglary (χ
2
=112.736, P=0.061). As a
result, over-dispersed models were fitted using negative binomial regression and the residential
burglary series fitted using Poisson regression. Because of the issue of the experiment-wise
error rate problem, the standard social science level of statistical significance (P<0.05) was
adjusted using a Bonferroni correction to P<0.0125 for the independent variables. In the
statistical model tables below, the βcoefficients are replaced with the incidence rate ratios to
ease interpretation. Regression coefficients in count models represent the change in the log of
expected counts of the response variable with a one unit change in the predictor, but rather than
interpret logged values we focus on the incidence rate ratios. Standard errors are also
converted.
Table 3: Regression results for the four studied crime types
Violent street felonies
a
Vehicle crime
a
Drug crimes
a
Residential burglary
b
Intercept 1.529 6.460* 1.147 0.572
(0.390) (0.898) (0.431) (0.217)
Average monthly temperature 1.0098 1.0099* 1.0046 1.0135
(0.004) (0.002) (0.006) (0.005)
Linear trend 1.0001 1.0069* 0.9892 1.0023
(0.0037) (0.0020) (0.0060) (0.0053)
Quadratic trend 1.00009 1.0001 0.9996 0.9999
(0.0001) (0.00007) (0.0002) (0.0002)
Casino 1.001 1.055 1.412 1.155
(0.328) (0.180) (0.772) (0.531)
Notes:*P<0.0125.
Incident rate ratios reported, with standard errors in parentheses. N=96 months with 16 months post intervention.
a
Negative binomial regression.
b
Poisson regression.
Johnson and Ratcliffe
12 © 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
Table 3 displays the results of negative binomial models estimating violent street felonies,
vehicle crime and drug crime in the casino patrol area. It also shows the Poisson regression
results for residential burglary. As can be seen from the results of the violent street felonies
regression, none of the variables achieve statistical significance. Changes over time were
unrelated to violent street felonies in the casino patrol area suggesting that the area had a
relatively stable violent street felonies trajectory over time. The constant, average monthly
temperature variable and linear trend variables were statistically significant for vehicle crimes.
These indicate that for every 1 degree Fahrenheit increase in monthly average temperature
there is an expected 1 per cent increase in the count of vehicle crimes in the casino area, and
that across the entire time series, controlling for other factors, there was a monthly 0.6 per cent
increase in expected vehicle crime counts. Importantly however, the opening of the casino had
no statistically significant impact on any of the crime types assessed.
Discussion
This study evaluated the effect of a new casino development on four crime types at the
neighborhood level. SugarHouse Casino opened its doors in the Fishtown section of
Philadelphia in September 2010. Results of multiple time series analyses using 96 months of
crime incident data and a WDQ analyses indicated mixed effects. The combined results indicate
that while violent street felonies increased at a rate slightly greater than violence in the control
area, this increase was not statistically significant when examined in the context of the longer
trend since 2004. Vehicle crime reduced in the target area; however, there was substantial
displacement and the reductions in vehicle crime were not statistically significant over the long
term. Both residential burglary and drug crime reduced in the casino area (again though, not
significantly from a statistical perspective) and there were reductions in these crimes in the
buffer areas. In summary, there is no evidence that the opening and operation of the casino had a
significantly detrimental effect on the immediate neighborhood in terms of vehicle crime, drug
activity, residential burglary or violent street felonies. These findings are contextualized below.
First, we should note that this is not a stand-alone quasi-experimental evaluation of the
introduction of a casino to a neighborhood, due to the additional complication of
the Philadelphia Police Department instigating a dedicated patrol to the neighborhood.
The additional patrolling from 14 assigned officers may have acted to provide additional
deterrence to any criminal activity. The dosage issue of these officers is an unresolved
question here, though the numbers would suggest a limited role for the police in any crime
suppression. Even if we consider the inclusion of the sergeant into any patrol activities, there
was limited ability to provide substantial round-the-clock policing. If officers worked a
40-hour week, then this averages out at approximately an additional 3.5 officers on duty at
any particular time in the patrol area. In reality, with refreshment breaks, vacation, sickness
and so on, the actual level of dedicated patrol activity is likely to be less. Whether this is
sufficient additional patrol for an area to have any impact cannot be tested here but it is clear
that any additional resources were modest at best.
The results do at least suggest that neighborhood concerns that the introduction of a
casino will herald significant increases in local crime rates appear unfounded. This first
examination of local crime rates at the sub-neighborhood level would appear to support the
notion that crime rates are largely unaffected by the introduction of a casino, or at least any
A partial test of the impact of a casino on neighborhood crime
13© 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
significant increases can be held in abeyance by the reassignment of modest police resources
from an existing pool of staff (the district was not assigned additional personnel for the
casino but instead reassigned officers from within the police district).
One further threat to the validity of this research worth mentioning is instrumentation.
Semi-structured interviews with casino management revealed that the inside of the
casino is supervised by the Pennsylvania State Police, while all outdoor areas fall under
the jurisdiction of the Philadelphia Police. The conclusions and results of analyses
presented here are relative to data derived from the Philadelphia Police Department. We
find this to be a minor limitation as we are most interested in linkages between the
presence of the casino and crime in the surrounding neighborhood, rather than activity
that takes place within the location. Relative to the availability of appropriate data, future
research should explore the nuances of casino-related crime. Specifically, separately
analyzing crime types occurring within the casino and specifically related to the nature of
casinos (such as thefts from dealers) versus crime associated with casinos occurring
outdoors (such as car thefts) would enable a richer understanding of casinos, crime and
deviance.
With regards to the crime counts, we have two different measures. It could be argued that
the WDQ results are meaningful on their own given that we are examining the total
population of crime events in the target, displacement and control areas, and thus there is no
need for a statistical test. But as Fotheringham and Brunsdon (2004) point out, statistical tests
are useful for not just sample inference but also process inference. The longitudinal time
series analysis was designed to determine if changes detected by the WDQ were indicative of
more than simply modest random variation in crime patterns. The statistical tests modeled no
statistically significant changes in crime counts due to the introduction of the casino. This
would suggest that the variations in crime counts after the opening of the casino detected
with the WDQ are within the parameters of the types of natural fluctuations that we would
expect from longitudinal crime series.
Therefore while additional crime opportunities may have increased the pool of potential
victims for robbery, and increased the number of potential offenders drawn to engage in
burglary or vehicle crime, the presence of added patrons to the neighborhood could also have
increased formal as well as informal guardianship. Formal in terms of casino security and
additional police patrols, and informal guardianship through the addition of more eyes on the
street. These opposing forces (opportunity and guardianship) may have both increased,
offsetting each other.
The displacement findings are interesting. In anticipation of the casino opening, the 26th
Police District commander created the special patrol district, to which were assigned
additional police officers. The increased police attention in the special patrol area may have
led to the displacement of vehicle crime to the surrounding area. Officers that were re-
assigned to the patrol area were not replaced in the rest of the district. It is possible that the
relative reduction in personnel outside of the casino area reduced patrol deterrence in the
displacement area, while suppressing crime in the target area.
In the current study we endeavored to analyze neighborhood level criminogenic effects of
a casino in an urban environment. Findings here do not settle the debate on casino and crime
linkages, but contribute to a growing body of knowledge and suggest a need for more
neighborhood level research. At the least, findings demonstrate that oft-stated community
concerns regarding local crime conditions with the addition of a casino to a neighborhood
Johnson and Ratcliffe
14 © 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
were not borne out by the SugarHouse Casino example. Any potential significant crime
increases either did not occur, or were effectively controlled by a reassignment of existing
local police resources.
Notes
1 Accessible localities were those ‘…immediately adjacent to the City or at intersections of major non-toll arterial
roads to New York and Philadelphia up to a distance of 30 miles …’ (Friedman et al, 1989, pp. 616–617).
2 Municipalities include Sioux City, IA; St. Joseph, MO; St. Louis City, MO; St. Louis County, MO; Biloxi, MS;
Alton, IL; Peoria, IL and East Peoria, IL (Stitt et al, 2003).
3 Casino hotspots are high crime areas, identified by a kernel density estimation analysis, to be within 1000 feet of a
casino (Barthe and Stitt, 2009a).
4 It may be difficult to tell who walked from the parking lot versus those who arrive by walking from the
surrounding community.
5‘RAMP was created by the Pennsylvania Liquor Control Board to help licensees and their employees serve
alcohol responsibly’. It teaches employees how to identify those under the legal drinking age, those visibly
intoxicated and suspend drinking service as necessary (Pennsylvania Liquor Control Board, 2012).
References
Albanese, J.S. (1985) The effect of casino gambling on crime. Federal Probation 49(2): 39–44.
Associated Press (2010) Philadelphia becomes largest U.S. City with a casino, USA Today, http://www.usatoday.
com/news/nation/2010-09-27-phillycasino27_ST_N.htm, accessed 11 August 2012.
Barthe, E. and Stitt, B.G. (2009a) Impact of casinos on criminogenic patterns. Police Practice and Research 10(3):
255–269.
Barthe, E. and Stitt, B.G. (2009b) Temporal distributions of crime and disorder in casino and non-casino zones.
Journal of Gambling Studies 25(2): 139–152.
Bowers, K. and Johnson, S. (2003) Measuring the geographical displacement and diffusion of benefit effects of
crime prevention activity. Journal of Quantitative Criminology 19(3): 275–301.
Bowers, K., Johnson, S., Guerette, R.T., Sommers, L. and Poynton, S. (2011) Spatial displacement and diffusion of
benefits among geographically focused policing initiatives. Campbell Systematic Reviews 3: 147. The Campbell
Collaboration.
Box, G.P. and Jenkins, G.M. (1976) Time Series Analysis: Forecasting and Control. San Francisco, CA: Holden-Day.
Brantingham, P.L. and Brantingham, P.J. (1995) Criminality of place: Crime generators and crime attractors.
European Journal of Criminal Policy and Research 3(3): 5–26.
Buck, A.J., Hakim, S. and Spiegel, U. (1991) Casinos, crime, and real estate values: Do they relate? Journal of
Research in Crime and Delinquency 28(3): 228–303.
Casino-Free Philadelphia (2012) Casino facts, http://www.casinofreephilly.org/casino-facts, accessed 15 August
2012.
CBS News Miami (2012) Anti-casino group study: Mega casinos will cause crime spike, http://miami.cbslocal.com/
2012/02/01/anti-casino-group-study-mega-casinos-will-cause-crime-spike/, accessed 17 August 2012.
Chamlin, M.B. and Myer, A.J. (2009) Disentangling the crime-arrest relationship: The influence of social context.
Journal of Quantitative Criminology 25(4): 371–389.
Chang, S. (1986) Impact of casinos on crime: The case of Biloxi, Mississippi. Journal of Criminal Justice 24(5):
431–436.
Chatfield, C. (1989) The Analysis of Time Series: An Introduction. London: Chapman and Hall.
Clarke, R.V. and Weisburd, D. (1994) Diffusion of crime control benefits: Observations on the reverse
of displacement. In: R.V. Clarke (ed.) Crime Prevention Studies. Vol. 2. Monsey, NY: Criminal Justice Press,
pp. 165–183.
Cohen, L.E. and Felson, M. (1979) Social change and crime rate trends: A routine activity approach. American
Sociological Review 44(4): 588–608.
A partial test of the impact of a casino on neighborhood crime
15© 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
Cornish, D. and Clarke, R. (1987) Understanding crime displacement: An applicaiton of rational choice theory.
Criminology 25(4): 933–947.
Cotti, C.D. and Walker, D.M. (2010) The impact of casinos on fatal alcohol-related traffic accidents in the United
States. Journal of Health Economics 29(6): 788–796.
Curran, J.J. (1995) The house never loses and Maryland cannot win: Why casino gaming is a bad idea: Joint
executive-legislative task force.
Degenhardt, L., Conroy, E., Gilmour, S. and Collins, L. (2005) The effect of a reduction in heroin supply in
Australia upon drug distribution and acquisitive crime. British Journal of Criminology 45(1): 2–24.
Eck, J. (1995) A general model of the geography of illicit retail marketplaces. In: D.E. Wiesburd and J.E. Eck (eds.)
Crime and Place. Vol. 4. Monsey, NY: Criminal Justice Press, pp. 67–93.
Fotheringham, A.S. and Brunsdon, C. (2004) Some thoughts on inference in the analysis of spatial data.
International Journal of Geographical Information Science 18(5): 447–457.
Friedman, J., Hakim, S. and Weinblatt, J. (1989) Casino gambling as a ‘growth pole’strategy and its effect on crime.
Journal of Regional Science 29(4): 615–623.
Gazel, R.C., Rickman, D.S. and Thompson, W.N. (2001) Casino gambling and crime: A panel study of Wisconsin
counties. Managerial and Decision Economics 22(1-3): 65–75.
Giacopassi, D. and Stitt, B.G. (1993) Assessing the impact of casino gambling on crime in Mississippi. American
Journal of Criminal Justice 18(1): 117–131.
Gottfredson, M. and Hirschi, T. (1990) A General Theory of Crime. Stanford, CA: Stanford University Press.
Grant, J.E., Kushner, M.G. and Kim, S.W. (2002) Pathological gambling and alcohol use disorder. Alcohol Research
& Health 26(2): 143–150.
Greenberg, D.F. and Roush, J.B. (2009) The effectiveness of an electronic security management system in a
privately owned apartment complex. Evaluation Review 33(1): 3–26.
Grinols, E.L. and Mustard, D.B. (2006) Casinos, crime, and community costs. Review of Economics and Statistics
88(1): 28–45.
Groff, E.R., Taylor, R.B., Elesh, D.B., McGovern, J. and Johnson, L. (2014) Permeability across a metropolitan area:
Conceptualizing and operationalizing a macrolevel crime pattern theory. Environment and Planning A 46(1): 129–152.
Koo, J., Rosentraub, M.S. and Horn, A. (2007) Rolling the dice? Casinos, tax revenues, and the social costs of
gaming. Journal of Urban Affairs 29(4): 367–381.
Krimmel, J.T. and Mele, M. (1998) Investigating stolen vehicle dump sites: An interrupted time series quasi-
experiment. Policing: An International Journal of Police Strategies and Management 21(3): 479–489.
Lawton, B.A., Taylor, R.B. and Luongo, A.J. (2005) Police officers on drug corners in Philadelphia, drug crime, and
violent crime: Intended, diffusion, and displacement impacts. Justice Quarterly 22(4): 427–451.
MacCleary, R. and Hay, R.A.J. (1980) Applied Time Series Analysis for the Social Science. London: Sage.
Miller, W.J. and Schwartz, M.D. (1998) Casino gambling and street crime. Annals of the American Academy of
Political and Social Science 556: 124–137.
National Gambling Impact Study Commission (1999) Final report Washington DC: National Gambling Impact
Study Commission.
Nichols, M.W., Stitt, B.G. and Giacopassi, D. (2000) Casino gambling and bankruptcy in new United States casino
jurisdictions. Journal of Socio-Economics 29(3): 247–261.
Nichols, M.W., Stitt, B.G. and Giacopassi, D. (2004) Changes in suicide and divorce in new casino jurisdictions.
Journal of Gambling Studies 20(4): 391–404.
Novak, K.J., Hartman, J.L., Holsinger, A.M. and Turner, M.G. (1999) The effects of aggressive policing of disorder
on serious crime. Policing: An International Journal of Police Strategies and Management 22(2): 171–190.
Pennsylvania Gaming Control Board (2011a) Pennsylvania gaming control board 2009–2010 annual report.
Harrisburg, PA.
Pennsylvania Gaming Control Board (2011b) (Producer) Who we are and what we do video. [video], http://
gamingcontrolboard.pa.gov/?popup=video, accessed 11 August 2012.
Pennsylvania Intergovernmental Cooperation Authority (2007) Staff report of the city of Philadelphia’sfive-year
financial plan for fiscal year 2008 −fiscal year 2012 Philadelphia, PA: Pennsylvania Intergovernmental
Cooperation Authority.
Pennsylvania Liquor Control Board (2012) Protecting your business with RAMP, http://www.portal.state.pa.us/
portal/server.pt?open=514&objID=612015&mode=2, accessed 11 August 2012.
Quddus, M.A. (2008) Time series count data models: An empirical application to traffic accidents. Accident Analysis
and Prevention 40(5): 1732–1741.
Johnson and Ratcliffe
16 © 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
Ratcliffe, J.H. (2012) The spatial extent of criminogenic places: A changepoint regression of violence around bars.
Geographical Analysis 44(4): 302–320.
Ratcliffe, J.H. and Breen, C. (2011) Crime diffusion and displacement: Measuring the side effects of police
operations. Professional Geographer 63(2): 230–243.
Ratcliffe, J.H., Taniguchi, T.A. and Taylor, R.B. (2009) The crime reduction effects of public CCTV cameras:
A multi-method spatial approach. Justice Quarterly 26(4): 746–770.
Reppetto, T.A. (1976) Crime prevention and the displacement phenomenon. Crime and Delinquency 22(2): 166–177.
Shadish, W.R., Cook, T.D. and Campbell, D.T. (2002) Experimental and Quasi-Experimental Designs for
Generalized Causal Inference. Boston, MA: Houghton-Mifflin Company.
Sorg, E.T. and Taylor, R.B. (2011) Community-level impacts of temperature on urban street robbery. Journal of
Criminal Justice 39(6): 463–470.
Stitt, B.G., Nichols, M. and Giacopassi, D. (2003) Does the presence of casinos increase crime? An examination of
casino and control communities. Crime and Delinquency 49(2): 253–284.
SugarHouse Casino (2011) Catch a free ride on the sugar express, http://www.sugarhousecasino.com/sugar-express/
index.php, accessed 11 August 2012.
Weisburd, D. and Green, L. (1995a) Measuring immediate spatial displacement: Methodological problems. In:
J. Eck and D. Weisburd (eds.) Crime Prevention Studies. Vol. 4. Monsey, NY: Criminal Justice Press,
pp. 349–361.
Weisburd, D. and Green, L. (1995b) Policing drug hot spots: The Jersey City drug market analysis experiment.
Justice Quarterly 12(4): 711–735.
Wilson, J.M. (2001) Riverboat gambling and crime in Indiana: An empirical investigation. Crime and Delinquency
47(4): 610–640.
A partial test of the impact of a casino on neighborhood crime
17© 2014 Macmillan Publishers Ltd. 0955-1662 Security Journal 1–17
A preview of this full-text is provided by Springer Nature.
Content available from Security Journal
This content is subject to copyright. Terms and conditions apply.
