SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature Review
  5. Empirical Analysis of VLTs, Casinos, and Crime Rates
  6. Methodology
  7. Results
  8. Summary and Conclusions
  9. References

Much research examines the positive and negative impacts of gambling in specific areas, including the relationship between gambling, such as casinos and electronic gaming, on crime. Since Grinols and Mustard, the academic literature finds a mixed relationship. The present research examines the relationship between video lottery terminals (VLTs) and casino gambling and crime in the province of Alberta from 1977 to 2008 using data from the Uniform Crime Reporting Survey. Estimates from a two-way fixed effect regression indicate little association between gambling and crime. However, some positive and negative crime-specific effects are found for both casinos and VLTs.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature Review
  5. Empirical Analysis of VLTs, Casinos, and Crime Rates
  6. Methodology
  7. Results
  8. Summary and Conclusions
  9. References

A small, growing literature examines the relationship between access to legal gambling and crime. Much of this evidence comes from the U.S. No consensus exists regarding the nature of this relationship; some studies found a positive association, whereas others found no association. The gambling–crime relationship has important public policy implications because communities across North America have been expanding access to legal gambling to increase the government revenues generated by these activities (Kearney 2005). However, increased crime associated with casinos could increase costs in the jurisdiction to fight the increase in crime (Eadington 1999).

Research on the relationship between access to legal gambling and crime occurs within the larger context of evaluating the overall costs and benefits to society generated by gambling (Kearney 2005). Walker (2007) argued that accounting for legal and justice costs (and benefits) of gambling based on government provision of judicial and police services is problematic. Thus, researchers should examine data from specific jurisdictions using information about individual crimes and not government financial accounts.

The present study examines the relationship between legal gambling, in the form of casinos and video lottery terminals (VLTs) in bars and taverns, in the Canadian Province of Alberta. Alberta represents an interesting setting, as a large number of casinos and VLTs are scattered throughout the province. The numbers of VLTs that are operated throughout the province are limited and have not changed over time. However, VLT locations changed as existing machines moved to different communities, providing a good empirical setting to examine the relationship between access to VLTs and crime. Finally, violent crime is less common in Canada compared to the U.S., where much of the previous research was based. The present research examines the relationship between access to VLTs and casinos over time in communities and nine different types of crime using data from the Canadian Uniform Crime Report (UCR) from 1977 to 2008. The sample period encompasses time both before and after legalized gambling existed in the province. Based on Walker's (2007) observations, the present research focuses only on estimating the statistical relationship between gambling and crime in Alberta, and does not attempt to estimate a monetary value of the benefits and costs of crime.

Results from a reduced form regression model of the determination of the incidence of nine types of crime find little statistical association between access to legal gambling and crime. However, the introduction of VLTs is associated with an increase in credit card fraud and a decrease in prostitution and shoplifting. The introduction of casinos is also associated with an increase in robberies and a decrease in shoplifting. The addition of lags and leads to the empirical model, like in Grinols and Mustard (2006), shows declines in the incidence rates of many of these types of crimes following the opening of a new casino in a community. These results provide new evidence on the relationship between legal gambling and crime, and show that the relationship is not necessarily positive.

Literature Review

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature Review
  5. Empirical Analysis of VLTs, Casinos, and Crime Rates
  6. Methodology
  7. Results
  8. Summary and Conclusions
  9. References

A large body of research exists on the overall costs and benefits generated by increased access to legal gambling, where crime represents an important source of costs. Kearney (2005) documented the increase in access to legal gambling over the past 30 years in North America and demonstrated the importance of assessing the costs and benefits of this increased access. Walker (2003) pointed out the problems inherent in any full cost–benefit accounting of gambling in society. Researchers generally focus on the relationship between legal gambling and specific costs and benefits, including the effect of casinos on property values (Wenz 2007), bankruptcy rates (Koo, Rosentraub, and Horn 2007), individuals' quality of life (Wenz 2008), and fatal alcohol-related traffic accidents (Cotti and Walker 2010).

This research is theoretically motivated by the routine activities and rational choice theories. Rey, Mack, and Koschinsky (2012) stated that these two theories are complementary when explaining crime. In rational choice, “Crime is assumed a priori to involve rational calculation and is viewed essentially as an economic transaction or a question of occupational choice” (Clarke and Cornish 1985: 156). Becker's (1968) seminal model of crime is rooted in rational choice theory. In his model, committing a crime is a utility maximizing choice made by individuals who compare the expected costs and benefits of committing a crime and engage in criminal behavior if the expected benefits outweigh the expected costs. Since Becker's theoretical model, crime has been extensively studied as an economic activity in the economics literature.

Routine activities theory addresses the environment that a crime is being committed (Rey, Mack, and Koschinsky 2012). Cohen and Felson (1979) defined routine activities as “any recurrent and prevalent activities which provide for basic population and individual needs, whatever their biological or cultural origins” (p. 593). Specifically dealing with crime, the theory predicts that a crime is likely to be committed due to the convergence of three factors in one location: a person who is likely to commit a crime, a target, and the absence of protection that would deter a person to commit a crime (Rey, Mack, and Koschinsky 2012). Routine activities theory is especially useful when examining casinos and access to legal gambling in the broader context of economic development. Policy makers in communities may use gambling as a way to attract tourists to an area, increasing the opportunities for crime (Giacopassi and Stitt 1993). Both rational choice and routine activities theories explain how the presence of legal gambling can affect crime and why the impact may differ across types of legal gambling and categories of crime. In Alberta, VLTs exist in bars and taverns located near other businesses and residential areas in small and large communities so many VLT gamblers may arrive on foot and gamblers may be difficult to distinguish from other patrons; casinos tend to be relatively large venues located away from other businesses; many gamblers drive and park at the facility and few non-gamblers are present. In terms of rational choice theory, VLT gamblers may have a different expected benefit than casino gamblers for criminals, as they can be difficult to distinguish from non-gambling bar and tavern patrons and may be on foot, reducing the expected benefit of some crimes such as robbery in contrast to casinos where gamblers arrive by car and park in large underground parking lots potentially carrying large amounts of cash.

In terms of routine activities theory, casino gamblers congregate in large numbers at casinos which provides an increase in the number of likely targets which may attract individuals more likely to commit certain crimes such as robbery but not others such as drug use or shoplifting. Rational choice theory also predicts that different crimes may have different expected costs, especially in light of the locational differences for VLTs and casinos. VLTs located in high-traffic areas may increase the likelihood of detection of a crime, increasing the total costs of commission, relative to casinos with large, often underground, parking lots. VLTs in bars and taverns may be located near other commercial establishments, increasing the ability to commit crimes such as shoplifting. Both theories predict that the relationship between different forms of legal gambling and different types of crime should not be uniform; VLT gambling might be expected to increase shoplifting, for example, if many bars with VLTs are located near convenience stores; casino gambling might be expected to increase robbery if the presence of large number of casino gamblers carrying large sums of cash attracted more individuals intent on committing robbery.

Gambling and crime

Typically, casinos are chosen to examine the relationship between legal gambling and crime. This relationship has considerable interest for many stakeholders because the initial development of casino gambling in the U.S. was linked to organized crime and crime continues to be a visible measure of the quality of life in a community. Early research by Giacopassi and Stitt (1993) examined the impact of casino gambling on crime in Biloxi, Mississippi, and found that larceny and motor vehicle theft were the only categories of crime associated with the introduction of casinos. However, no significant effect was found for violent crimes.

Grinols and Mustard (2006) performed an extensive statistical analysis of the relationship between casinos and crime in the U.S. They concluded that approximately 8 percent of the crimes occurring in the U.S. counties with casinos were attributable to the presence of casinos. The crimes affected by casino openings included several types of violent crimes (aggravated assault, robbery, rape), burglary, and auto theft. The crime rates increased 3 to 5 years after the opening of a casino in a county.

Grinols and Mustard's (2006) results attracted considerable attention. Walker (2008a,b) criticized Grinols and Mustard on methodological grounds, whereas Reece (2010) raised additional issues regarding the methodology and cast doubt on the conclusion that casinos cause crime to increase. He found that the presence of hotels near casinos, an omitted variable from the Grinols and Mustard study, made inviting targets for criminals committing crimes such as robbery, car theft, and prostitution. Reece (2010) also examined the temporal lag between the opening of a casino and the increase in crime, and showed that hotels were built and opened several years after casinos opened in communities, providing an alternative to Grinols and Mustard's explanation for the temporal lag.

Clark and Walker (2009) found a positive and significant relationship between the amount young adults lost while gambling and the likelihood of committing a crime. Wheeler, Round, and Wilson (2011) examined the relationship between expenditure on electronic gaming and crime in the state of Victoria in Australia. Their research grouped crimes into income-generating crimes (such as robbery, fraud, and theft) and non–income-generating crimes (defined as all other crimes committed). Increased spending on electronic gaming was associated with increased rates of income-generating and non–income-generating crimes in 1996, 2001, and 2006. Hyclak (2011) examined the relationship between the presence of casinos and crimes committed on nearby college campuses in four Midwestern U.S. states. Car thefts increased on campuses close to casinos, but robberies and burglaries did not.

Overall, the strong positive relationship between casinos and crime reported by Grinols and Mustard (2006) has not been found in other settings while this body of research casts considerable doubt on their results. Little attention has been paid to the relationship between other forms of legal gambling, such as electronic gaming, and crime. The lack of consensus about the relationship between legal gambling and crime in the literature suggests that additional research in other settings needs to be conducted. The present research focuses on the province of Alberta.

Legal gambling in Alberta

In 1980, Alberta's first private casino opened in Calgary. The Criminal Code of Canada was modified in 1985 to transfer legal gambling authority from federal control to the provinces, in addition to legalizing VLT and slot machine gambling. Seven years after the change, VLTs were introduced in Alberta. The province initially authorized 6,000 VLTs, and the number authorized has remained constant, although the location of VLTs, and the precise number in operation, has changed over time. In 2011, 5,694 VLTS were installed and operated at 1,030 bars and taverns. Although successful as a form of entertainment and an important source of gambling revenue, many communities viewed VLTs with hostility. In response to plebiscites, and despite the objections of retailers, VLTs were voted to be “removed” from seven communities in Alberta by its citizens. For example, in Fort McMurray, VLTs were eliminated from bars and lounges due to increased concerns related to problem gambling, and concentrated in the casino. They effectively remained in the community, but as slot machines (permitted in casinos since 1996). The most recent change in Alberta's gambling environment was the construction of First Nations casinos on tribal reserve land. Currently, five First Nations casinos exist in Alberta, all of which comply with the Alberta Gaming and Liquor Commission (AGLC) charitable gaming model, but with provisions that allow flexibility to provide additional financial resources in terms of revenues and jobs to addressing issues within First Nations' communities.

A wide array of legal gambling activities are available in Alberta, including casino gambling at both private charity casinos and casinos on First Nations reserve land with table games and slot machines, horse race gambling and slot machines at “racinos,” VLTs in bars and pubs, and other traditional legal gambling activities such as lottery and bingo. We focus on VLTs and casinos because of the variation in the availability of these forms of legal gambling over the period for which crime data exist; bingo, lottery outlets, and horse racing venues did not change much over the sample period, making them less interesting from an empirical perspective. Table 1 summarizes the availability of casino, racino, and slot machine gambling in the province in 2008. Note that the casinos in Kananaskis and Whitecourt opened in 2008 on First Nations reserve lands after the sample period ends, so these casinos are not included in the analysis.

Table 1. Casinos and Racinos in Alberta for 2008
CommunityCasinosRacinosSlot machines
  1. Source: Various websites, Alberta Gaming and Liquor Commission Website and Annual Reports.

Calgary74,591
Camrose1200
Cold Lake1150
Edmonton614,296
Fort McMurray1399
Grande Prairie11491
Kananaskis1300
Lethbridge11398
Medicine Hat1230
Red Deer2598
St. Albert1240
Whitecourt1250
Total24312,143

The AGLC has the responsibility for regulating gaming and managing the distribution of gambling revenues in Alberta. The AGLC also regulates the operation of casinos and the location and operation of VLTs. Under its oversight, a charitable gaming model was established to ensure that the proceeds and benefits of legalized gambling were returned to the citizens of Alberta. The charitable gaming model in Alberta allows charity and non-profit organizations to benefit from legal gambling in Alberta. Charitable gambling revenues are distributed to community organizations through two mechanisms: the Alberta Lottery Fund and event licenses granted to individual organizations for casino, bingo, and instant win gambling activities. This model could have an impact on the relationship between legal gambling and crime since gambling revenues explicitly benefit good causes in the province and could affect the expected costs and benefits of crime. Since VLT and casino location are tightly regulated, and the number of VLTs in the province has been constant, this can also mitigate potential correlation between the presence of legal gambling in a community and unobservable factors affecting crime rates as provincial regulators attempt to maximize gambling revenues generated for charitable causes.

Charitable gambling revenues generated by ticket lotteries, slot machines, keno, and VLTs are distributed to community groups through the Alberta Lottery Fund. In 2007–2008, $1.62 billion in gambling revenues were transferred to the Alberta Lottery Fund according to the AGLC Annual Report. This money is then paid out in the form of grants to community organizations—an estimated $465 million during 2007–2008. Access to charitable gaming revenues requires that a charitable, religious, or non-profit group apply for a license and provide volunteers from their membership to contribute labor for operating charity casinos.

Researchers from different backgrounds have analyzed the impacts of legalized gambling on communities. One common impact studied is the relationship between legalized gambling and crime. The results from previous research have been mixed, increasing the importance of studying the relationship in additional jurisdictions. The present research examines the province of Alberta due to the importance of gambling revenues within the province for both the government and the charitable organizations but also the wide availability for the citizens throughout the large province. Based on the history of the legalization of gambling in Canada and the change in regulation policy over the past decades, examining the relationship before and after the regulation change provides additional information to researchers and public policy makers.

Empirical Analysis of VLTs, Casinos, and Crime Rates

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature Review
  5. Empirical Analysis of VLTs, Casinos, and Crime Rates
  6. Methodology
  7. Results
  8. Summary and Conclusions
  9. References

Following the general practice in the literature, the present research estimates reduced form regression models that explain the determination of crimes committed in specific areas in Alberta over time. These models include variables that reflect access to legal gambling opportunities, in this case VLTs and casinos, in each area. Alberta has a relatively large number of casinos and horse racing tracks with slot machines, 27 by the end of the sample, and almost 6,000 VLTs located in bars and taverns throughout the province. Casinos and racinos opened over the last 30 years, and the number and location of VLTs has changed, providing variation in the availability of these forms of legal gambling across areas and over time.

Crime in Alberta

The crime data source used, the UCR, comes from Statistics Canada in association with the Canadian Centre for Justice Statistics (CCJS). The UCR is an annual census based on all incidents of crime reported to the policing community in jurisdictions in Canada. Detailed historical crime data are available for a large number of communities in Alberta in the UCR. The CCJS cooperates with various police agencies to collect crime statistics through the UCR. The data reflect reported crimes substantiated by police investigation. The crime data in the UCR are available annually over the period 1977–2008, providing an opportunity to examine crime before and after the casinos and VLTs were legally established in the province.

The present research uses data from the UCR1 survey, an aggregate survey for reporting incidents across jurisdictions throughout Canada. An “incident” is the basis for counting a reported crime and is defined as a set of connected crime-related events usually constituting an occurrence report and represents a single event that may include multiple crimes. To avoid double counting of crimes, an incident that involves the commission of multiple crimes is identified only once in the UCR, based on the most serious offence committed. Under the most serious offense procedure, an incident containing a non-violent crime, say breaking and entering, and a violent crime, say assault, will be recorded only as an assault; this procedure can result in undercounts of crimes related to gambling. The UCR data contain information on the number of criminal incidents, incidence rates (per 100,000 population), and the clearance rate—percent of incidents where a suspect was identified—of those incidents from 1977 to 2007.

The geographic unit of observation in the UCR to analyze the relationship between gambling and crime in Alberta is a Royal Canadian Mounted Police (RCMP) Division or a metropolitan police force jurisdiction. After eliminating observations from RCMP Divisions and municipal police forces with missing data, the sample included crime data from 78 communities in Alberta. Table 2 lists the communities and the number of observations from each community in the sample. The communities that have one or more casinos, and the eight communities that do not have VLTs are identified in Table 2. Note that the present research only analyzes data from the urban part of the police jurisdictions in the UCR. The UCR contains data on incidents from the rural part of RMCP Divisions in Alberta beginning in 2002, but does not contain information on the population of these rural areas. Since the literature clearly identifies population as an important determinant of crime (the larger the population in an area, the more potential criminals and victims), a thorough statistical analysis cannot be performed. While this limitation is recognized to limit the results somewhat, if the relationship between gambling and crime in rural areas is similar to the relationship in urban areas, then the results can be generalized to the entire province. In any event, the vast majority of crime incidents take place in the urban part of RCMP Divisions.

Table 2. Alberta Communities in the Sample
CommunityObservations% sample
  1. a

     No video lottery terminals in community.

  2. b

     Casino in community.

  3. Source: Uniform Crime Reporting Survey.

Airdrie311.45
Athabasca271.26
Banff140.65
Barrhead291.35
Beaumont140.65
Blairmorea30.14
Bonnyville311.45
Brooks311.45
Calgaryb311.45
Camroseb311.45
Canmore311.45
Cardstona281.31
Chestermere30.14
Claresholm281.31
Coaldale271.26
Cochrane311.45
Cold Lakeb311.45
Colemana20.09
Crowsnest Passa271.26
Devon311.45
Didsbury281.31
Drayton Valley311.45
Drumheller311.45
Edmontonb311.45
Edson311.45
Fairview281.31
Fort Macleod281.31
Fort McMurrayb311.45
Fort Saskatchewan311.45
Fox Creek281.31
Grande Cache281.31
Grande Prairieb311.45
Grimshaw281.31
Hanna281.31
High Level281.31
High Prairie281.31
High River311.45
Hinton321.49
Hobbemaa30.14
Innisfail311.45
Lac La Biche281.31
Lacombe311.45
Leduc311.45
Lesser Slave Lakea40.19
Lethbridgeb311.45
Louis Bulla210.98
Medicine Hatb311.45
Morinville311.45
Okotoks311.45
Olds311.45
Peace River311.45
Pincher Creek281.31
Ponoka311.45
Raymonda281.31
Red Deerb311.45
Redcliffa160.75
Redwater281.31
Rimbey291.35
Rocky Mt. House311.45
Sherwood Park311.45
Slave Lake311.45
Spruce Grove311.45
St. Albertb311.45
St. Paul311.45
Stettler311.45
Stony Plain311.45
Strathmore311.45
Swan Hills281.31
Sylvan Lake311.45
Taber311.45
Three Hills281.31
Valleyview271.26
Vegreville311.45
Vermilion281.31
Wainwright311.45
Westlock281.31
Wetaskiwin311.45
Whitecourtb311.45

Although Grinols and Mustard (2006) found that casino openings were related to increases in violent crimes in the U.S., violent crime rates in Canada are substantially lower, and research by Smith and Wynne (1999) reported that most of the crimes associated with gambling in Western Canada are non-violent in nature. The availability of legal gambling could affect the propensity to commit any crime to the extent that legal gambling changes either the total expected cost or benefit of committing a crime. According to Grinols and Mustard, local labor market conditions can change after the opening of a casino in a community, affecting the relative return to crime compared to work. Humphreys and Marchand (2012) developed evidence that opening new casinos had positive effects on local labor markets in Canada over this period. Since different crimes have different expected returns, labor market effects could have a differential impact on rate of commission of crimes. The focus of this analysis is on nine specific non-violent crimes identified in the UCR data: breaking and entering, credit card fraud, drug possession, illegal gambling, other fraud, prostitution, robbery, shoplifting over $5,000, and shoplifting under $5,000. In Canada, frauds are “Every one who, by deceit, falsehood or other fraudulent means, whether or not it is a false pretence within the meaning of this Act, defrauds the public or any person, whether ascertained or not, of any property, money or valuable security or any service.” The “other fraud” category includes writing bad checks, selling goods or services that were never produced, or other deceitful actions. This list of crimes includes many crimes associated with gambling in Western Canada identified by Smith and Wynne and other less serious crimes that could be plausibly linked to legal gambling.

Figure 1 shows the annual incidence rate per 100,000 persons for these nine crimes in Alberta over the period 1977–2008. The red vertical lines identify three important gambling-related events: the opening of casinos (1980), the introduction of VLTs (1992), and the introduction of slot machines in casinos (1996).

figure

Figure 1. Crime Incidence Rates in Alberta 1977–2008.

Source: Authors calculations based on data from the Uniform Crime Reporting Survey.

Download figure to PowerPoint

Figure 1 provides a general picture of the incidence rates and trends in crimes committed in Alberta. These are incidence rates so they indicate the number of crimes of each type reported to the police and determined to be genuine. Some of these crimes, including breaking and entering, drug possession, illegal gambling, and shoplifting, show clear downward trends in the incidence rates over the sample period. Others, including credit card fraud and robbery, show upward trends. There is also quite a bit of year-to-year variation in the sample period.

Detailed data on the number of VLTs in each UCR community in Alberta, as well as the dates of the opening of all casinos, were obtained from the AGLC. These data annually identify the number of VLT machines and casinos in each community. Eight communities in the sample (Blairmore, Cardston, Coleman, Crowsnest Pass, Hobbema, Lesser Slave Lake, Louis Bull, Raymond, and Redcliff) had no VLTs. One or more casinos exist in 10 communities. Since the crime data begin before the introduction of VLTs and casinos in the province, this analysis constitutes a “before and after” statistical analysis similar to Grinols and Mustard (2006). The results from the present research reflect conditional crime rate estimates before and after the introduction of VLTs and casinos, between communities with and without VLTs and casinos, and within communities with VLTs and casinos, since the number of VLTs and casinos in communities varies over the sample period.

The focus here is on the statistical relationship between crime and VLTs and casinos. While many other types of legal gambling exist, including lottery, sports betting, bingo, and horse racing, casino gambling and VLTs generally receive most of attention within the literature on gambling and crime. Casino gambling and VLTs, unlike bingo and horse racing, has increased in popularity in recent years in Alberta, and policy makers face pressure to increase the number of venues and machines in the province.

Table 3 shows the sample means for the nine crime incidence rates per 100,000 in population and the other key variables in the analysis. Illegal gambling is relatively rare in the province. Keep in mind that these are incidence rates, not the rate of commission of crimes. Every crime committed is not reported to authorities, so incident rates understate the rate at which crimes are committed.

Table 3. Sample Means, Crime Incidence Rates, and Other Variables, 1977–2008
VariableMean
  1. Source: Authors Calculations based on data from the Uniform Crime Reporting Survey.

  2. VLTs, video lottery terminals.

Breaking and entering incidence rate per 100,000 population1,095
Credit card fraud incidence rate per 100,000 population54
Drug possession incidence rate per 100,000 population397
Illegal gambling incidence rate per 100,000 population2
Other fraud incidence rate per 100,000 population295
Prostitution incidence rate per 100,000 population3
Robbery incidence rate per 100,000 population31
Shoplifting over $5,000 incidence rate per 100,000 population16
Shoplifting under $5,000 incidence rate per 100,000 population390
Community population29,214
Unemployment rate7.0
No. of VLTs in community31

The introduction of VLTs and the changes in the number of VLTs in communities over time provide variation in the access to VLT gambling in each community and variation within each community over time. The regression model exploits this variation to quantify the relationship between crime and legal gambling, conditional on other observable and unobservable factors that affect crime.

Methodology

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature Review
  5. Empirical Analysis of VLTs, Casinos, and Crime Rates
  6. Methodology
  7. Results
  8. Summary and Conclusions
  9. References

The present study analyzes the relationship between the opening of casinos in the province and the introduction of VLTs in bars and taverns and historical crime incidence rates in each of the 78 communities in the sample over the period 1977–2008 using reduced form regression models. This regression approach explains observed variation in crime incidence rates across communities and over time with observed variation in other factors that have been shown to affect crime rates in the literature, including economic factors such as the unemployment rate, demographic factors such as the population, and variation in gambling opportunities across communities and over time. The regression models control for unobservable heterogeneity in the communities and years in the sample, as well as the effect of factors known to affect crime such as the unemployment rate in the province and the population of each community, and province wide trends in crime.

A panel data model is estimated to explain observed variation in crime. The dependent variable (Yit) is the crime incidence rate per 100,000 population for the nine crimes identified above in the 78 UCR communities. The dependent variables vary both across the i = 1, 2, 3, …, N (= 78) communities over t = 1, 2, 3, …, T (= 31) years. Formally, the regression models estimated take the form:

  • display math(1)

and contain both a vector of explanatory variables that vary over cross-sectional units and time, EXit, and a gambling-related explanatory variable, Git, that also varies over the cross-sectional units and time. αi, αt, and γ are vectors of unknown parameters to be estimated. αi is a vector of cross-sectional unit specific intercepts that capture unobservable heterogeneity in the communities. αt is a vector of time-period specific intercepts that capture unobservable heterogeneity in each year in the sample. This could include the business cycle in the province, the regulatory environment, the effects of demographic changes, or other time varying effects. Estimates of αi and αt are not reported. β, the key unknown parameter of interest, captures the relationship between the gaming-related explanatory variable, Git, and the outcome variable.

eit is an unobservable equation error term that captures the effects of all other unobservable factors that affect the outcome variable of interest. By assumption, eit is a mean zero independent and identically distributed random variable with constant variance σe that is uncorrelated with αi, αt, Git, and EXit. Under this assumption, ordinary least squares (OLS) applied to equation (1) produces unbiased, efficient, and consistent estimates of the population parameters of interest. We use the White-Huber “sandwich” correction for heteroscedasticity, in the form of different σe's for different communities and also cluster correct the estimated standard errors at the community level, to account for potential within-community correlation of the unobservable equation error term (Greene 2000).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature Review
  5. Empirical Analysis of VLTs, Casinos, and Crime Rates
  6. Methodology
  7. Results
  8. Summary and Conclusions
  9. References

Tables 4 and 5 contain the parameter estimates, p-values, and other summary statistics from equation (1) using data on the number of casinos in each community as the gambling-related explanatory variable of interest. Recall that only 10 of the UCR communities (Calgary, Camrose, Cold Lake, Edmonton, Fort McMurray, Grand Prairie, Lethbridge, and Medicine Hat) had casinos present in one or more years in the sample period. This is a relatively simplistic measure of legal gambling opportunities, since it treats all casinos as equal in terms of their potential to affect crime rates. Since casino handle and patronage varies, a better approach would use a measure of the gambling activity that takes place inside casinos instead of a variable containing the number of casinos in each community. However, the data to perform this type of analysis are not available during the sample period. These results represent a local effect of casinos on crime because it assumes that the effect of the casino on crime does not extend beyond the nearby community.

Table 4. Two-way Fixed Effects Regression Results: Casinos and Crime I
Model12345
Variable/crimeB&ECredit card fraudDrugsIllegal gamblingOther fraud
  1. n = 2,166; p-values shown in parentheses below parameter estimates.

  2. Source: Authors Calculations based on data from the Uniform Crime Reporting Survey.

Population−6.269−0.069−3.111−0.059−2.994
(0.264)(0.855)(0.147)(0.180)(0.032)
Population20.00314−0.000005570.001730.00003770.00187
(0.322)(0.980)(0.156)(0.160)(0.027)
Unemployment rate−207−19.936.697−3.049−94.82
(0.092)(<0.001)(0.936)(0.020)(0.265)
Casino24.0619.288.020.3414.21
(0.741)(0.084)(0.767)(0.432)(0.459)
Time trend200.621.2−24.42.98102.3
(0.090)(<0.001)(0.760)(0.016)(0.212)
R20.6430.1990.4940.0960.206
Table 5. Two-way Fixed Effects Regression Results: Casinos and Crime II
Model1234
Variable/crimeProstitutionRobberyShoplifting < 5kShoplifting > 5k
  1. n = 2,166; p-values shown in parentheses below parameter estimates.

  2. Source: Authors calculations based on data from the Uniform Crime Reporting Survey.

Population0.5090.176−0.3130.00987
(0.002)(0.436)(0.058)(0.996)
Population2−0.0004−0.00020.00020.0004
(0.005)(0.229)(0.046)(0.734)
Unemployment rate−1.966−8.01423.5524.15
(<0.001)(0.301)(0.091)(0.741)
Casino5.4739.554−2.893−65.63
(0.078)(0.019)(0.105)(0.029)
Time trend1.5067.745−18.67−15.9
(0.002)(0.294)(0.159)(0.823)
R20.5730.6760.2360.541

From Tables 4 and 5, equation (1) does a somewhat mixed job in explaining the variation in the incidence of these nine crimes in communities in Alberta. Population is not related to the incidence rate of most of these crimes. Prostitution, credit card fraud, and illegal gambling respond to business cycle conditions, as captured by the provincial unemployment rate, but other crimes do not. Grinols and Mustard (2006) discussed how the commission of different crimes could respond differently to changes in the economic and legal environment. These models explain between 10 percent, for illegal gambling, and 68 percent, for robbery, of the observed variation in crime incidence across communities and over time. Recall that equation (1) also contains a vector of community fixed effects and a vector of year fixed effects to control for unobservable heterogeneity. These parameter estimates are not reported.

From Tables 4 and 5, the relationship between the presence of casinos in a community and crime in those communities is weak. The annual number of robberies committed in each year increases by 9.5 per 100,000 in population, and the number of shoplifting cases under $5,000 declines by about 65 in each year following the opening of a casino. The annual incidence of other crimes has no statistical association with the presence of casinos in the community. The positive relationship between casinos and robberies is consistent with Grinols and Mustard (2006), although the size of the effect is considerably smaller. Grinols and Mustard reported coefficient estimates on the order of about 20–80 for robberies.

It is important to keep in mind that these estimates are not causal, so the results do not mean that opening casinos caused the incidence rate of shoplifting to fall in communities; it simply means that the presence of casinos was statistically associated with lower incidence rates for shoplifting. Still, the finding that some type of crime falls after the opening of a casino is interesting, in that it suggests that opening a casino is not associated with increases in all forms of crime. One explanation for the negative relationship between casinos and shoplifting is that shoplifting has a “thrill” component and, following the introduction of casinos, individuals who would have satisfied this desire for a “thrill” by shoplifting instead satisfy it by gambling in casinos.

Another explanation for the observed negative statistical relationship between casinos and crime is that the regression model is mis-specified. There may be important variables that were omitted that affects the incident rate of crime and also happens to be correlated with the presence of casinos. For example, Reece (2010) showed the importance of hotels nearby casinos in explaining the relationship between casinos and crime. Data on the presence of hotels at the casinos in the sample were not available for the sample period. Alternatively, the regression model may fail to adequately account for the clear downward secular trend in shoplifting in Alberta after 1990 visible in Figure 1, despite the presence of a time trend variable in equation (1).

The present research also examines statistical association between the number of VLTs in a community and crime. VLT gambling differs from casino gambling in important ways, so the relationship between VLT gambling and crime can be expected to differ from the casino–crime relationship. Tables 6 and 7 show parameter estimates, p-values, and other regression summary statistics from estimation of equation (1) when the gambling variable was defined as the number of VLTs in a community. The significance of the estimated parameters on the population, unemployment, and time trend variables were identical to those in Tables 4 and 5.

Table 6. Two-Way Fixed Effects Regression Results: Video Lottery Terminals and Crime I
Model12345
Variable/crimeB&ECredit card fraudDrugsIllegal gamblingOther fraud
  1. n = 2,143; p-values shown in parentheses below parameter estimates.

  2. Source: Authors Calculations based on data from the Uniform Crime Reporting Survey.

Population−3.8180.399−2.563−0.049−2.742
(0.323)(0.392)(0.106)(0.200)(0.023)
Population20.00239−0.000280.001510.00003320.00166
(0.321)(0.314)(0.119)(0.172)(0.035)
Unemployment rate−318.5−20.28−17.6−2.105−120.7
(0.014)(<0.001)(0.832)(0.268)(0.041)
Number of VLTs−0.2530.057−0.02640.000440.0662
(0.232)(0.038)(0.583)(0.650)(0.250)
Time trend30721.41−1.1892.065127.2
(0.012)(<0.001)(0.988)(0.268)(0.041)
R20.6280.1970.4920.0950.204
Table 7. Two-Way Fixed Effects Regression Results: Video Lottery Terminals and Crime II
Model1234
Variable/crimeProstitutionRobberyShoplifting < 5kShoplifting > 5k
  1. n = 2,143; p-values shown in parentheses below parameter estimates.

  2. Source: Authors Calculations based on data from the Uniform Crime Reporting Survey.

Population0.8560.606−0.309−1.562
(0.003)(0.005)(0.012)(0.291)
Population2−0.00046−0.000330.000220.00124
(0.014)(0.028)(0.007)(0.186)
Unemployment rate−2.46−10.509.3239.55
(0.001)(0.093)(0.132)(0.683)
Number of VLTs−0.030−0.015−0.020−0.169
(0.022)(0.221)(0.005)(0.001)
Time trend1.7729.924−4.935−30.23
(0.001)(0.094)(0.410)(0.683)
R20.5890.4890.2360.538

The regression results in Tables 6 and 7 indicate little relationship between VLTs and crime in communities in Alberta since the introduction of VLTs in the early 1990s. There is no statistical association between the number of VLTs in communities and breaking and entering, drug possession, illegal gambling, fraud, and robbery. Credit card fraud is slightly higher in communities with VLTs, but the effect was small, since 100 additional VLTs were associated with an increase of about 5 additional incidences of credit card frauds per 100,000 population.

Interestingly, the association between the presence of VLTs and prostitution and the two types of shoplifting was negative. The incident rate of both types of shoplifting tended to decline after VLTs were introduced in communities, although the effects were small. In the case of shoplifting under $5,000, there were about two fewer incidences of these crimes per year for each additional 100 VLTs in a community. Again, these estimates are not causal, so the results do not mean that introducing VLTs caused the incidence rate of these crimes to fall in communities; it simply means that VLTs were statistically associated with lower crime. One explanation for the observed relationship between VLTs and prostitution is that some individuals have a “vice” budget that they spend on illicit behavior. When VLTs are introduced to a community, some individuals with preferences for illicit behavior may substitute VLT gambling for dealings with prostitutes. One explanation for the relationship between VLTs and shoplifting is that shoplifting has a “thrill” component in the total utility generated by committing this crime, and after the introduction of VLTs individuals who would have satisfied this desire for a “thrill” by shoplifting instead satisfy it by playing VLTs.

Another explanation for the observed negative statistical relationship between VLTs and crime is that the regression model is mis-specified. There could be some important variables that were omitted that affect the incident rate of crime and happen to be correlated with VLTs. Alternatively, the regression model may fail to adequately account for the clear downward secular trend in prostitution and shoplifting in Alberta after 1990 visible in Figure 1. The regression model contains a time trend variable that should capture this downward trend, but the trend in the model is linear and the actual trend could be non-linear. Alternatively, VLTs may have been placed in communities that experienced relatively large declines in prostitution and shoplifting by chance.

Robustness checks

For the results reported in Tables 4-7, several robustness checks were performed. Both Grinols and Mustard (2006) and Reece (2010) emphasized the importance of temporal leads and lags when estimating the relationship between access to legal gambling and crime. The inclusion of lags and leads variables reflecting access to legal gambling in an area is related to the potential for endogeneity in the observed relationship between access to legal gambling and crime. Walker (2008b) and Grinols and Mustard (2008) discussed the issue of endogenous siting of casinos in detail. The idea is straightforward: The error term in equation (1) captures all factors that affect crime in an area other than the explanatory variables on the right hand side of equation (1). There could be factors captured in this error term that are observable by decision makers who determine the location of casinos and other forms of legal gambling that affect the decision to put a casino in a specific area. The idea that casinos may be intentionally located in high-crime areas is one example of this problem; Grinols and Mustard addressed this point. If a specific area can “self-select” into the category of areas with a casino, and the self-selection depends on unobservable factors captured in the error term in equation (1), then OLS parameter estimates from equation (1) may suffer from econometric problems. Grinols and Mustard investigated this possibility by including leads and lags of the indicator variable for the presence of a casino in a county in their econometric model.

In Alberta, the regulation of casinos and the charity gambling model used in the province mitigate the potential that casinos are intentionally located in high-crime areas. Under the charity gambling model, casinos are viewed as an important source of funds for local charitable organizations, and these organizations are granted casino “licenses” that allows them to keep the profits generated from casino gambling (net of operation costs) on one or two specific dates in exchange for providing the casino with unskilled labor on those dates. Since the provincial casino regulators want to maximize the funds going to charitable organizations from a casino, and recognize that members of these charitable organizations will go to the casino to work, it is unlikely that these casinos would be intentionally placed in high-crime areas.

The province also regulates the location and quantity of VLTs in communities. The number of VLTs in communities changed over time as the AGLC shifted VLTs around in the province and some communities eliminated them. Conversations with AGLC staff indicated that changes in VLT location in the province reflect changes in the demand for VLT gambling in communities, and not community pressure to remove them. All of the observed reallocation of VLTs in the sample was from low- to high-revenue generating locations. It is unlikely that VLT location has any relationship to crime rates in the surrounding community.

However, casino location may still be influenced by unobservable factors captured in the equation error term in equation (1), and the impact of casinos on crime in an area may happen over time. To explore these factors, we added leads and lags of the variable indicating presence of a casino in a community, Git, to equation (1), like the specification used by Grinols and Mustard (2006) and re-estimated the regression model. Table 8 contains the parameter estimates and p-values on the casino indicator variables and other summary statistics for eight of the categories of crime analyzed here.1

Table 8. Regression Results—Leads and Lags
VariableBreaking/enteringCredit card fraudDrug possessionOther fraud
Parameter estimatep-valueParameter estimatep-valueParameter estimatep-valueParameter estimatep-value
  1. Source: Authors calculations based on data from the Uniform Crime Reporting Survey.

Lead 1 −40.70.3124.6390.6288.2420.77219.050.670
Lead 277.80.16917.180.0623.4290.9204.1010.889
Open562.30.16128.480.018226.450.080426.120.002
Lag 1−905.2<0.001−31.89<0.001−236.51<0.001−521.93<0.001
Lag 2−48.80.859−35.650.002−13.080.647−160.030.417
N/R21,8460.6461,8460.1921,8460.4991,8460.206
 ProstitutionRobberyShoplifting over $5kShoplifting under $5k
Estimatep-valueEstimatep-valueEstimatep-valueEstimatep-value
Lead 1−8.9410.008−1.4370.785−6.9040.024−32.350.105
Lead 210.080.0198.9460.0031.3220.605−25.370.339
Open0.0290.97627.57<0.001−46.240.054449.18<0.001
Lag 1−2.0710.015−17.79<0.001−15.160.023−381.18<0.001
Lag 2−1.8030.006−17.270.03781.280.029−124.290.241
N/R21,8460.5931,8460.4691,8460.2521,8460.532
Year effectsYesYesYesYesYesYesYesYes
Area effectsYesYesYesYesYesYesYesYes
Cluster correctionYesYesYesYesYesYesYesYes

From Table 8, very few of the parameter estimates on the leads are statistically significant; casinos do not appear to have been located in areas where crime was high in the years leading up to the opening of these casinos. The contemporaneous relationship between the opening of a new casino and the eight types of crime are similar to those reported in Tables 4 and 5; most are not statistically different from zero, some are positive and one is negative, indicating a mixed relationship. Interestingly, the estimated parameter on the first lag variable is negative and statistically different from zero in all cases. The incidence rate of all these types of crime declined in the year following the opening of a casino in communities in Alberta. These results differ markedly from those reported in Grinols and Mustard (2006), who found increases in crime rates after the opening of casinos in the U.S. counties. The second lags are generally not statistically different from zero, and those that are significantly different from zero are generally negative.

One interpretation of these results is that the presence of a new casino either decreases the expected benefits of committing crimes or increases the expected costs, reducing the incentive to commit crimes in the community. Grinols and Mustard (2006) discussed improvements in the local labor market, in terms of increased prospects for employment, as one mechanism through which this might happen. Humphreys and Marchand (2012) showed that new casinos in Canada had positive effects in local labor markets. However, these negative and significant parameter estimates may still reflect econometric problems. Adding leads and lags will not correct for endogeneity. Instrumental variables is the appropriate econometric correction for this problem. Unfortunately, we lack an instrument to identify the opening of a new casino. Also, the results reported in Tables 4-8 are not causal; they simply reflect correlation between expanding opportunities for legal gambling and crime in a community.

Since casinos draw patrons from all over the province, it is possible that the relationship between casinos and crime extends outside the immediate vicinity of casinos. To investigate this relationship, the present research aggregated the crime and population data to a larger area, the Census Division. Aggregating to Census Divisions reduced the number of cross-sectional units in the sample to 19. The results using more aggregated spatial areas showed no relationship between variation in crime incidence rates and the presence of a casino in Census Divisions in Alberta, suggesting that the association between casinos and crime was localized.

Summary and Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature Review
  5. Empirical Analysis of VLTs, Casinos, and Crime Rates
  6. Methodology
  7. Results
  8. Summary and Conclusions
  9. References

The present research used data from the Statistics Canada Uniform Crime Reporting Survey (UCRS) to analyze the relationship between access to legal gambling and crime in Alberta. The relationship between incidence rates for nine types of crimes in 78 Alberta communities and the opening of casinos in Alberta and the introduction of VLTs in bars and lounges was examined and provided the following conclusions about the impact of gambling on crime.

The results indicated little relationship between the presence of VLTs and crime since the introduction of VLTs in the early 1990s. There was no statistical association between the number of VLTs in communities and breaking and entering, drug possession, illegal gambling, fraud, and robbery. Credit card fraud was slightly higher in communities with VLTs, but the effect was small, since 100 additional VLTs were associated with an increase of about 5 incidences of credit card fraud per 100,000 population per year.

The association between the presence of VLTs and prostitution and shoplifting was negative. The incidence rate of these crimes tended to decline slightly after VLTs were introduced. These results are similar to what one might hypothesize using the routine activities theory as these crimes generally occur in areas of large economic activity. Bars and restaurants that have VLTs may not be located in areas where much economic activity occurs. Again, these estimates are not causal, so the results do not mean introducing VLTs caused the incidence rate of these crimes to fall in communities.

There was no association between the opening of casinos in communities and local crime incidence for breaking/entering, credit card fraud, drug possession, illegal gambling, and prostitution. There is a slight increase in the local incidence of robbery and, in contrast, a slight decrease in shoplifting over $5,000.

In general, the results here suggest a weaker association between access to legal gambling, in the form of VLTs and casinos, and crime in Alberta. Unlike previous research, the results here suggest a negative association between legal gambling and some types of crime. One reason for this difference may be the charity gambling model used in Alberta. Gamblers know that a portion of their losses at casinos, and in VLTs, will be returned to charitable organizations in their community, which could have an effect on the propensity to commit crimes. In the case of casino gambling, criminals may be less likely to prey on people at charity casinos since they are effectively making a donation to charitable organizations in the province. In the context of Becker's (1968) model of crime, potential criminals might perceive the expected utility from committing crimes against charity casino patrons as lower, or the expected costs as higher.

The negative association between the presence of VLTs and the “victimless” crimes of prostitution and shoplifting highlights the importance of the entertainment aspect of VLT gambling. Many of the communities in this sample are small rural towns or villages with limited entertainment options compared with larger cities. In such communities, VLTs may represent an important entertainment option for those interested in entertainment activities with an important “thrill” component. In this sense, VLT gambling could be a substitute for activities such as minor shoplifting.

In any event, the results here indicate that the association between access to legal gambling and crime does not have to be positive. For some types of crime, the association can be negative. This may be due to the charity gambling model used in Alberta, or it could reflect econometric issues. Additional research using techniques, like instrumental variables, that correct for confounding factors, like reverse causality, will help to assess the robustness of these results.

Note
  1. 1

    We omit the results for illegal gambling. All parameter estimates on casino indicator variables were not statistically different from zero for that category of crime, like in Table 4. We also included three lags of these variables. Those results were very similar to those reported in Table 8.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Literature Review
  5. Empirical Analysis of VLTs, Casinos, and Crime Rates
  6. Methodology
  7. Results
  8. Summary and Conclusions
  9. References
  • Becker, G.S. 1968. Crime and punishment: An economic approach. The Journal of Political Economy 76: 169217.
  • Clark, C., and D.M. Walker. 2009. Are gamblers more likely to commit crimes? An empirical analysis of a nationally representative survey of US young adults. International Gambling Studies 9: 119134.
  • Clarke, R.V., and D.B. Cornish. 1985. Modeling offenders' decisions: A framework for research and policy. Crime and Justice 6: 147185.
  • Cohen, L.E., and M. Felson. 1979. Social change and crime rate trends: A routine activity approach. American Sociological Review 44: 588608.
  • Cotti, C.D., and D.M. Walker. 2010. The impact of casinos on fatal alcohol-related traffic accidents in the United States. Journal of Health Economics 29: 788796.
  • Eadington, W.R. 1999. The economics of casino gambling. The Journal of Economic Perspectives 13: 173192.
  • Giacopassi, D., and B.G. Stitt. 1993. Assessing the impact of casino gambling on crime in Mississippi. American Journal of Criminal Justice 18: 117131.
  • Greene, W.H. 2000. Econometric analysis, 4th ed. Upper Saddle, NJ: Prentice-Hall.
  • Grinols, E.L., and D.B. Mustard. 2006. Casino, crime, and community costs. The Review of Economics and Statistics 88: 2845.
  • Grinols, E.L., and D.B. Mustard. 2008. Correctly critiquing casino-crime causality. Econ Journal Watch 5: 2131.
  • Humphreys, B.R., and J. Marchand. 2012. New casinos and local labor markets: Evidence from Canada. Working Papers 2012-16, University of Alberta, Department of Economics.
  • Hyclak, T. 2011. Casinos and campus crime. Economics Letters 112: 3133.
  • Kearney, M.S. 2005. The economic winners and losers of legalized gambling. National Tax Journal 58: 281302.
  • Koo, J.K., M.S. Rosentraub, and A. Horn. 2007. Rolling the dice? Casinos, tax revenues, and the social costs of gaming. Journal of Urban Affairs 29: 367381.
  • Reece, W.S. 2010. Casinos, hotels, and crime. Contemporary Economic Policy 28: 145161.
  • Rey, S.J., E.A. Mack, and J. Koschinsky. 2012. Exploratory space-time analysis of burglary patterns. Journal of Quantitative Criminology 28: 509531.
  • Smith, G., and H. Wynne. 1999. Gambling and crime in Western Canada: Exploring myth and reality. A report prepared for the Canada West Foundation.
  • Walker, D.M. 2003. Methodological issues in the social cost of gambling studies. Journal of Gambling Studies 19: 149184.
  • Walker, D.M. 2007. Problems in quantifying the social costs and benefits of gambling. American Journal of Economics and Sociology 66: 609645.
  • Walker, D.M. 2008a. Do casinos really cause crime? Econ Journal Watch 5: 420.
  • Walker, D.M. 2008b. Evaluating crime attributable to casinos in the U.S.: A closer look at Grinols and Mustard's “Casinos, crime, and community costs”. Journal of Gambling Business and Economics 2: 2352.
  • Wenz, M. 2007. The impact of casino gambling on housing markets: A hedonic approach. Journal of Gambling Business and Economics 1: 101120.
  • Wenz, M. 2008. Matching estimation, casino gambling and the quality of life. The Annals of Regional Science 42: 235249.
  • Wheeler, S.A., D.K. Round, and J.K. Wilson. 2011. The relationship between crime and electronic gaming expenditure: Evidence from Victoria, Australia. Journal of Quantitative Criminology 27: 315338.