Drugs and Gangs: An exploration of individual and organizational factors

INTRODUCTION

Prior to the 1990s, the methamphetamine problem in the United States was small and restricted to a few southwestern states. However, in recent years criminal justice, treatment, and public health officials have made numerous claims that methamphetamine use has increased dramatically—and several drug use indicators have confirmed their claims. First, the National Household Survey on Drug Abuse, which samples the general population, estimated that in 2000 about 8.8 million people had tried methamphetamines, compared to 1994 when only 1.8 million people were estimated to have tried the drug (Methamphetamine Abuse and Addiction, 2003). Second, reports from the Drug Abuse Warning Network (DAWN), which collects data on drug-related episodes from hospital emergency rooms in 21 metropolitan areas, have indicated that methamphetamine episodes increased 43% between 1999 and 2001 (Kissin and Ball, 2002). Third, results from the Monitoring the Future Program, which surveys 8^th, 10^th, and 12^th graders, reported that methamphetamine use has doubled in the last decade, with about 3% of high school students reporting having used methamphetamines in 1991 (Pennell et al., 1999), compared to roughly 6% in 2002 (Johnston et al., 2002). Last, data obtained from the Arrestee Drug Abuse Monitoring (ADAM) program, showed that methamphetamine use increased significantly in the 1990s among arrestees, particularly in the Southwest (Pennell, Ellett, Rienick, and Grimes, 1999).

In response, federal policymakers enacted several new statutes to more severely sanction methamphetamine manufacturers and users and provided funding aimed at curtailing the use, sale, and manufacturing of methamphetamine. For example, the Methamphetamine Control Act of 1996 doubled the federal penalty for possession of methamphetamines and increased the maximum prison sentence for possession of equipment used to manufacture methamphetamines from four to ten years (Wermuth, 2000: 428). In 1998, the Office of Community Oriented Policing Services (COPS Office) allocated about $4.5 million to six communities with methamphetamine problems to facilitate partnerships between police and community agencies for the purpose of reducing and controlling methamphetamine use (McEwen and Uchida, 2000). Similarly, local community and criminal justice agencies engaged in numerous activities to combat methamphetamine problems. For instance, metropolitan and county task forces were established to suppress methamphetamine use, sales, and production. County prosecutor’s offices created special programs to facilitate the prosecution of those involved in the methamphetamine trade. Community agencies, along with the National Guard, began to implement training programs focusing on the identification of methamphetamine labs and clean-up requirements, and school districts began to implement anti-methamphetamine curriculum programs to prevent youth from using methamphetamines (McEwen and Uchida, 2000).

Due to the heightened concern regarding methamphetamine use researchers from various disciplines have conducted studies to improve the overall knowledge of methamphetamine use and its market effects on communities. Researchers have focused on the history (Anglin et al., 2000), chemical structure (Kleven and Seiden, 1992), and physiological impacts of methamphetamines (Abadinsky, 1993). Other studies have examined the organization of the methamphetamine industry (Jenkins, 1992; 1994) and how that industry impacts the local environment (Centers for Disease Control, 2000). Despite this relatively large body of literature there has been fairly little research examining the causes and correlates of methamphetamine use—particularly in relation to the volume of research on other illicit drugs.

LITERATURE REVIEW

Several national drug use indicators have shown that methamphetamine use has increased over the past 15 years; however, there has been little research in the field of criminal justice and criminology on the causes and correlates of methamphetamine use. With this said, existing research from other disciplines examining illicit drug use in general have provided guidance on several factors that might influence methamphetamine use. In the sections below, we review prior research on the correlates of methamphetamine use. Because of the paucity of peer-reviewed research examining methamphetamine use in the field of criminal justice and criminology, we frame the present study using prior literature that examined the correlates of drug use in general. We also rely heavily on literature from the fields of public health and drugs and behavior to inform us on the causes and correlates of methamphetamine use. Combined, these bodies of literature suggest that methamphetamine use might be influenced by three conceptually distinct domains: individual-level factors, community-level factors, and local drug markets.

Individual-Level Factors and Methamphetamine Use

Much of the existing research of methamphetamine use has centered on the individual- level factors associated with such use. In the early 1990’s, methamphetamine users were said to be primarily White males between the ages of 18 and 35 years old. This body of research also reported that methamphetamine use was more common among individuals in blue-collar occupations such as truck driving, construction, and factory work (Miller, 1991; U.S. Congress, 1990; See also Jenkins, 1994). One of the first major initiatives toward a broader understanding of methamphetamine use was conducted at the 44^th meeting of the Community Epidemiology Work Group (CEWG) in 1989. Researchers from 21 cities reported on several data sources to consolidate findings and further examine the correlates of methamphetamine use. While many of the communities reported that Whites were still more likely to use methamphetamines, they also demonstrated that women were just as likely to use the drug as men and that younger people were beginning to use methamphetamines at higher rates (Mason, 1989; Miller and Thomas, 1989).

One of the largest studies to examine methamphetamine use was conducted by Oetting et al. (2000). Oetting et al. examined data obtained from over 600,000 students as part of the American Drug and Alcohol Survey. They reported that while female methamphetamine use had increased substantially from years past, males were still more likely to use the drug. They also reported that American Indians and Hispanics were most likely to use methamphetamines, followed by Asians, Whites, and African-Americans (Oetting et al., 2000). Glittenberg and Anderson (1999), who examined methamphetamine use in southern Arizona, further explained that much of the methamphetamine use by marginalized youthful populations has been a consequence of their economic plight. They explained that many young people are frustrated with their situation in life and use methamphetamines to cope with “the frustrations and anger associated with their lack of a future with hope” (p. 1987).

Perhaps the largest contribution to the study of methamphetamine use has developed through a focus on sexual behavior. This body of work, primarily conducted in the public health arena, has found a strong correlation between sexual risk-taking behavior and methamphetamine use. Researchers have reported that methamphetamine is perceived to have an aphrodisiac effect for users; resulting in stronger sexual excitement, longer duration of intercourse, and intensified orgasms (Beltran, Ostrow, and Joseph, 1993; Halkitis, 1998; Reback, 1997; Molitor, Ruiz, Flynn, Mikanda, Sun, and Anderson, 1999). An expansion of this research has found that methamphetamine users are more likely to have more sex partners (Kall, 1992), trade sex for money and drugs (Molitor et al., 1999), be gay (Halkitis, Parsons, and Strratt, 2001), and be at a higher risk of HIV transmission (Vittinghoff, Buchbider, Judson, McKiran, and MacQueen, 1998).

Perhaps the two most sophisticated studies to examine the correlates of methamphetamine use were conducted by Pennell et al. (1999) and Herz (2000). Both studies relied on urinalysis and self-report data from recently booked arrestees surveyed through the Arrestee Drug Abuse Monitoring (ADAM) program. Pennell et al. used data from five southwestern cities and Herz used data from several communities in Nebraska. As part of their studies, they compared the socio-demographic characteristics of methamphetamine users to non-methamphetamine users. The data showed that methamphetamine users were more likely to be White and live in private sector housing (versus public housing), but that there were no differences between the two groups with regard to educational level and age. Both researchers also examined the impact of methamphetamine use on criminal involvement. Pennell et al. reported that methamphetamine users were significantly more likely to have been arrested and served time in jail/prison in the past 12 months when compared to non-methamphetamine users. Likewise, in Nebraska, Herz found that methamphetamine users were 11% more likely to have been arrested on a felony charge and were 4% more likely to have prior offenses than non-methamphetamine users.

While the above work has made important contributions to our understanding of methamphetamine use, it has almost exclusively focused on individual-level factors such as socio-demographic variables (e.g., race, age, gender, income, and living arrangement) and behavior (e.g., sexual behavior, crime). However, substance abuse researchers have long argued that macro-level factors such as community characteristics and local drug markets play an important role on drug use.

Community-Level Factors and Methamphetamine Use

Much of the literature on crime and delinquency has been guided by social disorganization theory. Social disorganization theory posits that the structural and economic composition of neighborhoods is related to crime and delinquency (Bursik, 1988: 520; Skogan, 1990). Communities typified by economic deprivation, residential mobility, population density, and racial/ethnic heterogeneity can be characterized as socially disorganized. This body of literature strongly indicates that crime and delinquency are significantly associated with disorganized and impoverished neighborhoods (For full review see Bursik, 1988).

Only recently have researchers begun to examine the impact of community-level factors on drug use. These studies have typically focused on more youthful populations. Esbensen and Huizinga (1990) conducted one of the first studies examining the impact of community characteristics on drug use. Esbensen and Huizinga analyzed data from the Denver Youth Study, which collects self-report data on a wide range of youth behavior, including drug use. The authors matched this data with census data to examine the impact of community structural factors on marijuana use. The researchers relied on seven variables theoretically derived from social disorganization theory: family structure, ethnicity, socio-economic status, housing, residential mobility, marital status, and age composition. Their bivariate analyses showed that self-reported marijuana use was greater among youth who lived in areas with high concentrations of unemployed people, persons employed as labors or service workers, and in areas with higher per unit density.

To date, the most methodologically rigorous study to examine the relationship between community characteristics and drug use was conducted by Joon and Johnson (2001). The authors studied the impact of neighborhood disorder and individual religiosity on marijuana and hard drug use with self-report data from the National Youth Survey and a hierarchical modeling procedure. Their measures of neighborhood disorder were based on respondents’ perceptions of social and physical disorder in their neighborhood. Joon and Johnson reported that those who perceived more social and physical disorganization in their neighborhood were less likely to be closely bonded to family, school, and religious institutions and were more likely to use both marijuana and hard drugs.

Other studies have found that many of these same neighborhood characteristics associated with drug use are also risk factors for youth substance abuse (Hawkins, Catalano, and Miller, 1992). Over a number of years, Hawkins, Catalano, and their associates have surveyed 11 to 18 year-olds in Seattle, Washington as part of their Communities That Care Youth Study. The survey uses 350 self-report questions to study risk and protective factors among youth. Studies produced from the project have repeatedly reported that low neighborhood attachment, neighborhood disorganization, and neighborhood residential mobility are strongly associated with youth drug use (Arthur, Hawkins, Pollard, Catalano, and Baglioni, 2002).

As previously noted, studies examining the impact of neighborhood characteristics on drug use have primarily been focused on youth, marijuana, hard drugs, or drug use in general (i.e., aggregating all illicit drug types). Anecdotal evidence suggests that the types of neighborhood dynamics leading to use of marijuana or other drugs may not have the same relationship to methamphetamine use. For instance, methamphetamine use is most closely associated with Whites who are employed in low skill service industries. Such findings have led many criminal justice officials as well as policymakers to suggest that methamphetamine use is associated with neighborhood characteristics which vary significantly from those of socially disorganized communities.

Local Drug Markets and Methamphetamine Use

The important role of drug markets on drug use has similarly been documented. For example, a number of researchers have found that market factors including drug price (Eck, 1995; Stolzenberg and D’Alessio, 2002), availability (Loxley, 1998; Arthor et al, 2002), enforcement practices (Caulkins et al., 2000; McEwen and Uchida, 2000), and buyer-seller relations (Riley, 1997) have a significant impact on powder cocaine, crack, and heroin use. Interestingly, recent studies of methamphetamine drug markets have shown that such markets differ significantly from other drug markets.

Eck (1995) examined the differences between the methamphetamine, heroin, and cocaine drug markets in San Diego. He found that methamphetamine dealers typically sold through social networks, whereas cocaine and heroin dealers typically sold to strangers. He also found that while most dealers, regardless of the drug, sold out of apartment complexes, methamphetamine dealers were substantially more likely to sell out of single-family homes than cocaine or heroin dealers. He also reported that the locations where methamphetamines were sold had significantly fewer security measures, were more likely to be located in wealthier areas of the community, and were more spread out across a community than those locations where other drugs were sold. Eck argued that that methamphetamine dealers face fewer risks than other drug dealers, and as a consequence, “can lower cost(s), because risks are lower and thus sellers do not need as great a return to compensate them” (p. 83).

In their study of methamphetamine use among arrestees, Pennell et al. (1999) reported that the methamphetamine market is closed, with most users relying on one main source from which to obtain the drug. They found that when methamphetamine users’ sources are not available, users typically abstain from use rather than try to obtain the drug from another source. Arrestees also reported that most transactions were initiated over the phone and took place indoors rather than outdoors. Interestingly, most methamphetamine users reported obtaining the drug free (77%), typically from a friend. Their data illustrated that the amount paid for methamphetamine was similar from one city to another (ranging from $40-50).

THE PRESENT STUDY

The present study expands the research described above by identifying how individual-level, community-level, and drug market characteristics influence methamphetamine use across these jurisdictions. Although prior studies have monitored the trends in methamphetamine use and reported its increase over the years, few studies have focused on explaining methamphetamine use. Prior research in this area has either relied on merely describing the methamphetamine user (e.g., White, employed) by estimating bivariate statistics or examined individual-level factors as the only possible predictors of methamphetamine use. Studies that have relied on more statistically appropriate techniques to examine both individual and community-level data have been based on self-report data from samples of youth. Such prior studies have provided much insight to the understanding of methamphetamine use but have been limited in their ability to substantiate whether methamphetamine use is different from other forms of drug use. Further, it is still unknown whether methamphetamine use is linked with those socially disorganized community and structural dimensions that have been associated with other illicit drug use.

Based on findings from prior research on the relationship between individual and community dimensions and methamphetamine use, we examine the following research questions:

1) What is the relationship between individual-level factors (e.g., socio-demographic and legal factors) and methamphetamine use?

2) What is the relationship between community-level factors (e.g., unemployment rate, educational level, racial/ethnic composition) and methamphetamine use?

3) What is the relationship between methamphetamine drug markets and methamphetamine use?

4) How do individual and community-level predictors of methamphetamine use differ from those of other illicit drugs (e.g., marijuana, cocaine, heroin)?

Prior studies have found that the socio-demographic characteristics and criminal records of methamphetamine users are different from non-methamphetamine users (Reback, 1997; Molitor et al., 1999; Halkitis et al., 2001; Vittinghoff et al., 1998). However, the nature of this relationship is unclear. While some studies indicate that American Indians and Hispanics are more prone to use methamphetamine (Oetting et al., 2000; Glittenberg and Anderson, 1999), other studies have found that Whites use methamphetamine at much higher rates than non-Whites (Mason, 1989; Miller and Thomas, 1989)). In regard to other individual correlates, prior studies have found that users of methamphetamine are more likely to be employed and have more prior contacts with the criminal justice system than non-methamphetamine users (Miller, 1991; Pennell et al., 1999; Herz, 2000). This characterization suggests that methamphetamine users are a different type of user than those who are more likely to use crack or marijuana. By examining how individual-level factors influence methamphetamine use this study will determine the extent to which arrestees in these two jurisdictions mirror the profiles of methamphetamine users found in other studies.

Consistent with studies that stress the important role that community characteristics play in drug use, this study will examine how community-level factors influence methamphetamine use. The availability of arrestee residence zip code enables linking ADAM data to Census 2000 data and facilitates the examination of community-level data on methamphetamine use. Prior studies focused on community-level data have relied on self-report data of youthful populations and marijuana use (Esbensen and Huizinga, 1990). Such studies have found a direct relationship between marijuana users and communities characterized by high levels of unemployment. Other studies have identified a relationship between perceived social and physical disorganization and drug use (Joon and Johnson, 2001). The inclusion of community-level data in an examination of methamphetamine use will show whether characteristics associated with socially disorganized communities (e.g., high unemployment rates, low levels of educational, racial/ethnic mix) are in fact more or less likely to be associated with methamphetamine use. The inclusion of such variables will also enable the examination of the possible mediating effects that community factors may have on individual-level factors.

The relationship between drug markets and drug use has been substantiated by prior studies (Eck, 1995; Stolzenberg and D’Alessio, 2002; Loxley, 1998; Arthor et al, 2002; Caulkins et al., 2000; McEwen and Uchida, 2000; Riley, 1997). Interestingly, the few studies that have examined drug markets across various drugs have found that methamphetamine transactions take place indoors and are highly dependent on social networks. In essence, methamphetamine users obtain the drug not from open sources but receive it for free from their social networks (e.g., friends). Methamphetamine markets have also been shown to be more closed and prevalent in communities characterized by economic prosperity. In order to identify the relationship between drug markets and methamphetamine use, we consider how such factors as transaction location and unit price are associated with methamphetamine use.

Although the primary focus of this study is to examine methamphetamine use among arrestees, we also consider whether methamphetamine users differ from users of other drugs. Thus, we examine how the individual and community-level factors that significantly influence methamphetamine use also predict marijuana, cocaine, and opiate use. A comparison with other drugs will provide a more complete understanding of methamphetamine use than currently exists.

METHODOLOGY

In order to examine how individual, community-level data, and drug markets influence methamphetamine use we integrate two data sources and then examine each of our research objectives in turn.

Data

Data for this study comes from the Maricopa and Pima County Arrestee Drug Abuse Monitoring (ADAM) program from 2000–2003 (N= 6,042). Formerly funded by the National Institute of Justice, the ADAM program was designed to collect information from arrestees to monitor national drug use trends and provide local jurisdictions with information on drug use, treatment needs, and drug markets. Interview data and urine samples were collected within 48 hours of the arrest, providing data on socio-demographics, prior involvement with the criminal justice system, drug dependence and abuse measures, treatment experience(s), drug markets, and self-report and confirmed drug use (for a full review of the ADAM data, please see Hunt and Rhodes, 2001).[1] To incorporate community-level measures within the analyses, arrestees’ residential zip codes were used to link zip code-level data from the 2000 Census (U.S. Bureau of the Census, 2000, Summary Tape File 1 and 3). Arrestees in our sample represent 147 zip codes in Maricopa and Pima County.

Sample

ADAM data were collected from four booking facilities across both counties. Only male arrestees who provided a urine sample and had a valid residential zip code within either Maricopa or Pima Counties are included in the analyses.

Measures

We rely on urinalyses results from arrestees to measure four dependent variables. Methamphetamine use is coded as a binomial variable (positive = 1; negative = 0). In order to establish whether methamphetamine users differ from other drug users, we utilize the following multinomial measures created from a two-way classification of methamphetamine use and each of three other commonly-used drugs: methamphetamine/marijuana use (negative both = 1; positive both = 2; positive marijuana, negative methamphetamine = 3; positive methamphetamine, negative marijuana = 4), methamphetamine/cocaine use (negative both =1; positive both = 2; positive cocaine, negative methamphetamine = 3; positive methamphetamine, negative cocaine = 4), and methamphetamine/opiate use (negative both = 1; positive both = 2; positive opiates, negative methamphetamine = 3; positive methamphetamine, negative opiates = 4).

Predictors at the individual-level include race/ethnicity (dummy coded variables for Hispanic/Latinos, Blacks, Native Americans, with Whites as the omitted category) and age at time of arrest. We also include measures of education (less than high school education = 1; high school education = 0) and employment status (working = 0; not working = 1). Legal factors are measured by the inclusion of most serious offense at arrest (violent offense = 1; non-violent offense = 0) and prior arrest history (yes = 1; no = 0).

To account for potential county differences, we use a dummy coded measure of county (Maricopa = 1; Pima = 0).[2] We then rely on five community-level measures derived from Census data. Unemployment measures the percent of all residents in the labor force who reported being unemployed. Educational level is measured with two variables: percent of residents whose highest level of education is high school and whose highest level of education is college. A measure of racial and ethnic heterogeneity (Blau, 1977) was computed based upon racial/ethnic proportions of the population for each zip code:

Racial and Ethnic Heterogeneity = 1-[(P_White)² +(P_Hispanic)²+(P_Black)²+(P_{American
Indian})²+(P_{Hawaiian or
Pacific Islander})²+ (P_Asian)²+(P_{Multi-racial/ethnic})²+(P_Other)²]

This measure varies between 0 and 1, with higher values indicating greater racial/ethnic heterogeneity. The size of the Hispanic/Latino population in Arizona mandates the inclusion of a cultural indicator, Spanish-speaking households (percent of household who reported Spanish as the primary language spoken).

In order to capture how drug markets influence methamphetamine use, we examine acquisition methods, sources of purchases, location of purchases, dollar amounts of purchases, units of methamphetamine purchased, and frequency of purchases. Table 1 presents the dependent and independent variables used in the analyses along with the corresponding coding scheme.

Analytical Strategy

Given the nested structure of the data (i.e., arrestees within zip codes), a hierarchical modeling technique is used to analyze the data. Multi-level models allow for an examination of both individual and community-level effects on the dependent variable (Raudenbush and Bryk, 2002). Since the dependent variables are non-linear – one is dichotomous (methamphetamine use) and the others are multinomial in nature – hierarchical generalized linear models (HGLM) are used to estimate the impact of individual (Level 1) and community-level (Level 2) factors on drug use. Appendix A contains a full presentation of the models. Maps showing the percentage of drug use and values on the community-level measures across zip codes in these jurisdictions are included to help demonstrate the relationships between the community-level data and methamphetamine use.

Findings

Descriptive statistics of the individual-level data show that the average age of arrestees in this study was 31 (see Table 1). Among the four racial/ethnic groups under examination, the largest proportion of arrestees were White (44%), followed by Hispanics/Latinos (40.1%), Blacks (11%), and Native Americans (5.3%). The majority of cases in the study involved non-violent offenses (71.3%), and unmarried (77%), employed arrestees (75.9%) with a high-school education (68%) who had been arrested in the past (83.2%). Among all arrestees, 24% tested positive for methamphetamines, 41% for marijuana, 30% for cocaine, and 6% for opiates.

The Level 2 (i.e., community-level data) descriptive statistics represent the average value across zip codes (unweighted for number of residents). These statistics indicate that 25% of the population received a high school degree and 54% had received a college degree. The average rate of racial/ethnic heterogeneity across communities was .40. The unemployment measure revealed that the average unemployment rate of community members across zip codes was 6%. Spanish was the primary language spoken at the home in 20% of households.

Individual-Level Correlates of Methamphetamine Use

In order to determine whether the mean methamphetamine use of arrestees varied across communities (i.e., zip codes), an intercepts-only model was first estimated (see Table 2, Model 1). The significant random effects component for the intercept indicates that the rate of methamphetamine use varies across zip codes (p < .001). It is possible that the variation is simply due to the unequal distribution of arrestee characteristics across communities. To control for individual characteristics, a random-coefficient model including only the individual-level measures was estimated. Coefficients that did not vary across zip codes were constrained and Model 2 was estimated. Model 2 presents the estimates of both the fixed effects and random effects. The findings show that age has a negative effect on methamphetamine use and that Hispanic, Black, and Native American arrestees were all less likely to test positive for methamphetamine use than White arrestees. In particular, the odds for Hispanics testing positive for methamphetamines were .48 times (exp [-.751]) lower than the odds for Whites. The likelihood of testing positive was even lower for Black (.17 = exp [-1.799]) and Native American arrestees (.2 = exp [-1.63]) compared to Whites. Arrestees who were unemployed were 1.37 times (exp [.318]) more likely than employed arrestees to use methamphetamines. Violent offenders were .74 times (exp [-.306]) as likely to test positive than non-violent offenders. Reporting a prior arrest had a positive effect on methamphetamine use, with such arrestees being 3 times (exp [1.097]) more likely than arrestees with no prior arrests to use methamphetamines. Education and marital status had no effect on the likelihood of testing positive for methamphetamine. Because the Hispanic coefficient showed statistically significant variance, indicating that its effect on methamphetamine use varied across communities, it was left unconstrained and allowed to vary across zip codes. Further, though nearly cut in half, the variance in the intercept is still significant, revealing that differences in rates of methamphetamine use across zip codes can only partially be explained by the individual level characteristics included in the model.

Community-Level Correlates of Methamphetamine Use

Including community-level characteristics can help explain why the mean rate of methamphetamine use varies across communities. Thus, Model 3 incorporates the community measures as predictors of the intercept in each zip code. Level 1 estimates were similar to the estimates reported in Model 2. Of the five community-level indicators introduced in the model, four significantly influenced methamphetamine use. The percent of college educated residents, unemployed residents, and Spanish-speaking households all significantly influenced the mean rate of methamphetamine use. Unemployment rate of communities has a negative effect on methamphetamine use, indicating that arrestees from communities with a higher percentage of unemployed residents were less likely to use methamphetamines. The effect of the proportion of Spanish-speaking households and percent college education is also negative, revealing that arrestees from communities characterized by a higher percentage of Spanish-speaking households and higher percentage of college educated residents were also less likely to use methamphetamines. Finally, arrestees who reside in Maricopa County were more likely to test positive for methamphetamines than arrestees who reside in Pima County.

Drug Markets and Methamphetamine Use

Only those arrestees who indicated they had obtained methamphetamine in the past 30 days provided data on the methamphetamine market. Market information cannot be used as an individual level predictor because the market information is confounded with drug use (74% of arrestees who provided market information tested positive for methamphetamine, while market information is missing for those who did not obtain methamphetamine). An alternative strategy is to utilize aggregate measures of the market within specific communities (i.e., zip code). To utilize aggregate measures at the zip code level it was only possible to consider those transactions where arrestees’ obtained methamphetamine in their own neighborhoods (rather than outside of their neighborhoods). A review of the drug market indicators revealed that the number of cases meeting these requirements was relatively low, producing small cell counts in most zip codes. This limitation prohibited us from including these variables in the multilevel analyses. Measures at the county level are much more reliable than at the zip code level. Thus, we estimated descriptive statistics for Maricopa and Pima County and found minimal variation across drug markets indicators. In fact, the only statistically significant indicators across these counties were in the percentage obtaining methamphetamine in the past 30 days, number of days acquired in past 30 days, and percent for own use (see Table 3).

Individual and Community-Level Correlates of Marijuana, Cocaine, and Opiate Use Versus Methamphetamine Use

Our final research question was to identify whether the correlates of methamphetamine use differ from the correlates of other drug usage. Separate models were estimated to compare the predictors of methamphetamine use with marijuana (Table 4), cocaine (Table 5), and opiates (Table 6). As stated previously, the categorical nature of the dependent variable called for the use of a multinomial model. Such models estimate the log-odds of falling into each category versus a common comparison category (this can be visualized as a set of logistic regression models estimated simultaneously). In the models presented here, the common reference category is a positive UA for methamphetamine and negative UA for the other drug. Tables 4, 5, and 6 each present a single model, however, the primary coefficients of interest are presented in the first column of each table (the second and third columns of coefficients refer to testing positive for both drugs and testing positive for neither drug respectively). These coefficients represent the effect of the independent variables on the odds of testing positive for the other drug and not methamphetamine versus testing positive for methamphetamine and not the other drug . When significant coefficients are positive they indicate that the measure increases the odds of testing positive for the other drug versus methamphetamine. Negative coefficients indicate that the measure decreases the odds of testing positive for the other drug versus methamphetamine, or that the measure is more likely to lead to methamphetamine use than use of the other drug.

Results in Table 3 indicate that age has a negative effect on marijuana use; older offenders were less likely to use marijuana than methamphetamines. Hispanic, Black, and Native American arrestees were more likely to test positive for marijuana than methamphetamine when compared to White arrestees. Arrestees who had a prior arrest were less likely to test positive for marijuana than methamphetamines, while violent offenders were more likely to test positive for marijuana. Also, arrestees who were unemployed were less likely to test positive for marijuana than methamphetamines. Level 2 measures indicate that arrestees who reside in communities characterized by a higher percentage of college educated residents, unemployed residents, and Spanish-speaking households were more likely to test positive for marijuana than methamphetamines. The county measure shows that arrestees who reside in Maricopa County were less likely to test positive for marijuana than methamphetamine compared to arrestees who reside in Pima County.

Cocaine use predictors show that as age increases, arrestees were more likely to use cocaine than methamphetamines (see Table 4). Hispanic, Black, and Native American arrestees were more likely to test positive for cocaine than methamphetamines when compared to Whites. Arrestees who had a prior arrest were less likely to test positive for cocaine than methamphetamines. At the community level, measures indicate that arrestees who reside in communities characterized by higher unemployed rates and more Spanish-speaking households were more likely to test positive for cocaine than methamphetamines. Consistent with the marijuana analysis, arrestees who reside in Maricopa County were less likely to test positive for cocaine than methamphetamines compared to arrestees who reside in Pima County.

Table 5 contains the findings from the analysis of opiate use. They reveal that as age increases offenders were more likely to use opiates than methamphetamines. Hispanic, Black, and Native American arrestees were more likely to test positive for opiates than methamphetamines when compared to Whites. Arrestees who had a prior arrest were less likely to test positive for opiates than methamphetamines. Level 2 measures indicate that arrestees who reside in communities characterized by higher rates of college education, unemployment, and with more Spanish-speaking households were more likely to test positive for opiates than methamphetamine. Lastly, arrestees residing in Maricopa County were less likely to test positive for opiate use.

Maps of those arrestees who tested positive for the four drugs under examination and the significant community-level measures are found in Figures 1-18. Figures 1-8 display levels of positive urinalyses by drug and by zip code for Maricopa County and Pima County. Figures 9-18 present rates of methamphetamine use alongside values on the community-level measures. Drug usage rates are presented only for those zip codes in which at least 20 arrestees resided. Thus, several zip codes contain information on the measures derived from census data, but do not contain information on drug use.

Figure 1 presents rates of methamphetamine use across zip codes in Maricopa County. Overall rates of use tend to be higher than 30%, with only 5 of the zip codes represented having rates lower than 30%. The highest levels of methamphetamine use are in Glendale, north Phoenix, and the far west and east portions of the county. Arrestees residing in central Phoenix and Tempe zip codes had the lowest levels of methamphetamine use. Figure 2 presents the percent of arrestees testing positive for marijuana in Maricopa County by zip code. Note that the lowest level of usage in any zip code is 24% and that fewer patterns exist in marijuana usage by zip code. Marijuana tends to be scattered more throughout the Valley. Figure 3 presents the percent of positive cocaine tests by zip codes. Cocaine use shows the most variation in use across zip codes with rates in several zip codes less than 20% and rates in several other zip codes greater than 41%. Cocaine usage (either crack or powder) was highest in south Phoenix, south Scottsdale, and Tempe zip codes. Figure 4 displays the percent of positive opiate urinalyses. Overall rates of use are consistently much lower than other drugs across the valley, with only 5 zip codes exceeding 10%.

Figure 5 presents the rate of methamphetamine use across zip codes in Pima County. The maps for Pima County use the same categories as in Maricopa County to represent use. It is readily apparent that rates of methamphetamine usage are much lower in Pima County, with most zip codes exhibiting usage rates lower than 30%. Zip codes with higher rates of methamphetamine use tend to be located closer to the outskirts of Tucson (the primary city in Pima County). Figure 6 displays levels of marijuana use in Pima County. As in Maricopa County, marijuana use is more widely used across the county, with only a single zip code falling beneath 32% in Pima County. Figure 7 presents cocaine use across Pima County, which, like marijuana, tends to be used consistently across the county, though perhaps with higher rates in the southern and western areas shown. Rates of opiate use, shown in Figure 8, are rather consistent across the county, with only a single zip code falling below 4%. The highest rates tend to be in the western areas and selected central areas.

Figure 9 presents rates of methamphetamine use by percentage of residents with a college education in Maricopa County. Interestingly, the zip codes with the highest levels of college-educated residents do not contain enough arrestees to make reliable estimates of methamphetamine use. Also, it appears that zip codes with more college-educated residents have higher rates of methamphetamine use. However, as Table 2 showed, after controlling for individual and other community characteristics, the relationship between methamphetamine use and college education is negative. Figure 10 presents rates of methamphetamine use by percentage of residents with a high school degree as their highest education level in Maricopa County. Unlike college education, there does not appear to be a spatial pattern in this relationship. Figure 11 displays the relationship between methamphetamine usage rates and the percent of residents who are not working. The highest rates of methamphetamine use appear in those zip codes with the lowest levels on unemployment. Zip codes with more unemployment are mostly located in south Phoenix, where methamphetamine usage rates are not high. As shown in Table 2, this relationship does persist when individual and other community level characteristics are controlled. A similar pattern is seen in Figure 12, in which the zip codes with the most households speaking Spanish have lower rates of methamphetamine use. Like unemployment, zip codes with more households where Spanish is the primary language are located in southern Phoenix. Figure 13 presents rates of methamphetamine use by racial/ethnic heterogeneity within zip codes in Maricopa County. Though the most heterogeneous neighborhoods tend to be in the southern and western portions of the county, there does not appear to be an association between heterogeneity and methamphetamine use.

Figure 14 presents rate of methamphetamine use by percentage of residents with a college education in Pima County. Unlike Maricopa County where there appeared to be a relationship between the two measures, there does not appear to be a relationship in Pima County. Low levels of methamphetamine use are seen both in counties with high rates of college education and with low rates of college education. The relationship between methamphetamine use and rates of high school as the highest level of education can be seen in Figure 15. As with college education, there does not appear to be a relationship between the two measures in Pima County. Figure 16 presents rates of methamphetamine use by the unemployment rate across zip codes in Pima County. Because the rate of methamphetamine use is low all across Pima County, it is difficult to identify a clear relationship between methamphetamine use and unemployment rates in this instance. The relationship between the percent of households where Spanish is spoken primarily and methamphetamine use can be seen in Figure 17. Spanish speaking households tend to be focused in the southern and western areas of the county. Some of these zip codes have low levels of use, while others have higher rates of use. While it is possible that the two measures are related, the analysis in Table 2 revealed that after controlling for individual and other community characteristics, the relationship between Spanish speaking households and methamphetamine use was negative. Finally, Figure 18 presents rates of methamphetamine use by racial/ethnic heterogeneity in Pima County. As in Maricopa County, there does not appear to be a spatial relationship between these two measures.

Discussion

An extensive body of research has examined the patterns underlying illicit drug use. Much of this research has focused on individual-level factors, and the findings reported here reaffirm the importance of such factors in methamphetamine use within the criminally involved population. Perhaps more importantly, the findings from the present study demonstrate the importance of incorporating community-level factors along side individual factors in building a more complete understanding of the etiology of substance abuse. These findings demonstrate that predictors of methamphetamine use (at the individual and community-level) differ significantly from those of marijuana, cocaine, and opiate use. These findings have both practical and theoretical consequences. Identifying factors that are related to methamphetamine use enables criminal justice and public health agencies to expend resources where the greatest impact in methamphetamine use reduction is expected. Theoretically, our findings show that the community-level factors associated with socially disorganized communities are not significant correlates of methamphetamine use. These findings have important practical applications. For example, efforts to curb methamphetamine use in communities, based on the faulty assumption that social disorganization and methamphetamine use go hand in hand, may in fact prove ineffective and be a misuse of scarce resources. The insignificant role of racial/ethnic heterogeneity and significant but negative effects of community levels of unemployment, college education, and Spanish-speaking households on methamphetamine use demonstrates that the communities with higher methamphetamine use also tend to have relatively high degrees of social organization. Additionally, our findings show that the effects of these community-level measures on drug use vary across different types of drugs. Effective treatment delivery strategies and law enforcement suppression strategies will be more effective if they take into consideration community-level factors and reflect an understanding of the different patterns of drug use that are associated with different community characteristics. Our findings suggest, for example, that the expansion of treatment services in communities with individuals who use other drugs many not reach those who use methamphetamines.

The link between drug markets and drug use may play a less significant role given our comparison of the drug market indicators across counties. The few differences found between counties may reinforce the notion that traditional measures of drug markets (e.g., open markets, source type, and unit price) are less important in determining methamphetamine use. Given that the individual and community-level predictors of methamphetamine use in this study are significantly different from marijuana, cocaine, and opiate use, it is quite possible that the methamphetamine drug markets themselves also vary from other drug markets. Methamphetamine acquisition may take place in niches or communities where a segment of society identifies with a particular social/cultural norm that finds such use acceptable. As such, other niches within communities may find other drug use more acceptable than methamphetamine use. These niches may be based more on social, political, and cultural aspects.

County maps of drug use show that drug usage patterns tend to exhibit consistency across zip codes. Such maps can be useful to policymakers trying to reduce drug use, whether through suppression, education, or rehabilitation. Maps of community characteristics and drug use can help identify which communities in particular experience problems of drug use and provide insight to their capacity to draw on other community resources.

Future studies of methamphetamine use should direct focus to explaining why individuals choose this particular drug over other types of drugs and how social networks operating within drug-specific niches within the community influence and structure drug use. Particular methodologies may be more appropriate to examine the decision-making processes of users (e.g., more ethnographic studies). Also, exploring methamphetamine drug markets in a broader perspective will better facilitate the identification of links between methamphetamine use and drug markets. This may reveal more about how social relations among individuals mediate the acquisition of methamphetamines. Research should continue to examine the racial/ethnic dimensions of drug use given the changing demographic composition of communities. Lastly, future studies should rely on statistically appropriate techniques to examine drug use and recognize that incorporating measures that extend beyond individual socio-demographics are important to the study of drug use.

REFERENCES

Abadinski, H. (1993). Drug Abuse. Nelson-Hall: Chicago.

Anglin, D, Burke, C., Perrochet, B., Stamper, E. & Dawud-Noursi, S. (2000). History of the methamphetamine problem. Journal of Psychoactive Drugs, 32, 2: 137-141.

Arthur, M.W., Hawkins, J.D., Pollard, J.A., Catalano, R.F., & Baglioni, A.J. (2002). Measuring risk and protective factors for substance use, delinquency, and other adolescent problem behaviors: The Communities That Care Youth Survey. Evaluation Review 26, 6: 575- 601.

Beltran, E., Ostrow, D., & Joseph, J. (1993). Predictors of sexual behavior change among men requesting their HIV-1 antibody status. AIDS Education and Prevention, 5, 3: 185-195.

Blau, P.M. (1977). Inequality and Heterogeneity: A Primitive Theory of Social Structure. New

York: Free Press.

Bursik, R.J. (1988). Social disorganization and theories of crime and delinquency: Problems and prospects, 26, 4: 519-552.

Caukins, J.P., Dworak, M. & Feichtinger, G. (2000). Price-raising drug enforcement and

property crime: a dynamic model. Journal of Economics, 71, 3, 227-253.

Centers for Disease Control (2000). Public health consequences among first respondents to emergency events associated with illicit methamphetamine laboratories. Weekly, 49, 45: 1021-4.

Eck, J. E. (1995). A general model of the geography of illicit retail marketplaces. In J.

Eck & D. Weisburd (Eds.). Crime and Place, vol. 4 (pp. 67-94). Monsey: Criminal Justice Press.

Esbensen, F. & Huizinga, D. (1990). Community structure and drug use: from a social

disorganization perspective. Justice Quarterly, 7, 4, 691-709.

Glittenberg, J & Anderson, C. (1999). Methamphetamines: Use and Trafficking in the Tucson- Nogales Area. Substance Use and Misuse, 34, 4” 1977-1989.

Halkitis, P. (1998). Formative research of HIV+ men who have sex with men. Paper presented at the research colloquium series of the National Development and Research Institutes. New York, NY.

Halkitis, P., Parsons, J., & Strratt, M. (2001). A double epidemic: Crystal methamphetamine drug use in relation to HIV transmission among gang men. Journal of Homosexuality, 4, 2, 17-35.

Hawkins, J.D., Catalano, R.F., & Brewer, D.D. (1992). Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood. Psychological Bulletin, 112, 64-105.

Herz, D.C. (2000). Drugs in the Heartland: Methamphetamine Use in Rural Nebraska.

Washington DC: National Institute of Justice

Hunt, D. & Rhodes, W. (2001). Methodology Guide for ADAM. Washington ,DC: National

Institute of Justice.

Jenkins, P. (1992). The speed capital of the world: organizing the methamphetamine

industry in Philadelphia 1970-1990. CJPR, 6, 1, 18-39.

Jenkins, P. (1994). “The Ice Age” the social construction of a drug panic. Justice

Quarterly, 11, 1, 7-31.

Johnston, L. D., O’Malley, P. M., & Bachman, J. G. (2003). Montoring the Future

national results on adolescent drug use: Overview of key findings, 2002. (NIH Publication No. 03-5374). Bethesda, MD: National Institute on Drug Abuse.

Joon, S. J. & Johnson, B. R. (2001). Neighborhood disorder, individual religiosity, and

adolescent use of illicit drugs: a test of multilevel hypothesis. Criminology, 39, 1, 109-143.

Kall, K. (1992). Effects of amphetamine on sexual behavior of male IV drug users in Stockholm—A pilot study. AIDS Education Prevention, 4:6-17.

Kissin, W. & Ball, J. (2002). Club drugs, 2001 update. The DAWN Report, Oct., 1-4.

Kleven, M. & Seiden, L. (1992). Methamphetamine-induced neurotoxicity. The neurobiology of drug and alcohol addiction. Annals of the New York Academy of Sciences, 654: 292- 301.

Loxley, W. (1998). Weed or speed? Convicted drug dealers’ views of substitutions in the

western Australian illicit drug market to 1992. International Journal of Drug Policy, 9, 109-118.

Mason, T. (1989). Preliminary look at social and economic dynamics influencing drug

markets, drug use patterns, and HIV risk behaviors among injecting drug users in two Baltimore networks. Community Epidemiology Work Group, 108-117.

McEwen, T. & Uchida, C. D. (2000). An Evaluation of the COPS Office

Methamphetamine Initiative :Interim Report. Alexandria: Institute for Law and Justice.

Methamphetamine Abuse and Addiction (2003). National Institute on Drug Abuse: Washington DC. Can be found at

http://www.nida.nih.gov/ResearchReports/Methamph/methamph2.html.

Miller, M. A. (1991). Methamphetamine Abuse: Epidemiologic Issues and Implications.

Rockville, MA: National Institute on Drug Abuse.

Miller, M. A. & Tomas, J. M. (1989). Past and current methamphetamine epidemics.

Community Epidemiology Work Group, 58-65.

Molitor, F., Ruiz, J., Flynn, N., Mikanda, J., Sun, R., & Anderson, R. (1999). Methamphetamine use and sexual and injection risk behaviors among out-of-treatment injection drug users. American Journal of Drug and Alcohol Abuse, 25, 3: 475-493.

Oetting, E., Deffenbacher, J., Taylor, M., Luther, N., Beauvais, F., & Edwards, R. (2000). Methamphetamine use by high school students: Recent trends, gender, and ethnicity differences, and use of other drugs. Journal of Child and Adolescent Substance Abuse, 10, 1: 33-50.

Pennell, S., Ellett, J., Rienick, C. & Grimes, J. (1999). Meth Matters: Report on

Methamphetamine Users in Five Western Cities. Washington DC: National Institute of Justice.

Reback, C. (1997). The social construction of a gay drug. Report funded by the City of Los Angeles.

Riley, K. J. (1997). Crack, Powder Cocaine, and Heroin: Drug Purchase and Use

Patterns in Six U.S. Cities. Washington DC: Office of National Drug Control Policy.

Raudenbush, S.W. & Bryk, A.S. (2002). Hierarchical Linear Models: Applications and Data Analysis Methods, (2^nd Edition). Thousand Oaks: Sage Publications.

Skogan, W. G. (1990). Disorder and Decline: Crime and the Spiral of Decay in

American Neighborhoods. New York: The Free Press.

Stolzenburg, L. & D’ Alessio, S. J. (2002). A multilevel analysis of the effect of cocaine

price on cocaine use among arrestees. Journal of Criminal Justice, 550, 1-11.

United States Congress (1990). The Re-Emergence of Methamphetamine. Hearings before the Subcommittee on Narcotics Abuse and Control, US House of Representatives, 101^st Congress, First Session. October 24, 1989. Washington, DC: US Government Printing Office.

United States Congress (2002). Emerging threats, methamphetamine. Hearings before the Subcommittee on Criminal Justice, Drug Policy, and Human Resources, US House of Representatives, 107^th Congress, First Session. July 12, 2001. Washington, DC: US Government Printing Office.

Vittinghoff, E., Buchbinder, S., Judson, F., McKirnan, D., & MacQueen, K. (1998). Per-contact risk of HIV associated with four types of homosexual contact. Paper presented at the fifth annual Conference of Retroviruses and Opportunistic Infections, Chicago, IL.

Wermuth, L. (2000). Methamphetamine use: Hazards and social influences. Journal of Drug Education, 30, 4: 423-433.

Appendix A. Description of Models

Leve1 1 and 2 models for the examination of the dependent variables are presented below:

Level 1: (individual)

h_ij= b_0j + b_1j (X₁_ij – X_1j) + …. + b_pj (X_pij – X_pj) + e_ij (Equation 1)

Level 2: (community)

b_0j = g₀₀ + g₀₁w₁_j + …. + g₀_qw_qj + u₀_j(Equation 2)

Equation 1 includes variables at the individual level while Equation 2 includes zip code level measures. In Equation 1, h_ij represents the log-odds of the dependent variables (i.e., log-odds of methamphetamine use) for arrestee i in zip code j. In the Level 1 model, (X_pij – X_pj) refers to the independent variables measured at the individual level. The independent variables are centered around the Level 2 means (group mean centering) given the interest in examining potential differences across zip codes. The b coefficients in this model estimates the magnitude of the independent variables’ impact on the dependent variable. Subscript p represents the number of individual-level variables (error term (e_ij) represents the variation in error among arrestees). Subscript q represents the number of community-level variables. In Equation 2, b_0j represents the intercept in Equation 1. This model attempts to predict mean differences in the Level 1 outcome variable across zip codes. In the Level 2 model, w refers to the independent variables measured at the zip code level while the g coefficients represent their effects. Error term in the Level 2 model (u_pj) represents the error across zip codes.

[1] ADAM data contain sampling weights to correct for the unequal probability of selection during data collection. We were unable to incorporate these weights because no known software package (SAS, Stata, SPSS, HLM) will estimate the HGLM model while including sampling weights, without extensive programming. Stata comes the closest, and we were able to estimate a multinomial model that corrects for the nonindependence of observations within zip codes while also specifying sampling weights. These models differ from the final models presented in Tables 2-5 in that level-2 predictors could not be included in the model. We estimated a model for each dependent variable with and without the weights. The difference in coefficients and standard errors between the weighted and unweighted models are small and do not change the substantive interpretation of the models. We are confident that the results presented with the unweighted data offer a reasonable representation of the true relationships in the population.

[2] Our county measure could be considered as a third level of data, however, with only two level-3 units we are prohibited from computing variance in the county mean. Including this measure controls for the many possible differences across these counties not captured by the community level measures (e.g., political, social, and cultural differences) at the zip code level.

TABLE OF CONTENTS
	Page
INTRODUCTION	1
LITERATURE REVIEW	3
Individual-Level Factors and Methamphetamine Use	3
Community-Level Factors and Methamphetamine Use	6
Local Drug Markets and Methamphetamine Use	8
THE PRESENT STUDY	9
METHODOLOGY	12
Data	12
Sample	13
Measures	13
Analytical Strategy	15
Findings	15
Individual Level Correlates of Methamphetamine Use	16
Community Level Correlates of Methamphetamine Use	17
Drug Markets and Methamphetamine Use	18
Individual and Community-Level Correlates of Marijuana, Cocaine, and Opiate Use Versus Methamphetamine Use	19
Discussion	24
REFERENCES	27
APPENDIX A: Description of Models	32
APPENDIX B: Tables	33
APPENDIX C: Figures	41