Instructions No directly quoted material may be used in this project paper. Resources should be summarized or paraphrased with appropriate in-text and Resource page citations. Project 3 (Final Proj

Aggressive Crime, Alcohol and Drug Use, and Concentrated Poverty in 24 U.S. Urban Areas Avelardo Valdez, 1Charles D. Kaplan, 1and Russell L. Curtis, Jr. 2 1Graduate School of Social Work, University of Houston, Houston, Texas, USA2Department of Sociology, University of Houston, Houston, Texas, USA Abstract: The nexus between substance use and aggressive crime involves a complex interrelationship among mediating individual and community-level variables. Using multilevel logistic regression models, we investigate how com- munity-level concentration of poverty variables mediate the predictive relationships among individual level social attachment variables and substance use on aggressive crime in a large national sample of male arrestees (N¼20,602) drawn from 24 U.S.

urban areas. The findings support our hypothesis that individual social attachments to marriage and the labor force (education and employment) are the principal individual-level pathway mediating the substance abuse=aggression nexus. In the random intercept model, 3.17%of the variation not explained by the individual- level predictor variables is attributable to community-level variation in urban area female-headed households and households receiving welfare. This confirms our hypothesis that social structural conditions of an urban environment differentially expose persons to conditions that predict being arrested for an aggressive crime.

Our findings tend to counter the cultural theorists who argue for an indigenous culture of violence in inner-city ghettos and barrios.

Keywords:Aggressive crime, alcohol, arrestees, drugs, Drugs-violence nexus A common assumption in the U.S. is that substance use and violent crime is highly related. Upon closer observation, however, the association of these two behaviors at the individual, situational, and community-level Support for this research was funded by the National Institute on Drug Abuse (R24 DA07234). Special thanks are given to Donald Hedeker for advising on the initial statistical analysis.

Address correspondence to Avelardo Valdez, Graduate College of Social Work, Office for Drug and Social Policy Research, University of Houston, 237 Social Work Building, Houston, TX 77204-4013. E-mail: [email protected] The American Journal of Drug and Alcohol Abuse, 33: 595–603, 2007 CopyrightQInforma Healthcare ISSN: 0095-2990 print/1097-9891 online DOI: 10.1080/00952990701407637 595 is more complex and subtle. This article builds upon our previous research expanding its scope to include the variability of urban context, specifically concentrated poverty (1, 2). Utilizing a large national sample, we investigate here how the concentration of poverty mediates the rela- tionships among individual-level predictors, substance use, and violent crime in male arrestees (N¼20,602) drawn from 24 U.S. metropolitan areas.

ILLEGAL DRUGS, ALCOHOL, AND VIOLENT CRIME While the association of alcohol, drug use, and violent crime enjoys a long research history, it is only in recent years that direct measures of this relationship (e.g., physical drug tests and officially known crimes) using large quantitative data sets have been available. These studies have found that alcohol is consistently linked to aggressive and violent behavior (3, 4). In contrast, research on drug use and violence generally concludes, contrary to popular conceptions, that these relationships are unsyste- matic and=or weak (5, 6). Nonetheless, mediating individual-level charac- teristics such as age, gender, race, and ethnicity, and personality factors, for example, may be important in explaining the causal pathways from intoxication to aggression (7). As well, community-level risk factors using neighborhoods as the unit of analysis has been used to explain violence and crime with disadvantaged urban areas (8, 9).

We theorize that aggressive crime will vary systematically with the structural features of the urban environment. Our argument is that aggressive crime and violence is rooted in the structural differences among these metropolitan areas. That is, the higher the concentration of poverty, the higher the levels of aggressive crime. Moreover, on the individual level, the existence of social attachments, such as marriage, are important in deterring aggressive crimes. Our hypothesis is that alcohol and drug use will be significantly related to aggressive crime, but that specific individual-level social characteristics and community- level concentrated poverty variables will mediate this relationship.

PROCEDURES Sample and Measurement Our data are drawn from the 1992 Drug Use Forecasting (DUF) program conducted in 24 cities ranging from larger (Houston and Miami) to smaller (Ft. Lauderdale) cities, some with high Mexican-American (e.g., San Antonio), African-American (St. Louis), and other Hispanic 596 A. Valdez et al. (New York and Chicago) populations. In 1997, the program was reorganized and renamed the Arrestee Drug Abuse Monitor (ADAM) program. The 1992 national data set was used in this analysis similar to prior analyses we published. If we had chosen more recent data from the ADAM system, the interpretability of our earlier results would be confounded.

Female arrestees were excluded from our study because males are overwhelmingly more likely to be perpetrators of aggressive crimes (10, 11). The sample includes a wide range of racial and ethnic groups in this relatively young group of men with lower levels of education— the groups charged with the bulk of violent=aggressive crime in this coun- try. The DUF data combine measures of violent and=or aggressive actions and drug (urinalysis) and alcohol (self-reported) use with mea- sures of ethnicity, socioeconomic positions, age, and city for over 20,000 respondents. The validity of drug test data of arrestees has been demonstrated in numerous studies (12, 13). Over 90%of those arrestees approached agreed to be interviewed and over 80%of these consented to urine samples. The limitations of DUF methodology have also been recognized (14).

The type of crime (aggressive=nonaggressive) was based upon the charge for which the offender was booked and conceived as the depen- dent variable in the analysis. Aggressive crimes included extortion=threat, homicide, kidnapping, robbery, sex offenses (rape), assault, family offenses, obstruction of police, and disturbance of public peace. Non- aggressive crimes included burglary, prostitution, drug sale, weapons, flight from bench warrants, forgery, fraud, larceny=theft, probation=parole violation, stolen property, stolen vehicle, under the influence, drug pos- session, fare beating, liquor, obscenity, driving while intoxicated (DWI), and driving violations (not DWI). Alcohol consumption was obtained from self-reports with a cut-off point based on previous studies (15, 16). The DUF sociodemographic characteristics of the arrestees were also included.

Four community-level concentrated poverty variables were included: the percentages of high-school dropouts, unemployed males, households receiv- ing welfare, and female-headed households in that metropolitan area. These variables were calculated using information published in the 1990 U.S.

Census Survey and the procedures documented in the national Urban Underclass Database (17).

Statistical Analysis Due to the design of DUF data collection procedures, the sample has an unbalanced clustered structure. We used random-effects logistic regression models (RRM) in order to include a random cluster effect that Crime, Alcohol, and Drugs in 24 U.S. Cities 597 estimates the influence of the cluster on the outcomes of the individuals within the cluster (18, 19). Application of conventional statistical models that assume independent observations, such as linear regression and fixed-effects analysis of variance models, to clustered data tends to inflate the Type I error rate and produce significance tests that are too liberal.

Estimation of the parameters of the RRM was performed using the HLM program (20).

Three models were fit to predict aggressive=nonaggressive crime in the 24-city data. In all models a random urban area effect was included to account for the clustering of individuals within cities. Beside the ran- dom urban area effect, the base model included individual-level effects of an offender’s drug and alcohol use. An interaction term of drug use and alcohol use was included in a preliminary analysis. Although a trend was identified, the term was removed from the model for the sake of par- simony. The simple random effects model added socioeconomic covariate effects at the individual-level: employment status, level of education, marital status, ethnicity, income, and offender’s age. The random inter- cept model added covariate effects of the community-level concentrated poverty variables. In a preliminary analysis, a variable indicating if the city alone or the county would determine whether there was a mediating effect of differences in size among DUF metropolitan areas proved not to be significant and was excluded from the analysis. Through the examin- ation of the variance components of a null model and these three models, we determined which model best fitted the data.

RESULTS For the total sample, almost two-thirds of the offenders have been charged with nonaggressive crimes while around one third have been charged with aggressive crimes. Nearly 19%of the sample is of Hispanic-American origin, 23%Euro-American, and 58%African-American. The sample is relatively young and undereducated with the average age being 30 years old (SD¼8.877) and the majority (56%) not having completed high school. The majority of the sample is single (56%) with 30%being married and 14%divorced or separated. Sixty-three percent of the offender’s urine sample tested positive for some type of drug. Nearly 45%of the sample tested positive for cocaine, 26%for marijuana, and 7%for opiates. Only a small percentage of the sample tested positive for the other seven drugs.

Table 1 presents an overview of the four concentrated poverty community-level variables used in this study for the 24 DUF metropoli- tan areas. On the percentage of high-school dropouts, most cities were between the 13%and 17%range. St. Louis displayed the lowest rate 598 A. Valdez et al. on this measure at 7.86 while Houston showed the highest rate at 17.45.

On male unemployment most cities were in the range between 10%and 13%. St. Louis also had the lowest male unemployment rate (9.72) while Houston also had the highest (15.84). St. Louis consistently had the low- est rate on the variable of households receiving welfare (3.57) with Ft.

Lauderdale next in the ranking (3.97) while Detroit had the highest rate (16.04). This variable showed more variation than either high school drop out or male unemployment rates. Most cities were in the 8%to 11% range. The most variation was found in the variable of percentage of female-headed households. A wide range from a high of 41.58%for Atlanta to a low of 9.81%for Birmingham was distinguished.

Table 1.Percentages of underclass city-level indicator variables for 24 DUF metropolitan areas in 1992 Metropolitan areaHigh school— dropout (%)Male unemployment (%)Households receiving welfare (%)Female-headed households (%) 1. Atlanta 13.08 14.65 13.64 41.58 2. Birmingham 12.29 9.81 12.29 9.81 3. Chicago 17.04 11.69 14.36 31.05 4. Cleveland 10.62 10.64 10.45 22.10 5. Dallas 17.14 11.62 4.60 18.51 6. Denver 16.17 12.06 7.61 21.93 7. Detroit 14.95 11.22 16.04 29.81 8. Ft. Lauderdale 13.97 9.92 3.97 15.21 9. Houston 17.45 15.84 7.06 22.73 10. Indianapolis 17.23 10.48 5.82 20.63 11. Kansas City 15.66 12.37 9.22 24.93 12. Los Angeles 17.33 10.99 9.85 18.81 13. Miami 13.17 10.95 9.96 20.93 14. New Orleans 12.81 11.68 9.81 24.89 15. NY=Manhattan 13.08 11.10 10.94 30.43 16. Omaha 10.47 10.51 6.52 20.78 17. Philadelphia 15.05 15.03 13.98 31.77 18. Phoenix 15.02 11.12 4.94 14.40 19. Portland 13.49 12.32 6.54 17.48 20. San Antonio 11.42 12.53 8.40 19.40 21. San Diego 11.24 10.85 8.18 15.49 22. San Jose 10.97 10.67 6.38 14.16 23. St. Louis 7.86 9.72 3.57 14.42 24. Washington, D.C. 13.87 13.23 8.94 39.19 Metropolitan area¼county (instead of city). Crime, Alcohol, and Drugs in 24 U.S. Cities 599 Table 2 presents the conditional coefficient estimates and standard errors of the predictor variables on aggressive crime for the 3 models.

The effects of the drug and alcohol variables were robust across the three models. Specifically, a positive response on alcohol use increased the like- lihood of being charged with an aggressive crime, while a negative response on drug use increased the probability of being charged for a crime. Moreover, findings largely support our hypothesis that social attachments to marriage and the labor force are the principal individ- ual-level pathway mediating the substance abuse=aggression nexus. Test- ing negative on drugs is the strongest predictor for being arrested for an aggressive crime in our multilevel analysis. These findings tend to counter the cultural theorists who argue that there is an indigenous culture of violence in inner-city ghettos and barrios.

While not shown, the amount of variation attributable to the metro- politan area is statistically significant. In the random intercept model, 3.17%of the variation not explained by the individual-level predictor variables is attributable to community-level variation. This confirms our hypothesis that structural conditions of an urban environment differ- entially expose persons to conditions that predict being arrested for an aggressive crime.

DISCUSSION AND CONCLUSIONS We find for a large national sample of arrestees that testing positive for illegal drug use is negatively associated with aggressive crime and that, in contrast, self-reported frequent use of alcohol has strong and robust posi- tive effects. These results are consistent with our earlier research in Hous- ton, Dallas, San Antonio, as well as European national-level studies of aggressive behavior and substance use (1, 2, 21). The negative association of drug use on aggressive crime supports the less popular notion that illegal drug-related violence has less to do with intoxication (pharmaco- logical) and possibly more with other factors.

We found that the multilevel model provides the best fit of the 1992 DUF data. Two significant concentrated poverty variables in the model were significant in explaining variation in aggressive behavior across the 24 urban areas. The specific urban area profile of a high percentage of female-headed households with a corresponding low percentage of house- holds receiving welfare was found in our study to shape the urban context in which drug and alcohol use have robust effects on aggressive crime.

The additive effect of heavy drinking to this stressful social complex appears to further increase the odds of being arrested for an aggressive crime. Lastly, we also found that exposure to certain specific structural 600 A. Valdez et al. Table 2.Conditional coefficient estimates of fixed effects aand standard errors of the individual- and city-level variables on aggressive crime for the base, simple random intercept and random intercept models for 1992 DUF national sample (N¼20,602) Base Model Simple Random Intercept Model Random Intercept Model Variables Coefficient Std Error Coefficient Std Error Coefficient Std Error Individual-level variables Grand intercept .378 .076 .248 .097 .089 .587 Drug use .560 .031 .544 .032 .544 .032 Alcohol use .165 .032 .174 .033 .174 .033 Employment status .068 .037 .067 .037 High school education .036 .035 .034 .035 College education .109 .044 .109 .044 Single vs. married .308 .035 .309 .035 Divorced vs. married .358 .050 .357 .045 African-American vs. Euro-American .005 .040 .004 .041 Hispanics vs. Euro-American .052 .051 .047 .051 Legal income .100 .037 .100 .037 Age .009 .012 .009 .012 City-level variables %high school dropout.024 .029 %unemployed males .059 .059 %households receiving welfare .060 .020 %female-headed households.033 .014 p<.05; p<.01; p<.001.

aControlling for the random effect of metropolitan area.

601 conditions of concentrated poverty seems to be more salient than race in explaining the violence and substance abuse nexus.

Wilson (22) and others (11) argue that the constellation of these characteristics in low-income urban communities produces what they identify as ‘‘concentrated effects.’’ These communities are characterized by poverty, joblessness, welfare dependency, female-headed families, declin- ing marriage, illegitimate births, welfare dependency, and crime that result in multiple, interlocking social problems. The violence–substance use nexus as indicated by this study can be traced, in part, to the social disorganization that is associated with community-level factors of these cities.

Study Limitations One limitation of this study is that the DUF data is not representative of the general population. Further, it is not possible to precisely determine whether the periods of arrestee drinking and=or drug use overlapped pre- cisely with the period when the alleged crimes were committed. Our analysis was also limited by not breaking down the urine analysis mea- sure by specific illegal drugs. Another limitation is that there might be an overlap between the measures of alcohol and drug use. Despite their limitations, these data allow us to identify the specific pathways leading from the urban context to individual aggressive behavioral outcomes at a national level.

REFERENCES 1. Valdez A, Kaplan CD, Curtis RL, Yin Z. Illegal drug use, alcohol and aggressive crime among Mexican-American and White male arrestees in San Antonio. J Psychoactive Drugs 1995; 27(2):135–143.

2. Valdez A, Yin Z, Kaplan CD. A comparison of alcohol, drugs, and aggress- ive crime among Mexican-American, Black, and White male arrestees in Texas. Am J Drug Alcohol Abuse 1997; 23(2):249–265.

3. Parker RN, Cartmill RS. Alcohol and homicide in the United States 1934– 1995—Or one reason why U.S. rates of violence may be going down. J Crimi- nal Law and Criminology 1998; 88(4):1369–1398.

4. Weiner MD, et al. From Early to Late Adolescence: Alcohol Use and Anger Relationships. Journal of Adolescent Health 2001; 28(6):450–457.

5. Fagan J. Intoxication and Aggression. In Drugs and Crime, Tonry M, Wilson JQ, eds., Chicago, IL: University of Chicago Press, 1990; 241–320.

6. Miczek KA, De Bold JF, Haney M, Tidey J, Vivian J, Weertz, M. Alcohol, drugs of abuse, aggression, and violence. In Understanding and Preventing Violence. Reiss AJ, Jr., Roth JA, eds. Washington, D.C.: National Academy Press, 1994; 377–570. 602 A. Valdez et al. 7. Sampson RJ, Lauritsen JL. Violent victimization and offending: Individual-, situational-, and community-level risk factors. In Understanding the Prevent- ing Violence: Volume 3 Social Influences. Reiss AJ, Jr., Roth JA, eds.

Washington, D.C.: National Academy Press, 1994; 1–114.

8. Quane JM, Rankin BH. Neighborhood poverty, family characteristics, and commitment to mainstream goals: The case of african american adolescents in the inner-city. Journal of Family Issues 1998; 19(6):769–794.

9. Sampson RJ, Laub JH. Life-course desisters? trajectories of crime among delinquent boys followed to age 70. Criminology 2003; 41(3):555–592.

10. Pollock JM, Mullings JL, Crouch BM. Violent women: findings from the texas women inmates study. Journal of Interpersonal Violence 2006; 21(4):485–502.

11. Sampson RJ. Urban black violence: The effect of male joblessness and family disruption. American Journal of Sociology 1987; 93(2):348–382.

12. Goldkamp JS, Gottfredson MR, Weiland D. Pretrial drug testing and defend- ant risk. Journal of Criminal Law and Criminology 1990; 81(3):585–652.

13. Smith DA, Polsenberg C. Specifying the relationship between arrestee drug test results and recidivism. Journal of Criminal Law and Criminology 1992; 83(2):364–377.

14. Goldstein PJ, Brownstein HH, Ryan PJ. Drug-related homicide in New York: 1984 and 1988. Crime and Delinquency 1992; 38(4):459–476.

15. Cahalan D, Cisin IH, Crossley HM. American Drinking Practices: A National Study of Drinking Behavior and Attitudes. New Brunswick, NJ:

Rutgers Center for Alcohol Studies, 1969.

16. Clark WB, Midanik L. Alcohol use and alcohol problems among U.S. adults:

Results of the 1979 national survey. In Alcohol Consumption and Related Problems. Alcohol and Health Monograph No. 1., N.I.o.A.A.a. Alcoholism.

ed. Washington, DC: National Institute on Alcohol Abuse and Alcoholism, 1982; 4–13.

17. Kasarda JD. Urban Underclass Database: An Overview and Machine- Readable File Documentation. New York: Social Science Research Council, 1993.

18. Goldstein H. Multilevel Statistical Models. 2nd ed. New York: Halsted Press, 1995.

19. Hedeker D, Gibbons RD. A random-effects ordinal regression model for multilevel analysis. Biometrics 1994; 50:933–944.

20. Bryk AS, Raudenbush SW, Congodon RT. Hierarchical Linear Modeling (HLM). Chicago, IL: Scientific Software International, Inc., 1996.

21. Norstrom T. The impact of alcohol, divorce, and unemployment on suicide:

A multilevel analysis. Social Forces 1995; 74(1):293–314.

22. Wilson WJ. The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy. Chicago, Il: University of Chicago Press, 1987, 254. Crime, Alcohol, and Drugs in 24 U.S. Cities 603