Drug Courts and Net-Widening in U.S. Cities: A Reanalysis Using Propensity Score Matching

Abstract

Although drug courts were intended to reduce the justice system involvement of drug offenders, a recent study found evidence that drug courts were associated with increased (rather than decreased) arrests for minor misdemeanor drug offenses (Lilley, 2017; Walsh, 2011). However, the previous study did not utilize an equivalent comparison group and may have relied on a large sample size to generate findings. The current study tested the robustness of those findings by analyzing only cities with over 50,000 population, including four additional years of data, and utilizing a more equivalent comparison group that was propensity-matched to reduce the possibility that a preexisting difference may have generated higher arrest outcomes among drug court jurisdictions. Net-widening, arrest, and crime challenges and implications for drug court policies and law enforcement roles are also discussed.

Keywords

drug court arrest drug enforcement substance use policy

Introduction

Drug courts were created during a new phase of the “war on drugs” that coincided with a surge of crack cocaine use in U.S. cities and efforts to address the backlog of cases and crowded jail facilities (Bureau of Justice Statistics [BJS], 2012; Huddleston, Freeman-Wilson, & Boone, 2004). Although drug courts were intended to reduce the justice system involvement of drug offenders, a recent study found evidence that the presence of these specialized courts increased (rather than decreased) arrests for minor misdemeanor drug offenses (Lilley, 2017; Walsh, 2011). However, the previous study did not utilize an equivalent comparison group and may have relied on large sample size by including over 6,000 comparison jurisdictions¹ to generate findings. In contrast, the current study tested the robustness of those findings by analyzing only cities with over 50,000 population, including four additional years of data, and utilizing a more equivalent comparison group that was propensity-matched to reduce the possibility that a preexisting difference may have generated higher arrest outcomes among drug court jurisdictions.

Although it is clear from Figure 1 that arrests for drug use and possession increased after 1989 when drug courts were created, nationwide data also indicate that arrests for drug sale and other nondrug offenses declined during the same time period. Consequently, the extent to which drug courts actually facilitated or encouraged these additional arrests remains unclear. The primary objective of this study was to determine whether previous findings were robust to the exclusion of smaller jurisdictions and use of a more equivalent comparison group by conducting a series of longitudinal analyses that controlled for the influence of changes in the number of police officers and enforcement of minor offenses, as well as economic and demographic trends. The outcomes of this study may assist the design of future programs and aid decision makers in their efforts to understand the effects of recent drug policies.

Figure 1.

Nationwide trends in minor drug arrests, crime, and policing, 1990-2006.

Literature Review

Drug court critics have noted that the development of a new type of court to deal specifically with drug-related offenses results in an environment that inherently encourages arrests of drug users by creating a system that depends on a continuous flow of new arrestees (Gross, 2010; National Association of Criminal Defense Lawyers [NACDL], 2009). In addition, there is evidence that some drug court officials directly encouraged law enforcement officers to intensify their focus on drug offenses with one spokesperson noting, “if a cop makes five more arrests” because drug courts are available, that officer has “increased public safety” and is helping the community and families of the user (Hutlock, 2003, p. 20). To what extent did these observations translate into systemic increases in minor drug arrests or net-widening? A review of what is known and the limitations of recent scientific studies involving arrest and crime outcomes are detailed below.

In the scientific literature, the relationship between drug courts and arrest has differed substantially depending on whether the focus was the individual participant or the community at large. During the past 20 years, more than 150 individual studies have been conducted to assess recidivism and arrest outcomes associated with drug court participants (Wilson, Mitchell, & MacKenzie, 2006). The vast majority of these studies have found that individuals who successfully completed the program engaged in lower recidivism as measured by subsequent arrests for both drug and nondrug offenses. Nevertheless, systematic reviews of these studies have noted that most participant-level evaluations have been plagued with selection bias, lack of equivalent comparison, and other methodological problems (Government Accountability Office [GAO], 2005a; Mitchell, Wilson, Eggers, & Mackenzie, 2012).

Evaluations of Participant Outcomes

Many drug court evaluations have compared the outcomes of 12-month program graduates to those who dropped out or were removed for noncompliance (nongraduates; GAO, 2005a). For example, in a review of 30 prior studies, Belenko (1999) found that drug court dropouts were rearrested at rates 4 to 9 times higher than those who completed the program. However, from a net-widening perspective, drug courts must be held accountable for the arrest outcomes of those who drop out of treatment as well as those who successfully complete the program. Historically, a substantial 60% of participants have failed to complete the program according to a nationwide assessment by the U.S. GAO (2005a). In addition, several studies have found that individuals who failed to complete drug court programs were more likely to have extensive criminal histories and were more likely to report serious addiction to hard drugs, such as crack cocaine (Bowers, 2008; Miller & Shutt, 2001; Schiff & Terry, 1997; Senjo & Leip, 2001; Wolf, Sowards, & Wolf, 2003). These differences make group comparisons and assessment of overall net-widening outcomes more difficult.

Further complicating participant-level evaluation, California, Florida, and Arizona created laws that forbade incarceration for failing to complete drug treatment (“Florida: Drug Court Judges Get Sanction Authority,” 2006, p. 7; Florida House Bill 175, 2006; Hepburn & Harvey, 2007; Worrall et al., 2009). Thus, participant addicts who dropped out of these programs often remained unsupervised in the community until prior charges were reactivated or new arrests occurred. Consequently, from a broader community perspective, questions arise as to whether subsequent arrests of noncompliant dropouts might have contributed to the number of minor drug arrests in drug court jurisdictions. At present, no participant-level studies have been identified that assessed the overall arrest and community crime impacts of dropouts in drug court jurisdictions.

Another common problem with prior measurement of recidivism and arrests involves the “apples to oranges” comparison of participants who remained under drug court supervision with those who were not under any program restriction. One such study utilized the random assignment of treatment and nontreatment groups but then compared the new arrests of individuals while undergoing frequent drug testing, weekly meetings with program staff, and other restrictions to those who were free in the community, rather than assessing recidivism solely after program completion (Gottfredson, Najaka, & Kearley, 2003). As a result, equivalent comparison was substantially negated with regard to post-program recidivism outcomes. Interestingly, this study also found that after 24 months of drug court involvement, only 19% of participants had completed the drug court program. Other studies have found program dropout rates as high as high as 90% (Hepburn & Harvey, 2007). At present, the extent to which high dropout rates may have resulted in new or supplemental arrests in drug court cities is unknown.

Although program retention rates may have improved somewhat in recent years, measuring the overall arrest impacts of drug courts over time by evaluating individual participant outcomes has proven to be challenging. Procedural and legal differences among jurisdictions make consistent tallying of arrests from removed participants more difficult. In addition, given that some programs may be more strongly associated with net-widening than others and that drug court enrollments and jurisdictional populations differ in size, it may be difficult for assessments of graduates and nongraduates to identify average effects across jurisdictions. Finally, individual-level evaluations have not yet determined whether jurisdictions with drug court programs experienced changes in community perceptions regarding enforcement leniency. If community residents perceived that minor drug offenses would no longer lead to incarceration, increased public substance use, driving while intoxicated, or related behaviors might have occurred, resulting in higher arrest frequency. Fortunately, analyses at higher (macro or community) levels may be more suited to addressing many of these challenging methodological issues.

Context and Background for Macro-Level Analyses

In contrast with studies of individual participants, only a handful of macro-level studies of drug court jurisdictions have been conducted involving arrests, crime, and net-widening. Macro-level studies are designed to assess overall or net outcomes in communities where drug courts are implemented. Most often, aggregate patterns or changes in arrests, crime, or other dependent variables are measured before and after program implementation. However, to improve accuracy, macro-level studies must utilize techniques to control or limit other confounding factors that might influence arrest or crime outcomes, such as changes to employment rates, population demographics, and law enforcement activities. Consequently, macro-level research requires an awareness of drug court program context as well as national trends relating to crime and arrest.

In 1994, Congress allocated funding for drug court implementation after evidence emerged that the first experimental drug court program (created in Miami, 1989) appeared to reduce the backlog of regular criminal trials and the recidivism of participants (Bureau of Justice Assistance [BJA], 1993; GAO, 1997). Within 6 years, the number of funded adult drug courts grew from 14 to more than 350 and an estimated 220,000 individuals had enrolled in the program (Belenko, 2001; Turner et al., 2001). Since that time, drug courts have been implemented in over 1,600 jurisdictions. By 2002, nearly every city with over 100,000 residents was served by least one drug court (Lilley, 2017).

However, during approximately the same time period as the growth of drug court programs, the United States experienced a substantial and sustained decline in crime. More specifically, nationwide crime reached record highs around 1993 before declining to 30-year lows around 2001 (Levitt, 2004). As a result, assessments related to crime and felony arrests during this time period will be inherently biased downward. Thus, macro-level studies need to incorporate control variables and equivalent comparison locations to adequately measure nationwide crime and arrest trends in communities that were not directly influenced by drug court programs.

Macro- and Community-Level Studies

Crime and arrest patterns are inherently related. Consequently, though jurisdictional crime outcomes are not the primary focus of the current analysis, a brief review of crime-related studies is included to provide net-widening context. Three macro-level studies have assessed the relationship between drug courts and crime. Each of these studies utilized fixed effects regression of annual panel data during the 1990s and early 2000s (Lilley, 2013; Orrick, 2005; Zafft, 2014). Overall, findings were inconclusive with one study reporting no relationship (Orrick, 2005); another finding small, positive drug court-crime coefficients (Lilley, 2013); and a third reporting negative relationships with robbery, burglary, auto theft, and other felony offenses (Zafft, 2014). Despite the fact that two of these analyses utilized small convenience samples and one lacked controls for changes in economy, population, police force size, and enforcement activity, these studies provide an indication that drug court programs were not universally associated with crime increases. If the program were positively associated with crime, increases in arrests could be interpreted as artifactual.

The first macro-level study to examine the direct relationship between drug courts and arrests utilized fixed effects panel regressions among 63 drug court jurisdictions from 1990 through 2008 (Zafft, 2014). This study reported a consistent, positive relationship between drug court implementation and arrests for drug offenses. Consistent with net-widening, the positive coefficients were strongest when arrests for misdemeanor use or possession of marijuana were analyzed. However, this study did not include measures of police strength or enforcement activity. These omitted variables are important as the hiring of more officers might endogenously lead to increases in minor arrests. In addition, the sample in this study included only drug court jurisdictions and lacked any nonparticipant cities or counties. Thus, equivalent comparison to minimize the impact of nationwide trends in arrest and crime was not possible.

As noted in the “Introduction” section, a second study also found evidence that the presence of drug court programs was associated with increased arrests for minor misdemeanor drug offenses (Lilley, 2017). This study utilized fixed effects regression to analyze over 8,000 jurisdictions from 1990 to 2002 and included controls for policing, economic, and demographic population changes. Although every drug court city and county nationwide was included along with nonparticipant jurisdictions with populations over 10,000, no effort was made to create a matched or equivalent comparison group. Consequently, as the vast majority of nonparticipant jurisdictions (more than 5,000 of the 8,000 included) were small and exhibited lower starting and ending drug arrest counts, results might have been skewed toward a positive net-widening finding in drug court communities.

Need for the Current Study

Given that drug courts were intended to reduce the justice system involvement of drug offenders, recent macro-level findings of increased (rather than decreased) arrests for minor misdemeanor drug offenses are troubling (Lilley, 2017; Walsh, 2011; Zafft, 2014). During the past 20 years, participant-level studies have consistently found that post-program recidivism and arrests of graduates were reduced while arrests of nongraduate dropouts remained high (Mitchell et al., 2012). However, participant-level studies have not yet assessed broader program impacts on community resident perceptions or overall net-widening changes in jurisdictional arrest patterns.

More recently, a few macro-level, panel data studies analyzed total or net community arrest outcomes finding that drug court jurisdictions experienced higher numbers of misdemeanor drug arrests after implementation (Lilley, 2017; Zafft, 2014). The macro-level approach avoids procedural program differences and “apples to oranges” comparisons of participant recidivism (Lilley, 2013). Nevertheless, new methodological challenges are found in macro-level studies pertaining to sample size, control variables, and equivalent jurisdictional comparison. Given that the nationwide trend in arrests for minor drug possession increased among both drug court and nonparticipating jurisdictions while serious crime trended downward, methodologies that analyze more similar or equivalent comparison groups are needed.

This study will improve upon prior work by using a propensity stratification technique to generate two comparison groups that are more balanced with regard to sample size and more statistically similar with regard to economic, demographic, policing, and arrest patterns. In addition, by removing the potentially biasing influence of thousands of very small (mostly nonparticipating) locations, this study will test the robustness of previous findings by determining whether regression significance values were artifactually related to large sample size. Did drug courts in mid- to large-sized cities intentionally or unintentionally facilitate increases in minor arrests following implementation? This study will help to answer that question by conducting a series of more rigorous regression comparisons.

Method

The primary analyses in this study involved a series of weighted panel data regressions to assess changes in annual rates of arrest for drug use and possession as predicted by drug court implementation among U.S. cities with populations over 50,000 during the years 1990-2006.² All panel data regressions included fixed effects for both jurisdiction and year to measure changes within each jurisdiction over time³ while controlling for changes in policing, economic, and demographic conditions (Allison, 2005; Frees, 2004). This methodology has recently been utilized by researchers in analyses of the effects of police force size on crime (Marvel & Moody, 1996) and analyses of the effects of federal anti-crime programs (GAO, 2005b; Lilley & Boba, 2008; Worrall, 2008; Zhao, Scheider, & Thurman, 2002).

To improve on previous work and reduce the possibility that drug court cities were inherently different from those without a drug court, this study utilized a propensity score stratification technique (Lunceford & Davidian, 2004; Rosenbaum & Rubin, 1984). Conceptually, the purpose of generating a predicted probability or score was to identify cities with attributes that were statistically similar to drug court cities for use in subsequent comparative regressions. Thus, a logistic regression was implemented to predict the probability that each city would be served by a drug court during the 1996-1997 time period. As indicated in Table 1, these propensity years were selected because they represented the middle of the primary wave of drug court implementations across the nation and were the first years that provided a sufficient sample size (e.g., more than n = 200 drug court cities) from which to model drug court participation. In addition, the number of new drug court cities dropped dramatically (n = 12 new courts) during the 2002-2004 period, making analysis of comparative pre–post change less effective after 2006.⁴ As shown in Table 2, mid- to large-sized cities that were most likely to implement a drug court had higher rates of drug arrests as indicated by the odds ratio (OR; 2.85) but fewer disorder arrests (OR = 0.90). They also had higher proportions of non-White residents (OR = 409.82) with higher per capita income (OR = 40.14) and fewer police officers (OR = 0.58) when compared with cities that were less likely to implement drug courts. Because propensity models cannot measure all relevant city attributes and trends, cities with drug court implementation propensity scores below 0.50 were removed to reduce the likelihood that changes in arrests might arise from preexisting differences among drug court and nondrug court cities. The chi-square value (151.71) for the overall logit model was strongly significant, and 77.4% of the 568 mid- to large-sized cities were correctly classified. Propensity stratification resulted in 372 remaining cities, from which 115 never had a drug court as matched with 257 similar cities that had at least one functioning drug court by 1997.

Table 1.

Cities (Over 50,000) Served by Municipal or County Drug Courts (1990-2006).

Time period	New drug court	Cumulative drug court
1990-1992	35	35
1993-1995	108	143
1996-1998	182	325
1999-2001	99	424
2002-2006	12	436

Note. From 1990-2006, 279 cities over 50,000 population did not have a drug court.

Table 2.

Logit Model for the Development of Propensity Scores.

	Coefficient	SE	Wald	Significance	Odds ratio
Ln (drug arrest rate)	0.73	0.16	20.59	0.000	2.85
Ln (disorder arrest rate)	−0.34	0.12	7.91	0.004	0.90
Ln (officer rate)	−1.24	0.36	12.16	0.001	0.58
Ln (per capita income)	2.49	0.61	16.61	0.000	40.14
Non-White (%)	4.35	0.85	26.34	0.000	409.82
Ln employment rate	−0.45	0.45	1.02	0.314	1.53
Constant	−18.80	6.04	9.68	0.002

Note. During the 1996-1997 time period, 289 cities over 50,000 population had a functioning drug court, and 279 had no drug court. N = 568 annual observations; model chi-square = 151.71, p < .001; correctly classified = 77.4%.

Jurisdictional Arrest and Officer Data

Annual data pertaining to arrests and officers per capita from 1990 to 2006 originated from two databases produced by the Federal Bureau of Investigation (FBI). Jurisdictions that provided a full 12 months of arrest data during at least 8 years of the 1990-2006 time period were included in analyses. In addition, state police as well as campus, airport, and other special police agencies were removed because their jurisdictions overlap with other law enforcement agencies that have primary responsibility for policing these populations. Prior to propensity matching, the 17-year unbalanced panel comprised 568 cities with over 50,000 population and nearly 45% of officially reported arrest data in the United States.

The number of sworn officers in each jurisdictional agency is contained in the FBI’s Police Employee Master Files. These annual data include jurisdictional population and were utilized to create a measure of police force size that is expressed as the number of officers per 100,000 residents (officer rate). In addition, annual arrest data in each jurisdiction were obtained for 37 types of felonies and misdemeanors through the FBI’s Arrest Master Files (FBI, 2002).

Two arrest indices were created from the FBI Master Arrest Files to test whether drug courts were associated with changes in arrests for disorder and drug use in each jurisdiction. Both indices included only misdemeanor arrests. These types of misdemeanor arrests are indicators of police effort to address minor disorder and drug activity within the jurisdiction. The amount of enforcement effort can change over time as a result of internal processes such as changes in management as well as external influences related to anti-crime program funding and drug courts.

The first index, “disorder arrests,” was created by summing arrests for disorderly conduct, drunkenness, prostitution, curfew violation, runaways, and vagrancy. The second index, “drug use and possession arrests,” was created by summing arrests for drug abuse and minor drug possession. Possession of a large quantity of drugs is a felony offense that was excluded from the analyses so that the index reflected only personal drug use. Annual arrest totals for each jurisdiction were converted into rates per 100,000 residents.

Drug Court Implementation

Data pertaining to the timing and location of each drug court were requested from the National Drug Court Program Office (NDCPO). These data contain jurisdiction location, court name, starting date, ending date, and basic information about the type of court (e.g., adult, juvenile, city or county government). The accuracy and completeness of drug court operation dates was verified via data that were obtained from American University Drug Court Clearinghouse and the U.S. GAO⁵ (1997; Office of Justice Programs [OJP], 2001). For the panel data models, a dummy variable was created based on each drug court start date such that a positive or non-zero value indicated that the court was active. Many drug courts serve more than one city or police jurisdiction. In addition, many jurisdictions contain more than one drug court. Consequently, a binary dummy code value of “1” indicated that the jurisdiction was being served by at least one drug court. Drug court implementation data were linked with arrest rates and control variables via Federal Information Processing Standard (FIPS) codes that indicate state, county, and jurisdiction location.

Economic and Demographic Control Variables

Demographic and employment variables were included in analyses to control for contextual changes that occurred annually between 1990 and 2006 that may have affected crime rates. Annual county-level employment rates and per capita income were obtained from the Bureau of Economic Analysis. Annual changes in the percentage of population aged 15 to 24 years (population at risk for offending) and percentage of the population that was non-White (a proxy measure for disadvantage) were obtained from the U.S. Census intercensal estimates database and the National Center for Health Statistics (2006).

Panel Data Regression Procedures

There is no single econometric model that can seamlessly assess the relationship between drug court implementation and arrest rates without limitation. Consequently, to ensure that a thorough understanding of program impacts was obtained, a variety of alternative models were tested. For example, regressions that were weighted by city population, square root of population, and without weights were tested in addition to models that contained variables that were both log transformed and untransformed. Log transformation of variables facilitates comparison of the relative size of anti-crime program outcomes while minimizing outlying data. To affect logarithmic transformation, the formula log (1 + x) was utilized to address zero values. This approach has become the standard in crime policy studies (Wooldridge, 2000).

Regression analyses of crime and arrest rates frequently include heteroskedastic error. That is, regression estimates are more precise among jurisdictions with larger populations and less precise for smaller jurisdictions. An in-depth analysis of this issue was conducted by Hannon and Knapp (2003) who recommended that regressions be weighted by population and standard errors be adjusted through the implementation of a heteroskedasticity-consistent covariance matrix (HCCM) method. Consequently, all panel data regressions in this study utilized heteroskedasticity-robust standard errors and most were weighted by jurisdiction population or another appropriate value.⁶ This weighting method also produces more accurate nationwide estimates of program outcomes. Similar methods have been utilized in a variety of recent panel data studies pertaining to crime (GAO, 2005b; Levitt, 2002; Zhao et al., 2002).

Descriptive Statistics

Tables 1 and 3 list descriptive statistics pertaining to the predictor and outcome variables in this analysis. Between 1990 and 2006, 436 cities with over 50,000 population were represented by an active drug court program (Table 1). Most drug courts began operation after 1993 with the largest wave of drug courts (n = 182) beginning between 1996 and 1998. With regard to demographic and economic control variables (Table 3), the average proportion of population that was non-White among the 372 propensity-matched jurisdictions was 31% and the proportion of youth (ages 15-24 years) was 12% from 1990 to 2006. The mean per capita income was US$30,468 during the 17-year period, and the average rate of employment for residents of all ages in reporting jurisdictions was 53.27%. The law enforcement control variables indicate that the average jurisdiction employed about 187 officers per 100,000 residents, and the average number of annual arrests for disorderly conduct was 750 per 100,000 residents.

Table 3.

Descriptive Statistics for Dependent and Control Variables (1990-2006).

Variable	M	SD
Economic and demographic control variables
Officer rate	187.07	0.961
Disorder arrest rate	750.13	9.231
Non-White, %	0.31	0.001
Age 15-24, %	0.12	0.001
Per capita income	30,468	124.858
Employment rate	53.27	0.184
Drug use and possession arrest rate	526.85	4.837

Note. All rates pertaining to officers and crime are per 100,000 residents. Employment rate is per 100 residents. N = 6,002 annual observations among 372 propensity-matched cities.

The value of the dependent variable, misdemeanor drug use, and possession arrests averaged nearly 526 per 100,000 residents during the 1990-2006 time period. However, the drug arrest trend line in Figure 1 indicates that misdemeanor drug enforcement activity increased by more than 40% among all propensity-matched jurisdictions during the first 5 years of the decade. After a slight recession in 2002, drug use arrests again surged to record highs in 2006.

Figure 2 shows the individual trend lines for drug use arrests among jurisdictions that implemented drug courts (primarily after 1993) with propensity-matched cities that were never served by a drug court. Although arrests among drug court jurisdictions (~500 arrests per 100,000 residents) were already higher than nonparticipating jurisdictions (~325 arrests per 100,000 residents) during the initial years, the trends began to diverge around 1996. Throughout the remainder of the 1990s, arrests among drug court jurisdictions rapidly increased. In contrast, the trend line among nonparticipating jurisdictions also increased but remained flatter and more consistent.

Figure 2.

Nationwide trends in drug arrest rates per 10,000 residents, 1990-2006.

Results

During the 1990s, misdemeanor drug arrest activity increased by more than 40% among cities with over 50,000 population as drug courts were implemented in jurisdictions across the nation (Figure 1). Although these increases in drug possession arrests coincided with increases in the number of police officers, it is clear that arrests do not always follow police force size. During the same time period, other types of enforcement activity related to disorder and drug sales declined. These trends suggest that local police may have consciously chosen to increase arrests for misdemeanor drug use and possession offenses rather than focusing on sales or other offenses. If law enforcement activity simply provided a mirror reflection of criminal activity, for example, it would be difficult to understand how drug sales could decline while drug use substantially increased.

A series of panel data regressions were conducted to determine whether there was a relationship between drug court implementation and subsequent increases in minor drug arrests while controlling for employment, per capita income, demographic changes, and police presence. Regression models employed weighted city-level fixed effect analyses with controls for nationwide changes in drug arrests via the yearly trend variable. The initial baseline model (Table 4) included a measure of police force size while controlling for the amount of police enforcement activity toward minor offenses by measuring changes in minor nondrug (disorderly conduct) arrests as well as demographic and economic controls.

Table 4.

Primary Model of Drug Court Impacts on Drug Possession Arrests (1990-2006).

	Coefficient	Robust SE	t	Significance
Drug courts program	70.09	22.81	3.07	0.002
Ln (disorder arrest rate)	129.36	19.18	6.74	0.000
Ln (officer rate)	584.01	203.77	2.87	0.004
Ln (per capita income)	374.75	369.52	1.01	0.311
Age: 15 to 24 (%)	2,883.94	2,866.03	1.01	0.315
Non-White (%)	−937.14	618.50	−1.52	0.131
Ln employment rate	52.84	354.99	0.15	0.882
Year	−0.39	16.75	−0.02	0.981
Constant	−6,823.10	29,770.05	−0.23	0.819

Note. N = 6,002 annual observations (372 cities). R² = .245.

In this primary regression model, the drug court coefficient (70.09) indicated a strong positive relationship (p = .002) between drug court implementation and arrest for drug possession offenses. The coefficient for disorder arrests, a measure of police assertiveness toward other minor offenses (129.36), was also strongly associated with increased drug arrests (p < .001) as was the officer rate variable (584.01). In contrast, none of the economic or demographic control variables were associated with changes in drug possession arrest rates. Per capita income (p = .311) as well as the proportion of the population that was between ages 15 and 24 years (p = .315), and the percentage of the population that was non-White (p = .131) were all nonsignificant as was the employment rate (p = .882) in this regression model. These data indicate that drug court implementation, more police officers, and active police enforcement of minor nondrug offenses were all associated with increases in minor drug arrests.

In terms of effect size, these results indicate that drug court implementation was associated with a 16.8% overall increase in minor drug arrests among cities with over 50,000 population during the 1990-2006 time period. Using the nationwide average 1990 drug arrest rate (414.758 per 100,000 residents) as the starting point to estimate overall impacts, these data indicate that a typical jurisdiction size of 100,000 experienced an average increase of approximately 70 misdemeanor drug arrests (an annual jurisdiction-wide increase from 415 to 485) during each year that the drug court was active. By way of comparison, a 10% increase in officer force strength (approximately 15 officers) in a city of size 100,000 would yield about 55 additional minor drug arrests.

Three alternative models of drug court impacts on misdemeanor drug use and possession arrests were also tested (Table 5). Model 1 is identical to the regression from Table 4 and was included for comparison purposes. Model 2 used the same approach as the initial model but weighted the regression by the square root of the city population to minimize the influence of large cities. The drug court coefficient in this model (41.77) was positive and significant (p = .026) and indicates that the drug court influence was not solely driven by larger jurisdictions. Model 3 followed the same methodology but was weighted by the relative proportion of overall crime that each city contributed within each county. This model provided a better estimate of the impact of drug courts among cities with higher relative crime rates. For example, if a city reported 20% of the total crime in a county during the 1990 to 2006 time period, this city was weighted with a value of 20. However, as overall crime is correlated with population, the drug court coefficient in Model 3 (63.06) was quite similar to the primary model (p = .030). Finally, Model 4 utilized the same approach as the primary regression model but without any weighting of cities. Although this model underestimates the nationwide association of drug courts on arrest among mid- to large-sized cities, it provided a check on the robustness of these propensity-based regressions. The coefficient in Model 4 (22.19) was smaller but remained significant (p = .011).

Table 5.

Alternative Models of the Impact of Drug Court Implementation on Drug Use Arrests.

Alternate models	Coefficient	Robust SE	t	Significance
Model 1(Weighted by population)	70.09	22.81	3.07	0.002
Model 2(Weighted by square root population)	41.77	18.71	2.23	0.026
Model 3(Weighted by county crime percentage)	63.06	28.99	2.18	0.030
Model 4(Unweighted)	22.19	8.72	2.54	0.011

Note. N = 6,002 annual observations (372 cities).

When all four fixed-effect panel model outcomes were analyzed together, they indicated that drug courts were associated with increases in misdemeanor drug use and possession arrests among cities over 50,000 across the nation. Consistent with prior analyses where cities and counties of all sizes were included without propensity matching and with four fewer years of data, the drug court coefficient in Model 1 of this study was indicative of a moderate effect on possession arrests. This coefficient was slightly larger in size than the previous study which indicated that drug courts were associated with a 12% to 15% increase in possession arrests (Lilley, 2017). The significance of the drug court coefficient persisted across a variety of specifications with controls for demographic and economic influences as well as measures of police force size and enforcement activity. These models provide additional evidence that drug court implementation was associated with subsequent increases in arrests for misdemeanor drug use and possession, even among propensity-matched cities with over 50,000 residents across the nation.

Limitations

The fixed-effect regressions in this study allowed pre–post comparisons among drug court jurisdictions while including nonparticipating jurisdictions as a control for the overall nationwide trend in drug arrests. As such, the relative change in arrest rates was compared both within drug court jurisdictions (before and after) and between these jurisdictions and those that were never served by a drug court. However, given that drug courts were not implemented using experimental design, it is not possible to definitively differentiate causation from correlation. Nevertheless, this study utilized a drug court propensity score to improve upon prior work by selecting comparison cities that were similar to drug court cities in terms of demographics, drug arrests, police officers, and economic factors. In addition, by adding four more years of drug court and arrest data (2003-2006), this study improved upon prior work by providing a more comprehensive and robust analysis.

Discussion

Consistent with prior work, the findings of this study suggest that mid- to large-sized cities with drug courts substantially increased arrests for misdemeanor drug use and possession during the 1990-2006 time period. Using a more rigorous methodology to generate propensity-matched cities of similar size, demographic, economic, and law enforcement circumstances, drug courts were associated with a 16.8% increase in minor drug arrests during every year of operation. Four alternative regression models confirmed the robustness of the positive drug court–arrest relationship. Consequently, these findings raise a series of questions about the mechanisms behind increased misdemeanor arrests in drug court cities. Why did increased arrest occur when a goal of the program was to reduce criminal justice involvement of drug users? (Walsh, 2011). Could minor arrest increases have resulted from court-influenced actions by law enforcement or indirectly as an artifact of targeted resources? Are more arrests good or bad for community residents? Each of these questions is discussed below, followed by policy recommendations for improvements to the current drug court system.

Direct Effects of Law Enforcement Net-Widening

Could minor arrest increases have resulted from court-influenced actions by law enforcement officers? It is clear from the trends lines in Figures 1 and 2 that police in both drug court and nondrug court jurisdictions intentionally and selectively focused on drug use offenses while reducing enforcement of drug sale and disorderly conduct offenses. In an era that has been characterized as utilizing “broken windows” and “zero tolerance” policing methods, both groups of cities increased minor drug use arrests by around 40% (Gaston, 2016; Green, 1999). Consequently, drug court programming could not have been responsible for the vast majority of this change in focus.

Nevertheless, there is evidence from a series of recent studies and reports that the long-term dedication of resources toward relatively minor offenders may have exacerbated the nationwide focus on minor drug offenses, resulting in an unintended net-widening side effect (Brook, 2010; Gross, 2010 Hoffman, 2000, 2001; King & Pasquarella, 2009; NACDL, 2009). According to one report, the availability of drug courts has resulted in custodial and felony-level charges “for $10 and $20 hand-to-hand drug cases” that would not previously have advanced (Gross, 2010, p. 167; NACDL, 2009). Another researcher in an urban neighborhood observed that after drug courts were created “police began arresting many destitute people,” advising them to plead guilty to gain entry into the program (Brook, 2010, p. 26). In addition, according to a local judge, in the year following the creation of a drug court in Denver, the number of minor drug charges nearly tripled (Hoffman, 2000).

Despite these criticisms, it could be argued that the increased focus on drug use offenses was effective because it coincided with one of the largest and most sustained crime reductions in our nation’s history (Lilley, 2013). However, this argument may prove to be problematic if the majority of arrestees consisted of nonaddicted or recreational drug users. Although numerous studies have documented high levels of criminal involvement among seriously dependent drug users (Anglin & Speckart, 1988; Boyum & Kleiman, 2003; Goldstein, Bellucci, Spunt, & Miller, 1991; Inciardi & Pottieger, 1998), the vast majority of casual or infrequent drug users are not involved in serious crime (Boyum & Kleiman, 2003; Lipton & Johnson, 1998; Wei, Loeber, & White, 2004). Consequently, the crime control effectiveness of increased arrests for substance use likely depends upon whether the arrested individuals were frequent crime-involved daily users.

Indirect Effects of Dedicated Resources

Several recent studies have reported that substantial numbers of prior drug court participants were not chemically dependent (DeMatteo, Marlowe, Festinger, & Arabia, 2009; Hoffman, 2002; Shah et al., 2015). For example, it was discovered that 45% of drug court participants scored below the threshold for the mildest addiction on a severity index⁷ and that numerous participants provided drug-negative urine samples during initial screening (DeMatteo et al., 2009; Marlowe, Festinger, & Lee, 2003; Shah et al., 2015). In addition, prior research has found evidence that some drug courts made efforts to avoid clients with more serious drug and crime involvement while encouraging participation from drug users with minor or infrequent substance use to improve graduation success rates (Belenko, 1999; DeMatteo et al., 2009; Gross, 2010). Thus, a perceived need for continuous flow of misdemeanor arrestees with infrequent substance use may have been generated by some drug court programs.

Another potential indirect program effect stems from the fact in most programs, the majority of participants did not graduate or were removed for noncompliance (GAO, 2005a). Unfortunately, many of these nongraduates were not immediately incarcerated following removal and generated new arrests for substance use in the months following removal from the drug court program (Lilley, 2013). Were these new arrests of program dropouts substantial enough to account for the measured net-widening in recent studies? The answer is not yet known but merits further inquiry.

Are More Arrests Good or Bad for Community Residents?

If a primary goal of drug courts is to facilitate therapeutic assistance to individuals with substance use problems, an argument could be made that more arrests are better as this would mean that greater numbers of individuals are being helped. Furthermore, as completion of the drug court program generally results in removal of the arrest charge or conviction from the record of the defendant, it may appear that persons captured in the wider net were not harmed by the process. However, from an individual perspective, being arrested and formally charged with a crime is a traumatic experience, even if intentions are benevolent. Furthermore, there is evidence that the arrest and coerced treatment of minor drug offenders can have harmful effects that include increased future drug involvement and higher likelihood of chemical dependency (DeMatteo et al., 2009; Marlowe, 2012).

The 12-month long drug court program is time and resource intensive, which requires frequent contacts with court workers and attendance at group counseling and court disposition hearings. As a result, this process may inherently disadvantage individuals with limited control over work schedule and restricted access to transportation. In some program evaluations, for example, the drug court completion or graduation rate for disadvantaged Black participants has been one third to one half that of Whites (Belenko, 2001; Brook, 2010; Howard, 2016). Moreover, individuals who fail to complete the program often end up in jail, sometimes with longer terms of confinement than traditional adjudication (Gottfredson, Najaka, & Kearley, 2003; Nolan, 2003; Rempel et al., 2003).

At present, our knowledge is limited about the precise mechanisms of net-widening, the extent to which new arrestees consisted of infrequent or nonaddicted substance users, and whether the widened net disproportionately impacted disadvantaged groups. In addition, more work is needed to identify the benefits and harms associated with increased drug use arrests at both individual and community levels of analyses among drug court jurisdictions. The results of this study suggest that drug court officials should attempt to mitigate net-widening outcomes so that the criminal justice involvement of minor offenders is reduced in accordance with program goals.

Policy Recommendations

Despite the noted net-widening setbacks, there is a consensus that drug courts have helped 40% to 60% of participants to reduce both substance use and recidivism (Mitchell et al., 2012). Thus, if drug courts are used to increase the capacity of the justice system to identify and assist individuals with serious chemical dependency, the net benefit to the community could be enhanced. Numerous studies have suggested, for example, that frequent users of cocaine and heroin are more likely to be involved in other types of criminal activity, such as burglary, theft, shoplifting, and “con games” (Inciardi & Pottieger, 1994, 1998). Historically, however, individuals with serious chemical dependency have been more likely to be excluded or removed from drug court programs (Belenko, 1999; Gross, 2010).). Consequently, rather than excluding or removing individuals who are most “at risk” from the drug court program if they are not able to comply with rules or are unable to discontinue substance use on their own, a more rigorous system of intensive in-house treatment and monitoring could be implemented.

On the other end of the substance use spectrum is the casual or recreational user. In light of recent work indicating that many drug court participants are not chemically dependent and provide drug-negative urine samples during initial screening, it is reasonable to conclude that that a substantial portion of individuals captured in the wider enforcement net were infrequent or occasional users (DeMatteo et al., 2009; Shah et al., 2015). Consequently, this subgroup may not benefit from daily group counseling, frequent appearances before judges, or other forms of therapy. Although some drug court programs have begun to implement a separate process for arrestees that are not chemically dependent, the confirmatory findings of this reanalysis suggest that officials should consider universal implementation of an alternative program for this increasingly large participant subgroup.

To minimize net-widening, more direct communication with local law enforcement about drug court referral criteria is also recommended. In jurisdictions where alternative drug court pathways have not yet been created for casual users who are not involved in other forms of crime, police officials should be informed of program limitations. At present, police officers effectively function as the “front door” for admissions pertaining to minor drug offenses. However, officers are not typically experts in determining addiction severity. Thus, if their role is to be altered from mere enforcers of the law to facilitators of therapeutic jurisprudence, further assistance or training may be advisable. For example, drug courts could provide on-call program workers to assist police with evaluation and disposition of individuals who are involved in possession or use of small quantities illicit substances. Alternatively, training sessions could be created to help law enforcement officers differentiate between casual users, those with emerging chemical dependency, and those who rely on criminal activity to support substance use. By increasing coordination with local officials, the drug court system could be further leveraged to enhance public health, safety, and welfare in the broader community.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biographies

David R. Lilley is a criminal justice professor at the University of Toledo and previously worked as a senior justice analyst in Washington, DC. His current research involves quantitative analyses of crime control programs, drug policies, and analysis of national and global crime trends.

Megan C. Stewart is a criminal justice professor at the University of Toledo. She earned her PhD from the University of Cincinnati in criminal justice, with an emphasis in crime prevention and criminology. She previously served as a hotline counselor and rape crisis advocate in Butler County, Ohio. Her current research primarily focuses on violent victimization, violence against women, and stalking.

Kasey Tucker-Gail is a criminal justice professor at the University of Toledo and director of the Center for Student Advocacy and Wellness. She is a member of the Toledo Community Initiative to Reduce Violence (TCIRV), and her recent research includes personnel issues in law enforcement and victimization.

References

Allison

(2005). Fixed effects regression methods for longitudinal data using SAS. Cary, NC: SAS Institute.

Anglin

Speckart

(1988). Narcotics use and crime: A multisample, multimethod analysis. Criminology, 26, 197-233.

Belenko

(1999). Research on drug courts: A critical review 1999 update. National Center on Addiction and Substance Abuse, Columbia University, New York, NY.

Belenko

(2001). Research on drug courts: A critical review (2001 update). New York, NY: National Center on Addiction and Substance Abuse.

Bowers

(2008). Contraindicated drug courts. UCLA Law Review, 55, 783-794.

Boyum

Kleiman

(2003). Breaking the drug-crime link. The Public Interest, 152, 19-38.

Brook

(2010). Drug courts and the treatment of addiction: Therapeutic jurisprudence and neoliberal governance (Doctoral dissertation). Available from ProQuest Dissertations and Theses. (UMI No. 3428034)

Bureau of Justice Assistance. (1993). Assessment of the feasibility of drug night courts (NCJ 142415). Washington, DC: U.S. Department of Justice.

Bureau of Justice Statistics. (2012). Arrest in the United States, 1980-2009 (NCJ 239423). Washington, DC: U.S. Department of Justice.

10.

Conway

Levy

Vanyukov

Chandler

Rutter

Swan

Neale

(2010). Measuring addiction propensity and severity: The need for a new instrument. Drug and Alcohol Dependence, 1, 4-12.

11.

DeMatteo

Marlowe

Festinger

Arabia

(2009). Outcome trajectories in drug court: Do all participants have drug problems? Criminal Justice and Behavior, 36, 354-368.

12.

Federal Bureau of Investigation. (2002). Crime in the United States 2002: Uniform crime reports. Washington, DC: U.S. Department of Justice.

13.

Florida: Drug court judges get sanction authority. (June 30, 2006). Crime Control Digest, 40, 7.

14.

Florida House Bill 175. (2006). Drug court programs. Florida Senate. Retrieved from http://www.flsenate.gov/data/session/2006/House/bills/analysis/pdf/h0175c.JUVJ.pdf

15.

Frees

(2004). Longitudinal and panel data: Analysis and applications in the social sciences. Cambridge, UK: Cambridge University Press.

16.

Gaston

(2016). Race, neighborhood context, and drug enforcement: A mixed-method analysis of racial disparities in drug arrest (Doctoral dissertation). St. Louis: University of Missouri–St. Louis.

17.

General Accounting Office. (1997). Drug courts: Overview of growth, characteristics, and results. Washington, DC: Author.

18.

Goldstein

Bellucci

Spunt

Miller

(1991). Frequency of cocaine use and violence: A comparison between men and women (NIDA Research Monograph No. 110). Washington, DC: U.S. Government Printing Office.

19.

Gottfredson

Najaka

Kearley

(2003). Effectiveness of drug treatment courts: evidence from a randomized trial. Criminology and Public Policy, 2, 171-196.

20.

Government Accountability Office. (2005a). Adult drug courts: Evidence indicates recidivism reductions and mixed results for other outcomes. Washington, DC: Author.

21.

Government Accountability Office. (2005b). Community policing grants: COPS grants were a modest contributor to declines in crime during the1990s. Washington, DC: Author.

22.

Green

(1999). Zero tolerance: A case study of police policies and practices in New York city. Crime & Delinquency, 45, 171-187.

23.

Gross

(2010). The effects of net-widening on minority and indigent drug offenders: A critique of drug courts. University of Maryland Law Journal of Race, Religion, Gender & Class, 10, 161-177.

24.

Hannon

Knapp

(2003). Reassessing nonlinearity in the urban disadvantage/violent crime relationship: An example of methodological bias from log transformation. Criminology, 41, 1427-1448.

25.

Hepburn

Harvey

(2007). The effect of the threat of legal sanction on program retention and completion: Is that why they stay in drug court? Crime & Delinquency, 53, 255-280.

26.

Hoffman

(2000). The drug court scandal. North Carolina Law Review, 78, 1437-1534.

27.

Hoffman

(2002). The Denver drug court and its unintended consequences. In Nolan

(Ed.), Drug courts: In theory and in practice (pp. 67-88). Hawthorne, NY: Aldine de Gruyter.

28.

Howard

(2016). Race, neighborhood, and drug court graduation. Justice Quarterly, 33, 159-184.

29.

Huddleston

Freeman-Wilson

Boone

(2004). Painting the current picture: A national report card on drug courts and other problem solving court programs in the U.S. Alexandria, VA: National Drug Court Institute.

30.

Hutlock

(2003). Addressing concerns about drug courts: Despite political popularity, this treatment method still garners some criticisms. Behavioral Health Management, 23, 16-20.

31.

Inciardi

Pottieger

(1994). Crack-cocaine use and street crime. Journal of Drug Issues, 24, 273-292.

32.

Inciardi

Pottieger

(1998). Drug use and street crime in Miami: An (almost) twenty-year retrospective. Substance Use & Misuse, 33, 1839-1870.

33.

King

Pasquarella

(2009). Drug courts: A review of the evidence. Washington, DC: The Sentencing Project. Retrieved from http://www.sentencingproject.org/doc/dp_drugcourts.pdf

34.

Levitt

(2002). Using electoral cycles in police hiring to estimate the effect of police on crime: Reply. American Economic Review, 92, 1244-1250.

35.

Levitt

(2004). Understanding why crime fell in the 1990s: Four factors that explain the decline and six that do not. Journal of Economic Perspectives, 18, 163-190.

36.

Leymon

(2010). “Fixed” sentencing: The effects on imprisonment rates over time (Master’s dissertation). Eugene: University of Oregon.

37.

Lilley

(2013). Drug courts and community crime rates: A nationwide analysis of jurisdiction-level outcomes. Journal of Criminology, 2013, Article 571760. doi:10.1155/2013/571760

38.

Lilley

(2017). Did drug courts lead to increased arrest and punishment of minor drug offenses? Justice Quarterly, 34, 673-698.

39.

Lilley

Boba

(2008). A comparison of outcomes associated with two key law-enforcement grant programs. Criminal Justice Policy Review, 19, 438-465.

40.

Lipton

Johnson

(1998). Smack, crack and score: Two decades of NIDA-Funded drugs and crime research at NDRI 1974-1994. Substance Use & Misuse, 33, 1779-1815.

41.

Lunceford

J. K.

Davidian

(2004). Stratification and weighting via the propensity score in estimation of causal treatment effects: A comparative study. Statistics in Medicine, 23, 2937-2960.

42.

Marlowe

(2012). Alternative tracks in adult drug courts: Matching your program to the needs of your clients. Alexandria, VA: National Drug Court Institute.

43.

Marlowe

Festinger

Lee

(2003). The role of judicial status hearings in drug court. Offender Substance Abuse Report, 3, 33-46.

44.

Marvel

Moody

(1996). Specification problems, police levels, and crime rates. Criminology, 34, 609-646.

45.

Miller

Shutt

(2001). Considering the need for empirically grounded drug court screening mechanisms. Journal of Drug Issues, 31, 91-106.

46.

Mitchell

Wilson

Eggers

Mackenzie

(2012). Assessing the effectiveness of drug courts on recidivism: A meta-analytic review of traditional and non-traditional drug courts. Journal of Criminal Justice, 40, 60-71.

47.

National Association of Criminal Defense Lawyers. (2009). America’s problem-solving courts: The criminal costs of treatment and the case for reform. Washington, DC: Author.

48.

National Center for Health Statistics. (2006). National vital statistics system: Data files and documentation. Retrieved from http://www.cdc.gov/nchs/about/major/dvs/popbridge/datadoc.htm

49.

National Institute of Justice. (2015). Drug courts. Retrieved from http://www.nij.gov/topics/courts/drug-courts/pages/welcome.aspx

50.

Nolan

(2003). Redefining criminal courts: Problem-solving and the meaning of justice. American Criminal Law Review, 40, 1541-1565.

51.

Office of Justice Programs. (2001). Drug court activity update: Summary information on all programs and detailed information on adult drug courts. Washington, DC: Drug Court Clearinghouse and Technical Assistance Project.

52.

Orrick

(2005). Assessing the city-level impact of drug courts on crime rates, 1985-2000 (Master’s dissertation). College Park: University of Maryland.

53.

Rempel

Fox-Kralstein

Cissner

Cohen

Labriola

Farole

. . . Magnani

(2003). The New York State adult drug court evaluation: Policies, participants, and impacts. New York, NY: Center for Court Innovation.

54.

Rosenbaum

Rubin

(1984). Reducing bias in observational studies using subclassification on the propensity score. Journal of the American Statistical Association, 79, 515-524.

55.

Schiff

Terry

(1997). Predicting graduation from Broward County’s dedicated drug treatment court. The Justice System Journal, 19, 291-310.

56.

Senjo

S. R.

Leip

L. A.

(2001). Testing and developing theory in drug court: A four-part logit model to predict program completion. Criminal Justice Policy Review, 12, 66-87.

57.

Shah

DeMatteo

Keesler

Davis

Heilbrun

Festinger

(2015). Addiction severity index scores and urine drug screens at baseline as predictors of graduation from drug court. Crime & Delinquency, 61, 1257-1277.

58.

Turner

Longshore

Wenzel

Fain

Morral

Deschenes

. . . Taxman

(2001). National evaluation of 14 drug courts. Washington, DC: Rand Corporation.

59.

Walsh

(2011). Addicted to courts: How a growing dependence on drug courts impacts people and communities. Washington, DC: Justice Policy Institute. Retrieved from http://www.justicepolicy.org/research/2217

60.

Wei

Loeber

White

(2004). Teasing apart the developmental associations between alcohol and marijuana use and violence. Journal of Contemporary Criminal Justice, 20, 166-183.

61.

Wilson

D. B.

Mitchell

MacKenzie

D. L.

(2006). A systematic review of drug court effects on recidivism. Journal of Experimental Criminology, 2, 459-487.

62.

Wolf

Sowards

Wolf

(2003). Predicting retention of drug court participants using event history analysis. Journal of Offender Rehabilitation, 37, 139-162.

63.

Wooldridge

(2000). Introductory econometrics: A modern approach. Mason, OH: South-Western College Publishing.

64.

Worrall

(2008). The effects of local law enforcement block grants on serious crime. Criminology & Public Policy, 7, 325-350.

65.

Worrall

Hiromoto

Merritt

Jacobson

Iguchi

(2009). Crime trends and the effect of mandated drug treatment: Evidence from California’s Substance Abuse and Crime Prevention Act. Journal of Criminal Justice, 37, 109-113.

66.

Zafft

K. M.

(2014). Can drug courts improve public safety? Exploring the impacts of drug court on crime (Doctoral dissertation). University of Maryland, College Park.

67.

Zhao

Scheider

Thurman

(2002). Funding community policing to reduce crime: Have cops grants made a difference? Criminology & Public Policy, 2, 7-32.