Although a substantial number of studies have reported that drug courts reduced the recidivism of graduates (Wilson et al., 2006), a series of recent analyses suggested that drug courts and similar programs were associated with unintended crime outcomes in cities and counties across the nation (Lilley and Boba, 2008; Miethe et al., 2000; Peters et al., 2002; Worrall et al., 2009). Given that over 220,000 offenders participated in this alternative to incarceration and most did not successfully complete the drug court program, jurisdictional crime may have been impacted. A series of panel data analyses were conducted among more than 5,000 jurisdictions nationwide from 1995 to 2002 to assess the impact of drug court implementation grants on UCR Part I felony offenses. Consistent with prior findings, drug court implementation grants were associated with
In response to the crack epidemic and soaring crime, congress created the Drug Court Discretionary Grant in 1994 with the goal of expanding this alternative to incarceration to jurisdictions throughout the United States. Within six 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 [
Drug courts were created as an experimental program in Miami, Florida, in 1989 [
Drug court implementation grant dollars to local jurisdictions (1995–2001).
Year | Sum | Mean | Standard deviation |
---|---|---|---|
1995 | 560,750 | 18,667 | 12,220 |
1996 | 981,868 | 42,351 | 51,940 |
1997 | 1,577,550 | 51,272 | 90,421 |
1998 | 6,560,345 | 113,345 | 110,056 |
1999 | 9,780,375 | 101,255 | 90,590 |
2000 | 10,899,559 | 98,300 | 84,037 |
2001 | 8,971,893 | 101,030 | 79,452 |
| |||
All | 39,332,340 | 75,174 | 74,102 |
Payment observations and funding totals include drug court implementation grants from Byrne, LLEB, and the Drug Court Discretionary grants program.
The drug court program was designed to last approximately one year and generally consisted of outpatient treatment [
Despite the best efforts of many drug court counselors and officials, several studies have noted that retention of participants has been problematic. A nationwide assessment by the US Government Accountability Office [
Recent data suggest that more than 40 percent of participants who dropped out of treatment did so within 90 days after entry into the program [
Numerous studies indicate that drug court dropouts are more likely than graduates to report a serious addiction to hard drugs, such as crack cocaine [
Although the personal and criminal history backgrounds of drug court participants are extremely varied, recent nationwide assessments indicated that the vast majority (nearly 74 percent) of drug court participants had a felony conviction prior to admission [
Lastly, it is noteworthy that individual drug addiction and related offending can translate into measurable crime rate impacts in communities. In this regard, one recent study found strong evidence of a spatial link between drugs and violent crime at the neighborhood level [
A number of individual studies and reviews have been conducted to assess recidivism outcomes associated with this program. Systematic reviews and meta-analyses of these studies have noted that most evaluations have been of poor quality, suffer from selection bias, and lack equivalent comparison [
Any rigorous assessment of drug courts should separately assess the
Although prior evaluations have been incomplete, researchers that evaluated more than 100 individual (mostly program sponsored) studies have reached a general consensus that participants who successfully completed the twelve-month program recidivated at lower rates after graduation [
From a cost-benefit perspective, analysis of net, or total drug court effects, including postprogram recidivism among both graduates and nongraduates, is necessary. In this regard, it is possible that recidivism reductions among graduates in some drug courts may have been substantial enough to offset rearrests by nongraduates. Several individual cost-benefit analyses of varying quality have been conducted that report net benefits during treatment [
The Office of Management and Budget recently rated the drug court program as “not performing” because it lacked clear outcome measures pertaining to impact on public safety [
Based on the preponderance of studies of drug court graduates, declines in recidivism among the 220,000 participants could translate into measurable jurisdiction-level crime reductions in locations across the nation that received drug court implementation grants. Thus, one objective of this analysis was to provide a series of tests to determine whether drug court implementation funds were associated with reductions or increases in crime at the jurisdiction level during the 1995–2002 time period. An additional objective was to increase our understanding of the specific offenses that were most impacted and the extent or relative size of these effects. A final objective was to provide important outcome information to policy makers to aid in the design and implementation of future Department of Justice grants and to provide a starting point for criminal justice researchers in their efforts to identify mechanisms of crime reduction that result from federal funding.
This methodology included a series of panel data regressions to assess changes in annual rates of reported crime as predicted by drug court implementation grant funding during the years 1995–2002. As a result, the impact of annual drug court grant funds on burglary, larceny, auto theft, robbery, aggravated assault, rape, and murder was assessed over time among more than 5,000 jurisdictions. All panel data regressions included fixed effects for both jurisdiction and year to measure changes within each jurisdiction over time [
The fixed-effects panel data approach offered a substantial benefit over cross-sectional research in that jurisdictional differences that were stable over time were no longer problematic [
Similarly, trends that affected all agencies during this era were controlled for by including a numeric year fixed effect or variable in the model [
In this study, it was anticipated that drug court funding would correlate with funding from a variety of other justice grant programs that were active in each jurisdiction. Therefore, per capita funding amounts from COPS Hiring and MORE, Local Law Enforcement Block Grants (LLEBG), Weed and Seed, and the State Criminal Alien Assistance Program (SCAAP) were also included in the analyses. The resulting panel data models with controls for other federal grants allowed assessment of relationships over time using a methodology that is superior to studies that focus on a single grant program in isolation.
Annual rates of reported murders, rapes, aggravated assaults, robberies, burglaries, larcenies, and vehicle thefts from each jurisdiction were assessed using Uniform Crime Report (UCR) data that are collected by the Federal Bureau of Investigation [
State police, 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. The UCR program refers to state and special police agencies as “zero-population agencies” [
Data pertaining to annual drug court implementation and other grant fundings were obtained directly from the Office of Justice Programs (OJP) comptroller. (The database utilized in this study was developed by a team of criminologists, economists, and data analysts over a two-year period with federal funding. From 2003 to 2006, federal drug court funding was greatly reduced. However, nearly all courts continued to operate.) More specifically, these data contain annual payment amounts to all recipients of Justice grants from 1990 to 2001. Consequently, these data provided more precise and accurate information pertaining to the distribution of funds than prior studies of justice grants that used award data [
In addition to the grant funding independent variables, demographic and employment variables were included in this analysis to control for contextual changes that occurred annually during the 1990s 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 that was aged fifteen to twenty-four (population at risk for offending) and percentage of the population that was nonwhite (a proxy measure for disadvantage) were obtained from the US Census intercensal estimates database and the National Center for Health Statistics [
There is no single econometric model that can perfectly assess the relationship between drug court implementation funding and crime rates without limitation. Consequently, to ensure that a thorough understanding of these grant impacts was obtained, a variety of alternative models were tested. For example, regressions with and without lagged-dependent variables were tested as well as log transformed and untransformed specifications. Additionally, it was noted that drug court grants were first implemented at a time when crime was trending downward from peak levels in the early 1990s in many jurisdictions. Thus, the years prior to this downward trend (1990–1993) were included in one regression model.
To facilitate comparison of the relative size of grant outcome coefficients while minimizing outlying data, all variables in the primary models were log transformed. The resulting log-log models also allowed interpretation of effects as elasticities—the percent change in the outcome that is expected from a 1 percent in change in grant funding [
Prior analyses of justice grant funding distribution has suggested that a delay in expected outcomes of approximately one year from the time of the initial award is common due to program implementation requirements and federal payment distribution procedures [
Given that no single, reliable, and comprehensive list of all drug courts since the beginning of the program currently exists (partial lists of drug courts were obtained from the American University Drug Court Clearinghouse, the National Drug Court Program Office, and the US Government Accountability Office), drug court grants were utilized as a proxy measure for both the timing of drug court startup and relative program size. When compared with available lists of individual drug courts, per capita drug court funding was found to be significantly correlated with counts of total program participants (
There is a possibility that justice grant funding is sometimes influenced by recent changes in local crime rates [
Thirdly, to address longer-term crime trends that may have influenced grant acquisition, data from five years prior to the distribution of drug court implementation grants (1990–1994) were included in some models. Lastly, to address the possibility of “regression to the mean” or a longer-term endogenous relationship, growth quartiles were calculated for all jurisdictions by comparing the change in crime between 1985 and 1992. These years were chosen because they represent the start and end points for the crime increase that occurred in many agencies as indicated by national trends [
Regression analyses of crime 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 [
Tables
Descriptive statistics for independent and control variables (1995–2002).
Variable | Mean | Standard deviation |
---|---|---|
Independent variables | ||
Drug court grants (startup, Byrne, LLEB) | 0.02 | 0.24 |
COPS Hiring grant | 1.61 | 2.86 |
COPS MORE grant | 0.44 | 1.56 |
Weed and Seed grant | 0.09 | 0.55 |
State Criminal Alien Assistance grant | 0.24 | 4.94 |
Local Law Enforcement Block Grant | 1.14 | 4.05 |
Control variables | ||
Pct. nonwhite | 0.19 | 0.14 |
Pct. age 15–24 | 0.14 | 0.03 |
Per capita income | 29,312 | 11,777 |
Employment rate | 0.64 | 0.58 |
Independent variables are expressed in dollar amounts per capita.
Descriptive statistics for dependent variables (1995–2002).
Variable | Mean | Standard deviation |
---|---|---|
Dependent variables | ||
Murder rate | 6.9 | 9.2 |
Robbery rate | 190.8 | 234.8 |
Aggravated assault rate | 375.2 | 344.0 |
Rape rate | 34.4 | 28.1 |
Burglary rate | 866.4 | 544.3 |
Larceny rate | 2757.4 | 1695.2 |
Vehicle theft rate | 502.9 | 485.4 |
With regard to demographic control variables, the average proportion of population that was nonwhite among the 5,172 jurisdictions was 19 percent and the proportion of youth (ages 15–24) was 14 percent from 1995 to 2002. The mean per capita income was slightly over $29,000 during the eight-year period, and the average rate of employment for citizens of all ages in reporting jurisdictions was 64 percent.
Table
Drug court implementation grant funding to cities and counties nationwide was associated with
The estimated impact of the drug court and other justice grants on individual index crimes (1995–2002).
Variable | Murder | Robbery | Aggrav. assault | Rape | Burglary | Larceny | Vehicle theft |
---|---|---|---|---|---|---|---|
Drug courts | 0.071 (1.50) |
|
|
|
|
|
|
COPS Hiring | − |
− |
−0.004 (0.59 ) | 0.006 (0.069) |
|
−0.009 (1.68) |
|
COPS MORE | − |
−0.091 (6.32) | − |
− |
|
|
|
Weed and Seed | −0.027 (1.27) | − |
− |
− |
|
0.007 (1.15) | −0.005 (0.45) |
SCAAP | − |
− |
− |
− |
|
|
0.026 (1.78) |
Law Enforc. Block Grant | − |
− |
− |
− |
|
|
|
Per capita income | − |
− |
− |
− |
|
|
|
Age 15−24 |
|
|
|
|
|
−0.344 (0.87) |
|
Nonwhite |
|
−0.197 (0.439) | −0.944 (1.31) | − |
−0.834 (1.53) |
|
−0.165 (0.39) |
Employment rate | 0.146 (0.38) | −0.162 (0.41) | 0.354 (1.33) |
|
0.122 (0.61) | 0.112 (1.55) | 0.628 (1.49) |
Crime growth cells | − |
− |
− |
− |
|
|
|
Dependent var. (lagged) | − |
− |
− |
|
|
|
|
Constant | −0.066 (1.78) | −0.109 (2.90) | −0.084 (2.25) | −0.109 (2.54) | −0.035 (2.11) | 0.016 (1.37) | −0.062 (2.07) |
|
0.057 | 0.137 | 0.086 | 0.044 | 0.364 | 0.234 | 0.142 |
Coefficients that are significant at the 0.05 level are displayed in bold. Robust
Demographic and economic control variables in the model yielded, for the most part, expected results. Per capita income was negatively associated with burglary (−0.882), larceny (−0.571), and all other crime types. The percentage of the population that was young (ages 15 to 24) was positively associated aggravated assault (4.027), rape (5.152), burglary (4.185), and vehicle theft (9.245). The variables representing percent employed and percent nonwhite yielded mostly mixed and nonsignificant coefficients. Lastly, the control for prior crime growth from 1985 to 1992 indicated that jurisdictions with higher previous crime growth experienced greater crime declines for all offenses except burglary (0.001) and larceny (0.002).
With regard to effect size, the largest and most robust drug court coefficient was found for vehicle theft (0.239). When interpreted as an elasticity, this coefficient indicates that a 1 percent increase in drug court implementation funds was associated with a 0.24 percent increase in vehicle theft. Stated another way, a 10 percent increase in drug court funding was associated with approximately 13 additional vehicle thefts per 100,000 citizens among recipient jurisdictions. (To determine the effect size, a rate of 550 vehicle thefts per 100,000 from the baseline year (1995) was utilized. Using the mean annual amount drug of court funding ($75,174), a 1 percent increase in funding would be approximately $752 per year in recipient jurisdictions. The 10 percent increase figure assumes a linear relationship between funding and crime.) Drug court implementation grants were also associated with small increases in rape (0.112), robbery (0.079), and aggravated assault (0.069). It was noted, however, that the association between drug court funding and robbery and aggravated assault was less consistent in the alternative models that are detailed below.
Models 1–8 are alternative specifications of the model associated with Table Model 1: primary model (same as shown in Table Model 2: includes drug court funds with no other grants as control variables (1995–2002), Model 3: no lagged dependent variables included as predictors (1995–2002), Model 4: same as primary model but includes 13 years of data without prior crime growth cells (1990–2002), Model 5: includes only drug court jurisdictions (1990–2002), Model 6: annual drug court grant modeled as binary (received grant = 1) among jurisdictions over 50,000 population (1995–2002), Model 7: annual drug court grant models weighted by square root of population (1995–2002), Model 8: includes two lags of the dependent variable, weighted by square root of population (1995–2002).
During the analysis, seven alternative models were tested to ensure that findings were not solely due to one specific methodology (Table
Alternative models of the impact of drug court implementation grants on individual index crimes.
Alternate models | Murder | Robbery | Aggrav. assault | Rape | Burglary | Larceny | Vehicle theft | Observations |
---|---|---|---|---|---|---|---|---|
Model 1 | 0.071 (1.50) |
|
|
|
|
|
|
31,041 |
Model 2 | 0.025 (0.49) | 0.031 (0.84) | 0.042 (1.20) |
|
|
|
|
31,041 |
Model 3 | 0.070 (1.39) |
|
0.067 (1.80) |
|
|
|
|
31,041 |
Model 4 | 0.085 (1.74) |
|
|
|
|
|
|
46,366 |
Model 5 | 0.096 (1.81) |
|
0.002 (0.05) |
|
|
|
|
1,680 |
Model 6 | 0.020 (0.54) |
|
|
|
|
|
|
6,951 |
Model 7 | 0.061 (1.19) | 0.067 (1.83) | 0.051 (1.56) |
|
|
|
|
31,041 |
Model 8 | 0.033 (0.58) | 0.017 (0.38) | 0.046 (1.44) | 0.090 (1.75) |
|
|
|
31,041 |
Coefficients that are significant at the 0.05 level are displayed in bold. Robust
Jurisdiction and year fixed effects were included along with control variables from Table
Models 6–8 were conducted on a post hoc basis to provide additional tests of sensitivity and robustness to modeling changes. To determine whether the initial results could be artifacts of per capita grant dollars or large sample size, drug court grant dollars were recoded as a binary variable such that recipients of annual drug court funds in a given year were coded as “1” (see Table
When all models were analyzed together, a positive relationship between drug court implementation grants and crime was found in all models of funding dollars and crime for burglary, larceny, and vehicle theft. Additionally, robbery, aggravated assault, and rape showed positive coefficients that occasionally reached statistical significance. These findings were robust to changes in the years included in analyses, altering or removing different variables, and regressions of drug court jurisdictions separately. Overall, these regressions suggest that the positive relationship between
The drug court implementation grant program was the only major anticrime initiative that was consistently associated with crime
Were noncompliant drug court dropouts efficiently and immediately incarcerated after removal from the program and is there evidence that they actually committed crimes in substantial numbers after removal from the program? To help answer these questions, data from a two-year follow-up study of discharged participants in Southeast County drug court from 2002 to 2004 were analyzed. (This previous study compared recidivism among graduates and nongraduates within the home county only and thus, reflects an undercount of total recidivism. The two year study took place in a large, well-established drug court in a Southeastern State and was supervised by the first author as part of a separate project.) Of the 1,118 participants, 60 percent (677) did not successfully graduate. A substantial majority of these noncompliant dropouts (408) generated new arrest charges within the county jurisdiction after removal from the program. Additionally, 63 percent of the 1,841 new arrest charges were not drug or alcohol related and nearly one-half of these new charges were felonies. Most importantly, 52 percent of subsequent arrests (380 total) among drug court dropouts occurred within the first four months of removal from the program. The timing of all 837 postdrug court arrests among graduates and nongraduates is illustrated in Figure
Number of arrests per month after drug court (graduates versus dropouts).
To rule out the possibility that drug court jurisdictions experienced different preexisting crime trends, Figure
Index crime trends among drug court and other jurisdictions (1990–2002).
Given that funds for COPS grants, Weed and Seed, and other anticrime programs were not assigned using experimental design, it is not possible to definitively differentiate causation from correlation. Similarly, the extent to which drug court implementation grants may have acted as a sole cause of crime increases is not entirely clear in this initial assessment. Nevertheless, the key finding in this study is that drug court grants were the
Drug court grants were developed to aid the expansion of alternatives to incarceration for individuals with drug or alcohol addictions that were charged with nonviolent felonies or misdemeanors. Given the general consensus among researchers that drug courts reduced recidivism among many of the 220,000 nationwide participants [
There are a number of potential explanations for the finding that drug court grants were associated with jurisdictional crime increases. Chief among them is that a substantial majority of individuals who entered the drug court system during the 1995–2002 time period did not successfully complete the program [
Several studies have found that individuals who failed to complete drug court programs were more likely to have extensive criminal histories [
The possibility that active drug court participants may have continued to commit offenses during treatment must also be considered. In this regard, for example, one prior investigation found that drug court-approved halfway houses were being used for narcotics trafficking and prostitution [
Most drug courts were originally conducted as pretrial diversionary programs. Thus, although participants had been arrested and charged with offenses such as larceny, burglary, and auto theft, they often had not been convicted of these new charges when they entered the drug court system. As a result, the legal basis for immediate incarceration of noncompliant drug court participants was unclear and subsequently challenged in court [
In the current study, property offense increases such as vehicle theft, larceny, and burglary were most strongly associated with drug court implementation grants. This finding is consistent with prior research indicating that illegal drug use increases the instrumental needs of addicts, in part, because intoxication may lower their ability to perform at work or school [
Drug users are also at increased risk of victimization due to vulnerabilities that result from the process of obtaining, ingesting, and coping with pharmacological effects. One study, for example, indicated that 11 percent of regular drug users were victims of forcible sex crimes [
Further research is needed to identify the reasons that the drug court grant program was the only one of six Department of Justice grants that consistently demonstrated positive associations with crime in this analysis. Additional studies that separately analyze the outcomes of drug courts that immediately incarcerate noncompliant dropouts from courts that do not utilize this approach are also needed. More rigorous evaluations that include experimental design and cost benefit analyses of postprogram outcomes among both graduates and nongraduate dropouts would also be beneficial.
Numerous recent studies suggest that drug courts reduce the recidivism of individuals who successfully complete the program. Nothing in the current analysis contradicts that finding. What this first ever nationwide jurisdiction-level study does point out, however, is that the process or implementation of drug court programs from 1995 to 2002 may have led to unintended consequences for communities by resulting in small to moderate crime rate increases. Regardless of whether this positive crime association directly resulted from noncompliant individuals, drug court dropouts, or another related factor, it is clear that programmatic adjustments could improve the effectiveness of the drug court program with regard to community and jurisdiction-level outcomes.
Participants with extensive criminal histories or more serious addiction problems are at increased risk of failure and should be identified at the outset so that additional assistance can be provided. Additionally, drug court procedures should be designed to more quickly identify noncompliant individuals and ensure that they do not remain unsupervised and unaccountable in the community for extended periods of time (including those who are removed from the program). Moreover, the legal authority of drug court judges to incarcerate noncompliant individuals should be clearly specified so that persons who are diverted prior to trial for burglary, vehicle theft, or other offenses cannot find a loophole in the process. Alternatively, drug courts could be limited strictly to postconviction individuals.
Drug courts substantially differ with regard to how closely participants are monitored, responses to noncompliance, and handling of those who are removed from the program. At present, much remains unclear regarding the specific mechanisms by which drug court implementation funding was associated with crime increases. The good news regarding the findings of this analysis, however, is that these relatively minor but important programmatic adjustments would be simple and inexpensive to enact. If implemented in a manner that is consistent with both treatment and crime control objectives, these procedural adjustments have the potential to substantially reduce or eliminate unintended community level outcomes associated with the drug court program.