Measuring the Demand for Drugs
Illegal drug use is a covert behavior. Whether such use is ignored, tolerated, or aggressively deterred through law enforcement, it occurs outside the explicit framework of legal markets. Determining the prevalence of such use—defined as either the number of users or the quantity of drugs consumed—poses inherent challenges to both social scientists and epidemiologists. Interpreting the patterns is further complicated by the heterogeneity within the population of drug users. A substance such as marijuana is consumed in small quantities by many casual users, who may use it irregularly and rarely satisfy standard criteria for abuse or dependence.1 In contrast, most people who use heroin consume it regularly and frequently, and they are much more likely to satisfy the Diagnostic and Statistical Manual of Mental Disorders 4th Edition (DSM-IV) criteria for substance use disorders. This chapter describes the datasets that are available on drug use and its consequences in the United States. It assesses the strengths and weaknesses of each dataset and how it contributes to understanding of the demand for illegal drugs.
The National Household Survey of Drug Abuse and the National Survey of Drug Use and Health
The 1990-2001 National Household Survey of Drug Abuse (NHSDA) and its successor, the National Survey of Drug Use and Health (NSDUH), provide key data regarding the prevalence of substance use, abuse, and dependence and substance abuse treatment participation in a nationally representative sample of the noninstitutionalized U.S. population. These datasets include information regarding substance use, psychiatric disorders (including substance abuse and dependence), welfare receipt, and substance abuse treatment receipt during the 12 months prior to the survey interview.
Figure 3-1 shows changes in the percentage of respondents (aged 12 years and older) who reported that they had used cocaine or marijuana in the previous 30 days from 1979 to 2007. For marijuana, the prevalence of use fell sharply in the 1980s from a very high rate (13 percent) in the late 1970s, rebounded modestly in the 1990s, and has been relatively stable since 2002 at about 6 percent. For cocaine, the story is somewhat similar,
though the figures are much lower: in 2007, only 0.8 percent of respondents reported cocaine use in the past 30 days.2
For no other illegal drug are prevalence rates so high. Methamphetamine has become a major health and criminal justice problem in many parts of the country, as indicated by the numbers of treatment admissions and the percentage of arrestees testing positive for use of that drug; however, the prevalence of past month use among 18-25-year-olds, the highest use group, has never risen above 0.7 percent.
In recent years a new pattern of drug use has emerged that has generated considerable concern: the reported consumption of diverted pharmaceuticals, that is, prescription drugs (see, e.g., Compton and Volkow, 2006). In 2004, 6.2 percent of the population aged 12 and over reported nonprescribed use of a prescription drug in the previous 12 months. Among those aged 18-25, the rate was more than twice as high, 14.8 percent. Approximately 12 percent of those reporting use within the past 12 months reported that they had used more than twice per week over that period.
The NHSDA and NSDUH have many limitations that complicate trend analysis. Such analyses are particularly difficult when one seeks to compare current substance use patterns to those of the mid-1990s or earlier because of changes in survey methodology. The two surveys do not provide data for incarcerated individuals or those in residential treatment settings. They also do not provide chemical verification of survey responses. Other aspects of NHSDA and NSDUH design also suggest that these surveys provide poor coverage for the most criminally active segment of the drug-using population (see Fendrich et al., 2004; Gfroerer et al., 1997; Midanik, 1982; Midanik and Greenfield, 2003; Pollack and Reuter, 2006). NHSDA and NSDUH also face more general challenges that result in declining response rates and increased rates of refusal, which is true for many epidemiological studies over the past three decades (Galea and Tracy, 2007).
Perhaps most important, the surveys are of self-reported data and are therefore vulnerable to underreporting of substance use and other stigmatized characteristics and behaviors. There are known biases in reported substance use and in substance abuse treatment (see Midanik and Greenfield, 2003; Minkoff et al., 1997). NHSDA and NSDUH are known to underrepresent frequent users of cocaine and heroin and to underrepresent the overall volume consumed of both substances (National Institute on Drug Abuse, 1997; Office of National Drug Control Policy, 2001).
Analysis of the 2000 NHSDA illustrates these difficulties. Only 29 of 58,647 respondents reported at least weekly heroin use over the past 12 months. Accounting for the weighted and stratified nature of NHSDA, this number corresponds to an estimated 150,528 weekly heroin users3 in the United States. But this number represents approximately 16 percent of the estimated number of weekly heroin users as determined by a study done for the Office of National Drug Control Policy (ONDCP) (2001) in the same year.
NHSDA and NSDUH have captured a somewhat greater number of cocaine users. The 2000 NHSDA included 225 respondents who reported at least weekly powder or crack cocaine use (Office of National Drug Control Policy, 2001, Table 3). This number corresponds to an estimated 606,364 weekly cocaine users.4 However, this estimate is still less than ONDCP’s estimated number of chronic cocaine users.
Research by Fendrich and colleagues (2004) attempted to validate household survey responses of Chicago respondents through the use of biomarkers. The authors found that the majority of women who tested positive for heroin and cocaine in hair, urine, or saliva tests did not reveal their use of these substances. Responses regarding marijuana use appeared more complete in these data. Harrison (1995) and Harrison and Hughes (1997) documents these patterns in greater detail, showing that self-report bias increases with the social stigma associated with a specific substance and that self-administered questionnaires reduce, but do not eliminate, such underreporting. A more recent study by Harrison and colleagues (2007) of a large subsample of the NSDUH wave found that only 21 percent of those who tested positive for recent use of cocaine reported that in their questionnaires. When the NSDUH replaced the NHSDA in 2002, it included several survey design improvements. The survey now appears to capture a somewhat greater percentage of chronic substance users. An analysis of 2008 NSDUH data showed an estimated 173,839 weekly heroin users5 and an estimated 1,096,630 weekly cocaine users (powder or crack).6 The documentation for the survey specifically warns against performing trend analysis that compares NHSDA and NSDUH data because of the major changes in survey methodology. Among other things, increased payments to respondents reduced survey nonresponse, and so it appears that there was an increase in estimated drug use prevalence in the NSDUH (Substance Abuse and Mental Health Services Administration, 2003). Likely as a result of improved survey
methodology, the estimated prevalence of last-year cocaine use rose from 1.9 percent in 2001 to 2.5 percent in 2002 (almost a one-third increase), which is implausible in the light of the much more modest changes in the years before and after.
Survey methodology poses other obstacles to trend analysis in many variables. Over the 1990s, NHSDA used varying operational definitions of important demographic variables, including family income, welfare participation, and the age and number of dependent children in the household. We believe that we have constructed consistent subsamples for the committee’s trend analysis. However, NHSDA and NSDUH pose difficulties for trend analysis not found in more consistently implemented surveys, such as the Monitoring the Future (MTF) datasets used to track adolescent substance use.
Until the year 2000, NHSDA did not operationalize DSM-III-Revised criteria for abuse. NHSDA provides an inconsistent and incomplete measure of drug and alcohol dependence across survey years—a problem addressed in one-time surveys such as the 2002 National Epidemiologic Survey of Alcohol and Related Conditions and now the NSDUH, but not addressed in a consistently implemented annual survey.
Despite these limitations, NHSDA and NSDUH provide nationally representative individual data widely used for policy analysis, though the lack of state identifiers in public-use files has been a major hindrance to such analysis. We take this up in Chapter 5.
Monitoring the Future Survey
The MTF survey, which began in 1975 and continues, examines substance use and other behaviors for a nationally representative sample of approximately 50,000 8th, 10th, and 12th grade students in 420 schools across the United States. MTF provides a high-quality data source to scrutinize the prevalence of self-reported substance use among students enrolled in school. In particular, the survey has asked exactly the same core questions on drug use over its almost 35 years of operation (although initially covering only 12th graders), allowing for consistent data on the major measures. However, a key study limitation is that MTF surveys of high school seniors capture only those who remain in school: dropouts are thus not effectively captured in the survey. For 12th grade, dropouts constitute about 9 percent, although there are wide disparities by race and ethnicity (Child Trends Data Bank, 2010). The MTF survey methodology also undersamples students who are pursuing General Educational Development certification.
MTF technical materials suggest that the survey excludes between 15 and 20 percent of the pertinent cohort in the 12th-grade year (Bachman et
al., 2001). Given the strong correlation between substance use and limited educational attainment, this is an important concern, though one that has been acknowledged and subjected to analysis by MTF investigators (Bachman et al., 2001).
MTF, NHSDA, and NSDUH differ in methodology, and there are consistent differences in reported rate of drug use. Perhaps most importantly, MTF is administered in the classroom and provides respondents with greater anonymity than does the household survey. Thus it is not surprising that analyses comparing the reported rates for youth find higher rates in MTF than for a closely matched age group from NHSDA and NSDUH (Gfroerer, 1992). However, the time trends of the two surveys for the common age groups are so similar that we report only the MTF results for youth to show the greater variation in changes over time for this group in comparison with the broader population aged 12 and over.
Figure 3-2 shows the changes over time in prevalence of drug use among high school seniors. In this figure, for marijuana we use the prevalence of more intense use, namely daily use (“on at least 20 of the last 30 days”). The data again show the deep decline during the 1980s, following the upturn in the late 1970s, the recovery of rates during the 1990s, and the more recent stabilization and decline. Less restrictive measures of use, such as “any use during the past 12 months,” yield higher prevalence
estimates but display similar trends over time (Johnston et al., 2009). At its height, the prevalence of daily use exceeded 10 percent; at its nadir it was barely 2 percent. For cocaine the timing is different, but the pattern is similar. Figure 3-2 also shows that use of ecstasy, a matter of great concern at the end of the 1990s has now declined to very low rates, illustrative of a drug that is briefly popular and then fades from sight.
MTF also creates a panel of high school seniors each year, and the respondents are surveyed on their drug use for many years afterward. Some drug use data from these panels are reported in an annual report. However, the data are little used by scholars outside the Survey Research Center research group at Ann Arbor, which has itself made minimal use of the data. A National Research Council (2001) report commented on the loss of important information attributable to the restriction on access to the data. We take up this matter in Chapter 5.
The 1990-1992 National Comorbidity Study (NCS) was the first nationally representative survey to use a fully structured diagnostic interview to assess the prevalence and correlates of (then-DSM-III) psychiatric disorders, including substance use disorders.
The 2001-2003 Collaborative Psychiatric Epidemiology Surveys (CPES) replicated NCS methodologies. These surveys provide high-quality, nationally representative data to explore a wide range of DSM-IV defined psychiatric disorders, including lifetime and current substance use disorders. They also capture diverse physical and mental health measures, as well as respondents’ sociodemographic characteristics that are likely associated with both welfare receipt and substance use. CPES includes three distinct surveys, each of which is a weighted and stratified national probability sample of a specific population pertinent to policy debate (see Heeringa et al., 2004); see below.
One of them is the 2001-2003 National Comorbidity Study-Replication (NCS-R), an enhanced replication of the NCS. It provides a high-quality, nationally representative survey to explore a wide range of DSM-IV defined psychiatric disorders, including lifetime and current substance use disorders (see Degenhardt et al., 2007). NCS-R also provides data on diverse issues related to individual well-being (Alegria et al., no date). The survey explores such outcomes as homelessness and food insecurity that are of particular importance to very low-income populations. NCS-R also explores problem behaviors, such as fighting, vandalism, and theft. NCS-R examines a more diverse range of psychiatric disorders, with higher fidelity to DSM-IV criteria than is available from other national data sources (such as NHSDA and NSDUH). NCS-R also captures the
receipt of mental health and substance abuse services, along with important information regarding both the financing of services and respondents’ perceived barriers to service receipt.
The second of the CPES is the National Survey of American Life (NSAL). NSAL is a national household probability sample of 3,570 African Americans, 1,006 non-Hispanic whites, and 1,623 Afro-Caribbean adults (Heeringa et al., 2004). NSAL provides the most detailed information currently available on psychiatric disorders, well-being, and social performance of African and Afro-Caribbean Americans (see Ford et al., 2007; Jackson et al., 2004a, 2007a; Neighbors et al., 2007). The survey replicates the methodology and questions used in the NCS-R and it further explores questions of specific concern to populations of color (Jackson et al., 2004b; Pennell et al., 2004).
The third of the CPES is the National Latino and Asian-American Study of Mental Health (NLAAS). NLAAS also replicates the NCS-R methodology to provide the most detailed information currently available on psychiatric disorders, well-being, and social performance of Latino and Asian American adults in U.S. households (see Abe-Kim et al., 2007;(see Abe-Kim et al., 2007; Alegria et al., 2007; Chae et al., 2006; Chatterji et al., 2007; Nicdao et al., 2007; Pennell et al., 2004).
Like NSDUH, CPES likely undersamples individuals with severe mental illness or substance use disorders, precisely because each of the surveys is also a household sample. Because these surveys replicate the NCS methodology, it is possible to examine trend changes in a national sample.
Given the limits of household surveys, particularly in obtaining data on frequent cocaine, heroin, and methamphetamine users, there has been great interest in proxy indicators that might provide insight regarding the levels of use and changing size of this population. Proxy indicators include drug-related emergency department admissions and overdoses, alcohol-related traffic fatalities, admissions into substance abuse treatment, and toxicology screening of arrestees in major metropolitan areas. Each of these proxies captures some dimension of the social harms associated with substance use and fails to capture others. Showing how they jointly provide a picture of drug use remains an important task.
DATASETS FOR RESEARCH
In this section we describe the features and main strengths and weaknesses of the most pertinent datasets for the committee’s work on under-
standing demand. We do not summarize all datasets. For example, we have not discussed cohort studies of treatment participants such as Drug Abuse Treatment Outcome Study and the National Treatment Improvement Evaluation Study (NTIES), or studies of treatment participants affiliated with the Clinical Trials Network.
We also do not summarize other data collection and dissemination activities that are useful for drug policy formulation but that play a smaller role in academic research. For example, the Community Epidemiology Work Group (CEWG) provides a venue for policy makers and researchers to assemble diverse data to conduct and communicate ongoing community-level surveillance of drug use and related trends (National Institute on Drug Abuse, 2010). CEWG seeks to help policy makers and researchers identify emerging trends, characteristics of vulnerable populations, and the social and health consequences of substance use (Community Epidemiology Work Group, 2009). It is not itself an important source of data for research.
Several datasets provide specific information regarding the population of people who receive substance abuse treatment services. These datasets provide detailed clinical information, as well as administrative data concerning payment sources, entry characteristics of treatment clients, and characteristics of the treatment experience itself. These datasets also provide pre-post data regarding substance use, criminal offending, and other factors.
These datasets also have several limitations. They are not representative of the full population with substance use disorders, since the majority of those people do not receive treatment services. The most detailed datasets are also not generalizable to the full treatment population, since the underlying sample frames are not representative of the full population of treatment units.
National Treatment Improvement Evaluation Study
NTIES, conducted from 1992 to 1997, features a large sample size of substance abuse treatment clients across short- and long-term residential settings, methadone maintenance, and ambulatory outpatient interventions. NTIES has a higher follow-up response rate (82 percent) than any comparable client-level follow-up treatment survey (Flynn et al., 2001; Gerstein and Johnson, 2000, 2001). Funded by the Center for Substance Abuse Treatment (CSAT), NTIES is available for public use through the Inter-university Consortium for Political and Social Research (ICPSR) at the University of Michigan (see Gerstein et al., 1997)
NTIES was not designed to be nationally representative of treatment clients. It does not cover people who are out of contact with the substance
abuse treatment system. The sample universe is drawn from units supported by CSAT. Compared with nationally representative client surveys, NTIES included a high percentage of nonwhites and criminal justice clients (Zarkin et al., 2002).
Treatment Episodes Data Set
The Treatment Episodes Data Set-Admissions (TEDS-A) provides annual, individual-level data on the demographic characteristics and substance use disorders for 1.9 million annual client admissions to treatment facilities for substance use disorders. The data items collected include primary and secondary substances of abuse, treatment referral source, prior treatment episodes, age at first use, metropolitan area, and age. The 2005 TEDS-A included more than 640,000 treatment referrals from the criminal justice system, providing ample coverage of this key population of public health and law enforcement concern. Facilities that receive state funding (including federal funding through the substance abuse prevention and treatment block grant) for alcohol or drug disorders form the TEDS-A sample frame. In 1997, TEDS-A was estimated to cover about 67 percent of all substance abuse treatment clients. The system has been characterized by uneven participation by treatment units, particularly in the correctional system. Analyses at the state level can be seriously affected by these inconsistencies.
The Treatment Episode Data Set-Discharges (TEDS-D) is an administrative data system that provides annual client-level data on discharges from alcohol or drug treatment in the same public or private substance abuse treatment facilities that comprise the TEDS sample frame. TEDS-D began data collection in 2000, though data were only released for public use through ICPSR in September 2008 for 2006.
TEDS-D captures several variables that are critically important to policy makers and researchers. It provides basic admissions data, including primary, secondary, and tertiary drug of abuse; number of prior treatments; primary source of referral; employment status; whether methadone was prescribed in treatment; diagnosis codes; presence of psychiatric problems; living arrangements; source of income; health insurance; expected source of payment; substance(s) abused; route of administration; frequency of use; age at first use; pregnancy and veteran status; health insurance; and days waiting to enter treatment. It also provides useful discharge data, such as client length of stay, whether the client successfully completed treatment, and service modality at time of discharge.
TEDS-D features many of the strengths and weaknesses of the TEDS admission data. Investigators request data from all substance abuse treatment facilities that receive public funds. Although data are requested on
all clients, some facilities provide data only on clients whose treatment is financed by public funds. Data are collected on distinct admissions rather than distinct individuals. So some people may appear more than once in TEDS-D data. Moreover, a person who experiences a single treatment episode that involves multiple providers or care modalities may appear as multiple admissions and discharges in these data. Technical features of the data complicate comparisons of TEDS-D data across different states. Facility identifiers are stripped from TEDS-D.
TEDS-D appears to provide a rich set of client and program characteristics for future research, yet we are unaware of any research papers using these data. TEDS-D provides, and will provide, a valuable data source for researchers and policy makers who seek to examine trends in length of stay, treatment completion, and other key measures. Moreover, the data provides a resource for multivariate analysis of basic associations, such as differences in length of stay as a function of insurance type, referral source, and the sociodemographic characteristics of clients.
Although TEDS-D is not fully representative, it provides a large discharge-level national dataset with no close substitute in other available datasets. TEDS-D would be especially valuable if provisions were made to allow controlled research access to additional confidential data, such as identifiers of specific facilities that are linked with the National Survey of Substance Abuse Treatment Services dataset of the Substance Abuse and Mental Health Services Administration (SAMHSA). Such linkage would facilitate comparisons across space and time and also would facilitate improved multivariate analysis controlling for unit effects.
Arrestee Drug Abuse Monitoring Program and Drug Use Forecasting Series
The Arrestee Drug Abuse Monitoring (ADAM) Program and Drug Use Forecasting (DUF) series provides data on the prevalence of drug use among arrested and booked persons. Between 1987 and 1997, DUF collected data in 24 sites across the United States and expanded to 35 sites in 1998. Beginning in 2000, ADAM implemented a probability-based sampling strategy (although a number of studies had shown that the earlier data do not generate biased results). The sampling frame comprised all people arrested and booked on local and state charges in identified ADAM counties in the United States.
ADAM includes detailed, representative data regarding the severity of charges leading to arrest and booking; individuals’ contact with health care and substance abuse treatment systems; lifetime, 12-month, 30-day, and 72-hour experiences of substance use; and circumstances of drug purchases and sales. ADAM also includes voluntary urine test results.
Toxicology results are reported for a wide range of illegal substances. Data are available through SAMHSA.
ADAM was terminated in 2003, but ONDCP (Office of National Drug Control Policy, 2009) restarted it in 10 sites in 2007 and plans to continue it as an annual survey. The only information currently available for the public is some initial prevalence estimates released as a data appendix.
Drug Abuse Warning Network
The Drug Abuse Warning Network (DAWN) is a SAMHSA-funded public health surveillance system that monitors drug-related emergency department use, along with drug-related deaths and other health-related harms investigated by medical examiners and coroners. It began operation in 1972. Until 2002 it provided estimates for emergency department visits both nationally and for about 30 metropolitan areas, and it also provided estimates of deaths due to drug use for about 38 counties in metropolitan areas. Medical chart review data are available since 1994 in selected emergency departments.
In 2002 DAWN switched to a new data collection system, which differs in the kind of record abstracted and the sample of cities and facilities. Many hospitals that had previously reported refused to do so after the switch, partly because of privacy concerns raised by the 1996 Health Insurance Portability and Accountability Act of 1996 and partly because of increasing cost concerns. Only 220 of 550 eligible hospitals participate in the national panel.
By February 2010, almost 8 years after the switch in data collection, there were very few published reports available from the new DAWN. In early 2010 the most recent report of national emergency department data was for 2006; for more recent years the agency website contained only Excel files that could be downloaded and used to prepare tables a researcher might want. No subnational data were available. For drug-related mortality, the 2007 report was available, but it presented no national data, only figures for 10 states and 40 metropolitan areas.
DAWN had been little used in the past: since the new system was created, it has been difficult to use at all since no public-use data (apart from recent national emergency department data) have been made available. The loss of many hospitals from the sample that occurred during the redesign reduced the potential accuracy of the system.
These datasets, along with the others discussed above, provide information on the characteristics of drug users and their experience that adds to understanding of demand. However, the analytic task remains of showing how they should be put together for that purpose.
Public health investigators have also assembled useful datasets
regarding specific critical populations at risk. For example, one valuable recent dataset is the National HIV Behavioral Surveillance System (NHBS). Beginning in 2003, NHBS has collected data regarding sexual and drug use risk behaviors from men who have sex with men, injection drug users, and heterosexual adults in areas in which HIV is prevalent (Sanchez et al., 2006).
ESTIMATES OF PREVALENCE AND QUANTITIES USED
The population surveys described above, and the accompanying indicator series, have mainly been used to examine the prevalence and incidence of drug use in the general population. There have been minimal efforts to estimate the quantities of drug consumed or how much is spent by consumers (see, e.g., Bretteville-Jensen, 2006). One exception is the government estimates that have been published on three occasions (Office of National Drug Control Policy, 1995, 1997, 2001), covering the period since 1988 up to, in the most recent publication, 2000. These studies are, to our knowledge, the only efforts to estimate the number of persons using illegal drugs on a frequent and intense basis, the total quantity consumed by all users, and the amount spent in purchasing the drugs. These are three important measures of consumption and indicators of demand. The studies have been conducted by a research team at Abt Associates on behalf of ONDCP. They rely on integrating data from various data series—including NHSDA and NSDUH (for occasional use), DUF and ADAM (for chronic or heavy use), and DAWN and System to Retrieve Information from Drug Evidence (STRIDE)—and cover four drugs: (1) cocaine, (2) heroin, (3) marijuana, and (4) methamphetamines.
These estimates have been used by policy makers, yet they are based on complex analysis of datasets with important weaknesses that have not been subject to detailed research scrutiny. They show a decline in quantity consumed for cocaine, sharp during the period 1988 to 1991 and gradual thereafter; by 2000 the estimate was more than one-fourth lower than that in 1991: see Table 3-1. For heroin, the figures showed no clear trend but fluctuated over the period.
Changes in the estimates of total cocaine consumption over the period show the limited insight provided by NSDUH data. While the population survey shows fairly stable prevalence during the 1990s, the Abt estimates of chronic users (individuals who had used the drug on at least eight occasions in the previous 30 days) show a substantial decline. This estimate may be the consequence of a decline in the number of dependent users, though it is somewhat surprising since the decline in cocaine prices (discussed later in the chapter) would suggest that each user would consume more.
TABLE 3-1 Total Amount of Cocaine and Heroin Consumed, 1988-2000 (metric tons)
*Estimates for 2000 are projections.
SOURCE: Office of National Drug Control Policy (2001).
There is a minimal base of studies of quantities consumed by individual users (for a review of what is available for heroin, mostly from European studies of treatment samples, see Paoli et al., 2009, Appendix B). Kilmer and Pacula (2009) review studies for other drugs in trying to estimate the total size of the world markets for cocaine, heroin, amphetamines, and marijuana. Treatment samples are a troubling source for estimates of consumption levels, since treatment entry is frequently motivated by problems resulting from higher than usual consumption, so that reports of prior 30-day or 3-month consumption (as is typically collected) provide upwardly biased estimates; that same problem holds for arrestee samples. Questions about earlier periods before entry to treatment suffer from the problems of long-term recall, particularly serious for a population of frequent users of psychoactive drugs.
There is a still more fundamental problem for quantity estimates based on self-reports, namely that, as noted in Chapter 2, users do not know how much was in the package purchased. What the user knows is the cost of the purchase. Thus Abt’s estimates of total quantity are calculated by first estimating total expenditures from self-reports and then dividing that figure by an estimate of average price. Since both the expenditure and price data series are noisy, the result is considerable uncertainty about quantity estimates.
The NHSDA and NSDUH surveys include self-report data on quantity each year. These data generate implausibly low estimates of the total quantity, as would be expected from a survey that missed a large proportion of the users who are frequent users. The data are hardly used in research. It is unclear how one improves the estimation of quantity estimates at the individual level. We discuss this further in Chapter 5.
All the prices discussed in this section are adjusted for the consumer price index and for purity. We discuss data at the city level because there are such large differences in absolute prices, though there is considerable consistency in trends.
One of the most surprising observations about major drug markets over the last 30 years has been failure of increasingly stringent supply-side enforcement (as measured by the number of people imprisoned for offenses related to drug sales) to raise the prices of cocaine and heroin. Indeed, in spite of those stringent efforts, there have been marked price declines over the period.
Price data are generally drawn from STRIDE of the Drug Enforcement Administration (DEA). STRIDE records price and purity information from drug purchases undertaken by the DEA and a few local police
departments (most notably, in the District of Columbia) who use the DEA laboratory for testing drug seizures. STRIDE’s merits and drawbacks have been addressed by a number of authors, including a previous report of the National Research Council (2001; see also Horowitz, 2001; Rhodes and Kling, 2001). Caulkins (2007) provides further information on the proper use of STRIDE and its limitations (Arkes et al., 2008; Caulkins 2007). The most recent data, extending a price and purity series through 2007, were recently published by ONDCP (Office of National Drug Control Policy, 2008).
Figures 3-3 and 3-4 show inflation-adjusted price trends for heroin and powder cocaine, in Chicago; Washington, DC; Atlanta; and San Diego. Both figures show the average price per gram at the average purity offered. Heroin and cocaine displayed sharp price declines between 1983 and 1993, with much slower declines after that. There are occasionally short-lived spikes in prices, but none that has lasted for longer than a year. Price trends are also similar across cities, suggesting the difficulty of any cross-sectional time-series analysis that controls for city and year effects.
The real price of marijuana (for which only national estimates were available) was rather stable over the sample period: see Figure 3-5. However potency (the percentage of tetra hydrocannabinol [THC]) as measured by seizure samples rose over most of the period (National Drug Intelligence Center, 2008), so that one cannot determine what happened to potency-adjusted prices.
Figure 3-5 provides data for three levels of the market. One level is retail transactions, involving purchase of about one-tenth of an ounce at
$20 per gram (in recent years) and individual expenditures of approximately $50. Another level is low-level wholesale transactions, with purchases of about an ounce; the price has been about $10 per gram and expenditures would be around $250. The third level is transactions at the high wholesale level, involving about a pound and, in recent years, a price of $6 per gram and expenditure of $2,500. These data indicate how high a proportion of the final price of marijuana is accounted for the activities of lower level dealers.
In order to assess the effects of these price changes over time on consumption, it is important to pay attention to substances that are potential substitutes or complements to these drugs. The real price of beer and spirits also declined markedly over the same period. Real tobacco prices sharply increased, reflecting state and local excise tax increases, as well as price increases brought about by the tobacco master settlement agreement.
CHANGES IN DRUG MARKETS SINCE 1990
Drug markets have changed in many ways since 1990. In particular, the markets for cocaine and heroin now both involve much older buyers and sellers, and this change has profound consequences for how the markets operate and for their effects on society.
During the 1990s, the number of “chronic users” of cocaine and heroin showed steady decline according to the most recent estimate published by the ONDCP (Office of National Drug Control Policy, 2001).7 Yet the number of emergency department admissions and the number of deaths related to these drugs markedly rose. In the case of heroin, it was estimated that the total number of chronic users fell from 1,000,000 in 1990 to 800,000 in 1999 while the estimated number of emergency department admissions related to heroin rose from 33,000 to 84,000. Over this time period, the rate of emergency department admissions per heroin addict rose from about 3 per hundred to 10 per hundred. This is consistent with a population which, through aging, is increasingly subject to acute health problems (Scott et al., 2007).
Another manifestation of the aging phenomena may be the decline in crime despite continued high rates of detected crack use. Levitt (2004) argued that the receding of the crack epidemic was a major factor in explaining the decline in black youth homicides in the 1990s, just as the epidemic itself was a principal driver of the homicide rise in the
1980s. In a subsequent article (Fryer et al., 2005), Levitt and colleagues develop a crack index that summarizes diverse indicators of crack use. The index was flat through most of the 1990s, and the authors conjecture that the decline in homicide, in particular, arose from the creation of property rights—that is, established ownership of specific locations for selling drugs—in a stabilized market. The property rights hypothesis is an interesting one; we know of no evidence to directly test it. However, large urban policy initiatives, such as the Chicago Housing Authority’s Transformation project, may provide policy experiments to scrutinize this hypothesis (Jacob and Ludwig, 2006). A recent study of the Denver heroin market (Hoffer, 2006) points to the complexity of arrangements in these markets and the extent to which they are shaped by specific physical and social environments. In Denver, the open air heroin market settled in an area that had been occupied by a number of homeless men, some of whom were themselves heroin addicts. When Hoffer observed the market in the 1990s, these men had become important go-betweens for the more professional sellers, mostly illegal Mexican immigrants working for a Mexican drug gang, and the broader population of users in the city. The city cleaned up the area in the mid-1990s, partly to prepare for the new baseball stadium. This change made the area much less attractive both to customers and to the immigrant sellers; the locals moved from being go-betweens to active sellers themselves and forced the market to be reconfigured in a number of different ways.
Given that male violence declines with age, a simpler, compelling hypothesis for the changed linkage between aggregate measures of crack use and homicide may be found in the aging of the crack-using population, conjectured in MacCoun and Reuter (2001). This pattern is also consistent with prison inmate survey data, which show marked aging in the population of prison inmates who reported recent cocaine use at the time of their incarceration (Pollack, Reuter, and Sevigny, 2010). Prison inmate survey data also indicate sharply declining age profiles in violent offending among cocaine users (Pollack et al., 2010).
The contrasting trends in numbers and adverse consequences suggest that the overall number of drug users is just one of several variables that influence the health, employment, and crime consequences of substance use. The age of drug users, the duration and intensity of their drug use, and other factors play important roles. Similar insights apply to the supply side of illegal drug markets. The aging of drug sellers and the maturing of drug markets may be more important than the overall number of drug sellers in determining the social effects of these markets on local communities.
An influential study by Levitt and Venkatesh (2000), based on data collected in the early 1990s, examined the young and eager sellers will-
ing to work for low wages in the hope of succeeding to the position of a high-level dealer. These sellers, 15 years later, may form an aging cohort of cocaine-dependent sellers, who are advantaged by the fact that they take some of their return in the form of reduced-price drugs. More recently, youths may no longer be so readily tempted to enter into drug selling rather than completing school.
In this respect, data collected on juvenile arrestees in the District of Columbia since 1987 are of some interest. In the late 1980s more than 20 percent of juvenile arrestees tested positive for recent cocaine use; the comparable figure since about 2003 has been less than 4 percent (District of Columbia Pretrial Services Agency, 2009).
Given the chronic, relapsing nature of substance use disorders, these age patterns become especially important (Pollack et al., 2010). For example, Hser and colleagues (2001) found that the risk of incarceration for a cohort of heroin addicts they recruited in 1964 varied over the 33 years that they followed them. When the addicts were surveyed at the first follow-up in 1973-1974 at average age 37, 23 percent were incarcerated; in 1996-1997, at average age 57, only 14 percent of the survivors were incarcerated.
Most recently, Basu, Paltiel, and Pollack (2008) used data from NTIES to examine criminal offending among substance abuse treatment clients. These authors report that clients under the age of 25 were four times as likely to report that they had recently robbed someone with a weapon as were clients over the age of 30. Although by some measures older clients achieved better treatment outcomes, substance abuse treatment was most cost-beneficial when provided to the most criminally active population of male clients under 25, precisely because these younger drug addicts inflict such high costs on society through their criminal offending.
Recently, there has been some attention to the aging of the population being treated for drug dependence. Trunzo and Henderson (2007) show that, of those in treatment for drugs or drugs and alcohol, the number over age 50 quintupled in 13 years (1992-2005), while the total population in treatment rose only by 14 percent over about the same period (1993-2003). According to 2005 TEDS data, substance abuse treatment clients over the age of 50 have been using for a very long time (Trunzo and Henderson, 2007): the average duration of cocaine use was 20 years; the average duration of heroin use was 34 years.
These data indicate strong period effects in the reported initiation of some substances, though not others. Figure 3-6 shows the reported year of first use among patients recently admitted for heroin use disorders aged 50 or older in 2005: more than one-third of them initiated use between 1966 and 1971; more than three-fourths initiated use before 1980.
Figure 3-7 shows the most dramatic descriptive evidence of cohort
aging among in-treatment substance users. The figure, drawn from 1992 and 2006 TEDS data, displays changes in the age distribution of clients admitted for cocaine (smoked) disorders. In 1992 more than 50 percent of those entering treatment were 30 years old or younger; in 2006 that figure had dropped to 21 percent. At the same time, the percent over age 40 rose from 7 percent to more than 40 percent. These changes do not reflect the consequence of an epidemic of new use among the older population; rather, they represent the aging of those who were caught in the earlier epidemics.
Similar, although somewhat weaker evidence of aging can be found in DAWN emergency department data: see Figure 3-8. The population-adjusted rate of cocaine-related admissions hardly changed between 1994 and 2002 for age groups under 35. The rate increased by 75 percent for patients aged 35-44, and it more than doubled for those aged 45-54.
In the case of heroin, there is other evidence of a sudden elevation of initiation rates during the late 1960s and early 1970s, followed by a rapid decline to a much lower rate, a phenomenon first reported by Kozel and Adams (1986). Similarly, in an early 1990s sample of street heroin users, Rocheleau and Boyum (1994) also found evidence of much higher initiation rates in the early 1970s than in the following 15 years. For cocaine powder, the decline is less pronounced than that for heroin
(Rydell and Everingham, 1994). More recently, Caulkins and colleagues (2004) reported estimates of annual cocaine initiation using NHSDA and a variety of methods; all show a peak in 1980 followed by a decline of two-thirds in the next 5 years. For crack cocaine, the epidemic was still later, starting between about 1982 and 1986, depending on the city (Cork, 1999).
This phenomenon of sudden change in initiation has been the subject of a new class of epidemiologic models developed by Jonathan Caulkins and collaborators (e.g., Caulkins et al., 2004; Tragler et al., 2001). These authors use diverse data to document the long trajectory of drug epidemics. After the peak, the initiation rate does not return to its original zero level, but it does fall to a rate well below the peak. Under reasonable assumptions, the result is a flow of new users who do not fully replace those lost through desistance, death, or incarceration. Thus, the number of dependent users declines over time. Moreover, the drug-using population ages, with corresponding changes in the health, employment, and crime consequences of substance use.
This aging phenomenon is not restricted to the United States. Similar analyses of the aging heroin-dependent population can be found in Switzerland. For example, Nordt and Stohler (2006) show the same kind of sharp increase and decline in heroin initiation. They reference a similar pattern in Italy. However, data from England (De Angelis et al., 2004) and Australia (Law et al., 2001) show a much slower and less peaked epidemic of initiation. These findings are a reminder that epidemics represent social rather than biological contagion and so vary in shape over time and place, and they focus attention on what can be done to prevent new ones from taking hold. In addition to the formal modeling of epidemics of drug use, there is a substantial observational literature, often based on ethnographic research that describes the process of change; see, for example, Agar and Risinger (2002) on heroin, Hamid (1991) on crack, and Murphy and colleagues (2005) on ecstasy. Understanding what generates these sudden upsurges in particular places and particular times is a research issue of the greatest importance.
Economic models help to illuminate drug markets, but they leave many unsettled questions. Nationally representative survey data provide a useful resource to examine the determinants of occasional drug use, particularly among youth and young adults. The most socially costly forms of chronic substance abuse and dependence are not well captured in available survey data. Other epidemiological sources—including emergency department data and analysis of data from arrestees—provide a better, albeit indirect, window into these patterns.
Abe-Kim, J., D.T. Takeuchi, S. Hong, N. Zane, S. Sue, M.S. Spencer, H. Appel, E. Nicdao, and M. Alegria. (2007). Use of mental health-related services among immigrant and U.S.-born Asian Americans: Results from the National Latino and Asian American study. American Journal of Public Health, 97(1), 91-98.
Agar, M., and H.S. Reisinger. (2002). A heroin epidemic at the intersection of histories: The 1960s epidemic among African Americans in Baltimore. Medical Anthropology, 21, 189-230.
Alegria, M., J.S. Jackson, R.C. Kessler, and D. Takeuchi. (No date). Codebook for National Comorbidity Survey Replication (NCS-R) 2001-2003. Ann Arbor, MI: Inter-university Consortium for Political and Social Research.
Alegria, M., N. Mulvaney-Day, M. Woo, M. Torres, S. Gao, and V. Oddo. (2007). Correlates of past-year mental health service use among Latinos: Results from the National Latino and Asian American Study. American Journal of Public Health, 97(1), 76-83.
Arkes, J., R.L. Pacula, S.M. Paddock, J.P. Caulkins, and P. Reuter. (2008). Why the DEA STRIDE Data Are Still Useful for Understanding Drug Markets. Cambridge, MA: National Bureau of Economic Research.
Bachman, J.G., L.D. Johnston, and P.M. O’Malley. (2001). The Monitoring the Future project after twenty-seven years: Design and procedures. In Monitoring the Future Occasional Papers. Ann Arbor, MI: Institute for Social Research.
Basu, A., A.D. Paltiel, and H.A. Pollack. (2008). Social costs of robbery and the cost-effectiveness of substance abuse treatment. Health Economics, 17(8), 927-946.
Bretteville-Jensen, A.L. (2006). Drug demand—Initiation, continuation and quitting. The Economist, 154(4), 491-516.
Caulkins, J.P. (2007). Price and purity analysis for illicit drug: Data and conceptual issues. Drug Alcohol Dependence, 90(S1), S61-S68.
Caulkins, J.P., D.A. Behrens, C. Knoll, G. Tragler, and D. Zuba. (2004). Markov chain modeling of initiation and demand: The case of the U.S. cocaine epidemic. Health Care Management Science, 7(4), 319-329.
Chae, D.H., A.R. Gavin, and D.T. Takeuchi. (2006). Smoking prevalence among Asian Americans: Findings from the National Latino and Asian American study (NLAAS). Public Health Reports, 121(6), 755-763.
Chatterji, P., M. Alegria, M. Lu, and D. Takeuchi. (2007). Psychiatric disorders and labor market outcomes: Evidence from the National Latino and Asian American study. Health Economics, 16(10), 1,069-1,090.
Child Trends Data Bank. (2010). High School Dropout Rates. Available: http://www.childtrendsdatabank.org/?q=node/162 [accessed August 2010].
Community Epidemiology Work Group. (2009). Epidemiologic Trends in Drug Abuse: Proceedings of the Community Epidemiology Work Group. Available: http://www.drugabuse.gov/PDF/CEWG/CEWGJan09508Compliant.pdf [accesssed August 2010].
Compton, W.M., and N.D. Volkow. (2006). Major increases in opioid analgesic abuse in the United States: Concerns and strategies. Drug and Alcohol Dependence, 81(2), 103-107.
Cork, D. (1999). Examining space-time interaction in city-level homicide data: Crack markets and the diffusion of guns among youth. Journal of Quantitative Criminology, 15(4), 379-406.
De Angelis, D., M. Hickman, and S. Yang. (2004). Estimating long-term trends in the incidence and prevalence of opiate use/injecting drug use and the number of former users: Back-calculation methods and opiate overdose deaths. American Journal of Epidemiology, 160(10), 994-1,004.
Degenhardt, L., W.T. Chiu, N. Sampson, R.C. Kessler, and J.C. Anthony. (2007). Epidemiological patterns of extra-medical drug use in the United States: Evidence from the National Comorbidity Survey Replication, 2001-2003. Drug Alcohol Dependence, 90(2-3), 210-223.
District of Columbia Pretrial Services Agency. (2009). Drug Testing Statistics. Available: http://www.dcpsa.gov/foia/foiaERRpsaDT.htm#Juv [accessed May 2009].
Fendrich, M., T.P. Johnson, J.S. Wislar, A. Hubbell, and V. Spiehlter. (2004). The utility of drug testing in epidemiological research: Results from a general population survey. Addiction, 99(2), 197-208.
Flynn, P.M., D.D. Simpson, M.D. Anglin, and R.L. Hubbard. (2001). Comment on nonresponse and selection bias in treatment follow-up studies. Substance Use and Misuse, 36(12), 1,753-1,757.
Ford, B.C., K.M. Bullard, R.J. Taylor, A.K. Toler, H.W. Neighbors, and J.S. Jackson. (2007). Lifetime and 12-month prevalence of diagnostic and statistical manual of mental disorders, fourth edition disorders among older African Americans: Findings from the national survey of American life. American Journal of Geriatric Psychiatry, 15(8), 652-659.
Fryer, R.G., P.S. Heaton, S.D. Levitt, and K.M. Murphy. (2005). Measuring the Impact of Crack Cocaine. NBER Working Paper 11318. Cambridge, MA: National Bureau of Economic Research.
Galea, S., and M. Tracy. (2007). Participation rates in epidemiologic studies. Annals of Epidemiology, 17(9), 643-653.
Gerstein, D.R., and R.A. Johnson. (2000). Nonresponse and selection bias in treatment follow-up studies. Substance Use and Misuse, 35(6-8), 971-1,014.
Gerstein, D.R., and R.A. Johnson. (2001). In reply to Patrick M. Flynn and colleagues. Substance Use and Misuse, 36(12), 1,753-1,757.
Gerstein, D.R., A.R. Datta, J.S. Ingels, R.A. Johnson, K.A. Rasinski, S. Schildhaus, and K. Talley. (1997). National Treatment Improvement Evaluation Study (NTIES) Final Report. Chicago: National Opinion Research Center at the University of Chicago. Available: http://www.icpsr.umich.edu/SAMHDA/NTIES/NTIES-PDF/ntiesfnl.pdf [accessed February 2010].
Gfroerer, J. (1992). An overview of the National Household Survey on Drug Abuse and related methodological research. In Proceedings of the Section on Survey Research Methods (pp. 464-469). Alexandria, VA: American Statistical Association.
Gfroerer, J., D. Wright, and A. Kopstein. (1997). Prevalence of youth substance use: The impact of methodological differences between two national surveys. Drug and Alcohol Dependence, 47(1), 19-30.
Hamid, A. (1991). Ganja to crack: Caribbean participation in the underground economy in Brooklyn, 1976-1986. Part 1, Establishment of the marijuana economy. International Journal of the Addictions, 26(6), 615-628.
Harrison, L.D. (1995). The validity of self-reported data on drug use. Journal of Drug Issues, 25(1), 91-111.
Harrison, L.D., and A. Hughes (Eds.). (1997). The Validity of Self-Reported Drug Use: Improving the Accuracy of Survey Estimates. NIDA Research Monograph 167. NIH Pub. No. 97-4141. Washington, DC: U.S. Government Printing Office. Available: http://www.nida.nih.gov/pdf/monographs/monograph167/download167.html [accessed August 2010].
Harrison, L.D., S.S. Martin, T. Enev, and D. Harrington. (2007). Comparing Drug Testing and Self-Report of Drug Use Among Youths and Young Adults in the General Population. HHS Publication No. SMA 07-4249, Methodology Series M-7. Rockville, MD: Substance Abuse and Mental Health Services Administration, Office of Applied Studies. Available: http://www.oas.samhsa.gov/validity/drugTest.cfm [accessed August 2010].
Heeringa, S.G., J. Wagner, M. Torres, N. Duan, T. Adams, and P. Berglund. (2004). Sample designs and sampling methods for the Collaborative Psychiatric Epidemiology Studies (CPES). International Journal of Methods of Psychiatric Research, 13(4), 221-240.
Hoffer, L. (2006). Junkie Business: The Evolution and Operation of a Heroin Dealing Network. Belmont, CA: Thompson Wadsworth.
Horowitz, J.L. (2001). Should the DEA’s STRIDE data be used for economic analyses of markets for illegal drugs? Journal of the American Statistical Association, 96(456), 1,254-1,262.
Hser, Y.I., V. Hoffman, C.E. Grella, and M.D. Anglin. (2001). A 33-year follow-up of narcotics addicts. Archives of General Psychiatry, 58(5), 503-508.
Jackson, J.S., M. Torres, C.H. Caldwell, H.W. Neighbors, R.M. Nesse, R.J. Taylor, S.J. Trierweiler, and D.R. Williams. (2004a). The National Survey of American Life: A study of racial, ethnic and cultural influences on mental disorders and mental health. International Journal of Methods in Psychiatric Research, 13(4), 196-207.
Jackson, J.S., H.W. Neighbors, R.M. Nesse, S.J. Trierweiler, and M. Torres. (2004b). Methodological innovations in the National Survey of American Life. International Journal of Methods of Psychiatric Research, 13(4), 289-298.
Jackson, J.S., H.W. Neighbors, M. Torres, L.A. Martin, D.R. Williams, and R. Baser. (2007). Use of mental health services and subjective satisfaction with treatment among Black Caribbean immigrants: Results from the National Survey of American Life. American Journal of Public Health, 97(1), 60-67.
Jacob, B., and J. Ludwig. (2006). The Effect of Family Income on Youth Outcomes: New Evidence from a Housing Voucher Lottery. NBER Working Paper 14570. Cambridge, MA: National Bureau of Economic Research.
Johnston, L.D., P.M. O’Malley, J.G. Bachman, and J.E. Schulenberg. (2009). Trends in Annual Prevalence of an Illicit Drug Use Index in Grades 8, 10, and 12. Institute for Social Research 2009 [cited January 31, 2010]. Available: http://www.monitoringthefuture. org/data/09data.html [accessed August 2010].
Kilmer, B., and R.L. Pacula. (2009). Estimating the Size of the Global Drug Market: A Demand-Side Approach. Report 2. Santa Monica, CA: RAND.
Kozel, N.J., and E.H. Adams. (1986). Epidemiology of drug abuse: An overview. Science, 234(4,779), 970-974.
Law, M.G., M. Lynskey, J. Ross, and W. Hall. (2001). Back-projection estimates of the number of dependent heroin users in Australia. Addiction, 96(3), 433-443.
Levitt, S. (2004). Understanding why crime fell in the 1990s: Four factors that explain the decline and six that do not. Journal of Economic Perspectives, 18(1), 163-190.
Levitt, S.D., and S.A. Venkatesh. (2000). An economic analysis of a drug-selling gang’s finances. Quarterly Journal of Economics, 115(3), 755-789.
MacCoun, R.J., and P. Reuter. (2001). Drug War Heresies: Learning from Other Vices, Times, and Places. New York: Cambridge University Press.
Midanik, L. (1982). The validity of self-reported alcohol consumption and alcohol problems: Literature review. British Journal of Addiction, 77(4), 357-382.
Midanik, L.T., and T.K. Greenfield. (2003). Telephone versus in-person interviews for alcohol use: Results of the 2000 National Alcohol Survey. Drug and Alcohol Dependence, 72(3), 209-214.
Minkoff, H., T. Bauer, and T. Joyce. (1997). Welfare reform and the obstetrical care of immigrants and their newborns. New England Journal of Medicine, 337(10), 705-707.
Murphy, S., P. Sales, M. Duterte, and C. Jacinto. (2005). A Qualitative Study of Ecstasy Sellers in the San Francisco Bay Area. A research report submitted to the U.S. Department of Justice. Available: http://www.ncjrs.gov/pdffiles1/nij/grants/209267.pdf [accessed August 2010].
National Drug Intelligence Center. (2008). National Drug Threat Assessment 2009: Marijuana. Washington, DC: U.S. Department of Justice. Available: http://www.justice.gov/ndic/pubs31/31379/marijuan.htm [accessed August 2010].
National Institute on Drug Abuse. (1997). The Validity of Self-Reported Drug Use: Improving the Accuracy of Survey Estimates. Research Monograph 167. Washington, DC: National Institute on Drug Abuse.
National Institute on Drug Abuse. (2010). Community Epidemiology Work Group (CEWG). Available: http://www.drugabuse.gov/about/organization/CEWG/CEWGHome.html [accessed August 2010].
National Research Council. (2001). Informing America’s Policy on Illegal Drugs: What We Don’t Know Keeps Hurting Us. Committee on Data and Research for Policy on Illegal Drugs, C.F. Manski, J.V. Pepper, and C.V. Petrie (Eds.). Committee on Law and Justice and Committee on National Statistics, Commission on Behavioral and Social Sciences and Education. Washington, DC: National Academy Press.
Neighbors, H.W., C. Caldwell, D.R. Williams, R. Nesse, R.J. Taylor, K.M. Bullard, M. Torres, and J.S. Jackson. (2007). Race, ethnicity, and the use of services for mental disorders: Results from the National Survey of American Life. Archives of General Psychiatry, 64(4), 485-494.
Nicdao, E.G., S. Hong, and D.T. Takeuchi. (2007). Prevalence and correlates of eating disorders among Asian Americans: Results from the National Latino and Asian American Study. International Journal of Eating Disorders, 40, S22-S26.
Nordt, C., and R. Stohler. (2006). Incidence of heroin use in Zurich, Switzerland: A treatment case register analysis. Lancet, 367, 1,830-1,834.
Office of National Drug Control Policy. (1995). What America’s Users Spend on Illegal Drugs, 1988-1993. Washington, DC: Executive Office of the President.
Office of National Drug Control Policy. (1997). What America’s Users Spend on Illegal Drugs, 1988-1995. Washington, DC: Executive Office of the President.
Office of National Drug Control Policy. (2001). What America’s Users Spend on Illegal Drugs, 1988-2000. Washington, DC: Executive Office of the President.
Office of National Drug Control Policy. (2008). The Price and Purity of Illicit Drugs: 1981-2007. Washington, DC: Executive Office of the President.
Office of National Drug Control Policy. (2009). Arrestee Drug Abuse Monitoring (ADAM) II 2008 Annual Report. Washington, DC: Executive Office of the President.
Paoli, L., V.A. Greenfield, and P. Reuter. (2009). The World Heroin Market: Can Supply Be Cut? New York: Oxford University Press.
Pennell, B.E., A. Bowers, D. Carr, S. Chardoul, G.Q. Cheung, K. Dinkelmann, N. Gebler, S.E. Hansen, S. Pennell, and M. Torres. (2004). The development and implementation of the National Comorbidity Survey Replication, the National Survey of American Life, and the National Latino and Asian American Survey. International Journal of Methods of Psychiatric Research, 13(4), 241-269.
Pollack, H.A., and P. Reuter. (2006). Welfare receipt and substance-abuse treatment among low-income mothers: The impact of welfare reform. American Journal of Public Health, 96(11), 2,024-2,031.
Pollack, H.A., P. Reuter, and E. Sevigny. (2010). If Drug Treatment Works So Well, Why Are So Many Drug Users in Prison? Cambridge, MA: National Bureau of Economic Research.
Rhodes, W., and R. Kling. (2001). Should the DEA’s STRIDE data be used for economic analyses of markets for illegal drugs? Comment. Journal of the American Statistical Association, 96(456), 1,264-1,266.
Rocheleau, A.M., and D. Boyum. (1994). Heroin Uses in New York, Chicago, and San Diego. Washington, DC: U.S. Government, Office of National Drug Control Policy.
Rydell, C.P., and S.S. Everingham. (1994). Controlling Cocaine: Supply Versus Demand Programs. Santa Monica, CA: RAND.
Sanchez, T., T. Finlayson, A. Drake, S. Behel, M. Cribbin, E. DiNenno, T. Hall, S. Kramer, and A. Lansky. (2006). Human immunodeficiency virus (HIV) risk, prevention, and testing behaviors—United States, National HIV behavioral surveillance system: Men who have sex with men, November 2003-April 2005. Morbidity and Mortality Weekly Report, 55(SS06), 1-16.
Scott, G., S.D. Thomas, H.A. Pollack, and B. Ray. (2007). Observed patterns of illicit opiate overdose deaths in Chicago, 1999-2003. Journal of Urban Health, 84(2), 292-306.
Substance Abuse and Mental Health Services Administration. (2002). Emergency Department Trends from DAWN: Final Estimates 1994-2001. Available: http://dawninfo.samhsa.gov/ old_dawn/pubs_94_02/edpubs/2001final/default.asp [accessed August 2010].
Substance Abuse and Mental Health Services Administration. (2003). Results from the 2002 National Survey on Drug Use and Health: National Findings. Washington, DC: U.S. Department of Health and Human Services. Available: http://www.samhsa.gov/oas/ nhsda/2k2nsduh/Results/2k2Results.htm#1.2 [accessed August 2010].
Substance Abuse and Mental Health Services Administration. (2007). Substate Area Estimates of Substance Use Disorders for Maryland, 2005 and 2006. Rockville, MD: U.S. Department of Health and Human Services.
Substance Abuse and Mental Health Services Administration. (2008). Results from the 2007 National Survey on Drug Use and Health: National Findings. Available: http://www.oas. samhsa.gov/nsduh/2k7nsduh/2k7Results.pdf [accessed August 2010].
Tragler, G., J.P. Caulkins, and G. Feichtinger. (2001). Optimal dynamic allocation of treatment and enforcement in illicit drug control. Operations Research, 49(3), 352-362.
Trunzo, D., and L. Henderson. (2007). Older Adult Admissions to Substance Abuse Treatment: Findings from the Treatment Episode Data Set. Paper presented at the meeting of the American Public Health Association, Washington, DC. Available: http://apha.confex. com/apha/135am/webprogram/Paper160959.html [accessed August 2010].
Zarkin, G.A., L.J. Dunlap, J.W. Bray, and W.M. Wechsberg. (2002). The effect of treatment completion and length of stay on employment and crime in outpatient drug-free treatment. Journal of Substance Abuse Treatment, 23(4), 261-267.