Eight Questions for Drug Policy Research

By Mark A. R. Kleiman, Jonathan P. Caulkins, Angela Hawken, Beau Kilmer

Drug abuse—of licit and illicit drugs alike—is a big medical and social problem and attracts a substantial amount of research attention. But the most attractive and most easily fundable research topics are not always those with the most to contribute to improved social outcomes. If the scientific effort paid more attention to the substantial opportunities for improved policies, its contribution to the public welfare might be greater.

The current research agenda around drug policy concentrates on the biology, psychology, and sociology of drugtaking and on the existing repertoire of drug-control interventions. But that repertoire has only limited capacity to shrink the damage that drug users do to themselves and others or the harms associated with drug dealing, drug enforcement, and drug-related incarceration; and the current research effort pays little attention to some innovative policies with substantial apparent promise of providing improved results.

At the same time, public opinion on marijuana has shifted so much that legalization has moved from the dreams of enthusiasts to the realm of practical possibility. Yet voters looking to science for guidance on the practicalities of legalization in various forms find little direct help.

All of this suggests the potential of a research effort less focused on current approaches and more attentive to alternatives.

The standard set of drug policies largely consists of:

• Prohibiting the production, sale, and possession of drugs
• Seizing illicit drugs
• Arresting and imprisoning dealers
• Preventing the diversion of pharmaceuticals to nonmedical use
• Persuading children not to begin drug use
• Offering treatment to people with drug-abuse disorders or imposing it on those whose behavior has brought them into conflict with the law
• Making alcohol and nicotine more expensive and harder to get with taxes and regulations
• Suspending the drivers’ licenses of those who drive while drunk and threatening them with jail if they keep doing it

With respect to alcohol and tobacco, there is great room or improvement even within the existing policy repertoire for example, by raising taxes), even before more-innovaive approaches are considered. With respect to the currently illicit drugs, it is much harder to see how increasing or slightly modifying standard-issue efforts will measurably shrink the size of the problems.

The costs—fiscal, personal, and social—of keeping half a million drug offenders (mostly dealers) behind bars are sufficiently great to raise the question of whether less comprehensive but more targeted drug enforcement might be the better course. Various forms of focused enforcement offer the promise of greatly reduced drug abuse, nondrug crime, and incarceration. These include testing and sanctions programs, interventions to shrink flagrant retail drug markets, collective deterrence directed at violent drug-dealing organizations, and drug-law enforcement aimed at deterring and incapacitating unusually violent individual dealers. Substantial increases in alcohol taxes might also greatly reduce abuse, as might developing more- effective treatments for stimulant abusers or improving the actual evidence base underlying the movement toward “evidence-based policies.”

These opportunities and changes ought to influence the research agenda. Surely what we try to find out should bear some relationship to the practical choices we face. Below we list eight research questions that we think would be worth answering. We have selected them primarily for policy relevance rather than for purely scientific interest.

1) How responsive is drug use to changes in price, risk, availability, and “normalcy”?

The fundamental policy question concerning any drug is whether to make it legal or prohibited. Although the choice s not merely binary, a fairly sharp line divides the spectrum of options. A substance is legal if a large segment of he population can purchase and possess it for unsupervised “recreational” use, and if there are no restrictions on who can produce and sell the drug beyond licensing and routine regulations.

Accepting that binary simplification, the choice becomes what kind of problem one prefers. Use and use-related problems will be more prevalent if the substance is legal. Prohibition will reduce, not eliminate, use and abuse, but with three principal costs: black markets that can be violent and corrupting, enforcement costs that exceed those of regulating a legal market, and increased damage per unit of consumption among those who use despite the ban. (Total use related harm could go up or down depending on the extent to which the reduction in use offsets the increase in harmfulness per unit of use.)

The costs of prohibition are easier to observe than are its benefits in the form of averted use and use-related problems. In that sense, prohibition is like investments in prevention, such as improving roads; it’s easier to identify the costs than to identify lives saved in accidents that did not happen.

We would like to know the long-run effect on consumption of changes in both price and the nonprice aspects of availability, including legal risks and stigma. There is now a literature estimating the price elasticity of demand for illegal drugs, but the estimates vary widely from one study to the next and many studies are based on surveys that may not give adequate weight to the heavy users who dominate consumption. Moreover, legalization would probably involve price declines that go far beyond the support of historical data.

Furthermore, as Mark Moore pointed out many years ago, the nonprice terms of availability, which he conceptualized as “search cost,” may match price effects in terms of their impact on consumption. Ye t those effects have never been quantitatively estimated for a change as profound as that from illegality to legality. The decision not to enforce laws against small cannabis transactions in the Netherlands did not cause an explosion in use; whether and how much it increased consumption and whether the establishment of retail shops mattered remain controversial questions.

This ignorance about the effect on consumption hamstrings attempts to be objective and analytical when discussing the question of whether to legalize any of the currently illicit drugs, and if so, under what conditions.

2) How responsive is the use of drug Y to changes in policy toward drug X?

Polydrug use is the norm, particularly among frequent and compulsive users. (Most users do not fall in that category, but the minority who do account for the bulk of consumption and harms.) Therefore, “scoring” policy interventions by considering only effects on the target substance is potentially misleading.

For example, driving up the price of one drug, say cocaine, might reduce its use, but victory celebrations should be tempered if the reduction stemmed from users switching to methamphetamine or heroin. On the other hand, school based drug-prevention efforts may generate greater benefits through effects on alcohol and tobacco abuse than via their effects on illegal drug use. Comparing them to other drug-control interventions, such as mandatory minimum sentences for drug dealers, in terms of ability to control illegal drugs alone is a mistake; those school-based prevention interventions are not (just) illicit-drug–control programs.

But policy is largely made one substance at a time. Drugs are added to schedules of prohibited substances based on their potential for abuse and for use as medicine. Reformers clamor for evidence-based policies that rank individual drugs’ harmfulness, as attempted recently by David Nutt, and ban only the most dangerous. Ye t it makes little practical sense to allow powder cocaine while banning crack, because anyone with baking soda and a microwave oven can convert powder to crack.

Considerations of substitution or complementarity ought to arise in making policy toward some of the so-called designer drugs. Mephedrone looks relatively good if most of its users would otherwise have been abusing methamphetamine; it looks terrible if in fact it acts as a stepping stone to methamphetamine use. But no one knows which is the case.

Marijuana legalization is in play in a way it has not been since the 1970s. Various authors have produced social-welfare analyses of marijuana legalization, toting up the benefits of reduced enforcement costs and the costs of greater need for treatment, accounting for potential tax revenues and the like.

Yet the marijuana-specific gains and losses from legalization would be swamped by the uncertainties concerning its effects on alcohol consumption. The damage from alcohol is a large multiple of the damage from cannabis; thus a 10% change, up or down, in alcohol abuse could outweigh any changes in marijuana-related outcomes.

There is conflicting evidence as to whether marijuana and alcohol are complements or substitutes; no one can rule out even larger increases or decreases in alcohol use as a result of marijuana legalization, especially in the long run.

Marijuana legalization might also influence heavy use of cocaine or cigarette smoking. But again, no one knows whether that effect would be to drive cocaine or cigarette use up or down, let alone by how much. If doubling marijuana use led to even a 1% increase or decrease in tobacco use, it could produce 4,000 more or 4,000 fewer tobacco related deaths per year, far more than the (quite small) number of deaths associated with marijuana.

This uncertainty makes it impossible to produce a solid benefit/cost analysis of marijuana legalization with existing data. That suggests both caution in drawing policy conclusions and aggressive efforts to learn more about cross-elasticities among drugs prone to abuse.

3) Can we stop large numbers of drug-involved criminal offenders from using illicit drugs?

Many county, state, and federal initiatives target drug use among criminal offenders. Ye t most do little to curtail drug use or crime. An exception is the drug courts process; some implementations of that idea have been shown to reduce drug use and other illegal behavior. Unfortunately, the resource intensity of drug courts limits their potential scope. The requirement that every participant must appear regularly before a judge for a status hearing means that a drug court judge can oversee fewer than 100 offenders at any time.

The HOPE approach to enforcing conditions of probation and parole, named after Hawaii’s Opportunity Probation with Enforcement, offers the potential for reducing use among drug-involved offenders at a larger scale. Like drug courts, HOPE provides swift and certain sanctions for probation violations, including drug use. HOPE starts with a formal warning that any violation of probation conditions will lead to an immediate but brief stay in jail. Probationers are then subject to regular random drug testing: six times a month at first, diminishing in frequency with sustained compliance. A positive drug test leads to an immediate arrest and a brief jail stay (usually a few days but in some jurisdictions as little as a few hours in a holding cell). Probationers appear before the judge only if they have violated a rule; in contrast, a drug court judge participates in every status review. Thus HOPE sites can supervise large numbers of offenders; a single judge in Hawaii now supervises more than 2,000 HOPE probationers.

In a large randomized controlled trial (RCT), Hawaii’s HOPE program greatly outperformed standard probation in reducing drug use, new crimes, and incarceration among a population of mostly methamphetamine-using felony probationers. A similar program in Tarrant County, Texas (encompassing Arlington and Fort Worth), appears to produce similar results, although this has not yet been verified by an RCT, as has a smaller-scale program (verified by an RCT) among parolees in Seattle. Reductions in drug use of 80%, in new arrests of 30 to 50%, and in days behind bars of 50% appear to be achievable at scale. The last result is the most striking; get-tough automatic-incarceration policies can reduce incarceration rather than increasing it, if the emphasis is on certainty and celerity rather than severity.

The Department of Justice is funding four additional RCTs; those results should help clarify how generalizable the HOPE outcomes are. But to date there has been no systematic experimentation to test how variations in program parameters lead to variations in results.

Hawaii’s HOPE program uses two days in jail as its typical first sanction. Penalties escalate for repeated violations, and the 15% or so of participants who violate a fourth time face a choice between residential treatment and prison. No one is mandated to undergo treatment except after repeated failures. The results suggest that this is an effective design, but is it optimal? Would some sanction short of jail for the first violation—a curfew, home confinement, or community service—work as well? Are escalating penalties necessary and if so, what is the optimal pattern of escalation? Is there a subset of offenders who ought to be mandated to treatment immediately rather than waiting for failures to accumulate? Should cannabis be included in the list of drugs tested for, as it is in Hawaii, or excluded? How about synthetically produced cannabinoids (sold as “Spice”) and cathinones (sold as “bath salts”), which require more complex and costly screening? Would adding other services to the mix improve outcomes? How can HOPE be integrated with existing treatment-diversion programs and drug courts? How can HOPE principles best be applied to parole, pretrial release, and juvenile offenders?

Answering these questions would require measuring the results of systematic variation in program conditions. There is no strong reason to think that the optimal program design will be the same in every jurisdiction or for every offender population. But it’s time to move beyond the question “Does HOPE work?” to consider how to optimize the design of testing-and-sanctions programs.

4) Can we stop alcohol-abusing criminal offenders from getting drunk?

Under current law, state governments effectively give every adult a license to purchase and consume alcohol in unlimited quantities. Judges in some jurisdictions can temporarily revoke that license for those with an alcohol-related offense by prohibiting drinking and going to bars as conditions of bail or probation. However, because alcohol passes through the body quickly, a typical random-but-infrequent testing regiment would miss most violations, making the revocation toothless.

In 2005, South Dakota embraced an innovative approach to this problem, called 24/7 Sobriety. As a condition of bail, repeat drunk drivers who were ordered to abstain from alcohol were now subject to twice-a-day breathalyzer tests, every day. Those testing positive or missing the test were immediately subject to a short stay in jail, typically a night or two. What started as a five-county pilot program expanded throughout the state, and judges began applying the program to offenders with all types of alcohol-related criminal behavior, not just drunk driving. Some jurisdictions even started using continuous alcohol-monitoring bracelets, which can remotely test for alcohol consumption every 30 minutes. Approximately 20,000 South Dakotans have participated in 24/7—an astounding figure for a state with a population of 825,000.

The anecdotal evidence about the program is spectacular; fewer than 1% of the 4.8 million breathalyzer tests ordered since 2005 were failed or missed. That is not because the offenders have no interest in drinking. About half of the participants miss or fail at least one test, but very few do so more than once or twice. 24/7 is now up and running in other states, and will soon be operating in the United Kingdom. As of yet there are no peer-reviewed studies of 24/7, but preliminary results from a rigorous quasi-experimental evaluation suggest that the program did reduce repeat drunk driving in South Dakota. Furthermore, as with HOPE, there remains a need to better understand for whom the program works, how long the effects last, the mechanism(s) by which it works, and whether it can be effective in a more urban environment.

Programs such as HOPE and 24/7 can complement traditional treatment by providing “behavioral triage.” Identifying which subset of substance abusers cannot stop drinking on their own, even under the threat of sanctions, allows the system to direct scarce treatment resources specifically to that minority.

Another way to take away someone’s drinking license would be to require that bars and package stores card every would be to require that bars and package stores card every buyer and to issue modified driver’s licenses with nondrinker markings on them to those convicted of alcohol-related crimes. This approach would probably face legal and political challenges, but that should not discourage serious analysis of the idea.

There is also strong evidence that increasing the excise tax on alcohol could reduce alcohol-related crime. Duke University economist Philip Cook estimates that doubling the federal tax, leading to a price increase of about 10%, would reduce violent crime and auto fatalities by about 3%, a striking saving in deaths for a relatively minor and easy-to-administer policy change. There is also evidence that formal treatment, both psychological and pharmacological, can yield improvements in outcomes for those who accept it.

There is also strong evidence that increasing the excise linked. Among people with drug problems who are also crimtax on alcohol could reduce alcohol-related crime. Duke Uni- inally active, criminal activity tends to rise and fall with drug versity economist Philip Cook estimates that doubling the consumption. Reductions in crime constitute a major benfederal tax, leading to a price increase of about 10%, would efit of providing drug treatment for the offender population, reduce violent crime and auto fatalities by about 3%, a strik- or of imposing HOPE-style community supervision. ing saving in deaths for a relatively minor and easy-to-ad- Reducing drug use among active offenders could also minister policy change. There is also evidence that formal shrink illicit drug markets, producing benefits everywhere, treatment, both psychological and pharmacological, can yield from inner-city neighborhoods wracked by flagrant drug improvements in outcomes for those who accept it.

5) How concentrated is hard-drug use among active criminals?

Literally hundreds of substances have been prohibited, but the big three expensive drugs (sometimes called the “hard” drugs)—cocaine, including crack; heroin; and methamphetamine— account for most of the societal harm. The serious criminal activity and other harms associated with those substances are further highly concentrated among a minority of their users. Many people commit a little bit of crime or use hard drugs a handful of times, but relatively few make a habit of either one. Despite their relatively small numbers, those frequent users and their suppliers account for a large share of all drug-related crime and violence.

The populations overlap; an astonishing proportion of those committing income-generating crimes, such as robbery, as opposed to arson, are drug-dependent and/or intoxicated at the time of their offense, and a large proportion of frequent users of expensive drugs commit income-generating crime. Moreover, the two sets of behaviors are causally linked. Among people with drug problems who are also criminally active, criminal activity tends to rise and fall with drug consumption. Reductions in crime constitute a major benefit of providing drug treatment for the offender population, or of imposing HOPE-style community supervision.

Reducing drug use among active offenders could also shrink illicit drug markets, producing benefits everywhere, from inner-city neighborhoods wracked by flagrant drug dealing to source and transit countries such as Colombia and Mexico.

A back-of-the envelope calculation suggests the potential size of these effects. The National Survey on Drug Use and Health estimates users in the household population. The Arrestee Drug Abuse Monitoring Program measures the rate of active substance use among active offenders (by self-report and urinalysis). Two decades ago, an author of this article (Kleiman) and Chris Putala, then on the Senate Judiciary Committee staff, used the predecessors of those surveys to estimate that about three-quarters of all heavy (morethan-weekly) cocaine users had been arrested for a nondrug felony in the previous year.

Applying the Pareto Law’s rule of thumb that 80% of the volume of any activity is likely to be accounted for by about 20% of those who engage in it—true, for example, about the distribution of alcohol consumption—suggests that something like three-fifths of all the cocaine is used by people who get arrested in the course of a typical year and who are therefore likely to be on probation, parole, or pretrial release if not behind bars.

Combining that calculation with the result from HOPE that frequent testing with swift and certain sanctions can shrink (in the Hawaii case) methamphetamine use among heavily drug-involved felony probationers by 80%, suggests that total hard-drug volume might be reduced by something like 50% if HOPE-style supervision were applied to all heavy users of hard drugs under criminal-justice supervision. No known drug-enforcement program has any comparable capability to shrink illicit-market volumes.

By the same token, HOPE seems to reduce criminal activity, as measured by felony arrests, by 30 to 50%. If frequent offenders commit 80% of income-generating crime, and half of those frequent offenders also have serious harddrug problems, such a reduction in offending within that group could reduce total income-generating crime by approximately 15 to 20%, while also decreasing the number of jail and prison inmates.

The Kleiman and Putala estimate was necessarily crude because it was based on studies that weren’t designed to measure the concentration of hard-drug use among offenders. Unfortunately, no one in the interim has attempted to refine that estimate with more precise methods (for example, stochastic-process modeling) or more recent data.

6) What is the evidence for evidence-based practices?

Many agencies now recommend (and some states and federal grant programs mandate) adoption of prevention and treatment programs that are evidence-based. But the move toward evidence-based practices has one serious limitation: the quality of the evidence base. It is important to ask what qualifies as evidence and who gets to produce it. Many programs are expanded and replicated on the basis of weak evidence. Study design matters. A review by George Mason University Criminologist David Weisburd and colleagues showed that the effect size of offender programs is negatively related to study quality: The more rigorous the study is, the smaller its reported effects.

Who does the evaluation can also make a difference. Texas A&M Epidemiologist Dennis Gorman found that evaluations authored by program developers report much larger effect sizes than those authored by independent researchers. Yet Benjamin Wright and colleagues reported that more than half of the substance-abuse programs targeting criminal-justice programs that were designated as evidence-based on the Substance Abuse and Mental Health Services Administration’s (SAMHSA’s) National Registry of Evidence Based Programs and Practices (NREPP) include the program developer as evaluator. Consequently, we may be spending large sums of money on ineffective programs. Many jurisdictions, secure in their illusory evidence base, could become complacent about searching for alternative programs that really do work.

We need to get better at identifying effective strategies and helping practitioners sort through the evidence. Requiring that publicly funded programs be evaluated and show improved outcomes using strong research designs—experimental designs where feasible, well-designed historicalcontrol strategies where experiments can’t be done, and “intent-to-treat” analyses rather than cherry-picking success by studying program completers only—would cut the number of programs designated as promising or evidence-based by more than 75%. Not only would this relieve taxpayers of the burden of supporting ineffective programs, it would also help researchers identify more promising directions for future intervention research.

The potential for selection biases when studying druginvolved people is substantial and thus makes experimental designs much more valuable. Small is key. It avoids expense, and equally important, it avoids champions with bruised egos. It is difficult to scale back a program once an agency becomes invested in the project. Small pilot evaluations that do show positive outcomes can then be replicated and expanded if the replications show similarly positive results.

7) What treats stimulant abuse?

Science can alleviate social problems not only by guiding policy but also by inventing better tools. The holy grail of such innovations would be a technology that addresses stimulant dependence.

The ubiquitous “treatment works” mantra masks a sharp disparity in technologies available for treating opiates (heroin and oxycodone) as opposed to stimulants (notably cocaine, crack, and meth). A variety of so-called opiate-substitute therapies (OSTs) exist that essentially substitute supervised use of legal, pure, and cheap opiates for unsupervised use of street opiates. Methadone is the first and best-known OST, but there are others. A number of countries even use clinically supplied heroin to substitute for street heroin.

OST stabilizes dependent individuals’ chaotic lives, with positive effects on a wide range of life outcomes, such as increased employment and reduced criminality and rates of overdose. Sometimes stabilization is a first step toward abstinence, but for better and for worse the dominant thinking since the 1980s has been to view substitution therapy as an open-ended therapy, akin to insulin for diabetics. Either way, OST consistently fares very well in evaluations that quantify social benefits produced relative to program costs.

There is no comparable technology for treating stimulant dependence. This is not for lack of trying. The National Institute on Drug Abuse has invested hundreds of millions of dollars in the quest for pharmacotherapies for stimulants. Decades of work have produced many promising advances in basic science, but with comparatively little effect on clinical practice. The gap between opiate and stimulant treatment technologies matters more in the United States and the rest of the Western Hemisphere, where stimulants have a large market, than in the rest of the world, where opiates remain predominant.

There are two reactions to this zero-for-very-many batting average. One is to redouble efforts; after all, Edison tried a lot of filament materials before hitting on carbonized bamboo. The other is to give up on the quest for a chemical that can offset, undo, or modulate stimulants’ effects in the brain and pursue other approaches. For example, immunotherapies are a fundamentally different technology inasmuch as the active introduced agent does not cross the blood-brain barrier. Rather, the antibodies act almost more like interdiction agents, but interdicting the drug molecules between ingestion and their crossing the blood-brain barrier rather than interdicting at the nation’s border.

There is evidence from clinical trials showing that some cognitive-behavioral therapies can reduce stimulant consumption for some individuals. Contingency management also takes a behavioral rather than a chemical approach, essentially incentivizing dependent users to remain abstinent. The stunning finding is that, properly deployed, very small incentives (for example, vouchers for everyday items) can induce much greater behavioral change than can conventional treatment methods alone.

The ability of contingency management to reduce consumption, and the finding that even the heaviest users respond to price increases by consuming less, profoundly challenge conventional thinking about the meaning of addiction. They seem superficially at odds with the clear evidence that addiction is a brain disease with a physiological basis. Brainimaging studies let us see literally how chronic use changes the brain in ways that are not reversed by mere withdrawal of the drugs. So just as light simultaneously displays characteristics of a particle and a wave, so too addiction simultaneously has characteristics of a physiological disease and a behavior over which the person has (at least limited) control.

8) What reduces drug-market violence?

Drug dealers can be very violent. Some use violence to settle disputes about territory or transactions; others use violence to climb the organizational ladder or intimidate witnesses or enforcement officials. Because many dealers have guns or have easy access to them, they also sometimes use these weapons to address conflicts that have nothing to do with drugs. Because the market tends to replace drug dealers who are incarcerated, there is little reason to think that routine drug-law enforcement can reduce violence; the opposite might even be true if greater enforcement pressure makes violence more advantageous to those most willing to use it.

That raises the question of whether drug-law enforcement can be designed specifically to reduce violence. One set of strategies toward this end is known as focused deterrence or pulling-levers policing. These approaches involve lawenforcement officials directly communicating a credible threat to violent individuals or groups, with the goal of reducing the violence level, even if the level of drug dealing or gang activity remains the same. Such interventions aim to tip situations from high-violence to low-violence equilibria by changing the actual and perceived probability of punishment; for example, by making violent drug dealing riskier, in enforcement terms, than less violent drug dealing.

The seminal effort was the Boston gun project Ceasefire, which focused on reducing juvenile homicides in the mid-1990s. Recognizing that many of the homicides stemmed from clashes between juvenile gangs, the strategy focused on telling members of each gang that if anyone in the gang shot someone (usually a member of a rival gang) police and prosecutors would pull every lever legally available against the entire gang, regardless of which individual had pulled the trigger. Instead of receiving praise from colleagues for increasing the group’s prestige, the potential shooter now had to deal with the fact that killing put the entire group at risk. Thus the social power of the gang was enlisted on the side of violence reduction. The results were dramatic: Youth gun homicides in Boston fell from two a month before the intervention to zero while the intervention lasted. Variants of Ceasefire have been implemented across the country, some with impressive results.

An alternative to the Ceasefire group-focused strategy is a focus on specific drug markets where flagrant dealing leads to violence and disorder. Police and prosecutors in High Point, North Carolina, adopted a focused-deterrence approach, which involved strong collaborations with community members. Their model, referred to as the Drug Market Intervention, involved identifying all of the dealers in the targeted market, making undercover buys from them (often on film), arresting the most violent dealers, and not arresting the others. Instead, the latter were invited to a community meeting where they were told that, although cases were made against them, they were going to get another chance as long as they stopped dealing. The flagrant drug market in that neighborhood, as David Kennedy reports, vanished literally overnight and has not reappeared for the subsequent seven years. The program has been replicated in dozens of jurisdictions, and there is a growing evidence base showing that it can reduce crime.

A third approach recognizes the heterogeneity in violence among individual drug dealers. By focusing enforcement on those identified as the most violent, police can create both Darwinian and incentive pressures to reduce the overall violence level. This technique has yet to be systematically evaluated. This seems like an attractive research opportunity if a jurisdiction wants to try out such an approach.

An especially challenging problem is dealing-related violence in Mexico, now claiming more than 1,000 lives per month. It is worth considering whether a Ceasefire-style strategy might start a tipping process toward a less violent market. Such a strategy could exploit two features of the current situation: The Mexican groups make most of their money selling drugs for distribution in the United States, and the United States has much greater drug enforcement capacity than does Mexico. If the Mexican government were to select one of the major organizations and target it for destruction after a transparent process based on relative violence levels, U.S. drug-law enforcement might be able to put the target group out of business by focusing attention on the U.S. distributors that buy their drugs from the target Mexican organization, thereby pressuring them to find an alternative source. If that happened, the target organization would find itself without a market for its product.

If one organization could be destroyed in this fashion, the remaining groups might respond to an announcement that a second selection process was underway by competitively reducing their violence levels, each hoping that one of its rivals would be chosen as the second target. The result might be—with the emphasis on might—a dramatic reduction in bloodshed.

Whatever the technical details of violence-minimizing drug-law enforcement, its conceptual basis is the understanding that in established markets enforcement pressure can have a greater effect on how drugs are sold than on how much is sold. So violence reduction is potentially more feasible than is greatly reducing drug dealing generally.

Conclusion

Drug policy involves contested questions of value as well as of fact; that limits the proper role of science in policymaking. And many of the factual questions are too hard to be solved with the current state of the art: The mechanisms of price and quantity determination in illicit markets, for example, have remained largely impervious to investigation. Conversely, research on drug abuse can provide insight into a variety of scientifically interesting questions about the nature of human motivation and self-regulation, complicated by imperfect information, intoxication, and impairment, and engaging group dynamics and tipping phenomena; not every study needs to be justified in terms of its potential contribution to making better policy. However, good theory is often developed in response to practical challenges, and policymakers need the guidance of scientists. Broadening the current research agenda away from biomedical studies and evaluations of the existing policy repertoire could produce both more interesting science and more successful policies.